# REGULATION OF IMMUNE FUNCTION BY THE LYMPHATIC VASCULATURE

EDITED BY : Beth A. Jirón Tamburini, Amanda W. Lund and Timothy P. Padera PUBLISHED IN : Frontiers in Immunology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-325-8 DOI 10.3389/978-2-88963-325-8

#### 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

# REGULATION OF IMMUNE FUNCTION BY THE LYMPHATIC VASCULATURE

Topic Editors:

Beth A. Jirón Tamburini, University of Colorado Denver, United States Amanda W. Lund, Oregon Health & Science University, United States Timothy P. Padera, Massachusetts General Hospital, Harvard Medical School, United States

Citation: Tamburini, B. A. J., Lund, A. W., Padera, T. P., eds. (2020). Regulation of Immune Function by the Lymphatic Vasculature. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-325-8

# Table of Contents


Dante Alexander Patrick Louie and Shan Liao

*41 Leucocyte Trafficking via the Lymphatic Vasculature— Mechanisms and Consequences*

David G. Jackson


Morgan Campbell Hunter, Alvaro Teijeira, Riccardo Montecchi, Erica Russo, Peter Runge, Friedemann Kiefer and Cornelia Halin

*89 Mesenteric Lymphatic Alterations Observed During DSS Induced Intestinal Inflammation are Driven in a TLR4-PAMP/DAMP Discriminative Manner*

Matthew Stephens, Shan Liao and Pierre-Yves von der Weid


Laure Garnier, Anastasia-Olga Gkountidi and Stephanie Hugues

*124 The Interplay Between Lymphatic Vessels and Chemokines* Rae H. Farnsworth, Tara Karnezis, Simon J. Maciburko, Scott N. Mueller and Steven A. Stacker

#### *138 Comparative Transcriptomic Analysis Identifies a Range of Immunologically Related Functional Elaborations of Lymph Node Associated Lymphatic and Blood Endothelial Cells*

Stella J. Berendam, Alexander F. Koeppel, Nicole R. Godfrey, Sherin J. Rouhani, Amber N. Woods, Anthony B. Rodriguez, J. David Peske, Kara L. Cummings, Stephen D. Turner and Victor H. Engelhard

*155 The Antigen Processing and Presentation Machinery in Lymphatic Endothelial Cells*

Laura Santambrogio, Stella J. Berendam and Victor H. Engelhard

*164 Chronic Liver Disease in Humans Causes Expansion and Differentiation of Liver Lymphatic Endothelial Cells*

Beth A. Jirón Tamburini, Jeffrey M. Finlon, Austin E. Gillen, Michael S. Kriss, Kent A. Riemondy, Rui Fu, Ronald P. Schuyler, Jay R. Hesselberth, Hugo R. Rosen and Matthew A. Burchill


Beth A. Jirón Tamburini, Alan M. Elder, Jeffrey M. Finlon, Andrew B. Winter, Veronica M. Wessells, Virginia F. Borges and Traci R. Lyons

# Editorial: Regulation of Immune Function by the Lymphatic Vasculature

#### Beth Ann Jiron Tamburini 1,2 \*, Timothy P. Padera<sup>3</sup> \* and Amanda W. Lund<sup>4</sup> \*

*<sup>1</sup> Division of Gastroenterology and Hepatology, Department of Medicine, School of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, United States, <sup>2</sup> Department of Immunology and Microbiology, School of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, United States, <sup>3</sup> Edwin L. Steele Laboratories, Department of Radiation Oncology, Massachusetts General Hospital Research Institute, Harvard Medical School, Boston, MA, United States, <sup>4</sup> Department of Cell, Developmental, & Cancer Biology, Knight Cancer Institute, Oregon Health & Science University, Portland, OR, United States*

Keywords: lymphatic vasculature, lymphatic endothelial cell (LEC), leukocyte trafficking, lymphedema, immune regulation, cancer lymphatics, antigen presentation

**Editorial on the Research Topic**

#### **Regulation of Immune Function by the Lymphatic Vasculature**

#### Edited and reviewed by:

*Herman Waldmann, University of Oxford, United Kingdom*

#### \*Correspondence:

*Beth Ann Jiron Tamburini beth.tamburini@cuanschutz.edu Timothy P. Padera tpadera@steele.mgh.harvard.edu Amanda W. Lund lunda@ohsu.edu*

#### Specialty section:

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

Received: *21 September 2019* Accepted: *21 October 2019* Published: *05 November 2019*

#### Citation:

*Tamburini BAJ, Padera TP and Lund AW (2019) Editorial: Regulation of Immune Function by the Lymphatic Vasculature. Front. Immunol. 10:2597. doi: 10.3389/fimmu.2019.02597* The lymphatic vasculature is composed of a hierarchy of vessels that extend from peripheral tissues into lymph nodes (LN) and provide the critical route for leukocyte migration and antigen presentation that drive innate and adaptive immune responses. However, beyond physically connecting peripheral tissue to LNs, there is an expanding view of how the lymphatic vasculature influences immune responses. The lymphatic vasculature can modulate immune responses ranging from peripheral tolerance during homeostasis to protective immunity following infection or vaccination to tumor immune escape. Recent work demonstrates that lymphatic vessels actively regulate transport functions, express and secrete chemo-attractants to initiate leukocyte migration, and directly interact with leukocytes to inform their behavior, activation, and establishment of memory. The immunological impact of how the lymphatic vasculature is influenced by tissue microenvironments and ongoing regional inflammatory processes is only beginning to be appreciated, but is of critical relevance to the role lymphatic vessels play in diseases. This Research Topic, "Regulation of Immune Function by the Lymphatic Vasculature," brings together 17 articles that present our current understanding of and provide new insight into the necessary role the lymphatic vasculature plays in regulating inflammation and immunity.

Lymphatic vessels mediate the exit of multiple leukocyte types from peripheral tissue and LNs, including dendritic cells, neutrophils, T cells and B cells. A comprehensive review by Jackson presents the molecular mechanisms that determine leukocyte homing and transmigration to and through peripheral lymphatic vessels while Hampton and Chtanova describe the immunological consequences of leukocyte exit in disease. Farnsworth et al. complement these insights with an extensive review of the interplay between lymphatic vessels and chemokines, and how the dynamic display of various chemokine ligands in response to microenvironmental stimuli influences the migratory nature of leukocytes. Interesting new data presented by Campbell et al. describe novel MHC-dependent interactions between T cells and dendritic cells within peripheral lymphatic capillaries. These exciting data suggest that within lymphatic capillaries, dendritic cells may interact with circulating T cells prior to entry into the LN. How the kinetics of leukocyte egress from peripheral tissues contributes to disease remains a critical question moving forward and raises the possibility that lymphatic vessels may help to pattern both the initiation and resolution phases of peripheral tissue inflammation.

**5**

Given the multiple roles lymphatic vessels play in regulating leukocyte trafficking and function, the field has long considered how lymphatic function, or dysfunction, contributes to the perturbation of immune homeostasis. Intriguingly, lymphatic vessel expansion is associated with autoimmune, chronic inflammatory, and malignant contexts. However, it remains unclear whether lymphatic remodeling is a cause of or rather a biproduct of tissue inflammation. Schwager and Detmar provide an extensive review of the mechanisms that regulate inflammatory lymphangiogenesis and discusses implications for disease and Schwartz et al. review the current literature surrounding the contribution of lymphatic dysfunction to autoimmunity. Providing new insights, Stephens et al. demonstrate that TLR4 signaling during intestinal inflammation is required for mesenteric lymphatic vessel expansion and leakiness, which leads to associated changes in dendritic cell migration to draining LNs. Though inhibition of TLR4 signals restored lymphatic function, there was no change in tissue inflammation, suggesting lymphatic vessel dysfunction lies downstream of inflammatory insult. Interestingly, Tamburini et al. also describe disease-dependent changes in lymphatic endothelial cell (LEC) transcriptional states in human chronic liver disease. Whether these changes cause inflammation or are the response to inflammation remains to be carefully determined.

The interplay between lymphatic and immune dysfunction is perhaps most apparent in the context of lymphedema. As lymphedema progresses, tissue fluid accumulation, and stasis cause multiple tissue changes, including fibrosis, adipose tissue deposition, and chronic inflammation. These tissue changes, along with impairment in the ability to transport antigen and antigen presenting cells to lymph nodes, drives progressive deterioration in local immune function. Kataru et al. and Yuan et al. provide back to back reviews highlighting the cellular and molecular mechanisms that occur during lymphedema and their impact on immunity in mouse models and outcomes for patients.

In this Research Topic, and in the field in general, much emphasis is placed on peripheral lymphatic vessel remodeling, however, Lucas and Tamburini discuss the molecular mechanisms of LEC expansion and contraction in LNs draining inflamed tissues and the role of LN LECs in regulating antigen presentation vs. antigen exchange. Louie and Liao further review the role of subcapsular macrophages in the LN, whose position and survival depend on LN LECs, in innate pathogen defense. Interestingly, the LN LEC population may also be altered in the context of chronic disease. Tay et al. present new data that describes reduced lymphocyte egress from LNs during hypercholesterolemia, which results in LN hypertrophy. These changes in lymphocyte trafficking were likely driven at least in part by altered CCL21 and S1P gradients, signaling molecules generated by LN LECs.

LECs, particularly those found in LNs, were recently demonstrated to harbor intrinsic antigen-presentation capabilities and to archive antigens in order to influence T cell activation. Antigen-dependent and independent interactions between LECs, T cells and DCs may both preserve peripheral tolerance and promote protective immunity depending on context. These observations have continued to fuel the concept that LECs directly influence immunity. This hypothesis is now further supported by transcriptional data from Berendam et al. that describe modules of immune transcripts within LN LECs that support the novel functionality of LN LECs and provide the basis for future mechanistic studies. These data are summarized and integrated with the existing knowledge by Santambrogio et al. in order to describe the machinery found within LECs that may facilitate antigen scavenging and presentation.

Finally, the contribution of lymphatic vessels to tumor progression remains an area of intense interest. While historical paradigms associate lymphatic vessels with regional tumor metastasis, new work integrates the potential effects of lymphatic vessel remodeling on anti-tumor immunity and immunotherapy. Garnier et al. provide a comprehensive review of the existing framework for lymphatic vessel involvement in anti-tumor immunity, describing both the impact of lymphatic vessel remodeling on immunity and responses to local inflammation. Tamburini et al. demonstrate that the microenvironment of postpartum mammary gland involution potentiated immunotherapy leading to reduced tumor volume, increased immune cell activation and decreased lymphatic vessel density compared to tumors in nulliparous hosts. Lymphatic vessels are high expressors of PD-L1 in involuting mammary glands and may contribute to this response, consistent with recent published data also demonstrating a functional role for PD-L1 on tumorassociated lymphatic vessels.

With this cumulative work, our field continues to show that lymphatic vessels are active components of host immunity. The mechanisms through which lymphatic vessels synchronize immune expansion and contraction in both LNs and peripheral tissues provide therapeutic opportunities for immunomodulation and will ultimately make a significant impact on our understanding of health and disease.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

BT was supported by the University of Colorado Department of Medicine Outstanding Early Career Scholar and RNA Biosciences Clinical Scholar Award and National Institutes of Health NIAID (R01 AI121209). AL was supported by the OHSU Knight Cancer Center support grant from the National Institutes of Health (NIH P30-CA069533), Department of Defense Peer Reviewed Cancer Research Program (W81XH-15- 1-0348), Melanoma Research Alliance, V Foundation for Cancer Research (V2015-024), American Cancer Society (RSG-18-169- 01-LIB), Cancer Research Institute CLIP Program, National Institutes of Health NCI (R01CA238163), and is a Cancer Research Institute Lloyd J. Old STAR. TP was supported by the National Institutes of Health (R01CA214913, R01HL128168, and R21AI135092).

## ACKNOWLEDGMENTS

We thank the authors for contributing to this Research Topic.

**Conflict of Interest:** 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 © 2019 Tamburini, Padera and Lund. 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(s) 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.

# Lymph Node Lymphatic Endothelial Cell Expansion and Contraction and the Programming of the Immune Response

#### Erin D. Lucas 1,2 and Beth A. J. Tamburini 1,2 \*

*<sup>1</sup> Division of Gastroenterology and Hepatology, Department of Medicine, School of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, United States, <sup>2</sup> Department of Immunology and Microbiology, School of Medicine, University of Colorado Anschutz Medical Campus, Aurora, CO, United States*

#### Edited by:

*Anne Fletcher, Monash University, Australia*

#### Reviewed by:

*Muriel Moser, Free University of Brussels, Belgium Lianjun Zhang, Suzhou Institute of Systems Medicine (ISM), China*

\*Correspondence:

*Beth A. J. Tamburini beth.tamburini@ucdenver.edu*

#### Specialty section:

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

Received: *01 November 2018* Accepted: *08 January 2019* Published: *25 January 2019*

#### Citation:

*Lucas ED and Tamburini BAJ (2019) Lymph Node Lymphatic Endothelial Cell Expansion and Contraction and the Programming of the Immune Response. Front. Immunol. 10:36. doi: 10.3389/fimmu.2019.00036* Lymphatic endothelial cells (LECs) form the structure of the lymphatic vessels and the sinuses of the lymph nodes, positioning them to be key players in many different aspects of the immune response. Following an inflammatory stimulus, LECs produce chemokines that recruit immune cells to the lymph nodes. The recruitment of immune cells aids in the coordination of both LEC and lymph node expansion and contraction. More recent data has demonstrated that to coordinate LEC division and death, cell surface molecules, such as PD-L1 and interferon receptors, are required. During homeostasis, LECs use PD-L1 to maintain peripheral tolerance by presenting specific peripheral tissue antigens in order to eliminate tissue specific responses. LECs also have the capacity to acquire, present, and exchange foreign antigens following viral infection or immunization. Here we will review how lymph node LECs require immune cells to expand and contract in response to an immune stimulus, the factors involved and how direct LEC-immune cell interactions are important for programming immunity.

Keywords: lymphatic endothelial cell, lymph node expansion, PD-L1, apoptosis, immune tolerance, lymph node contraction, dendritic cell, interferon

### INTRODUCTION

Lymphatic endothelial cells (LECs) are a specialized subset of endothelial cells that comprise lymphatic vessels in the tissue and lymph node (LN). LECs interact with innate and adaptive immune cells both in the tissue and in the LN. LECs have the capacity to produce chemokines in order to recruit immune cells to the LN. Of the chemokines that LECs produce, CCL21 has been implicated in the recruitment of dendritic cells (DC), which in turn promotes LN expansion (1–4). Regulation of LN LEC division and death during LN expansion and contraction is a complicated process to which innate immune cells, adaptive immune cells, and specific signaling molecules contribute. Furthermore, LN LECs not only receive signals from immune cells, but also provide signals to the adaptive immune system to regulate peripheral tolerance and protective immunity. In this review we will highlight how LN LEC interactions and signaling regulate LECs in the LN in response to an inflammatory insult and how LECs program the immune response.

**8**

#### REGULATION OF LN LEC DIVISION BY THE INNATE IMMUNE SYSTEM DURING INFLAMMATION

During an inflammatory response, the LN must expand to allow for the rapid influx and division of responding lymphocytes. To do this, several coordinated processes in the LN occur: (1) the secretion of chemokines and cytokines and thus the recruitment of innate immune cells; (2) the relaxation of the fibroblastic reticular cell (FRC) network; (3) the division of the stromal cells in the lymph node; (4) the adaptive immune response and (5) the contraction of the LN.

Between 0 and 24 h following an inflammatory stimulus both type 1 and type 2 interferon (IFN) production is increased, which inhibits LEC division (5) (**Figure 1A**). Why LEC division is inhibited at this time point is unclear, however this time point coincides with increased expression of CCL19 and 21 by LN stromal cells (17, 18). Dendritic cells (DC) are recruited to the LN through interactions between CCR7 and CCL19 and 21 (19–21) (**Figure 1A**). Following DC recruitment to the LN, LEC division is initiated. CD11c+ DCs have been shown to lead to LEC proliferation through LEC-DC contact, a process that is ablated following CD11c+ cell depletion (6). Further work showed that DCs regulate the relaxation of the FRC network through the interaction of C-type lectin like 2 (CLEC-2) with podoplanin (PDPN) on the FRCs (22–24). CLEC-2 binding inhibits PDPN signaling, resulting in FRC elongation and increased LN elasticity (24). PDPN expression by LECs and binding by CLEC-2 also elicits the expansion of LECs in the LN (25, 26). In addition to PDPN engagement, DC initiation of LEC proliferation also occurs by inducing vascular endothelial growth factor (VEGF) production by FRCs in the lymph node (8, 9). Early LEC proliferation appears to be independent of T and B cells, as transfer of bone marrow derived dendritic cells into a mouse that lacks T or B cells can elicit LEC expansion at early timepoints following an immune stimulus (6). Thus, DC-stromal interactions are part of the initial step in the expansion of the LN following an inflammatory insult (**Figure 1B**).

In addition to DCs, macrophages have an important role in regulating lymphangiogenesis in response to inflammation through the production of VEGF, specifically, VEGF-C, VEGF-D, and to a lesser extent, VEGF-A (7). These CD11b+ macrophages accumulate around lymphatic vessels in the draining LN following inflammation in the skin, resulting in LEC proliferation (7). Importantly, following clodronate macrophage depletion, lymphangiogenesis was markedly decreased in the draining LN, demonstrating that macrophages participate in inflammationinduced LEC proliferation (7) (**Figure 1B**). These data are consistent with macrophages inducing lymphangiogenesis in non-lymphoid organs such as the cornea, peritoneum, and skin (27–29). Neutrophils have been shown to participate in LEC division in the skin (30), both through the production of VEGF-D and by increasing the bioavailability of VEGF-A via MMP-9 and heparanase. However, whether neutrophils or other innate cells contribute to LN LEC expansion is still unclear (30).

Although the mechanisms by which innate immune cells influence LN LEC expansion during an inflammatory response have been fairly well-studied, less is known about the transcription and signaling that occur within the LEC. The primary signals that LECs receive to induce division include VEGF receptor (VEGFR) engagement as described above. However, other factors are involved, including IL-7 which is important for LEC remodeling (11) (**Figure 1C**). Perhaps not surprisingly, the transcriptional program that ensues in LECs sorted from LNs at 12 h after lipopolysaccharide stimulation suggests that the LECs recruit immune cells through chemokine expression (CXCL9 and CCL5) (10). LECs also increase IFN inducible gene expression (Mpeg1, Lcn2, Irf7, IFI44, and Ly6a among others) at this time point (10). How LN LECs are transcriptionally regulated during inflammation to directly control LEC division could be through the immediate downregulation of genes that regulate cell division, Ccna2 and Klhl9 (10). This downregulation of genes involved in division could be due to the response to IFNγ induced by lipopolysaccharide and may be part of the mechanism behind how IFNα or IFNγ inhibit LEC division (5, 14) (**Figures 1A,C**). Furthermore, following sorting of LEC populations 6 days after polyI:C injection, non-dividing [(programed death ligand 1(PD-L1hi)] LECs express more CXCL4 (an angiostatin), while dividing (PD-L1lo) LECs express more growth and differentiation factor 10 (GDF10) and integrin beta 1 (ITGB1), both of which are important for angiogenesis (5, 31–36). Intriguingly, VEGFR3 expression by LECs in the LN was unchanged. However, as LECs generally express high levels of VEGFR3 (10, 37), it seems likely that upregulation of VEGFR3 may not be required to induce LEC division, but instead that LECs require the upregulation of VEGF as described in detail above.

PD-L1 has also been shown to be involved in determining which LECs divide. In both mice that are PD-L1 deficient and mice in which the non-hematopoietic cells lack PD-L1, LEC division was significantly increased at 6 days after polyI:C injection (5). The mechanism behind how PD-L1 could regulate division was at least partially attributed to lost CXCL4 expression in Pdl1−/<sup>−</sup> LECs (5) (**Figure 1C**). Although CXCL4 was identified as a potential downstream target of PD-L1, much work needs to be done to determine the signaling pathway of PD-L1, and how it regulates division. These new findings indicate that PD-L1 may have a primary function in coordinating LN LEC expansion and survival during inflammation. Therefore, LECs recruit immune cells, receive signals from DCs and macrophages to divide (**Figure 1B**), change their transcriptional profile and divide based on expression of PD-L1 (**Figure 1C**) during the early phase (0–48 h) of an inflammatory response.

### REGULATION OF LEC EXPANSION BY THE ADAPTIVE IMMUNE SYSTEM DURING INFLAMMATION

While DCs and macrophages contribute to LEC division at early timepoints during an immune response (**Figure 1B**), B cells have been shown to influence LEC division at the peak of the immune

FIGURE 1 | Regulation of LN LEC expansion and contraction during an immune response. (A) Control of initiation of LEC expansion by innate immunity: 0–24 h. At early timepoints following an immune stimulus, type 1 IFN production inhibits LEC expansion. DCs traffic to the LN in response to CCL19 and CCL21. (B) Control of initiation of LEC expansion by innate immunity: 24–48 h. DCs induce the production of VEGF from FRCs to initiate LEC division. CD11b+ macrophages accumulate along lymphatic vessels in the LN, producing VEGF (5–9). (C) Intrinsic mechanisms controlling LEC expansion. LECs produce IL-7, which increases LEC division. PD-L1 expression on LECs inhibits LEC expansion, likely through increasing expression of CXCL4 which is a negative regulator of cell division. PD-L1 expression is controlled by type 1 IFN, resulting in PD-L1 upregulation at early timepoints following an immune stimulus, as well as the expression of other IFN inducible genes. LECs induce expression of chemokines, including CXCL9 and CCL5 following an immune stimulus (5, 10, 11). (D) Control of LEC expansion by adaptive immune cells: 4–7 days. At later timepoints during an immune response B cells interact with FRCs, and then produce VEGF to increase LEC expansion. T cells contribute to LEC expansion. (E) Control of LEC contraction by adaptive immune cells: 8–14 days. T cells produce type 2 IFN to inhibit LEC expansion and induce LEC apoptosis (8, 12–16).

response (**Figure 1D**). Following immunization with complete Freund's adjuvant, B cell recruitment to the lymph node was required for LEC expansion. In a mouse model where B cells lack L-selectin, an adhesion molecule necessary for lymphocyte migration across high endothelial venules in the LNs, LEC expansion was impaired due to the loss of VEGF-A production in the follicle (12). Intriguingly, utilizing in vitro modeling, this group also showed that activated B cells likely produce VEGF-A in the LN only during inflammation (12). Indeed, another study found that inducing the expression of VEGF-A by B cells led to an increase in LN lymphangiogenesis, as well as enlargement of the LN (13). Recently, Dubey et al. showed B cells interact with lymphotoxin-beta receptor (LTβR) on FRCs which results in the production of B cell activating factor (BAFF). In combination with IL-4, production of BAFF causes B cells to produce VEGF-A and C (16). Together, these data suggest B cell production of VEGF-A or C can influence LN LEC expansion, but may not be required (15) (**Figure 1D**).

Others have shown that in addition to B cells, T cells are also involved in LN and LEC division. First, the lack of both B and T cells led to an almost complete loss of vascular-stromal expansion at later timepoints following complete Freund's adjuvant (8). When only T cells were absent, LEC proliferation was impaired, but surprisingly the absence of T cells did not affect total LEC numbers after complete Freund's adjuvant (8). Other work has also shown a role for T cells in regulating LEC expansion. In a mouse lacking endogenous T or B cells, T cell receptor transgenic T cell transfer did not lead to LEC expansion after immunization, unless the transferred T cells were activated with their cognate antigen (15). Thus, a functional T cell response, in the absence of B cells, is enough to induce LEC expansion following immunization. These data highlight the importance of the adaptive immune response in regulating LEC expansion during late time points (4–7 days) after an inflammatory stimulus (**Figure 1D**).

#### LEC APOPTOSIS AND LN CONTRACTION DURING RESOLUTION OF THE IMMUNE RESPONSE

While LEC expansion is important for coordinating the immune response, LEC contraction must also occur during the resolution of the immune response. Very little has been done to understand how this process occurs, however, in an athymic mouse, LN lymphatic vessel density is dramatically increased (14). This hypertrophy of lymphatic vessels is reduced by IFNγ production by T cells (14). Furthermore, when IFNγ was absent, lymphatic vessel regression did not occur as it normally does during LN contraction (14). This suggests that the production of IFNγ by T cells may be important for inhibiting lymphatic growth and/or promoting LEC apoptosis (**Figure 1E**). Interestingly, recent data looking at stromal cells, including LECs, 15 days after lymphocytic choriomeningitis virus, showed increased expression of the chemokines CXCL9 and CXCL10, as well as the activation marker Nur77 (38). While lymphocytic choriomeningitis virus was cleared by this time, LECs remain activated. This could be a process in which LECs recruit IFNγ producing cells until the regression of the lymphatic vasculature and LN size returns to normal.

