# TOLEROGENIC ANTIGEN-PRESENTING CELLS – MODULATING UNWANTED IMMUNE RESPONSE AT THEIR CORE

EDITED BY : John Isaacs and Catharien Hilkens PUBLISHED IN : Frontiers in Immunology

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-176-6 DOI 10.3389/978-2-88963-176-6

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# TOLEROGENIC ANTIGEN-PRESENTING CELLS – MODULATING UNWANTED IMMUNE RESPONSE AT THEIR CORE

Topic Editors: John Isaacs, Newcastle University, United Kingdom Catharien Hilkens, Newcastle University, United Kingdom

Citation: Isaacs, J., Hilkens, C., eds. (2019). Tolerogenic Antigen-Presenting Cells – Modulating Unwanted Immune Response at Their Core. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-176-6

# Table of Contents

*06 Tolerogenic Dendritic Cells as a Promising Antigen-Specific Therapy in the Treatment of Multiple Sclerosis and Neuromyelitis Optica From Preclinical to Clinical Trials*

Georgina Flórez-Grau, Irati Zubizarreta, Raquel Cabezón, Pablo Villoslada and Daniel Benitez-Ribas


Mohamed B. Ezzelarab, Lien Lu, William F. Shufesky, Adrian E. Morelli and Angus W. Thomson

*96 Tolerogenic Dendritic Cells in Solid Organ Transplantation: Where do we Stand?*

Eros Marín, Maria Cristina Cuturi and Aurélie Moreau


Patty Sachamitr, Alison J. Leishman, Timothy J. Davies and Paul J. Fairchild

*192 Nanoparticle-Based Modulation and Monitoring of Antigen-Presenting Cells in Organ Transplantation*

Jordi Ochando and Mounia S. Braza

*198 Monitoring T-Cell Responses in Translational Studies: Optimization of Dye-Based Proliferation Assay for Evaluation of Antigen-Specific Responses*

Anja Ten Brinke, Natalia Marek-Trzonkowska, Maria J. Mansilla, Annelies W. Turksma, Karolina Piekarska, Dorota Iwaszkiewicz-Grześ, Laura Passerini, Grazia Locafaro, Joan Puñet-Ortiz, S. Marieke van Ham, Maria P. Hernandez-Fuentes, Eva M. Martínez-Cáceres and Silvia Gregori

*213 Tolerance Through Education: How Tolerogenic Dendritic Cells Shape Immunity*

Matthias P. Domogalla, Patricia V. Rostan, Verena K. Raker and Kerstin Steinbrink

*227 The Immunomodulatory Potential of tolDCs Loaded With Heat Shock Proteins*

Willem van Eden, Manon A. A. Jansen, A Charlotte MT de Wolf, Irene S. Ludwig, Paul Leufkens and Femke Broere


*253 Update on Dendritic Cell-Induced Immunological and Clinical Tolerance* Carolina Obregon, Rajesh Kumar, Manuel Antonio Pascual, Giuseppe Vassalli and Déla Golshayan

*271 Regulating Immunogenicity and Tolerogenicity of Bone Marrow-Derived Dendritic Cells Through Modulation of Cell Surface Glycosylation by Dexamethasone Treatment*

Kevin Lynch, Oliver Treacy, Jared Q. Gerlach, Heidi Annuk, Paul Lohan, Joana Cabral, Lokesh Joshi, Aideen E. Ryan and Thomas Ritter

### *287 Dexamethasone and Monophosphoryl Lipid A Induce a Distinctive Profile on Monocyte-Derived Dendritic Cells Through Transcriptional Modulation of Genes Associated With Essential Processes of the Immune Response*

Paulina A. García-González, Katina Schinnerling, Alejandro Sepúlveda-Gutiérrez, Jaxaira Maggi, Ahmed M. Mehdi, Hendrik J. Nel, Bárbara Pesce, Milton L. Larrondo, Octavio Aravena, María C. Molina, Diego Catalán, Ranjeny Thomas, Ricardo A. Verdugo and Juan C. Aguillón

### *300 Clinical Tolerogenic Dendritic Cells: Exploring Therapeutic Impact on Human Autoimmune Disease*

Brett Eugene Phillips, Yesica Garciafigueroa, Massimo Trucco and Nick Giannoukakis

# Tolerogenic Dendritic Cells as a Promising Antigen-Specific Therapy in the Treatment of Multiple Sclerosis and Neuromyelitis Optica From Preclinical to Clinical Trials

*Georgina Flórez-Grau1,2, Irati Zubizarreta2 , Raquel Cabezón1,2, Pablo Villoslada2 and Daniel Benitez-Ribas <sup>1</sup> \**

*1Department of Immunology, Hospital Clinic i Provincial, Barcelona, Spain, 2Neuroimmunology Group, Institut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain*

#### *Edited by:*

*Natalio Garbi, Universität Bonn, Germany*

#### *Reviewed by:*

*Muriel Moser, Free University of Brussels, Belgium Isis Ludwig-Portugall, Universität Bonn, Germany*

> *\*Correspondence: Daniel Benitez-Ribas dbenitezr@clinic.cat*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 30 October 2017 Accepted: 09 May 2018 Published: 31 May 2018*

#### *Citation:*

*Flórez-Grau G, Zubizarreta I, Cabezón R, Villoslada P and Benitez-Ribas D (2018) Tolerogenic Dendritic Cells as a Promising Antigen-Specific Therapy in the Treatment of Multiple Sclerosis and Neuromyelitis Optica From Preclinical to Clinical Trials. Front. Immunol. 9:1169. doi: 10.3389/fimmu.2018.01169*

The identification of activated T-lymphocytes restricted to myelin-derived immunogenic peptides in multiple sclerosis (MS) and aquaporin-4 water channel in neuromyelitis optica (NMO) in the blood of patients opened the possibility for developing highly selective and disease-specific therapeutic approaches. Antigen presenting cells and in particular dendritic cells (DCs) represent a strategy to inhibit pro-inflammatory T helper cells. DCs are located in peripheral and lymphoid tissues and are essential for homeostasis of T cell-dependent immune responses. The expression of a particular set of receptors involved in pathogen recognition confers to DCs the property to initiate immune responses. However, in the absence of danger signals different DC subsets have been revealed to induce active tolerance by inducing regulatory T cells, inhibiting pro-inflammatory T helper cells responses or both. Interestingly, several protocols to generate clinical-grade tolerogenic DC (Tol-DC) *in vitro* have been described, offering the possibility to restore the homeostasis to central nervous system-related antigens. In this review, we discuss about different DC subsets and their role in tolerance induction, the different protocols to generate Tol-DCs and preclinical studies in animal models as well as describe recent characterization of Tol-DCs for clinical application in autoimmune diseases and in particular in MS and NMO patients. In addition, we discuss the clinical trials ongoing based on Tol-DCs to treat different autoimmune diseases.

Keywords: tolerogenic dendritic cells, dendritic cells, immunotherapy, immunosuppression, multiple sclerosis, Neuromyleitis optica

### INTRODUCTION

Multiple Sclerosis (MS) is a chronic autoimmune inflammatory disease affecting the central nervous system (CNS) (1). Nowadays, there are 2.3 million affected people worldwide, being the most frequent age of diagnosis between 20 and 40 years old (2). Additionally, the studies determine that MS is more frequent in women and in northern locations. There are different subtypes of MS which

**Abbreviations:** APCs, antigen presenting cells; APL, altered peptide ligand; AQP4, aquaporin-4; CNS, central nervous system; DCs, dendritic cells; Mo-DCs, monocyte-derived DCs; MS, multiple sclerosis; NMO, neuromyelitis optica; PPMS, primaryprogressive MS; RA, rheumatoid arthritis; RRMS, relapsing-remitting MS; SPMS, secondary-progressive MS; Tol-DCs, tolerogenic DCs.

are based on their clinical phenotype (3). These subtypes are: The primary-progressive MS (PPMS) which is a disabling subtype from the beginning, the relapsing-remitting type (RRMS) that is characterized by clinical relapses without progression of disability and finally, the secondary-progressive subtype that appears about 20 years after RRMS.

The MS diagnosis is summarized in the revised 2010 Mc Donald criteria which is included in **Table 1** (4). Although the cause of the immune deregulation is unknown, there are evidences that implicate Th1 and Th17 lymphocytes in the pathophysiology of MS (5–10). Furthermore, it was supported by studies performed in experimental models of MS either knocking out or blocking using monoclonal antibodies for IL-17 or IL-23 resulted in a suppression of the activity of this disease (11, 12). Other authors have described that memory T-cells are activated in the periphery by different processes that can be promoted by environmental or genetic factors. These activated cells cross the blood–brain barrier, penetrate to CNS where they are locally reactivated (9, 13).

First-line therapies for MS include injectable treatments such as IFN-β, and glatiramer as well as oral therapies such as teriflunomide and dimethyl-fumarate. Second-line therapies include fingolimod, and the intravenous natalizumab, which present higher levels of efficacy in reducing the relapse rate; however, it has potential severe side effects. Moreover, Alentuzumab, Cladribine, and Ocrelizumab were recently added as approved therapies, and they are in progress of being defined in the pyramid of the MS therapy. All these mentioned treatments are systemic immunomodulatory or immunosuppressive treatments with risks of adverse events.

Neuromyelitis optica (NMO) is an inflammatory disease affecting the CNS (14) with similar physiopathology as MS, but is considered an autoimmune astrocytopathy. NMO is a rare disease which presents with incidence between 0.05 and 0.4/100,000 (15, 16). About 70% of the patients diagnosed with NMO shows the presence of anti-aquaporin-4 (AQP4) antibody as well as specific T-lymphocytes in the bloodstream or CSF which suggest the pro-inflammatory role of these cells (17). Importantly, the detection of anti-AQP4 antibodies is related with more severe disease (14). Recently, among seronegative patients, anti-(MOG) antibodies have been described as the pathological antibody (18). This disease has its own international consensus diagnostic criteria (19), defining the NMO spectrum disorder (NMOSD) concept (**Table 2**). Different MS drugs such as natalizumab or finolimob have been evaluated in NMO resulting in exacerbation of relapses (20). Immunomodulatory or immunossuppressant therapies are used for label in NMOSD (e.g., azathioprine, mycophenolate, cyclophosphamide, or rituximab) (21). Furthermore, several monoclonal antibodies are in clinical trials to evaluate their efficacy and safety, as tocilizumab, satralizumab, eculizumab, or aquapuromab for example (22). Based in the unmet need of achieving higher levels of efficacy and/or better safety profile, antigen-specific therapies are being considered as a potential treatment for MS and NMO (19).

### DENDRITIC CELLS (DCs)

Dendritic cells act as a link between innate and adaptive immune responses. Their main function is to capture and process exogenous antigens in the peripheral tissues to present them to T-cells after migration to the draining lymph nodes. In addition, they polarize immune responses by promoting both pro- and antiinflammatory immune responses depending on the presence of danger signals associated to the antigens (**Figure 1**) (24, 25).



Diagnostic criteria for NMO spectrum disorder (NMOSD) with aquaporin-4 (AQP4)-IgG.


Diagnostic criteria for NMOSD without AQP4-IgG or NMOSD with unknown AQP4-IgG status.

	- (a) At least 1 core clinical characteristic must be optic neuritis, acute myelitis with LETM, or area postrema syndrome
	- (b) Dissemination in space (2 or more different core clinical characteristics) (c) Fulfillment of additional MRI requirements, as applicable

Core clinical characteristics.


Additional MRI requirements for NMOSD without AQP4-IgG and NMOSD with unknown AQP4-IgG status.


Dendritic cells are located in peripheral tissues (skin and mucosa) and remain in an immature state (iDCs) until they interact with the antigens. After cells activation, DCs initiate a maturation process in which mature DCs (mDCs) lose capacities for antigen uptake in favor of acquiring stimulatory properties for the activation of naïve T-cells and the development of effector T-cells (27). Maturation process involves different processes and physiological changes in DCs, which are illustrated in **Figure 2** (28).

Due to their immunological functions and the availability of clinical-grade reagents, immunogenic DCs have been safely used in clinical trials to potentiate immune response against tumors or infectious diseases (30). However, only a few studies recently published have taken advantage of their specific tolerogenic properties to treat Type 1 diabetes, rheumatoid arthritis (RA) and Crohn's disease patients (25, 31, 32).

### HUMAN DCs SUBSETS

Dendritic cells can be sub-classified based on anatomical location, origin, and function. In humans, different DC subsets have been identified in blood, spleen and skin and in non-lymphoid tissues.

Adapted from: O'Neil et al. (26). TLRs: toll-like receptors, HSP: Heat shock proteins.

Each DC subset presents different specialization in T-cell priming and induction of immune responses, although their functions can partially overlap (33).

In peripheral blood, DCs that express Human Leukocyte Antigen—antigen D Related (HLA-DR) and lineage negative fraction are divided into two main groups: conventional myeloid DCs (cDCs) and non-conventional plasmacytoid DCs (pDC). Within myeloid DCs two main subsets have been identified based on their surface marker expression: CD1c/BDCA-1 cDCs and CD141/BDCA-3 cDCs. However, recently new DC subset classification has been described (CD16 and DC5) (23). Circulating DCs represent a little fraction of total circulating peripheral blood mononuclear cells (PBMCs) as they account for less than 1% of PBMCs (24, 34).

In the skin two different subsets of DCs can be found. Langerhans cells (LCs) which contributes to immune surveillance and CD14 DCs, which are involved in tolerance induction (35, 36).

From all the different DC subsets above mentioned, the BDCA-1, pDCs, LCs, and CD14 have been described to generate both immunogenic and suppressive functions (**Figure 3**). BDCA-1 have the capacity to produce IL-10 in response to *E. coli* and potentially contribute to suppress immune responses. Recently, a particular subset of BDCA-1 (BDCA1-CD14<sup>+</sup>) has been shown to act as immunosuppressive cells in certain types of tumor environment and may hamper anti-cancer DCs vaccines (37, 38) Moreover, in an steady state, pDCs are able to induce tolerogenic immune responses by inducing T-cell anergy and promoting T-reg cells development. They have been found to be infiltrated in tumors activating Tr1 cells (33, 39, 40). LCs, apart from respond to intracellular pathogens and viruses under inflammatory conditions are in charge to maintain epidermal health and tolerance to commensals from the skin, while retaining the

ability to respond to selected pathogens (40–42). Finally CD14 DCs also have the ability to generate T-regs through the elevated IL-10 production (43, 44).

To sum up, BDCA-1, pDCs, LCs, and CD14 have been shown to present immunoregulatory effects. However, deeper characterization of this tolerogenic profile and mechanisms needs to be performed.

### TOLEROGENIC DCs (Tol-DCs) AND MECHANISMS OF TOLERANCE INDUCTION

As described in the previous section, DCs play a crucial role in the initiation of immune responses and also in maintaining the immune tolerance. DCs present both foreign antigens as well as endogenous antigens derived from tissues. For this reason, the immune system is able to distinguish between innocuous and harmful antigens to avoid autoimmune or undesired immune responses (45). Several studies point that a key factor for DCs to initiate immunity or tolerance is the maturation stage of DCs (25). It is generally accepted that in absence of danger signals provided by infection or inflammation, DCs remain in an immature state which will induce tolerance by deleting or inducing apoptosis of self-antigen-specific T-cells (25, 46). However, other several mechanisms to explain how DCs induce tolerance have been proposed. Some authors have reported that low expression of MHC molecules and co-stimulatory receptors on DC surface fail to stimulate T-cells sufficiently, thus resulting in T-cell anergy (47–49). Currently, it has been demonstrated that the expression of single immunoglobulin IL-1 related receptor, which is lower in iDCs, has a role in maintain low levels of costimulatory molecules and in the regulation of T-reg cell expansion (50). Furthermore, it is well established that the expression of certain molecules such as PD-L1 rather than promote activation signals to T-cells, they induce T-cell anergy (28, 51, 52). Moreover, some authors demonstrated that suboptimal antigen presentation, together with

indoleamine 2,3-dioxygenase (IDO) or Fas-L (CD95L) expression by DCs leads to inhibition of T-cell proliferation and T-cell deletion, respectively. Finally but not the least, the production of the potent anti-inflammatory cytokine IL-10 by DCs is crucial for peripheral tolerance induction. IL-10 acts on a wide variety of immune cells and it has been clearly involved in T-reg as well as Tr1 induction (38). In the steady state, peripheral T-reg cells rise from peripheral CD4<sup>+</sup>CD25<sup>−</sup>FOXP3<sup>−</sup> T cells that are exposed to antigen in the presence of transforming growth factor-β as well as IL-10 without IL-6 or IL-1β, which promotes the up-regulation of FOXP3 (17) (**Figure 4**). Recent developments carried on by Agrawal et al., have shown that C-lectin receptor (CLEC-2) upregulation in DCs, is associated with T-reg induction. Moreover, they have also described that platelet growth factor is able to induce IL-10 production by DCs and in consequence T-reg cell induction (53).

In consequence, major efforts have been focused on *in vitro* generation of Tol-DCs. In this regard, different immunosuppressive drugs, such as corticosteroids, cyclosporine, tacrolimus, rapamycin, deoxyspergualin, vitamin D3 (vitD3), mycophenolate mofetil, and sanglifehrin A, have been successfully used to modulate DCs differentiation and function. Thus, several protocols that include the generation of monocyte-derived DCs in the presence of corticosteroids and a defined maturation cytokine cocktail (including TNF-α, IL-1β, IL-6, and PGE2) or lipopolysaccharide (LPS) activation in order to boost their tolerogenic properties, have been described to generate Tol-DCs *in vitro* (54, 55).

Tolerogenic DCs present an intermediate phenotype between iDCs and mDCs regarding costimulatory molecules, a pronounced shift toward anti-inflammatory versus pro-inflammatory cytokine production (high amounts of IL-10 versus low levels of IL-12p70 and IL-23) and a reduced capacity to stimulate T-cells response. In addition Tol-DCs present an increment of IL-10 production upon Gram-negative bacterial interaction which represents a relevant factor to induce tolerance due to the potent antiinflammatory role of IL-10 (**Figure 5**) (56–58).

The role of *in vitro* generated Tol-DCs as potential immunomodulatory and immunosuppressive agents have been evaluated by different groups (44, 60, 61). The first experimental data to objectify the potential of human Tol-DC to induce tolerance in MS, was the induction of T-cell hyporesponsiveness by Tol-DC from MS patients. The results obtained shown that only Tol-DCs (vitD3) derived from RRMS patients, induced hyporesponsiveness in autologous antigen-specific T-cells restricted to myelinderived peptides and produced higher levels of IL-10 and reduced levels of TNF-α compared to healthy controls, making the tolerogenic potential of these autologous Tol-DCs may be an effective tool to re-establish tolerance in RRMS patients and set up the basis for the ongoing clinical trials (62). In addition, a critical consideration for Tol-DC application in immunotherapy is the phenotype stability once the cells are injected into the patients. It has been demonstrated that *in vitro* generated Tol-DCs have a stable tolerogenic profile after LPS stimulation as they produce higher amounts of IL-10 and as well as they are able to induce antigen-specific T-cell hyporesponsiveness (58, 63).

In summary, Tol-DCs generated *ex vivo* using immunosuppressive agents, induced T regulatory cells through different mechanism such as lower expression of co stimulatory molecules, expression of inhibitory receptors and IL-10 production.

### Tol-DC Therapy in the Animal Model of MS

Animal models are the first step in the development of new therapies, and antigen-specific therapies are not an exception to this rule. Over the past several decades animal models have been used to understand different aspects of human MS. There are three different animal models of MS that are the most commonly used: (1) the experimental autoimmune encephalomyelitis (EAE), (2) viral induced models, and (3) toxin-induced models of demyelination (6).

In addition to the *in vitro* demonstration of the capacity of Tol-DC to induce immune tolerance, the role of Tol-DCs has been evaluated in the EAE model. The critical role of mDCs and pDCs in the chronic pathogenesis of EAE in Lewis rats described by Miller and colleagues makes this model extremely relevant to study positive and negative regulatory pathways involved in MS and other chronic autoimmune diseases (64). Wang et al. demonstrated the involvement of CD11b<sup>+</sup> and CD11c<sup>+</sup> DCs in the generation of both T-regs and Tr1 cells, by depleting DCs they observed that tolerance effect disappeared (65). In consequence, the induction of DCs with a regulatory profile is a key mechanism underlying auto antigen-induced tolerance (64). It is interesting to highlight that studies performed in EAE induced in Lewis rats demonstrated that the maturation state as well as the route of administration influence on the induction of tolerance by these DCs which is in concordance with the *in vitro* performed studies (65, 66). Moreover, different authors have described that the administration of Tol-DCs generated with different immunosuppressive agents such as vitD3 or estriol induced a decrease of the incidence of the disease as well as they promoted the induction of regulatory T-cells though higher levels of IL-10 production (63, 67).

In addition, other authors have performed comparisons regarding the use of immunosuppressive oral drugs such as vitD3 and (for 20 days after EAE induction) or pretreating DCs before EAE induction. The results obtained were similar in both cases: significant improvement of clinical severity and an increase of regulatory CD4<sup>+</sup> Foxp3<sup>+</sup> cells and increased IL-10 levels in lymph nodes from treated animals suggesting that DCs are the main target of tolerogenic effect of vitamin D. Some studies pointed out that in the absence of DCs during the priming process of autoreactive T-cells leads to a unidirectional deficiency of cell generation which results in a fulminant attack against CNS (65, 66, 68). Different studies using DCs to induce tolerance have been performed in EAE animal models of mice and rats and they are summarized in **Table 3**.

In addition, Tol-DCs have also been generated for another disease models such as type I diabetes T1D by using a combination of both dexamethasone and vitD3. This generated Tol-DCs presented a stable phenotype and a high capacity to induce T-reg cells (73). Moreover, other protocols, such as DC treatment with CD40, CD80, and CD86 antisense oligo nucleotides or even low doses of GM-CSF has also been reported although in some cases partial loss of tolerance have been reported.

The critical part is that after being culture, all generated Tol-DCs have to present different characteristics: (a) low levels of co stimulatory molecules, (b) stability when challenges with maturation stimuli and also produce IL-10, (c) lower activation of T-cells (73).

Overall, different protocols for Tol-DCs in preclinical studies has been shown to be beneficious to treat different autoimmune



EAE*,* experimental autoimmune encephalomyelitis*.*

diseases, in particular for EAE induction the use of vitD3 or corticosteroids is the most extended.

### Therapeutic Application of Tol-DCs in Type I Diabetes, RA and Crohn's Disease

Following the encouraging results obtained from different *in vitro* and preclinical studies in animal models, Tol-DCs are revealed as a promising therapy for autoimmune diseases and transplantation (32). Consequently, in 2011, the first phase I clinical trial with Tol-DCs was conducted at the University of Pittsburgh. The trial enrolled 10 insulin-dependent diabetic patients, and administrated control DCs to three patients and immunosuppressive DC (iRNA for CD40, CD80, and for CD86) to seven patients. The treatment was safe and well tolerated. There were no changes in insulin requirements, hematology assessments or blood immune cell population levels in both groups, showing a slight increase of CD4<sup>+</sup>CD25+++ FoxP3<sup>+</sup> T cells in immunosuppressive DC group. All treated patients had normal immune responses to vaccination and alloantigen stimulation *in vitro* (74). Thus, a double-blinded, placebo-controlled cross-over phase II trial is planned to start in Diabetes mellitus type 1 in 24 patients with a recent onset of the disease, inducing tolerability of DC with antisense DNA targeting CD40, CD80, and CD86 (NCT02354911).

Among autoimmune arthritis, two trials have been published recently. In the first one, a unique intradermal administration of "Rheumavax" (autologous DCs modified with a nuclear factor κb inhibitor exposed to 4-citrullinated peptide antigens), was studied in a phase I clinical trial of RA patients. They observed a significant increased ratio of regulatory to effector T cells and a reduction of IL-15, which is a relevant pro-inflammatory cytokine. Moreover, in a more clinical level they found a decrease of DAS28 which is a clinical scale for RA severity together with no disease flares (75). Furthermore, in 2017, results from AUTODECRA trial (NCT01352858) came out resulting a safe and well tolerated therapy with no target knee flares, but with no significant clinical and immunomodulatory changes in serum (76).

Importantly, other clinical trials have been recently reported in other autoimmune diseases such as Crohn's disease. In Crohn's disease, our institution conducted a phase I clinical trial to demonstrate the safety of intraperitoneal administration of autologous Tol-DCs in refractory patients. The immune monitoring studies showed an increase of circulating T-regs and a decrease of IFN-γ production after T-cell activation (31). Regarding organ transplantation, two trials are ongoing. A phase I clinical trial, open-label and non-controlled, in liver transplantation is aimed to assess the safety of Tol-DCs therapy in this type of patients (NCT03164265). The ONEatDC study, aims to assess if Tol-DC administration before renal transplantation is beneficial to reduce immunosuppression needs (NCT02252055).

Overall, the encouraging results obtained in above mentioned clinical trials, of an increase immunosuppressive activity, drawn Tol-DCs as a potential tool to modulate autoinflammatory diseases in the coming years.

### ANTIGEN-SPECIFIC THERAPIES IN MS AND NMO

In the recent years, several strategies to modulate antigenspecific T-cells have been evaluated in therapeutic clinical trials for patients with MS and NMO. Among the advantages to use antigen-specific therapies, they lack of general immunosuppression and its side effects as infections and cancer, as well as the lack of metabolic activity that activates self-reactive T cells, the induction of tolerance to a specific antigen without changing the general immunity (77). The use of DC to induce immune tolerance is also pursuit in patients with MS and NMO. In this sense, a phase I trial to assess the safety of Tol-DC in MS and NMO patients in an ascending dose of intravenous administration of the DCs (NCT02283671) has been performed at our institution and the results are under evaluation. In addition, two more clinical trials are ongoing (NCT02618902) and (NCT02903537), which will provide precious information about safety, modulation of immune response and clinical efficacy.

Several approaches to induce antigen-specific tolerization have been evaluated as DNA vaccination of myelin protein, peptides inoculation, altered peptide ligand (APL) administration to modify TCR recognition, autologous myelin-reactive T cells administration, HLA/MOG recombinant construct administration and autologous PBMCs coupled with myelin-peptides administration, Tol-DCs with myelin-peptides administration (78). Specifically, myelin-peptides approaches are based in a myelin relevant immunodominant peptide administration, like administration of the synthetic peptide itself like MBP, MOG, or PLP, administration of APL or the administration of a region of TCR-peptide complex.

Antigen-specific therapeutic approaches have been demonstrated in the majority of the phase I clinical trials to be safe and well tolerated. However, a trial conducted at NIH with APL induced disease exacerbation and the trial was stopped due to safety issues (79). The concept of APL is based in the administration of modified peptides by introducing some amino acids in substitution in specific positions relevant to link with the TCR, but without changing the MHC binding part. This strategy is aimed to inhibit the inflammatory T cell response, as acts as partial agonist or as antagonist. A phase II trial using MBP83–99 was interrupted as three out of eight participants presented relapses during the clinical trial, that were considered as inflammatory activation as MRI controls showed disease worsening, and this was correlated with MBP specific T cell expansion in blood and CSF samples (80). Two more trials with APLs were done afterward, without objectifying exacerbations of the disease activity (81).

DNA vaccination aims to induce tolerance using heterotopic expression of some antigens, for example using whole human MBP protein. The BHT-3009 molecule is a union of the whole MBP molecule, a human cytomegalovirus promoter and an altered plasmid. In two clinical trials it was demonstrated safe and gadolinium-enhancing lesions were fewer in the treated groups comparing with placebo groups; although, there were significant improvement in clinical outcomes. Immunologically, a decrease in IFN-γ production and T cell proliferation by MBP, PLP, and MOG specific T-cells was observed (82). In another trial, reduction of autoreactive T cells was demonstrated with this approach, creating a proof of concept of the possible efficacy of DNA vaccination (80).

The vaccination with T-cell consists in the administration of activated and irradiated MPB-specific T-cell lines and clones (attenuated autologous T-cells). Phase I and phase II clinical trials have been done, with no relevant side effects, but without significant clinical improvement in treated group comparing with placebo group (83).

Other antigen-specific tolerization approach studied in MS was the antigen-coupled cell tolerance, based on inactivated autologous PBMCs chemically linked with myelin relevant peptides. After proving reduction of onset and severity as well as preventing epitope spreading in EAE, this approach was evaluated in humans. In 2013, a phase I clinical trial (ETIMS trial) was published where antigen-specific tolerance induced with inactivated PBMCs coupled with six immunodominant myelin-peptides was safe, with some immunological promising results to objectify clinical significance (78). Significant advantage of this approach is that the tolerization to several myelin relevant peptides derived from three different antigens (MBP, MOG, and PLP) simultaneously is aimed to prevent the epitope spreading situation.

### REFERENCES


To synthetize, there are different antigen-specific therapies that have been asses in MS patients. The majority has been presented as safe and well tolerated with encouraging data regarding the clinical benefits.

### CONCLUSION AND FUTURE PERSPECTIVES

Antigen-specific tolerance in autoimmune diseases is a therapeutic approach that is currently been evaluated in MS and NMO as well as in other autoimmune diseases. Different reports have demonstrated that DCs are powerful therapeutic tools to modify the immune response and restore the immune tolerance in animal models and in preclinical data. Most importantly, the use of Tol-DCs in clinical trials is being safe in several phase I clinical trials (type I Diabetes, RA and Crohn's disease) showing in some of the studies promising clinical and immunomodulatory results.

In MS several reports have revealed the therapeutic effect of Tol-DCs in ameliorating EAE in animal model. These results highlight the importance of DCs in the homeostasis control and open new avenues for an innovative therapeutic indication for human patients. A major challenge is to translate all these results obtained in animal models to humans. For that reason, it will be crucial to correlate clinical efficacy with modulation of immunological parameters and also to define the optimal administration route, dose of cells, tolerogenic treatments and the potential tolerogenic effect of circulating DCs.

From the studies conducted so far, several important considerations have been raised, application of Tol-DCs in humans is safe and well tolerated without remarkable side effects and showing promising immunological and clinical results. However, phase II and/or III clinical trials including control (placebo) group will bring some light about the clinical efficacy of this therapy in MS/ NMO patients. In addition, more studies are needed to evaluate the real effectiveness and the possibility to use Tol-DC as a real treatment for autoimmune diseases.

### AUTHOR CONTRIBUTIONS

GF-G, IZ, and RC wrote the manuscript and designed figures. PV and DB-R revised the manuscript.


**Conflict of Interest Statement:** PV is an employee of Genentech. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer IL-P and handling Editor declared their shared affiliation.

*Copyright © 2018 Flórez-Grau, Zubizarreta, Cabezón, Villoslada and Benitez-Ribas. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Carl Engman1 , Yesica Garciafigueroa1 , Brett Eugene Phillips1 , Massimo Trucco1,2 and Nick Giannoukakis1,2\**

*<sup>1</sup> Institute of Cellular Therapeutics, Allegheny Health Network, Pittsburgh, PA, United States, 2Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, PA, United States*

Dendritic cells (DC) are important in the onset and severity of inflammatory bowel disease (IBD). Tolerogenic DC induce T-cells to become therapeutic Foxp3+ regulatory T-cells (Tregs). We therefore asked if experimental IBD could be prevented by administration of bone marrow-derived DC generated under conventional GM-CSF/IL-4 conditions but in the presence of a mixture of antisense DNA oligonucleotides targeting the primary transcripts of CD40, CD80, and CD86. These cell products (which we call AS-ODN BM-DC) have demonstrated tolerogenic activity in preventing type 1 diabetes and preserving beta cell mass in new-onset type 1 diabetes in the NOD mouse strain, in earlier studies. In addition to measuring efficacy in prevention of experimental IBD, we also sought to identify possible mechanism(s) of action. Weight, behavior, stool frequency, and character were observed daily for 7–10 days in experimental colitis in mice exposed to dextran sodium sulfate (DSS) following injection of the AS-ODN BM-DC. After euthanasia, the colons were processed for histology while spleen and mesenteric lymph nodes (MLNs) were made into single cells to measure Foxp3+ Treg as well as IL-10+ regulatory B-cell (Breg) population frequency by flow cytometry. AS-ODN BM-DC prevented DSS-induced colitis development. Recipients of these cells exhibited significant increases in Foxp3+ Treg and IL-10+ Breg in MLN and spleen. Histological examination of colon sections of colitis-free mice remained largely architecturally physiologic and mostly free of leukocyte infiltration when compared with DSS-treated animals. Although DSS colitis is mainly an innate immunity-driven condition, our study adds to the growing body of evidence showing that Foxp3+ Treg and IL-10 Bregs can suppress a mainly innate-driven inflammation. The already-established safety of human DC generated from monocytic progenitors in the presence of the mixture of antisense DNA targeting the primary transcripts of CD40, CD80, and CD86 in humans offers the potential to adapt them for clinical IBD therapy.

Keywords: dendritic cells, immune hyporesponsiveness, autoimmunity, tolerogenic dendritic cells, regulatory immune cells, regulatory B-cells, regulatory T-cells, retinoic acid

### INTRODUCTION

While the major target for dendritic cell (DC) therapy, relying on their powerful immunostimulatory ability, has been malignancy (1), the long-sought objective of using the other aspect of DC which is their capability to induce immune hyporesponsiveness, clinically took its first step forward in the last decade in a phase I safety trial humans (2). In addition to this first-in-concept trial for

#### *Edited by:*

*John Isaacs, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Muriel Moser, Free University of Brussels, Belgium Femke Broere, Utrecht University, Netherlands Catharien Hilkens, Newcastle University, United Kingdom*

*\*Correspondence:*

*Nick Giannoukakis ngiannou@wpahs.org*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 27 November 2017 Accepted: 10 April 2018 Published: 03 May 2018*

#### *Citation:*

*Engman C, Garciafigueroa Y, Phillips BE, Trucco M and Giannoukakis N (2018) Co-Stimulation-Impaired Bone Marrow-Derived Dendritic Cells Prevent Dextran Sodium Sulfate-Induced Colitis in Mice. Front. Immunol. 9:894. doi: 10.3389/fimmu.2018.00894*

**16**

type 1 diabetes, accumulating encouraging preclinical data using different embodiments of tolerogenic DC to treat various other autoimmune conditions have made possible a number of other clinical trials. These include trials in the space of rheumatoid arthritis, multiple sclerosis, and intestinal bowel disease (3). Even though the different DC populations differ in the methods used to generate them *ex vivo*, what they appear to share in common is a mechanism that results in the increase in T-regulatory cells, a feature that is inherent in naturally occurring DC that are shaped *in vivo* into a tolerogenic state.

Endogenous DC are mainly found in the immature state and orchestrate tolerance largely by maintaining and promoting the frequency and activity of mainly CD4+ CD25+ regulatory T-cells (Tregs) (4). Immature, co-stimulation impaired DC are known to actively induce the differentiation and proliferation of Foxp3+ Tregs (4–11) [reviewed in Ref. (12, 13)]. This mechanism underlies peripheral tolerance to autoantigens and hyporesponsiveness to alloantigens in transplantation studies (7–11, 14, 15). Co-stimulation-impaired DC and DC engineered to produce cytokines promoting Foxp3+ Tregs successfully prevent, attenuate, and reverse autoimmunity and facilitate allograft survival (7–11, 14, 15).

We showed for the first time that DC generated from monocytic progenitors in the presence of the mixture of antisense DNA targeting the primary transcripts of CD40, CD80, and CD86 were safe in humans (2). In addition, data from this first-in-human trial demonstrated that some recipients of these DC began to exhibit C-peptide positivity during and slightly beyond the cell treatment cycle. This is noteworthy given that these patients were C-peptide negative during screening and baseline testing. Whether this could anticipate potential benefits is currently unknown and will have to be established in phase II trials.

One of the notable characteristics of the DC generated from monocytic progenitors in the presence of the mixture of antisense DNA targeting the primary transcripts of CD40, CD80, and CD86 used in the phase I type 1 diabetes safety trial is their ability to produce retinoic acid (RA) (16, 17). RA and other retinoids have been shown to regulate autoimmunity in rheumatoid arthritis, experimental encephalomyelitis, and type 1 diabetes (18–20). RA, acting *via* the RA receptor, affects the transcription of Foxp3, IL-17, and RORγt, thereby participating in the local homeostasis of Tregs through the balance of Tregs:TH17 cells (21, 22). RA, in fact, has been shown to attenuate experimental colitis by increasing the numbers of Tregs and inhibiting the generation of TH17 cells (22, 23). RA-producing DC are, in fact, naturally found in the mucosa (24, 25), and their role is suggested to be one of maintenance of a stable immunoregulatory state preventing the exacerbation of gut inflammation (24, 25). There is evidence that such RA-producing DC also express CD103 and, at least in the mucosa and more recently in the pancreas, CD103+ DC exert a tolerogenic effect (26–29) even though they can be immunostimulatory under specific conditions (30–33). Tolerogenic DC that express CD103 act *via* their ability to induce Foxp3 expression in T-cells (28, 34–42), especially in the presence of TGF-β in an RA-dependent manner (22, 43–46). Under homeostatic conditions, gut CD103+ DC constitutively migrate to the mesenteric lymph node (MLN) (47). Gut CD103+ DC preferentially support antigen-induced spontaneous differentiation of Foxp3+ Tregs from naive precursors. Furthermore, CD103+ DC isolated from the MLN of ovalbumin-fed mice activate and drive naive DO11.10 CD4+ T cells to express Foxp3 (48). Intestinal CD103+ DC were shown to efficiently differentiate *in situ* into tolerogenic DC (43–45, 48, 49). Thus, adoptive immunotherapy for inflammatory bowel disease (IBD) could become clinically relevant since DC that prevent and reverse T1DM exhibit features similar to gut tolerogenic CD103+ DC; they are stably immature, costimulation-impaired, and express the RA-metabolizing enzyme ALDH1A2 which together convert immunosuppressive progenitors of Foxp3+ Tregs into highly suppressive Foxp3+ Tregs.

Various approaches to generate tolerogenic DC for use in mouse models of IBD have been demonstrated. Curcumin treatment of *in vitro*-generated bone marrow-derived DC resulted in the expression of ALDH1 as well as IL-10 and these DC, acting *via* induction of Tregs and Tr1 cells, inhibited colitis *in vivo* (50). Pedersen et al. used IL-10-conditioned bone marrow-derived DC exposed to an enterobacterial extract to suppress colitis severity and weight loos in SCID mice adoptively transferred with CD4+ CD25− colitogenic T-cells (51). Vasoactive intestinal peptideconditioned bone marrow DC showed efficacy in the TNBS model of murine colitis (52). This study was the first to show that anatomic area selection for DC administration was relevant in facilitating the accumulation of the DC into the MLNs, where the most important antigen presentation and activation of Th1/ Th17 cells takes place (53). A popular approach to generating tolerogenic DC has been the combination dexamethasone/vitamin D3 conditioning of bone marrow DC (54–57), and these DC were shown to suppress colitis in the CD4+ CD25− colitogenic T-cell transfer SCID model (58). Although these antigen-agnostic approaches were effective, some studies suggest that provision of IBD-relevant antigen improves therapeutic outcomes [e.g., by provision of carbonic anhydrase I; (59)].

Although these studies were concurrent with our research in the area of type 1 diabetes, as well as a phase I clinical trial using dexamethasone-generated autologous DC in refractory Crohn's disease having been initiated (http://clinicaltrials.gov identifier NCT02622763), given that bone marrow-derived DC generated in the presence of a mixture of antisense DNA oligonucleotides targeting the CD40, CD80, and CD86 primary transcripts (which we term *AS-ODN BM-DC*) shown to mobilize Tregs and regulatory B-cells (Bregs) in the NOD mouse strain, cells that are critical in maintaining tolerance also in the intestinal tissues, we considered that AS-ODN BM-DC could also be useful to treat IBD, and more importantly in a severe model of murine colitis. The additional rationale to consider our these DC is also underlied by the data showing their production of RA (16) which was a contributing factor to the differentiation of B-cells into IL-10+ Bregs and the proliferation of existing IL-10+ Bregs (16). Given the accumulated evidence that Bregs are also potent regulators of colitis (60–64), that the deficiency of Bregs in mice results in exacerbated arthritis with increased frequency of TH17 cells and decreased Foxp3+ Tregs (65), we have now tested the efficacy of AS-ODN BM-DC to treat IBD using the dextran sodium sulfate (DSS) colitis mouse model and to determine the degree of Treg and/or Breg involvement.

## MATERIALS AND METHODS

### Animals

All mice were maintained in a specific pathogen-free environment, and experiments were conducted in line with specific protocols approved by the Allegheny Health Network IACUC.

### Human Blood

We purchased human complete blood from a commercial source (Grifols) from which we generated DC (see below). The blood products were obtained from a normal adult individual with no reported acute or chronic health conditions or disease.

### Generation of Murine DC

Two types of DC were generated for the purposes of this research endeavor: (i) DC from bone marrow progenitors (which we term BM-DC) and (ii) DC from bone marrow progenitors that were cultured in the presence of a mixture of antisense DNA oligonucleotides targeting the CD40, CD80, and CD86 primary transcripts (which we term *AS-ODN BM-DC*). Both DC populations were generated from bone marrow progenitors from 7- to 8-week-old C57BL/6 mice (Bar Harbor, ME, USA) in 6-day cultures with GM-CSF and IL-4 using previously published protocols (BM-DC) (66, 67). The DC generated in the continuous presence of a mixture of phosphorothioate DNA oligonucleotides targeting the primary transcripts of CD40, CD80, and CD86 (AS-ODN BM-DC) are immunosuppressive. The cells generated at the end of the 6-day culture in the presence of only GM-CSF and IL-4 (BM-DC; no antisense oligonucleotides) are mostly DC; however, there are some undifferentiated monocytic precursors. To generate the AS-ODN BM-DC, the same antisense oligonucleotide sequences and backbone chemistry used in the study by Machen et al. were used in this study (67). BM-DC served as control cell populations in this study. The phenotype and characteristics of the AS-ODN BM-DC have been published elsewhere (66–69). Prior to adoptive transfer of the AS-ODN BM-DC into mice, and for each such experiment, we verified that the general phenotype and functionality of these cells conformed to that which we have previously shown [(67); i.e., low cell surface expression of CD40, CD80, CD86, and the ability to suppress the proliferation of allogeneic leukocytes *in vitro*]. For this, we compare the mean fluorescence intensity of CD40, CD80, and CD86 in day 6 AS-ODN BM-DC to BM-DC using flow cytometry (see below). Table S1 in Supplementary Material also provides other characteristics of BM-DC and AS-ODN BM-DC. To determine the functional phenotype of BM-DC and AS-ODN BM-DC, we added splenocytes from freshly isolated spleen of allogeneic mice (Balb/c) to BM-DC or AS-ODN BM-DC-containing IFNγ ELISPOT assay plates (ELISPOT-PLUS, MabTech) for 72 h as recommended by the manufacturer. Results of these two verifications, representative of routine outcomes, are shown in Figures S1A,B in Supplementary Material.

### DSS Colitis/Treatment of Mice With BM-DC or AS-ODN BM-DC

Following a standard DSS induction protocol (70, 71), mice were randomly placed into three groups (*n* = 4 mice per group; two independent study cohorts totaling *n* = 8 mice per treatment group): DSS, DSS+ BM-DC recipients, and DSS+ AS-ODN BM-DC recipients. Three days prior to exposure to DSS, mice were injected with 2 × 106 BM-DC or AS-ODN BM-DC intraperitoneally (i.p.) in a minimal volume of sterile endotoxin-free PBS or the PBS vehicle only as control. All mice were then switched to drinking water containing 3.5% DSS to which they had *ad libitum* access for 5 days. On day 3 of exposure to DSS, a second injection of 2 × 106 moDC, iDC, or PBS vehicle i.p. was administered. Mice were euthanized 7–10 days after the initiation of DSS exposure.

### Measurements/Assessment of Colitis

Mice were weighed on the day before DSS exposure and then every day thereafter until euthanasia. Colitis was assessed by weight loss, stool consistency, fecal blood, and anal prolapse. Upon euthanasia, colons were harvested, flushed, and fixed for histopathological and immunofluorescence assessment. Concurrently, the MLNs and spleen were collected, made into single cells in preparation for flow cytometric measurements.

### Flow Cytometry

FACSCalibur/FACSAria with DIVA support (BD Biosciences) or Influx workstations with species-specific antibodies, nonoverlapping fluorophores, and appropriate isotype controls were used for flow-sorting and FACS analyses. Cells were antibody stained either after pre-enrichment for specific populations over magnetic columns (Miltenyi Biotec) or stained as freshly isolated single cells from MLNs or spleen *in vitro*.

To measure Tregs, we used the detection system that includes the FJK-16s Foxp3-specific antibody, CD4-FITC clone RM4-5, and CD25-APC clone PC61.5 (eBioscience). For B-cell population characterization and FACS analysis, the following antibodies were used (all from BD Biosciences): B220 (clone RA3-6B2), CD19 (clone 1D3), CD5 (clone 53-7.3), and CD1d (clone 1B1). IL-10-producing cells were identified following positive selection along IL-10 surface adsorption using a commercial magnetic isolation method (Miltenyi Biotec product #130-090-435, Auburn, CA, USA). Characterization of these cells as Bregs was then confirmed by FACS with the B-cell antibodies listed above.

To measure the frequency of DC producing RA, with or without the expression of CD103, we first stained single splenocytes or MLN cells with the ALDEFLUOR reagent (StemCell Technologies, BC, Canada) (72, 73) with parallel control cell cultures treated with *N*,*N*-diethyl-amino-benzaldehyde (DEAB), an inhibitor of all ALDH isozymes and therefore endogenous background non-specific fluorescence. Subsequently, we stained with a CD103-specific antibody (clone 2E7, Biolegend, CA, USA) and measured the frequency of CD103+ ALDEFLUOR+ cells by flow cytometry. True ALDEFLUOR fluorescence was taken as the measurement in the ALDEFLUOR reagent-treated cells minus the measurement in the DEAB-treated cells.

BM-DC and AS-ODN BM-DC accumulation inside the MLNs following i.p. injection was measured post-administration of the cells pulsed *in vitro* with fluorescent nanoparticles (Fluospheres; Thermo Fisher). Cells were injected within 5 h of confirmed nanoparticle uptake. 3–72 h later, the MLNs were harvested and single cells were stained with fluorescence-tagged CD45 (clone 30-F11, BD Biosciences) and CD11c (clone Rea754, Miltenyi Biotec) antibodies. The percentage of fluorescent nanoparticle+ cells inside a CD45+ CD11c+ gate was considered to represent the number of exogenously administered BM-DC or AS-ODN BM-DC that accumulated into the tissue.

Prior to adoptive transfer into mice, CD40, CD80, and CD86 surface levels on BM-DC and AS-ODN BM-DC were measured using the following antibody clones directly conjugated with non-overlapping excitation/emission fluors: CD40 (clone 3/23), CD80 (clone 16-10A1), and CD86 (clone GL1). These antibodies were purchased from BD Biosciences (San Jose, CA, USA) and titered before use.

### Histology/Immunocytochemistry

The colons of mice were cut into proximal, middle, and distal segments. After fixation in 4% paraformaldehyde (Sigma-Aldrich, MO, USA) for 3–4 h, tissues were transferred to 30% sucrose (Sigma-Aldrich, MO, USA) overnight, and then embedded in Tissue-Tek OCT (Fisher Chemicals, NJ, USA). 10-µm frozen sections were cut. For H&E staining, frozen sections were dried at room temperature, and staining was then conducted with a commercially available kit (Frozen Section Staining Kit; Thermo Fisher Scientific, NJ, USA). For H&E-based inflammation assessment, each colon segment was scored individually, and these scores were summed to reach a total score for the entire colon. Histological scores were assigned as follows: 0, normal; 1, ulcer or cell infiltration limited to the mucosa; 2, ulcer or limited cell infiltration in the submucosa; 3, focal ulcer involving all layers of the colon; 4, multiple lesions involving all layers of the colon, or necrotizing ulcer larger than 3 mm in length. Thus, the total possible histologic score is 12. Scoring was performed by a pathologist blinded to the treatment of the mouse.

### Detection of Human IL-37 in DC Culture *In Vitro*

Two populations of DC were generated from freshly obtained PBMC of a healthy volunteer as described previously (16). One population of DC was generated in the presence of GM-CSF/IL-4 and served as a control cell population. The other was generated in the presence of GM-CSF/IL-4 (which we term conventional PBMC DC; CP-DC) and a mixture of antisense DNA oligonucleotides targeting the primary transcripts of CD40, CD80, and CD86, which we term tolerogenic human DC (TH-DC). These two DC populations were used in a phase I clinical trial in established type 1 diabetic patients and shown to increase the frequency of human Bregs *in vivo* and *in vitro* (2, 16) *via* RA production (16). 1 × 105 CP-DC or TH-DC were cultured for 18 h in the presence or absence of 2 μg/mL LPS. The culture supernatants were collected and IL-37 was detected by a human-specific ELISA (R&D Systems, catalog # DY1975). The concentration of the cytokine in cell-free serum-containing medium was taken to represent control.

### Statistical Analyses

Two-tailed *t*-tests were used to determine the statistical relevance of the differences in the means of *in vitro* outcomes where replicates were considered (e.g., replicate cell culture wells in multi-well plates). When comparing the differences between two groups of mice, one-tailed ANOVA with Dunnett's *post hoc* test was conducted or repeated-measures Kruskal–Wallis test, depending on the experimental objective. Differences in the colitis score in the colons of different groups of mice was determined by one-way MANOVA.

A *p* value of <0.05 was considered to indicate statistical relevance to the differences in the outcomes in all statistical tests listed above.

## RESULTS

### BM-DC and AS-ODN BM-DC Prevent DSS-Induced Colitis

In **Figure 1**, we show the median weights and the range (error bars) of the mice in each of the three DSS treatment groups (no DC, BM-DC, and AS-ODN BM-DC). These observations were consistent among the two treatment cohorts which represented two independently conducted experiments. Those mice that were not treated exhibited significant weight loss and typical symptoms associated with DSS colitis (evidence of blood in feces as well as anal prolapse). By contrast, the AS-ODN BM-DC and BM-DC treatments were effective in significantly preventing weight loss. There was no statistically distinguishable difference in the outcomes in mice treated with AS-ODN BM-DC or BM-DC. We did not observe blood in stools in the DC-treated mice.

### Increased Frequency of Foxp3**+** Tregs in Colitis-Free DC Recipients

Given the evidence that tolerogenic DC promote the differentiation of T-cells into Foxp3+ Tregs while preventing conversion of gut T-cells into effector TH17-type cells (48, 74, 75), we hypothesized that the beneficial outcomes of the AS-ODN BM-DC treatment in the DSS-exposed mice could be associated with increased Foxp3+ Treg in the MLN and possibly other lymphoid organs into which the exogenously injected DC could potentially accumulate. In **Figure 2**, we demonstrate that Foxp3+ Tregs are increased in frequency as a % of total cells in the MLN. The analysis shown in **Figure 2** was conducted on cells obtained from tissue collected 5 days following DC administration. The increase in cell number was evident as early as 3 days following DC administration (data not shown). Similar results were obtained when measuring the frequency of Tregs in spleen from identically treated mice (**Figure 2B**). There were no apparent differences in frequency of Tregs in the analyzed tissues between mice treated with AS-ODN BM-DC or BM-DC.

## Increased Frequency of B10 Bregs in Colitis-Free DC Recipients

Accumulating data indicate that B-cells can act in a suppressive manner and a number of these B-cells, although with some differences in phenotype (76, 77), can transfer protection and improve experimental arthritis, lupus, and colitis in mice (78–80). We have presented evidence that immature DC, including our AS-ODN BM-DC, directly increase the prevalence of

or PBS vehicle 3 days before DSS exposure followed by a second DC (or PBS vehicle) injection 3 days following DSS exposure. Each graph shows the outcome in three groups of four mice. The bars represent the SD of *n* = 4 mice in each treatment group. Two mouse cohorts independently treated are shown. At each time point, represented in the graphs by an asterisk, the difference in weights between the BM-DC/AS-ODN BM-DC treatment arms, and the control mouse arm was statistically significant (determined by one-tailed ANOVA with Dunnett's *post hoc* test; *p* < 0.01 in study cohort 1, top graph panel, and *p* < 0.05 in study cohort 2, bottom graph panel). BM-DC indicates treatment of mice with GM-CSF and IL-4-generated cells from bone marrow progenitors and AS-ODN BM-DC indicates treatment with BM-DC generated in the presence of GM-CSF/IL-4 with the antisense DNA oligonculeotides.

the "B10" Breg population (79, 80) *in vitro* and *in vivo* (16, 17). We measured the frequency of B10 Bregs in the MLN and the spleen of mice pre-treated with BM-DC and AS-ODN BM-DC prior to DSS colitis induction. In **Figure 3**, we show that B10 Bregs increased in frequency as a % of total B-cells (% of CD19+ B220+ cells) in MLN but not in spleen (data not shown). In fact,

one-way ANOVA).

and MLN, the difference in the means between the BM-DC/AS-ODN BM-DC and control mice (DSS alone or untreated) were statistically significant (*p* < 0.01,

DC treatment had no effect on the frequency of B10 Bregs in spleen of any treatment group, including DSS induction on its own (data not shown). The analysis shown in **Figure 3** was conducted on cells obtained from tissue collected 5 days following DC administration. The increase in cell number was evident as early as 3 days following DC administration (data not shown). Even though there are no apparent differences in the frequency of Bregs in the tissues analyzed between AS-ODN BM-DC and BM-DC recipients, on a per-cell basis, the density of IL-10 in the AS-ODN BM-DC recipients was significantly greater than that in the BM-DC recipients (Figure S2 in Supplementary Material).

### BM-DC and AS-ODN BM-DC Accumulate Inside the MLNs After i.p. Injection

To confirm that BM-DC and AS-ODN BM-DC accumulate inside the MLNs of DSS-treated mice, we pulsed the DC with fluorescent nanoparticles *in vitro* (Fluospheres). Within 5 h of pulsing, a time when a maximal number of nanoparticles was phagocytosed by the DC, the cells were resuspended in sterile PBS and injected i.p. In Figure S3 in Supplementary Material, we show that BM-DC and AS-ODN BM-DC accumulated inside the MLNs as early as 3 h following administration (shown in figure). Accumulation was maximal by 3 days (data not shown). There were no statistically distinguishable differences in MLN-accumulated cells between BM-DC and AS-ODN BM-DC recipients.

### Increased Frequency of CD103**+** ALDEFLUOR**+** DC in Colitis-Free DC Recipients

Although BM-DC and AS-ODN BM-DC express ALDH and produce RA *in vitro* (16, 17), we hypothesized that exogenous administration of these DC could change the endogenous DC phenotype in the spleen and the MLN of treated mice. We therefore measured the frequency of total DC expressing ALDH (CD11c+ ALDEFLUOR+) as well as the frequency of CD103+ ALDEFLUOR+ cells as a function of total splenocytes or MLN single cells in DSS colitis mice treated with BM-DC or AS-ODN BM-DC. In **Figure 4**, we show that CD11c+ ALDEFLUOR+ cell frequency was significantly increased in mainly the AS-ODN BM-DC-recipients. The differences in CD103+ ALDEFLUOR+ cells between BM-DC or AS-ODN BM-DC and no DC recipients were statistically significant in the splenic population (bottom graph, **Figure 4B**). Although we observed similar differences in ALDEFLUOR+ CD11c+ cells in the MLN, we were unable to verify the presence of CD103+ cells that co-stained consistently with ALDEFLUOR in the MLN of these mice (data not shown).

### Colitis-Free DC Recipients Exhibit Inflammation-Attenuated Colon Architecture

H&E staining of representative sections of tissue from control, BM-DC, and AS-ODN BM-DC-treated mice suggested that DC significantly attenuated inflammation (**Figure 5A**). In **Figure 5B**, we summarize the scoring of inflammation in all treated mice.

### DISCUSSION

Many studies confirm the tolerogenic capacity of immature DC (81–83). Clinical applications of these DC have long been sought for transplantation tolerance and as a method to treat autoimmunity; however, the stability of the immature state *in vivo*, once the cells have been administered, has acted as a conceptual barrier to clinical translation. Our successful phase I clinical trial in established T1DM human volunteers with co-stimulation impaired, tolerogenic DC (2), together with the outcomes of more recent clinical safety trials using other variations of tolerogenic DC (84–86) should compel a reassessment of this barrier. In preclinical and ongoing studies in the NOD mouse model of T1DM, as well as a number of transgenic strains, we have discovered that DC (human and mouse) generated in the presence of antisense DNA oligonucleotides targeting the CD40, CD80, and CD86 primary transcripts increase the frequency of suppressive immune cells including Foxp3+ Tregs (66, 67, 69) and novel Bregs (16, 17). Based on the evidence demonstrating that RA-generated Tregs are therapeutic for IBD and that tolerogenic DC producing RA upregulate the number of Foxp3+ Tregs, we predicted that AS-ODN BM-DC (2, 67) could be beneficial in IBD as well. It is worth noting that BM-DC are inherently immature and immunosuppressive on their own. The rationale behind our approach to generate these DC in the presence of the antisense oligonucleotides targeting CD40, CD80, and CD86 was to ensure that these major co-stimulation proteins are stably knocked down *in vivo*. Unconditioned BM-DC, exogenously administered into an inflammatory environment such as that in T1DM, can rapidly differentiate into potent immunostimulatory cells characterized by high-level surface expression of CD40, CD80, and CD86 (87, 88).

In previous studies, we demonstrated that AS-ODN BM-DC stimulate the proliferation of existing Bregs concomitantly with the differentiation of B-cells into Bregs *in vivo* in NOD mice (17) and we also showed that the human counterpart to the AS-ODN BM-DC population generated from peripheral blood monocyte progenitors (TH-DC) also achieved the same outcome *in vitro*, in human co-cultures (16). Herein, we implicate B10 Bregs as responsive to BM-DC and AS-ODN BM-DC administration *in vivo* in the DSS colitis model of IBD. Although B-cells have been traditionally viewed as effector-type immune cells, mainly producing antibody and serving as accessory antigen-presenting cells, accumulating evidence supports their immunosuppressive ability. IL-10 production appears to be a defining feature of immunosuppressive B-cells. Two major populations of B-cells uniquely adapted to act as specific regulatory, immunosuppressive cells have been identified and characterized (89, 90). Even though IL-10 expression is the main feature of these Bregs, its production is not a *conditio sine qua non* for immunosuppression as we and others have reported elsewhere (91, 92) (and unpublished observations). Bregs, especially the B10 population suppress inflammation in experimental autoimmune encephalomyelitis, collagen-induced arthritis, and colitis (93–95). In a spontaneous model of murine colitis, the prevalence of B10 Bregs increases at the peak of inflammation and suppresses the disease by attenuating IL-1 and STAT3-mediated processes of immune reactivity (93). In another model of colitis, in TCRalpha-deficient transgenic mice, B-cell deficiency exacerbates disease and only CD40 ligand-activated B-cells can adoptively transfer protection and suppress the colitis inflammation (96). Although it is not yet clear where Bregs act to suppress the inflammation, evidence suggests that B-cells isolated from MLN are stable suppressors of colitis, even though splenic marginal zone B-cell exhibit a plasticity of suppressive ability when adoptively co-transferred with Gαi2-deficient CD3+ T-cells into Rag2-deficient mice (97). Interestingly, in murine models of colitis as well as in lupus, very few marginal zone splenic B-cells are found within the inflammation area further supporting a lymph node-source of suppressive B-cells. Our data are compatible with such a possibility where stably suppressive Bregs within the MLN are mobilized following their interaction with tolerogenic DC. Alternatively, endogenous, intralymphatic DC differentiate into tolerogenic DC upon encounter with the exogenously administered DC, in an RA-dependent manner. That Foxp3+ Tregs and B10 Bregs are increased in frequency coordinately inside the MLN following BM-DC and AS-ODN BM-DC administration (which produce RA) leads us to propose a model whereby DC are central in converting T-cells and B-cells into suppressive cells which then migrate into the inflamed colon structures to prevent or attenuate inflammation. This model is in the process of being tested in our laboratory.

Although the underlying mechanisms of B10 Breg and Foxp3+ Treg increase in the spleen and MLN are currently unclear in these experiments in the DSS colitis model, previous studies have outlined two non-mutually exclusive pathways concerning DC-stimulated increases in Treg numbers. DC can directly promote the proliferation of naturally occurring Tregs inside the lymph nodes (98–100). However, a second mechanism appears to be more common and this involves the conversion of resting naive T-cells that either do not express, or express low

levels of Foxp3, into suppressive Tregs (101). These adaptive, or induced, Tregs exhibit some plasticity in suppressive ability and depending on the presence or absence of cytokines like IL-10 or TGF-β, can revert to non-suppressive cells (101). RA and TGF-β coordinately provide a third mechanism, especially in IBD, effectively blocking the conversion of naive T-cells in the periphery into TH17-type cells and instead directing them into potently suppressive Foxp3+ Tregs (74, 75, 102). Accumulating data in our lab suggest that BM-DC and AS-ODN BM-DC (AS-ODN BM-DC > BM-DC) administration results in increased Foxp3 immunoreactivity through the entire colon tissue (unpublished observations). Together with the increased splenic and MLN complement of Foxp3+ Tregs, a significant tolerogenic state is established *in vivo* and this, along with the increase in B10 Bregs, could be a powerful suppressant of the most acute and damaging experimental model of colitis; DSS. Although it is not currently clear in the data from the study herein, how AS-ODN BM-DC compel an increase in the frequency of Bregs, our mechanistic studies in NOD mice could provide some insight (16, 17). There, AS-ODN BM-DC stimulate the proliferation of existing Bregs together with the differentiation of B-cells into Bregs. We propose that similar mechanisms underlie the AS-ODN BM-DC effect in the DSS colitis model, although these will have to be formally demonstrated in ongoing studies. If, and how the DC treatment affects less acute and less disruptive IBD models (e.g., adoptive transfer of CD4+ CD25− T-cells into SCID mice) remains to be determined. That AS-ODN BM-DC suppressed the severity of DSS colitis, which is mainly a non-T-cell-driven inflammatory syndrome, raises the intriguing possibility that tolerogenic DC can suppress the ability of innate immune cells to cause autologous tissue pathology and to even impede their ability to stimulate adaptive immune responses. This is not unprecedented (103, 104). Also, in an already-discovered and described mechanism, tolerogenic DC-stimulated Tregs can directly affect the

Figure 5 | AS-ODN BM-DC treatment preferentially preserves colon architecture with significant protection from inflammation in dextran sodium sulfate (DSS)-exposed mice. (A) H&E staining of colons resected from DSS-exposed mice treated with BM-dendritic cell (DC) or AS-ODN BM-DC. Representative sections are shown at two magnifications (×5 and ×20). Untreated, DSS-exposed mice exhibit inflammatory as well as significant tissue architecture disruption. Even though BM-DC treatment does not prevent inflammatory foci formation, the architecture of the tissue remains mostly intact. AS-ODN BM-DC treatment preserves colon architecture with evidence of significant protection from inflammation. (B) Colitis inflammation in resected colons of DSS-treated mice administered BM-DC or AS-ODN BM-DC was scored in a blinded manner. The bars in the graph represent the mean score of all colon sections (representing four mice per treatment arm) assessed. The error bars show the SEM. The differences in scores between the BM-DC/AS-ODN BM-DC and control (DSS colitis) mouse colons were statistically significant (*p* < 0.05, one-way MANOVA).

function of innate immune cells including in IBD (105–118). Another, more recently described mechanism could involve IL-37. We propose that AS-ODN BM-DC (and their human counterpart generated from peripheral blood monocytic progenitors) could suppress DSS colitis through either or all of the above mechanisms. For example, they can directly interfere with innate immune cell activity (e.g., ability to produce chemokines), they can stimulate the proliferation and differentiation of Tregs and Bregs that can then go on to impede innate immune cell activity, and they could even trigger the production of IL-37 in a proinflammatory environment. IL-37 will then act directly on innate immune cells, possibly *via* the IL-18 signaling pathway, to dampen their activity. This would affect the severity of both innate-driven inflammation and the impedance of triggering of the adaptive arm. This also is not without precedent. Luo et al. have shown that DC expressing IL-37 are tolerogenic (119) and subsequent to that discovery, Dinarello and colleagues demonstrated the suppressive effects of IL-37 on innate inflammation (120). Indeed, in preliminary experiments, we have discovered that human DC generated from PBMC in the presence of GM-CSF/ IL-4 (CP-DC) as well as CP-DC generated with the addition of the mixture of the antisense DNA oligonucelotides targeting the primary transcripts of CD40, CD80, and CD86 (TH-DC; refer to Section "Materials and Methods") produce IL-37 (Figure S4 in Supplementary Material). Interestingly, when stimulated with LPS, CP-DC produced less IL-37 in culture, however, when TH-DC were stimulated with LPS the amount of IL-37 produced was slightly greater (albeit not statistically significant; Figure S4 in Supplementary Material). It is worth noting that IL-37 production *in vitro* and very likely *in vivo* may exhibit interindividual variation and initial observations suggest that this is so (BP, CE, MT, NG; manuscript in preparation). Whether such variation could be associated with autoimmunity is unknown. Since stimulation of T-cell differentiation into Tregs and/or proliferation of existing Tregs by AS-ODN BM-DC cannot account for the rapid effects we observe in the DSS colitis model, we hypothesize that the second mechanism of innate immune suppression potentially underlies our observations in the DSS colitis model. This possibility is currently under investigation.

Onji and colleagues have reported that carbonic anhydrase I-pulsed bone marrow-derived DC generated *in vitro* with GM-CSF/IL-10/TGF-β inhibited colitis progression *via* rebalancing the Foxp3+/TH17 T-cell ratio inside the MLN (59). Pedersen and colleagues have also shown that tolerogenic DC pulsed with fecal extract suppressed colitis development (51). Early data suggest that, in addition to Foxp3 Tregs, cecal bacterial extract-pulsed DC protect from experimental colitis by generating Tr1 Tregs (52, 121). Our findings agree with those of these investigators, with the benefit that our approach does not rely on antigen-pulsing or preconditioning of the DC with immunosuppressive cytokines. More importantly, our AS-ODN BM-DC have been successfully translated clinically (using leukapheresissourced monocytic progenitors) without any safety issues (2). Despite the promising data presented herein, we have not yet determined whether BM-DC or AS-ODN BM-DC can "reverse," or inhibit already-established colitis. This step along with validation of the approach in at least two other mouse models of colitis are important milestones before considering autologous tolerogenic DC therapy for IBD. Intriguingly, Badami and colleagues have demonstrated a significant reduction in suppressive Foxp3+ Treg frequency in duodenal biopsies of T1DM patients compared with healthy controls. Furthermore, the patients exhibited an impairment in peripheral blood-derived CD103+ DC to convert CD4+ CD25− T-cells into Foxp3+ Tregs, unlike DC from healthy normal controls (122). These data, along with the accumulating evidence demonstrating a central role for RA-producing DC in regulating Tregs and their frequency, and a significant association between T1DM and IBD, further compel a more incisive investigation into the RA-producing DC/Treg/ Breg axis in the gut and the consideration of stably suppressive tolerogenic DC therapy as a cell-based, personalized medicine approach toward attenuating or completely reversing colitis and possibly other IBDs.

Our data show that the measurable outcomes between BM-DC and AS-ODN BM-DC are not different, even though they are significantly different when compared with untreated mice. This is not surprising as BM-DC are well known to be inherently tolerogenic (123–127). The original work using GM-CSF/IL-4 to propagate a DC population from bone marrow progenitors, two decades ago, clearly established that BM-DC, as functionally immature cells in their ability to stimulate significant T-cell proliferation, induced donor-specific hyporesponsiveness to alloantigens in transplantation models (123–127) and also were able to prevent the onset of autoimmune disease, T1D in particular (128–131). Given that the experimental conditions in all these studies used animals in specific pathogen-free facilities, and well-controlled environmental conditions, the reality in natural environments is expected to be much different; whereas the probability of conversion of the BM-DC *in vivo* into immunostimulatory cells under a controlled environment is low, the reality in the wild would predict that the metastable state of BM-DC would be sensitive to stimuli that confer to them a powerful immunostimulatory capacity. For example, a pathogenic enteric infection is expected to convert exogenously administered BM-DC, which accumulate into the gastrointestinal lymph nodes, into proinflammatory DC. Instead, the AS-ODN BM-DC are designed specifically to be co-stimulation impaired, even in conditions that can stimulate a maturation process. The same rationale underlies other approaches to generate tolerogenic DC; to maintain the cells in a state where, even though they may migrate through, or accumulate inside an immunostimulatory environment, the *ex vivo* conditioning maintains at least one major feature that maintains the balance in favor of a tolerogenic state.

### ETHICS STATEMENT

All mice were maintained in a specific pathogen-free environment, and experiments were conducted in line with specific protocols approved by the Allegheny Health Network IACUC. The human DC were generated from commercially purchased blood (exempt from IRB requirements).

### AUTHOR CONTRIBUTIONS

CE, YG, and BP conducted the experimental work. MT critically reviewed the manuscript for important intellectual content. NG was responsible for the design, the interpretation, and the overall compilation of this manuscript.

### FUNDING

This work was supported by the Henry Hillman Endowed Chair, and the John Rangos Endowment (to MT) in his previous position at the University of Pittsburgh, and by grants to YG from the Juvenile Diabetes Research Foundation (17-2012-348) and the National Institutes of Health (AI 124783-01).

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00894/ full#supplementary-material.

Figure S1 | Verification of AS-ODN BM-dendritic cell (DC) functionality *in vitro*. (A) Day 6 culture AS-ODN BM-DC (referred to also as AS-ODN DC in the graph) express significantly less surface CD40, CD80, and CD86 compared with paired BM-DC (cells generated from the same batch of bone marrow progenitors) measured by flow cytometry. The bars show the mean fluorescence intensity (MFI) corresponding to the levels of each of the indicated co-stimulatory proteins. The error bars show the SEM of triplicate measurements in 1 × 104 cells. The differences between the means are statistically significant when analyzed by two-tailed *t*-test. These outcomes are representative of the results obtained when characterizing the DC before administration in mice. (B) AS-ODN BM-DC do not stimulate IFNγ production in allogeneic mixed leukocyte culture compared to BM-DC *in vitro*. Two replicate co-cultures are shown (labeled 1 and 2); BM-DC (row of wells on the left side) and AS-ODN BM-DC (row of wells on the right side). These outcomes are representative of the results obtained when characterizing the DC prior to administration in mice.

Figure S2 | Regulatory B-cells (Bregs) retrieved from the mesenteric lymph node (MLN) of AS-ODN BM-dendritic cell (DC) recipients exhibit increased levels of IL-10 protein on a per-cell basis. The graph represents the geometric mean fluorescence intensity of the flow cytometry-measured events shown in the histograms in the panels in the middle of Figure 3A. These histograms correspond to IL-10 in permeabilized Bregs obtained from the MLN of PBS-injected mice alone, or mice that were dextran sodium sulfate (DSS)-treated alone, or DSS+ BM-DC or DSS+ AS-ODN BM-DC (2 × 106 cells per mouse) i.p. The analysis for IL-10 events is performed in cells gated into CD11c− CD19+ B220+ populations. The error bars represent the median of four mouse recipients per treatment group. The differences between AS-ODN BM-DC and BM-DC recipients are statistically significant as shown in the graph (repeatedmeasures, Kruskal–Wallis test).

Figure S3 | BM-dendritic cell (DC) and AS-ODN BM-DC accumulate inside the mesenteric lymph node (MLN) of dextran sodium sulfate (DSS)-treated mice following i.p. injection. Flow cytometry analysis to measure the frequency of exogenously administered DC inside the MLN. The panels are representative of an analysis conducted on single cells from freshly collected MLN at 3-h post-injection. The graph represents the frequency (Fluosphere+ CD45+ CD11c+ gated cells as a percentage of total cells) of the DC measurable from the single cells of freshly collected MLN 3-h post-DC injection. The data in the graph are shown as medians in the frequency of cells from the MLN of individual mice (*n* = 3 per treatment group) together with the range. There are no statistically distinguishable differences in accumulation of BM-DC compared with AS-ODN BM-DC; however, the differences in DC accumulation between mice that were administered DC vs. those that were not was statistically relevant (one-way ANOVA).

Figure S4 | Human PBMC-derived dendritic cell (DC) generated in the presence of GM-CSF/IL-4 alone (CP-DC) as well as with a mixture of antisense DNA oligonucleotides targeting the primary transcripts of CD40, CD80, and CD86 [tolerogenic human DC (TH-DC)] produce IL-37 *in vitro*. A human-specific IL-37 ELISA (R&D Systems) was used to measure the concentration of the cytokine in the culture supernatants of 1 × 105 CP-DC and TH-DC DC that remained naive or were stimulated with 2 μg/mL LPS overnight (18 h). The graph shows the means of quadruplicate wells of supernatants collected 18 h following DC plating (naive cells) or 18 h following LPS stimulation. The error bars represent the SEM. The differences in IL-37 produced between naive and LPS-stimulated moDC are statistically significant (*p* < 0.02, one-way ANOVA). The differences in IL-37 production between naive TH-DC and LPS-stimulated TH-DC are not statistically significant, even though there is a trend suggestive of more cytokine production by LPS-stimulated iDC. IL-37 production between naive CP-DC and TH-DC are not-significant. IL-37 was not detected in serum-containing, cell-free medium (data not shown).

### REFERENCES


could effectively induce antigen-specific regulatory T cells in vivo. *Transplant Proc* (2013) 45(9):3408–13. doi:10.1016/j.transproceed.2013.07.056


that parallels mouse regulatory B10 cells. *Blood* (2011) 117(2):530–41. doi:10.1182/blood-2010-07-294249


T cell-dependent inflammatory responses. *Immunity* (2008) 28(5):639–50. doi:10.1016/j.immuni.2008.03.017


peripheral induction of antigen-specific CD4+ CD25+ Foxp3+ regulatory T cells. *Eur J Immunol* (2011) 41(5):1420–34. doi:10.1002/eji.201040930


and suppresses dextran sulfate sodium-induced colitis. *Mucosal Immunol* (2015) 8(6):1297–312. doi:10.1038/mi.2015.20


**Conflict of Interest Statement:** NG and MT hold equity in Diavacs Inc., which has licensed the intellectual property (all patents) related to the methods in the preparation of the tolerogenic DC (AS-ODN BM-DC) described herein, as well as the use of the AS-ODN BM-DC and the human counterpart in autoimmune diseases, including colitis and related inflammatory bowel disease. They also serve as members of the Scientific Advisory Body. Diavacs did not review nor did it contribute to any aspect of the research presented in, or the preparation/editing of this manuscript.

The reviewer CH and the handling Editor declared their shared affiliation. The reviewer CH declared a past co-authorship with one of the authors NG to the handling Editor.

*Copyright © 2018 Engman, Garciafigueroa, Phillips, Trucco and Giannoukakis. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Inmaculada Serrano† , Ana Luque† and Josep M. Aran\**

*Immune-Inflammatory Processes and Gene Therapeutics Group, Institut d'Investigació Biomèdica de Bellvitge (IDIBELL), L'Hospitalet de Llobregat, Barcelona, Spain*

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Philippe Blancou, University of Nice Sophia Antipolis, France Muriel Moser, Free University of Brussels, Belgium*

> *\*Correspondence: Josep M. Aran jaran@idibell.cat*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 20 December 2017 Accepted: 10 April 2018 Published: 30 April 2018*

#### *Citation:*

*Serrano I, Luque A and Aran JM (2018) Exploring the Immunomodulatory Moonlighting Activities of Acute Phase Proteins for Tolerogenic Dendritic Cell Generation. Front. Immunol. 9:892. doi: 10.3389/fimmu.2018.00892*

The acute phase response is generated by an overwhelming immune-inflammatory process against infection or tissue damage, and represents the initial response of the organism in an attempt to return to homeostasis. It is mediated by acute phase proteins (APPs), an assortment of highly conserved plasma reactants of seemingly different functions that, however, share a common protective role from injury. Recent studies have suggested a crosstalk between several APPs and the mononuclear phagocyte system (MPS) in the resolution of inflammation, to restore tissue integrity and function. In fact, monocyte-derived dendritic cells (Mo-DCs), an integral component of the MPS, play a fundamental role both in the regulation of antigen-specific adaptive responses and in the development of immunologic memory and tolerance, particularly in inflammatory settings. Due to their high plasticity, Mo-DCs can be modeled *in vitro* toward a tolerogenic phenotype for the treatment of aberrant immune-inflammatory conditions such as autoimmune diseases and allotransplantation, with the phenotypic outcome of these cells depending on the immunomodulatory agent employed. Yet, recent immunotherapy trials have emphasized the drawbacks and challenges facing tolerogenic Mo-DC generation for clinical use, such as reduced therapeutic efficacy and limited *in vivo* stability of the tolerogenic activity. In this review, we will underline the potential relevance and advantages of APPs for tolerogenic DC production with respect to currently employed immunomodulatory/immunosuppressant compounds. A further understanding of the mechanisms of action underlying the moonlighting immunomodulatory activities exhibited by several APPs over DCs could lead to more efficacious, safe, and stable protocols for precision tolerogenic immunotherapy.

Keywords: acute phase proteins, inflammation, monocyte-derived dendritic cells, tolerance, immunotherapy

### INTRODUCTION

In the superior organisms, inflammation is considered as an evolutionarily conserved, physiological response of the vascularized tissue against external physical, chemical, and biological insults, or internal threats such as metabolic stress. This complex, exquisitely fine-tuned and coordinated process is engaged with the final goal of restoring the homeostasis and repair/regenerate the damaged tissues in a relatively short-time window (1). Whether the insult persists, chronic undesirable inflammation ensues and is associated with a variety of pathologies such as autoimmune processes or vascular diseases. Innate immune cells with the capacity for antigen presentation, that is, specialized antigen-presenting cells (APCs) such as monocytes/ macrophages and dendritic cells (DCs), are key players in all phases of inflammation (2). Thus, APCs are involved in the initial sensing of noxious agents through recognition of dangerassociated molecular patterns (DAMPs), including pathogenassociated molecular patterns (PAMPs), in the amplification of the defense/protection by locally attracting other immune cells through the vasculature and, finally, are essential effector cells in the resolution of inflammation. All these events are orchestrated mainly by DCs, endowed with high plasticity to bridge innate and acquired immune responses within the inflammatory program (3, 4). Local DAMPs/PAMPs detection by pattern recognition molecules (PRMs), notably the toll-like receptor (TLR) family of proteins, in these cells initiates an adaptive immune process leading to the activation and expansion of antigen-specific effector T lymphocytes in the secondary lymphoid organs (5). Conversely, the absence of pro-inflammatory stimuli or engagement of particular immunoreceptors, such as co-inhibitory receptors (PD-L1, PD-L2, B7-H3, ILT3, etc.) or other tyrosine-based inhibitory motif-containing receptors by a variety of signals maintain DCs in an "immature-like" state. These, "immature" DCs are able to elicit generalized or antigen-specific unresponsiveness/tolerance in central lymphoid organs or in the periphery, promoting the further stimulation of T cells (Treg) able to regulate or suppress other T cells (6). Such actions are crucial to maintain or return to immune homeostasis and to prevent autoimmune responses.

Another inherent aspect of the innate immunity elicited in wounded hosts (particularly those severely injured by trauma or microbial infection), in parallel to the advent of the above-described cellular or acquired immune response, is the prompt occurrence of a prominent non-specific immune-inflammatory response involving systemic physiological and metabolic alterations and affecting tissues/organs distant to the injured site, namely, the acute phase response (7). Thus, immunological stress induces a pro-inflammatory cytokine "storm," diffusing into the circulation and alerting the liver, which in turn reinforces a protective response through coordinated, cytokine-driven transcriptional changes in hepatocytes, leading to the secretion of a variety of molecules that limit tissue injury and participate in host defense, termed acute phase proteins (APPs), such as the prototypical C-reactive protein (CRP), serum amyloid P (SAP), and serum amyloid A (SAA). These proteins have been traditionally explored as diagnostic/ prognostic biomarkers reflecting the presence and intensity of inflammation during infection or injury. Indeed, while most APPs have been traditionally viewed as having a pro-inflammatory function, for example, in immune cell recruitment for efficient pathogen clearance (8), more recent studies are suggesting that a variety of APPs, depending on the microenvironment and through molecular mechanisms not yet completely understood, are able to interact directly with mononuclear phagocytes inducing a regulatory phenotype to these cells.

Mirroring the recent success and increasing importance of cellular immunotherapy strategies for cancer, in the last years a substantial effort has been devoted to generate DCs from blood precursors with tolerogenic features for the treatment of autoimmune diseases, allergy, and transplantation. As the first phase I adoptive tolerogenic DC therapy clinical trials are being concluded, preliminary lessons learned include the overall safety of tolerogenic DC administration, although also highlight present limitations regarding its efficacy. Thus, important current challenges to overcome for a more effective therapeutic outcome include the achievement of antigen-specific tolerogenic responses and, particularly, the maintenance of a "stable" tolerogenic phenotype of the infused DCs regardless of the inflammatory microenvironment that they may confront. Therefore, more progress has to be achieved on the thorough characterization, using both in *in vitro* functional readouts and preclinical assays, of tolerogenic DCs generated through alternative immunomodulatory inducers able to increase their clinical performance in immune-inflammatory pathologies.

In this review, we will consider the potential of APPs as novel immunomodulators. We will overview the current knowledge regarding the interaction of relevant APPs with phagocytes, fundamentally monocytes, and monocyte-derived DCs (Mo-DCs), resulting in a bias toward immune tolerance. A better understanding of the crosstalk between the innate and the adaptive immune systems in homeostasis and inflammatory pathology, taking into account the unique roles of both APPs and DCs, may support therapeutic benefits of APP-induced tolerogenic DCs for transplantation and autoimmunity.

### THE ACUTE PHASE RESPONSE AT THE CROSSROADS BETWEEN INNATE AND ADAPTIVE IMMUNITY

The immediate innate body defense against acute illnesses, that is, the acute phase response, features both, hepatic and extra-hepatic overproduction and release, typically within 24–48 h after the initial insult, of a variety of seemingly biochemically and functionally unrelated APPs into the circulation. In fact, phagocyte sentinels (macrophages, DCs, and neutrophils) sensing eminently damaged, stressed or infected cells, elicit a local pro-inflammatory response, and seek further help by secreting pro-inflammatory cytokines such as IL-6, IL-1, IL-8, TNF-α, and IFN-γ, and releasing a large assortment of "alarmins." These key mediators travel through the circulation, induce neuroendocrine and behavioral changes (fever, hyponatremia, anorexia, somnolence, and lethargy), and reach the liver, whose most abundant cell type, the hepatocytes, hold also the capability to act as immunological agents and have a central role in the systemic innate immune response through the intravascular secretion of APPs (9). Indeed, APPs conform up to 40 different proteins whose serum concentration increase (positive APPs) or decrease (negative APPs) at least 25% in response to inflammation (10). Positive APPs include soluble PRMs [CRP, SAP, SAA, lipopolysaccharide binding protein, complement components, and α1 acid glycoprotein (AAG)], hemostasis factors (fibrinogen, plasminogen, prothrombin, and plasminogen activators), binding/transport proteins [haptoglobin (Hp), hemopexin, and ceruloplasmin], and antiproteases [α1-antichymotrypsin (AAC), antithrombin (AT), α1-antitrypsin (AAT), and α2-macroglobulin (α2M)]. These proteins participate in host defense (e.g., attracting inflammatory cells, inactivating proteolytic enzymes, activating complement, opsonizing, and clearing infectious agents) and limit tissue injury (scavenging free radicals and modulating the host's immune response). Conversely, negative APPs comprise albumin, AT, transferrin, transthyretin, transcortin, and retinolbinding protein (8). It has been suggested that reduced albumin production enhances the amino acids "pool" available for positive APP production, and that decreased transferrin production could protect the host by starving microorganisms of the iron required for growth and virulence expression (11).

Based on their degree of response to inflammatory stimuli, APPs can be grouped as strong (more than 100-fold increase in blood levels; CRP, α2M, SAA), moderate (2–10-fold increase; haptoglobulin, fibrinogen, AAT), or weak (up to twofold increase; C3, ceruloplasmin). While strong APPs usually increase abruptly within the first 24–48 h after an acute inflammatory event, and further experience a quick decline related to their relatively short half-life, moderate to weak APPs are more likely present during chronic inflammatory processes. According to the differential regulation of their synthesis by cytokines, positive APPs can also be classified in type I and type II. Type I are induced by IL-1-like pro-inflammatory cytokines (SAA, CRP, C3, AAG, and SAP), and type II are induced by IL-6-like cytokines (fibrinogen, Hp, AAC, AAT, and α2M). In turn, the production of hepatic APPs may also be influenced by other cytokines and by hormones (insulin, dexamethasone, glucagon, and/or epinephrine) (12). Thus, at the level of the organism, the complex neuroendocrine-immunological axis seems to efficiently modulate the acute phase response through various feedback loops (13). For instance, cytokines released from monocytes/macrophages activated locally through noxious inflammatory agents stimulate the brain to release stressresponse neuropeptides such as corticotropin (ACTH), which acts into the adrenal glands inducing glucocorticoid production. Glucocorticoids can downregulate pro-inflammatory cytokines (IL-1, TNF-α).

Due to their stability in the circulation compared with cytokines, which are cleared from the circulation within a few hours, several APPs have been extensively used as diagnostic/prognostic biomarkers because their increased/decreased levels reflect the presence and intensity of inflammation during infection or injury, remaining unchanged for 48 h or longer. Nevertheless, although presenting high sensitivity, the diagnostic value of APPs is being questioned due to their low specificity (14).

### INFLAMMATORY DCs IN INFLAMMATION

Relevant features of the acute phase response are an increase in the number of peripheral leukocytes and the dilation and leakage of the vasculature through the release of inflammatory mediators such as reactive oxygen species, arachidonate metabolites, and pro-inflammatory cytokines and chemokines (15). Pro-inflammatory cytokines activate and mobilize blood cell precursors in both bone marrow and peripheral blood (16–18). Moreover, stimulated endothelial cells allow the extravasation and migration of circulating leukocytes. Among these, Mo-DCs have been appealing due to: (1) their influence on adaptive immune function and rapid accumulation in the inflammatory focus and (2) their easy *ex vivo* isolation, amplification, and manipulation. Mo-DCs arise from monocyte precursors both *in vitro* and *in vivo* (19, 20). Monocytes are recruited to sites of inflammation, having a major role in the protective immune response of the host (21). For instance, local differentiation of monocytes into inflammatory macrophages and DCs is induced in response to natural killer cell-produced IFN-γ (22). In fact, by depletion of tissue-resident cell populations it has been shown that circulating monocyte precursors in the blood can replenish functionally specialized macrophages and DCs (23), which reinforces the concept of blood monocytes as reservoirs that can be utilized on demand, particularly in inflammatory processes where monocyte recruitment is strongly increased. Accordingly, monocytes have been shown to migrate to inflammatory sites and differentiate into DCs in various murine models of inflammation (24, 25). Sequential trafficking and/or differentiation of the different monocyte subsets to the sites of inflammation is likely modulated by diverse mechanisms (26–28). Following tissue damage, classical monocytes (human: CD14++CD16<sup>−</sup>; mouse: Ly6C<sup>+</sup>CCR2highCX3CR1low) appear to be recruited within the first few hours, after their egression from the bone marrow being modulated by the CCR2–CCL2/CCL7 axis (29). Once in the inflammatory milieu, they differentiate into DCs and macrophages and exert a potent pro-inflammatory immune response through high-level production of IL-1β and TNF-α, among other protective functions (30–32). When the progression of the immune-inflammatory response is not halted, the prolonged action of classical inflammatory monocytes may result in tissue damage and drive autoimmunity (33). Several days after the initial damaging insult, acute inflammation enters in a resolution phase where the classical monocyte levels are reduced and progressively replaced by intermediate [CD14<sup>+</sup>(+) CD16<sup>+</sup>] and non-classical (human: CD14<sup>+</sup>CD16++; mouse: Ly6C<sup>−</sup>CCR2low CX3CR1high) monocytes, which relay on the CX3CR1–CX3CL1 axis to accumulate in the damaged tissue and, after DC/macrophage differentiation, secrete anti-inflammatory cytokines (IL-10, TGF-β) that counteract tissue injury and promote wound healing (34). Certainly, it has been suggested that, in response to inflammatory stimuli, patrolling non-classical CD16-expressing monocytes could leave the blood vessels and function as DC precursors (35). Thus, these inflammatory Mo-DCs seem to hold unique features influenced by the microenvironmental status of the inflamed tissue, boosting more potent immune responses DCs derived from classical monocytes, and better immune tolerance DCs generated from non-classical monocytes (36).

Monocytes from human or mouse peripheral blood or bone marrow are widely utilized to generate *in vitro* large amounts of Mo-DCs upon differentiation, typically with IL-4 and GM-CSF (37), allowing comprehensive mechanistic studies regarding their key role in the immune-inflammatory processes at the molecular level and to initiate DC therapy approaches in the clinic. In fact, a comparative transcriptional profiling has revealed that human DCs isolated from inflammatory fluids are the *in vivo* counterpart of *in vitro*-generated Mo-DCs from CD14<sup>+</sup> monocytes (38), in the same way that murine inflammatory DCs share equivalent developmental and functional features to *in vitro* GM-CSF/ IL-4-induced BM-DCs (39).

Monocyte-derived cells have been deemed essential for inducing protective Th1 cell-mediated immunity following both pathogen infection and non-infectious conditions (40, 41), and may acquire DC-specific functions such as cross-presentation (41, 42).

Conversely, DCs play a key role in tolerance, whether participating in the negative selection of autoreactive T cells in the thymus (central tolerance) (43), or limiting effector T cells through deletion or anergy and, instead, promoting Treg differentiation (peripheral tolerance). A variety of mechanisms are orchestrated by DCs to induce tolerance and suppress inflammatory responses against innocuous stimuli, including the overexpression of inhibitory immunoreceptors (e.g., PD-L1, B7H, and CD80/86), the ligand-activated transcription factor aryl hydrocarbon receptor, the pore-forming cytolytic protein perforin, and the release and/ or control of several immunomodulatory mediators, such as antiinflammatory cytokines (IL-10, IL-27, and TGF-β), indoleamine 2,3-dioxygenase (IDO) metabolites, retinoic acid, vitamins A and D, ATP, and adenosine [see Ref. (44, 45), and references therein]. The regulatory function of DCs is determined by their maturation/activation status (46). Hence, tolerogenic DCs hold an "immature" or "semi-mature" state.

A myriad of recent studies has reported the *in vitro* generation of monocyte-derived "permissive," "tolerogenic," "regulatory," "alternatively activated," or "maturation-resistant" cell types (47), although most attention has been focused on Mo-DCs. This is being achieved by incubation with a variety of different biological or pharmacological agents such as cytokines (IL-10, TNF-α, IFN-γ, TGF-β, IL-21, and thymic stromal lymphopoietin), immunosuppressant drugs (dexamethasone, tacrolimus, and mycophenolate), organic molecules (vitamin D3, salycilate, vasoactive intestinal peptide, intravenous immunoglobulin, and hepatocyte growth factor), other agents (pathogen products, mesenchymal stem cells), or their combinations, or by genetic engineering (48–50). Mimicking the *in vivo* circumstances, the resulting tolerogenic Mo-DCs are characterized essentially by reduced surface expression of co-stimulatory molecules (CD80, CD86, and CD40) and maturation markers (CD83), increased expression of inhibitory receptors (ILT3, PD-L1, and PD-L2), reduced or null production of pro-inflammatory cytokines (IL-12, TNF-α, IFN-γ, and IL-8) and, conversely, increased production of anti-inflammatory cytokines (IL-10, TGF-β), even in the presence of inflammation (51–54). Thus, the main features of these cells would be to present a state of unresponsiveness through hampering key activation/ maturation pathways such as the pro-inflammatory NF-κB pathway, and to support the differentiation and maintenance of different types of Treg cells.

### TOLEROGENIC ACTIONS OF APPs ON DCs

There are clear evidences showing that the acute phase response can directly influence the differentiation of DCs toward a tolerogenic state. In sepsis, an overwhelming systemic inflammatory response syndrome, an expansion of intermediate monocytes has been detected in the circulation (55). Monocytes from sepsis patients preferentially differentiated into alternative CD1a<sup>−</sup> DCs, holding increased capacity to induce Foxp3<sup>+</sup> Treg cells, when compared with monocytes from healthy individuals in which classical monocytes predominated (56). On the other hand, the hepatic APPs SAA and Cxcl1/KC cooperatively promoted myeloid-derived suppressor cell (MDSC) mobilization, accumulation and survival, reversed dysregulated inflammation, and restored survival of mice deficient for gp130 (the signaling receptor shared by IL-6 family cytokines) undergoing polymicrobial sepsis (57). Thus, hepatocytes may also modulate innate immune cells through the acute phase response, for example, by recruitment and promotion of MDSC function.

Accordingly, it is not unreasonable to consider a number of APPs, acting either systemically or locally in a restricted time window coinciding with a parallel increase of monocyte recruitment toward the inflammatory focus, as a part of a protective network to restrain the harmful consequences of continued overinflammation. That is, APPs could directly exert a feedback loop redirecting the differentiation of these inflammatory monocytes to regulatory or tolerogenic DCs, in an attempt to regain homeostasis and maintain tissue integrity through the resolution of the immune-inflammatory response (**Figure 1**). We will now focus in representative APPs and APP-related proteins that are able to induce tolerogenesis through modulation of Mo-DC differentiation and/or maturation.

Soluble PRMs are a heterogeneous group of molecules (collectins, ficolins, pentraxins, and other complement components) belonging to the humoral arm of innate immunity that have been proposed to represent the functional ancestor of antibodies (58). They share basic functions with the membrane-bound PRMs from DCs, such as the recognition of "non-self " and "modified self " and, additionally, play an important role in opsonization and complement activation. In the last years, several studies have evidenced that APPs, particularly soluble PRMs, acting directly in the early stages of monocyte differentiation mediated by GM-CSF/IL-4 (a faithful *in vitro* model for the generation of inflammatory DCs), are able to confer a tolerogenic phenotype and function to the ensuing Mo-DCs, although the detailed molecular mechanisms of APPs action over DCs are still not known for most of them.

In the next paragraphs, we will address the state of understanding and arguments regarding APP-mediated tolerogenic DC generation and functional outcome, according to common features currently defining tolerogenic DCs.

### Pentraxins

Pentraxins constitute a superfamily of evolutionarily conserved multimeric and multifunctional proteins sharing an 8-amino acid "pentraxin domain" (HxCxS/TWxS, where "x" is any amino acid) in their carboxy terminus. Based on the primary structure of the promoter, pentraxins are divided into short pentraxins (CRP and SAP) and long pentraxins (PTX3) (58). Both CRP and SAP are homooligomeric proteins arranged in a ~25 kDa subunit pentameric radial symmetry and hold 51% amino acid sequence identity. They constitute the main APPs in human and mouse, respectively, are produced by hepatocytes and have wide capacity for pathogen recognition, phagocytosis, and cytokine secretion through interaction with Fcγ receptors (59). Moreover, CRP and SAP are able to regulate the activation of the complement system by interaction with C1q, ficolins, C4b-binding protein (C4BP) and factor H, favoring efferocytosis and preventing the onset of autoimmune diseases (60, 61).

Strong inflammatory stimuli (infection, lupus nephritis, etc.) trigger the presence of inflammatory monocytes in the blood, which are actively recruited to inflamed tissues, and differentiated to inflammatory DCs, having the ability to stimulate naïve T cells. Under these conditions, besides its function as complement inhibitor, the increased presence of C4BP(β−) in the blood would act in one or both ways upon engaging one or some, as yet unknown, cell surface receptor(s): 1) reducing transendothelial migration and accumulation of the inflammatory monocytes into the inflamed tissue and 2) inducing a tolerogenic phenotype in the recruited inflammatory DCs, which would led to: (a) inhibition of T cell proliferation and differentiation into Th1, Th2, and/or Th17 cells depending on the inflammatory microenvironment, (b) decreased pro-inflammatory cytokine secretion (IL-12, TNF-α, IFN-γ, etc.), (c) reduced migration to the lymph nodes, and conversely, to: induction of anti-inflammatory cytokine release (IL-10, TGF-β, etc.), and (d) Treg generation within the inflamed tissues (3).

C-reactive protein has been shown to transform biological functions of Mo-DCs toward a tolerogenic phenotype. Interestingly, when CRP was added at the early stage of Mo-DC differentiation from CD14<sup>+</sup> monocytes, it downregulated surface expression of DC-SIGN and the antigen uptake molecules CD205 and CD206, resulting in reduced endocytosis capacity (62, 63). Moreover, LPS-mediated Mo-DC maturation was also impaired, through downregulation of co-stimulatory molecules CD80 and CD86, and of the maturation marker CD83, inhibition of allogeneic T cell proliferation and decreased production of pro-inflammatory cytokines (IL-12, IL-8, IL-6, TNF-α, MIP-1α, MIP-1β, and MCP-1). These effects seemed to be mediated through the immunoreceptor FcγRII/CD32, which is downregulated during differentiation into Mo-DCs. Conversely, another study reported just the opposite, that is, CRP was able to activate Mo-DCs through upregulation of DC activation markers (CD40, CD80, CD83, and CCR7) and induced allogeneic T cell proliferation and IFN-γ production (64). Nevertheless, in that case the pulsation of Mo-DCs was started at day 6 of culture, once the Mo-DCs were fully differentiated. These results evidence the restricted tolerogenic activity window characterizing CRP, at the initial steps of Mo-DC differentiation. Analogously, human SAP has been reported to bind strongly to monocytes but weakly to differentiated Mo-DCs (65). SAP also inhibits neutrophil recruitment and monocyte to fibrocyte differentiation, in part, by binding to the FcγRs (66, 67), and polarizes macrophages toward an immunoregulatory phenotype through PI3K/Akt-ERK signaling (68). Thus, SAP regulates key components of the innate immune system and inflammation.

Pentraxins is a multimeric 340 kDa glycoprotein with a complex quaternary structure (elongated, with a large and a small domain interconnected by a stalk region) composed of two tetramers linked by interchain bridges to form an octamer. PTX3 expression is induced in a variety of cell types (particularly in phagocytes) by inflammatory cytokines, TLR agonists or pathogens, binds to a wide range of microorganisms, and plays a relevant role in host defense and inflammation (69), for example, by regulating leukocyte recruitment (70). Moreover, analogously to CRP and SAP, PTX3 is also able to modulate the activation of the complement system by binding C1q, ficolins, mannose-binding lectin (MBL), and the complement regulators C4BP and FH, and increases phagocytosis in an FcγRII-dependent manner. Hence, PTX3 binds to apoptotic cells and recruits C4BP, limiting complement activation and an exacerbated inflammatory response (71). In this context, PTX3 reduces the release of TNF-α and IL-10 by LPSchallenged Mo-DCs, and consistently inhibits the upregulation of membrane molecules (CD86, HLA-ABC, HLA-DR) on an inflammatory cell surface induced by LPS. Moreover, PTX3 also induces macrophages to secrete anti-inflammatory cytokines such as TGF-β and IL-10 (72), modulates LPS-induced inflammatory response and attenuates liver injury (73).

### Complement Components

The evolutionarily conserved complement system, in addition to its crucial function in the innate defense against common pathogens, holds also a key regulatory non-immunogenic role in the "silent" clearance of immune complexes from the circulation and apoptotic cells from damaged tissues, in close crosstalk with the mononuclear phagocyte system (74). We have recently discussed the "non-canonical" activities of a variety of complement effectors and modulators able to transform DCs toward a tolerogenic phenotype (75). Thus, we will instead focus here on the functional outcome of a few representative complement components directly interacting with Mo-DCs.

In addition to their central role as complement cascade initiators for microbial phagocytosis and killing, it is becoming evident that both, complement cascade initiators such as mannosebinding lectin (MBL) and soluble complement inhibitors such as C4BP, are able to promote an immunomodulatory and antiinflammatory environment by direct interaction with DCs and other immune cells.

MBL, the prototypic initiator of the lectin pathway of complement activation, belongs to the collectin family and, through its carbohydrate-recognition domains, is able to bind to oligosaccharides (mannose, *N*-acetyl-glucosamine) on the pathogen surface (76). DCs from MBL-deficient individuals showed increased IL-6 production and poor allogeneic T cell responses, features of pathogen-stimulated DCs, which could be reversed by *in vitro* addition of MBL (77). In fact MBL, at supraphysiological concentrations, influences the phenotype and function of DCs by attenuating LPS binding to immature DCs and their further maturation and pro-inflammatory cytokine production (IL-12, TNF-α), while preventing allogeneic T lymphocyte proliferation (78). Moreover, MBL not only attenuates LPS-induced Mo-DC maturation, but also affects early Mo-DC differentiation from CD14<sup>+</sup> monocytes, yielding Mo-DCs with tolerogenic features (low MHC-II, CD80 and CD40 expression, increased IL-10 and IL-6 secretion, and reduced T cell alloproliferation), and being possibly mediated by members of the STAT family (79).

Among the complement inhibitors, the regulator of the classical and lectin pathways of complement activation C4BP has a complex oligomeric structure. The major C4BP isoform, C4BP(β+), has an heterooligomeric radial structure (570 kDa). It is composed by seven identical 70 kDa modular α-chains (responsible for the complement inhibitory activity, and for pentraxin, heparin, DNA, and pathogen binding, among others), and a single 40 kDa β-chain (high-affinity binding site for anticoagulant vitamin K-dependent Protein S, allowing a strong interaction with apoptotic/necrotic cells) (80, 81). The minor C4BP isoform, C4BP α7β0 or C4BP(β−), holds the same oligomeric structure and complement inhibitory function than C4BP(β+), but lacks the β-chain. Under acute phase conditions (poly-traumatisms, sepsis) the levels of circulating C4BP(β−) isoform increase significantly as a consequence of the differential hepatic regulation of the α- and β-chains by pro-inflammatory cytokines (82). Thus, C4BP(β−) is a genuine APP. We have shown that the C4BP(β−) isoform, but not the C4BP(β+) isoform, by direct interaction with Mo-DCs through as yet unknown receptor(s), only in the early stages of monocyte to Mo-DC differentiation, is able to confer an anti-inflammatory, tolerogenic phenotype to these cells, retaining a high-endocytic activity, and morphological features of immaturity. Upon LPS priming, these C4BP(β−)-treated Mo-DCs featured low-surface expression of CD83, CD80, and CD86, inhibition of pro-inflammatory IL-12, TNF-α, IFN-γ, IL-6, and IL-8 production and, instead, increased expression of anti-inflammatory IL-10 and TGF-β, reduced CCR7 expression and chemotaxis, and promoted Treg expansion. Moreover, C4BP(β−) induced tolerogenic DCs with increased viability and yield when compared with the immunomodulator vitamin D3, and similarly prevented T cell alloproliferation (83).

Although perhaps not a *bona fide* APP, C1q, the recognition unit from the classical pathway of complement activation and major component of the C1 complex, binds to various APPs including CRP, SAP, and PTX3, thereby regulating the classical complement pathway. C1q has also been recognized to modulate cellular functions within the adaptive immune response (84). Certainly, C1q has even been proposed as a tolerogenic DC marker because relevant immunomodulatory agents such as dexamethasone, IL-10, or vitamin D3 are able to induce at least a 10-fold overexpression of C1q at both mRNA and protein levels in Mo-DCs (85). The regulatory effects of C1q on monocyte/ DC precursors could be mediated by gC1qR, occurring within a narrow timeframe of monocyte to Mo-DC transition and being influenced by the microenvironment. Accordingly, while in the presence of danger signals C1q would recognize and bind antigens through its globular head domains, leading to activation of a pro-inflammatory immune response in immature Mo-DCs, in the absence of danger signals C1q would maintain immature Mo-DCs in a tolerance state through gC1qR (86). Certainly, gC1qR ligation on the surface of Mo-DCs suppresses TLR4-induced IL-12 production through PI3K pathway activation (87). Furthermore, an alternative mechanism of C1q-mediated immunomodulation involves high-affinity binding between C1q and the inhibitory immunoreceptor LAIR-1, which inhibits monocyte-to-Mo-DC differentiation (88). More recently, this interaction has been refined through the characterization of a tri-molecular engagement encompassing C1q-CD33/LAIR-1 crosslinking (89).

### Hemoglobin- and Iron-Binding Proteins

Essential cellular processes, such as energy generation, DNA replication, oxygen transport, and protection from oxidative stress are dependent on iron. Since bacterial pathogens also require iron for replication and infection, iron sequestration strategies from vertebrates constitutes a significant form of nutritional innate immunity (90). Thus, in homeostatic healthy conditions iron is largely intracellular and sequestered within ferritin. Conversely, acute inflammatory processes such as infection, include the release of lactoferrin from secondary granules contained within polymorphonuclear leukocytes. Furthermore, hemoglobin released by physiological and pathological hemolysis is captured by Haptoglobin (Hp). All together, these proteins ensure a virtually free iron environment in vertebrate tissues.

Ferritin is a major tissue iron-binding protein with a molecular weight of 500 kDa, whose main function is to store iron in a soluble non-toxic form, protecting the cell from iron-mediated redox reactions. The levels of this APP remain elevated in many chronic inflammatory diseases such as periodontitis (91). Ferritin is composed of 24 subunits consisting of heavy (H) and light (L) chains, and may heterooligomerize forming isoferritins depending on the proportions of H and L chains (92). The immunosuppressive effects of cancer cell supernatants, such as melanoma supernatants, correlated with their content of H-ferritin. Accordingly, H-ferritin has been shown to inhibit anti-CD3-stimulated lymphocyte proliferation, probably mediated by increased IL-10 production (93). Importantly, H-ferritin is also able to induce a semi-mature, tolerogenic phenotype on Mo-DCs featuring increased expression of CD86 (B7-2) and B7-H1, and the activation of IL-10-producing Treg cells (94).

Lactoferrin, also known as lactotransferrin, is produced in a number of tissues and is frequently found in mucosal secretions and neutrophil secretory granules (95). This 80 kDa iron-binding glycoprotein is an important component of innate immunity, and holds a key role in the protection of mucosal surfaces from microbial infections (96). Lactoferrin also modulates innate and adaptive immune-inflammatory responses, including cytokine production, promotion of T and B cell maturation, and enhancement of delayed-type hypersensitivity against defined antigens. Moreover, it has been suggested that lactoferrin might exert adjuvant activity, enhancing DC function to promote generation of antigen-specific T cells (97). In contrast, bovine lactoferrin (bLF) seems to play an opposite role. Thus, Mo-DCs differentiated in the presence of bLF showed a fully tolerogenic or immunomodulatory behavior [potent anti-inflammatory activity, high-endocytic capacity, increased expression of molecules with negative immunoregulatory functions (ILT3, PD-L1, IDO, and SOCS3), CCL1 production, and impaired capacity to undergo activation and to promote Th1 responses]. bLF is internalized and seems to reach the nucleus, although the molecular details mediating the bLFmediated transcriptional regulation of Mo-DC differentiation are still unknown (98, 99).

Hp is the major hemoglobin-binding protein in plasma. This APP, whose hepatic expression is induced by inflammatory mediators such as IL-6-type cytokines, interacts with free hemoglobin neutralizing and restricting its oxidative damage to various organs (100). Hp has been suggested to exert immunomodulatory effects constituent with suppression of lymphocyte function (101). During physiological and pathological hemolysis, the Hp-CD163-heme oxygenase (HO-1) pathway efficiently scavenges and circumvents hemoglobin/heme-induced toxicity. This pathway plays an anti-inflammatory role in phagocytes, and the resulting heme metabolites, such as bilirubin, reinforce its cytoprotective and anti-inflammatory efficacy (102). Hp seems also to prevent epidermal Langerhans cells from spontaneously undergoing functional maturation in the skin, inhibiting their capacity to activate autologous T cells *in vitro* (103).

### Other APPs and APP-Related Proteins

Serum amyloid A is an APP produced mainly by the hepatocytes, but also by other cell types such as macrophages, smooth muscle cells, chondrocytes, epithelial cells, and adipocytes, under pro-inflammatory stimuli (58). SAA interacts with Gramnegative bacteria and, through its opsonic activity, increases their phagocytosis and the production of TNF-α and IL-10 by phagocytes (104). SAA has also been recently shown to be involved in the expression of the "alarmin" IL-33 by monocytes and macrophages (105). Notably, it has been recently shown that SAA-stimulated monocytes (HLA-DRhi HVEMlo) most resemble immature Mo-DCs, and are able to drive Treg proliferation (106). SAA is also a chemoattractant for immature Mo-DCs through formyl peptide receptor like 1/formyl peptide receptor 2 (107). Furthermore, mice lacking SAA3, an acutely expressed isoform found in non-primate mammals, develop metabolic dysfunction, and exacerbated pro-inflammatory responses from innate immune cells. Particularly, bone marrow-derived DCs from SAA3(−/−) mice produce increased levels of IL-1β, IL-6, IL-23, and TNF-α in response to LPS compared with cells from wild-type mice (108). Thus, endogenous SAA3 likely modulates metabolic and immune homeostasis.

α1-Antitrypsin, a member of the SERPIN superfamily of protease inhibitors, is a major inhibitor of the neutrophil-derived serine proteases [neutrophil elastase (NE), cathepsin G, and proteinase 3]. It has a primary anti-inflammatory role by irreversible binding and inactivation of NE, protecting the lung against the destructive effects of NE released by degranulating neutrophils during inflammation (109). AAT is predominantly produced by the liver, and its secretion, increased under acute phase conditions, is mediated by pro-inflammatory cytokines (IL-6, IL-1β, and TNF-α) (12). Recent studies have reported tolerogenic activities of AAT that are difficult to explain solely by serine-protease inhibition or by its anti-inflammatory actions (110). Circulating AAT is bound to lipoprotein particles (LDL and HDL) and docks onto lipid-rafts. Thus, TLR2 and TLR4 contained in lipid-rafts from macrophages and DCs are downregulated by AAT (111). Moreover, AAT induces a tolerogenic phenotype on DCs characterized by low levels of CD40, CD86, and MHC class II, increased production of IL-10 and enhanced generation of Tregs through a so far unknown mechanism. These tolerogenic DCs maintain, nevertheless, the inflammation-driven cell migration capacity (112). In fact, AAT monotherapy has been shown to induce tolerance in islet allograft and kidney transplantation (113, 114), graft-versus-host disease (115), improved islet function in type I diabetes (116), and attenuated lupus nephritis (117).

Fibrinogen is synthesized mainly by hepatocytes, and its level increases substantially during infections and inflammatory conditions. This 340 kDa glycoprotein, made up of two identical subunits joined together by disulfide bonds, functions as a blood coagulation factor, supporting platelet aggregation, and fibrin cloth formation at the site of vessel injury. Fibrinogen had an Mo-DC maturation effect comparable with poly I:C, TNFα, and PGE2, but it failed to induce IL-12 production (118). On the other hand, it has been recently reported that fibrinogen cleavage products generated by protease allergens, through induction of IL-13 production by mast cells, increased the number of TH2-favorable (PD-L2<sup>+</sup>) DCs in allergic asthma (119). Interestingly, another member of the fibrinogen-related protein superfamily, soluble fibroleukin or fibrinogen-like protein 2 (sFGL2), highly inducible by IFN-γ and with features of APP, has a 50 kDa weight and is highly expressed in cytotoxic T cells and Tregs upon activation (120). sFGL2 seems to act as an immunosuppressor, repressing the proliferation of alloreactive T lymphocytes and the maturation of DCs (121, 122). Thus, by binding to FcγRIIB and FcγRIII, sFGL2 can adjust the antigen presentation ability of APCs. Accordingly, the levels of Th2 cytokines and the activity of DCs have found to be increased in FGL2-deficient mice (123).

### IMMUNOTHERAPEUTIC POTENTIAL OF APPs FOR TOLEROGENIC DC INDUCTION

Pharmacological immunosuppression has gone mainstream of past and, still, current therapeutic strategies to prevent transplant rejection and to restore autoantigen tolerance in autoimmune disorders. Yet, the downside of the immunosuppressive regimens is the appearance of numerous and often severe side effects and increased risk of infection as a consequence of the general suppression of the host immune system (124–126). Thus, the attractive concept of using DCs, central orchestrators of other immune cells, with the aim to modulate immune-inflammatory responses that have gone awry while leaving protective immunity intact is becoming gradually a reality in the clinical setting. In fact, in addition to being explored in experimental animal models of autoimmune diseases such as collagen-induced arthritis (127, 128), diabetes (129, 130), and experimental autoimmune encephalomyelitis (EAE) (131), along with experimental graft rejection after transplantation (132, 133), tolerogenic DCs have recently been, and are currently being tested in phase I clinical trials for alloimmune (transplantation, graft-versus-host disease) and autoimmune processes (type I diabetes, rheumatoid arthritis, multiple sclerosis, and Crohn's disease), and allergy, and there are ongoing collaborative efforts to harmonize/standardize tolerance-inducing therapies for upcoming trials (134–136).

The identification, characterization and, most notably, isolation and amplification of genuine regulatory or tolerogenic DCs populating a given healthy or diseased tissue, much like MDSCs, has proven a daunting task and a real limitation when planning to adoptively transfer them for therapeutic benefit. Therefore, due to the high plasticity of mononuclear myeloid cells such as monocytes, well-established *ex vivo* protocols of monocyte to Mo-DC expansion and differentiation, relying in the use of inflammatory cytokines (GM-CSF and IL-4) have become instrumental to adopt Mo-DCs as therapeutic cell products for clinical use (137).

A central aspect for the successful clinical application of tolerogenic Mo-DC relates to their development and manufacture. As previously stated, a variety of agents have been employed *in vitro* to skew Mo-DCs toward a tolerogenic or regulatory phenotype (notably vitamin D3, immunosuppressive drugs-like dexamethasone, or NF-κB inhibitors), opposing their "natural" tendency to be activated *in vivo* in a pro-inflammatory environment. Yet limited efficacy has been reported in terms of disease outcome, although most trials have noted an increase in Treg levels in the recipient's blood during tolerogenic Mo-DC administration (138–140). Clearly, maintaining tolerogenic Mo-DCs in an activation- or maturation-resistant state is a fundamental requirement for a successful tolerogenic Mo-DC therapy, because unstable tolerogenic Mo-DCs able to reverse back *in vivo* to an immunogenic phenotype in contact with a pro-inflammatory microenvironment could aggravate the pathology. Indeed, semi-mature DCs, considered tolerogenic in *in vitro* assays, may become immunogenic when administered *in vivo* (141, 142). Furthermore, it has been recently reported that continuous treatment of DCs during their differentiation from bone marrow cells (10-day treatment) with the histone deacetylase inhibitor suberoxylanilide hydroxamic acid generated tolerogenic DCs that, however, were not stable and, therefore, inefficacious when administered in mice with EAE (143). Hence, the possibility to anticipate and modulate the stability of tolerogenic Mo-DCs *in vivo*, particularly in the pro-inflammatory allo- or autoimmune environments in which these cells are applied, would enhance their therapeutic efficacy. As yet, there is not enough mechanistic knowledge to ascertain which stimuli guarantee the induction of stable tolerogenic Mo-DCs adapted to particular *in vivo* situations. Still, in tolerogenic Mo-DC conditioning protocols, establishing the appropriate timing and intensity of the tolerogenic reagent treatment, its toxicity, as well as the migration capacity of the resulting conditioned cells are crucial aspects to take into account for a successful Mo-DC-based immunotherapy (134). For example, both rapamycin- and dexamethasone-conditioned cell cultures have been shown to markedly reduce DC recovery (144, 145). Importantly, none of the APPs proved cytotoxic in the Mo-DC cultures, most likely because of the wide range of physiological concentrations that these proteins are able to reach in serum, fluctuating between homeostatic and acute phase conditions. Thus, APP-derived tolerogenic Mo-DCs might overcome some of the limitations of the current tolerogenic Mo-DCs employed for immunotherapy approaches regarding consistency, safety, and efficacy (124, 146).

On the other hand, the common tolerogenic moonlighting activity of APPs over Mo-DCs seems surprising, given the variety of different physiological functions ascribed to these proteins. Remarkably, nearly all of them share a complex, oligomeric, and


APPs for Tolerogenic DC Generation

Serrano et al.


multi-modular structure (**Table 1**), providing flexibility in their capacity for binding, with a different grade of specificity, a variety of humoral and/or cellular determinants, including different receptors in the surface of Mo-DCs. This feature might constitute an advantage for the fine-tuning of the desired tolerogenic phenotype on Mo-DCs.

Another key aspect of APP action, as outlined in previous sections, relates to the simultaneous presence and increased levels of both APPs and Mo-DCs under overwhelming immuneinflammatory conditions such as the acute phase response, which incites a physiological crosstalk between these humoral (APPs) and cellular (Mo-DCs) systems with the common goal of providing protection and progress toward the resolution of inflammation. In this regard, APPs probably contribute *in vivo* to mononuclear phagocyte switching toward an anti-inflammatory mode aimed at restoration of tissue integrity and function. Consequently, the APP interaction with Mo-DCs will be safer and effective over a wide range of concentrations, according to the significantly increased blood levels reached by APPs under acute phase conditions. On the other hand, most APPs have been shown to act over a narrow window within the Mo-DC differentiation and/or maturation program, limiting also their hypothetical toxicity, if any, and increasing their specificity compared with some of the current immunosuppressive/immunomodulatory agents, which need to be present over the full differentiation/maturation program to induce a tolerogenic outcome into Mo-DCs. Furthermore, several APPs and, particularly, all soluble PRMs, operate only in the early stages of monocyte to Mo-DC differentiation, while IL-10, for example, is active on Mo-DCs up to their terminal differentiation, when Mo-DCs downregulate the IL-10 receptor (147). This restricted activity of APPs at the beginning of the Mo-DC differentiation program may induce a more permanent and stable modification of the Mo-DC tolerogenic phenotype than that achieved by immunomodulators/immunossupressors influencing Mo-DCs late in their differentiation process, or by agents affecting only their maturation/activation status. These last agents may be more prone to be influenced by the pro-inflammatory microenvironment that the tolerogenic Mo-DCs face upon clinical administration. Comparing the performance and outcome of *in vitro* assays, APPs seem to hold a tolerogenic activity over Mo-DCs (low expression of co-stimulatory molecules, low production of pro-inflammatory cytokines, increased release of anti-inflammatory cytokines, low T cell alloproliferation and, instead, increased Treg generation, etc.) at least as efficient as the agents (rapamycin, dexamethasone and/or vitamin D3, NF-κB inhibitors, etc.) currently employed to generate clinical-grade tolerogenic Mo-DCs for the induction or restoration of immune tolerance in autoimmune pathologies and transplantation (148). Furthermore, given the broadened presence that APPs can reach in serum under acute phase conditions and the present hurdles facing adoptive DC-based immunotherapy (time-consuming, expensive, and arduous to implement in the current regulatory environment), the direct *in vivo* administration of APPs, either naked or complexed with nanoparticles, may become a useful and efficacious alternative in inflammatory pathologies. In fact, nanoparticle formulations for DC-specific receptor targeting (DEC205, DC-SIGN, CD40, CD11c, etc.) are being used in preclinical assays and phase I clinical trials as vaccines for oncoimmunotherapy (149–151).