While not directly regulating LEC contraction, PD-L1 does appear to specifically control LEC survival. These findings predict that PD-L1 may determine which LECs undergo apoptosis during LN contraction (5) (**Figure 2A**). This is consistent with other data showing that PD-L1 can act as a negative regulator of apoptosis in other endothelial cells (43), a process which may be hijacked by cancer cells (44–46). As such, loss of the cytoplasmic domain of PD-L1 in cancer cells resulted in increased apoptosis, from either T cell mediated killing, administration of a chemotherapeutic agent, or interferon beta cytotoxicity (44–46). Although the cytoplasmic domain of PD-L1 is relatively short, it appears that there are at least two signaling domains that help regulate inhibition to apoptosis in response to type 1 interferon, and mutation of these domains can sensitize cancer cells to interferon alpha/beta cytotoxicity (46). While these studies were done in the context of cancer cells which hijack normal cellular functions, recent data suggests that the expression of PD-L1 by LECs, and the regulation of cellular division and survival, may be a normal physiologic role for PD-L1. Further work is needed to determine the precise signaling pathways by which PD-L1 regulates survival, and if this process differs between endothelial cells and cancer cells (**Figure 2A**).

### LECS BALANCE OPPOSING ROLES DURING AN IMMUNE RESPONSE

While LEC expansion and contraction in the lymph node is important for the immune response, LECs also have a major role in programming the adaptive immune response. As stated above, some LECs in the lymph node express PD-L1 at high levels (47). The role of PD-L1 expression by other cells has been well-described as being inhibitory for T cell activation when programmed death-1 expressed on T cells binds to PD-L1 (48–51). LECs are involved in the maintenance of peripheral T cell tolerance, via expression and presentation of tissue specific antigens, such as tyrosinase (39, 40, 42, 47). Loss of PD-L1 expression by LECs that express tyrosinase results in autoimmune vitiligo (39, 40). In addition, LECs are capable of inducing CD4+ T cell tolerance, through major histocompatibility complex (MHC) class 2. Interestingly, while LECs do not express functional MHC class 2 (42), they are able to acquire loaded MHC class 2—self antigen peptide complexes from DCs (41). These LECs then present the self-antigen to CD4+ T cells which results in anergic self-antigen specific T cells (41). Loss of PD-L1 or loss of LEC acceptance of MHC class 2 complexes leads to autoimmunity (41, 42) (**Figure 2A**). Furthermore, in tissue lymphatics, PD-L1 also plays a major role in inducing tumor tolerance by LECs (52, 53).

LECs also have a role in the maintenance of protective memory through the archiving of foreign antigens following both immunization and viral infection (15, 26). LECs that acquire antigens during an inflammatory response do not present the foreign antigen directly. Instead, LECs archive antigens, and hand off the antigen to migratory DCs either directly or via LEC apoptosis (**Figure 2B**). The migratory DCs (26) then present the antigen to CD8+ T cells, improving the effector T cell response upon re-challenge (15). Importantly, antigen archiving is decreased without an inflammatory signal, and other groups have shown that LECs can present foreign antigens in a tolerizing manner when inflammation does not occur, in a process called foreign antigen scavenging

(54). In this case, LECs directly present the antigen to CD8+ T cells and the CD8+ T cells are deleted (54). These recent findings illustrate the flexibility of LECs in programming immune responses in the LN and highlight differences between inflammatory and non-inflammatory responses.

What mechanisms regulate the different functions of LECs during the transition from homeostasis to an immune response are not fully understood. An inflammatory signal is needed to prevent LECs from presenting foreign antigen in a tolerizing manner, and a role for LEC expansion has been described in antigen archiving. Following immunization, LECs will acquire, but will not archive antigen unless the LECs are expanding (15). These data indicate that there may be a role for LEC expansion in regulating the function of the lymphatic network during an acute, inflammatory response. What controls LEC expansion during a memory immune response is virtually unknown. However, it seems likely that the immediate production of cytokines, such as IFNγ by memory T cells, would inhibit LEC expansion, similar to the role of effector T cell production of IFNγ inducing LN contraction (14). Interestingly, PD-L1 expression is controlled by both type 1 and type 2 IFN (5), therefore it is possible that, in the absence of division, the upregulation of PD-L1 may be a mechanism to prevent improper activation of auto-reactive T cells and B cells while the LN is preparing to respond to the inflammatory insult (**Figure 2**).

### SUMMARY

Further research is needed to fully explore novel regulators of LEC expansion and contraction, including PD-L1 and CXCL4. How LECs function to both maintain peripheral tolerance and promote protective immunity is also not wellunderstood. Understanding these processes, and how LECs can determine the fate of the immune response, are likely very important in the prevention of autoimmunity as well as the development of a strong memory response. Therefore, future studies of LN LECs during an active immune response may lead to novel therapeutic targets in a wide range of diseases.

## AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work was supported by a grant from the National Institutes of Health (R01 AI121209) to BT and the National Institutes of Health (T32 AI007405) to EL.

#### REFERENCES


#### ACKNOWLEDGMENTS

We would like to thank Dr. Matthew Burchill, Dr. Colm Collins, and Dr. Anne Tye for critical reading of the manuscript.


nuclear expression in breast cancer cells: role of B7-H1 as an anti-apoptotic molecule. Breast Cancer Res. (2010) 12:R48. doi: 10.1186/bcr2605


**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 © 2019 Lucas and Tamburini. 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(s) 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.

# Modulation of Immunity by Lymphatic Dysfunction in Lymphedema

Yinan Yuan1†, Valeria Arcucci 1†, Sidney M. Levy <sup>1</sup> and Marc G. Achen1,2 \*

<sup>1</sup> Tumour Angiogenesis and Microenvironment Program, Peter MacCallum Cancer Centre, Melbourne, VIC, Australia, <sup>2</sup> Sir Peter MacCallum Department of Oncology, University of Melbourne, Parkville, VIC, Australia

The debilitating condition known as secondary lymphedema frequently occurs after lymphadenectomy and/or radiotherapy for the treatment of cancer. These therapies can damage lymphatic vessels leading to edema, fibrosis, inflammation and dysregulated adipogenesis, which result in profound swelling of an affected limb. Importantly, lymphedema patients often exhibit impaired immune function which predisposes them to a variety of infections. It is known that lymphadenectomy can compromise the acquisition of adaptive immune responses and antibody production; however the cellular mechanisms involved are poorly understood. Here we discuss recent progress in revealing the cellular and molecular mechanisms underlying poor immune function in secondary lymphedema, which has indicated a key role for regulatory T cells in immunosuppression in this disease. Furthermore, the interaction of CD4<sup>+</sup> T cells and macrophages has been shown to play a role in driving proliferation of lymphatic endothelial cells and aberrant lymphangiogenesis, which contribute to interstitial fluid accumulation in lymphedema. These new insights into the interplay between lymphatic vessels and the immune system in lymphedema will likely provide opportunities for novel therapeutic approaches designed to improve clinical outcomes in this problematic disease.

#### Edited by:

Amanda W. Lund, Oregon Health and Science University, United States

#### Reviewed by:

Pauliina Hartiala, Turku University Hospital, Finland Michael Detmar, ETH Zürich, Switzerland

#### \*Correspondence:

Marc G. Achen marc.achen@petermac.org

†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: 31 October 2018 Accepted: 11 January 2019 Published: 29 January 2019

#### Citation:

Yuan Y, Arcucci V, Levy SM and Achen MG (2019) Modulation of Immunity by Lymphatic Dysfunction in Lymphedema. Front. Immunol. 10:76. doi: 10.3389/fimmu.2019.00076 Keywords: immune function, inflammation, lymphedema, regulatory T cells, T-helper cells

### INTRODUCTION

The lymphatic system is a highly structured vascular network, important for interstitial fluid homeostasis, immune surveillance and lipid absorption, which consist of distinct types of lymphatic vessels. Interstitial fluid is absorbed by highly permeable initial lymphatics and transported by lymphatic pre-collectors to lymphatic collectors, which converge to form lymphatic trunks that ultimately transport lymph to the venous system via lymphatic ducts. Each type of lymphatic vessel is anatomically specialized for its function (1), but all lymphatic vessels share the feature of being lined by a single layer of lymphatic endothelial cells (LECs). Lymphatic vessels can undergo a variety of remodeling processes in development and disease, including lymphangiogenesis (the growth of new lymphatic vessels), which have important implications for lymphatic biology and immune function (2). Much progress has been made over recent years in defining the effects of lymphatic remodeling, lymphangiogenesis and LECs on immune function, particularly in the setting of cancer [for example see (3–5)]. More specifically, the establishment of tumor-associated immunity is thought to depend on lymphatic vessel remodeling and drainage. Further, there is emerging evidence that LECs are important for the maintenance of peripheral tolerance, modulating effector T cell responses and influencing leukocyte function (6). Here we review the role of lymphatic vessels in modulating immunity in secondary lymphedema, a prevalent condition caused by lymphatic dysfunction, which involves remodeling of lymphatic vessels and compromised immune function.

### SECONDARY LYMPHEDEMA: CLINICAL ASPECTS AND PATHOPHYSIOLOGY

Secondary lymphedema is a chronic disease characterized by the accumulation of interstitial fluid in tissues due to damaged lymphatic vessels, leading to swelling, and dysfunction of limbs (7). It is an acquired condition that is etiologically distinct from primary, or hereditary, lymphedema, which is a rare disease caused by intrinsic abnormalities of lymphatic function due to defects in genes involved in the growth and development of the lymphatic vasculature (8). Secondary lymphedema is a slow but progressive condition which can be caused by trauma, infection and inflammation. Globally, the most common cause is lymphatic filariasis, due to lymphatic vascular invasion by filarial nematodes, which has been estimated to afflict 68 million people in 73 countries worldwide (9). However, disruption of the lymphatic vasculature due to surgical interventions (e.g., lymphadenectomy) and/or radiotherapy for breast cancer is the most common cause in the developed world (7), with an incidence of lymphedema of 21% among women who were diagnosed with breast cancer (10). The condition can occur not only in breast cancer patients, but in patients with any cancer types which require lymph node dissection or radiotherapy treatment, such as head and neck, genitourinary and gynecological cancers, and melanoma (11). The onset of secondary lymphedema can be highly variable and has been reported to occur immediately postoperatively or up to 30 years post-treatment in the context of breast cancer, and it is not clear what determines a patient's predisposition to develop the disease. Historically, secondary lymphedema has been considered underdiagnosed, and robust epidemiological data have been scant, however, it is clearly a relatively prevalent condition with between two and five million people estimated to suffer from it in the United States (12).

Secondary lymphedema can be highly debilitating both physically and psychologically for patients due to the reduced quality of life associated with limb discomfort, anxiety, depression, sexual dysfunction, and social isolation. Current treatment choices for lymphedema include massage, manual lymph drainage (13), remedial exercise (14), compression bandaging (15), electrophysical modalities (including lowlevel laser therapy and electrical stimulation) (16), elevation techniques, exercise programs, and dietary/weight loss interventions (17, 18). There are also a range of surgical options such as liposuction (19), various forms of vascular anastomosis (20), lymph node transplantation (21) and other regional tissue transfer procedures. Unfortunately these treatments have not proven to be reliably curative as they have limited efficacy in controlling the disease, and many do not address the cause. Notably, there are no molecular-based therapies for the condition although therapeutics targeting inflammation (ketoprofen) (22) or promoting lymphangiogenesis [Lymfactin and Ubenimex also known as Bestatin (23, 24)] are being tested for treatment of lymphedema in clinical trials programs-see ClinicalTrials.gov for further information about these trials (ClinicalTrials.gov identifiers: NCT02257970, NCT02700529, and NCT02994771). Given the absence of curative treatment options and prevalence of the condition, secondary lymphedema is considered an important unmet clinical need in medicine.

The pathological features of secondary lymphedema include edema, inflammation, dermal fibrosis and formation of fat tissue, and patients often exhibit impaired immune function predisposing them to a variety of infections (25). These features are thought to further restrict lymphatic function in lymphedematous tissue thereby establishing a vicious pathophysiological cycle (26). The types of infections observed in secondary lymphedema include cellulitis involving the deeper dermis and subcutaneous fat, erysipelas involving the superficial dermis and lymphangitis involving the superficial dermal lymphatics. Soft-tissue infections associated with secondary lymphedema can lead to sepsis and, on occasions, death (27). Therefore, patients can require lifelong prophylactic antibiotic therapy. Given the clinical management of lymphedemaassociated infections can be highly problematic, it is important to understand how the immune response is impaired by the lymphatic injury which underlies secondary lymphedema. Such understanding could provide opportunities for prevention or improved treatment of secondary lymphedema.

### IMMUNOLOGICAL VULNERABILITY OF SECONDARY LYMPHEDEMA

Extensive clinical literature and experience has made it clear that lymphedematous tissue is immunologically vulnerable. Not only infections, but also neoplasms and immune-related disorders, such as neutrophilic dermatosis and toxic epidermal necrolysis, occur more frequently than in normal tissue [for example see (28)]. Chronic secondary lymphedema is typically characterized by an altered abundance of immune cells. Clinical studies have demonstrated increased numbers of lymphocytes, plasma cells, macrophages, dendritic cells, and neutrophils in the affected skin and subdermal tissue of lymphedema patients (29, 30). Such immune cell populations can be important for development of lymphedema, for example a CD4<sup>+</sup> cell inflammatory response and T-helper 2 (Th2) cell differentiation can contribute to key pathological changes including fibrosis and lymphatic dysfunction (**Figure 1**) (31). The accumulation of macrophages and lymphocytes in lymphedematous tissue can be induced by lymphatic fluid stasis in animal models of lymphedema (32). Importantly, there is evidence from animal models that lymphatic vascular defects can be associated with inadequate humoral immunity (33) which is consistent with clinical studies in lymphedema patients which showed that vaccination in lymphedematous tissue was associated with significantly decreased antibody titres (34).

Lymphedematous tissue constitutes a highly abnormal environment from the perspective of immune function. It is to be expected that chronic lymph stasis in lymphedema would impair local immune surveillance by restricting the trafficking of immunocompetent cells in lymphedematous tissues. Further, the irregular accumulation of immune mediators (cytokines and chemokines) in lymphedematous tissue could be an initiating factor promoting activation of LECs and immune cells (35). It is

well-known that lymphatics can respond to immune mediators produced by macrophages, including TNF-α, IL-1β, and IFNγ (36), and are able to produce immune mediators which regulate macrophage function such as IL-6 and CCL21 (35, 37). It has also been shown that macrophages produce inducible nitric oxide synthase which causes a reduction of lymphatic vessel contraction (38) and thus may contribute to accumulation of tissue fluids and impairment of antigen transport to lymph nodes in lymphedema (**Figure 1**). In summary, it is likely that lymphatic activation perpetuates abnormal activation of macrophages, and visa-versa, which could contribute to immune dysfunction and abnormal inflammation in lymphedema. Nevertheless, the cellular mechanisms which cause immune deficits in lymphedema have begun to emerge only recently.

### MECHANISMS OF IMMUNE DYSFUNCTION IN SECONDARY LYMPHEDEMA

A recent study exploring the function of T cells in lymphedema demonstrated a major increase in regulatory T cells (Tregs) in the lymphedematous extremity, compared to contralateral control tissue, in patients with breast cancer-related lymphedema (39). This finding was replicated in a mouse model of axillary lymph node dissection which showed increased infiltration of both CD4<sup>+</sup> T cells and Tregs. In this model, it was shown that Treg proliferation was localized to the tissue distal to the area of lymphatic injury caused by the surgery. Further analyses suggested that the Tregs downregulated local tissue inflammation post-lymphatic injury, and inhibited acquisition of T-cell-mediated immune responses (39). The loss of draining lymph nodes was also thought to diminish these responses. In addition, Tregs impaired bacterial phagocytosis, regulated humoral responses and compromised dendritic cell activation in this model after lymphatic injury (39). Overall, Tregs impaired both innate and adaptive immune responses, and depletion of these cells restored immune-mediated responses, indicating an important role for these cells in local immunosuppression in lymphedema (39) (**Figure 1**).

Immune function has been studied in transgenic mice expressing a soluble form of the lymphangiogenic receptor VEGFR-3 in skin (K14-VEGFR-3-Ig mice). These mice lack small dermal lymphatic vessels, develop lymphedema, and provide a model in which to monitor immune responses in the setting of lymphatic insufficiency (40). These mice produced lower antibody titres in response to dermal immunization, which was not due to compromised function of B cells, but was thought to be due to physiological differences in antigen transport to draining lymph nodes (33). T cell responses to dermal vaccination were delayed in these mice, although these responses were nevertheless robust. T-cell-mediated contact hypersensitivity (CHS) responses were strong, but the ability of these transgenic mice to induce CHS tolerance in the skin was impaired (33). The mice also exhibited hallmarks of autoimmunity, including antibody deposits in the skin, which supports the concept that lymphatic drainage to lymph nodes is important for maintaining immune tolerance against peripheral antigens. These findings provide mechanistic insight into how compromised lymphatic drainage in lymphedema plays a role in regulating humoral immunity and peripheral tolerance (33).

The effect of re-introducing lymph nodes, post-lymphatic damage, on immune responses and development of secondary lymphedema was monitored by Huang et al. in a mouse model of lymphatic ablation and popliteal lymph node dissection (41). Lymph node transplantation in this model led to a decreased accumulation of perilymphatic inflammatory cells, increased dendritic cell trafficking from the periphery to the inguinal node, and markedly improved adaptive immune responses. These changes were accompanied by decreases in hindlimb swelling and fibroadipose tissue deposition, as well as a pronounced lymphangiogenic response. The findings from this model may have clinical relevance for improving immune function postlymphatic damage, given lymph node transfer is being used in human patients and is being developed in animal models in combination with lymphangiogenic growth factor therapy (42–44).

### EFFECTS OF IMMUNE CELLS ON LYMPHEDEMA PATHOPHYSIOLOGY

The involvement of CD4<sup>+</sup> T cells in lymphedema pathogenesis was studied by Ogata et al. who employed a mouse model of lymphedema based on ligating the major collecting lymphatic vessels in the skin of the abdomen and removing the associated axillary lymph node (30). This model exhibited excessive generation of immature lymphatic vessels that was essential for the early emergence of edema and the subsequent development of lymphedema pathology. CD4<sup>+</sup> T cells interacted with macrophages to promote lymphangiogenesis in this model, and both lymphangiogenesis and edema were reduced in macrophage-depleted or CD4<sup>+</sup> T-cell-deficient mice. From a mechanistic perspective, Th1 and Th17 cells activated macrophages to produce the lymphangiogenic growth factor VEGF-C, which likely drove the aberrant lymphangiogenesis. Inhibition of this mechanism suppressed both early lymphangiogenesis and development of lymphedema (30). Macrophages have also been reported to restrict fibrosis as depletion of these cells in a mouse model of secondary lymphedema significantly increased fibrosis, and impaired lymphatic transport, decreased VEGF-C expression and promoted Th2 differentiation (45). Th2 cells may also be involved in lymphedema pathogenesis as neutralization of two cytokines produced by these cells, IL-4 and IL-13, in a mouse model of secondary lymphedema promoted lymphatic function and restricted fibrosis (31). The role of CD4<sup>+</sup> T cells was also studied in a mouse model of secondary lymphedema by use of adoptive transfer techniques in CD4-deficient mice that underwent excision of skin and lymphatics in the tail or dissection of popliteal lymph nodes (46). This study revealed naïve CD4<sup>+</sup> T cells were activated in skin-draining lymph nodes and then migrated to lymphedematous skin. These activated cells promoted fibrosis and inflammation, and inhibited lymphangiogenesis and lymphatic function. Importantly, use of a sphingosine-1-phosphate receptor modulator to block release of T cells from lymph nodes prevented lymphedema in the mouse tail model employed in this study (46). It is now clear that CD4<sup>+</sup> T cells play a major role in the development of secondary lymphedema (**Figure 1**), at least in animal models, although the effect of these cells on lymphangiogenesis in lymphedema differed in different mouse models and therefore requires further clarification.

In a separate study, RNA sequencing of lymphedematous mouse skin suggested an upregulation of many T cell-related networks (47). More specifically, upregulation of Foxp3, a transcription factor specifically expressed by Tregs, indicated a potential role for these cells in lymphedema, consistent with findings discussed in the previous section. While global deletion of CD4<sup>+</sup> cells restricted lymphedema development in the mouse tail lymphedema model used in this study, targeted depletion of Tregs led to exacerbated edema associated with increased infiltration of immune cells and a mixed Th1/Th2 cytokine profile (47). Conversely, expansion of Tregs in the mouse model restricted lymphedema development. Therapeutic use of adoptively transferred Tregs upon lymphedema establishment reversed the major hallmarks of lymphedema such as edema, fibrosis and inflammation, and promoted lymphatic drainage (47). These findings on the role of Tregs are supported by the study of Garcia Nores et al. which showed that depletion of Tregs up-regulated local tissue inflammation after lymphatic injury (39). However, this study also showed that Tregs can locally impair adaptive immunity and clearance of bacteria after lymphatic injury. While it is clear that the number of Tregs in lymphedematous tissue is increased compared to normal tissue, the functional significance of these cells for development of lymphedema may differ depending on the relative importance of inflammation vs. adaptive immunity in the lymphedema model employed. This issue needs to be considered when assessing if Treg application could be a potential new therapeutic approach for treating lymphedema.

### CONCLUDING REMARKS

Recent studies using animal models and clinical samples have established that immune function is significantly compromised in secondary lymphedema, and demonstrated that a variety of T-cell-related networks are up-regulated in this condition. Tregs, in particular, are increased in abundance in lymphedematous tissue and are thought to compromise immune function in this disease by promoting immunosuppression, although they can make a positive contribution by reducing the degree of inflammation. Further analysis of Treg function in secondary lymphedema is required to establish whether or not modulating the levels or function of these cells could be beneficial for prevention or treatment of this condition. This may need to be pursued in large animal models, as opposed to mice, to give a clearer picture of how targeting immune cells might be beneficial for lymphedema patients.

### AUTHOR CONTRIBUTIONS

YY and MA conceived the topic and outlined the paper. YY, VA, SL, and MA wrote the paper.

#### FUNDING

This work was supported by a Program Grant from the National Health and Medical Research Council of Australia, and the Operational Infrastructure Program of the Victorian Government.

#### ACKNOWLEDGMENTS

Thanks to Steven Stacker for helpful discussions.

## REFERENCES


47. Gousopoulos E, Proulx ST, Bachmann SB, Scholl J, Dionyssiou D, Demiri E, et al. Regulatory T cell transfer ameliorates lymphedema and promotes lymphatic vessel function. JCI Insight (2016) 1:e89081. doi: 10.1172/jci.insight.89081

**Conflict of Interest Statement:** MA is a shareholder in Opthea Ltd. and an Inventor on patents assigned to Vegenics Ltd.

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.

Copyright © 2019 Yuan, Arcucci, Levy and Achen. 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(s) 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.

# Inflammation and Lymphatic Function

#### Simon Schwager and Michael Detmar\*

*Institute of Pharmaceutical Sciences, Swiss Federal Institute of Technology, ETH Zurich, Zurich, Switzerland*

The lymphatic vasculature plays a crucial role in regulating the inflammatory response by influencing drainage of extravasated fluid, inflammatory mediators, and leukocytes. Lymphatic vessels undergo pronounced enlargement in inflamed tissue and display increased leakiness, indicating reduced functionality. Interfering with lymphatic expansion by blocking the vascular endothelial growth factor C (VEGF-C)/vascular endothelial growth factor receptor 3 (VEGFR-3) signaling axis exacerbates inflammation in a variety of disease models, including inflammatory bowel disease (IBD), rheumatoid arthritis and skin inflammation. In contrast, stimulation of the lymphatic vasculature, e.g., by transgenic or viral overexpression as well as local injections of VEGF-C, has been shown to reduce inflammation severity in models of rheumatoid arthritis, skin inflammation, and IBD. Strikingly, the induced expansion of the lymphatic vasculature improves lymphatic function as assessed by the drainage of dyes, fluorescent tracers or inflammatory cells and labeled antigens. The drainage performance of lymphatic vessels is influenced by vascular permeability and pumping activity, which are influenced by VEGF-C/VEGFR-3 signaling as well as several inflammatory mediators, including TNF-α, IL-1β, and nitric oxide. Considering the beneficial effects of lymphatic activation in inflammation, administration of pro-lymphangiogenic factors like VEGF-C, preferably in a targeted, inflammation site-specific fashion, represents a promising therapeutic approach in the setting of inflammatory pathologies.

Keywords: lymphatic vessels, lymphangiogenesis, inflammation, inflammatory bowel disease, arthritis, psoriasis, skin, inflammatory mediators

## INTRODUCTION

Inflammation is a defensive reaction of the organism against pathogens or irritants. It is characterized by the five cardinal symptoms of rubor (redness), calor (increased heat), tumor (swelling), dolor (pain), and functio laesa (impaired function), which are mostly mediated by the expansion and activation of blood vessels. Inflammation is commonly associated with the formation of new blood (angiogenesis) and lymphatic (lymphangiogenesis) vessels from the pre-existing vascular networks. Interestingly, while the activation of the blood vasculature has been reported to aggravate inflammation severity in a variety of disease models (1–3), lymphatic vessels generally appear to exert beneficial effects, possibly by improving the clearance of extravasated fluid, thus reducing edema formation and levels of pro-inflammatory mediators as well as numbers of immune cells.

#### Edited by:

*Beth Ann Tamburini, University of Colorado Denver, United States*

#### Reviewed by:

*Sophia Ran, Southern Illinois University School of Medicine, United States Girdhari Lal, National Centre for Cell Science (NCCS), India*

#### \*Correspondence:

*Michael Detmar michael.detmar@pharma.ethz.ch*

#### Specialty section:

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

Received: *01 November 2018* Accepted: *06 February 2019* Published: *26 February 2019*

#### Citation:

*Schwager S and Detmar M (2019) Inflammation and Lymphatic Function. Front. Immunol. 10:308. doi: 10.3389/fimmu.2019.00308*

**21**

This review provides an overview of studies investigating the role of lymphatic expansion and function in common inflammatory diseases such as skin inflammation, inflammatory bowel disease (IBD) and rheumatoid arthritis (RA). In addition, the known effects of inflammatory mediators on the lymphatic vasculature and commonly used mouse models are described.

The lymphatic vasculature is a hierarchically structured, one-way circuit composed of initial capillaries, which lack a continuous basement membrane and smooth muscle cell coverage, draining into larger, smooth muscle cell-covered collectors and ultimately lymph nodes. In the setting of inflammation, the lymphatic system is critically important, as it is needed to ensure tissue fluid homeostasis by draining the larger amounts of extravasated fluid originating from increasingly leaky, inflammatory blood vessels. Indeed, an increased interstitial fluid pressure has been found to lead to the dilation of initial lymphatic vessels, thus facilitating the entry of fluid and inflammatory cells into the lymphatic vasculature and thereby removal from the inflamed tissue (4). In addition, lymphatic vessels are crucial for immune surveillance, as they serve as main transport routes for cells and inflammatory mediators to lymph nodes, where immune responses are mounted.

The most-thoroughly characterized signaling axis involved in lymphatic expansion and development consists of the vascular endothelial growth factor receptor 3 (VEGFR-3) and its ligands VEGF-C and VEGF-D. VEGFR-3 is part of the receptor tyrosine kinase family and is expressed widely in vascular endothelial cells during embryonic development, but becomes strongly restricted to lymphatic endothelial cells (LECs) in the adult organism under physiological conditions (5).

VEGF-C is the main ligand of VEGFR-3 and induces proliferation and migration of endothelial cells (6, 7). It undergoes extensive post-translational proteolytic processing, which also regulates the molecule's binding properties. Fully processed VEGF-C binds VEGFR-3 and, albeit with a lower affinity, VEGFR-2 (8). A mutated form of VEGF-C in which the cysteine 156 is replaced with a serine (VEGF-C156Ser) selectively binds VEGFR-3 (9).

VEGF-D has been reported to induce proliferation of endothelial cells (10). In mice, VEGF-D exclusively binds VEGFR-3, while fully processed human VEGF-D may also bind VEGFR-2 (11).

In order to study the role of lymphatic vessels in different pathologies, various mouse lines with a modified VEGFR-3 signaling axis have been generated. In K14-VEGF-C mice, VEGF-C is overexpressed under the control of the keratin-14 promoter, resulting in elevated levels of the growth factor in the skin and an enlarged dermal lymphatic vascular network (12). A similar lymphatic hyperplasia has been observed in mice transgenic for VEGF-D (K14-VEGF-D) (13). In contrast, mice overexpressing a soluble form of VEGFR-3 in the skin (K14-VEGFR-3-Ig mice) lack dermal lymphatic vessels and develop edema in the feet and skin (14).

Apart from promoting or inhibiting lymphatic vascular expansion, the clearance capacity of lymphatic vessels is subject to regulation by various signals. Drainage performance is influenced by vascular permeability and pumping activity of lymphatic vessels. Mediators inducing increased lymphatic vessel permeability include TNF-α, IL-1β, histamine, and the VEGF-C/VEGFR-3 axis (15–17). Lymphatic contractions and thereby pumping are negatively regulated by various inflammatory mediators, including prostaglandins, histamine, and nitric oxide (NO), while VEGF-C has enhancing effects (18–21). NO regulates lymphatic vessel function via its effects on lymphatic smooth muscle cells leading to vasodilation. It is produced constitutively by the endothelial nitric oxide synthase (eNOS) under physiological conditions. In inflammation, however, its levels are elevated due to the higher expression of inducible nitric oxide synthase (iNOS) on immune cells and inflamed endothelium, which has been linked to reduced lymphatic contraction frequency (22).

### THE LYMPHATIC VASCULATURE IN INFLAMMATORY DISEASES

#### Skin Inflammation

A wide range of skin pathologies including psoriasis, atopic dermatitis, rosacea, and UV damage are characterized by pronounced and often prolonged inflammation. The lymphatic vasculature is often aberrant in inflamed skin; in human psoriatic plaques for example, lymphatic vessels are dilated and tortuous (23–25). Nevertheless, lymphatic dysregulation in the human disease has attracted comparatively little attention.

Multiple mouse models have been established to facilitate the study of these diseases in general and the role of the lymphatic vasculature in particular. A common model are K14-VEGF-A transgenic mice which overexpress VEGF-A under the control of the keratin-14 promoter, resulting in chronically elevated levels of said growth factor in the skin and a concomitant expanded, leaky blood vasculature. Homozygous mice spontaneously develop a chronic skin inflammation at the age of 6 months (26). In hemizygous mice, a contact sensitizer (e.g., oxazolone) can be used to trigger a contact hypersensitivity reaction (CHS), leading to a similar chronic inflammatory skin disease (24).

In wild-type mice, skin inflammation may be elicited by inducing CHS, exposure to UVB radiation, injection of bacterial antigens like LPS or application of pro-inflammatory agents such as tetradecanoylphorbolacetate (TPA) or imiquimod (27).

Using these models, skin inflammation has been extensively studied in mice and the lymphatic vasculature has been demonstrated to be functionally impeded in UVB-irradiated, chronically inflamed ear skin. Evans blue injected into the inflamed skin stained strongly dilated lymphatic vessels that were extremely leaky, indicating reduced drainage capacity (1).

#### Stimulation of Lymphatic Vessels in Skin Inflammation

Activating the lymphatic vasculature in the setting of skin inflammation has been associated with reduced disease severity (summarized in **Table 1**). In K14-VEGF-A mice that had been crossed with K14-VEGF-C mice and were undergoing chronic CHS of the ear skin, the lymphatic vasculature was expanded and the inflammation, as assessed by edema formation,

#### TABLE 1 | Effects of lymphatic vessel stimulation in inflammatory diseases.


*DSS, dextran sulfate sodium; LN, lymph node; LPS, lipopolysaccharide; LTA, lipoteichoic acid; LV, lymphatic vessel; MDP, muramyl dipeptide; TPA, tetradecanoylphorbolacetate.*

inflammatory cell infiltrate, and altered epidermal proliferation or differentiation, was significantly reduced compared to control inflamed K14-VEGF-A mice. Strikingly, the vascular expansion was accompanied by an improved lymphatic clearance function. Local injections of VEGF-C156Ser had similar disease-alleviating effects, indicating that VEGFR-3- rather than VEGFR-2-mediated signaling is mainly responsible for the observed anti-inflammatory effects (28). In agreement with this observation, local injections of VEGF-C156Ser also triggered a strong lymphangiogenic response and reduced inflammatory ear

#### TABLE 2 | Effects of lymphatic vessel inhibition in inflammatory diseases.


*DSS, dextran sulfate sodium; LN, lymph node; LPS, lipopolysaccharide; LTA, lipoteichoic acid, LV, lymphatic vessel; MDP, muramyl dipeptide.*

swelling and CD11b-positive immune cell infiltration in UVBirradiated ear skin inflammation (32).

These findings are in line with a different study investigating the role of macrophages and lymphatic vessels in cutaneous inflammation. K14-VEGF-C mice that were subjected to lipopolysaccharide (LPS)- or lipoteichoic acid (LTA)/muramyl dipeptide (MDP)-induced skin inflammation presented with an expanded dermal and lymph node lymphatic vasculature. In addition, inflammatory tissue swelling and skin reddening were reduced. While no difference in FITC-dextran clearance was found, inflammatory cell migration to the draining lymph nodes and the drainage of fluorescently labeled antigen was significantly accelerated in K14-VEGF-C mice. These effects appeared to be dependent on macrophages, as clodronate-mediated depletion of these cells reduced lymphangiogenesis and delayed inflammation resolution (29). An enhanced lymphatic drainage function due to lymphatic stimulation has also been reported in other studies, e.g., after repeated application of TPA to the back skin of K14- VEGF-C transgenic mice, in which a lymphatic-specific, nearinfrared tracer was cleared more rapidly than in wild-type mice (30). Similarly, in a study of acute skin inflammation, both K14- VEGF-C and, to a lesser extent, K14-VEGF-D transgenic mice had improved clearance of Evans blue out of UVB-irradiated ear skin (31). Moreover, these mice also had less inflammatory edema and reduced epidermal thickening in oxazolone- and UVB-induced skin inflammation. The reduction in inflammation was generally more pronounced in VEGF-C transgenic mice than in VEGF-D transgenic animals, indicating stronger antiinflammatory effects of VEGF-C (31).

#### Inhibition of Lymphatic Vessels in Skin Inflammation

In contrast to stimulation of the lymphatic vasculature, inhibiting lymphatic vessels has been shown to aggravate skin inflammation in several studies (summarized in **Table 2**). Antibody-mediated blocking of VEGFR-3 strikingly reduced the number of lymphatic vessels in the inflamed ear skin of K14-VEGF-A mice during a CHS reaction. At the same time, tissue swelling, epidermal thickening, keratinocyte proliferation and the numbers of CD8- and CD11b-positive cells were significantly increased, indicating a more severe inflammatory phenotype. Interestingly, blocking VEGFR-2 alone or in combination with VEGFR-3 alleviated inflammation, indicating that VEGFR-2-mediated inhibition of blood vessels is beneficial in skin inflammation and outweighs the detrimental effects of VEGFR-3 inhibition (28). Similarly, adenoviral overexpression of a soluble VEGFR-3 strongly reduced lymphangiogenesis in mice undergoing LPS- or LTA/MDP-induced skin inflammation, resulting in delayed inflammation-resolution, slower clearance of FITCdextran as well as FITC-labeled LPS, and reduced migration of inflammatory cells from the skin to the draining lymph nodes (29). Systemic, antibody-mediated inhibition of VEGFR-3 also led to increased edema formation and CD11b-positive cell numbers in UVB-irradiated ear skin (36).

#### Inflammatory Bowel Disease

The term inflammatory bowel disease (IBD) comprises Crohn's disease (CD) and ulcerative colitis (UC), which are characterized by a chronic inflammation of the digestive tract. While UC generally affects the colon and presents with superficial ulcerations of the mucosa and submucosa, CD may occur at any location in the gastrointestinal tract and often causes transmural inflammation. As in the case of skin inflammation, research has long been focused on changes in the blood vasculature and VEGF-A has been suggested as an important mediator of IBD (39, 40).

In human patients suffering from IBD, lymphangiogenesis, lymphatic vessel obstruction, dilation, and submucosal edema are commonly observed (41–44) and abnormalities in the lymphatic vasculature had already been recognized during the original characterization of CD (45). In addition to morphological alterations, the functionality of IBD-associated lymphatic vessels is reduced. A study in patients with CD employed injections of the lymphatic-vessel-staining Patent Blue V dye in the inflamed colon and demonstrated morphological aberrations and functional impairment of the lymphatic vasculature, which could be correlated with disease severity. Strikingly, following surgical intervention and inflammation regression, lymphatic vessel appearance reverted back to normal, indicating that lymphatic vessel function may be involved in IBD pathogenesis in humans (46). In line with this, a lower density of lymphatic vessels could be linked to an increased risk of CD recurrence (47).

A multitude of studies have been performed in mouse models of IBD, the two most-commonly used being IL-10 knockout mice and dextran sulfate sodium (DSS)-induced colitis. IL-10 deficient mice spontaneously develop colitis at the age of 10– 12 weeks, most likely due to the lacking anti-inflammatory and immunosuppressive activity of IL-10 (48, 49).

DSS-induced colitis relies on administration of the detergent DSS in drinking water, which damages the intestinal epithelium, most strongly in the distal colon, and compromises its barrier function, making the underlying tissue accessible to bacteria and associated substances. In order to model acute inflammation, mice are commonly given DSS for a certain amount of time (e.g., a week), for chronic inflammation, mice receive multiple cycles of DSS and intermittent regular drinking water (50, 51).

#### Stimulation of Lymphatic Vessels in Inflammatory Bowel Disease

Akin to skin inflammation, inducing the lymphatic vasculature is generally correlated with a reduction in inflammation severity (summarized in **Table 1**).

In IL-10 knockout mice as well as in animals undergoing DSS-induced colitis, adenoviral delivery of VEGF-C significantly increased lymphatic vessel density and was associated with a reduction in bodyweight loss and disease severity as assessed by stool consistency and presence or absence of fecal blood. Moreover, histological analyses revealed decreased submucosal tissue edema and inflammatory cell infiltration, while the proliferation of LECs was greatly increased. Quantification of Evans blue clearance out of inflamed distal colon tissue revealed an enhanced lymphatic drainage function, which was also reflected in an improved clearance of fluorescently labeled antigen-coated beads and an augmented inflammatory cell migration from the inflamed tissue to the draining lymph nodes. Similar to the observations in skin inflammation, depletion of macrophages by clodronate largely abolished the protective effects of VEGF-C (33). It has been suggested that VEGF-C may influence the cytokine balance in the inflamed colon. Indeed, in vitro experiments have shown VEGF-C to induce the upregulation of IL-10 by bone marrow-derived macrophages (33). In line with this, increased levels of IL-10 in combination with a reduction of IL-9, which is associated with intestinal barrier disruption, have been reported upon treatment with adenovirally delivered VEGF-C in mice undergoing DSS-induced colitis (52, 53).

#### Inhibition of Lymphatic Vessels in Inflammatory Bowel Disease

Blocking VEGFR-3 resulted in a worsened colitis in IL-10 knockout mice as well as DSS-treated animals in terms of the histological score (summarized in **Table 2**). Animals of both models presented with strongly reduced lymphatic vessel density and LEC proliferation upon VEGFR-3 inhibition. At the same time, lymphatic clearance of Evans blue and bacterial antigen as well as inflammatory cell mobilization to the draining lymph nodes were significantly reduced (33).

In a different, independent study, IL-10 knockout mice were treated with a blocking antibody to VEGFR-3. This resulted in enlarged and tortuous lymphatic vessels in the colon, increased submucosal edema and a higher leukocyte infiltration in the inflamed tissue as well as a higher disease severity score (37).

#### Rheumatoid Arthritis

Rheumatoid arthritis (RA) is a chronic inflammatory disease affecting the joints and characterized by episodic flares (54). In its chronic stage, RA is commonly associated with lymphadenopathy and a decrease in lymphatic drainage function, as shown for example by tracking the drainage of intradermallyinjected, radioactively labeled albumin from the forearm (55). Lymphangiogenesis is also commonly observed in the joints of human RA patients and has been reproduced in mouse models of the disease (56, 57).

Commonly used mouse models of rheumatoid arthritis include TNF-α transgenic mice and K/B × N mice. The former overexpress human TNF-α and spontaneously develop chronic progressive joint inflammation at the age of ∼4 weeks (58). K/B × N mice model the autoimmunity aspect of RA and are based on a mouse line transgenic for a T cell receptor specific for bovine ribonuclease. After breeding onto the NOD background, accidental recognition of a NOD-derived antigen triggers the onset of joint inflammation at 4 weeks after birth (59).

Lymphatic function has mainly been studied in these animals and a two-phase model has been proposed [reviewed in (60)]. In an initial "expansion" phase during joint inflammation, lymphangiogenesis and popliteal lymph node expansion with or without increased lymphatic vessel contractions limit the inflammatory response (57, 61, 62). During the following "collapse" phase, popliteal lymph nodes shrink and lymphatic vessel contractions as well as lymphatic drainage function decrease significantly. At the same time, the joint inflammation increases in severity (61, 63–65). Blocking TNF-α signaling resulted in increased lymphatic contractions and reduced joint inflammation (66).

Similar changes in lymph node characteristics have also been reported in human patients, where lymph node hypertrophy could be observed in the vast majority of patients suffering from active RA, while healthy individuals and patients in remission showed no lymph node alterations (67).

#### Stimulation of Lymphatic Vessels in Rheumatoid Arthritis

Stimulating the lymphatic vasculature has been associated with reduced disease severity in animal models of RA (summarized in **Table 1**). Adeno-associated viral (AAV) delivery of VEGF-C in the inflamed ankle joints of 6-week-old TNF-α transgenic mice partially reversed the inflammation-associated increase in synovial volume and significantly improved leg mobility. Histological analyses revealed that mice treated with VEGF-C had less cartilage and bone destruction than animals injected with a control vector. In chronic arthritis (mice at 5 months of age), lymphatic drainage of indocyanine green (ICG) out of the footpad was strongly decreased in TNF-α transgenic compared to wild-type mice. AAV-mediated delivery of VEGF-C significantly improved the clearance of ICG out of the paws and increased the number of lymphatic vessels in the pannus of the inflamed joint (34).

In an alternative approach, based on the observation that increased levels of NO in inflammation reduce lymphatic pumping, lymphatic vessel function was studied using inhibition of NOS. Local application of L-N6-(1-iminoethyl)lysine 5 tetrazole-amide (L-NIL), a moderately selective inhibitor of iNOS (68), in TNF-α transgenic mice with collapsed lymph nodes restored lymphatic contractions and strongly improved ICG transport from the footpad to popliteal lymph nodes, while Nωnitro-l-arginine methyl ester (L-NAME), an unspecific inhibitor of both eNOS and iNOS was not associated with beneficial effects (35). Although the impact on disease severity in these mice was not assessed in the study, it provides evidence that selective inhibition of iNOS might offer an alternative and clinically relevant approach for RA therapy.

#### Inhibition on Lymphatic Vessels in Rheumatoid Arthritis

Inhibiting the lymphatic vasculature led to worsened inflammation in mouse models of arthritis (summarized in **Table 2**). Injecting TNF-α transgenic mice that had developed joint inflammation with a VEGFR-3-blocking antibody for 2 months significantly reduced the number of lymphatic capillaries in the draining popliteal lymph nodes and inflamed ankles. Blocking VEGFR-3 also aggravated inflammation of the knee and ankle joints, as the increase in synovial volume over time as well as its absolute size were elevated in these animals compared to IgG-treated controls. Similarly, histological analyses of hematoxylin-eosin-stained sections revealed exacerbated inflammation after VEGFR-3 inhibition. Akin to the effects observed in chronic skin inflammation, blocking VEGFR-2 was associated with a reduced inflammatory reaction, as assessed by synovial volume and histological scoring. Lymphatic drainage function, as assessed by tracking the ICG signal in paws and draining popliteal lymph nodes following injection into the footpad, was dramatically reduced upon blocking VEGFR-3 (38).

#### The Effect of Inflammatory Mediators on the Lymphatic Vasculature

Inflammatory lymphangiogenesis is mostly mediated by VEGF-A and VEGF-C which are produced by keratinocytes and stromal cells like fibroblasts as well as immune cells, most importantly macrophages (69–71). Indeed, several inflammatory mediators have been found to induce VEGF-C transcription (72–74).

Macrophages are of critical importance, as demonstrated in a model of IBD and LPS-induced skin inflammation, where depletion of macrophages aggravated the inflammation (29, 33). While VEGFs are important for inflammation-induced lymphangiogenesis, there are many additional factors at play. IL-17, a crucial cytokine in the pathogenesis of psoriasis for example, has been shown to induce lymphangiogenesis in vitro and in cornea micropocket assays (75), and IL-8 promoted lymphangiogenesis in cell culture experiments and in an animal model of lymphedema (76). Similarly, inhibition of TGF-β, which mediates anti-inflammatory effects, supported lymphangiogenesis in a mouse model of peritonitis and in lymphedema (77, 78). In line with this, cytokines characteristic for TH2 cells like IL-4 and IL-13, which are often linked to inflammation resolution, inhibited lymphangiogenesis (79). Interestingly, several inflammatory mediators have anti-lymphangiogenic activity. Interferon-γ (IFN-γ), which is produced by activated T cells, decreased lymphatic vessel formation of both human and murine lymphatic endothelial cells in vitro as well as in mouse lymph nodes (15, 80). Likewise, TNF-α inhibited capillary formation and proliferation of mouse LECs, while IL-1β had no consistent effects on proliferation, but reduced barrier function of LECs (15). Indeed, inflammatory mediators not only influence lymphangiogenesis, but also impact lymphatic function more directly. Prostaglandins, IL-1β, IL-6, and TNF-α reduced lymphatic pumping frequency (81, 82). Similarly, inflammatory mediators affect lymphatic vessel permeability, as demonstrated in vitro by assessing the effect of a wide array of inflammatory mediators on rat lymphatic endothelial cell monolayers, where IL-6, TNF-α, and IFN-γ strikingly increased the permeability, probably by reducing vascular endothelial (VE)-cadherin expression (83). Few studies have addressed lymphatic vessel permeability in vivo, but results of those that have showed impaired barrier function as well as pronounced leakiness and reported VEGF-A as important mediator of these effects, possibly by signaling via VEGFR-2 (1, 84).

It is important to consider that cytokines and growth factors often have pleiotropic effects, making it challenging to distinguish between direct and indirect mechanisms. IL-17 for example has been reported to induce VEGF-D expression, thereby triggering lymphangiogenesis indirectly (75). The wide array of signaling

molecules involved in inflammation as well as their different and often pleiotropic effects on the lymphatic vasculature result in a highly complex network of signals which is still incompletely understood.

### CONCLUSIONS

The lymphatic vasculature represents a crucial, although often under-appreciated, player in inflammation. Lymphatic vessels serve as the main transport route for inflammatory mediators, fluid, antigen and immune cells, thus playing a pivotal role in inflammation initiation and resolution. Indeed, it has been controversial whether expansion of the lymphatic endothelium contributes to inflammation by facilitating transport of leukocytes to lymph nodes and mounting of immune responses, or whether lymphatic vessels support inflammation resolution by draining inflammatory mediators and cells from the site of inflammation. However, in recent years, a number of studies detailed above have reported alleviated inflammation severity following activation and/or expansion of the lymphatic vasculature (depicted for skin inflammation in **Figure 1**), thus indicating that promoting the lymphatic vasculature supports inflammation resolution and may represent a valid therapeutic approach. It should be considered, however, that VEGF-C/VEGFR-3 signaling itself might also account for some of the anti-inflammatory effects observed in VEGF-C transgenic mice, as it has been shown to reduce the production of pro-inflammatory cytokines and protect mice from septic shock (85).

Interestingly, the lymphatic vasculature is also affected by established standard therapies used for the treatment of inflammatory diseases, e.g., in RA, where blocking TNFα resulted in an increased lymphangiogenic response and increased lymphatic contractions in the inflamed tissue (66, 86). Other therapies aimed at blocking certain cytokines (e.g., IL-17 in psoriasis) may also exert parts of their antiinflammatory effects by modulating the lymphatic vasculature. Curiously, some anti-inflammatory agents have been associated with anti-lymphangiogenic activity. Glucocorticoids reduced lymphangiogenesis in cornea inflammation and chronic airway inflammation mediated by M. pulmonis infection (87, 88). In addition, prostaglandin E2, whose biosynthesis is inhibited by cyclooxygenase (COX)-blocking non-steroidal anti-inflammatory drugs (NSAIDs), has been reported to induce VEGF-C expression and lymphangiogenesis in the setting of lung cancer (73). Coherently, inhibition of COX-2 reduced tumor-induced lymphangiogenesis (89). A possible explanation for these findings could be that potent therapeutic agents inhibit inflammation strongly enough to also reduce the concomitant inflammation-induced lymphangiogenesis. Moreover, while prostaglandin E2 interferes with lymphatic expansion, it has also been reported to inhibit lymphatic function (81). Therefore, glucocorticoids and NSAIDs may improve lymphatic clearance despite reducing lymphangiogenesis. However, further studies are needed to thoroughly investigate these possibilities.

It is important to consider that immunomodulatory properties of the lymphatic endothelium, which have received increasing attention over the last decade, may explain the observed anti-inflammatory effects of lymphatic vessel induction at least partially. A good example is the receptor D6, which is highly expressed by lymphatic endothelial cells and scavenges inflammatory cytokines. Mice deficient for D6 suffered from more severe skin inflammation and colitis compared to wild-type animals (90, 91), hence, lymphatic expansion may increase the levels of D6 and accordingly lower the levels of inflammatory mediators in the inflamed tissue, resulting in reduced disease severity. However, the immunomodulatory roles of the lymphatic vasculature are outside the scope of this review.

Although VEGF-C has been associated with antiinflammatory effects in a variety of diseases as described above, its biological roles are highly complex and may be organ- and disease-dependent. In the setting of experimental obesity for example, transgenically overexpressed VEGF-C induced pro-inflammatory macrophage chemotaxis, increased weight gain and worsened metabolic parameters such as insulin resistance (92). In contrast, blockade of VEGF-C and VEGF-D by overexpression of a soluble form of VEGFR-3 reduced macrophage infiltration and improved insulin sensitivity in diet-induced obesity (93). Similarly, in tumor studies, VEGF-C has been reported to induce tumor lymphangiogenesis and stimulate the migration of macrophages (94), which may explain

#### REFERENCES


the observed increase in tumor metastasis in VEGF-C transgenic mice (95).

Applying VEGF-C in these diseases might be counterproductive and these findings therefore highlight the complexity of VEGF-C biology and emphasize the necessity of thoroughly evaluating possible beneficial and detrimental effects of VEGF-C in individual pathologies.