Nevertheless, presently the most important drawback for the use of APPs as tolerogenic agents lies in the fact that the detailed molecular mechanisms of action of APP-mediated transformation of Mo-DCs toward a tolerogenic phenotype are not known for most of these proteins. Thus, current efforts employing highthroughput genomics and proteomics approaches will certainly dissect cell surface-interacting partner(s) and relevant signaling and metabolic pathways underlying APP-mediated programming and distinctive functional outcome of the ensuing tolerogenic Mo-DCs (152, 153).

### CONCLUSION AND PROSPECTS

For a successful tolerogenic immunotherapy, Mo-DC conditioning must regulate antigen-specific immune responses in the intrinsically complex pro-inflammatory environments evolving in autoimmune disorders and transplantation, sustaining the development of immunological memory toward tolerance. Thus, it is critically important to thoroughly test the performance of novel tolerance-inducing agents regarding the potency and durability of the ensuing tolerogenic Mo-DC phenotype.

Besides being proposed as useful biomarkers for a variety of inflammatory pathologies, recent studies have proposed that APPs play important roles in tissue homeostasis and repair following overwhelming immune-inflammatory processes, probably in close interaction with inflammatory monocytes and DCs. In fact, APPs are able to generate tolerogenic Mo-DCs *in vitro* with the desired regulatory features (increased expression of immunomodulatory molecules, enhanced production of anti-inflammatory cytokines, and Treg generation) and low immunogenicity (**Table 1**), comparable with the currently used clinical tolerogenic Mo-DC-inducing immunomodulatory/immunosuppressive agents. Although the precise mechanism of action of tolerogenic Mo-DC skewing induced by most APPs is still unknown, these proteins may prove useful alternatives to overcome the present limitations for a more efficacious, safe, and stable Mo-DC-based tolerogenic immunotherapy. In this regard, attractive attributes of APPs include a physiological basis regarding their interaction with Mo-DCs in the context of the acute phase response, and a wide range of action due to the own intrinsic features of APPs, which would ensure reduced toxicity at the cellular level and increased safety upon *in vivo* administration. Moreover, the narrow activity window in the early stages of monocyte to Mo-DC differentiation shown by

### REFERENCES


several APPs, notably soluble PRMs, should increase specificity and, more importantly, may contribute to a more stable tolerogenic phenotype. APPs targeting differentiating Mo-DCs could turn these cells unresponsive to the *in vivo* pro-inflammatory microenvironment present in autoimmune or alloimmune conditions and, therefore, refractory to Mo-DC maturation. Furthermore, taking into account novel findings, such as the proteomic characterization of tissue-/disease-specific posttranslational modifications of APPs (14), or the influence of the clinical status of the Mo-DC recipient (154) may fine-tune the tolerogenic potential of APPtreated Mo-DCs, e.g., their ability to modulate T cell responses. Nevertheless, additional research should help clarify whether some APPs, particularly pentraxins, or complement activators, are able to maintain their induced tolerogenic DCs in a stable and functional state upon administration in complex pathological tissue contexts, because of the dual protective and pro-inflammatory role played by these multifaceted molecules in physiology.

Definitely, although further work is warranted to establish which method, or perhaps combination of methods, is most suitable to generate tolerogenic Mo-DCs in the clinical setting, APPs may contribute, either on their own, combined with currently employed immunomodulators/immunosuppressants (155), and/or with recently proposed tolerogenic DC boosters such as minocycline (156), or reinforcing the tolerogenic properties of iPSC-derived CD141<sup>+</sup> DCs holding enhanced capacity for antigen cross-presentation (157), to the design of tailored protocols to induce or re-establish immunological tolerance in different clinical settings including allogeneic transplantation and autoimmune diseases.

### AUTHOR CONTRIBUTIONS

JA wrote the manuscript. IS and AL added additional insights. The final version was proofread and edited by all authors.

### FUNDING

JA received support from Ministerio de Economía y Competitividad (Madrid, Spain) [grants FIS-ISCIII PI13/01490 and PI16/00377, co-funded by FEDER funds/European Regional Development Fund (ERDF)-a way to build Europe-], from "La Marató de TV3" Foundation (grant 12/1210), and from Generalitat de Catalunya (grant 2014SGR541). JA is sponsored by the "Researchers Consolidation Program" from the SNS-Dpt. Salut Generalitat de Catalunya (Exp. CES06/012).


interleukin 18 in response to hepatitis C virus replication. *Gastroenterology* (2014) 147(1):209–20.e3. doi:10.1053/j.gastro.2014.03.046


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflict of interest.

*Copyright © 2018 Serrano, Luque and Aran. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Michela Comi, Giada Amodio and Silvia Gregori\**

*San Raffaele Telethon Institute for Gene Therapy (SR-Tiget) San Raffaele Scientific Institute IRCCS, Milan, Italy*

The prominent role of tolerogenic dendritic cells (tolDCs) in promoting immune tolerance and the development of efficient methods to generate clinical grade products allow the application of tolDCs as cell-based approach to dampen antigen (Ag)-specific T cell responses in autoimmunity and transplantation. Interleukin (IL)-10 potently modulates the differentiation and functions of myeloid cells. Our group contributed to the identification of IL-10 as key factor in inducing a subset of human tolDCs, named dendritic cell (DC)-10, endowed with the ability to spontaneously release IL-10 and induce Ag-specific T regulatory type 1 (Tr1) cells. We will provide an overview on the role of IL-10 in modulating myeloid cells and in promoting DC-10. Moreover, we will discuss the clinical application of DC-10 as inducers of Ag-specific Tr1 cells for tailoring Tr1-based cell therapy, and as cell product for promoting and restoring tolerance in T-cell-mediated diseases.

Keywords: interleukin-10, dendritic cells, tolerance, DC-10, T regulatory type 1 cells

### INTRODUCTION

Interleukin (IL)-10 is a powerful anti-inflammatory cytokine that plays an essential role in dampening immune responses and in preventing chronic inflammatory pathologies (1). Deficiency or aberrant expression of IL-10 or IL-10 receptor (IL-10R) enhance inflammatory responses to microbial challenge and lead to the development of inflammatory bowel disease (2–4) and several autoimmune diseases [as reviewed in Ref. (5, 6)]. Some pathogens can harness the immunosuppressive capacity of IL-10 to limit host immune responses, leading to persistent infection [as reviewed in Ref. (7)].

Human IL-10 was cloned (8) from a tetanus toxin-specific CD4<sup>+</sup> human T-cell clone isolated from peripheral blood of a patient with severe combined immunodeficiency successfully transplanted with fetal liver and thymus, who spontaneously developed tolerance (9). From its discovery, IL-10 has been demonstrated to be produced by almost all leukocytes, including all T cell subsets, monocytes, macrophages, dendritic cells (DCs), B and natural killer (NK) cells, mast cells, neutrophils, and eosinophils [reviewed in Ref. (10)]. In addition, epithelial cells and keratinocytes can also secrete IL-10 in response to infection or tissue damage as well as tumor cells (11, 12).

Interleukin-10 upon interaction with IL-10R regulates the expression of several genes resulting in the downregulation of pro-inflammatory mediators, the inhibition of antigen (Ag) presentation, and the upregulation of immune-modulatory molecules. Overall, IL-10 modulates antigenpresenting cells (APCs), inhibits, directly and indirectly, effector T cell proliferation and cytokine production, and promotes regulatory cell differentiation [reviewed in Ref. (13, 14)].

### *Edited by:*

*Luis Graça, Universidade de Lisboa, Portugal*

### *Reviewed by:*

*Muriel Moser, Université libre de Bruxelles, Belgium Ana Izcue, Uniklinik RWTH Aachen, Germany*

> *\*Correspondence: Silvia Gregori gregori.silvia@hsr.it*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 15 November 2017 Accepted: 20 March 2018 Published: 06 April 2018*

#### *Citation:*

*Comi M, Amodio G and Gregori S (2018) Interleukin-10-Producing DC-10 Is a Unique Tool to Promote Tolerance Via Antigen-Specific T Regulatory Type 1 Cells. Front. Immunol. 9:682. doi: 10.3389/fimmu.2018.00682*

**47**

Here, we present an overview on the role of IL-10 in promoting the differentiation of myeloid regulatory DCs, focusing on the induction of a subset of human tolerogenic (tol) DCs, termed DC-10. Moreover, we discuss the role of DC-10 in modulating T cell responses *in vitro* and *in vivo* and the current clinical application of DC-10 for cell-based therapeutic approaches.

### IL-10 AND MODULATION OF MYELOID CELLS

Interleukin-10 signaling in monocytes/macrophages and DCs converges, *via* several mechanisms, to regulate nuclear transcriptional events, inducing the initiation of homeostatic and anti-inflammatory programs. IL-10 interacts with a tetrameric receptor consisting of two IL-10Rα and two IL-10Rβ subunits. IL-10Rα binds IL-10, while IL-10Rβ, interacting with accessory molecules, mediates intracytoplasmic signals (14). IL-10/IL-10R interaction leads to phosphorylation of Janus kinase 1 (JAK1) associated with IL-10Rα and of Tyrosine Kinase 2 (TYK2), associated with IL-10Rβ. These kinases further phosphorylate two tyrosine residues located on the intracellular domain of IL-10Rα that act as temporary docking sites for STAT3 and STAT1 (15). Phosphorylated STATs homo/hetero-dimerize and translocate into the nucleus, where they bind to STAT-responsive genes (1, 16). Although the mechanisms underlying the IL-10/ STAT3-mediated responses are still to be fully understood, it has become evident that both IL-10 and STAT3 are required for anti-inflammatory responses. In macrophages, one of the major effects of IL-10/STAT3-mediated signaling is the transcription inhibition of up to 20% of the LPS-induced genes (17). This antiinflammatory activity is mediated primarily by STAT3 that, upon nuclear translocation, promotes the expression of specific genes, including those encoding for transcription factors, the ultimate effectors of the IL-10-mediated anti-inflammatory responses (18). Among molecules involved in inhibiting activation of myeloid cells, BCL3 has been shown to suppress LPS-induced TNF-α expression by inhibiting NF-kB (19), and NFIL3 has been demonstrated to specifically target a distal enhancer of *Il12b* and repress IL-12p40 expression (20, 21). IL-10/STAT3-mediated signal in macrophages promotes the expression of suppressor of cytokine signaling 3 (SOCS3) (22), a member of the SOCS protein family that plays important roles in the negative regulation of cytokine signaling pathways (23) (**Figure 1**). Although both IL-10 and IL-6 promote *via* STAT3 the expression of SOCS3, its inhibitory effects are restricted to IL-6R-mediated signaling (16). This evidence indicates that SOCS3 plays a role in regulating the pro-inflammatory effects of IL-6 (24).

In macrophages, upon activation with LPS or TNF-α, IL-10 prevents the activation and nuclear translocation of the classical NF-kB by inhibiting IkB kinase (IKK) activity (25–27), and hampers NF-kB DNA binding (28). This mechanism has been applied also to *in vitro* differentiated myeloid DCs, in which pre-treatment with IL-10 results in NF-kB inhibition that correlates with suppression of IKK and Akt activities (29). Similarly, the addition of IL-10 during TLR-mediated activation of monocyte-derived DCs hinders MyD88 signaling, leading to the downregulation of NF-kB family members c-Rel and p65, and interferon regulatory factor (IRF)-3 and IRF-8, an effect mediated by the inhibitory activity on the PI3K/Akt pathway (30). The IL-10-mediated inhibition of the PI3K/Akt signaling pathway leads also to the activation of the glycogen synthase kinase 3 beta (GSK3beta) and of the downstream microphthalmia-associated transcription factor (MITF) that translocates to the nucleus and drives the expression of the inhibitory molecule glycoprotein (GP) NMB (30) (**Figure 1**). At steady state and upon activation of myeloid cells, IL-10 signaling induces the selective nuclear translocation of NF-kB p50/p50, overall preventing the expression of several pro-inflammatory mediators, including IL-6 and MIP-2α (27). Interestingly, in activated macrophages, BCL3, a member of the IkB protein family localized in the nucleus and tightly associated with NF-kB p50 (31), acts to repress the transcription of proinflammatory cytokines, and positively regulates the expression of IL-10 (32).

An additional effect of IL-10 in myeloid cells is the downregulation of MHC class II (33, 34) and costimulatory molecules (35) expression (**Figure 1**). The mechanism of IL-10-mediated deregulation of MHC class II expression involves the transport inhibition of mature and peptide-loaded MHC class II complex to the plasma membrane (36). These IL-10-mediated effects are completely reversed by blocking STAT3 (37), although the role of STAT3 in these mechanisms has not been fully elucidated.

Interleukin-10 regulates at post-transcriptional levels, *via* micro (mi)RNAs, the expression of pro-inflammatory cytokines (38). IL-10 inhibits the expression of LPS-induced miR155, allowing the expression of SH-2 containing inositol 5′ polyphosphatase 1 (SHIP-1), which in turn negatively regulates PI3K-mediated activation of NF-kB and MAPK, and switches off the pro-inflammatory response (39). On the contrary, upon LPS stimulation, IL-10 rapidly and transiently enhances miR146b and sustains miR187 expression in myeloid cells. miR187 acts as negative modulator of LPS responses by directly limiting TNF-α production at posttranscriptional level and by reducing IL-6 and IL-12p40 transcription *via* silencing the transcription factor IkB (40) (**Figure 1**). miRNAs have been also involved in regulating IL-10 expression upon LPS-mediated activation: upregulation of miR21 indirectly increases IL-10 production *via* downregulation of programmed cell death 4 (41). Overall, these evidences indicate that, through a complex network of miRNAs, IL-10 drives anti-inflammatory responses by upregulating miR146b and miR187 and by downregulating pro-inflammatory miRNAs, such as miR155.

In summary, IL-10 directly and indirectly, *via* inducing STAT3 responsive genes and/or modulating NF-kB and MAPK activities, inhibits pro-inflammatory cytokine gene transcription in activated myeloid cells, and the expression of MHC class II and costimulatory molecules, overall preventing the ability of myeloid cells to efficiently present Ags to T cells and to activate effector T cells [reviewed in Ref. (7)].

Besides being an anti-inflammatory mediator, IL-10 promotes the expression of several tolerogenic molecules in human monocytes, macrophages, and DCs, including IL-10 itself (15), heme-oxygenase (HO-1) (42, 43), and immunoglobulin-like transcript 3 (ILT3) and ILT4 (44). HO-1 is a protein of heme

Figure 1 | IL-10-mediated modulation of myeloid cells. IL-10 binds to a tetrameric receptor consisting of two IL-10Rα and two IL-10Rβ subunits. 1. IL-10/IL-10R interaction leads to JAK1 and TYK2 phosphorylation and the consequent STAT3 and STAT1 phosphorylation. P-STATs, and in particular P-STAT3, dimerizes and translocates to the nucleus, where it promotes the transcription of specific molecules (i.e., SOCS3) or transcription factors (i.e., BCL3 and NFIL3), and inhibits the transport of MHC class II to the plasma membrane. 2. IL-10 signaling inhibits LPS-mediated activation of IKK that in turn prevents NF-kB-p65/p50 nuclear translocation and the expression of pro-inflammatory cytokine. In parallel, IL-10 promotes the selective NF-kB-p50/p50 nuclear translocation, which concurs in downregulating pro-inflammatory cytokine expression, and, in association with BCL3, promotes IL-10 expression. 3. IL-10 inhibits PI3K/Akt pathway that prevents LPS-mediated activation of MyD88, resulting in the inhibition of the expression of IRF-3 and IRF-8. 4. IL-10-mediated inhibition of PI3K/Akt pathway leads to GSK3β and MITF activation, responsible for the upregulation of the transcription of GPNMB. 5. IL-10 downmodulates LPS-induced expression of miR155, which directly inhibits SHIP1 and favors the negative regulation of TLR4 signaling by counteracting PI3K activity. 6. IL-10 enhances LPS-mediated induction of miR146b and miR187, which post-transcriptionally regulate mRNA encoding for TNF-α and reduce IL-6 and IL-12p40 transcription *via* inhibition of the transcription factor IkB. TYK, tyrosine kinase; JAK, Janus kinase; PI3K, phosphoinositide 3-kinase; Akt, protein kinase B; MyD88, myeloid differentiation primary response 88; STAT, signal transducer and activator of transcription; IKK, IkB kinase; NF-kB, nuclear factor kappa-light-chain-enhancer of activated B cells; GSK3β, glycogen synthase kinase 3 beta; MITF, microphthalmia-associated transcription factor; MAPK, mitogen-activated protein kinase; SHIP1, SH-2 containing inositol 5′ polyphosphatase 1; Bcl3, B-cell lymphoma 3-encoded protein; NFIL3, nuclear factor interleukin 3; TFs, unknown transcription factors; SOCS, suppressor of cytokine signaling; GPNMB, glycoprotein NMB; HO-1, heme-oxygenase-1; ILT, immunoglobulin-like transcript; IRF, interferon regulatory factor; IL-10R, IL-10 receptor.

degradation pathway playing a central role in tissue homeostasis and protection against oxidative stress (42). HO-1 is involved in the polarization of anti-inflammatory macrophages, which in turn acquire the ability to secrete high levels of HO-1 (45, 46). In human DCs, HO-1 inhibits their ability to stimulate allogeneic (allo) T cells and promotes their suppressive effects (43). ILT3 and ILT4 display a long cytoplasmic tail containing immunereceptor tyrosine-based inhibitory motifs that upon binding to HLA class I molecules transduce a negative signal through the recruitment of the tyrosine phosphatase SHP-1. This leads to inhibition of NF-kB activation and, consequently transcription of genes encoding for costimulatory molecules (47, 48). Finally, IL-10 upregulates the transcription of the non-classical HLA class I molecule HLA-G (49, 50), one of the ILT4 ligands (47) with known immune-modulatory functions.

Overall, IL-10 *via* several mechanisms regulates activation and function of myeloid cells, thereby playing an important role in modulating immune responses in healthy and pathological conditions.

### IL-10-MEDIATED MODULATION OF MONOCYTE-DERIVED DCs

Interleukin-10 has been repetitively applied to modulate *in vitro* differentiation of monocyte-derived DCs with contradictive results (51–53). Allavena et al. (51) demonstrated that IL-10 prevents DC differentiation by promoting a macrophage-like cell phenotype, whereas other studies reported that monocytes treated with IL-10 express markers associated with DCs (52, 53). Our group demonstrated that monocyte-derived DCs generated in the presence of IL-10 are a distinct subset of DCs with regulatory activities [(54), see next paragraph].

Interleukin-10 has been also applied to regulate already differentiated monocyte-derived immature (55) or matured (56, 57) DCs. In both settings, DCs exposed to IL-10 treatment express reduced levels of MHC class II and costimulatory molecules, show decreased Ag-presenting capacity, and become regulatory cells with the ability to promote anergic T cells (55, 56) with suppressive activity *in vitro* (57). More recently, it has been demonstrated that DCs matured in the presence of IL-10, termed IL-10-induced DCs, consist of two phenotypically and functionally distinct populations: CD83highCCR7+ and CD83lowCCR7<sup>−</sup> cells. The former cells display a strong migratory activity toward secondary lymphoid organs, have a stable phenotype, and induce *in vitro* T regulatory (Treg) cells with high suppressive activity. Based on these observations, the authors indicate that CD83high CCR7+ IL-10-induced DCs are promising candidates for cellbased approaches to induce/restore tolerance *in vivo* (58).

### DC-10 A DISTINCT POPULATION OF HUMAN TOLEROGENIC DENDRITIC CELLS (tolDCs)

DC-10 are an inducible subset of human tolDCs characterized by the ability to secrete high levels of IL-10 in the absence of IL-12, and by the expression of a specific set of markers including CD14, CD16, CD11c, and CD11b, but not CD1a, M-DC8, or CD68 (54). Despite being generated from precursors in the presence of IL-10, DC-10 are mature cells expressing CD80, CD86, and HLA class II molecules. Importantly, DC-10 express a bunch of tolerogenic molecules such as ILT2, ILT3, ILT4, and HLA-G. Functional assays showed that, although DC-10 have a low stimulatory activity, they promote T cell anergy and induction of allo-specific T regulatory type 1 (Tr1) cells (50, 54, 59, 60). Tr1 cells are a subset of CD4<sup>+</sup> T cells that co-express the integrin alpha2 subunit (CD49b) and the lymphocyte-activation gene 3 (LAG-3) (61), and secrete IL-10, TGF-β, variable amounts of IFN-γ and low/no IL-2, IL-4, and IL-17. Tr1 cells suppress immune responses *via* the secretion of IL-10, TGF-β, and of granzyme B [as reviewed in Ref. (13, 62)]. We demonstrated that DC-10 promote Tr1 cell differentiation *via* the IL-10-dependent ILT4/HLA-G pathway (54). Interestingly, DC-10-mediated induction of Tr1 cells is associated with high HLA-G expression (50).

DC-10 are present in peripheral blood and secondary lymphoid organs of healthy subjects and accumulate in human decidua in the first trimester of pregnancy (63). Interestingly, in peripheral blood of pregnant and non-pregnant women, the frequency of DC-10 is comparable, suggesting that either DC-10 migrate into decidua during pregnancy or are induced within the endometrium. Human decidua microenvironment is enriched in GM-CSF and IL-10 (64), both known to promote DC-10 differentiation, thereby decidual DC-10 can be either *de novo* induced from monocytes or derived from the conversion of resident decidual APCs. In the decidua of women with early miscarriage, DC-10 frequency is low (65), suggesting that in an inflammatory microenvironment differentiation of DC-10 is impaired. In line with this conclusion, in women with preeclampsia a subset of decidual CD14<sup>+</sup>DC-SIGN<sup>+</sup> APCs with reduced HLA-G and ILT4 expression and impaired ability to promote Tregs *in vitro* have been identified. The authors speculated that the reduced IL-10 levels observed in preeclampsia may lead to reduced HLA-G and ILT4 expression and impaired tolerogenic activity of these CD14<sup>+</sup>DC-SIGN<sup>+</sup> APCs (66).

An altered frequency of DC-10 has been reported in peripheral blood of cancer patients. In patients affected by acute myeloid leukemia, a significantly higher frequency of DC-10 compared with that observed in healthy donors was described. Interestingly, the percentage of DC-10 is higher in patients with HLA-G-expressing blasts compared with patients with HLA-G negative blasts (67). Even though the primary source of HLA-G was unclear, it was postulated that the presence of HLA-G-expressing DC-10 is involved in sustaining the expression of HLA-G on blasts contributing to inhibition of the immune system promoting tumor immune-escape. According to this hypothesis, an increased frequency of DC-10 expressing high levels of HLA-G has been identified in peripheral blood of patients with gastric cancer. Interestingly, the percentage of HLA-G<sup>+</sup>DC-10 strongly associates with advanced disease stage (68).

Overall, these studies indicate that DC-10 represent a subset of regulatory DCs contributing to IL-10-mediating tolerance and immune-escape.

### DC-10 AS INDUCERS OF Ag-SPECIFIC Tr1 CELLS

Dc-10 have entered the clinical arena as inducers of Ag-specific Tr1 cells for tailoring Treg-based cell therapy. We established and validated in Good Manufacturing Practice (GMP) conditions an efficient and reproducible *in vitro* method to generate, with minimal cell manipulation, allo-specific Tr1 cells (69, 70). Indeed, stimulation of T cells with allo-DC-10 induces a population of allo-specific Tr1 cells actively suppressing allo-specific effector T cells (50, 54, 59, 60). Recently, two improved GMP-compatible protocols using DC-10 have been developed for generating Tr1 cells for cell-based therapy. The first method generates allospecific Tr1 cells (named T-allo10 cells, Bacchetta and Roncarolo, ClinicalTrials.gov identifier: NCT03198234) by culturing purified CD4<sup>+</sup> T cells isolated from hematopoietic stem cell donor with patient-derived DC-10 in the presence of IL-10 (**Figure 2**). T-allo10 cells will be used as Tr1-based cell therapy in leukemia pediatric patients to prevent graft-versus-host disease (GvHD) (ClinicalTrials.gov identifier: NCT03198234). In the second protocol, CD4<sup>+</sup> T cells isolated from patients on dialysis are cultured with donor-derived DC-10 in the presence of IL-10 to generate donor-specific Tr1-enriched cell medicinal product (named T10 cells) (**Figure 2**). T10 medicinal products will be injected in kidney transplant recipients to prevent graft rejection (60).

Stimulation of Th2 cells isolated from house dust mite allergic patients with autologous *in vitro* differentiated DC-10 pulsed with

the allergen promotes their conversion of into IL-10-producing T cells (59). Moreover, DC-10 differentiated from monocytes of healthy subjects and peanut allergic patients and pulsed with relevant allergen induced the differentiation of peanut-specific Tr1 cells (71).

These findings indicate that patient-derived DC-10 can be *in vitro* pulsed with a given Ag and used to generate Ag-specific Tr1 cells for Treg-based cell approaches aim at restoring tolerance in allergy and autoimmune diseases.

### DC-10-BASED CELL THERAPY

The prominent role of DCs in promoting T-cell tolerance and the development of a GMP-compatible method to generate tolDC products allow their clinical application. Thus far, the few clinical trials performed demonstrated the safety and feasibility of tolDCbased cell therapies in the settings of autoimmunity and transplantation (72, 73). Nevertheless, the stability of the infused tolDC products and the maintenance of their tolerogenic properties *in vivo* remain open issues to be tackled for improving the safety and the efficacy of these therapies. Moreover, due to the increasing number of tolDCs that have been described, the optimal subset to be used as medicinal product is still to be defined. A comparative analysis of different populations of *in vitro* differentiated tolDCs examining their stability, cytokine production profile, and suppressive activity indicated that IL-10-modulated mature DCs are the best-suited cells for tolDC-based therapies (74, 75).

The observation that DC-10 are functionally more efficient than IL-10-modulated mature DCs in inducing hyporesponsiveness in allo-specific T cells (59) suggests that DC-10 represent a good alternative for cell-based approaches. Moreover, DC-10 are stable, since upon LPS stimulation, they maintain unaltered transcription profile and phenotype, and importantly the ability to induce Tr1 cells (76). DC-10 stability has been confirmed also *in vivo*, as their adoptive transfer modulates human T cell responses in a humanized mouse model. More recently, we demonstrated that DC-10 modulate allo iNKT cell induction and functions (Wu, under revision), indicating a broaden immunoregulatory function of DC-10, not limited to the CD4<sup>+</sup> T cell compartment. The potency, stability, and widespread immunoregulatory activity of DC-10 make feasible their application in clinical setting. Specifically, autologous DC-10 pulsed with a given Ag and allo-DC-10 can be infused in patients to restore tolerance in autoimmune diseases and allergy and to prevent allograft rejection and GvHD, respectively (**Figure 2**).

### CONCLUSION AND PERSPECTIVES

The discovery that DC-10 can be generated *in vitro* and induce Ag-specific Tr1 cell differentiation prompt their development as a tool for clinical approaches aimed at promoting/restoring Ag-specific tolerance in immune-mediated diseases. Protocols to generate alloAg-specific Tr1 cells with DC-10 for adoptive Tr1 based cell therapy have been developed and validated in GMP and are currently using in clinical applications. We believe that DC-10 represent a good candidate for DC-based therapies as they modulate effector immune responses, including pathogenic T cells, while leading to long-term tolerance *via* the *in vivo* induction of Ag-specific Tr1 cells. Studies in humanized mouse models are ongoing to establish the best route and dose of administration, lifespan, and homing kinetic of DC-10 and will be instrumental to design clinical protocols to test the safety and efficacy of DC-10 based cell therapy.

### AUTHOR CONTRIBUTIONS

MC and GA wrote the manuscript. SG designed, supervised the drawing of the manuscript, and wrote the manuscript.

### REFERENCES


### ACKNOWLEDGMENTS

The authors thank Dr. Daniele Avancini (Mechanisms of Peripheral Tolerance Unit, SR-TIGET, San Raffaele Scientific Institute, Milan, Italy) for critical reading of the manuscript. This work was supported by grant to SG from the Italian Telethon Foundation (TGT17G01) Associazione Italiana per la Lotta contro il Cancro (AIRC) project AIRC IG 18540, and from COST (European Cooperation in Science and Technology) and the Action BM1305 A-FAACT (http://www.afactt.eu) and the COST Action BM1404 Mye-EUNITER (http://www.myeeuniter.eu). COST is part of the EU Framework Programme Horizon 2020.


inhibits TLR-mediated activation of antigen-presenting cells. *Leukemia* (2009) 23:535–44. doi:10.1038/leu.2008.301


macrophages, and dendritic cells involved in antigen processing. *J Exp Med* (1997) 185:1743–51. doi:10.1084/jem.185.10.1743


regulatory T-cell induction in preeclampsia. *Am J Pathol* (2012) 181:2149–60. doi:10.1016/j.ajpath.2012.08.032


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Comi, Amodio and Gregori. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Tolerogenic Transcriptional Signatures of Steady-State and Pathogen-induced Dendritic Cells

*Emilia Vendelova1 , Diyaaeldin Ashour1 , Patrick Blank <sup>2</sup> , Florian Erhard1 , Antoine-Emmanuel Saliba3 , Ulrich Kalinke2 and Manfred B. Lutz1 \**

*<sup>1</sup> Institute for Virology and Immunobiology, University of Würzburg, Würzburg, Germany, 2 Institute for Experimental Infection Research, TWINCORE, Centre for Experimental and Clinical Infection Research, a joint venture between the Helmholtz Centre for Infection Research and the Hannover Medical School, Hannover, Germany, 3Helmholtz Institute for RNA-Based Infection Research (HIRI), Würzburg, Germany*

Dendritic cells (DCs) are key directors of tolerogenic and immunogenic immune responses. During the steady state, DCs maintain T cell tolerance to self-antigens by multiple mechanisms including inducing anergy, deletion, and Treg activity. All of these mechanisms help to prevent autoimmune diseases or other hyperreactivities. Different DC subsets contribute to pathogen recognition by expression of different subsets of pattern recognition receptors, including Toll-like receptors or C-type lectins. In addition to the triggering of immune responses in infected hosts, most pathogens have evolved mechanisms for evasion of targeted responses. One such strategy is characterized by adopting the host's T cell tolerance mechanisms. Understanding these tolerogenic mechanisms is of utmost importance for therapeutic approaches to treat immune pathologies, tumors and infections. Transcriptional profiling has developed into a potent tool for DC subset identification. Here, we review and compile pathogen-induced tolerogenic transcriptional signatures from mRNA profiling data of currently available bacterial- or helminth-induced transcriptional signatures. We compare them with signatures of tolerogenic steady-state DC subtypes to identify common and divergent strategies of pathogen induced immune evasion. Candidate molecules are discussed in detail. Our analysis provides further insights into tolerogenic DC signatures and their exploitation by different pathogens.

Keywords: tolerogenic dendritic cells, transcriptional profiling, steady-state dendritic cells, bacteria, mycobacteria, helminths, immune evasion

## TOLEROGENIC DENDRITIC CELLS (DCs)

Tolerogenicity of DCs is an intrinsic functional definition for this cell type and their induction of T cell anergy, regulatory T cells and T cell deletion have been reported (1). All major DC subsets have been described to exert tolerogenic functions. Tolerogenic DCs were first described *ex vivo*, showing that UV-irradiated Langerhans cells induced T cell anergy (2). Spontaneous or UV-induced apoptotic cell death represents a source of self-antigens employed by DCs for tolerance induction. Steady-state mechanisms to maintain self-tolerance rely on the uptake of apoptotic material and its tolerogenic presentation (3–6). The ability to generate tolerogenic DCs *in vitro* facilitated their subsequent use for adoptive cell therapy in mice. However, *in vitro* generated immature DCs injected to protect from allo-transplant rejection matured, as indicated by their upregulation of B7-1 and B7-2 molecules, an unwanted phenomenon that was hypothesized to dampen the DCs tolerogenicity (7). Later, this hypothesis was confirmed by generating immature and maturation-resistant DCs in the

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Eva Martinez Caceres, Universitat Autònoma de Barcelona, Spain Nathalie Cools, University of Antwerp, Belgium*

*\*Correspondence:*

*Manfred B. Lutz m.lutz@vim.uni-wuerzburg.de*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 24 October 2017 Accepted: 06 February 2018 Published: 28 February 2018*

#### *Citation:*

*Vendelova E, Ashour D, Blank P, Erhard F, Saliba A-E, Kalinke U and Lutz MB (2018) Tolerogenic Transcriptional Signatures of Steady-State and Pathogen-Induced Dendritic Cells. Front. Immunol. 9:333. doi: 10.3389/fimmu.2018.00333*

**55**

same transplantation model, which dramatically extended the allograft survival time from 22 days to more than 120 days (8). Thus, maturation resistance was considered as a hallmark of tolerogenic DCs to maintain their immaturity. Several protocols have been developed to achieve maturation resistance, mostly using maturation inhibitors such as IL-10, TGF-β, dexamethasone, or vitamin D3 alone or in combinations (1). Reports on the transcriptional profiling of such DCs treated with tolerogenic substances followed and have been described elsewhere (9, 10).

Here, we analyzed transcriptional data sets deposited on public databases from steady-state migratory DCs (ssmDCs) and functionally similar spontaneously matured GM-CSF-derived bone marrow DCs (BM-DCs) as tolerogenic DC sources. Since ssmDCs act in a tolerogenic manner, despite their partial maturity, they phenotypically resemble much more mature DCs than non-migratory, immature DCs do. Therefore, they represent a more similar DC phenotype for our comparison of tolerance markers. We then analyzed transcriptional datasets of DCs treated with substances known to cause inflammation, including pathogen-derived molecules. The comparisons concentrated on bacteria or bacterial products but also included helminths, known as masters of immune evasion, but excluded protozoa and viruses. Candidate tolerogenic molecules that were highly upregulated by selected inflammatory or pathogenic stimuli in DCs are then discussed individually and compiled in tables.

### TOLEROGENIC MARKERS IDENTIFIED FOR STEADY-STATE AND PATHOGEN-EXPOSED DCs

### Self-tolerance versus Microbial Immune Evasion

Dendritic cells residing in peripheral tissues at an immature stage act as immune sensors for pathogens. Pathogens, danger or inflammatory signals convert DCs into a mature/activated state which enables their migration into the draining lymph nodes. Subsequent stimulation of T cell immunity occurs by DC presentation of pathogen-derived antigens in the context of costimulation and proinflammatory cytokine production (11). In contrast, during homeostasis lymphoid organ-resident DCs and ssmDCs contribute to immune tolerance, thus controlling unwanted T cell responses against harmless or self-antigens (12).

Most microbes, especially those causing chronic infections, are evolutionarily well-adapted to their host. Such adaptation results in a balance between a pathogen-induced protective immune response and immune tolerance mechanisms that prevent microbial elimination. Infections with non-adapted microbes either kill the host rapidly or the microbe is immediately cleared by the host's immune response. In both cases, the microbes cannot replicate and spread to another host. A successful microbe induces a chronic and preferably asymptomatic infection. This can be achieved by exploitation of the host's immune tolerance mechanisms during pathogen–host coevolution.

Here, we analyzed public data in a comparative manner including tolerogenic and anti-inflammatory mRNA signatures of (1) steady-state DCs, (2) helminth-exposed DCs, (3) mycobacteriaexposed DCs, and (4) defined *in vitro* generated murine GM-CSF BM-DCs and human monocyte-derived DCs (MoDCs) treated with different inflammatory or pathogen-derived stimuli.

### Transcriptional Signatures of Tolerogenic Migratory DCs under Steady-State Conditions

To identify tolerogenic DC signatures after pathogen stimulation, we first sought to identify comparative DC subsets known for their tolerogenic function as a reference dataset. While CCR7<sup>−</sup> resident DCs appear at an immature stage, CCR7<sup>+</sup> ssmDCs undergo a homeostatic maturation process reaching a semi-mature stage, which is characterized by low expression of MHC II and costimulatory molecules, such as CD40 and CD86, and the absence of proinflammatory cytokine production (13–16). In several respects, steady-state plasmacytoid DCs (pDCs) resemble resident CD4<sup>+</sup> or CD8α+ conventional DCs (cDCs) of cutaneous lymph nodes and spleen (**Figure 1**). After pathogen-induced maturation DCs upregulate MHC II, CD40 and CD86 molecules on their surface (14, 15). Depending on the stimulus, mature RelB+++, RelA+++, and cRel+++ DCs differ qualitatively in the production of the proinflammatory cytokines IL-6, TNF, IL-1β, IL-12p70, IL-23, or type-I interferon, while RelB+++, RelA<sup>+</sup>, and cRel<sup>+</sup> ssmDCs induce Tregs by their release active TGF-β+ from its latent form of surface-bound latency-associated peptide (LAP) molecules (14–16). While tolerogenic functions of ssmDCs have been described by many authors, the demonstration of T cell tolerogenicity by immature lymph node-resident DCs is much less understood (17). Thus, due to their increased maturity, we

selected the tolerance markers of ssmDCs for comparison with pathogen-induced DCs.

In ssmDCs increased transcription of *Cd274 (*PD-L1), *CD200, Socs2, Relb, Ccl5*, and *IL12b* was observed as compared with pDCs and resDCs (**Figure 1**). Their enhanced transcription was observed in all three subsets of semimatured CCR7<sup>+</sup> ssmDCs but not immature resident DCs of lymph nodes (**Figure 1**). In addition, high levels of *CD83, Cd150, Aldh1a2 (Raldh2,) Adora2a*, and *Itgb8* were found in ssmDCs (**Table 1**). Of these 11 molecules, 6 were also found in spontaneously matured BM-DCs (**Table 1**). The individual roles and mechanisms of tolerogenicity are explained below or referred to in **Tables 1**–**5**. Although the extent to which GM-CSF-derived BM-DCs resemble cDCs is still a matter for debate (18), the tolerogenic signatures observed in spontaneously matured BM-DCs (19) are strikingly similar to those observed in ssmDCs (14–16) (**Table 1**).

### Tolerogenic Signatures of DCs Induced by Helminths

Due to evolutionary pressure, phylogenetically distinct parasitic worms—collectively termed "helminths"—convergently evolved the ability to manipulate their host's immune systems. In nearly all cases, the antihelminth type 2 immunity of M2 macrophages and T helper cell 2 (Th2) cells fails to eliminate the worms (59, 60); hence helminths persist within their hosts for years. Helminths often exploit the host's immune regulation machinery with DCs being major targets (59, 61, 62).

Type 2 immunity, in contrast to type 1, is promoted by weaker costimulation and/or absence of proinflammatory and polarizing cytokines such as IL-12p70 and IL-23 (13, 63). Moreover, the DC potential to induce type 2 immunity can be associated with tolerogenic mechanisms such as IL-10 secretion (63). Phenotypic maturation of DCs occurs after recognition of pathogen-associated molecular patterns (PAMPs) frequently inducing canonical NF-κB signaling (involving classical IκBα, -β, and -ε, NF-κB1 p50, RelA, and c-Rel). In contrast, recognition of helminths and their products by DCs results only in partial maturation resulting in low levels of costimulatory molecules at the surface and poor release of proinflammatory cytokines (64). It is believed that the non-canonical NF-κB pathway (Nfκb2/p52, RelB) not only direct cell development (65) but also might play a role in the regulation of immune tolerance (14, 66–68). Transcriptomic analyses of human DCs treated with Brugia malayi revealed upregulation of RELB and NFκB2 (24) and RelB in DCs isolated from mice after infection with *Nippostrongylus brasiliensis* (27) or *Schistosoma mansoni* eggs (28) (**Table 1**). This was similar to what has been observed in ssmDCs which induced Foxp3<sup>+</sup> Tregs from naive T cells (14). In line with this hypothesis, Lacto-N-fucopentaose III, a carbohydrate found in *S. mansoni egg antigen*, has been shown to activate the alternative NF-κB pathway in DCs (69). Thus, non-canonical NF-κB activation in the absence of low activity of canonical RelA and cRel may be characteristic for tolerogenic DCs in helminth infections.

The activation status and cytokine release of DCs fine-tunes the polarization of different T cell-effector and regulatory mechanisms. Suppressor of cytokine signaling (SOCS) proteins play decisive roles in innate immune cell signaling. They modify the polarization of immune responses by negative regulation of cytokine signals (70, 71). Different helminth species promote upregulation of *Socs2* and *Socs3* (24, 27) (**Table 1**), which may skew immune responses toward a Th2-biased anti-inflammatory phenotype. Indeed, it was shown that SOCS3-transduced DCs express low levels of MHC II and CD86 molecules on their cell surface and produced high levels of IL-10 but low levels of proinflammatory cytokines such as IL12p70. They thereby induced Th2-cell differentiation in mice supporting allergic Th2 responses but impairing Th1/Th17 development by means of immune deviation toward Th2 as shown in the autoimmune model EAE (72, 73). As described above, tolerogenic ssmDCs express *Socs2* (**Table 1**). Therefore, induction of *Socs2* during helminth infection might even inhibit Th2 differentiation and instead support a tolerogenic environment (27, 74). It is not clear whether helminths induce *Socs* expression directly or through indirect cell mechanisms such as host-derived cytokines. For example, anti-inflammatory *Il27* is expressed in DCs after immunization with *Nippostrongylus brasilienis* (27) (**Table 1**). IL-27 induces expression of *Socs3* in mouse and human cells leading to induction of IL-10-producing Tr1 cells (75).

Different DC populations exposed to helminths induce expression of the regulatory cytokines *IL12b* and *IL-10* (27, 28). CD103<sup>+</sup> migratory mature DCs from *N. brasiliensis* and *S. mansoni* infected mice significantly upregulate IL12b (27), also expressed in ssmDCs (**Figure 1**; **Table 1**).

Among others, *Cd200* and *Cd274* (PD-L1) were upregulated in DCs from *N. brasiliensis* immunized mice (**Table 1**). As detailed below, PD-L1 transmits inhibitory signals to PD-1 (CD279) on T cells. This interaction modifies TCR signaling, results in anergy or functional inactivation of T cells and is currently used for anticancer "checkpoint" inhibitory therapies (76, 77). PD-L1 expression would certainly support the chronicity of helminth infection. Suppression of T cell responses by PD1 during helminth infections has mainly been attributed to macrophages expressing PD-L1 and/or PD-L2 (78–80). Although the role of PD-L1 on DCs was not experimentally addressed, it may play a similar role.

Gene expression profiling using microarray or RNA sequencing technologies has been widely used to reveal cellular processes involved in host immune responses to different pathogens. Transcriptomic meta-analyses characterizing host immune responses against helminths have shown robust effects on immune gene signatures across different species (62). However, the common tolerogenic gene signature of DCs during helminth infection has not been addressed. Despite the fact that transcriptional profiling of DCs would improve our understanding of helminth effect during infection, the available helminth-related datasets are limited and further studies are required.

### Tolerogenic Markers Expressed after Infection with *Mycobacterium tuberculosis* (Mtb)

During coevolution with the human immune system, Mtb has developed multiple immune evasion strategies (81). To address whether Mtb is able to induce tolerogenic gene signatures in DCs, we analyzed transcriptional profiles of human DCs infected with Mtb and evaluated those for known tolerogenic markers.


Table 1 | Tolerogenic genes upregulated more than log2-fold by DCs matured during steady state, inflammation, or by pathogens.

*Steady-state migratory DCs/spontaneously matured BM-DCs.*

*LPS/TNF/CT-matured DCs.*

*Mycobacteria-matured DCs.*

*Helminth-matured DCs.*

Table 2 | Common transcripts induced under all six conditions (TNF, CT, LPS, each human and mouse data from Figure 2) and for which anti-inflammatory or tolerogenic functions have been reported.


Table 3 | Tolerogenic transcripts induced specifically by LPS (human and mouse data from Figure 2) and for which anti-inflammatory or tolerogenic functions have been reported.


Table 4 | Tolerogenic transcripts induced specifically by CT (human and mouse data from Figure 2) and for which anti-inflammatory or tolerogenic functions have been reported.


Monocyte-derived DCs infected with Mtb or BCG highly upregulated the two tolerogenic genes *IDO-1* and *IL27*. *IDO-1* upregulation was detected already 8 h after infection of human MoDCs, whereas *IL27* transcripts were detected only upon Mtb, but not BCG, infection (25). Others showed upregulation of *RELB*, *CD83*, and *HLA-G* in MoDCs after 16 h of Mtb infection (24). The tolerogenic function of *RELB* is discussed below. *CD83* might also confer a regulatory function, as indicated by inhibition of T-cell proliferation that was mediated by the soluble form of the CD83 protein (58). Finally, HLA-G has been shown to induce human MoDC tolerogenicity *via* its CD85b/ILT4 ligand in huILT4-transgenic mice, inducing anergy and suppressor T cells (82). Hence, expression of *IDO*, *IL27*, *RELB*, *CD83*, and *HLA-G*

Table 5 | Tolerogenic transcripts induced specifically by TNF (human and mouse data from Figure 2) and for which anti-inflammatory or tolerogenic functions have been reported.


(**Table 1**) by DCs might promote tolerogenic responses in Mtb infection.

### Tolerogenic Signatures of Murine and Human DCs Upregulated by Selected Inflammatory or Pathogenic Stimuli: TNF, Cholera Toxin, Lipopolysaccharide (LPS)

Transcriptional profiles of DCs stimulated *in vitro* under tolerogenic conditions have been reviewed before (10). Early transcriptional profiling work revealed that expression profiles of some cytokines are tightly regulated with time kinetic mRNA profiling revealing clear insights. IL-10 production stimulated by *Escherichia coli* LPS was only induced after 6 h in the DC cell line D1, but not earlier or later, whereas mRNA for TGF-β1 or IL-12p40 was detectable in time windows of more than 18–20 h after stimulation (83). DC cell line D1 showed IL-12p40 induction with LPS but not TNF (84) as reported for murine BM-DCs and human MoDCs (85). The fact that only two tolerogenic markers were identified in D1 cells may indicate a general limitation of obtaining transcriptional data from cell lines.

Of note, LPS stimulated DCs produce immunogenic Th1 polarizing IL-12p70, formed by the p35/p40 heterodimer (*Il12a* and *Il12b* genes), but the protein amounts of IL-12p40 secreted by DCs are typically 50–100 times higher than the amount measured for IL-12p70. Similarly, the IL-23 heterodimer secretion, composed of p19/p40 (*Il23a Il12b genes*) is much lower than p40 by cholera toxin stimulation of DCs (22, 23). This opens space for speculation on a counterbalancing and thereby tolerogenic role for excessive IL-12p40 production.

Dendritic cell maturation induced by inflammatory or microbial products triggering DAMPs or PAMPs, respectively, direct polarized Th1, Th2, or Th17 responses. Previously, we performed transcriptional profiling of murine GM-CSF generated BM-DCs and human MoDCs. Selected *in vitro* maturation protocols for induction of Th1 responses by LPS, Th2 by TNF and Th17 by cholera toxin (CT-DCs) were applied to both human and mouse DCs for the same time period of 6 h (22, 23) (GEO data bases GSE106080). Among the clearly immunogenic transcriptomic signatures, we also identified additional molecules at the protein level that exert tolerogenic immune functions. These include IL-10 production by LPS-DCs (86), Tr1 induction by Trypanosomamatured or TNF-DCs after three injections (22, 87) and Foxp3<sup>+</sup> Treg induction *via* TGF-β plus CTLA-2, a newly identified tolerogenic molecule from CT-DCs (23).

It remains a subject for debate whether the tolerogenic signature observed after infection has evolved as protective mechanism by the host or is actively induced by the pathogen. Pathogens aim to prevent their elimination and also the host aims to survive. If a pathogen cannot be eliminated, the host has to develop a protection strategy including the prevention of immunopathology. Excessive immune responses may be more deleterious than microbial pathogenicity in the host, as observed in sepsis. Thus, host-intrinsic negative feedback regulation of immune stimulation may be advantageous. To address this in our analyses, we included TNF as a non-pathogen-derived inflammatory stimulus. Interestingly, four tolerogenic genes showed increased mRNA transcription overlapping between TNF, CT and LPS stimulation (**Figure 2**) (**Tables 1** and **2**).

Pathogens and inflammatory mediators induce numerous mechanisms of immunity in DCs. Additionally, molecules with tolerogenic or anti-inflammatory functions are induced. Mouse BM-DCs and human MoDC generated with GM-CSF (±IL-4) result from conversion of classical human CD14<sup>+</sup> or mouse Ly-6Chigh monocytes into DCs, in a human-mouse interspecies comparison. As expected, common proinflammatory genes such as *Il-1β*, *Il-6*, and *Cox2* (*Ptgs2*) were upregulated under all six conditions. Furthermore, four gene transcripts*: Il12b*, *Ido-1*, *Cd150* (*Slamf1*), and *Inhba* (coding for Inhibin/Activin) with reported anti-inflammatory or tolerogenic function were upregulated under all 6 conditions by stimulation with TNF, CT, or LPS of both human MoDCs or mouse BM-DCs (**Figure 2** arrows, **Table 2**). Taken together, as these four genes were also upregulated by TNF, this tolerogenic DC response may reflect a host-initiated protection mechanism to avoid immunopathology rather than a purely pathogen-driven strategy.

Besides the common tolerogenic genes upregulated by all three stimuli, additional tolerogenic transcripts were found by the individual stimuli LPS (**Table 3**), CT (**Table 4**), and TNF (**Table 5**). These data indicate that microbial adaptation to the host and induction of tolerogenic signatures by LPS and CT not only share mechanisms of tolerogenicity but also differ in their strategies of immune evasion. Therfore, LPS selectively upregulates mRNA for adenosine A2a receptor, optineurin, and Slamf7/CD319, while CT induces higher transcription of thrombospondin-1 (TSP1) and Vegfa indicating divergent tolerance strategies (**Tables 3** and **4**).

Since tolerogenic signatures of differentially stimulated human MoDCs and mouse BM-DCs were strikingly similar (**Table 2**), we asked whether also distinct differences exist between DC from the two species. Surprisingly, very few genes were selectively upregulated by human MoDCs but remained unaltered or downregulated in murine BM-DCs and vice versa (**Figures 2B,C**). Among those, no tolerogenic genes appeared. Interestingly, differences in the expression of *Gitr* (*Tnfrsf18*) were found, confirming known differences in expression and function of GITR in mice and humans on DCs (91). Thus, with respect to LPS sensing and transcriptional responses, human MoDCs and murine BM-DCs are remarkably similar.

### THE ROLE OF SELECTED TOLEROGENIC MOLECULES IN HOMEOSTASIS AND IMMUNE EVASION

### Il12b

*Il12*b codes for IL-12p40 protein forming homo- and heterodimers. Two heterodimers can be formed with p40: p35/p40 that are linked *via* a disulfide bond to form IL-12p70 and p19/p40 to form IL-23. The release of IL-12p70 by DCs plays a pivotal role in the induction of Th1 responses (92, 93) while IL-23 supports Th17 generation (94, 95). However, the p40 monomer and especially the homodimer (p40)2 have been shown to strongly inhibit IL-12-dependent T cell or Th1 responses *in vitro* and *in vivo* (29, 30, 96), mainly by competing with IL-12p70. Interestingly, the total serum IL-12, and the ratio of IL-12p40/IL-12p70 increased with age in healthy individuals compared to IL-12p70 levels (97). This observation likely contributes to impaired immunity in the elderly. The expression of *IL12b* by ssmDCs is observed only in the CD103<sup>+</sup> Langerin<sup>+</sup> CD11blow subset (15), and is significantly higher on ssmDCs when compared to lymphoid organ-resident DCs (**Figure 1**) (**Table 1**). Since *IL12a* mRNA coding for IL-12p35, is undetectable or at very low levels in any of the subsets under steady-state conditions, this may point to a tolerogenic role of p40 homodimers as described.

### Relb

RelB is an NF-κB/Rel transcription factor family member associated with both tolerogenic and immunogenic functions (98). The RelB-p50 heterodimer has been associated with inflammatory and immunogenic responses (68). In this case, it functions through the RelA-NF-κB canonical pathway and cooperates with the cRel-p50 heterodimer (65). cRel is specifically required for IL-12p70 production (99). On the other hand, the RelB-p52 heterodimer, which functions through the NF-κB non-canonical pathway, was shown to be important for organogenesis of lymphoid organs (100), for normal development of splenic CD4+ and CD8+ (101, 102) and ssmDCs (14). RelB, but absence of (or extremely low levels) of RelA or cRel, is expressed by migratory DCs both under steady-state conditions and upon immune activation (14, 15) (**Figure 1**). In the peripheral lymph nodes of p52−/− mice, the ssmDC subsets were severely reduced while the resident DCs were not affected. In contrast, p50−/− mice did not show a specific preference for migratory or resident DCs and both were equally reduced (14). Additionally, RelB-deficient animals show a severe pathological phenotype characterized by inflammatory infiltrates into multiple organs, which is caused by hyper activity of conventional T cells (100). RelB+ ssmDCs have been shown to be either critical for conversion of naive T cells into Foxp3+ iTreg (14, 103), or for maintaining the homeostatic Foxp3+ natural Treg pool (16). Together, the available data indicate that moderate RelB expression in DCs alone is associated with lymphoid organogenesis and tolerogenic functions, whereas

LPS stimulated human DCs [1 = (85); 2 = (88); 3 = (89); 4 = GSE106080] or mouse DCs [1 = (90); 2 = (22)]. Only genes with probe sets on each of the microarrays used were retained and *z*-scores were computed as in Panel (A). In Panel (B), genes with *z*-score > 2 in at least two human experiments and *z*-score < 0 in both mouse experiments are shown. Panel (C) depicts genes with *z*-score > 2 in both mouse experiments and <0 in at least two human experiments.

MoDCs additional IL-4 was added. Murine data are from Ref. (22, 23), human data obtained from GEO data bases (GSE106080). (B,C) Expression signatures of

coexpression of RelB with RelA and cRel at high levels in DCs marks immunogenic functions.

### CC Chemokine Ligand 5 (Ccl5)

The *Ccl5* gene encodes CCL5, also known as RANTES, has been described as a gene expressed by activated T cells, macrophages, eosinophils, fibroblasts, epithelial cells as well as certain types of tumor cells. CCL5 plays an important role in the migration of different leukocytes toward inflammatory sites where it acts through its binding to CCR1, CCR3, or CCR5 (104). One interesting observation is that certain types of tumors express high levels of CCL5, which is a predictor of a poor prognosis (105, 106). Blocking of CCL5 can redirect myeloid-derived suppressor cells (MDSCs) and thereby improve antitumor immunity (107). CCL5 has been shown to be important for the generation of CD11b<sup>+</sup>/ Gr-1<sup>+</sup> MDSCs and its absence alters their differentiation and their immunosuppressive capacity (108). CCL5 release by NKT cells was required for the recruitment of antigen-specific CD8+ regulatory T cells and TGFβ-dependent tolerogenic antigen-presenting cells in order to mediate tolerance in the immune-privileged anterior eye chamber (109). Given the higher *Ccl5* expression by ssmDCs relative to resident DCs it will be interesting to uncover its precise function in these cell types (**Figure 1**) (15).

### IL-10

Several TLR ligands, including LPS, induce IL-12p70 release from DCs to induce Th1 immunity and, in parallel, release of IL-10 (110). Listeria infection in neonates induces CD8α+ DCs to release IL-10 (111). The suppressive effect of IL-10 on Th1 responses is indirect *via* DCs or macrophages (112) and seems to control IFN-γ release but not proliferation of Th1 clones *in vitro* (113). This IL-10 production has been suggested to serve as a selfcontrol mechanism to avoid Th1-mediated immunopathology (114) but also as a means of microbial immune evasion (115, 116). IL-10 can inhibit the differentiation of monocytes into Mo-DCs (117). Others found DC-to-DC effects by observing CpG-activated cDC-derived IL-10 blocked pDC release of type I interferons (118). Persistent production of IL-10 may then facilitate the conversion of Th1 (or Th2) responses into a IL-10<sup>+</sup> Foxp3<sup>−</sup> regulatory T cell response of the Tr1 type (119), similar to what had been observed for harmless antigen application and steady-state transport and Tr1 induction by lung DCs (120). The detailed regulation of IL-10 production (121) or its role of IL-10 for Tr1 cell induction has been reviewed elsewhere (122). However, although all this indicates an important role of IL-10 in immune tolerance, remarkably in none of the data sets analyzed herein (**Figure 2**; **Table 1**) was IL-10 identified as part of the tolerogenic transcriptional signature in DCs. The reasons for this may depend on delayed gene transcription kinetics or epigenetic regulation, thus identifying a limitation of tolerogenic transcriptional profiling.

## TGF-**β**/Itgb8

Foxp3 is the major transcription factor directing functions of thymus-derived natural Foxp3<sup>+</sup> Tregs, but also peripherally induced Foxp3<sup>+</sup> iTregs (123). Therefore, the production or employment of TGF-β by tolerogenic DCs for Treg generation or maintenance is of interest. TGF-β inhibits the maturation of BM-DC (124). However, murine BM-DCs produce soluble TGF-β when stimulated by *Lactobacilli* (125) and its release may be under the control of GITR (91). GM-CSF cultured BM-DCs lack the surface expression of LAP which can bind TGF-β in a latent form before it can be released for Treg induction (126). Therefore, they are unable to mediate iTreg conversion from naive CD4<sup>+</sup> T cells *in vitro* without addition of exogenous TGF-β (23). In contrast, lymph node DCs express LAP and the partially matured ssmDCs do so at even higher levels when compared with immature resident DCs (14).

The release of active TGF-β from its latent form is the critical event in TGF-β biological activity. The integrins αVβ6 (Itgav, Itgb6) (127), αVβ8 (Itgav, Itgb8) (128), and *TSP1* (43) have been described to mediate non-proteolytic release of TGF-β, while metalloproteinase 9 (*MMP9*) performs proteolytic release (129). The activity of integrin αVβ8 has been shown as a key mechanism to prevent autoimmunity by maintaining Treg activity (130). Thus, these genes might be better markers for transcriptional signatures of TGF-β activity, although not identified in any of the RNA profiling data sets analyzed here (**Figure 1**). This indicates that not all important tolerogenic molecules are transcriptionally regulated and can be identified in such studies. A broader tolerogenic transcriptional signature was also identified for the subset of incompletely matured XCR1<sup>+</sup> ssmDCs *ex vivo*, including the upregulation of TGF-β2 (15).

### Cd150/Slamf1

*Cd150* is upregulated on activated lymphoid and myeloid cells and acts *via* homotypic interaction (131). It represents the main human receptor for measles virus has been shown to inhibit DC functions (**Table 2**). Interestingly, the SH2D1A gene encoding for the SLAM-associated adapter protein to mediate SLAM signaling is mutated on X-linked immunodeficiency patients and responsible for the observed uncontrolled T and B lymphocyte proliferation after an EBV infection (132, 133). These data indicate that intact SLAM acts as an immune control molecule to prevent over activation of adaptive immunity during EBV infection.

## Indoleamine 2,3-dioxygenase (*Ido*)

IDO is an enzyme catabolizing l-tryptophan. Deprivation of this essential amino acid in the environment of proliferating T cells results in metabolic starvation, apoptosis and thus inhibition of the T cell responses (134). Interestingly, in pDCs a TGF-βdependent tolerogenic function of IDO has been reported that is independent of its enzymatic activity (135). IDO also plays a decisive role in establishment of LPS tolerance *via* control of the aryl hydrocarbon receptor signaling (136).

### Inhba

The genes *INHBA/Inhba* encode for the Inhibin-βA or ActivinβA protein. Inhibin-βA forms homo- or heterodimers with other inhibin/activin family members to form the protein complexes Activin A (βA/βA homodimer), Inhibin B (α/βA heterodimer), or Inhibin AB (βA/βB heterodimer). They all belong to the TGFβ family (126) and many of the TGF-β family members influence DC development and function (137). Inhibition of DC maturation has been reported for Activin A and Inhibin A (38). Activin A may cooperate with TGF-β to increase generation of Foxp3<sup>+</sup> induced regulatory T cells (iTregs) (39). Why Inhibin/Activin and not directly TGF-β are targets of immune evasion at the transcriptional level requires further investigation.

### Il27

IL-27 protein belongs to the IL-12 cytokine family and is a heterodimeric protein consisting of IL-27p28 and the Epstein-Barr virus-induced gene 3 (EBI3) (138). This cytokine is expressed early upon activation of antigen presenting cells. It has been shown to induce the initial step in Th-1 differentiation of naive CD4 T-cells by STAT1 dependent induction of T-bet (139). Besides this immunogenic function, several studies have analyzed the regulatory function of IL-27 during infection with various different pathogens (140). Infection with Mtb IL-27 was described to suppress T-cell responses by the reduction of TNF, IL-12p40, and IFN-γ expression and to inhibit T-cell recruitment and proliferation (141). Furthermore, IL-27 can induce the expression of IL-10 in activated CD4<sup>+</sup> effector T-cells and thus reduce antimycobacterial activity (116).

### Socs2

Suppressor of cytokine signaling proteins play important roles in both the maintenance of homeostasis and the resolution of inflammation (71). Recent evidence suggests that SOCS2 plays a role in immune regulation. Similar to SOCS1 and SOCS3, also SOCS2 regulates pattern recognition receptor signaling in both human and murine DCs by counterregulating their activation (142). *Socs2*<sup>−</sup>*/*<sup>−</sup> mice showed uncontrolled Th1 responses to *Toxoplasma gondii*, due to generalized proinflammatory responses to the infection (143). Besides innate immunity, SOCS proteins balance T helper cell polarization. SOCS1 and SOCS3 support Th17 cell generation by inhibiting Th1 differentiation while Th2 differentiation is regulated by SOCS3 (72, 73). SOCS2 was recently shown to play a major role in inhibiting the development of Th2 cells and Th2-associated allergic responses (74). However, whether SOCS expression in DCs is responsible for observed effects in T cells was not investigated by these studies. Here, we identified *Socs2* transcript elevation in all ssmDCs and spontaneously matured BM-DCs (**Table 1**) and upon *in vitro* exposure of DCs with *N. brasiliensis* and *S. mansoni*, further suggesting its important role in immune regulation (27, 28).

### Cd274

*Cd274* encodes programmed death ligand-1 (PD-L1) which delivers inhibitory signals *via* PD1 into T cells to regulate the delicate balance between immune defense and tissue-damage. PD-L1 is constitutively expressed or upregulated after activation on wide hematopoietic and non-hematopoietic cells and affect the responses against self and foreign antigens (76)*.* Unsurprisingly, to evade immunity, microbes and tumors exploit the PD1/PD-L pathway which may act in concert with other immunosuppressive signals to establish chronic infection and tumor survival (76)*.* Evidence that PD1/PD-L1 pathway is one of the main factors of tumor immune escape in humans is provided by the strategy of PD1/PD-L1 blockade. In addition to PD-L1 expression by tumors, myeloid DCs infiltrating tumors also express PD-L1. PD-L1 blockade improves myeloid DC-mediated antitumor immunity in several types of cancer (144). The blockade of this so called "checkpoint" has already been applied to clinical cancer therapy (145)*.*

## DISCUSSION AND FUTURE PERSPECTIVES BY SINGLE-CELL RNA-SEQ

The identification and the definition of DCs based on morphology, functional studies and surface markers have been subjected to many controversies and transcriptional studies have played a pivotal role in characterizing DC ontology (18). Disentangling DCs from monocytes and macrophages and understanding how DCs plasticity is shaped after stimulation or pathogen sensing remain technologically challenging because transcriptomics applied to a population of cells assumes a strict homogeneity among the cells, which does not reflect the biological reality. Genome-wide transcriptomics at the single-cell level (single-cell RNA-seq) is emerging as a powerful tool to phenotype cells and is elevating biased bulk approaches and profiling methods restricted to selected surface markers (146, 147). The revolution of singlecell RNA-seq lies in that cellular identities are no longer bounded by a restricted number of signals, but instead are inferred in an unbiased manner from an array of expressed genes. Single-cell RNA-seq can capture thousands of transcripts (148) to assess a cellular identity and enables profiling how a single-cell responds to stimulus. The response of *in vitro* differentiated DCs stimulated with three pathogenic components at the single-cell level (149, 150) revealed a dramatic difference between individual cells. The analysis demonstrated the existence of "gene modules" indicating the differential activation of gene circuits between cells where some cells are prone to mounting a precocious response, acting as "leaders" of an antiviral response. Furthermore, combining genome editing with CRISPR/Cas9-based technologies and single-cell RNA-seq helped to uncover the regulatory network controlling DC response to LPS (151). As a proof-of-concept the perturbation of *Rela, Irf9*, and *Cebpb* facilitated the decoupling of antiviral and inflammatory pathways. Such approaches, termed CRISPR-seq or Peturb-seq, are not only restricted to *in vitro* cultures, but can uncover the complexity of DC regulatory circuits *in vivo*. Notably, this approach has been used to resolve the contribution of STAT-1/2-dependent antiviral genes to myeloid cell function (151). Future applications of single-cell RNA-seq technologies should include in-depth studies of DCs exposed to pathogens, revealing their immunogenic and tolerogenic signatures.

## CONCLUSION

Activation-associated changes enabling DCs to activate adaptive immune responses are well understood. More recently, the scientific community has given greater attention to the counterregulation of these activation processes due to the clinical success of the checkpoint inhibitors, especially to the PD-1/PD-L1 molecules. Understanding of the tolerogenic mechanisms limiting inflammation is of utmost importance for therapeutic approaches that target immune pathologies, tumors and infections. As such, transcriptional profiling of tolerogenic DCs may provide insights into strategies allowing homeostasis and exploitation of own regulatory machinery by tumors and microbes.

Here, we aimed to uncover tolerogenic signatures of infla-matory or pathogen-matured DCs that included known tolerogenic markers of non-inflammatory ssmDCs. The present study addresses mainly transcriptomic studies as performed by microarray technologies of inflammatory or candidate bacteriaor helminth-induced DC signatures. This offered only a limited ability to fully identify all tolerance-associated mRNA species. However, our analysis revealed tolerogenic and anti-inflammatory genes among the expected expression of inflammatory genes. We reviewed the tolerogenic signatures of DCs exposed to different stimuli from both *in vitro* and *in vivo* studies across different host tissues and DCs subsets of man or mouse. Surprisingly, all pathogens analyzed here seem to use a rather restricted pool of target molecules for immune evasion. In the future, the possibility to quantify minute amounts of RNA species from single cells will enable analysis of much more complex regulatory networks in a wide variety of DC subsets.

## AUTHOR CONTRIBUTIONS

All authors contributed by writing parts of the text and edited the final version of the text. Figures and tables were generated by EV, DA, FE, and ML.

### REFERENCES


### FUNDING

The Helmholtz Institute for RNA-based Infection Research (HIRI) supported this work with a seed grant through funds from the Bavarian Ministry of Economic Affairs and Media, Energy and Technology (grant allocation nos. 0703/68674/5/2017 and 0703/89374/3/2017). Furthermore, this publication was funded by the German Research Foundation (DFG) and the University of Wuerzburg in the funding programme Open Access Publishing. We are grateful to Richard Brown for text editing.


type 1–polarizing program in dendritic cells. *Nat Immunol* (2005) 6:769–76. doi:10.1038/ni1223


single-Cell RNA-Seq. *Cell* (2016) 167(1883–1896):e1815. doi:10.1016/j. cell.2016.11.039

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Vendelova, Ashour, Blank, Erhard, Saliba, Kalinke and Lutz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Mechanisms of Tolerance induction by Dendritic Cells *In Vivo*

### *Hitoshi Hasegawa\* and Takuya Matsumoto*

*Department of Hematology, Clinical Immunology and Infectious Diseases, Ehime University Graduate School of Medicine, Toon, Japan*

Dendritic cells (DCs) are a heterogeneous population playing a pivotal role in immune responses and tolerance. DCs promote immune tolerance by participating in the negative selection of autoreactive T cells in the thymus. Furthermore, to eliminate autoreactive T cells that have escaped thymic deletion, DCs also induce immune tolerance in the periphery through various mechanisms. Breakdown of these functions leads to autoimmune diseases. Moreover, DCs play a critical role in maintenance of homeostasis in body organs, especially the skin and intestine. In this review, we focus on recent developments in our understanding of the mechanisms of tolerance induction by DCs in the body.

Keywords: dendritic cells, immune tolerance, regulatory T cells, development, thymus, skin, intestine

#### *Edited by:*

*John Isaacs, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Hans Acha-Orbea, University of Lausanne, Switzerland Raymond John Steptoe, The University of Queensland, Australia*

> *\*Correspondence: Hitoshi Hasegawa hitoshih@m.ehime-u.ac.jp*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 08 October 2017 Accepted: 07 February 2018 Published: 26 February 2018*

#### *Citation:*

*Hasegawa H and Matsumoto T (2018) Mechanisms of Tolerance Induction by Dendritic Cells In Vivo. Front. Immunol. 9:350. doi: 10.3389/fimmu.2018.00350*

### INTRODUCTION

Dendritic cells (DCs) represent a heterogeneous population derived from distinct hematopoietic lineages of bone marrow origin, being characterized by specific homing patterns and specialized immune functions (1–4). DCs play a pivotal role in immune responses and tolerance. Efficient priming of T cells by DCs leading to immune responses requires additional signals from the pro-inflammatory environment that can be sensed by DCs through specific pattern recognition receptors including toll-like receptors (TLRs) (mature DCs; mDCs). In contrast, lack of T cell priming in the absence of pro-inflammatory stimuli initially led to the characterization of DCs as potentially tolerogenic immature bystanders under steady-state conditions (immature DCs; iDCs). Semi-mDCs induced by apoptotic cells, by a special cytokine environment such as IL-10 and TGFβ, or by pharmacological agents also show tolerogenic properties (5–7). Tolerogenic DCs (tolDCs) in the body play an essential role in central and peripheral tolerance, resulting in resolution of ongoing immune responses and prevention of autoimmunity. DCs promote immune tolerance through negative selection of autoreactive T cells and generation of regulatory T cells (Tregs) in the thymus during acquisition of central tolerance. They also limit the differentiation of effector T cells and promote that of Tregs in the periphery through various mechanisms. Breakdown of these functions leads to autoimmune diseases. The skin and intestine act as large barrier organs to the external environment, being exposed to a wide range of environmental antigens such as foods, commensal bacteria, and pathogens. In both organs, DCs fulfill a crucial role in the balance of immune responses, leading to homeostasis and prevention of unnecessary inflammation (8). Accordingly, it is important to analyze the role of DCs in the mechanism of immune tolerance. This review presents an overview of our current understanding of the mechanisms of tolerance induction by DCs in the body.

### DC ORIGIN, DIFFERENTIATION, AND SUBSETS

Dendritic cells originate from CD34<sup>+</sup> hematopoietic progenitor cells in the bone marrow, which then differentiate further *via* common macrophage/DC progenitors into the monocyte/

macrophage lineage or common DC progenitors (CDP) (**Figure 1A**) (9). CDPs give rise to both plasmacytoid DCs (pDCs) and pre-conventional DC (cDC) progenitors. Fmslike tyrosine kinase 3 ligand (FLT3L) and its receptor, FLT3, have an instructive role in the commitment of hematopoietic progenitors to the DC-restricted lineage and their subsequent development (10, 11). FLT3L is sufficient to drive DC differentiation from mouse and human precursors, since expression of FLT3 is restricted to the DC lineage (11). Before they migrate into the bloodstream, pDCs complete their last step of maturation in the bone marrow before they migrate into the blood stream. Pre-cDC progenitors then migrate through the vascular system to their final locations in tissues or lymphoid organs, before completing their differentiation into iDCs comprising two distinct cDC subsets, CD8α+/CD103<sup>+</sup> DCs [conventional DCs 1 (cDC1s)] and CD11b<sup>+</sup> DCs [conventional DCs 2 (cDC2s)] (3). On the other hand, monocyte-derived DCs (moDCs) can differentiate from CD14<sup>+</sup> monocytes under the influence of a combination of stimuli, including GM-CSF, TNF-α, and IL-4, during tissue inflammation (12, 13). DCs are more numerous in lymphoid organs and epithelia and can express various molecular markers depending on their location. Therefore, cDC1s, cDC2s, and pDCs are present in different tissues. **Figure 1B** shows the cDC cluster to which each cell type belongs. In this context, it is necessary to consider the phenotype and specific location of DCs when addressing their function in particular tissues (9).

In mice, lymphoid organ-resident CD8α+ DCs and migratory tissue-resident CD103<sup>+</sup> DCs have a common origin (9). Their development is dependent on FLT3L, inhibitor of DNA binding protein 2, the transcription factor interferon regulatory factor 8 (IRF8), and the basic leucine zipper transcription factor ATFlike 3 (BATF3) (9). Functional and phenotypic comparison has shown that the human counterpart of murine CD8α+/CD103<sup>+</sup>

DCs is CD141 (BDCA-3)-positive DCs (14). CD8α+/CD103<sup>+</sup> DCs share common receptors such as chemokine receptor XCR1 and lectin receptor CLEC9A (15–17). CD8α+/CD103<sup>+</sup> DCs are responsible for efficient cross-presentation of antigen and stimulation of CD8<sup>+</sup> T cell immunity through secretion of IL-12, thus promoting Th1 differentiation (18, 19). In contrast, in the non-inflamed intestine, CD103<sup>+</sup> DCs in the lamina propria express high levels of TGF-β and retinaldehyde dehydrogenase 2 (RALDH2), leading to induction of Tregs (20). Therefore, CD8α+/CD103+ DCs induce either mucosal tolerance or crosspresentation-dependent CD8<sup>+</sup> T cell immunity on the basis of the local microenvironment.

In mice, CD11b<sup>+</sup> DCs are present in all major lymphoid and non-lymphoid organs. Development of CD11b<sup>+</sup> DCs depends on various transcription factors including neurogenic locus notch homolog protein 2, V-Rel avian reticuloendotheliosis viral oncogene homolog B, and IRF4 (9). The human counterpart of murine CD11b<sup>+</sup> DCs is CD1c (BDCA-1)-positive DCs (21). CD11b<sup>+</sup> DCs in the spleen express CD4<sup>+</sup> and can be subdivided according to their expression of the endothelial cell-selective adhesion molecule (22). Splenic CD11b<sup>+</sup> DCs show higher expression of MHC class II than CD8α+ DCs and can present antigen more effectively to CD4<sup>+</sup> T cells in both the steady state and during inflammation (23). In contrast, CD11b<sup>+</sup> DCs in the skin and CD11b<sup>+</sup>CD103<sup>+</sup> DCs in the lamina propria are reported to induce Treg differentiation through retinoic acid (RA) metabolism (20, 24, 25). Both CD8α+/CD103<sup>+</sup> DCs and CD11b<sup>+</sup> DCs induce tolerance or CD4<sup>+</sup> T cell proliferation according to the local microenvironment.

Murine pDCs are defined as CD11c<sup>+</sup>, MHC-II<sup>+</sup>, B220<sup>+</sup>/ CD45R+, BST2+, and SiglecH+ cells and depend on the transcription factor E2-2 for their development (26). pDCs express high levels of TLR7 and 9, which when ligated by viral products stimulate secretion of a large amount of type I IFN. pDCs can

Hasegawa and Matsumoto Tolerance Induction by DCs *In Vivo*

upregulate the expression of MHC class II, allowing the induction of T cell proliferation. On the other hand, murine pDCs induce differentiation of T cells into regulatory type 1 T (Tr1) cells (27). Naïve T cell stimulation using CpG oligonucleotidestimulated human pDCs has been reported to give rise to Tregs with suppressive properties (28). Phenotypic markers of mouse and human DC subsets are summarized in **Table 1**.

### TOLERANCE INDUCTION IN THE THYMUS AND PERIPHERY

### Central Tolerance

Dendritic cells together with medullary thymic epithelial cells (mTECs) have a critical role in inducing central tolerance in the thymus by elimination of self-antigen-reactive thymocytes and generation of Tregs (2, 29). This is supported by the fact that mice lacking DCs show marked accumulation of CD4<sup>+</sup> thymocytes without negative selection, leading to fatal autoimmunity (30). Three thymic DC subsets contribute to central tolerance: resident DCs (CD8α+ SIRPα−), migratory DCs (CD8α− CD11b<sup>+</sup> SIRPα+), and pDCs (CD11cint CD45RAint). Resident DCs that develop from thymic lymphoid precursors are the most abundant subset (>50%) and are localized mainly in the medulla (31, 32). They contribute to the elimination of autoreactive thymocytes by presenting broadly expressed selfantigens and by cross-presenting both blood-derived antigens and tissue-specific antigens from mTECs (33, 34). On the other hand, migratory DCs and pDCs develop in the periphery and migrate to the corticomedullary perivascular space, which is freely permeable to circulating antigens *via* CCR2/α4 integrin and CCR9/α4 integrin, respectively (35, 36). Through this strategic location, migratory DCs and pDCs effectively capture and present blood-derived antigens. All of the DC subsets contribute to immune tolerance by presenting self-antigens and inducing negative selection of thymocytes with high affinity for self-antigens. Then, resident DCs provide immature T cells with a distinct self-antigenic repertoire, while migratory DCs and pDCs specialize in the presentation of peripheral antigens.

Thymic DCs are also important for the development of Tregs. Resident and migratory DCs are able to induce Tregs from thymocytes *in vitro* through different mechanisms (37, 38). Resident DCs promote Treg survival *via* their expression of CD70, while CD70-deficient migratory DCs effectively induce Tregs through an undefined pathway (38). Thymic stromal lymphopoietin (TSLP) expressed by Hassall's corpuscles in the thymus medulla induces the tolerogenic phenotype on bone marrow-derived DCs, rendering them capable of converting naïve T cells into functional Tregs *in vitro* (39, 40). However, TSLP receptor-deficient mice have a normal number of Tregs in the thymus, suggesting that TSLP signaling is not essential for Treg development (29, 41). Thymic pDCs can also induce Tregs (42, 43) that are more efficient producers of IL-10 than those induced by other thymic DCs. These findings show that all of the DC subsets in the thymus are essential for the maintenance of central tolerance.

Recently, it has been examined how mTECs and CD8α+ resident DCs contribute to thymic tolerance using mice depleted of mTECs and/or resident DCs (44). Although mice depleted of resident DCs were normal and those depleted of mTECs developed liver inflammation, depletion of both resident DCs and mTECs resulted in multiorgan autoimmunity. Depletion of mTECs significantly reduced the production of thymic Tregs, but there was no additional effect on thymic Tregs when both mTECs and resident DCs were absent. Both CD4<sup>+</sup> and CD8<sup>+</sup> T cells in the thymus were increased in mice depleted of both mTECs and resident DCs. These results suggest that mTECs and resident DCs act to prevent autoimmunity through thymic T cell depletion in a cooperative manner, whereas mTECs have a non-redundant role in the production of thymic Tregs. Thus, mTECs and resident DCs have a unique role in tolerance induction that cannot be compensated for by remaining migratory DCs and pDCs.

The lymphotoxin β receptor (LTβR), a member of the TNF receptor superfamily, is a key regulator of thymic microenvironments and intrathymic tolerance, and its expression is detectable in multiple mTEC subsets (45, 46). The relationship between LTβR and coordination of mTECs and DCs for negative selection and Treg development has been recently investigated using LTβR-deficient mice (47). In LTβR-deficient mice, the thymic DC pool size was decreased due to reduced numbers of both pDCs and thymic cDCs, especially migratory DCs. In addition, LTβR-deficient mice showed a greater reduction in the numbers of CD4<sup>+</sup>CD8<sup>−</sup> thymocytes and caspase-3<sup>+</sup>CD5<sup>+</sup>CD69<sup>+</sup> thymocytes, representing cells undergoing negative selection, although they showed no change in Treg generation relative to control mice. These findings indicate that LTβR controls thymic tolerance by regulating the frequency and makeup of intrathymic DCs required for effective thymocyte negative selection rather than Treg generation.

### Peripheral Tolerance

Although thymic selection efficiently removes most self-antigenreactive T cells, some remain and migrate into the periphery. Therefore, peripheral tolerance is crucial for maintenance of immune homeostasis throughout life. Tregs of thymic origin and peripheral DCs are crucial in inducing tolerance to antigens under steady-state conditions (**Figure 2**) (1, 48). The tolDC population consists of iDCs (naïve DCs) and alternatively activated DCs (semi-mature) that exhibit resistance to maturation in the presence of an inducing signal (5, 48). iDCs derived from bone marrow constitutively migrate throughout the periphery and lymphatic systems and become distributed in peripheral tissues. iDCs are poorly immunogenic as they show low surface expression of costimulatory molecules and have only modest levels of MHC class II (1, 48). A major functional characteristic of iDCs is their capacity for endocytosis and phagocytosis, including both foreign antigens and apoptotic cells, which occurs continuously in the steady state. The maintenance of DCs in an immature state, due to the absence of maturation stimuli, is associated with tolerance through induction of T cell deletion, anergy, and polarization toward a regulatory phenotype (4). Antigen-loaded iDCs in draining secondary lymphoid organs are more effective at inducing antigen-specific Treg populations than lymphoid-resident DCs *in vivo* (49). This supports a role for migratory iDCs in promoting peripheral tolerance under


steady-state conditions. Furthermore, repetitive stimulation of T cells with iDCs can convert naïve T cells to Tregs (50, 51). Uptake of apoptotic cells polarizes DCs to a tolerogenic state, resulting in the promotion of T cell anergy and death and induction of Tregs *via* TGF-β1 secretion (52, 53). These data indicate that apoptotic cells are likely an insufficient stimulus for full DC maturation.

Dendritic cell subsets that differentiate through TLR ligands or in a specific cytokine environment might have involvement in tolerance, rather than in T cell activation (1, 48). This DC type has a semi-mature phenotype with reduced expression of MHC class II and costimulatory molecules in comparison to fully mDCs. Semi-mDCs differentiate in the presence of IL-6 or by stimulation with TLR ligands at low concentrations (54, 55). Stimulation of iDCs with TLR2 or TLR4 ligands at low concentration with the commensal bacterium *Bacteroides vulgatus*, which colonizes the intestinal tract, leads to secretion of IL-6, but not IL-12 or TNF-α (56). These DCs themselves differentiate into semi-mDCs through an autocrine loop, and exposure of iDCs to IL-6 (paracrine loop) triggers their differentiation to semi-mDCs. Furthermore, tolerogenic semi-mDCs are induced in the presence of IL-10 or TNF-α alone (57–60).

Several studies have demonstrated which DC subtypes contribute to peripheral Treg induction by combining methods of antigen delivery to DCs with diverse genetic mouse models lacking specific DC subtypes (1). Targeting of antigens to CD8α+/CD103<sup>+</sup> DCs using recombinant chimeric antibodies such as DEC205, CLEC9A, and langerin results in the induction of peripheral Tregs (49, 61–63). Moreover, peripheral Treg induction is impaired through a reduction in the proportion of CD8α+/CD103<sup>+</sup> DCs in BATF3-deficient mice and IRF8-deficient mice; both are transcription factors that are required for the development of CD8α+/CD103<sup>+</sup> DCs (64, 65). In contrast, Treg induction is restored in mice deficient in IRF4, a transcription factor that governs CD11b<sup>+</sup> DCs development. These data indicate that CD8α+/CD103<sup>+</sup> DCs rather than CD11b<sup>+</sup> DCs contribute to peripheral Treg induction.

Tolerogenic DCs show expression of immunomodulatory molecules and produce immunosuppressive factors such as IL-10, TGF-β, IL-35, and indoleamine 2,3-dioxygenase (IDO), resulting in T cell anergy and apoptosis and induction of Tregs (2, 48). The following section outlines these mechanisms.

### MECHANISMS OF IMMUNE TOLERANCE BY DCs

### T Cell Anergy

Anergy is a hyporesponsive state in which T cells remain inactive under conditions where immune activation would be undesirable, thus ensuring recognition of self-antigens and maintenance of a steady state (66). Anergy is induced in T cells that recognize antigen in the absence of costimulatory signals resulting from binding of CD28 on their surface to its ligand, CD80/CD86, on DCs. Consequently, IL-2 production is blocked, and T cells are unable to proliferate the same antigen (5, 67). Anergy can also be induced by coinhibitory signals such as programmed cell death-1 (PD-1) receptor and cytotoxic T lymphocyte antigen 4 (CTLA-4) (68, 69). PD-1 binds to PD-1 ligand (PD-L1) and PD-L2 on DCs, whereas CTLA-4 interacts with CD80/CD86 on DCs.

Several studies have shown that tolDCs can induce antigenspecific anergy through various mechanisms (70–73). tolDCs generated with IL-10 induced hyporesponsiveness of tetanus toxin (TT)-specific CD4<sup>+</sup> T cell clone toward restimulation with TT-pulsed DCs (70). This inhibition of T cell proliferation was due not to release of soluble inhibitor factors from tolDCs but to a cell contact mechanism. Tuettenberg et al. have demonstrated that induction of anergy in CD4<sup>+</sup> T cells by IL-10-modulated tolDCs was based on cell-to-cell contact through interaction of inducible T-cell costimulatory (ICOS)–ICOS ligand (ICOS-L) (71). Torres-Aguilar et al. showed that tolDCs generated with different combinations of the cytokines IL-10, TGF-β, and IL-6 induced anergy of TT-specific CD4<sup>+</sup> T cells through thrombospondin-1 expression and production of prostaglandins and adenosine by tolDCs (72). Recently, Rodriguez et al. have reported that interaction of the dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin with pathogens triggers specific signaling events that modulate DC maturation and activity, resulting in induction of T-cell anergy (73).

Induction and maintenance of T-cell anergy depend on activation of ubiquitin ligases of E3 family: Casitas B-lineage lymphoma-b (Cbl-b), Itchy homolog E3 ubiquitin protein ligase (Itch), and gene related to anergy in lymphocytes (GRAIL) (74). These enzymes act mainly through induction of proteolysis of molecules involved in TCR signaling (66, 67). T cells from Cbl-b- or Itch-deficient mice were hyperreactive and produced an increased amount of IL-2 (75–78). GRAIL was upregulated

in anergic CD4<sup>+</sup> T cells (79). In addition, the expression of these ubiquitin ligases in anergic T cells is associated with transcriptional factors, early growth response (Egr) type 2 and 3 (80). Blockade of Egr2 and Egr3 is resistant to anergy induction, while the transgenic expression of these factors suppresses TCR signaling (81, 82). In addition, anergic T cells could also act as Tregs and IL-10-producing Tr1 cells (83–86).

### Clonal Deletion

Clonal deletion, which involves the elimination of T cells through apoptosis, is an important process for maintenance of self-tolerance in the periphery (87). Apoptotic pathways can be triggered by extrinsic (receptor-dependent) and intrinsic (mitochondriadependent) stimuli (88, 89). Both pathways involve a cascade of caspases whose activation commits cells to a death outcome. The extrinsic apoptosis pathway is initiated by binding of death receptors such as Fas and TNF receptor. Ligation of FasL, TNF, or TNF-related apoptosis-inducing ligand (TRAIL) to death receptors results in activation of caspase 8 and downstream caspases and, ultimately, cell death. tolDCs can also induce naïve and memory T cell apoptosis through interaction between FasL and Fas (90), TRAIL interaction with TRAIL receptors (91), and tryptophan catabolism due to IDO production (92, 93).

Tolerogenic DCs induce extensive T cell apoptosis in a manner dependent on interaction between DC FasL and Fas expressed by the target lymphocytes. Recently, a new immunosuppressive action of DCs through the Fas signal has been reported (94). Splenic stroma-educated tolDCs expressed a high level of Fas, and Fas ligation was able to promote the inhibition of CD4+ T cell proliferation by tolDCs more significantly. Furthermore, Fas ligation preferentially induced tolDCs to produce IL-10. In addition, activated T cells promoted the secretion of more IL-10 by tolDCs through FasL. This shows that, at least from activated T cells, the Fas signal can promote the immunosuppressive action of Fas-expressing tolDCs, providing a new path for regulation of adaptive immunity by tolDCs. The cellular and molecular mechanisms of Fas-independent apoptosis of T cells induced by DCs have also been investigated by *in vitro* and *in vivo* analyses in MRL/*lpr* mice (95). This has revealed that FAS-independent T cell apoptosis can be induced by direct interaction between TRAIL receptor 2 on T cells and TRAIL on Fas-deficient DCs in MRL/*lpr* mice.

Indoleamine 2,3-dioxygenase is a rate-limiting enzyme that catalyzes the degradation of tryptophan into various metabolites, which subsequently inhibit T cell proliferation by impairing the cell cycle machinery and promoting apoptosis (48, 92, 93, 96, 97). IDO is not expressed constitutively in DCs and requires induction by various mediators including IFN-γ, TGF-β, and endotoxin (97). In rodents, CD103+ DCs in mesenteric lymph nodes (MLNs) and intestinal mucosa are known to express IDO. When IDO activity is inhibited, Th1 and Th17 cells are induced *in vivo*, preventing the development of Tregs that are specific for oral antigens (98). In contrast, tryptophan starvation increases the expression of the inhibitory receptors, immunoglobulin-like transcript 3 (ILT3) and ILT4, on DCs, leading to upregulation of Treg function. This phenomenon is associated with the GCN2 kinase-mediated stress response pathway (99).

Galectins, a family of β-galactoside-binding proteins, are expressed on DCs and also induce apoptosis of T cells (100–103). Especially, galectin 9 preferentially induces apoptosis of activated CD4<sup>+</sup> T cells through the calcium–calpain–caspase 1 pathway (101). Galectin 9 is a ligand of T cell immunoglobulin- and mucin domain-containing molecule 3 (Tim-3) expressed in Th1 cells, and the galectin 9-induced cell death in Th1 cells is dependent on Tim-3 (104).

The intrinsic apoptosis pathway can be triggered by various stimuli such as gamma irradiation, pathogens, steroid hormone, and reactive oxygen radicals and by costimulatory blockade with CTLA-4 (88, 89, 105, 106). This pathway is induced by a change in mitochondrial membrane potential provoked by the Bcl-2 family of proteins (89). Cytochrome *c* is then released by the mitochondria, binds to the apoptotic protease-activating factor 1, and forms an apoptosome that triggers the activation of caspase 9, leading to cell death. Bcl-xL and Bcl-2 impair intrinsic apoptosis by maintaining mitochondrial integrity (88). It has been reported that Bcl-xL transgenic mice were resistant to induction of transplantation tolerance through costimulatory blockade, whereas a Bcl-2/Bcl-xL inhibitor (ABT-737), in combination with costimulatory blockade and donor bone marrow cells, induced complete peripheral deletion of alloreactive T cells (105, 107, 108). On the other hand, the Bcl-2 family protein Bim present on mitochondrial membranes is involved in TCRinduced apoptosis, since deficiency of Bim impairs apoptosis of autoreactive thymocytes and mature T cells (109, 110). Taken together, these findings indicate that the intrinsic apoptosis pathway plays a critical role in not only peripheral T-cell homeostasis but also central tolerance.

### Induction of Tregs

Tolerogenic DCs can induce several subtypes of Tregs such as CD4<sup>+</sup>CD25<sup>+</sup>Foxp3<sup>+</sup> T cells and Tr1 cells. This can be achieved through a number of mechanisms, including direct cell–cell contact-dependent signaling *via* surface molecules, as well as by alteration of Treg fate *via* secretory proteins (3). DCs are known to mediate Treg generation *via* several surface molecules, including CD80/CD86 (111, 112), ICOS-L (113), ILT3, and ILT4 (114) and PD-L1 or PD-L2 (115–117). Tolerance can be induced by presentation of MHC class II antigen by DCs without any additional costimulatory signal such as CD80/CD86 and ICOS-L or in combination with a coinhibitory signal such as PD-L1/2 and ILT3/4. Furthermore, ligation of CD80/CD86 by CTLA-4 drives Treg differentiation, whereas insufficient ligation of CD80/CD86 by CD28 leads to tolerance induction. ICOS-L expressed by DCs binds to its receptor on T cells and maintains the homeostasis of Tregs.

Recently, it has been demonstrated that DCs require B- and T-lymphocyte attenuator (BTLA), an immunoglobulin domain superfamily protein, to induce Tregs (64). BTLA is specifically expressed in DEC205<sup>+</sup>CD8α+ DCs. Anti-BTLA antibody, which prevents BTLA binding to its ligand, the herpes virus entry mediator (HVEM), expressed on T cells, dramatically reduces Treg conversion. In addition, in BTLA-deficient mice, Treg induction is also decreased. BTLA mediates the upregulation of CD5 expression in T cells through HVEM engagement-increased phosphorylation of mitogen-activated protein kinase kinase (MEK). MEK increases the expression of the Cd5-positive regulator ETS1 and inhibits the expression of the Cd5-negative regulator TCF-3. CD5 is expressed on all T cells and is a well-established negative regulator of TCR signaling. CD5 promotes Treg conversion in response to self and tolerizing peripheral antigens by blocking the activation of mechanistic target of rapamycin (118).

Dendritic cells secrete many factors that are known to induce tolerance and Treg generation. IL-10, produced in the surrounding milieu under tolerogenic conditions, can trigger the development of iDCs into semi-mature tolDCs in peripheral tissues. In turn, these tolDCs acquire the ability to generate IL-10 and migrate to neighboring lymphoid organs, where IL-10 produced by DCs regulates the development and proliferation of CD4<sup>+</sup>CD25<sup>+</sup>Foxp3<sup>+</sup> T cells and Tr1 cells (48, 117). IL-10 also plays a pivotal role in regulating the expression of immune-inhibitory molecules. IL-10 upregulates the surface expression of ILT3 and ILT4 (114), PD-L1 (119), and CD95L (120) on DCs, leading to regulatory function and apoptosis.

TGF-β promotes the conversion of peripheral naïve T cells to Tregs through induction of Foxp3 expression (121). Similarly, several studies have demonstrated that DCs promote extrathymic Treg differentiation in a TGF-β-dependent manner (72, 122). Inhibition of T cell-specific TGF-β signaling *via* expression of a dominant-negative TGFβRII blocks the differentiation of Tregs (62). Coculture of Tregs with DCs results in secretion of IL-10, IL-27, and TGF-β by DCs, leading to the differentiation of Tr1 cells (123). DC-derived IL-27 suppresses the secretion of IL-1β and IL-23, induces the production of IL-10, and blocks Th17 differentiation (124). Through activation of STAT1 and STAT3, DC-derived IL-27 drives the transcription of IL-10 and activates the IL-10 promoter, thus inducing Tr1 differentiation (125). Moreover, IL-27 induces expression of the immunoregulatory molecule CD39, leading to suppression of T cell responses and autoimmunity (126). Gut-located DCs are a major source of RA, which promotes the generation of Tregs, while simultaneously inhibiting Th17 cells (127, 128).

Tolerogenic DCs secreted an anti-inflammatory cytokine, IL-35, and its production was enhanced upon stimulation with IFN-γ, LPS, or CD40 ligand (129). Conversely, IL-35 induced the conversion of cDCs to tolDCs (130). In addition to tolDCs, IL-35 is also secreted from Tregs and regulatory B cells (131–134). IL-35 is a member of the IL-12 family, consisting of IL-12α subunit p35 and IL-27β subunit Epstein–Barr virusinduced gene 3 and contributes to controlling homeostatic proliferation by suppressing T-cell proliferation and function (131, 132, 135). IL-35 could induce naïve T cells to differentiate into IL-35-producing Foxp3-induced Tregs, which maintain self-tolerance and promote infectious tolerance (131). IL-35 also plays an essential role in the balance between Th17 cells and Tregs through suppression of Th17 differentiation (132, 136). Moreover, a recent study has reported that mice vaccinated with IL-35-producing DCs showed promotion of tumor growth and amelioration of autoimmune encephalitis (130). Taken together, these findings suggest that IL-35 plays a significant role in the regulation of immune tolerance.

Plasmacytoid DCs induce Treg differentiation in the peripheral lymph nodes (137). Although, in the steady state, pDCs express very low levels of MHC class II and costimulatory molecules, activated pDCs upregulate MHC class II and migrate to the T cell area to induce Treg generation. Type I IFN and IL-10 produced by pDCs contribute to Treg generation. pDCs can also produce IDO and express PD-L1, and this is correlated with an increase of Treg numbers (115, 138).

Regulatory T cells are also able to affect DC function. Mutual interaction between DCs and Tregs is required for maintenance of immune tolerance: tolDCs induce Tregs, and conversely Tregs prepare DCs for an immunosuppressive role, thus extending the immunosuppressive function of Tregs. For example, IL-10 and TGF-β locally secreted from Tregs are able to suppress the maturation of DCs and render them tolerogenic (139). Another pivotal role of Tregs is their immunosuppressive effect when in contact with DCs. Recently, individual Treg–DC interaction events in lymph nodes have been examined *in vivo* using imaging techniques (140). Endogenous Tregs exhibited enhanced adhesion to antigen-presenting DCs, thus mediating the activation of conventional CD4<sup>+</sup> T cells (T conv cells) in draining lymph nodes. Subsequent experiments using adoptive transfer of Tregs and MHC class II-deficient DCs have demonstrated that this increased Treg–DC adhesion can be promoted only by exposure to IL-2 without requiring MHC recognition. Importantly, physical contact with polyclonal Tregs significantly reduces the ability of DCs to form stable conjugates with cognate T conv cells *in vivo*, resulting in impaired T cell priming. These results suggest that Tregs of any TCR specificity can suppress DCs in a contact-dependent and MHC class II-independent manner. Moreover, the dynamic cytoskeletal components underlying contact-dependent Treg-mediated DC suppression have been analyzed using imaging (141). This revealed that Tregs, rather than T conv cells, exhibited strong intrinsic adhesiveness to DCs. This adhesion of Tregs caused sequestration of Fascin-1, an actin-bundling protein essential for the formation of immunological synapses and skewed Fascin-1-dependent actin polarization in DCs toward the Treg adhesion zone. This sequestration caused DCs to become lethargic, leading to reduced T cell priming. Mechanisms of immune tolerance by tolDCs are summarized in **Table 2**.

### INDUCTION OF IMMUNE TOLERANCE BY DCs IN THE SKIN AND INTESTINE

### Skin

The skin is the largest barrier organ separating the internal milieu from the external environment. It is exposed to not only physical stress but also a huge number of environmental antigens, including chemicals, commensal bacteria, and pathogens. Therefore, the immune system of the skin must detect and discriminate between these diverse antigens and induce appropriate

#### Table 2 | Mechanisms of DC-induced immune tolerance.

#### T cell anergy


#### Clonal deletion (apoptosis)


#### Induction of Tregs


#### Other suppressive mechanisms

#### • Mutual interaction between DCs and Tregs

*CTLA-4, cytotoxic T lymphocyte antigen 4; PD-1, programmed cell death-1; PD-L1, programmed cell death-1 ligand; ICOS-L, inducible T-cell costimulatory ligand; DC-SIGN, dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin; TRAIL, TNF-related apoptosis-inducing ligand; IDO, indoleamine 2,3-dioxygenase; Tim-3, T cell immunoglobulin- and mucin domain-containing molecule 3; ILT3, immunoglobulin-like transcript 3; BTLA, B- and T-lymphocyte attenuator; HVEM, herpes virus entry mediator; RA, retinoic acid; Tregs, regulatory T cells; DCs, dendritic cells.*

tolerogenic or protective responses (142). The skin consists of two anatomically distinct layers, the epidermis and dermis, which are separated by a basement membrane. Langerhans cells (LCs), expressing the C-type lectin langerin (CD207), represent the sole tissue-resident DC population in the epidermis, while several subsets of DCs are resident in the dermis, including CD103<sup>+</sup> cDCs, CD11b<sup>+</sup> cDCs, and CD103<sup>−</sup> CD11b<sup>−</sup> cDCs. In addition, during inflammation, moDCs are recruited to the dermis (8). LCs are very motile, although most abundant in the spinous layer of the epidermis. LCs constantly migrate from the skin to draining lymph nodes even under steady-state conditions. In general, LCs induce effector-type immunity to pathogens and foreign proteins (143, 144). On the other hand, recent evidence suggests that LCs might be involved in peripheral tolerance induction. In a murine model of contact hypersensitivity (CHS), it has been demonstrated that the absence of LCs leads to an increase in the number of hapten-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cells. This has revealed a mechanism of immune regulation in the skin whereby interplay with CD4<sup>+</sup> T cells enables LCs to suppress antigen-specific responses through IL-10 production (145). Another CHS study involving experimental depletion and adoptive transfer has demonstrated that LCs confer protection against CHS development through a mechanism involving both anergy and deletion of allergen-specific CD8<sup>+</sup> T cells and activation of a T cell population identified as ICOS<sup>+</sup>CD4<sup>+</sup>Foxp3<sup>+</sup> Tregs (146). In a *Leishmania* infection model in mice, it has been demonstrated that the absence of LCs leads to reduced Treg immigration, indicating a suppressive role of epidermal LCs through promotion of Tregs (147). Recently, the use of a transgenic mouse model has facilitated analysis of the immune functions of LCs *in vivo* without any alteration in the complex composition of skin DC subsets (148). When ovalbumin was presented by steadystate LCs or by activated LCs, they developed antigen-specific CTL tolerance due to an increase in Tregs or the CTL memory response, respectively. This decision-making depends on the condition of the presenting LCs.

All dermal cDCs are derived from hematopoietic stem cellderived progenitor cDCs, pre-cDCs, that continuously repopulate the dermis. Mainly, four subsets of cDCs are resident in the dermis: langerin<sup>+</sup>CD103<sup>+</sup>, langerin<sup>−</sup>CD11b<sup>+</sup>, langerin<sup>−</sup>CD11b<sup>−</sup> and CD103<sup>−</sup>CD11b<sup>−</sup> cDCs (8). Langerin<sup>+</sup>CD103<sup>+</sup> cDCs include 10–20% dermal DCs and express XCR1. Langerin<sup>+</sup>CD103<sup>+</sup> cDCs efficiently cross-present viral and self-antigens, and mice deficient in langerin<sup>+</sup>CD103<sup>+</sup> show impaired priming of CD8<sup>+</sup> T cells (18, 148, 149). On the other hand, langerin<sup>+</sup>CD103<sup>+</sup> cDCs are capable of generating Tregs (150). When langerin<sup>+</sup>CD103<sup>+</sup> cDCs were depleted in Lang-DTR mice, anti-DE205-mediated antigen-specific delivery to DCs was no longer able to induce antigen-specific Tregs, resulting in loss of immune tolerance (49, 150). Human BDCA-3<sup>+</sup> DCs, the counterpart of murine langerin<sup>+</sup>CD103<sup>+</sup> cDCs, have been shown to produce large amounts of IL-10 and to present self-antigens and induce Tregs (151). Langerin<sup>−</sup>CD11b<sup>+</sup> cDCs include 70–80% dermal DCs. Langerin<sup>−</sup>CD11b<sup>+</sup> cDCs can prime naïve CD4<sup>+</sup> T cells to undergo Th2 differentiation and play a role in the immune response through IL-23/IL-17 signaling (152–156). RA-producing CD11b<sup>+</sup> cDCs can induce Tregs upon migration to draining lymph nodes (157). Interestingly, it has recently been demonstrated that targeted deletion of IKKβ, a major activator of NF-κB, in DCs prevents the accumulation of skin migratory DCs in draining lymph nodes under steady-state conditions, thus compromising Treg conversion (158). Thus, NF-κB signaling appears to play a critical role in immunity and tolerance, as NF-κB is a key regulator of TLR-induced DC maturation and production of pro-inflammatory cytokines. Taken together, the evidence suggests that LCs and dermal DCs play a pivotal role in not only the immune response but also immune tolerance in the skin.

### Intestine

In the intestinal tract, IgA antibody production and the immune response by effector T cells are induced against pathogenic microorganisms. On the other hand, immune tolerance is also induced to avoid unnecessary inflammatory responses to food antigens and commensal bacteria. DCs play a critical role in the intestinal immune response and immune tolerance. In the intestinal mucosa, DCs are scattered diffusely throughout the intestinal lamina propria, within gut-associated lymphoid tissues including Peyer's patches and solitary intestinal lymphoid tissues (SILT), and also in intestinal draining lymph nodes such as MLNs (159, 160). Migration of intestinal DCs plays an important role in immune surveillance and homeostasis of the gut. Migratory intestinal DCs can be derived from three different sites: the lamina propria, Peyer's patches, and SILT presented within the small intestinal mucosa.

The murine small intestinal lamina propria contains at least three distinct populations of cDCs: CD103<sup>+</sup>CD11b<sup>−</sup>, CD103<sup>+</sup>CD11b<sup>+</sup>, and CD103<sup>−</sup>CD11b<sup>+</sup> DCs. These three cDC subtypes are able to migrate *via* afferent lymphatics to the draining MLNs, a process that requires CCR7 signaling (161). cDCs in the lamina propria acquire antigens by handover, either from epithelial goblet cells or CX3CR1high macrophages (162, 163). CD103<sup>+</sup>CD11b<sup>+</sup> cDCs in the lamina propria migrate into the epithelium and capture pathogenic bacteria presented in the gut lumen by extending their dendrites (164). Upon antigen uptake, lamina propria cDCs enter the T cell zone of gut-draining MLNs for DC–T cell interaction. Double negative CD103<sup>−</sup>CD11b<sup>−</sup> cDCs, which might exclusively originate from Peyer's patches and SILT, have also been reported to carry antigens *via* afferent lymphatics to MLNs (165).

For adaptive immunity, migratory DC subsets derived from the lamina propria show some specialization in the generation of distinct Th cell subsets. CD103<sup>+</sup>CD11b<sup>+</sup> cDCs activated with TLR produce high levels of IL-6 and induce IL-6-dependent Th17 differentiation, while CD103<sup>+</sup>CD11b<sup>−</sup> and CD103<sup>−</sup>CD11b<sup>+</sup> cDCs can drive Th1 differentiation rather than CD103<sup>+</sup>CD11b<sup>+</sup> cDCs (165–168). On the other hand, cDCs in the lamina propria and MLNs, especially the population of CD103<sup>+</sup> cDCs, play a central role in enforcing tolerance to food antigens and commensal bacteria under steady-state conditions. Many studies have shown that intestinal CD103<sup>+</sup> cDCs highly induce Tregs through a mechanism mediated by TGF-β and RA (20, 159, 169). Tregs are induced by TGF-β, and RA enhances Treg induction only in the presence of TGF-β. TGF-β is secreted by intestinal CD103<sup>+</sup> cDCs, Tregs, and intestinal stromal cells. Intestinal CD103<sup>+</sup> cDCs highly express RALDH2, which convert vitamin A to RA, resulting in RA production from CD103+ cDCs. In addition to Treg induction, RA also induces gut-homing receptors, CCR9 for the small intestinal chemokine CCL25 and α4β7 integrin for the mucosal vascular addressin, MAdCAM1. In addition, it has recently been demonstrated that RA acts cell intrinsically in developing guttropic pre-mucosal DCs to trigger differentiation and drive the specialist role of intestinal CD103<sup>+</sup>CD11b<sup>−</sup> and CD103<sup>+</sup>CD11b<sup>+</sup> cDCs (170). Overall, the evidence suggests that through production of RA and TGF-β, intestinal CD103+ cDCs induce the differentiation of Tregs and home them into the intestinal mucosa to control tolerance.

A conditional knockout approach allowing the deletion of specific subsets of CD103<sup>+</sup> cDCs has demonstrated that intestinal CD103<sup>+</sup>CD11b<sup>−</sup> cDCs possess the greatest capacity to induce Treg differentiation, while CD11b<sup>+</sup> DC subsets are rather inefficient (65). Moreover, PD-L1 and PD-L2 expression by MNL DCs has been implicated in the induction of oral tolerance *via* regulation of the Treg compartment (171). Comparison among four distinct DC subsets in MLNs—CD103+CD11b+PD-L1high, CD103<sup>+</sup>CD11b<sup>−</sup>PD-L1high, CD103<sup>+</sup>CD11b<sup>−</sup>PD-L1low, and CD103−CD11b+PD-L1int—has shown that CD103+CD11b−PD-L1high DCs have a high capacity to induce Treg differentiation through TGF-β signaling (172). It has been reported that αvβ8 integrin, an activator of latent TGF-β, is expressed preferentially on

CD103<sup>+</sup>CD11b<sup>−</sup> cDCs in the lamina propria and MLNs and that αvβ8 integrin-expressing DCs induce Tregs *via* TGF-β activation (173). However, it remains unclear whether CD103<sup>+</sup>CD11b<sup>+</sup> PD-L1high DCs and αvβ8 integrin-expressing DCs represent the same population. Under steady-state conditions, CD103<sup>+</sup> cDCs in the lamina propria are tolerogenic. In contrast, under inflammatory conditions, CD103<sup>+</sup> cDCs in the MLNs are immunogenic. MLN CD103<sup>+</sup> cDCs from colitic mice have been shown to trigger Th1 responses with high levels of IL-6 production (174). During intestinal inflammation, MLN CD103+ cDCs acquire these proinflammatory properties with no ontogenetic changes. Therefore, as well as DCs in the skin, CD103<sup>+</sup> cDCs in the intestine can also be immunogenic or tolerogenic due to the microenvironment.

Some nutrients other than vitamin A are known to exert notable effects on intestinal tolerance. Tryptophan is a dietary element required for the IDO-dependent tolerogenic effects of intestinal DCs (175). Dietary tryptophan is metabolized into agonists for the aryl hydrocarbon receptor (AhR) through a series of cooperative biochemical reactions catalyzed by enzymes provided by gut commensal bacteria and the host (176). Tryptophan-derived AhR ligands induce the production of IL-10 and IL-27 by DCs, favoring the generation of Tregs and Tr1 cells. Diet-derived lipid mediators can activate peroxisome proliferator-activated receptor γ (PPARγ), and DCs exposed to PPARγ can be tolerogenic (7, 177). The gut mucosa is permeated by a complex nervous system and therefore exposed to the local release of neurotransmitters. Vasoactive intestinal peptide (VIP) is produced by intestinal enteroendocrine and immune cells and acts as a vasodilator and regulator of epithelial permeability. VIP suppresses LPS-induced DC maturation and promotes the differentiation of Tregs and Tr1 cells (178, 179). Taken together, these findings suggest that metabolites provided by the diet and gut flora act in concert with endogenous signals to regulate the ability of DCs to control T cell responses and tissue homeostasis. Mechanisms of induction of tolDCs in the intestine are summarized in **Figure 3**.

## CONCLUSION

Dendritic cells play a pivotal role in immune tolerance and homeostasis in the body. In this review, we present an overview of our current understanding of the mechanisms of tolerance induction by DCs in the body: DC origin, differentiation, and subsets; tolerance induction in the thymus and periphery; mechanisms of immune tolerance by DCs; and induction of immune tolerance by DCs in the skin and intestine. However, since analysis of the abovementioned mechanisms in health and disease is still insufficient, further studies are needed. A thorough understanding of the mechanisms that control immune tolerance will guide the development of novel strategies for the treatment of autoimmunity.

## AUTHOR CONTRIBUTIONS

HH designed the study and drafted manuscript. TM edited the manuscript.

## FUNDING

This work was supported by grants-in-aid (15K09551) for scientific research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

### REFERENCES


regulatory T cell development via the CD27-CD70 pathway. *J Exp Med* (2013) 210(4):715–28. doi:10.1084/jem.20112061


dendritic cells drive mucosal T helper 17 cell differentiation. *Immunity* (2013) 38(5):958–69. doi:10.1016/j.immuni.2013.03.009


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Hasegawa and Matsumoto. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Donor-Derived regulatory Dendritic cell infusion Maintains Donor-reactive cD4**+**cTla4hi T cells in non-human Primate renal allograft recipients Treated with cD28 co-stimulation Blockade

#### *Edited by:*

*Yair Reisner, Weizmann Institute of Science, Israel*

#### *Reviewed by:*

*Thomas Wekerle, Medizinische Universität Wien, Austria Baojun Zhang, Duke University, United States*

> *\*Correspondence: Angus W. Thomson thomsonaw@upmc.edu*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 31 October 2017 Accepted: 29 January 2018 Published: 19 February 2018*

#### *Citation:*

*Ezzelarab MB, Lu L, Shufesky WF, Morelli AE and Thomson AW (2018) Donor-Derived Regulatory Dendritic Cell Infusion Maintains Donor-Reactive CD4+CTLA4hi T Cells in Non-Human Primate Renal Allograft Recipients Treated with CD28 Co-Stimulation Blockade. Front. Immunol. 9:250. doi: 10.3389/fimmu.2018.00250*

*Mohamed B. Ezzelarab1 , Lien Lu1 , William F. Shufesky1 , Adrian E. Morelli1,2 and Angus W. Thomson1,2\**

*1Department of Surgery, Starzl Transplantation Institute, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States, 2Department of Immunology, University of Pittsburgh School of Medicine, Pittsburgh, PA, United States*

Donor-derived regulatory dendritic cell (DCreg) infusion before transplantation, significantly prolongs renal allograft survival in non-human primates. This is associated with enhanced expression of the immunoregulatory molecules cytotoxic T-lymphocyte-associated antigen (Ag) 4 (CTLA4) and programmed cell death protein 1 (PD1) by host donor-reactive T cells. In rodents and humans, CD28 co-stimulatory pathway blockade with the fusion protein CTLA4:Ig (CTLA4Ig) is associated with reduced differentiation and development of regulatory T cells (Treg). We hypothesized that upregulation of CTLA4 by donor-reactive CD4+ T cells in DCreg-infused recipients treated with CTLA4Ig, might be associated with higher incidences of donor-reactive CD4+ T cells with a Treg phenotype. In normal rhesus monkeys, allo-stimulated CD4+CTLA4hi, but not CD4+CTLA4med/lo T cells exhibited a regulatory phenotype, irrespective of PD1 expression. CTLA4Ig significantly reduced the incidence of CD4+CTLA4hi, but not CD4+CTLA4med/lo T cells following allo-stimulation, associated with a significant reduction in the CD4+CTLA4hi/CD4+CTLA4med/lo T cell ratio. In CTLA4Ig-treated renal allograft recipient monkeys, there was a marked reduction in circulating donor-reactive CD4+CTLA4hi T cells. In contrast, in CTLA4Ig-treated monkeys with DCreg infusion, no such reduction was observed. In parallel, the donor-reactive CD4+CTLA4hi/CD4+CTLA4med/lo T cell ratio was reduced significantly in graft recipients without DCreg infusion, but increased in those given DCreg. These observations suggest that pre-transplant DCreg infusion promotes and maintains donor-reactive CD4+CTLA4hi T cells with a regulatory phenotype after transplantation, even in the presence of CD28 co-stimulation blockade.

Keywords: regulatory T cells, dendritic cells, co-stimulation blockade, renal allografts, non-human primates

## INTRODUCTION

Despite major advances in clinical organ transplantation, an important limitation remains patients' life-long dependency on immunosuppressive drugs with associated increased risk of morbidity and mortality (1). Based on compelling pre-clinical evidence, regulatory immune cell therapy offers considerable potential for the development of protocols that may promote clinical transplant tolerance (2–5). Dendritic cells (DC) are uniquely well-equipped integrators and regulators of innate and adaptive immunity. They can promote antigen (Ag)-specific tolerance (3, 6–8) as well as regulate memory T cell (Tmem) responses (9–11), a major barrier to the induction of transplantation tolerance (12–15).

Recently, testing of CD28 co-stimulation blockade (Co-SB) using belatacept [cytotoxic T lymphocyte Ag 4:Ig (CTLA4Ig)] and a calcineurin inhibitor-free regimen (steroids and mycophenolate mofetil) in kidney transplant recipients, has resulted in an increased incidence of acute cellular rejection within 1 year after transplantation, despite superior graft function (16, 17). While allo-reactive Tmem are known to be Co-SB-resistant (18, 19), there is recent evidence that CTLA4Ig may reduce donor-reactive regulatory T cells (Treg)/Tmem ratios after transplantation and prevent Treg-dependent transplantation tolerance (20, 21).

We have reported previously (22) using a pre-clinical nonhuman primate (NHP) model, that administration of donorderived, maturation-resistant, regulatory dendritic cells (DCreg) 1 week before transplantation, can significantly prolong major histocompatibility complex (MHC)-mismatched renal allograft survival in CTLA4Ig-treated recipients. Although no increase in regulatory T cells (Treg) was detected after transplantation, we observed increased Treg to Tmem ratios in the graft recipients given DCreg infusion. This was associated with upregulation of the co-inhibitory molecule CTLA4 (CD152) and programmed cell death protein 1 (PD1) by host CD4<sup>+</sup> and CD8<sup>+</sup> T cells in response to donor but not third party stimulation, suggesting donor-specific regulation of T cell responses (23).

Cytotoxic T-lymphocyte-associated Ag 4 is a critical negative regulator of T cell responses (24). It is expressed constitutively in Treg but only upregulated in conventional T cells after activation. CTLA4 deficiency in mice results in a lethal lymphoproliferative disease (25, 26), hence it is essential for maintaining T cell tolerance to self-Ags (27). In rodent models of organ transplantation, CTLA4 blockade accelerates acute rejection (28). Also, importantly, donor-reactive CD4<sup>+</sup>CD25<sup>+</sup> Treg expressing high levels of CTLA4 promote allograft survival (29), an effect that is dependent on exposure to donor Ag before transplantation. Further, there is recent evidence that CTLA4 plays a critical role in Treg suppressive function (30). Moreover, CTLA4 can promote T cell suppressive function, even in the absence of forkhead box p3 (Foxp3) expression (31).

We hypothesized that upregulation of CTLA4 by rhesus allo-reactive CD4<sup>+</sup> T cells might be associated with increased incidences of Treg, and that pre-transplant infusion of donorderived DCreg might promote and maintain donor-reactive Treg after renal transplantation, despite host treatment with CTLA4Ig. In this study, we examined the relationship between expression of CTLA4 and a Treg phenotype by allo-stimulated rhesus monkey CD4+ T cells. We also investigated the influence of CD28 Co-SB on Treg development both *in vitro* and in CTLA4Ig-treated kidney allograft recipient monkeys, with or without DCreg infusion.

### MATERIALS AND METHODS

### Experimental Animals

Indian male juvenile rhesus macaques (*Macacca mulatta*; 5–7 kg), obtained from the NIAID-sponsored colony (Yemasse, S.C.) were maintained in the Non-Human Primate Research Facility of the Department of Laboratory Animal Resources at the University of Pittsburgh School of Medicine. All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Experiments were conducted according to the guidelines set forth in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Specific environment enrichment was provided.