Considering all available data, the induction of lymphangiogenesis and activation of the lymphatic vasculature in the setting of inflammation appears to represent a potent therapeutic approach. It is therefore striking that this strategy has not been explored more thoroughly, let alone exploited clinically. A major obstacle has been the lack of clinically feasible delivery systems of lymphangiogenic factors. In a recent study, however, a targeted F8-VEGF-C fusion protein that specifically accumulates in the inflamed tissue was characterized and shown to reduce inflammation in two mouse models of skin inflammation, possibly filling this therapeutic gap (96).

#### AUTHOR CONTRIBUTIONS

SS designed and wrote the manuscript. MD designed and revised the manuscript.

#### FUNDING

Work in the authors' laboratory was supported by Swiss National Science Foundation grant 310030\_166490, and Advanced European Research Council Grant LYVICAM.


arthritis by accelerating local lymphatic drainage in mice. Arthritis Rheum. (2011) 63:2318–28. doi: 10.1002/art.30421


endothelial growth factor-C up-regulation: a novel mechanism of lymphangiogenesis in lung adenocarcinoma. Cancer Res. (2004) 64:554–64. doi: 10.1158/0008-5472.CAN-03-1301


**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 © 2019 Schwager and Detmar. 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(s) 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.

# Lymph Node Subcapsular Sinus Macrophages as the Frontline of Lymphatic Immune Defense

Dante Alexander Patrick Louie and Shan Liao\*

*Department of Microbiology, Immunology and Infectious Diseases, The Snyder Institute for Chronic Diseases, Cumming School of Medicine, University of Calgary, Calgary, AB, Canada*

Lymphatic vessels collect and transport lymph and pathogens to the draining lymph node (LN) to generate proper immune protection. A layer of macrophages that strategically line the LN subcapsular sinus (SCS) is directly exposed to the afferent lymph and are denoted as SCS macrophages. These macrophages are the frontline of immune defense that interact with lymph-borne antigens. The importance of these macrophages in limiting the spread of pathogens has been demonstrated in both viral and bacterial infection. In anti-microbial responses, these macrophages can directly or indirectly activate other LN innate immune cells to fight against pathogens, as well as activate T cells or B cells for adaptive immunity. As the first layer of immune cells embracing the tumor-derived antigens, SCS macrophages also actively participate in cancer immune regulation. Recent studies have shown that the LNs' SCS macrophage layer is interrupted in disease models. Despite their importance in fighting the spread of pathogens and in activating anti-tumor immunity, the mechanism and the immunological functional consequences for their disruption are not well-understood. Understanding the mechanism of these macrophages will enhance their capability for therapeutic targeting.

#### Edited by:

*Amanda W. Lund, Oregon Health and Science University, United States*

#### Reviewed by:

*Ferdinando Pucci, Oregon Health and Science University, United States Santiago F. Gonzalez, Institute for Research in Biomedicine (IRB), Switzerland*

#### \*Correspondence:

*Shan Liao liaos@ucalgary.ca*

#### Specialty section:

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

Received: *21 November 2018* Accepted: *11 February 2019* Published: *28 February 2019*

#### Citation:

*Louie DAP and Liao S (2019) Lymph Node Subcapsular Sinus Macrophages as the Frontline of Lymphatic Immune Defense. Front. Immunol. 10:347. doi: 10.3389/fimmu.2019.00347* Keywords: subcapsular sinus macrophage, CD169, lymph node (LN), free-floating antigens, virus, bacteria, cancer

## INTRODUCTION

The lymphatic system consists of two major parts: lymphatic vessels and lymph nodes (LNs). Lymphatic vessels are present throughout the body, acting as a road map for immune surveillance. These vessels are responsible for collecting interstitial fluid, soluble proteins, peptides, metabolites, invading pathogens, and immune cells in the tissue, then transporting the collected contents to the draining LNs via afferent lymphatic vessels (1–3). Initial lymphatic vessels (also named lymphatic capillaries) have discontinuous junction molecules which are highly permeable, and permit easy access of fluid and other content from peripheral tissues (4). Initial lymphatic vessels congregate to contractile lymphatic vessels, also known as collecting lymphatic vessels. Collecting lymphatic vessels direct lymph to the LN. Once in the LN, free-floating antigens, migrating antigen-presenting cells, and resident LN immune cells meet to initiate immune activation. After immune surveillance in the LN, efferent lymphatic vessels return lymph and activated immune cells to the circulation in order to enter the site of pathogen invasion for immune protection.

The transport of tissue-originated antigen-loaded antigen-presenting cells via lymphatic vessels has been largely studied. Migrating dendritic cells enter lymphatic vessels through the portals formed by the discontinuous basement membrane between adjacent endothelial cells, and is

**32**

dependent on CCR7 expression on dendritic cells (DC) and chemokines CCL19 and CC21 expressed on lymphatic endothelial cells (5–8). However, not all antigens transported in lymphatics are loaded on dendritic cells. Some free-floating lymph-borne antigens can travel with lymph to the LN. The importance of how LN-resident antigen-presenting cells react to free-floating antigens in lymph has been gaining more interest in the past decade. As lymph enters the LN, fluid fills the sinus lumen. Lining the floor of the sinuses are sinus macrophages that directly embrace the lymph coming from the afferent lymphatic vessels (**Figure 1**). These macrophages sample the freefloating antigens in the afferent lymph within several minutes after administration of model antigen tracers or pathogens (9, 10). Larger molecules and particles, such as viruses and bacteria, are captured by sinus macrophages (11–15) and sinus DCs (16). Smaller antigens, such as ovalbumin (OVA), can be captured by sinus macrophages and DCs. Additionally, smaller antigens can enter the LN conduits and are sampled by the LN conduit-associated DCs (17, 18). The first wave of DC activation occurs several hours before tissue-originated antigen-bearing DCs enter the LN, acting as another layer of protection in the event pathogens evade detection at the site of invasion (18– 21). In fact, even in the absence of tissue-migrating antigenpresenting cells, the LN-resident antigen-presenting cells are capable of generating a protective immune response against invading pathogens (16, 22, 23). Therefore, LN sinus resident macrophages function as a frontline of immune protection to lymph-borne pathogens.

During cancer lymphatic metastasis, metastatic tumor cells and tumor-derived antigens travel through lymphatic vessels to the tumor draining lymph node. Metastatic tumor cells were observed to first accumulate at the subcapsular sinus (24). LN metastatic tumor cells can invade the LN blood vessels as early as 2 days post-injection and spread to distant organs from the tumor draining LN (25, 26). Subcapsular sinus macrophages are the first layer of immune cells that are exposed to the metastatic tumor cells and tumor-derived antigens coming from the afferent lymphatic vessels. Studies in this field can reveal exciting new prospects when it comes to developing cancer immunotherapy. We reviewed the literature on how these macrophages are responsible for activating an immune response to the invading pathogens or tumor-derived antigens, as well as how the interruption of these macrophages in the LN is associated with disease.

#### LN SINUS MACROPHAGES

Sinus macrophages are not uniform across the entire LN; they can be subdivided into two major populations: the subcapsular sinus (SCS) macrophages and the medullary sinus macrophages according to their anatomical location in the LN (**Figure 1**). There are also sinus dendritic cells that sparsely populate the subcapsular sinus. Functionally, both sinus macrophages and DCs can acquire pathogen or particles from the passing lymph in the SCS. The sinus macrophages differ phenotypically, as SCS macrophages express

Mac1 (CD11b/CD18), Siglec-1 (CD169), but lack the expression of F4/80, a murine macrophage marker (27). On top of that, a small proportion of the SCS CD169<sup>+</sup> cells are CD169+CD11c+, indicating their DC phenotype (12, 16). Yet researchers still title these cells as macrophages, because, despite lacking the common F4/80 murine macrophage marker, SCS macrophage differentiation depends on the "macrophage colony-stimulating factor" cytokine, also known as CSF-1 (27–29). On the other hand, the phenotype of medullary sinus macrophages is more indicative of their macrophage characterization as they express F4/80 and Mac1. Some of the medullary sinus macrophages also express CD169 at a relatively lower level, therefore SCS macrophages are specifically distinguished as CD169+F4/80−, while medullary sinus macrophages are CD169+F4/80<sup>+</sup> or CD169low/−F4/80<sup>+</sup> (9).

Sinus macrophages also differ from each other functionally. Classically activated macrophages, known as M1 macrophages, typically produce pro-inflammatory cytokines, mediate pathogen resistance, and contribute to tissue destruction (30). This largely describes the medullary sinus macrophages, given their high lysozyme content and ability to process antigens, but no evidence has been shown for their capability to produce pro-inflammatory cytokines (31, 32). In contrast, SCS macrophages show relatively low phagocytic activity, but have demonstrated the ability to produce pro-inflammatory cytokines, namely type I interferon's (27, 33, 34). Therefore, while both sinus macrophages exhibit components of M1 macrophage function, a consensus on their categorization has yet to be reached in the field.

The origin and development of SCS macrophages has been studied to better understand their function. As stated earlier, the CSF-1/CSF-1 receptor signaling interaction is pivotal for the presence of SCS macrophages. Transgenic mice with a recessive osteopetrotic mutation (op/op) demonstrate a CSF-1 deficiency and show a significant reduction in SCS macrophages. Similarly, anti-CSF-1 receptor treatment to block the CSF-1 ligand from binding to CSF-1 receptor significantly depleted SCS macrophages, while medullary sinus macrophages remained intact (28). However, while medullary sinus macrophages are unaffected by blocking CSF-1/CSF-1 receptor interaction, CSF-1 receptor deficient mice show a significant depletion of F4/80<sup>+</sup> macrophages, indicating the requirement of CSF-1 receptor activation for F4/80<sup>+</sup> macrophage development (29). In addition to CSF-1, SCS macrophages appear to need the lymphotoxin signal for their development. Lymphotoxin receptor LTβR is shown to be present on the surface of both SCS macrophages and medullary sinus macrophages, however chimeric mice lacking the LTβR (ltbr−/−) only show a deficiency in SCS macrophages (27). The activation of LTβR on SCS macrophages largely depends on LTα1β2, the ligand for LTβR, present on LN B cells that are located just underneath the SCS in the LN. µMT mice, which lack mature B cells in the LN, show significantly fewer macrophages with the SCS phenotype (CD169+F4/80−) and an abundance of the medullary sinus phenotype (CD169+F4/80+) (34). Furthermore, by ablating lymphotoxin signaling with LTβR-Ig, a soluble lymphotoxin receptor that blocks downstream signaling, a similar deficiency in the SCS macrophage phenotype can be found in wild-type mice as the µMT mice. Medullary sinus macrophages appeared unaffected by lymphotoxin signaling blockade (34). Based on these observations, while medullary sinus macrophages rely on CSF-1 receptor signaling for their development, SCS macrophages require CSF-1 receptor and LTβR for their development and the maintenance of their phenotype.

#### SCS MACROPHAGES PREVENT LYMPH-BORNE PATHOGEN SYSTEMIC SPREADING

Because SCS macrophages directly embrace pathogenic particles arriving from afferent lymphatic vessels, SCS macrophages have been widely studied in antimicrobial immunity, including anti-viral and anti-bacterial responses (**Figure 2A**). Studies on the function of SCS macrophages has first been demonstrated in preventing virus from spreading from the LN to the blood circulation or other organs after subcutaneous infection. Multiphoton intravital microscopy showed CD11b+CD169+MHCII<sup>+</sup> macrophages located on the floor of the popliteal SCS functioning as a "flypaper" to capture fluorescently labeled vesicular stomatitis virus (VSV) particles after a subcutaneous injection at the footpad (11). This observation extends to different viruses, such adenovirus, vaccinia virus and murine cytomegalovirus (MCMV), as luciferase-labeled MCMV is limited to the LN for several days before spreading systemically (11, 35). Artificially depleting the SCS macrophages prior to VSV challenge led to a significant reduction in animal survival and a marked increase in viral titers found in the brain and spinal cord (33).

The "flypaper" function of SCS macrophages is also applicable to lymph-borne bacteria. Fluorescently labeled Pseudomonas aeruginosa, an extracellular bacterium, was found in the LN parenchyma and blood 8 h post-injection when the macrophages were depleted, while bacteria were limited to the SCS when the macrophage layer was intact (36). More specifically, lipid antigens, such as lipopolysaccharide found on bacteria, has also been shown to localize with the SCS macrophages (37). This further defines the "flypaper" function of the SCS macrophages as it is not only preventing a systemic spread, but specifically limits pathogens to the SCS in the LN. Restricting pathogens to the SCS is at least partially achieved by the expression of CD169 on the macrophages, as CD169 interacts with α2,3-linked sialic acids expressed on the surface of cells or microbes. Biotinylated exosomes specifically bound to SCS macrophages on tissue sections while biotinylated bovine serum did not, suggesting the CD169<sup>+</sup> macrophages retain extracellular vesicles and microbes rather than free flow proteins at the sinus (38).

However, evident by their minimal phagocytosis function and failure to adequately process the self-quenching DQ Green probe, SCS macrophages are poorly phagocytic and cannot clear the microbes directly (27). Instead, these macrophages ensure enough immune stimulation by supporting replication of captured pathogens. Fluorescently labeled VSV was robustly replicated in wild-type LNs, while mice lacking the SCS macrophages showed no virus replication (33, 34). Without the immune protection generated by this layer of macrophages, viruses may invade deeper into the parenchyma and infect LN neurons or LN fibroblasts, and eventually disseminate into other organs (33, 35). The mechanism of SCS restricting virus spreading does not apply to all types of viral infection. Capture of the influenza virus alternatively depends on medullary sinus DCs to generate durable B cell responses (39). Like the subcapsular sinus macrophages, medullary sinus macrophages recruit additional immune cells to clear their targeted pathogen. While both SCS macrophages and medullary sinus macrophages demonstrate early activation marker CD69 to UV-inactivated influenza virus, medullary sinus macrophages have been shown to be preferentially activated through the secretion of IFN-β, which further induces IL-1α expression, leading to the expression of dendritic cell and monocyte chemoattractant, MCP-1 (40). Understanding the mechanism of how influenza virus escape SCS macrophages and alternatively activate medullary sinus macrophages may help influenza vaccine design.

FIGURE 2 | Function of the subcapsular sinus macrophage layer in normal and inflamed lymph nodes. (A) Lymph-borne free floating particles and pathogens travel with lymph and enter the lymph node subcapsular sinus via the afferent lymphatics. Subcapsular sinus macrophages are the first layer of cells in the draining lymph node that capture and retain lymph-borne pathogens from entering the lymph node parenchyma likely via the interaction between CD169 and its ligand, α2,3-linked sialic acids, expressed on the surface of cells or microbes. After pathogen capture, SCS macrophages can relay the antigen to B cells just underneath the SCS to prime B cell and humoral responses. SCS macrophage activation produces different types of cytokines to recruit and communicate with other immune cells, such as NK cells, γδ T cells, non-classical CD8<sup>+</sup> T cells, neutrophils, monocytes, T cells etc. to combat the invading pathogens. The SCS macrophage layer prevents pathogen from invading the lymph node parenchyma or systemic spreading. (B) In an inflamed LN during diseased condition, the SCS macrophage layer is interrupted, allowing pathogen to invade the lymph node parenchyma or systemic spreading. The immunological consequence of disrupting SCS macrophage appears contraversial in different types of infection or in cancer progression. The reason behind SCS macrophage layer disruption remains unclear as well.

Once activated, SCS macrophages function by communicating with other LN resident lymphocytes or recruiting other cells to the SCS to provide rapid and robust anti-microbial responses to lymph-borne antigens. As B cells reside directly underneath the SCS macrophage layer, the early studies exploring the function of SCS macrophages identified that activating these cells attract B cells from the follicles to the SCS. SCS macrophages then relay captured antigens to B cells using a complement-dependent and -independent pathway (41). Multiphoton intravital microscopy visualized the accumulation of virus serotype-specific B cells at the SCS depending on the virus challenge, indicating their migration is highly selective (11, 42). Co-stimulatory molecule CD86 was upregulated and B cell receptors were internalized within 6 h after virus challenge, indicative of the activation of B cells. B cell activation after viral challenge failed in the LN when SCS macrophages were depleted with clodronate liposome (CLL). After the early activation, B cells migrate to the boundary between the B and T cell zones of the LN (43). Here, an interaction occurs between the primed B cell and the helper T cells, causing a proliferation of B cells and germinal center formation. Upon macrophage depletion, the antiviral B cells remained spread in the LN and take a much longer time to migrate to T-B cell border. However, depleting the macrophage layer only delayed rather than completely prevented B cell activation or humoral responses, raising the question about the exact function of these cells (11).

Antigen challenge also recruits innate immune cells to the LN SCS. In the case of lipid antigens, α-galactosylceramide was used to coat 200 nm silica particles to stimulate immune cell activation. iNKT cells migrate toward the SCS and are arrested within a few hours. Three days post-injection, the LNs had been inflamed and the number of iNKT cells present in the LN were 10-fold higher than normal (37). Modified vaccinia virus Ankara, a viral vector, was shown to induce NK cell motility and transition from the interfollicular zone and outer T cell zone to the SCS. Depleting the SCS macrophages with CLL reduced the NK cell accumulation and activation normally triggered by virus challenge (44). Similar to the viral challenge, parasite infection with Toxoplasma gondii causes NK cell accumulation in the SCS. However, depletion of SCS macrophage with CLL did not reduce the proportion of NK cells, only suppressed NK cell activation during T. gondii infection (45). Parasitic challenges such as QS-21, an adjuvant component of malaria, colocalized with SCS macrophages. Depleting these macrophages using CLL reduced monocyte, neutrophil, and dendritic cell recruitment to the draining LN (46). However, while neutrophil recruitment to the LN occurred in response to Staphylococcus aureus infection, SCS macrophage depletion via CLL did not change neutrophil movement to SCS (47). Based on these results, it is apparent that SCS macrophage activation recruits and activates different types of immune cells to the SCS when responding to different types of lymph-borne microbes.

SCS macrophages appear to produce different types of cytokines to aid in their function, which potentially explains the different immune cell recruitment against lymph-borne pathogens. In response to lymph-borne virus pathogens, infected SCS macrophages produce interferon-α. Activated SCS macrophages additionally recruit plasmacytoid DCs to the SCS to express type I interferon to initiate anti-viral immunity (33). Mice lacking the SCS macrophages show lower levels of both interferon-α and interferon-β mRNA levels in their LNs, which is correlated with a lower survival rate after viral challenge (34). In the case of bacterial infection, the number of interferonγ producing lymphoid cells increased 4 h after subcutaneous P. aeruginosa infection (36). SCS macrophages are not necessarily the only cell population to express interferon-γ, as the NLR dependent inflammasome activation in these cells enhances other lymphoid cells, such as NK cells, γδ T cells, and nonclassical CD8<sup>+</sup> T cells, to produce interferon-γ. Absence of the SCS macrophages significantly reduces cytokine production and limits the recruitment of important innate immune cells to restrict bacteria spread (36). This could also adversely affect the initiation of the adaptive immunity as inflammasome activation induces an influx of innate immune cells and T cells (48).

### INTERRUPTION OF SCS MACROPHAGES IN DISEASES

While the early activation of SCS macrophages has been studied extensively, disease models pose a different perspective. It appears that pathogen-induced inflammation can disrupt the SCS macrophage layer (**Figure 2B**). Recruitment of neutrophils or NK cells appear to interrupt the SCS macrophage layer, but the mechanism and the function of SCS macrophage dissociation from the SCS remains unclear (45, 49). Using either CpG or LPS, the SCS macrophages were able to dissociate from the SCS, leading to disrupted protective layer. The observed macrophage dissociation appears to be CCR7 dependent, as CCR7-deficient animals seem resistant to inflammation-induced SCS macrophage disruption. Transferring activated bone marrow derived DCs to the LN is sufficient to disrupt SCS macrophage layer, indicating DC activation may cause SCS macrophages to dissociate from the SCS. However, since a subpopulation of CD169<sup>+</sup> macrophages are CD11c+, whether the CCR7 dependent SCS macrophage dissociation is only restricted to CD169+CD11c<sup>+</sup> cell or all SCS CD169<sup>+</sup> cells remains unclear. The mechanism of SCS macrophage dissociation from the SCS remains to be clarified. When SCS macrophages were dissociated during inflammation, B cells were incapable of receiving the antigen and showed diminished activation as measured by germinal center formation and immunoglobulin production (50). These results appear to contradict experimental SCS macrophage depletion, where only early cognate B cell migration to SCS or to the border of T cell and B cell zone is affected, but does not prevent total B cell activation (11, 15). Influenza vaccination is capable of inducing lymph node subcapsular sinus and medullary sinus macrophage necrosis 12 h post-injection. However, the necrosis was independent of neutrophil or NK cell recruitment. Virus challenge activated TLR7 and Myd88, causing necrosis of the subcapsular sinus, but not the medullary sinus, macrophages (40).

Recent interest has sparked over these macrophages in the context of anti-tumor immunity. LN metastases are a key component in patient prognosis. Metastatic tumor cells present in the sentinel lymph node were able to spread systemically via the lymph node blood vessels or the efferent lymphatic vessels (25, 26). SCS macrophages in the tumor draining LN directly interacts with metastatic tumor cells or tumor-derived antigens coming from the afferent lymphatic vessels. The idea that SCS macrophages can limit the spread of cancer, similar to how they limit the spread of lymph-borne microbes, has developed into a relatively new field of study. Clinical studies have determined a correlation between CD169<sup>+</sup> macrophage density in human sentinel LNs and a favorable tumor prognosis. Consistent between multiple different types of tumors, indicators for the favorable prognosis often include a lower number of LN metastases and increased CD8<sup>+</sup> T cell tumor infiltration, reflecting the SCS macrophages' functions of limiting cancer spread and immune activation (51–53). To activate the antitumor CD8<sup>+</sup> T cell response, SCS macrophages are capable of capturing irradiated tumor cells. Like microbes, subcutaneously injected apoptotic tumor cells travel to the LN and are captured mainly by CD169<sup>+</sup> macrophages. Then, activated SCS macrophages recruit and prime anti-tumor CD8<sup>+</sup> T cells at the SCS. Mice with their CD169<sup>+</sup> cells depleted in a CD169-DTR model were incapable of activating CD8<sup>+</sup> T cells or rejecting tumor cells after a vaccination with irradiated tumor cells (12).

However, a growing tumor and its complex tumor microenvironment significantly changes the function of SCS macrophages. Instead of capturing tumor cells as seen with injected irradiated tumor cells, growing melanoma tumors deposit tumor-derived antigens into B cell follicles in patients (54). The accumulation of fluorescent tumor-derived antigen in the follicular dendritic cells in the germinal centers was observed using B16F10 melanoma. Depletion of the SCS macrophages ablated tumor-derived antigen accumulation in the follicular dendritic cells, demonstrating the necessity for SCS macrophages in depositing tumor-derived antigens into the B cell follicle (54). However, a recent publication has shown a contrasting observation; depletion of SCS macrophages increases tumor-derived exosome penetration deep into the B cell follicles and enhances B cell activation as measured by plasma immunoglobulin levels (15). In this study, growing tumors appear to disrupt the SCS macrophages in the tumordraining LN and permits tumor-derived exosome entry into the B cell follicle. The increase in B cell response was correlated with a larger tumor size, suggesting the SCS macrophages are necessary to limit a pro-tumor B cell response (15). Because of these contradicting studies, further studies are required to reveal the function of SCS macrophage in anti-tumor immunity in the tumor-draining LN.

In contrast to the disruption of the macrophage layer seen in microbial infection, inflammation, or melanoma mentioned above, a recent study showed that inflammatory bowel disease increases the CD11b+CD169<sup>+</sup> macrophages in the draining mesenteric LN. Depletion of the CD169<sup>+</sup> macrophages in a CD169-DTR model showed reduced symptoms of inflammation, indicating that these macrophages promoted inflammation in the inflammatory bowel disease model (55). Whether these changes depend on the anatomical location of the LN or the disease models remains to be investigated.

#### CONCLUSION AND PROSPECTIVE

The lymphatic system collects invading bacterial and viral pathogens and drains them to the LN for efficient processing and clearance. In this process, the LN sinus macrophages are among the first immune cells that interact with lymph-borne pathogens. With the evidence from different models, it is clear that SCS macrophages are essential for the response against lymph-borne pathogens. Unlike typical macrophages, the SCS macrophages are incapable of breaking down pathogens. The SCS macrophages appear to diversify its ability to target and initiate specific immune responses to a variety of lymph-borne pathogens by relaying antigens to B cells, producing cytokine signaling cascades to cause influx of dendritic cells, neutrophils,


NK cells, or in some conditions, presenting antigens to T cells. Using cytokine production and immune cell recruitment, SCS macrophages can mount an early immune response against freefloating pathogens and prevent their LN invasion or systemic spreading. While there is a consensus that SCS macrophages limit the systemic dissemination of pathogens, there does not appear to be a universal mechanism for their action (**Table 1**). The requirement of SCS macrophages appear to be more critical for innate immunity, since depletion of SCS macrophages allow pathogens to escape the draining LN and spread systemically, adversely affecting survival rate. Surprisingly, although SCS macrophages appears critical to relay antigens to B cells, artificial depletion of SCS macrophages did not substantially interrupt the overall anti-microbial adaptive immune responses, except several hours of delay in the induction of adaptive immunity.