### Donor Leukapheresis and DCreg Generation

Leukapheresis and generation of donor-derived DCreg from circulating monocytes were performed as described (22, 32). Briefly, prospective transplant donors underwent cytokine treatment comprising granulocyte-macrophage colony-stimulating factor (GM-CSF) for 4 days, followed by granulocyte (G)-CSF for 4 days. Leukapheresis was performed using a dedicated COBE® Spectra Apheresis System (Lakewood, CO, USA). Leukapheresis products were processed and stored in liquid N2 until needed for DCreg generation. DCreg were generated from purified CD14<sup>+</sup> cells in recombinant human (rhu) GM-CSF + rhu IL-4 over 7 days, with the addition of Vitamin D3 on days 1 and 5, and rhu IL-10 on day 5 as previously described (32).

### DCreg Infusion, Renal Transplantation, and Immunosuppression

Renal transplantation was performed as described (22). Briefly, bilateral nephrectomy of native kidneys was performed before graft insertion and recipient pairs—i.e., control (no DCreg infusion; *n* = 4) and experimental (DCreg infusion; *n* = 4) received MHC-mismatched kidney grafts from the same donor. In the experimental group (**Table 1**), DCreg (3.5–10 × 106 /kg) were infused intravenously, 7 days before transplantation. All recipients in the control and DCreg groups received CTLA4Ig (abatacept; Bristol-Myers Squibb; Princeton, NJ, USA; 12.5 mg/ kg i.v.) on day −7 and day −4, to further minimize risk of host sensitization. Four recipients (two in each group) received shortterm co-stimulation blockade (CTLA4Ig; 20 mg/kg on days −1, 0, 2, 4, 7, and 10), while two recipients (two in each group) received long-term Co-SB (CTLA4Ig; 20 mg/kg on days −1, 0, 3, 7, 10, 14, 21, and 28, then 10 mg/kg on days 35, 42, 49, and 56). Intramuscular rapamycin (LC laboratories, Woburn, MA, USA) was given daily, starting on day −2 for 6 months. Whole blood trough levels were measured twice weekly and maintained between 10 and 15 ng/ml for the first month, between 5 and



*a Details of short-term and long-term administration of cytotoxic T-lymphocyte-associated antigen (Ag) 4:Ig (CTLA4Ig) are explained in Section "Materials and Methods." bData published originally in The American Journal of Transplantation (22).*

*c Donor-derived DCreg infused on day* −*7.*

10 ng/ml for the subsequent 4 months, and between 1 and 5 ng/ml for the sixth month. Immunosuppressive therapy was withdrawn completely at 6 months (22).

### Mixed Leukocyte Reactions (MLRs)

Peripheral blood mononuclear cells (PBMC) were isolated from normal rhesus monkeys for *in vitro* studies. Unlabeled or carboxyfluorescein succinimidyl ester (CFSE; Molecular Probes, Eugene, OR, USA)-labeled PBMC were used as responders and CD2<sup>+</sup> T cell-depleted allogeneic irradiated PBMC as stimulators, at 1:1 ratio. In some MLRs, CTLA4Ig was added (1 µg or 100 µg/ ml) at the start of the culture. PBMC were also isolated before and after transplantation [post-operative days (POD) 28–56, unless otherwise specified], and co-cultured with either donor or third party cells. Data were acquired using an LSR II flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) and analyzed with FlowJo software (Tree Star, San Carlos, CA, USA).

### Phenotypic Analysis of Allo-Reactive T Cells

The following fluorochrome-labeled monoclonal antibodies were used as described (22, 33) for cell surface or intracellular staining of rhesus T cells: CD3 PerCP-Cy5.5, CD4 APC-H7, CD28 APC-H7, CD127 (IL-7Rα) PE, CD45RA PE-Cy7, CTLA4/CD152 APC, and CTLA4/CD152 VB450 (all from BD Biosciences, San Jose, CA, USA), CD8α AF700, CD25 AF700, and Foxp3 VB421 (all from Biolegend, San Diego, CA, USA), and PD1/CD279 PE (from eBioscience, San Diego, CA, USA). Following surface staining for CD3, CD4, CD8, CD25, CD28, CD127, and PD1, cells were fixed and permeabilized for 45 min at 4°C using Fixation/ Permeabilization buffer (eBioscience™; ref 00-5123-43). After fixation/permeabilization, cells were stained for CTLA4 and Foxp3. No antibodies were added to the co-cultures.

## Immunofluorescence Staining of Kidney Allografts

Tissues were collected from one recipient with no DCreg infusion (control group) on the day of euthanasia and from one recipient with DCreg infusion (experimental group) on POD 28 by open biopsy of the kidney graft. Tissues were embedded in O.C.T. (Miles), snap-frozen, and stored at −80°C. Cryostat sections (8–10 µm) were mounted on slides pre-coated with Vectabond (Vector) then fixed in 96% ethanol and allowed to dry. Sections were blocked successively with 5% goat serum and an avidin/ biotin blocking kit (Vector). Next, sections were incubated with anti-human CD4 Ab (Dako; Clone 4B12, 1:100, overnight, 10°C), followed by Alexa Fluor 555-goat anti-mouse IgG (Molecular Probes, 1:400, 1 h, RT). The slides were then blocked with mouse irrelevant IgG1 (BD Biosystems, 1:100, 1 h, RT) and incubated successively with biotin anti-human CTLA4 (CD152) (clone BNI3, BD Biosystems, 1:100, 1 h, RT), followed by Streptavidin Dylight 488 (Jackson Immunoresearch, 1:400, 1 h, RT). Cell nuclei were stained with DAPI (Molecular Probes).

### Statistical Analyses

The significances of differences between groups were determined using Kruskal–Wallis one-way analysis of variance or Mann–Whitney *U* test, as appropriate. Significance was defined as *p* < 0.05.

### RESULTS

### CTLA4 and PD1 Expression by CD4**<sup>+</sup>** T Cells in Renal Allografts 1 Month Post-Transplant

We have shown previously (22) that Tmem in rhesus renal allograft recipients given DCreg before transplant upregulate CTLA4 and PD1 expression in response to *ex vivo* donor but not third party stimulation. Additionally, graft-infiltrating CD8<sup>+</sup> T cells were characterized by higher expression of CTLA4 and PD1 (33). Here, we hypothesized that graft-infiltrating CD4<sup>+</sup> T cells in monkeys given DCreg infusion would also express high levels of CTLA4 and PD1. Thus, we examined the expression of CTLA4 and PD1 by graft-infiltrating CD4+ T cells 28 days posttransplant in monkeys given no DCreg infusion or DCreg infusion (**Figure 1**). With no DCreg infusion, graft-infiltrating CD4<sup>+</sup> T cells showed minimal CTLA4 and PD1 expression. In contrast, strong expression of CTLA4 and PD1 by graft-infiltrating CD4<sup>+</sup> T cells was observed in the recipient given DCreg infusion before transplantation. This observation is consistent with the

upregulation of both CTLA4 and PD1 by circulating T cells from DCreg-infused graft recipients following *ex vivo* donor stimulation 28 days post-transplant (22).

### CTLA4 and PD1 Upregulation Correlates with Increased Regulatory CD4**+** T Cell Marker Expression Following Allo-Stimulation *In Vitro*

While the inhibitory molecules CTLA4 and PD1 are considered markers of T cell activation, exhaustion, and regulation (34, 35), they are also associated with the induction of Treg (36, 37). We examined their expression by normal rhesus CD4<sup>+</sup> T cells together with Treg phenotype analysis based on CD25, CD127, and Foxp3 expression following their allo-stimulation for 5 days in CFSE-MLR. In comparison with non-proliferating cells, proliferating allo-reactive CD4<sup>+</sup> T cells significantly upregulated CTLA4, PD1, CD25, and Foxp3 expression (*p* < 0.05), but significantly downregulated CD127 expression (*p* < 0.05) (**Figure 2**). These observations indicate that upregulation of CTLA4 and PD1 by allo-reactive CD4<sup>+</sup> T cells is associated with an increased Treg phenotype.

### High CTLA4, but Not PD1 Expression Is Associated with a Regulatory CD4**+** T Cell Phenotype in Normal Rhesus Monkeys

Cytotoxic T-lymphocyte-associated Ag 4 expression is critical for optimal Treg function and their ability to suppress T cell responses to allo-Ag (38, 39). In addition, Foxp3 is known to upregulate CTLA4 expression (40). We investigated whether

Foxp3hi, and CD127lo, as well as CTLA4 and PD1 were evaluated in carboxyfluorescein succinimidyl ester (CFSE)-mixed leukocyte reaction. Responder peripheral blood mononuclear cells (PBMC) were co-cultured with allogeneic, T cell-depleted PBMC for 5 days. (B) Combined data from four individual monkeys are shown. Values are means + 1SD. \**p* < 0.05; only significant values are shown.

expression of CTLA4 and/or PD1 by rhesus CD4+ T cells correlated with a higher frequency of Treg after allo-stimulation. As shown in **Figure 3A**, the incidence of CD25hiFoxp3hi CD4<sup>+</sup> T cells was significantly higher in the CTLA4<sup>+</sup>PD1<sup>−</sup> and CTLA4<sup>+</sup>PD1<sup>+</sup> populations than in the CTLA4−PD1+ and CTLA4−PD1− populations. Thus, irrespective of PD1 expression, CTLA4 expression was associated with a higher incidence of CD25hiFoxp3hi Treg.

Next, we evaluated the expression of Treg markers (Foxp3, CD25, and CD127) in relation to CTLA4 and PD1 expression by total (proliferating and non-proliferating) CD4<sup>+</sup> T cells (**Figure 3B**). CD4+CTLA4hi cells exhibited significantly higher levels of Foxp3 and CD25 than CD4<sup>+</sup>CTLA4med/lo and CD4+CTLA4neg T cells. Conversely, CD127 was expressed at the lowest level by CD4<sup>+</sup>CTLA4hi T cells compared with CD4+CTLA4med/lo and CD4<sup>+</sup>CTLA4<sup>−</sup> T cells. In contrast, CD4<sup>+</sup>PD1hi cells did not express higher levels of Foxp3 or CD25 than PD1med/lo or PD1neg CD4<sup>+</sup> T cells. On the other hand, CD4<sup>+</sup> PD1<sup>−</sup> T cells expressed higher levels of CD127 than PD1hi and PD1med/lo CD4<sup>+</sup> T cells. These data indicate that only high CTLA4 expression is associated with a Treg phenotype in normal rhesus monkeys.

### CD28 Co-SB Reduces CTLA4 Expression More Markedly Than PD1 Expression after Allo-Stimulation *In Vitro*

Next, we examined whether CD28 Co-SB could reduce CTLA4 and PD1 expression by allo-stimulated rhesus CD4<sup>+</sup> T cells. As shown in **Figure 4**, in the absence of CTLA4Ig, both PD1 and

Irrespective of PD1 expression, CTLA4+PD1+ and CTLA4+PD1− CD4+ T cells exhibit a high incidence of dual CD25hiFoxp3hi expression. Combined data from eight different monkeys are shown in the graph (right). (B) Treg were characterized based on expression of forkhead box p3 (Foxp3), CD25, and CD127 and evaluated in CD4+CTLA4hi, CD4+CTLA4med/lo, CD4+PD1hi, CD4+PD1med/lo T cell populations after allo-stimulation in normal rhesus monkeys. Graphs show means + 1 SD from three independent experiments; \**p* < 0.05; \*\**p* < 0.01; only significant values are shown.

CTLA4 expression was upregulated significantly by total (proliferating and non-proliferating) CD4<sup>+</sup> T cells. In the presence of CTLA4Ig, reduced CD4<sup>+</sup> T cell proliferation was associated with concentration-dependent reductions in the percentage of CTLA4<sup>+</sup>CD4<sup>+</sup> T cells. In contrast, the incidence of PD1<sup>+</sup>CD4<sup>+</sup> T cells not affected significantly.

### CD28 Co-SB Reduces Allo-Reactive CD4**+**CTLA4hi T Cells *In Vitro*

CD28 co-stimulation is essential for Treg differentiation, function, and homeostasis (41–43). In humans, Co-SB with CTLA4Ig decreases the incidence of circulating Treg (44, 45) and impairs Treg expansion (21). Thus, we questioned whether reduced CTLA4 expression by rhesus CD4<sup>+</sup> T cells following allo-stimulation in the presence of CTLA4Ig was due mainly to enhanced suppression of the CTLA4hi population. Following allo-stimulation in MLR, the frequencies of CD4<sup>+</sup>CTLA4med/lo and CD4+CTLA4hi among total CD4+ T cells were increased significantly (**Figures 5A,B**). However, these increases were more pronounced for CD4<sup>+</sup>CTLA4hi (*p* < 0.01) than CD4<sup>+</sup>CTLA4med/lo T cells (*p* < 0.05) (**Figure 5B**). The addition of CTLA4Ig during allo-stimulation significantly reduced the frequencies of both

CD4+CTLA4hi and CD4+CTLA4med/lo populations in a concentration-dependent manner (**Figure 5B**; left). At CTLA4Ig 100 µg concentration, the frequencies of both populations were comparable with non-proliferating cells (**Figure 5B**; left). However, the percent reduction in CD4<sup>+</sup>CTLA4hi T cells by CTLA4Ig was more marked than the percent reduction in CD4<sup>+</sup>CTLA4med/lo T cells (*p* < 0.001) (**Figure 5B**; right).

While the reduction in CTLA4 expression mediated by CTLA4Ig could be attributed to reduced proliferation of both CD4<sup>+</sup>CTLA4hi and CD4+CTLA4med/lo T cell populations, a more pronounced reduction in CD4<sup>+</sup>CTLA4hi (*p* < 0.01) than CD4<sup>+</sup>CTLA4med/lo (*p* < 0.05) cell proliferation was observed (**Figure 5C**). To confirm this observation, we examined the ratio of proliferating allo-stimulated CD4<sup>+</sup>CTLA4hi to CD4+CTLA4med/lo

Figure 5 | Cytotoxic T-lymphocyte-associated antigen (Ag) 4:Ig (CTLA4Ig) reduces CD4+CTLA4hi more than CD4+CTLA4med/lo T cells following allo-stimulation *in vitro*. (A) CTLA4Ig significantly reduces the incidence and proliferation of CD4+CTLA4hi T cells following allo-stimulation *in vitro*. Carboxyfluorescein succinimidyl ester (CFSE)-labeled responder peripheral blood mononuclear cells (PBMC) were co-cultured with allogeneic, T cell-depleted PBMC for 5 days, in the absence or presence of CTLA4Ig (1 or 100 µg/ml). CFSE dilution of CTLA4hi and CTLA4med/lo CD4+ T cells and the percentages of total CTLA4hi and CTLA4med/lo CD4+ T cells were determined after gating on CD4+ cells. (B) Mean percentages of total CTLA4hi and CTLA4med/lo CD4+ T cells. (D) Percent reductions in CTLA4hi and CTLA4med/lo CD4+ T cells in the presence of CTLA4Ig are shown on the right. (C) Mean values of CFSE dilution of CTLA4hi and CTLA4med/lo CD4+ T cells (left). Ratios of CTLA4hi to CTLA4med/lo CD4+ T cell proliferation following allo-stimulation in the presence or absence of CTLA4Ig (right). Graphs represent data from five independent experiments; bars represent means + 1 SD. \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; only significant values are shown.

T cells in the presence or absence of CTLA4Ig. CD4<sup>+</sup>CTLA4hi/ CD4<sup>+</sup>CTLA4med/lo T cell ratios were reduced significantly by CTLA4Ig in a concentration-dependent manner (**Figure 5D**). These observations indicate that CD28 Co-SB during allostimulation of normal rhesus CD4<sup>+</sup> T cells *in vitro* may enhance the incidences of CD4<sup>+</sup>CTLA4med/lo but not CD4+CTLA4hi T cells.

### DCreg Infusion Maintains Donor-Specific CD4**+**CTLA4hi T Cell Proliferation in CTLA4Ig-Treated Renal Allograft Recipients

Our *in vitro* data suggest that CD28 Co-SB may be associated with a reduction in allo-reactive CD4<sup>+</sup>CTLA4hi T cells following transplantation. In our earlier study (22), rhesus renal allograft recipients received CTLA4Ig together with rapamycin monotherapy, either without [control (CTRL) group] or in combination with pre-transplant donor-derived DCreg infusion (DCreg group). Since we observed upregulation of CTLA4 expression by host T cells in response to donor but not third party stimulation in recipients with DCreg infusion (22), we hypothesized that DCreg infusion before transplantation, would promote donor-reactive CD4<sup>+</sup>CTLA4hi T cells after transplantation. We examined the proliferation of host CD4<sup>+</sup>CTLA4hi and CD4+CTLA4med/lo T cells in response to *ex vivo* donor or third party allo-Ag stimulation, before and 1 month after transplantation (**Figure 6**).

Before transplantation, proliferation of CD4<sup>+</sup>CTLA4hi T cells in response to either donor or third party stimulation was similar in the CTRL and DCreg groups. In the CTRL group, the proliferation of CD4<sup>+</sup>CTLA4hi T cells in response to donor stimulation was reduced markedly after transplantation (*p* = 0.057), but only modestly reduced in response to third party stimulation (**Figure 6A**). In contrast, in graft recipients given DCreg, proliferation of CD4<sup>+</sup>CTLA4hi T cells in response to donor stimulation was not reduced after transplantation, whereas proliferation induced by third party cells was reduced modestly. Since there was no reduction in CD4<sup>+</sup>CTLA4med/lo T cell proliferation in either group after transplantation compared with before transplantation (not shown), we evaluated the ratio of CD4<sup>+</sup>CTLA4hi to CD4<sup>+</sup>CTLA4med/lo proliferating T cells (as in **Figure 5B**) in response to donor stimulation in both groups (**Figure 6C**). With no DCreg infusion (CTRL), the proliferating CD4<sup>+</sup>CTLA4hi/ CD4<sup>+</sup>CTLA4med/lo T cell ratio in response to donor stimulation was reduced markedly after transplantation (*p* < 0.05). On the other hand, this ratio was slightly increased in graft recipients given DCreg infusion. Notably, in response to donor stimulation, the CD4<sup>+</sup>CTLA4hi/CD4+CTLA4med/lo ratio was significantly higher in the DCreg-treated recipients than in the CTRL group (*p* < 0.05).

These observations indicate that, while CD28 Co-SB (with CTLA4Ig) can significantly reduce rhesus allo-reactive CD4<sup>+</sup>CTLA4hi T cells both *in vitro* and in CTLA4Ig-treated renal allograft recipients, pre-transplant donor DCreg infusion prevents reduction of donor-specific CD4<sup>+</sup>CTLA4hi T cells after transplantation.

### DISCUSSION

We have reported previously (22) that a single systemic infusion of donor-derived DCreg 1 week before transplantation in combination with CD28 Co-SB (CTLA4Ig), results in prolonged renal allograft survival in a robust, MHC-mismatched pre-clinical NHP rhesus macaque model. Furthermore, graft recipients given DCreg showed selective attenuation of donor-reactive Tmem, associated with enhanced expression of the T cell co-inhibitory molecules CTLA4 and PD1 upon *ex vivo* stimulation of Tmem with donor but not third party cells.

Cytotoxic T-lymphocyte-associated Ag 4, a CD28 homolog, is upregulated by T cells after activation and negatively regulates immune responses (24, 27, 46). Rodent studies have validated the significance of CTLA4 expression for the promotion of allograft tolerance. Thus, blockade of the interaction between CTLA4 and B7 molecules accelerates graft rejection (28). Of particular relevance to our previous study in NHP (22), exposure of CTLA4-expressing CD4<sup>+</sup> T cells to donor Ag is essential for the prevention of effector T cell responses and the promotion of transplant tolerance (47, 48). CTLA4 is expressed by natural and inducible Treg and contributes to their suppressive function (49, 50). Indeed, CTLA4 expression is critical for optimal Treg function and for the suppressive effect of these cells on T cell responses to allo-Ags (38, 39). While Foxp3 is known to upregulate CTLA4 expression (40), it has been argued that both Foxp3 and CTLA4 can independently promote immune tolerance (30).

CD28 co-stimulation is not only required for T cell activation and effector function, but it is also critical for Treg generation and for sustaining a balance between effector and Treg cells. CD28 co-stimulation and CTLA4 interaction with CD80 and CD86 are known to reduce CD4<sup>+</sup> (51) and CD8<sup>+</sup> (52) Th17 differentiation (53). CD28 Co-SB with CTLA4Ig (belatacept) was approved by the FDA in 2011 for kidney transplantation, but its use has been associated with higher rates of acute cellular rejection, despite superior renal graft function (16, 54). While allo-reactive Tmem are known to be CD28 Co-SB-resistant (18, 19) since they do not require co-stimulation (55, 56), some potentially unfavorable

renal allograft recipients with DCreg infusion (DCreg group) before transplantation and on POD 28 and POD 56. (C) Mean values of four graft recipients from the CTRL group (upper left) and four recipients from the DCreg group (upper right). In all recipients, ratios of percent proliferation of CTLA4hi to CTLA4med/lo CD4+ T cells in response to donor stimulation before and after transplantation are shown (below). \**p* < 0.05; only significant values are shown.

effects of CTLA4Ig on regulation of allo-reactive T cell responses are being recognized. Thus, recent reports indicate that CTLA4Ig can prevent Treg-dependent transplantation tolerance (20, 21). Moreover, in patients with rheumatoid arthritis (44) and following kidney transplantation (45), treatment with CTLA4-Ig has been shown to decrease the incidence of circulating Treg. These observations imply potential limitations of CTLA4Ig-based therapies for transplantation.

In our NHP renal allograft model (22), immunosuppression based on CD28 Co-SB using CTLA4Ig did not lead to increased Treg frequency, with or without DCreg infusion. However, we did observe increased Treg/Tmem ratios in the blood of graft recipients with DCreg infusion, in association with upregulation of CTLA4 and PD1 expression by Tmem following their stimulation with donor but not third party Ag. This suggests that, when Tmem encounter donor Ag following infiltration of the graft, they may upregulate PD1 and CTLA4, which in turn may control their activation/survival. Indeed, similar to graft-infiltrating CD8<sup>+</sup> T cells, graft-infiltrating CD4<sup>+</sup> T cells also upregulated CTLA4 and PD1 expression in DCreg-infused recipients (**Figure 1**).

In this study, we examined the relationship between CTLA4 and/or PD1 expression by allo-reactive T cells and Treg *in vitro* and observed that only CD4<sup>+</sup> T cells with high CTLA4 expression exhibited a regulatory phenotype, i.e., CD25hi Foxp3hi CD127lo, while CTLA4med/lo CD4<sup>+</sup> T cells did not (**Figure 4**). On the other hand, there was no correlation between PD1 expression and Treg phenotype. Interestingly, while the frequencies of both CTLA4hi and CTLA4med/lo CD4<sup>+</sup> T cells increased following allo-stimulation, CTLA4Ig reduced CTLA4hi more than CTLA4med/lo CD4<sup>+</sup> T cells (**Figure 5**), resulting in the ratio of proliferating CD4<sup>+</sup>CTLA4hi to CD4<sup>+</sup>CTLA4med/lo T cells being reduced significantly in a CTLA4Ig concentration-dependent manner (**Figure 5**). This suggests that in normal rhesus, blocking CD28 co-stimulation during allo-stimulation reduces CD4<sup>+</sup>CTLA4hi T cells in favor of CD4<sup>+</sup>CTLA4med/lo T cells.

While in NHP allograft recipients, we found no increase in the frequency of circulating Treg, with or without DCreg infusion (22), there was a significant reduction in the incidence of circulating CD4<sup>+</sup>CTLA4hi T cells in the control group following donor Ag stimulation post-transplant, while in recipients given DCreg infusion, the incidence of CD4+CTLA4hi T cells was modestly elevated (**Figure 6**). Moreover, the ratio of CD4<sup>+</sup>CTLA4hi to CD4<sup>+</sup>CTLA4med/lo T cell proliferation in response to donor stimulation post-transplant was significantly higher in DCregtreated than in control graft recipients. These observations suggest that an unfavorable influence of CTLA4Ig on allo-reactive Treg was averted in graft recipients given DCreg infusion. Indeed, combination of donor-derived DCreg with CTLA4Ig can result in long-term murine organ allograft survival, mediated at least in part, by CD4<sup>+</sup> Treg (57).

Our observations indicate that high CTLA4 expression by donor-reactive T cells correlates with host immune regulation and is associated with better graft outcomes after transplantation.

### REFERENCES


In a recent report (58), high levels of CTLA4 expression correlated with augmentation of CD4<sup>+</sup> Th17 Tmem responses in renal allograft recipients given CD28 Co-SB immunosuppression. In our NHP study, we have confirmed the regulatory phenotype of CD4<sup>+</sup>CTLA4hi T cells, compared with CD4+CTLA4med/lo T cells. Although we did not assess Th17 expression in this study, donorreactive CD4<sup>+</sup>CTLA4med/lo T cells may correlate with Th17 CD4<sup>+</sup> Tmem, particularly in the presence or absence of DCreg infusion.

Our observations provide further insight to the limitations of CD28 Co-SB in renal allo-transplantation. They are consistent with the view that, while CTLA4Ig efficiently prevents effector T cell responses to donor Ags, this may come at the expense of regulatory mechanisms that favor donor-specific Treg and attenuate donor-specific Tmem, and hence increased rates of acute cellular rejection. Our findings also suggest that DCreg infusion before renal transplantation is associated with preservation of donor-specific T cell regulation that may otherwise be compromised by CD28 Co-SB.

### ETHICS STATEMENT

All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Experiments were conducted according to the guidelines set forth in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Specific environment enrichment was provided.

### AUTHOR CONTRIBUTIONS

ME participated in research design, writing of the paper, performance of the research, and data analysis. LL participated in performance of the research and data analysis. WS participated in performance of the research. AM participated in research design and data analysis. AT participated in research design, data analysis, and writing of the paper.

### FUNDING

This study was supported by National Institutes of Health (NIH) grants U01 AI51698 and U19 AI131453, part of the NIH NHP Transplantation Tolerance Study group and sponsored by the NIAID and NIDDK.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Ezzelarab, Lu, Shufesky, Morelli and Thomson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Tolerogenic Dendritic Cells in Solid Organ Transplantation: where Do we Stand?

### *Eros Marín1,2, Maria Cristina Cuturi1,2 and Aurélie Moreau1,2\**

*1Centre de Recherche en Transplantation et Immunologie UMR1064, INSERM, Université de Nantes, Nantes, France, 2 Institut de Transplantation Urologie Nephrologie (ITUN), CHU Nantes, Nantes, France*

Over the past century, solid organ transplantation has been improved both at a surgical and postoperative level. However, despite the improvement in efficiency, safety, and survival, we are still far from obtaining full acceptance of all kinds of allograft in the absence of concomitant treatments. Today, transplanted patients are treated with immunosuppressive drugs (IS) to minimize immunological response in order to prevent graft rejection. Nevertheless, the lack of specificity of IS leads to an increase in the risk of cancer and infections. At this point, cell therapies have been shown as a novel promising resource to minimize the use of IS in transplantation. The main strength of cell therapy is the opportunity to generate allograft-specific tolerance, promoting in this way long-term allograft survival. Among several other regulatory cell types, tolerogenic monocyte-derived dendritic cells (Tol-MoDCs) appear to be an interesting candidate for cell therapy due to their ability to perform specific antigen presentation and to polarize immune response to immunotolerance. In this review, we describe the characteristics and the mechanisms of action of both human Tol-MoDCs and rodent tolerogenic bone marrow-derived DCs (Tol-BMDCs). Furthermore, studies performed in transplantation models in rodents and non-human primates corroborate the potential of Tol-BMDCs for immunoregulation. In consequence, Tol-MoDCs have been recently evaluated in sundry clinical trials in autoimmune diseases and shown to be safe. In addition to autoimmune diseases clinical trials, Tol-MoDC is currently used in the first phase I/II clinical trials in transplantation. Translation of Tol-MoDCs to clinical application in transplantation will also be discussed in this review.

Keywords: autologous tolerogenic dendritic cells, transplantation, cell therapy, clinical trial, safety, mechanisms

### INTRODUCTION

More than half a century has passed since the first successful renal transplantation at the Peter Bent Brigham Hospital in Boston. The procedure performed by Joseph Murray's team showed for the first time the surgical feasibility of solid organ transplantation, at least between identical twins (1). Parallel to this achievement, research on immunosuppressive drugs (IS) demonstrated that 6-mercaptopurine (6-MP), a drug already used to treat acute lymphocytic leukemia, was able to impair immune response (2). These novel concepts of feasibility of solid organ transplantation and immunosuppressive treatment to avoid graft-versus host disease opened the doors for unrelated organ transplantation. Over the following years, advances in IS research led to the replacement of

#### *Edited by:*

*Daniel Hawiger, Saint Louis University, United States*

#### *Reviewed by:*

*Alain Le Moine, Université libre de Bruxelles, Belgium Bruce Milne Hall, University of New South Wales, Australia*

*\*Correspondence: Aurélie Moreau aurelie.moreau@univ-nantes.fr*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 31 October 2017 Accepted: 30 January 2018 Published: 19 February 2018*

#### *Citation:*

*Marín E, Cuturi MC and Moreau A (2018) Tolerogenic Dendritic Cells in Solid Organ Transplantation: Where Do We Stand? Front. Immunol. 9:274. doi: 10.3389/fimmu.2018.00274*

6-MP, which is highly toxic, by cyclosporine, leading to an increase in one-year graft survival (3). Nowadays, more specific IS are being used to treat post-transplanted patients, such as mophetil mycophenolate, a B and T-cell proliferation inhibitor; tacrolimus, a B and T-cell activation inhibitor (4), and monoclonal antibodies, such as basiliximab, an IL2Rα (CD25) blocking antibody (5). However, although IS treatments favor allograft survival, these treatments are also associated with an increased risk of cancer and infections associated to the immunosuppressive state (6). Moreover, IS primarily prevents the acute rejection of allografts, whereas their efficacy in chronic rejection remains difficult to predict (7). A novel and promising strategy to minimize drugs treatment and control of chronic rejection is to combine reduced amounts of IS with immunoregulatory cell therapy in solid organ transplantation.

Cell therapy for solid organ transplantation could be performed with mesenchymal stem cells (MSC), regulatory macrophages (Mreg), tolerogenic monocyte-derived dendritic cells (Tol-MoDCs), and regulatory T (Treg) and B (Breg) cells (8). The common characteristic between these different cells is that they have been already tested in transplantation models in animals showing a benefit in terms of safety and graft survival. For example, MSC have been shown to delay heart allograft rejection (9). In humans, several clinical trials with MSC have been performed in kidney and liver transplantation (10). Among them, a large trial was carried out to compare MSC to anti-IL2Rα therapy. In this study, the authors showed a lower incidence of acute rejection and a better estimated renal function at 1 year compared to the anti-IL2Rα receiving cohort (11). On the other hand, Mreg have been shown to increase fully allogeneic allograft survival in non-immunosuppressed mice (12). Additionally, Mreg were tested in a clinical trial in living donor renal transplantation. In this study, two patients were treated with Mreg prior to transplantation followed by low doses of tacrolimus. The outcomes of this trial showed that Mreg-treated patients displayed a stable graft function after tacrolimus weaning (13). Tolerogenic bone marrow-derived DCs (Tol-BMDCs) have demonstrated to increase heart, skin, and pancreatic islet allograft survival in combination with IS (14–16). Regarding lymphoid cells, Treg therapy has been shown to be safe and effective in a pilot study in living donor liver transplantation. Indeed, 6 from 10 initial patients in this study were able to stop the immunosuppressive therapy (17). In the context of the ONE study consortium, clinical trials with Treg, Mreg, type 1 Treg cells (Tr1), and Tol-MoDCs are currently performed in living donor kidney transplantation in order to evaluate and compare the safety of these cells in transplantation (www.onestudy.org) (18). In this review, we will focus on both Tol-MoDCs and Tol-BMDCs and their translation to the clinical trial with an emphasis on their characteristics, mechanisms, and safety.

### DENDRITIC CELLS

Dendritic cells were discovered by Steinman and Cohn back in 1973 (19, 20). However, the first clinical trial with DC therapy was carried out in 1995 in advanced melanoma patients (21). The reason to use these cells in cell therapy resides in their capacity to present antigens to T cells and to polarize the immune response; in other words, to link the innate and adaptive response (22). DCs are potent antigen presenting cells (APC), able to induce either immunity or tolerance. The first studies about the functions and characteristics of DCs demonstrated that DCs were strong stimulators of T cell response in allogeneic MLR. Additionally, the authors demonstrated the capacity of these cells to induce antigen-specific proliferation (23). Over the following years, different subsets with different ontogenies and functions have been characterized in DCs, such as conventional DC (cDC), plasmacytoid DC (pDC), Langerhans cells (LC), and inflammatory DCs. cDC commonly located in lymphoid tissues and nonlymphoid tissues are able to present antigen through major histocompatibility complex class II (MHC class II) in rodent and humans. Moreover, cDC can cross-present antigens *via* MHC class I (24). pDC, located usually in peripheral organs, are able to induce T-cell proliferation. However, pDCs are usually known to secrete high amounts of type I interferon (IFN) upon viral infection. Inflammatory DCs, also named MoDCs are derived from monocytes that infiltrate lymphoid and nonlymphoid organs as a consequence of inflammation or infection. Finally, LCs are DC skin-resident cells with the capacity to migrate to skin-draining lymph nodes. Unlike cDC, pDC, and MoDC that share the same precursor (monocyte-DC common precursor), the ontogeny of LC go back to the prenatal origin (25).

Nowadays, it has been demonstrated that the orchestration of all these DC subsets is essential for an adequate physiological response against threats, but also for the preservation of selftolerance. In fact, it has been demonstrated that the ablation of cDC, pDC, and LCs in a model of transgenic CD11c-CRE mice, leads to a spontaneous autoimmunity (26).

### *Ex Vivo* Generated Tolerogenic DCs

Nowadays, rodent DCs are derived from bone marrow cells, whereas human DCs are derived from monocytes for both immunosuppressive and other therapies. Monocytes are used in humans for convenient reasons as they are more abundant than other DC precursors, and can be also manipulated *ex vivo*. From a pragmatic point of view, DCs can be differentiated *in vitro* as immunogenic or tolerogenic cells depending on the protocol. Immunogenic DCs are characterized by a high expression of costimulatory molecules, such as CD80 and CD86, a production of pro-inflammatory cytokines, such as IL1β, IL-12, and tumor necrosis factor-α (TNFα) and the ability to stimulate T-cell proliferation. In counterpart, tolerogenic DCs weakly express costimulatory molecules, are resistant to maturation, produce immunomodulatory cytokines, such as IL-10 and transforming growth factor-β (TGFβ) and impair T-cell proliferation (**Figure 1**). Both DCs are known to express common markers, such as CD11c, CD11b, or MHC Class I and Class II molecules (27).

As it has been previously mentioned, *in vitro* derived DC can be manipulated *ex vivo* in order to design more accurate therapies. For example, these cells can be loaded with target peptides, such as synthetic nanopeptides of MAGE-1 protein in order to direct immune response against human melanoma cells (21). On the other hand, they can be treated with inhibiting

molecules associated to antigen presentation, in order to prevent pro-inflammatory response (28).

Due to this versatility and functional duality, *in vitro* derived DCs have already been used in immunogenic therapy, such as in infections (29) and cancer therapy (30), and immunosuppressive therapy, such as in allergy (31), autoimmunity (32), immunization (33), and more recently in transplantation (34).

GM-CSF is a growth factor related with bone marrow precursor mobilization and DC differentiation (35). However, the role of GM-CSF in tolerance remains unclear as its administration improves some diseases, such as myasthenia gravis, type 1 diabetes (T1D), and colitis, but its depletion improves experimental autoimmune encephalomyelitis (EAE), arthritis, nephritis, and psoriasis in rodent models (36). GM-CSF is a cytokine indispensable for *in vitro* DC generation, which is used both for immunogenic or tolerogenic DC differentiation. This dual role of GM-CSF is dichotomized by the concentration of the cytokine. Indeed, low doses of GM-CSF are associated to tolerogenic phenotypes, whereas high amounts of GM-CSF lead to immunogenic phenotypes (37).

Moreover, there is not a single standardized method to generate Tol-MoDC from monocytes in humans or Tol-BMDCs in rodents apart from GM-CSF and IL-4. Protocols to induce human and rodent tolerogenic DCs usually include several other factors, such as cytokine cocktails, organic molecules, or even clinically approved and experimental drugs (38). For example, IL-10 and TGF-β, two well-known immunomodulatory molecules, have been shown to maintain the immature phenotype of DCs (39, 40). Human Tol-MoDCs generated with IL-10 spontaneously secrete high amounts of IL-10 and are able to impair T-cell proliferation and induce Tr1 cells (41). Similarly, Tol-MoDCs generated with IL-10 and TGF-β from monocytes obtained from T1D patients was able to induce tolerance to insulin antigens. These cells express several DC markers, such as CD83, CD1a, MHC II, but not CD14 (42). Regarding small organic molecules, such as 1α,25-dihydroxyvitamin D3 (1α,25(OH)2D3 Vit D3), and prostaglandin E2 (PGE2) have been shown to induce Tol-MoDCs (38). Immature DCs treated with Vit D3 are resistant to maturation upon lipopolysaccharide (LPS) stimulation and impair allogeneic T-cell proliferation. In this study, the authors showed that Vit D3 treated MoDCs downregulated CD1a and CD14 markers (43). However, another study demonstrated that Vit D3 differentiated Tol-MoDCs express DC-SIGN (CD209), CD14, but not CD1a (44). PGE2 induces the expression of indoleamine 2,3 dioxygenase (IDO) by DC leading to a production of kynurenine that plays a role in Treg generation and allogeneic response inhibition (45). Tol-MoDCs can also be differentiated in the presence of dexamethasone (Dex) and rapamycin (Rapa). A comparative study determined that both Dex-DCs and Rapa-DCs were able to impair T-cell proliferation, but unlike Dex-DCs, Rapa-DCs displayed a mature DCs phenotype and were not able to produce IL-10 upon LPS stimulation (46). Phenotypically, it has been shown that Dex-DCs have a low expression of CD1a and CD14 and they express CD209 (44). On the other hand, it has been shown that Tol-BMDCs differentiated with Rapa are phenotypically characterized by the expression of CD11b, CD11c, CCR7, and have a low expression of MHC ClassII (47). Furthermore Dex-DCs stimulated with a cytokine cocktail (IL-6, TNFα, IL-1β, and PGE2) have been administered in patients suffering from refractory Crohn's disease. An increase of Treg cells and a decrease of interferon-γ (IFN-γ) in blood were observed following DC injection (48). Other protocols to generate TolDCs, include genetic tools, concretely antisense oligonucleotides (AS-ODN). A study performed in nonobese diabetic (NOD)-mice showed that the injection of TolDCs modified using AS-ODN anti-CD40, CD80, and CD86 delayed diabetes onset (28).

Among these different methods, our group has adopted a protocol to generate tolerogenic DCs from mouse bone marrow cells with low doses of GM-CSF, excluding IL-4 from the classic protocol (16). This protocol, previously described by Lutz et al. (49), allowed obtaining Tol-BMDCs expressing low levels of MHCII, CD40, CD80, and CD86, and displaying resistance to maturation upon LPS stimulation. Furthermore, these Tol-BMDCs impaired allogeneic T-cell proliferation. Lutz et al. demonstrated that these cells were able to increase graft survival following a fully allogeneic vascularized heterotopic cardiac allograft, whereas we highlighted the potential of Tol-BMDCs in minor antigen skin graft survival. Alternatively, this protocol was adopted in human to generate Tol-MoDCs from blood monocytes, resulting in an equivalent profile (49). Nowadays, we are performing a first phase I/II clinical trial in kidney transplantation using Tol-MoDCs generated with low doses of GM-CSF as described previously (50). Altogether, the common phenotypical observation after tolerogenic DC differentiation showed that due to the heterogeneity of differentiation protocols it is not possible to describe a unique phenotype for these cells. However, the most common markers observed on tolerogenic DCs are CD11c and low expression of MHCII. On the other hand the expression of DC markers CD209 and CD1a, monocyte/macrophage marker CD14, and macrophage marker CD11c are variable.

## TOLEROGENIC DC SOURCE

Unlike other diseases or conditions, transplantation involves the allorecognition between the two parts, the graft and the host. Allorecognition refers to an immune response against allogeneic peptides or against MHC molecules (51). The alloresponse could be differentiated depending on the nature of the interaction by direct, indirect, and semi-direct pathways. In the direct pathway, recipient T cells are activated following presentation of allogeneic MHC molecules by donor DCs and this pathway is associated with acute rejection. Indirect pathway refers to the processed allopeptides presentation by recipient DCs to autologous T cells and is usually associated to chronic rejection. On the semidirect pathway, intact donor MHC molecules are transferred to recipient DCs through cell-to-cell contacts; the cells are then able to stimulate autologous T cells (52). Therefore, in order to avoid these types of rejection two strategies were considered: the infusion of donor-specific antigens in order to generate antigen-specific regulatory cells or in contrast, the minimization of the risk of transfer allogeneic molecules in order to avoid sensitization.

The first alternative is currently used clinically in kidney transplantation. Indeed, donor-specific transfusion (DST) is a procedure in which recipients receive a donor-specific blood transfusion in order to generate tolerance to donor antigens. A study performed in living donor kidney transplantation comparing recipients receiving DST or not, in addition to immunosuppressive therapy, showed a reduction in patients with acute rejection and an increase in patients with optimal renal function at 1 and 10 years after transplantation in the DST group (53). On the other hand, the presence of allogeneic molecules in transplantation is unavoidable and even if the efficacy of DST has been demonstrated, sensitization against HLA can occur and appears as a risk for allograft rejection (54). For this reason, the safety and efficiency of donor and recipient DCs have been discussed in DC-based therapy in transplantation.

As it has been previously mentioned, the work performed by Lutz et al. showed that Tol-BMDCs generated with low dose of GM-CSF induced an increase in allograft survival in recipient CBA mice receiving a cardiac allograft from donor B10 mice and pretreated with donor Tol-BMDCs for 7 days before the transplantation. This prolongation of allograft survival was achieved until day 100 for 70% of mice, meanwhile the mice pretreated with donor Tol-BMDCs receiving a third-party allograft from NZW mice or DC generated with GM-CSF and IL-4 increased graft survival only in 20% of mice. Moreover, in this study the authors showed that T cells cultured with allogeneic Tol-BMDCs remained unresponsive after polyclonal restimulation. These results implied that this unresponsiveness was specific (55). Another study performed by DePaz et al. in rats using donor BMDCs generated with low doses of GM-CSF showed that Tol-BMDC therapy in combination with antilymphocyte serum (ALS) was able to increase rat cardiac allograft survival in 50% of rats up to 200 days. In the same way as the previous work, the authors showed that T cells purified from transplanted mice receiving Tol-BMDCs therapy and ALS were unresponsive to donor antigens, indicating an induction of antigen-specific tolerance (56). Nevertheless, a later study using donor Tol-BMDCs or apoptotic bodies from donor Tol-BMDCs, showed that tolerance was mediated by the presentation of donor peptides (from donor cells or apoptotic bodies) by recipient DC, that inhibits CD4+Tcell activation and favors Treg expansion (57). Altogether these studies demonstrate the similarities of donor Tol-BMDC therapy with DST therapy. Both therapies have been shown to be partly efficient, but on the other hand, the risk of sensitization (including the development of alloantibodies) still remains. Therefore, the use of autologous tolerogenic DCs appears as a better alternative at least in terms of safety because it avoids the risk of sensitization.

In order to determine if autologous tolerogenic DCs shared a closer efficacy with donor tolerogenic DCs in transplantation, several studies have been performed. In 2005, a study performed by our team demonstrated that rat Tol-BMDCs (corresponding to the adherent fraction of rat BMDCs generated with GM-CSF and IL-4) displayed an immature phenotype were maturation resistant and were able to prolong cardiac allograft survival. Interestingly, autologous Tol-BMDCs were more efficient than donor DCs in delaying graft rejection. In this study, autologous Tol-BMDCs were injected the day before the transplantation suggesting that this time of administration was sufficient to pre-treat patients before the intervention (58). We then demonstrated that rats receiving heart allograft and treated with autologous Tol-BMDCs in combination with suboptimal doses of LF15-0195, an nuclear factor-κB (NF-κB) inhibitor, achieved definitive allograft acceptance. Moreover, we demonstrated that this tolerance was donor-specific (59). These results combined demonstrated that autologous Tol-BMDCs are even more efficient than donor Tol-BMDCs and due to its source, conceptually safer.

### PROFILING TOLEROGENIC DC THERAPY

### Combined Therapy

Previous results have shown that tolerogenic DC therapy could be improved by the addition of a complementary treatment such as ALS. However, more specific drugs have been used showing an improvement of tolerogenic DC therapy.

LF15-0195 is a NF-κB blocking agent that was previously reported to increase cardiac allograft survival in rats in short-term treatment (60). Moreover, this compound impairs the maturation of DCs (31). In combination, autologous Tol-BMDCs with a suboptimal dose of LF15-0195 induced tolerance to cardiac allograft in 92% of treated rats compared to autologous Tol-BMDCs alone, LF15-0195 alone, or rats treated with Rapa with or without autologous Tol-BMDCs. In order to determine whether this tolerance was specific, donor, recipient, and thirdparty skin transplantations were performed in tolerant rats. Our results showed that tolerant rats do not reject donor skin graft, but reject third-party skin graft for 16–18 days after transplantation (59). Another efficient combined therapy in transplantation was anti-CD3 antibody. Indeed, it has been demonstrated that the use of monoclonal antibody anti-CD3 leads to an increase of pancreatic islet, skin, and cardiac allograft survival in transplantation models and led to remission in T1D in autoimmune disease models (61, 62). Our results show that the combination of anti-CD3 antibody and autologous Tol-BMDCs therapy led to an increase of pancreatic islet allograft survival, associated with a decrease in CD4<sup>+</sup>/CD8<sup>+</sup> T cell frequency, and an increase in Treg frequency. The relevance of this increased CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg frequency and its contribution to allograft survival in this model was demonstrated by the depletion of CD25<sup>+</sup> T cells with anti-CD25 antibody (15). We then confirmed the strong potential of autologous Tol-BMDCs and anti-CD3 therapy to prolong allograft survival in a model of minor antigen mismatch skin transplantation. In this model, our group found an increase in regulatory CD8<sup>+</sup>CD11c<sup>+</sup> T cells associated with this combined therapy (16). Rapa is another drug that demonstrated an improvement of efficacy in collaboration with Tol-BMDCs in transplantation. Indeed, the injection of donor Tol-BMDCs generated with Rapa and pulsed with donor antigens followed by post-operative low doses of Rapa in heart transplantation mouse model demonstrated an increase of allograft survival. This allograft survival was related to an increase in donor-specific CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> Treg in the graft. To ensure the specific regulatory activity of these induced Treg, the authors performed an adoptive transfer of purified CD4<sup>+</sup> T cells from treated mice to naïve mice receiving heart allograft. Adoptive Treg transfer resulted in an increase in allograft survival, indicating that tolerance was induced by this combined therapy (47). These results altogether demonstrated that autologous Tol-BMDC therapy in combination with specific drugs increased its potency.

### Administration Route and Efficacy in Non-Human Primates (NHP)

In terms of therapeutic effects, Tol-BMDCs have been shown to be efficient and safe in rodents. To ensure its safety profile for clinical trial, several works have been performed in NHP. The first study using tolerogenic DCs therapy, performed in a kidney transplantation model in NHP, showed an increase of median graft survival compared to the control group. In this study, rhesus macaques were co-treated with CTLA4-Ig and donor Tol-BMDCs generated with Vit D3 and IL-10, 7 days before the transplantation. This study demonstrated for the first time the safety and the efficiency of intravenously (IV) injected Tol-MoDCs in transplantation in NHP (63). More recently, the same authors demonstrated similar results in kidney transplantation models in NHP using autologous Tol-MoDCs pulsed with allogeneic cell membranes from donor monocytes. In this study, the authors showed an increase of graft median survival in the group treated with pulsed Tol-MoDCs compared to unpulsed Tol-MoDCs group. This improvement in allograft survival was associated with the hyporesponsiveness of T cells to donor antigens resulting in a decrease in systemic IL-17 (64). In addition, other studies have demonstrated the safety and efficacy of Tol-BMDCs in NHP notably in gene therapy. Indeed, we demonstrated the benefits of autologous Tol-BMDCs therapy to reduce immune response against a transgene product in NHP. In this study, autologous Tol-BMDCs were injected IV or intradermally (ID) in order to determine the best administration route. Our results highlighted the superiority of IV route to favor immune tolerance (65). Furthermore, several clinical trials have already been performed, confirming that ID (32, 66), intraperitoneal (48), and IV administration routes were safe and well tolerated in humans.

### TOLEROGENIC DC CELLULAR AND MOLECULAR MECHANISMS

Once it has been confirmed that tolerogenic DCs improve allograft survival in rodent models and are safe in humans, the remaining question is to determine the cellular and molecular mechanisms of these cells in transplantation. To understand tolerogenic DC mechanisms (in Tol-BMDC and Tol-MoDC), it is first crucial to define the complexity of solid organ transplantation. Due to the invasiveness of the surgical procedure and the implantation of a foreign organ, even from a close source, different types of immune and non-immune cells are involved in the physiological response following transplantation. This physiological response against allograft will lead in some cases, to three expected types of rejection. The earliest one is the hyperacute rejection, in which, pre-existing recipient antidonor antibodies will react against allograft over the hours following the transplantation. This type of rejection is rare thanks to the control of HLA donor/recipient compatibility. The acute rejection is led by cellular and humoral response against allograft. This type of rejection is usually bypassed by the use of IS. Finally, the chronic rejection is led by cellular and humoral response and associated with memory cells. Chronic rejection is nowadays the main cause of rejection (67). Due to the complexity of the different types of rejection, TolDC therapy in transplantation has been evaluated on these different parameters: the migration to graft and lymphoid organs, the capacity to induce specific regulatory cells, and the ability to impair cellular and humoral response.

### Migration

It is well known that DCs have migratory skills that allow reaching different organs in order to exert different functions, depending on the maturation state. At the immature state, DCs express chemokine receptors, such as CCR2, CCR5, CCR6, CXCR4, and CXCR3 and are attracted by inflamed tissues expressing chemokines, such as CCL2, CCL5, and CCL20. At the inflammation site, DCs become mature due to the stimuli provided by the microenvironment and the antigen intake. Following their maturation, DCs overexpress CCR7 allowing them to migrate to the lymphatic system and reach the lymph nodes through CCL19 and CCL21 chemoattraction, where they present antigens to T cells. In the lymph nodes, a certain percentage of DCs will migrate to other lymphoid organs, such as spleen, thymus, and bone marrow (68). In a recent study performed in an EAE model, *in vivo* imaging of pulsed Tol-BMDCs generated with GM-CSF and VitD3 showed that these cells reached the liver and the spleen at 24 h after IV injection and remain stable for 7 days. A small amount of cells were also found in lymph nodes, thymus, and bone marrow (69). In order to support the importance of migration, DCs transduced with lentiviral vectors coding for CCR7 and IL-10 genes, prolonged cardiac allograft survival in mice, but this delay of rejection did not occur when DCs were transduced with IL-10 or CCR7 only. In this study, the authors also showed that DC transduced with CCR7 were able to migrate to LN and spleen (70). Additionally, in order to expose donor and recipient DC dynamics, a study was performed using intravital imaging in ear skin graft model in mice. In this work, authors showed that after transplantation donor dermal DCs migrate from allograft and are replaced by host DC. After donor antigen intake, these recipient DCs migrate to lymph nodes in order to present antigens to CD8<sup>+</sup>T cells and prime anti-allograft response. This work suggested the dynamics of DC immunotherapy *in vivo* (71).

### T Cell Inhibition

Even if tolerogenic DCs are able to migrate to lymphoid organs, the goal is to avoid the exacerbated proliferation of T cells in those organs and in the long term, the memory, and humoral responses. Conveniently, two common effects between the different works performed with tolerogenic DC therapy in transplantation have been observed: a decrease in the frequency of T cells in spleen, lymph nodes, and graft and an unresponsiveness of splenic T cells in contact to alloantigens (15, 58). This decrease of T-cell proliferation could be related to several tolerogenic DC molecules that lead to apoptosis, anergy, or hyporesponsivenes. There are many proposed mechanisms used by tolerogenic DCs to explain their tolerogenic activity, including contact-dependent and contact-independent mechanisms. Contact-dependent mechanisms include molecules, such as programmeddeath-ligand 1 (PD-L1), Fas-Ligand (Fas-L), inducible T-cell costimulator-ligand (ICOS-L), but also other molecules, such as immunoglobulin-like transcript-2 (ILT-2), ILT-3, ILT4, HLA-G, and others. Contact-independent mechanisms could be classified into immunomodulatory cytokines, such as IL-10 and TGF-β, or enzymes that generate immunomodulatory molecules or related to nutrient deprivation, such as IDO, heme-oxygenase-1 (HO-1), inducible nitric oxide synthase (iNOS), and arginase 1 (Arg1) (**Figure 2**) (72).

### Contact-Dependent Mechanisms

Contact-dependent mechanisms refer to those mechanisms that need contact between lymphocyte and DCs. The inhibition of proliferation through anergy, hyporesponsiveness, or apoptosis and the differentiation of regulatory cells depend in part of the combination of surface molecules and signal integration between both cells (73). As the different types of tolerogenic DCs have different combinations of inhibitory molecules, the following description is based uniquely on contact-dependent mechanisms observed in transplantation models no matter of tolerogenic DCs type.

Inducible T-cell costimulator-ligand, expressed in immature DC, could interact with ICOS expressed by T cells in order to induce a hyporesponse which is not recovered after restimulation (74). However, a recent study in NHP in kidney transplantation using a combinatorial therapy with belatacept and ICOS-Ig human Fc fusion protein, showed no improvement of allograft survival (75).

Another well-known immunomodulatory molecule related to allograft survival and present in DCs is PD-L1. The blockade of PD-L1 accelerates skin allograft rejection in a similar way to that of anti-CTLA4 treatment (76). Similarly, the use of anti-PD-L1 antibody accelerates heterotropic cardiac allograft rejection, abrogating the effect of cytotoxic T-lymphocyte associated protein-4 (CTLA-4)-Ig (77). Moreover, a recent study showed that DCs transfected with adenovirus coding for PD-L1 was able to induce an increase of kidney allograft survival in fully mismatched rats. This improvement was associated with impairment in CD8<sup>+</sup> T-cell proliferation and a decrease in pro-inflammatory cytokine production (78).

Interaction between ILT-2/ILT-4 and HLA-G in tolerogenic DC, has been shown to impair allogeneic T-cell proliferation. Nevertheless, ILT4-HLA-G pathway is more related to Treg generation (79).

Fas-ligand is another contact-dependent molecule that impairs T cell response *via* the induction of apoptosis. A study using BMDCs transfected with pBK-CMV coding for Fas-L demonstrated that these cells were able to improve cardiac allograft survival in a mouse model and to inhibit allogeneic MLR proliferation through apoptosis induction (80).

### Immunomodulatory Cytokines

Cytokines related to tolerogenic DCs, such as IL-10, TGF-β, and others have been associated with several immunomodulatory functions, such as DCs impairment of maturation, inhibition of T-cell proliferation, and regulatory cell induction.

IL-10 is a well-known immunomodulatory cytokine that has been shown to be essential for the differentiation of several regulatory populations. IL-10 activates the tyrosine kinase IL-10 receptor leading to an activation of signal transducer and activator of transcription 3. This allows an activation of the suppressor of cytokine signaling 3 that inhibits NF-κB translocation leading to a hyporesponsiveness to pro-inflammatory stimuli (81). IL-10 is expressed by tolerogenic DCs under different dynamics depending on the type of tolerogenic DCs. For example, it has been shown that MoDCs generated with IL-10 spontaneously secrete IL-10 (82). However, other Tol-MoDCs require proinflammatory stimulation to produce IL-10, such as Dex- and VitD3- generated TolDCs (44). IL-10 leads to a state of anergy of human CD4<sup>+</sup>T cells in allogeneic MLR and also after polyclonal stimulation with anti-CD3 antibody (83).

TGF-β, for its part, is a pleiotropic cytokine related to immunosuppression. In one hand, TGF-β impairs both CD4<sup>+</sup> and CD8<sup>+</sup> T cell differentiation, activation, and proliferation, and in the other hand, it promotes Treg expansion. In fact, it has been shown that the lack of TGF-β signaling leads to the development of autoimmune inflammatory disease due to an uncontrolled CD4<sup>+</sup> activation (84). Moreover, it has been shown that Tol-BMDCs secrete TGF-β and this expression plays a crucial role in tolerance induction tolerance in several models (85). In cardiac allograft model in rat, the induction of tolerance by LF15-0195 is associated with an increase in *tgfb* expression in allograft of tolerant rats. Moreover, the adoptive transfer of splenocytes from tolerant rats to syngeneic rats receiving cardiac allograft and treated with Rapa in the presence or absence of anti-TGF-β blocking Ab showed that the tolerance was transferred and partially mediated through TGF-β (86).

Apart from classical immunomodulatory molecules, some other cytokines are potentially involved in tolerogenic DCs mechanisms. Among these cytokines, two of them share the Epstein–Barr virus-induced gene 3 (EBI3) monomer, IL-35, and IL-27. IL-35, a heterodimer of EBI3 and IL12p35, is related to immunosuppressive activity. Il-35 is mainly secreted by Treg although several studies demonstrated that APCs are also able to produce this cytokine. In fact, it has been shown that IL-35, but not other IL-12 members, is produced by Tol-BMDCs generated with Dex. In this study, the authors showed that the silencing of *Il12a* (IL-12p35) partially impaired the inhibitory effect of Tol-BMDC toward CD4<sup>+</sup> T cells (87). On the other hand, IL-27 is a heterodimer composed by EBI3 and IL27p28 that acts through IL-27R (gp130 and WSX1). IL-27 impaired several pro-inflammatory functions leading to a reduced effector T-cell response, a control of neutrophil migration, and an impairment of oxidative burst (88). Nevertheless, it has been suggested a dual role for IL-27 as it displayed a suppressive role in EAE model (89), but enhanced CD8<sup>+</sup> T cell anti-tumor activity in other models (90). In transplantation, IL-27 has an important relevance combined with TGF-β1. It has been demonstrated that the overexpression of IL-27 through injection of AAV-IL27 combined with Rapa improved cardiac allograft survival (86). However, monomeric function of EBI3 has been related also with tolerogenic potential in Tol-BMDCs. In fact, in heart allograft rodent model, our work highlighted that mice treated with autologous Tol-BMDCs and low dose of IS displayed an increase of splenic TCRαβ+CD3<sup>−</sup>CD4<sup>−</sup>NKRP1<sup>−</sup>DN T cells expressing high amounts of IFNγ. The increase of this double-negative regulatory population and the allograft survival were related to the EBI3-expressing autologous Tol-DCs. We showed that *in vivo* blockade of either EBI3 or IFN-γ leads to allograft rejection, demonstrating that these molecules are playing a critical immunoregulatory role in this model of allograft tolerance (14).

### Nutrient Deprivation and Other Mechanisms

On the other side, other mechanisms involving interaction between cells or nutrient competition have been observed in transplantation models for several years. These mechanisms open a new perspective on the understanding of graft microenvironment. Among these distinct mechanisms, IDO, iNOS, Arg1, and HO-1 have been related to the impairment of T-cell proliferation.

Inducible nitric oxide synthase and Arg1 are two enzymes commonly associated with macrophages. iNOS is an enzyme that metabolizes arginine and produce nitric oxide (NO) and citrulline, while arginase metabolizes arginine to ornithine and urea. Usually iNOS is known as a M1 macrophage marker and it is induced by pro-inflammatory stimuli, such as IFN-γ. The production of NO by macrophages is usually associated with pro-inflammatory response because this molecule belongs to the Reactive Nitrogen Species (RNS) family that is able to peroxidize membrane lipids in order to eliminate the inflammatory agent. On the other hand, the production of ornithine by M2 macrophages leads to the synthesis of L-Proline, which is essential for collagen production in the resolution of the inflammation (91, 92). However, it has been shown in DCs that these molecules are related to the inhibition of T-cell proliferation. To verify the implication of L-arginine in tolerance and transplantation, several studies were performed. In transplantation, a study demonstrated that the hypoproliferation of T cells isolated from grafted rats treated with Tol-BMDCs was induced by iNOS. Indeed, the use of an iNOS inhibitor (L-NMMA) allowed recovery of T-cell proliferation in treated mice (58). These results showed that iNOS was involved in allograft survival in this model. Similarly, another study demonstrated the relevance of L-arginine metabolism through iNOS and Arg1 in Tol-BMDCs. In this work, tolerogenic DCs were differentiated with retinoic acid (RA) and pulsed with OVA peptide in order to induce *in vivo* lymphoproliferation. The authors showed that Inos−/− RA do not display tolerogenic potential *in vivo* in the presence of OT-II cells. This study corroborated the results observed in transplantation models (93).

Indoleamine 2,3 dioxygenase and tryptophan metabolism, have been suggested as essential factors to inhibit T, B, and NK proliferation and to induce regulatory cells. Paradoxically, it has been shown that IDO is also essential for pro-inflammatory differentiation of DCs (94). A study performed by transfecting human DCs with adenovirus coding for IDO demonstrated that these cells were able to impair T-cell proliferation. Moreover, the study showed that this effect was led by the production of several metabolites of the Kynurenine pathway including kynurenine, 3-hydroxykynurenine, 3-hydroxyanthranilic acid, but not anthranilic acid nor quinolinic acid (95). Similarly, recent findings demonstrated that IDO<sup>+</sup>BMDCs improved heart allograft survival in rodent models associated with an impairment of CD4<sup>+</sup> response and an increase of apoptosis (96).

Heme-oxygenase-1 is an enzyme that catalyzes the conversion of Fe-Protoporphyrin-IX (Heme group) to biliverdin, ferrous ion, and carbon monoxide (CO) (97). CO is usually associated with protective anti-apoptotic effect in a large range of cells, but in lymphocytes, it is usually associated with impaired proliferation and impaired production of inflammatory cytokines (98, 99). The use of a HO-1 inductor (cobalt protoporphyrin, CoPP) or HO-1 product CO, was already tested in pancreatic islet allograft in mice. Both, the pretreatment of allograft or the pretreatment of recipient with CO or CoPP result in an improvement of allograft survival. Moreover, the delay of graft rejection was even more significant when both recipient and allograft were treated (100). Like IDO, HO-1 expression is associated to DC maturation. Indeed, HO-1 is expressed in immature DCs, but not in mature DC. Our group demonstrated that immature DCs stimulated with the HO-1 inductor CoPP preserve an immature phenotype with a low production of IL-12p70, a high expression of IL-10, and were able to impair allogeneic T-cell proliferation in humans and rats (101). Based on these results and the observation that Tol-BMDCs expressed HO-1, we then investigated the role of HO-1 in the protective effect of Tol-BMDCs in our transplantation model of heart allograft in rats. Our results highlighted that the co-treatment of grafted rats with ATDC and an HO-1 inhibitor (tin protoporphyrin IX, SnPP), impaired the beneficial effect of autologous Tol-BMDC treatment. These results suggest that HO-1 is involved in the improvement of allograft tolerance mediated by autologous Tol-BMDCs in this model (102).

Other molecules, such as thrombospondin-1 (TSP-1), PGE2, and adenosine could also influence the tolerogenic potential of tolerogenic DCs in transplantation. To test the role of these molecules in tolerance, a study was performed to compare human Tol-MoDCs differentiated with IL10, IL10/TGF-β, and IL10/IL-6. The results demonstrated that only Tol-MoDCs generated with IL10/ TGF-β lost the suppressive potential *in vitro* in the presence of ARL67156 (CD39 inhibitor) or Indomethacin (PG inhibitor synthesis). However, IL-10 and TSP-1 inhibitors impaired tolerogenic potential in IL10 differentiated-DCs and IL10/IL6-DCs (103).

In conclusion, different types of tolerogenic DCs have different types of immunosuppressive mechanisms to elicit T-cell hypoproliferation.

### REGULATORY CELL INDUCTION

### Induction of CD4**+** Treg Cells

Nowadays, the main goal in post-transplantation therapy is to avoid chronic rejection. To be efficient in the long term, it is essential to induce regulatory cells. Different types of regulatory cells induced or expanded by tolerogenic DCs were described in several animal models and were also observed in the first clinical trials. Among them, the main ones are Tr1 cells, induced CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg, CD8<sup>+</sup>Treg, CD3<sup>+</sup>CD4<sup>−</sup>CD8<sup>−</sup>Treg (104) and Breg (**Figure 3**) (32).

The important role of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi T cells has already been demonstrated in transplantation. Indeed, it was highlighted that the transplantation of skin allografts from tolerant mice onto new recipients, receiving donor or third-party skin allografts leads to the transfer of tolerance. In this study, the authors demonstrated that the donor allograft was not rejected while the third-party one was, meaning that tolerance was led by specific mechanisms (105). CD4+CD25+FoxP3hi Treg are usually associated with several suppressive molecules, such as CTLA-4 and lymphocyte-activation gene 3 (Lag3) that trigger a signal to DCs in order to impair antigen presentation. CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg are also associated with the production of granzyme B and immunomodulatory molecules such as IL-10, TGF-β, and IL-35. Apart from classical contact mechanisms, CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg also compete with effector T cells for IL-2. The deprivation of IL-2 leads to an inhibition of proliferation and apoptosis in effector CD4 T cells (106, 107). Other mechanisms such as the production of adenosine through CD39 and CD73 have also been described (108). In transplantation models, several groups showed that tolerogenic DCs lead to the induction of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg. For example, a study using Tol-BMDCs generated with Rapa have been shown to favor CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg population. In this study, the injection of syngeneic Rapa-DCs pulsed with donor antigens induced tolerance to heart allograft. The adoptive transfer of T cells from tolerant mice to syngeneic mice transplanted with heart allograft from the same source promote an increase in allograft survival due to the transfer of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg (47). Moreover, our recent studies in pancreatic islet allograft transplantation demonstrated that CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg were increased in spleen, lymph nodes and graft of mice treated with autologous Tol-BMDCs and anti-CD3. As mentioned above, this Treg induction was essential for graft prolongation (15).

Other Treg-cell types commonly observed in tolerogenic DC therapy are Tr1 and Tr1-like cells (104). Tr1 are associated with a high expression of IL-10 after specific stimulation and the expression of Lag3 and CD49b markers (109). These Tr1 cells could be induced by Tol-MoDCs generated with IL-10 through the HLA-G/ILT4 pathway (41). Furthermore, it has been shown that Tol-MoDC generated with VitD3 stimulate the generation of Tr1-like cells with a high expression of IL-10 and are able to impair allogeneic T-cell proliferation (110). Interestingly, these Tr1-like cells are induced by contact with Tol-MoDCs notably by PDL-1/PD1 interaction (44). Tr1 have been shown to play an important role sustaining graft CD4<sup>+</sup>CD25<sup>+</sup>FoxP3hi Treg from the spleen through the expression of IL-10 in pancreatic islet allograft (111). These results indicated a network between different tolerogenic populations in order to prolong allograft survival. Another study demonstrated that Tr1-like cells (IL-10+FoxP3−CTLA-4+CD25hiEgr2+ cells) could be differentiated from anergic IL-10−FoxP3−CTLA-4+CD25+Egr2+T cells following their interaction with immature DCs (112). These Tr1-like cells were able to inhibit T-cell proliferation *in vivo* and *in vitro* in an antigen-specific manner (112).

Another CD4 T cell regulatory population potentially associated with tolerogenic DCs are the iTR35 cells. iTR35 are regulatory cells that suppress through IL-35 production but not through IL-10 nor TGF-β. Interestingly, these cells do not express FoxP3. iTR35 are generated *in vitro* with IL-10 and IL-35 but *in vivo* they are present in models such as intestine infection and cancer (113). IL-35 is highly expressed on human Dex induced-tolerogenic DCs after pro-inflammatory stimulation with IFN-γ, CD40-L, or LPS (87). However, the role of IL-35 secreting tolerogenic DCs and iTr35 differentiation *in vivo* remains a conjecture today.

### Induction of Non-CD4**+** Regulatory Cells

Apart from CD4 regulatory cells, there are other regulatory populations involved in TolDC therapy in transplantation such as CD8 Treg and Breg. CD8 Treg cells are less characterized than CD4<sup>+</sup> regulatory cells but they are known to express IL-10 and TGF-β (114). In mice and humans, splenic CD8+CD122+PD-1+ population is associated to an increased allograft survival (115) and also to an anti-inflammatory and suppressive function in other models (116). Moreover, there are several works that have demonstrated a link between tolerogenic DCs and CD8 Treg induction. In humans, a study performed in 2002 showed that antigen-specific CD8 T cells with suppressive activity are generated in healthy volunteers treated with immature DCs pulsed with influenza matrix peptide (117). Another study performed in NHP showed that animals treated with CTLA4-Ig and donor Tol-BMDCs prior to kidney transplantation developed an increased proportion of donor-specific EomesoderminloCTLA4hiCD8<sup>+</sup> T cells. This population is associated with an improvement in allograft survival (118). In our experiments, an increase of CD8<sup>+</sup>CD11c<sup>+</sup> T cells was observed in a model of allograft skin transplantation in mice treated with autologous Tol-BMDCs and low doses of anti-CD3 antibody. The adoptive transfer of CD8<sup>+</sup> T cells purified from these animals was able to prolong allograft survival in new transplanted mice. These results suggest that CD8<sup>+</sup>CD11c<sup>+</sup> T cells induced by autologous Tol-BMDCs could be regulatory cells (16).

Although B cells are well known to promote allograft responses, there is growing evidence that in some circumstances B cells also contribute to the maintenance of transplant tolerance (119). Different populations of regulatory B cells have been described from immature state to plasma cells. Breg effects were described to be mediated by immunomodulatory cytokines such as IL-10, IL-35, and TGF-β, contact-dependent mechanisms, cytotoxic activity mediated by Granzyme B and also by immunoglobulin secretion (120). In transplantation, the ability of B cells to delay graft rejection has already been demonstrated in different rodent transplantation models (121, 122) Furthermore, studies from our team and others demonstrated that the adoptive transfers of splenic B cells from tolerant animals (either total B cells or B cell subsets) were able to delay graft rejection both in heart transplantation in rats and in a mouse model of skin transplantation (123, 124). Other reports highlighted the induction of Breg following Tol-MoDC therapy. Interestingly, in the first phase I clinical trial with Tol-MoDC therapy in type 1 diabetic patients, an increase of B220<sup>+</sup>CD11c<sup>+</sup> population was observed in the blood of patients treated with Tol-MoDCs modified with ODN anti-CD40/CD80/CD86 during the first 6 weeks. This phenotype coincides with a regulatory population (32). Additionally, the same authors demonstrated the contribution of suppressive B cells to control the development of T1D in NOD mice after Tol-BMDC treatment. In this study, the authors suggested that the expansion of pre-existing IL-10<sup>+</sup> B cells and the "*de novo*" generation from CD19<sup>+</sup> B cells could be mediated by the secretion of RA-DCs from Tol-BMDCs (125). However, the link between tolerogenic DCs, regulatory B cells, and allograft tolerance remains unclear.

Altogether these results show that tolerogenic DCs are able to induce regulatory cells leading to a regulatory network that could improve the allograft acceptance.

### WHERE DO WE STAND?

From the first DCs vaccines back in 1995 (33) until today, the expectation on DCs therapy have increased due to the safety and potential demonstrated in animal models and in humans. Nowadays, four clinical trials using Tol-MoDCs in autoimmune diseases have already been completed (32, 48, 66, 126) (**Table 1**).


First clinical trial was performed in insulin-requiring T1D patients. In this clinical trial, seven patients received Tol-MoDCs modified with ODN anti-CD40/80/86 and three were treated with unmodified Tol-MoDCs. An increase in B220 B cells and no adverse effects were observed (32). The second clinical trial using Tol-MoDCs was performed in rheumatoid arthritis patients. In this study, 18 HLA-positive RA patients were divided into two cohorts, patients from the first one received a low dose of Tol-MoDC (one million cells) and the others received high dose (five million cells). Tol-MoDCs used in this study were modified with an NF-κB inhibitor and pulsed with four citrullinated peptides. No adverse effects were observed. Additionally, the authors observed an increase in circulating Treg cells and a decrease in IL-6 expression in T cells in response to vimentin447–455 Cit450 (66). The third clinical trial using Tol-MoDCs was performed in patients suffering from refractory Crohn's disease. In this clinical trial, 12 patients were divided in 6 cohorts, receiving 2, 5, or 10 million Tol-MoDCs in a single dose or biweekly. Despite that no adverse effects were observed in most patients, three of them withdrew the study due to worsening of the disease. Additionally, the authors found an increase in Treg cells and a decrease in IFN-γ in blood (48). Finally, the most recent clinical trial using Tol-MoDCs was performed in rheumatoid and inflammatory arthritis. In this study, 13 patients were divided in four cohorts, receiving 1, 3, or 10 millions cells and three patients receiving saline solution. Tol-MoDCs used in this clinical trial were differentiated using Dex and vitD3 and loaded with autologous synovial fluid. The outcome of this study showed that the treatment was safe and feasible. Moreover hypertrophy, vascularity and synovitis were stable in all cohorts and in placebo-treated patients. Nevertheless, two patients that have received 10 millions cells showed a decrease in synovitis score (126). Apart from these studies, there are many other ongoing clinical trials focused on other pathologies, such as allergy or multiple sclerosis. Among the ongoing clinical trials using Tol-MoDCs, we supervise a phase I/II clinical trial in kidney transplantation at Nantes university hospital (NCT02252055). This trial will evaluate the safety of autologous Tol-MoDCs in patients receiving living donor kidney transplantation and a minimized immunosuppression. In this trial, autologous Tol-MoDCs are generated in the presence of low-dose GM-CSF as the only cytokine used. These Tol-MoDCs are characterized by a weak capacity to stimulate allogenic T cells and a suppression of the proliferation of stimulated T cells. Furthermore, they are resistant to maturation stimuli. Patients receive their Tol-MoDCs

### REFERENCES


the day before transplantation by intravenous route at a dose of one million/kg [for review (34)]. The team of Angus Thomson also evaluate the potential of Tol-MoDC in transplantation. In this trial, patients receive donor-derived Tol-MoDCs one-week prior to liver transplantation (NTC03164265) [for review (127)]. Due to the outcomes of these clinical trials, at least in terms of safety and biological effect, Tol-MoDC therapy appears more and more as an interesting strategy to treat several diseases. However, more clinical trials must be performed in order to find out the adequate dose, injection conditions, and associated drugs to efficiently treat patients.

### CONCLUSION

Tolerogenic DCs have a solid background that corroborates their usefulness in transplantation, but also to treat autoimmunity and allergy diseases. Despite the different methods to generate them and the different models used, the common features of tolerogenic DCs converge in a low expression of costimulatory and presentation molecules, a maturation resistance, a high expression of immunomodulatory molecules, a low expression of pro-inflammatory molecules, and an impairment of T-cell proliferation. Moreover, tolerogenic DCs induce regulatory populations that are related to the protection of allograft in the long term. More importantly tolerogenic DCs have been proved to be safe supporting the feasibility of this cell therapy in humans. Finally, results confirming the efficacy and safety of autologous Tol-MoDC in humans in transplantation will be evaluated in the following years.

### AUTHOR CONTRIBUTIONS

All authors contributed in discussing the topic and wrote the manuscript.

### FUNDING

The work performed in the INSERM U1064 and presented in this review was funded by IMBIO-DC, Fondation Progreffe, DHU Oncogreffe, The ONE Study (FP7-260687), and BIODRIM (FP7-305147) European Union seventh Framework Programs. The work of INSERM U1064 was also supported by funds from IHU-CESTI (Investissement d'Avenir ANR-10-IBHU-005, Région Pays de la Loire and Nantes Métropole) and the Labex IGO project (n° ANR-11-LABX-0016-01).


genotype-positive rheumatoid arthritis patients. *Sci Transl Med* (2015) 7:290ra87. doi:10.1126/scitranslmed.aaa9301


127. Thomson AW, Zahorchak AF, Ezzelarab MB, Butterfield LH, Lakkis FG, Metes DM. Prospective clinical testing of regulatory dendritic cells in organ transplantation. *Front Immunol* (2016) 7:15. doi:10.3389/fimmu.2016.00015

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Marín, Cuturi and Moreau. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Tatjana Nikolic, Leiden University, Netherlands Nick Giannoukakis, Allegheny Health Network, United States*

#### *\*Correspondence:*

*David P. Funda funda@biomed.cas.cz*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 17 November 2017 Accepted: 01 February 2018 Published: 16 February 2018*

#### *Citation:*

*Funda DP, Goliáš J, Hudcovic T, Kozáková H, Špíšek R and Palová-Jelínková L (2018) Antigen Loading (e.g., Glutamic Acid Decarboxylase 65) of Tolerogenic DCs (tolDCs) Reduces Their Capacity to Prevent Diabetes in the Non-Obese Diabetes (NOD)-Severe Combined Immunodeficiency Model of Adoptive Cotransfer of Diabetes As Well As in NOD Mice. Front. Immunol. 9:290. doi: 10.3389/fimmu.2018.00290*

*David P. Funda1 \*, Jaroslav Goliáš1 , Tomáš Hudcovic2 , Hana Kozáková2 , Radek Špíšek3,4 and Lenka Palová-Jelínková3,4*

*<sup>1</sup> Institute of Microbiology of the Czech Academy of Sciences, v.v.i., Prague, Czechia, 2 Institute of Microbiology of the Czech Academy of Sciences, v.v.i., Nový Hrádek, Czechia, 3SOTIO a s., Prague, Czechia, 4Department of Immunology, 2nd Medical School, Charles University, Prague, Czechia*

Tolerogenic DCs (tolDCs) are being researched as a promising intervention strategy also in autoimmune diseases including type 1 diabetes (T1D). T1D is a T-cell-mediated, organ-specific disease with several well-defined and rather specific autoantigens, i.e., proinsulin, insulin, glutamic acid decarboxylase 65 (GAD65), that have been used in animal as well as human intervention trials in attempts to achieve a more efficient, specific immunotherapy. In this study, we have tested tolerogenic DCs for their effectiveness to prevent adoptive transfer of diabetes by diabetogenic splenocytes into non-obese diabetes (NOD)-severe combined immunodeficiency (NOD-SCID) recipients. While i.p. application of tolDCs prepared from bone marrow of prediabetic NOD mice by vitamin D2 and dexamethasone significantly reduced diabetes transfer into the NOD-SCID females, this effect was completely abolished when tolDCs were loaded with the mouse recombinant GAD65, but also with a control protein—ovalbumin (OVA). The effect was not dependent on the presence of serum in the tolDC culture. Similar results were observed in NOD mice. Removal of possible bystander antigen-presenting cells within the diabetogenic splenocytes by negative magnetic sorting of T cells did not alter this surprising effect. Tolerogenic DCs loaded with an immunodominant mouse GAD65 peptide also displayed diminished diabetes-preventive effect. Tolerogenic DCs were characterized by surface maturation markers (CD40, CD80, CD86, MHC II) and the lipopolysaccharide stability test. Data from alloreactive T cell proliferation and cytokine induction assays (IFN-γ) did not reveal the differences observed in the diabetes incidence. Migration of tolDCs, tolDCs-GAD65 and tolDCs-OVA to spleen, mesenteric- and pancreatic lymph nodes displayed similar, mucosal pattern with highest accumulation in pancreatic lymph nodes present up to 9 days after the i.p. application. These data document that mechanisms by which tolDCs operate *in vivo* require much better understanding for improving efficacy of this promising cell therapy, especially in the presence of an antigen, e.g., GAD65.

Keywords: type 1 diabetes, cell therapy, autoantigen, tolerogenic, dendritic cells, glutamic acid decarboxylase 65, non-obese diabetes-severe combined immunodeficiency mouse, non-obese diabetes mice

### INTRODUCTION

Type 1 diabetes (T1D) is a T cell-mediated disease, in which both CD4 and CD8 T cells are necessary and sufficient in precipitating the disease by targeting very specifically relatively small volume of highly specialized beta-cells within Islets of Langerhans (1, 2). Non-obese diabetes (NOD) mice and NOD-severe combined immunodeficiency (NOD-SCID) mice represent well studied and frequently used animal models of T1D. While the NOD spontaneous mouse model allows to study the natural course of the disease in its complexity, NOD-SCID mice with adoptively cotransferred diabetogenic NOD splenocytes are often used for testing regulatory/protective capacity of various cell subsets in a shorter timeframe (3).

Dendritic cells (DCs) are highly effective specialized antigen-presenting cells (APCs) and central regulators of immune responses—they are important for induction of effector immune responses, but depending on their developmental stage or environment/culture conditions, also promote tolerance by various mechanisms: T-cell anergy, T-cell deletion, induction of different subsets of Tregs, such as CD8<sup>+</sup> Tregs, Foxp3<sup>+</sup> Tregs, or Tr1 cells (4), or even induction of a Th2 shift (5) or Bregs (6).

The first animal study dealing with DCs in T1D prevention showed that DCs isolated from pancreatic and not other lymph nodes lowered diabetes incidence when reinjected to 4-week-old NOD mice (7). It has been shown that bone marrow-derived tolerogenic DCs (tolDCs) generated in the presence of GM-CSF and IL-4 in an antigen-nonspecific manner displayed diabetespreventive properties (8, 9). Antigen-nonspecific treatment with a pegylated TLR7 ligand *in vivo* induced tolDCs and decreased diabetes in NOD mice (10). Administration of DCs prepared in the presence of interleukin 10 (IL-10) with (11) or without (12) antigen supply both prevented diabetes and insulitis in NOD mice. In addition, tolDCs pulsed with apoptotic bodies containing beta-cell antigens decreased diabetes and insulitis in a transgenic NOD model of accelerated diabetes (13). While data from Pujol-Autonell et al. documented that reverting diabetes in already diabetic animals might be difficult (14), genetically engineered bone marrow-derived DCs transduced with IL-4 were able to prevent diabetes in 12-week-old prediabetic NOD mice with advanced insulitis (15).

Thus, tolDCs represent a promising strategy in T1D prevention at high-risk individuals or even treatment of the disease. The first human phase I trial of autologous tolDCs in T1D was completed (16, 17) and another, based on proinsulin-loaded tolDCs, has been opened (18).

Apart from the efficacy of tolDCs to suppress the disease in animal models, preferably also at later stages before or even after clinical onset of T1D, several other important parameters must be taken into account, such as their stability, survival, expression of costimulatory and homing molecules, migration, dying pathway, antigen-specificity or requirement, and optimal application route (4, 19). We have been involved in testing and optimizing tolDC protocol based on GM-CSF and IL-4 cell culture with added dexamethasone and vitamin D2 followed by activation of tolDCs by lipopolysaccharide (LPS) analog monophosphoryl lipid A (MPLA). This protocol was developed according to the good manufacturing practice standards for preparation of human tolDCs that are stable under inflammatory conditions (20). Indeed, it would be desirable to make this protocol antigenspecific by using safely a beta-cell specific antigen for targeting the pathological immune reaction more effectively, as it has been researched in experimental autoimmune encephalomyelitis (EAE) (21, 22) or experimental arthritis (23, 24), but less clear-cut in case of T1D (8, 9, 11, 13).

Thus, the initial aim of this study was to test this human tolDC protocol in NOD-SCID mice in an antigen-specific manner by using mouse recombinant glutamic acid decarboxylase 65 (GAD65) naturally processed by tolDCs. Surprisingly, GAD65-loaded tolDCs (tolDCs-GAD65) while keeping their surface characteristics as well as their allogeneic proliferative and cytokine induction properties lost their diabetes-preventive effect. Diabetes incidence was also assessed in the NOD mouse model. Some possible mechanisms, other antigens, culture conditions as well as migration patterns are addressed or excluded in this study.

### MATERIALS AND METHODS

The minimum information about tolerogenic antigen-presenting cells (MITAP) checklist was followed for the preparation of this manuscript (25).

### Animals

Female NOD, NOD-SCID, and C57BL/6 mice were purchased from Taconic (Albany, NY, USA) whereas female C57BL/6 mice were obtained from the animal facility of the Institute of Physiology, Czech Acad. Sci., Prague, Czech Republic and used in experiments as described below at 6–13 weeks of age. The mice were maintained in the specific pathogen-free animal facilities under standard light- and climate-controlled conditions, fed standard Altromin 1414 diet, and water was provided *ad libitum*. All experiments were approved by our

**Abbreviations:** cDC, control matured bone marrow-derived dendritic cell; iDC, immature bone marrow-derived dendritic cell; ILN, systemic inguinal lymph node; MLN, mesenteric lymph node; pept, GAD65-immunodominant peptide no. 35; PLN, pancreatic lymph node; tolDC, tolerogenic dendritic cell.

institutional animal ethics office (Laboratory Animal Care and Use Committee of the Institute of Microbiology v.v.i., Academy of Sciences of the Czech Republic, approval ID: 94/2006 and 244/2009) in strict accordance with the Federation of European Laboratory Animal Science Associations guidelines. Endpoint criteria were established to minimize suffering and ensure animal welfare.

### DC Generation

Mouse bone marrow dendritic cells were generated from femur and tibia of 8- to 10-week-old female NOD mice, which were surgically removed postmortem. The bone marrow was flushed with a syringe/needle combination. Erythrocytes were lysed using red blood cell lysing buffer (Sigma-Aldrich, St. Louis, MO, USA), isolated cells were washed and counted for absolute live cells quantity but without documenting their morphology. The fresh isolated cells were subsequently cultured (37°C, 5% CO2) in Petri cell-culture dishes (90 mm in diameter) in complete prewarmed l-glutamine containing RPMI-1640 (Lonza, Verviers, Belgium) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco-Life Technologies, Paisley, UK), 100× diluted NEM-Non-essential amino acid (Sigma-Aldrich), 1 µM sodium pyruvate (Sigma-Aldrich), 50 IU/mL penicillin, 50 µg/mL streptomycin, and 50 µM 2-β-mercaptoethanol. In other experiments, cells were cultured in serum-free (SF) CellGro medium (CellGenix, Freiburg, Germany) with the same supplements as for RPMI-1640 except FBS. Cells were plated at a density of 4 × 106 cells/10 mL in the corresponding medium in the presence of GM-CSF (20 ng/mL) and IL-4 (4.5 ng/mL; PeproTech, Rocky Hill, NJ, USA) for 6 days. Fresh medium (10 mL) was added on day 3. At day 6, half (10 mL) of the medium was harvested, collected cells were counted, and resuspended in 10 mL of fresh medium and added back into the culture. Thus half of the medium was replaced with fresh one. Tolerogenic DCs were induced by adding dexamethasone (1 µM; Medochemie, Limassol, Cyprus) and vitamin D2 (1.5 ng/mL; Zemplar, AbbVie, North Chicago, IL, USA) on day 6, whereas immature DCs (iDCs) and control matured DCs [control matured bone marrow-derived dendritic cell (cDCs)] were generated without these tolerogenic factors. For antigen loading of tolDCs the mouse recombinant GAD65 was obtained from Sino Biological, Beijing, China, whereas Ovalbumin EndoGrade (OVA) was purchased from Hyglos, Regensburg, Germany and immunodominant peptide no. 35 of GAD65 sequence (purity 97%) from ThinkPeptides, Oxford, UK. At day 7, nonadherent cells were collected, washed, counted, and plated at a density of 1 × 106 cells/mL in fresh medium on a 6-well plate (4 × 106 cells per well). TolDCs were left unpulsed or loaded with GAD65 (tolDC-GAD65, 2 or 1 µg/mL), OVA (tolDC-OVA, 1 µg/mL), or GAD65-immunodominant peptide no. 35 (tolDC-pept, 1 µg/mL). After 4 h, all types of tolDCs as well as cDCs were finally activated with 2 µg/mL VacciGrade MPLA from S. minnesota R595 (MPLA; InvivoGen, Toulouse, France) for 22 h. At the end of the cell cultivation, the nonadherent cells were collected, counted for absolute live cells, and in the same culture medium immediately processed for follow-up experiments.

### Adoptive Transfer and Diabetes Monitoring

7- to 8-week-old NOD-SCID females were used as recipients in adoptive cotransfer experiments. Diabetogenic splenocytes were isolated from 12- or 13-week-old prediabetic NOD female mice. At day 8 of a cell culture, 3 × 106 of live tolDCs, tolDCs-GAD65, tolDCs-OVA, or tolDCs-pept were resuspended in Phosphate Bovine Saline (PBS, Lonza) together with 5 × 106 live diabetogenic splenocytes (erythrocytes were lysed with red blood cell lysing buffer and cells washed twice in PBS) and injected i.p. (left side of the belly) in a final volume of 300 µL of PBS. The Control group was injected with 5 × 106 diabetogenic splenocytes in PBS. In another experiment, T cell-enriched splenocytes were prepared by a negative selection using EasySep Mouse T cell Enrichment Kit (Stemcell Technologies, Vancouver, BC, Canada). The T cell-enriched fraction had purity >92%. Equivalent of 33% of 5 × 106 whole splenocytes, i.e., 1.65 × 106 T cells were mixed with 3 × 106 tolDCs or tolDCs-GAD65, and injected i.p. in a final volume 300 µL of PBS to NOD-SCID mice. The same procedure was used for adoptive cotransfer of tolDCs generated in SF CellGro medium. All recipient NOD-SCID mice were monitored for diabetes incidence weekly for min of 12 weeks. Tolerogenic DCs as well as tolDCs-GAD65 and tolDCs-OVA were also tested in the spontaneous NOD mouse model by a single dose of 3 × 106 of live cells injected i.p. at age of 4 weeks. NOD mice were monitored for diabetes incidence from 8 weeks until age of 310 days. The diabetes onset was monitored once weekly from tail vein blood with the glucometer Freestyle Lite (Abbott Diabetes Care Ltd., Witney, UK) and diagnosis of diabetes was based on two consecutive blood glucose readings >12 mM in 3 days. The first reading was then used as the date of diabetes onset. Neither a NOD-SCID nor a NOD mouse displayed transient glycemia over 12 mM in this study. The glycemia measurement values for individual NOD mice are provided in the Table S1 in Supplementary Material.

### Flow Cytometry

Cells were stained with the following fluorochome-conjugated monoclonal antibodies: anti-CD3 (145-2C11), CD4 (GK1.5), CD8a (53-6.7), CD11c (N418), CD40 (1C10), CD80 (16-10A1), CD86 (GL1), MHC class II (I-A/I-E) (MS/114.15.2), CD103 (2E7), c-kit (ACK2), IL7Rα (A7R34), CCR5 (HM-CCR5), CCR7 (3D12), IFN-γ (XMG1.2), AnnexinV, and Fc block CD16/CD32 (Thermo Fisher Scientific, Waltham, MA, USA). Cells were incubated in PBS containing 2% FBS when stained for surface markers. Propidium iodide or Hoechst33342 were used for exclusion of dead cells or to assess the proportion of dead cells. HEPES buffer was used for AnnexinV staining. Cells were washed three times, then stained and kept in the HEPES buffer for flow cytometry analysis. For intracellular detection of IFN-γ on day 5 of allogeneic coculture, cells were restimulated *in vitro* with 25 ng/mL of phorbo-12-myristate-13-acetate and ionomycin (1 µg/mL, Sigma-Aldrich) for 4 h in the presence of monensin (2 µM, Thermo Fisher Scientific). Cells were first stained for surface markers then fixed/permeabilized with the Cytofix/Cytoperm kit (Thermo Fisher Scientific) following the manufacturer's instruction. Unstimulated cells cultured in the presence of monensin were used as controls. Isotype control antibodies were included to determine the amount of nonspecific binding. Data were acquired by LSR II flow cytometer (BD Bioscience, San Diego, CA, USA) and analyzed using FlowJo software (Tree Star, Ashland, OR, USA).

### Stability Test and Cytokine Production

Stability test was carried out by stimulation of iDC, cDC, and all derived types of tolDCs with 1 µg/mL LPS from *Escherichia coli* 0111:B4 (LPS; Sigma-Aldrich) for additional 24 h. Cells were seeded in 96-well U-bottom plates at density 2 × 105 /200 μL. Unstimulated cells were used as controls. Expression of surface maturation markers (CD40, CD80, CD86, MHC II) was measured on live cells by flow cytometry. IL-10 and IL-12p70 were measured in tissue culture supernatants after MPLA activation or after the stability test (24 h LPS at 1 µg/mL) by ELISA DuoSet kits (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's instructions.

### Allogeneic DC/T Cell Cultures

To assess alloproliferative responses splenocytes isolated from C57BL/6 female mice were labeled for 5 min with 3 µM 5,6-carboxylfluorescein diacetate succinimidyl ester (CFSE; Thermo Fisher Scientific) and according to the manufacturer's instructions. Labeled splenocytes were seeded in 96-well U-bottom plates at a concentration 2 × 105 /200 μL and cocultured in complete RPMI-1640 medium with 2 × 104 /200 μL (10:1 ratio) iDCs, cDCs, or different types of tolDCs for 3 and 5 days at 37°C and 5% CO2. CFSE dilution in live CD3<sup>+</sup> T cells was assessed by flow cytometry. Unstained splenocytes cocultured with DCs were used in the same setting for allogeneic induction of IFN-γ after 5 days. The expression of IFN-γ was assessed by intracellular staining in CD3<sup>+</sup>CD4<sup>+</sup> cells and flow cytometry.

### Autologous CD8**+** T Cell-Mediated *In Vitro* Killing of tolDCs

Splenocytes were isolated from NOD females and CD4<sup>+</sup> or CD8<sup>+</sup> were isolated by negative magnetic selection using EasySep Mouse CD4<sup>+</sup> and CD8<sup>+</sup> T Cell Isolation Kit (Stemcell Technologies) according manufacturer's instructions. Dendritic cells (iDCs, cDCs, tolDCs, tolDCs-GAD65, and tolDCs-OVA) were cocultured with autologous CD4<sup>+</sup> or CD8<sup>+</sup> splenic T cells at a 1:1 ratio and at a concentration 1 × 106 /mL in complete RPMI-1640 medium for 4, 8, 12, and 24 h. DCs without T cells were cultured as controls. In another experiment dendritic cells and autologous splenocytes were cultured in the same ratio (3:5) as used for i.p. administrations to NOD-SCID mice. Cell death was measured by AnnexinV and Hoechst33342 staining on CD3- CD11c<sup>+</sup> cells by flow cytometry.

### DC Migration

Unloaded tolDCs and tolDCs-GAD65 were generated from 8-week-old-NOD females. The PKH26 Red Fluorescent Cell Linker kit (Sigma-Aldrich) was used for labeling DCs for *in vivo* migration experiment according to the manufacturer's protocol. Briefly, cells were washed in RPMI-1640 medium without FBS, resuspended in Diluent Solution C (1 mL of diluent C/1 × 107 cells), and stained with PKH26 for 5 min in room temperature with periodic mixing. The staining reaction was stopped by addition of FBS-supplemented RPMI-1640 medium and in which cells were then washed three times. Labeled tolDCs and tolDCs-GAD65 were applied i.p. (left side of the belly) to 6-week-old NOD female mice at dose of 5 × 106 cells. The control group was injected with unlabeled tolDCs. Cell suspensions from spleen, mesenteric lymph nodes (MLNs), pancreatic lymph nodes (PLNs), and systemic inguinal lymph nodes (ILNs) were prepared after 3, 5, 7, 9, and 12 days and live PKH26<sup>+</sup>CD11<sup>+</sup> were detected by flow cytometry.

### Statistical Analysis

All data are expressed as the mean ± SEM values. Statistical analyses were performed using GraphPad Prism 4 (GraphPad Software, La Jolla, CA, USA). The unpaired *t*-test and one-way ANOVA followed by Tukey's multiple comparison posttest were used for evaluation of data from multiple measurements from two or multiple groups. Differences were considered statistically significant when *p*-value was <0.05. The cumulative diabetes incidence was assessed using the Kaplan–Meier estimation and contingency tables. Log-rank test and Chi-square test were used for comparisons between. The *p*-values were compensated for multiple comparison (Bonferroni) of survival curves for each experiment.

### RESULTS

### Tolerogenic DCs but not tolDCs-GAD65 Prevent Diabetes in the Adoptive Transfer Model of NOD-SCID Mice

In this initial experiment we tested whether i.p. administration of autoantigen-loaded tolDCs could increase the efficacy of the diabetes prevention by tolDCs in the NOD-SCID model of adoptive transfer of diabetes. Surprisingly, we found out that tolDCs that were cultured from day 7 with mouse GAD65 (2 µg/mL) completely lost their diabetes-preventive properties when cotransferred with diabetogenic splenocytes from 13-week-old prediabetic NOD females to NOD-SCID recipients (*n* = 12). As shown in **Figure 1**, i.p. application of 3 × 106 unloaded tolDCs together with 5 × 106 diabetogenic splenocytes led to substantial, although not statistically significant, reduction (75% to 42%) of diabetes onset in NOD-SCID recipients compared to the Control group injected with only diabetogenic splenocytes. In contrast, i.p. administration of tolDC exposed to mouse GAD65 (tolDC-GAD65) not only did not reduce diabetes incidence but, on the contrary, led to a more rapid onset and higher diabetes incidence than in the Control group receiving only diabetogenic splenocytes (**Figure 1**). Thus, the diabetes incidence in tolDC-GAD65 group was significantly higher than in group cotransferred with unloaded tolDCs (*p* = 0.0159). There was no difference in diabetes incidence between the tolDC-GAD65 and Control groups (*p* = 0.6651) (**Figure 1**).

The tolDCs and tolDCs-GAD65 were prepared by a standard 8-day protocol using RPMI-1640 medium supplemented with 10% FBS. Dexamethasone and vitamin D2 were added on day 6, mouse GAD65 on day 7. Dendritic cells were activated for the last 22 h by 2 µg/mL of VacciGrade MPLA. This unexpected result led us to repeat this initial experiment and carry out several modifications in order to clarify this phenomenon.

### Both GAD65 and OVA Abrogate Diabetes-Preventive Properties of tolDCs and This Effect Does Not Seem to Be due to the Presence of APCs within Diabetogenic Splenocytes

To confirm and extend the data from **Figure 1**, we have repeated these groups in a second adoptive cotransfer experiment in NOD-SCID mice. In order to clarify whether the effect is autoantigen (i.e., GAD65) specific we have also included a group of tolDCs loaded with a naturally processed control protein—OVA (tolDC-OVA). These tolDCs were again prepared in serumsupplemented RPMI-1640 medium, but this time loaded with a lower dose (1 µg/ml) of GAD65 or OVA on day 7, and activated by 22-h culture with 2 µg/ml of VacciGrade MPLA. Similar to previous experiment, unloaded tolDCs again markedly lowered diabetes incidence in NOD-SCID mice compared to the Control group injected with diabetogenic splenocytes alone (**Figure 2**), but the difference was not statistically significant due to multiple comparisons of six groups (*n* = 8). Tolerogenic DCs pulsed with GAD65, but also OVA, both failed to substantially lower the 100% diabetes transfer found in the Control group. The tolDC–OVA group displayed slightly lower diabetes incidence than the tolDC-GAD65 group, but this disease prevention was

Figure 2 | Antigen-loaded tolDCs fail to lower induction of diabetes in non-obese diabetes (NOD)-severe combined immunodeficiency (NOD-SCID) mice by transfer of diabetogenic NOD splenocytes but also by transfer of splenic T cells. Diabetogenic splenocytes (5 × 106 per mouse) were isolated from 13-week-old prediabetic NOD females (*n* = 11). T cells were enriched (cell purity >92%) by negative selection (EasySep T cell Enrichment kit, Stemcell Tech.), and equivalent of 33% of splenocytes, i.e., 1.65 × 106 T cells per mouse were used for diabetes induction in NOD-SCID recipients. Tolerogenic DCs, glutamic acid decarboxylase 65 (GAD65)- (1 µg/mL) or OVA- (1 µg/mL) loaded tolDCs were generated from bone marrows of 8- to 10-week-old NOD females by cultivation in the presence of GM-CSF and IL-4 followed by additions of dexamethasone/vitamin D2 and stabilized by monophosphoryl lipid A (MPLA). Diabetogenic splenocytes or enriched T cells (groups marked as T cell+) were resuspended in phosphate bovine saline (PBS) together 3 × 106 tolDCs and injected i.p. (left side of the belly) in a volume of 300 µL PBS to 8-week-old NOD-SCID female recipients (*n* = 8). Diabetogenic splenocytes in PBS were used as the Control group. Data are presented as cumulative diabetes incidence in NOD-SCID recipients, *p*-values were compensated for multiple comparisons.

not at all statistically significant when compared to the Control group (**Figure 2**).

Next, we wanted to test whether other APCs present within the preparation of whole diabetogenic NOD splenocytes may be responsible for an aberrant immunogenic presentation of previously processed antigen (GAD65) by tolDCs. Thus, instead of whole splenocytes, we have used T-cell fraction, prepared by a negative T-cell enrichment (Stemcell Tech.) with T-cell purity over 92%. Approximately equivalent dose of T cells (1.65 × 106 ) as present within the 5 × 106 splenocytes was used for disease induction in the NOD-SCID recipients (T cell<sup>+</sup> groups).

However, by using T cell-enriched splenocytes (T cell<sup>+</sup>) we observed again the same diabetes-preventive pattern of tolDCs and tolDCs-GAD65. Unloaded tolDCs substantially lowered the effect of diabetogenic T cells (100–50%) and slightly more compared to the whole diabetogenic splenocytes (100–37.5%), whereas autoantigen-loaded tolDCs-GAD65 again failed to prevent diabetes in NOD-SCID recipients. In fact, the course of the disease induction was closest between the Control and tolDC-GAD65 (T cell<sup>+</sup>) groups (**Figure 2**).

These data document, that APCs within diabetogenic splenocytes probably play an unimportant role in abolishing the disease preventive effect when mixed with tolDCs that naturally processed an autoantigen (GAD65). This experiment also showed that the decrease of diabetes-preventive potential of tolDCs does not seem to be GAD65-specific. Although OVA-loaded tolDCs prevented diabetes onset in two of eight mice, this reduction does not represent a substantial disease protection compared to the Control group (**Figure 2**).

## GAD65-Peptide Loaded tolDCs Failed to Prevent Diabetes in NOD-SCID Recipients and Serum-Free Conditions Do Not Alter the Effect of Antigen-Loaded tolDCs

In order to determine in the literature (26) addressed effect of FBS on antigen-specific tolerance induction by tolDCs as well as to reflect the fact that DCs for human trials are prepared in SF conditions, we have compared the diabetes-preventive capacity of unloaded tolDCs cultured in serum-supplemented vs. SF conditions as well as retested effects of tolDCs-GAD65 and tolDCs-OVA in SF conditions (**Figure 3A**). Compared to the Control group of NOD-SCID mice (*n* = 8) administered with only 5 × 106 of diabetogenic splenocytes, antigen-unloaded tolDCs again markedly reduced diabetes incidence in serumsupplemented conditions (100–50% in both **Figures 3A,B**). The difference was even more remarkable (100–50%) due to the course of the diabetes onset and thus statistically significant in SF conditions (*p* = 0.026) (**Figure 3A**).

On the other hand, antigen-loaded tolDCs-GAD65 and tolDCs-OVA prepared in SF conditions did not substantially decrease diabetes incidence in the NOD-SCID recipients, but this time tolDC-GAD65 and tolDC-OVA groups (**Figure 3A**) displayed a reversed pattern as in **Figure 2**. Thus, the presence of FBS did not alter disease-preventive properties of tolDCs. Last but not least, we have also tested not a whole protein, but the GAD65-immunodominant peptide no. 35 (pept) that is identical in murine and human GAD65 (27). As presented in **Figure 3B**, this immunodominant peptide added to tolDCs at concentration of 1 µg/mL and naturally processed by DCs (tolDC-pept) also failed to statistically significantly prevent diabetes in the NOD-SCID recipients in both SF and serum-supplemented conditions. Thus, in all three experiments only antigen-unloaded tolDCs exhibited substantial effects on disease prevention using the NOD-SCID mouse model of adoptive cotransfer of diabetes.

### Diabetes- Preventive Effect of tolDCs and Antigen-Loaded tolDCs in NOD Mice

To further assess the effect of antigen loading and serum-free conditions on the diabetes-preventive properties of tolDCs, selected groups of tolDCs were tested also in the spontaneous NOD mouse model of T1D. While control animals displayed diabetes incidence of 87.5% at age of 310 days, the group injected with a single dose of 3 × 106 unloaded tolDCs showed substantially reduced diabetes incidence to 50%, *n* = 16 (**Figure 4**). Similar effect was observed when using unloaded tolDCs prepared in serum-free conditions, that lowered diabetes incidence to 43.8%, but this group displayed a faster initial onset of diabetes. On the other hand, antigen-loaded tolDC-GAD65 and tolDC-OVA prepared in serum-supplemented conditions did not substantially decreased diabetes incidence compared to the Control group (75 and 62.5% diabetic animals, respectively) (**Figure 4**). Thus, although the comparison of the multiple groups with the Control group was not statistically significant (*p* values corrected for multiple comparisons), data obtained in NOD mice paralleled those from the NOD-SCID model of adoptive cotransfer of diabetes (**Figures 1**, **2** and **3A,B**).

### No Substantial Differences in Phenotype and Stability of tolDCs and Antigen-Loaded tolDCs

Expression of surface maturation markers on DCs such as costimulatory (CD80, CD86), activation (CD40) and antigen-presenting (MHC II) molecules was carried out by flow cytometry. As shown in **Figure 5A**, unloaded tolDCs as well as antigen-loaded tolDCs-GAD65, tolDCs-OVA, and tolDCs-pept exhibited similar pattern of these markers reflecting their immature to semi-matured phenotype in both serumsupplemented and SF conditions. The CD40 expression was substantially increased in cDCs compared to both iDCs and all

Figure 3 | Serum-free cultured tolDCs and GAD65-loaded tolDCs in diabetes prevention using the non-obese diabetes (NOD)-severe combined immunodeficiency (NOD-SCID) model of adoptive transfer of diabetes. Dendritic cells were generated from bone marrows of 8- to 10-week-old NOD females by cultivation in the presence of GM-CSF and IL-4 followed by additions of dexamethasone/vitamin D2 and final maturation with monophosphoryl lipid A (MPLA). (A) Tolerogenic DCs loaded with 1 µg/mL of glutamic acid decarboxylase 65 (GAD65) (tolDC-GAD65) or OVA (tolDC-OVA) were prepared in SF medium, whereas tolDCs were cultured in both serum-supplemented (10% fetal bovine serum RPMI-1640) and SF media (tolDC SF). (B) In another experiment, unloaded tolDCs were compared to tolDCs loaded with 1 µg/mL of the GAD65-immunodominant peptide no. 35 (tolDC-pept) and prepared in both serum-supplemented and SF media. Diabetogenic splenocytes (5 × 106 per mouse) from 12-week-old prediabetic NOD females (*n* = 10) and above listed groups of tolDCs (3 × 106 ) were mixed and applied i.p. (left side of the belly) in a volume of 300 µL phosphate bovine saline (PBS) to 7-week-old NOD-SCID female recipients (*n* = 8). Diabetogenic splenocytes in PBS were used as the Control group in both experiments. Data are presented as cumulative diabetes incidence in NOD-SCID recipients and *p*- values were compensated for multiple comparisons (A) Control PBS vs. tolDC SF: \**p* = 0.026.

groups of tolDCs in serum-supplemented conditions (*p* < 0.01, *p* < 0.001), whereas the differences were not significant in SF conditions due to a greater variability in cDCs values. There were no differences in CD80 expression, whereas CD86 was significantly upregulated on cDCs compared to iDCs and tolDCs in both serum-supplemented and SF conditions (*p* < 0.05, *p* < 0.001). A similar pattern was seen in expression of MHC II, where again cDCs exhibited statistically significantly increased levels compared to iDCs and all groups of tolDCs, irrespective of serum conditions (*p* < 0.05, *p* < 0.001, *p* < 0.001). Although not significant, of note is a slight increase of MHC II expression in all groups of tolDCs generated by our protocol compared to iDCs in both culture conditions. There were no differences between the serum-supplemented and SF cultures except a remarkably lower CD80 expression in all types of DCs cultured in SF conditions. The replacement of vitamin D2 by vitamin D3 in our protocol did not alter the pattern of maturation markers of tolDCs cultured in serum-suppl. conditions (Figure S1 in Supplementary Material). To conclude, antigen-unloaded tolDCs as well as antigen-loaded tolDCs-GAD65, tolDCs-OVA, and toldDCs-pept exhibited the same pattern of decreased maturation markers compared to cDCs in serum-supplemented and SF conditions.

We have examined expressions of other markers related to migration, mucosal homing, or induction of regulatory T-cell responses in antigen-loaded tolDCs-OVA and tolDCs-GAD65 and unloaded tolDCs. We found no substantial difference in expression of CCR5, CCR7, CD103, IL-7Rα, and c-kit among the tolDC-GAD65, tolDC-OVA and unloaded tolDC groups (some data shown in **Figure 5B**). None of the surface markers was differentially expressed among the three groups of tolDCs.

In order to determine whether tolDCs are resistant to additional maturation stimuli, the stability of unloaded tolDCs and antigen-loaded tolDCs was tested by 24-h restimulation with 1 µg/mL LPS, while 24 h-left unstimulated cells we used as controls. **Figure 5C** shows that irrespective of antigen loading, both tolDCs as well as tolDCs-GAD65, tolDCs-OVA and tolDCspept were refractory to LPS restimulation as documented by CD40 and even more CD86 expressions. Small increase in CD80 and MHC II expression could be noted after the prolonged LPS exposure; however, it was similar for all groups of tolDCs tested. While cDCs expressed already high levels of the examined maturation markers that could be only moderately further increased by LPS, iDCs showed substantial upregulation of CD40 and CD86 and to a lesser extent also CD80 and MHC II. Thus, all groups of tolDCs displayed a stable phenotype, irrespective of antigen loading.

We have also assessed secretion of IL-10 and IL-12 at the end of our DC-generation protocol, i.e., after final MPLA activation as well as after 24 h LPS restimulation (**Figure 5D**). Compared to cDCs both antigen-unloaded tolDCs as well as antigen-loaded tolDC-GAD65 and tolDC-OVA produced similar and distinct levels of IL-10 on day 8 after MPLA activation that were markedly increased by LPS restimulation (**Figure 5C**). No detectable IL-12 was found in all tolDC culture conditions, whereas cDCs produced 21.4 ± 1.1 pg/ml of IL-12 after MPLA activation (data not shown).

### Allogeneic Proliferative Responses and INF-**γ** Induction by tolDCs and Antigen-Loaded tolDCs

The tolerogenic properties of tolDCs were evaluated by allogeneic proliferative T-cell responses. Splenocytes from C57BL/6 mice were cocultured with all groups of tested DCs in 10:1 ratio for 3 and 5 days and proliferation of CD3<sup>+</sup> cells was assessed by flow cytometry. As shown in **Figures 6A,B**, the cDCs induced strong proliferative responses on day 3 and day 5 (9.28 ± 3.38 and 34.53 ± 0.51%, respectively). On the other hand, cocultures with iDCs and all groups of tolDCs—i.e., unloaded tolDCs, tolDCs-GAD65, tolDCs-OVA, and tolDCs-pept led to substantially reduced and comparable levels of proliferation of allogeneic splenic T cells. Thus, both iDCs and all groups of tolDCs induced statistically significantly lower proliferation of T cells on day 5 (*p* < 0.001). A similar pattern could be noted on day 3, but not statistically significant due to a greater variation of values in the cDC group (**Figure 6A**).

To further assess the tolerogenic properties of the tested DCs, we have determined induction of IFN-γ by CD4+ T cells in allogeneic settings. After 5 days, both cDCs and iDCs induced high proportion of IFN-γ producing CD4<sup>+</sup> T cells (**Figures 6C,D**), whereas all groups of tolDCs were characterized by statistically significantly reduced percentages of IFN-γ-producing CD4<sup>+</sup> T cells at level of *p* < 0.001, *p* < 0.001 for iDCs and *p* < 0.001 for cDCs (**Figure 6C**). In conclusion, all groups of tolDCs were characterized by a similar and statistically significant reduction of allogeneic proliferative T-cell responses and decreased IFN-γ induction.

autoantigen-loaded tolDCs-GAD65. (C) Stability test of above listed types of DCs was carried out by additional 24 h culture (Control) or restimulation with 1 µg/mL lipopolysaccharide (LPS). Changes in expression of maturation markers CD40, CD80, CD86, and MHC II were assessed by flow cytometry and are displayed as MFI. (D) Interleukin 10 (IL-10) release after MPLA activation or 24 h LPS restimulation (stability test) by iDCs, cDCs, tolDCs, tolDCs-GAD65, and tolDCs-OVA. Data are expressed as means from two to four parallel cell cultures.

### Dying Pattern of Antigen-Loaded tolDCs Compared to Unloaded tolDCs When Exposed to Autologous CD8**+** or CD4**<sup>+</sup>** Diabetogenic Splenocytes

In this experiment, we have tested the hypothesis that the opposite effect of tolDCs compared to antigen-loaded tolDCs-GAD65 or tolDCs-OVA on diabetes incidence may be due to an increased killing of antigen-loaded tolDCs by CD8<sup>+</sup> or CD4<sup>+</sup> T cells from within the cotransferred autologous diabetogenic splenocytes.

Thus, iDC, tolDCs, tolDCs-GAD65, and tolDCs-OVA were cocultured with autologous CD8<sup>+</sup> or CD4<sup>+</sup> splenic T cells for 4, 8, 12, and 24 h. The purity of enriched CD4<sup>+</sup> and CD8<sup>+</sup> T cells by negative magnetic selection in this experiment was 94 and 80%, respectively (**Figure 7B**). The flow cytometry analysis of iDCs, tolDCs, and tolDCs-GAD65 or tolDCs-OVA documented relatively small changes in proportions of live non-apoptotic DCs (CD3- CD11c+AnnexinV−Hoechst33342−) among controls cultivated for additional 4, 8, 12, and 24 h (**Figures 7A,C**). There were also no statistically significant differences in percentage of live

Figure 6 | Allogeneic T cell proliferation and IFN-γ production by stimulation with tolDC vs. antigen-loaded tolDCs. (A) Allogeneic proliferative responses of immature bone marrow-derived dendritic cells (iDCs), control matured bone marrow-derived dendritic cells (cDCs), tolDCs, tolDCs-GAD65, tolDCs-OVA, and tolDC-pept were assessed by coculture of CFSE-labeled splenocytes (6- to 8-week-old C57BL/6 females) with DCs (8-week-old non-obese diabetes females) at 10:1 ratio for 3 and 5 days. Proliferation was measured as CFSE dilution in live CD3+ cells by flow cytometry. Splenocytes cultured alone were used as a control. All experiments were carried out in the serum-supplemented RPMI-1640 medium. Data are expressed as mean percentage of CFSElowCD3+ cells ± SEM of four experiments, \*\**p* < 0.01, \*\*\**p* < 0.001. (B) Example of proliferation analysis by the flow cytometry of CFSE-labeled CD3+ splenocytes. (C) Induction of INF-γ in allogeneic CD4+CD3+ T cells was measured after 5 days of coculture with iDCs, cDCs, tolDCs, tolDCs-GAD65, tolDCs-OVA, and tolDCs-pept, following 4-h restimulation with phorbo-12-myristate-13-acetate/ionomycin by intracellular staining and flow cytometry analysis. Data are expressed as mean percentage of CD4+CD3+ cells ± SEM of four experiments, \*\**p* < 0.01, \*\*\**p* < 0.001. (D) Example flow cytometry data of allogeneic induction of IFN-γ by DCs within CD3+CD4<sup>+</sup> splenocytes.

non-apoptotic DCs among the 4, 8, 12, and 24 h time points in general (**Figure 7A**). Compared to unloaded tolDCs, cocultures with CD8+ T cells led to slightly lowered percentage of live nonapoptotic tolDCs-GAD65 4 (3.23%), 8 (2.7%), 12 (4.0%), and also 24 h (5.5%), whereas a more substantial decrease was observed in the tolDC-OVA group after 4 (8.6%), 8 (5.3%), 12 (7.8%), and also 24 h (7.1%) (**Figures 7A,C**). Lower percentages of live non-apoptotic antigen-loaded tolDCs-GAD65 and tolDCs-OVA were also present after cocultures with CD4<sup>+</sup> T cells (**Figure 7A**). Although the differences were not statistically significant, these percentages may reflect some CD8- and CD4-mediated killing of tolDCs-OVA or tolDCs-GAD65. However, similar but much less pronounced pattern of differences could also be seen in the control DCs cultured without autologous CD4<sup>+</sup> or CD8<sup>+</sup> T cells (**Figures 7A,C**). In another experiment, differences in percentage of live non-apoptotic antigen-unloaded tolDCs and antigen-loaded tolDCs-GAD65 and tolDCs-OVA after 4, 8, 12, and 24 h were also assessed in cocultures with whole autologous splenocytes at the ratio 3:5, to better reflect settings used for i.p. administrations to NOD-SCID mice. Similarly to data shown in **Figure 7A**, slightly lower percentage of antigen-loaded tolDCs-GAD65 and tolDCs-OVA compared to unloaded tolDCs were detected (Figure S2 in Supplementary Material), suggesting that some killing of antigen-loaded tolDCs may occur.

### Migration of Antigen-Loaded tolDCs-GAD65 and tolDCs after i.p. Administration in NOD Mice

The difference in the diabetes-preventive properties of unloaded tolDCs compared to antigen-loaded, e.g., tolDCs-GAD65 and tolDCs-OVA might be due to an altered migration pattern and/or trafficking dynamics of these cells after i.p. application. To test this hypothesis, we have investigated migration pattern of tolDCs in NOD mice. Tolerogenic DCs were prepared from bone marrows of NOD mice, then tolDCs and antigen-loaded tolDCs-GAD65 were labeled with the PKH26 dye for *in vivo* tracking and injected i.p. at dose of 5 × 106 to 6-week-old NOD mice. Not surprisingly, the flow cytometry analysis revealed labeled PKH26<sup>+</sup>CD11c<sup>+</sup> cells first (on day 3) in the spleen of NOD mice injected either with tolDCs or tolDCs-GAD65; 0.148 and 0.162%, respectively (**Figure 8**). While the number of the few PKH26<sup>+</sup>CD11c<sup>+</sup> cells within spleen declined on day 7 in both tolDC and tolDC-GAD65 injected animals, there was a prolonged accumulation of the PKH26<sup>+</sup> tolDC and tolDC-GAD65 cells in mucosal, MLNs from day 3 (0.426 and 0.308%), and 5 (0.381 and 0.213), to day 7 (0.220 and 0.221), while declining on day 9 (**Figure 8**). A similar pattern of trafficking to spleen and MLNs, but with a peak of accumulation of PKH26<sup>+</sup>CD11c<sup>+</sup> cells in MLNs on day 5 was observed in a second (smaller scale) experiment (data not shown).

In both experiments, the highest percentage of PKH26<sup>+</sup>CD11c<sup>+</sup> cells was found in the pancreas draining PLNs, irrespective of GAD65 antigen loading. As shown in **Figure 8**, the proportion of PKH26<sup>+</sup> tolDCs and tolDCs-GAD65 in PLNs was 0.400 and 0.586 on day 5, 0.707 and 0.578 on day 7, and 0.381 and 0.426 on day 9, respectively. Only very few PKH26<sup>+</sup>CD11c<sup>+</sup> positive cells were detected in PLNs on day 12, perhaps also due to the DC survival time *in vivo*. PKH26<sup>+</sup>CD11c<sup>+</sup> cells were practically absent in the control ILNs (only data from day 7 and 9 are shown in **Figure 8**).

In conclusion, we have not found any differences in the migration pattern of diabetes-preventive tolDCs compared to tolDCs-GAD65, that are not effective in the disease prevention in the NOD-SCID model of adoptive transfer of diabetes. Both tolDCs and tolDCs-GAD65 appeared first in the spleen (day 3), stayed longer in MLNs and from day 5 accumulated most in the pancreas draining PLNs, where they stayed for up to 9 days, with the highest accumulation of the PKH26<sup>+</sup>CD11c<sup>+</sup> cells on days 5 and 7. Both types of these tolDCs equally trafficked within the mucosal lymphoid compartment, preferentially to PLNs, and were not found in ILNs.

### DISCUSSION

In this study, we have shown that antigen-unloaded tolDCs consistently decreased diabetes transfer in the NOD-SCID model of adoptive cotransfer of diabetes (**Figures 1**–**3**). On the other hand, tolDCs loaded with GAD65, with GAD65-immunodominant peptide no. 35, but also with a control protein OVA, all decreased their diabetes-preventive properties. Regardless their functional differences in the diabetes cotransfer model of NOD-SCID mice, the unloaded tolDCs and all three groups of antigen-loaded tolDCs displayed tolerogenic phenotype with IL-10 secretion and very similar effects in alloreactive T cell assays as well as remained stable after restimulation with LPS (**Figures 5** and **6**).

Unlike in animal models of EAE (22, 28) as well as in the models of experimental arthritis (24, 29), in which most of the tolDC-treatments are carried out in an autoantigen-specific manner, this is different in the field of T1D. Several studies documented diabetes prevention by antigen-nonspecific tolDCs (10, 15, 30, 31). While Tai et al. showed that tolDCs (prepared without final activation) in the presence of GM-CSF

cocultured for 4, 8, 12, and 24 h. Dendritic cells cultured without splenic T cells were used as controls. Percentage of live nonapotopic cells was measured by flow cytometry as DCs (gated according to the FSC, SSC and CD3−CD11c+ parameters) stained double negative for Hoechst33342− and AnnexinV−. Data are expressed as mean ± SEM of two to three experiments, \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001. (B) Example of CD4+ and CD8+ T cell enrichment by negative magnetic selection. (C) Example of flow cytometry analyses of CD3−CD11c+AnnexinV−Hoechst33342− cells at the 4 h timepoint.

in combination with IL-10, but not IL-4, prevented diabetes in NOD mice (12), two other studies documented that antigennonspecific tolDCs prepared in GM-CSF + IL-4 (without final activation) prevented diabetes in NOD recipients (8, 9). Our diabetes-preventive unloaded tolDCs were prepared in the presence of GM-CSF + IL-4, further treated with dexamethasone

Figure 8 | *In vivo* migration of PKH26-labeled tolDCs and tolDCs-GAD65. Bone marrow-derived dendritic cells were prepared from 8-week-old non-obese diabetes (NOD) mice. Tolerogenic DCs and tolDCs-GAD65 (1 µg/mL) were labeled with fluorescent PKH26 dye and 5 × 106 cells were applied i.p. (left side of the belly) to 6-week-old NOD females. Unlabeled tolDCs were used as a negative control. FACS detection of PKH26+ cells was carried out on cell suspensions from spleen, mesenteric lymph nodes (MLNs), pancreatic lymph nodes (PLNs), and systemic inguinal lymph nodes (ILNs) after 3, 5, 7, 9, and 12 days (three mice per group for day 12, five mice per group for all other timepoints). Following doublets exclusion cells were gated according to the FSC-A and SSC-A parameters and dead cells were excluded by Hoechst 33258. PKH26+ tolDCs and tolDCs-GAD65 are displayed as percentage of live CD11c+ cells (1–2 × 106 events per sample). Example of a larger (no. of time points) of two independent experiments.

and vitamin D2 and stabilized by a final activation with MPLA. On the other hand, not so many studies addressed T1D prevention by antigen-specific tolDCs. Tolerogenic DCs loaded with apoptotic bodies from the NIT-1 beta-cell line were effectively used in diabetes prevention in the RIP-IFN-β transgenic NOD mice (13); however, by this approach tolDCs are not exposed to a single antigenic entity. Lo et al. reported peptide-pulsed tolDCs preventing diabetes in NOD mice by using a peptide from an ignored GAD65 sequence and nonstabilized immature DCs (32). Peptide-specific approach was nicely documented in the humanized transgenic HLA-DR4 mouse model, in which i.d. application of GM-CSF + IL-4 + Vitamin D3-generated, LPS-activated tolDCs loaded with the proinsulin peptide C19- A3 reversed the break of tolerance as documented by decreased proliferation and peptide-specific induction of IL-10 (33). Finally, i.p. application of GM-CSF + IL-10-generated (without MPLA activation) tolDCs loaded with immunodominant insulin B chain peptides prevented diabetes in NOD mice by *in situ* induction of Foxp3<sup>+</sup> Tregs, when cultured in autologous mouse serum (11).

This study as well as their previous paper (26) addressing the effect of nonautologous (FBS) vs. autologous sera on tolDC mechanisms of action (Th2 shift vs. induction of Tregs) together with data by Feili-Hariri et al. (9), who reported no additional beneficial effect of antigen-specific tolDCs compared to unloaded tolDCs (both cultured in FBS medium) pointed toward a nonspecific shift to Th2 immune responses probably due to presentation of FBS-related antigens. Thus, we tested our tolDC protocol also in SF conditions by using the CellGro medium, which is often used for DC cultures in human trails. As shown in **Figure 3**, we found no significant differences in SF settings, i.e., all antigen-loaded tolDCs (tolDCs-GAD65, tolDCs-OVA, tolDCs-pept) were again ineffective, whereas unloaded tolDCs led to a reduction of diabetes. Tolerogenic DCs cultured in SF conditions displayed slightly better diabetes-preventive effect (**Figure 3A**).

Several protocols generating stable tolerogenic DCs have been reported in the literature. These include, e.g., longer protocols based on the use of vitamin D3 such as an 8-day protocol with a continuous presence of vitamin D3 (34), a 10-day protocol with vitamin D3 and dexamethasone added only during last 16 h of LPS activation (24), vitamin D3 added from day 0 a dexamethasone on day 1 (35) or shorter one with dexamethasone added on days 3 and 6 and vitamin D3 on day 6 during final maturation (36), or a protocol using dexamethasone alone (37). In this study, we have used a protocol of tolDCs prepared in the presence of both GM-CSF and IL-4, added vitamin D2/dexamethasone and stabilized by final activation with MPLA. Our previous papers from overlapping groups of authors documented that human tolDCs prepared by this protocol (GM-CSF + IL-4 with added vitamin D2/dexamethasone and MPLA activation, in SF medium) led to induction of stable antigen-specific (GAD65) T cell hyporesponsiveness as well as induction of suppressive Tregs. Unloaded tolDCs prepared by this protocol prevented diabetes in the NOD-SCID model (38). This protocol also suppressed proliferation and induced IL-10 producing Tregs in a human allogeneic system. Good stability of these tolDCs in inflammatory environment was controlled by multiple signaling pathways including p38 MAPK, ERK1/2, mTOR, STAT3 and mTOR-dependent glycolysis (20). In the study by Sochorová et al. (39) vitamin D2 or vitamin D3 was added during LPS-induced activation of tolDCs. These tolDCs exhibited similar tolerogenic properties compared to tolDCs generated in the presence of vitamin D3 or vitamin D2 from the beginning of the cultivation (see also Figure S1 in Supplementary Material).

As shown in **Figure 5**, all groups of the NOD tolDCs were stable after 24 h restimulation with 1 µg/mL of LPS, secreted IL-10 (especially after restimulation with LPS) and displayed an immature phenotype in both serum-supplemented and SF conditions. In both culture conditions, tolDCs as well as all groups of antigen-loaded tolDCs displayed slightly increased expression of MHC II than iDCs. This increase of MHC II expression is not disadvantageous if tolDCs are stable. A similar pattern in MHC II expression was reported in dexamethasone-induced human tolDCs (40) or dexamethasone/ vitamin D3 mouse tolDCs in experimental arthritis (24). The only notable difference between the tolDCs cultured in serumsupplemented and SF conditions was substantially decreased expression of CD80 in SF (**Figure 5A**). A comparative study on human clinical grade tolDCs (activated by a cytokine mix) showed that both dexamethasone and vitamin D3 produced stable tolDCs that suppressed allogeneic proliferation and IFNγ induction by T cells (40). Similarly, García-Gonzales et al. (37) presented data on dexamethasone and MPLA activated stable human tolDCs with reduced allogeneic proliferation and IFN-γ induction (a 5-day protocol, dexamethasone added for last 48 h, without a vitamin D, 24 h MPLA activation). Our tolDCs, irrespective of antigen loading, displayed comparable parameters (**Figures 5A,C,D** and **6**).

Several studies documented that DC cultivation with GM-CSF and IL-4 is favorable over GM-CSF alone, resulting in better tolerogenic properties and a more mature phenotype of tolDCs (9, 41, 42). The importance of IL-4 for induction of an increased stimulatory potential of DCs was documented by Wells et al. (43). GM-CSF + IL-4 generated DCs transduced with IL-4 were able to prevent diabetes in NOD mice with advanced insulitis (44). Gene array analyses revealed several differences including increased expression of costimulatory molecules, CD200, Ym-1 (marker of alternative macrophage activation), and different pattern of cytokine and chemokine expression by GM-CSF + IL-4 DCs (45). NOD tolDCs generated by GM-CSF + IL-4 and vitamin D3 were shown to induce Foxp3<sup>+</sup> Tregs and IL-10 expression *in vitro* (34).

Our observation that antigen-loaded tolDCs are not so effective is not a complete surprise among studies dealing with tolDC therapy in animal models of T1D. There are scattered reports of antigen-loaded tolDCs being ineffective in the disease prevention, however, these findings were not much discussed or followed up. In 1999, Feilli-Hariri M et al. showed that tolDCs prepared in GM-CSF alone were less effective than tolDCs cultured in GM-CSF + IL-4, however in both cultures tolDCs pulsed with a mixture of 3 peptides (2 of GAD65 and 1 from hsp60 sequences) were less effective than unloaded control tolDCs (9). Later, Machen et al. reported that tolDCs prepared in GM-CSF + IL-4 and by antisense oligonucleotides against DC's surface costimulatory molecules reduced diabetes incidence in NOD mice, but not if they were coadministered with a lysate from the NIT-1 β cell line (8). *In vivo* stimulation of DCs by PEGylated TLR7 ligand (1Z1) delayed and reduced diabetes as well as insulitis when they were transferred to prediabetic NOD mice. However, in the Figure S2 in Supplementary Material, the authors also show that 1Z1 treated DCs pulsed with GAD65 peptide 515–524 significantly increased insulitis in NOD mice (9 weeks after transfer) compared to both control animals with no transfer of cells but also compared to mice treated with 1Z1 DCs only (10). Finally, in the study addressing the effect of FBS on mechanisms of tolDC-action in NOD mice, only GM-CSF and IL-10 generated tolDCs pulsed with 2 insulin B chain peptides prevented diabetes in NOD mice. When splenocytes from these protected animals were cotransferred with diabetogenic splenocytes to NOD-SCID recipients, they caused a more rapid 100% diabetes onset compared to controls receiving only diabetogenic splenocytes (11). Indeed the NOD-SCID model of adoptive cotransfer of diabetes and the spontaneous NOD mouse model differ, e.g., the presence of self T and B cells or possibly a lower proportion of transferred Tregs within the diabetogenic splenocytes that may alter the effect of tolDCs in the NOD-SCID model. Another mechanism to consider is a homeostatic expansion of transferred diabetogenic lymphocytes in immunodeficient settings. On the other hand, Machen et al. (8) documented that NOD mice injected first with NOD T cells followed by administration of tolDCs or control DCs displayed a significant increase in the number of total splenic CD4<sup>+</sup>CD25<sup>+</sup> and CD25<sup>+</sup>CD62L<sup>+</sup> regulatory cells only in the group injected with tolDCs. This finding together with the absence of differences in the prevalence and numbers of single CD4<sup>+</sup> or single CD8<sup>+</sup> cells between NOD-SCID groups treated with tolDCs and control DCs argues against homeostatic expansion as the basis of the increased prevalence of the CD4<sup>+</sup>CD25<sup>+</sup>CD62L<sup>+</sup> cells. Since many studies were performed in the spontaneous NOD mouse model of T1D, we also tested unloaded and antigenloaded tolDCs in NOD mice. Similar diabetes-preventive effect of antigen-unloaded but not antigen-loaded tolDCs (not only GAD65- but also OVA-loaded tolDCs) on diabetes prevention was documented (**Figure 4**). The above listed (8–11) scattered evidence about less effective or ineffective antigen-loaded tolDCs in the literature is also derived from the NOD mouse model.

We also addressed the possibility of bystander antigen presentation by APCs present within the diabetogenic splenocytes and cotransferred with antigen-loaded tolDCs to NOD-SCID recipients. Yet, by using purified diabetogenic T cells instead of whole splenocytes, we only observed a tendency for a bit more effective diabetes prevention by unloaded tolDCs (**Figure 2**). Another mechanism to consider was killing of antigen-loaded tolDCs by autologous CD8<sup>+</sup> or CD4<sup>+</sup> T cells present within the diabetogenic splenocytes. Although not statistically significant, decreased percentage of live non-apoptotic OVA- and to a lesser extent GAD65-loaded tolDCs compared to unloaded tolDCs was detected in cocultures with autologous CD8<sup>+</sup> and CD4<sup>+</sup> T cells (**Figure 7**). However, there is a possibility that antigens from dying antigen-loaded donor tolDCs are presented *in vivo* by recipient APCs (DCs) in an immunogenic fashion as reported by Smyth et al. (46), even in autologous settings. In NOD mice, these mechanisms could be further enhanced by a defect in tolerance induction by CD8<sup>+</sup>DCs that express higher levels of CD40 (47).

Our next experiment showed preferential mucosal migration of both unloaded tolDCs as well as antigen-loaded tolDCs-GAD65 and tolDCs-OVA with highest accumulation in PLNs (**Figure 8**). PLNs were shown to play a critical role in priming beta-cell-specific immune responses as, e.g., removal of PLNs prevented diabetes development in NOD mice (48), T cells from BDC2.5 T cell receptor transgenic mice, that are specific for a natural beta-cell antigen, proliferate exclusively in PLNs before onset of insulitis (49) and increased migration of mature tolDCs generated by vitamin D3 to PLNs of NOD mice was reported (34). NOD's PLNs were also described to harbor increased number of merocytic mcDCs that induce T-cell activation and break T-cell tolerance to beta-cell antigens (50). Although we found no differences in surface expression of DC markers related to migration and mucosal homing between the unloaded tolDCs and antigen-loaded tolDCs-OVA or tolDCs-GAD65, of note is that our tolerogenic protocol as well as protocol based on dexamethasone alone followed by MPLA activation (37) led to increased CCR7 expression (**Figure 5B**). It has been reported that CCR7 (51) but not CD103 (52) is critical for mucosal (MLNs) homing of DCs.

Turner et al. have published interesting data documenting the importance of antigen dose on induction of Foxp3<sup>+</sup> Tregs (low dose, 0.4 µM) or Foxp3<sup>−</sup>CD4<sup>+</sup> T cells (high dose, 40 µM) in relation to weak and strong activation of Akt/mTor TCR signaling pathway. This effect was modulated by IL-6 and was present not only in GM-CSF + IL-4 tolDCs, but also in immature DCs cultured in GM-CSF only (53). The effect of a dosage might explain the decreased diabetes-preventive properties not only of GAD65- and pept-loaded tolDCs but also of tolDCs loaded with the control protein OVA (**Figures 2** and **3A**). These findings may correspond with a pattern of diabetes incidence of GAD65-loaded tolDCs presented in **Figures 1** and **2**. While the higher dose of 2 µg/mL of GAD65 led to 100% diabetes transfer and a more rapid onset of diabetes (**Figure 1**), tolDCs loaded with the lower dose of 1 µg/mL (**Figure 2**) precipitated lower (seven of eight mice) and slower transfer of diabetes than Controls. The antigen dose of 1–2 µg/mL in our experiments is, however, lower than in other studies using antigen-loaded tolDCs for T1D prevention that ranged from 10 µg/mL (11) to 3 × 60 μg/mL (9) or 10 µM (32). In animal models of EAE and rheumatoid arthritis (RA) doses of peptides ranging from 5 to 50 µg/mL and 1 to 50 µg/mL, respectively, were used for pulsing tolDCs (42). Nevertheless, controlling the outcome of a tolDC therapy by antigen doses alone would be a difficult task. Less protective effect of antigen-loaded tolDCs could be also due to a combination of factors, e.g., the dose, killing by autologous CD8<sup>+</sup> and CD4<sup>+</sup> T cells or presentation of antigens from dying antigen-loaded tolDCs by recipient APCs (46) and/ or by homeostatic expansion of diabetogenic lymphocytes in the NOD-SCID model.

Although antigen-specific tolDCs are widely used in animal models of autoimmune diseases such as EAE or RA (42), it should be noted that T1D may differ in some aspects from other autoimmune diseases. Development of T1D seems to be more related to impaired, "missing" regulatory immune responses in genetically predisposed individuals (54). Interestingly, T1D does not fulfill one important of the Rose-Witebsky's criterions of autoimmune diseases—induction of the disease by immunization with an autoantigen (55).

In conclusion, while tolDCs represent a very promising strategy for prevention or even early cure of T1D, our data together with previously published scattered evidence suggest that antigen loading decreases the disease-protective effect of tolDCs in animal models of T1D. *In vivo* testing of tolDCs is important as multiple factors may influence their therapeutical effects. Further studies are needed to shed more light on the mechanisms of antigen-specific tolDCs in animal models of T1D.

### REFERENCES


### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of our institutional animal ethics office (Laboratory Animal Care and Use Committee of the Institute of Microbiology v.v.i., Academy of Sciences of the Czech Republic, approval ID: 94/2006 and 244/2009) in strict accordance with the Federation of European Laboratory Animal Science Associations guidelines.

### AUTHOR CONTRIBUTIONS

All authors (DF, JG, TH, HK, RŠ, LP-J) contributed to the study concept, data acquisition and drafting the work. DF wrote the paper. DF, LP-J, and JG performed data analysis and interpretation of the data. LP-J and RŠ contributed to the critical revision of the manuscript. DF and LP-J are the guarantors of this work.

### ACKNOWLEDGMENTS

Authors wish to thank Jan Svoboda from the Cytometry and Microscopy Facility of the Institute of Microbiology of the Czech Academy of Sciences, Prague for excellent technical assistance and support.

### FUNDING

This work was in part supported by the Institutional Research Concept RVO 61388971 and grant 15-24487S from the Grant Agency of the Czech Republic.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00290/ full#supplementary-material.

diabetes-preventive properties to nonobese diabetic mouse dendritic cells. *J Immunol* (2004) 173:4331–41. doi:10.4049/jimmunol.173.7.4331


transduced to express IL-4 in NOD mice. *Clin Immunol* (2008) 127:176–87. doi:10.1016/j.clim.2007.12.009


55. Rose NR, Bona C. Defining criteria for autoimmune diseases (Witebsky's postulates revisited). *Immunol Today* (1993) 14:426–30. doi:10.1016/0167- 5699(93)90244-F

**Conflict of Interest Statement:** LP-J, and RŠ are named inventors in a related patent, "Tolerogenic Dendritic Cells, Methods of Producing the Same, and Uses Thereof " PCT/EP2015/074536 which describes methods for the preparation of stable semi-mature tolerogenic DC. The other authors have no financial conflicts of interest.

*Copyright © 2018 Funda, Goliáš, Hudcovic, Kozáková, Špíšek and Palová-Jelínková. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Phosphatidylserine-liposomes Promote Tolerogenic Features on Dendritic cells in human Type 1 Diabetes by apoptotic Mimicry

*Silvia Rodriguez-Fernandez1 , Irma Pujol-Autonell1 , Ferran Brianso2,3, David Perna-Barrull1 , Mary Cano-Sarabia4 , Sonia Garcia-Jimeno4 , Adrian Villalba1 , Alex Sanchez 2,3, Eva Aguilera5 , Federico Vazquez <sup>5</sup> , Joan Verdaguer6,7, Daniel Maspoch4,8 and Marta Vives-Pi1,7\**

*<sup>1</sup> Immunology Section, Germans Trias i Pujol Research Institute, Autonomous University of Barcelona, Badalona, Spain, 2Statistics and Bioinformatics Unit, Vall d'Hebron Research Institute, Barcelona, Spain, 3Department of Genetics, Microbiology and Statistics, University of Barcelona, Barcelona, Spain, 4Catalan Institute of Nanoscience and Nanotechnology, CSIC and The Barcelona Institute of Science and Technology, Bellaterra, Spain, 5Endocrinology Section, Germans Trias i Pujol University Hospital, Badalona, Spain, 6Department of Experimental Medicine, University of Lleida & IRBLleida, Lleida, Spain, 7CIBERDEM, ISCiii, Madrid, Spain, 8 ICREA, Barcelona, Spain*

#### *Edited by:*

*John Isaacs, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Bert A. 'T Hart, Biomedical Primate Research Center, Netherlands Elisabetta Padovan, University Hospital of Basel, Switzerland*

#### *\*Correspondence:*

*Marta Vives-Pi mvives@igtp.cat*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 26 September 2017 Accepted: 29 January 2018 Published: 14 February 2018*

#### *Citation:*

*Rodriguez-Fernandez S, Pujol-Autonell I, Brianso F, Perna-Barrull D, Cano-Sarabia M, Garcia-Jimeno S, Villalba A, Sanchez A, Aguilera E, Vazquez F, Verdaguer J, Maspoch D and Vives-Pi M (2018) Phosphatidylserine-Liposomes Promote Tolerogenic Features on Dendritic Cells in Human Type 1 Diabetes by Apoptotic Mimicry. Front. Immunol. 9:253. doi: 10.3389/fimmu.2018.00253*

Type 1 diabetes (T1D) is a metabolic disease caused by the autoimmune destruction of insulin-producing β-cells. With its incidence increasing worldwide, to find a safe approach to permanently cease autoimmunity and allow β-cell recovery has become vital. Relying on the inherent ability of apoptotic cells to induce immunological tolerance, we demonstrated that liposomes mimicking apoptotic β-cells arrested autoimmunity to β-cells and prevented experimental T1D through tolerogenic dendritic cell (DC) generation. These liposomes contained phosphatidylserine (PS)—the main signal of the apoptotic cell membrane—and β-cell autoantigens. To move toward a clinical application, PS-liposomes with optimum size and composition for phagocytosis were loaded with human insulin peptides and tested on DCs from patients with T1D and control age-related subjects. PS accelerated phagocytosis of liposomes with a dynamic typical of apoptotic cell clearance, preserving DCs viability. After PS-liposomes phagocytosis, the expression pattern of molecules involved in efferocytosis, antigen presentation, immunoregulation, and activation in DCs concurred with a tolerogenic functionality, both in patients and control subjects. Furthermore, DCs exposed to PS-liposomes displayed decreased ability to stimulate autologous T cell proliferation. Moreover, transcriptional changes in DCs from patients with T1D after PS-liposomes phagocytosis pointed to an immunoregulatory prolife. Bioinformatics analysis showed 233 differentially expressed genes. Genes involved in antigen presentation were downregulated, whereas genes pertaining to tolerogenic/ anti-inflammatory pathways were mostly upregulated. In conclusion, PS-liposomes phagocytosis mimics efferocytosis and leads to phenotypic and functional changes in human DCs, which are accountable for tolerance induction. The herein reported results reinforce the potential of this novel immunotherapy to re-establish immunological tolerance, opening the door to new therapeutic approaches in the field of autoimmunity.

#### Keywords: immunotherapy, autoimmunity, human type 1 diabetes, liposomes, tolerance, dendritic cells

**Abbreviations:** DC, dendritic cell; PS, phosphatidylserine; T1D, type 1 diabetes.

## INTRODUCTION

Type 1 diabetes (T1D) mellitus is a metabolic disease caused by loss of tolerance to self and consequent autoimmune destruction of insulin-producing pancreatic β-cells (1). When β-cell mass decreases significantly, the individual's endogenous production of insulin is no longer able to meet metabolic demands, leading to overt hyperglycemia. Upon diagnosis, patients with T1D require exogenous insulin administration, and although this treatment has allowed them to survive, long-term complications due to glycemic imbalances are bound to arise (2, 3). T1D usually appears during childhood or adolescence, and its incidence is increasing an average of 4% per year (4). Despite knowing that both genetic and environmental factors contribute to its development, triggering events remain elusive. The autoimmune attack against β-cells is led by a mild leukocytic infiltrate—insulitis—consisting of dendritic cells (DCs), macrophages, T and B lymphocytes, and natural killer cells, which gradually advance through the islets (5). The first islet-infiltrating cells are DCs (6), which orchestrate the loss of tolerance to β-cell autoantigens, insulin being a key autoantigen in human T1D (7).

The first step to revert T1D would be to arrest the pathological recognition of β-cell autoantigens. Many immunotherapies have prevented and even reverted T1D in animal models (8), but clinical trials have corroborated how challenging T1D prevention and reversion is, and most of them have been unsuccessful (9). In this scenario, the development of new therapies to halt autoimmunity in T1D has become an urgent biomedical matter. An ideal immunotherapy should restore tolerance to β-cells, avoiding systemic side effects, and allow islet regeneration. One of the most efficient physiological mechanisms for inducing tolerance is apoptosis, a form of programmed cell death lacking inflammation. The uptake of apoptotic cells by professional phagocytes such as macrophages and immature DCs (iDCs) is named efferocytosis (10). The exposure of "*eat-me*" signals on the apoptotic cell surface is what promotes their specific recognition and phagocytosis. Phosphatidylserine (PS), a phospholipid usually kept in the inner leaflet of the plasma membrane, is a relevant signal for efferocytosis. This molecule is recognized by multiple distinct receptors on antigen presenting cells, including members of the TIM family, Stabilin-2, integrins, CD36, CD68, among others, as well as by soluble receptors that in turn bind to membrane receptors (11). After the capture, the apoptotic cell is processed, thus prompting the release of anti-inflammatory signals and presentation of autoantigens in a tolerogenic manner by DCs (12). Failure of this mechanism, owing to an increase of apoptotic β-cells or defects in efferocytosis, contributes to the loss of tolerance to self in the context of T1D (13).

Our group demonstrated that DCs acquired a tolerogenic phenotype and functionality after engulfment of apoptotic β-cells, and that they prevented T1D when transferred to non-obese diabetic (NOD) mice (14, 15). However, since finding a substantial source of autologous apoptotic β-cells for T1D immunotherapy would be impossible, we conceived an immunotherapy based on biomimicry that consisted of liposomes—phospholipid bilayer vesicles—displaying PS in their surface and containing autoantigenic peptides, thus resembling apoptotic cells. Indeed, apoptotic mimicry performed by PS-liposomes successfully restored tolerance to β-cells in experimental autoimmune diabetes, preventing disease development and decreasing the severity of insulitis (16). Moreover, by only replacing the autoantigenic peptide encapsulated within PS-liposomes, we confirmed the potential of this immunotherapy to prevent and ameliorate experimental autoimmune encephalomyelitis, the experimental model of multiple sclerosis (17). In both cases, we demonstrated that phagocytosis of autoantigen-loaded PS-liposomes induced a tolerogenic phenotype and functionality in DCs, expansion of regulatory T cells and release of anti-inflammatory mediators that are responsible for arresting the autoimmune attack to target cells. Therefore, PS-liposomes could constitute a platform serving as a physiological and safe strategy to restore peripheral tolerance in antigen-specific autoimmune diseases. Liposomes, already used clinically as drugs deliverers for antitumor drugs and as vaccines (18), have the advantage of being safe and biocompatible, customizable, easily large-scale produced, and standardizable.

Aiming for the clinical potential of this strategy, we have encapsulated human insulin peptides to assess the effect of PS-liposomes in human DCs from patients with T1D and control subjects *in vitro*. We herein report that PS-liposomes are efficiently captured by human DCs, thus eliciting transcriptomic, phenotypic, and functional changes that point to tolerogenic potential. This immunotherapy constitutes a promising strategy to arrest autoimmune aggression in human T1D, benefiting from the co-delivery of tolerogenic signals and β-cell autoantigens.

### MATERIALS AND METHODS

### Patients

Patients with T1D (*n* = 34) and control subjects (*n* = 24) were included in this study. All patients with diabetes fulfilled the classification criteria for T1D. Inclusion criteria were 18–55 years of age, a body mass index (BMI) between 18.5 and 30 kg/m2 and, for patients with T1D, an evolution of the disease longer than 6 months. Exclusion criteria were: being under immunosuppressive or anti-inflammatory treatment, or undergoing pregnancy or breastfeeding. For the RNA-sequencing (RNA-seq) experiment, we selected 8 patients of the 34 that participated in the study, but BMI was limited to a maximum of 24.9 kg/m2 , duration of the disease was restricted to a maximum of 5 years (in order to minimize the effect that long-term hyperglycemia could have on genetic and/or epigenetic profiles) and the presence of other chronic diseases became an exclusion criterion. All study participants gave informed consent, and the study was approved by the Committee on the Ethics of Research of the Germans Trias i Pujol Research Institute and Hospital.

### Cell Separation and Generation of DCs

Peripheral blood mononuclear cells (PBMCs) were obtained from 50 ml blood samples of control subjects and patients with T1D by means of Ficoll Paque (GE Healthcare, Marlborough, MA, USA) density gradient centrifugation. Monocytes were further magnetically isolated using the EasySep Human CD14 Positive Selection Kit (STEMCELL Technologies, Vancouver, BC, Canada) following the manufacturer's instructions. Once CD14 purity in the positively selected fraction was >70%, monocytes were cultured at a concentration of 106 cells/ml in X-VIVO 15 media (Lonza, Basel, Switzerland), supplemented with 2% male AB human serum (Biowest, Nuaillé, France), 100 IU/ml penicillin (Normon SA, Madrid, Spain), 100 µg/ml streptomycin (Laboratorio Reig Jofré, Sant Joan Despí, Spain), and 1,000 IU/ ml IL-4 and 1,000 IU/ml GM-CSF (Prospec, Rehovot, Israel) to obtain monocyte-derived DCs. After 6 days of culture, DC differentiation yield was assessed by CD11c-APC staining (Immunotools, Friesoythe, Germany) and cell viability was determined by annexin V-PE (Immunotools) and 7aad staining (BD Biosciences, San Jose, CA, USA) using flow cytometry (FACS Canto II, BD Biosciences). The negatively selected fraction of PBMCs was cryopreserved in Fetal Bovine Serum (ThermoFisher Scientific, Waltham, MA, USA) with 10% dimethylsulfoxide (Sigma-Aldrich, Saint Louis, MO, USA) and stored for later use.

### Peptide Selection and Preparation of Liposomes

Thinking in a future clinical application of liposomes, the two chains of insulin were selected to be encapsulated separately in order to avoid possible biological effects of insulin. A and B chains of insulin contain well-known β-cell specific target epitopes in human T1D (19). Peptide A corresponds to the whole human insulin A chain (21 aa, N-start-GIVEQCCTSICSLYQLENYCN-C-end), and peptide B is the whole human insulin B chain (30 aa, N-start-FVNQHLCGSHLVEALYLVCGERGFFYTPKT-C-end) (Genosphere Biotechnologies, Paris, France). Peptides were >95% pure and trifluoroacetic acid was removed. Liposomes consisted of 1,2-dioleoyl-sn-glycero-3-phospho-l-serine (sodium salt) (Lipoid, Steinhausen, Switzerland), 1,2-didodecanoyl-sn-glycero-3-phosphocholine (Lipoid), and cholesterol (Sigma-Aldrich). Liposomes were prepared using the thin film hydration method from a lipid mixture of 1,2-dioleoyl-sn-glycero-3-phosphol-serine, 1,2-didodecanoyl-sn-glycero-3-phosphocholine and cholesterol at 1:1:1.33 molar ratio, respectively, as described (20). Liposomes without PS were generated as controls with 1,2-didodecanoyl-sn-glycero-3-phosphocholine and cholesterol at 1:1 molar ratio. All liposomes were produced under sterile conditions and at a final concentration of 30 mM. Lipids were dissolved in chloroform and the solvent was removed by evaporation under vacuum and nitrogen. The lipids were hydrated with the appropriate buffer (phosphate buffered saline or 0.5 mg/ml solution of peptide A or peptide B) and the liposomes obtained were homogenized to 1 µm by means of an extruder (Lipex Biomembranes Inc., Vancouver, BC, Canada). Peptide encapsulation efficiencies were calculated according to the equation: encapsulation efficiency (%) = [(Cpeptide,total-Cpeptide,out)/Cpeptide,total] ×100, where Cpeptide,total is the initial peptide A or peptide B concentration and Cpeptide,out is the concentration of non-encapsulated peptide. To measure the Cpeptide,out, liposome suspensions were centrifuged at 110,000 *g* at 10°C for 30 min. The concentration of non-encapsulated peptide was assessed in supernatants by PIERCE BCA protein assay kit (ThermoFisher Scientific). In addition to PS-rich liposomes loaded with insulin peptides [PSA-liposomes (*n* = 3) and PSB-liposomes (*n* = 3) encapsulating peptide A or peptide B, respectively], fluorescent-labeled liposomes with PS (empty fluorescent PS-liposomes, *n* = 4) and without PS (empty fluorescent PC-liposomes, *n* = 4) were also prepared using lipid-conjugated fluorescent dye Oregon Green 488 1,2-dihexadecanoyl-snglycero-3-phosphoethanolamine (Invitrogen, Carlsbad, CA, USA) and following the aforementioned methodology. Particle size distributions and stability—expressed as zeta potential (ζ)—were measured by dynamic light scattering using Malvern Zetasizer (Malvern Instruments, Malvern, UK) in undiluted samples. Liposome morphology and lamellarity were examined by cryogenic transmission electron microscopy (cryo-TEM) in a JEOL-JEM 1400 microscope (Jeol Ltd., Tokyo, Japan).

### Phagocytosis Assay

To assess whether DCs were able to phagocyte liposomes, DCs were co-cultured with 100 µM of empty fluorescent PS-liposomes (*n* = 5 for control subjects and *n* = 10 for patients with T1D) or empty fluorescent PC-liposomes (*n* = 6 for control subjects and *n* = 9 for patients with T1D) at 37°C from 5 min to 24 h. As control, the same assay was performed at 4°C to confirm that liposomes were captured by an active mechanism of phagocytosis. Cells were extensively washed in cold phosphate buffered saline to remove all liposomes attached to the cell membrane. Liposome uptake was determined by flow cytometry (FACSCanto II, BD Biosciences).

### Assessment of DCs Phenotype after Liposome Uptake

Although insulin chains were encapsulated separately, DCs were stimulated by a mixture of PSA-liposomes (50%) and PSBliposomes (50%), in order to assess the effect of the whole insulin molecule as autoantigen. Thus, DCs from control subjects (*n* ≥ 5) and patients with T1D (*n* ≥ 8) were co-cultured with 1 mM of liposomes (PSAB-DCs) for 24 h in the presence of 20 µg/ml human insulin (Sigma-Aldrich), and their viability and phenotype were analyzed by flow cytometry (FACSCanto II, BD Biosciences). The sample number stated (*n*) is referred to the minimum number of control subjects and patients included in each experiment. As controls, DCs were either cultured with 20 µg/ml human insulin (Sigma-Aldrich) to obtain iDCs or adding a cytokine cocktail (CC) consisting of tumor necrosis factor (TNF)α (1,000 IU/ ml, Immunotools), IL-1β (2,000 IU/ml, Immunotools) and Prostaglandin E2 (PGE2, 1 µM, Cayman Chemical, Ann Arbor, MI, USA) for 24 h to obtain mature DCs (mDC). Moreover, PSAB-DCs were cultured after phagocytosis with CC for 24 h in order to assess the response in front a pro-inflammatory stimulus (mPSAB-DCs). Phenotyping was performed as follows: DCs were stained with 7aad (BD Biosciences) and monoclonal antibodies to CD11c-APC, CD25-PE, CD86-FITC, HLA class I-FITC, HLA class II-FITC, CD14-PE and CD40-APC (Immunotools), CD36-APCCy7, TIM4-APC, αvβ5 integrin-PE, CD54-PECy7, TLR2-FITC, CXCR4-APCCy7, CCR2-APC, DC-SIGN-APC (Biolegend, San Diego, CA, USA) and CCR7-PECy7 (BD Biosciences). Corresponding fluorescence minus one staining was used as control. Data were analyzed using FlowJo software (Tree Star Inc., Ashland, OR, USA).

### T Cell Proliferation Assays

Autologous T lymphocyte proliferation (*n* = 12 for control subjects and *n* ≥ 12 for patients with T1D) was assessed by exposing PBMCs to the different conditions of DCs used in this study. Briefly, PBMCs from the same donor were thawed and stained with 0.31 µM CellTrace Violet (21) (ThermoFisher Scientific) according to manufacturer's instructions. The PBMCs were then co-cultured with iDCs, PSAB-DCs, mPSAB-DCs or mDCs at a 10:1 ratio (105 PBMCs:104 DCs). For each donor, 105 PBMCs were cultured in basal conditions as a negative control or with Phorbol 12-Myristate 13-Acetate (50 ng/ml, Sigma-Aldrich) and Ionomycin (500 ng/ml, Sigma-Aldrich) as a positive control. After 6 days of co-culture, proliferation was assessed in the different T cell subsets with CD3-PE, CD4-APC and CD8-FITC staining (Immunotools) by flow cytometry (FACS LSR Fortessa, BD Biosciences). Data were analyzed using FlowJo software (Tree Star Inc.).

### Cytokine Production

The Human Th1/Th2/Th17 kit (CBA system; BD Biosciences) was used to assess cytokine production. Culture supernatants from DCs and from T cell proliferation assays (*n* ≥ 3 for control subjects and *n* ≥ 3 for patients with T1D) were collected and frozen at −80°C until use. IL-2, IL-4, IL-6, IFN-γ, TNF, IL-17A, and IL-10 were measured. Data were analyzed using CBA software. The production of Human TGF-β1 by DCs after PSAB-liposome uptake was determined by ELISA (eBioscience, San Diego, CA, USA).

### RNA-Seq of DCs before and after Liposome Phagocytosis

Dendritic cells obtained from patients with T1D (*n* = 8) were cultured in basal conditions (iDCs) or with 1 mM of PSAliposomes and PSB-liposomes (PSAB-DCs) at 37°C for 4 h. Cells were then harvested from culture wells using Accutase (eBioscience), and viability and DC purity were assessed with 7aad (BD Biosciences), annexin V-PE and CD11c-APC (Immunotools) staining by flow cytometry (FACS Canto II, BD Biosciences). Liposome capture control assays were performed for every sample (see above *Phagocytosis Assay* section) to confirm phagocytosis. Supernatant was removed and cell pellets were stored at −80°C until use. RNA was extracted using RNeasy Micro Kit (QIAGEN, Hilden, Germany) and following manufacturer's instructions. RNA purity, integrity and concentration were determined by NanoDrop (ND-1000 Spectrophotometer, ThermoFisher Scientific) and 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA). Afterward, 1 µg of total RNA was used to prepare RNA libraries following the instructions of the NebNext Ultra Directional RNA Library Prep Kit (New England Biolabs, Ipswich, MA, USA). Library quality controls were assessed using a TapeStation 2200 (Agilent High Sensitivity Screen Tape) and a narrow distribution with a peak size of approximately 300 bp was observed in all cases. Libraries were quantified by qPCR using a QC KAPA kit (Hoffman-LaRoche, Basel, Switzerland) sequenced in a NextSeq 500 genetic analyzer (SBS-based sequencing technology, Illumina, San Diego, CA, USA) in a run of 2 × 75 cycles and a high output sequencing mode. Twenty million reads were obtained and analyzed for each sample. Fastq files coming from Illumina platform were merged and basic quality controls were performed with FASTQC and PRINSEQ tools. Paired-end (forward-reverse) sample merging was carried out with software CLCBio Genomics Workbench® version 8.5 (22), following the RNA-seq analysis pipeline found in CLCBio manuals. Read alignment and mapping steps to only gene regions were performed using CLCBio software against the human genome (Homo sapiens GRCh38 assembly, at both gene- and transcript-level tracks). The same software, with default options, was used to normalize counts by applying standard "Reads Per Kilobase of transcript per Million reads mapped" method. The remaining steps of the analysis were carried out with scripts and pipelines implemented with R software (23). The selection of differentially expressed genes (DEGs) was performed using the linear model approach implemented in the limma Bioconductor package (24), with previous log2-transformation of the normalized data. Adjusted *p* values of ≤0.125, taking into account multiple testing with the False Discovery Rate method, were considered significant. Therefore, genes with a *p* value <0.0013 and Log2 of fold change >0.05 were considered upregulated, whereas those with Log2 of fold change <0.05 were considered downregulated. Experimental data have been uploaded into European Nucleotide Archive (EBI, https://www.ebi.ac.uk/ena; accession number: PRJEB22240). DEGs were categorized using Ingenuity Pathway Analysis Software (QIAGEN), Protein Analysis Through Evolutionary Relationships Classification System (25), REACTOME Pathway database (26) and Gene Ontology Biological Process database (27). Furthermore, R software (23) was used to generate a gene heatmap of DEGs.

### Quantitative RT-PCR

To validate transcriptome results, DCs obtained from patients with T1D (*n* ≥ 4) and control subjects (*n* ≥ 3) were cultured and pelleted in three conditions: iDCs, PSAB-DCs and mDCs. RNA was isolated using RNeasy Micro Kit (QIAGEN), and was reversetranscribed with a High Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific). cDNA synthesis reactions were carried out using random hexamers (0.5 mg/ml, BioTools, Valle de Tobalina, Madrid, Spain) and reverse transcriptase Moloneymurine-Leukemia-virus (200 U/ml, Promega, Madison, WI, USA). Quantitative RT-PCR assays were performed with TaqMan universal assay (ThermoFisher Scientific) on a LightCycler® 480 (Roche, Mannheim, Germany) using the following TaqMan assays: *CYTH4* (Hs01047905\_m1), *GIMAP4* (Hs01032964\_m1), *HPGD* (Hs00960590\_m1), NFKB inhibitor alpha (*NFKBIA*) (Hs00153283\_m1), *PLAUR* (Hs00958880\_m1), *TNFAIP3* (Hs00234713\_m1), tumor necrosis factor superfamily member 14 (*TNFSF14*) (Hs00542477\_m1), and *VEGFA* (Hs00900055\_m1). Relative quantification was performed by normalizing the expression for each gene of interest to that of the housekeeping gene *GAPDH* (Hs02758991\_g1), as described in the 2–ΔCt method (28).

### Statistical Analysis

The statistical analysis was performed using Prism 7.0 software (GraphPad software Inc., San Diego, CA, USA). Analysis of variance (ANOVA) was used for comparisons with several factors. For comparisons of unpaired data, a non-parametric Mann-Whitney test was used; for paired comparisons, a nonparametric Wilcoxon test was used. A *p* value ≤ 0.05 was considered significant.

### RESULTS

### Patients with T1D and Control Subjects Display Similar Features

Thirty-four patients with T1D (50% female, 50% male) from the Germans Trias i Pujol Hospital and 24 control subjects (45.8% female, 54.2% male) met the inclusion and exclusion criteria and were included in the study (**Table 1**). Age of control subjects was 30.46 ± 8.18 years (mean ± SD), while that of patients with T1D was 32.54 ± 8.96 years; BMI was 23.90 ± 2.87 kg/m2 and 23.80 ± 3.11 kg/m2 , respectively. Patients with T1D had been diagnosed at 20.79 ± 9.90 years, had a duration of disease of 11.75 ± 9.70 years, and a hemoglobin A1c level of 7.66 ± 1.26%. Control subjects did not significantly differ from patients with T1D in terms of age or BMI. Within the 34 patients, we selected 8 for the RNA-seq analysis, 50% female and 50% male, with a more stringent inclusion and exclusion criteria. Their age was 29.75 ± 5.85 years and their BMI was 22.50 ± 2.00 kg/m2 . They had been diagnosed with T1D at 27.13 ± 7.43 years, had a

Table 1 | Data from the control subjects and patients with T1D recruited for the study.


*Data presented as mean* ± *SD; p value calculated from Mann–Whitney test. T1D: type 1 diabetes; BMI, body mass index; NA, not applicable.*

duration of the disease of 2.50 ± 1.98 years and a hemoglobin A1c level of 7.36 ± 2.07%. Specific information on each subject can be found in **Table 2**.

### DC Differentiation Efficiency Is Similar in Patients with T1D and Control Subjects

Monocytes were isolated magnetically from PBMCs. The yield of monocyte isolation—calculated as the percentage of the absolute number of CD14+ cells in the positively isolated fraction related to the absolute number of CD14<sup>+</sup> cells in PBMCs was 54.95 ± 24.97% (mean ± SD) for control subjects and 56.62 ± 18.12% for patients with T1D. The percentage of purity of CD14<sup>+</sup> cells in the isolated fraction was 80.59 ± 10.18% for control subjects and 79.33 ± 7.56% for patients, and viability was 95.06 ± 4.14 and 94.92 ± 3.60%, respectively. The efficiency of differentiation to DCs at day 6 was 87.96 ± 6.61% for control subjects and 86.92 ± 6.86% for patients. No statistically significant differences were found when comparing these parameters between both groups. Data are detailed in **Table 3**.

### PS-Liposomes Show Multivesicular Vesicle Morphology and Encapsulate Insulin Peptides

Liposomes were characterized in terms of diameter, polydispersity index (PdI), surface charge (ζ-potential) and efficiency of peptide encapsulation (**Table 4**). All liposomes had a final lipid concentration of 30 mM. All liposomes were large to guarantee efficient phagocytosis, displaying a diameter superior to 690 nm. The presence of PS molecules in liposomes was confirmed by the negative charge measured at the liposome surface by ζ-potential (−38 mV). Regarding specific features of PSA-liposomes (*n* = 3), the mean diameter was 690 ± 29 nm (mean ± SD), the PdI was 0.40 ± 0.28 and the ζ-potential was −38.57 ± 6.76 mV. The mean of peptide A (human insulin A chain) encapsulation efficiency was 39.74 ± 22.10%. As for PSBliposomes (*n* = 3), they had a mean diameter of 788 ± 264 nm, the PdI was 0.52 ± 0.42 and the ζ-potential was −37.50 ± 7.16 mV, and the mean of peptide B (human insulin B chain) encapsulation efficiency was 93.19 ± 0.92%. Differences in peptide encapsulation efficiency (PSA *vs.* PSB) are due to amino acid composition and different solubility of insulin chains A (21


Table 2 | Data from the patients with T1D included in the RNA-seq experiment.

*Data presented as mean* ± *SD.*

*T1D, type 1 diabetes; BMI, body mass index.*

aa) and B (30 aa) in phosphate buffered saline media. B chain is more positively charged than A chain at neutral pH, resulting in a higher encapsulation efficiency in negatively charged liposomes. PSA-liposomes and PSB-liposomes presented multivesicular vesicle morphology when cryo-TEM analysis was performed (**Figure 1**).

Fluorescent-labeled PS-liposomes (*n* = 4) showed a diameter of 836 ± 217 nm, a PdI of 0.32 ± 0.06 and a ζ-potential of −38.90 ± 2.52 mV. Their PS-free counterparts, fluorescent PC-liposomes (*n* = 4), had a diameter of 1665 ± 488 nm, a PdI of 0.32 ± 0.09 and a ζ-potential of −7.60 ± 2.68 mV (**Table 4**).


*Yield: % of the absolute number of CD14*+ *cells in the positively isolated fraction related to the absolute number of CD14*+ *cells in PBMCs. Purity: % of CD14*+ *cells in the isolated fraction. Viability: % of Annexin V*– *7aad*–  *cells. Differentiation efficiency: % of CD11c*+ *cells. Data presented as mean* ± *SD; p value calculated from Mann–Whitney test. T1D, type 1 diabetes.*

### Human DCs Display Optimal Kinetics of PS-Liposomes Phagocytosis without Affecting Viability

A time course analysis was performed to determine PS-liposomes uptake kinetics (**Figure 2A**, upper left panel). The capture of empty fluorescent PS-liposomes by DCs was significantly higher at 37°C when compared to 4°C (*p* < 0.001), coming from either control subjects (*n* = 5) or patients (*n* = 10). This result is immunologically crucial and demonstrates that DCs engulf liposomes by an active mechanism of phagocytosis. PS-liposomes uptake kinetics were identical between control subjects and patients with T1D.

To indirectly assess the role of PS in phagocytosis, the same analysis was performed replacing PS-liposomes with PC-liposomes (**Figure 2A**, lower left panel). The percentages of empty PC-liposomes phagocytosis by DCs from control subjects (*n* = 6) and patients (*n* = 9) were significantly higher at 37°C when compared to 4°C starting at 2 h (*p*< 0.0001). The kinetics of the capture did not differ between control subjects and patients. When comparing uptake kinetics of PS- and PC-liposomes (Figure S1 in Supplementary Material), statistically significant differences were found, as expected. The presence of PS significantly accelerated phagocytosis in the first 2 h of co-culture (*p* < 0.05) both in control subjects and patients. In preliminary experiments, each type of liposome was tested in several sizes (diameter range 505–2,138 nm), and similar kinetics of capture were observed,



*Data presented as mean* ± *SD.*

Figure 1 | PS-liposomes display multivesicular and multilamellar morphology. Cryogenic transmission electron microscopy images of (A) PSA-liposomes (left) and (B) PSB-liposomes (right). Bar = 0.2 μm.

Figure 2 | Liposomes are efficiently phagocyted by dendritic cells (DCs) and preserve a high viability. (A) Uptake of liposomes fluorescently labeled with lipid-conjugated fluorescent dye Oregon Green 488 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine by DCs. Upper panel: time course of the capture of fluorescently-labeled PS-liposomes by DCs obtained from control subjects (white circles, *n* = 5) and patients with type 1 diabetes (T1D) (black circles, *n* = 10) at 37°C (continuous line) and at 4°C (discontinuous line). Results are mean ± SEM. Comparisons between phagocytosis by control subjects DCs at 37 and 4°C showed significant differences [++++*p* < 0.0001, two-way analysis of variance (ANOVA)]; also, significant differences were found when comparing phagocytosis in patients with T1D at 37 and 4°C (\*\*\**p* < 0.001, \*\*\*\**p* < 0.0001, Two-way ANOVA). No differences were found when comparing PS-liposomes uptake by DCs from control subjects and patients with T1D (Two-way ANOVA). Lower panel: time course of the capture of fluorescently-labeled PC-liposomes by DCs obtained from control subjects (white squares, *n* = 6) and patients with T1D (black squares, *n* = 9) at 37°C (continuous line) and at 4°C (discontinuous line). Results are mean ± SEM. Comparisons between phagocytosis by control subjects DCs at 37 and 4°C showed significant differences (++++*p* < 0.0001, Two-way ANOVA); also, significant differences were found when comparing phagocytosis in patients with T1D at 37 and 4°C (\*\**p* < 0.01, \*\*\*\**p* < 0.0001, Two-way ANOVA). No differences were found when comparing PC-liposomes uptake by DCs from control subjects and patients with T1D (Two-way ANOVA). (B) Viability of DCs from control subjects (upper panel, white symbols, *n* ≥ 6) and patients with T1D (lower panel, black symbols, *n* ≥ 6) assessed by annexin V and 7aad staining. Triangles represent immature DCs (iDCs), circles represent iDCs cultured with PSA-liposomes and PSB-liposomes (PSAB-DCs), squares represent mature PSAB-DCs (mPSAB-DCs) and upside-down triangles represent mature DCs (mDCs). MDCs were induced by culture with cytokine cocktail.

independently of liposome size (Figure S2 in Supplementary Material), thus confirming that PS is the key factor in accelerating phagocytosis.

The viability of the different conditions of DCs (iDCs, PSAB-DCs, mPSAB-DCs, and mDCs) was assessed to determine liposome toxicity. The mean viability for each condition was always >90%, both in DCs obtained from control subjects (*n* ≥ 6) (**Figure 2B**, upper right panel), and patients with T1D (*n* ≥ 6) (**Figure 2B**, lower right panel).

## PS-Liposomes Uptake Regulates the Phenotypic Maturation of Human DCs

Changes in DCs phenotype were determined in control subjects (*n* ≥ 5) and patients with T1D (*n* ≥ 8). The membrane molecules assessed were: PS-receptors (CD36, TIM4, and αvβ5 integrin), antigen-presentation molecules (HLA-ABC and HLA-DR), adhesion molecules (CD54), costimulation molecules (CD40 and CD86), activation molecules (CD25), chemokine receptors (CCR7, CCR2, and CXCR4), and pattern recognition receptors (TLR2, CD14, and DC-SIGN). **Figure 3** shows the relative Median of Fluorescence Intensity referred to mDCs.

PS-receptors CD36, TIM4 and αvβ5 integrin were expressed in iDCs. After liposome uptake, PSAB-DCs from patients decreased CD36 expression (*p* < 0.05) and upregulated TIM4 expression (*p* < 0.05) in comparison to iDCs, but PSAB-DCs presented a higher expression of CD36 and TIM4 than mDCs from control subjects and patients (*p* < 0.05). Moreover, PSAB-DCs from patients had increased levels of TIM4 in comparison to PSAB-DCs from control subjects (*p* < 0.05). The expression of αvβ5 integrin was higher in PSAB-DCs than in mDCs in patients (*p*< 0.05). As expected, CD36 and TIM4 were downmodulated in mDCs (*p* < 0.05), and αvβ5 integrin showed the same tendency.

Regarding HLA molecules, HLA-ABC was expressed similarly in iDCs and mDCs from both groups, and decreased in PSAB-DCs from patients after liposome capture (*p* < 0.05) —and control subjects displayed the same tendency. Concerning HLA-DR, iDCs showed a lower expression of this marker when compared to mDCs (*p* < 0.001). After liposome phagocytosis (PSAB-DCs), the low HLA-DR levels were preserved. As for the expression of adhesion molecule CD54, it was lower in iDCs from patients in comparison to mDCs (*p* < 0.05), and control subjects displayed the same tendency. After liposome uptake, no changes in CD54 expression were observed in PSAB-DCs when compared to iDCs, but mDCs displayed increased levels of CD54 in comparison to PSAB-DCs (*p* < 0.05). The expression of CD54 was higher in mPSAB-DCs exposed to a maturation stimulus despite the uptake of liposomes (*p* < 0.05).

Expression of costimulatory molecules CD40 and CD86 was lower in iDCs than in mDCs (*p* < 0.01). Liposome phagocytosis did not increase the expression of these molecules in PSAB-DCs. Moreover, PSAB-DCs presented lower levels of these markers when compared to mDCs (*p* < 0.0001), and even when exposed to pro-inflammatory stimulus (mPSAB-DCs) in comparison to mDCs (*p* < 0.05). Regarding the expression of activation marker CD25, it was lower in iDCs when compared to mDCs (*p*< 0.0001). Upregulation of CD25 was observed after liposome uptake in PSAB-DCs from control subjects (*p* < 0.01), but remained unaltered in patients. PSAB-DCs from both groups presented CD25 downmodulated when compared to mDCs (*p* < 0.001). Furthermore, DCs loaded with liposomes and exposed to proinflammatory stimulus (mPSAB-DCs) displayed lower levels of CD25 than mDCs in patients with T1D (*p* < 0.01).

Chemokine receptors CCR7 and CCR2 were expressed in iDCs. After liposome capture, the expression of both molecules increased in patients with T1D (*p* < 0.05). CCR7 was upregulated in PSAB-DCs when compared to mDCs in patients (*p* < 0.05), and control subjects displayed the same tendency. CXCR4, overexpressed in mDCs in comparison to iDCs (*p* < 0.05), was maintained low after liposome engulfment (PSAB-DCs). The expression of CXCR4 was higher in DCs exposed to a maturation stimulus despite the uptake of liposomes (mPSAB-DCs) (*p* < 0.05).

Pattern recognition receptors were assessed in DCs. TLR2 expression was similar in all experimental conditions, despite showing a tendency to increase after liposome phagocytosis (PSAB-DCs) in patients. CD14 was similarly expressed in iDCs and mDCs, but liposome uptake and maturation stimulus induced downregulation of this marker (*p*< 0.05). DC-SIGN, expressed in iDCs, displayed a tendency to be downmodulated after liposome capture (PSAB-DCs), especially in controls, which was more marked after a pro-inflammatory stimulus. Nonetheless, this marker showed a tendency to remain higher in PSAB-DCs than in mDCs in patients.

In terms of cytokine secretion by DCs from patients (*n* ≥ 3) and control subjects (*n* ≥ 3) (**Figure 4**), IL-6 was released in low amounts after liposome phagocytosis and, as expected, its secretion increased after maturation stimulus. TNF-α was not increased after liposome uptake and its secretion increased in pro-inflammatory conditions. Liposome engulfment maintained a high profile of TGF-β1 secretion both in control subjects and patients, and tended to decrease in mPSAB-DCs and mDCs, although non-significant. Regarding IL-10 production, PSAB-DCs displayed a tendency to increase the secretion, although non-statistically significant, in patients with T1D. IL-2, IL-17A, and IFN-γ were not detected in any condition of the assay (data below the detection limit). IL-4 was not considered as it was used in culture media for DC differentiation.

### PS-Liposomes Uptake Impairs DCs Ability to Stimulate Autologous T Cell Proliferation

DCs derived from patients with T1D (*n*≥ 12) and control subjects (*n* = 12) induced similar levels of autologous T cell proliferation (**Figure 5**). As expected, CD4<sup>+</sup> T cell proliferation induced by mDCs was higher than proliferation induced by iDCs in both groups (*p* < 0.01). CD8<sup>+</sup> T cell proliferation induced by mDCs was higher than proliferation induced by iDCs in control subjects (*p* < 0.05), but not in patients. Importantly, the capture of PSAB-liposomes by iDCs did not increase autologous CD4<sup>+</sup> and CD8<sup>+</sup> T cell proliferation, both in patients and control subjects. Moreover, a significant decrease of CD8<sup>+</sup> T cell proliferation induced by PSAB-DCs from patients was observed after liposome capture, when compared to iDCs (*p* < 0.01). This effect was reverted after DCs maturation.

In terms of cytokine production, PBMCs co-cultured with PSAB-DCs displayed a cytokine profile (IL-6 and IFN-γ) similar to iDCs (**Figure 5**). Interestingly, PBMCs showed a tendency to increase IFN-γ and IL-6 secretion when co-cultured with mPSAB-DCs or mDCs, respectively, only in patients with T1D. IL-2, IL-4, IL-10, IL-17A, and TNF-α were not detected in any condition of the assay (data below the detection limit).

### Transcriptional Changes in DCs from Patients with T1D after PS-Liposomes Phagocytosis Point to an Immunoregulatory Prolife

RNA-seq analysis was performed in DCs from 8 patients with T1D (**Table 2**) in order to identify transcriptional changes after the capture of PS-liposomes. Phagocytosis was verified by flow cytometry using fluorescent liposomes. After 4 h of co-culture, 73.88 ± 11.57% (mean ± SD) of DCs were positive for fluorescent signal.

Figure 3 | Capture of PSA-liposomes and PSB-liposomes regulates dendritic cell (DCs) phenotype. Relative CD36, TIM4, Integrin αvβ5, HLA-ABC, HLA-DR, CD54, CD40, CD86, CD25, CCR7, CCR2, CXCR4, TLR2, CD14, and DC-SIGN membrane expression in DCs obtained from control subjects (white bars, *n* ≥ 5) and patients with type 1 diabetes (T1D) (black bars, *n* ≥ 8). Bars represent immature DCs (iDCs), iDCs after the capture of PSA-liposomes and PSB-liposomes (PSAB-DCs), mature PSAB-DCs (mPSAB-DCs), or mature DCs (mDCs), 24 h after culture. MDCs were induced by culture with cytokine cocktail. Data presented as mean ± SD of relative Median of Fluorescence Intensity (MFI), this being MFI of each culture condition referred to their respective mDCs control. Significant differences were found when comparing culture conditions in the same group of subjects (\**p* ≤ 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001, Wilcoxon test), and when comparing the same culture conditions between patients with T1D and control subjects (+*p* < 0.05, Mann–Whitney test).

iDCs after the capture of PSA-liposomes and PSB-liposomes (PSAB-DCs), mature PSAB-DCs (mPSAB-DCs), or mature DCs (mDCs). MDCs were induced by culture with cytokine cocktail. Data presented as mean ± SD. Significant differences were found when comparing the different conditions in the same group of subjects (\**p* < 0.05, Wilcoxon test), and differences were not found when comparing the same culture conditions between patients with T1D and control subjects (Mann–Whitney test).

Integrity of the isolated RNA material was assessed for each sample, being optimal for RNA-seq experiment: RIN 9.0 ± 0.56 (mean ± SD). Bioinformatics analysis of the RNA-seq experiment revealed that only 233 of 22,711 genes detected were differentially expressed between iDCs and PSAB-DCs (*p* value < 0.0013, adjusted *p* value < 0.1254). Of these 233 genes, 203 (87.12%) were downregulated and the remaining 30 (12.88%) were upregulated, and 224 corresponded to protein-coding genes. Despite the heterogeneous basal transcriptomics of DCs from eight patients, gene expression was clearly altered toward a similar profile after PS-liposomes phagocytosis (Figure S3 in Supplementary Material).

We analyzed several categories and molecules related to DC function (Table S1 in Supplementary Material). DEGs were mainly related to metabolism, gene expression, immunoregulation, signal transduction, molecule transport, post-translational protein modification, cytokine signaling, cell cycle, vesicle-mediated processes, DNA replication and repair, antigen processing and presentation, apoptosis, and cytoskeleton organization (**Table 5**). Due to the immunotherapeutic potential of PS-liposomes, DEGs involved in tolerance were analyzed in detail. DEGs linked to the immune system were primarily downregulated and involved in antigen processing and presentation (*KBTBD6*, *BTK*, *CDC23*, *UBE2E3*, *CD1D*), regulation of the immune response (*DAPP1*,

Figure 5 | Capture of PSA-liposomes and PSB-liposomes affects dendritic cells (DCs) functionality. (A) Relative autologous proliferation of CD3+CD4+ and CD3+CD8+ subsets induced by DCs obtained from control subjects (white bars, *n* = 12) and patients with type 1 diabetes (T1D) (black bars, *n* ≥ 12). Autologous peripheral blood mononuclear cells were stained with CellTrace Violet (CTV) and co-cultured at 1:10 ratio for 6 days with each condition of DCs, and proliferation was measured as the percentage of CTVlow cells. Bars represent immature DCs (iDCs), iDCs after the capture of PSA-liposomes and PSB-liposomes (PSAB-DCs), mature PSAB-DCs (mPSAB-DC), or mature DCs (mDCs). MDCs were induced by culture with cytokine cocktail. Data presented as mean ± SD of relative proliferation induction, this being the percentage of CTVlow cells in each co-culture condition referred to that of their respective mDCs control. Significant differences were found when comparing culture conditions in the same group of subjects (\**p* ≤ 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001, Wilcoxon test), and differences were not found when comparing the same culture conditions between patients with T1D and control subjects (Mann-Whitney test). (B) IL-6 and IFN-γ secretion (pg/ ml) assessed in supernatants of autologous proliferation co-culture with cells from control subjects (white bars, *n* ≥ 3) and patients with T1D (black bars, *n* ≥ 3). Bars represent immature DCs (iDCs), iDCs after the capture of PSA-liposomes and PSB-liposomes (PSAB-DCs), mature PSAB-DCs (mPSAB-DC), or mature DCs (mDCs). MDCs were induced by culture with cytokine cocktail. Data presented as mean ± SD. No differences were found when comparing the different conditions in the same group of subjects (Wilcoxon test), or when comparing the same culture conditions between patients with T1D and control subjects (Mann–Whitney test).

*GIMAP4*, *SLAMF6*) and cytokine signaling relevant in the interaction between T cells and DCs (*SOCS2*, *TNFRSF11A*). However, although very few genes were upregulated after PS-liposomes phagocytosis, these were related to the prevention of DC maturation [*TNFSF14*, *TNFAIP3*, *VEGFA*, *SHB*, leukocyte associated immunoglobulin-like receptor 1 (*LAIR1*), *NFKBIA*]. Also, genes related to apoptosis were downregulated in DCs after liposome uptake (*BLCAP*, *PMP22*, *LMNB1*).

Validation by qRT-PCR of the selected gene targets (*CYTH4*, *GIMAP4*, *HPGD*, *NFKBIA*, *PLAUR*, *TNFAIP3*, *TNFSF14*, and *VEGFA*) confirmed the RNA-seq results, when tested in DCs from 10 patients with T1D (**Figure 6**). As expected, gene expression analysis in DCs from 9 control subjects showed the same pattern. Regarding mDCs—from 4 patients with T1D and 3 control subjects—we observed a seemingly different gene expression pattern when compared to PSAB-DCs — and with iDCs. Genes upregulated by PS-liposomes, such as *CYTH4* and *TNFSF14*, tended to be downregulated in mDCs; other genes tended to be differentially expressed in mDCs (*NFKBIA*, *PLAUR*, *TNFAIP3*, *GIMAP4*, *VEGFA*, and *HPGD*) in comparison to the other conditions.

### DISCUSSION

Apoptosis is a key factor in the maintenance of immunological tolerance. The uptake of apoptotic cells, through a process called efferocytosis, results in tolerogenic presentation of autoantigens inducing specific tolerance rather than autoimmunity (14). Table 5 | DEGs in dendritic cells (DCs) from patients with type 1 diabetes (T1D) after PSA-liposomes and PSB-liposomes phagocytosis.


*DEGs, differentially expressed genes.*

Figure 6 | Quantitative RT-PCR validates the RNA-seq results. Relative gene expression of 8 selected targets in immature dendritic cells (iDCs), after phagocytosis of PSA-liposomes and PSB-liposomes (PSAB-DCs), and in mature dendritic cells (mDCs), analyzed by quantitative RT-PCR. Gene expression signals were normalized to *GAPDH*. Bars show the mean ± SD of gene expression in control subjects (white bars, *n* ≥ 3) and patients with type 1 diabetes (T1D) (black bars, *n* ≥ 4). Statistically significant differences were found when comparing the different conditions in the same group of subjects (\**p* < 0.05, Wilcoxon test), and differences were not found when comparing the same culture conditions between patients with T1D and control subjects (Mann–Whitney test).

Therefore, the inherent immunomodulatory properties of apoptotic cells can be useful to design innovative immunotherapies. Based on this tolerogenic potential of efferocytosis, we generated liposomes that mimic apoptotic β-cells. At present, liposomes are clinically used mainly as vehicles for drugs (29–31), but they can be designed to modulate immune responses. This liposome-based immunotherapy resembles apoptotic cells and acts through the immunosuppressive signal of PS (32) and tolerogenic autoantigen presentation. These large PS-liposomes, after phagocytosis, are effective in restoring self-tolerance in experimental autoimmune diseases (16, 17) by their interaction with DCs and the arrest of the autoimmune reaction. To explore the clinical potential of this strategy, we have determined the effect of PS-liposomes loaded with human insulin peptides in DCs from patients with T1D. This effect has been assessed in several aspects: phagocytosis, phenotypic changes, effect on T cell proliferation and cytokine profile.

Regarding phagocytosis, lipid membrane composition is crucial for rapid engulfment by DCs, as demonstrated using PSand PC-liposomes. As expected, the presence of PS accelerated phagocytosis of liposomes both in control subjects and patients with a dynamic typical of apoptotic cell clearance. PS-liposomes were more efficiently engulfed by DCs than the equivalent ones without PS in the first 2 h of co-culture, reaching plateau after 6 h. When encountering PS-liposomes, DCs are deceived into sensing that they are actual apoptotic cells that need to be rapidly efferocyted in order to avoid secondary necrosis that could contribute to autoimmunity (33, 34). Moreover, the preservation of DCs viability proved that PS-liposomes are not toxic, as reported for other types of liposomes (29–32).

The second aspect was the assessment of DCs phenotype. After liposome engulfment, PSAB-DCs maintained high levels of PS-receptors that mediate this uptake, when compared to mDCs, pointing to the preservation of phagocytosis ability in tolerogenic DCs (tolDCs). Interestingly, the upregulation of TIM4 expression observed in PSAB-DCs from patients might contribute to a positive feedback of phagocytosis. Upon maturation, PS-receptors were downmodulated correlating with the phagocytic capacity of mDCs, as described (35). The expression pattern of molecules involved in antigen presentation (HLA, costimulatory and adhesion molecules) in PSAB-DCs concurs with a tolerogenic function, both in patients and control subjects. The expression of CD25 activation marker, linked to DCs activation and autoimmunity (36, 37), confirmed the intermediate activation status of PSAB-DCs after phagocytosis. Also, the chemokine receptors expression pattern supports DCs ability to drive their migration to secondary lymphoid tissues (38, 39), and moreover, the high CCR7 expression is associated with induction of tolerance after efferocytosis (40). Additionally, the expression of pattern recognition receptors was not altered by liposomes, as described for human DCs (32). This phenotype is similar to the previously observed in mice (16, 17). Of note, RNA-seq analysis reinforces these results. Furthermore, upon liposome capture, the immunomodulatory cytokine TGFβ-1 was secreted, a reported effect driven by PS (34) that could suppress DC maturation and define the T cell response afterward. As expected, liposome capture did not induce IL-6 nor TNF-α secretion by DCs, but maturation did. Overall, the results point to the tolerogenic effect of these vesicles, which act on re-establishing self-tolerance. We observed minor phenotypic differences between DCs from patients and control subjects, which could be due to epigenetic changes caused by autoimmunity and metabolic dysregulation (41–43).

The third aspect was the analysis of autologous T cell proliferation induced by PSAB-DCs. In agreement with DCs phenotype, T cell proliferation induced by PSAB-DCs was similar or even lower than the induced by iDCs, both in patients and control subjects. Interestingly, in patients with T1D, there was a significant reduction in CD8<sup>+</sup> T cell subset proliferation induced by tolDCs when compared to iDCs. This effect could be related to a reduction of the T cell cytotoxic activity, the most important effector response in human T1D (44, 45). In fact, after efferocytosis, DCs present apoptotic cell autoantigens to cognate T cells in the absence of costimulation, favoring tolerance to self (12, 13). It is reasonable to think that liposomes mimicking apoptotic cells will cause a similar effect, Additional studies using tetramers would be relevant to determine the antigen-specificity of the T cells involved in tolerance induction, even in pro-inflammatory conditions, in which T cells seem to proliferate more vigorously. Cytokines produced during the autologous T cell proliferation assay induced by PSAB-DCs discard a Th1 and Th17 profile, which could be detrimental in the induction of tolerance. In fact, IFN-γ, which is involved in a Th1 response, and IL-6 secretion, which partially contributes to induce Th17 response in T1D (46), remain poorly secreted in co-cultures of PBMCs with PSAB-DCs. Interestingly, higher amounts of IL-6 and IFN-γ tend to be produced by mDCs from patients with T1D when compared to controls. This feature could reflect the ongoing autoimmune reaction, present in peripheral blood from patients (47).

One of the obstacles of tolerogenic therapies in human disease is the heterogeneity of the *ex vivo*-generated tolDCs, which vary depending on the source, the manufacturing protocols, and the timespan of the experiment. Work is in progress to define and standardize a set of phenotypical and functional characteristics of tolDCs (48). To date, tolDCs are accepted as maturation-reluctant cells with low expression of antigen-presenting and costimulatory molecules and a tolerogenic-skewed cytokine profile (49). In this sense, one of the advantages of direct administration of the liposomes reported herein would be the generation of tolDCs *in vivo*, avoiding *ex vivo* cell manipulation. Our previous results in mice demonstrate this hypothesis (16, 17). However, a global picture of changes induced by PS-liposomes phagocytosis would grant a better understanding of tolerogenicity.

To fully characterize the immunomodulatory effects of liposomes, transcriptomic analysis was performed in DCs. Eight patients with a short T1D duration were selected in order to minimize the influence of long-term hyperglycemia on immune response, as reported (41–43). RNA-seq revealed a set of DEGs that avoid DCs maturation and contribute to tolerogenic antigen presentation. One of the most hyperexpressed genes was the vascular endothelial growth factor (VEGF) A (*VEGFA*), involved in cytokine signaling after efferocytosis (50), iDCs recruitment and maturation inhibition (51). VEGF increases the expression of the *TNFSF14* gene (52), also upregulated by PS-liposomes (53). In turn, TNFSF14 upregulates the production of TGF-β by phagocytes (54), and upon interacting with its ligand in T cells, TNFSF14 regulates T cell proliferation (55), inducing local immunosuppression (56). Supporting this fact, apoptotic cell clearance has been described to inhibit inflammation *via* TGF-β and VEGF production (34). Additionally, VEGFA enhances the expression of indoleamine 2,3-dioxygenase (*IDO*) (57), which in turn codifies for an immunomodulatory enzyme expressed in tolDCs (58). Moreover, the hematopoietically expressed homeobox (*HHEX*) gene, a repressor of VEGF signaling (59), is downregulated after PS-liposomes phagocytosis, whereas an inductor of VEGF expression, activating transcription factor 4 (*ATF4*) (60), is upregulated. Furthermore, the hyperexpressed SH2 domain containing adaptor protein B (*SHB*) gene codifies for a protein that regulates VEGF-dependent cellular migration (61), Th2 polarization and T regulatory cell induction (62). Other upregulated tolerogenic genes, such as the TNF alpha induced protein 3 (*TNFAIP3*) and the *LAIR1*, can inhibit DC maturation and their deficiency causes autoimmune and autoinflammatory diseases (63–66). In the same way, the hyperexpression of the *NFKBIA* gene would contribute to inhibit DC maturation and T cell activation (67, 68). Regarding cytokine signature, our results agree with those found in phenotypic and functional experiments. A relevant cytokine for tolerance induction is TGF-β1, secreted after efferocytosis (69). After PS-liposomes capture, DCs showed a biological increase of TGF-β1 transcription, although non-significant, probably due to the short timespan of the experiment (70). In fact, TGF-β1 was found in culture supernatants 24 h after PS-liposomes phagocytosis. Also, the immunoregulatory interferon lambda receptor 1 (*IFNLR1*) gene is one of the few overexpressed in DCs after PS-liposomes uptake. This receptor induces tolDCs that promote regulatory T cell expansion (71). Unexpectedly, the *TNF*-encoding gene was upregulated in DCs after liposome phagocytosis, and the same tendency was observed in protein secretion. Nevertheless, this behavior was very different to that observed in mDCs, which secreted higher amounts of TNF. The increase of TNF gene expression in our RNA-seq agrees with the upregulation of *TNFSF14* and *TNFAIP3* genes. Furthermore, a critical role for TNF has been reported in human tolDCs in the induction of antigen-specific regulatory T cells (72). Also, our previous results showed that murine tolDCs upregulated TNF-gene expression after efferocytosis (14). Overall, these results are consistent with the pleiotropic effects of TNF. Furthermore, in our previous research, PGE2 was found to be crucial in tolerance induced by PS-liposomes in mice (16). Strikingly, this pathway does not seem to be upregulated in human DCs, probably due to divergences between mice and men. Nevertheless, our data indirectly point to the involvement of the PGE2 pathway in human DCs: first, the downregulation of the hydroxyprostaglandin dehydrogenase 15-(NAD) (*HPGD*) gene, involved in PGE2 degradation, and second, a biological upregulation (although non-significant) of the peroxisome proliferator activated receptor gamma (PPARG), a gene induced by prostaglandins which is a negative regulator of pro-inflammatory cytokines (73). Furthermore, PGE2 has been described to stimulate the synthesis of VEGF (74). In summary, comparative transcriptome studies identify the whole molecular features of tolDCs rather than describe a simple state of maturation or lack thereof in terms of phenotype and function (75). Further studies are required to find a common signature of tolerogenicity, a fact hindered by individual differences of human DCs and the heterogeneous results obtained with different agents used to promote tolDCs. Our findings describe the specific gene signature of PS-liposomes-induced tolDCs. Their genomic program, which drives a different functionality than those of iDCs and mDCs, contributes to dissect the complexity of tolerance regulation.

Perhaps not so peculiarly, most of the alterations found in DCs after PS-liposomes capture are also physiopathological strategies used by tumor cells to escape immune surveillance. Small vesicles rich in PS are released by tumor cells and act as immunosuppressive agents to inhibit tumor antigen-specific T cells (76). Tumor cells can induce immunological tolerance using mechanisms characteristic of apoptotic cell clearance, and PS-liposomes seem to make use of the same pathways to achieve similar effects.

The use of PS-liposomes filled with autoantigens is an innovative strategy to arrest autoimmunity by restoring tolerance to self. As a whole, our results support the tolerogenic behavior of DCs, induced by the phagocytosis of PS-liposomes, and suggest that, in the context of autoimmunity, they could act silencing potential autoreactive T cells. This process could possibly be an active silencer, and not only a lack of maturation of DCs. In summary, here we unveil a picture of efferocytosis mimicry that leads to phenotypic and functional changes in human DCs, accountable for tolerance induction. The herein reported results reinforce the potential of this biocompatible immunotherapy to re-establish immunological tolerance, opening the door to new therapeutic approaches in the field of antigen-specific autoimmune disorders.

### ETHICS STATEMENT

This study was carried out after the approval and in strict accordance with the recommendations of the guidelines of Germans Trias i Pujol Ethical Committee. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

## AUTHOR CONTRIBUTIONS

SR-F, IP-A, MC-S, FV, JV, and MV-P designed the study and interpreted the data; SR-F, DM, JV, and MV-P wrote the manuscript; SR-F, IP-A, DP-B, MC-S, SG-J, and AV performed the experiments; IP-A, SR-F, DM, and MV-P supervised the experiments; SR-F, EA, FV, and MV-P selected the patients and supervised the collection of blood samples; and SR-F, IP-A, FB, and AS analyzed the data. All authors revised the work and gave final approval of the version to be published.

### ACKNOWLEDGMENTS

We are grateful to the physicians and nurses working at the Endocrinology Section and the Clinical Investigation Unit (UPIC) from the Germans Trias i Pujol University Hospital, and to Dr. J. Carrascal and Ms. B. Quirant-Sanchez, from Immunology Section, for collecting blood samples and data, respectively. We are indebted to all blood donors who participated in this study and made it possible. We would also like to acknowledge Mr. M. Fernandez for technical assistance with flow cytometry, Dr. P. Armengol and Ms. A. Oliveira for their support in the RNA-seq experiment, and Dr. C. Prat for microbiological controls (IGTP). Special thanks to Ms. D. Cullell-Young for English grammar assistance.

### FUNDING

This work has been funded by a grant from the Spanish Government (FIS PI15/00198) co-financed with the European Regional Development funds (FEDER), by Fundació La Marató de TV3 (28/201632-10), by Catalan AGAUR (project 2014 SGR1365) and by CERCA Program/Generalitat de Catalunya. CIBER of Diabetes and Associated Metabolic Diseases (CIBERDEM) is an

### REFERENCES


initiative from Instituto de Salud Carlos III. ICN2 acknowledges the support of the Spanish MINECO through the Severo Ochoa Centers of Excellence Program, under grant SEV-2013-0295. This work has been supported by positive discussion through A FACTT network (Cost Action BM1305: www.afactt.eu). COST is supported by the EU Framework Program Horizon 2020. SR-F is supported by the Agency for Management of University and Research Grants (AGAUR) of the Generalitat de Catalunya.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at https://www.frontiersin.org/articles/10.3389/fimmu.2018.00253/ full#supplementary-material.

phosphatidylserine-liposomes to arrest autoimmunity in type 1 diabetes. *PLoS One* (2015) 10(6):e0127057. doi:10.1371/journal.pone.0127057


76. Kelleher RJ Jr, Balu-Iyer S, Loyall J, Sacca AJ, Shenoy GN, Peng P, et al. Extracellular Vesicles present in human ovarian tumor microenvironments induce a phosphatidylserine-dependent arrest in the T-cell signaling cascade. *Cancer Immunol Res* (2015) 3(11):1269–78. doi:10.1158/2326-6066. CIR-15-0086

**Conflict of Interest Statement:** IP, MC, JV, DM, and MV are inventors in a patent (WO2015107140) that describes the use of autoantigen-encapsulating liposomes for the prevention or treatment of autoimmune disorders.

*Copyright © 2018 Rodriguez-Fernandez, Pujol-Autonell, Brianso, Perna-Barrull, Cano-Sarabia, Garcia-Jimeno, Villalba, Sanchez, Aguilera, Vazquez, Verdaguer, Maspoch and Vives-Pi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# *Trichinella spiralis* excretory– secretory Products induce Tolerogenic Properties in human Dendritic cells *via* Toll-like receptors 2 and 4

*Nataša Ilic1 , Alisa Gruden-Movsesijan1 , Jelena Cvetkovic1 , Sergej Tomic1 , Dragana Bozidar Vucevic2 , Carmen Aranzamendi3,4, Miodrag Colic1,2, Elena Pinelli4† and Ljiljana Sofronic-Milosavljevic1 \*†*

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Daniel Benitez-Ribas, Hospital Clínic de Barcelona, Spain Elisabeth Zinser, Universitätsklinikum Erlangen, Germany Xin Ping Zhu, Capital Medical University, China*

#### *\*Correspondence:*

*Ljiljana Sofronic-Milosavljevic sofronic@inep.co.rs*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 29 September 2017 Accepted: 04 January 2018 Published: 24 January 2018*

#### *Citation:*

*Ilic N, Gruden-Movsesijan A, Cvetkovic J, Tomic S, Vucevic D, Aranzamendi C, Colic M, Pinelli E and Sofronic-Milosavljevic L (2018) Trichinella spiralis Excretory– Secretory Products Induce Tolerogenic Properties in Human Dendritic Cells via Toll-Like Receptors 2 and 4. Front. Immunol. 9:11. doi: 10.3389/fimmu.2018.00011*

*<sup>1</sup> Institute for the Application of Nuclear Energy, University of Belgrade, Belgrade, Serbia, 2Medical Faculty of the Military Medical Academy, University of Defence, Belgrade, Serbia, 3Groningen Biomolecular Science and Biotechnology Institute (GBB), University of Groningen, Groningen, Netherlands, 4Centre for Infectious Disease Control Netherlands, National Institute for Public Health and the Environment (RIVM), Bilthoven, Netherlands*

*Trichinella spiralis*, as well as its muscle larvae excretory–secretory products (ES L1), given either alone or *via* dendritic cells (DCs), induce a tolerogenic immune microenvironment in inbred rodents and successfully ameliorate experimental autoimmune encephalomyelitis. ES L1 directs the immunological balance away from T helper (Th)1, toward Th2 and regulatory responses by modulating DCs phenotype. The ultimate goal of our work is to find out if it is possible to translate knowledge obtained in animal model to humans and to generate human tolerogenic DCs suitable for therapy of autoimmune diseases through stimulation with ES L1. Here, the impact of ES L1 on the activation of human monocyte-derived DCs is explored for the first time. Under the influence of ES L1, DCs acquired tolerogenic (semi-matured) phenotype, characterized by low expression of HLA-DR, CD83, and CD86 as well as moderate expression of CD40, along with the unchanged production of interleukin (IL)-12 and elevated production of IL-10 and transforming growth factor (TGF)-β, compared to controls. The interaction with DCs involved toll-like receptors (TLR) 2 and 4, and this interaction was mainly responsible for the phenotypic and functional properties of ES L1-treated DCs. Importantly, ES L1 potentiated Th2 polarizing capacity of DCs, and impaired their allo-stimulatory and Th1/ Th17 polarizing properties. Moreover, ES L1-treated DCs promoted the expansion of IL-10- and TGF-β- producing CD4+CD25hiFoxp3hi T cells in indolamine 2, 3 dioxygenase (IDO)-1-dependent manner and increased the suppressive potential of the primed T cell population. ES L1-treated DCs retained the tolerogenic properties, even after the challenge with different pro-inflammatory stimuli, including those acting *via* TLR3 and, especially TLR4. These results suggest that the induction of tolerogenic properties of DCs through stimulation with ES L1 could represent an innovative approach for the preparation of tolerogenic DC for treatment of inflammatory and autoimmune disorders.

Keywords: *Trichinella spiralis*, excretory–secretory products, dendritic cells, tolerance, immunomodulation, tolllike receptors

### INTRODUCTION

One of the permanent challenges in immunology is overcoming the rising problem of losing the delicate balance, provided by the innate immunity, reflected in responding to foreign antigens while remaining tolerant to self-antigens. If this balance is altered due to an increased inflammatory response and diminished tolerance, it can result in autoimmune diseases such as type I diabetes, multiple sclerosis, rheumatoid arthritis, and inflammatory bowel disease (1). Currently available treatments for those diseases do not provide cure or a long-term remission. They usually include immunosuppressive drugs or biological agents which slow down the disease progress but can cause serious adverse effects (2). Given that available therapy cannot restore self-tolerance and provides only temporary remission, efforts are being made in order to develop new therapeutic approaches that would enable the restoration of tolerogenic immune response and silencing of autoimmune processes (3, 4). Dendritic cells (DCs), key antigen-presenting cells, possess the capacity for a fine tuning of the immune response and represent a good candidate as an immunotherapeutic tool (5–8), since their plasticity provides the opportunity to reverse the autoimmune process by mediating restoration of self-tolerance (9).

From the beginning of twenty-first century, a number of results obtained in animal model systems provided evidence that DCs could be treated in a way to acquire tolerogenic properties and that such DCs have a potential to mitigate autoimmune diseases like autoimmune diabetes (10, 11), collagen-induced arthritis (12, 13), experimental autoimmune encephalomyelitis (EAE) (14–16), experimental autoimmune myasthenia gravis (17, 18), and experimental autoimmune uveoretinitis (19, 20). It has been demonstrated that the tolerogenic status of DCs depends on the applied stimuli and maturation conditions. The doctrine that, by default, immature DCs are considered tolerogenic while mature DCs are immunogenic and induce effector responses (21) has been modified by the fact that even mature DCs could have tolerogenic properties (1, 22). Nevertheless, tolerogenic phenotype of DCs usually refers to semi-matured cells with low to intermediate expression of MHC II, as well as co-stimulatory molecules CD80, CD86, and CD40, with elevated production of interleukin (IL)-10 but decreased production of IL-12 (23–25). Different agents proved to be potent inducers of tolerogenic DCs phenotype. Among them are vitamin D3 (26, 27), corticosteroids (28), rapamycin (29), IL-10 (30, 31), or other cytokines (32). The observed phenomenon that tolerogenic DCs could modulate the course of autoimmune disease in terms of reducing the clinical signs and the severity of the disease, directed research from animal model system toward human DCs. Some of these investigations, that showed success *in vitro*, have turned into phase I clinical trials over time (5, 33). Nevertheless, a search for agents able to induce stable tolerogenic DCs is an ongoing story.

It is well known that parasite antigens have the potential to modulate the host immune response *via* DCs by inducing T helper (Th)2 and regulatory response while simultaneously inhibiting Th1 and Th17 response (34) and some of the investigated parasitic antigens showed the capacity to induce tolerogenic DCs phenotype (35, 36). Still, the results considering the impact of parasitic products on human DCs, their tolerogenic properties and the potential of these tolerogenic DCs to modulate the immune response, as well as the mechanisms employed in this phenomenon, are scarce. Potential candidates for the induction of tolerogenic DCs are excretory–secretory (ES L1) antigens of *Trichinella spiralis* muscle larvae. ES L1 antigens are a complex mixture of molecules, released by this parasite into the circulation during the chronic phase of the infection, which can activate regulatory network elements as guardians of homeostasis. Through the action of these products, mediated mainly by DCs, the parasite suppresses the host immune response against itself in order to survive, but it also mitigates the unwanted immune responses like those to autoantigens and allergens (37). Several studies, including our own (38), preformed in mouse model system, showed that ES L1 antigens of *T. spiralis* muscle larvae, or its components (39) possess the ability to induce the semi-matured DCs, which are able to induce the expansion of regulatory T cells (Tregs) *in vitro*. Our work using the rat model system also demonstrated that upon treatment with ES L1 antigens, DCs acquire semi-mature status and an increased capacity to induce Th2 and regulatory immune response both *in vitro* and *in vivo* (40). Also, ES L1-treated DCs, if applied prophylactically, showed considerable ability to modulate the outcome of EAE in Dark Agouti rats by activating and maintaining anti-inflammatory and regulatory immune response while alleviating pro-inflammatory response (16). This was reflected in the enhanced production of IL-4, IL-10, and transforming growth factor (TGF)-β, as well as in diminished production of interferon (IFN)-γ and IL-17, both on systemic level and in the target tissue (CNS). Also, the data obtained in this study indicated that the increased proportion of Foxp3<sup>+</sup> Tregs on systemic level and in CNS was associated with the amelioration of EAE. Moreover, the applied DCs managed to maintain such immunological profile throughout the disease, which indicates that ES L1-induced tolerogenic properties of DCs are functionally stable. Those results suggest that the immunomodulatory properties of ES L1-treated DCs are worth further research and the present study was designed to translate the knowledge obtained in animal model system on humans. However, considerable differences in immune system exists between human and rodents (41), especially in DCs populations (42). Therefore, it is critical to investigate whether ES L1 antigens could induce similar tolerogenic properties of human DCs as well. Here, we found for the first time that *T. spiralis* ES L1 antigens indeed possess the ability to establish stable tolerogenic human DCs *in vitro,* which could be potentially useful to modulate autoimmune diseases in humans.

### MATERIALS AND METHODS

The minimum information about tolerogenic antigen-presenting cells checklist was followed for the preparation of this manuscript (43).

### Ethics Statement

Animal experiments were performed according to institutional guidelines and were approved by the local Institutional Animal Care and Use Committee of the Institute for the Application of Nuclear Energy.

Samples of human peripheral blood were obtained from healthy volunteers after written informed consent in accordance with the Declaration of Helsinki and approval by the Ethical Board of the Institute for the Application of Nuclear Energy.

### Antigen Preparation

Adult male Wistar rats, aged 10–12 weeks, were obtained from Military Medical Academy (Belgrade, Serbia) and were housed under standard conditions in animal room with access to food and water *ad libitum*. The rats were used for maintaining of *T. spiralis* strain (ISS 161). Muscle larvae were recovered by digestion of the carcasses in pre-warmed gastric juice (44), and kept under controlled conditions (37°C, 5% CO2) in complete Dulbecco's modified Eagle medium (DMEM) (Sigma), for 18 h (45). ES L1 antigens were obtained by dialysis and concentration of the culture supernatants to at least 4.2 mg/ml in sterile phosphate buffered saline (PBS). Potential endotoxin contamination in ES L1 antigens was neutralized using SERVA Blue PrepProtein Endotoxin ExMicroKit (AMS Biotechnology, UK) according to the manufactures guidelines. Endotoxin levels in ES L1 preparation, in the highest concentration used in the experiments (200 µg/ml), were lower than 0.5 EU/ ml [the limit provided by the US Food and Drug Administration guidelines (46)], as detected by the Limulus Amoebocyte Lysate turbidimetric test. The quality of ES L1 products was checked by *Trichinella* ELISA test (INEP, Serbia). Namely, ES L1 antigens were adhered to microtiter plates and their immunoreactivity was analyzed using reference sera with pre-defined titer of anti-*T. spiralis* specific antibodies, as described previously (47).

### Cell Isolation and Culture

Peripheral blood mononuclear cells (PBMCs) from healthy blood donors were isolated by density gradient centrifugation on Lymphoprep gradient (Carl Roth). The monocytes were isolated by negative magnetic-activated cell sorting (MACS) from PBMCs, using the Monocyte Isolation kit (Myltenil Biotec), following their cultivation in CellGro DC medium (Cell Genix), supplemented with 100 ng/ml of human recombinant GM-CSF and 20 ng/ml of human recombinant IL-4 (both from R&D Systems) (complete DC growth medium). CD3<sup>+</sup> T cells were isolated from PBMCs by negative MACS, using the Pan T cell isolation kit (Miltenyi Biotec), and they were used as responders in allogeneic coculture experiments with monocyte-derived DCs. The purity of CD14<sup>+</sup> monocytes and CD3<sup>+</sup> T cell populations was usually higher than 90%, as evaluated by flow cytometry analysis (CyFlow Cube 6, Sysmex Partec GmbH, Görlitz, Germany) (data not shown).

To differentiate immature DCs, monocytes were cultivated in 24-well plate, 0.5 × 106 /well, in the complete DC growth medium for 5 days, refreshing the medium on Day 4. To assess the effect of ES L1 on DCs differentiation, the cells were cultivated during this period in the presence of 10, 50, or 200 µg/ml of ES L1 antigen, or without ES L1 (control DCs). The impact of ES L1 on the maturation of DCs was determined by adding ES L1 antigens to the culture of immature DCs on Day 4, for 48 h. To induce mature DCs, the cells were stimulated with LPS from *Escherichia coli* (500 ng/ml, Sigma-Aldrich) and human recombinant IFN-γ

(50 ng/ml, R&D Systems) on Day 5, for the next 24 h. In some experiments, LPS (500 ng/ml) or polyinosininc:polycytidylic acid [Poly (I:C), 10 µg/ml] were used instead of LPS/IFN-γ, to investigate the stability of phenotypical and functional characteristics of DCs acquired after the pulsing with ES L1. To determine the role of toll-like receptors (TLR)2 and TLR4 on the DCs status induced by ES L1, immature DCs were treated with blocking antibodies against TLR2 and/or TLR4 (10 µg/ml each, both from BioLegend), or isotype control antibody (anti-rat IgG, eBioscience), 1 h before the treatment with ES L1. The phenotype of DCs was checked using flow cytometry, whereas DC culture supernatants were used for the cytokines analyses. DCs used for functional assays were extensively washed to prevent the transfer of free ES L1 or stimuli to the cocultures.

### Flow Cytometry

Dendritic cells cultivated with different agents were stained after the culture with the following antibodies/reagents: immunoglobulin (Ig)G1a negative control—peridinin–chlorophyll– protein complex (PerCP), IgG1 negative control—phycoerythrin (PE), IgG1 negative control—fluorescein isothiocyanate (FITC), IgG1a negative control—PE cyanine (Cy)5, IgG1 negative control—allophycocyanin (APC), anti-CD83-FITC, anti-CD86- PE, anti CD40-APC, anti-CCR7-FITC, anti-TGF-β-PeCy5, anti-TLR4:PeCy5 (eBioscience), anti-CD1a-PE (Biolegend), anti-IL12p40/p70-PE, anti-indolamine 2,3 dioxygenase (IDO)- 1-APC, anti-Ig-like transcript (ILT)3-PE (R&D Systems), anti-IL-10-FITC (AbD Serotec), anti-CD14-FITC, and anti-HLA-DR-PerCP (Miltenyi). The viability of DCs after the culture with ES L1 antigens was determined by 7-aminoactinomycin D (7AAD; Invitrogen) staining. For T cells staining, the following antibodies were used: anti-forkhead box (Fox)P3-PE, anti-CD4- FITC, anti-CD4-PeCy5, anti-TGF-β-PE, anti-IL-4-PE, anti-IL-10-PE (eBioscience), anti-CD25-PeCy5 (BD Pharmigen), anti-IL-10-FITC (AbD Serotec), IL-17A-PerCP (Biolegend), anti-IFNγ-FITC (R&D Systems), and appropriate negative isotype controls, as indicated.

For surface labeling, the cells were washed once in PBS containing 2% FCS and 0.1% Na-azide and then incubated with primary Abs for 30–60 min at 4°C. Intracellular staining was conducted after the surface staining, by using the flow cytometry fixation and permeabilization kit I (R&D). Intracellular staining of IFN-γ, IL-4, IL-10, IL-17, and TGF-β in T cells, was carried out after the 4-h activation of T cells with phorbol-12-myristate-13-acetate (20 ng/ml, PMA) and ionomycin (500 ng/ml) in the presence of monensin (2 µM). For each analysis, more than 5,000 cells were gated according to their specific side-scatter (SSC)/ forward-scatter (FSC) properties, thereby avoiding the cells with low SSC/FSC properties (predominantly dead cells), as indicated. Signal overlap between the FL channels was compensated before each experiment using single labeled cells, and non-specific fluorescence was determined by using the appropriate isotype control antibodies and fluorescence minus one controls.

### T Cell Proliferation Assay

In mixed leukocyte reaction assays, DCs treated with the different stimuli, as well as non-treated DCs, were cultured with MACS-purified allogenic T cells in complete RPMI 1640 medium containing 10% FCS (Capricorn Scientific), l-glutamine, 2 mercaptoethanol (50 µM, Sigma), and antibiotics (penicillin, streptomycin, gentamicin, 1% each, ICN Galenika). T cells (1 × 105 /well of 96-well plate) were cocultivated with different number of DCs (0.5 × 104 , 0.25 × 104 , 0.125 × 104 ) for 5 days in 96-well round-bottom plates. The blank controls were T cells cultivated separately. To measure the level of T cell proliferation, the cells were pulsed for the last 18 h of coculture with 3H-thymidine (1 μCi/well, Amersham). The radioactivity was measured by β-scintillation counting (LKB-1219 Rackbeta, Finland).

## Activation of T Cells by Allogenic DCs

The capacity of ES L1-stimulated DCs to instruct the polarization of T cells was analyzed in allogenic stimulation assay. DCs (0.5 × 104 /well, 96-well round-bottom plate) were cultivated with naive allogenic T lymphocytes in 1:20 ratio, for 6 days. To detect cytokines in the supernatants of allogenic stimulation assay, DC/T cell cocultures were treated with PMA (20 ng/ml) and Ionomicin (500 ng/ml) for the last 4 h of coculture, followed by the harvesting of cell-free supernatants. The levels of cytokines produced in DCs cultures (IL-10, IL-12p70, TGF-β), as well as in DC/T cell cocultures (IL-4, IL-10, IFN-γ, IL-17, TGF-β), were measured in cell-free supernatants by sandwich ELISA Kits (R&D). Additionally, T cells were primed with DCs (2 × 103 / well of 96-well round-bottom plate) at a 1:50 DC/T cell ratio for 3 days and then expanded with IL-2 (2 ng/ml, R&D) for two more days (T primed cells, Tpr). In some experiments 1-methylthryptophane (1-MT, 0.5 mM, Sigma), as an IDO-1 inhibitor (48), was added in the priming cocultures to assess the role of IDO-1 in the induction of Tregs. The Tregs were identified by flow cytometry based on their expression of CD4, CD25, FoxP3, IL-10, and TGF-β by flow cytometry.

### Suppressor Assay

T cells primed with DCs (Tpr.) were extensively washed and then cocultivated with allogeneic PBMCs responder cells. PBMCs, 1 × 105 cells/well, were cocultured in 1:2 and 1:4 cell ratios with Tpr. (0.5 × 105 and 0.25 × 105 cells/well, respectively), in 96-well round-bottom plates. In order to assess the proliferation of responder cells, PBMCs were pre-stained with 1 µM CFSE (Invitrogen) for 20 min, at 37°C, at density of 1 × 107 cells/ml. The proliferation of cells was stimulated with 8 µg/ml phytohemaglutinine (PHA), and after 5 days of coculture, the cells were harvested, stained with 20 µg/ml propidium iodide to exclude dead cells from the analysis, and the proliferation of responder T cells was measured by flow cytometry. The Proliferation Index (PI), i.e., the average number of cells derived from the initial cell, was calculated using proliferation fit statistics in FCS Express 4 (*De Novo* Software, Glendale, CA, USA).

### Investigation of Pattern Recognition Receptors (PRRs) Activation by ES L1

HEK-Blue™ reporter cell lines expressing either TLR2, TLR3, TLR4, TLR5, TLR7, nucleotide-binding oligomerization domainlike receptors (NOD)1, or NOD2 (InvivoGen, San Diego, CA, USA) (49) were used to evaluate the potential activation of these PRRs by ES L1. The stimulation of HEK-Blue™ cells by PRR agonists triggers the intracellular signaling events leading to the activation of nuclear factor (NF)-κB and activating protein (AP)-1 and the production of secreted alkaline phosphatase (SEAP). Therefore, the SEAP activities in supernatant correspond to the activation of the specific PRRs.

HEK-Blue™ hTLR2 (hkb-htlr2), HEK-Blue™ hTLR3 (hkbhtlr3), HEK-Blue™ hTLR4 (hkb-htlr4), HEK-Blue™ hTLR5 (hkbhtlr5), HEK-Blue™ hTLR7 (hkb-htlr7), HEK-Blue™ hNOD1 (hkb-hnod1), and HEK-Blue™ hNOD2 (hkb-hnod2) cells were cultured in Dulbecco's Modified Eagle's culture medium (DMEM; Gibco, Thermo Fisher Scientific), supplemented with 100 U/ml penicillin, 100 µg/ml streptomycin, 0.3 mg/ml l-glutamine (Gibco; Pen–Strep–Glu 100×; Thermo Fisher Scientific, Rockford, IL, USA) and 10% (v/v) heat-inactivated HyClone FBS (30 min at 56°C) (Thermo Fisher Scientific) at 37°C in 5% CO2/95% air. Cell lines (2.5 × 104 /well, 96-well flat-bottom plates) were cultivated in 150 µl of culture medium. Cells were treated with 50 µl of various concentrations of ES L1 antigens (50, 5, and 0.05 µg/ml) or PRRs agonists: Pam3CSK4 (10 ng/ml) as TLR2 ligand, Poly (I:C) HMW (1 µg/ml) as TLR3 ligand, LPS (1 ng/ml) *E. coli K12* as TLR4 ligand, FLA-ST (50 ng/ml) as TLR5 ligand, Imiquimod (10 µg/ml) as TLR7 ligand, iE-DAP (25 µg/ml) as NOD1 ligand, and MDP (10 µg/ml) as NOD2 ligand (all from InvivoGen). After 22–24 h incubation, supernatants were collected for determination of SEAP activity. QUANTI-Blue™ medium (180 µl) substrate and cell culture supernatant (20 µl) were added per well on a flatbottom 96-well ELISA plate. After 4 and 24 h of incubation at 37°C in 5% CO2, the levels of SEAP activity were analyzed by measuring the absorbance at 649 nm on a BioTek EL808 microplate reader and Gen5 Data Analysis Software (BioTek, VT, USA).

### Statistical Analyses

Data are presented either from representative experiments or as mean ± SD of at least three experiments carried out with different DCs donors. Statistical analysis of the data was conducted using repeated measures ANOVA, followed by a Tukey's posttest in PRISM5 (Graphpad software), and for *p* < 0.05, the differences were considered significant statistically.

## RESULTS

### ES L1 Antigens Impair the Maturation of DCs, without Affecting Their Differentiation

Human monocytes cultured in complete DC growth medium *in vitro* differentiate into immature DCs, by upregulating CD1a and diminishing CD14 expression (50). To assess the influence of ES L1 antigens on differentiation of DCs, monocytes were differentiated with GM-CSF/IL-4 in the presence or absence of different doses of ES L1 antigens (10, 50, or 200 µg/ml), applied on Day 0. To exclude the possibility that the immunomodulatory effects of ES L1 antigens were due to their cytotoxicity, we first analyzed the percentage of dead (7AAD<sup>+</sup>) cells after 4 days of cultivation with ES L1. It was observed that the percentage of 7AAD<sup>+</sup> DCs cultivated with 10 or 50 µg/ml was similar to the values obtained in control DCs cultures. In contrast, 200 µg/ml ES L1 increased the percentage of 7AAD<sup>+</sup> cells up to 27% (**Figures 1A,B**). However, none of the applied ES L1 doses induced more than 30% of dead DCs which, according to ISO 10933-5 (51), could be interpreted as the lack of cytotoxicity. The phenotypic analysis of DCs suggested that ES L1 did not interfere with DCs differentiation irrespective of the dosage applied, since the percentage of CD1a and CD14 on ES L1-treated DCs was similar to that on un-treated control DCs (**Figures 1C,D**).

FIGURE 1 | Effects ES L1 antigens on differentiation of dendritic cells (DCs). (A–E) Immature DCs were generated from monocytes in GM-CSF/interleukin-4 supplemented medium in the presence of ES L1 (10, 50, and 200 µg/ml) antigens, or their absence (control), during 5 days. (A) Representative histograms from the analysis of 7AAD+ (dead) cells after the culture are shown and, (B) the summarized results from three different donors are shown as mean% ± SD. (C) Representative plots for CD1a and CD14 expression on DCs after 5 days of culture with ES L1 are shown, and (D) the summarized results from three different donors are presented as mean% ± SD. (E) DCs differentiated in the presence or absence of ES L1, were stimulated with LPS/IFN-γ on Day 5, and the expression of CD83, CD86, CD40, and HLA-DR was analyzed by flow cytometry after 24 h. The results collected with three different DC donors are shown as mean ± SD (see also Figure S1 in Supplementary Material, showing a representative experiment). \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.005 compared to corresponding control DCs, or as indicated (one-way ANOVA with Tukey's posttest).

To evaluate whether the differentiation of DCs with different ES L1 doses affect their subsequent maturation, DCs were stimulated with LPS/IFN-γ on Day 5 of culture, for the next 24 h. It was observed that differentiation with higher doses of ES L1 (50 or 200 µg/ml) impaired severely the LPS/IFN-γ-induced upregulation of CD83, CD86, CD40, and HLA-DR, whereas the cultivation with 10 µg/ml ES L1 impaired only CD40 upregulation (**Figure 1E**; Figure S1 in Supplementary Material). Although 200 μg/ml of ES L1 had the strongest effect, it was not significantly different from the effect of 50 µg/ml ES L1.

### Human DCs Acquire Stable Semi-Mature Phenotype upon Stimulation with ES L1 Antigens

To assess whether ES L1 antigens exhibit similar effects on DCs maturation when added at latter periods of cultivation, the cells were pulsed with ES L1 antigens on Day 4 of cultures and then stimulated with LPS/IFN-γ on Day 5 for the next 24 h, or left un-stimulated. Considering the negligible effects of 10 µg/ml of ES L1 on LPS/IFN-γ-induced DCs maturation and significant immunomodulatory effect of 50 µg/ml of ES L1 without an increased cytotoxicity, further experiments were carried out with 50 µg/ml of ES L1 to monitor the functional significance of ES L1 on DCs maturation. The analysis of maturation markers on DCs, 48 h after the treatment with ES L1, showed that the expression of HLA-DR, CCR7, CD83, and CD86 were somewhat upregulated, but not significantly, compared to the control immature DCs (**Figure 2A**; Figure S2 in Supplementary Material). ES L1 treatment significantly upregulated only the expression of CD40 (**Figure 2A**; Figure S2 in Supplementary Material), all of which suggested that ES L1 induce a semi-mature phenotype of DCs. Similar results were obtained when the treatment with ES L1 lasted only 24 h (data not shown). In contrast to ES L1-treated DCs, the expression of all surface markers after the 24h-stimulation with LPS/IFN-γ was significantly upregulated, as expected for type 1 inflammatory DCs induced by this cocktail (52). However, DCs treated with 50 µg/ml of ES L1 before LPS/IFN-γ stimulation displayed an impaired upregulation of maturation markers (**Figure 2A**), confirming that the effect of ES L1 was limited to the maturation stage of DCs.

To investigate the effect of ES L1 antigens on DCs' cytokines production, the percentages of IL-10, IL-12, and TGF-β positive DCs were determined by intracellular staining (**Figures 2B,C**), and the cytokine levels were analyzed in DCs culture supernatants (**Figure 2D**). Both analyses showed that the levels of IL-12 in ES L1-treated DCs were unchanged compared to control immature DCs, i.e., ES L1 failed to induce the production of IL-12 by DCs. On the other hand, ES L1 significantly increased the percentage of IL-10<sup>+</sup> DCs and the levels of IL-10 in DCs culture supernatants (**Figures 2C,D**). The percentage of TGF-β+ DCs also increased significantly after the ES L1 treatment, compared to control DCs, but this was not followed by increased levels of TGF-β in the culture supernatants (**Figures 2B–D**).

To investigate the functional stability of ES L1-treated DCs, the cells were challenged with different maturation stimuli 24 h after the treatment with ES L1 antigens. The stimulation of DCs with LPS/IFN-γ for 24 h, expectedly, induced a strong upregulation of surface maturation markers (**Figure 2A**; Figure S2 in Supplementary Material), IL-12 expression, and its production by DCs. However no such upregulation occurred if DCs were treated with ES L1 before the challenge with LPS/IFN-γ (**Figures 2B–D**). The expression of TGF-β in ES L1-treated DCs, after the challenge with LPS/IFN-γ, remained significantly higher compared to control LPS/IFN-γ-matured DCs. Curiously enough, the levels of TGF-β in DCs culture supernatants did not correspond to the percentages of DCs expressing TGF-β intracellularly, which could be a consequence of the autocrine uptake of the produced TGF-β (53). On the other hand, the expression of IL-10 by ES L1-treated DCs matured with LPS/ IFN-γ and its production levels were not significantly different from the values obtained with control LPS/IFN-γ-matured DCs (**Figures 2B–D**).

Considering the inhibitory effects of IFN-γ on IL-10 production by DCs (54), we wondered whether the use of LPS alone, as a maturation stimulus, would exhibit different effects on the phenotype and cytokines production by ES L1-treated DCs. ES L1-treated DCs exhibited similar resistance to LPS-induced phenotypic maturation and IL-12 expression, as when LPS/ IFN-γ was used as a stimulus, while the expression of IL-10 in ES L1-treated LPS-matured DCs was significantly higher compared to control LPS-matured DCs (Figure S3 in Supplementary Material). The mitigated effect of LPS/IFN-γ or LPS alone on the phenotype and cytokine production by ES L1-treated DCs was not a consequence of ES L1-induced downregulation of TLR4 on DCs, since its expression was not changed significantly 48 h after the ES L1 treatment (Figure S4 in Supplementary Material).

Since LPS and LPS/IFN-γ stimulation includes a TLR4 dependent induction of DCs maturation (55), we also tested whether ES L1-treated DCs display similar maturation resistance to other TLR stimuli, such as the stimulation *via* TLR3 with its agonist, Poly (I:C). The data showed that ES L1 impaired significantly the upregulation of HLA-DR, CD40, and IL-12 by DCs, but not CD83 and CD86, upon the Poly (I:C) stimulation. Additionally, ES L1-treated Poly (I:C)-matured DCs produced significantly higher levels of IL-10 compared to control Poly (I:C)-matured DCs (Figure S5 in Supplementary Material). These data suggested that the treatment of DCs with ES L1 impeded full maturation of DCs by different maturation stimuli, showing the stability of ES L1-treated DCs.

### DCs Treated with ES L1 Acquire Limited Allo-Stimulatory Capacity and Provoke Th2 and Regulatory Responses

To assess the capacity of human DCs to influence the proliferation of T cells after the treatment with ES L1, coculture of DCs and allogeneic MACS-purified CD3<sup>+</sup> T cells was performed, and the T cell proliferation was determined based on 3H-thymidine incorporation level. ES L1-treated DCs had a weaker capacity to stimulate T cell proliferation compared to both immature and LPS/IFN-γ-matured DCs (**Figure 3A**). The limited T cell proliferation capacity was preserved even when ES L1-treated DCs were challenged with LPS/IFN-γ.

FIGURE 2 | The maturation capacity of dendritic cells (DCs) treated with ES L1 antigens. (A–D) Immature DCs were treated with ES L1 antigens (50 µg/ml) on Day 4 of culture for 24 h and then additionally activated or not with LPS/interferon (IFN)-γ for the next 24 h, followed by flow cytometry analysis. (A) The percentages of CD83, CD86, HLA-DR, CCR7, and CD40 expression by DCs are shown as mean ± SD from four different experiments (see also Figure S2 in Supplementary Material for a representative experiment). (B) Representative analysis of interleukin (IL)-10, IL12p40/p70, and transforming growth factor (TGF)-β expression within DCs are shown, and (C) the summarized results are presented as fold change of cytokines expression as mean ± SD from seven different experiments. (D) The levels of IL-10, IL-12p70, and TGF-β (picograms/milliliter) in DCs culture supernatants were measured by ELISA test. \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.005 compared to control, or as indicated (one-way ANOVA with Tukey's posttest).

FIGURE 3 | Allo-stimulatory and T helper polarizing capacity of ES L1-pulsed dendritic cells (DCs). (A–D) DCs treated with ES L1 antigens (50 µg/ml) and/or LPS/ interferon (IFN)-γ were washed thoroughly and then cocultured with magnetic-activated cell sorting-purified allogenic T cells (Tly) (1 × 105 /well) for 6 days in two DC:T cell ratios (1:10 and 1:20). (A) The proliferation in cocultures was measured by 3H-thymidin incorporation assay, and Tly cultivated alone were used as a blank control. (B) The concentration of indicated cytokines were determined in the supernatants of PMA/ionophore-treated DC:T cell cocultures at 1:20 cell-to-cell ratio, respectively, by specific ELISA tests. (C) The percentage of cytokines expression measured intracellularly by flow cytometry, within the T cells cocultivated with DCs as in (B) and treated with PMA/Ionophore/monensin for the last 4 h, are shown as mean% ± SD of four experiments with different DCs donors. \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001 compared to control, or as indicated by line (one-way ANOVA with Tukey's posttest). (D) The analysis of the intracellular cytokines in T cells was carried out after the surface staining of CD4, as indicated on representative dot plots collected from two experiments.

The Th polarization capacity of ES L1-treated DCs was assessed by measuring the levels of IFN-γ, IL-17, IL-4, IL-10, and TGF-β in DCs/T cell cocultures (**Figure 3B**) and by analyzing the expression of these cytokines within gated CD4<sup>+</sup> and CD4<sup>−</sup> (more than 98% CD8<sup>+</sup>) T cell population (**Figures 3C,D**) after the stimulation with PMA/ionophore. The results suggested that ES L1 potentiated DC-mediated induction of Th2 and regulatory type of immune response, as judged by the increased IL-4, IL-10, and TGF-β expression by T cells and their production in DC/T cell cocultures. The increased capacity of ES L1-treated DCs for the induction of regulatory cytokines in T cells was detected in both CD4<sup>+</sup> and CD8<sup>+</sup> T cell population (**Figure 3D**). After the challenge with LPS/IFN-γ, the capacity of ES L1-treated DCs to induce TGF-β-producing T cells increased even more. Their capacity to induce Th2 cells remained significantly higher than the capacity of control LPS/IFN-γ-matured DCs, whereas the capacity to induce IL-10-producing T cells was similar to the capacity of control mature DCs. Additionally, ES L1-treated DCs displayed an impaired potential to induce Th17 cells in the absence of stimuli, and an impaired Th1/Th17 polarization capacity after their challenge with a strong Th1/Th17 polarizing cocktail, LPS/IFN-γ, which was confirmed by the analyses of both soluble products and intracellular cytokines expression in T cells (**Figures 3B–D**). Taken together, ES L1-treated DCs displayed a clear shift in polarizing capacity, away from Th1/Th17, toward increased regulatory and Th2 type of immune response.

### DCs Pulsed with ES L1 Antigens Have Tolerogenic Properties

Since DCs upon the treatment with ES L1 antigens acquire semi-mature phenotype, limited allo-stimulatory capacity, and shift the immune response toward Th2 and regulatory type, we presumed that ES L1-treated DCs possess tolerogenic properties. In addition to reduced pro-inflammatory phenotype and functions, the tolerogenic DCs express tolerogenic markers, such as IDO-1 and ILT-3, and display an increased capacity to induce Tregs (56, 57). Therefore, we first analyzed the expression of IDO-1 and ILT-3 by DCs under the influence of ES L1 antigens, in the presence or absence of additional stimulation with LPS/ IFN-γ (**Figures 4A,B**). The results showed that ES L1 induce the expression of both markers on DCs, but not significantly. However, the expression of IDO-1 was significantly increased after the challenge of ES L1-treated DCs with LPS/IFN-γ. The expression pattern of ILT-3 showed similar trend, but the differences were not significant statistically.

The percentage of Tregs in coculture with DCs was determined based on their high expression of CD25 and FoxP3 within CD4<sup>+</sup> T cells, to distinguish them from transiently activated CD25<sup>+</sup>Foxp3<sup>+</sup> Th cells. ES L1-treated DCs induced a significant increase in the percentage of CD4<sup>+</sup>CD25hiFoxp3hi compared to control cells (twofold), and retained that ability even after LPS/IFN-γ challenge (**Figures 4C,D**). The observed expansion of Tregs was IDO-1 dependent, as it was demonstrated that the addition of 1-MT in the cocultures decreased significantly the percentages of CD4<sup>+</sup>CD25hiFoxp3hi cells (**Figures 4C,D**). All CD4<sup>+</sup>CD25hi T cells were positive for TGF-β, but it was found that ES L1-treated DCs significantly increased the expression of TGF-β within CD4<sup>+</sup>CD25hi T cells, compared to the control DCs, and this ability of ES L1-treated DCs was even more pronounced after their challenge with LPS/IFN-γ (**Figures 4E,F**). Moreover, the expression of IL-10 within CD4<sup>+</sup>CD25hi T cell population was fourfold higher upon priming with ES L1-treated DCs than after the priming with control DCs. However, such an upregulation of IL-10 within CD4+CD25hi T cells was not observed when DCs were stimulated additionally with LPS/IFN-γ (**Figures 4E,F**).