Several studies have shown infection induces SCS macrophage dissociation from the SCS (**Figure 2**). CpG or LPS induces SCS macrophage migration deeper into the LN parenchyma, which impairs B cell responses to a secondary infection (50). Tumor progression induces SCS macrophage dissociation from the SCS in the tumor draining lymph node and results in B cell activation and tumor growth (15). Why the effect of dissociated SCS macrophages on subsequent immune protection appears contradictory between infectious diseases and cancer progression remains unclear. As it is now clear that inactivated influenza virus causes macrophage necrosis, one interpretation could be that challenge of microbes or materials mimicking microbial products causes more severe damage to SCS macrophages when compared to tumor derived antigens. Another possibility is that microbial product challenge lasts several hours, while infection or tumors may continuously deliver antigens for several days. Additionally, disease induced SCS macrophage dissociation also differs from the artificial SCS macrophage depletion as the former did not complete abrogate the SCS

#### REFERENCES


macrophage layer and some of these macrophages are relocated deeper into the LN parenchyma. Most studies use CLL and/or diphtheria toxin (DT) in a CD169-DTR to deplete SCS macrophages (**Table 1**) and both may cause off-target cell death. Additionally, the induced cell death may impact the function of immune cells in the LN. Thus, the mechanisms that cause SCS macrophage dissociation would substantially impact the immune protection to a subsequent challenge, such as secondary infection or continuous tumor-derived antigen delivery. More studies are required to understand why the SCS macrophages leave their position after stimulation and what is the immunological consequence of SCS macrophage dissociation from the SCS.

Currently, the mechanisms of how SCS macrophages participate in fighting against lymph-borne pathogens are better studied. The role of SCS macrophages in anti-tumor immunity in the tumor draining LN is still young. The collective literature in anti-microbial studies suggest future studies center around how SCS macrophage communication with other immune cells at different stages of tumor progression could provide pivotal insights into the development of immunotherapy.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work is supported by the startup up fund to SL, provided by Dianne & Irving Kipnes Foundation, Canadian Institutes of Health Research (CIHR, PJT-156035), and the Canada Foundation for Innovation (CFI, 32930) to SL.


vesicle-B cell interactions. Science. (2016) 352:242–6. doi: 10.1126/science. aaf1328


a fatal viral infection independent of adaptive immunity. Immunity. (2012) 36:415–26. doi: 10.1016/j.immuni.2012.01.013


**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 © 2019 Louie and Liao. 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(s) 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.

# Leucocyte Trafficking via the Lymphatic Vasculature— Mechanisms and Consequences

#### David G. Jackson\*

MRC Human Immunology Unit, Radcliffe Department of Medicine, MRC Weatherall Institute of Molecular Medicine, University of Oxford, Oxford, United Kingdom

The lymphatics fulfill a vital physiological function as the conduits through which leucocytes traffic between the tissues and draining lymph nodes for the initiation and modulation of immune responses. However, until recently many of the molecular mechanisms controlling such migration have been unclear. As a result of careful research, it is now apparent that the process is regulated at multiple stages from initial leucocyte entry and intraluminal crawling in peripheral tissue lymphatics, through to leucocyte exit in draining lymph nodes where the migrating cells either participate in immune responses or return to the circulation via efferent lymph. Furthermore, it is increasingly evident that most if not all leucocyte populations migrate in lymph and that such migration is not only important for immune modulation, but also for the timely repair and resolution of tissue inflammation. In this article, I review the latest research findings in these areas, arising from new insights into the distinctive ultrastructure of lymphatic capillaries and lymph node sinuses. Accordingly, I highlight the emerging importance of the leucocyte glycocalyx and its novel interactions with the endothelial receptor LYVE-1, the intricacies of endothelial chemokine secretion and sequestration that direct leucocyte trafficking and the significance of the process for normal immune function and pathology.

#### Edited by:

Amanda W. Lund, Oregon Health & Science University, United States

#### Reviewed by:

Cornelia Halin, University of Zurich, Switzerland Alvaro Teijeira, ETH Zürich, Switzerland

\*Correspondence: David G. Jackson david.jackson@imm.ox.ac.uk

#### Specialty section:

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

Received: 02 February 2019 Accepted: 21 February 2019 Published: 14 March 2019

#### Citation:

Jackson DG (2019) Leucocyte Trafficking via the Lymphatic Vasculature— Mechanisms and Consequences. Front. Immunol. 10:471. doi: 10.3389/fimmu.2019.00471 Keywords: lymphatic, trafficking, transmigration, dendritic cell, endothelium, chemokine, hyaluronan, LYVE-1

### INTRODUCTION

The lymphatics form an extensive network that facilitates the drainage of plasma leaked from the peripheral vasculature and its re-uptake by the venous circulation for maintenance of fluid homeostasis (1, 2). Moreover, they constitute an essential compartment of the immune system, providing conduits for the trafficking of antigen loaded dendritic cells (DCs), memory and regulatory T cells (TMEM and TREG) and neutrophils to draining lymph nodes (dLNs) in the process of immune activation, modulation and peripheral tolerance (3–8). In addition, they mediate the clearance of macrophages that remove pathogens and tissue debris during resolution of tissue inflammation and infection, and are exploited by microbial pathogens such as Group A streptococci, Salmonella, Brucella and M. tuberculosis, and parasitic nematodes that use the lymphatics for host colonization and systemic dissemination (9–13). Understanding the mechanisms by which cells enter and exit lymphatics in tissues and lymph nodes and their detailed choreography will therefore be essential to understanding how such processes help regulate immunity, and how they might be manipulated for therapeutic intervention.

**41**

The lymphatic network is quite distinct from the blood vasculature in terms of both structure and physiology. Notably, the lymphatics start as blind-ended capillaries that are freely permeable to fluids, and have discontinuous overlapping junctions pre-adapted to cell transit, unlike the conventional tight junctions that seal most blood vessels (8, 14, 15). In addition, unlike the blood circulation, cell trafficking in most afferent lymphatics involves intravasation rather than extravasation, in keeping with their role in accommodating the passage to dLNs of tissue resident leucocytes and transient immune cell populations recruited from the circulation (3–5, 8, 16–18). Moreover, during entry to lymphatic capillaries, leucocytes are exposed to the very low shear rates associated with interstitial fluid flow, as distinct from the high shear rates experienced during extravasation from blood capillaries (19). Reflecting such different environments, some of the molecular mechanisms for leucocyte entry and trafficking in the lymphatics are quite different to those in blood vessels. Nevertheless, as will be apparent from this present review, some are broadly similar. Just as extravasation of leucocytes from blood is triggered by inflammation and the induced expression of dedicated chemokines and adhesion molecules in the vascular endothelium, so too is the entry of most leucocytes to afferent lymph vessels. Indeed, as discussed later, certain key adhesion molecules are shared by both vasculatures.

In the following sections, I describe the latest findings on how leucocytes exploit chemotactic and adhesive mechanisms to enter and migrate within lymphatic vessels, as well as exit the lymphatic sinuses in dLNs to fulfill their various immune functions. I begin with an outline of the major leucocyte populations that migrate in lymph, and the characteristic architecture of lymphatic endothelial junctions. Based largely on knowledge gained from studies on DCs, I go on to provide a detailed account of the key steps in lymphatic trafficking from interstitial migration, lymphatic entry, and intraluminal crawling, to transit within downstream dLNs **(Figure 1)**. In my discourse, I highlight the newly discovered role of the lymphatic endothelial HA receptor LYVE-1 in lymphatic entry and its functional relationship with other more ubiquitous adhesion receptors in endothelial transit, and the possibility of their cooperation in a "lymphatic synapse." I also describe the coordinated triggering of chemokine release by transmigrating DCs, and some of the additional mechanisms employed by neutrophils and certain T cell populations during lymphatic transit. Lastly, I speculate on how knowledge of lymphatic trafficking mechanisms might be exploited in the future to develop new therapies for immune and inflammatory disorders.

#### LEUCOCYTE POPULATIONS THAT TRAFFIC VIA LYMPH

Classic cannulation studies carried out in domestic animals and applicable also to mice and humans showed the major cell populations migrating in normal afferent lymph are T cells (80–90%), followed by antigen presenting DCs and very small numbers of B cells which together account for most of the remaining 10–15%. Most of the T cells are antigen-experienced CD4<sup>+</sup> CD45RO<sup>+</sup> effector memory (TEM) cells, recently redefined as the recirculating memory (TRCM) subset (20, 21), which, having entered the extra-lymphoid tissues from blood, engage in immune surveillance for cognate antigens before exiting via the afferent lymphatics to dLNs where they modulate recall immune responses (22–25). Notably, lymphocytes of the CD4 subset in afferent lymph outnumber those of the cytotoxic CD8 subset by some 5× fold (26–28), which mostly remain immotile as tissue-resident (TRM) cells. Furthermore, more recent cell tracking studies using photoconvertible Kaede mice have revealed that a significant proportion (25%) of the CD4 population are FOXP3<sup>+</sup> TREGs, thus uncovering a previously unrecognized role for the lymphatics in conveying these important immunoregulatory cells. By comparison, only low numbers of naïve T cells are usually present in afferent lymph, and despite the fact these can be shown to enter lymphatic capillaries after adoptive transfer in mice, their normally low frequency in tissue means they rarely do so in vivo. The likely functional significance of TEM cell migration in afferent lymph may be to allow the amplification and polarization of immune responses in the dLNs and maintenance of the T cell memory pool, as well as enabling the re-entry of these lymphocytes to the circulation to target pathogen dissemination in further tissue sites (7, 21).

The second most numerous leucocyte population in afferent lymph are DCs, which ferry endocytosed antigens from the tissues, primarily for immune priming in dLNs. Although small numbers of DCs migrate in lymph under steady state conditions to maintain peripheral tolerance to self-antigens (29– 33), the majority are mobilized by inflammation, which induces a program of differentiation and the expression of appropriate chemokine receptors for vessel entry (34–36) (see below). DCs, more than any other cell type, have been the subject of studies into the mechanisms of lymphatic trafficking, not least because they are normal tissue residents whose migration can be readily monitored by dye uptake in experimental mice (37). Moreover, in laboratory animals the lymph migrating DCs are almost as numerous as T cells, owing to the fact the pathogen-low environment in which they are bred and maintained generates a smaller pool of circulating memory cells (38).

During inflammation, the numbers of T cells and DCs in afferent lymph increase several fold along with an increase in lymph vessel permeability and the rate of lymph flow (8, 34). Furthermore, the afferent lymph can contain type I and II macrophages which are recruited to inflamed tissues for the clearance of debris and remodeling of the extracellular matrix (ECM) and which utilize the lymphatics for subsequent exit in a process that is becoming increasingly appreciated as critical for resolution and the return to normal homeostasis (39–41). In addition, this lymph contains subsets of neutrophils that are rapidly recruited to the tissues during sepsis and trauma and subsequently exit via the inflamed lymphatics to dLNs (42, 43). Although the majority of neutrophils in tissues are shortlived (T1/<sup>2</sup> 6–12 h) and undergo early apoptosis before removal by macrophage efferocytosis (44–46), the lymph-migrating cells have an extended lifespan (47). Most notably they can transport phagocytosed pathogens such as Leishmania, H. pylori and M.

FIGURE 1 | Key steps in leucocyte migration through the lymphatic system. Beginning in the interstitium, motile leucocytes including recirculating (TRCM) cells and neutrophils recruited from the blood, in addition to antigen loaded tissue resident DCs and macrophages are mobilized through inflammation to migrate toward initial lymphatic capillaries for access to dLNs by means of chemokine-directed amoeboid movement and in some cases integrin-mediated adhesion to fibronectin. The leucocytes then transit the blind-ended lymphatic capillaries at discrete sites within their discontinuous button-like endothelial junctions through receptor-mediated adhesion and chemokine directed squeezing, or in the case of neutrophils via lipoxin mediated chemorepulsion (detailed in Figures 2,4, respectively). Although not depicted, it is also possible that leucocytes can enter downstream vessels through conventional zipper-like junctions, and exit at intermediate stages to the surrounding tissues. On entering the capillary lumen, the transmigrated cells crawl in a semi-directional manner along the endothelial surface, guided by gradients of immobilized CCL21 established under low shear flow, using adhesion to ICAM-1 for traction. On entering downstream valved collectors, the leucocytes are now conveyed by passive transport, propelled by lymph flow generated by smooth muscle contraction. Ultimately, the migrating cells arrive within the SCS of dLNs, where they immediately transmigrate to the underlying cortex to initiate or modify immune responses in the case of DCs and likely neutrophils and macrophages, or continue to the medullary sinuses where they transit to the cortex to either remain there or pass to the efferent lymphatics to re-enter the blood circulation in the case of TMEMs. Although not shown in the Figure, naïve T and B cells also recirculate directly from the blood through lymph nodes at specialized high-endothelial venules to enter the cortex and re-exit through efferent lymph.

bovis BCG to dLNs where they can influence the polarity of protective T cell responses through cytokine release and crosstalk with DCs, thus bridging the gap between innate and adaptive immunity (48–51). Indeed, neutrophils can migrate via lymph more rapidly than any other leucocyte populations, reportedly arriving in the ipsilateral dLNs some 12–72 h earlier than either DCs or macrophages (52–56).

In contrast to afferent lymph, the leucocyte population present in the efferent lymphatics that exit from the lymph node hilum are mostly naïve T and B cells. Having entered the lymph nodes through high endothelial blood venules in a separate circuit to probe for antigens presented by DCs in the cortex and paracortex, these are ultimately returned to the circulation through the subclavian vein (36). Notably, during the onset of infection or inflammation, the efflux of this recirculating population from the lymph node is halted transiently (3–4 days) so as to prolong their residence time and thus increase the efficiency of immune recognition (57).

#### THE DISTINCTIVE ARCHITECTURE OF INITIAL LYMPHATIC VESSELS

As already mentioned, the afferent lymphatics initiate as blindended capillaries that branch and merge with larger collecting vessels, emptying their contents into dLNs before exiting as efferent vessels that reconnect either directly, or via other intervening nodes to the venous blood **(Figure 1)**. In keeping with their fluid draining function, the blind-ended capillaries have only a rudimentary basement membrane (BM), and lack any

VCAM-1 expressed within the cups bind to integrin ligands on the DC surface activated by local secretion of CCL21 and likely provide the necessary traction for diapedesis. The precise details of the interplay between LYVE-1 and these receptors are not yet known.

investment by actin-containing smooth muscle cells. Moreover, the endothelial cells that make up the first few millimeters of these initial capillaries have a distinctive oakleaf shape that allows them to interdigitate and form loose discontinuous junctions **(Figure 2**), quite unlike the continuous junctions of endothelia in most blood vessels (14, 58–61). As revealed by electron microscopy (EM) and confocal imaging studies of such capillaries in mice, the alternating membrane flaps making up these structures are pinned at their sides by discrete assemblages some 3µm wide and spaced 3µm apart, that contain the adherens-junction protein VE-cadherin and the tight junction proteins Claudin-5, ZO-1 (zonula occludens-1), ESAM (endothelial selective adhesion molecule) and JAM-A (14). In contrast, the flaps remain free at their tips, where they guard openings of ∼0.5–1µm, that are decorated by CD31 and the lymphatic endothelial HA receptor LYVE-1 as detailed below (14). It is through these openings that DCs appear to enter the lymphatic capillaries. Importantly, the process is far from passive, as the dimensions of the DCs are many times greater than the gaps through which they must enter, and hence transit requires pushing and squeezing via intimate contact with the endothelium (14, 62–65). While the rationale for such an elaborate arrangement of buttons and flaps is not fully clear, it likely represents a compromise between the conflicting requirements of high vessel permeability for fluid uptake and a more restrictive barrier for regulating leucocyte entry. Indeed, for the experimenter, such architecture poses problems for establishing in vitro models to study vessel entry, as primary cultured lymphatic endothelial cell (LEC) monolayers form only partial surrogates of discontinuous junctions (66), and full authentication requires the application of whole animal models or crawl-in assays with tissue explants.

In contrast to the initial capillaries, downstream pre-collector and valved collector vessels lack buttons and instead adopt conventional tight or "zipper" like junctions that allow the formation of a more fully sealed vasculature (14, 15). The collectors are also covered by smooth muscle cells that promote contractile pumping and which convey migrating leucocytes to the dLNs with minimal leakage. Of note, these zippered junctions constitute the default state of capillaries in the developing embryo and early neonate, and button-like junctions emerge only during the later neonatal period, co-incident with the establishment of full immune function (67). Curiously, the nascent lymphatics generated by lymphangiogenesis in chronically inflamed tissues also have conventional zippers, raising the possibility that vessel entry via zippers rather than buttons may be permissible in certain contexts (15).

### MIGRATION THROUGH INTERSTITIAL MATRIX AND THE PERI-LYMPHATIC BASEMENT MEMBRANE

In order to access the initial lymphatic capillaries for exit from the tissues and onward trafficking to lymph nodes, leucocytes must first migrate distances of several hundred microns through the surrounding interstitial matrix, a variably dense complex of type I collagen fibrils, fibronectin, hyaluronan (HA), and heparan sulfate proteoglycans (HSPGs) (68). The rate of such interstitial migration is thought to be comparable with that of subsequent migration through afferent lymph capillaries [4.4 vs. 5.7 µm/min in the case of DCs (69) and see below] and the time taken for DCs to navigate to a lymphatic capillary from the point of initial mobilization has been estimated as ∼1 h (70). In mice, under conditions of normal homeostasis, the numbers of migrating cells in sites such as skin are small and the populations consist mostly of immature dermal DCs and effector TMEM cells en route to lymph nodes during immune surveillance. However, in response to inflammation, as deduced from in vivo studies using skin contact sensitizing agents and complete Freund's adjuvant (CFA), this traffic increases significantly. For practical reasons, most studies of leucocyte interstitial migration in the context of lymphatic trafficking have focused on DCs in the mouse dermis and epidermis. Although sessile in the steady state, these cells are subsequently induced to differentiate and crawl toward LVs, in response to inflammatory signals received from cytokines, prostaglandins and leukotrienes released from keratinocytes activated by microbial products and other Tolllike receptor (TLR) ligands (71). The movement of activated DCs is further stimulated by the increase in interstitial fluid flow that is characteristic of tissue inflammation. Nevertheless, the major driving force for leucocyte interstitial migration under resting and inflamed conditions is chemotaxis, directed primarily by the chemokine CCL21 secreted radially from lymphatic endothelium and its G-protein coupled receptor CCR7 on both migrating DCs and T cells (32, 36, 57, 72). Elegant studies by Weber and Sixt using single cell tracking in mouse ear skin revealed DC migration begins in a random fashion until the cells reach a distance of ∼90µm from an initial capillary, at which point their migration becomes directional and persistent (73). This directionality of motion is imposed by fixed, haptotactic gradients of CCL21 secreted constitutively from lymphatic vessel endothelial cells and sequestered by collagen and HSPGs in the surrounding interstitial matrix via the positively charged chemokine N-terminal domain (73). Moreover, the gradients fall away with distance from the vessel periphery in such a way as to provide a polarity that can be sensed by DCs in the typical size range of 15–50µm. Importantly, DC migration under both resting and inflamed conditions is independent of β2 integrin/ICAM-1 adhesion, as demonstrated by seminal studies in mice that compared intact and pan-integrin knockout cells (74). Migration proceeds instead by "amoeboid movement" whereby CCL21 directs the pushing and squeezing of DCs through the 3D collagen matrix, primarily by triggering actin polymerization and actomyosin contraction at the leading and trailing edges via the small GTPases Rac1 and Rac2 and nuclear contraction via the Rho associated kinase ROCK (69, 74, 75). To date few studies have addressed the mechanisms of interstitial migration employed by leucocyte populations other than DCs. Although in the case of CD4<sup>+</sup> effector TMEM, such migration in CFA inflamed ear dermis was also found to be independent of β2 integrin/ICAM-1 adhesion (76), it was nevertheless reliant on guidance via β1and β3 integrin-mediated adhesion to matrix fibronectin, laid down in association with parallel oriented collagen fibers (77). In the looser fibronectin-rich ECM network that is characteristic of inflamed tissues, such adhesion may be a greater requirement for the chemokine driven migration of T cells due to their smaller size and inability to extend dendritic processes.

Lastly, before engaging with the external surface of initial lymphatic capillaries, migrating leucocytes must traverse the surrounding BM, a rudimentary structure comprised mainly of type IV collagen and the non-network forming laminin isoform α4, before they can transmigrate the endothelium to enter the vessel lumen. In contrast to the more close-knit BM of blood vessels (78), the BM of lymphatic capillaries is sparse and highly perforated by gaps of ∼1µm diameter that are completely devoid of any ECM components. Making up 30% of the vessel surface, these gaps offer points through which migrating cells can access the underlying vessel endothelium with the aid of physical expansion (to ∼2µm). Studies of inflamed mouse ear skin indicate that those gaps which overlie sites of vessel entry are marked by discrete deposits of sequestered CCL21 which direct the transit of DCs by chemotaxis/haptotaxis rather than simply chemokinesis. Indeed, migrating DCs have been observed to extend cellular processes toward these CCL21 puncta and apparently make physical contact with them (71). Furthermore, detailed imaging of such events in ex vivo crawl-in assays with dermal tissue explants (79) suggested these invadopodialike protrusions enable DCs to transiently expand the BM portals by physical squeezing without the parallel requirement for proteolytic re-modeling observed during leucocyte transit of blood vascular BMs. On traversing the BM the migrating leucocytes then encounter the lymphatic endothelium next to their interdigitating flaps, which bend inwards to accommodate entry to the vessel lumen (79).

#### TRANSMIGRATION OF THE VESSEL ENDOTHELIUM AND ENTRY TO THE LYMPHATICS

#### Integrin-Mediated Adhesion via Endothelial Microvilli

Similar to leucocyte transmigration of blood vessel endothelium, transmigration of lymphatic endothelium involves prior adhesive interactions between β2 integrins and their Ig superfamily counter receptors. Of note, early in vivo studies using mouse models of skin inflammation and contact hypersensitivity showed that the endothelial leucocyte adhesion receptor ICAM-1 is involved in the migration of epidermal Langerhans cells to skin dLNs (80, 81). Nevertheless, the requirement for adhesion receptors in lymphatic transmigration has been disputed, and one particularly prominent study using panintegrin (Int−/−) deficient mice reported that under normal resting conditions, DCs can enter lymphatic vessels and migrate to dLNs independently of all integrins, just as documented for interstitial migration, by means of chemokine-directed amoeboid motion (74). Importantly however, these studies tracked the migration of exogenous DCs that had been adoptively transferred to a non-inflamed dermis and hence reported a mode of trafficking that proceeds only quite rarely in normal tissues (74). Indeed, other studies that tracked endogenous DC trafficking in the uninflamed skin of CD11c YFP+/VE-Cadherin Cre/Rosa26 Fl RFP<sup>+</sup> chimeric mice detected few if any cells entering dermal lymphatic capillaries, and such entry was observed only after adoptive transfer into tissue inflamed by either contact hypersensitization, exposure to adjuvants (CFA) or treatment with bacterial LPS (71). Hence, an integrinindependent mode of lymphatic vessel transmigration likely applies only to the relatively small minority of immature DCs that traffic constitutively in immune surveillance.

In contrast, in inflamed tissues where DCs migrate more extensively via lymph, transmigration is indeed dependent on integrin-mediated adhesion mechanisms. As evidenced by studies with cultured primary LECs, and confirmed by transcriptional profiling of lymph vessel endothelium isolated from inflamed mouse skin, exposure to inflammatory cytokines and contact sensitizing agents results in rapid (T1/<sup>2</sup> ∼3 h) upregulation of the key integrin counter-receptors ICAM-1 and VCAM-1 together with various other adhesion molecules associated with leucocyte endothelial transit such as E-selectin, and a range of different chemokines that are attractants for DCs, monocytes, lymphocytes and neutrophils (82, 83). Furthermore, ICAM-1 and VCAM-1 blocking mAbs have been demonstrated to impair adhesion and transmigration of bone marrow DCs (BMDC) across inflamed LEC monolayers in vitro, as well as entry and trafficking of endogenous DCs to dLNs in vivo, both in murine models of skin hypersensitivity and dermal vaccine-induced T-cell immunity (82, 84). Likewise, in CD45.2 mice, LFA-1 blocking mAbs were shown to impair trafficking of injected CD45.1 DCs to dLNs from TNFα treated footpads (85). Most notably, it has been observed using in vitro confocal imaging of murine BMDCs engaged in transit across inflamed lymphatic endothelium, that ICAM-1 and VCAM-1 are concentrated within finger-like projections somewhat analogous to the transmigratory cups described originally in blood vascular endothelium (86–89) (**Figure 2** and see below). These extend around the transmigrating cells, facilitating their adhesion via activated (mAb 24+ve) forms of β2 integrin (LFA-1) present in complementary projections on the DC surface (85). Reportedly, similar ICAM-1 lined membrane protrusions were also observed around transmigrating DCs in vivo both in mouse ear and human dermal tissues.