Taken together, the above results suggested that ES L1 indeed potentiated the tolerogenic phenotype and functions of DCs, which induced Tregs in IDO-1-dependent manner, especially after the challenge with LPS/IFN-γ.

### T Cells Primed by ES L1-Treated DCs Exhibit Suppressive Activity

To investigate whether an increased percentage of CD4<sup>+</sup>CD25hiFoxp3hi Treg have any functional significance, the T cell population primed with DCs (Tpr) was cocultivated with CFSE-labeled PBMCs (responders) in the presence of PHA. A flow cytometry analysis of the responder cells' proliferation clearly suggested that T cells primed with ES L1-treated DCs exhibited a stronger capacity to inhibit the proliferation of responder cells in both 1:2 and 1:4 Tpr:responder cell ratios (**Figures 5A,B**). Similar results were obtained with T cells primed with LPS/ IFN-γ-stimulated ES L1-treated DCs, as demonstrated by their increased potential to suppress the responder cells proliferation in 1:2 cell-to-cell ratio. The results suggested that ES L1-treated DCs retained the ability to induce suppressive T cell populations even in inflammatory environment. Since the percentage of Tregs increased significantly after the priming with ES L1-treated DCs, the observed suppressive activity of the primed T cell population may be ascribed to the effects of Tregs as well.

### ES L1 Antigen Activates TLR2 and TLR4

The data about the receptors that recognize *T. spiralis* antigens are scarce. Having in mind that PRRs are a key element of the innate immune system and have important role in detection of pathogens and subsequent activation of DCs, the interaction between different PRRs (TLRs and NODs) with ES L1 antigens was investigated. The study was performed on HEK-Blue™ reporter cell lines expressing the individual TLRs (TLR2, -3, -4, -5, -7), NOD1, or NOD2 receptors. The activation of PRRs was indicated by SEAP activation in culture supernatants of stimulated cells. All used HEK-Blue™ cell lines were treated with ES L1 antigens, while the cells pulsed with corresponding TLR or NOD agonists were used as positive controls, and the cells cultivated only in medium were used as negative control (**Figure 6**).

Stimulation of HEK-hTLR2 cells with ES L1 induced an increased level of SEAP activity after 4 and 24 h (**Figures 6A,B** respectively) compared to negative control, indicating that ES L1 activates TLR2-mediated signaling events. ES L1 treatment of HEK-hTLR4 cells resulted in a significantly increased level of alkaline phosphatase activity after 4 h (**Figure 6C**) compared to control, and after 24 h (**Figure 6D**), the level of enzyme activity was similar to the one observed with TLR4 agonist (LPS, 1 ng/ml), suggesting that ES L1 also activated TLR4. HEK-hTLR3, HEKhTLR5, HEK-hTLR7, HEK-hNOD1, and HEK-hNOD2 cell lines cultivated in the presence of ES L1 antigens, after 4 h (data not shown) and after 24 h, showed no statistically significant change in SEAP activity, compared to control cells, unlike the cells treated with specific PRR ligands as described (**Figures 6E–I**). The obtained results suggested strongly that ES L1 antigens engage TLR2 and TLR4 and activate NF-κB and AP-1-mediated signaling events by these receptors, while TLR3, TLR5, TLR7, NOD1, and NOD2 receptors do not participate in ES L1-mediated effects.

FIGURE 4 | Tolerogenic properties and functions of ES L1-treated dendritic cells (DCs). (A) DCs treated with ES L1 antigens (50 µg/ml) and/or LPS/interferon (IFN)-γ, as described, were analyzed for IDO-1 and immunoglobulin-like transcript (ILT)-3 expression, as indicated on representative histograms. (B) The summarized results on IDO-1 and ILT-3 expression on three different DC donors is shown as mean% ± SD. (C) DCs treated as in (A) were washed thoroughly and then cocultivated with magnetic-activated cell sorting-purified allogenic T cells (1 × 105 /well) (DC:T cell ratio 1:50), for 3 days and then re-stimulated with interleukin (IL)-2 (2 ng/ml) for another 3 days, all in the presence or absence of 1-MT (0.5 mM). Representative analysis of CD25hiFoxP3hi cells within CD4+ T cells from one experiment is shown, and (D) the summarized results are shown as the mean fold change in regulatory T cells % ± SD from four different experiments. (E) Representative analysis of transforming growth factor (TGF)-β and IL-10 within CD4+CD25hi T cell population is shown, and (F) the summarized results collected from four different experiments are shown as mean fold change in mean fluorescence intensity (MFI) (TGF-β) or % (IL-10) within CD4+CD25hi T cell population ± SD. \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.005 compared as indicated by line (one-way ANOVA with Tukey's posttest).

(one-way ANOVA with Tukey's posttest).

## TLR2 and TLR4 Are Involved in ES L1- Mediated Induction of Semi-Mature DCs

Considering that ES L1 antigens activate TLR2 and TLR4, the next step was to investigate the relevance of this interaction for the development of semi-mature DCs by ES L1 antigens. For this purpose, TLR2 and TLR4 receptors were blocked individually or simultaneously using specific monoclonal blocking antibodies, prior to DCs treatment with ES L1 antigens. In the presence of the specific blocking antibodies, the expression of DCs surface markers upon ES L1 treatment was significantly changed compared to control DCs treated with ES L1 antigens in the presence of isotype control antibody. Namely, the expression of CD83 was significantly lower, while CD86, HLA-DR, and CD40 expression were significantly higher compared to control ES L1-treated DCs (**Figures 7A,B**).

Additionally, the differences in cytokines production by DCs were analyzed in the presence of specific blocking or isotype control antibodies upon the ES L1 treatment (**Figure 7C**). In case when TLRs were specifically blocked either individually or simultaneously, the anti-inflammatory properties of ES L1-treated DCs were diminished, i.e., the concentration levels of IL-12p70 were significantly higher and the levels of IL-10 and TGF-β were significantly lower compared to those observed in DCs treated with ES L1 and isotype control antibody. The treatment of DCs with isotype control antibody, as well as the blocking of receptors without subsequent treatment with ES L1 antigens, resulted in the same DCs maturation profile as observed with non-stimulated DCs (data not shown). The results obtained by individual blocking of TLR2 and TLR4 indicated that both receptors are involved in ES L1-driven DC phenotype. Simultaneous blocking of two receptors gave no phenotypic or functional changes compared to individual blocking, suggesting the absence of synergistic effect.

### DISCUSSION

A number of studies have dealt with *in vitro* generation of DCs from monocytes that can be manipulated to acquire tolerogenic properties (35, 56, 58–60). These cells were shown to have an immature or semi-mature phenotype, low production of inflammatory and increased production of anti-inflammatory cytokines, the ability to present antigens in a tolerogenic form, and consequently, an increased capacity to induce Tregs. Tolerogenic DCs are considered as a promising tool for development of cell-based therapy applicable in the treatment of autoimmune diseases, chronic inflammation, and transplantation

FIGURE 6 | Interaction of ES L1 antigen with pattern recognition receptors (PRRs) on HEK-Blue™ cell lines. HEK-Blue™ cell lines transfected with a single specific human PRR (TLR2, 3, 4, 5, 7, NOD1, and 2) were treated with ES L1 or PRR agonists for 24 h, followed by the analyses of secreted alkaline phosphatase (SEAP) levels (OD 649 nm) released in culture medium at two time points (4 and 24 h after the substrate addition). Pam3CSK4 (10 ng/ml), ES L1 (5 µg/ml)—incubation period 4 (A) and 24 h (B); LPS (1 ng/ml) *Escherichia coli K12*, ES L1 (5 µg/ml)—incubation period 4 (C) and 24 h (D); Poly (I:C) HMV (1 µg/ml), ES L1 (5 µg/ml) incubation period 24 h (E); FLA-ST (50 ng/ml), ES L1 (5 µg/ml)—incubation period 24 h (F); imiquimod (10 µg/ml), ES L1 (5 µg/ml)—incubation period 24 h (G); iE-DAP (25 µg/ml) ES L1 (50 µg/ml)—incubation period 24 h (H); MDP (10 µg/ml), ES L1 (50 µg/ml)—incubation period 24 h (I). Results are shown as mean ± SD from three different experiments \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001 compared with control (medium) (one-way ANOVA with Tukey's posttest).

therapy (61). Our previous results obtained on animal model system demonstrated the capacity of *T. spiralis* products to induce the development of DCs with tolerogenic properties (40), which successfully ameliorated autoimmune disease in animal models

experiments. \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.005 as indicated by line (one-way ANOVA with Tukey's posttest).

(16). However, further development of cell therapies based on ES L1 antigens require comprehensive studies on human DCs model system, which have not been conducted so far. Here, we showed for the first time that *T. spiralis* ES L1 antigens represent

by ELISA test, and transforming growth factor (TGF)-β expression (% fold change) determined by flow cytometry, are shown as mean ± SD of three different

a promising new tool for the generation of human tolerogenic DCs *in vitro*. In contrast to other protocols for generation of DCs, such as those which include vitamin D3 (26, 27) or IL-10 (31), DCs cultivated with ES L1 antigens from Day 0, even with high doses, displayed an unaltered differentiation pattern, i.e., complete downregulation of CD14 and upregulation of CD1a in the presence of GM-CSF and IL-4. These results suggested that the effects of ES L1 antigens are restricted to differentiated DCs population, but not on their monocyte precursors. Although the physiological significance of this finding is still unclear and require independent investigation, such property of ES L1 could have evolved as a result of tight regulation of the host immune response by the parasite. This kind of modulation is aimed specifically to prevent the host immune response against the parasite itself, but it also mitigates the unwanted immune responses like those to autoantigens and allergens, consequently increasing the chances of host survival (37). Upon contact with ES L1 antigens, human monocyte-derived DCs acquired semi-mature phenotype, characterized by: low expression of HLA-DR, co-stimulatory molecules and CCR7; moderate expression of CD40; a clear anti-inflammatory cytokine profile, i.e., low production of IL-12 along with the enhanced production anti-inflammatory/regulatory cytokines, IL-10 and TGF-β, even after the challenge with pro-inflammatory stimuli. In line with this, an increased ratio of IL-10/IL-12 in DCs was found to contribute significantly to their tolerogenic functions and increased capacity to expand Tregs and downregulate differentiation of Th1 and Th17 cells (50, 61). CCR7 expression by IL-10-producing DCs was shown critical for their migration to lymphoid organs and polarization of naïve T cells into Tregs (62–64). Somewhat increased expression of CCR7 by ES L1-treated DCs and their tolerogenic potential are in line with these findings. High production of IL-10 and TGF-β could be responsible for the decreased expression of co-stimulatory molecules and blocking of IL-12 production by DCs, as suggested by other findings (59). Also, synergistic effects of IL-10 and TGF-β were shown to result in enhanced tolerogenic properties of DCs and their higher potential to induce IL-10-producing Tregs (65), which is in line with the properties of ES L1-treated DCs observed in this study.

The stability of tolerogenic DCs and preservation of antiinflammatory phenotype in inflammatory surrounding is the most important issue when creating cell-based therapies for autoimmune diseases (61). TLRs were shown to be critically involved in the recognition of damage-associated molecular patterns (DAMPs), driving the unwanted inflammatory response in autoimmune diseases (66). In addition to TLR3, which was shown to be involved in the recognition of self-RNAs released from necrotic synovial fluid cells in rheumatoid arthritis patients (67), TLR4 has been implicated in the recognition of various DAMPs in different autoimmune processes (68). In line with this, ES L1-treated DCs displayed the resistance to the maturation induced by both TLR3 (Poly I:C) and TLR4 (LPS) agonist in particular. In contrast to Poly (I:C) and LPS, known to induce predominantly Th1 (56) and a mixed Th1/Th2 response (69), respectively, LPS/IFN-γ cocktail is known to induce inflammatory type 1 DCs (52), which are able to induce both Th1- and Th17-mediated immune response (50). Both, Th1 and Th17 cells were shown to be critically involved in the pathogenesis of chronic inflammation and autoimmunity (70). DCs maturated in the presence of ES L1 retained their tolerogenic properties regardless of activation with LPS/IFN-γ since, no noticeable change in the expression of surface markers and cytokine profile was observed. ES L1 antigens affected the maturation and cytokines production of DCs in a way that resembles tolerogenic DCs obtained under the influence of some immunosuppressive drugs (27, 28, 60) or immunosuppressive cytokines (65), but presumably *via* different mechanisms.

ES L1-treated DCs, even after the challenge with LPS/IFN-γ, exhibited a reduced capacity to induce T cell proliferation in mixed leukocyte reactions, which is in accordance with their observed semi-mature phenotype and tolerogenic functions. The experiments were performed using cells from healthy individuals. Due to HLA-restriction, the allogeneic coculture system applied in our study includes the response of a limited naïve T cell population that are able to respond to allogeneic MHC complexed with self-peptides (71). This model system may better reflect the potential response to ES L1-treated DCs *in vivo,* compared to models with a polyclonal stimulation of T cells, especially since recent findings suggested that *T. spiralis*-secreted components are structurally related to some human cell components (39), so only a limited repertoire of naïve T cells may be able to respond. Moreover, these DCs impaired inflammatory Th1 and Th17 cell responses manifested as both reduced production of IFN-γ and IL-17 cytokines and the percentage of Th1 and Th17 cells, respectively. Lower Th1 polarizing capacity of ES L1-treated LPS/IFN-γ-matured DCs might be a consequence of their low capacity to produce IL-12, a known Th1-polarizing factor (72). The fact that ES L1-treated DCs enhanced TGF-β production by T cells may also explain diminished production of Th1 polarization capacity and reduced allogenic proliferation, since TGF-β was shown to be critically involved in both processes (73). Although it is known that TGF-β contributes to regulatory as well as to Th17 type of immune response, there is evidence that low levels of TGF-β promote Th17 response, while increased TGF-β influences the elevated Foxp3 expression, hence promoting the expansion of Tregs (74). Since IFN-γ and IL-17 contribute in the genesis of autoimmune diseases (75, 76), the capacity of ES L1-induced tolerogenic DCs to reduce the production of these cytokines could favor the potential applicability of these DCs in the treatment of autoimmune diseases. The observations, that secretory products from different parasitic worms heavily skew the immune response toward Th2 type *via* DCs while inhibiting Th1 and Th17 responses (34), are in line with the finding that ES L1-primed DCs induced Th2 type response, as indicated by significantly expanded Th2 cells and elevated IL-4 production. In addition to lowering Th1 response, an increased Th2 polarization could also be related to moderately increased expression of CD40 on DCs upon stimulation with ES L1, as it was shown that CD40 is critically involved in the induction of Th2 cells by DCs, especially during helminths infection (77).

Besides the suppression of inflammatory immune response and the enhancement of Th2 type of response, ES L1-treated DCs demonstrated the ability to induce the expansion of CD4<sup>+</sup>CD25hiFoxp3hi T regulatory cells. This expansion was reduced upon addition of 1-MT, an IDO-1 inhibitor (48) during the coculture, suggesting that Tregs inducing capacity of ES L1-treated DCs is IDO-1 dependent. IDO-1-mediated actions on the induction of Tregs include both deprivation of tryptophan and kynurenine-dependent induction of Tregs *via* aryl hydrocarbon receptor (78). An increased expression of IDO-1 in DCs additionally treated with LPS/IFN-γ could be explained by the fact that IFN-γ is a strong inducer of IDO-1 (79). The results from these experiments also suggested that IDO-1 is more involved in the induction of Tregs by ES L1-treated DCs matured with LPS/IFN-γ than DCs treated with ES L1 only, as a stronger inhibition of Treg induction was obtained with the former. Therefore, DCs treated with ES L1 could have utilized additional mechanisms besides IDO-1, such as IL-10- or ILT-3-mediated induction of Tregs, both of which were shown to be involved in the induction of Tregs by tolerogenic DCs (61). Tregs are involved in the suppression of effector T-cell activity and maintenance of immunologic self-tolerance, as underlying processes in the modulation of autoimmune diseases (80). The finding that Treg-inducing capacity of ES L1-treated DCs is retained, even after the challenge with strong pro-inflammatory stimuli, is in agreement with previous findings on tolerogenic DCs primed by immunomodulatory molecules or immunosuppressive drugs (27, 81). Tregs induced by ES L1-treated DCs upregulated the production of IL-10 which may be the consequence of high IL-10/IL-12 ratio observed in ES L1-treated DCs that induced the expansion of Tregs, as previous findings on the mechanisms of Tregs induction suggest (61, 82). The same population of Tregs showed an elevated expression of TGF-β, probably due to the increased production of IL-10 and TGF-β by DCs (61). Although both TGF-β and IL-10 exhibit immunosuppressive functions, they were also shown to negatively regulate each other (83, 84). These findings could partially explain why Tregs induced by ES L1-treated DCs expressed predominantly IL-10, whereas Tregs induced by LPS/IFN-γ-matured ES L1-treated DC produced predominantly TGF-β. It is also possible that different mechanisms of tolerogenic induction were triggered after the maturation of ES L1-treated DCs, but these hypotheses require further investigations. A relevant criterion for the evaluation of optimal DCs properties for tolerance induction is their capacity to induce Treg cells that are able to inhibit allogeneic T cell proliferation, and this property was observed in some tolerogenic DCs like those primed by IL-10 (31, 57). We showed here that T cells primed with ES L1-treated DCs successfully suppressed the proliferation of allogeneic PBMCs, probably due to increased prevalence of Tregs cells in resulting T cell population, as well as other suppressive T cell population induced by ES L1-treated DCs.

Using HEK-Blue™ reporter cells lines expressing individual TLR or NOD-like receptors, we identified for the first time TLR2 and TLR4 as receptors that interact and induce intacellular signaling after ligation of ES L1 antigens. These results are consistent with the studies that have shown the interaction of TLR2 or TLR4 with other helminth antigens. For example, lipid fractions and lysophosphatidylserine of helminth *Schistosoma mansoni* and lipid of *Ascaris lumbricoides* reacted with TLR2 on DCs and mediate their differentiation into cells that induce Th2 and regulatory immune responses (85, 86). Lacto-*N*fucopentaose III of *S. mansoni* and ES 62 glycoprotein secreted by *Acanthocheilonema vitae*, were shown to activate TLR4 receptor and lead to consequent polarization of T cell responses toward Th2 (87, 88). Even though the microbial and helminthic products can engage the same TLRs, they can, most probably in combination with other PRRs, initiate different downstream signaling pathways that could lead to different immune response polarization. The precise underlying mechanisms are yet to be investigated.

The critical relevance of both TLR2 and TLR4 interaction with ES L1 antigens for the induction of tolerogenic properties in human DCs was demonstrated by blocking these two receptors before adding ES L1 antigens. Tolerogenic properties of ES L1-stimulated DCs were compromised when TLR2 and TLR4 were blocked, indicating, for the first time, that *T. spiralis* ES L1 antigens mediate phenotypical and functional maturation of DCs mainly *via* TLR2 and TLR4. The importance of both TLR2 and TLR4 was also demonstrated for the interaction with *S. mansoni* antigens (85, 89). Signaling *via* most TLRs normally results in the production of pro-inflammatory cytokines by DCs (90). However, helminth (and some microbial) products trigger Th2 and regulatory responses *via* interaction with TLRs (91). It is frequently suggested that stimulation of TLR2 is associated with Th2 immune response, while activation of DCs delivered through TLR4 results in Th1 type of response. ES L1 antigens interacted with both receptors on DCs, which resulted in the induction of anti-inflammatory responses. Moreover, ES L1 rendered DCs poorly responsive to TLR4-mediated induction of maturation, not *via* suppression of TLR4 expression, since we showed that ES L1 did not alter the expression of TLR4, but rather *via* modulation of downstream signaling events, and/or engagement of some other receptors expressed on DCs. Since ES L1 antigens are complex mixture of molecules, it is reasonable to assume that other PRRs are also involved in ES L1 recognition and binding. Indeed, it was revealed that ES L1 components are ligands for C-type lectin receptors, mannose receptor (92), and DC-SIGN (manuscript in preparation). Binding to lectin receptors could modulate intracellular signaling triggered by TLRs and affect DCs response in that way. Although we did not address this issue here, the results obtained with simultaneous blocking of TLR2 and TLR4 indicated that other receptors are involved as well. This presumption could also be supported by the finding that ES L1 did not affect DC differentiation, while exerted an impact on DCs maturation. The possible explanation for this phenomenon could be that completely differentiated DCs express different surface molecules compared to monocytes, including DC-SIGN, which could be important for the ES L1 impact on DCs. These findings open the possibilities for future research in trying to understand the mechanisms by which ES L1 induce toloregenic DCs.

In conclusion, this study revealed that the treatment of human monocyte-derived DCs with *T. spiralis* ES L1 antigens could be a promising new strategy for the development of stable tolerogenic DCs, with an increased capacity to suppress the inflammatory immune response while favoring the expansion of highly potent IL-10- and TGF-β- producing Tregs. Bearing in mind that the major scientific efforts are made to develop cellbased therapies that promote tolerance in humans, and although more investigation on mechanisms underlying the induction of tolerogenic DCs by ES L1 are needed, those tolerogenic DCs may present a potentially new tool for the treatment of inflammatory disorders.

### ETHICS STATEMENT

Animal experiments were performed according to institutional guidelines and were approved by the local Institutional Animal Care and Use Committee of the Institute for the Application of Nuclear Energy. Samples of human peripheral blood were obtained from healthy volunteers after written informed consent in accordance with the Declaration of Helsinki and approval by the Ethical Board of the Institute for the Application of Nuclear Energy.

### AUTHOR CONTRIBUTIONS

LS-M, MC, NI, AG-M, ST, and EP participated in the design of the study. NI, AG-M, ST, DV, JC, and CA participated in data acquisition and analysis. NI, AG-M, and ST prepared the manuscript. LS-M, MC, ST, DV, and EP participated in data interpretation and manuscript revision. LS-M and EP supervised this study and they have equally contributed. All authors gave final approval of the version to be published; and agreement to be accountable for all aspects of the work in ensuring that questions related to the

### REFERENCES


accuracy or integrity of any part of the work are appropriately investigated and resolved.

### ACKNOWLEDGMENTS

The authors greatly appreciate the donation of blocking antibodies against TLR2 and/or TLR4 (BioLegend) provided by A. Inic Kanada, OCUVAC, Medical University of Vienna, Vienna, Austria. The authors gratefully acknowledge support of JC by a FEBS short-term fellowship to characterize the interaction between different human pattern recognition receptors (PRRs) and *Trichinella spiralis* excretory–secretory muscle larvae antigens (ES L1).

### FUNDING

This work was supported by the Ministry of Education, Science and Technological Development, Republic of Serbia (projects No. 173047, 175102 and 173050), a grant (MFVMA/9/16-18) of the Military Medical Academy, University of Defence, Belgrade, and Strategic Program from the National institute for Public Health and the Environment, the Netherlands, grant RIVM-S/112001.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2018.00011/ full#supplementary-material.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a past co-authorship with one of the authors LS-M.

*Copyright © 2018 Ilic, Gruden-Movsesijan, Cvetkovic, Tomic, Vucevic, Aranzamendi, Colic, Pinelli and Sofronic-Milosavljevic. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Maxime De Laere1 , Judith Derdelinckx1,2, Mari Hassi1 , Mari Kerosalo1 , Heidi Oravamäki1 , Johan Van den Bergh1 , Zwi Berneman1,3 and Nathalie Cools1 \**

#### *Edited by:*

*Joanna Davies, San Diego Biomedical Research Institute, United States*

#### *Reviewed by:*

*Maria Cecilia G. Marcondes, San Diego Biomedical Research Institute, United States Jonathan D. Katz, Cincinnati Children's Research Foundation, United States*

#### *\*Correspondence:*

*Nathalie Cools nathalie.cools@uza.be*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 07 July 2017 Accepted: 19 December 2017 Published: 23 January 2018*

#### *Citation:*

*De Laere M, Derdelinckx J, Hassi M, Kerosalo M, Oravamäki H, Van den Bergh J, Berneman Z and Cools N (2018) Shuttling Tolerogenic Dendritic Cells across the Blood– Brain Barrier In Vitro via the Introduction of De Novo C–C Chemokine Receptor 5 Expression Using Messenger RNA Electroporation. Front. Immunol. 8:1964. doi: 10.3389/fimmu.2017.01964*

Frontiers in Immunology | www.frontiersin.org

*<sup>1</sup> Laboratory of Experimental Hematology, Faculty of Medicine and Health Sciences, Vaccine & Infectious Disease Institute (VAXINFECTIO), University of Antwerp, Wilrijk, Belgium, 2Department of Neurology, Antwerp University Hospital, Edegem, Belgium, 3Center for Cell Therapy and Regenerative Medicine, Antwerp University Hospital, Edegem, Belgium*

The use of tolerance-inducing dendritic cells (tolDCs) has been proven to be safe and well tolerated in the treatment of autoimmune diseases. Nevertheless, several challenges remain, including finding ways to facilitate the migration of cell therapeutic products to lymph nodes, and the site of inflammation. In the treatment of neuroinflammatory diseases, such as multiple sclerosis (MS), the blood–brain barrier (BBB) represents a major obstacle to the delivery of therapeutic agents to the inflamed central nervous system (CNS). As it was previously demonstrated that C–C chemokine receptor 5 (CCR5) may be involved in inflammatory migration of DCs, the aim of this study was to investigate CCR5-driven migration of tolDCs. Only a minority of *in vitro* generated vitamin D3 (vitD3)-treated tolDCs expressed the inflammatory chemokine receptor CCR5. Thus, messenger RNA (mRNA) encoding CCR5 was introduced by means of electroporation (EP). After mRNA EP, tolDCs transiently displayed increased levels of CCR5 protein expression. Accordingly, the capacity of mRNA electroporated tolDCs to transmigrate toward a chemokine gradient in an *in vitro* model of the BBB improved significantly. Neither the tolerogenic phenotype nor the T cell-stimulatory function of tolDCs was affected by mRNA EP. EP of tolDCs with mRNA encoding CCR5 enabled these cells to migrate to inflammatory sites. The approach used herein has important implications for the treatment of MS. Using this approach, tolDCs actively shuttle across the BBB, allowing *in situ* down-modulation of autoimmune responses in the CNS.

Keywords: tolerogenic dendritic cells, C–C chemokine receptor 5, messenger RNA electroporation, migration, blood–brain barrier, multiple sclerosis

### INTRODUCTION

Multiple sclerosis (MS) is a chronic autoinflammatory disease of the central nervous system (CNS), mediated by myelin-reactive T cells that escape central and peripheral tolerance mechanisms and induce inflammation and tissue damage within the CNS (1, 2). During the last two decades, several new and increasingly efficacious therapeutics have become available for the treatment of MS (3). However, this higher treatment efficacy is associated with a more hazardous adverse event profile (4), and none of the currently approved treatments is successful in completely halting MS. In addition, as the disease progresses, these therapeutics become less effective. The blood–brain barrier (BBB) represents a major hurdle in the treatment of this neuroinflammatory disorder. Previous authors hypothesized that during the progressive phases of MS, inflammation is trapped behind an intact BBB and hence is not accessible to immunomodulatory agents (5–7). Finding ways to improve the access of therapeutic agents to the CNS would undoubtedly and markedly improve the treatment outcome in progressive forms of MS.

Major advancements in current knowledge of immunology, together with increased understanding of the processes underlying MS and mechanisms contributing to immune tolerance, have led to the emergence of immune-regulatory cell therapy as a promising strategy to restore tolerance in MS (8, 9). Toleranceinducing dendritic cells (tolDCs) or tolerogenic dendritic cells have a unique ability to steer the host immune response toward tolerance induction (10). In general, tolDCs can be defined as maturation-resistant DCs, characterized by low to intermediate expression levels of major histocompatibility complex (MHC) class II and costimulatory molecules (11). They mediate tolerance by inducing T-cell anergy, deleting autoreactive T cells, and/or inducing and expanding the population of regulatory T cells (11). In early clinical trials, tolDC-based therapies were proven to be safe and well tolerated for the treatment of autoimmune diseases (12–15). The efficacy of this treatment approach remains to be determined in further clinical studies, and factors that affect the efficacy of tolDC-based therapies are not yet fully understood.

The migratory capacity of tolDCs may influence the potential clinical use of these cells. It can be reasoned that *in vivo* efficacy of tolDC-based therapies will depend not only on their potency (i.e., ability to induce tolerance) but also on their probability of encountering T cells and thus their ability to reach target organs (i.e., lymph nodes and CNS) in MS. DC migration to lymph nodes is mainly determined by C-C chemokine receptor 7 (CCR7) (16). CCR5, on the other hand, is a key molecule involved in guiding DCs to the site of inflammation (17). Some studies reported that expression levels of the CCR5 ligands CCL3, CCL4, and CCL5 were upregulated in lesions and cerebrospinal fluid of patients with MS (18–22). We (23) and others (24) demonstrated that circulating DCs of MS patients expressed increased levels of CCR5. Based on these findings, we hypothesized that the expression of CCR5 on tolDCs might drive DC migration to an inflamed CNS.

In animal model studies, the presence of steady-state or tolerogenic DCs in the CNS suppressed experimental autoimmune encephalomyelitis (EAE) (25–27). Mechanisms underlying this tolerance induction included preferential secretion by DCs of the immunomodulatory cytokines interleukin-10 (IL-10) and transforming growth factor-β, in addition to skewing of the T-cell response by favoring the development of T-helper 2 cells and regulatory T cells, while restraining T-helper 17 cell development. In these studies, DCs were either cultured *in vitro* and injected intracerebrally (27), rendered tolerogenic in the CNS *in situ* by hepatocyte growth factor selectively overexpressed by neurons (25), or implicated in the induction of tolerance after intravenous injection of an autoantigen peptide of myelin oligodendrocyte glycoprotein (26). Previously, we reported a culture protocol for the generation of vitamin D3 (vitD3)-treated tolDCs (28). Our data showed that vitD3-treated tolDCs of MS patients displayed a semi-mature phenotype and an anti-inflammatory cytokine profile. In addition, vitD3-treated tolDCs induced antigen-specific T-cell hyporesponsiveness, supporting the clinical potential of these cells in correcting the immunological imbalance inherent in MS. However, it remains to be determined to what extent *in vitro*-generated tolDCs migrate to an inflamed CNS, especially as this requires transmigration across the BBB. Unger et al. (29) reported that tolDCs downregulated CCR5 expression upon proinflammatory stimulation, suggesting that inflammatory trafficking of these cells might be suboptimal. This prompted us to study the CCR5-driven migratory capacity of tolDCs *in vitro* in a previously optimized and characterized model of the BBB (30). We hypothesized that the CCR5 driven migratory capacity of these cells could be optimized by introducing CCR5 protein expression using messenger RNA (mRNA) electroporation (EP). Ultimately, endowing tolDCs with the capacity to migrate to an inflamed CNS by introducing *de novo* CCR5 protein expression will allow optimal exploitation of their tolerogenic capacity. Active shuttling of cells across the BBB would allow for targeted *in situ* down-modulation of autoimmune responses by tolDCs.

### MATERIALS AND METHODS

### *In Vitro* Generation of Monocyte-Derived Dendritic Cells

Peripheral blood from healthy donors was obtained from buffy coats provided by the Red Cross donor center (Red Cross-Flanders, Mechelen, Belgium). Peripheral blood mononuclear cells were isolated by density gradient centrifugation (Ficoll Pacque PLUS, GE Healthcare, Amsterdam, the Netherlands). From the peripheral blood mononuclear cell fraction, monocytes were purified by CD14<sup>+</sup> immunomagnetic selection (CD14 Reagent, Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer's instructions. The CD14-depleted cell fraction [i.e., peripheral blood lymphocytes (PBLs)] was cryopreserved in fetal bovine serum (Thermo Fisher Scientific, Erembodegem, Belgium) supplemented with 10% dimethyl sulfoxide (Sigma-Aldrich, Bornem, Belgium) and stored at −80°C for later use in an allogeneic mixed leukocyte reaction. CD14<sup>+</sup> monocytes were cultured *in vitro* at a density of 1–1.2 × 106 /ml and differentiated into DCs in culture medium consisting of Iscove's modified Dulbecco's medium (IMDM) with l-glutamine (Thermo Fisher Scientific), supplemented with 200 IU/ml of granulocyte-macrophage colonystimulating factor (Gentaur, Brussels, Belgium), 250 IU/ml of IL-4 (Miltenyi Biotec), 2% human AB (hAB) serum (Thermo Fisher Scientific), 10 µg/ml of gentamicin (Thermo Fisher Scientific), and 1 µg/ml of amphotericin B (Thermo Fisher Scientific). TolDCs were differentiated under the same conditions, except for the addition of 2 nM 1,25(OH)2-vitamin D3 (vitD3, Calcijex, Abbott Laboratories, IL, USA) to the culture medium. On day 4 of culture, DCs were subjected to a proinflammatory cytokine cocktail by the addition of 1,000 IU/ ml of IL-1β (Miltenyi Biotec), 1,000 IU/ml of tumor necrosis factor-α (Miltenyi Biotec), and 2.5 µg/ml of prostaglandin E2 (Pfizer, Elsene, Belgium) to obtain mature control and tolDCs. For tolDC cultures, vitD3 was replenished on day 4. The cells were cultured in a humidified atmosphere with 5% CO2 at 37°C. On day 6, DCs were harvested for use in further experiments.

The study was approved by the ethics committee of Antwerp University Hospital and the University of Antwerp (15/50/543) and followed the tenets of the Declaration of Helsinki.

### Messenger RNA EP

The complementary DNA sequence of human *CCR5* (accession number U54994) was modified for optimal codon use (Figure S1 in Supplementary Material) and subcloned into a pST1-plasmid vector under the control of a T7 promotor and with the addition of a poly(A)tail (GeneArt, Thermo Fisher Scientific). After transformation in *Escherichia coli* and linearization of the circular DNA plasmid, mRNA transcripts were generated using a T7 *in vitro* transcription kit (mMessage mMachine T7 kit, Ambion, Life Technologies), according to the manufacturer's protocol. mRNA was resuspended at a concentration of 1 µg/µl, aliquoted, and stored at −20°C.

Messenger RNA EP of DCs was performed as previously described (31, 32). In brief, the cells were resuspended in Opti-MEM (Thermo Fisher Scientific), and a 200 µL aliquot of this cell suspension containing 2–10 × 106 cells was transferred into a 0.4-cm cuvette (Immunosource, Schilde, Belgium). Next, 10 µg of mRNA were added. EP was performed using a Gene Pulser Xcell™ electroporation system (Bio-Rad, Temse, Belgium) with a time constant protocol at 300 V for 7 ms. EP of cells without the addition of mRNA (mock EP) was performed as a control. Immediately after EP, the cells were transferred into fresh DC culture medium. For tolDCs, 2 nM vitD3 was added to the cell culture medium. After a 30-min resting phase, the cells were washed and resuspended in warm IMDM supplemented with 5% hAB serum. Following an additional resting period of 90 min, the cells were washed again, resuspended in IMDM supplemented with 1% hAB serum, and used in further experiments.

### Flow Cytometric Phenotyping

Flow cytometric analysis of the expression of CCR5 by DCs was performed 2, 4, 24, 48, and 72 h after EP. *CCR5* mRNAelectroporated, mock-electroporated, and nonelectroporated DCs were stained with a phycoerythrin-cyanin 7-labeled anti-CCR5-antibody (BD Pharmingen, Erembodegem, Belgium) or an isotype-matched control antibody (BD Pharmingen). LIVE/ DEAD® Fixable Violet Dead Cell Stain (Thermo Fisher Scientific) was added to assess cell viability. The indicated percentages of CCR5-positive cells were within the living DC population (i.e., gated for DCs based on light scatter properties and negative for LIVE/DEAD® Fixable Violet Dead Cell staining). Flow cytometric measurements were performed using a Cyflow ML flow cytometer (Partec, Münster, Germany). The results were analyzed using FlowJo software (Tree Star, Ashland, OR, USA).

The phenotype of DCs was characterized using the following fluorochrome-labeled mouse antihuman monoclonal antibodies: anti-CD83-fluorescein isothiocyanate (Life Technologies), anti-CD80-phycoerythrin (BD Pharmingen), antihuman leukocyte antigen (HLA)-DR-peridinin chlorophyll (BD Biosciences), anti-CD86-fluorescein isothiocyanate (BD Pharmingen), and anti-CCR5-phycoerythrin-cyanin 7. Isotype-matched control monoclonal antibodies were used to determine nonspecific background staining. For analytical flow cytometry, at least 104 events were acquired using a FACScan flow cytometer (BD). The indicated percentages were within the DC population based on light scatter properties. All the results were analyzed using FlowJo software.

### *In Vitro* BBB Model

The *in vitro* BBB model was constructed as described previously (30). In brief, human primary astrocytes (Sanbio, Uden, the Netherlands) were seeded at a density of 15,000 cells/cm2 on the poly-l-lysine-coated underside of a transwell (24-well format) with 3.0-µm pore size (Greiner Bio-one, Vilvoorde, Belgium) and allowed to adhere for 2 h. Subsequently, the inserts were transferred into a well filled with EGM-2-MV medium (Lonza, Verviers, Belgium) with 2.5% fetal bovine serum. hCMEC/D3 endothelial cells (Tébu-bio, Le Perray-en-Yvelines, France) were seeded onto the insert's collagen-coated upper side at a density of 25,000 cells/cm2 . Cultures were maintained in EGM-2-MV medium in 5% CO2 at 37°C. Three days after initiating the coculture, the growth medium was replaced by EBM-2-plus medium, consisting of EBM-2 medium (Lonza), supplemented with 1.4 µM hydrocortisone (Pfizer), 1 ng/ml of basic fibroblast growth factor (Thermo Fisher Scientific), 10 µg/ml of gentamicin, 1 µg/ml of amphotericin-B, and 2.5% fetal bovine serum. EBM-2-plus medium was replenished every other day. Migration assays were performed between days 10 and 13 of culture.

### Migration Assay

Chemotaxis of DCs was studied 2 h after EP or at an equivalent time point for nonelectroporated DCs using 3.0-μm-sized pore transwells and an *in vitro* BBB model. DCs (2 × 105 ) were added to the upper compartment of both the transwell and *in vitro* BBB model. The basolateral compartment contained 25 ng/ml of CCL4 and 25 ng/ml of CCL5 in IMDM, supplemented with 1% hAB serum. DCs were subsequently allowed to migrate for 4 h in the transwell assays or for 24 h in assays using the *in vitro* BBB model. The negative control consisted of 2 × 105 DCs added to the upper compartment, while no chemokines were added to the basolateral compartment. As a positive control, 2 × 105 DCs were added directly to the basolateral compartment. At the indicated time points, DCs were collected from the basolateral compartment. After resuspension in a fixed volume of 200 µl, they were counted using a BD FACScan flow cytometer. Events were acquired at a fixed flow rate for exactly 120 s. The results were analyzed using FlowJo software. The percentage migration was calculated as follows:


### RNA Isolation and Quantitative Real-time Polymerase Chain Reaction (qPCR)

For analysis of the gene expression profile of nonelectroporated and electroporated tolDCs and control DCs, total RNA was isolated. The cells were disrupted and homogenized using guanidine-thiocyanate-containing lysis buffer. Total RNA was isolated using an RNeasy microkit (Qiagen, Antwerp, Belgium). The RNA concentration was determined by measuring absorbance at 260 nm using a Nanodrop spectrophotometer (Wilmington, DE, USA). Reverse transcription of the obtained RNA into cDNA was performed using an iScript™ Advanced cDNA Synthesis Kit (Bio-Rad). Subsequently, SYBR® Green technology was used for relative mRNA quantification by qPCR in a CFX96 C1000 thermal cycler (Bio-Rad). qPCR reactions were conducted at 95°C for 30 s, followed by 40 cycles at 95°C for 5 s, and at 60°C for 30 s. All primer sets were obtained from Bio-Rad; validation data are shown in Table S1 in Supplementary Material. qPCR was performed in triplicate, and resulting mRNA levels were normalized to levels of the reference genes beta-actin and phosphoglycerate kinase 1. Melt curve analysis was performed to confirm the specificity of the amplified product. Bio-Rad CFX manager v3.1 was used for data processing and analysis.

### Allogeneic Mixed Lymphocyte Reaction

To assess the allogeneic T-cell stimulatory capacity of DCs, the cells were cocultured with allogeneic PBLs in a 1:10 ratio. Nonstimulated responder PBLs served as a negative control, and allogeneic PBLs stimulated with 1 µg/ml of phytoheamagglutinin (Sigma-Aldrich) were used as a positive control. Cocultures were performed in IMDM supplemented with 5% hAB serum at 37°C. After 6 days in coculture, the secreted level of interferon-γ (IFN-γ) in the cell culture supernatant was determined in duplicate as

a measure of allo-stimulatory capacity using a commercially available enzyme-linked immunosorbent assay (ELISA) kit (PeproTech, NJ, USA). In addition, IL-10 secretion was measured in the supernatant using a U-PLEX assay (Meso Scale Discovery, MD, USA), according to the manufacturer's instructions.

### Statistical Analysis

Data were analyzed using Graphpad Prism software version 5.01 (Graphpad, San Diego, CA, USA), except for qPCR data, which were analyzed using CFX Manager software, version 3.1 (Bio-Rad). Comparison of nonelectroporated, mock-electroporated, and *CCR5* mRNA-electroporated tolDCs was performed by a repeated measures one-way ANOVA, followed by Tukey's multiple comparisons test. For data that were not normally distributed according to the Kolmogorov–Smirnov test, the Friedman test, with Dunn's multiple comparison test was performed. Comparison of CCR5 expression levels at several time points after EP in mock-electroporated, *CCR5* mRNA-electroporated, and nonelectroporated tolDCs was performed using a two-way repeated measures ANOVA, with *post hoc* Bonferroni tests. For qPCR results, differences were considered significant when *p* < 0.01. For other data, statistical significance was considered at the 5% level. Data are shown as mean ± SEM. The number of biological replicates is indicated in the figure or table legend.

### RESULTS

### TolDCs Displayed Limited CCR5-Driven Migratory Capacity

Only a minority of *in vitro* generated vitD3-treated tolDCs expressed CCR5 (i.e., 12.96 ± 2.02% on average). This translated

toward CCR5 ligands CCL4 and CCL5 in a transwell chemotaxis assay (mean ± SEM of seven replicates). \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001. (C) Schematic overview of the tolDC migration experiment using an *in vitro* blood–brain barrier (BBB) model. (D) Although nonelectroporated and mock-electroporated tolDCs displayed only a limited capacity to transmigrate through the BBB *in vitro* in response to CCL4 and CCL5, EP of tolDCs with *CCR5* mRNA increased their transmigratory capacity in response to chemokines added basolaterally in the *in vitro* BBB model (mean ± SEM of six replicates). \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001.

into marginal chemotaxis of tolDCs. Only 0.10 ± 0.05% of tolDCs migrated in response to the chemokines CCL4 and CCL5.

### mRNA EP Resulted in a Marked Increase of CCR5 Expression in tolDCs

To increase CCR5 protein expression, tolDCs were electroporated with mRNA encoding CCR5. Using flow cytometric analysis of CCR5 protein expression at consecutive time points after EP, an incremental increase was detected in CCR5 expression levels from 2 to 48 h following EP, after which expression decreased again (**Figure 1**). CCR5 expression levels of *CCR5* mRNAelectroporated DCs were significantly higher as compared with those of both nonelectroporated and mock-electroporated DCs 4 h (36.84 ± 5.89 vs. 2.87 ± 0.60 and 4.75 ± 0.79%, respectively; *p* < 0.05), 24 h (55.42 ± 10.09 vs. 7.09 ± 3.04 and 5.04 ± 1.74%, respectively; *p* < 0.001), and 48 h (59.52 ± 9.59 vs. 14.44 ± 9.48 and 14.08 ± 7.86%, respectively; *p* < 0.001) after EP.

### CCR5 mRNA-Electroporated tolDCs Demonstrated Increased CCR5-Driven Migration *In Vitro*

To investigate whether elevated CCR5 expression translated into a higher capacity to migrate *in vitro*, chemotaxis of tolDCs across transwells in response to the CCR5 ligands CCL4 and CCL5 was studied. Although only 0.22 ± 0.11% of nonelectroporated tolDCs and 0.11 ± 0.10% of mock-electroporated tolDCs migrated toward a CCL4 and CCL5 gradient, 2.59 ± 0.37% of *CCR5* mRNAelectroporated tolDCs showed chemokine-mediated migration (*p* < 0.05 and *p* < 0.01, respectively) (**Figures 2A,B**). Likewise, there was an 18-fold increase in CCR5-driven transmigration of tolDCs across an *in vitro* BBB model following *CCR5* mRNA EP of tolDCs. Only 0.22 ± 0.16% of nonelectroporated tolDCs and 0.35 ± 0.35% of mock-electroporated tolDCs succeeded in transmigrating across the *in vitro* BBB model. In contrast, 4.98 ± 1.24% of tolDCs electroporated with *CCR5* mRNA crossed the BBB in response to CCL4 and CCL5 (*p* < 0.05) (**Figures 2C,D**).

### mRNA EP Did Not Affect the Tolerogenic Phenotype and Function of tolDCs

To ensure that the semi-mature phenotype of tolDCs was unaffected by mRNA EP, the expression of DC maturation markers, as well as that of molecules involved in antigen presentation, was investigated (**Table 1**). Mock or mRNA EP did not affect the proportion of control DCs or that of tolDCs expressing CD80, CD83, CD86, and HLA-DR. In addition, the level of protein expression per cell, as assessed by mean fluorescence intensity, was not significantly affected for the membrane molecules expressed by tolDCs following mRNA EP. In control DCs, a modest but significant decrease in protein expression levels was observed after EP for all molecules tested. However, expression levels of CD80, CD83, CD86, and HLA-DR were still significantly higher in *CCR5* mRNA-electroporated control DCs as compared to tolDCs. Interestingly, neither mock nor mRNA EP affected mRNA expression levels of LILRB4 and TLR2, two established regulators of tolerogenicity, in vitD3-treated tolDCs (33). Normalized expression levels of both markers remained significantly higher in tolDCs as compared with those of control DCs (**Table 1**).

Functionally, tolDCs maintained their capacity to induce T-cell hyporesponsiveness following mRNA EP (**Figure 3A**). No differences were observed in the level of secreted IFN-γ in the supernatant of PBLs stimulated with nonelectroporated tolDCs as compared with that of PBLs stimulated with either mock- or *CCR5* mRNA-electroporated tolDCs. In contrast, the levels of IL-10 secreted in the coculture supernatant were significantly higher when mock- or *CCR5* mRNA-electroporated tolDCs were

Table 1 | The semi-mature phenotype and tolerogenic messenger RNA (mRNA) expression profile of tolerance-inducing dendritic cells (TolDCs) was not affected by mRNA electroporation.


*Mean* ± *SEM of six replicates (protein expression) and three replicates (mRNA expression).*

<sup>+</sup>*Denotes a statistically significant difference from nonelectroporated DCs, within control DC or tolDC conditions.*

*\*Denotes a statistically significant difference from the corresponding (i.e., non-EP, mock EP, or CCR5 EP) control DC conditions.*

*\*/*+*p* < *0.05, \*\*/*++*p* < *0.01, \*\*\*/*+++*p* < *0.001.*

*MFI, mean fluorescence intensity; RNE, relative normalized expression.*

cocultured with PBLs as compared with cocultures of PBLs with corresponding control DCs (*p* < 0.05 and *p* < 0.001, respectively) (**Figure 3B**).

### DISCUSSION

TolDC-based therapies represent a promising strategy for the future treatment of autoimmune diseases, such as MS. DCs are key players in maintaining the balance between immunity and tolerance by priming T-cell responses in an antigen-specific manner (34). This makes them ideal vehicles for modulating detrimental autoimmune reactions in a disease-specific way, without compromising immune surveillance and host-protective mechanisms. Although previous research confirmed the safety and tolerability of tolDC treatment in patients with type I diabetes (13), rheumatoid arthritis (12, 35, 36), and Crohn's disease (15), the efficacy of tolDC-based treatments in human autoimmune diseases remains to be determined. In this regard, it can be envisaged that the ability of *in vitro* generated tolDCs to downmodulate an ongoing pathological immune response *in vivo* will critically

(B) Levels of IL-10 in the supernatant of PBLs cocultured with control or tolDCs. Cocultures of PBLs with mock- or *CCR5* mRNA-electroporated tolDCs contained higher levels of IL-10 as compared with cocultures of PBLs with the corresponding control DCs [mean ± SEM of four replicates (IFN-γ) or six replicates (IL-10)]. \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001.

depend on their ability to reach both secondary lymphoid organs and the site of inflammation. In MS, therapeutic access to the CNS is hindered by the BBB. The aforementioned could explain, at least in part, poor treatment responses typically observed in MS, especially in progressive disease stages, in which the BBB is hypothesized to encapsulate inflammation within the CNS (6, 7).

The BBB is also a major obstacle for tolDCs administered peripherally, making it difficult for them to reach the CNS for *in situ* down-modulation of ongoing inflammation. Following cell tracking and imaging of intravenously administered vitD3 treated tolDCs, Mansilla et al. (37) found only a transient and low level signal from labeled tolDCs in the brain of EAE mice. In the present study, only a minority of *in vitro* cultured vitD3-treated tolDCs expressed CCR5, despite having a maturation-resistant phenotype (28) and CCR5 being mainly expressed by immature DCs (38). Accordingly, the cells exhibited only marginal chemotaxis to a CCL4 and CCL5 gradient, with less than 0.2% of tolDCs on average displaying chemokine-driven migration. To increase the migratory potential of tolDCs, the cells were transfected with mRNA encoding the CCR5 protein according to a previously optimized protocol for mRNA EP of *in vitro* generated DCs (31, 32). Following mRNA EP, CCR5 protein expression reached its zenith 48 h after EP. Increased CCR5 expression resulted in higher *in vitro* migratory capacity of tolDCs in response to CCL4 and CCL5. Interestingly, mRNA electroporated tolDCs also displayed a higher capacity to transmigrate through the BBB *in vitro* in response to these chemokines. The number of cells needed to achieve a therapeutic effect *in vivo* is not known. Previous research showed that only 2–4% of the total administered population of immune-stimulatory DCs reached lymph nodes following *in vivo* migration but that this low number of cells was sufficient to elicit an antigen-specific immune response *in vivo* (39–42).

The finding that responsiveness of tolDCs to CCR5 ligands can be boosted is of particular relevance for the treatment of MS, as previous studies confirmed that these chemokines were upregulated in the CNS of MS patients (18–22). Moreover, we and others showed that CCR5 ligands were actively transported across the BBB (30, 43, 44). In this way, they provide traffic cues for circulating immune cells to enter the inflamed CNS. Previous research demonstrated that the presence of DCs with tolerogenic properties in the CNS delayed, prevented, or ameliorated EAE (25–27). Therefore, we hypothesize that *CCR5* mRNA-electroporated DCs will outperform nonmodulated tolDCs in terms of efficacy due to their acquired capacity to reach the site of inflammation. However, this hypothesis remains to be tested in *in vivo* models of neuroinflammation.

Besides being implicated in MS pathogenesis, CCR5 ligands drive immune cell accumulation in affected tissue in several other autoinflammatory and immune-mediated diseases. Researchers reported elevated levels of these chemokines in the inflamed synovium of rheumatoid arthritis patients (45), pancreatic islets of type I diabetic patients (46), and intestines of patients with inflammatory bowel disease (47). Hence, modulation of migratory capacity of tolDCs driven by CCR5 may also be advantageous in the treatment of these diseases. The approach described herein can also be applied to enhance the expression of and migration directed by other chemokine receptors, making it possible to tailor the migratory capacity of therapeutically administered cell populations to the chemokine expression profile associated with a specific target organ or disorder. For example, migration of tolDCs to lymph nodes is mainly driven by CCR7. Its ligands, CCL19 and CCL21, are highly expressed by lymph node fibroblastic reticular cells (48–50) and lymphatic endothelial cells (51). They guide mature DCs and specific T-cell subsets to T-cell zones of lymph nodes, coordinating their colocalization for subsequent interaction. CCR7 expression is upregulated on DCs by a maturation stimulus. Likewise, the expression of this chemokine receptor on tolDCs is upregulated after a proinflammatory challenge, albeit expression levels on tolDCs remain significantly lower as compared with those on mature DCs (29, 52, 53). This translates into reduced migratory capacity toward CCL19 and CCL21 *in vitro*. Similarly, introducing CCR7 expression in tolDCs using the proposed approach of chemokine receptor mRNA EP could overcome the limited lymphoid homing capacity of tolDCs.

RNA can act as both a pathogen-associated and damageassociated molecular pattern (54–56). Intracellular introduction of RNA by means of EP could thus lead to DC activation. In agreement with previous findings showing that mature monocytederived DCs were not activated by electroporated mRNA (57), we showed that mRNA EP did not affect the semi-mature phenotype, tolerogenic gene expression signature, or allo-stimulatory capacity of vitD3-treated tolDCs.

In conclusion, this is the first study to show that enhancing CCR5 expression of tolDCs using mRNA EP endowed these cells with CCR5-driven migratory capacity. This enabled the cells to migrate to inflammatory sites, even when this required crossing of functional barriers, such as the BBB. Importantly, both the tolerogenic phenotype and function of tolDCs were unaffected by the process of mRNA EP. These findings represent an important step forward in the development of a next generation of cell-based tolerance-inducing therapies for the treatment of immune-mediated disorders.

### AUTHOR CONTRIBUTIONS

All authors have contributed substantially to this work, have approved the manuscript, and agreed with its submission.

### ACKNOWLEDGMENTS

The authors would like to thank Gitte Slingers (Hasselt University) and Dr. Gudrun Koppen (Flemish Institute for Technological Research) for their excellent guidance and technical support in performing the MSD analysis.

### FUNDING

This work was supported by a BOF-GOA grant no. PS 28313 of the Special Research Fund (BOF) from the University of Antwerp, Belgium. Further support was provided through the Methusalem Funding Program from the University of Antwerp, by an applied biomedical research project of the Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT-TBM 140191) and by the Belgian Charcot Foundation. This work has been supported by positive discussion through the A FACTT network (Cost Action BM1305: www.afactt.eu). COST is supported by the EU Framework Programme Horizon 2020. Judith Derdelinckx holds a SB Ph.D. fellowship from the Research Foundation Flanders (FWO-Vlaanderen).

### REFERENCES


### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01964/ full#supplementary-material.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer MM and handling editor declared their shared affiliation.

*Copyright © 2018 De Laere, Derdelinckx, Hassi, Kerosalo, Oravamäki, Van den Bergh, Berneman and Cools. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Directed Differentiation of Human Induced Pluripotent stem cells into Dendritic cells Displaying tolerogenic Properties and resembling the cD141**+** subset

#### *Edited by:*

*John Isaacs, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Hans Acha-Orbea, University of Lausanne, Switzerland Simon Milling, University of Glasgow, United Kingdom*

> *\*Correspondence: Paul J. Fairchild paul.fairchild@path.ox.ac.uk*

#### *†Present address:*

*Patty Sachamitr, Peter Gilgan Centre for Research and Learning, The Hospital for Sick Children, Toronto, ON, Canada*

> *‡ These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 30 October 2017 Accepted: 15 December 2017 Published: 08 January 2018*

#### *Citation:*

*Sachamitr P, Leishman AJ, Davies TJ and Fairchild PJ (2018) Directed Differentiation of Human Induced Pluripotent Stem Cells into Dendritic Cells Displaying Tolerogenic Properties and Resembling the CD141+ Subset. Front. Immunol. 8:1935. doi: 10.3389/fimmu.2017.01935*

*Patty Sachamitr†,‡, Alison J. Leishman‡ , Timothy J. Davies and Paul J. Fairchild\**

*Sir William Dunn School of Pathology, University of Oxford, Oxford, United Kingdom*

The advent of induced pluripotent stem cells (iPSCs) has begun to revolutionize cell therapy by providing a convenient source of rare cell types not normally available from patients in sufficient numbers for therapeutic purposes. In particular, the development of protocols for the differentiation of populations of leukocytes as diverse as naïve T cells, macrophages, and natural killer cells provides opportunities for their scale-up and quality control prior to administration. One population of leukocytes whose therapeutic potential has yet to be explored is the subset of conventional dendritic cells (DCs) defined by their surface expression of CD141. While these cells stimulate cytotoxic T cells in response to inflammation through the cross-presentation of viral and tumor-associated antigens in an MHC class I-restricted manner, under steady-state conditions CD141+ DCs resident in interstitial tissues are focused on the maintenance of homeostasis through the induction of tolerance to local antigens. Here, we describe protocols for the directed differentiation of human iPSCs into a mixed population of CD11c+ DCs through the spontaneous formation of embryoid bodies and exposure to a cocktail of growth factors, the scheduled withdrawal of which serves to guide the process of differentiation. Furthermore, we describe the enrichment of DCs expressing CD141 through depletion of CD1c+ cells, thereby obtaining a population of "untouched" DCs unaffected by cross-linking of surface CD141. The resulting cells display characteristic phagocytic and endocytic capacity and acquire an immunostimulatory phenotype following exposure to inflammatory cytokines and toll-like receptor agonists. Nevertheless, under steady-state conditions, these cells share some of the tolerogenic properties of tissue-resident CD141+ DCs, which may be further reinforced by exposure to a range of pharmacological agents including interleukin-10, rapamycin, dexamethasone, and 1α,25-dihydoxyvitamin D3. Our protocols therefore provide access to a novel source of DCs analogous to the CD141+ subset under steady-state conditions *in vivo* and may, therefore, find utility in the treatment of a range of disease states requiring the establishment of immunological tolerance.

Keywords: induced pluripotent stem cell, dendritic cell, regulatory T cell, directed differentiation, tolerance, CD141

#### Sachamitr et al. Tolerogenic DCs from iPSCs

### INTRODUCTION

Through their unrivaled capacity for antigen processing and presentation, dendritic cells (DCs) are uniquely equipped to engage naïve T cells in dialog, implicating them in the genesis of all immune responses (1). As such, DCs are responsible for defining the outcome of antigen recognition, either ensuring robust immunity to a microbial challenge or pacifying deleterious autoimmune responses through the induction and maintenance of immunological tolerance. Which of these diametrically opposed outcomes prevails depends primarily on the context in which antigen presentation by DCs occurs, steady-state conditions promoting the maintenance of tolerance, while ongoing inflammation favors immunity (1). These properties have made DCs attractive therapeutic agents for intervening in the progression of an immune response, inspiring numerous clinical trials for vaccination to poorly immunogenic tumor associated antigens (TAAs) as the basis for cancer immunotherapy (2). Furthermore, the clinical application of DCs has recently extended beyond vaccination to the induction of antigen-specific tolerance for the treatment of autoimmune diseases as diverse as diabetes (3, 4), multiple sclerosis (5), and rheumatoid arthritis (6, 7) as well as the prevention of allograft rejection (8, 9). While these trials have shown a good safety profile (3), they have yet to demonstrate significant efficacy: for instance, recent analyses of over 54 clinical trials for melanoma revealed objective response rates of less than 10% (10).

Such disappointing outcomes may be attributed in part to the identity of the DCs employed in clinical trials which, for pragmatic reasons, are most commonly differentiated *in vitro* from the patient's own peripheral blood monocytes which may be subsequently matured by exposure to inflammatory cytokines or treated with a range of pharmacological agents such as interleukin (IL) 10, dexamethasone, rapamycin, and 1α,25-dihydroxyvitamin D3 (VD3), widely demonstrated to restrain their immunogenicity and render them more tolerogenic (11). Although ease of access confers a significant advantage on monocyte-derived DCs (moDCs), they are known to exhibit substantial donor-to-donor variation, which may be exacerbated by exposure of patients to long-term chemotherapy or immune suppression. Furthermore, moDCs display poor capacity for the cross-presentation of soluble or cellular antigens to MHC class I-restricted CD8<sup>+</sup> T cells. Antigen cross-presentation is not only a requirement for induction of the cytotoxic T lymphocyte (CTL) responses essential for the clearance of an established tumor (2) but has also been strongly implicated in the maintenance of "cross-tolerance" among CD8<sup>+</sup> T cells under steady-state conditions (12). The use of alternative subsets of DCs with proven capacity for the cross-presentation of soluble and cellular antigens may, therefore, provide a rational alternative to the widespread use of moDCs for immunotherapy.

In the human, conventional DC (cDC) belong to two distinct subsets, identified by their surface expression of CD1c or CD141. These subsets derive from a common progenitor which fails to give rise to monocytes or plasmacytoid DCs, formally distinguishing them from either lineage (13). CD141<sup>+</sup> DCs were recently shown to exhibit superior capacity for antigen crosspresentation (14–17). Furthermore, they may be defined by their co-expression of toll-like receptor (TLR) 3, Clec9A and the chemokine receptor, XCR1 and have been shown to be critical for eliciting responses to tumor and viral antigens without requiring either direct infection or endogenous expression of TAAs (18). To perform such a function, CD141<sup>+</sup> DCs are highly endocytic and phagocytic, permitting their efficient acquisition of both soluble and cellular antigens (19). Through cross-presentation of acquired antigen in concert with IL-12 secretion, CD141<sup>+</sup> DCs induce the activation of CTL to which they are attracted by virtue of their secretion of XCL1, the only known ligand of the XCR1 receptor (20). While such responses are commonly initiated in the secondary lymphoid organs in response to inflammation, CD141<sup>+</sup> DCs have also been found in non-lymphoid tissues including the skin, lung, kidney, and liver (21, 22) where they constitute the most abundant subset (18). In these anatomical locations, CD141<sup>+</sup> DCs have been shown to perform an essential regulatory role in the steady-state in order to maintain tissue homeostasis. In the skin, for example, CD141<sup>+</sup> DCs have been shown to express a distinctive CD14<sup>+</sup> CD1a<sup>−</sup> CD207<sup>−</sup> phenotype and constitutively secrete the anti-inflammatory cytokine IL-10 (23). Their capacity for expansion of CD4<sup>+</sup> regulatory T cells (Tregs) *in situ* was shown to reinforce tissue homeostasis and actively antagonize local inflammatory responses (23). The tolerogenicity of tissue-resident CD141<sup>+</sup> DCs and their proven capacity for antigen cross-presentation may, therefore, provide a compelling rationale for their use in immunotherapies aimed at intervening in the progression of deleterious immune responses. Nevertheless, such plans have so far been confounded by the complexities of their distribution *in vivo*.

Although CD141<sup>+</sup> DCs may be isolated from peripheral blood, these cells are thought to represent immature precursors of their tissue-resident counterparts (21). Furthermore, they represent the smallest subset of DCs in the peripheral circulation, constituting 0.03% of mononuclear cells. Consequently, a single leukapheresis has been estimated to yield as few as 3 × 105 cells following purification, posing a significant barrier to their downstream clinical application (24). Various strategies have sought to overcome these limitations: culture of human CD34<sup>+</sup> hematopoietic progenitor cells with a cytokine cocktail supplemented with the aryl hydrocarbon receptor antagonist StemRegenin 1 (SR1) promoted the *ex vivo* expansion of CD141<sup>+</sup> DCs but showed no specificity for this subset, resulting in the simultaneous expansion of both plasmacytoid and CD1c<sup>+</sup> DCs (25). Using an alternative approach, Ding and colleagues showed that NOD/SCID mice humanized using hematopoietic stem cells purified from cord blood, responded to administration of FLT3- Ligand by the generation of large numbers of both CD1c<sup>+</sup> and CD141<sup>+</sup> DCs (24). Nevertheless, such an approach is impractical for the purposes of scale-up and is incompatible with the generation of autologous cells, essential for their application to the induction of tolerance. Furthermore, the administration of FLT3- Ligand to healthy volunteers as a way of accessing autologous material resulted in the preferential expansion of cells expressing CD1c (26). Given the potential therapeutic benefits of harnessing the immunoregulatory properties of steady-state CD141<sup>+</sup> DCs, we have, therefore, sought to overcome their paucity in peripheral blood and difficulties in their expansion from precursors *ex*  *vivo*, by directing their differentiation from established lines of pluripotent stem cells.

We have previously demonstrated the feasibility of differentiating populations of primary DCs from both mouse and human embryonic stem cells (ESCs) (27, 28), thereby offering access to potentially unlimited numbers of cells, amenable to quality control. The advent of induced pluripotency and the derivation of induced pluripotent stem cells (iPSCs) under cGMP conditions (29) has led various groups to adapt our protocols developed using ESCs to the differentiation of DCs from human iPSCs (30, 31): nevertheless, such populations of iPSC-derived DCs (ipDCs) appear to belong predominately to the CD1c<sup>+</sup> subset (31). We have, therefore, recently optimized our protocols for use with patient-specific iPSCs and have reported the directed differentiation of DCs which, in addition to CD1c<sup>+</sup> cells, include a substantial population of CD141+ DCs capable of the crosspresentation of melanoma antigens to naïve peripheral blood T cells (32). Given the tractability of iPSCs for genome editing, this novel source offers opportunities for the introduction of subtle phenotypic or functional traits that might enhance the utility of the downstream cell therapy product or gain insight into aspects of the biology of this rare and inaccessible cell type in humans (2). Indeed, Sontag and colleagues used CRISPR/ Cas9-mediated genome editing of human iPSCs to generate a cell line deficient in the interferon regulatory factor 8 (IRF8) transcription factor and showed that differentiation of CD141<sup>+</sup> DCs was selectively compromised, while production of the CD1c<sup>+</sup> subset was largely preserved, providing clear evidence for a critical role for IRF8 in guiding lineage commitment toward the cross-presenting DC subset (33).

Given that iPSCs may serve as a source of autologous CD141<sup>+</sup> DCs, we have investigated whether this novel population might also show utility in pacifying deleterious immune responses under steady-state conditions and whether a tolerogenic phenotype may be further reinforced by exposure to defined pharmacological agents. Here, we describe in detail the protocols we have developed for the *in vitro* differentiation of CD141<sup>+</sup> DCs from human iPSCs, together with their subsequent enrichment and characterization. Their responsiveness to pharmacological agents known to reinforce the tolerogenic phenotype suggests new avenues for their use in the treatment of numerous disease states requiring the induction of immunological tolerance.

### OVERVIEW OF THE PROCEDURE

The use of human iPSCs as a novel source of potentially tolerogenic DCs expressing CD141 involves three distinct phases: (i) progressive differentiation of iPSCs *via* early mesoderm, through cells of the hematopoietic lineage, to committed DC precursors, (ii) modulation of the resulting ipDCs to reinforce their intrinsic tolerogenicity, and (iii) enrichment of the CD141<sup>+</sup> subset. **Figure 1A** illustrates the timelines involved, together with the cytokine cocktail required to effect each stage of the differentiation pathway. In summary, iPSCs are expanded in culture during routine passage until the approximate number of cells required for differentiation is achieved. The iPSCs are harvested at 80–85% confluency (**Figure 1B**, top left) using 0.02% ethylenediaminetetraacetic acid (EDTA) and mechanical scraping to generate small colony fragments. These are subsequently plated in ultra-low attachment (ULA) plates in mTesR1 medium supplemented with recombinant human Granulocyte Macrophage Colony-Stimulating Factor (rhGM-CSF), Bone Morphogenetic Protein 4 (rhBMP-4), Stem Cell Factor (rhSCF), and Vascular Endothelial Growth Factor (rhVEGF). The cultures are fed routinely every 2–3 days with differentiation medium consisting of XVIVO-15 supplemented with the appropriate cytokines. Guided differentiation of the cells is achieved by the stepwise withdrawal of growth factors, starting with BMP-4 on day 5, VEGF on day 14, and SCF on day 19 of culture, leaving only GM-CSF to sustain DC precursors and immature DCs, whose terminal commitment is subsequently reinforced by the addition of IL-4.

Using this protocol, clusters of differentiating iPSCs may be observed on day 3 of culture, where they later give rise to structures known as embryoid bodies, which imperfectly recapitulate some of the earliest stages of embryogenesis (**Figure 1B**, top center). Around days 14–16, macrophage-like cells may be observed in the differentiation cultures. Upon appearance of these cells, the medium is supplemented with IL-4, the concentration of which increases progressively with each subsequent feed, starting with 25 ng/ml and increasing to a maximum concentration of 100 ng/ml. DC precursors and immature DCs accumulating around embryoid bodies (**Figure 1B**, top right) are normally harvested between days 21 and 26 by gentle pipetting and are subsequently plated on cell-bind plates in XVIVO-15 medium supplemented with GM-CSF and IL-4 alone (**Figure 1B**, bottom left). Under these conditions, any contaminating macrophages adhere strongly to the plastic, while immature DCs remain in suspension and are recognizable by their cytoplasmic protrusions (**Figure 1B**, bottom center), which tend to become more prominent over time (**Figure 1B**, bottom right).

In order to promote a tolerogenic phenotype, cultures of ipDCs are further supplemented with pharmacological agents previously proven to modulate the function of human moDCs, such as rapamycin, dexamethasone, VD3, or the anti-inflammatory cytokine IL-10 (11). While VD3 is added to cultures on days 0 and 3 after harvesting, dexamethasone, rapamycin, and IL-10 are added from day 3 onward. After 5 days, ipDCs may be additionally matured by exposure to a cocktail of inflammatory cytokines for 48 h, after which they may be harvested by gentle pipetting to resuspend the lightly adherent cells. The purity of cDCs obtained using our protocol may be determined as a function of CD11c expression using standard flow cytometry (**Figure 2A**). Although the proportion of CD11c<sup>+</sup> cells may vary significantly between experiments, in our hands, the median percentage of cells expressing CD11c in 16 consecutive experiments was 85.5% (**Figure 2B**). However, in experiments yielding a purity below 60% (**Figure 2A**), cDCs may be enriched by labeling with monoclonal antibodies specific for CD11c and using magnetic bead separation techniques to isolate the labeled cells (**Figure 2C**).

Our attempts at purification of CD141<sup>+</sup> ipDCs using protocols for their positive selection have been hampered by significant levels of cell death following cross-linking of CD141. To avoid this issue, CD141<sup>+</sup> cells may be enriched by depletion of CD1c<sup>+</sup>

illustrating the morphology of colonies and individual cells during the differentiation process. Top left: colony of iPSCs cultured on matrigel in mTeSR-1 medium showing optimum morphology. Top center: early embryoid bodies on day 3 of culture on ultra-low attachment (ULA) plates in mTESR-1 supplemented with the full combination of growth factors. Top right: DC precursors at day 22 of culture accumulating around a single EB, from which they were originally released. Bottom left: DC precursors following harvesting onto cell bind plates to permit the adherence of macrophage-like cells. Bottom center and right: high magnification photomicrographs of fully differentiated ipDCs displaying characteristic DC morphology consisting of protrusions and veils of cytoplasm.

cDCs, with which they share a common progenitor (13). Removal of CD1c+ cells from cultures may likewise be achieved by separation using magnetic microbeads (**Figure 2D**). Protocols for each phase of the differentiation process outlined above, together with the reagents required, are described in detail below.

### MATERIALS

### Cell Lines

Protocols for the maintenance and passage of existing iPSC lines are now well-established and have been reported in detail elsewhere (34). While we describe here the directed differentiation of CD141<sup>+</sup> DCs from human iPSCs displaying some of the properties of the CD141<sup>+</sup> subset described *in vivo*, the outcome of the protocols we describe is entirely dependent on the quality and status of the parent cell line: failure to maintain iPSCs under optimal conditions may have adverse effects on their subsequent differentiation capacity and may result in the progressive accumulation of mutations or karyotypic abnormalities for which the culture conditions may serve as a selection pressure. It is advisable, therefore, to submit cells for routine karyotyping and to replace cell cultures with an earlier passage, should abnormalities be observed that might threaten the integrity of the iPSC line. While our original experiments made use of the human iPSC line C15 derived from human dermal fibroblasts (35) (a kind gift from Lee Carpenter and Suzanne Watt, University of Oxford), the reproducibility of our data has since been verified using numerous human iPSC lines derived from both healthy volunteers and patients suffering from various disease states.

Figure 2 | Purification of CD11c+ and CD141+ subsets of iPSC-derived DC (ipDCs) by magnetic bead separation. (A) FACS plot showing typical forward (FSC) and side scatter (SSC) of ipDCs obtained at the end of the differentiation procedure and the proportion of CD11c+ cells, which would normally suggest the need for their subsequent purification. (B) Percentage of CD11c+ cells obtained from 16 independent experiments. Each symbol represents an individual experiment, while the black line denotes the median (median = 85.55; SD = 11.94). (C) Enrichment of ipDCs from cultures containing lower proportions of CD11c+ cells: ipDCs were labeled with CD11c-biotin and purified using anti-biotin microbeads. CD11c expression is shown before and after purification. (D) Enrichment of "untouched" CD141+ ipDCs by depletion of CD1c+ cells using microbead separation. Co-expression of CD11c and CD141 is shown before and after enrichment, the quadrants being set according to non-specific staining with appropriately matched isotype controls. FACS plots are representative of three independent experiments.

IMPORTANT! For all studies involving human subjects, ethical approval should first be sought from the appropriate ethical review body. In the United Kingdom, recruitment of patients requires approval from the local NHS National Research Ethics Service (NRES) and should only be conducted following the receipt of informed consent.

### Reagents

### Cell Culture Media and Reagents


### Extracellular Matrix

• Matrigel (BD, cat. no. 356231)

### Cell Detachment


### Cytokines and Growth Factors


### Antibodies and Microbeads


### Equipment


### SET-UP

### Materials

### Differentiation Medium

Differentiation medium is composed of XVIVO-15 supplemented as outlined in **Table 1**. Since human iPSCs may be adversely affected by the routine use of antibiotics, their addition to the medium should be avoided if possible. Consequently, it is essential to rigorously maintain sterile technique when culturing and passaging iPSCs and to filter sterilize medium after the addition of individual components.

### Matrigel Stock

Thaw a 10 ml vial of matrigel on ice (this may take up to 4 h). Once thawed, add an equal volume of ice-cold Knockout DMEM to the matrigel solution using a pipette that has been kept at 4°C. Aliquot 1 ml into 1.8 ml Eppendorf tubes, placed on ice. Store at −80°C.

IMPORTANT! Note that matrigel rapidly solidifies above 4°C and must, therefore, be kept on ice at all times. To minimize solidification while aliquoting, all tips, pipettes, vials, and racks should be cooled to 4°C prior to use. During the aliquoting


procedure, vials must be kept on ice and transferred to −80°C for storage as soon as possible.

### 0.02% EDTA Solution (wt/vol)

Dissolve 1 g of EDTA in 500 ml of PBS to make 0.2% solution. Add 6N NaOH dropwise, while stirring until EDTA has dissolved. If necessary, adjust pH to 7.0 using 1M HCl. Autoclave to sterilize. Prepare a 1:10 dilution in PBS to make 0.02% solution of EDTA.

### Y-27632 [Rho-Associated Kinase (ROCK) Inhibitor]

Dissolve 1 mg of Y-27632 in 314 µl of PBS (pH 7.2) to prepare a 10 mM stock solution. Store at −80°C for up to 3 months. This solution should be diluted 1:1,000 to yield a final concentration of 10 nM.

### Granulocyte Macrophage Colony-Stimulating Factor (GM-CSF)

Dissolve 50 µg of lyophilized rhGM-CSF in 1 ml sterile PBS + 0.1% HSA to produce a stock solution of 50 µg/ml. Aliquot into 50 µl and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Vascular Endothelial Growth Factor (VEGF)

Dissolve 50 µg of lyophilized rhVEGF in 1 ml of sterile PBS + 0.1% HSA to produce a stock of 50 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Stem Cell Factor (SCF)

Dissolve 50 µg lyophilized rhSCF in 1 ml sterile PBS + 0.1% HSA to produce a stock solution of 50 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Bone Morphogenetic Protein 4 (BMP-4)

Dissolve 50 µg of lyophilized rhBMP-4 in 500 µl sterile PBS + 0.1% HSA to produce a stock solution of 100 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Prostaglandin E2 (PGE2)

Dissolve lyophilized PGE2 in DMSO to produce a stock solution of 5 mg/ml. Aliquot into 100 µl aliquots to serve as a 10× stock solution and store at −80°C. When required, thaw a single aliquot and dilute in 900 µl of PBS + 0.1% HSA to produce a working stock of 500 µg/ml.

IMPORTANT! DMSO is toxic and can penetrate the skin. Direct contact should, therefore, be avoided by wearing appropriate gloves.

### IL-1**β**

Dissolve 50 µg rhIL-1β in 1 ml sterile PBS + 0.1% HSA to produce a stock of 50 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Interferon-**γ**

Dissolve 1 mg IFN-γ in 2.8 ml sterile PBS + 0.1% HSA and dilute solution to 10 ml with PBS + 0.1% HSA to produce a stock solution of 100 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Tumor Necrosis Factor-**α**

Dissolve the lyophilized rhTNF-α in sterile PBS + 0.1% HSA to produce a stock solution of 100 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Interleukin-4

Dissolve 500 µg of lyophilized rhIL-4 in 500 µl sterile PBS + 0.1% HSA. Dilute to 5 ml with PBS + 0.1% HSA to produce a stock of 100 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### Rapamycin

Dissolve in 99% ethanol to produce a stock of 1 mg/ml and distribute into 100 µl aliquots. Store at −20°C and avoid freezethaw cycles. For further use after thawing, dilute a 100 µl vial with 900 µl of differentiation medium to produce a working stock that should be further diluted 1:1,000 in cultures to yield a final concentration of 100 ng/ml.

### 1**α**,25-Dihydroxyvitamin D3 (VD3)

Dissolve in 99% ethanol to produce a 100 µM stock. Distribute into 10 µl aliquots and store at −20°C. Avoid freeze-thaw cycles.

### Dexamethasone

Dissolve in 99% ethanol to produce a 10 mM stock. Distribute into 50 µl aliquots and store at −80°C avoiding freeze-thaw cycles.

### Interleukin-10

Dissolve in sterile PBS + 0.1% HSA to produce a stock solution of 50 µg/ml. Aliquot into 50 and 100 µl aliquots and store at −80°C. Avoid freeze-thaw cycles.

### 2-Mercaptoethanol

Dilute 70 µl of 2-ME in 20 ml of PBS to make a 1,000× stock solution.

IMPORTANT! 2-ME is highly toxic. Avoid inhalation and all contact with skin. Always use a fume hood to prepare a stock solution.

### Rinsing Buffer

Dilute stock EDTA in PBS to yield a final concentration of 0.02% (w/v). Keep buffer refrigerated at 4°C and place on ice while in use.

### Column Buffer

Prepare a 0.5% (w/v) solution of BSA in 0.02% (w/v) EDTA by carefully sprinkling the powder onto the surface and allowing it to dissolve slowly over time. Keep buffer refrigerated at 4°C and place on ice while in use.

### Equipment

### Preparation of Matrigel-Coated 6-Well Plates

For the routine passage of human iPSCs, 6-well tissue culture plates may be coated with matrigel prior to use as follows and stored at 4°C for up to 2 weeks:


IMPORTANT! Since matrigel solidifies above 4°C, it must be kept on ice at all times. To minimize solidification during the coating procedure, all tips, pipettes, vials, and racks used must be cooled to 4°C prior to use. Knockout DMEM must also be kept at 4°C. Take care to avoid creating bubbles, as these can result in uneven coating of the wells.

### DETAILED PROTOCOL OF THE PROCEDURE

### (A) Expansion of human iPSCs (6–7 days)

	- (i) Add Y-27632 to mTesR1 to produce a final concentration of 10 µM. Filter sterilize using a 0.22 µm syringe filter.

 IMPORTANT! Human iPSCs are especially sensitive to dissociation into a single cell suspension. The ROCK inhibitor Y-27632 has been shown to protect cells from dissociation-induced apoptosis (36) and is, therefore, routinely added during passaging. Y-27632 may also be added to medium upon thawing of the iPSC line in order to enhance viability but should be removed once the cells have adhered to matrigel.

	- (iv) Gently transfer the cell clusters suspended in mTesR1 into a 50 ml falcon tube using a 10 ml pipette. Wash the well with 1 ml of mTesR1 containing Y-27632 to collect any remaining clusters of cells. Cell clusters from the same passage can be pooled together from multiple wells.
	- (v) Top up the tube containing the clusters with an appropriate volume of mTesR1 containing Y-27632 to achieve a 1:12 dilution of the original cell suspension.
	- (vi) Pipette the suspension gently to ensure that clusters do not settle to the bottom of the tube and dispense 2 ml into each well of a fresh matrigel-coated 6-well plate, gently agitating the plate to ensure even distribution of the clusters.
	- (vii) Incubate the plate in a humidified incubator at 37°C, 5% CO2.

 IMPORTANT! The iPSCs should be expanded until the number of wells required for differentiation is achieved. As a rough estimate, three wells of iPSCs in a 6-well plate provide sufficient material to establish a single well of embryoid bodies in a 6-well ULA plate.

### (B) Set-up of cultures for the differentiation of DCs (45 min)

All reagents used to establish differentiation cultures should be maintained at room temperature.

	- (i) Aspirate the culture medium from a single well of iPSCs. Wash the well with 1 ml of PBS. Remove the PBS and add 1 ml of TrypLE express. Incubate the plate at 37°C for 5 min or until dissociated into a single cell suspension.
	- (ii) Add 1 ml of Knockout DMEM to the well and fully dissociate the cells by pipetting up and down with a Gilson pipette and a 1,000 µl pipette tip.
	- (iii) Transfer the cells to a 1.8 ml Eppendorf tube. Mix 20 µl of cell suspension with 20 µl of trypan blue and count the number of cells using a standard

hemocytometer. Calculate the total number of cells in one well.


 IMPORTANT! Differentiation cultures are initially established in mTesR1 medium, to which the cells have become accustomed during routine culture. Setting up the cultures in XVIVO-15 differentiation medium causes substantial cell death and may lead to failure of the differentiation process. XVIVO-15 is, therefore, introduced gradually by using it to replace mTeSR1 during routine feeding of the cultures, a process which appears to be better tolerated by iPSCs.


## (C) Maintenance of differentiation cultures (30 min every 2 days)

Differentiation cultures should be fed regularly every 2–3 days from day 2 of culture until they are harvested around days 21–24. However, if the culture medium consistently shows signs of exhaustion, the frequency of feeding should be increased. Toward the end of the differentiation, cultures are fed every 2 days. The growth factors in the differentiation medium are removed progressively until only GM-CSF remains, causing the concentration of each growth factor to decrease through the course of differentiation, according to a pre-defined schedule (**Figure 1A**).


 IMPORTANT! Cultures may contain a significant amount of cell debris during the early stages of differentiation, which is entirely normal.

(5) Around days 10–14 of culture, the appearance of small, round cells of hematopoietic origin should be observed. From days 14 to 18, macrophages with characteristic "fried egg" morphology and firm adherence to the tissue culture plate may start to appear. Numbers of macrophages may vary significantly between differentiations, even when using the same iPSC line. At the point of their appearance, add rhIL-4 to the differentiation cultures. IL-4 is introduced gradually, starting at 25 ng/ml which, in subsequent feeds, can be increased to 50, 75, and finally, 100 ng/ml as the number of DC precursors and immature DCs begins to increase.

### (D) Harvesting DC precursors and immature DCs (25 min)

(1) By days 21–24, significant numbers of DC precursors and immature DCs should be visible in the wells, frequently forming a "halo" surrounding individual embryoid bodies (**Figure 1B**, top right). Adherent macrophages may also be visible, although the addition of rhIL-4 appears to limit their numbers while further promoting the differentiation of DCs.

 IMPORTANT! The timing of events may vary significantly between experiments and even between wells cultured in parallel as part of the same experiment. Although we routinely harvest cultures around day 24, it is not uncommon to wait until day 30 for sufficient DC precursors and immature DCs to be available for harvesting.


 IMPORTANT! Cell-bind plates are used at this stage to encourage any macrophages that may have been carried over during harvesting to adhere.

### (E) Pharmacological modulation and maturation of ipDCs (7 days)

	- (i) For modulation with VD3, add 4 µl of 100 µM VD3 stock to each well containing 4 ml of medium to produce a final concentration of 100 nM. VD3 is added on days 0 and 3 following the plating of DCs onto cell-bind plates.
	- (ii) For modulation with dexamethasone, add 40 µl of 10 mM dexamethasone stock to each well containing 4 ml of medium to produce a final concentration of 100 µM. Dexamethasone is added on day 3 of culture following the transfer of DCs to cell-bind plates.
	- (iii) For modulation with rapamycin, dilute the 1 mg/ ml stock solution 1:1,000 in medium, of which

Table 2 | Pharmacological agents used for modulating dendritic cell function, showing their stock concentrations, final concentrations in culture, and the days on which they are added.


40 µl are added to each well in order to produce a final concentration of 10 ng/ml. Rapamycin is added on day 3 following the harvesting of DCs onto cell-bind plates.

	- (i) Determine the number of wells to be matured and for each well transfer 0.5 ml of XVIVO-15 differentiation medium containing 50 ng/ml rhGM-CSF and 100 ng/ml rhIL-4 to a 15 ml falcon tube.
	- (ii) Add rhTNF-α, PGE2, rhIL-1β, and rhIFN-γ to the medium to produce a stock 9 times the final concentration required. Filter sterilize using a 0.22 µm syringe filter.
	- (3) After 48 h, harvest immature and mature DCs by gently pipetting cultures up and down using a 10 ml pipette with the pipette controller set to low speed, so as to remove non-adherent and weakly adherent DCs, while leaving behind firmly adherent macrophages.

### (F) Purification of CD11c+ ipDCs (1.5 h)

	- (i) Pass the cells through a 70 µm cell filter to remove debris and clusters of cells.
	- (ii) Centrifuge DCs at 300 *g* for 10 min and carefully aspirate and discard the supernatant.
	- (iii) Resuspend 107 cells in 100 µl of column buffer and transfer them to a sterile Eppendorf tube.
	- (iv) Add 10 µl of biotinylated CD11c monoclonal antibody to 100 µl of cell suspension. Mix gently by pipetting up and down several times and incubate at 4°C for 10 min, either by placing on ice or in a refrigerator. If cell yields exceed 107 cells, scale up the volumes of buffer and antibody accordingly.

 IMPORTANT! Work quickly and keep the cells cold to prevent capping and shedding of bound antibody.

	- (i) Resuspend the cell pellet in 80 µl of column buffer and add 20 µl of anti-biotin microbeads to 107 cells. If working with more cells, scale up the volumes of buffer and microbeads accordingly.
	- (ii) Mix the cell suspension and microbeads by gently pipetting up and down several times and incubate at 4°C for 15 min, either by placing on ice or in a refrigerator.
	- (iii) Add 1 ml of cold column buffer to the tube to wash the cells. Centrifuge the cell suspension at 300 *g* for 10 min. Discard the supernatant and repeat this step 2 further times.
	- (iv) Resuspend the cell pellet in 500 µl of cold column buffer.
	- (v) Place a fresh MS column, with a maximum capacity of 107 cells, in the magnetic field of a magnetic separator.
	- (vi) Pass 500 µl of rinsing buffer through the column.
	- (vii) Add the 500 µl of cell suspension to the column and collect flow-through in a 15ml falcon tube: this represents the unlabeled cell fraction.
	- (viii) Wash the column by allowing 500 µl of rinsing buffer to flow through while it is still attached to the magnetic separator. Discard the eluent.
	- (ix) Remove the column from the magnetic separator. Add 1 ml of rinsing buffer to the column and immediately flush out the microbead-labeled cells by gently depressing the plunger. Collect the eluent in a fresh 15 ml falcon tube. This represents the purified fraction of CD11c<sup>+</sup> cells.

IMPORTANT! It is advisable to assess the purity of the population, preferably using flow cytometric analysis. Fluorescently labeled streptavidin will displace microbeads from the surface of the cells since its affinity for biotin is orders of magnitude higher than that of the anti-biotin monoclonal antibody, thereby permitting the percentage of CD11c<sup>+</sup> cells to be determined. Typically, a single round of purification yields a population enriched to ~90% purity (**Figure 2C**).

### (G) Purification of ipDCs by depletion of CD1c+ cells (1.5 h)

(1) After harvesting, estimate the total number of DCs obtained from a differentiation using trypan blue exclusion, as described in B step 1 (iii).

	- (i) Pass the cells through a 70 µm filter to remove cell clumps.
	- (ii) Centrifuge DCs at 300 *g* for 10 min. Carefully aspirate and discard the supernatant.
	- (iii) Resuspend the cells to a density of 107 cells in 200 µl of cold column buffer and transfer to a 15 ml falcon tube.
	- (iv) Add 10 µl of the FcR Blocking Reagent provided in the kit and 10 µl of CD1c-biotin to the cell suspension. Gently mix the cells and antibody by pipetting up and down several times and incubate at 4°C, either on ice or in a refrigerator for 15 min. If working with more cells, scale up volumes of buffer and antibody accordingly.

IMPORTANT! Work quickly, keeping the cells cold and using pre-cooled solutions to prevent capping and shedding of bound antibody.

	- (i) Resuspend cell pellet in 400 µl of column buffer.
	- (ii) Add 10 µl of anti-biotin microbeads to a maximum of 107 cells. If working with more cells, scale up the volumes of buffer and antibody accordingly.
	- (iii) Mix the cell suspension and microbeads by aspirating up and down gently a couple of times and incubate at 4°C for 15 min, either on ice or in a refrigerator.
	- (iv) Add 4 ml of cold column buffer to the tube to wash the cells. Centrifuge the cell suspension at 300 *g* for 10 min and discard the supernatant. Repeat this step 2 further times.
	- (v) Resuspend cell pellet in 500 µl of cold column buffer.
	- (vi) Prepare the MS column as described in F steps 3 (v)–(vi).
	- (vii) Add the cell suspension to the column and collect the flow-through in a 15-ml falcon tube.
	- (viii) Wash the column by adding 500 µl of rinsing buffer to the column, while it is still attached to the magnetic separator. Collect the eluent in a fresh 15 ml falcon tube and combine with the eluent from step (vii). This represents the unlabeled cell fraction that contains the CD141<sup>+</sup> cells.
	- (ix) Remove the column from the magnetic separator and flush out the microbead-labeled cells by gently depressing the plunger. This fraction contains the CD1c<sup>+</sup> subset.

IMPORTANT! It is advisable to assess the purity of either population before use, preferably by flow cytometry. A single round of negative selection typically enriches the CD141<sup>+</sup> subset to >70% purity (**Figure 2D**). Although this may be further improved by additional rounds of separation, such purity is generally at the expense of cell yields, which may decrease substantially. It is essential, therefore, to determine the cell numbers and level of purity required for each application and plan experiments accordingly.

## TIMING

Timings will vary depending on the magnitude of the differentiation culture and are, therefore, expressed as the time required for the handling of a 6 well plate.

	- Steps 3–5: 30 min every 2 days
	- Step 2: 10 min Step 3: 5–10 min
	- Step 1: 10 min
	- Step 2: 35 min Step 3: 45 min

### EXPECTED RESULTS

The application of our protocols to human iPSCs typically yields DCs displaying some of the features of the CD141<sup>+</sup> subset within approximately 24 days of culture, that may be enriched through negative selection of CD1c<sup>+</sup> cells to yield an "untouched" population, unaffected by cross-linking of surface CD141 (**Figure 2D**). Flow cytometry reveals that, in addition to CD141, these cells constitutively express TLR3 and the chemokine receptor XCR1 (**Figure 3A**) as reported previously (32), suggesting that they are analogous to the subset of DCs endowed with cross-presentation capacity (14–17, 32). Interestingly, these cells express barely detectable levels of CD1a and CD207, distinguishing them from dermal DCs and Langerhans cells, respectively, but consistently express both CD14 and CD209 (**Figure 3A**), recently found to be co-expressed by some populations of dermal DCs (37) and to define "regulatory" DCs in the skin (23), an indication that CD141+ ipDC may fail to perfectly recapitulate all properties of the conventional CD141<sup>+</sup> subset *in vivo*. Consistent with their capacity for antigen presentation, CD141<sup>+</sup> ipDCs

Figure 3 | Phenotypic and functional characterization of human iPSC-derived DC (ipDCs). (A) Histograms depicting the expression of lineage markers by ipDCs gated on the CD11c+ population. Levels of expression of individual markers are shown as red histograms, while non-specific staining by isotype controls is shown as gray histograms. (B) Maturation of ipDCs in response to a cocktail of inflammatory cytokines showing upregulation of MHC class II, CD40, CD54, and CD86 compared to immature cells. Red histograms represent levels of expression of markers by CD11c-gated ipDCs, while isotype controls are shown in black. All FACS plots are representative of three independent experiments. (C) ipDCs display increased immunostimulatory capacity upon maturation as evidenced by enhanced proliferation of naïve CD4+ T cells co-cultured in triplicate with mature (red line) or immature cells (black line). Data are representative of three independent experiments. Statistical significance was determined using the Mann–Whitney *U* test. (D) Interleukin (IL)-10 secretion by ipDCs in response to maturation stimuli. Immature ipDCs were cultured in triplicate with or without a cocktail of pro-inflammatory cytokines consisting of PGE2, tumor necrosis factor (TNF)-α, IL-1β, and interferon (IFN)-γ. Equivalent numbers of monocyte-derived DCs (moDCs) were cultured in parallel for 18 h and levels of IL-10 quantified from culture supernatants by standard ELISA. Plots are representative of at least three independent experiments. Statistical analysis was performed using parametric *T* tests with Welch's correction (\*\**p* < 0.01). (E) Secretion of IL-12p70 in response to immunological challenges. Immature ipDCs were cultured in triplicate either alone, with a cocktail of pro-inflammatory cytokines or with maturation cocktail further supplemented with toll-like receptor (TLR) agonists and soluble CD40L. Controls consisted of equivalent numbers of moDCs cultured in parallel. Levels of IL-12p70 were quantified from culture supernatants by standard ELISA. Data are representative of three independent experiments, and statistical analysis was performed using parametric *T* tests with Welch's correction. (\**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; \*\*\*\**p* < 0.0001).

constitutively express the co-stimulatory molecules CD40, CD54, and CD86, which are upregulated upon exposure to a cocktail of inflammatory cytokines (**Figure 3B**). MHC class II is likewise upregulated upon maturation, provoking the activation and proliferation of naïve, CFSE-labeled allogeneic T cells in the mixed leukocyte reaction (MLR) (**Figure 3C**). Similar to CD141<sup>+</sup> DCs in the skin (23), ipDCs constitutively secrete levels of IL-10 which are substantially higher than those produced by control populations of moDCs cultured in parallel (**Figure 3D**). Importantly, IL-10 has been shown to interfere with the initiation of Th1 responses (38) and to favor the polarization of naïve T cells toward a Treg phenotype (39, 40). Furthermore, upon maturation in response to inflammatory cytokines, ipDCs fail to secrete IL-12 required for Th1 polarization and CTL activation (41, 42), indeed, only maximal stimulation with a combination of inflammatory cytokines, TLR agonists and CD40 cross-linking is capable of eliciting significant IL-12 secretion (**Figure 3E**), reminiscent of reports of interstitial DCs including the CD141<sup>+</sup> subset (18, 21). Importantly, the equivalent treatment of moDCs in parallel cultures consistently yields substantially higher levels of the pro-inflammatory cytokine (**Figure 3E**).

CD141<sup>+</sup> DCs *in vivo* are characterized by their marked capacity for uptake and processing of both soluble and cellular antigens (19) and their chemotaxis in response to CCL19 and XCL1. Consistent with this remit, CD141+ ipDCs display significant capacity for the phagocytosis of fluorescently labeled latex beads which is abrogated upon fixation, more than 50% of immature cells being shown to phagocytose multiple beads over a 3 h incubation period, their propensity for phagocytosis decreasing following maturation, as previously reported (43, 44) (**Figure 4A**). DQ-OVA is a derivative of ovalbumin conjugated with boron-dipyrromethene, a photostable, pH insensitive dye which fluoresces following proteolytic cleavage (45) and therefore serves as a measure of antigen processing activity. Incubation of ipDCs expressing CD141 with DQ-OVA consistently reveals both significant uptake and processing of the substrate which is inhibited at 4°C and progressively lost upon maturation (**Figure 4B**).

In our hands, the migratory capacity of ipDCs is consistent with that reported previously for the CD141<sup>+</sup> subset, as determined using electrical cell-substrate impedance sensing. The xCELLigence Real-Time Cell Analyzer measures electrical impedance caused by the migration of cells through pores 8 µm in diameter in a filter, in which is embedded a gold microelectrode: the resulting arbitrary units of cell index provide a measure of the number of cells migrating across the filter in real time (46). Congruent with their expression of CCR7 (**Figure 3A**), ipDCs consistently migrate in response to a gradient of rhCCL19, known to guide DCs from interstitial tissues to the secondary lymphoid organs *in vivo* (47) (**Figure 4C**). Furthermore, ipDCs uniquely respond to the chemokine XCL1 in a dose-dependent manner (**Figures 4C,D**), confirming the functionality of surface XCR1. In contrast, moDCs cultured in parallel, respond reliably to CCL19 but do not migrate in response to XCL1 (**Figures 4E,F**), consistent with their failure to express the *XCR1* gene (16). *In vivo*, XCL1 is predominantly secreted by CD8<sup>+</sup> T cells and acts as a chemo-attractant that is highly specific for CD141+ DCs, thereby enhancing the crosspresentation of antigen to the MHC class I-restricted T cell repertoire (48).

Exposure to high levels of UV light promotes the local synthesis of VD3 within the skin which is known to be processed to its active form by resident DCs (49), potentially contributing to their regulatory function in the steady state. Accordingly, addition of VD3 to cultures of ipDCs, concomitant with their exposure to a maturation cocktail of proinflammatory cytokines, results in the further upregulation of CD14 (**Figure 5A**), the resulting CD14hiCD141<sup>+</sup> phenotype having been identified previously as indicative of regulatory function (23). Furthermore, ipDCs exposed to VD3 during differentiation show resistance to maturation, as evidenced by the failure to upregulate MHC class II and costimulatory molecules, while showing marked expression of the inhibitory receptors programmed death ligand-1 (PD-L1), PD-L2, and immunoglobulin-like transcript (ILT)-3 (50–53) (**Figure 5A**). Since tolerance is, in essence, an *in vivo* phenomenon, the tolerogenicity of ipDCs can be determined unequivocally only from the outcome of future clinical trials. Nevertheless, *in vitro* correlates have been shown to have predictive value, especially in mouse models in which allograft rejection has been prevented by the administration of "regulatory" DCs differentiated

Figure 4 | iPSC-derived DCs (ipDCs) show marked phagocytic and endocytic capacity and migrate in response to physiological stimuli. (A). Representative histograms and bar chart showing phagocytosis of 2 µm diameter fluorescently labeled beads by CD11c+ ipDCs over a 3 h incubation period (red histograms and bars). Non-specific binding of beads was assessed using fixed ipDCs (grey histograms and black bars). Data are representative of three independent experiments consisting of triplicate cultures. Statistical analysis was performed using parametric *T* test with Welch's correction. (B) Representative histograms and bar chart showing the endocytosis and proteolysis of DQ-OVA over a 30 min incubation period by CD11c+ ipDCs (red histograms and bars). Negative controls consisted of ipDCs incubated with DQ-OVA at 4°C (grey histograms and black bars). Data are representative of three independent experiments consisting of triplicate cultures. Statistical analysis was performed using parametric *T* test with Welch's correction. (C) Chemotaxis of ipDCs in response to CCL19 and XCL1 over a 4 h period measured in real time using the xCELLigence Real-Time Cell Analyzer. (D) Comparison of delta cell index (max–min of cell index) for each chemokine compared to negative controls, incubated in the absence of added chemokines. (E) Chemotaxis of control monocyte-derived DCs (moDCs) cultured in parallel, in response to CCL19 and XCL1 measured in real time over a 4 h period. (F) Comparison of delta cell index for cultures of moDCs. All plots are representative of three independent experiments. Data represent the mean ± SD. Statistical analysis was performed using parametric *T* tests with Welch's correction (\**p* ≤ 0.05; \*\**p* ≤ 0.001).

Figure 5 | VD3 treatment of iPSC-derived DCs (ipDCs) reinforces a regulatory phenotype. (A) Representative histograms showing the impact of VD3 on expression of cell surface markers including co-stimulatory molecules and the inhibitory receptors PD-L1, PD-L2, immunoglobulin-like transcript (ILT)3, and ILT4. Levels of expression of individual markers are shown as red histograms, while non-specific staining by appropriately matched isotype controls is shown in grey. (B) Reduced immunostimulatory capacity of ipDCs following exposure to VD3 (red line) compared with untreated controls (black line), as determined by proliferation of naïve CD4<sup>+</sup> T cells in the allogeneic mixed leukocyte reaction. (C) Polarization of naïve CD4+ T cells toward a regulatory T cell (Treg) phenotype following 5 days' co-culture with either untreated or VD3-treated ipDCs followed by the addition of 75 ng/ml rhIL-2 for a further 2 days. Treg commitment was assessed by the upregulation of FoxP3 and surface CTLA-4. (D) Polarization of naïve CD4+ T cells toward a Tr1 phenotype, characterized by secretion of interleukin (IL)-10. Mature untreated and VD3-treated ipDCs were co-cultured with CD4+ T cells for 5 days followed by a 2-day treatment with 75 ng/ml of rhIL-2. On day 7, co-cultures were treated with 10 µg/ml of Brefeldin A, 700 ng/ml ionomycin, and 20 ng/ml phorbol 12-myristate 13-acetate for 5 h before being stained for intracellular IL-10.

from iPSCs (54, 55). In these studies, the administered DCs showed decreased capacity for effector T cell priming *in vitro* and polarization of responding T cells toward a Treg phenotype. Accordingly, ipDCs conditioned by exposure to VD3 consistently display reduced stimulatory capacity in the MLR compared to untreated controls (**Figure 5B**). Furthermore, in co-cultures with naïve peripheral blood T cells, VD3-treated ipDCs promote a modest increase in commitment of responding T cells toward a Treg phenotype, defined by co-expression of FoxP3 and CTLA-4 (**Figure 5C**), but elicite a substantial increase in Tr1 cells (56–58), as evidenced by the appearance of T cells stained positively for intracellular IL-10 (**Figure 5D**).

Although VD3 has been used successfully to modulate the activity of moDCs for use in clinical trials (59), we consistently find that its use alters the morphology of ipDCs while substantially increasing their adherence to plastic, even when using ULA plates. Furthermore, the yield of ipDCs is significantly reduced in the presence of VD3 compared to cultures differentiated in its absence (**Figure 6A**), a finding which has prompted us to explore the use of other pharmacological agents known to induce a tolerogenic phenotype (11). Treatment with dexamethasone compromises both the yield and viability of ipDCs (**Figures 6A,B**), while rapamycin has little discernible effect on their propensity for Treg induction (**Figure 6C**). In contrast, IL-10 is compatible with acceptable yields and viability, while modestly enhancing the polarization of naïve allogeneic T cells toward a Treg phenotype (**Figure 6C**). Indeed, our results suggest that IL-10 may warrant further investigation as the agent of choice for reinforcing the tolerogenicity of ipDCs expressing CD141, either alone or in combination with a low dose of VD3, proposed as a conditioning regimen in forthcoming clinical trials of moDCs for the modulation of allograft rejection (8).

### POTENTIAL PITFALLS AND ARTIFACTS

While the advent of induced pluripotency provides an inexhaustible source of rare and inaccessible cell types with therapeutic potential, it is important to recognize the various drawbacks to their use. First, all cell types differentiated from pluripotent stem cells display a phenotype reminiscent of the fetal or neonatal period, a finding that has confounded the therapeutic use of cell types as diverse as hepatocytes (60) and cardiomyocytes (61). The "fetal" phenotype is especially evident among cells of the hematopoietic lineage: erythrocytes, for example, systematically fail to enucleate or progress beyond the expression of fetal hemoglobin to adult isoforms (62), greatly limiting their clinical utility. DCs differentiated from human ESCs or iPSCs likewise display hallmarks of a fetal phenotype: for instance, ipDCs secrete more abundant IL-10 than moDCs (**Figure 3D**) and fail to secrete IL-12, except in response to a combination of the most potent immunological stimuli (**Figure 3E**), a phenotype they share with moDCs isolated from neonates, which have been shown to actively repress expression of the p35 subunit of IL-12 (63, 64). Human fetal DCs have likewise been shown to suppress secretion of pro-inflammatory cytokines, additionally expressing arginase-2, whose capacity to deplete local l-arginine inhibits TNF-α secretion. Such a phenotype confers on fetal DCs the ability to induce abundant Treg cells, essential for the maintenance of maternal tolerance toward the developing fetus (65).

In addition to issues related to their unconventional provenance, the phenotype of numerous cell types differentiated from iPSCs has been shown to be influenced by the "epigenetic memory" they display for the cell type of origin, which may persist for many passages (66, 67). Given that human dermal fibroblasts remain the cell type of choice for reprogramming to pluripotency, as was the case for the C15 cell line described here (35), vestiges of the gene expression profile of the source cell type may confound the phenotypic analysis of differentiated cell types. In particular, many lineage-specific markers may be expressed at lower levels than anticipated for the equivalent cell type *in vivo*, a possible explanation for the low levels of expression of CCR7 and XCR1 by CD141<sup>+</sup> ipDCs (**Figure 3A**). Such findings emphasize the need for functional assays, such as chemotaxis, for characterization purposes (**Figures 4C,D**), rather than reliance on phenotype alone. This potential cause of artifacts is most evident in the context of MHC class II expression by ipDCs: given that dermal fibroblasts actively repress MHC class II expression, which is known to be epigenetically controlled (68), ipDCs differentiated from them have been shown in both mouse and man to express these molecules at unconventionally low levels (30, 69), albeit remaining responsive to maturation stimuli and at sufficient levels to fulfill their function as professional antigen presenting cells.

Together, vestiges of a fetal phenotype and the epigenetic memory of iPSCs suggest that few, if any, cell types differentiated from iPSCs are identical to their *in vivo* counterparts. The advent of single-cell RNA-seq that has proven such a powerful technique for clarifying lineage relationships between cell types of hematopoietic origin (70), may help to further illuminate the extent of similarity or difference between CD141<sup>+</sup> ipDCs and the conventional CD141<sup>+</sup> subset *in vivo* and determine whether a greater allegiance to the CD14<sup>+</sup> CD141<sup>+</sup> subset of "regulatory" DCs described by Chu and colleagues (23) can be substantiated

### REFERENCES


at the level of gene expression. While such caveats are important constraints when exploiting iPSCs to probe the molecular biology of precisely defined cell types through genome editing of the parent cell line (33), it need not undermine the therapeutic potential of the DCs differentiated from them which rely wholly on their functional capacity.

### ETHICS STATEMENT

Experiments reported in this study were carried out in accordance with the recommendations of the NRES Committee South Central—Oxford C (REC Reference: 09/H0606/5+5) following the receipt of written informed consent from all subjects in accordance with the Declaration of Helsinki.

### AUTHOR CONTRIBUTIONS

PF conceptualized, planned, and supervised the study. PS, AL, and TD performed the experiments and analyzed the data. PS, AL, and PF wrote the manuscript which was approved by TD.

### ACKNOWLEDGMENTS

The authors are grateful to Professor David Greaves and Drs. Asif Iqbal and Daniel Regan-Komito for their help with chemotaxis assays, Thirushan Wigakumar for his assistance with the preparation of the figures and to Dr Kumaran Shanmugarajah for helpful discussions.

### FUNDING

PS was the recipient of a Clarendon Scholarship and a Keith Murray Scholarship, while AL was supported by an MRC Capacity Building Studentship awarded to the Oxford Stem Cell Institute. This work was funded by grants from the Medical Research Council (UK) (Grant: G0802538), an MRC Confidence in Concept award (Grant: MC\_PC\_15029), and the Rosetrees Trust (Grant: A1372).

genotype-positive rheumatoid arthritis patients. *Sci Transl Med* (2015) 7:290ra87. doi:10.1126/scitranslmed.aaa9301


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as creating a potential conflict of interests. The derivation and use of CD141<sup>+</sup> DCs from human iPSCs is the subject of a patent application (WO2012/12720601) on which PF is a named inventor.

*Copyright © 2018 Sachamitr, Leishman, Davies and Fairchild. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

*Jordi Ochando\* and Mounia S. Braza\**

*Department of Oncological Sciences, Icahn School of Medicine at Mount Sinai, Immunology Institute, New York, NY, United States*

Donor-specific unresponsiveness while preserving an intact immune function remains difficult to achieve in organ transplantation. Induction of tolerance requires a fine modulation of the interconnected innate and adaptive immune systems. Antigen-presenting cells (APCs) predominate during allograft rejection and create a highly inflammatory context where allospecific T cells are primed. Currently, the available protocols to prevent allograft rejection include a cocktail of drugs that are efficient in the short-term, but with severe long-term side effects and considerable toxicity. Consequently, better and less burdensome strategies are needed to promote indefinite allograft survival. Targeted delivery of immunosuppressive drugs that prevent the alloimmune response may address some of these problems. Nanoparticle-based approaches represent a promising strategy to negatively modulate the alloresponse by specifically delivering small compounds to APCs *in vivo*. Nanoparticles are also used as integrating imaging moieties to monitor inflammation for diagnostic purposes. Therefore, nanotechnology approaches represent an attractive strategy to deliver and monitor the efficacy of immunosuppressive therapy in organ transplantation with the potential to improve the clinical treatment of transplant patients.

Keywords: nanoparticles, innate immune system, transplantation immunology, tolerance, therapeutics

## INTRODUCTION

Transplantation is a life-enhancing therapeutic option for tens of thousands of patients with end-stage organ failure. Outstanding short-term outcomes in organ transplantation have been achieved by pharmacologic immunosuppression. Despite these accomplishments, the detrimental effects' life-long continuous immunosuppression compromise long-term allograft survival (1, 2). Immunosuppressive combination therapies are not specific and often toxic, resulting in the deterioration of the patient quality of life and severe side effects, including infections and malignancies (3, 4).

Novel therapeutic approaches that target the adaptive immune response have been developed, but the long-term transplant outcomes remain suboptimal. This underlines the need for additional approaches to develop tolerance-inducing protocols. Allograft tolerance induction in murine models

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Pradyumna Kumar Mishra, ICMR-National Institute for Research in Environmental Health, India Andreina Schoeberlein, University of Bern, Switzerland*

#### *\*Correspondence:*

*Jordi Ochando jordi.ochando@mssm.edu; Mounia S. Braza mounia.braza@mssm.edu*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 29 September 2017 Accepted: 11 December 2017 Published: 22 December 2017*

#### *Citation:*

*Ochando J and Braza MS (2017) Nanoparticle-Based Modulation and Monitoring of Antigen-Presenting Cells in Organ Transplantation. Front. Immunol. 8:1888. doi: 10.3389/fimmu.2017.01888*

Frontiers in Immunology | www.frontiersin.org

**Abbreviations:** APC, antigen-presenting cell; DC, dendritic cell; FDA, Food and Drug Administration; IFN, interferon; IL, interleukin; HDL, high-density lipoprotein; HDL-NPs, high-density lipoprotein nanoparticles; PET, positron emission tomography; PLGA, poly(lactic-co-glycolic acid).

cannot be fully explained by mechanisms that target only the adaptive immunity (5, 6). Recent work revealed how the innate immune system, especially monocytes and macrophages, reacts to allogeneic non-self and critically influences the adaptive immune response (7–10). As a result, therapeutic approaches that target myeloid cells *in vivo* and deliver immunomodulatory agents that prevent activation of the adaptive immune response represents a largely unexplored approach to promote indefinite allograft survival.

In this mini review, we first discuss the current state and perspectives of nanotherapy in transplantation by focusing on nanoparticles, particularly for modulation and immunosuppressive drug delivery to antigen-presenting cells (APCs). We then introduce the synthetic high-density lipoprotein (HDL) nanoparticles (HDL-NPs), which represent an emerging and very promising nanotherapeutic option to be exploited in organ transplantation. In addition, we describe nanoparticle-based imaging approaches that are being evaluated for graft immune monitoring and transplant rejection diagnosis. We finally raise several outstanding questions about the use of nanoparticles in organ transplantation to conclude that this technology represents an additional therapeutic option to prevent transplant rejection and promote organ acceptance.

### APCs AS A THERAPEUTIC TARGET FOR IMMUNOSUPPRESSIVE THERAPY IN TRANSPLANTATION

Circulating and tissue-specific monocytes, macrophages, and dendritic cells (DCs) are APCs that activate strong cellular and humoral immune response against the transplanted organ. Non-self recognition by the innate immune system is certainly required for this response; however, it is still unclear what other mechanisms are involved in the early steps leading to APC maturation. It has been hypothesized that dying graft cells release "danger" molecules that directly induce APC maturation and that then initiate the adaptive alloimmune (11). Fadi Lakkis laboratory demonstrated that the "danger" signal associated with dying cells is not sufficient to initiate alloimmune response but that innate recognition of allogeneic non-self is required (9). By analyzing the innate immune response in either syngeneic or allogeneic grafts, it was demonstrated that only allogeneic grafts induced persistent differentiation of recipient monocytes into mature DCs that expressed interleukin 12 (IL-12) and stimulated T-cell proliferation and interferon γ (IFN-γ) production (9, 11). Altogether, these findings underline the importance of alloantigen innate recognition by APCs in initiating graft rejection and in maintaining a pro-inflammatory context. More recently, the Lakkis laboratory uncovered the mechanisms underlying non-self allorecognition and demonstrated that donor polymorphism in the gene encoding the signal regulatory protein α recognition by recipient CD47 elicits the innate immune response (12).

While monocyte-derived cell accumulation in transplanted organs has long been recognized as a feature of allograft rejection (13), recent data suggest that monocyte-derived macrophages inhibit graft-reactive immune responses (14) and mediate the induction of transplantation tolerance (10). This suggests that the functional properties (stimulatory or suppressive) of allograftinfiltrating APCs dictate the outcome of the transplanted organ. In this respect, circulating stimulatory (Ly-6Chi) monocytes contribute to leukocyte recruitment and consequently to acute organ rejection (15), while suppressive (Ly-6Clo) macrophages are responsible for the long-term allograft survival (10). These findings indicate that the innate immune system is not just an innocent bystander in the allograft immune response and that its modulation is required for tilting the immune balance in favor of the homeostasis status and of long-term allograft survival.

### NANOPARTICLE-BASED MODULATION OF APCs FOR TRANSPLANTATION TOLERANCE

Drug-loaded nanoparticles represent a promising tool in organ transplantation to circumvent the limitations of conventional approaches by a localized, sustained, and controlled delivery of bioactive agents. Engineering nanoparticles for modulating the innate immune system in transplantation is an emerging field that provides new insights into the basic immunobiology of graft rejection/tolerance. The therapeutic aim is to deliver antigens and immune modulatory agents through specific myeloid derived cell targeting, thus allowing a better control on the innate immune response to induce transplantation tolerance (**Figure 1A**).

Targeting DCs with nanoparticles harboring antibodies or small compounds is one of the most promising strategies to negatively regulate the immune response after transplantation. Delivering antigen to specific DC receptors may result in the production of regulatory cytokines and the induction of negative costimulatory pathways that promote tolerogenic responses. C-type lectin receptors that are responsible of antigen presentation, such as mannose receptor and DEC-205 (16), have been previously used for immune cell activation (17, 18). Interestingly, antigen delivery by the same nanoparticles in the absence of adjuvant induces suppressive immune responses, leading to a tolerogenic phenotype (19). This represents a potential strategy to inhibit activated CD4 and CD8 T cells that mediate transplant rejection. Furthermore, transplant recipient mice treated with nanoparticle-encapsulated immunosuppressive drugs, such as rapamycin, tacrolimus, and mycophenolic acid, prolong allograft survival. PLGA nanoparticles have been developed to deliver rapamycin to increase the suppressive activity of myeloid cells. The resulting nanoparticles have a better efficacy in comparison to free drug in terms of antiproliferative (20), and inhibitory effects on the maturation of DCs (21). In a mouse model of skin transplantation, Goldstein and colleagues successfully delivered mycophenolic acid loaded-PLGA nanoparticles to myeloid cells, which prolonged allograft associated with upregulation of programmed death ligand-1 (22). Using a similar mouse model of skin graft transplantation, treatment with a mixture of rapamycinand tacrolimus-loaded nanomicelles was shown to effectively target multiple immune cell subsets in the lymph node, with a prolonged allograft survival (23). Moreover, locally controlled

Figure 1 | Toward nanomedicine in transplantation. Conventional organ transplant treatment requires continuous immunosuppressive drugs to provide therapeutic benefit that results in several side effects, including toxicity. Due to their structural stability and gradual drug release capacity, nanoparticle-based strategies could be used to reduce drug doses, minimize toxicity, and induce long-term allograft tolerance. Among the nanomaterials currently being developed, many are studied as drug delivery and imaging agents. (A) Myeloid cells can be targeted by using nanoparticles (in green) with the aim of modulating the early steps of the immune response. Nanoparticles deliver immunosuppressive drugs and/or antigens that result in a tolerogenic environment through the upregulation of anti-inflammatory mediators, such as IL-10, TGF-β, and the Dendritic Cell-Specific Intercellular adhesion molecule-3-Grabbing Non-integrin (DC-SIGN), and the downregulation of pro-inflammatory mediators, such as TNF-α and CD40. This will promote the formation and maintenance of myeloid cells with suppressive activity to reduce the alloreactive T cell response and concomitantly induce regulatory T cells (Treg) and long-term tolerance. (B) Among the various polymers synthesized for formulating polymeric nanoparticles, poly(lactic-co-glycolic acid) (PLGA) is the most popular with several interesting properties such as controlled and sustained release, low cytotoxicity, biocompatibility with tissues and cells, and a targeted delivery. A schematic representation of PLGA-based nanoparticles (PLGA-NP) is included in this figure. The entrapped drug is distributed throughout the polymer matrix and the particles surface is covered with a cationic surfactant such as didodecyldimethylammonium bromide. (C) Nanoparticles can also be used as moieties for positron emission tomography (PET), magnetic resonance imaging (MRI), and X-ray imaging and monitor graft function in patients. These non-invasive imaging approaches could be applied for diagnosis and prognostic purposes.

and sustained release of corticosteroids using a biodegradable nanoparticle system after corneal transplantation prevents graft rejection in rats (24). In conclusion, these studies provide a comprehensive *in vitro* and *in vivo* evidence for the superiority of PLGA encapsulated immunomodulatory drugs over the soluble form and its potential in organ transplantation.

In summary, nanoparticles are used for the delivery of low dose immunosuppressive agents in conjunction with antigens to prevent specific immune responses. These studies mostly used nanotherapies based on the Food and Drug Administration-approved poly(lactic-co-glycolic acid) (PLGA) nanoparticles. While biodegradable, PLGA nanoparticles are large (≥100 nm in diameter), tend to aggregate, and are taken up by all phagocytic cells in a non-specific manner. The nanoparticle size is a critical factor for uptake and retention in the lymphoid secondary organs, since small nanoparticles (≤25 nm) are taken up more efficiently and retained for longer periods (25). Altogether, nanotherapeutic specific targeting of the APCs represents a promising strategy to inhibit the upstream steps of transplant rejection and to generate a durable donor-specific tolerant state.

### HDL-NPs AS NANOCARRIERS FOR DRUG DELIVERY TO APCs IN TRANSPLANTATION

High-density lipoproteins are natural, small dynamic nanoparticles that have immuno protective function through macrophage targeting (26). They are being exploited in atherosclerosis, as a nanotherapeutic option (27) and are also used for targeting tumor-associated macrophages and as a cancer therapeutic tool (28, 29). Since HDL-NPs exhibit high specificity toward myeloid cells, they deliver immune modulatory drugs to APCs *in vivo* (30). Apolipoprotein A-I (apoA-I) is the main amphipathic lipoprotein associated with HDL-NPs and defines the size and shape of these nanoparticles (28, 31). HDL-NPs preferentially interact with receptors that are highly expressed by myeloid cells, including ATP-binding cassette receptor A1 and scavenger receptor type B-1 (32). As a result, HDL-NPs represent an attractive approach to *in vivo* target myeloid cells in transplant recipients. Their ability to incorporate therapeutic agents can be used to specifically deliver immunosuppressive drugs to the innate immune system and prevent the immune alloreactivity, thus promoting longterm allograft survival. Since the biodistribution of HDL-NPss is tightly dependent on their composition, the number of apoA-I molecules, their purity, and ratio relative to other nanoparticle components, such as phospholipids, need additional investigation for optimal results. Ultimately, HDL-NPs synthesis should be adapted to each disease to provide the best and most specific tissue and cell targeting tool (33).

### NANOPARTICLES FOR APCs MONITORING IN TRANSPLANTATION: IMAGING APPROACHES

Besides their use as drug delivery carriers, nanoparticles can also be used to image a biological process. Pioneer approaches to imaging transplant rejection used radiolabeled anti-myosine antibody Fab fragments as a non-invasive detection of human cardiac transplant recipient rejection (34). Besides, magnetic resonance imaging (MRI) was used for repetitive imaging of transplanted hearts because it combines high spatial resolution with the ability to measure heart function while avoiding radiation exposure (35). Indeed, *in vivo* electrocardiographically gated MRI has been reported as a sensitive, non-invasive modality for the detection and the grading of cardiac transplant acute rejection, which correlates with the T2 relaxation times value (35). Even though gadolinium-based contrast agent play an important role in molecular and cellular imaging (36), most MRI cellular studies rely on the superior sensitivity of superparamagnetic or ultrasmall superparmagnetic iron oxide nanoparticles for imaging contrast. MRI-sensitive iron oxide approach exploits the phagocytic capacity of myeloid cells, specifically macrophages to monitor allograft rejection (37, 38).

More recently, nanoparticles were used to visualize macrophages *in vivo* and for assessing their absolute number, flux rate, and functional state in different tissues and models (39–41). Radiolabeled and dextran crosslinked nanoparticles have been used as a macrophage-specific imaging agent for positron emission tomography (39, 42, 43). Furthermore, magnetic nanoparticles could also be used as probes for MRI to examine the function of immune cells in humans. However, using MRI for *in vivo* cell quantitation in organs is often complicated and needs more concentrated magnetic materials than radiolabeled nanoparticles (44). PEGylated gold nanoparticles and other nanoparticle-based contrast agents have been used also for X-ray computed tomography (CT) (45). Differently from the other described imaging techniques, X-ray CT requires a high concentration of nanoparticles to follow the macrophage populations. Therefore, different nanoparticle platforms can be used in personalized clinical care to provide diagnostic and prognostic information as well as for quantifying the treatment efficacy of transplant patients (**Figure 1B**).

## CONCLUDING REMARKS AND OUTSTANDING QUESTIONS

The use of nanoparticles represents a promising therapeutic strategy to target APCs *in vivo* and negatively modulate the immune response in organ transplant recipients. Nanoparticles are capable to induce antigen-specific myeloid cells with suppressive function that promote regulatory T cells expansion (46). Therefore, the immunosuppressive effects of nanoparticles loaded with donor antigens are ultimately transplant- and patient-specific. In addition, assays that evaluate the robustness of this nanotherapeutic approach and potentially distinguish between tolerant and nontolerant patients need to be optimized. This could be in part be monitored using gene expression profiling of the patient's blood, urine, or transplant biopsy as previously reported. As the final clinical objective is to maintain graft function and intact host defenses, a patient-specific genetic tolerogenic signature could be used to determine the frequency and dose of the nanotherapeutic treatment of each patient.

Protocols using nanoparticles for imaging in transplantation need to be optimized for their clinical application as a non-invasive approach to characterize and monitor the allograft function. While some animal models are being developed that evaluate the efficacy of nanoparticles in organ transplantation, much work is yet to be done to translate the results from bench to bedside. In addition, the precise mechanisms of action and the long-term effects of nanoparticles have not been fully elucidated yet. Although drug-loaded nanoparticles have demonstrated lower toxicity than the soluble form, the potential long-term toxicity and side effects of nanoparticles are not fully known. Interestingly, drug-loaded nanoparticles could be used as a combination therapy with other induction therapy strategies, such as thymoglobulin and interleukin-2 (IL-2) receptor antibodies (47). In this respect, it is important to test whether combined approaches that use drug-loaded nanoparticles are optimized in a mechanism-independent fashion and to determine the potential synergistic effects. Collectively, the use of nanoparticles as a targeted delivery approach that modulates APCs *in vivo* represent an innovative therapeutic protocol to prevent undesirable immune responses and promote long-term organ acceptance in transplant recipients direct translation into the clinical practice.

## AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

### ACKNOWLEDGMENTS

This work was supported by the COST Action BM1305: Action to Focus and Accelerate Cell Tolerogenic Therapies (A FACTT),

### REFERENCES


the COST action BM1404: European Network of Investigators Triggering Exploratory Research on Myeloid Regulatory Cells (Mye-EUNITER), and the Mount Sinai Recanati/Miller Transplantation Institute career development funds. Part of the artwork present in **Figure 1** is adapted from Servier Medical Art (http://smart.servier.com) and used under a Creative Commons Attribution 3.0 Unported License (CC BY 3.0) current version 4.0 published on November 2013.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 Ochando and Braza. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Monitoring t-cell responses in translational studies: optimization of Dye-Based Proliferation Assay for Evaluation of Antigen-specific responses

### *Edited by:*

*John Isaacs, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Thomas Wekerle, Medical University of Vienna, Austria Milica Vukmanovic-stejic, University College London, United Kingdom*

#### *\*Correspondence:*

*Anja Ten Brinke a.tenbrinke@sanquin.nl*

#### *†Present address:*

*Maria P. Hernandez-Fuentes, UCB Pharma, Slough, United Kingdom*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 23 August 2017 Accepted: 08 December 2017 Published: 21 December 2017*

#### *Citation:*

*Ten Brinke A, Marek-Trzonkowska N, Mansilla MJ, Turksma AW, Piekarska K, Iwaszkiewicz-Grzes´ D, Passerini L, Locafaro G, Puñet-Ortiz J, van Ham SM, Hernandez-Fuentes MP, Martínez-Cáceres EM and Gregori S (2017) Monitoring T-Cell Responses in Translational Studies: Optimization of Dye-Based Proliferation Assay for Evaluation of Antigen-Specific Responses. Front. Immunol. 8:1870. doi: 10.3389/fimmu.2017.01870*

*Anja Ten Brinke1,2\*, Natalia Marek-Trzonkowska3 , Maria J. Mansilla4 , Annelies W. Turksma1,2, Karolina Piekarska3 , Dorota Iwaszkiewicz-Grzes*´ *<sup>5</sup> , Laura Passerini6 , Grazia Locafaro6 , Joan Puñet-Ortiz4 , S. Marieke van Ham1,2, Maria P. Hernandez-Fuentes7†, Eva M. Martínez-Cáceres4 and Silvia Gregori <sup>6</sup>*

*1 Department of Immunopathology, Sanquin Research, Amsterdam, Netherlands, 2 Landsteiner Laboratory, Academic Medical Centre, University of Amsterdam, Amsterdam, Netherlands, 3 Laboratory of Immunoregulation and Cellular Therapies, Department of Family Medicine, Medical University of Gdansk, Gdan* ́ *sk, Poland,* ́ *<sup>4</sup> Immunology Division, Department of Cellular Biology, Germans Trias i Pujol University Hospital and Research Institute, Physiology, and Immunology, Universitat Autònoma Barcelona, Barcelona, Spain, 5 Department of Clinical Immunology and Transplantology, Medical University of Gdansk, Gdan* ́ *sk, Poland,* ́ *6San Raffaele Telethon Institute for Gene Therapy (SR-Tiget), Division of Regenerative Medicine, Stem Cells and Gene Therapy, IRCCS San Raffaele Scientific Institute, Milan, Italy, 7MRC Centre for Transplantation, King's College London, London, United Kingdom*

Adoptive therapy with regulatory T cells or tolerance-inducing antigen (Ag)-presenting cells is innovative and promising therapeutic approach to control undesired and harmful activation of the immune system, as observed in autoimmune diseases, solid organ and bone marrow transplantation. One of the critical issues to elucidate the mechanisms responsible for success or failure of these therapies and define the specificity of the therapy is the evaluation of the Ag-specific T-cell responses. Several efforts have been made to develop suitable and reproducible assays. Here, we focus on dye-based proliferation assays. We highlight with practical examples the fundamental issues to take into consideration for implementation of an effective and sensitive dye-based proliferation assay to monitor Ag-specific responses in patients. The most critical points were used to design a road map to set up and analyze the optimal assay to assess Ag-specific T-cell responses in patients undergoing different treatments. This is the first step to optimize monitoring of tolerance induction, allowing comparison of outcomes of different clinical studies. The road map can also be applied to other therapeutic interventions, not limited to tolerance induction therapies, in which Ag-specific T-cell responses are relevant such as vaccination approaches and cancer immunotherapy.

Keywords: tolerance, monitoring, proliferation, antigen-specific, T cells, transplantation, autoimmune diseases, immune-therapies

### INTRODUCTION

The induction of antigen (Ag)-specific tolerance in transplanted or autoimmune disease patients is a pre-eminent goal in precision medicine. Progressively, several tolerance-inducing strategies are entering the clinical arena with immune-modulatory drugs, including novel therapeutic antibodies (1, 2) and cell therapies with regulatory T cells (Tregs) or tolerogenic Ag-presenting cells (tolAPCs).

**198**

Hence, the need for *in vitro* assays to evaluate the immunological mechanisms responsible for failure or success of these therapies is becoming critical. It may discriminate Ag-specific tolerance induction from general immune suppression and potential loss of pathogen-specific immunity. In addition, assessment of Ag-specific memory for tolerance may allow identification of patients in whom tapering of immunosuppression is likely to be safe, thus minimizing risks of adverse effects resulting from the ongoing treatments.

To evaluate Ag-specific responses *in vitro*, peripheral blood mononuclear cells (PBMCs) are the most widely used cells due to their relative convenient accessibility. Various methods to monitor Ag-specific responses have been developed, including measurement of cytokine production of Ag-responding T cells with enzyme-linked immunosorbent assay (ELISA) or enzymelinked immunospots (ELISpots), or analysis of T-cell proliferation based on 3 H-thymidine incorporation. Technical developments in the field of flow cytometry opened new possibilities for analysis and characterization of cell sub-populations and their Ag-specific responses using fluorescent dye dilution (3, 4) and flow cytometric assay of specific cell-mediated immune response in activated whole blood (FASCIA) (5, 6). Cell permeant dyes, such as carboxyfluorescein diacetate succinimidyl ester (CFSE), cell trace violet (CTV), and violet proliferation dye 450 (VPD-450), enabled more specific analysis of cell proliferation over several days. Since the dyes are divided equally between daughter cells (7), the number of cell divisions of the proliferating cells can be visualized, thus allowing the theoretical enumeration of Ag-specific cells. Moreover, dividing cells can phenotypically be characterized using antibodies specific for surface markers and/ or intracellular cytokines (8). This multiplies the information to be obtained from a single functional assay. It still remains to be defined, however, whether analysis with proliferation dyes is sensitive enough to evaluate the induction of tolerance in transplantation settings and autoimmunity, where the numbers of autoAg-specific cells are generally very low (9, 10).

Only recently, the use of dye proliferation to monitor Ag-specific T-cell responses has been introduced in clinical practice. Responsiveness to insulin in a small number of children in randomized clinical study, Pre-POINT study, demonstrated the value of combining proliferation dye with analysis for specific T helper profiles. The analysis demonstrated that the observed insulin- and pro-insulin-specific proliferating CD4<sup>+</sup> T cells acquired a Treg phenotype (11). Similarly, Ag-specific T-cell proliferation in response to Derp1 in small cohort of patients undergoing dust mite allergen-specific immunotherapy was used to demonstrate the ability of the treatment to promote unresponsiveness in allergen-specific T helper cells (12). These studies highlight the limitation of applying dye proliferation assay, as T-cell responses could only be evaluated in a fraction of treated patients. However, these examples also indicate that dye proliferation assay can be a valuable tool to better dissect the effect of a given therapy, since in combination with gene profile or phenotypical analysis (i.e., FOXP3 expression or intracytoplasmic staining for cytokines) can help to grasp the mechanism underlying tolerance induction.

Studies correlating transplant outcome with *in vitro* functional studies have been mostly non-conclusive [as reviewed in Ref. (13)]. In a trial of allo-specific tolerance induction, however, an absence of proliferation to the donor was observed in those patients that could continue with immunosuppression withdrawal (14). Besides in a study focused on finding a biomarker signature to detect renal transplant tolerance in humans the comparison of *in vitro* T-cell function between spontaneously tolerant kidney transplant recipients and non-tolerant recipients demonstrated that the best correlation to the clinical status was obtained with donor-specific IFNγ-ELISpot assays (15).

With the aim to join forces in development and implementation of tolerance-inducing cell products, such as Tregs and tolAPCs, a European network action to focus and accelerate cell-based tolerance-inducing therapies (A FACTT, www. afactt.eu) was initiated in 2014 under the umbrella of European Cooperation in Science and Technology (COST). By creating a forum for the exchange and integration of knowledge and expertise, A FACTT aims to minimize overlap and maximize comparison of the diverse tolerance-inducing cell products, but also to create consensus on monitoring parameters, immunemonitoring assays and establish minimum information models (16–18). Therefore, within A FACTT, we have determined the critical steps of a dye-based proliferation assay to monitor Ag-specific T-cell responses useful for assessing the results of tolerance-inducting therapies, since assay harmonization to monitor tolerance induction is essential to compare outcomes of different clinical studies.

In the current study, we propose a road map for the execution and analysis of dye-based proliferation assays for high-sensitivity monitoring of T-cell responses specific for alloAgs, pathogenderived exogenous Ags, and autoAgs. This approach will be of pivotal importance for defining effects of tolerance-inducing strategies for transplantation and autoimmune diseases.

### MATERIALS AND EQUIPMENT

### Subjects

Human peripheral blood was obtained from healthy donors upon informed consent in accordance with local ethical committee approval and with the Declaration of Helsinki.

Human peripheral blood was obtained from four multiple sclerosis (MS) patients from the Multiple Sclerosis Unit, Germans Trias I Pujol University Hospital (Badalona, Spain) upon

**Abbreviations:** A FACTT, a European network action to focus and accelerate cellbased tolerance-inducing therapies; Ag, antigen; allo, allogeneic; APC, antigenpresenting cell; auto, autologous; CFSE, carboxyfluorescein diacetate succinimidyl ester; cpm, counts per minute; COST, European Cooperation in Science and Technology; CTV, cell trace violet; DC, dendritic cell; DI, division index; ELISA, enzyme-linked immunosorbent assay; ELISpot, enzyme-linked immunospot; FASCIA, flow cytometric assay of specific cell-mediated immune response in activated whole blood; FBS, fetal bovine serum; GAD, glutamic acid decarboxylase; HS, human serum; OVA, ovalbumin; mAb, monoclonal antibody; MBP, myelin basic protein; MOG, myelin oligodendrocyte glycoprotein; MS, multiple sclerosis; PBMCs, peripheral blood mononuclear cells; Pen/strep, penicillin-streptomycin; PF, precursor frequency; PI, proliferation index; PLP, proteolipid protein; SEB, Staphylococcal enterotoxin B; SI, stimulation index; Tregs, regulatory T cells; tolAPCs, tolerogenic antigen-presenting cells; tolDC, tolerogenic dendritic cells; T1D, type 1 diabetes; VPD-450, violet proliferation dye 450.

informed consent in accordance with local ethical committee approval and with the Declaration of Helsinki. No patient had clinical exacerbations or was receiving corticosteroid or disease modifying treatments at the moment of the sample collection.

### Cell Preparations

Peripheral blood mononuclear cells were isolated from buffy coats obtained from healthy volunteer blood donor by Ficoll-Uropoline, Ficoll-Hypaque, or Lymphoprep gradient centrifugation and were either used fresh or after storing them in liquid nitrogen. CD4<sup>+</sup> T cells were separated by negative selection (StemCell Technologies or Miltenyi Biotec) according to manufacturer's instructions, with a resulting purity of >95%. Dendritic cells and CD3-depleted PBMC were prepared as previously described (19, 20).

### Dye-Labeling and Proliferation Polyclonal Stimulation

Responder cells were washed twice with warm (37°C) PBS to remove serum that affects staining. Then, cells were suspended in warm (37°C) PBS at a concentration of 5 × 106 cells/ml and labeled with various concentrations of CFSE (Invitrogen, USA) or VPD-450 (BD Biosciences, USA) at 37°C for 15 min. Each 5 min cells were vortexed to provide uniform staining. Subsequently, cells were washed with warm (37°C) PBS and then with culture medium (X-VIVO 20; Lonza) supplemented with 10% fetal bovine serum (FBS) and antibiotics pen/strep. After this step, cells were suspended in fresh medium (X-VIVO 20, 10% FBS, pen/strep) and incubated for 24 h. After this time, the labeled cells were collected, washed with fresh medium, counted, seeded on 96-well plates (1 × 105 cells/well), and stimulated with magnetic beads coated with anti-CD3 and anti-CD28 antibodies (Invitrogen) in 1:1 cell:bead ratio. In parallel, stained and not stimulated cells, as well as unstained and stimulated cells, were seeded in the same concentration as controls. After 96 h, cells were collected and stained with 7-AAD (20 min, at RT). Sample of unlabeled and stimulated cells was stained with anti-CD45 V450 mAb (BD Horizon, USA) or with anti-CD45-FITC mAb (BD Biosciences). After viability check, one well of CSFE or VPD-450 stained stimulated cells was mixed with one well of stained unstimulated cells and one well of unstained stimulated cells labeled with anti-CD45 V450 mAb or with anti-CD45 FITC mAb when CFSE or VPD-450 was used, respectively. Immediately after this step, cells were analyzed with flow cytometer (LSRFortessa; BD Biosciences).

### Staphylococcal Enterotoxin B (SEB) Stimulation

Thawed PBMCs (10 × 106 cells/ml) were incubated with 2 µM of VPD-450 (BD Biosciences) for 7 min at RT in the dark. Afterward, cells were washed twice with medium and resuspended in IMDM 5% human serum (HS) (Sanquin), pen/strep at a concentration of 1 × 106 cells/ml, plated in 24-well plate, and stimulated with 1 µg/ml SEB (Sigma Aldrich). After 4 or 5 days, cells were harvested, washed with PBS, and stained with near-IR dead cell stain (Thermo Fischer Scientific Inc.) for 30 min at RT in the dark. Subsequently, cells were stained with anti-CD3-BUV496, anti-CD4-Ag-presenting cell (APC), and anti-CD8-BUV805 (BD Biosciences, USA) and analyzed with flow cytometer (LSRFortessa; BD Biosciences). Data were analyzed using the FlowJo software (V10).

### AlloAg and Pathogen-Specific Ag Stimulation

Fresh PBMCs and purified CD4<sup>+</sup> T cells were labeled eFluor® 670 (10 µM) (eBioscience) and incubated for 10 min at 37°C in the dark. The labeling of cells was stopped by adding 4–5 volumes of cold FBS (Lonza) and incubating the cells on ice for 5 min. Then, cells were washed and resuspended in culture medium: X-VIVO15 medium with 5% HS (BioWhittaker-Lonza), pen/ strep (BioWhittaker). To evaluate the allo-specific proliferative response, labeled PBMCs or CD4<sup>+</sup> T cells were used as responder cells (R, 105 cells/well). As stimulators (S), either autologous (auto)/allogeneic (allo) CD3-depleted PBMCs (APC) at [R:S] ratio of [1:1] or auto/allo mature dendritic cell (mDC) at a [10:1] ratio were used. Cells were cultured for 4–6 days in 200 µl of X-vivo 5% HS in 96-well round-bottom plates. To evaluate the Ag-specific proliferative response labeled PBMCs (2 × 105 cells/ well) were plated in 96-well flat-bottom plates and stimulated either with heat-inactivated *Candida albicans* spores (5 × 106 spores/well, kindly provided by L. Romani, University of Perugia) or with tetanus toxoid (5 µg/ml; Enzo Life Sciences), or with total protein extract from a cell line infected with *Varicella zoster* Virus (2.5 µg/ml; Advanced Biotech) in a final volume of 200 µl of X-vivo (BioWhittaker-Lonza) 5% HS. For live/dead cell discrimination, PBMC or CD4<sup>+</sup> T cells were stained with Pacific Blue™ Succinimidyl Ester (ThermoFisher) at a final concentration of 0.1 µg/ml, according to manufacturer's instructions. Proliferated cells were counterstained with anti-CD3 Pacific Orange (clone UCHT1), anti-CD4 Pecy7 (clone SK3, BD Bioscience), and anti-CD8 APC-Cy7 (clone SK1, BD) mAbs by 15 min incubation at RT in PBS 2% FBS. Cells were washed with PBS 2% FBS and fixed with 0.25% formaldehyde. Flow cytometry analyses were performed with FCS Express 4 [*De Novo* Software; (https://www. denovosoftware.com/site/manual/proliferation\_statistics.htm)], and the frequency of precursors was calculated according to the automatic proliferation fit statistics, as described in the manufacturer's instructions. Alternatively, after 3, 4, or 5 days, cells were pulsed for 16 h with 1 μCi/well 3 H-thymidine.

### AutoAg Stimulation

Fresh PBMCs were labeled with VPD-450 (BD Horizon). A total of 8 × 106 cells/ml were stained with 1 µM VPD-450, 14 min at 37°C, in dark. After two washing steps with PBS, cells were resuspended in 1 ml of RPMI (Sigma-Aldrich) supplemented with 10% FBS, pen/strep (Cepa and Normon, respectively) and 2 mM l-glutamine (Sigma-Aldrich). A total of 1.5 × 105 PBMCs in 200 µl/well (5 wells/patient) were cultured in 96-well roundbottom plates for 7 days at 37°C in the presence of 5 µM of 7 myelin peptides [myelin oligodendrocyte glycoprotein (MOG) 1–20, MOG35–55, PLP139–154, myelin basic protein (MBP) 13–32, MBP83–99, MBP111–129 and MBP146–170]. Non-stimulated PBMC and 25 ng/ml phorbol 12-myristate-13-acetate (PMA) plus 250 ng/ml Ionomycin (Sigma-Aldrich) stimulated blood Ten Brinke et al. Optimization of Dye-Based Proliferation Assay

sample were used as negative and positive control, respectively. After 7 days of culture, PBMCs were stained with CD3-V500 (BD Horizon), CD45-APC, 7-AAD (actinomycin D), and CD4- FITC/CD8-PE (BD Biosciences), acquired with FACS Canto II (BD Bioscience) and analyzed using the FlowJo software. Alternatively, to analyze cell proliferation using 3 H-thymidine incorporation, after 7 days of cell culture (1.5 × 105 PBMCs/well, 60 wells/patient) with myelin peptides, cells were pulsed for 18 h with 1 μCi/well 3 H-thymidine.

### Ovalbumin (OVA) Stimulation for Frequency Calculation Responder Cells

BALB/C mice were purchased from Harlan (UK) and DO11.10 naive mice were bred in house and maintained in pathogenfree facilities (mice care was in accordance with institutional guidelines). Naive CD4+ T cells were isolated from splenocytes and peripheral lymph node lymphocytes by incubation with MoAbs: CD8 (56-3.72) and MHC class II (MS/114.15-2), locally produced hybridomas, CD16/32 and B220 (Becton Dickinson BD-Pharmingen), and CD25 (BD-Pharmingen), followed by negative selection using magnetic beads coated with sheep-anti rat IgG antibody (Dynal). Efficacy of depletion was measured by flow cytometry, and in all cases, CD4+ fractions were >85% pure; for DO11.10, the MoAb KJ126 (Caltag) was used. Cell calculations in our population were adjusted using this percentage.

### Stimulator Cells

CHO cells doubly transfected with mouse CD86 and H-2Ad were used as stimulators. CHO cells were maintained in culture as previously described (21), and prior to culture, they were incubated with 30 µg/ml Mitomycin C (Kiowa) for 1 h at 37°C, extensively washed and irradiated at 100 Gy. These cells were used to present OVA peptide 323–339 (Sigma) in the context of H-2Ad. All experiments with murine samples were performed in RPMI 1640 (Sigma) supplemented with pen/strep (Gibco), l-Glutamine 2 mM (Gibco), 10 mM HEPES (Gibco), 2β-mercaptoethanol (Gibco), and 10% fetal calf serum (SeraQ).

### CFSE Labeling for Frequency Calculation

BALB/C and DO11.10 CD4 T cells were labeled independently with CFSE (Molecular Probes, Leiden, The Netherlands) as follows: 2 × 107 cells were incubated with 1 µM of CFSE for 3 min at RT, washed extensively and were left overnight at 37°C 5% CO2 in culture medium. Known numbers of DO11.10 cells into BALB/C were mixed as above and incubated for 96 h at 37°C, 5% CO2 with stimulator cells and 0.5 µg/ml OVA323–339 peptide. Cells incubated in the absence of peptide were used as negative controls. Cells stimulated with 200 pg/ml phorbol dibutyrate (Sigma) and 1 µM Ionomycin (Sigma) were used as positive controls. Before acquisition in the flow cytometer, cells were labeled with mouse CD4-APC (Caltag), the clonotypic marker KJ126-PE (Caltag), and 20 ng/ml of propidium iodide (Sigma), thus enabling gating of the clonotypic receptor-expressing live CD4+ T cells. This way background proliferation of BALB/C cells to CHO stimulators was easily eliminated.

Flow cytometry analysis was performed before the beginning of the culture and after 3 days using Cell Quest and a FACScalibur (BD). Absolute counts of dividing precursors are achieved using Perfect-Count Micrsopheres (Cytognos) as per manufacturer's instructions. An absolute number of successful proliferative precursors can thus be obtained, by referring this number to the number of seeded cells in the well the frequency is easily calculated. Frequencies are given as 1 in "*n*" number of cells obtained as mean and standard deviation of the duplicated cultures.

### ELISpot Assay for Frequency Calculation

A commercial set of reagents was used (AID), and manufacturer's instructions were followed. Spots were enumerated with an ELISpot reader (AID). Plate was prepared as follows: duplicates at five "1/10" dilutions of mixed responder cells (according to the mix prepared to have 10–1,000 DO11.10 in the well) in 100 µl of medium were seeded. Irradiated and mitomycin-treated stimulator cells (50 × 104 in 50 µl) were added to all wells. OVA323–339 peptide at a final concentration of 0.5 mg/ml was also added to the necessary wells. Results are given as mean frequency and standard deviation calculated from the five dilutions in the format of 1 in "*n*" number of cells. Background spots of IL-2 production from negative control well (BALB/C cells and CHO stimulators) per dilution were subtracted from experimental IL-2 spots.

### Fascia

A total of 1 ml of PBS diluted (1/10) whole blood was stimulated with 50 µM of seven myelin peptides for 7 days at 37°C and 5% CO2. Non-stimulated and 25 ng/ml PMA plus 250 ng/ml Ionomycin-stimulated blood samples were used as negative and positive controls, respectively. After 7 days of culture, blood cells were stained with anti-CD3-V500 (BD Horizon), anti-CD45- APC, 7-AAD, and CD4-FITC/CD8-PE (BD Biosciences), and after lysis of erythrocytes, samples were acquired with FACS FACSVerse (BD Biosciences) and analyzed using the FACS Diva software (BD Biosciences). Number of proliferating cells was calculated following the protocol and formulas established in the Karolinska University Hospital (5).

### IFN-**γ** Production

Allogeneic-Ag-specific responses: labeled PBMCs (105 cells/well) were activated with irradiated (600 rad) auto or allo CD3-depleted PBMCs (APC) (2 × 105 cells/well) at a responder cells:stimulators ratio of 1:1. Alternatively, labeled PBMCs (1 × 105 cells/well) were stimulated with auto or allo mDC (104 cells/well) at a responder cells:stimulators ratio of 10:1 for the indicated time points in a final volume of 200 µl of X-VIVO15 medium with 5% HS (BioWhittaker-Lonza) and pen/strep (BioWhittaker) in 96-well round-bottom plates. Supernatants were harvested after 4 and 5 days of culture, and levels of IFN-γ were determined by ELISA according to the manufacturer's instructions (BD Biosciences).

Pathogenic Ag-specific responses: labeled PBMCs (2 × 105 cells/well) were left inactivated or stimulated with *C. albicans* spores (106 /well heat-inactivated spores generously provided by Prof. L. Romani, University of Perugia, Italy) or tetanus toxoid at 5 µg/ml, or in the presence of total protein extract from a cell line infected with *V. zoster* Virus (5 µg/ml) in a final volume of 200 µl of medium (96-well round-bottom plates). Supernatants were harvested after 3, 4, and 5 days of culture and levels of IFN-γ.

### Statistics

Analysis was performed using the GraphPad Prism 5.0 software. The correlation between the different parameters analyzed was evaluated by the non-parametric Spearman's rank correlation analysis.

## STEPWISE PROCEDURES

Functional *in vitro* assays to monitor frequency and phenotype of Ag-specific T-cell responses using fluorescent dye dilution depend on prolonged cell culture and proliferation of Ag-specific cells within the cultures. In these assays, responder cells are labeled by fluorescent dye and upon Ag-specific stimulation the dye is divided equally between daughter cells and the number of cell divisions of the proliferating cells can be visualized, allowing the theoretical enumeration of Ag-specific cells (Figure S1 in Supplementary Material). Here, we outline the critical steps required for establishing and analyzing an appropriate dye-based proliferation assay in a road map (**Figure 1**). Several technical constrains need to be taken into consideration in the setup of the assay, as they will have a definitive impact on the results of functional *in vitro* assays. Although not the focus of this study, choices of culture medium, serum lot—if used—and storage of reagents are obvious parameters that will affect the results of these assays.

An appropriate culture medium that ensures ample nutrient availability for the cells throughout the whole period of culture should be taken into consideration. Serum remains the preferred source of nutrients; however, most cultures are performed with 5–10% of heat-inactivated fetal calf or bovine serum. If serum is to be added to cultures each lot needs to be tested in all of the stimulations to be used to ensure the best signal-to-noise ratio. We suggest reserving a large amount of the serum lot to acquire consistency across a project. Recently, serum-free media have become popular (22) and ensure consistency of results

perform the analyses; (3) outlining the assay by defining the number of cells/wells to be put in culture, selection of responder and stimulatory cells, duration of culture, including additional staining for comprehensive analyses; (4) sample acquisition by delineating the gating strategy, i.e., live versus dead cells and number of events to acquire; and (5) data analyses.

independent of serum lot. Testing of the above-described variables is beyond the scope of the present work.

### Proliferation Dye Selection: Defining Optimal Dye and Labeling Concentration

Dye-based proliferation assay requires optimization in the laboratory where it will be performed, as for most other cellbased assays. Of specific importance is to optimize PBMC labeling with the dye of choice (23). Different proliferation dyes can be used, i.e., CFSE, CTV, VPD-450, and eFluor® 670. The staining intensity of responder cells should be as high as possible to obtain a broad analysis window of cell division (optimal fluorescence difference between specifically labeled cells and autofluorescenceof unlabeled cells). In this process of optimization, dye toxicity is an important issue to be taken into account. Toxicity of the labeling procedure is essential to be avoided by defining the optimal tolerable dye concentration, which can be monitored through a live/dead staining after culturing of the labeled cells. We recommend not only to focus on the induced toxicity, measurable by live/ dead staining, but also to determine the responsiveness of the PBMCs to stimulation as one of the quality controls. Polyclonal stimulation with anti-CD3 and anti-CD28 mAbs can be used to verify the impact of labeling on T-cell proliferation. As an alternative, stimulation with mitogens (PHA and PKW), or superAgs (SEB), can be used.

An important aspect that should be taken into account during optimization of the labeling procedure is the difference in dye fluorescence intensity between labeled and unlabeled cells, which affects the extent of analysis window. It is therefore clear that the choice of dye and labeling concentration for dye-based proliferation assays should be the result of a clear validation and analysis. In **Figure 2,** an example of such a dye selection is shown in which 3 concentrations of CFSE (1, 5, and 10 µM) and VPD-450 (1, 2, and 5 µM) were compared to label and analyze isolated total CD4<sup>+</sup> T cells. After 4-day culture, stimulated labeled cells were harvested and analyzed by flow cytometry, using live/dead staining and optimization of flow cytometer settings for each dye and each dye concentration to allow maximal separation between (auto-fluorescent) unlabeled cells and specifically labeled cells. The use of 5 and 10 µM CFSE led to higher signal intensities (undivided cells reach fifth decade on *X*-axis, green histograms) and better separation of division peaks than 1 µM concentration (**Figure 2A**). However, CFSE concentrations ≥5 µM were associated with relatively high cell toxicity (25 and 37% dead cells for 5 and 10 µM concentrations, respectively) as compared with 1 µM solution (13% dead cells). In addition, ≥5 µM CFSE concentrations decreased T-cell responsiveness: a lower percentage of

Figure 2 | Dye-based proliferation assay: dye selection and optimization of the concentration. Freshly isolated CD4+ T cells were labeled with different concentrations of CFSE (1, 5, and 10 µM) (A) and VPD-450 (1, 2, and 10 μM) (B), seeded at 1 × 105 cell/well and stimulated with magnetic beads coated with anti-CD3 and anti-CD28 antibodies (cell:bead ratio 1:1) for 4 days. The histograms correspond to unlabeled and stimulated cells (blue, auto-fluorescence), labeled and stimulated cells (pink) and labeled and unstimulated cells (green). The three cell populations were treated, cultured separately, and mixed before the analysis. Prior to the analysis and before mixing the cells, the unstained cells were labeled with anti-CD45 V450 or anti-CD45 FITC antibodies when CFSE or VPD-450 were used, respectively, to detect any overlap between stained and unstained cells. CFSE- and VPD-450-labeled and not stimulated cells were significantly smaller than labeled and stimulated cells; thus, they were discriminated according to the low values of FSC and SSC parameters. In the upper right corner of each histogram, only the proliferation of labeled responders is depicted and % of proliferating cells is presented. For each population only viable (7-AAD−), cells are presented. The following parameters are shown: PF = precursor frequency and PI = proliferation index.

responding cells (precursor frequency; PF; pink histograms) was observed for cells stained with 5 and 10 µM CFSE solutions than for 1 µM CFSE concentration. In addition, high dye concentrations decreased the average number of divisions of responding cells (proliferation index; PI). Along the same line, labeling with 5 µM of VPD-450 also led to higher fluorescence intensity than 1 and 2 µM solutions of VPD-450 (**Figure 2B**). Unlike for CFSE, cells stained with 1, 2 and 5 µM of VPD-450 showed similar viability (13, 18, and 17% of dead cells, respectively), but cell responsiveness (% of proliferating cells, PF and PI) was significantly lower for 5 µM than for 1 and 2 µM concentrations. When results for both dyes were compared, no differences in the dead cell frequency (7-AAD<sup>+</sup>), % of proliferating cells, and PI for 1 µM concentrations were observed. However, staining with 1 µM solution of VPD-450 resulted in a better peak separation than that observed for the same concentration of CFSE. Staining with 2 µM of VPD-450 further improved peak separation (pink histograms) with negligible impact on number of responding cells (PF) and PI. Therefore, for this assay setup, VPD-450 would be chosen as labeling dye at a concentration of 1 or 2 µM. Described analyses underline the importance of a designated assay for the selection of the proliferation dye, as a particular dye and/or its applied concentration may affect not only cell viability but also the proliferative responsiveness of the cells. Toxicity and labeling intensity are influenced by dye concentration, presence or absence of serum or other proteins during the labeling procedure, and the length and the temperature of labeling (24). In general, most laboratories choose to label PBMCs in a protein-free medium, since the used dyes covalently bind to free amines in proteins and hereby labeling of proteins in the medium is prevented. However, Quah and Parish (23) optimized the labeling in a protein rich medium by using high concentrations of dye and described optimal labeling, with low toxicity.

Importantly, every laboratory should perform this selection using their procedures, media, reagents and machines. Thus far, no specific indications regarding the best dye are available and each dye should be carefully tested in each particular setting.

### Peaks Resolution: Autofluorescence versus Cell Proliferation

Proper selection of the dye and its labeling concentration together with dye-optimized flow cytometer setup enable optimal separation of positive signal of responding cells (maximally divided labeled cells) from autofluorescence of the unlabeled cells (**Figure 2**, blue histograms). Efforts to optimize the signal-tonoise ratio are crucial to distinguish the separate peaks of dividing cells allowing reliable calculation of PFs of the responding cells.

Therefore, to perform optimal proliferation dye-based assays, we recommend to test and validate the most appropriate dye and its concentration with specific flow cytometer setup (**Figure 1**).

### Assay Setup

### Definition of the Number of Cells/Wells to Seed

When setting up culture conditions for Ag-specific T-cell enumeration, it is very important to use ≥10-fold more cells per culture than the expected frequency of responders to reliably monitor T-cell responses. Thus, if an Ag-specific response is to be measured in naive human individuals, at least 1 million of the responder cells need to be seeded, as the frequency of many Ag-specific naive T cells is in the order of 1:100,000. In contrast, if a subject has already been exposed to a given Ag, it is likely that 200,000 cells will be enough for detection of a response as the frequency of Ag-specific memory T cells is significantly higher (10, 25–27).

The threshold of sensitivity of the dye proliferation assay to reliably analyze low frequency T-cell responses is often questioned. The sensitivity of the dye proliferation assay was compared to ELISpot by using mouse TCR-transgenic CD4<sup>+</sup> T cells specific for OVA (DO11.10 cells) (**Figure 3**). DO11.10 cells were seeded at different known concentrations together with CD4<sup>+</sup> T cells from naive BALB/c cells, hereby knowing exactly the expected frequency of Ag-specific T cells to be found in the cultures. T cells were stimulated with OVA peptide-loaded CHO cells that expressed mouse CD86 (21), and the PFs were determined either by IL-2 ELISpot (**Figure 3A**) or by CFSE proliferation assay (**Figure 3B**). Results of both assays were in good correlation with the frequency of Ag-specific T cells in the culture (**Figure 3C**). Notably, at the lower frequencies of Ag-specific T cells (1/104 or 1/105 ) the measured frequencies were higher than expected, falling under the 45° line. Thus, maybe over-estimating the number of Ag-specific T cells in these settings. Overall, both methods are sensitive enough to reflect differences in frequency of Ag-specific T cells between samples.

When dealing with the low frequency of responder cells, we recommend calculating the number of cells to seed in culture as well as the number of events to be acquired should be calculated according to the estimated frequency of the putative Ag-specific cells present in the culture (4).

### Proper Selection of the Responder, Stimulator Cells and Optimal Time Point to Monitor Proliferation

An important issue that has to be taken into consideration during optimization of dye-based proliferation assays is the proper selection of the responder and stimulator cells to be used (**Figure 1**). Responder cells can be either total PBMCs or purified CD4<sup>+</sup> or CD8<sup>+</sup> T cells; however, the use of total PBMCs could be theoretically more informative, since it will allow studying the response of different lymphocyte sub-populations (i.e., CD4, CD8, effector, naive, and memory T-cell subsets), and analysis of activation markers (28). Furthermore, (allo)Ag-specific T cells and dye-based proliferation assay can be combined with the intracellular staining for cytokines upon *in vitro* re-stimulation (8, 29, 30) or for transcription factors, such as FOXP3 (31, 32), overall obtaining additional information regarding proportions of different cell subsets, including Tregs within the Ag-specific cell pool present in the culture.

Another important point to take into consideration during optimization of the proliferation dye assay is the selection of the proper time point to visualize effective proliferation and optimal separate division peaks. Examples of how the source of stimulatory cells, the purity of the responder cell population and the timing may impact the *in vitro* detection of alloAg-specific T-cell responses is provided in the Section "Anticipated Results."

mAbs, enabling gating of the DO11.10 CD4+ T cells. (C) Pearson correlation of observed frequencies by ELISpot and CFSE dilution against expected values of three independent experiments are presented. In accurate assays, it would be expected, within a tolerable error, that the measurements fall on a 45° line through the origin.

### Acquisition Setup

To analyze proliferation dye data, a gating strategy focusing on living cells and number of acquired events is recommended. Of note, in case, the proliferation dye-based assay is used to analyze alloAg-specific T-cell responses, where allo PBMCs/APCs are added as stimulators to the culture, it is very important to distinguish between proliferation of responder and stimulator cells (both negative for proliferation dye fluorescence). To this end, different approaches can be used including labeling of the responder and stimulator cells with different dyes or depleting CD3<sup>+</sup> lymphocytes from stimulator PBMCs.

## Optimal Parameters for the Analysis of Proliferation Dye Data

Results obtained by performing a dye-based proliferation assay can be depicted and interpreted in several manners (**Figure 1**). Results of proliferation dye assays are generally presented as percentage of cells showing dye dilution (% of proliferating gate, Figure S1 in Supplementary Material). The latter is the easiest and the most often used parameter to present proliferation data. However, this parameter is affected by both the number of cells responding to a given stimulus (PF) and the number of divisions of dividing cells (PI) and, therefore, gives limited insight in the dynamics of cell proliferation and reactive T-cell frequencies. Obviously, this result is affected by several parameters, including actual percentage of cells responsive to stimulus (also named progenitor cells or PF), number of divisions of the dividing cells, and occurrence of cell death. Thus, this parameter is good for general comparison between different samples but is difficult to interpret and may be not sufficient for monitoring Ag-specific responses, since it does not directly reflect real percentage of Ag-specific T cells present in culture. Alternatively, results of proliferation dye can be depicted as (i) PF (proportion of cells with reactivity to a specific Ag or mitogen within the starting population), (ii) division index (DI; average number of divisions of all cells, including undivided cells); and (iii) PI (average number of cell divisions of responding cells). It is recommended to present the PF and PI, since the DI is affected by both the PF and the PI. The calculation of these parameters for each dyebased proliferation culture can be determined by operator or by using flow cytometry analysis software (33, 34). In **Figure 4,** an example of different ways to depict proliferation dye data is given. Although in both conditions the percentage of proliferating cells was similar (70.1 vs. 71.4%), the different values for the PIs (1.6 vs. 1.9, Table S1 in Supplementary Material) of both cultures showed that the cells had not proliferated to the same extent in the two conditions. The PFs calculated for the two conditions also differed to some extent (PF, 38.4 vs. 30.7%, Table S2 in Supplementary Material).

Together, this example underlines the limitations of analysis of percentage of proliferating cells and concomitantly indicates that the combined use of PI and PF is most informative to analyze results from proliferation dye-based assays for immuno-monitoring. Importantly, percentage of proliferating cells is not an informative parameter when it exceeds 60%, since the ability to distinguish biological variations becomes difficult (34). When monitoring the effect of a tolerizing therapy, depicting both PF and PI separately, when possible, will give more information regarding the mode of tolerance induction. As a drawback, the calculation of PF is not always possible due to the lack of visible separate division peaks. This possibility can occur when autoAg-specific T-cell proliferation is monitored or when alloAg-specific proliferation of cells isolated from patients under immunosuppressive regimens. In the latter case, software peak prediction can be used. Nevertheless, as shown in this study, in some situations also computational prediction cannot be applied. In this case, the only remaining option is to present percentage of proliferating wells, as shown in the present study. Besides, as for all functional assay, it has to be taken into account that the outcomes of the assay will be affected by cell death in the culture, and is a reflection of the surviving cells.

### ANTICIPATED RESULTS (PITFALLS, ARTIFACTS, AND TROUBLESHOOTING)

To evaluate the mechanisms underlying failure or success of tolerogenic therapies in transplantation or autoimmunity, monitoring of Ag-specific immune responses is critically important. The precise enumeration and phenotypic analysis of Ag-specific T cells remains technically difficult, mainly due to their low frequency (26, 27, 35). Therefore, a sensitive, reproducible, and reliable method to enumerate and analyze Ag-specific T cells in treated subjects is important. Several approaches have been proposed and tested to identify Ag-specific T cells, including proliferative responses and cytokine production profiles. In this study, we focus on the use of T-cell proliferation for detection and analysis of Ag-specific T cells in PBMCs. We provide evidences that a dye-based proliferation assay is as sensitive as other currently used methods for enumerating low frequency Ag-specific T cells.

### Sensitivity of the Dye Proliferation Assay Compared to Other Approaches

We evaluated the sensitivity of dye-based assays in human samples, with unknown frequencies of responder cells in comparison with 3 H-thymidine incorporation and IFN-γ release. We evaluated the cellular response to alloAgs and pathogenderived exogenous Ags (*C. albicans*, tetanus toxoid, and *V. zoster* Virus) (**Figures 5** and **6**, respectively). For alloAg-specific T-cell responses, a comparison of alloAg-specific proliferation induced by allo CD3-depleted PBMCs (allo APC) or allo monocyte-derived mDCs (allo mDC) is shown (**Figure 5**). As negative control, auto APC or mDC (auto mDC) was used. In this example, the allo proliferative response induced by allo APC was equally good as that induced by allo mDC. CD3-depleted PBMCs may be considered as preferred stimulator source, as their generation is much less laborious and costly than *in vitro* generated dendritic cell (DC) from allo monocytes, and tend to induce less auto background proliferation than allo mDC (**Figures 5A,B**). Comparison of total PBMCs and isolated CD4<sup>+</sup> T cells as source of responder cells (**Figures 5B,C**, respectively) showed that total PBMCs may be the preferred choice, since the proliferative response was comparable to that of purified CD4<sup>+</sup> T cells, and they are easier and less expensive to obtain. Of note, in case of very low expected frequencies of Ag-specific T cells, purification of the CD4<sup>+</sup> T-cell pool may be advisable to increase the relative frequency of the specific T cells in culture (**Figure 5C**). In the case of pathogen-derived Ags, precursors' frequencies were very low (**Figure 6**) and separate division peaks were not visible (Figure S2 in Supplementary Material); the precursors' frequency calculation relied on the software peak prediction.

As expected, proliferation in response to alloAgs or pathogenderived exogenous Ags can also be detected by 3 H-thymidine incorporation (**Figures 5D** and **6C**) or IFN-γ release (**Figures 5E** and **6D**), although these two read outs did not allow to evaluate the proliferation or cytokine secretion specifically by CD4<sup>+</sup> or CD8<sup>+</sup> T cells within PBMCs. Interestingly, while the proliferation dye dilution and 3 H-thymidine incorporation correlated well both in the case of alloAg-specific T-cell responses and the responses to pathogen-derived Ags (**Figures 5F** and **6E**), correlation between the proliferation dye dilution and IFN-γ release was less evident in the case of alloAg-specific T-cell responses but present for pathogen-derived Ag-specific responses (Figure S3 in Supplementary Material). From the above experiments, it can be concluded that dye proliferation assay is suitable to detect T cell specific for alloAgs or pathogen-derived exogenous Ags. For optimal readout of alloAg responses 4–5 days of stimulation is suitable (**Figure 5**), while for pathogen-derived exogenous Ag responses 7 days of culture is required (**Figure 6**). The assay time is longer compared to other techniques (i.e., H-thymidine incorporation or cytokine profiles) (3). This time is mandatory for small population of Ag-specific T cells to reach numbers detectable and quantifiable with dye-based assays (4, 36).

The comparison of proliferation data obtained with dyebased proliferation assay and 3 H-thymidine incorporation for alloAg- and nominal Ag-specific T-cell response gave good concordance. These results are in line with previous reports (3, 37–39). However, a less correlation was observed between proliferation dye dilution and IFN-γ production, specifically in allo mixed lymphocyte reaction. This result may be related to NK-cell activation when total allo APC is used as stimulatory cells. To avoid this possibility, the use of monocye-derived DCs would be recommended. Moreover, it has to be considered that in the proposed examples as well as in the present study, correlation is observed when high frequency of Ag-specific T cells is present in the peripheral blood and strong antigenic responses are analyzed. This is less evident when Ag-specific T cells are less frequent, as in patients with autoimmune disorders, or a less immunogenic response is studied (40).

Detection of reactive T cells against autoAgs requires highly sensitive techniques due to the low frequency of these autoreactive T cells in peripheral blood (26, 27). For this reason, as indicated above high numbers of cells (i.e., 1.5 × 105 PBMCs/well) in several replicates should be seeded. To test the sensitivity of the proliferation dye assay in the detection of autoAg-specific T-cell responses, we tested the response of PBMC from multiple sclerosis (MS) patients to a mix of seven myelin peptides by comparing the proliferation dye (VPD-450), with 3 H-thymidine incorporation and FASCIA. As depicted in **Figure 7A** (3), 3H-thymidine incorporation is highly sensitive for detecting autoAg-specific T cells, since all patients tested exhibited increased proliferation ≥25% of analyzed wells compared to the mean of non-stimulated controls (**Figure 7A**). Analysis of proliferation dye dilution indicated that separate division peaks were not visible upon autoAg-specific stimulation, making it impossible to calculate PF, even with the aid of the software peak prediction program. Therefore, as alternative the frequency of autoAg-specific T cells was calculated as the percentage of positive wells defined by considering replicates showing ≥1.5 stimulation index (SI, % proliferating cells stimulated/% proliferating cells non-stimulated). Compared to 3 H-thymidine incorporation, the VPD-450 dilution assay generated a similar% of positive replicates (≥20% positive wells) in three out of four MS patients, while for one patient all replicates were positive in the dye-based assay. In parallel, FASCIA was also

Figure 5 | Dye-based proliferation assay to analyze alloAg-specific T-cell responses. Freshly isolated PBMCs (A,B,D,E) or purified CD4+ T lymphocytes (C) (from *n* = 3 healthy donors) were activated in the presence of either irradiated autologous or irradiated allogeneic CD3-depleted PBMCs (APC) (*n* = 2 allogenic stimulators) at a 1:1 responder: stimulator ratio. Alternatively, cells were stimulated with either autologous or allogenic mature DC (mDC) at a 10:1 responder:stimulator ratio for the indicated time points. The proliferative response was evaluated with proliferation dye (efluor-670; 10 µM) (A–C,F) (3), H-thymidine incorporation (D,F), or IFN-γ secretion (E). Mean ± SEM of precursor frequencies in the starting population (gated on CD3+ T lymphocytes) (A), means ± SEM of the percentage of proliferating CD3+ T cells (B,C), means ± SEM of cpm (D), mean ± SEM of IFN-γ concentration (E) are plotted. Each filled dot represents an independent responder-stimulator mismatch. Open dots represent responder-stimulator autologous controls (C). Cutoff for positive response was set as stimulation index (SI) vs. matched autologous stimulators >2. Red dots indicate responder-stimulator mismatch with SI <2 (non-responders). (F) Correlation between detection of alloAg-specific response by 3 H-thymidine incorporation and dye-based proliferation was evaluated by Spearman's rank correlation analysis (non-parametric). The plots show cpm at day 4/5 vs. precursor frequency of CD3+ T cells detected at day 6. Each dot represents an independent responder-stimulator (CD3-depleted PBMCs) match (including both auto- and allo-stimulators and both PBMC and purified CD4+ T cells as responders) (*n* = 18 independent determinations for cells derived from three healthy donors). The line represents the linear regression; coefficients and p values of the correlation are reported in the graphs. APC: antigen-presenting cells; mDC: monocytederived mature (LPS activated) dendritic cell; cpm: counts per minute; allo: allogeneic; auto: autologous.

performed by stimulating fresh whole blood from the same MS patients with the mix of myelin peptides. The analysis of CD4<sup>+</sup> blast cells showed a SI ≥1.3 in all samples (**Figure 7B**), like the SI ≥1.5 detected in positive auto-reactive wells using VPD-450 dilution assay. These examples indicate that dye-based proliferation assay is sensitive enough to detect T cells specific for a given Ag, including blood samples from patients with autoimmune disease in whom the frequency of autoAg-specific T cells in peripheral blood is generally low.

The latter results are compliant with those of Zafranskaya et al. (41), who compared 3 H-thymidine incorporation and CFSE-based assay for assessing MOG-reactive T cells in healthy donors, untreated MS patients and IFN-β-treated patients. Remarkably, data from MS patients contrast with those obtained in type 1 diabetes (T1D) patients. It was demonstrated that CFSE-based proliferation assay was more sensitive than 3 H-thymidine incorporation to study auto-Ag-specific reactivity in T1D patients (3). In this study, all tested patients had a

stimulated with *Candida albicans* spores (solid circles) or tetanus toxoid (triangles), or in the presence of total protein extract from a cell line infected with *Varicella zoster* Virus (squares) for the indicated time points. The proliferative response was evaluated by proliferation dye (efluor-670, 10 µM) of CD4+ T cells (A) and as frequency of precursors (B) (3), H-thymidine incorporation (C), or IFN-γ secretion (D). Each dot represents PBMC unstimulated or stimulated with a nominal Ag. Cutoff for positive response was set as stimulation index (SI) vs. autologous non-stimulated cells >2. Red symbols indicate donors with SI <2 (non-responders). Ag: antigen; cpm: counts per minute. Cells from ≥3 healthy donors were tested for each time point. (E) Correlation between detection of pathogen-specific Ag-specific response by 3 H-thymidine incorporation and dye-based proliferation was evaluated by Spearman's rank correlation analysis (non-parametric). The plots show cpm at day 5/6 vs. precursor frequency of proliferating CD3+ T cells at day 7. Each dot represents an independent responder PBMC unstimulated or stimulated with pathogen-specific Ag (*C. albicans* or tetanus toxoid or *V. zoster* Virus) (12 independent experiments were performed with cells derived from three healthy donors). The lines represent the linear regression; coefficients and *p* values of the correlation are reported in the graphs.

detectable response to glutamic acid decarboxylase (GAD), an autoAg in T1D, with CSFE dilution assay, that was revealed only in half of the patients by 3 H-thymidine incorporation. Moreover, Segovia-Gamboa et al. (42) detected GAD- and insulin-specific responses with a CFSE-based assay using auto DCs loaded with Ag as stimulators for memory CD4<sup>+</sup> T cells. We cannot exclude that the discrepancy between analyses performed in T1D patients and our data depends on the frequency of auto-reactive T cells in these patients, duration, and stage of the disease, or the immunogenicity of the autoAg. Data obtained with FASCIA assays are promising, although the assay is somewhat less sensitive than dye proliferation assay in detecting auto-reactive CD4<sup>+</sup> T cells. Nevertheless, it can be an alternative in cases PBMC isolation maybe complicated as, for instance, in pediatric patients. Although analysis of autoAg-specific responses is challenging, we believe that dye-based proliferation assays represent a good choice for the enumeration of autoAg-specific T cells, since they allow for measurement of additional phenotypical and cell function related parameters critical for a better description of auto-reactive T cells and their activation.

protein (MBP) 13–32, MBP83–99, MBP111–129, and MBP146–170] for 7 days. The proliferative response was evaluated by dye-based proliferation (VPD-450, 1 µM) or 3 H-thymidine incorporation (A), or flow cytometric assay of specific cell-mediated immune response in activated whole blood (FASCIA) (B). A total of 5 wells (VPD-450) or 60 wells (3 H-thymidine) were analyzed. The % of autoAg-reactive wells/replicates (wells showing increased 3 H-thymidine incorporation compared to the mean of non-stimulated wells) (black bars) and the % of autoAg-reactive wells/replicates from VPD-450 dilution assay (wells exhibiting a SI = stimulated wells/ non-stimulated wells ≥1.5) (gray bars) are shown (A). The number of CD4+ blast cells was determined by FASCIA (B).

In accordance with previous results reported using CFSEbased assay (43, 44), we showed that dye-based proliferation assay is suitable to detect autoAg-specific T cells in peripheral blood of MS patients. Moreover, the sensitivity of dye-based proliferation assay is comparable to that of 3 H-thymidine incorporation in detecting auto-reactive T cells in MS patient's PBMC.

### CONCLUDING REMARKS

Dye-based proliferation assays, in contrast to other approaches, offer the possibility to retrieve additional information additional to the overall proliferative response. First, the frequency of Ag-specific T-cell precursors in the starting population can be determined, which is not the case, for instance, for proliferation analyzed by 3 H-thymidine incorporation. Furthermore, dyebased proliferation assays provide insights in the dynamics of proliferation and phenotype of the cells at different stages of proliferation within a PBMC culture. Tolerizing immunotherapy can induce Ag-specific tolerance *via* several mechanisms: (allo) Ag-specific T cells can be deleted or become anergic, and this will lower the PF (45, 46). Alternatively, the tolerizing therapy may cause the (allo)Ag-specific T cells to respond to a lower extent, leading to a restraint on cell division, while not affecting PFs. In conclusion, tracking Ag-specific T-cell responses with dye dilution represents a valuable tool to monitor tolerance induction in human. Strict attention to setup and validation of the culture conditions should be given before execution of the study while taking into consideration the disease and the type of Ag under assessment. We believe that this is the first step to harmonize the monitoring of tolerance induction, which will enable the comparison of immunological mechanisms responsible for the clinical outcomes of different tolerance-inducing studies. In addition, a well-designed and validated dye proliferation assay can be applied to other therapies aimed at increasing Ag-specific T-cell responses such as vaccination and cancer immunotherapy.

### ETHICS STATEMENT

Human peripheral blood was obtained from healthy donors upon informed consent and approval by local ethical committee (Sanquin Amsterdam, Medical University of Gdańsk, San Raffaele Scientific Institute) and in line with the Declaration of Helsinki. Human peripheral blood was obtained from four MS patients from the Multiple sclerosis Unit, Germans Trias I Pujol University Hospital (Badalona, Spain) upon informed consent in accordance with local ethical committee approval and with the Declaration of Helsinki. No patient had clinical exacerbations or was receiving corticosteroid or disease modifying treatments at the moment of the sample collection. Studies with mice were approved and in accordance with guidelines from King's College London, UK.

## AUTHOR CONTRIBUTIONS

AB, NM-T, SH, MF, EM-C, and SG wrote the article. AB, NM-T, AT, EM-C, and SG designed and planned experiments. NM-T, MM, AT, KP, DI-G, LP, GL, JP-O, and MF performed and analyzed experiments.

### ACKNOWLEDGMENTS

All AFACTT members are acknowledged for their inspiring discussions and contributions during the AFACTT meetings.

### FUNDING

European grant for European cooperation in science and technology (Action BM1305: Action to Focus and Accelerate Cell-based Tolerance-inducing Therapies; http://www.afactt. eu) funded the networking activities which resulted in this paper. COST is supported by the EU Framework Program Horizon 2020. NM-T acknowledges the following sources of funding: Polish Ministry of Science and Higher Education (grant no. IP2011 033771) and National Centre of Science, Poland (funding decision no. DEC-2011/01/D/NZ3/00262). DI-G acknowledges the following sources of funding: National Centre for Research and Development (grants LIDER/160/L-6/14/NCBR/2015). MF acknowledges the following sources of funding: MRC (grants G0801537/ID: 88245 and MR/J006742/1MR/J006742/1); the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London—the views expressed are those of the author(s) and

### REFERENCES


not necessarily those of the NHS, the NIHR or the Department of Health; EU, HEALTH-F5-2010-260687 and FP7-HEALTH-2012-INNOVATION-1 project number 305147: BIO-DrIM. EM-C and MM research is supported in part by IWT-TMB Grant 140191 (Vlaanderen, Belgium) and FIS PI14/01175, integrated in the National Plan I+D+I and co-financed by ISCIII-Subdirección general de Evaluación y el Fondo europeo de Desarrollo Regional (FEDER). SG acknowledges the following sources of funding: grant from the Italian Telethon Foundation "Comitato Telethon Fondazione Onlus," Core grant TIGET TGT16G01 and the Italian Association for Cancer Research project IG 2013 N 14105 (Associazione Italiana per la Ricerca sul Cancro, or AIRC).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/articles/10.3389/fimmu.2017.01870/ full#supplementary-material.

T cell tolerance in children with allergic asthma. *Cell Mol Immunol* (2017). doi:10.1038/cmi.2017.26


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 Ten Brinke, Marek-Trzonkowska, Mansilla, Turksma, Piekarska, Iwaszkiewicz-Grzes´, Passerini, Locafaro, Puñet-Ortiz, van Ham, Hernandez-Fuentes, Martínez-Cáceres and Gregori. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Tolerance through education: How Tolerogenic Dendritic Cells Shape immunity

*Matthias P. Domogalla1,2\*, Patricia V. Rostan1,2, Verena K. Raker1,2 and Kerstin Steinbrink1,2*

*1Department of Dermatology, Division for Experimental and Translational Research, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany, 2Research Center for Immunotherapy, University Medical Center of the Johannes Gutenberg-University Mainz, Mainz, Germany*

Dendritic cells (DCs) are central players in the initiation and control of responses, regulating the balance between tolerance and immunity. Tolerogenic DCs are essential in the maintenance of central and peripheral tolerance by induction of clonal T cell deletion and T cell anergy, inhibition of memory and effector T cell responses, and generation and activation of regulatory T cells. Therefore, tolerogenic DCs are promising candidates for specific cellular therapy of allergic and autoimmune diseases and for treatment of transplant rejection. Studies performed in rodents have demonstrated the efficacy and feasibility of tolerogenic DCs for tolerance induction in various inflammatory diseases. In the last years, numerous protocols for the generation of human monocyte-derived tolerogenic DCs have been established and some first phase I trials have been conducted in patients suffering from autoimmune disorders, demonstrating the safety and efficiency of this cell-based immunotherapy. This review gives an overview about methods and protocols for the generation of human tolerogenic DCs and their mechanisms of tolerance induction with the focus on interleukin-10-modulated DCs. In addition, we will discuss the prerequisites for optimal clinical grade tolerogenic DC subsets and results of clinical trials with tolerogenic DCs in autoimmune diseases.

Keywords: tolerogenic dendritic cells, regulatory T cells, immunotherapy, tolerance, nanoparticles

## INTRODUCTION

The antigen-specific induction of immunological tolerance in the context of autoimmune and allergic diseases, which are driven by undesired immune responses against the body's own or foreign antigens, has long been described as ultimate solution for the treatment of excessive immune activation. Nowadays, common treatment options are life-long, systemic immune suppression, which however may lead to serious side effects like chronic infections or malignant transformation. Therefore, various cell types have been investigated to establish permanent antigen-specific immune tolerance toward the causative triggers. Dendritic cells (DCs) as key players in controlling immune responses by either inducing immunity or establishing tolerance through interaction with multiple immune cells seem to be excellent candidates for the re-establishment of permanent antigen-specific tolerance. Since their discovery in 1973 by Ralph M. Steinman, several *in vitro* protocols have been established for the generation of potent, stable tolerogenic DCs whereof some have recently been used for the treatment of transplantation rejection, autoimmune and allergic disorders *in vivo*. In addition, to avoid *ex vivo* generation and modulation of DCs, DC-specific *in vivo* targeting, e.g., by antibodies or nanoparticle-based approaches, which can directly deliver immunomodulatory drugs

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Carolina Llanos, Pontificia Universidad Católica de Chile, Chile Ljiljana Sofronic-Milosavljevic, Institute for the Application of Nuclear Energy (INEP), Serbia*

#### *\*Correspondence:*

*Matthias P. Domogalla matthias.domogalla@ unimedizin-mainz.de*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 29 September 2017 Accepted: 27 November 2017 Published: 11 December 2017*

#### *Citation:*

*Domogalla MP, Rostan PV, Raker VK and Steinbrink K (2017) Tolerance through Education: How Tolerogenic Dendritic Cells Shape Immunity. Front. Immunol. 8:1764. doi: 10.3389/fimmu.2017.01764*

to DCs, have emerged as a promising tool. In this review, we will outline the different protocols for generation of tolerogenic DCs, their mechanisms of tolerance induction, and summarize their use in preclinical and clinical settings.

### ROLE OF DCs IN IMMUNITY AND TOLERANCE

Recognition of DCs as professional antigen-presenting cells has come a long way. Antonio Lanzavecchia once stated that DCs seemed "too rare to be relevant" (1). With the Steinman lab pioneering DC immunology in the 1980s, the field started to expand rapidly and apart from their function in induction and maintenance of immunity, they also became relevant as promising candidates for immunotherapy with regards to tolerance induction.

Some refer to DCs as "nature's adjuvants" highlighting their central role in the induction of immune responses. DCs populate almost all body surfaces in order to serve as sentinels detecting pathogens either by membrane-bound toll-like receptors (TLRs) or within the cytosol through nucleotide-binding oligomerization domain-like receptors (NLR) (2, 3). They do not kill the pathogen directly but use an even more sophisticated approach that induces long-lasting antigen-specific responses sufficiently bridging innate and adaptive immunity. By utilizing a proteolytic machinery (endolysosomal and proteosomal), they partially degrade antigens to peptides to subsequently display peptide/ major histocompatibility (MHC) complexes on their surface (4). Although other cells such as macrophages and B cells are also able to present antigens *via* MHC, DCs are the only cell type to activate naïve T cells and to induce antigen-specific immune responses in all adaptive immune cells (4). They can for instance directly induce antibody production by presenting intact antigen to antigen-specific B cells without engaging T cells (5). DCs take a guiding role in immune responses as they interrogate, interpret, and transmit the nature of the antigenic stimulus, thereby shaping even T cell polarization *via* different intracellular signaling pathways (6).

Immature DCs (iDCs) are predominantly found in the peripheral tissues where they patrol and extensively take up large quantities of membrane-bound or soluble antigen by macropinocytosis and phagocytosis. However, at an immature state, DCs are inefficient in displaying MHC/peptide complexes on their surface as, e.g., their lysosomal activity is attenuated (3). The ability to channel MHC/peptide complexes to the surface increases upon engagement of pathogen recognition receptors such as TLRs or NLRs, which drive DC maturation (7). DCs change their capacity from antigen accumulation to T cell activation within only 1 day. Expression of chemokine receptors [C–C chemokine receptor (CCR) 1, CCR2, CCR5, CCR6, and C–X–C chemokine receptor (CXCR) 1] facilitates immature DC recruitment to the site of inflammation. Activation of DCs results in CCR6 downregulation and CCR7 and CXCR4 upregulation directing DCs toward the lymph node (8, 9).

Dendritic cell maturation, however, has a high degree of plasticity meaning that differentiated mature DCs (mDCs) can easily convert to tolerogenic DCs. This has been shown, e.g., by a group that stimulated activated DCs with pro-inflammatory interferon-γ (IFN-γ), which promoted the expression of indoleamine 2,3-dioxygenase (IDO) leading the respective DCs to acquire tolerogenic potential (10).

The original concept of tolerance induction by DCs is attributed to low amounts of surface MHC and co-stimulatory molecules such as cluster of differentiation (CD) 80 and CD86 found on iDCs. In contrast, the CD80/CD86high expressing mature DC counterpart would rather activate effector T cells. However, in an uninfected individual, maintenance of self-tolerance is ensured by a continuous input of short-lived DCs that provide self-antigens in the lymphatic tissues. Notably, DCs isolated in the cold from germ-free mice show expression of co-stimulatory molecules and activate T cells to enter cell cycle (11). This indicates that the original view of tolerance induction is highly dependent on DCs' mutual state of development and activation, as well as the surrounding microenvironment of cytokines and growth factors.

Dendritic cells in the thymus establish (central) self-tolerance by the display of self-antigens to developing T cells inducing T cell negative selection or Treg differentiation (12). Induction of peripheral T cell anergy and apoptosis, attenuation of effector and memory T cell responses, and the generation and activation of regulatory T cell (Treg) subpopulations has been attributed to a variety of tolerogenic DC subtypes (13–16).

Dendritic cell subtypes in humans can be characterized by anatomical localization and respective function. In steady state, blood DCs are immature precursors of tissue or lymphoid organ DCs. Epithelial tissues contain non-lymphoid or migratory DC subtypes (17). Lymphoid tissues harbor resident DC populations, which lack migratory capacities and play a role in retrieval of antigen and maintenance of antigen-specific immune responses (e.g., follicular DCs that recycle and "store" antigen for prolonged B cell activation in lymph node germinal centers) (18). Upon pathogen encounter and subsequent inflammatory state, the DC content of tissue and lymphoid organs is altered. Steady state DCs are diluted by CD14<sup>+</sup> classical monocytes and precursors of inflammatory DCs. Blood DCs might also enter tissues *via* CD62 ligand (L) and CXCR3 expression, which allows extravasation (19).

Site-specific appearance contaminating monocytes/macrophages and diverse inflammatory stimuli hinder a distinct phenotypical characterization of DC subsets. However, DCs were originally defined by their characteristic dendritic morphology and extraordinary capacity for antigen presentation and T cell priming (20, 21). These classical or conventional DCs (cDCs) are now classified into two main subsets, the CD11b<sup>+</sup> and CD8<sup>+</sup>/ CD103<sup>+</sup> cDCs in mice and the corresponding blood dendritic cell antigen (BDCA)-1<sup>+</sup> (CD1c<sup>+</sup>) and BDCA-3<sup>+</sup> (CD141<sup>+</sup>) cDCs in humans. Beyond these subsets, however, a significant functional, genetic, and phenotypic diversity of DCs has been recently appreciated. There have been huge recent efforts by the scientific community to identify strategies in order to align DC phenotypes in a tissue and cross-species specific manner *via* flow cytometry. A set of lineage-imprinted markers recently published by Guilliams et al. is sufficient to differentiate between human plasmacytoid DCs [pDC: CD45<sup>+</sup>CD11clow human leukocyte antigen-D related (HLADR)high interferon response factor (IRF) 8highIRF4mid], conventional type 1 (cDC1s), and conventional type 2 DCs (cDC2s) (**Figure 1**). The authors provide a smart strategy to identify human cDC1s within (monocyte/macrophage excluded) CD14<sup>−</sup>CD16<sup>−</sup> cells as cell adhesion molecule (CADM)1highCD172alowCD11cmid/highCD26high cells and cDC2s as CADM1lowCD172ahighCD1chighCD11chigh cells validated by means of mass spectrometry and even on transcription factor level. This strategy is robust even under inflammatory conditions, in different tissues and allows identification of the same DC subset in macaques, humans, and mice (22).

A sophisticated identification strategy will allow for a more profound analysis of DC fates in mice and humans with regard to immunological functions of DCs in immunity and tolerance.

### GENERATION AND SUPPRESSIVE MECHANISMS OF TOLEROGENIC DCs

During classical immune responses, after encountering an antigen in combination with a danger signal, DCs upregulate the expression of co-stimulatory molecules, lymph node-homing receptors plus MHC molecules, and start the secretion of pro-inflammatory cytokines (21). Those processes enable DCs to migrate to the lymph nodes and initiate the activation of naïve T cells. Full T cell activation requires a three step signaling process. First, the binding of the T cell receptor (TCR) to its cognate antigen, which is presented on MHC molecules, second, the engagement of CD28 with co-stimulatory molecules like members of the B7 protein family CD80 and CD86, and third, interaction of DC-secreted cytokines with appropriate respective cytokine receptors (23).

In contrast, tolerogenic DCs exploit several immunosuppressive mechanisms to induce tolerance (**Figure 2**). Tolerogenic DCs often display an immature or semi-mature phenotype that is characterized by low expression of co-stimulatory and MHC molecules and altered cytokine production. Presentation of low levels of antigen without co-stimulation leads to T cell anergy (24) and promotion of regulatory T cell differentiation *in vitro* and *in vivo* (25–27). TCR signaling in combination with co-stimulation results in activation of the transcription factors nuclear factor of activated T cells (NFAT), activator protein (AP)-1, and nuclear factor "kappa-light-chain-enhancer" of activated B-cells (NF-κB) that subsequently induce a transcriptional program resulting in T cell activation (28). It is not yet exactly clear how absence of co-stimulation results in a transcription profile that

CD26highCD11cmid-high as cDC1s or CADM1highC172ahighCD1chighCD11chigh cDC2s. In humans and mice, DC cell fate can be additionally identified on the level of transcription factors: DCs in general are dependent on flt-3. cDC1 development depends on BTAF3 and high levels of IRF8, whereas cDC2 evolution is dependent on IRF4 but independent of BATF3.

h

favors Treg induction but impaired CD28-induced activation of the rat sarcoma/mitogen activated protein kinase (Ras/MAPK) pathway results in deficient AP-1 activation. In the absence of AP-1, NFAT proteins, possibly in combination with other transcription factors or by forming dimers, may subsequently initiate a transcriptional program that cumulates in T cell anergy and Treg induction (29). However, recent studies demonstrated that phenotypically mDCs are also capable of inducing Tregs, indicating that the phenotype does not necessarily determine the immunogenic or regulatory function of DCs (13, 30, 31). Furthermore, secretion of anti-inflammatory cytokines like interleukin (IL)-10 and transforming growth factor-β (TGF-β) and reduced expression of pro-inflammatory cytokines by DCs critically contribute to tolerance induction. Production of IL-10 by tolerogenic DCs is indispensable for regulatory function in multiple settings (32–34) and DC-released TGF-β is important for tolerance induction as DC-specific ablation of the TGF-β activating integrin αvβ8 (Itgb8) results in autoimmunity and colitis as demonstrated in transgenic CD11c-Cre/Itgb8fl/fl mice (35). Moreover, TGF-β secretion by tolerogenic DCs is important for the regulation of TH17 responses in neuro-inflammation as shown in CD11cDNR mice, which is a dominant-negative form of TGF-β receptor II resulting in diminished TGF-β signaling (36). Furthermore, IL-10 and TGF-β, which are secreted by tolerogenic DCs in the tumor microenvironment, facilitate and reinforce tumor escape (37). In addition, several immunosuppressive features of tolerogenic DCs rely on induction of apoptosis in responding T cells including Fas cell surface death receptor (FasL/Fas) interactions (38) and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)/TRAILR engagement (39). Tolerogenic DCs may also express various inhibitory receptors like for example programmed cell death ligand (PDL)-1, PDL-2 (40, 41), inhibitory Ig-like transcripts (ILT) (42, 43), and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) (44), which act on T cells by dampening TCR signaling and competing with CD28, respectively. Tolerogenic DCs also alter T cell responses by modulation of metabolic parameters for example by the release of IDO and the induction of heme oxygenase-1 (HO-1) to control levels of tryptophan and carbon monoxide. IDO facilitates the degradation of tryptophan to N-formylkynurenin leading to reduced T cell proliferation (45, 46), whereas HO-1 inhibits hemoglobin, resulting in production of carbon monoxide, which leads to reduced DC immunogenicity (47, 48). In addition, tolerogenic DCs are also capable of producing retinoic acid (RA) (49), inducing Treg differentiation (50). Shedding of CD25 by DCs and subsequent deprivation of IL-2 was recently proposed as additional immunosuppressive mechanism for the suppression of effector T cell proliferation (13).

Several human tolerogenic DC subsets have been characterized *in vitro* based on their tolerogenic capacities. iDCs display minimal expression of co-stimulatory molecules and no secretion of inflammatory cytokines, demonstrating the aforementioned optimal requirements for tolerance induction, which has also been demonstrated *in vitro* (51). However, iDCs are unstable and may differentiate into immunogenic DCs under inflammatory conditions (13, 52).

Therefore, many protocols have been established to generate stable human DCs with tolerogenic capacities *in vitro* (**Figure 3**). The opportunity to genetically modify human DCs has been exploited to directly induce tolerogenic properties by the recombinant expression of FasL (53), PD-L1 plus TRAIL (54), or IDO (55) all of which lead to the induction of T cell apoptosis and suppression of effector T cell function, respectively. Additionally, DCs can genetically be engineered to secrete enhanced levels of IL-10 (56) or TGF-β (57) resulting in broad immunosuppression.

Furthermore, human tolerogenic DCs are induced by various immunosuppressive drugs (**Figure 3**) that are often systemically used to control excessive immune responses like corticosteroids, rapamycin, cyclosporine, or by acetylsalicylic acid (58). For instance, the corticosteroid dexamethasone is capable of inducing tolerogenic DCs that exhibit low expression of co-stimulatory molecules combined with highly expressed inhibitory receptors ILT-2 and ILT-3 and secrete large amounts of IL-10 and IDO resulting in the induction of T cells with regulatory capacities (15, 59–61). In a similar way, the immunosuppressive drug rapamycin, which inhibits mechanistic target of rapamycin persuades human DCs to express a stable tolerogenic phenotype with reduced expression of MHC and co-stimulatory molecules in combination with a high ILT-3 and ILT-4 expression, leading to Treg generation *in vitro* and *in vivo* (15, 59, 62–64). Furthermore, in the presence of acetylsalicylic acid, DCs downregulate the expression of co-stimulatory molecules, whereas inhibitory molecules like ILT-3 and PD-L1 are upregulated resulting in Treg induction (65).

In addition, incubation with the immunosuppressive cytokines TGF-β and IL-10 alone or in combination facilitated the generation of a tolerogenic DC phenotype (**Figures 2**–**4**) (48, 66). For instance, TGF-β dampens the antigen-presenting capabilities of DCs by downregulation of MHC and co-stimulatory molecules and upregulation of PDL-1 resulting in T cell anergy (15, 36, 59, 67, 68). The comprehensive tolerogenic properties of IL-10 induced tolerogenic DCs will be discussed in detail in the next chapter. However, other bioderivates are also capable of inducing DCs with tolerogenic function (**Figure 3**). In the presence of hepatocyte growth factor, DCs express various tolerance-inducing

molecules like IDO, IL-10, TGF-β and TRAIL, which may induce T cell with regulatory functions (48, 69). Furthermore, treatment of human DCs with vitamin D3 (VitD3) triggers a tolerance-inducing phenotype that is characterized by enhanced IL-10 secretion, augmented IDO production and the expression of PDL-1 and Trail. Resulting DC populations are either capable of inducing antigen-specific T cell apoptosis or expansion of Tregs (59, 60, 70–72). The immunoregulatory neuropeptide vasoactive intestinal peptide prevents full maturation of DCs and induces high IL-10 secretion (73). Furthermore, a recent study by Olivar et al. demonstrated tolerogenic capacities of human DCs that are generated in the presence of the complement factor H resulting in reduced expression of co-stimulatory molecules on DCs, enhanced IL-10 and TGF-β gene expression and induction of forkhead box P3 (FOXP3)<sup>+</sup> Tregs (74).

Since a comparative study by Boks et al. in 2012 identified human IL-10-modulated tolerogenic DCs (IL-10 DCs) as the most potent candidates for antigen-specific induction of tolerance *in vivo* (15), their generation and suppressive mechanisms will be highlighted in the next chapter.

### MECHANISM OF TOLERANCE INDUCTION BY HUMAN IL-10-MODULATED TOLEROGENIC DCs

Interleukin-10 DCs have been playing a pivotal and central role in the research field of tolerogenic DCs for over two decades. However, a variety of *in vitro* protocols exist for the generation of IL-10 DCs leading to a huge amount of data that are challenging to compare (**Figure 4**).

Interleukin-10-modulated DCs are usually generated from monocytes when cultured in the presence of IL-4 and GM-CSF to induce iDCs. The two most prominent protocols add IL-10 either during the whole culture (referred to as DC10s in the following) or at a later time point together with a maturation stimulus (referred to as IL-10 DCs) (13–15, 24, 75–79).

Gregori and colleagues generated DC10s using the first mentioned protocol and characterized them as CD14<sup>+</sup> CD16 <sup>+</sup>CD11c <sup>+</sup>CD11b <sup>+</sup>CD40 <sup>+</sup>HLA-DR <sup>+</sup>CD80 <sup>+</sup>CD83 <sup>+</sup> CD86<sup>+</sup>CD163<sup>+</sup> and CD1a<sup>−</sup>CD1c<sup>−</sup>CD68<sup>−</sup>CD115<sup>−</sup>MDC8 DCs. In contrast, mDCs are CD14<sup>−</sup> and CD16<sup>−</sup>, but express comparable amounts of CD80, CD83, CD86, and HLA-DR (77). It can be argued that the high expression of CD14 and CD16 indicate a macrophage-like phenotype (80). However, the lack of expression of the monocyte marker CD115 and the macrophage marker CD68 in combination with the constitutive expression of CD83, a DC-associated molecule, identified DC10s as immune cells of the DC lineage. Moreover, they show a DC-like morphology and are capable of driving naïve T cells to develop into antigen-specific Tr1 cells (77). DC10s also express the co-inhibitory molecules ILT-2, ILT-3, ILT-4, and HLA-G, and their capability to induce anergic Tr1 cells is ILT-4 dependent (77). Amodio et al. stated that the expression of HLA-G on DC10s is donor dependent and correlates with the expression of ILT-4 and with the frequency of Tr1 induction (78). These Tr1 cells are CD49b<sup>+</sup> and LAG3<sup>+</sup> and secrete high levels of IL-10, low IL-2, no IL-4, no IL-17, and variable amounts of IFN-γ (81–83). After LPS and IFNγ stimulation, mDCs and DC10s secrete comparable amounts of pro-inflammatory IL-6, but neither secrete the TH1-inducing cytokine IL-12, yet, DC10s produce slightly higher amounts of TNFα and considerably more IL-10 (77). The major function of tolerogenic DCs, the induction of Tregs, was found to be dependent on IL-10 secretion (77), similar to the upregulation of HLA-G on CD4<sup>+</sup> T cells through stimulation with DC10s (78).

Gregori et al. have also identified the *in vivo* counterpart of the *in vitro* generated DC10s as naturally occurring DC10s in humans*.* They were sorted from peripheral blood from healthy volunteers as CD14<sup>−</sup>CD11c<sup>+</sup>CD83<sup>+</sup> and analyzed for their poststimulation cytokine profile. In accordance with the findings from monocyte-derived DCs, their IL-12 secretion was negligible, but they produce relevant amounts of IL-6 and TNFα. Most importantly, their IL-10 levels were significantly increased compared to iDCs and mDCs (77).

Figure 4 | Phenotype of monocyte-derived interleukin (IL)-10 dendritic cells (DCs) obtained by different protocols. Immuno-activating and -inhibitory surface molecules as well as secreted signaling molecules and the T cell response are depicted. Arrows indicate up/downregulated or unchanged expression or secretion by human IL-10-modulated tolerogenic DCs compared to mature DC. IL-10 DCs are generated by addition of the immunosuppressive cytokine during the maturation step at the end of the culture, whereas DC10 are obtained by incubation with IL-10 during the entire culture period.

A slightly different phenotype can be observed, when IL-10 DCs are generated using the latter previously mentioned protocol in which IL-10 is added for the last 2 days of the culture during the maturation step (**Figure 4**). Here, presence of IL-10 prevents full DC maturation, indicated by intermediate expression of the co-stimulatory molecules CD80 and CD86, as well as the DC maturation marker CD83 (13, 75). The tolerogenic phenotype is further established through the increased expression of the co-inhibitory molecules ILT-3 and ILT-4 (13, 15). This IL-10 triggered surface marker modulation is dependent on glucocorticoid-induced leucin zipper, a transmembrane molecule, which blocks NF-κB, MAPK, and AP-1 (84). The results for the expression of HLA-DR and CD14 on IL-10 DCs are contradictory. Our studies revealed that the whole IL-10 DC population shows an intermediate HLA-DR expression and that a subpopulation of IL-10 DCs are CD14<sup>+</sup> (24), whereas other groups found that HLA-DR expression is comparable to mDCs and IL-10 DCs are exclusively CD14<sup>−</sup> (85). In comparison with mDCs, IL-10 DCs stimulate a reduced T cell activation and are capable of inducing an antigen-specific anergy in CD4<sup>+</sup> or CD8<sup>+</sup> naïve T cells (24, 75, 76). The induction of anergy is associated with the increased expression of the MAPK p38 and its effector molecules MAPKactivated protein kinases 2 and 3 (86) as they upregulate the expression of the cyclin-dependent kinase inhibitor 1B (p27Kip1), leading to a cell cycle arrest in the G1 phase (79). The induced Tregs in turn have the ability to efficiently suppress syngeneic effector CD4+ and cytotoxic CD8+ T cells in a cell-to-cell contactdependent and antigen-specific manner (15, 24, 79).

IL-10 DCs were identified as the most suitable candidate for DC-mediated tolerance-vaccination therapies as was shown by a comprehensive study by Boks et al. They compared five protocols for *ex vivo* induction of human tolerogenic DCs (through VitD3, dexamethasone, TGF-β, rapamycin, and IL-10) with regard to prerequisites for clinical applications in humans such as potent migratory capacity, sufficient Treg induction, and the stability of the tolerogenic phenotype under inflammatory conditions to guarantee the safety of the therapy (13, 15). The protocol using IL-10 for tolerogenic DC generation was shown to be superior as compared to the other tested protocols, with respect to the stability of the tolerogenic phenotype and the suppressive capacity of the induced Tregs (15). Boks et al. also revealed that co-maturation was indispensable for the stability of the phenotype and for the migratory capability in all protocols tested (15).

However, in their study, IL-10 DCs displayed a limited migratory capability due to a reduced CCR7 expression (15). This was confirmed by another comparative study by Adnan et al., which compared tolerogenic DC protocols in a similar way. However, they also showed that IL-10 DCs induce higher numbers of IL-10<sup>+</sup>CD4<sup>+</sup> Tregs than tolerogenic DCs generated with other protocols [involving protein kinase C inhibitor (PKCI), VitD3, dexamethasone, TGF-β, rapamycin, and peroxisome proliferatoractivated receptor γ + all-trans RA] and that the Tregs induced by both IL-10 and PKCI-treated tolerogenic DCs exhibited a higher suppressive capacity compared to Tregs induced by other tolerogenic DC protocols (13, 15, 59, 84, 85). In accordance with that, among all protocols tested by Boks et al., only IL-10 DC-induced Tregs exhibited a significantly enhanced suppressive function, compared to other tolerogenic DCs. Therefore, Boks et al. concluded that IL-10 DCs are the most suitable candidates for tolerogenic DC-based therapies for allergic and autoimmune diseases and transplantation rejections (15).

Recent investigations of our own laboratory refined this thesis by identifying two subpopulations of the human tolerogenic IL-10 DCs, distinguishable by the expression of CCR7 and CD83 (CD83highCCR7+ and CD83lowCCR7<sup>−</sup> IL-10 DCs) (13). Both IL-10 DC subsets were capable of inducing Tregs, but the CD83high IL-10 DC-induced Tregs exhibited a significantly enhanced suppressive capacity. It is evident from the proliferation, cytokine production, and surface makers that Tregs induced by CD83high IL-10 DCs exhibit a more activated phenotype compared to Tregs induced by CD83low IL-10 DCs. In addition, the tolerogenic phenotype of the CD83high IL-10 DC population was found to be extremely stable in the presence of IL-1β, IL-6, and TNFα, mimicking an inflammatory environment (13). In contrast to mDCs, IL-10 DCs and predominantly the CD83high subpopulation express increased amounts of membrane-associated and soluble CD25, the latter of which was found to play a role in the suppression of T cell proliferation (13). CD25 is known to exert seemingly contradicting functions: the membrane-bound molecule may be involved in the stimulation of T cells, whereas the soluble form attenuates T cell proliferation by trapping IL-2 (87, 88).

However, most importantly, dependent on their high expression of CCR7, the CD83high IL-10 DCs displayed a pronounced migratory capability that is superior to that of CD83low or unsorted IL-10 DCs (13, 81, 89). Therefore, in conclusion, the tolerogenic characteristics of the most promising population of tolerogenic DCs, IL-10 DCs can be further improved by sorting for CD83high IL-10 DCs.

### NANOPARTICLE-BASED *IN VIVO* INDUCTION OF TOLEROGENIC DCs

The above discussed protocols greatly expanded the knowledge of tolerogenic DC biology and enabled scientists to generate tolerogenic DCs that are stable under inflammatory conditions and may be used for antigen-specific clinical application. However, for this purpose, DC precursors need to be isolated from the patient's blood, modulated *ex vivo* and re-injected into the patient, which is a time-consuming and expensive process. In addition, recent data suggest that monocyte-derived DCs, which are used in such immunotherapeutic approaches may rather be allocated to the family of monocytes, which have less T cell stimulatory capacities than DCs *in vivo* (90, 91). Therefore, nanoparticle-based drug delivery systems that enable directed cell-type specific targeting *in vivo* in combination with delivery of multiple drugs in one formulation have emerged as another promising approach in DC-based immunotherapy.

For cell type-specific targeting, nanoparticles can be chemically conjugated to antibodies, peptides, carbohydrates, or cytokines that address receptors that are preferentially expressed on DCs (92–95). For instance, targeting of human DCs *in vivo* with subsequent antigen presentation and robust humoral and cellular responses can be achieved by antibodies against the c-type lectin receptor DEC205 as shown in a recent phase 1 clinical trial (96). In addition, other possible receptors that have been used to specifically target DCs include DC-SIGN, the mannose receptor, Fc receptors, CD40, or CD11c (93, 97, 98). Even though most approaches focus on induction of immunity for example in the context of tumor immunotherapy, cell-type-specific nanoparticle delivery is also a promising strategy to prevent excessive immune responses and induce DCs with tolerogenic capacity. For instance, polymeric synthetic nanoparticles that target DCs have been used to induce OVA-specific tolerance by delivery of rapamycin (99). In a similar approach, Zhang et al. were able to prevent antibody formation against substituted factor VIII (100). Intriguingly, Clemente-Casares et al. generated nanoparticles that target disease relevant peptides toward MHC II molecules, which subsequently trigger the expansion of antigen-specific Tr1 cells and regulatory B cells in different autoimmune disease models such as type 1-diabetes, inflammatory bowel disease, rheumatoid arthritis, and multiple sclerosis resulting in alleviation of disease symptoms (101).

### USE OF TOLEROGENIC DCs IN CLINICAL APPLICATIONS

Over the last decades, numerous trails with DC-based immunotherapies have been conducted using activated, mDCs to stimulate antitumor immune responses, and some have shown objective clinical benefits in patients with different types of cancer, including prostate cancer or malignant melanoma (102–104). Currently, several immunotherapeutic approaches are being studied using tolerogenic human DCs for treatment of inflammatory, autoimmune, and allergic diseases as well as transplant rejections (58, 105) (**Figure 5**).

In contrast to standard immunosuppressive therapies, which often do not specifically target the cause of disease and are accompanied by severe side effects, *ex vivo* generated tolerogenic DCs may be an attractive therapeutic approach to induce, enhance, or restore (antigen-specific) tolerance. After loading with exogenous or endogenous antigens, one major advantage of tolerogenic DC vaccination is their capability to act in an antigen-specific manner.

Evidence from several rodent models clearly showed the efficacy of tolerogenic DCs in the fields of inflammatory, autoimmune, and allergic disorders and transplantation medicine (58, 105). To translate the results into the human system, several *ex vivo* studies have been performed as proof of principle experiments demonstrating that human tolerogenic DCs efficiently inhibit disease-related immune responses, e.g., by induction of Tregs or T cell anergy and apoptosis. With regard to allergic diseases, *ex vivo* models have shown that human tolerogenic IL-10 DCs from atopic donors suppressed TH2 immune responses by induction of FOXP3<sup>+</sup> Tregs and dexamethasone-induced tolerogenic DCs activated IL-10 producing Tregs, specific for the latex Hev b 5 antigen, in rubber latex allergic patients (106, 107). Since tolerogenic DCs are also a promising tool to restore tolerance to specific tissue-derived autoantigens, several *ex vivo* studies have been conducted with tolerogenic DCs obtained from patients suffering from autoimmune disorders (58). Tolerogenic VitD3-treated DCs derived from precursor

cells of multiple sclerosis patients and loaded with myelin peptides induced a stable and antigen-specific hyporesponsiveness of autologous T cells (58, 108), which was shown to be TGF-β-dependent (109). Studies in type 1 diabetes patients revealed that tolerogenic DCs, generated either in the presence of vitamin D or of IL-10/TGF-β, and loaded with the pancreatic islet antigen glutamic acid decarboxylase 65 rendered antigenspecific T cells hyporesponsive toward a second challenge with fully competent, antigen-loaded DCs (66, 110). Furthermore, monocyte-derived DCs were obtained from systemic lupus erythematosus (SLE) patients, treated with dexamethasone/ rosiglitazone and loaded with self-antigens. Those tolerogenic DCs can modulate CD4<sup>+</sup> T cell activation and are a suitable tool for antigen-specific immunotherapy in SLE (111).

Although safety and feasibility of DC-based studies in general have already been shown, there are still a lot of open questions regarding the DCs manufacturing protocols, the route of application, the numbers of DCs, and the frequency and time points of injections. In addition, the characteristics of tolerogenic DCs, including the phenotype, migratory capacity, stability under inflammatory conditions, and the mode of action (induction/ activation of regulatory T and B cells, T cell anergy and apoptosis induction, interaction with other immune cells) have to be investigated and different protocols have to be compared with regard to these properties. Aiming to joint efforts in translating tolerogenic DCs into the clinic by harmonizing protocols and defining functional quality parameters, international co-operations in science, and technology network have been initiated (112, 113).

One major concern in the context of tolerogenic DC-based immunotherapies is the stability of the tolerogenic phenotype under inflammatory conditions as DCs express several patternrecognition receptors and receptors for growth factors and cytokines, which can be stimulated in an inflammatory environment. Therefore, clinical grade tolerogenic DCs must be intensively tested for a robust, stable phenotype to exclude a loss of the regulatory function and a switch to an immunostimulatory phenotype of the differentiated DCs, leading to an (antigenspecific) immune activation rather than to the intended immunosuppressive reaction. Comparative studies revealed that most of the tolerogenic *ex vivo* generated DC populations (by use of, e.g., IL-10, TGF-β, VitD3, rapamycin, dexamethasone, PKCI as described above), exhibit the aforementioned stable phenotype (15, 59). However, both reports demonstrated that tolerogenic IL-10 DCs showed the most powerful tolerogenic properties in terms of Treg induction with strong suppressive capacities. Another important feature is the CCR7-directed migratory capacity of tolerogenic DCs toward secondary lymphatic organs, resulting in the induction and generation of T cell-mediated immunosuppression. A recent study (as mentioned above) revealed that IL-10 DCs are consisting of two different populations, CD83highCCR7+ IL-10 DC and CD83lowCCR7<sup>−</sup> IL-10 DC subpopulations, both exhibiting tolerogenic properties, resulting in Treg induction (13). However, sorting of IL-10 DCs into these two subsets ascertained a significantly improved migratory capacity of the CD83highCCR7+ IL-10 DC subpopulation compared to CD83lowCCR7<sup>−</sup> IL-10 DCs, and to the non-separated IL-10 DC population as well. The stable phenotype, efficient CCR7-directed migration, and, in particular, pronounced tolerogenic capacity to induce Tregs with high suppressive activity of IL-10 DCs is a prerequisite for clinical grade DCs considered for vaccinations studies in humans.

Regarding the route of DC administration, different applications have been used in humans. Tolerogenic DCs have been injected intraperitoneally in patients suffering from Crohn's disease (114), intradermally in diabetes, and rheumatoid arthritis patients (115, 116), subcutaneously in rheumatoid arthritis patients (117) and *via* arthroscopic injections in joints of patients with rheumatoid or inflammatory arthritis (118). In all studies, the route of administration has been well tolerated without any signs of toxicity. Likewise reports of intravenous injections of tolerogenic DCs into nonhuman primates revealed their safety (119).

The first attempt to apply tolerogenic DCs to humans was undertaken by Ralf Steinman's group in 2001 (120, 121). They showed that subcutaneous applications of human immature tolerogenic DCs (2 × 106 ), generated in the presence of IL-4 and GM-CSF and pulsed with antigens, into healthy subjects was well tolerated and suppressed antigen-specific CD8<sup>+</sup> T cell responses up to 6 months. Thus, they pioneered to demonstrate the tolerogenic potential of DCs in humans *in vivo.*

Several protocols for tolerogenic DCs have been tested in phase I trials with highly encouraging results from a safety point of view and in terms of adverse effects such as allergic reactions, exacerbations of autoimmunity, and pro-inflammatory immunity (114–118) (**Table 1**).

The first clinical trial with tolerogenic DCs was carried out in 10 patients suffering from diabetes type 1 in 2011. They were injected intradermally four times at 2-week intervals with 1 × 107 autologous DCs which have been either un-manipulated (controls) or have been treated with antisense oligonucleotides targeting CD40, CD80, and CD86 to silence these surface molecules. DC treatment was well tolerated without any adverse effects and did not induce autoantibody production (115). Analysis of the immune response revealed no alterations with exception of increased IL-4/IL-10 levels and elevated frequencies of a regulatory B220<sup>+</sup>CD11c<sup>+</sup> B cell population. Importantly, the patients did not lose their ability to mount T cell responses, e.g., to pathogens, demonstrating the absence of a general immune suppression.

Another clinical trial was conducted to analyze the impact of tolerogenic DCs in nine patients suffering from Crohn's disease (114). Here, under an escalating protocol tolerogenic, DCs (treated with dexamethasone and VitD3) were intraperitonally injected in once or biweekly intervals, respectively. The DC vaccination was well tolerated and did not induce adverse effects from week 1 to 12 and in a follow-up up to 12 months.

In the field of rheumatoid arthritis, three trials have been published to date. In one study, 12 patients were subcutaneously injected with a low (0.5 × 107 ) or high dose (1.5 × 107 ) of autologous DCs for five times at 2- to 4-week intervals (117) (**Table 1**). The tolerogenic DCs were pulsed with protein arginine deiminase 4, heterogeneous nuclear ribonucleoprotein A2/B1 (RA33), citrullinated filaggrin, and vimentin antigens (=CreaVax-RA). The authors observed only a few patients with grade 1 or 2 adverse effects, but a combination of a significant decrease in autoantibody levels and a good-to-moderate EULAR response at 14 days after initiation of the trial, which was more pronounced in the DC high-dose group. Bell et al. reported the results of another dose escalation trial (AUTODEKRA trial) with rheumatoid arthritis patients who were intra-articularly treated with tolerogenic DCs (1 × 106 , 3 × 106 , or 10 × 106 ), generated in the presence of dexamethasone and VitD3 and loaded with autologous synovial fluid as source of autoantigens (118). No target knee flares and other severe side effects were observed. The authors did not find any trends in disease activity scores (DASs) or in consistent alteration of immune parameters in the peripheral blood; however, patients with the highest dose exhibited an improvement of the clinical symptoms.

In the study of Benham et al., tolerogenic DCs were generated in the presence of an NF-κB inhibitor, resulting in CD40 deficient but highly CD86 expressing tolerogenic DCs, which were administered to rheumatoid arthritis patients (116). For an antigen-specific immune response, DCs were pulsed with four different citrullinated peptide antigens (Rheumavax). 18 patients were injected intradermally with a single dose of the tolerogenic DCs (either 1 × 106 or 5 × 106 ). Evaluation of the patients after 1, 3, and 6 months revealed that the vaccination was well tolerated and no side effects in form of (auto-) inflammatory reactions have been observed. The authors found a reduction in effector T cells and several inflammatory mediators and an increased regulatory to effector T cell ratio in the patients. In addition, the DAS was decreased within 1 month in vaccinated patients with active rheumatoid arthritis, indicating the biological and clinical activity of this therapy.



Further phase I/II studies are under way in the fields of allergic diseases (allergic asthma), autoimmunity (Crohn's disease, diabetes type 1, rheumatoid arthritis, and multiple sclerosis), and transplantation medicine (kidney transplantation) (https:// clinicaltrials.gov).

A multitude of protocols has been developed to generate human tolerogenic DCs that can be tailored to induce specific tolerance. These innovative and attractive tools represent a promising therapeutic approach to treat inflammatory, autoimmune and allergic diseases, and transplant rejections. However, there is a high need to define optimal vaccination protocols and to identify the underlying immune mechanism of tolerance induction by human DCs in more detail. In this context, high-throughput approaches, e.g., in form of genomics and proteomics, will be of great help to analyze critical pathways contributing to programming and function of human tolerogenic DCs (122). Furthermore, next-generation tolerogenic DC vaccines should be integrated into future combinatorial immunotherapy regimes, including biologicals, nanoparticles, and *in vivo* targeting of DCs. So, it was demonstrated that combination of tolerogenic DCs with CTLA-4Ig strengthen their tolerogenic effect (123).

### CONCLUSION

Dendritic cells are the most potent professional antigenpresenting cells of the immune system and bridge innate and adaptive immunity by interacting with a large number of different cell types, thereby initiating and regulating adaptive immune responses. Hence, DCs are promising targets for immunotherapy either for initiating immunity as for example desired for the clearance of pathogens or antitumor immunotherapy or for the objective to alleviate unwanted and excessive immune responses in allergic and autoimmune disorders. Multiple *ex vivo* protocols have been established to induce stable tolerogenic human DCs exhibiting numerous different mechanisms to dampen immune responses. Those DCs may be used for antigen-specific induction of tolerance *in vivo,* which would be exceptionally beneficial for the therapy of allergic and autoimmune disease or in transplantation medicine. Progress in the fields of improved immunization protocols, genome editing, expression of recombinant proteins, and nano-dimensional drug delivery may contribute to overcome obstacles and to open up new unexpected approaches to

### REFERENCES


improve the promising therapeutic option of DC vaccination for the future.

### AUTHOR CONTRIBUTIONS

All authors wrote and critically read the manuscript.

### FUNDING

This work was supported by the German Research Foundation (DFG): STE791/9-1, CRC 1066/B6, TR156 A4/C5, and by the German Cancer Aid (110631), and by intramural grants (all to KS). MD was supported by a fellowship of the Max Planck Graduate Center (MPGC) Mainz.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationship that could be construed as a potential conflict of interest.

*Copyright © 2017 Domogalla, Rostan, Raker and Steinbrink. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# The Immunomodulatory Potential of tolDCs Loaded with Heat Shock Proteins

*Willem van Eden1 \*, Manon A. A. Jansen1 , A Charlotte MT de Wolf1 , Irene S. Ludwig1 , Paul Leufkens <sup>2</sup> and Femke Broere1*

*1Department of Infectious Diseases and Immunology, Faculty of Veterinary Medicine, Utrecht University, Utrecht, Netherlands, 2 Trajectum Pharma, Utrecht, Netherlands*

Disease suppressive T cell regulation may depend on cognate interactions of regulatory T cells with self-antigens that are abundantly expressed in the inflamed tissues. Heat shock proteins (HSPs) are by their nature upregulated in stressed cells and therefore abundantly present as potential targets for such regulation. HSP immunizations have led to inhibition of experimentally induced inflammatory conditions in various models. However, re-establishment of tolerance in the presence of an ongoing inflammatory process has remained challenging. Since tolerogenic DCs (tolDCs) have the combined capacity of mitigating antigen-specific inflammatory responses and of endowing T cells with regulatory potential, it seems attractive to combine the anti-inflammatory qualities of tolDCs with those of HSPs.

#### *Edited by:*

*Shohei Hori, The University of Tokyo, Japan*

#### *Reviewed by:*

*Irun R. Cohen, Weizmann Institute of Science, Israel Silvia Gregori, San Raffaele Telethon Institute for Gene Therapy (SR-Tiget), Italy*

> *\*Correspondence: Willem van Eden w.vaneden@uu.nl*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 30 October 2017 Accepted: 16 November 2017 Published: 30 November 2017*

#### *Citation:*

*van Eden W, Jansen MAA, de Wolf ACMT, Ludwig IS, Leufkens P and Broere F (2017) The Immunomodulatory Potential of tolDCs Loaded with Heat Shock Proteins. Front. Immunol. 8:1690. doi: 10.3389/fimmu.2017.01690*

Keywords: self-tolerance, autoimmunity, heat shock protein, regulatory T cell, peptide

### INTRODUCTION

Regulatory T cells (Tregs) downmodulate unwanted immune responses. The induction or expansion of Tregs with the use of antigen-loaded tolerogenic dendritic cells (tolDCs) is a novel and attractive therapeutic possibility. Tregs are predominantly immunosuppressive CD4<sup>+</sup> T cells, selected in the thymus on the basis of relatively high-affinity interactions with self-antigens (1). These cells are called natural Tregs (nTregs). Alternative populations of Tregs may become induced by peripheral antigen presentation in tolerance promoting tissues leading to what is called peripheral or induced Tregs (pTreg). Comparisons between T cell receptor (TCR) repertoires of thymus-derived nTregs and gut-residing pTregs have not led to direct conclusions on the relative significance of either nTregs or pTregs. Some studies, such as those for the colon, have reported a relatively unique nature of TCRs present on gut Tregs (2), whereas others have emphasized the presence of shared TCRs between thymic and colon Tregs (3, 4). Nonetheless, in both situations, it is proposed that gut-residing Tregs expand after recognition of microbiota-associated microbial antigens. The cognate interactions with these foreign antigens would fit well with a unique nature of colon Treg TCRs, whereas a driving activity by self-cross-reactive microbial antigens would be more compatible with the shared TCR idea.

Microbial heat shock proteins (HSPs) are antigens with a well-established tolerance promoting capacity. Since the identification of mycobacterial HSP60 as the driving antigen for modulatory T cells in the adjuvant arthritis model (5), immunization with microbial HSPs, mainly HSP60 and HSP70, was shown to inhibit disease in various inflammatory models (6–8). Subsequent analysis of the specificities of the microbial antigen responding T cells led to the hypothesis that conservation of HSPs was critical in the tolerance promoting character of HSPs (9). T cells with specificity for van Eden et al. Therapeutic Tolerance by HSP-Loaded tolDC

conserved microbial sequences are cross-reactive with (mammalian) self-HSPs and therefore have potential to regulate by targeting a regulatory effect to upregulated self-HSP in inflamed tissues (6, 10). A possible sequence of mechanistic events (see **Table 1**) would be that T cells, or Tregs in this case, are selected in the thymus on the basis of self-HSP recognition (**Table 1**). For example, HSP70 is abundantly expressed in normal thymic epithelial cells (11). In addition, HSP70 epitopes were found on thymic dendritic cells (DCs) which demonstrates that these epitopes are presented in the healthy human thymus (12). For these reasons, central tolerance may be assisted by HSP-driven thymic-positive selection. In the periphery, HSP recognizing T cells are maintained or expanded in the tolerizing gut environment through recognition of cross-reactive microbiota HSPs. Several of the following factors could add to the efficiency of this immune imprinting of HSP reactivity at the Treg level. When microbes are being taken up by macrophages or DCs lining the gut, phagocytosis and exposure of the ingested bacteria to the hostile intracellular environment of the phagocyte will lead to a microbial HSP upregulating stress response (13). In addition, the continuous contact with a variable set of microbiotaassociated bacterial species may emphasize the driving nature of just the conserved and therefore repeatedly encountered bacterial sequences. Similarly, stress in host tissues as seen in inflammation will also lead to the enhanced expression of self-HSPs and thereby enhance attractiveness as a target for T cell regulation.

In the next sections, we will explain how the anti-inflammatory effects of HSPs can work in synergy with the cell therapy approach with tolDC.

### DIFFERENT APPROACHES TO INDUCE tolDCs AND THEIR FUNCTION IN EXPERIMENTAL MODELS

Experimental disease models have been used to explore the abilities of tolDCs to induce Tregs and their possible therapeutic application *in vivo* (14–16). Among these studies, tolDCs were also tested in an arthritis model. In this study, tolDCs were generated *in vitro* from murine bone marrow with dexamethasone and 1α,25-dihydroxyvitamin D3 (the active form of vitamin D3). Subsequently, they were pulsed with collagen type II (17). These tolDCs, showing a semi-mature phenotype, were able to reduce T cell proliferation and diminish arthritis severity. Unpulsed tolDCs were not able to reduce arthritis. This suggests that antigen is needed to suppress disease. Choosing the right antigen is important since most autoimmune diseases are caused by a

TABLE 1 | Mechanistic sequence of events leading to anti-inflammatory activity of heat shock protein (HSP)-specific regulatory T cell (Tregs).

deviant lymphocytic response against a self-antigen. Whereas rheumatoid arthritis (RA) affects one body component (the joint), other autoimmune diseases such as systemic lupus erythematosus (SLE) affect multiple organs. Can tolDC therapy also be applied in these type of autoimmune diseases? And could we induce tolDCs *in vivo* if the autoantigen is unknown?

Systemic lupus erythematosus is a multisystem autoimmune disease in which autoantibodies play an important role. Another feature of SLE is dysregulation of DCs as these cells continuously display a mature phenotype with high expression of costimulatory molecules and chemokine receptors (18, 19). Because dysregulated DCs are involved in the pathogenesis of SLE, DC therapy could contribute to the welfare of SLE patients. Monocytederived DCs (moDCs) from SLE patients were isolated as a first step in investigating the possibility of tolDC therapy. The moDCs were stimulated with iC3b-opsonized apoptotic cells or the combination of dexamethasone and vitamin D3. In both cases, the DCs gained tolerogenic properties (20, 21). This indicates that SLE DCs can be modified. Up until now, no *in vivo* studies have been performed with *in vitro* generated tolDCs in SLE but blocking NF-κB activity in DC-induced tolerogenic characteristics. SLE mice treated with these NF-κB blockers showed a reduction in circulating autoantibodies (22). These results suggest that inducing tolDCs *in vivo* could be a solution in SLE. The same group showed that this method is also successful in experimental immune encephalomyelitis (EAE) (23). Antigen-specific effects of NF-κB activity-blocked DCs in EAE were studied with MOG as the autoantigen. In this model also bone marrow-derived tolDCs loaded with MOG peptide (40–55) were found to reduce disease by the induction of Treg (15).

Next to using NF-κB blockers in SLE and EAE, they have also been used to generate tolDCs *in vitro* with regard to RA. More specifically, addition of the NF-κB inhibitor Bay11-7082 to bone marrow- or peripheral blood-derived DCs caused a lower expression of costimulatory molecules and weak stimulation of T cells (24, 25). These DCs, when pulsed with antigen and injected into mice, attenuated inflammatory arthritis via the induction of Tregs. To test if this could also be achieved *in vivo* without modulating DCs *in vitro*, solely liposomes containing antigen and a NF-κB inhibitor were infused into arthritic mice. Arthritic symptoms were reduced only when the antigen was codelivered with the NF-κB inhibitor. Merely the delivery of antigen or NF-κB inhibitor did not reduce arthritis (26). This shows that liposomes carrying both antigen and NF-κB inhibitor can target DCs *in vivo* to induce antigen-specific tolerance.

Other drug delivery systems have also been used to influence the status of a DC *in vivo*. DC stimulation with intranasal antigen encapsulated by polylactic-*co*-glycolic acid (PLGA) nanoparticles resulted in increased antigen uptake of the DCs and induction of CD4+FoxP3+ cells. Furthermore, PLGA nanoparticle treatment was tested in a delayed type hypersensitivity (DTH) model and an arthritis model. In the DTH model, the mice were treated with PLGA nanoparticles or control and subsequently sensitized with ovalbumin (OVA) in combination with incomplete Freund's adjuvant and after 24 h challenged with OVA. PLGA nanoparticle treatment resulted in a reduced sensitivity reaction, whereas the controls did not (27). Next to

<sup>(1)</sup> HSP expression in thymic epithelial cells

<sup>(2)</sup> Loading of HSP peptides into MHC class II of positively selecting thymic epithelial cells

<sup>(3)</sup> Repertoire of HSP-specific Tregs expanded and maintained by crossrecognition of conserved microbial (microbiota) HSP peptides in the gut

<sup>(4)</sup> HSP overexpression due to inflammation (stress) in tissues

<sup>(5)</sup> Selective targeting of HSP-specific Tregs to inflamed tissues

this, nasal application with PLGA nanoparticles encapsulating mB29a (a mammalian HSP70 peptide) reduced arthritis severity for 30 days after disease development, suggesting that chronic inflammatory responses can also be modulated by tolDC (27). All in all, substantial evidence has been collected in preclinical models for an effective tolerance promoting effect of tolDC in autoimmunity.

### A Treg-INDUCING HSP70 PEPTIDE

Since the autoantigen in many autoimmune diseases is unknown, surrogate autoantigens could be used to restore tolerance. Among mammalian HSP, the HSP70 family of proteins contains some of the most stress-inducible HSPs, besides constitutive family members. In addition, some HSP70 family members are involved in chaperone-mediated autophagy (CMA). CMA contributes to maintenance of cellular homeostasis by facilitating recycling of degraded proteins and by eliminating abnormal or damaged proteins. HSP70, in combination with HSP90, is responsible for the targeting of proteins to the lysosome during CMA. MHC class II elution studies have shown that autophagy, as a consequence of cell stress, contributes to preferential loading of MHC II with HSP70 peptides (28). In the latter study, nutrient-deprived human HLA-DR4<sup>+</sup> B cells were used for the analysis. Among the more abundant peptides present in the elution profile of the stressed B cells, there was a peptide that had been previously discovered by us as a dominant T cell epitope in Balb/c mice previously (29). This peptide was discovered, when immunizations with mycobacterial HSP70 were found to protect against disease in the proteoglycan-induced arthritis (PGIA) model. This particular peptide, called B29, triggered disease protective T cell responses and consisted of a highly conserved sequence (29). This mycobacterial HSP70-B29 peptide had mammalian homologs (counterparts), called mB29a, mB29b, and mB29c in both constitutive and stress-inducible HSP70 family members. Exactly the mB29b variant was present in the elution profile of these HLA-DR4<sup>+</sup> B cells. Interestingly enough, the same mB29b variant was also reported to be present in the MHC-II clefts of human thymic antigen-presenting cells (12).

Our interest in the B29 peptides developed from the observations that nasal application of the peptide-suppressed PGIA in mice (29). Follow-up experiments made clear that B29 and its mammalian homologs were capable of inducing Tregs. Immunizations with B29 or an ovalbumin peptide (pOVA) as a control were performed, and CD25<sup>+</sup> T cells were sorted by FACS from the responding CD4<sup>+</sup> T cell population. Adoptive transfer of these CD25<sup>+</sup> T cells led to reduction of PGIA in recipients, whereas CD25− T cells did not. Also CD25+ and CD25− populations obtained from pOVA immunized animals were not having any effect on arthritis. The cell numbers needed for reaching these effects were relatively low: 3 × 105 cells sufficed for prevention of disease by adoptive transfer prior to disease induction, whereas only 1 × 106 T cells were needed to suppress ongoing disease.

By the use of a congenic T cell marker (CD90.1), exclusively present on the transferred T cells, we were able to track the transferred T cells *in vivo* (13). Our transferred, disease suppressing, T cells were still found present 50 days after transfer in the spleen, draining lymph nodes, joints, bone marrow, and blood. In addition, they were found to have kept their Treg phenotype, as they expressed CD25, Foxp3, NRP-1 (neuropilin 1), and lymphocyte activation gene-3 (LAG-3). When we infused a depleting CD90.1 antibody during the phase of disease suppression, after transfer of the CD25<sup>+</sup> T cells from B29 immunized donors, disease relapsed, reaching a severity identical with that in CD25<sup>−</sup> T cell-transferred animals. Altogether, these experiments showed the potential of conserved HSP70 peptides to induce a Treg response, which is long-lived and actively engaged in suppression of disease. When we sorted the LAG-3 positive CD4<sup>+</sup> T cells from our CD25<sup>+</sup> population and transferred these cells before induction of PGIA, we prevented induction of disease with the very small number of 4,000 CD4<sup>+</sup> T cells (29). As far as we know, this has never been seen in mouse models before, and it may indicate the superiority of antigen-selected Tregs as compared to non-antigen-selected Tregs. Furthermore, it also shows the strong potential of targeting Tregs to HSPs since the pOVA-induced Tregs never suppressed disease. Herewith, the example of HSP70 peptide B29 shows the potential of HSP peptides to effectively modulate immunity by the induction of Tregs.

### HSP-MEDIATED DC MODULATION

Given the protective effects of mycobacterial HSP70 in arthritis models, the immune modulatory activities of this molecule were also studied in RA. Bonorino and her group have demonstrated that mycobacterial HSP70 is capable of inducing IL-10 production in cells obtained from the inflamed synovium of arthritis patients. In addition, this was found in peripheral blood mononuclear cells from RA patients and healthy controls. Besides this, TNF-α and INF-γ production in these cells decreased and IL-10 production was raised. Cell-separation studies showed that the cells that produced IL-10 were monocytes (30).

When mouse bone marrow-derived DC were exposed to mycobacterial HSP70, maturation markers MHC class II and CD86 remained suppressed, indicating a tolerogenic phenotype (31, 32). Also in the presence of lipopolysaccharide (LPS), HSP70 reduced the upregulation of these markers. As in human cells, mycobacterial HSP70 induced IL-10 and not TNF-α. Altogether, it was concluded that DC maturation was halted by mycobacterial HSP70. Furthermore, LPS-free mycobacterial HSP70 was seen to inhibit phytohemagglutinin-induced T cell proliferation and was not seen to induce CD86 expression on splenic DCs *in vivo*, whereas LPS did (31). Although various alternative receptors were claimed to act as cellular receptors for HSP70, the signaling leading to IL-10 production in these studies may have involved TLR2 triggering with MyD88 activation and ERK phosphorylation (33).

When HSP70-treated DCs were tested in the proteoglycan (PG)-induced arthritis model, suppression of disease induction was seen when the treated DCs were loaded in addition with PG (32). And interestingly, when OVA-specific (TCR transgenic) T cells were cotransferred together with OVA-pulsed HSP70 treated DCs, the OVA-specific T cells were producing increased levels of IL-10 when re-stimulated *in vitro* with OVA. This regulatory cytokine induction in DCs was seen for both mycobacterial and mammalian HSP70 (32).

Thus, it may be concluded that part of the tolerance-promoting effects of HSP is mediated through their capacity to induce a regulatory mode in DCs.