The transmigration of T cells across lymphatic endothelium also involves integrin-mediated adhesion. Accordingly, in studies employing antibody blockade, ICAM-1 and its β2 integrin ligand LFA-1 were shown to be functionally required for T cell adhesion and transmigration of TNFα treated murine LECs in vitro, as well as entry to lymphatic vessels and trafficking to dLNs in the inflamed skin of oxazolone and adjuvant treated mice in vivo (76). Indeed, in an analogous manner to DCs, ICAM-1 was observed to be distributed in microvillar projections around CD4<sup>+</sup> T cells and ionomycin/PMA-activated peripheral blood mononuclear cells adhering to a lymphatic endothelium (85). Hence it appears that similar mechanisms are employed by both CD4 T cells and DCs for lymphatic vessel entry through extension of ICAM/VCAM enriched endothelial microvilli (19, 90). Furthermore, as detailed later in this review, the initial stages of neutrophil entry involves adhesion through the β2 integrins LFA-1 and Mac-I and most likely their main lymphatic endothelial counter-receptor ICAM-1, as evidenced from experiments using adoptive transfer of Int−/<sup>−</sup> neutrophils and receptor blocking mAbs (42, 43, 56, 91). Indeed, it was clearly shown that administration of either β2 integrin or ICAM-1 blocking mAbs impaired the entry of GFPlysM labeled neutrophils to the initial lymphatic capillaries in mice, causing the cells to logjam at the vessel periphery (56), in a manner reminiscent of DCs given similar blockade in oxazolone-treated mouse skin (17, 82). Such harnessing of integrins, ICAM-1 and VCAM-1 for leucocyte adhesion to the basolateral surface of lymphatic vessel endothelium and subsequent transendothelial migration is in many ways analogous to their involvement in the abluminal crawling of newly extravasated leucocytes on the pericyte surface of blood vessels (92) and stands as an example of how comparable mechanisms can operate in the two vasculatures albeit in reverse orientation.

#### Transit via Hyaluronan and LYVE-1 Transmigratory Cups

In addition to integrins and their counter-receptors, it has recently been demonstrated that transmigration of DCs requires critical involvement of LYVE-1 (93) within the button-like junctions of initial lymphatic capillaries (14, 15) **(Figure 2)**, and its engagement with the large mucopolysaccharide ligand HA present on the DC surface (19, 90). Closely related to the leucocyte receptor CD44 (94, 95), LYVE-1 contains a conserved lectin-like HA-binding domain, termed the Link module, at its N-terminus (19, 96). Although HA is a ubiquitous component of perivascular ECMs (97), it can also be synthesized by DCs and other leucocytes including macrophages and T cells as a surface glycocalyx (98–100), and it is the selective interaction of LYVE-1 with this latter structure in preference to ambient HA that facilitates DC transmigration (90, 101). Such specificity is possible because of the strict avidity-dependent nature of LYVE-1: HA interactions. Because the receptor binds only a short 8–20 saccharide region of HA with low affinity (K<sup>D</sup> 125µM), it therefore relies on homo-dimerization and clustering, as well as a high ligand density to achieve the multiplicity of co-ordinate binding interactions required for tethering of the

polymer chains (96, 101, 102), a phenomenon that has been termed superselectivity (103). Intriguingly, in vitro confocal imaging studies using primary murine LEC monolayers have shown that LYVE-1 is recruited along with ICAM-1 and VCAM-1 to the transmigratory cups that form upon initial DC contact with endothelium, and furthermore that engagement of LYVE-1 with the HA glycocalyx is actually critical for their formation (90). Consistent with a functional role for the receptor in DC transmigration, LYVE-1 HA blocking mAbs also impaired both the adhesion and transit of DCs across LEC monolayers in vitro. Similar features of LYVE-1 transmigratory cups have been observed in vivo, during DC transit of lymphatic capillaries in oxazolone sensitized mouse skin. Moreover, interference with LYVE-1 mediated DC interactions in such studies by LYVE-1 gene deletion, antibody blockade or DC HA glycocalyx depletion resulted in the characteristic logjamming of endogenous and adoptively transferred DCs on the basolateral surface of dermal lymphatics, and impaired their capacity to prime antigen specific T cells in dLNs, mirroring the effects seen with β2 integrin blockade (90). Notably, besides DCs, LYVE-1 also mediates adhesion and transmigration of macrophages across lymphatic endothelium through similar mechanisms (102). Indeed, in a murine model of myocardial infarction where damage-inducing M1 macrophages infiltrate the ischaemic myocardium and are subsequently cleared via cardiac lymphatics (104, 105), the process is blocked by Lyve1 deletion, which delays resolution and leads to fibrotic scarring (41). Whether or not T cells also engage LYVE-1 in such structures has yet to be determined. However, in common with DCs and macrophages these also have a capacity for HA biosynthesis, and hence may well assemble a similar HA surface glycocalyx (98). Curiously, another receptor, CLEVER-1, containing an HA-binding "Link" domain related to LYVE-1 has been reported to mediate CD4 and CD8 T cell trafficking through afferent murine skin lymphatics (see **Table 1**) as well as transmigration across monolayers of lymphatic endothelium in vitro (115–117). However, the mode of action of CLEVER-1 is distinctly different to LYVE-1, as the Link module was shown to be non-functional and the site mediating transmigration instead found to reside within a distant EGF repeat region (116, 118).

Importantly, the role of LYVE-1 in lymphatic transmigration extends beyond merely supporting cell adhesion. For example, it has been demonstrated that engagement of the receptor can transduce signals for endothelial junctional relaxation, in particular the phosphorylation and detachment of VEcadherin located within the button-like foci of initial capillaries from which LYVE-1 is selectively excluded (14, 106). Hence, engagement with the DC glycocalyx and un-buttoning of the VE-cadherin lined junctions at vessel entry sites may well promote coalescence of the alternating endothelial flaps and redistribution of LYVE-1 to form a single continuous interface for leucocyte diapedesis (2, 19). In addition, the large contour lengths of HA polymers, which can extend to several microns, likely mask access to the more compact integrins [extracellular domains ∼20 nm (107)] on the underlying DC surface, and allow the polysaccharide to make primary contact with the capillary endothelium. It has also been postulated that the low affinity of LYVE-1 HA-binding supports crawling of DCs along the vessel surface toward junctional portals through the inherently low friction of the interaction (19). This role as lubricant is supported by physicochemical studies of LYVE-1 HA binding mechanics at the single molecule level using atomic force microscopy, which indicate the individual interactions are weak and that they rupture collectively under the low forces experienced in interstitial flow [see (19)]. It contrasts markedly with the behavior of CD44, a receptor tuned for leucocyte capture in post-capillary venules, which forms bonds that are stronger and detach sequentially in a Velcro (hook and loop) like fashion in response to the higher forces experienced in blood flow (108, 109). Nevertheless, the transit of cells through lymphatic endothelium must involve traction, and if this is not provided by LYVE-1, then it is likely that DCs use both HA and integrin-based adhesion either on different faces of the cell, or in sequential fashion during diapedesis.

Undoubtedly, many other adhesion molecules located in and around the buttoned junctions of lymphatic endothelium contribute to leucocyte transmigration, and may even specify the entry of discrete leucocyte populations. A number of such receptors including Mannose receptor (110–114), ALCAM (CD166, Activated Leucocyte Cell Adhesion Molecule) (119), L1CAM (CD171) (120), 4-1BB (CD137) (121), CD99, and CD31 (PECAM-1) (122) have already been implicated in the process from various in vitro and in vivo studies (**Table 1**). However, the precise functional roles played by each of these receptors, and how they are individually choreographed during lymphatic trafficking have yet to be elucidated.

#### DIRECTIONAL GUIDANCE OF LEUCOCYTE TRANSMIGRATION BY CHEMOKINES

#### The Key Roles of CCL21 and CCR7

In concert with adhesion receptors, critical cues for the guidance of leucocytes during the process of lymphatic vessel entry are provided by chemokines synthesized and secreted in the main by underlying LECs. However, the emerging view is that these may operate as much by inducing the transient arrest of migrating leucocytes at endothelial junctions as by guiding their migration along conventional chemotactic gradients. CCL21 released from the endothelium has been identified as the primary chemokine controlling the entry of DCs to afferent lymph, based initially on the findings from elegant studies in mice showing that CCL21 neutralizing mAbs or CCR7 gene deletion decreased or delayed DC migration from the dermis to dLNs (32, 72, 123, 124). Curiously, mice express two separate genes for CCL21 that encode a lymph node isoform CCL21ser and an afferent vessel isoform CCL21leu (125, 126), and it has logically been assumed (though not formally proven) that the latter (CCL21leu) controls the lymphatic entry step (32, 72, 123, 124, 126). Although a naturally occurring genetic deletion of CCL21ser in the plt/plt mouse line compromises DC trafficking via lymph (125–127), this likely reflects a more distal defect in either entry or retention in downstream dLNs.

In addition to DCs, CCL21 is also the primary chemokine driving the entry of T cells to afferent lymph vessels in the


periphery, in particular the antigen-experienced CD4<sup>+</sup> and CD8<sup>+</sup> TEM population that exits from the circulation to patrol the inflamed tissues. Approximately 50% of this population in the skin of humans and mice are CCR7+, and as confirmed from cannulation studies in sheep, almost all the T cells that migrate in afferent lymph express CCR7 and respond chemotactically to CCL21 (23, 128). Furthermore, parallel studies with CFSE labeled T cells in gene deficient mice have indicated their entry to the afferent lymphatics in both dermis and lung and trafficking to dLNs is almost entirely dependent on expression of CCR7 (20, 23). These migratory CCR7<sup>+</sup> T cells which can ultimately re-enter the blood via the thoracic duct are clearly distinct from the CCR7<sup>−</sup> T effector (TEM) population that remains resident within the tissues as sentinels, and have since been defined phenotypically as a (CCR7int/<sup>+</sup> CD62Lint CD69<sup>−</sup> CD103+/<sup>−</sup> E-selectin ligand+) recirculating memory (TRCM) cell subset, based on their tracking in photoconvertible fluorescent Kaede mice (21). Indeed, they also include the important immunosuppressive CD4<sup>+</sup> T regulatory cell (TREG) as well as inflammation-associated Th1 and Th17 cell subsets (129, 130). Curiously however, the importance of CCL21/CCR7 for T cell entry and trafficking is diminished in chronic as compared to acute inflammation (130), and hence it is likely that other inflammation-induced chemokines become involved at later time points (see below).

It has also been reported that CCL21 chemotaxis helps direct the entry of neutrophils to afferent lymphatics. Accordingly, in a study of skin inflammation evoked by topical CFA administration in mice, the lymph migrating neutrophil population was identified exclusively as CCR7<sup>+</sup> and trafficking to dLNs was decreased almost 4× fold in CCR7−/<sup>−</sup> animals (55). Likewise, the entry of neutrophils to the cremaster muscle lymphatics following TNFα treatment, which induces CCL21 release, was reported to be almost completely (>97%) inhibited in CCR7−/<sup>−</sup> mice (43). Whether CCL21 is the primary chemokine in every context is however open to question, as another study of neutrophil migration in S. aureus treated mice found that neutrophil entry to the dermal lymphatics was directed by CXCL12, on the basis of inhibition by the CXCR4 receptor antagonist AMD3100 (91).

#### Leucocyte-Induced Chemokine Release

Synthesis of CCL21 is markedly upregulated in murine and human lymphatic endothelium in response to inflammation (35, 66, 131) whereby the chemokine accumulates in intracellular storage vesicles in readiness for secretion, notably at the basolateral surface of the endothelium where leucocytes transmigrate (132). Intriguingly, in a recent seminal study of DCs by Vaahtomeri et al it was reported that leucocytes themselves can trigger such secretion from lymphatic vessels through a contact dependent mechanism in which the transmigrating cells extend filopodia toward the endothelium, provoking a Ca2<sup>+</sup> flux that triggers disassembly of cortical actin and the exocytosis of pre-stored CCL21 from trans Golgi vesicles along linear microtubule tracks, for fusion with the plasma membrane (**Figure 3**) (65). As visualized using EM, the exocytosed CCL21 is then retained focally in the form of minute puncta at the basolateral surface of the endothelium close to intercellular junctions where they are thought to induce local arrest of DCs through β2 integrin activation. Such confined release may well avoid the desensitization of CCR7 on the migrating cells that might otherwise be evoked by formation of a conventional transendothelial CCL21 gradient (133). Diapedesis—the actual transit process, then proceeds by CCL21-driven re-arrangement of the DC actomyosin cytoskeleton that allows the cell to push and squeeze through the flap-like protrusions between adjacent oakleaf shaped endothelial cells in the initial capillary junctions (65), aided by Semaphorin 3A induced signaling via RhoA and ROCK for contraction of the uropod (134). The CCL21 secreted focally in response to DC contact is thus functionally distinct from the CCL21 that is secreted homeostatically for interstitial migration. Interestingly, the DC adhesion-induced CCL21

FIGURE 3 | Local secretion of CCL21 from lymphatic endothelium triggered by DC contact. Contacts with the basolateral surface of lymphatic endothelium via filopodia extended from the migrating DCs trigger an influx of Ca2<sup>+</sup> that leads to disassembly of the adjacent cortical actin network and enables translocation of endothelial trans-Golgi vesicles containing pre-stored CCL21 to the plasma membrane along parallel microtubule tracks. Fusion of the vesicles with the plasma membrane, most likely orchestrated by Ca2<sup>+</sup> binding synaptotagmins, leads to the local release of CCL21 which is retained on the plasma membrane surface close to the endothelium:DC interface, where it directs DC transit through β2 integrin activation and actomyosin-mediated pushing and squeezing, upon the formation of LYVE-1 and ICAM-1/VCAM-1 lined transmigratory cups (see text and Figure 2). The identity of the adhesion molecules involved in making initial contact between DC and lymphatic endothelium are not currently known.

exocytotic mechanism described by Vaahtomeri et al. does not appear to be initiated by integrins. Indeed, the identities of the receptors on DCs and endothelium responsible have yet to be determined. Moreover, it is likely that the process of DC-induced CCL21 secretion orchestrates the assembly of LYVE-1, ICAM-1, and VCAM-1 enriched endothelial transmigratory cups, as independent studies reported that corresponding structures consistently formed close to CCL21 puncta in LEC monolayers in vitro, and their assembly was blocked by CCL21 neutralizing antibody (76, 85, 90, 135). It remains to be determined whether other leucocyte populations such as CD4<sup>+</sup> T cells can trigger local CCL21 secretion in a similar manner to DCs and whether the coupling of chemokine release and transmigratory cup formation is a general phenomenon for vessel entry.

### Additional Inflammatory Chemokines and Exosome—Mediated Secretion

Besides CCL21, LECs synthesize a variety of other chemokines including CCL1, CCL2, CCL5, CCL20, CXCL12, and CX3CL1 that are chemotactic for T cells, DCs and monocytes expressing the G-protein coupled receptors CCR8, CCR2, CCR5, CCR4, and CX3CR, respectively, and CXCL1, CXCL2, CXCL5, and CXCL8 that are chemotactic for neutrophils that express CXCR1 and CXCR2. In common with CCL21 these are all upregulated by exposure of the endothelium to inflammatory cytokines or other inflammatory stimuli (82, 83). In most cases however, the mechanisms underlying their release and extracellular localization in relation to junctional entry sites are not so well-understood. Amongst those chemokines that have been studied in any significant detail, CXCL12 (SDF-1) has been reported to direct the entry of DCs and epidermal Langerhans cells to lymphatic vessels in mice and migration to dLNs as assessed by FITC skin painting (136, 137). More specifically, its receptor CXCR4 was shown to be highly expressed by lymph-migrating MHC class II<sup>+</sup> DC in skin and coadministration of a synthetic CXCR4 antagonist (4-F-benzoyl-TN14003) impaired their migration to dLNs and capacity to promote T cell proliferation after contact hypersensitization with the hapten DNBS (136). Nevertheless, the authors concluded that the role of CXCL12/CXCR4 axis, although operating in parallel for DC migration was subordinate to that of CCL21 and CCR7. Furthermore, CXCL12 does not appear to control the exit of T cells to afferent lymphatic vessels in inflamed skin, despite the fact they express CXCR4 and exhibit responsiveness to the chemokine in vitro (138). Additionally, CCL1, which regulates DC and tumor cell transit across the LN SCS via its primary receptor CCR8, has also been implicated in directing transmigration of monocytes and monocyte derived DCs across inflamed peripheral lymphatics (139–141).

More recently, CX3CL1 was shown to promote both in vitro transmigration and in vivo lymphatic entry of CX3CR<sup>+</sup> monocyte-derived DCs in the skin of oxazolone-hypersensitised mice in parallel with CCL21 (142). Unique amongst chemokines, CX3CL1 is synthesized as a membrane-anchored molecule that is subsequently cleaved by proteases including the disintegrin and metalloproteases ADAM10 and ADAM17 to generate a conventional soluble chemoattractant, and it is this form that is released basolaterally from cytokine-activated endothelium (142). Moreover, as reported within the last few months (143), CX3CL1 is also secreted from lymphatic vessels in CD9<sup>+</sup> and CD63<sup>+</sup> exosomes that form halos around the periphery of lymphatic vessels in inflamed mouse and human tissues. Intriguingly, these exosomes which carry the membraneanchored form of CX3CL1 on their surface can elicit cellular protrusions in monocyte-derived DCs and promote their transmigration across human dermal LEC (HDLEC) monolayers, in co-operation with CCL21, as well as entry to intact lymphatic vessels in ex vivo exposed skin (143). Why lymphatic vessels should employ two such different modes of CX3CL1 release is unclear. However, as exosomes act as vehicles for the release of chemokines other than CX3CL1 it is possible they direct the entry of multiple different inflammatory leucocyte populations.

Interestingly, most of the remaining chemokines such as CCL2, CCL5, CCL20, and the neutrophil chemokines CXCL2, CXCL5, and CXCL8 are preferentially secreted from the luminal rather than the basolateral face of lymphatic endothelium (17, 56, 66, 132, 142), unlike CCL21 and CX3CL1, and hence it is currently unclear how they might regulate leucocyte entry to afferent lymphatic vessels. Indeed, the reduced trafficking of epidermal Langerhans cells and CD8<sup>+</sup> dermal DCs to dLNs observed in CCR2−/<sup>−</sup> and to a lesser extent in CCR5−/<sup>−</sup> mice

Based on a combination of in vitro studies with HDLECs and in vivo studies in mice, the model depicts the guidance of neutrophils toward the basolateral surface of endothelium following inflammation-induced release of the chemokine CXCL8 (IL-8) which transmits signals through its G protein coupled receptors CXCR1 and CXCR2 to activate the neutrophil β2 integrin LFA-1 and engage its endothelial counter-receptor ICAM-1. This in turn triggers local secretion of the matrix metalloproteinases MMP8 and MMP9 and neutrophil elastase (which also bind to the endothelial surface) and exocytosis of the chemorepellent 12(S)-HETE that together promote transient junctional retraction and neutrophil transmigration.

appeared to result from their accumulation inside rather than outside dermal lymphatics and hence it is more likely that such chemokines direct intraluminal crawling rather than initial vessel entry (144).

Finally, the steady-state levels of secreted chemokines on and around lymphatic capillaries are regulated by a group of "atypical" chemokine receptors present in lymphatic endothelium that lack signaling capacity and act primarily as chemokine scavengers. This group which includes ACKR1 (Duffy antigen), ACKR2, formerly known as D6, ACKR3 (CXCR7), and ACKR4 (CCRL1) bind and internalize inflammatory CC chemokines and prevent their inappropriate accumulation on the surface of lymphatic capillaries while also helping to establish the polarity of their gradients (145). Accordingly, the action of ACKR2 which scavenges the inflammatory CC chemokines CCL2, CCL3, CCL4, and CCL5 is thought to aid in the preferential entry of mature activated CCR7<sup>+</sup> DCs via CCL21, largely by preventing an accumulation of CCR2/CCR5 macrophages at the vessel surface that might otherwise block their access (146, 147). It has also been posited that ACKR4 helps preserve the responsiveness of DCs to CCL21 during entry to dermal lymphatics in inflamed tissues by scavenging CCL19 released from stromal cells and preventing a build-up of the chemokine that might otherwise de-sensitize CCR7 and lead to DC stasis (148).

### Chemoattractive Guidance for T Cells via Sphingosine-1-Phosphate and Lymphotoxins

Besides conventional chemokines, T cells also engage two other chemoattractive pathways for vessel entry, that function cooperatively with CCL21 and may impinge at least partly on the integrin: ICAM/VCAM mediated transmigratory mechanism described above. The first involves the chemotactic lipid sphingosine 1 phosphate (S1P), best known as a regulator of lymphocyte exit from lymph nodes but which appears also to be a negative regulator of CD4<sup>+</sup> T cell entry to initial lymphatics in inflamed peripheral tissues. Normally present at high concentrations in lymph and low concentrations in lymph nodes, the resulting gradients of S1P direct transit of T cells bearing the G-protein coupled receptor S1PR to efferent lymph (149, 150). However, in a murine alloantigeninduced model of inflammation, increased synthesis of S1P in peripheral tissue or administration of the S1PR1 functional antagonist FTY720 (fingolimod) signals retention and arrest of CD4<sup>+</sup> T cells, which logjam around the basolateral surface of initial lymphatic capillaries (151). Moreover, as shown using in vitro transmigration assays with monolayers of the lymphatic endothelial like cell line SVEC4-10, S1P treatment of T cells blocked their transit and caused their arrest through β2 integrinmediated adhesion to ICAM-1 and VCAM-1 (151). Whether this S1P/S1PR driven chemotactic mechanism also operates in conjunction with CCL21/CCR7 to positively regulate T cell transmigration and whether ICAM-1 and VCAM-1 are recruited to structures similar to or distinct from DC transmigratory cups during S1P mediated transmigration/arrest remains uncertain.

Secondly, the main immunoregulatory TREG population that migrates between tissues and lymph nodes to maintain peripheral tolerance and immune suppression have been shown to employ lymphotoxin (LT), a lymphokine member of the TNF superfamily, to transit inflamed lymphatic vessels by engaging its signal transducing receptor LTβR in lymphatic endothelium (152, 153). Best known for its role in inducing lymph node neogenesis via lymphoid tissue inducer (LTi) cells, the trimeric lymphotoxin molecule is expressed on the surface of TREG as a membrane-bound LTα1β<sup>2</sup> heterotrimer. Moreover, genetic deletion of LTα in murine TREGs or treatment with soluble LTβR Ig fusion protein was shown to disrupt their entry to lymphatic capillaries in mouse skin, while leaving the transit of naïve CD4 and CD8 cells unaffected (153). More specifically, the interaction between LT and LTβR in vitro induces the rapid extension of VCAM-1 enriched lamellipodia-like protrusions akin to transmigratory cups in lymphatic endothelium, that appear to facilitate TREG transmigration (153). Curiously however, the VCAM counter-receptor involved in LT-mediated TREG transit has not been identified. While RNA array data suggest that TREGs express much higher levels of LTα than other T cell types (154), it remains unclear whether this mechanism is used by other T or B cell populations for lymphatic transmigration, or indeed whether the LTβR is redistributed to TREG transmigratory structures in lymphatic endothelium together with LYVE-1 and VCAM-1 must await further investigation.

#### TRANSMIGRATION VIA CHEMOREPULSION—THE UNUSUAL MECHANISM USED BY NEUTROPHILS

In comparison to other leucocyte populations, neutrophils deploy a particularly unique and complex mechanism to enter lymphatic capillaries. While studies in mice harboring bacterial infections have shown that entry involves adhesion via β2 integrins, like that of DCs and T cells, more detailed in vitro studies using inflamed human LEC monolayers and mouse tissue explants revealed that such adhesion is just the first in a coordinated series of events that induce release of neutrophil elastase and the matrix metalloproteinases MMP8 and MMP9 and focal secretion of the arachidonate-derived chemorepellant lipid 12-hydroxyeicosatetraenoate (12(S)HETE), that together evoke local endothelial junctional retraction (**Figure 4**) (56, 155). These studies indicated the openings created by such retraction were transient and resolved spontaneously without any attendant cell death. Moreover, they served as portals for enhanced transit of successive waves of neutrophils which formed loosely attached swarms over the LEC monolayers resembling those previously described in lymph nodes of infected mice. Notably, the rate of in vitro neutrophil transmigration was some 10× fold higher than that of DCs, even in neutrophil preexposed LEC monolayers, suggesting they exclusively target such portals for entry without inducing significant endothelial damage (56). The contact dependent nature of this neutrophil-induced activation process is curiously reminiscent of the contact-induced secretion of CCL21 by transmigrating DCs. Indeed, neutrophil adhesion to human LEC monolayers triggered the secretion of several chemokines including CXCL2, CXCL5, and CXCL8 from the basolateral surface that could potentially direct in vitro neutrophil transit, although experiments with neutralizing mAbs revealed only CXCL8 (IL-8) fulfilled this role (56). Whether a murine IL-8 ortholog guides neutrophil transmigration similarly in mice along with the recently reported actions of CCL21 and CXCL12 (43, 55, 91, 136, 137) remains to be determined. Importantly, neutrophils as distinct from other lymph migrating leucocyte populations do not synthesize an HA glycocalyx, and hence cannot engage LYVE-1 transmigratory cups for vessel entry. The likely significance of the alternative transmigration mechanism is that it offers a far more rapid mode of lymphatic entry than through the button junctions of initial capillaries, in keeping with the primary function of these cells in the rapid response to tissue injury. It is also noteworthy that an analogous lipoxin-mediated process of endothelial retraction has also been described for lymphatic metastasis of tumors in mice, and certain human cancers (156).