### COINDUCTION OF ENDOGENOUS HSPs

Cell stress, caused by environmental factors or endogenous factors such as accumulation of unfolded proteins in the cytosol, is the primary trigger for upregulation of endogenous HSPs in cells. Such upregulation of HSPs lead, among others as the result of stress induced autophagy as discussed earlier, to further routing of HSP peptides to the MHC class II-binding grooves for recognition by T cells.

One possibility for enhancing HSP expression during stress is through the application of the so-called HSP coinducers. These are compounds that help to enhance production of HSP during stress, but are not capable of initiating a stress response on their own. When heat shock factor has become activated, this transcription factor induces the HSP synthesis, and coinducers just boost the level of production, through as yet unknown mechanisms.

A first and well-studied HSP coinducer, which is rather selective for the coinduction of HSP70 and not the others, is geranylgeranylacetone (GGA), an acyclic polyisoprenoid. Originally, GGA was developed as an effective anti-gastric ulcer drug. It has been tested now in various inflammatory diseases, including experimental autoimmune uveoretinitis (34). In this model, oral GGA inhibited disease and local HSP70 mRNA expression in the eyes was transiently upregulated. The autoantigen-specific T cell proliferation was also suppressed in GGA-treated mice (34).

A more recent example of an effective HSP coinducer is carvacrol, an essential oil present in Oregano species. Carvacrol was known to have antibacterial activity but upon testing on mammalian cells, carvacrol was found to have strong stress

T cells. (5) The cells are re-introduced into the patient ③ (remission allows for better tolerance induction). (6) The epitope is presented to Tregs by the tolerogenic DCs ④, to activate the regulatory T cells. (7) The patient now has a Treg repertoire that naturally suppresses inflammation ⑤ (in the joint).

protein coinducing capacity (35). Carvacrol promoted HSP70 expression in human cell lines and mouse spleen cells during stress *in vitro* (caused by a raised temperature or exposure to arsenite). Upon intragastric administration, it resulted in raised HSP70 gene expression in Peyer's patches of mice *in vivo*. As a consequence, the same intragastric administration of carvacrol specifically promoted T cell recognition of endogenous HSP70, as demonstrated by amplified T cell responses to HSP70. It also systemically increased the number of CD4<sup>+</sup>CD25<sup>+</sup>FoxP3<sup>+</sup> T cells in the spleen, and almost completely suppressed PG-induced experimental arthritis.

Very similar to what was seen for HSP-treated DCs loaded with PG in the PGIA model, as mentioned in the previous section, heat stress (42.5°C for 1 h) and carvacrol treated DCs loaded with PG also protected against PGIA. In addition, these DCs induced expansions of Foxp3<sup>+</sup> cells *in vivo* (17).

Herewith, it can be concluded that apart from the loading of DCs with HSP peptides, it would be possible in theory to use upregulated cell-endogenous HSPs for getting HSP peptides presented in tolDCs.

### THE ATTRACTIVE POSSIBILITY OF HSP PEPTIDE-LOADED tolDCs

Anti-inflammatory interventions using antigen-loaded tolDCs are already being developed for several chronic inflammatory diseases, such as diabetes type I and RA. First clinical trials indicated safety and have suggested clinical benefits (36, 37). In these cases, presumed relevant autoantigens were used to load the DCs.

In multiple sclerosis (MS), tolerogenic moDCs from relapsing-remitting (RR) MS patients, loaded with myelin peptides as specific antigen, were studied. The RR-MS tolDCs expressed

### REFERENCES


a stable semi-mature phenotype and induced a stable antigenspecific hyporesponsiveness in myelin-reactive T cells from RR-MS patients *in vitro* (38).

A recent clinical trial in RA has attempted to induce tolerance for self-antigens present in the synovial fluids of inflamed joints (37). Nevertheless, the target group here was patients with active disease, which may have hindered the chances for real tolerance induction. In addition, not a well-defined antigen for immune monitoring was available in this study.

In the case of HSP-loaded tolDCs (see **Figure 1**), the use of a well-defined HSP antigen will help the exact monitoring of induced HSP-specific Tregs with defined specificity in clinical experiments. In addition, patients can be selected on the basis of their antigen-specific response profiles for inclusion in the trial. Given the fact that molecules such as HSP70 are upregulated in inflamed tissues, therapeutic tolerance can be achieved through the induction of HSP70-specific Tregs, which then *via* bystander suppression reduce inflammation irrespective of the disease or inciting autoantigens. Therefore, when shown to be effective, HSP peptide-loaded tolDCs could be of use for therapies directed toward inflammatory diseases in the broadest possible sense.

### AUTHOR CONTRIBUTIONS

WE wrote most of the paper. MJ wrote part of the paper. IL, CW, PL, and FB were essential for discussing the content of the paper.

### ACKNOWLEDGMENTS

We thank the Dutch Reumafonds for their support in the preclinical development of tolDCs loaded with HSP 70 peptides for the induction of tolerance.


can both prevent and treat autoimmune diabetes in NOD mice. *Mol Ther* (2010) 18:2112–20. doi:10.1038/mt.2010.146


arthritis. *Proc Natl Acad Sci U S A* (2012) 109:14134–9. doi:10.1073/pnas. 1206803109


**Conflict of Interest Statement:** WE and PL have shares in Trajectum Pharma. The co-authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 van Eden, Jansen, de Wolf, Ludwig, Leufkens and Broere. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

#### *Andre Lamurias1 \*, João D. Ferreira1 , Luka A. Clarke2 and Francisco M. Couto1*

*<sup>1</sup> LaSIGE, Faculdade de Ciências, Universidade de Lisboa, Lisboa, Portugal, 2BioISI: Biosystems & Integrative Sciences Institute, Faculdade de Ciências, Universidade de Lisboa, Lisboa, Portugal*

Tolerogenic cell therapies provide an alternative to conventional immunosuppressive treatments of autoimmune disease and address, among other goals, the rejection of organ or stem cell transplants. Since various methodologies can be followed to develop tolerogenic therapies, it is important to be aware and up to date on all available studies that may be relevant to their improvement. Recently, knowledge graphs have been proposed to link various sources of information, using text mining techniques. Knowledge graphs facilitate the automatic retrieval of information about the topics represented in the graph. The objective of this work was to automatically generate a knowledge graph for tolerogenic cell therapy from biomedical literature. We developed a system, ICRel, based on machine learning to extract relations between cells and cytokines from abstracts. Our system retrieves related documents from PubMed, annotates each abstract with cell and cytokine named entities, generates the possible combinations of cell–cytokine pairs cooccurring in the same sentence, and identifies meaningful relations between cells and cytokines. The extracted relations were used to generate a knowledge graph, where each edge was supported by one or more documents. We obtained a graph containing 647 cell–cytokine relations, based on 3,264 abstracts. The modules of ICRel were evaluated with cross-validation and manual evaluation of the relations extracted. The relation extraction module obtained an F-measure of 0.789 in a reference database, while the manual evaluation obtained an accuracy of 0.615. Even though the knowledge graph is based on information that was already published in other articles about immunology, the system we present is more efficient than the laborious task of manually reading all the literature to find indirect or implicit relations. The ICRel graph will help experts identify implicit relations that may not be evident in published studies.

#### Keywords: tolerogenic therapy, text mining, knowledge graph, cytokines, machine learning

### 1. INTRODUCTION

Tolerogenic cell therapies provide an alternative to conventional immunosuppressive treatments of autoimmune disease and address, among other goals, the rejection of organ or stem cell transplants (1). These therapies aim at modulating the pathological immune response with minimal effect on the immune system. Antigen-presenting cells (APCs) can be induced to control the immune response by targeting specific T cell responses, avoiding general suppression of the immune system (2). It is necessary to understand the underlying mechanisms of the immune system to develop

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Phillip Lord, Newcastle University, United Kingdom James A. Hutchinson, Universitätsklinikum Regensburg, Germany*

*\*Correspondence: Andre Lamurias alamurias@lasige.di.fc.ul.pt*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 27 September 2017 Accepted: 13 November 2017 Published: 29 November 2017*

#### *Citation:*

*Lamurias A, Ferreira JD, Clarke LA and Couto FM (2017) Generating a Tolerogenic Cell Therapy Knowledge Graph from Literature. Front. Immunol. 8:1656. doi: 10.3389/fimmu.2017.01656*

tolerogenic cell therapies. Cytokines are small peptides involved in cell signaling, which can be used to induce tolerance in APCs (3). Immune cells express cytokines and their respective receptors. High-throughput sequencing techniques have improved our knowledge about cell signaling, introducing a variety of information about how cytokines are used by the immune system. This information is important to understand and develop new methods to isolate, culture, and induce tolerance in APCs.

Biomedical information is often presented to the community through published literature, including information about human autoimmune diseases and therapies to treat them. There are knowledge bases aiming at organizing the findings provided by the literature through a single access point. Populating such knowledge bases is, therefore, important for biomedical research, in particular, because they allow computational methods to find patterns in the data, thus generating new hypotheses to be tested experimentally. If a cell produces the same cytokine receptors as another cell, and a new cytokine is found to interact with the first cell, it is plausible that new cytokine could also affect the second cell. This type of inference, also known as ABC model (4), is only possible if the results of many studies are analyzed together.

The scientific community has shown interest in curating databases about cells and cytokines. For example, the National Center for Biotechnology Information (NCBI) provides a compilation of several biomedical and genomic resources (5), including the Entrez Gene database (6). This database contains entries for the genes associated with cytokines, and each entry contains useful information about that cytokine, such as interactions, pathways, and gene ontology annotations. There are also resources specific to cytokine information. The Cytokine Reference is an online database of information on cytokines and receptors, compiled from the literature by experts (7). This database contains links to other databases such as MEDLINE and GenBank, and can be searched by cytokine, cell or disease. Another relevant database is the Cytokine & Cells Online Pathfinder Encyclopedia (COPE)1 , which focuses on the interactions between cell types through cytokines. The current version of COPE contains 45k entries, including a cell type dictionary of 3k entries. These efforts show the importance of information structures for cells and cytokines. Therefore, the development of computational methods to structure this information would benefit researchers working in this domain.

These computational methods require two conditions: (i) the information is readable by computers and (ii) it is comprehensive, encoding the up-to-date collective knowledge of the community. Both these tasks are currently subject to intensive research. Converting heterogeneous data formats to a common language and merging the data is one approach to the first task. For example, Bio2RDF converts heterogeneous data from several datasets into RDF, a standard data model based on the specification of links between data elements (8).

As for the second task, the information stored in many biomedical datasets is the result of manual processing of documents,

This manuscript presents the system, Identifying Cellular Relations (ICRel), that we developed, based on machine learning, to extract cell–cytokine relations from documents and generate a knowledge graph. ICRel was trained and evaluated with the immuneXpresso database to extract meaningful relations between cells and cytokines in documents. We did not aim at finding novel information, instead we demonstrate the utility of the system by studying the graph generated by ICRel, in particular, the nodes associated with APCs. Therefore, the contributions of this manuscript are: (i) the open source ICRel system that generates a cell–cytokine graph from biomedical abstracts and (ii) the knowledge graph obtained using ICRel on a set of documents relevant to tolerogenic antigen-presenting cell therapy. ICRel was able to identify cytokines associated with tolerogenic antigen presenting cells that were missing from the immuneXpresso database. The code and results obtained with ICRel are available at https://github.com/lasigeBioTM/ICRel.

### 2. MATERIALS AND METHODS

The objective of ICRel is to automatically generate a knowledge graph relevant to tolerogenic cell therapy from a given corpus. The system was written in Python 3.5 and its code is openly available.3 The methodology used can be adapted to other domains, by selecting an appropriate set of documents and reference database. **Figure 1** presents the pipeline of ICRel, describing the input and output of each module, whereas **Figure 2** provides an example of an abstract being processed by each module. The first module retrieves abstracts from PubMed into an internal database, according to a given query specified as input. The second module identifies named entities with an external tool, requiring one lexicon for each entity type to be identified. In this case, we had a lexicon for cell names and another for cytokines. The third module combines all cell–cytokine pairs identified within a sentence to generate instances for the machine learning

which is becoming less practical, since the number of published documents increases at a high rate. A more feasible approach is to use automatic text mining methods to process documents and generate a knowledge graph for a given topic. In a knowledge graph, nodes correspond to real world entities while edges represent relationships between the entities. A widely popular knowledge graph is the one integrated with Google search. This graph is generated from web documents, and organizes information about various topics, such as people, places, and works of art, to improve the quality of the search results delivered to the users.2 Recent works have demonstrated how biological knowledge graphs can be extracted from documents, based on protein–protein (9), miRNA–gene (10), and drug–target interactions (11). While these graphs provide important efforts to link the discoveries of various manuscripts, there is still a need for automatic methods that can create specialized graphs and update them as more works are published.

<sup>2</sup>https://www.google.com/intl/es419/insidesearch/features/search/knowledge. html.

<sup>3</sup>https://github.com/lasigeBioTM/ICRel.

<sup>1</sup>http://www.cells-talk.com

FIGURE 1 | Pipeline of the ICRel system. This first module (A) retrieves documents from PubMed, the second module (B) annotates cell and cytokine entities in each document using the Cell Ontology and Cytokine registry, the third module (C) combines the cells and cytokines mentioned in the sentence, and the fourth module (D) classifies each pair and generates the graph.

(A) shows these sentences, numbered and with cells and cytokines bolded manually. The second box (B) shows the entities recognized automatically, where the numbers at the start of each line represent the first and last character offset of the entity. The third box (C) shows the possible cell–cytokine combinations using the sentences shown. The fourth box (D) shows the confidence scores obtained with our system for those pairs. It should be noted that those scores were obtained using several documents and not just the example shown.

classifier and to calculate the pair frequency score. Finally, the fourth module classifies each pair, assigns a confidence score and generates a graph based on the pairs that were classified as positive. The remainder of this section describes in detail the data and methods used to develop this system.

### 2.1. Datasets

A previous study provided a database of interactions between cytokines and cells, named immuneXpresso (13). Although this database was generated using automatic information extraction methods, its contents were evaluated with two manually curated databases, regarding the interactions containing B cells. The authors obtained a 20% false negative rate and no false positives. Even though we have no other guarantee that all entries of this database are correct, we considered this database as a silver standard due to the evaluation scores reported by the authors. A gold standard would require each entry to be manually validated by different domain experts. Since we could not find a gold standard for cytokine–cell interactions in abstracts, we used this silver standard to train and evaluate our method using 5-fold cross-validation. In previous studies, this type of methodology has been shown to be useful for information extraction evaluations (14, 15).

Each entry of the immuneXpresso database represents an interaction between a cytokine and a cell found in the literature. The interactions are supported by one or more abstracts, and they have the following attributes: direction (cell to cytokine or vice-versa), sentiment (Positive, Negative or Unknown), number of articles, and e-score. The sentiment reflects if the interaction indicates upregulation (positive) or downregulation (negative). Each interaction can be found in the associated abstracts, in at least one sentence mentioning both the cytokine and cell. We retrieved these abstracts from PubMed and associated each entry with the respective abstracts. A total of 25,347 abstracts were considered for this silver standard.

Our main objective was to develop an automatic system to generate a knowledge graph about cellular tolerogenic therapies, focusing on those that use APCs. Hence, we retrieved a corpus of documents related to this topic using the MeSH term "Antigen-Presenting Cells," which should include most published abstracts with information relevant to our graph. We restricted this query to abstracts published from January 2015 to August 2017, to avoid overlapping with immuneXpresso, which has no abstracts published after 2015. Using this query, we obtained 3,264 abstracts, which were then annotated with cytokine and cell named entities. **Figure 2A** shows an excerpt of one of these abstracts. We expect that the information obtained by our system can be complementary to this database, which is not focused on any specific topic besides immunology. Furthermore, our system can automatically process new abstracts and add new relations to the graph.

### 2.2. Named Entity Recognition

Each abstract of our datasets contained named entities corresponding to concepts relevant to tolerogenic cell therapies. We were interested specifically in references to cells and cytokines in these abstracts. To this end, we established a lexicon of cell and cytokine names. The cell lexicon is based on the Cell Ontology (16) (version: 2017-07-29). We compiled all the concept labels and corresponding synonyms, resulting in a total of 8,503 terms. For cytokines, we used a cytokine registry,4 which includes several synonyms for each cytokine, corresponding to a total of 7,242 terms (version: November 2015). In both cases, each synonym was mapped to a reference string: Cell Ontology concept label in the case of cells and Entrez name in the case of cytokines. This way, we could associate the same entities mentioned across various documents through different synonyms, as long as those synonyms were considered in our lexicon.

We employed MER (17) to identify named entities in the abstracts. MER matches a list of terms (lexicon) to their mentions in the text, returning the characters of the entities found. For example, in the sentence "The dendritic cells were safely tolerated." MER would return the characters from 4 to 19, which correspond to the text "dendritic cells." **Figure 2B** shows an example of the output of MER for an abstract. This tool has the advantage of being easy to adapt to any entity type, it does not require annotated training data, and it is lightweight in terms of computational resources. We ran MER for each entity type (cell and cytokine) on each abstract. Due to its simplicity, MER has some limitations, for example, it is not able to use context to recognize entities, and it is susceptible to orthographic variations. To increase the number of entities recognized, we added plural variants of every cell name to the lexicon with the Python package inflect. This way, in the previous example, "dendritic cells" would be matched to the "dendritic cell" concept of the Cell Ontology, even if the text is not a perfect match. Furthermore, we removed common words such as "light" and "killer" from the cytokine lexicon, since these words could also appear in other contexts, for example, as part of "natural killer cell." We found these words by comparing the lexicon to a list of common English words. The main limitation of MER is that the lexicon may be incomplete and some references to cells and cytokines in the documents will be missed. However, by using a large corpus, our assumption is that only rare variants will not be identified since most journals recommend a specific nomenclature for cells and proteins.

### 2.3. Cell–Cytokine Relation Extraction

A classifier is a model capable of assigning labels to new data according to a specific function learned from the training data. Supervised machine learning algorithms learn to classify instances (in this case, pairs) by adjusting a function to the labels of each instance of the training set. Generally, these algorithms require the training data to consist of a matrix where each line corresponds to an instance and each column to a feature. We consider an instance to be a specific combination of cell and cytokine, while the features consist of the words used in sentences where that pair cooccurs. A classifier should be evaluated to understand how useful it can be to predict the labels of new data. This type of evaluation is done by comparing the real labels assigned by experts to the labels predicted by the classifier. **Figure 3A** shows the workflow of the training and evaluation process of a supervised machine learning approach using 5-fold cross-validation. Cross-validation consists of iteratively partitioning the dataset in folds, using all but one of the folds to train a classifier. This classifier is used to predict labels for the remaining fold, which are then compared to the original labels. In a 5-fold cross-validation, this process is repeated 5 times, and an average of the scores obtained in each iteration is used to estimate the quality of the classifier. Afterward, a classifier can be trained using the whole dataset.

We consider a knowledge graph to be a set of facts associated with a specific domain using the RDF data model, i.e., specified by predicate–verb–object triplets. In our case, the knowledge graph is constituted by cell–cytokine interactions, where the focus is

<sup>4</sup>http://immport.org/immport-open/public/reference/cytokineRegistry.

on the predicate and objects, which are cells and cytokines, with no specific order. An instance is any cooccurrence of a specific cell–cytokine pair within a sentence. We consider various types of relations, where a cell expresses a cytokine, or a cytokine affects the behavior of a cell. We are interested only in direct relations, where there are no intermediaries to the relation described. This includes cases of up- and downregulation, signaling, activation, and stimulation, for example. However, we are not interested in cases where the relation is negated (e.g., the cell does not express the cytokine) or hypothetical (e.g., the authors consider that a similar cell may express the same cytokine). For each pair, at least one sentence must explicitly state the existence of the relation for it to be considered a positive instance. That sentence may contain other information, such as the mechanism of the relation, experimental details or other cells and cytokines.

Distant supervision assumes that if a relation between two entities is stated in a database, it can be assumed that whenever those two entities cooccur in a document a relation between them is described (**Figure 3B**). We used distant supervision to generate a dataset for training since it is not easy to obtain labeled training data for most domains. For example, it would be assumed that every sentence in the abstract of the article (12) that mentions both dendritic cells and IL-12 is supporting that relation, including this sentence: "These dendritic cells were stimulated for another 48 h, and IL-12 p70 was measured by ELISA." Although this assumption does not take into account the semantics of the text, it has been shown that distant supervision can be useful to extract relations from documents (18). In this work, we adopted immuneXpresso as the reference database. As previously mentioned, this database was generated automatically, however, the authors report a high accuracy when compared to experimental data.

The machine learning algorithm used by ICRel, multi-instance learning (MIL), organizes instances in bags, which consist simply of sets of instances with a common property. All instances are negative if the bag label is negative, or at least one instance is positive if the bag label is positive. Therefore, there is no need to manually label the relations in the documents. This approach can be applied to relation extraction, assuming that the instances are potential relations and the bags contain instances of the same pair of entities. **Figure 2C** shows an example of the way the instances are organized in bags, where each line corresponds to a different bag. Each bag has a label, which can be positive if the database contains an entry establishing a relation between the two elements of the bag, or negative otherwise. Using a machine learning algorithm, a classifier can be trained to classify new instances. This classifier will assign a confidence score to each bag. It is a reasonable assumption that an interaction is stated in a single sentence, so we consider only pairs of entities mentioned within a sentence.

Besides the labels of each bag, the MIL algorithm uses a feature representation of each instance to train a classifier. In our case, the feature representation of each instance is based on a window of words around each entity of the pair. We used a context window of size three, meaning that at most three words before and after each entity were considered. Each word was represented by its lemma so that variations of the same root word did not affect the learning process. Words that were part of named entities were represented by their respective entity type, to avoid any bias toward specific entities, and words that appeared in less than 1% of the documents were not considered, to reduced noise caused by text artifacts.

Then, we generated tf-idf weights for each word, to obtain a vector representation of each instance. Tf-idf corresponds to the product between term frequency (tf) and inverse document frequency (idf), and it is used to estimate the relative importance of each word in a corpus. This is required since machine learning algorithms require numeric vectors. The weights generated during the training phase were also applied to new data. In summary, each document was converted to sets of instances (bags), with each instance corresponding to a feature vector obtained with tf-idf weighting.

We observed that only some sentences in each abstract described relations between cells and cytokines, while the other sentences presented other types of information, such as definitions or experimental parameters. This would be an issue to traditional approaches relation extraction because there is a larger proportion of negative pairs (no direct and explicit relation is described in the text) than positive. In our preliminary experiments, we found that often less than 10% of the pairs in a document are positive. Therefore, it was necessary to use an algorithm that takes into account the sparsity of the data. We tested variations of MIL and found that sparse MIL (sMIL) (19) provided the best results. This algorithm is based on support vector machines, with an adapted objective function to account for the reduced number of positive labels. This new cost function assumes that smaller positive bags are more informative, weighting the feature vector of each positive bag according to its number of instances.

Our system contains a classifier trained using all entries and documents from the immuneXpresso database, corresponding to about 25k abstracts, using the methods described above. ICRel extracts relations from documents by transforming the text into feature vectors and then applying this classifier. The trained classifier predicts the label of a bag but does not predict the individual label of its instances. This means that it is not possible to know the exact sentence where the interaction is described. However, this information is sufficient for our purposes, since we know that each extracted relation has at least one sentence supporting it.

We used two different measures to classify an instance: the confidence score assigned by the machine learning classifier, and the number of sentences associated with a pair, which we call the pair frequency. The classifier confidence score was based on the distance to the hyperplane given by the sMIL algorithm, as described in Ref. (20). The pair frequency was calculated as the number of abstracts where that pair cooccurs in a sentence divided by the total number of abstracts in the corpus. We expect that pairs mentioned in more documents are more likely to have been correctly identified. Both scores were used to study how precision and recall varies when using a threshold. As the threshold increases, recall should decrease while precision increases.

### 2.4. Knowledge Graph for Tolerogenic Cell Therapy

The proposed ICRel system can extract candidate entries to generate a cytokine-cell graph. Each candidate entry is supported by the sentences where it was found, a classifier confidence score and its frequency. **Figure 2D** shows an example of the final output of the ICRel system. Since each cell and cytokine entity was normalized to a reference database, we can associate relations described over many documents, even if the authors use various nomenclatures. Furthermore, since we used the Cell Ontology as the reference for cell names, its axioms can be explored to expand the graph.

To obtain a knowledge graph for tolerogenic cell therapy, we first obtained a set of 3,264 documents about APCs. This set of documents does not overlap with the documents used to train the classifier, which includes only documents published before 2015. The same documents should not be used for training and testing machine learning classifiers because the classifier will have a biased performance on the training documents, leading to an overestimation of the quality of the results. Instead, we can simply match the immuneXpresso relations with our graph to obtain more knowledge.

The extracted relations were imported to Cytoscape (21) to visualize the graph. The ICRel graph is an undirected bipartite graph where each edge corresponds to a cell–cytokine relation. We compared our graph to the one obtained with immune-Xpresso, by considering it also as an undirected graph. We computed standard properties of the two graphs, such as diameter and center nodes, with the Python package NetworkX (22). Furthermore, since our system is focused on obtaining information about tolerogenic cell therapies, we explored the information contained by each graph relevant to this type of therapy.

We considered that a manual evaluation of the automatically generated knowledge graph was necessary to estimate the quality of the information. We sampled a set of 60 edges to be manually validated by three human curators. Each curator validated 30 edges, with a set of 15 edges common to all three, to calculate the interannotator agreement. Each curator accepted an edge if there was at least one sentence supporting it in the corpus, and rejected otherwise. We asked to classify the cause of each rejection to understand the sources of error of our graph. The interannotator agreement was measured using Fleiss' kappa, an adaptation of Cohen's kappa for multiple annotators (23). The classifications of the curators were used to estimate the accuracy of the graph.

### 3. RESULTS

The silver standard described in Section 2.1 is composed of 25,347 abstracts and a total of 4,445 cell–cytokine relations, without considering direction or any other attribute. The silver standard did not contain any information about entities mentioned in the abstracts that did not participate in cell–cytokine relations. We identified 185,243 cells and 189,457 cytokines mentions in these abstracts, which we then used to extract relations using the distant supervision approach. Considering that only 26,357 cell and 25,946 cytokines mentions exist in the immuneXpresso database, we identified about seven times more entities. Notice that these numbers refer to total mentions, i.e., any cell or cytokine may be mentioned more than once across the abstracts. We obtained a precision of 0.366 and recall of 0.853 when comparing with this silver standard. We estimate that the low precision is due to entities that do not participate in interactions, and, as such, are not considered in the silver standard used. For our objective, it is more important to recognize most of the cell and cytokines mentioned in the abstracts because the relation classifier will train and identify new relations based on those entities. Therefore, a recall of 0.853 indicates that most of the cell and cytokine names were identified.

We ran a 5-fold cross-validation on the silver standard documents to evaluate the performance of our system. We randomly divided the documents into 5 partitions and iteratively trained a classifier on the documents and respective relations of 4 partitions and tested on the documents of the other one. Then we compared the relations obtained on each iteration with the silver standard, to calculate precision and recall. Using the classifier confidence score of each prediction, we can use it as a threshold to observe how it affects precision and recall. We compared this approach with only using the pair frequency, which was given by the number of documents where the cell and cytokine appeared within a sentence divided by the total number of documents. For both cases, we tested several threshold values and calculated precision, recall and F-measure assuming that only pairs with scores above the threshold were predicted as positive. **Table 1** compares the confidence score calculated by the classifier with the pair frequency, at the threshold where the highest F-measure was obtained. **Figure 4** shows the precision-recall curve obtained by ranking the pairs by classifier confidence or pair frequency. In this figure, we can see that for the same recall values, the distant supervision approach has higher precision than the frequency approach, hence it can provide higher quality results. At the highest recall values, the precision of the frequency approach is slightly higher, and for maximum recall, the precision is the same in both cases since the only difference is how the pairs are

TABLE 1 | Results obtained with cross-validation on the immuneXpresso silver standard using the classifier confidence score and pair frequency at the threshold where the highest F-measure was obtained.


ranked. However, the classifier confidence score has a larger area under the curve (0.881 vs. 0.850). The area under the PR curve is used as an estimate of the quality of a classifier in cases where the distribution of the labels is skewed (24).

We generated a graph from the immuneXpresso database to compare with the graph generated using ICRel. This graph is composed of cell–cytokine relations found automatically in 25k abstracts from 1988 to 2015, resulting in 432 nodes and 2,495 edges. The authors of this database provided other properties for each relation, such as direction and degree. However, since our system did not provide this type of information, we considered all interactions regardless of their properties.

The ICRel graph contains 212 nodes and 647 edges, extracted from 3,264 abstracts. Each edge is supported by at least one sentence from these abstracts, with an average of 2.87 sentences per edge. Furthermore, each edge has a confidence value given by the classifier. We calculated the Pearson correlation between this confidence value and the number of sentences associated with the two nodes. We obtained a correlation of 0.666, which indicates that while the two variables are positively correlated, this correlation is not very strong. The diameter of this graph is 7, which is one edge larger than the immuneXpresso graph. Overall, the immuneXpresso graph contains more nodes and edges, which is expected since it was derived from a larger number of documents than the ICRel graph. **Figure 5** presents an overview representation of the ICRel graph, while **Table 2** provides a comparison between the two graphs. The files used to generated the graph are provided as supplementary material. Data Sheet 1 is a table where each line is an edge of the graph and the PubMed IDs of the documents are included, whereas Data Sheet 2 contains the sentences which support each of the edges.

Regarding the manual evaluation of the graph, the accuracy obtained was of 0.615. We obtained a kappa score of 0.600, which can be considered an adequate level of agreement (25). In the following section, we summarize the most common sources of error found in this evaluation.

### 4. DISCUSSION

Our work demonstrates how text mining solutions can be used to automatically generate a knowledge graph relevant to tolerogenic cell therapy. A reference database is required to train a classifier based on a specific type of relation. Due to the lack of databases about immunological therapies, we could only train and evaluate our system on immuneXpresso. As such, we were also limited in terms of type of relation to extract, since it had to be a relation described in that database. However, cytokines have been shown to be therapeutic agents in various diseases such as diabetes mellitus and multiple sclerosis. Cytokines also have important roles in the production of APCs (3). It is relevant to understand the relation described in the literature between cells and cytokine since these could suggest novel approaches to tolerogenic cell therapy. Our graph contains these relations and can be integrated with other sources of information through the unique identifiers provided by the Cell Ontology or Entrez databases.

We compared the confidence score given by our classifier with a frequency-based approach, where the ranking score is given by

TABLE 2 | Comparison of ICRel and immuneXpresso graphs in terms of number of nodes, edges, abstracts used, and diameter.


the frequency of a cell–cytokine pair in the corpus. We found that the score given by the classifier is more accurate than the pair frequency. This is also supported by the low correlation between the classifier confidence and number of sentences supporting that pair (0.666). Our system learns how to classify relations using the context words as features. A cell–cytokine pair may be mentioned in multiple documents, but if the context words used are not similar to other positive pairs, it will not be classified as such. This is the main advantage of machine learning methods, along with the possibility of improving the classifier with more validated data.

Most of the processing time necessary to run our system consists of training the classifier. This part of the process takes more time and memory as more documents are considered for training since each document introduces new words and entities. In our case, the training itself took about 1 day. However, once the classifier is trained, a new set of documents can be processed relatively quickly.

## 4.1. Comparison between ICRel and immuneXpresso Graphs

The main point of comparison of our graph is the one created by Shen-Orr et al. (13), which we refer to as the immuneXpresso graph. This graph is larger than ours, containing more nodes and edges. However, it is important to consider that immuneXpresso was created using a more generic set of documents, that were retrieved using the keywords "Immunology and Allergy" and "General Science," from a span of about 50 years. We demonstrated the usefulness of our system by generating a knowledge graph focused on one particular subject and using only abstracts published in the past two years. We expect that the number of relations extracted by our system would increase with a larger set of documents. Our assumption is that a more limited and focused set of documents should result in a graph with more relevant information to the subject of study.

We first compared the information stored in each graph in general terms. As shown in the Section 3, despite the difference in size, both graphs have a similar diameter. The diameter corresponds to the shortest distance between the two most distant nodes of a graph. As an example, **Figure 6** shows a subgraph containing the union of the longest paths of each graph with at least three nodes in common. There are three edges in this subgraph that are shared between the two graphs (T cell < - > IL4, IL4 < - > T-helper 2 cell and T-helper 2 cell < - > IL13). These associations that exist in both graphs show that ICRel can extract well studied cell–cytokines relations, while in Section 4.2 we show examples of extracted relations from recent articles that could not be found in the immuneXpresso graph.

Comparing the relations described by each graph, we can observe various differences. The nodes in the center of the immuneXpresso graph (the center is the set of nodes whose distance to any other node is less or equal to the radius) are all cytokines (TGFB and TNG) while the ICRel graph has two cytokines (IL-6 and CSF2) and two cells (dendritic cell and T-cell) in the center. Dendritic cells are APCs, while T-cells can be targeted by APCs. Both cytokines CSF2 and IL-6 are also relevant to APCs since the former is used to differentiate APCs and the latter is produced by dendritic cells.

To better understand the degree of novelty of ICRel we divided its edges in four categories: (i) edges in common with the immuneXpresso graph; (ii) edges where the nodes existed in the immuneXpresso graph but were not connected; (iii) edges containing only one node that existed in the immuneXpresso graph; and (iv) edges where the two nodes did not exist in the immuneXpresso graph. **Table 3** shows the total of edges for each of these categories.

The two graphs have 132 nodes and 195 edges in common. The top five nodes that were in these edges were T cells (36), macrophages (20), TNF (19), CSF2 (17), and dendritic cells (15). Considering only nodes that were common to both graphs, ICRel found 178 new relations. For example, ICRel identified a relation between mononuclear cells and CSF2, supported by six documents.

The ICRel graph has 76 nodes (23 cells, 53 cytokines) that were not in the other graph. Of the new cytokines identified,

TABLE 3 | Degree of novelty of ICRel vs. immuneXpresso.


most were actually genes coding cytokine receptors. However, we believe that these are as relevant to understand cell–cytokine relations as the cytokines themselves. A cell that produces a cytokine receptor is intrinsically associated with that cytokine. We found that 14 of the 76 new nodes were actually in the immuneXpresso database under different synonyms. For example, we identified the expressions "alpha interferon" and "interferon-alpha," but we were not able to associate with IFNA, which is how it is represented in immuneXpresso. These synonyms should be considered in future analysis to facilitate the integration of different knowledge graphs.

The ICRel graph contains 256 edges with one new node, and 18 where the two nodes were new. The top five nodes of this category were T cells (27), dendritic cells (25), FLT3 (16), CCR7 (16), and monocytes (16). While the immuneXpresso graph contained many edges with T cells and dendritic cells, ICRel identified even more cytokines related to those cells. The FLT3 receptor is associated with the differentiation of dendritic cells, which might explain why our graph contains more edges with this cytokine receptor. CCR7 is a cytokine receptor annotated with the Gene Ontology term "positive regulation of dendritic cell antigen processing and presentation," which was recognized by our system due to an entry in the cytokine registry that we used.

### 4.2. Manual Evaluation

We manually evaluated a partition of the ICRel graph to understand how a classifier trained on the immuneXpresso dataset would perform on a different corpus. This evaluation was performed by three researchers, who we refer to as curators, who read the sentences associated with 60 relations and determined if the cell–cytokine relation was supported by the text. The curators were given the same description of what was considered a relation, similar to the one presented in Section 2.3. We observed that the curators did not agree in some cases, leading to an interannotator agreement of 0.600, based on 15 relations. Since this value represented only a moderate agreement, we analyzed the cases where the curators disagreed. Our system considered both cytokine and cytokine receptors, and it was not clear to the curators which one was relevant. For example, one of the sentences contained the following text: "Flt3 ligand (Flt3L)"; our system recognized both FLT3LG and FLT3 and as cytokines, while FLT3 is actually a cytokine receptor. It is reasonable to assume that a cell associated with FLT3LG is also associated with its receptor, however, since it is not explicitly stated in the sentence, it caused ambiguity among the curators.

The accuracy obtained with the manual evaluation of the graph was of 0.615. The most common errors were indirect relation between the cytokine and cell, i.e., whenever there is a third element that affects both cytokine and cell. For example, consider the pair (CXCL2, memory T cell) in the sentence "(…) perivascular macrophages that are activated by IL-1a produced by keratinocytes and dDCs that are attracted by these macrophages through **CXCL2** signaling, both of which are essential for the efficient activation of **memory T cells** *in situ*." Although both elements of the pair are mentioned in the sentence, there is not a direct relation described, instead, they are both directly associated with keratinocytes and dDCs.

Another common source of error is the incorrect recognition of named entities, both cytokines and cells. For example, in every sentence mentioning "granulocyte macrophage colonystimulating factor," macrophage was recognized as a cell entity. The cytokine registry we used to generate a list of synonyms contained some entries that were too ambiguous to be used by our system, such as acronyms that correspond to normal words. Although we were able to remove most of these synonyms, some cytokine synonyms stayed in the lexicon and generated named entity recognition errors. This is the case of immunoglobulin M (IgM), which was recognized as CD40LG since IGM is a synonym of that cytokine.5 These errors are hard to prevent since it is not possible to have complete knowledge of which synonyms have multiple meanings. One possible solution to this problem consists in computing the semantic similarity of all entities of an abstract and using that value to exclude outliers. Assuming named entity recognition errors would have low similarity to the other entities, this method could improve the precision of our graph (26). In the previous example, we expect that immunoglobulin M and CD40LG would have low similarity to the other entities of that abstract.

To identify if the graph contains information relevant to APCs, we evaluated manually the edges containing the node "professional antigen-presenting cell." In the ICRel graph, this node is connected to 10 nodes: CCL19, CCL21, CCL5, CCR7, CSF2, CXCL12, IFN1, IL12, TGFB1, and TNF. Two of these cytokines (CSF2 and IL12) also appear associated with APCs in immuneXpresso. The ICRel graph contains the more generic IFN1, which includes two cytokines that appear associated with APCs in immuneXpresso (IFNA and IFNG). We confirmed the relations between APCs and its respective cytokines in the articles from where they were extracted (**Table 4**). By carefully analyzing the articles or the sentences provided in the supplementary material Data Sheet 2, it is possible to obtain more details about these relations. For example, Bryce et al. (27) explain the roles of CCL19 and CCL21 in the migration of APCs to lymph nodes. Since our system identifies both cytokines and their receptors, it also identified a relation between CCR7, a chemokine receptor, and APCs. Even though CCR7 is associated with APCs, as explained in this article, it is out of the scope of the knowledge graph, which consists of cell–cytokines relations (28). show that CXCL12 and CCL5 are relevant to the recruitment of APCs in early vitiligo. Although this is not directly related to tolerogenic therapies, understanding the mechanisms of APCs in disease can lead to new methods to generate and modulate the action of these cells. Further improvements could be added to ICRel in order to extract other attributes of each relation, such as directionality, temporality and magnitude. For example, by adapting the methods that we recently developed to classify the type, polarity, degree and modality of clinical events (29).

To understand whether our method was able to find relations that were not yet well studied, we compared the cytokines associated with APCs and dendritic cells on ICRel and immuneXpresso (**Table 4**). ImmuneXpresso was generated using abstracts up to 2015, excluding that year. Only 2 of the 10 cytokines from ICRel were also found in immuneXpresso. Seven cytokines were found to be associated with APCs in articles from recent years. One cytokine receptor (CCR7) was also found to be associated with APCs and dendritic cells by our system. Our system as able to correctly extract this new information and organize it in a knowledge graph. We also studied the edges containing the node "dendritic cell," which is a type of professional APC. The ICRel graph contains 64 edges associated with dendritic cells, of which 49 were not found in immuneXpresso. Dendritic cells and APCs had 7 edges in common in the ICRel graph (IFN1, CCR7, IL12, CSF2, TNF, CCL5, and CCL19). Comparing to the immuneXpresso graph, we can see that most of the cytokines associated with dendritic cells were found to be associated with APCs by ICRel.

TABLE 4 | Cytokines and receptors identified by ICRel as being associated with APCs.


*The second column indicates the reference of the abstract where that relation was found. The third and fourth columns indicate if that cytokine was associated with APCs or dendritic cells in ICRel and immuneXpresso respectively.*

<sup>5</sup>https://www.ncbi.nlm.nih.gov/gene/959.

TABLE 5 | Relations of tolerogenic APC types found by the ICRel system.


Since there is no overlap in the source documents, this means that while these cytokines were first reported to be associated with dendritic cells, other APCs types have also been studied, such as epidermal Langerhans cells (27) and macrophages (33).

We found that immuneXpresso lacked information about specific tolerogenic cell types, given that the version of the Cell Ontology used did not contain them. Thus, we added a list of 13 tolerogenic APC types to the lexicon so that relations containing these cells could also be detected. This led to the identification of 8 relations containing tolerogenic APCs (**Table 5**). The majority of these relations included myeloid-derived suppressor cells (MDSC). The system identified relations between MDSC and TNF, TNFRSF1A, and TNFRSF1B. While TNFRSF1A and TNFRSF1B are actually cytokine receptors, the article that mentions them (source article) describes the effects of gene deletion of both the cytokine and the receptors in carcinogenesis (35). The relation between MDSC and IL10 was extracted from a review article about the role of these cells in inflammatory diseases (36). Another relation extracted was between tolerogenic dendritic cells and TGFB1. In this case, the source article establishes the importance of TGFB1 in immunotherapies using tolerogenic dendritic cells (37).

### 4.3. Conclusion and Future Directions

Due to its initial stage, there is a lack of openly available databases about tolerogenic cell therapy. Although commercial databases such as COPE and Cytokine Reference exist, these depend on manual curation. It is time-consuming to manually develop and then update databases with newly found information from published articles. Our ICRel system presents a solution to this issue, by using machine learning to automatically generate a knowledge graph of cell–cytokine relations. Using the knowledge graph, experts can then find more facts to store in their own databases, or help them formulate new hypotheses that need further study. Our system obtained higher precision values when compared to a frequency based approach.

We demonstrated the usefulness of the system by focusing on antigen presenting cells relevant to tolerogenic cell therapy. There have been various advancements in our understanding of immune mechanisms and pathways that are dysregulated in autoimmune diseases, and active in transplant rejection, contributing to advancements in tolerogenic therapies. A better organization of the current knowledge about this process would benefit the development of new treatments and clinical trials. The knowledge graph contained relations between APCs that were found only in recent articles, thus showing how our system can lead to a more complete information structure on this topic. Furthermore, we identified multiple associations between specific tolerogenic APCs and cytokines. We believe that our proposed system has a large potential to help practicing cell biologists or cell therapy experts in identifying relevant relationships that can only be found by exploring various scientific articles in an integrated way. It was not our aim to find novel or specialized information but rather show the feasibility of the system and to use examples for guiding practitioners and experts on how to take advantage of it.

The work presented in this manuscript has two major applications. The first is information retrieval systems that can use the information from our graph to integrate various sources of information. This is the case of Bio2RDF (8), which stores several biomedical databases, such as KEGG, PubMed, and HGNC, in RDF format. The Bio2RDF project is an effort to link the entries of these databases using normalized URIs. Since our system matches each cytokine to the Entrez database and each cell to the Cell Ontology, it should be simple to integrate our graph with other databases for information retrieval. Another major application is recommendation systems. It is useful for a researcher working with a specific group of cell lines to know which other cells could also fit in that group. There are various methods to provide this type of recommendation, one of them consisting in exploring the structure of the graph to compute similarity measures. A recommender system could then suggest cells that interact with the same cytokines as the cells in the group. By integrating with external sources, it would be possible to suggest cytokines associated with specific diseases, chemicals or genes.

### AUTHOR CONTRIBUTIONS

Conceptualization and methodology: AL and FC. Funding acquisition, project administration, and supervision: LC and FC. Investigation, validation, writing, review, and editing: AL, JF, LC, and FC. Software: AL. Visualization and writing original draft: AL, LC, and FC.

### ACKNOWLEDGMENTS

We thank the reviewers for their valuable comments and suggestions and in particular the feedback from reviewer 2 that helped us make the manuscript more appealing to immunologists.

### FUNDING

This work was supported by the Portuguese Fundação para a Ciência e Tecnologia (http://www.fct.pt/) through the PhD Grant ref. PD/BD/106083/2015 to AL, UID/MULTI/04046/2013 (BioISI) to LC, and UID/CEC/00408/2013 (LaSIGE) to AL, JF, and FC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01656/ full#supplementary-material.

### REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer PL and handling Editor declared their shared affiliation.

*Copyright © 2017 Lamurias, Ferreira, Clarke and Couto. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Translating Mechanism of Regulatory Action of Tolerogenic Dendritic Cells to Monitoring endpoints in Clinical Trials

*Jessica S. Suwandi1 , Tatjana Nikolic1 and Bart O. Roep1,2\**

*1Department of Immunohematology and Blood Transfusion, Leiden University Medical Center, Leiden, Netherlands, 2Department of Diabetes Immunology, Diabetes & Metabolism Research Institute, Beckman Research Institute, City of Hope, Duarte, CA, United States*

Tolerogenic dendritic cells (tolDCs) have reached patients with autoimmune and inflammatory disease, at least in clinical trials. The safety of tolDCs as intervention therapy has been established, but the capacity to modulate autoimmune response *in vivo* remains to be demonstrated. Studies have revealed a diversity of regulatory mechanisms that tolDCs may employ *in vivo*. These mechanisms differ between various types of modulated tolDC. The most often foreseen action of tolDCs is through regulatory polarization of naïve T cells or activation of existing regulatory T cells, which should ultimately diminish autoimmune inflammation. Yet, selection of a target autoantigen remains critical to expedite tissue specific tolerance induction, while measuring immune modulation incited by tolDCs *in vivo* provides a great challenge. We will discuss the regulatory action of different types of tolDCs and the possible methods to monitor immunological efficacy endpoints for the next generation clinical trials.

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Willem Van Eden, Utrecht University, Netherlands Fang-Ping Huang, University of Hong Kong, Hong Kong*

> *\*Correspondence: Bart O. Roep B.O.Roep@lumc.nl*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 26 September 2017 Accepted: 06 November 2017 Published: 22 November 2017*

#### *Citation:*

*Suwandi JS, Nikolic T and Roep BO (2017) Translating Mechanism of Regulatory Action of Tolerogenic Dendritic Cells to Monitoring Endpoints in Clinical Trials. Front. Immunol. 8:1598. doi: 10.3389/fimmu.2017.01598*

Keywords: tolerogenic dendritic cells, monitoring endpoints, clinical trials, autoimmune diseases, regulatory action, antigen specific, regulatory T cells, immune metabolism

### INTRODUCTION

The regulatory properties of dendritic cells (DCs) have been subject of study throughout the last decade (1–5). The ability of DCs to orchestrate the immune system makes them interesting candidates for therapeutic application. In autoimmune diseases where the physiological state of self-tolerance is lost, tolerogenic dendritic cells (tolDCs) could aid in restoring the immunological balance. Several modulating actors have proved to induce DCs with stable regulatory capacity and researchers have since developed clinical grade tolDCs suitable for clinical trials (6–8). Phase I clinical trials with tolDCs are ongoing or have been completed in patients with type 1 diabetes (T1D), rheumatoid arthritis (RA), Crohn's disease and multiple sclerosis proving tolDC vaccination safe and well tolerated, encouraging next generation trials to verify the therapeutic efficacy (9–13). While disease amelioration is the goal in the long run, immunological changes may be detectable more promptly and understanding the regulatory mode of action of tolDCs is essential to define immunological efficacy endpoints.

Variation in the methods used for culture makes comparison of tolDCs difficult, may therefore lead to diversity and inconsistency when comparing results from clinical trials evaluating different tolDCs in different diseases or conditions. In addition, tolDCs with desired therapeutic efficacy have not been identified yet. Current actions such as that of Action to Focus and Accelerate Cell-based Tolerance-inducing Therapies (http://www.afactt.eu) have generated minimum information models to report and interpret data on the quality and preclinical efficacy of tolDCs (14, 15). This may enable the comparison of treatment effects of tolDCs generated with other methods. In this review, we consider regulatory actions of tolDCs and discuss the methods to monitor these *in vivo* as immunological efficacy endpoints for future clinical trials, whether they are described as a common feature or shown only for a certain type of tolDC. Using similar immunomonitoring strategies in different trials could also help answering the question whether the variation in the culture methods translates into variable functional properties.

### PHENOTYPICAL CHARACTERISTICS AND CYTOKINE PROFILE OF tolDCs— MEDIATORS FOR TOLEROGENIC FUNCTION

Several approaches have been tested to induce maturation resistant tolDCs *in vitro* (2, 7, 16, 17). Common features of tolDCs presumed to mediate tolerogenic functions include low antigen presentation capacity, reduced co-stimulatory signals, expression of inhibitory molecules and an anti-inflammatory cytokine profile. Co-stimulatory signals such as CD80, CD86, and CD40 in addition to antigenic stimulation are key to adequate T cell activation and absence thereof leads to unresponsiveness, i.e., anergy and activation of regulatory T cells (Tregs) (18, 19). The balance between pro- and anti-inflammatory cytokines IL-12 and IL-10 is important for tolerance. IL-12 is central in the induction of T helper 1 cells (Th1) and high IFN-γ production. By contrast, IL-10 reduces the antigen-presenting function of DCs, inhibiting Th1 responses (20). Furthermore, presence of IL-10 is a requisite for the induction of a subset of Tregs (type 1 Treg), while Forkhead box P3 (Foxp3) demethylation is dispensable, rather than a condition *sine qua non* (21–23).

An overview of phenotype and functions of clinically applied tolDCs is provided in **Table 1**, showing variations of the abovementioned common traits as well as unique features that may initiate regulation through distinctive mechanisms. Most tolDCs show reduced expression of co-stimulatory molecules and HLA-DR, while expressing inhibitory molecule PD-L1 (8, 10, 24, 25). tolDCs treated with antisense oligonucleotides against co-stimulatory molecules CD40, CD80, and CD86 (antisense tolDC), and NF-kB inhibitor (NF-kB tolDC) demonstrate low TNF and IL-10 production (10, 26). By contrast, tolDCs induced with combined dexamethasone and vitamin A or vitamin D3 show high production of IL-10 (8, 24, 25, 27). Gene and protein expression data revealed CD52 as candidate marker specifically for VitD3-Dex-modulated tolDC (28) and MERTK was identified in Dex-VitA tolDCs as a specific molecule involved in the negative regulation of T cell activation (29), yet these markers remain to be validated in other tolDCs. Although efforts have been made to find molecules underlying tolDC function, common regulators of tolerogenicity have not been found (28, 30).

While the knowledge about ligands and soluble mediators help us understand how tolDCs shape immune response and may be utilized as clinical release criteria for *in vitro* generated tolDCs, none of them have proved to be unique to serve as a biomarker of tolDCs *in vivo*, while their efficacy to achieve therapeutic efficacy remains to be confirmed.

### HYPORESPONSIVENESS OF EFFECTOR CD4 AND CD8 T CELLS

A common trait of tolDCs is the suppression of effector T cells (**Table 1**) (2). tolDCs inhibit T cell proliferation either directly by inducing anergy or apoptosis, or through the induction of Tregs. Death receptor ligands such as PD-L1 function as direct negative regulator of T cell response. tolDCs treated with VitD3 delete T cells antigen specifically with co-ligation of PD-1 (32). Another mechanism through which Dex-VitA tolDCs inhibit T cell proliferation is through MERTK. MERTK is a family of TAM tyrosine kinase receptors and directly inhibits T cell activation through competition of PROS1 on the surface of T cells, which drives autocrine proliferation (29). Furthermore, VitD3- Dex tolDCs inhibit naïve CD8 T cell proliferation and induce anergy in memory CD8 T cells. However, this effect is countered by cytotoxic killing of tolDCs presenting CD8 epitopes (33). Whether other tolDCs similarly affect CD8 T cells, needs to be verified.

Altogether, tolDCs are capable of inhibiting T cell proliferation through different mechanisms. This common feature is ideal to utilize as efficacy endpoint in clinical trials. *In vivo* alterations of CD4+ T cell responses can be determined with a lymphocyte stimulation test (LST) and enzyme-linked immunosorbent spot assay (34), which quantifies antigen-specific T cell proliferation and cytokine secretion in human peripheral blood mononuclear cells. The LST was proven valid in predicting graft survival in pancreatic islet transplantation, since increase of proliferation was associated with a rapid failure of islet grafts (35). Effects on T cell populations could be further assessed through quantification of effector CD4 T helper subsets (Th1, Th2, and Th17) and CD8 T cells by flow cytometry. Moreover, using quantum dot nanotechnology (Qdot) it is possible to detect and quantify autoreactive CD8 T cells (36). *In vivo* signs of T cell modulation were already observed in the NF-kB tolDC trial by a reduction of CD4+ CD25+ CD127+ effector T cells (10).

*In vivo*, tolDCs could alter different T cell subsets with the potential to influence overall disease outcome as affected subsets may have specific pathophysiological relevance for a particular autoimmune disease. The inflammatory reaction in RA and Crohn's disease is mediated by T helper 1 and 17 (Th1 and Th17) cells secreting pro-inflammatory cytokines IFN-γ, IL-17, and IL-22 (37, 38). By contrast, autoreactive CD4 T helper cells contribute to T1D pathogenesis but cytotoxic CD8 T cells are the main offenders, destroying the insulin producing beta cells (39–41). Therefore, harmonizing assays and following changes in multiple T cell subsets in response to tolDC treatment could enable comparison and correlation to clinical outcomes in different trials.


#### Table 1 | Characteristics of clinically applied tolDCs.

*Evidence from preclinical studies (light blue) and clinical studies (dark blue).*

↓*, Low expression/secretion;* ↑*, high expression/secretion; T1D, type 1 diabetes; RA, rheumatoid arthritis; IA, inflammatory arthritis; Crohn's, Crohn's disease; tolDC, tolerogenic dendritic cell; Tregs, regulatory T cells; CTLA-4, cytotoxic T lymphocyte-associated protein 4; Foxp3, Forkhead box P3.*

## INDUCTION OF Tregs

Perhaps the most important and diverse mechanism of tolDCs is the induction of Tregs, which has been demonstrated *in vitro* and *in vivo* (17). These induced Tregs are suspected to suppress pathogenic autoimmune processes by effector T and B cells involved in a multitude of autoimmune diseases. So far, several Treg populations have been described, and tolDCs can induce or activate various Tregs depending on the DC modulating agent. Naturally produced thymic Tregs (nTregs) are defined using the high and stable expression of transcription factor Foxp3 and represent the best described Treg subset next to CD4+ Foxp3− type 1 Tregs (Tr-1) producing high IL-10. *In vitro*, NF-kB tolDCs and Dex-DCs promote CD4+ CD25+ Foxp3+ Tregs, while tolDCs modulated by Dex plus VitD3 also induced Tr-1 like Tregs (23, 24, 31). This is in line with the thought that Tr-1 Treg induction is dependent on IL-10 production by tolDC (**Table 1**) (21, 22). Membrane bound TNF and PD-L1 are other factors involved in the induction of antigen-specific Tregs and may contribute to the capacity of VitD3-Dex tolDCs to induce heterogeneous Treg subsets which suppress through distinct mechanisms such as killing of monocytes and inhibition of naïve or effector T cell proliferation (23, 42, 43). Indeed, tolDCs generated by VitD3- Dex induce at least three different types of Tregs (23). Whether these features are shared by Tregs induced by different types of tolDCs, and whether the variety of Tregs induced by tolDCs extends to other types remains to be investigated. Induction of Tr-1 like Tregs seems preferred over nTreg induction, since the antigen specificity of the latter is undefined. nTregs may therefore suppress any effector T cell response, including those against cancer, whereas Tr-1 cells with defined specificity will exclusively exert their action when their cognate target of choice (e.g., islet autoantigen) is recognized.

Similar to determining effector T cell responses, the quantification and qualification of Tregs in patients in relation to tolDC treatment is an essential tool in all trials. Indeed, an upregulation of CD4+ CD25+ Foxp3+ Tregs was observed in humans injected with Dex-VitA tolDCs (12). Most studies were limited to measuring circulating Foxp3+ Tregs which possibly underestimates the therapeutic effect, since Tr-1 Tregs induced by IL-10 producing tolDCs need not express Foxp3 (21–23). Another problem with simply looking at Foxp3 expression is that this transcription regulator is also transiently expressed in activated T cells, thus CD4+ CD25+ Foxp3+ cells represent a mixture of both Tregs and activated effector T cells (44, 45). Lastly, an observed increase of CD4+ CD25+ Foxp3+ cells could be indicative of an expansion of dedicated nTregs or newly induced Tregs from naïve T cells in the periphery (45, 46). Additional markers such as ICOS, PD-1 and cytotoxic T lymphocyte-associated protein 4 (CTLA-4) could help us describe suppressive cells in response to tolDC action. Yet, due to a lack of a common marker for all Tregs, measuring the suppressive capacity of T cells in a suppression assay remains the only valid method to determine whether Tregs are present (47, 48).

### INFECTIOUS TOLERANCE AND LINKED SUPPRESSION

It becomes increasingly clear that the interaction between Tregs and DCs is bidirectional since DCs induce Tregs, which in turn impact DC development reducing co-stimulatory ligands and stimulation of suppressive molecules (49, 50). CTLA-4 expression on Tregs modulates DCs by scavenging the co-stimulatory ligands on DCs through the process of trans-endocytosis (51). Tregs induced by VitD3-Dex tolDCs stimulate the expression of inhibitory B7-H3 and ICOS ligand (i.e., B7-H2) on inflammatory DCs upon cognate interaction, which thereafter induced IL-10 producing T cells with other antigen specificities causing infectious tolerance (50). Another molecule described on modified DCs is B7-H4, which is up-regulated under influence of IL-10 secretion by CD4+ CD25+ Tregs (52). Hence, tolDC can exert infectious tolerance through the capacity of Tregs to induce linked suppression and potentially modulate other DCs *in vivo* to acquire tolerogenic phenotype and function. In this way, induced Tregs can augment the suppressive capacity of tolDCs by transferring regulatory properties to other inflammatory DCs. So far, this complementary action is proven in VitD3-Dex tolDCs but is yet to be validated in other, such as antisense and NF-kB tolDCs. *In vivo*, analysis of DCs acquiring the expression of inhibitory molecules from the B7 family (B7-H2, B7-H3, and B7-H4) or a spreading of tolerance to antigens other than that carried by tolDC-vaccine may be an additional lead to monitoring of tolerance induction in the trials and create legacy of tissue specific immune regulation beyond the lifetime of the injected tolDCs.

### IMPORTANCE OF ANTIGEN-SPECIFIC TOLERANCE INDUCTION

The ultimate goal of tolDC therapy is the induction of targeted tolerance, thereby impeding autoimmune inflammation in the affected lesion. Addition of one or more target antigen(s) will guide tolDCs to address effector cells which is desirable to induce disease-relevant immunomodulation. For this purpose, established disease-associated autoantigens are necessary. This may be a straight-forward approach in the case of T1D and multiple sclerosis where tissue specific antigens are identified as suitable targets (53–57). Since tolDCs induce Tregs that act through linked suppression, regulation will not be limited to the antigen to which the Tregs were generated, but spread to all other specificities presented by residency DCs in the lesion or draining lymph nodes. However, in some autoimmune diseases, specific autoimmunity-inducing antigens are unknown or associated antigens are not tissue specific. Citrullinated antigens and deadcell-related epitopes associated with RA and systemic lupus erythematodes, respectively are present throughout the body, which obscures the desire to induce specific tolerance (3, 58). Antigens involved in Crohn's disease also remain unidentified despite great efforts (12, 37). In the latter case, application of tolDCs will rely on the migratory capacity of tolDCs to the pathogenic lesion and local uptake of proteins and presentation in tolerogenic context. In whichever way it may be achieved, the antigen specificity of tolerance induction is essential to avoid general immune suppression and should be closely monitored, for example by measuring the proliferative response against pathogens included in the childhood immunization program.

### SUPPRESSION OF B CELLS

A rarely studied effect of tolDCs is the regulation of B cells, as suggested by preliminary clinical data, but the clinical relevance of such B cell modulation *in vivo* needs to be confirmed. Patients with RA treated with citrullinated peptide loaded NF-kB tolDCs showed reduced anti-CCP IgA/IgG levels, which correlated with clinical improvement (10). Similarly, a significant reduction of antigen-specific autoantibodies was observed in another clinical trial with tolDCs in RA patients (13). The mechanism through which tolDCs regulate B cells is still undiscovered. DCs play an important role in the function of B cells through transferring antigens to naïve B cells and initiation of antigen-specific antibody responses. In addition, DCs provide B cells with isotype-switch signals and promote B cell proliferation and survival through CD40 (59, 60). It is plausible that tolDCs lack the capacity to stimulate B cells resulting in reduced activity of plasma cells or regulate B cell activity indirectly by inducing Tregs. Inhibition of B cell function may not be equally important in all autoimmune diseases as the role of B cells in the pathophysiology of T1D is largely elusive, and islet specific antibody titration does not correlate with disease progression (55, 61, 62). Yet, regardless of whether B cells are pathogenic, Bregs may still prove valuable in disease modulation (62, 63).

More recent data show evidence of tolDC involvement of Breg induction (11). B cells with suppressive activity (Bregs) have been described in the past, but their biology is just beginning to unravel. The so-called Bregs regulate through promotion of Treg development and suppression of effector CD4 and CD8 T cells (64). The phenotype of Bregs could be characterized by the expression of various surface markers (CD19, CD21, and CD23) and the expression of IL-10. Recipients of antisense tolDC vaccination showed an increase of IL-10 producing Bregs in peripheral blood, these Bregs inhibited allogeneic T cell proliferation *in vitro* independent of Tregs (11, 65). Dex-VitD3 tolDCs increased a population of CD19+ IL-10+ Bregs *in vitro* (24). The underlying mechanism of Breg induction is largely unknown, and IL-10 may be involved. More specifically in the case of antisense tolDC, the antisense oligonucleotide mixture may stimulate expression of CD40L and IL-7 on tolDCs and drive Breg induction (24, 65).

Dependent on the pathophysiology of the disease in question, quantification of B cell populations or measuring of disease-specific antibody titers could be relevant. The potential role of tolDCs in Breg induction should be further explored in other types of tolDCs and may proof relevant as additional player with regulatory property.

### POTENTIAL METABOLIC EFFECTS OF tolDCs

Gene expression data and proteomics have revealed considerable changes in metabolic pathways in tolDCs induced by VitD3 or VitD3-Dex (16, 25, 66, 67), which might affect the microenvironment where tolDCs exert their tolerogenic function. Interestingly, tolDCs induced by other agents such as dexamethasone alone or rapamycin did not show similar metabolic changes (68). The increase of metabolic rate through upregulation of oxidative phosphorylation while maintaining or enhancing glycolysis (28, 68), may be a phenomenon similar to the so-called Warburg effect (69). This will result in enhanced glucose uptake and fermentation to lactic acid and may be a target for *in vivo* monitoring upon tolDC treatment.

The ability of tolDC to switch from aerobic respiration to anaerobic glycolysis may have several functional implications. It is presumed to enhance tolDC longevity and resistance to metabolic stress in inflammatory milieu, where low oxygen and glucose levels prevail. Experiments *in vitro* showed that while

glucose was essential in the metabolic programming of VitD3 tolDC, the regulatory phenotype remained stable in hypoxic and hypoglycemic conditions after induction (68). The increase in oxidative phosphorylation activity may surge reactive oxygen species (ROS) as byproduct and cause damage to cells with no effective antioxidative machinery in the close proximity (25). The enhanced glucose throughput may cause nutrient deprivation, which can activate intracellular metabolic sensors such as mTOR controlling the homeostatic proliferation of Tregs (70). Thus, metabolic changes inside tolDCs may result in an immune suppressive effect through nutrient deprivation supporting Treg proliferation and the secretion of ROS damaging immune cells in the proximity. It is difficult to envisage how such effects may be monitored *in vivo* as the effect may be local and easily compensated to non-detectable changes in the circulation. Still it may be interesting to explore this uncharted field of immunometabolism as potential functional activity of tolDCs *in vivo*.

### CONCLUDING REMARKS

Application of tolDC therapy in the clinical setting is an exciting progression toward specific tolerance induction in patients with autoimmune diseases. We now face the challenge to establish the efficacy of tolDC therapy. Results from phase I clinical trials using tolDCs show preliminary effects regarding immune regulation *in vivo*. In this review, we evaluated the regulatory mechanisms of different types of tolDCs to find potential immunological efficacy endpoints, which are summarized in **Figure 1**.

### REFERENCES


Lymphoproliferative assays measuring the response to diseaseassociated antigens provide an elegant method to grasp a view of antigen-specific T cell modification, whereas examining affected immune subsets such as Tregs may prove a holy grail that requires appropriate assay improvements. Features that particular tolDCs exert, such as the induction of Bregs may be further explored in other tolDCs, to assess whether these are unique to certain types of tolDCs or common assets. A better understanding of the phenotypical properties of the different tolDC and affected immune cells will provide essential information for choosing the preferred type of tolDC and designing appropriate monitoring endpoint. Therefore, harmonizing assays and following changes in multiple T cell subsets in response to tolDC therapy could enable comparison and correlation to clinical outcomes in different trials.

### AUTHOR CONTRIBUTIONS

All authors contributed equally to the design and writing of this review.

### FUNDING

The authors' studies were supported by the Dutch Diabetes Research Foundation, Stichting DON, the European Commission (INNODIA-115797, EE-ASI-305305, NAIMIT-241447), the Dutch Arthritis Foundation (LLP-16), and the Wanek Family Project for Type 1 Diabetes.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling editor declared a past coauthorship with one of the authors TN.

*Copyright © 2017 Suwandi, Nikolic and Roep. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Update on Dendritic Cell-induced immunological and Clinical Tolerance

*Carolina Obregon1‡, Rajesh Kumar1†‡, Manuel Antonio Pascual1,2, Giuseppe Vassalli 3,4 and Déla Golshayan1,2\**

*1Department of Medicine, Transplantation Centre and Transplantation Immunopathology Laboratory, Service of Immunology, Centre Hospitalier Universitaire Vaudois, University of Lausanne, Lausanne, Switzerland, 2Department of Surgery, Transplantation Centre, Centre Hospitalier Universitaire Vaudois, University of Lausanne, Lausanne, Switzerland, 3Département coeur-vaisseaux, Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland, 4 Fondazione Cardiocentro Ticino, Swiss Institute of Regenerative Medicine (SIRM), Lugano, Switzerland*

#### *Edited by:*

*John Isaacs, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Raymond John Steptoe, The University of Queensland, Australia Hans Acha-Orbea, University of Lausanne, Switzerland*

#### *\*Correspondence:*

*Déla Golshayan dela.golshayan@chuv.ch*

#### *†Present address:*

*Rajesh Kumar, Surgical Oncology Research Laboratory, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States*

*‡ These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 12 July 2017 Accepted: 26 October 2017 Published: 20 November 2017*

#### *Citation:*

*Obregon C, Kumar R, Pascual MA, Vassalli G and Golshayan D (2017) Update on Dendritic Cell-Induced Immunological and Clinical Tolerance. Front. Immunol. 8:1514. doi: 10.3389/fimmu.2017.01514*

Dendritic cells (DCs) as highly efficient antigen-presenting cells are at the interface of innate and adaptive immunity. As such, they are key mediators of immunity and antigenspecific immune tolerance. Due to their functional specialization, research efforts have focused on the characterization of DCs subsets involved in the initiation of immunogenic responses and in the maintenance of tissue homeostasis. Tolerogenic DCs (tolDCs) based therapies have been designed as promising strategies to prevent and control autoimmune diseases as well as allograft rejection after solid organ transplantation (SOT). Despite successful experimental studies and ongoing phase I/II clinical trials using autologous tolDCs in patients with type 1 diabetes, rheumatoid arthritis, multiple sclerosis, and in SOT recipients, additional basic research will be required to determine the optimal DC subset(s) and conditioning regimens for tolDCs-based treatments *in vivo*. In this review, we discuss the characteristics of human DCs and recent advances in their classification, as well as the role of DCs in immune regulation and their susceptibility to *in vitro* or *in vivo* manipulation for the development of tolerogenic therapies, with a focus on the potential of tolDCs for the treatment of autoimmune diseases and the prevention of allograft rejection after SOT.

Keywords: tolerogenic dendritic cells, autoimmune diseases, immunotherapy, solid organ transplantation, tolerance

## INTRODUCTION

Dendritic cells (DCs) are at the interface of innate and adaptive immunity and, thus, are key mediators of immunity and tolerance. Importantly, DCs constitute a heterogeneous population that comprises multiple subsets exhibiting distinct functional specializations that vary according to their origin, maturation state, location, and environmental conditions (1, 2). In their immature state, DCs mainly traffic and reside in peripheral tissues where they can capture antigens and process them into major histocompatibility complex (MHC):peptide complexes. DCs undergo maturation not only after microbial infection but also in response to damage-associated molecular patterns (DAMPs) and pro-inflammatory cytokines produced as a result of tissue injury. This is the cornerstone for the initiation of effective adaptive immune responses (3, 4). Extensive experimental data over the years have highlighted the possibility of generating "tolerogenic DCs" (tolDCs) that are maturation-resistant *in vitro,* express low levels of T-cell costimulatory molecules, and a have a reduced capacity to produce pro-inflammatory cytokines. These tolDCs mediate antigen-specific T-cell hyporesponsiveness and promote the expansion and/or induction of regulatory T cells (Treg). Thus, the potential for tolDCs to dampen immune responses may be used clinically, e.g., in autoimmune diseases and after solid organ transplantation (SOT). In this review, we briefly describe human DC subsets and the immune regulatory mechanisms mediated by these cells. We then discuss how DCs may be manipulated in the perspective of tolerogenic immune therapies.

### HUMAN DC SUBSETS AND FUNCTIONAL SPECIALIZATION

Dendritic cells represent a heterogeneous cell population arising from bone marrow-restricted precursors identified in mice and humans (5). In humans, common DC progenitors give rise to plasmacytoid DCs (pDCs) and intermediate precursors of conventional DCs (pre-cDCs) that are pre-committed to become either CD1c<sup>+</sup> (BDCA-1) or CD141<sup>+</sup> (BDCA-3) conventional DCs (cDCs) (6). The HLA-DR<sup>+</sup>CD14<sup>−</sup>CD11b<sup>−</sup> fraction of human peripheral blood mononuclear cells (PBMCs) comprises the CD1c+ DC subset (characterized by CD172α and IRF4 expression), the CD141high DC subset (characterized by Clec9A, XCR1, IRF8, and TLR3 expression), and the pDC subset (identified by BDCA-2, BDCA-4, and CD123bright expression) (**Table 1**).

### Peripheral Blood DCs

Peripheral blood DCs are likely the precursors of DCs found in peripheral tissues and lymphoid organs. In peripheral tissues, there is evidence for high phenotypic heterogeneity of DCs, contrasting with the well-defined phenotypic expression of blood DCs. In addition to the BDCA population and Langerhans cells, other subpopulations of tissue DCs can be distinguished by the expression of langerin, CD1a, and CD14 (24, 25); however, these markers are promiscuously expressed making it difficult to unambiguously discriminate peripheral tissue DC subpopulations (24). For example, studies in patients with allergic asthma show that most of the lung CD1c<sup>+</sup> (BDCA-1<sup>+</sup>) DCs also express CD141 (BDCA-3) (26). Recently, using gene expression profiling and mass cytometry analysis, a set of lineage-imprinted cell-surface markers, such as CD172α/IRF4 and XCR1/IRF8, were identified, allowing for a better discrimination between CD1c<sup>+</sup>CD14<sup>−</sup> DC and CD141<sup>+</sup>CD14<sup>−</sup> DC subsets in human tissues (7).

### Lymphoid Organs DCs

Multiple subsets of DCs have been found in lymphoid organs; however the distinction between migratory and lymphoid organsresident DCs still requires further investigation. CD1c<sup>+</sup>CD14<sup>−</sup> DC, CD141<sup>+</sup>CD14<sup>−</sup> DC, and pDC subsets have been found in the human spleen, tonsils, and axillary and pulmonary lymph nodes (LNs) (8–10, 27). These subsets likely correspond to the resident DC population. In axillary LNs, pDCs have been reported to

Table 1 | Characteristics of blood human dendritic cells (DCs) and monocyte-derived DCs (ModDCs) subsets. CD1c CD141 Plasmacytoid DCs (pDCs) ModDCs Reference Surface expression, intracellular markers or transcriptional markers Lin−, MHC II<sup>+</sup> CD1c+ (BDCA-1+) CD11chi, CD11b<sup>−</sup> CD4+, CD2<sup>+</sup> CD45RO<sup>+</sup> CD172α+ IRF4<sup>+</sup> Lin−, MHC II<sup>+</sup> CD141hi (BDCA-3hi/ Thrombomodulin) CD11c+,CD11b<sup>−</sup> CD4+, CD2<sup>−</sup> TLR3<sup>+</sup> Clec9A+ (DNGR-1+) XCR1<sup>+</sup> IRF8<sup>+</sup> Lin−, MHC II<sup>+</sup> CD303+ (BDCA-2+) CD304+ (BDCA-4+/ Neuropilin-1) CD11c−, CD11b<sup>−</sup> CD123<sup>+</sup> CD45RA CD4<sup>+</sup> ILT7<sup>+</sup> TLR7+, TLR9<sup>+</sup> CD1c+ (BDCA-1+) CD14<sup>+</sup> CD11c+, CD11b<sup>+</sup> DC-SIGN<sup>+</sup> Additional markers expressed in tissue ModDCs: CD1a<sup>+</sup> CD206<sup>+</sup> FcεRI<sup>+</sup> FLT3<sup>+</sup> IRF4<sup>+</sup> Zbtb46<sup>+</sup> (6–17) Frequency in peripheral blood (% peripheral blood mononuclear cell) 0.2 ± 0.1% 0.02 ± 0.01% 0.2 ± 0.1% ≈0.29 ± 0.2% (18, 19) Functional specialization Excel in CD4+ T-cell priming. Th1 and Th17 polarization. Cross-presentation of soluble antigens to CD8+ T cells. Secretion of Type-I IFN (poly I:C) Type-I IFN secretion in response to viral infections. Liver and respiratory tract pDCs promote tolerance. Naïve and memory CD4+ T-cell stimulation. Th17 polarization. DC-10, skin CD141+CD14+ DCs and CD1c+CD14+ DCs in melanoma patients promote tolerance. (6, 9, 20, 21) Mouse equivalent CD4+, CD11b<sup>+</sup> (lymphoids) CD11b+ (tissues) CD172α+ IRF4<sup>+</sup> CD8α DCs (lymphoids) CD103 DCs (tissues) XCR1<sup>+</sup> IRF8<sup>+</sup> B220 Siglec H BST2 (PDCA1) Ly6C<sup>+</sup> CD172α+ DC-SIGN<sup>+</sup> CD206<sup>+</sup> FcεRI<sup>+</sup> (7, 22, 23)

*BDCA, blood dendritic cells antigen; DNGR-1, dendritic cell natural killer lectin group receptor-1; IFN, interferon; ILT, immunoglobulin-like transcript; Lin, lineage; MHC II, major histocompatibilty complex class II; TLR, toll-like receptor.*

localize in the paracortex (27). The CD141<sup>+</sup>CD14<sup>−</sup> DC subset, characterized by Clec9A expression, is mostly distributed around the LN cortex (inner and outer) (25). By contrast, CD1c<sup>+</sup>CD14<sup>−</sup> DCs have been reported to localize within the T-cell zone in close proximity to the B-cell zone (10). Interestingly, in axillary and pulmonary LNs, but not in the spleen and tonsils, a high frequency of a HLA-DR+ cells with the CD141+CD14+DC-SIGN<sup>+</sup>CD206<sup>+</sup>CD1c<sup>+</sup>CCR7int/low phenotype has been reported, suggesting that this subset could be related to a migratory subset. This population is found in the diffuse T-lymphocyte regions of the LN paracortex (10, 27).

### Monocyte-Derived DCs (ModDCs)

*In vitro* experiments have documented that monocytes are important precursors of DCs (28, 29). However, it has been difficult to properly identify ModDCs *in vivo* due to common features shared by cDCs, monocytes and macrophages. Recent data suggest that a ModDCs subset may exist in humans (10–12, 25, 30). For example, studies in steady-state conditions described a subpopulation of cells expressing CD1c<sup>+</sup>CD14<sup>+</sup>HLA-DR<sup>+</sup> in both blood and bronchoalveolar lavage fluid (BALF) (10, 18). Although it was demonstrated that blood CD1c<sup>+</sup>CD14<sup>+</sup> cells have monocytic features, these cells have increased antigenpresenting ability and a different gene signature compared to monocytes (18). Interestingly, in non-diseased lung tissue CD1c<sup>+</sup>CD14<sup>+</sup> populations were shown to be enriched for the gene signatures of ModDCs described in the literature, which includes the expression of *ZBTB46*, *IRF4*, and *FLT3* genes (10). During inflammation, CD1c<sup>+</sup>CD14<sup>+</sup> cells have been reported in the BALF from sarcoidosis patients co-expressing CD141, CD123, and DC-SIGN, or in synovial fluid from rheumatoid arthritis (RA) patients and carcinomatous ascites from untreated cancer patients co-expressing CD1a, FcεRI, CD172a, and CD206 (11, 12). These cells were enriched for the ModDC signature and functionally ModDC from ascites showed an important capacity to polarize naive T cells into Th17 cells as well as to stimulate memory CD4 T cells to produce IL-17 (11).

In the past few years, additional DC subsets were associated with the induction of immune tolerance; however, their precise ontogeny and phenotype remains to be fully established. Gregory and co-workers described a DC subset expressing HLA-DR<sup>+</sup>CD14<sup>+</sup>CD16<sup>+</sup> receptors in human blood, which was able to induce type 1 regulatory T (Tr1) cells through the release of IL-10; hence, its name DC-10 (31). Furthermore, the presence of a DC subset expressing HLA-DR<sup>+</sup>CD141<sup>+</sup>CD14<sup>+</sup> was reported in skin dermis. This subset exhibited a potent inhibitory activity on skin inflammation.

### Functional Specialization of DCs

In terms of function, DCs can exhibit an immature phenotype at steady-state or a mature phenotype upon exposure to inflammatory stimuli. Immature DCs have a unique immune surveillance function. At this stage, DCs express low levels of MHC and costimulatory molecules such as CD80/B7.1, CD86/B7.2, CD40, OX40L, inducible T-cell costimulatory ligand, as well as low expression of adhesion molecules such as intercellular adhesion molecule-1 (ICAM-1/CD54) (32). Interestingly, at steady-state tissue CD1c<sup>+</sup>CD14<sup>−</sup> DCs exhibit a higher activation state, e.g., higher expression levels of CD80, CD83, CD86, and CD40 compared with their blood counterparts (22, 30).

Quiescent immature DCs can mature and become activated in local tissues in the presence of pathogen-associated molecular patterns or DAMPs in the context of sterile injury (e.g., autoimmunity or ischemia/reperfusion) and local inflammatory mediators (IFN-α, IL-1β, IL-6, TNF-α, or CD40L/CD154). Within the context of this maturation process, DC function is regulated by a core set of genes controlled by NF-κB and IFNmediated signaling (33). In this process, immature DCs evolve from an antigen-capturing mode to an antigen-processing and antigen-presenting mode by upregulating MHC molecules and costimulatory molecules along with chemokine receptors. This allows them to migrate to specialized lymphoid organs, release the corresponding polarizing cytokines, and initiate specific adaptive immune responses.

Regarding the fate and function of human DCs, both unstimulated CD1c<sup>+</sup>CD14<sup>−</sup> and CD141<sup>+</sup>CD14<sup>−</sup> DCs from blood, non-lymphoid, and lymphoid tissues were shown to be more immunogenic than pDCs, with an increased capacity to process and present soluble foreign antigens, including transplantderived alloantigens, as immunogenic MHC:peptide complexes to CD4<sup>+</sup> T cells (25, 34–36). It has been reported that both blood CD1c<sup>+</sup> DCs and CD141<sup>+</sup> DCs efficiently induce Th1 polarization in allogeneic co-culture assays, the latter with increased release of IFN-γ upon maturation (9). CD141<sup>+</sup> DCs were also shown to be more efficient at inducing Th2 cells compared to CD1c<sup>+</sup> DCs (20). By contrast, both CD1c<sup>+</sup> and CD141<sup>+</sup> DCs derived from lymphoid tissues efficiently induced Th1 and Th2 responses (21). In lung tissues, CD1c<sup>+</sup> DCs were shown to have a great capacity to induce Th17 responses following *A. fumigatus* challenge (37). In addition to their capacity to induce effector CD4<sup>+</sup> T cells, all DC subsets isolated from lymphoid tissues were able to efficiently cross-present soluble antigens to CD8<sup>+</sup> T lymphocytes (21). CD141<sup>+</sup> DCs are referred to as "human cross-presenting DCs" due to their functional homology with mouse CD8α+ DCs (38, 39), in particular with respect to the expression of TLR3 which promotes cross-priming and is required for the production of large amounts of IFN-λ upon TLR3 ligation (13).

In comparison to cDCs, pDCs have a similar distribution in peripheral blood and lymphoid organs, but are present at lower numbers in tissues (30, 40). In the immature state, pDCs express lower levels of costimulatory molecules but multiple pattern recognition receptors that are important for type-I IFN secretion, including intracellular TLR7 and TLR9 (14). In the presence of infectious or inflammatory stimuli, pDCs traffic to lymphoid organs and sites of inflammation. While their role was first described in response to viral infections via the recognition of nucleic acids, tissue-resident gut and airway pDCs have been shown to exert a pivotal role in oral and mucosal tolerance (15, 16, 41). In experimental mouse models of SOT as well as in clinical liver transplantation, pDCs were associated with the generation of alloantigen-specific Treg promoting prolonged allograft survival (42–45). Activated pDCs could also induce CD8<sup>+</sup> Treg in *in vitro* co-cultures (46).

Overall, these studies highlight the diverse responses of DCs depending on their origin. *Bona fide* cDC were demonstrated to have an inherent capacity to induce immunogenic responses, while pDCs and some immature ModDC subsets participate in tolerance induction. In humans, however, the functional specialization of DCs in the polarization of T cells appears to be less sharply defined compared to mice. The nature and the intensity of the stimuli, as well as the local environment, play an important role in determining the functional specialization of human DCs.

### MECHANISMS OF IMMUNE REGULATION BY DCs

Dendritic cells play an important role in the maintenance of immune homeostasis and self-tolerance under steady-state conditions. Their significant role in the induction and maintenance of tolerance has been demonstrated in experimental models. Constitutive or conditional depletion of cDCs was shown to break self-tolerance of CD4+ T cells, leading to spontaneous development of lethal autoimmunity manifested by splenomegaly, neutrophilia, autoantibody formation, and an increased frequency of Th1 and Th17 effector cells (47, 48).

### Surface Molecules Expressed on tolDCs

In the absence of local inflammation, DCs remain immature with low surface expression of MHC class II and costimulatory molecules, reflecting their participation in the maintenance of peripheral immune tolerance. Indeed, some DC subsets, such as CD103<sup>+</sup> DCs in mice and pDCs, blood DC-10, and skin CD141<sup>+</sup>CD14<sup>+</sup> DCs in human, exhibit inherent tolerogenic properties including the ability to induce Treg and/or promote T-cell hyporesponsivness to antigenic stimuli (15, 31, 49, 50). In addition to low expression of MHC class II and costimulatory molecules, tolDCs overexpress inhibitory molecules such as HLA-G, programmed death ligand (PD-L)-1 and PD-L2, and galectins that contribute to their tolerogenic potential. HLA-G is a non-classical MHC class I antigen that plays an important role in materno-fetal tolerance. Through interactions with inhibitory receptors expressed on maternal NK cells (killer cell immunoglobulin-like receptor, KIR) and T cells (ILT2 and ILT4), the expression of HLA-G on fetal cells protects them against maternal alloreactive cytotoxic cells (51). HLA-G engagement of the human inhibitory receptor ILT4 overexpressed on DCs in transgenic mice promoted longterm survival of allogeneic skin grafts, in part as a result of the downregulation of MHC class II and costimulatory molecules, leading to the induction of Treg and hyporesponsiveness of the alloreactive T-cell repertoire (52). Expression levels of PD-L1 and PD-L2 on DCs increase during DC maturation. These ligands can interact with the inhibitory receptor PD-1 expressed on activated T cells and Treg, thus contributing to T-cell homeostasis (53). Galectins have been identified as important regulators of T cells and DCs (54–56). Galectin-1 was shown to inhibit T-cell effector functions by promoting growth arrest and apoptosis of activated T cells (57, 58), and by blocking pro-inflammatory cytokines secretion by DCs (59). Moreover, galectins are overexpressed in the microenvironment of tumors and have been implicated in their immune escape. In *in vivo* models, DCs constitutively expressing galectin-1 delayed the onset of autoimmune diabetes in mice (60). Conversely, galectin-1-deficient mice experienced accelerated rejection of skin allografts (61).

### Immunomodulatory Molecules Secreted by tolDCs

Tolerogenic DCs were shown to secrete molecules, such as transforming growth factor-beta (TGF-β), IL-10, and indoleamine 2,3-dioxygenase (IDO), which favor a tolerogenic environment and the induction and/or expansion of Treg. TGF-β is a pleiotrophic cytokine involved in multiple cellular functions, including growth, differentiation, proliferation, remodeling, apoptosis, and immune homeostasis. TGF-β is secreted in a latent form complexed with latent TGF-β binding protein and latency-associated peptide. tolDCs were shown to play a crucial role in both the release of TGF-β and activation of the latent TGF-β protein complex (62, 63). TGF-β has been also involved in the induction of Foxp3 expression and peripheral conversion of conventional naïve CD4<sup>+</sup> T cells into induced Treg (iTreg) in the presence of IL-2 (64, 65). Through their constitutively high expression of the inhibitory receptor cytotoxic T-lymphocyte antigen-4 (CTLA-4/CD152), Foxp3<sup>+</sup> Treg ligate B7.1/2 expressed on mature DCs and outcompete costimulatory CD28 unregulated on effector T cells (Teff) (66). The interaction between B7.1/2 and CTLA-4 was shown to promote the expression of IDO by DCs, a potent regulatory molecule that catalyzes the degradation of tryptophan required for Teff functions (67–72). In addition, tryptophan catabolites, such as kynurenine, quinolinic acid, and 3-hydroxyanthranilic acid exhibit direct immunosuppressive properties (73, 74).

### Function of tolDCs

Mouse CD103<sup>+</sup> DCs as well as both human and mouse pDCs were shown to mediate oral tolerance through an IDO-dependent mechanism (49). In human blood, DC-10 induce and expand Tr1 cells through the release of IL-10 and TGF-β (31, 75). DC-10 identified as CD11b<sup>+</sup>CD11c<sup>+</sup>CD14<sup>+</sup>CD16<sup>+</sup>CD83<sup>+</sup>HLA-DR<sup>+</sup> cells that do not express CD1a and CD1c. Furthermore, DC-10 expresses the inhibitory receptors ILT2, ILT3, ILT4, and HLA-G. Despite concomitant high surface levels of costimulatory molecules (CD40, CD80, and CD86), DC-10 exhibit potent tolerogenic activity (31). Another DC population characterized by CD1c<sup>+</sup>CD14<sup>+</sup>CD16<sup>−</sup> expression was found in the blood of melanoma patients and exhibited immunosuppressive functions by suppressing T-cell proliferation in an antigen-specific manner. It was suggested that this DC subtype modulated T-cell responses through the expression of PD-L1 (18). A similar phenotype was described in minced lung tissues. The frequency of the CD1c<sup>+</sup> subset (including the CD1c<sup>+</sup>CD14<sup>+</sup> fraction) was increased in patients with chronic obstructive pulmonary disease, suggesting that these cells may be involved in the enhanced susceptibility of these patients to infections. Indeed, this subset favored the generation of IL-10-secreting CD4+ T cells and mediated immunosuppression through IL-10, IL-27, and ICOS-L (76). In skin dermis, resident CD141<sup>+</sup>CD14<sup>+</sup> DCs were shown to produce large amounts of IL-10 and were able to induce Treg (50).

Besides their role in controlling peripheral immune responses, DCs play a role in the maintenance of central tolerance, as trafficking peripheral DCs can home to the thymus and promote negative selection of antigen-reactive T cells, thus contributing to a safe peripheral T-cell repertoire (77, 78). By presenting antigens directly within the thymus, DCs and particularly resident pDCs influence the generation of natural Foxp3+ Treg, a process mediated by the IL-7-related molecule, thymic stromal lymphopoietin (TSLP), which is secreted in the thymic medulla (79–81).

### GENERATION OF tolDCs

Dendritic cell-based therapeutic approaches are being explored with the aim to reestablish self-tolerance in autoimmune diseases, and to promote alloimmune tolerance after SOT. Several strategies for the generation of tolDCs are being explored (**Figure 1**). These include treatment with pharmacologic agents or cocktails of immunomodulatory cytokines, genetic engineering, and exposure to apoptotic cells. Research groups have developed protocols to generate and expand antigen-specific tolDCs *in vitro*. Most of these *in vitro* conditioning regimens aim to stabilize the immature state of DCs, even in the presence of strong inflammatory challenges [e.g., lipopolysaccharide (LPS)]. The resulting tolDCs also express and/or secrete immunomodulatory molecules that favor the development and expansion of Treg (82, 83). After adoptive transfer *in vivo*, maturation-resistant tolDCs may, therefore, promote peripheral tolerance mainly by inducing antigenspecific T-cell hyporesponsiveness and an immuno-regulatory microenvironment.

### Pharmacologic Interventions for tolDC Induction

Various pharmacological agents have been used to generate tolDCs, including immunosuppressive drugs, cyclic AMP inducers, chemicals, cytokines, and growth factors (**Table 2**) (84, 85). Many of these agents were primarily studied for their inhibitory

(in the case of autoimmune diseases or transplantation) and can be further pulsed *in vitro* with specific antigens (peptides, donor cell lysates, apoptotic cells). tolDCs can regulate Teff responses by various mechanisms: 1. Fas/FasL pathway-mediated deletion; 2. Production of IDO which degrades the essential amino acid tryptophan through kynurenine pathway, causing starvation of Teff. The production of IDO is favored by reversed signaling via interaction between CD80/CD86 on DCs and CTLA-4 on regulatory T cell (Treg). 3. Surface expression of inhibitory molecules and secretion of regulatory mediators. Abbreviations: CTLA-4, cytotoxic T-lymphocyte antigen-4; IDO, indoleamine 2,3-dioxygenase; iTreg, induced regulatory T cell; PBMC, peripheral blood mononuclear cells; tolDC, tolerogenic dendritic cell; Teff, effector T cell; Tmem, memory T cell; Tnaive, naïve T cell; tTreg, thymic-derived regulatory T cell.

Table 2 | Pharmacologic interventions to induce tolerogenic DCs (tolDCs).


*AMP, adenosine monophosphate; DSG, 15-deoxyspergualine; GM-CSF, granulocytemacrophage colony-stimulating factor; HLA, human leukocyte antigen; IL, interleukin; TGF, transforming growth factor-beta.*

effects on T-cell activation and proliferation. As such, some of them are currently used in the clinical treatment of autoimmune diseases and in the prevention of allograft rejection after SOT.

Dexamethasone, a potent immunosuppressant, blocks the differentiation and maturation of DCs and enhances their death by apoptosis (92, 109). Using rat bone marrow-derived DCs and human GM-CSF/IL-4-induced ModDCs, we demonstrated that pretreatment with dexamethasone-induced selective expansion of Treg and T-cell alloantigen-specific hyporesponsiveness in re-challenge experiments (110). Dexamethasone was reported to have synergistic effects with other drugs to induce tolDC, in particular 1,25-dihydroxyvitamin D3 (VitD3) (111). While traditionally known for its role in the regulation of calcium and bone homeostasis, VitD3 and its receptor were also described to regulate innate and adaptive immune responses. Exposure to VitD3 inhibited the expression of MHC class II, CD80, and CD86 on DCs with a high ratio of PD-L1/CD86, while reducing the production of pro-inflammatory cytokines, such as IL-12 and IL-23, and increasing that of TGF-β and IL-10. Moreover, VitD3 favored Treg development and blocked B-cell proliferation and differentiation toward antibody-producing plasma cells (112).

Inhibitors of the mammalian target of rapamycine (mTOR) pathway engage FK506-binding protein 12 forming a complex that blocks mTOR, but not calcineurin, resulting in non-specific inhibition of cell cycle progression and, therefore, of T- and B-cell proliferation. *In vitro* assays, together with experimental and clinical data, suggest that immunosuppression based on mTORinhibitors may favor the induction of peripheral tolerance. In rodent models, the adoptive transfer of rapamycin-conditioned alloantigen-pulsed DCs resulted in prolonged cardiac and skin allograft survival (87, 113). Moreover, rapamycin was shown to facilitate peripheral deletion of alloreactive Teff by promoting activation-induced cell death in experimental transplantation models, while selectively expanding human Foxp3<sup>+</sup> Treg and Tr1 cells both *in vitro* and *in vivo* (114–117).

Various other immunosuppressive drugs and biologic agents were described to generate tolDCs *in vitro* including mycophenolic acid (MPA) formulations (90, 118), deoxyspergualin (DSG) and its analogs (93), aspirin (98, 99), retinoic acid (117), and prostaglandin E2 (119). These substances mainly interfere with NF-κB pathway-mediated DCs maturation and the capacity of DCs to produce IL-12p70 (84, 85).

Bone marrow-derived DCs exposed *in vitro* to the immunomodulatory cytokines IL-10, TGF-β, or low-dose GM-CSF in the absence of IL-4 exhibit low expression levels of costimulatory molecules and pro-inflammatory cytokines with minor changes in MHC class-I and -II molecules. These conditioned DCs were less immunogenic when co-cultured with CD4<sup>+</sup> and CD8<sup>+</sup> T cells, while promoting the expansion of Treg (both natural Treg and iTreg) and antigen-specific T-cell hyporesponsiveness *in vivo* upon re-challenge (108, 120).

Some of these agents also directly promote the differentiation of tolDCs *in vivo*. For example, local or systemic presence of cytokines such as IL-10, TGF-β, and even IFN-γ-induced differentiation of monocytes into tolDCs and promoted FoxP3<sup>+</sup> Treg (121). The eye is a known locally immune-privileged site and the aqueous humor constitutively contains molecules that maintain DCs in an immature state, such as TGF-β2, IDO, the neuropeptide α-melanocyte stimulating hormone (α-MSH) and FasL (122). The epithelium also plays a critical role in dampening inflammation through the release of epithelial-derived factors, including prostaglandin E2, TSLP, retinoic acid, and TGF-β, which are able to promote tolDC (119, 123).

Cobalt protoporphyrin is an inducer of heme oxygenase (HO)-1, an intracellular enzyme that catalyzes the degradation of heme, resulting in the production of biliverdin and carbon monoxide. HO-1 expression is induced by local oxidative stress. Experimentally, HO-1 upregulation was protective in the context of inflammatory processes and after allogeneic SOT (124). Besides maintaining HO-1 expression on human DC, cobalt protoporphyrin prevents their maturation and promotes the secretion of regulatory cytokines (124, 125). Interestingly, HO-1 is also able to inhibit the activation of T, B, and NK cells (126).

### Genetic Engineering of DCs

We and others have used gene transfer technology to generate tolDCs *via* increased expression of immunomodulatory molecules such as IL-10, TGF-β, CTLA-4, IDO, PD-L1, or ligands for receptors resulting in T-cell deletion such as CD95/Fas (TNF-family related death receptor) and TNF-related apoptosis-inducing ligand (127–129). Using gene therapy approaches, recombinant adenovirus vectors typically achieve great transfection efficiencies but are limited by pro-inflammatory effects leading to DCs maturation (130). Methods using recombinant retrovirus vectors and, recently, non-viral gene transfer methods such as nucleoporation may induce lower degrees of DCs maturation (131). As an example, genetically engineered DCs over-expressing IDO regulated T-cell alloresponses *in vitro*, while IDO adenovirus-mediated gene transfer into the donor heart attenuated acute rejection of MHC-mismatched cardiac allografts in rats (132). Compared to pharmacological conditioning, genetic engineering of DCs using retroviral or lentiviral vectors offers the advantage of a potentially more stable cell phenotype and function *in vivo*. Another interesting approach that was described is the *in vivo* transfer of antigen-encoding bone marrow progenitor cells which, at steadystate, prevent antigen-specific sensitization and promote T-cell tolerance mostly by deletional mechanisms (133–135).

### DC Exposure to Apoptotic Cells

Exposure of DCs to early apoptotic cells down-modulates their stimulatory functions (136). DC internalization of apoptoticcells-associated molecular patterns selectively leads to decreased production of pro-inflammatory cytokines (e.g., IL-1, TNF-α, IL-6, and IL-12), while enhancing the secretion of IL-10 and TGF-β (137). This process also limits upregulation of MHC class II and costimulatory molecules such as CD40, CD80 and CD86, hence maintaining DCs in an immature state (138). This mechanism contributes to self-tolerance and was exploited to induce tolerance to alloantigens in SOT. Following intravenous injection, donor allogeneic apoptotic cells were rapidly internalized in the spleen by red pulp macrophages and marginal zone DCs and were able to prolong cardiac allograft survival in rodents (139). Interestingly, T-cell depleting therapies used in clinical SOT, such as anti-CD3 or anti-CD52 monoclonal antibodies, were shown to induce T-cell apoptosis *in vivo.* This effect was associated with TGF-β secretion by DCs and subsequent expansion of Treg (140). Thus, administration of apoptotic cells or direct induction of apoptosis *in vivo* could be used to promote the generation of tolDCs *in vivo*. Of note, the induction of apoptosis *in vivo* must be carefully controlled to prevent simultaneous activation of the necrotic cell death mechanism and subsequent release of necrotic cell-associated antigens. These antigens were demonstrated to be able to stimulate CD141+CD14− DCs through the Clec9A receptor favoring antigen cross-presentation and, hence, CD8<sup>+</sup> T-cell responses (8).

### THERAPEUTIC USE OF tolDCs IN AUTOIMMUNE DISEASES

In Europe and North America, 5% of adults, of whom two-thirds are females, suffer from autoimmune diseases, the most prevalent pathologies being type 1 diabetes (T1D), psoriasis, RA, inflammatory bowel disease (IBD), and multiple sclerosis (MS).

### Type 1 Diabetes

Type 1 diabetes is due to a breakdown of self-tolerance and is mainly orchestrated by CD4<sup>+</sup> and CD8<sup>+</sup> autoreactive T cells that activate B cells, resulting in the production of autoantibodies specific for pancreatic islets β-cell antigens with progressive immune-mediated destruction of the β-cell mass and insulin insufficiency (141). Insulin treatment increases the life expectancy of T1D patients; however, it often fails to prevent T1D-associated cardiovascular and renal complications with increased morbidity and mortality. This opens the way to immune-based interventions aiming at restoring immune tolerance and preventing early T1D either by targeting autoreactive T cells (142) or by modifying the immunogenicity of DC. As T1D is a progressive and not a relapsing-remitting autoimmune disease, there is only a window of opportunity treatment period at the onset of the disease, in order to reinstitute self-tolerance and preserve the function of the existing β-cells mass.

In non-obese diabetic (NOD) mice, GM-CSF treatment prevented the development of diabetes primarily by inducing tolDCs and Treg (143). Another therapeutic option for T1D consists of transplantation of pancreatic islets or whole pancreas; however, such approaches require long-term immunosuppression to prevent allograft rejection and reoccurrence of autoimmunity. In an experimental model of syngeneic pancreatic islets transplantation in NOD mice, both transient TGF-β expression within islets and transplantation of islets grafts containing TGF-β-conditioned tolDCs reduced the activation of islet autoantigen-specific T cells in graft-draining LNs, resulting in prolonged graft survival (144). These results support the notion that TGF-β-induced tolDCs could be an effective strategy to restore peripheral tolerance within the context of an established autoimmune disease. Aside from generating tolDC, TGF-β promoted the survival of thymicderived Treg and the differentiation of iTreg (64, 145, 146).

Other strategies using tolDCs for the prevention or treatment of T1D currently are being explored in experimental models. Antisense oligonucleotides have been used to specifically downregulate costimulatory molecules, resulting in DCs with an immature phenotype. A single injection of bone marrow-derived DCs engineered *ex vivo* with a mixture of antisense oligonucleotides targeting the CD40, CD80, and CD86 primary transcripts significantly delayed the onset of diabetes in syngeneic NOD recipients. The beneficial effect of these tolDCs was partly mediated by an increase in Treg (147). Based on these encouraging experimental data, efforts have been made at translating the use of autologous tolDCs in patients with new-onset T1D, aiming to prevent disease progression, or even to revert established disease (148, 149). A phase I clinical study showed that intradermal injection of autologous monocyte-derived costimulationimpaired tolDCs (10 × 106 cells every 2 weeks for a total of 4 administrations), treated *ex vivo* with antisense oligonucleotides, was safe and well tolerated in patients with established T1D (150). A phase II follow-up clinical trial (ClinicalTrials.gov identifier NCT02354911) using DCs isolated from patients with recent-onset T1D is ongoing. Potential therapeutic success will be evaluated through the improvement of the glycemic control as evidence for a preserved β-cell mass. A similar clinical study using costimulation-impaired tolDCs is currently registered (identifier NCT01947569). This clinical trial includes a sequential open-label, phase-IB safety assessment and a randomized, double-blind, phase-IIA efficacy trial aiming at maintaining and improving residual β-cell function in new-onset T1D patients.

### Psoriasis

Psoriasis is a chronic inflammatory skin disease mainly characterized by abnormal keratinocyte proliferation and differentiation causing thickening of the epidermis (151). The psoriatic skin shows a prominent infiltration of neutrophils in the epidermis, together with macrophages, DCs, and T cells in the dermis. The successful use of cyclosporine A, a drug that inhibits early T-cell activation by blocking the TCR-downstream calcium–calcineurin pathway, highlights the role of T cells in the pathogenesis of the disease (152). Psoriasis has been associated with impaired Treg suppressive functions, resulting in overproduction of pro-inflammatory cytokines such as TNF-α and IFN-γ, as well as IL-17 and IL-22 produced by Th1 and Th17 effector cells, respectively (151, 153–155). The neuropeptide α-MSH is a well known mediator of skin pigmentation and has recently been shown to exert anti-inflammatory and immunomodulatory activities (106). Treatment with α-MSH ameliorated psoriasislike skin inflammation, in part by suppressing the proliferation and effector function of Th17 cells. The beneficial effect of α-MSH was shown to be mediated by tolDCs and functional iTreg (107).

### Rheumatoid Arthritis

Rheumatoid arthritis, an autoimmune disease associated with chronic joint inflammation and destruction, is characterized by infiltration of innate immune cells (neutrophils, monocytes, NK cells and DCs) as well as T and B cells in the synovial compartment (156). The current treatment of RA includes immunosuppressive drugs such as corticosteroids, cytokine antagonists (anti-TNF-α), costimulation blockade (CTLA-4 Ig), and B-cell depleting monoclonal antibodies. More recently, the potential of DC-mediated immunomodulation for the treatment of RA was investigated. Clinical-grade tolDCs have been generated from monocytes of patients with RA by conditioning them *ex vivo* with VitD3 and dexamethasone. The resulting tolDCs exhibited reduced costimulatory molecules expression, low production of pro-inflammatory cytokines, and impaired capacity to stimulate antigen-specific T cells. Importantly, the phenotype and functional characteristics of tolDCs generated from RA patients were comparable to those generated from healthy individuals. These tolDCs remained stable in the absence of immunosuppressive drugs even after further challenge with pro-inflammatory mediators (157). Interestingly, the tolDCs exhibited high cell-surface expression of TLR2 compared to mature immunogenic DC. The use of this marker should be considered in future immunotherapeutic protocols to assess the quality and stability of tolDCs produced *ex vivo*. An ongoing registered phase I randomized, placebo-controlled trial (identifier NCT01352858, AutoDECRA) aims to generate autologous tolDCs to be injected (single dose) into the knee joint of patients suffering from RA. This clinical study will evaluate the effect of injected tolDCs on both the local and the systemic disease activity.

### Inflammatory Bowel Disease

The gut mucosa is constantly exposed to food antigens, pathogens, and commensal microorganisms, and holds the largest mass of lymphoid tissues in the body. The interplay between the intestinal epithelium and the local innate and adaptive immune system is crucial to the maintenance of immune homeostasis and oral tolerance. IBD (Crohn's disease and ulcerative colitis) is mainly a consequence of loss of peripheral tolerance to otherwise harmless bacterial flora with dysregulated T-cell function in response to local intestinal antigens (158). Current treatments include corticosteroids, azathioprine, and 6-mercaptopurine, as well as anti-TNF-α and anti-α4β7 integrin therapies in severe cases. T-cell activation and effector function is dependent on the microenvironment in which antigen presentation occurs; hence, tolDCs may have a potential for reestablishing intestinal immune regulation (159). CD103<sup>+</sup> DCs are found in the human gut under normal conditions (160). These cells express low levels of CD40, TLR2 and TLR4, secrete IL-10 but not IL-12, and produce retinoic acid and IDO that promote the differentiation of iTreg and local immune regulation (49, 161). Peripheral tissue-resident pDCs exert a pivotal role in oral and mucosal tolerance (15, 16). An aberrant pDC distribution and effector function was described in the mesenteric LNs and inflamed mucosa of patients with Crohn's disease compared to healthy individuals (162). Moreover, immature peripheral blood pDCs and cDCs were reduced during flares in IBD patients (163). A phase I randomized clinical study (identifier NCT02622763 TolDecCDintra) currently evaluates the safety and clinical efficacy of autologous tolerogenic ModDCs injected into the intestinal lesions identified by endoscopy in patients with refractory Crohn's disease.

G-CSF therapy has been shown to be beneficial in Crohn's disease patients. The benefit was associated with increased numbers of pDCs in the gut mucosa and induction of IL-10 production (164). IL-10 is a crucial immunoregulatory cytokine in the gut, as documented by severe spontaneous IBD in IL-10-knockout mice (165). Protocols are under development to produce tolDCs under clinical-grade conditions for IBD patients by conditioning ModDCs with IL-10, together with a cocktail of other cytokines and prostaglandin E2 (166). The generated tolDCs display a semi-mature phenotype (intermediate expression of CD80 and CD86, MHC class II low), produce IL-10 with low levels of IL-12p70, IL-23, and TNF-α, and remain stable even in proinflammatory conditions. These data suggest that the strategy based on using autologous DCs (derived from patients with autoimmune diseases) may indeed be feasible for future immune therapies. Although the initial monocyte population is selected from patients with an overt inflammatory disease, the cells can be conditioned to acquire beneficial tolerogenic properties *ex vivo*.

### Multiple Sclerosis

Multiple sclerosis, a chronic inflammatory disease of the central nervous system (CNS), is predominantly a T-cell-mediated autoimmune disease characterized by leukocyte infiltration into the CNS, demyelination, and axonal loss. Besides current strategies targeting T and/or B cells, tolDCs may represent a potential therapeutic approach. Myelin peptide-loaded tolDCs were generated *ex vivo*, exhibiting a stable semi-mature phenotype and an anti-inflammatory cytokine profile. These tolDCs induced antigen-specific hyporesponsiveness in myelin-reactive T cells isolated from relapsing-remitting MS patients (167). IFN-β is an immunomodulatory agent used in the treatment of MS (168). Using healthy donors as well as MS patients PBMCs, it was shown that *in vitro* treatment with IFN-β enhanced PD-L1 expression on monocytes and DCs, inhibited antigen-specific CD4<sup>+</sup> T-cell activation, and increased Treg numbers. In addition, serial *in vivo* measurements in MS patients before and 6 months after initiation of IFN-β therapy revealed a significant increase in PD-L1 mRNA (169). Sex hormones such as estrogens can modulate immune responses, contributing to the observed difference in the incidence of autoimmune diseases between males and females. Although the overall incidence of autoimmune diseases is higher in women compared with men, estrogens were protective in the experimental autoimmune encephalomyelitis (EAE) animal model of MS, even during pregnancy (170, 171). *In vivo* exposure to estriol (E3), a pregnancy-specific estrogen, induced tolDCs (E3 tolDC) characterized by increased expression of the inhibitory molecules PD-L1, PD-L2, B7-H3, and B7-H4, as well as mediators such as IL-10 and TGF-β, along with decreased expression of IL-12, IL-23, and IL-6. The transfer of E3 tolDCs to mice prior to active induction of EAE prevented the development of the disease. The protective effect was associated with immune deviation from pathogenic Th1/Th17 cells to a Th2 response (172). Two phase I clinical trials are currently registered on the ClinicalTrials. gov website, assessing the feasibility and safety of tolDCs loaded with myelin peptide in patients with MS (identifier NCT02618902 and NCT02903537). The TOLERVIT-MS (NCT02903537) trial involves VitD3-induced tolDCs used at increasing doses, starting from 5 × 106 cells. Interestingly, different routes of administration will be explored for optimal efficacy, such as intravenous, intradermal or direct intranodal cell injection into cervical LNs.

### THERAPEUTIC POTENTIAL OF tolDCs IN SOT

In the absence of adequate immunosuppression after SOT, the recognition of donor alloantigens by recipient T cells initiates a strong immune response leading to alloimmunization and allograft rejection. As T cells play a central role in alloresponses, immunosuppressive regimens have been historically developed to target T cells, whereas recent attention has also been devoted to the roles of B cells and alloantibodies (173). The development of immunosuppressive drugs has led to decreased rates of acute rejection after SOT, but their long-term administration is associated with side-effects including cardiovascular and renal toxicities, as well as infections and tumors (174, 175). Moreover, current regimens have limited effects on T- and B-cell memory responses and may interfere with the induction and expansion of donor-specific Treg (83, 176). Therefore, inducing sustained donor-specific tolerance with minimal drug exposure remains an important goal to improve long-term outcomes in transplantation medicine.

Compared to immune responses to autoantigens or pathogens, SOT constitutes a unique situation whereby DCs present antigens to alloreactive T cells through three different mechanisms: the direct, indirect and semi-direct pathways (177). Recipient T cells can be activated by donor antigens either as intact allogeneic MHC:peptide complexes presented by donor mature DCs that have migrated out of the allograft (direct allorecognition) or as donor-derived MHC:peptide complexes that have been processed and presented by recipient DCs (indirect allorecognition) (178, 179). The semi-direct pathway involves the presentation by recipient DCs of intact donor MHC:peptide complexes that have been captured from donor cell membranes or exosomes (177). Therefore, DCs from either donor or recipient origin can be considered for the development of immunotherapeutic protocols in SOT.

### Donor-Derived tolDC

Experimental models have illustrated the potential of donorderived tolDCs to promote peripheral transplantation tolerance through the induction of donor-specific T-cell hyporesponsiveness, deletion, and/or regulation of alloreactive T cells (180–182). Costimulation-deficient tolDCs generated from donor bone marrow in the presence of GM-CSF, without IL-4, significantly prolonged the survival of MHC-mismatched heart allografts in mice when delivered (2 × 106 cells intravenously) 1 week before transplantation. The effect was only partially antigen-specific, as third-party tolDCs also prolonged graft survival, albeit to a lesser extent (median graft survival time 22 vs. 16.5 days, respectively; vs. 9.5 days in control non-treated mice) (180). However, *in vivo* maturation of the injected tolDC occurred, as evidenced by upregulation of CD80 and CD86, partially explaining the limited efficacy of tolDCs *in vivo*. Based on these results, a newer approach evaluated donor-derived tolDCs generated in the presence of GM-CSF and TGF-β, delivered in conjunction with CD40–CD40L costimulation blockade 1 week before transplantation. This strategy resulted in extended allograft survival (181). In a preclinical non-human primate (NHP) model, the infusion of donor-derived tolerogenic ModDCs (3.5–10 × 106 cells/kg) in combination with B7-CD28 costimulation blockade (CTLA-4 Ig given at day −7 and up to 8 weeks after transplantation) and rapamycin (started on day −2 and tapered over 6 months) significantly prolonged renal allograft survival (median graft survival time 113.5 vs. 39.5 days in controls) (86). In this study, tolDCs generated *in vitro* with VitD3 and IL-10 expressed low MHC class II and costimulatory molecules, high levels of PD-L1, and were resistant to inflammation-induced maturation. *Ex vivo* immune monitoring demonstrated regulation of donor-reactive memory T cells in tolDC-treated NHP compared to controls. Importantly, no adverse events, and particularly no significant allosensitization, were observed in the recipients.

Donor-derived tolDCs could also contribute to the induction of donor-specific central tolerance after SOT. A recent study evaluated the thymus-homing ability of DCs and their tolerogenic function. FMS-related tyrosine kinase 3 ligand (Flt3L) is a cytokine that synergizes with other growth factors to stimulate the proliferation and differentiation of hematopoietic progenitor cells. Flt3L-induced bone marrow-derived DCs (FLDCs), but not GM-CSF-induced DCs, were able to traffic to recipient thymus after systemic injection. Infusion of allogeneic donor-derived FLDCs induced clonal deletion of both CD4 and CD8 single-positive alloreactive thymocytes, leading to donorspecific central tolerance and prolonged survival of donor skin grafts (183).

### Recipient-Derived tolDCs

It should be emphasized that the use of recipient autologous DCs appears to be more feasible than that of donor DCs, as tolDCs can be prepared from peripheral blood of the recipient before SOT. Recipient tolDCs could be generated and stored while the patient is on the waiting list, and loaded with donor-derived antigens (HLA peptides, donor cell lysates) at the time of transplantation. Taking advantage of linked suppression mechanisms, it could be sufficient to generate recipient tolDCs expressing some but not all donor alloantigens (184). Importantly, the use of recipient DCs allows for the generation of tolDCs with the potential to induce indirect pathway CD4+ T-cell hyporesponsiveness and donor-specific Treg, while also controlling anti-donor humoral responses, which could have protective effects against chronic rejection (185–188). Recipient DCs pulsed with donor allopeptides and injected into the thymus in combination with one dose of anti-lymphocyte serum 7 days before transplantation induced donor-specific tolerance to cardiac (189) and pancreatic islets (190) allografts in rats. In an experimental model, the infusion of recipient DCs expressing donor intact MHC class I antigens selectively prevented indirect pathway alloreactive CD4<sup>+</sup> T cells activation and the generation of alloantibodies, resulting in long-term survival of MHCmismatched skin allografts (191). Interestingly, recipient bone marrow-derived tolDCs (generated in the presence of low-dose GM-CSF) infused one day before transplantation significantly prolonged cardiac allograft survival, even in the absence of prior *in vitro* pulsing with donor antigens. Although recipients showed reduced anti-donor cellular and humoral responses *ex vivo,* the effect of these tolDCs was not antigen-specific (192). This study suggests that, while alloantigen-specific immune regulation is desirable in the setting of SOT, recipient autologous tolDCs may also modulate the microenvironment in a way that favors allograft survival. The ONE study consortium aims at promoting clinical tolerance in living-donor renal transplant recipients (193). Within this consortium, a multicenter phase I/II safety/efficacy trial (ClinicalTrials.gov identifier NCT02252055) is currently evaluating the administration of autologous tolDCs on top of the standard immunosuppressive regimen (tacrolimus-MPA-corticosteroids).

### FEASIBILITY AND SAFETY OF DC-BASED IMMUNOTHERAPY

### Phenotype and Function of Generated tolDCs

In recent years, an improved understanding of the mechanisms that govern central and peripheral immune tolerance, along with a more precise characterization of DC subsets, has opened the door to their therapeutic use in autoimmune diseases and SOT (**Table 3**). There are, however, some caveats that need to be addressed in order to safely translate and evaluate DC-based immunotherapy to the clinical arena. One major hurdle is the generation of high amounts of tolDCs of reproducible quality using clinical-grade procedures. Cells isolation, culture, and preconditioning protocols need to be optimized to generate tolDCs with stable phenotypes and functions. A better identification of subset-specific markers of human tolDCs would be valuable for safe delivery to patients. Another major concern is the stability of *ex vivo* generated tolDCs in a pro-inflammatory environment *in vivo*. While many studies support a maturationresistant state of human tolDCs produced *in vitro* using various conditioning regimens, the *in vivo* environment may still induce maturation toward undesired immunogenic DCs (180, 194, 195). In the context of SOT, infused donor tolDCs have been shown to be short-lived, being eliminated mainly by recipient NK cells. Autologous recipient DCs could then process and present donorderived MHC:peptide alloantigens in an immunogenic context, leading to allosensitization and accelerated allograft rejection (196). However, this aspect remains controversial as there is evidence suggesting that, while donor tolDCs indeed rapidly die after infusion and are re-processed by splenic recipient DC,


*NP, not provided; T1DM, type 1 diabetes mellitus; NIAID, National Institute of Allergy and Infectious Diseases; NIDDK, National Institute of Diabetes and Digestive and Kidney Diseases.*

this may not lead to allograft rejection because alloantigens are delivered in the context of tolDCs (197). Nonetheless, tolDCbased "negative vaccination" most likely will need to be combined with immunomodulatory drugs *in vivo* to harness strong immune activation, as demonstrated in NHP transplantation models (86, 198). Overall, additional preclinical and clinical trials are now needed to demonstrate the feasibility and safety of tolDC-based immunotherapy in humans. For instance, in a clinical study antigen-loaded autologous tolDCs were injected intradermic in healthy volunteers. While the treatment was well tolerated, it also resulted in antigen-specific regulation of Teff (199, 200).

### Source of DCs and Administration Protocols

The source of DCs is a central aspect of tolDC-based strategies*.* In the context of autoimmunity or end-stage organ disease, a possible effect of the disease on DCs population and function needs to be addressed. Specifically, it must be demonstrated that autologous tolDCs from patients are as stable as tolDCs derived from healthy individuals (157). The number of cells to be delivered, the most appropriate timing of injection, and the route of administration also need to be carefully evaluated. Antigen-specificity of DC-based immunomodulation is another open issue with respect to clinical applications. Targeted regulation of antigen-specific T-cell responses would avoid generalized immunosuppression and impaired immune surveillance leading to infections or the development of malignancies. However, specific autoantigens that are responsible for T-cell priming have not been identified in some autoimmune diseases such as IBD, and peripheral DCs used for immunotherapy may not present some tissue-specific antigens. It also should be considered that, in deceased donor SOT, donor alloantigens are not known until the day of transplantation.