#### INTRALUMINAL CRAWLING

Having transited the lymphatic endothelium, migrating leucocytes enter the vessel lumen and begin their onward journey to the dLNs. In initial capillaries, which are noncontractile and lack smooth muscle investment, the rate of lymph flow has been variably estimated as up to 200–300 µm/min (157, 158), some 2–3 orders of magnitude slower than in blood capillaries and sinusoids, and only marginally exceeding that in the interstitium (69). It is not until these capillaries merge into downstream contractile collectors that the flow rates even approach those of blood vessels. Surprisingly, intravital imaging studies revealed that DCs migrate within initial capillaries (6–8 µm/min) at an even slower rate than lymph itself, and that the majority of newly transmigrated cells are not conveyed by passive flow, but rather crawl along the luminal surface of the lymphatic capillaries until they enter downstream collectors (69–71, 159). Moreover, the crawling leucocytes exhibit semi-random patterns of migration, frequently changing direction before moving downstream (160). Indeed, using time-lapsed microscopy of YFP tagged DCs in the ear tissue of lymphatic reporter mice, it was confirmed that the rate and directionality of intraluminal crawling is almost completely unaffected by changes in lymph flow and even proceeds in its absence (160). Instead, crawling was shown to be driven by chemotaxis/haptotaxis, guided by physical gradients of CCL21 sequestered on the luminal surface of the vessel, as evidenced by confocal and immune EM imaging of mouse dermis, and by the demonstration that its downstream directionality in mouse dermal lymphatics was abrogated by CCL21 blocking mAbs or CCR7 gene deletion in the case of both endogenous and adoptively transferred DCs (160). The spontaneous establishment of such intraluminal gradients was elegantly demonstrated in vitro by flow chamber experiments with transfected LECs which affirmed that under levels of shear close to those of afferent lymph (0.015 dynes/cm<sup>2</sup> ), fluorescent CCL21 secreted from the luminal surface of the endothelium underwent re-binding downstream to form authentic, directionally oriented gradients for DC migration (160) (see also **Figure 1**). The macromolecules responsible for sequestering the CCL21 gradient on the endothelium likely include the lymphatic marker podoplanin (161) that reportedly binds the chemokine with high affinity (K<sup>D</sup> 70 nM) in Biacore analyses (162), and HSPGs, whose digestion or specific knockdown has been shown to disrupt CCL21-dependent DC adhesion to LEC monolayers under flow (163–165). Thus, lymph flow generates the chemotactic gradients which drive intraluminal crawling, rather than conveying cells through physical propulsion. The necessary traction for crawling is provided by β2 integrin-mediated adhesion to ICAM-1 on the luminal surface of the endothelium, whose expression is upregulated in inflammation and whose blockade by mAbs was shown to reduce DC crawling velocity in vitro (69). Like DC migration in the interstitium, efficient intraluminal crawling of DCs in inflamed lymphatics also depends on signaling via the Rho-associated protein kinase ROCK for dissociation of β2 integrin from ICAM-1 and uropod retraction in a continual process of adhesion and detachment from the endothelium (69).

As regards other leucocyte populations, semi-directional crawling behavior has also been observed for CD4<sup>+</sup> T cells migrating in the lumen of initial dermal lymphatic capillaries, as evidenced in a recent study that used time-lapsed imaging to track CD2 DsRed fluorescent lymphocytes in Prox1 GFP mice reporter mice (76). Furthermore, the speed of T cell intraluminal crawling and the degree of motility were both markedly increased in the inflamed dermal lymphatics of skin contact hypersensitised mice, supported again by integrin-mediated adhesive interactions with ICAM-1 on the inner surface of the endothelium (76). Just as described for DCs, these T cells became detached from the luminal endothelium as the initial capillaries merged with downstream collectors where they were passively drawn into lymph flow by vessel pumping. Neutrophils also crawl within the lumen of initial lymphatic capillaries in a mostly downstream direction toward lymphatic collectors, and at broadly comparable velocities (mean 6–13 µm/min) to DCs and T cells. In common with these other leucocyte populations, migration was shown to be directed by haptotactic gradients of CCL21 sequestered on the capillary floor in the direction of lymph flow, with traction provided by β2 integrin mediated ICAM-1 adhesion (91). The faster recruitment of neutrophils to dLNs compared to DCs and T cells via afferent lymph may therefore be the product of more rapid mobilization in the tissues, the ability to translocate stored CCR7 from intracellular vesicles rather than relying on de novo synthesis (43), and a more efficient mode of endothelial transmigration.

On passing from initial capillaries to the smooth muscle invested pre-collectors and collectors, migrating leucocytes encounter a large increase in lymph flow rate (>1 mm/min) that likely renders intraluminal crawling redundant, and hence it is thought the cells are conveyed toward downstream lymph nodes by passive lymph flow.

#### ARRIVAL AND TRANSIT AT LNS

The major destination for leucocytes migrating through tissue lymphatics is the lymph node, an organ the size of a small bean in most peripheral tissues. With the exception of recirculating TRCM cells which reside transiently before returning to the blood circulation via efferent lymphatics, most leucocytes reaching the nodes proceed no further and having fulfilled their immune function, ultimately die there. In either case the cells arrive into the subcapsular sinus (SCS), a labyrinthine compartment continuous with the afferent lymphatics that is situated just beneath the outer capsule of the node. From there, they transit across the SCS endothelium to access the T and B cell-rich cortex, in the case of DCs and neutrophils to prime or re-activate T cell immune responses, and in the case of TMEM and TREGs to influence or downregulate such responses, before they egress and circulate back to the tissues (see **Figure 1**). While exit from the nodes is known to be directed by the sphingosine 1-phosphate (S1P) receptor S1PR on recirculating cells and S1P in efferent lymph (150, 166, 167), it has often been assumed that cell entry from afferent lymph is a more passive process. However, it is becoming increasingly clear that the SCS endothelium represents a checkpoint for nodal entry that is regulated both by chemokines and adhesion receptors.

In one particularly informative study using mouse eGFP tagged leucocytes microinjected into pre-nodal (popliteal) lymphatics, it was found that DCs and CD4<sup>+</sup> T cells used separate routes to transit across the SCS (168). Whereas, DCs invariably crossed directly through the floor of the SCS to the cortical zones, T cells instead continued to the adjoining medullary sinuses, previously considered as exit routes from the nodes, before transmigrating to the underlying parenchyma (168). Moreover, DC transmigration across the SCS was shown to be directed by CCR7/CCL21 dependent chemotaxis, whereas CD4<sup>+</sup> T cells required this chemokine receptor pair only for subsequent haptotactic crawling within the underlying parenchyma (168). Curiously however, when CD4<sup>+</sup> T cells were co-injected into mice, the former then switched to the SCS route for transmigration, suggesting that DCs in some way remodel the endothelium during transit. More recent work has revealed that the role of CCL21 in directing DC transit can also be aided or augmented by other chemokines. Notably CCL1 released from the floor of the SCS was reported to induce the entry of monocyte-derived DCs to the LN parenchyma via its cognate receptor CCR8 (139). More recently, in Th2 immunized mice, the alternative CCR8 ligand CCL8 released from CD169<sup>+</sup> SCS macrophages was shown to potentiate transit of DCs across the SCS by enhancing CCR7 signal transduction (141). Significantly, polarized expression of the alternative CCL19/CCL21 scavenging receptor CCRL1 (ACKR4) in the ceiling of the SCS was reported to maintain the gradient of CCL21 inside the SCS that drives DC transit, as its deletion led to trapping of DCs within the SCS lumen (169).

Knowledge of the contribution made by adhesion receptors in leucocyte transmigration across the SCS floor is still rather sketchy in comparison with peripheral lymphatics. Curiously, a key role has been identified for the transmembrane protein PLVAP (aka Pal-E, MECA 32) that forms diaphragm-like structures in the SCS and certain blood vessels and which primarily regulates the size selective entry of macromolecules to the underlying cortical fibrillar conduit network for presentation by LN HEVs (170). Notably, deletion of PLVAP was found to result not only in the uncontrolled entry of small macromolecules to the cortex, but also the enhanced transmigration of injected splenic T cells. Based on in vitro studies with isolated SCS endothelium, it was concluded that PLVAP diaphragms guard entry portals remote from VE-cadherin buttoned junctions, through which T cells transmigrate by extension of their leading edges (170). Whether PLVAP plays a direct or indirect role in the process and also mediates transit of DCs and other leucocytes remains unknown. More recently, transcriptional profiling has identified further candidate receptors that might regulate the differential transit of leucocytes across SCS and medullary sinuses. In particular, the macrophage scavenger receptor MSR1 (CD203) was found to be selectively expressed in SCS, where it was shown to mediate T cell adhesion in ex vivo assays with frozen sections. Like PLVAP however, the receptor appears to act as a regulator of SCS transit rather than a gatekeeper, as the process was enhanced not retarded in MSR−/<sup>−</sup> mice (171). Studies have yet to determine whether transit through SCS or other LN sinuses involves leucocyte adhesion via β2 integrins in resting or inflammatory conditions, or interactions between the leucocyte HA glycocalyx and LYVE-1 as in the case of initial lymphatic capillaries. Nevertheless, the observation that transit of DCs is accompanied by marked morphological changes in the floor of the SCS endothelium, including modulation of LYVE-1 and realignment of the SCSlining CD169<sup>+</sup> macrophages is indicative that the process is complex and that these and other adhesion receptors may well play contributory roles (36). Finally, it should be stressed that dLN sinuses and the surrounding afferent lymphatic network undergo considerable expansion following antigen challenge or inflammation in peripheral tissues, through a process of lymphangiogenesis driven by VEGF-A released primarily by B cells and macrophages arriving through afferent lymph (172– 174). This is accompanied by a transient increase in LN size, cellularity and lymph flow that induces DC mobilization and migration, and markedly enhances DC transit into the deep underlying LN cortex and paracortex for lymphocyte activation (174–176). It is highly likely that such changes also enhance the entry of other migrant leucocyte populations to dLNs and it is hoped that future research in this area will yield much needed insights into the underlying mechanisms.

#### SUMMARY AND CONCLUSIONS

Over the past decade, research using new techniques and animal models for tracking and imaging cell migration, combined with the efforts of a wide interdisciplinary community of interested scientists, has led to huge advances in our understanding of leucocyte trafficking in the lymphatic system and its immune significance. As this pace of advance seems set to continue in the immediate future, we can anticipate that the resulting mechanistic insights will translate into new targets and therapies for immune disorders and even the treatment of lymph metastasising cancers.

As outlined in this text, we now have detailed insight into the first key step in such trafficking, the entry of cells to the lymphatic vessels. In the case of DCs, this has revealed an intricate and closely co-ordinated mechanism in which physical contact of migrating leucocytes with lymphatic endothelium triggers the local exocytosis of CCL21 and formation of LYVE-1<sup>+</sup> transmigratory cups which envelop the migrating cells and promote their transit into the vessel lumen. Moreover, parallel observations that transmigrating T cells and macrophages elicit the formation of similar endothelial protrusions containing ICAM-1, VCAM-1 and/or LYVE-1, and reliance on an HA glycocalyx or β2 integrin adhesion, raise the possibility that lymph-migrating leucocytes exploit a common mechanism for vessel entry (76, 85, 101). Indeed, this could be considered as a form of "lymphatic synapse," through which appropriate input from other chemokine receptors or signaling components such as lymphotoxin/LTβR and S1P/S1PR1 might direct selective entry of TREGs or other lymph migrating cell populations including Natural Killer (NK) cells and innate lymphoid cells (ILCs). However, the notion of a synapse may not apply to neutrophils which use a unique mechanism of integrin-dependent proteolysis and lipoxinmediated endothelial retraction to "invade" lymphatic vessels. This unusually specialized process may have evolved to enable these professional phagocytes to exit almost instantaneously from sites of infection and reach the dLNs well-ahead of slower migrating DCs. In comparison, we know little about how the docking and adhesion of transmigrating leucocytes is choreographed and how the many "accessory" adhesion molecules including the Mannose receptor, ALCAM, CLEVER-1, CD31, CD99, and others (**Table 1**) integrate with key cup-forming components such as ICAM-1, VCAM-1, and LYVE-1. Though most leucocytes that employ such cups have been observed to enter the afferent lymphatics at button junctions in initial vessels, it seems unlikely that these are the only junctional types to allow entry. During inflammation-induced lymphangiogenesis for example, newly sprouting vessels assemble zippered rather than buttoned junctions, and in chronic inflammation, when leucocyte traffic via lymph is markedly increased, zippers replace buttons. It will be interesting to determine whether leucocytes have the ability to enter through zipper junctions and whether the process involves different molecular mechanisms to buttons and transit via a transcellular or paracellular route. A further priority will be to ascertain whether these mechanisms of lymphatic entry are universally applicable or vary between tissue beds such as the intestines, brain and central nervous system (177), given that most of our current insight has been gained from studies on mouse dermis, due to its greater accessibility.

Recent research has also provided surprising insight into how leucocytes, having entered the initial lymphatics, migrate within the vessel lumen toward downstream lymph nodes. Rather than being conveyed by passive lymph flow, it is now apparent that DCs, CD4<sup>+</sup> T cells and neutrophils actively crawl along the internal surface of initial vessels using guidance from CCL21 and transient integrin-mediated adhesion. Yet why such mechanisms should have evolved to deliberately slow the downstream progress of antigen presenting and immune effector cells is unclear. Might the intimate contact with endothelium imposed by intraluminal crawling enable for example en route uptake of foreign antigens or maturation signals by DCs, or MHC-mediated antigen presentation to recirculating T cells or TREGs for immune tolerance? Could it provide a platform for interactions between different leucocyte populations themselves? Or might intraluminal crawling provide DCs or T cells with the option to exit and re-enter lymphatic vessels prior to reaching dLNs in order to sample or respond to antigen in the surrounding tissues? Though not yet reported in initial lymphatics vessels, there is evidence that migrating leucocytes exit collectors in adipose tissue during bacterial infection, and that DCs which interdigitate the vessel endothelium to regulate vessel permeability and sample the surrounding tissue can subsequently detach and migrate to dLNs (178).

With the many insights into lymphatic trafficking and its immune consequences that have been gained from basic research, there is increasing scope for clinical translation and the development of new immune based therapies based on migration blockade. Judicious targeting of integrins, ICAM-1, chemokine receptors, S1P and particularly lymph-specific adhesion molecules such as LYVE-1 by appropriate blocking mAbs may be envisaged as therapeutic strategies for transplant rejection, to prevent DC migration from engrafted tissues and consequently activation of alloimmune responses in host dLNs. This is particularly applicable in the case of corneal allografts, where such therapies could be applied locally, thus avoiding offtarget effects associated with systemic antibody administration. In the case of TREGs, the inclusion of lymphotoxin blockade to impair their migration to lymph nodes for immune suppression might also provide an adjunct to checkpoint inhibitors for tumor immunotherapy. As a corollary, our understanding of the main factors regulating DC migration from inflamed tissues could be exploited to optimize vaccine delivery. For many years the notion of using DCs as adjuvants for adoptive cancer immunotherapy has fuelled efforts to enhance their maturation ex vivo using TLR ligands and inflammatory cytokines and to optimize antigen loading and presentation. Combining these approaches with pre-conditioning of vaccination sites

#### REFERENCES


to boost lymphatic vessel density, the efficiency of DCs for vessel entry and nodal transport should lead to much greater clinical efficacy. Likewise, boosting the exit of macrophages via lymphatics could aid in the resolution of inflammation and tissue recovery in conditions such as myocardial infarction where their delayed removal results in a failure of cardiomyocyte replenishment, and an increase in tissue scarring and fibrosis (41, 179–181).

#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and has approved it for publication.

#### FUNDING

DJ gratefully acknowledges financial support in the form of Unit Funding (Grant MC\_UU\_00008/2) from the UK Medical Research Council.

#### ACKNOWLEDGMENTS

The author thanks all his previous and present research team members for their seminal contributions to this work, and colleagues including Prof. Paul Riley, Dr. Cornelia Halin, Dr. Peter Baluk, and Prof. Ralf Richter for continued collaborations and helpful discussions.


phenotype and migratory pattern. J Immunol. (2013) 190:970–6. doi: 10.4049/jimmunol.1202805


of epidermal Langerhans cells to regional lymph nodes. Cell Immunol. (1994) 158:389–99. doi: 10.1006/cimm.1994.1285


the lymphatics by activating myosin II. Nat Immunol. (2010) 11:594–600. doi: 10.1038/ni.1885


nodes to suppress the alloimmune response. Immunity. (2009) 30:458–69. doi: 10.1016/j.immuni.2008.12.022


vasculature identifies candidates for manipulation of cell traffic. Proc Natl Acad Sci USA. (2016) 113:10643–8. doi: 10.1073/pnas.1602357113


**Conflict of Interest Statement:** The author declares 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 © 2019 Jackson. 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(s) 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.

# Regulation of Immune Function by the Lymphatic System in Lymphedema

Raghu P. Kataru, Jung Eun Baik, Hyeung Ju Park, Itay Wiser, Sonia Rehal, Jin Yeon Shin and Babak J. Mehrara\*

Division of Plastic and Reconstructive Surgery, Department of Surgery, Memorial Sloan Kettering Cancer Center, New York, NY, United States

The lymphatic vasculature has traditionally been thought to play a passive role in the regulation of immune responses by transporting antigen presenting cells and soluble antigens to regional lymph nodes. However, more recent studies have shown that lymphatic endothelial cells regulate immune responses more directly by modulating entry of immune cells into lymphatic capillaries, presenting antigens on major histocompatibility complex proteins, and modulating antigen presenting cells. Secondary lymphedema is a disease that develops when the lymphatic system is injured during surgical treatment of cancers or is damaged by infections. We have used mouse models of lymphedema in order to understand the effects of chronic lymphatic injury on immune responses and have shown that lymphedema results in a mixed T helper cell and T regulatory cell (Treg) inflammatory response. Prolonged T helper 2 biased immune responses in lymphedema regulate the pathology of this disease by promoting tissue fibrosis, inhibiting formation of collateral lymphatics, decreasing lymphatic vessel pumping capacity, and increasing lymphatic leakiness. Treg infiltration following lymphatic injury results from proliferation of natural Tregs and suppresses innate and adaptive immune responses. These studies have broad clinical relevance since understanding how lymphatic injury in lymphedema can modulate immune responses may provide a template with which we can study more subtle forms of lymphatic injury that may occur in physiologic conditions such as aging, obesity, metabolic tumors, and in the tumor microenvironment.

Keywords: lymphatic vessels, immune function, Th2 type T cells, inflammation, fibrosis

### THE LYMPHATIC SYSTEM DIRECTLY AND INDIRECTLY REGULATES IMMUNE RESPONSES

The lymphatic system is comprised of a series of blind ended, single cell thick initial lymphatic vessels that drain progressively into successively larger vessels and eventually return interstitial fluid back to the systemic circulation. Lymphatic flow is regulated by coordinated pumping of smooth muscle cells that partially envelop collecting lymphatics and compressive forces from surrounding skeletal muscles. One-way valves in collecting lymphatics ensure forward flow of interstitial fluid and prevent reflux when a segment of the collecting vessel located between two valves contracts (1).

In addition to draining interstitial fluid, the lymphatics system is responsible for lipid and fatty acid absorption and is an important regulator of cholesterol metabolism (2). The lymphatic system

#### Edited by:

Beth Ann Tamburini, University of Colorado Denver, United States

#### Reviewed by:

Shan Liao, University of Calgary, Canada Epameinondas Gousopoulos, ETH Zürich, Switzerland

> \*Correspondence: Babak J. Mehrara mehrarab@mskcc.org

#### Specialty section:

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

Received: 23 December 2018 Accepted: 21 February 2019 Published: 18 March 2019

#### Citation:

Kataru RP, Baik JE, Park HJ, Wiser I, Rehal S, Shin JY and Mehrara BJ (2019) Regulation of Immune Function by the Lymphatic System in Lymphedema. Front. Immunol. 10:470. doi: 10.3389/fimmu.2019.00470

**60**

also regulates immune responses by transporting bacteria, foreign antigens, particulate matter, exosomes, and immune cells to regional lymph nodes and lymphoid structures (3). Regulation of immune responses occurs at multiple levels and is both active and passive in nature. Active mechanisms of immune response regulation by the lymphatics includes regulation of immune cell entry and migration through the lymphatic system by LEC cytokine, chemokine, and adhesion molecule expression. In addition, LECs modulate immune responses and regulate autoimmunity by transferring self-antigen to DCs (4), or by directly inducing T cells tolerance using their PD-L1 molecule or MHC II-self antigen peptide complex that acquired from DCs (5–7). The lymphatic system can also control immune responses indirectly by modulating the rate at which antigens and cells are delivered to regional lymph nodes by regulating lymphatic vessel tone and pumping (8–11).

Given the important role of the lymphatic system in a wide range of physiologic processes, it is not surprising therefore that abnormalities in lymphatic function have been implicated in inflammatory disorders (12, 13), immune tolerance (14), metabolic abnormalities such as obesity and metabolic syndrome (2), cardiovascular disease including hypertension and atherosclerosis (15), cancer growth and metastasis (16– 18), infectious diseases (19, 20), and septic shock (21). Genetic, iatrogenic, traumatic, or infectious abnormalities of the lymphatic system cause severe complications including lymphedema, chylous ascites, chylothorax, and lymphatic vascular anomalies. Recent findings suggest that many of these abnormalities are related not only to changes in lymphatic fluid transport function, but also lymphatic regulation of immune responses.

#### LYMPHATIC FUNCTION IS VARIABLE AND CAN REGULATE IMMUNE RESPONSES

Lymphatic function is highly variable clinically and modulated by numerous factors including chronic inflammation, tumors, external stimuli such as radiation, age, obesity, and metabolic dysfunction. For example, reports published in the late 1990s showed that aging results in structural changes in the lymphatic system including loss of elasticity, reduced smooth muscle coverage, decreased number of mesenteric collecting vessels, and decreased mesenteric lymphatic flow (22, 23). More recent studies have shown that aging results in ultrastructural changes in collecting lymphatics resulting in tissue degeneration and loss of extracellular matrix components, decreased expression of contractile and regulatory proteins, and increased lymphatic vascular permeability (24, 25). These structural changes, together with changes in gradients of eNOS, iNOS, and histamine significantly decrease aging lymphatic vessel contraction, interstitial fluid transport function, transport of pathogens to regional lymph nodes, and clearance of macromolecules from the central nervous system (26, 27). Similar changes in lymphatic function have been reported in obesity. For example, obesity results in structural and physiologic changes in the lymphatic system including increased lymphatic leakiness, decreased collecting vessel contractility, and decreased lymph node size and changes in lymph node architecture (2). Obese patients have decreased clearance of interstitial fluid as compared to lean individuals (28), obesity increases the risk of developing lymphedema after surgery (29), and severe obesity can lead to the spontaneous development of lymphedema (30). Interestingly, obesity induced lymphatic abnormalities decrease adaptive immune responses and are reversible with treatments that promote lymphangiogenesis and increase lymphatic transport (31). These findings are important because they suggest that common comorbid conditions have significant effects on the lymphatic system and these changes in turn significantly modulate immune responses.

Variability in lymphatic function resulting from aging, obesity, or metabolic syndrome may play a key role in immune responses to solid tumors and provide a rationale for the fact that these comorbid conditions increase the risk of tumor development and metastasis. Solid tumors such as melanoma and breast cancer are surrounded by abnormal, leaky lymphatics with impaired lymphatic transport function. Tumor, draining lymph node lymphangiogenesis and increased VEGF-C expression by inflammatory cells increase tumor growth and metastasis. Lymphatic vessel density and VEGF-C expression correlates with cytotoxic T cell infiltration and expression of immunosuppressive factors (iNOS, IDO, Arg-1) in patients with melanoma indicating a possibility of LECs playing a dual role in promoting and hindering anti-tumor responses (32). These changes are associated with increased risk of local/regional tumor recurrence and decreased survival. Intradermal implantation of melanoma in mice that lack dermal lymphatics due to transgenic expression of K-14 VEGFR3-Ig results in more rapid tumor growth locally, decreased distant metastasis, and decreased inflammatory cell infiltration, and impaired dendritic cell migration to regional lymph node (33). Interestingly, K-14- VEGFR3-Ig mice had impaired tumor specific immune responses after vaccination. Lymphatic endothelial cell (LEC) presentation of tumor antigens on major histocompatibility complex proteins (MHCI or MHCII) in the context of PD-L1 (checkpoint molecule programmed death ligand 1) and the absence of co-stimulatory molecules results in suppression of T cell mediated immune responses by decreasing T cell activation and proliferation and increasing apoptosis (4, 34). Taken together, these findings suggest that tumor lymphatics regulate tumor immune response and modulate the tumor microenvironment (35).

### LYMPHEDEMA RESULTS IN CHRONIC INFLAMMATION

Lymph node dissection for cancer treatment is the most common cause of lymphedema development in Western Countries. Because lymphedema in this scenario develops secondary to surgical injury, this type of lymphedema is referred to as secondary lymphedema. Patients with secondary lymphedema develop progressive fibroadipose deposition in the affected limb and have an increased risk of developing infections and secondary malignancies. These pathologic changes cause significant morbidity and decrease quality of life (36). It is estimated that 20–40% of patients who undergo treatment for solid malignancies such as breast cancer, melanoma, gynecological or urologic tumors, or sarcomas go on to develop lymphedema (37). Because these cancers are common, there is a large number of patients who are diagnosed with lymphedema annually. This fact, together with the fact that lymphedema is a life-long disease and survival following cancer treatment has significantly improved, is responsible for the increasing number of patients who suffer from lymphedema. Although estimates of the number of patients who suffer from secondary lymphedema are variable and range between 5 and 10 million individuals, it is important to note that even the most conservative estimates make lymphedema among the most common chronic disorders and the most common long-term complication of cancer treatment.