In an attempt to limit the workload and costs of *ex vivo* generation of tolDCs, while also circumventing the uncertainty regarding their stability, these cells could be directly generated *in vivo*. As discussed, the administration of apoptotic cells or *in vivo* induction of apoptosis could be a first option in this regard (136, 201). Another possibility may involve the administration of specific immunomodulatory cytokines or drugs. Infliximab, a chimeric monoclonal antibody that neutralizes both soluble and membrane-bound TNF-α, is an effective treatment in autoimmune diseases such as psoriasis, RA and IBD, which can lead to long-term remission. Besides its inhibitory effects on T-cell activation and homing, infliximab was shown to impair the differentiation and antigen-presenting capacity of ModDCs (202) and to restore the suppressive function of previously compromised Treg in patients with RA (203). CTLA-4 Ig, a fusion protein that blocks B7-CD28 costimulation and is currently used in RA patients as well as in SOT, can also promote tolDC (204, 205). The combination of adoptive transfer of tolDCs and CTLA-4 Ig resulted in extended survival of MHC-mismatched heart allografts in mice (206). In an experimental model, CTLA-4 Ig suppressed collagen-induced arthritis by inducing tolDCs and expanding Treg. This effect was abrogated by anti-TGF-β treatment (207). Other drugs that can be used to modulate DCs functions are listed in **Table 2**. Finally, antigens could be directly delivered to steady-state quiescent DCs *in vivo* by specifically targeting DC-restricted endocytic receptors (DEC-205) with monoclonal antibodies (208), aiming at inducing antigen-specific T-cell hyporesponsiveness. Altogether, it will be important to take advantage of some currently used drugs or biological agents in order to promote an environment that favors tolDCs.

### Overcoming Memory Responses

Based on their central role in immune activation, it is very tempting to consider immunotherapy using tolDCs for tolerance induction to specific antigens. Indeed, the type of DC and the cytokine microenvironment at the time of antigen presentation are major components in the regulation of T-cell responsiveness. While many *in vitro* and *in vivo* experimental data support the potential of tolDCs in regulating the priming of naïve T cells, less evidence exist regarding memory T cells. This is an important issue as the human immune repertoire harbors antigen-specific as well as cross-reactive long-lived memory T cells and antibodies which could represent an obstacle for clinical translational of tolDCs-based immunotherapy in chronic autoimmune diseases as well as in SOT (209, 210). There are, however, some experimental and preclinical studies that illustrate the potential of tolDCs in controlling memory T cells (108, 133, 134), suggesting that tolDCs would be advantageous compared to Treg-based immunotherapy or more conventional immunosuppression that are less efficient in the presence of preexisting memory responses (211). In an experimental allergic airway disease model, tolDCs were shown to inhibit allergenspecific memory Th2 responses and airway inflammation in sensitized hosts (135). In a clinically relevant NHP model of MHC-mismatched renal transplantation with minimal immunosuppression, infusion of donor-derived tolDCs 1 week before transplantation prolonged allograft survival, with no evidence of host sensitization. This therapeutic effect was associated with selective attenuation of donor-reactive memory T-cell responses (86, 198).

### Migration of tolDCs

Although significant progress has been made in the generation of tolDCs, the capacity of DCs to access LNs and encounter T cells, as well as the survival of treated DCs remains a critical concern. It is known that immature DCs traffic poorly to LNs. Different approaches have been made to overcome this issue. Genetically engineered DCs that express CCR7 and IL-10 showed improved migration ability to T-cell zones of secondary lymphoid organs and promoted prolonged heart allograft survival in a mouse model (212). Interestingly, in the setting of cancer vaccines development, the direct intra-lymphatic delivery approach is being evaluated in a clinical setting. DCs injected into a lymphatic vessel showed a prolonged half-life compared to DCs injected intravenously and were rapidly detected in LNs where they elicited a strong T-cell response (213, 214). The technique proved to be feasible and safe for short-term delivery of DCs. However, research is still needed to assess the efficacy of this route and the clinical relevance in the context of tolerance induction.

### CONCLUSION AND PERSPECTIVES

The field of immune regulation has become increasingly complex with the identification of both lymphoid and non-lymphoid cells that are involved in the induction and maintenance of immune tolerance. DCs are at the cornerstone of adaptive immune responses and, therefore, represent an appealing tool for immunoregulatory therapies in autoimmune diseases as well as in SOT. While extensive data have been generated in animal models, questions remain regarding the distribution and function of human DCs subsets *in vivo*. Clinical protocols need to be optimized for the safe generation of tolDCs, ensuring stable phenotypes and immunomodulatory functions. Interestingly, although both exogenously transferred or *in vivo* induced tolDCs may have a short half-life, they were shown to induce a regulatory environment (regulatory cell-subsets and cytokines) that could promote a more prolonged maintenance of peripheral tolerance (215). Finally, tolDCs immunotherapy has a unique potential for inducing antigen-specific central tolerance.

Whether or not tolDCs can maintain their function and regulate memory T-cell and B-cell responses in a strong inflammatory environment still remains to be demonstrated in more stringent preclinical models (209, 216, 217). In the clinical setting,

### REFERENCES


tolDC-based therapy, therefore, may not be sufficient *per se* to promote tolerance and may need to be used in conjunction with immunomodulatory drugs. While further preclinical and clinical studies are needed before tolDC-based immunotherapy can be successfully translated to the clinic, the quest for modalities to induce immune tolerance has clearly improved our understanding of human DC biology. The production and use of tolDCs in the upcoming years will continue to represent a challenging and exciting road, which, ultimately, may improve various immunemediated clinical pathologies.

### AUTHOR CONTRIBUTIONS

All authors contributed to writing the article. The final version of the manuscript was reviewed and approved by all authors.

### ACKNOWLEDGMENTS

This work was supported by the Fondation Pierre Mercier pour la Science, the Fondation Medi-CAL Futur, the Fondation Lausannoise pour la Transplantation d'Organes (to DG), the Fondation Vaudoise de Cardiologie (to GV), as well as an unrestricted grant from Astellas (to DG and MP).


and the inflamed setting. *Annu Rev Immunol* (2013) 31:563–604. doi:10.1146/ annurev-immunol-020711-074950


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 Obregon, Kumar, Pascual, Vassalli and Golshayan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Regulating Immunogenicity and Tolerogenicity of Bone Marrow-Derived Dendritic Cells through Modulation of Cell Surface Glycosylation by Dexamethasone Treatment

*Kevin Lynch1 , Oliver Treacy1 , Jared Q. Gerlach1,2, Heidi Annuk <sup>2</sup> , Paul Lohan1 , Joana Cabral1 , Lokesh Joshi <sup>2</sup> , Aideen E. Ryan1,3 and Thomas Ritter1 \**

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*Lenka Palová Jelínková, Charles University, Czechia Maja Wallberg, University of Cambridge, United Kingdom*

*\*Correspondence: Thomas Ritter thomas.ritter@nuigalway.ie*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 15 September 2017 Accepted: 13 October 2017 Published: 30 October 2017*

#### *Citation:*

*Lynch K, Treacy O, Gerlach JQ, Annuk H, Lohan P, Cabral J, Joshi L, Ryan AE and Ritter T (2017) Regulating Immunogenicity and Tolerogenicity of Bone Marrow-Derived Dendritic Cells through Modulation of Cell Surface Glycosylation by Dexamethasone Treatment. Front. Immunol. 8:1427. doi: 10.3389/fimmu.2017.01427*

*1School of Medicine, Regenerative Medicine Institute (REMEDI), National University of Ireland Galway, Galway, Ireland, 2Glycoscience Group, NCBES National Centre for Biomedical Engineering Science, National University of Ireland Galway, Galway, Ireland, 3Discipline of Pharmacology and Therapeutics, School of Medicine, National University of Ireland, Galway, Ireland*

Dendritic cellular therapies and dendritic cell vaccines show promise for the treatment of autoimmune diseases, the prolongation of graft survival in transplantation, and in educating the immune system to fight cancers. Cell surface glycosylation plays a crucial role in the cell–cell interaction, uptake of antigens, migration, and homing of DCs. Glycosylation is known to change with environment and the functional state of DCs. Tolerogenic DCs (tDCs) are commonly generated using corticosteroids including dexamethasone, however, to date, little is known on how corticosteroid treatment alters glycosylation and what functional consequences this may have. Here, we present a comprehensive profile of rat bone marrow-derived dendritic cells, examining their cell surface glycosylation profile before and after Dexa treatment as resolved by both lectin microarrays and lectin-coupled flow cytometry. We further examine the functional consequences of altering cell surface glycosylation on immunogenicity and tolerogenicity of DCs. Dexa treatment of rat DCs leads to profoundly reduced expression of markers of immunogenicity (MHC I/II, CD80, CD86) and pro-inflammatory molecules (IL-6, IL-12p40, inducible nitric oxide synthase) indicating a tolerogenic phenotype. Moreover, by comprehensive lectin microarray profiling and flow cytometry analysis, we show that sialic acid (Sia) is significantly upregulated on tDCs after Dexa treatment, and that this may play a vital role in the therapeutic attributes of these cells. Interestingly, removal of Sia by neuraminidase treatment increases the immunogenicity of immature DCs and also leads to increased expression of pro-inflammatory cytokines while tDCs

**Abbreviations:** iDC, immature bone marrow-derived dendritic cell; tDC, tolerogenic bone marrow-derived dendritic cell; niDC, neuraminidase-treated bone marrow-derived immature dendritic cell; ntDC, neuraminidase-treated bone marrowderived tolerogenic dendritic cell; mDC, mature bone marrow-derived dendritic cell (LPS-treated); MLR, mixed lymphocyte reactions; Sia, sialic acid; Dexa, dexamethasone.

are moderately protected from this increase in immunogenicity. These findings may have important implications in strategies aimed at increasing tolerogenicity where it is advantageous to reduce immune activation over prolonged periods. These findings are also relevant in therapeutic strategies aimed at increasing the immunogenicity of cells, for example, in the context of tumor specific immunotherapies.

Keywords: tolerogenic dendritic cells, glycosylation, dexamethasone, immunogenicity, tolerogenicity, sialic acid, autoimmunity, cell therapy

### INTRODUCTION

Dendritic cells are professional antigen-presenting cells, which are a component of the innate immune system which induce adaptive immune responses (1). Dendritic cells (DCs) were first described by Steinman and Cohn in 1973 (2) and were subsequently identified to be potent activators of the immune system when employed in mixed lymphocyte reactions (MLRs) (3). DCs are a heterogeneous population classified in different subsets dependent on the origin (4). DCs have been extensively investigated for potential use as a cellular therapy due to their ability to maintain peripheral tolerance, which is of importance in the field of transplantation and autoimmunity. Since mature DCs are potent activators of the T-cell responses, pharmacological approaches have been used to maintain DCs in a maturation resistant state (5–7). The glucocorticoid dexamethasone (Dexa) has been widely used in this context (8–11). Glucocorticoids are potent immunosuppressive drugs that are used in clinical regimens to treat both Th1- and Th2-mediated inflammatory diseases including allograft rejection (12). Dexa is known to exert potent effects on many immune cells including DCs (8, 13). It has been consistently described in the literature that Dexa has inhibitory effects on the development of immature DCs (iDCs) (5, 8, 12, 14), and that it also impairs lipopolysaccharide (LPS) (TLR4) stimulation of DCs, which would otherwise lead to their maturation (mDCs) (15–17). In addition to this, Dexa-treated DCs have a reduced capacity to activate naïve T lymphocytes by interfering with Signals 1–3 important for T-cell activation (17).

In the context of transplantation, preclinical experiments suggested the potential therapeutic use of both donor and recipientderived tolerogenic DCs to prevent organ graft rejection (18). In a rat model, we have recently shown that pretreatment of donor DCs with Dexa *ex vivo* prevents the maturation of DCs and prolongs rat corneal allograft survival upon injection in corneal transplant recipients (13). However, the mechanisms of how tolerogenic DCs engage with other immune cells and exert their immunomodulatory effects are not completely understood. Despite this, tolerogenic DCs have been already tested in humans suffering from various diseases. As of this writing, there are currently eight tolerogenic DC cell therapies listed in Phase I/ II clinical trials for treatment of autoimmune disease and graft rejection (https://clinicaltrials.gov. September 2017, search for key words tolerogenic DCs), which highlights the importance and urgency of understanding the mechanisms associated with the therapeutic effect.

Glycosylation is one of the most vital and frequent forms of posttranslational modification and is involved in the function of many immune associated molecules. Some of the functions of glycosylation include, but are not limited to, protein folding and molecular trafficking to the cell surface (19–23). Glycosylation has also been implicated in the stability of proteins and protection from proteolysis (24). All immune cells are coated by a glycocalyx composed of a complex assortment of oligosaccharides (glycans), of which one frequent terminal component is sialic acid (Sia). Sias are a broad family of negatively charged, 9-carbon monosaccharides that are exposed to the cellular microenvironment and are involved in communication and in cellular defense (25). It has been reported that a typical somatic cell surface presents millions of Sia molecules (26) and also that they have long been noted to be important in immune cell behavior (27). It has been suggested that Sias can play important roles in both acting as a recognizable molecule for cellular interactions but also as a biological shield preventing receptors on cells recognizing their ligands (28). Large amounts of Sias on the cell surface of immune cells will result in an overall negative charge, which can have biophysical effects, such as the repulsion of cells from each other and subsequently disrupting cellular interactions (29).

Since immune cell interactions form the basis of immune responses, glycosylation is, therefore, likely to play a major role in dictating these responses. However, there is a significant knowledge gap as to how glycosylation modulates immune responses. Currently, little information exists on how DC glycosylation patterns change after Dexa treatment. Here, we present a comprehensive profile of bone marrow-derived DCs (BMDCs), examining their cell surface glycosylation before and after Dexa treatment as resolved by both lectin microarrays and lectincoupled flow cytometry.

In this work, the composition of the glycocalyx of both iDCs and tolerogenic DCs (tDCs) was altered using neuraminidase (sialidase) treatment and the functional consequences in immunogenicity and inhibition of T-cell proliferation were observed. We show that Sia is upregulated on tDCs contributing to the tolerogenic state of tDCs. However, removal of Sia leads to increased stimulatory activity of iDCs leading to enhanced T-cell activation and proliferation. These findings have important implications in strategies aimed at increasing tolerogenicity where it is advantageous to reduce immune activation over prolonged periods. These findings are also relevant in therapeutic strategies aimed at increasing the immunogenicity of cells, for example, in the context tumor specific immunotherapies.

## MATERIALS AND METHODS

### Animals

All animals used in experiments were accommodated in an accredited animal housing facility under a license granted by the Department of Health, Ireland, and were approved by the Animals Ethics Committee of the National University of Ireland, Galway. Bone marrow used in the generation of BMDCs was isolated from male Dark Agouti (DA, RT-1avl) rats at 8–14 weeks of age. For the allogeneic MLRs, male Lewis (LEW, RT-1l ) rats served as a source of lymphocytes, isolated from both the cervical and mesenteric lymph nodes and spleen. DA and LEW rats were obtained from Harlan Laboratories (Bicester, UK).

### Isolation and Generation of iDCs and tDCs

Immature DCs were generated using an adapted version of the protocol, which has been previously described (13) (Figure S1 in Supplementary Material). Briefly, on day 0, male DA rats of the specified age were sacrificed and the tibia and femur were surgically removed postmortem. The epiphyses were cut and the bone marrow was flushed from the long bones with a syringe/needle combination. The erythrocytes were removed from the suspension by lysis using a standard red blood cell lysis buffer (Sigma-Aldrich, Dublin, Ireland). After erythrocyte lysis, the cells were washed in RPMI-1640 (Gibco, Grand Island, NY, USA) medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mmol/L l-glutamine, 100 mmol/L nonessential amino acids, 1 mmol/L sodium pyruvate, 100 U/mL penicillin, 100 µg/mL streptomycin, and 55 µmol/L 2-β-mercaptoethanol (2β-ME) (Gibco). Cells were resuspended at a concentration of 1.5 × 106 /mL and plated at a concentration of 4.5 × 106 per well of a 6-well plate. The culture medium was supplemented with 5 ng/mL rat granulocyte-macrophage colony-stimulating factor (GM-CSF) (Invitrogen, Paisley, UK) and 5 ng/mL rat IL-4 (Peprotech EC, London, UK). Cells were incubated under standard cell culture conditions (37°C at 5% CO2) and, on the third day of culture, half of the medium from each well was harvested and cells were resuspended in fresh medium supplemented with rat GM-CSF and IL-4 and added back into the culture. On the fifth day, the supernatant was exchanged with fresh supplemented growth medium to remove dead granulocytes and lymphocytes. In experiments requiring tDCs, Dexa (Sigma-Aldrich) was added to the culture at 10<sup>−</sup><sup>6</sup> mol/L at this point. On the seventh day of culture, half of the medium was again removed and replaced with fresh supplemented medium (Dexa was added as required). To generate mDCs, LPS (1 µg/mL; Sigma-Aldrich) was added 24 h before the cells were cultured. Cultures were maintained until day 10 and then gently pipetted off the bottom of the wells for the *in vitro* assays.

### Neuraminidase Treatment

To produce neuraminidase-treated iDCs and tDCs (niDCs and ntDCs), BMDCs were harvested on day 10 of culture and 2 × 105 /mL were treated with 400 U/mL of recombinant *Clostridium perfringens* neuraminidase (P0720S, New England Biolabs, Ipswitch, MA, USA) in phosphate buffered saline (PBS) supplemented with 1 mM MgCl2 (Sigma-Aldrich), 1 mM CaCl2 (Sigma-Aldrich), and 1% bovine serum albumin (Sigma) for 90 min at 37°C.

### RNA-Isolation and RT-PCR

RNA was exacted from iDCs, tDCs, mDCs, niDCs, and ntDCs on day 10 using Bioline Isolate II RNA mini kits according to manufacturer's protocols. All cDNA was produced using RevertAidTM H Minus Reverse Transcriptase (Thermo Fisher Scientific, MA, USA) with random primers. For primer sequences of GAPDH, TNF-α, IL-12p40, inducible nitric oxide synthase (iNOS) IL-10, IDO, IL-6, and IL-1β, see Table S1 in Supplementary Material. All samples were normalized to expression of the house-keeping gene GAPDH and made relative to iDCs. All quantitative realtime PCR was performed according to the standard program using a real-time PCR system (StepOne Plus, Applied Biosystems, Thermo Fisher Scientific).

### Flow Cytometry

Cells were characterized by flow cytometry using the monoclonal antibodies (mAbs) CD11b/c-APC, CD80-PE, CD86-PE, MHCI-FITC, and MHCII-PE (BioLegend, San Diego, CA, USA). For analysis of the glycocalyx, lectins from *Maackia amurensis* (MAL II, indicating α2-3 Sia) and *Sambucus nigra* (SNA-I, indicating α2-6 Sia) were used (Vector Labs). Lectins were biotin conjugated. PE-streptavidin was used for detection. Negative controls for non-specific fluorescence were used, these consisted of PE-streptavidin staining solutions in the absence of the lectin conjugated to biotin. Lectins were prepared in lectin staining buffer (PBS containing 1% FBS, 1 mmol/L CaCl2, and 2 mmol/L MgCl2) and resuspended in FACS buffer (PBS containing 2% fetal calf serum and 0.01% NaN3, all from Sigma-Aldrich) before analysis using a FACS Canto II (BD Biosciences, Oxford, UK).

For analysis of the assays involving lymphocytes from the lymph nodes and spleen, the following mAbs were used CD3/PE, CD8/PE-Cy7, CD4/APC (BioLegend), and CD25/FITC (eBioscience, San Diego, CA, USA). Prior to staining, cells were washed with FACS buffer. mAbs were diluted in 50 µL FACS buffer, added to the cells, and incubated for 15 min at 4°C. To remove any unbound antibodies, the cells were washed three times with FACS buffer. The cells were then filtered through a nylon mesh (40 µm) before analysis in the cytometer.

### Mixed Lymphocyte Reaction/T Cell Proliferation Assays

Lymphocytes were isolated from the spleen and lymph nodes of LEW rats. T cells were washed with phosphate-buffered saline and stained in prewarmed (37°C) CellTrace™ Violet (CTV) phosphate-buffered saline staining solution (Invitrogen, Carlsbad, CA, USA) as per manufacturer's instructions. 2 × 105 CTV-stained T cells were stimulated at a 1:1 ratio with antirCD3/anti-rCD28-labeled beads in supplemented RPMI 1640 media. Assays were incubated at various BMDC: T-cell ratios in a humidified incubator for 4/5 days at 37°C following which T-cell proliferation and CD4 and CD8 expression were assayed by flow cytometry (mAbs CD4-APC and CD8α-PE-Cy7; Biolegend). T-cell proliferation, activation, and differentiation were analyzed using a FACS Canto II.

## Membrane Protein Extraction and Labeling

Membrane proteins were extracted from iDCs, tDCs, niDCs, and ntDCs using a commercial protein extraction kit (Mem-Per®, Thermo Fisher Scientific). Proteins recovered from 106 cells were labeled with 100 µg (10 mg/mL in DMSO) Alexa Fluor® succinimidyl ester 555 dye (Thermo Fisher Scientific) as per the manufacturer's instructions. Labeled protein was separated from unconjugated dye with Bio-Gel® P6 (Bio-Rad Laboratories, Dublin, Ireland).

### Lectin Microarray Construction and Sample Interrogation

Lectin microarrays were constructed essentially as described previously in Ref. (30). Forty-four lectins (Table S2 in Supplementary Material) sourced from multiple vendors were diluted to 0.5 mg/mL in PBS supplemented with 1 mM of respective haptenic sugar to maintain binding site integrity (see Table S2 in Supplementary Material) and printed on Nexterion® H (Schott, Mainz, Germany) functionalized glass substrates using a sciFLEXARRAYER S3 non-contact spotter (Scienion, Berlin, Germany). During printing, relative humidity and temperature were maintained at 62% (±2%) and 20°C, respectively. Following printing, slides were incubated in a humidity chamber overnight at 20°C to ensure completion of covalent conjugation. Unoccupied functional groups were deactivated by 1 h incubation with 100 mM ethanolamine in 50 mM sodium borate, pH 8. Finished slides were washed with PBS with 0.05% Tween-20 (PBS-T) three times for 3 min and once with PBS for 3 min, centrifuged dry (450 × *g*, 5 min), and stored at 4°C with desiccant until use.

Labeled cellular proteins were incubated with finished microarrays following extensive optimization as described in Ref. (30). All processes were carried out with limited light exposure. Samples were applied to microarrays using an 8-well gasket slide and incubation cassette system (Agilent Technologies, Cork, Ireland). 70 µL of each labeled glycoprotein at 0.5 mg/mL, in incubation buffer [TBS-T; Tris-buffered saline (TBS; 20 mM Tris– HCl, 100 mM NaCl, pH 7.2, supplemented with 1 mM CaCl2, and 1 mM MgCl2) with 0.05% Tween®-20], was applied to each well of the gasket. A total of 18 technical replicates were carried out for iDC and tDC profiling (encompassing samples of five biological replicates). Each microarray slide was loaded into a cassette with an accompanying gasket slide and placed in a rotating incubation oven (23°C, approximately 4 rpm) for 1 h. Incubation cassettes were disassembled under TBS-T, and microarrays were washed in a Coplin jar twice in TBS-T for 2 min each and once with TBS for 2 min. Microarrays were dried by centrifugation (450 × *g*) and imaged immediately using an Agilent G2505B microarray scanner at 5 µm resolution (532 nm laser, 100% laser power, 90% PMT).

### Microarray Data Extraction and Analysis

Data extraction and analysis was performed essentially as previously described (30, 31). In brief, raw intensity values were extracted from high-resolution \*.tif files using GenePix Pro v6.1.0.4 (Molecular Devices, Berkshire, UK) and a proprietary \*.gal file (containing feature spot addresses and identities) using adaptive diameter (70–130%) circular alignment based on 230 mm features. Numerical data were exported as text to Excel (Version 2010, Microsoft, Dublin, Ireland). Local backgroundcorrected median feature intensity data (F543median-B543) was analyzed. The median value, derived from data from six replicate spots per subarray, was handled as a single data point for graphical and statistical analyses.

Lectin microarray intensity values were normalized to the median total intensity value for all features across all subarrays. The significance of difference between relative intensity data (\**p*< 0.05, \*\**p*< 0.01, \*\*\**p*< 0.001, \*\*\*\**p*< 0.0001) was evaluated for each set of replicates on a lectin-by-lectin basis using a standard Student's *t*-test (two-tailed, two sample unequal variance). Unsupervised, hierarchical clustering of lectin-binding data was performed with Hierarchical Clustering Explorer v3.0 (http:// www.cs.umd.edu/hcil/hce/hce3.html). For clustering analysis, previously, normalized data were imported directly and clustered with the following parameters: no pre-filtering, complete linkage, Euclidean distance. Principal component analysis (PrCA) of previously normalized and pre-filtered data (those lectins which demonstrated *p* < 0.01 or better in the above *t*-tests, 15 in total) was performed using Minitab version 16.1.1 (Minitab, Inc., State College, PA, USA).

### Statistical Analysis

Data were analyzed using the software package FlowJo v10 (Tree Star, Ashland, OR, USA). All data were analyzed with Graphpad Prism V6 software (Graphpad Software, CA, USA) and are expressed as mean ± SEM unless otherwise indicated. Comparisons among multiple groups were made with one-way ANOVAs followed by Tukey's multiple comparisons test. Data sets with two groups were analyzed using an unpaired *t*-test. Differences were considered statistically significant when *p-*value was <0.05.

### RESULTS

### Dexamethasone Treatment of BMDC Induces a Tolerogenic Phenotype

Dexamethasone treatment of DCs has been reported to generate tolerogenic DCs (tDCs) (32). To generate iDCs, bone marrow was flushed from the long bones of the tibia and femur of DA rats and cultured in medium supplemented with GM-CSF, IL-4, and Dexa (for tDCs) as required (Figure S1A in Supplementary Material). Following isolation, cell surface characterization was performed using flow cytometry by gating on the CD11b/c population (**Figure 1A**). tDC generation did not result in any significant changes in cell size (**Figure 1B**, i) but the number of cells harvested from wells that were treated with Dexa was significantly lower than that of wells that were Dexa-free (**Figure 1B**, ii). This may be due to Dexa-induced apoptosis of the DCs, which has been reported by other groups (33). While lower numbers of cells were obtained from tDC wells, after harvesting and washing of the

#### Figure 1 | Continued

Isolation, generation, and characterization of immature DCs (iDCs), tolerogenic DCs (tDCs), and stimulated DCs (mDCs). Bone marrow was flushed from the femur and tibia of 8- to 14-week-old DA rats and cultured in IL-4 and GM-CSF cultured media for 10 days (Figure S1 in Supplementary Material). (A) Representative gating strategy. Cells were selected according to size and granularity (i) followed by live/dead discrimination based on Sytox blue negative cells (live) (ii). After single cell selection (iii), cells were selected by CD11b/c (APC) positivity (iv). (B) Changes in cell size (*n* = 3) (i), the number of cells harvested (*n* = 8) (ii), and viability of iDCs to tDCs (*n* = 4) (iii) was compared. (C) Both immature DCs (iDCs) and tDCs were analyzed by flow cytometry for their cell surface expression of MHC I (FITC), MHC II (PE), CD 80 (PE) and CD 86 (PE). Representative histograms and bar charts displaying relative fluorescence intensity (RFI) for flow cytometric analysis of DC cell surface. Median fluorescence intensities were established relative to iDCs. (D) The mRNA expression of interleukin 6 (IL-6), Indoleamine 2,3-dioxygenase (IDO), interleukin 1 beta (IL-1β), inducible nitric oxide synthase (iNOS), and IL-12p40 was analyzed in iDCs and tDCs. Normalized to GAPDH and fold change relative to iDCs. Error bars: mean ± SEM \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 one-way ANOVA, Tukey's multiple comparisons test.

cells, no significant changes in viability was noted (**Figure 1B**, iii). We also analyzed the expression levels of the costimulatory molecules CD80/CD86 and the major histocompatibility complex class I and II molecules (MHCI/II) as an indicator of the maturation status of generated iDCs and tDCs (**Figure 1C**). The expression levels of CD80, CD86, MHC I, and MHC II indicate that the iDCs display a semi-mature phenotype. However, when the cells were treated with Dexa, a significant reduction in the expression level of MHC II was observed with no changes in MHC I (**Figure 1C**). To mature iDCs or tDC *in vitro*, LPS was added to the cultures (1 µg/mL) for 24 h. A significant increase in both CD80/CD86, MHC I and MHC II was noted. However, tDCs following LPS treatment showed significantly reduced expression levels of CD80/CD86 and MHC I/II molecules compared to stimulated iDCs indicating a phenotype that is maturation resistant. iDC and tDC populations were also assessed for expression of pro- and anti-inflammatory markers with and without Dexa-treatment by qRT-PCR (**Figure 1D**). Results indicate that LPS stimulation of iDCs leads to an increase in mRNA expression of pro-inflammatory molecules such IL-6, IL-12p40, and iNOS. In contrast, tDCs are less sensitive to TLR4 stimulation compared to mDCs, indicated by no observed increases in IL-6, IL-12-p40, and iNOS after LPS treatment. Higher levels of IDO mRNA, which is known as a marker in tolerogenic cells, is present in LPS-treated tDCs when compared to mDCs. Interestingly, IL-1β mRNA expression does not seem to be regulated by Dexa, as LPS stimulation leads to a profound increase, which cannot be blocked by Dexa. All together these data indicate that Dexa treatment of iDCs leads to the generation of a tolerogenic DC phenotype with reduced expression of markers of immunogenicity and reduced expression of pro-inflammatory molecules but increases in immunoregulatory molecules.

### tDC Generation Modulates the Glycocalyx by Significantly Increasing Levels of **α**2-6-Linked Sia

Changes in DC glycocalyx after induction of tolerogenic phenotype have not been investigated. To address this knowledge gap, lectin microarray profiling of proteins extracted from the membranes of iDCs and tDCs and lectin-coupled flow cytometry of intact iDCs and tDCs was undertaken.

Comparisons of all lectin microarray replicate profiles were made by unsupervised hierarchical clustering. This clustering approach revealed two major clusters with separation at 53% minimum similarity (**Figure 2A**). With the complete linkage method employed, two untreated iDC replicates were placed into the tDC group while only three of the iDC replicates, two from biological set 2 and one from set 5 (**Figure 2A**), showed outlier behavior and were excluded from the major cluster containing the balance of the iDC replicate data. However, the well-defined separation of the vast majority of the iDC and tDC replicates into two groups (**Figure 2A**, Group 1 and 2) supports the solidity of the subtle profile differences and also the high level of reproducibility for the lectin profiling method in distinguishing membrane glycoprotein samples from iDCs and tDCs.

Median values obtained from normalized lectin microarray profile data (*n* = 18) for iDCs and tDCs were broadly similar with only small, but significant, differences in intensities noted at a subset of the lectin panel (**Figure 2B**). The general profiles of tDC glycoproteins remained similar to those of iDCs across lectin features. Furthermore, the lectin profiles displayed no obvious signs of cell stress as evidenced by a lack of elevation of signals suggesting increased endoplasmic reticulum- and proximal Golgi-associated glycan structures (i.e., increased evidence of high mannose structures). However, SNA-I showed a consistent intensity increase with tDC surface glycoproteins (*p* = 2 × 10<sup>−</sup>10) relative to iDCs, which is in line with previous findings from our group (13). PrCA performed using the 15 lectins, which demonstrated *p* < 0.01 (SNA-II, BPA, PNA, DSA, LEL, SNA-I, RCA-I, CPA, ECA, LTA, UEA-I, EEA, GS-I-B4, MPA, and VRA) revealed a division of replicate lectin profiles dominated by distinct groups containing iDCs or tDCs with minimal overlap and further reinforced the ability of these lectins to distinguish untreated iDCs from tDCs (**Figure 2C**). In short, these lectin microarray profiles demonstrate that the glycocalyxes of the iDC and tDCs are distinct. These changes were validated using lectin-coupled flow cytometry. The increase in SNA-I binding suggests an increase in quantity or better accessibility to α2-6-linked with no significant change suggested for α2-3-linked Sia (MAL-II) confirmed lectin microarray findings (**Figure 2D**).

### Neuraminidase Treatment of iDCs and tDCs Modulates Levels of **α**2-6-Linked Sia and Alters Expression Levels of Immunogenicity Markers

Sia has long been reported to be important in DC biology (28). Considering the dramatic increase observed after Dexa treatment confirmed by both flow cytometry and lectin microarray (**Figures 2B–D**), we cleaved Sia using neuraminidase to study phenotypical and functional changes upon removal. iDCs and tDCs were treated with neuraminidase (designated niDC and ntDC, respectively) and lectin binding profiles for SNA-I and MAL-II were analyzed using flow cytometry. Both niDCs and ntDCs showed a significant reduction in SNA-I binding intensities and trend decreases MAL-II binding intensities suggesting the successful removal of α2-6-linked and α2-3-linked Sia, respectively (**Figure 3A**, i–iv). Based on these results, we further investigated if the removal of Sia resulted in a detectable increase of the expression of MHC I, MHC II, CD80, and CD86 immunogenicity markers after treatment with neuraminidase. niDCs (**Figure 3B**, i) had small but significant increases in MHC II and CD86 expression when compared to iDCs. MHC I showed

\*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 one-way ANOVA, Tukey's multiple comparisons test.

a trend increase in expression on niDCs compared to iDCs, and there was no change in CD80 expression after treatment with neuraminidase. ntDCs (**Figure 3B**, ii) displayed a significant increase in both MHC I and MHC II with no changes in CD80 and a trend increase in CD86 after neuraminidase treatment. niDC and ntDC populations were also assessed for expression of pro- and anti-inflammatory markers by qRT-PCR (**Figure 3C**). Although there was some sample-to-sample variation, our data indicate that neuraminidase treatment of iDCs leads to dramatic increases in pro-inflammatory mRNA expression of IL-6, IL-1β, iNOS, TNF-α, and IL-12-p40. However, ntDCs are protected from this strong increase in pro-inflammatory cytokine expression in the case of iNOS and IL-12-p40, but mRNA levels of IL-6, IL-1β, and TNF-α are increased. Interestingly, levels of anti-inflammatory IL-10 are lost after neuraminidase treatment in both iDCs and tDCs. In summary, these results indicate that neuraminidase treatment reduces Sia on the cell surface of both iDCs and tDCs and leads to the stimulation of pro-inflammatory cytokine mRNA expression, which can be largely inhibited by Dexa treatment.

### Neuraminidase Treatment Alters Immunomodulatory Properties of iDCs and tDCs

Considering that the removal of Sia altered the immunogenic phenotype of both iDCs and tDCs, we further analyzed the effects of neuraminidase treatment on iDCs and tDCs through *in vitro* allogeneic coculture experiments. iDCs or tDCs from DA rats were treated with neuraminidase and cocultured with allogeneic lymphocytes. The immunogenic potential or the ability of niDCs and ntDCs to induce the proliferation and/or the activation of allogeneic lymphocytes was analyzed by T-cell proliferation assays (**Figure 4A**). Responder LEW rat T cells were analyzed based on their co-expression of CD3<sup>+</sup>CD4<sup>+</sup> or CD3<sup>+</sup>CD8<sup>+</sup> (**Figure 4B**). Proliferation of lymphocytes was measured using CellTraceTM Violet (CTV) and activation of lymphocytes was measured using CD25 as an activation marker. DA iDCs (**Figure 4C**, i) and tDCs (**Figure 4C**, ii) did not induce an allogeneic response as indicated by a lack of changes in LEW CD3<sup>+</sup>CD4<sup>+</sup> or CD3<sup>+</sup>CD8<sup>+</sup> T cell proliferation when compared to unstimulated lymphocytes alone. Additionally, we observed no significant changes in CD3<sup>+</sup>CD4<sup>+</sup>CD25 or CD3<sup>+</sup>CD8<sup>+</sup>CD25 expression (data not shown) supporting our data on reduced immunogenicity of iDCs and tDCs. However, niDCs (**Figure 4C**, i) significantly stimulated both CD3<sup>+</sup>CD4<sup>+</sup> and CD3<sup>+</sup>CD8<sup>+</sup> T cell proliferation when compared to both unstimulated lymphocyte controls and iDCs. This indicates the importance of Sia in the maintenance of an iDCs phenotype. While ntDCs (**Figure 4C**, ii) show a trend increase to stimulate CD3+CD8+ T cells, there were no significant changes noted (**Figure 4C**). To eliminate the possibility of cell death as a potential cause of this increase in proliferation, we assessed cell death using Sytox Blue. We observed that iDCs have less cell death after neuraminidase treatment than tDCs (Figure S2 in Supplementary Material) enabling us to exclude this possibility. Finally, we investigated if niDCs and ntDCs can regulate the proliferation of stimulated T cells. LEW T cells were labeled with CTV, stimulated with CD3/CD28 labeled beads, and cocultured with niDCs and ntDCs (**Figure 5A**) and CD3<sup>+</sup>CD4<sup>+</sup> and CD3<sup>+</sup>CD8<sup>+</sup> proliferation was measured by flow

#### Figure 3 | Continued

Neuraminidase treatment of immature DCs (iDCs) and tolerogenic DCs (tDCs) decreases levels of α2-6-linked Sia significantly and alters phenotype. iDCs or tDCs were treated with neuraminidase for 90 min at 37°C to cleave Sia residues on the surface. (A) α2-6-linked Sia (SNA-I) (i) and α2-3-linked Sia (MAL-II) (ii) was measured on iDCs and niDCs after neuraminidase treatment (*n* = 3). α2-6-linked Sia (SNA-I) (iii) and α2-3-linked Sia (MAL-II) (iv) was measured on tDCs and ntDCs after neuraminidase treatment (*n* = 3). (B) Both iDCs (i) and tDCs (ii) were analyzed by flow cytometry for their expression of MHC I (FITC), MHC II (PE), CD 80 (PE), and CD 86 (PE) after neuraminidase treatments. (C) The mRNA expression of interleukin 6 (IL-6), interleukin 1 beta (IL-1β), inducible nitric oxide synthase (iNOS), tumor necrosis factor alpha (TNF-α), interleukin subunit beta (IL-12p40), and interleukin 10 (IL-10) was analyzed in iDCs, niDCs, tDCs, and ntDCs. Normalized to GAPDH and fold change relative to iDCs. Representative histograms and bar charts displaying relative fluorescence intensity (RFI) for flow cytometric analysis of DC cell surface. Median fluorescence intensities were established relative to iDCs in the case of niDCs and tDCs in the case of ntDCs. Error bars: mean ± SEM \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 one-way ANOVA, Tukey's multiple comparisons test. Data sets with two groups were analyzed using an unpaired *t*-test.

cytometry. Neuraminidase treatment completely abrogates the T cell inhibitory effect of iDCs leading to full restoration of T cell proliferation (**Figure 5A**, i). Interestingly, Dexa treatment is not sufficient to enable iDCs to inhibit the proliferation of activated T cells as no differences were observed between tDCs and ntDCs (**Figure 5B**, ii). In summary, these data indicate that the removal of Sia from iDCs increases the immunogenicity by its ability to stimulate CD4 and CD8 T cell proliferation, which can be prevented by Dexa treatment. In contrast, neuraminidase treatment completely restores the proliferation of polyclonally activated T cells, which cannot be prevented by Dexa treatment.

## DISCUSSION

Organ transplantation is often considered as the only therapeutic option for patients with life-threatening organ disease and is now performed on a routine basis. Due to incompatibilities between donor and recipient MHC-molecules, patients are required to take immunosuppressive drugs to prevent the destruction of the transplanted organ by the recipient's immune system. Immunosuppressive drug regimens are associated with severe side effects long term (34, 35). As a result, alternative immunosuppressive treatment strategies have been researched and developed including the use of therapeutic DCs in the treatment of autoimmune diseases and in the prevention of allograft rejection. DCs promote central and peripheral tolerance through various mechanisms, such as T cell anergy, inhibition of memory T cell responses, and clonal deletion amongst others (36). These characteristics form the basis of the use of DCs in the induction of tolerance. iDCs even have displayed the ability to convert naïve conventional T cells to regulatory T cells (Tregs) both *in vitro* (37, 38) and *in vivo* (39)*.* As shown here, and as shown by others, iDCs in non-inflammatory conditions display a poor immunogenic phenotype. One of the major barriers for use of iDCs in cellular therapies is that they respond to inflammatory stimuli, exemplified here by TLR4 (LPS) stimulation. In the context of autoimmunity and transplantation, iDCs are bound to encounter inflammatory environments if employed in therapeutic regiments. A potential solution to overcome this is the use of tDCs, which are maturation resistant.

Using tDC cellular therapies for the treatment of organ transplantation looks promising (18). tDCs are now routinely generated using different induction protocols, including the use of corticosteroids such as Dexa (11, 14, 15, 17, 40) and, in fact, we have recently shown in a rat model of corneal transplantation that Dexa generated tDCs significantly prolonged allograft survival without the need for additional immunosuppression (13). In this manuscript, we generate tDCs using Dexa and we characterize their maturation resistant phenotype by analyzing the expression of the immunogenicity markers MHCI, MHCII, CD80, and CD86 before and after TLR4 stimulation. We also analyze the expression of several immunomodulatory cytokine mRNAs. Dexa generated, maturation resistant, tDC have been well characterized by us (13, 32) and by other groups (17). However, to our knowledge, little is known on how Dexa induction of tDCs may affect the glycosylation profile of these cells and what functional consequences this may have. Glycosylation changes are not routinely assayed, but are likely to play crucial roles in iDC and tDC biology.

We describe here for the first time, using both lectin microarray and flow cytometry, that generation of tDCs by Dexa treatment leads to significant alterations in the cell surface glycosylation profile when compared to iDCs. We noted highly significant changes in lectin binding for α2-6-linked Sia (SNA-I) with no significant changes in lectin binding for α2-3-linked Sia (MAL-II). Interestingly, Jenner et al. (41) when comparing human iDCs with iDCs matured with a cytokine cocktail (IL-6, IL-1β, TNF-α, and prostaglandin E2) noted decreased α2-6-linked Sia with no changes in α2-3-linked Sia on the more immunogenic DC. This study also showed that Tregs have higher levels of α2-6-linked Sia when compared to activated conventional T cells. This suggests a possible link between α2-6-linked Sia content and tolerogenicity, where the increased α2-6-linked Sia may potentially serve as ligands for inhibitory sialic acid-binding proteins (Siglecs) on the surface of effector cells (41). In fact, hyper-sialylated antigens loaded onto DCs were recently shown to impose a regulatory program in the DCs. This resulted in the inducement of Tregs *via* Siglec-E and the inhibition of effector T cells (42).

Looking more closely at the lectin microarray analysis, other differences in lectin profiles observed here also hint at significant changes in the total abundance or potential branching alterations of underlying oligosaccharide structures, particularly N-acetyllactosamine (LacNAc), which may have occurred because of Dexa treatment. The relationship of responses among the 15 lectins (SNA-II, BPA, PNA, DSA, LEL, SNA-I, RCA-I, CPA, ECA, LTA, UEA-I, EEA, GS-I-B4, MPA, and VRA) which demonstrated the most significant differences between untreated iDCs and tDCs may hold further clues as to the nature of these variations in the glycocalyx, and it is possible that a portion of

#### Figure 4 | Continued

Neuraminidase treatment alters immunogenic properties of immature DCs (iDCs) and tolerogenic DCs (tDCs). To test the immunomodulatory properties of iDCs, niDCs, tDCs, and ntDCs, they were placed into MLRs for 5 days. (A) Schematic representation of experimental design. DA iDCs, niDCs, tDCs, and ntDCs were placed in cocultures for 5 days with allogeneic LEW lymphocytes isolated from the spleen and lymph nodes. (B) Representative gating strategy. Cells were selected according to size and granularity (i) followed by live/dead discrimination based on Sytox AADvancedTM negative cells (live) (ii). After single cell selection (iii) cells were selected by CD3 (PE) positivity (iv). Further selected by CD4 (APC) and CD8 (PE-CY7) and proliferation was measured by successive generations of CellTraceTM Violet positive cells. (C) The ability of iDCs, niDCs, tDCs, and ntDCs to stimulate allogeneic LEW T-cells was analyzed using unstimulated splenocytes/lymphocytes as a negative control (*n* = 3). (i) Representative histograms and bar charts displaying CD4+ and CD8+ T cell proliferation following a 5-day coculture with iDCs and niDCs. (ii) Representative histograms and bar charts displaying CD4+ and CD8+ T cell proliferation following a 5-day coculture with tDCs and ntDCs. Error bars: mean ± SEM \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 one-way ANOVA, Tukey's multiple comparisons test.

such variations exist among the membrane glycolipid structures as well as membrane proteins, which were analyzed here. With extracted glycoproteins, only one of the three lectins on the microarray, which has been reported to be indicative of Sia presence, SNA-I, demonstrated a significant intensity increase for tDCs. This was also demonstrated by lectin coupled flow cytometry showing how highly regulated Sia metabolism is in DCs. However, responses at lectins, which bind to structures, which are the most frequent attachment points for sialylation, those which bind to galactose (Gal) or *N*-acetylgalactosamine (GalNAc) (SNA-II, BPA, PNA), and those which bind to the associated disaccharide Type II LacNAc (RCA-I, CPA, ECA) or poly-LacNAc (LEL), are particularly interesting because the expected relationship of higher SNA-I binding and simultaneously lower Gal/GalNAc and LacNAc lectin binding did not consistently hold true across the lectin microarray profiles for DCs. The binding profiles and behavior of SNA-I and MAL-II in these experiments strongly infer quantitative differences between iDC and tDC surface Sia content; however, absolute quantitation will ultimately require chromatographic (e.g., HPLC) or chromatography-conjugated mass spectrometric analysis (LC-MS).

Because of the reported importance of Sias in DC pattern recognition (41, 43), endocytosis/phagocytosis (44–47), antigen presentation (48), migration (28, 49–52), and T cell interactions (28, 53). But also, considering that α2-6-linked Sia was the most significantly increased change after tDC generation by Dexa, we choose to investigate Sia's importance in iDC and tDC immunogenicity in an allogeneic setting, which would have potential implications in iDC and tDC cellular therapies.

For this, we removed Sia from the surface of the cells by enzymatic digestion using neuraminidase (sialidase). These experiments showed that Sia is involved in maintaining the tolerogenic phenotype of both iDCs and tDCs, as removal of Sia resulted in an increase in immunogenicity markers and increases in proinflammatory TH1 mRNA transcripts notably IL-6, IL-1β, iNOS (iDCs only), TNF-α, and IL-12p40 (iDCs only) with significant decreases in anti-inflammatory or tolerogenic IL-10. In experiments where neuraminidase-treated human monocyte derived DCs were cultured with ovalbumin (45) or *Escherichia coli* (44), there were reported increases in immunogenicity markers and cytokine gene expression also. Here, we show that even after Dexa treatment and tDC generation the removal of Sia from the cell surface results in increases in both cell surface immunogenicity markers and TH1 pro-inflammatory cytokine gene expression, underpinning the importance of Sia in a non-immunogenic phenotype.

In the context of allogeneic cell therapy for the treatment of autoimmune diseases and in the prevention of allograft rejection, it is important that the cell therapy itself does not elicit a deleterious immune response. In unstimulated allogeneic co-cultures using LEW responder lymphocytes, we show that iDCs and tDCs are non-immunogentic and do not elicit either CD3<sup>+</sup>CD4<sup>+</sup> nor CD3<sup>+</sup>CD8<sup>+</sup> proliferation. This attribute makes them ideal candidates in DC cellular therapies. We show that removal of Sia from iDCs is sufficient enough to stimulate the allogeneic responders, again showing the importance of Sia in a non-immunogenic phenotype. This may indicate that the removal of Sia uncaps underlying structures, which are then recognized as a signal for T-cell proliferation or that the Sias may act as ligands for inhibitory Siglecs on the surface of effector cells and once removed, this inhibitory effect is lost. Sia removal of tDCs did not induce CD3<sup>+</sup>CD4<sup>+</sup> proliferation, but we noted a trend increase in CD3<sup>+</sup>CD8<sup>+</sup> proliferation. Interestingly, this indicates that, despite the increase of immunogenicity markers and the transcript increase in several pro-inflammatory mRNAs, Dexa treatment of iDCs was sufficient to keep the cells, at least partially, in a non-immunogenic state.

In CD3/CD28 stimulated (hyper stimulated) allogeneic co-cultures using LEW responder lymphocytes, we show that iDCs had an impressive ability to supress stimulated allogeneic lymphocytes. Sia is critical in maintaining this suppressive ability as when it was absent we observed complete restoration of T cell proliferation for both CD3<sup>+</sup>CD4<sup>+</sup> and CD3<sup>+</sup>CD8<sup>+</sup> populations. These results are supported by the fact that Crespo et al. (45) showed increased T-lymphocyte proliferation in autologous mixed lymphocyte cultures using human monocyte-derived DCs where the lymphocytes were stimulated with tetanus toxoid, inactivated with mitomycin C, and cocultured with neuraminidase monocyte-derived DCs. Interestingly, we showed that tDCs do not have the ability to suppress hyperstimulated allogeneic lymphocytes to the same extent as iDCs. Sia removal had little effect on tDCs suppressive ability and did not exaggerate proliferation. Together, these experiments highlight that the tolerogenic properties between iDCs and tDCs are not inherently the same and understanding these characteristics and limitations will inform us on how to optimize therapy strategies.

The findings outlined here could also have numerous implications for our understanding of DC phenotype and function in the tumor microenvironment. Efficient induction of antitumor

Figure 5 | Neuraminidase treatment alters T-cell suppression properties of immature DCs (iDCs) and tolerogenic DCs (tDCs). To test the T-cell suppression properties of iDCs, niDCs, tDCs, and ntDCs, they were placed into stimulated MLR cultures for 4 days. Splenocytes/lymphocytes were stimulated with CD3/CD28 beads. (A) Schematic representation of experimental design. DA iDCs, niDCs, tDCs, and ntDCs were placed in cocultures for 4 days with CD3/CD28-stimulated allogeneic LEW lymphocytes isolated from the spleen and lymph nodes. (B) The ability of iDCs, niDCs, tDCs, and ntDCs to suppress CD3/CD28 stimulated allogeneic T-cells was analyzed using stimulated splenocytes/lymphocytes as a positive control. Error bars: mean ± SEM \**p* < 0.05, \*\**p* < 0.01, \*\*\**p* < 0.001, \*\*\*\**p* < 0.0001 one-way ANOVA, Tukey's multiple comparisons test. (i) Representative histograms and bar charts displaying stimulated CD4+ and CD8+ T cell proliferation following a 4-day coculture with iDCs and niDCs. (ii) Representative histograms and bar charts displaying stimulated CD4+ and CD8+ T cell proliferation following a 4-day coculture with tDCs and ntDCs.

responses requires that DCs in the tumor undergo proper maturation and activation (54). Understanding DC activation is important both in terms of their role in regulating immune responses locally in the tumor microenvironment (55), and also their use in *ex vivo* cellular and vaccination strategies to induce tumor specific immune responses.

In the context of tumor vaccination strategies using DCs, the required response is to induce tumor-specific effector T cells that can eliminate tumor cells specifically and that can induce immunological memory to control tumor relapse. Our findings suggest that Dexa, a common component of chemotherapy regimens, could suppress DC maturation and activation, their ability to present antigen (56), as well as their ability to induce T cell proliferation and activation. Interestingly, our data indicate that these potent Dexa-induced effects could be somewhat reversed in the presence of a neuraminidase, suggesting a key role for sialylation in Dexa generated tDCs. Removal of sialic acid has also previously been shown to increase tumor antigen-specific T cell responses (48). Our data also show that as well as a more potent ability to induce CD8+ T cell activation. In terms of modulating the tumor microenvironment directly, local delivery targeted approaches using sialyltransferase inhibitors delivered either to the tumor or the local lymph nodes could be exploited. In terms of *ex vivo* generated DCs for either cellular therapy or in vaccination strategies, treatment of DCs with sialyltransferase inhibitors could be sufficient to allow efficient priming of T cells systemically. As DCs provide an essential link between innate and adaptive immunity, these findings could have important implications in our understanding of the suppressive mechanisms within the tumor microenvironment that hinder adaptive antitumor immune responses and potential mechanisms by which they could be overcome.

Together, these results highlight the importance of Sia's in DC biology, especially in the context of iDC allogeneic cellular therapy. While the precise implications of increased or decreased Sia expression on iDCs and tDCs remain to be elucidated *in vivo*, we show here strong evidence that supports a function of Sia in the therapeutic aspects of DC cellular therapies. Identification of the molecular mechanisms and factors, which are regulated by Sia's are important to exploit this phenomenon in the clinic. This study points toward the potential of DC surface sialylation as a therapeutic target to improve and diversify DC-based therapies and treatments. In the context of disease, cell glyco-engineering could have positive implications in the treatment of autoimmunity, DC-based vaccines, the tumor microenvironment, and transplant biology.

### ETHICS STATEMENT

All animals used in experiments were accommodated in an accredited animal housing facility under a license granted by the Department of Health, Ireland, and were approved by the Animal Ethics Committee of the National University of Ireland, Galway.

### REFERENCES


### AUTHOR CONTRIBUTIONS

Substantial contributions to the conception or design of the work (KL, TR, LJ); or the acquisition (KL, OT, HA, PL), analysis (KL, OT, HA, JG, JC), or interpretation of data for the work (KL, JG, AR, TR); and drafting the work or revising it critically for important intellectual content (KL, JG, TR, AR); and final approval of the version to be published (KL, OT, JG, HA, PL, JC, LJ, AR, TR); and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved (KL, OT, JG, HA, PL, JC, LJ, AR, TR).

### ACKNOWLEDGMENTS

National University of Ireland flow cytometry core facilitated by Dr. Shirley Hanley. Functional genomics core facilitated by Dr. Enda O'Connell.

### FUNDING

This work is supported by Science Foundation Ireland (12/TIDA/ B2370 and 12/IA/1624) and European Cooperation in Science and Technology (COST) for the AFACTT project (Action to Focus and Accelerate Cell-based Tolerance-inducing Therapies; BM1305).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01427/ full#supplementary-material.

Table S1 | Table listing forward primers, reverse primers, and probes for interleukin 6 (IL-6), indoleamine 2,3-dioxygenase (IDO), interleukin 1 beta (IL-1β), inducible nitric oxide synthase (iNOS), and IL-12p40, tumor necrosis factor alpha (TNF-α), GAPDH and interleukin 10 (IL-10).

Table S2 | Lectin names and common major binding ligands. Table listing the abbreviations, the kingdom, species, common name, and major binding ligand of the lectins used in lectin micro array profiling.

Figure S1 | Isolation and generation of immature DCs and tolerogenic DCs.

Figure S2 | Resultant cell death due to neuraminidase treatment of immature DCs (iDCs) and tDCs. Both iDCs and tDCs were treated with neuraminidase for 90 min. These cells were then washed and placed into culture for 48 h. Cells were prepared for flow cytometry as previously described and stained with the live/dead indicator sytox blue (n = 1, technical replicate of 2).


doi:10.1002/1521-4141(200007)30:7<1807::AID-IMMU1807>3.0. CO;2-N


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling editor declared a past co-authorship with one of the authors TR.

*Copyright © 2017 Lynch, Treacy, Gerlach, Annuk, Lohan, Cabral, Joshi, Ryan and Ritter. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Dexamethasone and Monophosphoryl lipid a induce a Distinctive Profile on Monocyte-Derived Dendritic cells through Transcriptional Modulation of genes associated With essential Processes of the immune response

*Paulina A. García-González 1,2†, Katina Schinnerling1,2†, Alejandro Sepúlveda-Gutiérrez <sup>3</sup> , Jaxaira Maggi 1,2, Ahmed M. Mehdi4 , Hendrik J. Nel <sup>4</sup> , Bárbara Pesce1 , Milton L. Larrondo5 , Octavio Aravena1 , María C. Molina1 , Diego Catalán1,2, Ranjeny Thomas <sup>4</sup> , Ricardo A. Verdugo3 \* and Juan C. Aguillón1,2\**

*1Programa Disciplinario de Inmunología, Facultad de Medicina, Instituto de Ciencias Biomédicas (ICBM), Universidad de Chile, Santiago, Chile, 2Millennium Institute on Immunology and Immunotherapy, Santiago, Chile, 3Programa de Genética Humana, Facultad de Medicina, Instituto de Ciencias Biomédicas (ICBM), Universidad de Chile, Santiago, Chile, <sup>4</sup> Translational Research Institute, University of Queensland Diamantina Institute, Woolloongabba, QLD, Australia, 5Banco de Sangre, Hospital Clínico de la Universidad de Chile, Santiago, Chile*

There is growing interest in the use of tolerogenic dendritic cells (tolDCs) as a potential target for immunotherapy. However, the molecular bases that drive the differentiation of monocyte-derived DCs (moDCs) toward a tolerogenic state are still poorly understood. Here, we studied the transcriptional profile of moDCs from healthy subjects, modulated with dexamethasone (Dex) and activated with monophosphoryl lipid A (MPLA), referred to as Dex-modulated and MPLA-activated DCs (DM-DCs), as an approach to identify molecular regulators and pathways associated with the induction of tolerogenic properties in tolDCs. We found that DM-DCs exhibit a distinctive transcriptional profile compared to untreated (DCs) and MPLA-matured DCs. Differentially expressed genes downregulated by DM included MMP12, CD1c, IL-1B, and FCER1A involved in DC maturation/ inflammation and genes upregulated by DM included JAG1, MERTK, IL-10, and IDO1 involved in tolerance. Genes related to chemotactic responses, cell-to-cell signaling and interaction, fatty acid oxidation, metal homeostasis, and free radical scavenging were strongly enriched, predicting the activation of alternative metabolic processes than those driven by counterpart DCs. Furthermore, we identified a set of genes that were regulated exclusively by the combined action of Dex and MPLA, which are mainly involved in the control of zinc homeostasis and reactive oxygen species production. These data further support the important role of metabolic processes on the control of the DC-driven regulatory immune response. Thus, Dex and MPLA treatments modify gene expression in moDCs by inducing a particular transcriptional profile characterized by the activation of tolerance-associated genes and suppression of the expression of inflammatory genes, conferring the potential to exert regulatory functions and immune response modulation.

Keywords: tolerogenic dendritic cells, immune regulation, dexamethasone, transcriptome, tolerance induction

#### *Edited by:*

*Piotr Trzonkowski, Gdan´ sk Medical University, Poland*

#### *Reviewed by:*

*Lenka Palová Jelínková, Second Faculty of Medicine, Charles University, Czechia Nick Giannoukakis, Allegheny Health Network, United States*

#### *\*Correspondence:*

*Ricardo A. Verdugo raverdugo@u.uchile.cl; Juan C. Aguillón jaguillo@med.uchile.cl*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 06 September 2017 Accepted: 03 October 2017 Published: 23 October 2017*

#### *Citation:*

*García-González PA, Schinnerling K, Sepúlveda-Gutiérrez A, Maggi J, Mehdi AM, Nel HJ, Pesce B, Larrondo ML, Aravena O, Molina MC, Catalán D, Thomas R, Verdugo RA and Aguillón JC (2017) Dexamethasone and Monophosphoryl Lipid A Induce a Distinctive Profile on Monocyte-Derived Dendritic Cells through Transcriptional Modulation of Genes Associated With Essential Processes of the Immune Response. Front. Immunol. 8:1350. doi: 10.3389/fimmu.2017.01350*

## INTRODUCTION

Dendritic cells (DCs) are a heterogeneous group of specialized antigen-presenting cells with the capacity to orchestrate specific immune responses according to the antigen they encounter and the environmental signals derived from the local milieu (1). After antigen capture and processing, DCs undergo a complex process of differentiation to mature DCs, which express high levels of surface peptide–HLA complexes and co-stimulatory molecules. Mature DCs also produce inflammatory cytokines and T-cellattracting chemokines enabling the induction of Th1, Th2 or Th17 responses (2, 3). DCs can also differentiate into tolerogenic DCs (tolDCs), capable of inducing anergy or deletion of effector T cells, and/or differentiation and proliferation of regulatory T-cell (Treg) subsets. Regulation may result from various processes, including deficient antigen presentation, reduced co-stimulatory molecules, expression of inhibitory molecules, and/or secretion of anti-inflammatory cytokines such as IL-10 and TGF-β (4, 5). The ability of DCs to modulate T-cell responses has made them an interesting target of study for the immunotherapy of autoimmune diseases, since these cells are supposed to induce and maintain immune tolerance to harmless or self-antigens (6, 7).

Differentiation of DCs from peripheral blood monocytes using GM-CSF and IL-4 is a useful approach to obtain large numbers of DCs *in vitro* to study their function and biology. This approach is also used to generate tolDCs *in vitro* by adding immune modulators such as immunosuppressant drugs [dexamethasone (Dex), rapamycin, aspirine, rosiglitazone, tacrolimus] (8–12); anti-inflammatory cytokines (IL-10 and TGF-β) (13–15); natural compounds (vitamin D3, retinoic acid, and curcumin) (8, 16, 17); the JAK inhibitor tofacinib (18); and the NF-kB inhibitor BAY11- 7082 (19). All strategies lead to DCs with regulatory capacities, although some features may vary between protocols. Our group has described a protocol for tolDC generation from peripheral blood monocytes of healthy controls (20) and rheumatoid arthritis (RA) patients (21) using Dex to induce a tolerogenic phenotype and subsequent activation with the non-toxic lipopolysaccharide (LPS) analog monophosphoryl lipid A (MPLA) to confer lymph node homing capacity and stability. These cells, herein termed Dex-modulated and MPLA-activated DCs (DM-DCs), expressed low levels of CD83, CD86, and CD40, secreted high levels of IL-10 and TGF-β and low levels of IL-12, and stimulated T-cell proliferation and cytokine production at low levels in allogeneic and autologous cultures (20, 22).

While we generally understand the cellular mechanisms by which tolDCs modulate T-cell responses and induce tolerance, the molecular switches determining tolDC differentiation and function are still poorly known. The knowledge of molecular regulators and pathways could be of great benefit for searching targets for effective cellular therapies. However, only few studies have attempted to identify specific molecules or processes involved in tolerogenic functions of monocyte-derived DCs (moDCs) using whole-genome transcriptomic or proteomic analyses (23, 24). Most studies focus on vitamin D3-modulated moDCs (25, 26). Studies of Dex-treated moDCs comprised only proteomic approaches and focused on the identification of potential tolDC markers (27, 28).

We recently compared tolDCs derived from monocytes of healthy controls and RA patients at phenotypic, functional, and transcriptional levels (21) and demonstrated that Dex and MPLA treatments removed disease-associated features of moDCs to yield a uniform signature. Here, we describe a genome-wide differential expression study of tolDCs derived from monocytes of healthy controls in which we elucidate molecular processes that drive DC differentiation toward a tolerogenic profile in response to Dex and MPLA treatments. We found that DM-DCs exhibit a transcriptional profile that distinguishes them from other moDC subsets, characterized by the upregulation of several genes related to immunoregulatory functions and biological processes that could be involved in tolerance induction.

### MATERIALS AND METHODS

### Blood Samples

A total of 10 buffy coat samples from healthy controls were used for microarray analysis and phenotypic and functional studies. An additional 10 buffy coat samples were used to confirm differential expression of genes by qRT-PCR and flow cytometry. All subjects signed an informed written consent and all procedures were approved by the Ethics Committees for Research in Human Beings from the Faculty of Medicine and from the Clinical Hospital of the University of Chile. Demographic characterization of healthy controls is detailed in Table S1 in Supplementary Material.

### Generation of moDC Subsets

Human moDCs were generated from monocytes as previously described (20). Monocytes were isolated from peripheral blood of 10 healthy individuals by negative selection using RosetteSep Human Monocytes enrichment cocktail (Stemcell Technologies, Vancouver, BC, Canada) according to manufacturer's instructions. Monocytes were cultured at 2 × 106 cells/ml in serum-free AIM-V medium (Gibco BLR, Grand Island, NY, USA), supplemented with 500 U/ml of recombinant human GM-CSF and IL-4 (eBioscience, San Diego, CA, USA) for 5 days at 37°C and 5% CO2. At day 3, culture medium was replenished and cells were incubated with Dex (Sigma-Aldrich, St. Louis, CO, USA) at a final concentration of 1 µM [Dex-modulated DCs (D-DCs)]. At day 4, cells were stimulated with 1 µg/ml of cGMP-grade MPLA (Avanti Polar Lipids Inc., Alabaster, AL, USA) (DM-DCs). Unstimulated cells (DCs) and MPLA-matured DCs (M-DCs) generated in the absence of Dex were used as controls of immature and mature DCs, respectively. On day 5, cells were harvested and characterized by flow cytometry.

### Flow Cytometry

Antibodies used for analysis were anti-human CD80 FITC (clone 2D10.4), CD83 FITC (clone HB15e), CD40 PE (clone 5C3), CD86 PE (clone IT2.2), IDO1 PECy7 (clone eyedio), CD4 PECy7 (clone OKT4), IFN-γ APC (clone 4S.B3) (eBioscience); CD11c BUV395 (clone B-ly6), CD83 BUV737 (clone HB15e) (BD Biosciences); CD86 BV650 (IT2.2), CD163 BV605 (clone GHI/61), CD1c BV510 (clone L161), MERTK BV421 (clone 590H11G1E3), CD32 APC (clone FUN-2), ZBTB16/PLZF PE (clone Mags.21F7) (BioLegend) and TLR2 Alexa Fluor 700 (clone 383936), JAG1 Fluorescein (clone 188331), and FPR2 APC (clone 304405) (R&D Systems). Prior to antibody staining, cells were labeled with fixable viability dye eFluor 780 (eBioscience). Cells were resuspended in PBS supplemented with 10% of fetal bovine serum (FBS) (HyClone Thermo Scientific, Logan, UT, USA), stained with specific antibodies, fixed with IC fixation buffer (eBioscience), and resuspended in FACSFlow buffer (Becton Dickinson, San Diego, CA, USA) for subsequent analysis. Data were acquired on a FACSAria III with FACSDiva v6.1.3 software (both Becton Dickinson) and analyzed by FlowJo software (Treestar, USA).

### Cytokine Production

A total of 1 × 105 DCs were incubated for 24 hours with or without CD40L-transfected irradiated NIH3T3 cells at 1:1 ratio in AIM-V medium, in 96-well U bottom plates (BRAND, Wertheim, Germany). Supernatants of cocultures with NIH3T3 cells or T cells were recovered and stored at −80°C until quantification of IL-10, IL-12p70, and IFN-γ by ELISA (eBioscience).

### CD4**+** T Cell-Stimulatory Capacity of DCs

CD4<sup>+</sup> T cells were isolated by negative selection using RosetteSep Human T-cell enrichment cocktail (Stemcell Technologies) and labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE).

For the assessment of antigen-specific CD4<sup>+</sup> T-cell activation, DCs were loaded with 1 µg/ml tuberculin purified protein derivative (Staten Serum Institute, Copenhagen, Denmark) 4 hours prior to activation with MPLA and co-cultured with autologous CD4<sup>+</sup> T cells at a DC:T-cell ratio of 1:2 in RPMI medium (HyClone Thermo Scientific) with 10% FBS in 96-well U bottom plates for 6 days (20). CD4<sup>+</sup> T cells alone and stimulated with anti-human CD3 mAb (clone OKT3; 0.65 μg) (eBioscience) were used as negative and positive controls, respectively. Supernatants were collected to assess cytokine secretion. For intracellular IFN-γ detection, 50 ng/ml phorbol-12-myriastate-13-acetate (Sigma-Aldrich), 1 µg/ml ionomycin (Sigma-Aldrich), and 1 µg/ml brefeldin-A (eBioscience) were added for the last 5 hours of culture. Proliferation and IFN-γ production of CD4<sup>+</sup> T cells were analyzed by flow cytometry.

### RNA Isolation and Microarray Analysis

RNA was isolated from 5 × 105 DCs on day 5 using total RNA isolation RNeasy Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. Yield and quality of RNA samples were evaluated with NanoDrop 1000 spectrophotometer (Thermo Scientific, Waltham, MA, USA) and RNA integrity (RIN score) was analyzed with Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) or LabChip GX/ GX II (Caliper LifeSciences, Hopkinton, MA, USA). A total of 40 samples, corresponding to 10 healthy donors under four experimental conditions were considered for microarray analysis (Figure S1 in Supplementary Material). All RNA samples used for microarrays showed A260/A280 values between 1.8 and 2.2, and RIN scores >7. RNA samples were reverse transcribed, amplified, and labeled using an Illumina® TotalPrep™ RNA Amplification Kit, and cDNA was hybridized onto Illumina Human HT-12 v4 BeadChips (Illumina, San Diego, CA, USA), covering the whole human genome. Expression data were extracted with GenomeStudio Project Software from Illumina.

### Confirmation of Gene Expression by qRT-PCR

cDNA was prepared from moDCs RNA samples using the Superscript II Reverse Transcriptase kit (Thermo Fisher Scientific). Quantitative RT-PCR was performed in Stratagene Mx300P, using Brilliant II SYBR Green QPCR Master Mix (Agilent Genomics) with primer sets from IDT. The housekeeping genes *GAPDH* and *r18S* were used as internal controls and target gene expression was normalized to untreated DCs. Primer sequences for each target gene are described in Table S2 in Supplementary Material.

### Data Exploration and Statistical Analyses

For flow cytometry and qPCR data, Friedman repeated measures test and Dunn's *post hoc* test were used for data comparison between moDC culture conditions. Analyses were performed using Prism 5.01 software (Graphpad, San Diego, CA, USA).

Microarray data were log transformed followed by quantile normalization using the preprocess Core package v1.28.0 from Bioconductor. Differentially expressed (DE) genes in modulated DCs relative to unstimulated DCs were identified with the Maanova package v1.36.0 *t*-test for gene pairwise comparisons (29), and *p*-values were adjusted using false discovery rate (FDR) method. Genes with adjusted *p*-value ≤0.05 were considered differentially expressed and reported. K-means clustering of residual values of DE genes between DM-DCs and DCs was performed using the cluster package and a *K* value of 6 to maximize cluster distance and minimize distance between clustered genes.

Overrepresentation of pathways and biological functions was assessed using Ingenuity Pathway Analysis (Ingenuity Systems, Qiagen, Hilden, Germany).

## RESULTS

### Modulation of moDCs with Dex and MPLA Induces a Distinctive Transcriptional Profile

First, we confirmed that treatment with Dex and MPLA during moDC differentiation induces a tolerogenic phenotype on these cells. As previously described (20), DM-DCs expressed low levels of CD86, CD80, CD40, and CD83 and produced low levels of IL-12 and high levels of IL-10 relative to M-DCs. T-cell proliferation and IFN-γ production in response to antigen-exposed DM-DCs were significantly reduced when compared with T cells stimulated with DCs or M-DCs (Figure S2 in Supplementary Material). Subsequently, we used RNA from the same moDC preparations to conduct whole-genome analysis through microarray technology, using a total of 40 moDC samples, corresponding to four different DC subsets, i.e., unstimulated (DC), M-DCs, D-DCs, and Dexmodulated/MPLA-activated DCs (DM-DC), differentiated from monocytes from 10 healthy individuals. We defined statistically significant differences between samples by a FDR value of 0.05 or lower. Since all four differentiation protocols led to moDCs with different phenotypic characteristics, we tested whether these differences could also be found at transcriptional level. A principal component analysis was used to explore the data by projecting samples onto the major orthogonal components or genes expression. The first two dimensions separated moDCs samples by their differentiation state (unstimulated, MPLA, Dex, or Dex plus MPLA) (**Figure 1**). The first component (*X*-axis) naturally clustered samples as M-DCs < DCs < (D-DCs:DM-DCs). D-DCs and DM-DCs, which together could be associated with a tolerogenic potential, could not be distinguished on component 1. The second component (*Y*-axis) clustered samples by their grade of activation, with both moDC subtypes treated with MPLA projecting toward positive values, i.e., DCs < D-DCs < (M-DCs:DM-DCs). Thus, each protocol induces a unique transcriptional profile in moDCs that distinguishes them from other moDC subtypes.

### Dex and MPLA Treatment Modulates Genes Associated With Cell Movement, Signaling, and Metabolism

Given their tolerogenic phenotype, we next focused on the transcriptional effects of combined treatment with Dex and MPLA treatment on moDCs. Considering *p*-values (corrected using FDR) 0.05 or lower to define statistically significant differences in gene expression, we identified 259 DE transcripts in DM-DCs compared to the unstimulated control (DCs). The scavenger receptor CD163, several MT1 (metallothionein 1) isoforms and MT2A, C1QTNF1, ADORA3, S100A9, and both isoforms of Fc receptor for IgG FCGR2A/CD32 and FCGR2B/CD32B (**Table 1**; Table S3 in Supplementary Material) were among the genes most upregulated by Dex and MPLA treatments; while CD1b, FCER1A (high affinity I Fc fragment of IgE receptor subunit alpha polypeptide), and MMP12 (Matrix metallopeptidase 12) were the most downregulated genes (**Table 1**; Table S3 in Supplementary Material).

A pathway enrichment analysis of the DE genes found on DM-DCs done with ingenuity pathway analysis (IPA) identified cell movement, cell signaling, and metabolism as main functions modulated by Dex and MPLA (**Figure 2A**; Table S4 in Supplementary Material), assembling regulated genes into 11 networks representing significantly enriched biological functions, mostly associated with cellular movement, growth, immune cell trafficking, and metabolism (**Figure 2A**; Table S4 in Supplementary Material). The main regulated genes in DM-DCs such as CD163, ADORA3, FCGR2A, CD1c, CD1b, and MMP12 belong to these networks. In terms of canonical pathways and consistent with functional enrichment analysis, Dex and MPLA treatments affected cell adhesion and diapedesis (*p*-value 8.52e−3) (**Figure 2B**), modulating the expression of 10 genes, including the upregulation of chemokines such as CCL8, CCL18, CCL23, and CXCL5. Interferon signaling (*p*-value 7.49e−3) was predicted to be an active process, with a positive *z*-score (2.449) and high expression levels of the IFN-inducible genes IFI6, IFITM1, IFITM2, and IFITM3 (Table S4 in Supplementary Material). Processes associated with DC maturation and activation were inhibited, and several molecules associated with these processes such as CD80, CD83, CD1c, ACTA2, ACTG1, TMBS10, AP-1, and RAP1GAP were downregulated (Table S4 in Supplementary Material). Complement system pathway was highly modulated, with increased expression of C1QA, C1QB, C1QC, and CFB. IL-10 signaling (*p*-value 2.93e−3), along with TLR (*p*-value 4.11e−2), iNOS (*p*-value 4.65e−2), and p38MAPK signaling (*p*-value 1.31e−2), which also signal through IRAK and STAT1, was enriched processes on DM-DCs according to



IPA knowledge database (**Figure 2B**; Table S4 in Supplementary Material). Upstream regulator analysis revealed GILZ, STAT1, FOXO3, STAT3, SMARCA4 and CEBPD to be amongst the main transcriptional regulators of many genes from this dataset (Table S4 in Supplementary Material).

### The Transcriptional Program of DM-DCs Is Regulated by an Interplay between Dex and MPLA

The 259 DE genes in DM-DCs *versus* untreated DCs were further grouped into six clusters according to their expression pattern using K-means clustering of residual values (clusters C1–C6; **Figure 3**). Clusters 1 and 3 contain 61 DE genes (Cluster 1 = 20 genes; Cluster 3 = 41 genes) that correspond mainly to DC differentiation and maturation, including the DC markers CD1c, DC-SCRIPT/ZNF366, co-stimulatory molecule CD80, and other molecules involved in DC maturation and inflammation such as CD83, RAP1GAP, NDRG2, CD1b, FCER1A, CCL22, and MMP12 (**Figure 3**; Table S3 in Supplementary Material). Dex treatment alone or combined with MPLA leads to downregulation of these genes (**Figure 3**). Clusters 2 and 5 comprise 79 genes upregulated by MPLA (Cluster 2 = 44 genes; Cluster 5 = 35 genes), regardless of the presence of Dex. These clusters contain most genes related to IFN signaling and granulocyte and agranulocyte adhesion and diapedesis, in addition to IL-1B, STAT1, and IDO1. Cluster 4 contains 75 genes, which are upregulated by Dex irrespective of MPLA addition. This cluster includes many genes associated to inhibition of DC activation and maturation such as FCGR2B, C1Q, and MAFB, as well as genes related to anti-inflammatory responses of DCs leading to IL-10 production (TSC22D3/GILZ and MAP3K8/TPL-2) and modulation of T-cell activation and/ or expansion of Treg-cell populations such as JAG1 (Jagged 1), MERTK (receptor tyrosine kinase Mer), TBXAS1, and SEMA4A, thus contributing to the tolerogenic profile of DM-DCs (**Figure 3**; Table S3 in Supplementary Material). Additionally, Cluster 4 is characterized by the expression of membrane receptors (MERTK, FCGR2A, FCGR2B, and SEMA4A), signaling proteins (MAP3K8, S100A9, and JAG1) and transcriptional regulators (TSC22D3/GILZ) as well as molecules related to the complement system (C1QA, C1QB, C1QC, and CFH) and cell adhesion and migration (CCL13, CCL18, CCL23, SH3PXD2B, and ADORA3) (**Figure 3**; Table S3 in Supplementary Material). IL-10 signaling and complement system are among the pathways overrepresented in this group.

Finally, Cluster 6 contains 44 genes, which were regulated in response to a synergistic effect between Dex and MPLA. Metallothioneins, proteins involved in stress response, heavy metal scavenging, and immunosuppression were highly represented in this group, including several MT1 isoforms (MT1A, MTE, MT1F, MT1G, MT1H, MT1M, and MT1X) and MT2A. Enrichment of several molecules involved in cell adhesion and diapedesis (CXCL5 and MMP19), T-cell migration (THBS1 and TNFRSF6B), and production of reactive oxygen species (ROS) (FPR1, FPR2, NCF1, and SLAMF1) was also found in this cluster as well as molecules associated with a low inflammatory state (C1QTNF1, SLC39A8/ZIP8 and IRAK3), suggesting that genes in this cluster also contribute to the regulatory features of DM-DCs in addition to the genes of Cluster 4. Functional and pathways analysis of genes from this cluster showed that cellular movement and metabolic processes, particularly ROS metabolism is highly represented (Table S5 in Supplementary Material).

### Dex and MPLA Treatment Promotes the Upregulation of Genes Related to the Modulation of Biological Processes That Control Immune Responses

Biological functions related to cellular movement, growth, and proliferation, cell-to-cell signaling, and free radical scavenging were found to be upregulated. Canonical pathway analysis also provided agranulocyte/granulocyte adhesion and diapedesis, T helper differentiation, and DC maturation as main pathways regulated in our dataset (**Figure 2**; Tables S4 and S5 in Supplementary Material). Since these functions were highly enriched in the two clusters that are potentially involved in DM-DCs tolerogenic features (Clusters 4 and 6), we further analyzed interactions between genes contained in these clusters. Functional annotations involving activation and proliferation of T lymphocytes were highly represented in several networks and predicted to be inhibited in DM-DC, while chemotaxis of T lymphocytes and synthesis of ROS, also highly represented, were predicted to be activated in these cells (**Figure 4**). Most downregulated genes in DM-DCs, which are associated with T-cell activation and proliferation, are also involved in DC maturation and activation, while upregulated genes interacting in these networks encode inhibitory membrane receptors or transcriptional regulators that lead to inhibition of effector T-cell activation and

promote differentiation of Tregs (FCGR2B, MT1, IDO1, CCL18, JAG1, and MERTK). This is further supported by upregulation of genes involved in the activation of ROS production (FPR1 and NCF1 and NAMPT) and recruitment of naive and Tregs (CCL18, CCL23, and SAA1). Many of the genes that were upregulated in response to Dex and MPLA together (Cluster 6) also participate in the main metabolic processes in DM-DCs, mainly ROS production and zinc homeostasis (Table S5 in Supplementary Material), contributing to the particular metabolic profile of these tolDCs.

The differential expression of several genes of interest that are modulated on DM-DCs such as CD163, JAG1, IDO1, MERTK, MT1F, FPR1, and CD32, which might participate in the main processes enriched on DM-DCs, was confirmed through realtime PCR (**Figure 5**) and were shown to correlate with protein levels determined by flow cytometry (**Figure 6**).

### DISCUSSION

Here, we demonstrate, by whole-genome transcriptomic analysis of different moDC subtypes, that DM-DCs exhibit a distinctive transcriptional program that potentially endows them with regulatory functions through modulation of chemotactic responses, cell-to-cell signaling and interaction, and metabolic processes.

Glucocorticoids, including Dex, have been widely used for tolDC generation and have been demonstrated to inhibit DC maturation and inflammation (30, 31). Since activation of tolDCs has been shown to increase lymph node homing, antigen presentation, and stability against other inflammatory modulators (32), we activated Dex-treated DCs with MPLA (DM-DCs), a non-toxic clinical grade analog of LPS, which also signals via toll-like receptor 4 and exhibits potent immune-stimulatory capacity (33, 34). Glucocorticoid receptors are transcription factors that suppress the pro-inflammatory program induced by TLRs and in turn potentiate TLR-mediated anti-inflammatory responses such as IL-10 secretion (35, 36). Our group has previously shown that Dex and MPLA induce a tolerogenic profile in moDCs from both healthy controls and RA patients (21, 22). The resulting DM-DCs exhibit characteristic tolDCs features, with low expression of co-stimulatory and maturation markers, high production of anti-inflammatory cytokines, and reduced

capability to promote effector T-cell responses. Consistent with their phenotypic properties, we found that modulation of moDCs with either stimulus alone or a combination of both also induces a distinctive transcriptional profile, which allows separation of moDCs samples according to differentially expressed genes.

Several methallothioneins (MTs) (MT1G, MT1H, MT1E, MT2A, MT1M, MT1A, and MT1X) appear in this study as some of the most upregulated genes within the 259 DE transcripts of DM-DCs. These are small cysteine-rich metal-binding proteins involved in the regulation of homeostasis of zinc and other heavy metals at cytoplasmic level (37). MTs can also affect different cellular processes such as gene expression, apoptosis, proliferation, and differentiation (38) and have been described to suppress collagen-induced arthritis via induction of TGF-β and reduction of pro-inflammatory modulators such as TNF (39) and to promote the expansion of Treg (40, 41). Besides MTs, another molecule involved in zinc transport, the zinc importer SLC39A8/ ZIP8, is also highly upregulated in DM-DCs. Zinc is known to act as a modulator of immune responses through its availability, and zinc deficiency affects the immune system leading to increased inflammation and inflammatory diseases such as RA (38). In DCs, zinc supplementation has been shown to interfere with maturation, by inhibiting the upregulation of MHCII and co-stimulatory molecules (42), as well as to induce the expression of the tolerogenic markers PD-L1, IDO1, and CD103 (43). Thus, a tight regulation of zinc concentration is required and zinc may contribute to the immunoregulatory functions of DM-DCs, since several regulators of intracellular zinc concentration are overexpressed in these cells.

Cluster analysis of DE genes in DM-DCs (with respect to DCs) revealed six different expression patterns, highlighting genes associated with a single modulatory agent. Genes induced by MPLA were enriched in molecules involved in IFN signaling as well as DC differentiation and maturation-related genes, which were in turn downregulated by Dex treatment, confirming previous reports that Dex impairs DC maturation and induces an immature-like DC phenotype (31, 44). MPLA stimulation also induced the expression of the regulatory molecule IDO1, which was similar in DM-DCs and M-DCs. This finding was to be expected since IDO upregulation has already been described in mature DCs, in particular in TLR-stimulated DCs (45–47). It does, however, differ from the work of Danova et al. (48), whom reported a weak IDO expression in MPLA-treated DCs. Differences between both studies may be explained by differences in the generation protocols as well as the detection techniques used. Dex also induced several molecules associated with regulation of immune responses involving Treg and effector T-cell functions (JAG1, TBXAS1, and MERTK) (49, 50), DC differentiation and

(B) Network interaction analysis of genes from Cluster 6 (see Figure 3), modulated synergically by dexamethasone and MPLA.

function (IRAK3, GILZ, C1Q, and STAB1) (28, 51, 52) and suppression of inflammatory signaling (MAP3K8/TPL-2, FCGR2B, and VDR). Functional enrichment analysis showed that Dex treatment of moDCs also induced the expression of genes related to the complement system pathway. Of great interest among these genes is C1Q, previously described to be upregulated in tolDCs and proposed as potential marker of tolerogenicity (28). C1Q has been demonstrated to be a potent modulator of DCs, which suppresses DC differentiation and activation through engagement of the inhibitory receptor leukocyte-associated Ig-like receptor 1, limiting the activation of immune responses (53, 54). Moreover,

C1Q was shown to inhibit T-cell activation and pro-inflammatory cytokine production, while enhancing IL-10 secretion (55).

Additionally, we identified a group of genes that were induced by a synergistic effect of Dex and MPLA. These genes include anti-inflammatory mediators (SLC39A8/ZIP8, CCL18, and C1QTNF1/CTRP1) and molecules involved in the regulation of T-cell function (MT1, THBS1/TSP-1, and TNFRSF6B/DcR3). Interestingly, several DE transcripts within this cluster are associated with production of reactive oxygen and nitrogen species (FPR1, FPR2, NCF1, and SLAMF1), and accordingly, IPA analysis of our dataset predicted the activation of this biological function.

Furthermore, metabolic changes seemed to play an important role in DM-DCs, since processes associated with free radicals, in particular ROS production and fatty acid metabolism, were enriched in the DM-DCs dataset. Despite usually being considered pro-inflammatory, ROS participate in many physiological processes. Excessive ROS drives inflammation and oxidative damage, while low ROS amounts were shown to suppress immune responses (56, 57). Effector T cells exhibit impaired proliferation and increased apoptosis in response to sustained pro-oxidant conditions, whereas Treg is less sensitive to this effect and retain their suppressive function (58). Correspondingly, ROS production is one of the strategies used by Treg to suppress effector T cells (58, 59). Macrophages have been demonstrated to suppress T-cell responses by producing ROS and induce Treg in a ROS-dependent manner (60). Dex has been previously shown to increase ROS production in macrophages and moDCs (61). Here, we show that Dex treatment alone induces NCF1 and PDK4, which are both involved in ROS production and we demonstrate that MPLA activation after Dex-mediated modulation of moDCs leads to the upregulation of several genes involved in ROS metabolism and production. It has been previously described that modulation of moDCs is accompanied by changes in cellular metabolism and that tolDCs show a different metabolic profile than pro-inflammatory DC subsets (26). This catabolic and highly energetic metabolic profile of tolDCs may be due to higher energy demands required for suppressive functions (62). Therefore, in DM-DCs, regulation of ROS production and zinc homeostasis could be crucial to the regulatory function of DM-DCs.

Another hallmark of DM-DCs is the regulation of chemokine expression. In particular, the upregulation of Treg and naive T-cell attractants (CCL17, CCL18, CCL23/MIP-3, and CXCL9) and chemokines associated with the recruitment of monocytes and granulocytes (CCL13/MCP-4, CCL26, and CXCL5) as well as the downregulation of expression of chemokines attracting effector T cells might account for the potential to recruit Treg subsets to sites of inflammation to promote tolerance.

Of note, this is the first work investigating the molecular basis of tolerogenic features of moDCs modulated with Dex and alternatively activated with MPLA. We have demonstrated that besides inhibiting DC maturation and inflammation, Dex and MPLAs treatment jointly induce a distinctive transcriptional profile in moDCs mainly regulating pathways involving cellular chemotactic responses, cell-to-cell signaling and interaction, as well as zinc and ROS metabolism, favoring the recruitment and proliferation of Treg while inhibiting effector T-cell responses.

### REFERENCES


Our results indicate that there is a broad spectrum of immunoregulatory properties of tolDCs beyond the already described mechanisms depending on direct DC–T-cell contact and antiinflammatory cytokine secretion and thus provides novel targets for immunotherapeutic strategies based on tolDCs.

### ETHICS STATEMENT

All subjects gave written consent according to the Declaration of Helsinki, and all procedures were approved by the Ethics Committees of the Faculty of Medicine and the Clinical Hospital of University of Chile. Consent forms are held by the authors and are available for review by the Editor-in-Chief.

### AUTHOR CONTRIBUTIONS

All authors read and approved the final version of the manuscript. JA and RV had full access to all the data from the study and take responsibility for the integrity of the information and the accuracy of the data analysis. PG-G and KS contributed equally. PG-G, KS, JA, RV, and RT participated in study conception and design. JM and ML were responsible of recruitment and sample collection of healthy subjects. PG-G, KS, AM, HN, JM, BP, and OA participated in data acquisition. PG-G prepared the manuscript. PG-G, KS, AS-G, RV, JA, DC, RT, MM, and DC participated in analysis and interpretation of data and manuscript revision.

### FUNDING

This work was funded by Fondecyt-Chile 1140553 and Millenium Institute on Immunology and Immunotherapy P09/016-F.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://www.frontiersin.org/article/10.3389/fimmu.2017.01350/ full#supplementary-material.

TABLE S1 | Demographic characterization of donors.


TABLE S5 | Functional and pathway analysis of DM-DCs genes from Cluster 6.


a first step towards reproducibility and standardisation of cellular therapies. *PeerJ* (2016) 4:e2300. doi:10.7717/peerj.2300


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2017 García-González, Schinnerling, Sepúlveda-Gutiérrez, Maggi, Mehdi, Nel, Pesce, Larrondo, Aravena, Molina, Catalán, Thomas, Verdugo and Aguillón. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Clinical Tolerogenic Dendritic Cells: Exploring Therapeutic Impact on Human Autoimmune Disease

*Brett Eugene Phillips1 , Yesica Garciafigueroa1 , Massimo Trucco1,2 and Nick Giannoukakis1,2\**

*1Allegheny Health Network Institute of Cellular Therapeutics, Allegheny General Hospital, Pittsburgh, PA, United States, 2Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, PA, United States*

Tolerogenic dendritic cell (tDC)-based clinical trials for the treatment of autoimmune diseases are now a reality. Clinical trials are currently exploring the effectiveness of tDC to treat autoimmune diseases of type 1 diabetes mellitus, rheumatoid arthritis, multiple sclerosis (MS), and Crohn's disease. This review will address tDC employed in current clinical trials, focusing on cell characteristics, mechanisms of action, and clinical findings. To date, the publicly reported human trials using tDC indicate that regulatory lymphocytes (largely Foxp3+ T-regulatory cell and, in one trial, B-regulatory cells) are, for the most part, increased in frequency in the circulation. Other than this observation, there are significant differences in the major phenotypes of the tDC. These differences may affect the outcome in efficacy of recently launched and impending phase II trials. Recent efforts to establish a catalog listing where tDC converge and diverge in phenotype and functional outcome are an important first step toward understanding core mechanisms of action and critical "musts" for tDC to be therapeutically successful. In our view, the most critical parameter to efficacy is *in vivo* stability of the tolerogenic activity over phenotype. As such, methods that generate tDC that can induce and stably maintain immune hyporesponsiveness to allo- or disease-specific autoantigens in the presence of powerful pro-inflammatory signals are those that will fare better in primary endpoints in phase II clinical trials (e.g., disease improvement, preservation of autoimmunitytargeted tissue, allograft survival). We propose that pre-treatment phenotypes of tDC in the absence of functional stability are of secondary value especially as such phenotypes can dramatically change following administration, especially under dynamic changes in the inflammatory state of the patient. Furthermore, understanding the outcomes of different methods of cell delivery and sites of delivery on functional outcomes, as well as quality control variability in the functional outcomes resulting from the various approaches of generating tDC for clinical use, will inform more standardized *ex vivo* generation methods. An understanding of these similarities and differences, with a reference point the large number of naturally occurring tDC populations with different immune profiles described in the literature, could explain some of the expected and unanticipated outcomes of emerging tDC clinical trials.

Keywords: tolerogenic dendritic cells, autoimmune disease, autoimmunity, clinical therapeutics, type 1 diabetes, Crohn's disease, rheumatoid arthritis, multiple sclerosis

#### *Edited by:*

*Catharien Hilkens, Newcastle University, United Kingdom*

#### *Reviewed by:*

*David William Scott, Uniformed Services University of the Health Sciences, United States Raymond John Steptoe, The University of Queensland, Australia*

> *\*Correspondence: Nick Giannoukakis ngiannou@wpahs.org*

#### *Specialty section:*

*This article was submitted to Immunological Tolerance and Regulation, a section of the journal Frontiers in Immunology*

*Received: 29 August 2017 Accepted: 25 September 2017 Published: 12 October 2017*

#### *Citation:*

*Phillips BE, Garciafigueroa Y, Trucco M and Giannoukakis N (2017) Clinical Tolerogenic Dendritic Cells: Exploring Therapeutic Impact on Human Autoimmune Disease. Front. Immunol. 8:1279. doi: 10.3389/fimmu.2017.01279*

### INTRODUCTION

Autoimmune diseases are characterized by the loss of tolerance to self-antigens resulting in the immune system targeting a wide range of tissues leading to impaired function, tissue eradication, and clinical morbidity and mortality. Many of the current therapeutics manage symptoms of a general inflammatory state, even if they target specific molecules on inflammatory cells and/ or their secreted products (e.g., immunokines). Autoimmunity requires ongoing, often lifelong treatment. While systemic immunosuppressives are still the mainstay of treating most autoimmune conditions, biologic-based immunotherapies selectively targeting specific molecules and pathways have become part of the treatment approach, although their side effects often cause more problems than they intend to solve. Cell therapy has been a sought after alternative, or adjunctive approach for at least two decades, since the discovery of tolerogenic dendritic cells (tDC) and with the more recent characterization of T-regulatory cells (Tregs) (1–9). In this review, we will summarize the current tDCbased clinical trials, as well as those that are planned for the treatment of autoimmune diseases. We will point out the common features and the common mechanisms that they share in their functional outcomes and also highlight some key questions that remain to be answered to ensure that these cells remain stably tolerogenic *in vivo*.

Dendritic cells are considered to be the body's "professional" antigen-presenting cells (10–15) and they regulate adaptive immunity and maintain immune homeostasis in the periphery (16). When DC express low levels of surface proteins, collectively referred to as co-stimulation molecules (e.g., CD86, CD40, OX-40), produce little to no IL-12p70, and exhibit low nuclear factor kappa-light-chain-enhancer of activated B-cells (NF-κB) transactivational activity, they are referred to as "immature" (17–20). DC reside inside peripheral tissues throughout the body in this state under normal conditions and they acquire either draining tissue antigens or migrate through the tissues and stromal structures, acquiring antigens through phagoendocytic mechanisms (e.g., trogocytosis) (21–24). They remain as immature cells until the time they encounter a pro-inflammatory environment. When antigens are acquired in an environment of inflammation, such as an infection, DC undergo a series of maturation steps that increase the expression and cell surface level of major histocompatibility complex (MHC) class II molecules for antigen presentation concurrent with the upregulation of co-stimulation molecules, and production of IL-12p70 that together act in concert to stimulate the division and functional polarization of T-cells (25–28). Mature DC do this consequent to their accumulation inside the lymph nodes or lymphoid structures that drain the site from which they acquired the antigens. There, inside the lymphoid organs, they present those antigens to the T-cell receptor on naïve T-cells. A series of secondary interactions with co-stimulatory molecules fully activate T-cells (29, 30). Antigens presented in this fashion are typically foreign, but in autoimmune diseases self-antigens are presented to potentially autoreactive T-cells leading to targeted destruction of tissues (31).

Dendritic cells that acquire antigens but do not receive signals to undergo maturation maintain their immature state and can also present antigens to naïve T-cells in secondary lymphoid organs. In the absence of co-stimulation, these DC usually induce a state of anergy in target T-cells leading to peripheral tolerance. Immature DC further facilitate peripheral immune tolerance by maintaining populations of naturally occurring thymic Tregs and/or induce naïve T-cells to differentiate into peripheral Tregs as they also shift differentiated T-helper (Th) cell phenotypic and functional activity balance toward cell populations representing the Th2 side (1, 32–37). This outcome is usually a consequence of IL-10 gene activation and immunokine production by the DC instead of IL-12p70, which augments the Th2 subpopulation and, in a paracrine feedback manner, inhibits DC maturation (38). While autologous Tregs therapy is an alternative approach to treating autoimmune disease, it is limited by questionable stability of the administered cells *in vivo* (39–41), polyclonality (42–44), and concerns about systemic dissemination of the cells since they are administered intravenously. From a manufacturing perspective, the volume of blood currently needed to generate an injectable cell product (approximately 400 ml per patient) can be prohibitive. Instead, the advantages of tDC lie in their multiple mechanisms to treat disease that involve anergy of autoreactive T-cells, activation of different regulatory lymphocyte populations, dynamic antigen acquisition *in vivo* and presentation to autoreactive T-cells to induce hyporesponsiveness, and migration into lymphoid regions draining the disease target. Over the past 20 years, much research has been invested toward the characterization of these immature DC and into methods that can generate them *in vitro* from hematopoietic progenitors and maintain them stably in an immature state capable of possibly restoring tolerance *in vivo* in autoimmune diseases (2, 9, 17, 45–52).

### TYPE 1 DIABETES (T1D) MELLITUS

Type 1 diabetes is a disease that leads to the progressive loss of pancreatic beta cells and insulin production. Insulin replacement is the only and current gold standard of therapy, but even rigorous control of blood glucose levels fails to prevent the development of diabetic complications (53). These complications include neuropathy, nephropathy, vision loss, and cardiovascular disease which are associated with high morbidity and mortality (54). Devices for the delivery of insulin may mimic how insulin is secreted and could potentially reduce diabetes-related complications (55, 56), but they do not address the underlying autoimmune pathology, nor is insulin release fully coupled to

**Abbreviations:** APC, antigen-presenting cells; Bregs, B-regulatory cells; DAS28, disease activity scores 28; DC, dendritic cells; Dex, dexamethasone; GM-CSF, granulocyte macrophage colony-stimulating factor; HLA-DR, human leukocyte antigen-antigen D related; IFNγ, interferon gamma; IκBα, nuclear factor kappalight-chain-enhancer of activated B-cells inhibitor, alpha; IL, interleukin; MHC, major histocompatibility complex; MITAP, minimum information about tolerogenic antigen-presenting cells; MPA, monophyosphoryl lipid A; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B-cells; RA, rheumatoid arthritis; RALDH, retinaldehyde dehydrogenase; T1D, type 1 diabetes; TCR, T-cell receptor; tDC, tolerogenic dendritic cells; Th, T-helper cells (1,2, or 17); TNFα, tumor necrosis factor alpha; Tregs, T-regulatory cells.

second-to-second fluctuating glucose levels. Autoimmunity suppression is also a hurdle for the implementation of islet transplants that, while reducing or delaying the clinical outcome of complications, would come under the same rejection by leukocytes even with the application of drugs to prevent tissue rejection (57–60).

Tolerogenic dendritic cells are a potential therapy for the treatment of new onset T1D to prevent the further destruction of pancreatic beta cells. Loss of beta cell mass can reach 80% by time of diagnosis (61), making the therapeutic window small, but feasible. The first tDC clinical trial for the treatment of autoimmune disease was for T1D (clinicaltrials.gov identifier: NCT00445913) (62). Monocytes were isolated by leukapheresis and grown *ex vivo* in the presence of granulocyte macrophage colony-stimulating factor (GM-CSF) and interleukin-4 (IL-4) for 6 days. Cells for the treatment arm of the study were cultured with a mixture of antisense oligonucleotides targeting the primary transcripts of the CD40, CD80, and CD86 co-stimulatory molecules at a concentration of 3.3 µM each oligonucleotide. Cells proven to exhibit reduced expression of these co-stimulation proteins (by flow cytometry) and passing the viability and sterility screen were given to patients in four treatments of 1.0 × 107 cells, where each round of administration was 2 weeks apart. Each round of treatment was divided into four intradermal injection sites proximal to the expected anatomical location of the pancreas in an effort to enhance DC migration to the pancreatic and peri-pancreatic lymph nodes, based on known and suspected lymphatic drainage fields. All tDC were from thawed cryopreserved cell stocks. Ten patients were recruited for the phase I study; 3 patients in the control arm and 7 in the tDC treatment arm. Safety for patients was assessed in-trial (12 months).

The tDC were well tolerated without any adverse events noted. Two novel findings resulted from the study. First, the tDCtreated arm displayed a transient elevation of B220+ CD11c− B cells that, during the study, appeared to contain a subpopulation of B-regulatory cells (Bregs). The presence of Bregs and the effect of the tDC on their generation was demonstrated in a follow-up study (63). The second finding was that, in 4/7 patients who were insulin C-peptide negative at baseline, there was a conversion to C-peptide positivity to sub-physiological concentrations in 3/7, but to physiological levels in one patient, during the tDC administration cycle. C-peptide is the cleavage product of proinsulin as it matures into insulin during its biosynthesis and secretory phases inside the pancreatic beta cells and is used as a surrogate marker for functional beta cells (64). However, this trial's intent was to assess safety of the tDC and in spite of these findings, there was no attempt to determine if insulin dosage could be adjusted. Patients recruited in this study were diabetic and insulin-requiring for a minimum of 5 years and, therefore, should not have been expected to harbor significant residual beta cell mass. The emergence of detectable C-peptide during the tDC treatment cycle suggests restoration of insulin production from remaining islets or possible new islet formation. There were no significant differences in other measurements between control and tDC treatment arms (e.g., in cytokine serum concentrations or cell population number other than Bregs), even though a subtle, albeit statistically insignificant increase in Tregs number were detected in tDC-treated patients.

### RHEUMATOID ARTHRITIS (RA)

Rheumatoid arthritis is an inflammatory disease that targets the cartilage of the joint articulations, with the highest rate of occurrence in small joints of the hands and feet (65). Chronic inflammation further results in loss of bone mass, tendon inflammation, and rupture associated with airway and cardiovascular complications (66). Current treatment strategies require continuous treatment with anti-inflammatory drugs and biologics. These, however, fail to maintain remission over the life of the disease. With an RA global incidence rate as high as 1% of adults (67), there is a large patient population that could benefit from tDC therapeutics.

### Rheumavax RA Study

The first-in-human trial for the treatment of RA generated tDC by NF-κB inhibition (clinicaltrials.gov identifier: NCT00396812) (68). The transcription factor NF-κB controls gene expression of genes involved in many pro-inflammatory pathways, making it a target of choice for anti-inflammatory drugs (69). Inhibition of NF-κB prevents DC maturation, reduces the expression of CD40 and human leukocyte antigen–antigen D related (HLA-DR, a class II MHC molecule), and confers tolerogenic properties to DC including induction of T-cell anergy (70, 71). Isolated monocytes were grown in the presence of IL-4, GM-CSF, and 2–2.5 µM of the NF-κB inhibitor Bay 11-7082 for 48 h. DC were further prepared in a 3-h exposure to citrullinated peptides of aggrecan, vimentin, collagen type II, and a and b fibrinogen which are putative RA autoantigens (72) as anti-citrullinated protein antibodies are found in 50–80% of patients over the lifetime of the disease (65). Preloading tDC with disease-specific autoantigens increases the likelihood of their presentation to T-cells inside the inflamed joint-draining lymph nodes, thus disrupting the cycle of autoreactive T-cell activation. The resulting generated tDC displayed a 5% reduction in the mean fluorescence intensity (flow cytometric measurement) of CD40 and a 17% reduction in HLA-DR when assessed by flow cytometry (68). Patients were given a single intradermal injection of 1.0 × 106 or 5.0 × 106 tDC.

The treatment was generally well tolerated and deemed safe. General trends indicated a 25% decrease in pro-inflammatory T-cells (CD4+ CD25+ CD127+) and 25% increase in antiinflammatory Treg (CD4+ CD25+ high CD127−) within 1 month of treatment. Circulating levels of the inflammation marker C-reactive protein (CRP) were significantly decreased in patients receiving the high cellular dose. Similarly, cytokine expression profiles for IL-15, CXCL1, CXCL11, IL-29, and peptide YY were reduced in patients receiving the high dose Rheumavax treatment. Disease activity scores 28 (DAS28), a common metric used for the evaluation RA severity, were decreased in a portion of the patients.

### Newcastle University RA Study

The second RA tDC trial was conducted at the University of Newcastle and used dexamethasone (Dex) and vitamin D3 for tDC generation (clinicaltrials.gov identifier: NCT01352858) (73). Dex is a synthetic glucocorticoid that has a range of powerful anti-inflammatory effects in the clinical setting (74). Dex inhibits the NF-κB pathway through a number of mechanisms. The most prominent includes increased nuclear factor kappa-light-chain-enhancer of activated B-cells inhibitor, alpha (IκBα) expression which binds and retains the RelA subunit of NF-κB inside the cytoplasm preventing transcriptional activities inside the nucleus (75, 76). tDC grown in the presence of Dex exhibit decreased expression of co-stimulation proteins CD40 and CD86 and the DC maturation marker CD83, along with decreased class II MHC expression and IL-12p70 production (71, 77–79). These tDC produced high concentrations of the immunosuppressive IL-10 immunokine (80). Similar alterations in DC surface and cytokine expression profiles can also result with vitamin D3 treatment *in vitro* (81–83). Interestingly, vitamin D3 deficiency is associated with RA and poorer clinical outcomes (84, 85). Generation of tDC with both Dex and vitamin D3 has an additive effect on IL-10 production levels (26, 86).

In this trial, monocytes were isolated by density centrifugation followed by microbead selection of CD14 expressing cells. Monocytes were grown in culture for 7 days in the presence of 50 ng/ml IL-4 and 50 ng/ml GM-CSF; with the addition of 1 µM Dex on day 3 and day 6, 0.1 nM vitamin D3 on day 6, and 1.0 µg/ml monophyosphoryl lipid a (MPA) on day 6. Cells were then cocultured with synovial fluid collected from inflamed joints of study patients allowing for unique autoantigen loading specific to each patient. The patient-specific tDC were characterized with reduced CD40, CD83 surface levels and decreased IL-12p70 production while maintaining high concentrations of secreted IL-10 (73, 77). After tDC passed sterility testing, patients received a single injection of saline, 1.0 × 106 , 3.0 × 106 , or 1.0 × 107 cells into the affected knee joint. The treatment was deemed safe with no worsening knee flares and a reduction in symptoms of patients treated with the high dose. Peripheral blood immune T-cell populations (CD4+ IL-10+, CD4+ FoxP3+, CD4+ IFNγ+, CD4+ IL-17+) and cytokines production levels [IL-10, interferon gamma (IFNγ), IL-17, IL-6, tumor necrosis factor alpha (TNFα)] were unaltered.

### CROHN'S DISEASE

Crohn's disease is an autoimmune disease of the gastrointestinal (GI) tract that can affect tissues from the mouth to the anus (87). Common symptoms include abdominal pain, bloody diarrhea, inflammation, weight loss, and bowel blockage (87, 88). Current treatments are designed to manage the symptoms, but disease flare-ups are common. There are no specific therapies against the underlying autoimmunity. A single phase I clinical trial has been reported as completed, testing the safety of tDC (European Clinical Trials Database number 2007-003469-42) (89).

The immunologic space of the intestine is exposed to a high number of foreign antigens provided by intestinal flora. The breakdown of immune control is mediated by inappropriate activation of Th1 and Th17 cells and the loss of retinaldehyde dehydrogenase (RALDH)-positive DC. This DC subpopulation may be the reason vitamin A was incorporated into the tDC generation process for this trial. Vitamin A deficiency is prevalent in patients with Crohn's disease and correlates with disease severity (90). Conventional CD103+ CD11b+ intestinal DC convert vitamin A to retinoic acid through expression of RALDH which is atypical of DC found in draining lymph nodes (91). DC-generated retinoic acid maintains tolerance to GI tract cells and tissues through enhanced CD4+ T cell recruitment to the intestine and differentiation into FoxP3+ T-cells and Th17 from existing CD4+ T-cell populations (1, 26, 92, 93). Furthermore, generation of retinoic acid-producing DC naturally inside the disease-affected tissues as a consequence of administration of retinoic acid-producing tDC could establish an ongoing "feed forward" type of tDC generation and stabilization cycle in the patient's intestinal epithelial cells. This clinical trial relies on proximal tDC delivery, but mentions that future methods may switch to direct delivery of tDC into intestinal lesions (89).

For the generation of tDC in this trial, monocytes were obtained by leukapheresis. Cells were cultured in 500 UI/ml IL-4 and 800 UI/ml of GM-CSF for 7 days; 1 µM of Dex and 1 nM vitamin A starting on day 3; and the cytokines IL-1β, IL-6, TNFα, and prostaglandin E2 for the final day (89, 94). The cell products exhibited elevated CD80 and CD86, and low CD83 expression. MERTK, a glucocorticoid-induced receptor that is prevalent in tDC was also expressed at high levels. Production of IL-10 was detected in the cells with no detectable IL-12p70 or IL-23 in the cell culture media. Allogenic mixed lymphocyte reactions performed in the presence of tDC resulted in low T-cell proliferation and IFNγ production. tDC were administered to Crohn's patients by intraperitoneal injection in six different treatment arms based on the number of administered cells (2.0 × 106 , 5.0 × 106 , 1.0 × 107 ) and number of injections (one dose or three doses spread out every 2 weeks). These tDC were well tolerated. One-third of the patients completing the study showed a clinical improvement based on a Crohn's disease activity index. Th1 and Th17 cell populations were unchanged in numbers in circulation, but there was a significant increase in circulating Tregs (CD4+ CD25+ Foxp3+) 12 weeks after injection when compared to baseline. Isolated T-cells stimulated with CD3 antibody secreted less IFNγ suggesting that the tDC had established some form of immune hyporesponsiveness in the patients.

A second clinical trial has been initiated for Crohn's disease using tDC (clinicaltrials.gov identifier: NCT02622763); however, at this time very few details are known about the methods of tDC generation.

### MULTIPLE SCLEROSIS (MS)

Multiple sclerosis is an autoimmune disease that results in the demyelination of neurons in the central nervous system as well as in the peripheral nervous system. Demyelination is mediated by autoreactive T-cells activated by self-antigen presentation by DC. A number of drugs and biologics are being used to inhibit various immune pathways (95), and tDC are currently being used in two phase I clinical trials. To date, the results of these trials have not been yet published. The first trial (clinicaltrials. gov identifier: NCT02283671) utilizes tDC generated in the presence of IL-4, GM-CSF, and Dex. These cells are pre-loaded with myelin self-peptides and are administered intravenously in three injections each 2 weeks apart. The second trial (clinicaltrials.gov identifier: NCT02618902) considers tDC generated in the presence of vitamin D3 and similarly preloads cells with myelin self-peptides. Patients will receive 5.0 × 106 , 1.0 × 107 , or 1.5 × 107 cells spread over five intradermal injection sites in the subclavicular region. This will be the highest administered dose of tDC described in current tDC clinical trials, which was probably informed by the safety reports of previous tDC trials. Similar to retinoic acid, vitamin D levels are lower in patients with MS than healthy individuals. Relapse of MS symptoms are also associated with lower vitamin D levels when compared to MS patients that are currently in intermission (96, 97). Generation of tDC from healthy and MS patients in the presence of vitamin D3 results in reduced tDC IL-12 and IL-23 cytokine secretion,

inhibited maturation, and increased CD83/decreased CD80 cell surface expression (95).

### DISCUSSION

Tolerogenic dendritic cells have, or are currently being tested in phase I clinical trials for T1D, RA, MS, and Crohn's disease, with additional considerations aiming at lupus (98) and facilitating allogeneic tissue and organ transplantation (9, 99–101). tDC generation relies on the use of IL-4 and GM-CSF to differentiate monocyte progenitors, and these cytokines remain the central feature shared among all the tDC generation methods. The differences, however, lie in the additional factors added in the cell cultures from the time of monocyte seeding to the last changes in media prior to tDC harvest (e.g., putative autoantigens, vitamin D3, immunosuppressives like Dex and NF-κB inhibitors, antisense oligonucleotides targeting co-stimulation) (**Table 1**). To what extent these conditions change cellular effectiveness and mechanism of action of tDC to confer their potentially


*Cell generation displays the reagents used in tDC preparation (not including shared IL-4 and GM-CSF components) and cell characterization displays surface markers and cytokine secretion profiles of pre-injected cells. Table entries marked as "X" are values that were not assessed within a given trial. Arrows indicate a change for a given value, but were not present in all patients within a study, exist at specific time points that may not be maintained for the duration of the study, or failed to reach significance in some studies. MS studies are still underway and unpublished. The information provided derives from clinicaltrial.gov entries for these registered clinical trials and is current as of August 29, 2017.*

beneficial effects is unclear at present. Nevertheless, most tDC share one mechanistic feature: increased regulatory lymphocytes (e.g., Foxp3+ Tregs and Bregs) in the peripheral blood of patients during administration (62, 68, 89).

Another difference among the tDC used in clinical trials lies in the dose level administered and site of cell delivery in the body. This last point is relevant in the mechanism of tDC since affected tissues and focal points of inflammation differ among autoimmune diseases. The majority of tDC clinical trials to date deliver tDC proximal to the site of inflammation, with the desired goal of tDC migration into the local draining lymph node. Draining lymph nodes adjacent to the site of inflammation have a great preponderance of activated self-reactive T-cell populations to target for anergy (102). The clinical studies described so far have used between 1 and 5 injection sites per cell treatment cycle, targeting one or more pertinent lymph nodes such as the cervical lymph nodes in the MS study (clinicaltrials.gov identifier: NCT02618902). An alternative approach is to directly introduce tDC into the site of inflammation. Direct administration of tDC to lesion sites in Crohn's disease was not attempted but suggested for future study. This would address a different mode of action, where the vitamin A-generated tDC could potentially restore a lost intestinal subpopulation of tDC specific to Crohn's disease. Targeting "niche" tDC populations may require the need for the generation of tDC with more restricted immunosuppressive phenotypes. While the Newcastle University RA study introduced tDC directly at the site of inflammation, the intended goal was still for the migration of tDC to local draining lymph nodes. Even though the technique is more invasive than intradermal administration, the introduction of tDC producing IL-10 may have the added benefit of local immunosuppression at the point of inflammation. This consideration is balanced by the possibility that local inflammatory conditions may alter the introduced tDC phenotype to a more pro-inflammatory state.

Autoimmune diseases each have their own unique autoantigens and associated self-reactive T-cell populations. Preloading tDC with specific disease antigens enhances their ability to directly interact and inactivate self-reactive T-cells that cause tissue damage. The Rheumavax RA study loaded tDC with citrullinated peptides identified from 70% of RA patients who exhibit auto-antibodies to these targets. To further this strategy, they selected patients with high risk HLA alleles that have a strong association with citrullinated auto-antibody positivity. Unfortunately not all patients display uniform self-antigens for a given disease. T1D, for example, is associated with a range of self-antigens and auto-antibodies that are differentially expressed among patients and at different points during the disease. Even though there seems to be a general consensus about insulin and GAD65-derived peptide-pulsing tDC for T1D, antigen spreading that has occurred at the time of clinical disease may limit the autoreactive T-cell populations targetable, whereas other "late-antigen"-specific T-cells may in fact be driving autoimmunity after clinical onset. In an elegant study designed by the Newcastle University group, the RA trial overcame this potential limitation by collecting synovial fluid from inflamed joints of each patient. tDC were pre-exposed to autologous synovial fluid for antigen collection, and then given an additional chance to acquire patient-specific autoantigens through direct administration of tDC to the site of inflammation. If initial tDC therapeutics trials are successful, further studies may wish to look at the effectiveness of matching patient autoantigens despite the potential increase in manufacturing and quality control costs.

Currently, only four of the discussed clinical trials have been completed with reported outcomes (62, 68, 73, 89). Despite the different approaches used to generate the tDC in these trials, NF-κB inhibition is the central feature of 3 of these studies, with one study also including the use of vitamin D3. Generation of tDC with either NF-κB inhibitors or vitamin D3 promotes immature DC phenotypes with an additive effect when using both agents. The Newcastle University RA (Dex + Vit D3) and the Crohn's disease (Dex) trials both reported decreased CD83 expression, high CD86 expression, decreased IL-12 secretion, and elevated IL-10 secretion in their tDC products suggesting a possible tDC shared phenotype. Pre-activation of tDC with cytokines or lipid immune mediators is also shared between these two protocols. The *Rheumavax* RA (BAY 11-7082) study measured different parameters, but did report a divergent decrease in CD80 surface levels. In contrast, the T1D clinical trial directly intervened to reduce and maintain stably reduced co-stimulatory molecules CD40, CD80, and CD86 without the use of an NF-κB inhibitor, but other than demonstrating low IL-12 concentrations during stimulation *in vitro*, it did not further characterize the generated tDC beyond purity and sterility. Without full characterization of, at least, the immune phenotypes and functional immune activities, it will be difficult to compare the mechanisms of action among the different tDC to functionally identify their points of intersection (e.g., do all tDC promote Tregs, and how? Are key immunoregulatory immunokines produced by all tDC, and/or what are the immunokines that tDC elicit in common among the different Th cell populations?). The difficulty in comparing the characteristics of different clinical tDC does suggest the need for an uniform set of metrics for their description. This was the focus of the minimum information about tolerogenic antigen-presenting cells (103) initiative whose authors included members from a number of the completed and ongoing clinical trials.

Much of the current divergence in tDC phenotype and points of mechanistic intersection other than increased frequency of regulatory immune cells in the peripheral blood during treatment might also be due to the *ex vivo* upstream cell processing prior to the addition of GM-CSF/IL-4 (e.g., monocyte progenitors, contaminating granulocytes in the monocyte elutriation). An important question that needs to be addressed is the relevance of the tDC method and site of delivery (intravenous, subcutaneous, intradermal) on their effect and mechanism of action (direct or indirect) at the lymphoid organs draining the inflamed tissues and/ or the autoimmunity target tissues proper. Finally, it is important to determine if freshly generated versus thawed cryopreserved tDC are functionally different *in vivo*. Considering the limitations and adverse events encountered using biologic agents and the need to move past systemically acting immunosuppressives, the well-tolerated safety profile of tDC across a range of dose levels and administration sites, along with the evidence of increased regulatory cell frequency *in vivo* during treatment, strongly argues in favor of their further development, characterization, and consideration to fundamentally change how autoimmune diseases are treated, directly addressing the immune imbalance and moving away from disease and symptom management.

### REFERENCES


### AUTHOR CONTRIBUTIONS

BP and YG wrote the manuscript and MT and NG edited while adding additional insights. The final version was proofread and edited by NG.


**Conflict of Interest Statement:** NG and MT hold equity in Diavacs Inc., which has licensed the intellectual property concerning the tolerogenic dendritic cells noted in the review under the type 1 diabetes clinical trial (also referred to as the "Pittsburgh tolerogenic dendritic cells"). The other authors do not have any conflicts of interest, real or potential.

The handling editor declared a past co-authorship with the authors NG and MT.

*Copyright © 2017 Phillips, Garciafigueroa, Trucco and Giannoukakis. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*