Historically, development of lymphedema has been thought to be secondary to impaired development of collateral lymphatics that bypass the zone of injury. Indeed, this concept led to the multiple preclinical studies reporting on the use of exogenous lymphangiogenic growth factors as a therapeutic treatment for lymphedema (38–40). However, more recent studies have shown that although abnormal collateral lymphatic formation is a pathologic finding in patients with lymphedema, the clinical development of lymphedema may not be due to impaired production of lymphangiogenic cytokines such as VEGF-C (41, 42). In fact, patients with lymphedema have increased serum levels of VEGF-C (43) and transgenic mice that over-express VEGF-C have more severe pathologic changes of lymphedema in a tail model (44). These findings suggest that abnormalities in lymphangiogenesis alone are not enough to cause lymphedema. Rather, lymphatic injury appears to serve as an initiating factor setting into motion other pathologic changes that in some patients results in the development of lymphedema. This hypothesis is supported by the fact that lymph node dissection does not always cause lymphedema; instead only a subset of patients (about 1 in 3) who undergo this treatment go on to develop the disease. Further, the hypothesis that lymphatic injury is simply an initiating event that is necessary but alone insufficient to cause lymphedema is supported by the fact that the development of lymphedema in most cases occurs in a delayed fashion. Typically, patients who undergo lymph node dissection have minor swelling that resolves spontaneously 2–6 weeks after the initial surgery. In some patients however, this swelling recurs permanently 8–24 months later.

Recent studies from our lab and others have shown that lymphatic injury results in chronic inflammatory changes in the skin distal to the zone of injury and that this response, in turn regulates development of lymphedema by causing lymphatic leakiness, decreasing lymphatic pumping, increasing tissue fibrosis, and impairing development of collateral lymphatics. These inflammatory changes illustrate the important coordination of immune responses by the lymphatic system. The changes in inflammatory responses after significant lymphatic injury during surgery enable us to study the effects of more subtle forms of lymphatic injury as may occur in aging, obesity, metabolic dysfunction, or the tumor microenvironment. Thus, studying lymphedema is broadly relevant and may provide important insight into the role of the lymphatic system in regulating immune responses in other physiologic or pathologic events.

#### LYMPHATIC INJURY RESULTS IN UPREGULATION OF ENDOGENOUS DANGER SIGNAL MOLECULES

Danger-associated molecular patterns (DAMPs) are the endogenous cellular products released by stressed, damaged or cells undergoing necrosis that alarm and activate the innate immune system components. By activating innate immune system, DAMPs create a pro-inflammatory state in the damaged tissues with an intention of host defense. However, in excess DAMPs can be harmful due to continuous activation of innate immune reactions (45, 46). Earlier studies by our group revealed the spatial and temporal expression patterns of High mobility group box 1 (HMGB1) and Heat shock protein 70 (HSP-70), two of the well-studied DAMPs (47). Using a mouse tail model of lymphedema tissues and human lymphedema biopsy samples, these studies have shown that DAMP expression occurred along a spatial gradient relative to the site of injury with the highest expression occurring closest to the zone of lymphatic injury and decreasing more distally. DAMP expression was localized to virtually all tissue cells including LECs, blood endothelial cells, adipocytes, and other stromal cells. More importantly, the expression of DAMPs persisted chronically even 6 weeks post-surgery, a time period that is far longer than wound healing related to the initial surgery. Other studies have shown that HMGB1 promotes lymphangiogenesis in vivo and in vitro (48, 49). In support of this, we found that blockade of HMGB1 in the mouse tail lymphedema model inhibited inflammatory lymphangiogenesis.

DAMPs initiate innate immune responses by interacting with pattern recognition receptors (PRRs) such as Toll-like receptors (TLRs). To understand the role of DAMPs interaction with PRRs in lymphedema development, we previously studied lymphedema development in different TLR knockout mice (TLR 2, 4, and 9 KO) using a mouse tail model of the disease (47). Consistent with our findings with HMGB1 blockade resulting in impaired lymphangiogenesis, we found that TLR knockout mice had more severe lymphedema, decreased lymphatic transport, abnormal lymphatic structures, decreased number of lymphatic capillaries, increased collagen deposition and dermal fibrosis, and increased infiltration of T cells as compared with wildtype controls. Taken together, these studies indicate that lymphatic injury chronically activates DAMPs, that eventually activates TLRs. The net result of this DAMP-TLR cascade activation during lymphatic injury is regulation of inflammatory lymphangiogenesis and chronic inflammatory reactions.

#### ROLE OF MACROPHAGES DURING LYMPHATIC INJURY AND LYMPHEDEMA PROGRESSION

In similar lines with several other inflammation pathologies, macrophage recruitment and accumulation is significantly observed during lymphedema both in human biopsy (50, 51) and animal lymphedema models (50, 52). It is reported that macrophage recruitment is significantly high immediately after lymphatic injury compared to later stages of lymphedema (53). Macrophages seem to play multiple roles in lymphedema pathology based on results from several groups. Studies from our group shows that depletion of macrophages promotes impaired lymphatic function, infiltration of CD4<sup>+</sup> cells and aggravates fibrosis (50). In addition we have demonstrated that during lymphatic injury, M2-differentiated macrophages are responsible for initiation of superficial dermal lymphangiogenesis by secreting lymphangiogenic growth factors like VEGF-C (54). Furthermore, it is being reported that, macrophage induced VEGF-C production is positively influenced by prostaglandin E2 (55) or CD4<sup>+</sup> T cells and blocking of COX2, IFN-γ or IL-17 abrogates VEGF-C expression by macrophages (52) indicating interplay between inflammatory mediators and T cells with macrophages during lymphedema. Macrophages play a double-edged sword role in lymphedema pathology, because it plays important role in initial lymphangiogenesis post-lymphatic surgery transiently alleviating fluid accumulation (54). However, macrophages strongly express iNOS and are potential source of nitric oxide (NO) which attenuates collecting lymphatic vessels contraction and pumping significantly decreasing lymphatic function and eventually accumulation of lymph fluid and immunosuppression (9). Macrophages are also a major source of IL-6, a cytokine implicated in mediating chronic inflammation and adipose metabolism in lymphedema and found abundantly in lymphedema tissues (56, 57). Furthermore, macrophages are important source of TGF-β a major anti-lymphangiogenic cytokine that inhibits lymphangiogenesis and is copiously present in lymphedema tissues (58, 59). Taken together, these studies suggest that macrophages play a complex paracrine role in pathology of lymphedema regulating lymphangiogenesis, fibrosis and lymphatic function mostly through varied kinds of growth factor and cytokine secretion that have a dual impact on lymphatic endothelial cells.

### LYMPHATIC INJURY RESULTS IN ACTIVATION OF DENDRITIC CELLS IN THE SKIN AND MIGRATION TO REGIONAL LYMPH NODES

How do chronic inflammatory responses in lymphedema get initiated? We have studied this question using mouse models of lymphedema and adoptive transfer of labeled cells to track the homing, activation, and migration of inflammatory cells (60). Adoptive transfer using intravenous injection is useful tool since this approach can provide insight into the behavior of circulating and skin resident inflammatory cells. To study activation of chronic T cell responses, we injected labeled dendritic cells (DCs) since these leukocytes are powerful antigen presenting cells that regulate adaptive T cell inflammatory responses. Wildtype or CD4 knockout mice (CD4KO) underwent popliteal lymph node dissection (PLND) and allowed to recover. Two weeks following surgery, we adoptively transferred bone marrow DCs using intravenous injection and analyzed DCs in the skin and the inguinal lymph node (the next draining basin) using flow cytometry. Interestingly, we found that adoptively transferred DCs rapidly migrated to the lymphedematous skin (within 6 h of injection) where they expressed activation markers. Over the next 24 h, activated DCs migrated to the inguinal lymph node. Importantly, DC activation or migration was identical in wild-type and CD4KO mice suggesting that DC activation precedes chronic CD4<sup>+</sup> cell inflammatory reactions in lymphedema. These findings are supported by previous studies demonstrating that the lymphatics play a key role in regulating DC migration. Activated DCs upregulate cell surface expression of the chemokine receptor CCR7 (C-C chemokine receptor 7) whose ligands [CCL21 (C-C motif ligand 21) and CCL19] are expressed by LECs. Gradients of CCL21 guide DCs to initial lymphatics (61) and docking to CCL21 (62), and adhesion molecules such as intracellular adhesion molecule 1 (ICAM1), vascular cell adhesion molecule (VCAM1) expressed by LEC is required for entry into the vessel lumen (63). DCs enter the lymphatic vessel through gaps between LECs (64) and are guided to lymph nodes by gradients of CCL21 in lymphatic fluid (65) as well as passive lymphatic fluid flow (**Figure 1**).

### CD4<sup>+</sup> CELL INFLAMMATORY RESPONSES ARE NECESSARY FOR THE DEVELOPMENT OF LYMPHEDEMA

DCs activate naïve T cell in lymph nodes by presenting antigens in the context of co-stimulatory molecules. This process is also important for chronic T cell inflammatory reactions in lymphedema. Histological and flow cytometry analysis of tissue biopsies from patients with lymphedema as well as mouse models of lymphedema demonstrate that the predominant inflammatory cell infiltrate is comprised of CD4<sup>+</sup> cells. In fact, the severity of lymphedema correlates significantly with the degree of CD4<sup>+</sup> cell inflammatory response. T cell inflammatory responses are necessary for the development of lymphedema since nude mice (lack all T cells) or mice lacking CD4<sup>+</sup> cells (CD4KO) do not develop lymphedema after skin and lymphatic excision. Similarly, depletion of T cells using antibodies or topical treatment with tacrolimus, a medication that prevents T cell proliferation/differentiation prevents development of lymphedema in preclinical models and can be used to treat the disease once it has developed. In contrast, depletion of other inflammatory cell types such as cytotoxic T cells, B cells, and macrophages either has no significant effect or worsens the severity of lymphedema (66).

### T CELL ACTIVATION IN LYMPHEDEMA REQUIRES T CELL RECEPTOR ACTIVATION AND CO-STIMULATORY MOLECULE EXPRESSION

T cell activation and differentiation in lymphedema requires T cell receptor activation in the context of co-stimulatory

molecule expression by antigen presenting cells in regional lymph nodes (67). To demonstrate this concept, we used the adoptive transfer experimental approach to study T cell activation in lymphedema. In these experiments, CD4KO mice underwent PLND and 2 weeks later were intravenously injected with naïve CD4<sup>+</sup> cells harvested from wild-type mice or RAG2/OTII mice since these transgenic mice have clonal T cells that only express the ovalbumin T cell receptor (68). This experiment was therefore designed to determine if T cell receptor activation is necessary for skin homing of activated T cells in lymphedema. In other experiments, we tested the hypothesis that T cell receptor activation requires co-stimulatory molecule expression by antigen presenting cells by adoptively transferring naïve wild-type CD4<sup>+</sup> cells into either wild-type or transgenic mice lacking CD28 (a costimulatory molecule necessary for full T cell activation) (69). In these experiments, in contrast to our findings with adoptive transfer of DCs, we found that naïve T cells initially migrated to the ipsilateral inguinal lymph node (i.e., the next draining lymph node basin downstream from where the initial lymph node dissection was performed). Within the lymph node, both T cell receptor activation and costimulatory molecule activation were necessary for activation of CD4<sup>+</sup> cells following lymphatic injury. Thus, adoptively transferred CD4<sup>+</sup> cells harvested from RAG2/OTII mice were not activated in the inguinal lymph node and did not migrate to the lymphedematous skin. Similarly, adoptive transfer of wild-type CD4<sup>+</sup> cells to CD28 knockout mice failed to result

in T cell activation or T cell homing. Taken together, these findings suggest that T cells in lymphedema are activated in regional lymph nodes by antigen presenting cells in response to antigenic stimuli (60). Identification of T cell activating antigens in lymphedema is an active topic of study in our lab (**Figure 1**).

#### LYMPHEDEMA RESULTS IN A MIXED T HELPER CELL DIFFERENTIATION RESPONSE

Lymphatic injury results in a mixed T cell inflammatory reaction consisting of T helper 1 (Th1), T helper 2 (Th2), and Tregs (66). Lymphedematous skin from clinical biopsy specimens and mouse models of lymphedema are infiltrated with large numbers of CD4<sup>+</sup> cells that co-express interferon gamma (IFN-γ; putative Th1 cells) and CD4<sup>+</sup> cells that co-express interleukin 4 (IL4) or IL13 (putative Th2 cells). T cells in lymphedematous tissues tend to cluster around initial lymphatics and lymphatic collectors (60, 70). Using adoptive transfer experiments, we found that naïve CD4<sup>+</sup> cells are activated in regional lymph nodes and characterized by cell surface expression of Th1 (CD45+/CD4+/CCR5+/CXCR3+) and Th2 (CD45+/CD4+/CCR4+/CCR8+) cells. More importantly, we found that release of activated T cells from the lymph node via sphingosine 1 phosphate (S1P) signaling into the systemic circulation is necessary for the development of lymphedema. Treatment with an S1P inhibitor (FTY720) prevented release of activated T cells from the lymph node and prevented development of lymphedema in a mouse tail model of the disease.

Once activated T cells are released from the lymph node and actively home to lymphedematous skin by expressing skin homing cell surface receptors (71). CD4<sup>+</sup> cell migration to the skin in other inflammatory conditions is regulated by cell surface expression of chemokine receptors including cutaneous leukocyte antigen (CLA), cc chemokine receptor 4 (CCR4), CCR8, and CCR10 (72, 73). This fact, together with the finding that T cell inflammatory reactions are important regulators of lymphedema, suggests that the expression of skin homing receptors may also play an important role in the development of lymphedema. This hypothesis is supported by the finding that the expression of CLA ligand E-selectin (74) as well as other leukocyte adhesion molecules (ICAM1, VCAM1) is significantly increased in lymphedematous skin (60). Similarly, we have found that the expression of ligands for CCR4 [chemokine c-c motif ligand 17 (CCL17)] and CCR10 (CCL27) (75) is markedly increased in keratinocytes of lymphedematous skin. Thus, migration of activated T cells to lymphedematous skin is not random but rather a tightly coordinated active process that may enable us to design rational treatment options that may be useful for the treatment of this morbid disease (**Figure 1**).

### Th2 DIFFERENTIATION IS NECESSARY FOR PATHOLOGIC CHANGES IN LYMPHEDEMA

Lymphedema is characterized by fibro-adipose tissue deposition, impaired lymphatic pumping, lymphatic leak, and decreased formation of collateral lymphatics. Previous studies in our lab and others have shown that T cells in general, and Th2 cells in particular play a key role in these pathologic processes (52, 60, 66, 70, 76). In fact, we have hypothesized that lymphedema is simply fibrotic organ failure of the lymphatic system. This hypothesis is supported by the histological characteristics of lymphedema demonstrating progressive collagen deposition and encasement of initial lymphatics by thick collagen bundles (66, 77). In addition, clinical studies have shown that late stage lymphedema results in fibrosis of collecting lymphatics with resultant luminal obliteration and failure of the pump mechanism (78, 79). The fibrotic hypothesis of lymphedema also provides a rationale for the delayed onset of symptoms following surgery since the critical threshold of fibrosis necessary to become symptomatic takes time to occur. In addition, fibrosis is a common cause of organ failure affecting virtually every other organ system in one form or another. Similar to lymphedema, these diseases are progressive and eventually become irreversible with severe end organ injury.

Previous studies have shown that T helper cells play a key role in organ fibrosis in a variety of pathologic conditions including liver fibrosis, pulmonary fibrosis, and scleroderma (80–82). Although the inciting events causing fibrosis in these conditions is highly variable and the parenchyma in these organ systems is distinct, the cellular mechanisms that regulate fibrosis in these conditions appears to be conserved and dependent on chronic Th2 biased immune responses. Ordinarily, Th2 cells play an important role in responses to parasites, however, chronic Th2 biased inflammatory responses promote tissue fibrosis by increasing collagen deposition, decreasing collagen breakdown, and increasing expression of profibrotic growth factors such as IL4, IL13, and TGFb1 (83–85). Because Th1 immune responses often balance and oppose Th2 responses, in general Th1 biased responses are anti-fibrotic.

The regulation of organ fibrosis by chronic mixed Th1/Th2 inflammatory responses is referred to as the Th1/Th2 paradigm (86) and also appears to play a key role in the development of lymphedema. This hypothesis is supported by the fact that inhibition of Th2 differentiation with antibodies that neutralize IL4 or IL13, cytokines necessary for naïve T helper cell differentiation along the Th2 lineage, is highly effective in preventing the development of lymphedema in mouse models (70). Similarly, this treatment strategy is effective in reversing lymphedema once it has become established. Mice with impaired Th2 differentiation capacity do not develop lymphedema following lymphatic injury; in contrast, mice with impaired Th1 differentiation have a phenotype that is indistinguishable from wild-type littermates (71). Inhibition of Th2 differentiation markedly decreases accumulation of inflammatory cells in the skin, decreases collagen deposition and lymphatic fibrosis, reduces lymphatic leakiness, and preserves collecting lymphatic pumping capacity. Inhibition of Th2 responses decreases accumulation of perilymphatic inflammatory cells and markedly decreases expression of induced nitric oxide (iNOS) by perilymphatic inflammatory cells. This is important since increased iNOS expression in inflammatory conditions decreases lymphatic pumping capacity by decreasing gradients of endothelial derived nitric oxide expression by lymphatic cells (9). Finally, we have shown that T cell derived cytokines including IFN-γ, IL4, IL13, and TGF-β1 have potent anti-lymphangiogenic activity and impair LEC proliferation, differentiation, and migration (59, 87–89). Thus, Th2 mediated inflammatory responses impair lymphatic function by multiple mechanisms and play a central role in the pathology of lymphedema. More importantly, we have shown that other causes of lymphatic injury such as high fat diet induced obesity have a similar phenotype including peri-lymphatic accumulation of inflammatory cells, decreased lymphangiogenesis, lymphatic leaking, and impaired lymphatic pumping suggesting that our findings in lymphedema have broader physiologic relevance (90, 91).

Taken together, our findings in lymphedema suggest that lymphatic injury results in a mixed Th1/Th2 immune response secondary to T cell receptor mediated interactions with dendritic cells in regional lymph nodes and that these activated T cells migrate specifically to lymphedematous skin due to expression of cell surface receptors. Within the skin, TABLE 1 | The cellular and molecular factors in lymphedema development.


N.D. stands for not determined.

Th2 cells proliferate and regulate pathologic changes including fibrosis, lymphatic leakiness, impaired pumping, and decreased formation of collateral lymphatics that eventually result in lymphedema.

### HOW DOES LYMPHATIC INJURY REGULATE T REGULATORY CELL PROLIFERATION AND DIFFERENTIATION?

Tregs are immune cells that play a central role in regulating inflammatory responses, autoimmunity, and immune tolerance in a wide variety of physiologic settings. Tregs inhibit immune responses by a myriad of mechanisms including regulation of immune cell proliferation, apoptosis, and activation, production of cytokines, prevention of co-stimulation, and uptake of interleukin 2 (92, 93). These responses provide a homeostatic mechanism that prevents excessive inflammatory reactions. Tregs can be broadly divided into induced Tregs or natural Tregs; natural Tregs develop in the thymus from bone marrow derived T cell precursors (94). Induced Tregs, in contrast, develop from mature conventions T helper cells outside of the thymus, play an important role in the regulation of autoimmunity.

In addition to a mixed Th1/Th2 immune response, our lab and others have shown that lymphedema results in the accumulation of Tregs in lymphedematous tissues (66, 95, 96). Biopsy specimens of patients with unilateral upper extremity lymphedema demonstrated a nearly 6-fold increase in the number of Tregs in the lymphedematous skin (95). Using a mouse model of axillary lymph node dissection (ALND), we showed that the majority of Tregs present in the forelimb skin distal to the zone of lymphatic injury are proliferating, natural Tregs (CD4+/FoxP3+/Nrp-1+). In contrast, we found no changes in the number of induced Tregs in the skin and no changes in any Treg population in the blood or the spleen suggesting that Treg activation and proliferation was localized to the forelimb skin rather than systemic changes. Depletion of Tregs using diphtheria toxin treatment in Fox-P3-diphtheria toxin receptor (FoxP3-DTR) transgenic mice significantly increased the number of Th1 and Th2 cells in the forelimb tissues. In addition, this treatment increased the number of infiltrating macrophages (CD11b+F4/80+), neutrophils (Ly-6G+), and activated DCs (CD11c+MHCII+CD86+). Consistent with the immunosuppressive effects of Tregs in general, we found that Treg depletion improved T cell and B cell mediated immune responses after sensitization of the forelimb skin distal to the zone of lymphatic injury. Moreover, we found that Treg depletion increased bacterial phagocytosis and removal as compared with control mice after injection of heat inactivated bacterial particles in the forelimb skin. Other recent studies have shown that depletion of Tregs in mouse models of lymphedema results in increased severity of lymphedema while adoptive transfer of Tregs ameliorates the phenotype (96). Taken together, these findings suggest that Treg infiltration following lymphatic injury acts to suppress chronic inflammatory responses and may be homeostatic in nature. In addition, chronic infiltration of Tregs in lymphedematous tissues and subsequent suppression of immune responses may provide a rationale for the increased risk of infections and developing secondary malignancies in patients with lymphedema. These two papers in combination reveal the duality in Tregs function in lymphedema pathology by modulating one common factor namely inflammation. More importantly, these findings clearly show that lymphatic function can regulate Treg migration, proliferation, and differentiation.

#### REFERENCES


#### CONCLUSIONS

The lymphatic system, acting via direct and indirect mechanisms, is an important regulator of immune responses (**Table 1**). Lymphatic injury occurring either as a result of iatrogenic causes or secondary to physiologic changes such as obesity, tumor formation, metabolic syndrome, or infection can modulate immune response by regulating trafficking of antigen presenting cells, decreasing transport of particulate matter or antigens, regulating T cell differentiation, and modulating immunosuppressive immune responses. These changes may modulate the severity of the underlying condition and, in some cases, may promote the development of a vicious cycle of events. Thus, understanding the mechanisms regulating immune modulation by the lymphatic system is an important goal and has broad biologic relevance.

#### AUTHOR CONTRIBUTIONS

BM, RK, JB, HP, and IW did the literature search, complied, and wrote the manuscript. SR and JS proof read the manuscript. RK and JB prepared the figures and tables.

#### FUNDING

The review was supported by NIH R01 HL111130-01 grant and P30 CA008748 institutional grant.


lymphedema. Am J Physiol Heart Circ Physiol. (2014) 306:H1426–34. doi: 10.1152/ajpheart.01019.2013


study in a human cadaver. Plast Reconstr Surg. (2007) 120:982–91. doi: 10.1097/01.prs.0000277995.25009.3e


**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 © 2019 Kataru, Baik, Park, Wiser, Rehal, Shin and Mehrara. 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(s) 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.

# Lymphatic Function in Autoimmune Diseases

Noa Schwartz 1,2, Madhavi Latha S. Chalasani <sup>1</sup> , Thomas M. Li <sup>1</sup> , Zhonghui Feng<sup>1</sup> , William D. Shipman1,3 and Theresa T. Lu1,2,4,5 \*

*<sup>1</sup> HSS Research Institute, Hospital for Special Surgery, New York, NY, United States, <sup>2</sup> Division of Rheumatology, Department of Medicine, Hospital for Special Surgery, New York, NY, United States, <sup>3</sup> Weill Cornell Tri-Institutional MD-PhD Program, New York, NY, United States, <sup>4</sup> Division of Pediatric Rheumatology, Department of Medicine, Hospital for Special Surgery, New York, NY, United States, <sup>5</sup> Department of Microbiology and Immunology, Weill Cornell Medicine, New York, NY, United States*

Lymphatic vessels are critical for clearing fluid and inflammatory cells from inflamed tissues and also have roles in immune tolerance. Given the functional association of the lymphatics with the immune system, lymphatic dysfunction may contribute to the pathophysiology of rheumatic autoimmune diseases. Here we review the current understanding of the role of lymphatics in the autoimmune diseases rheumatoid arthritis, scleroderma, lupus, and dermatomyositis and consider the possibility that manual therapies such as massage and acupuncture may be useful in improving lymphatic function in autoimmune diseases.

Keywords: lymphatics, autoimmune disease, rheumatoid arthritis, scleroderma, systemic lupus erythematosus, lymphatic massage

#### Edited by:

*Timothy Padera, Massachusetts General Hospital and Harvard Medical School, United States*

#### Reviewed by:

*Edward Schwarz, University of Rochester, United States Shan Liao, University of Calgary, Canada*

> \*Correspondence: *Theresa T. Lu LuT@HSS.edu*

#### Specialty section:

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

Received: *15 December 2018* Accepted: *26 February 2019* Published: *20 March 2019*

#### Citation:

*Schwartz N, Chalasani MLS, Li TM, Feng Z, Shipman WD and Lu TT (2019) Lymphatic Function in Autoimmune Diseases. Front. Immunol. 10:519. doi: 10.3389/fimmu.2019.00519*

## INTRODUCTION

As early as the Fourth century, Aristotle described lymphatic vessels as fibers positioned between blood vessels and nerves, containing colorless liquid (1). While our understanding of the lymphatic system has advanced since Aristotle's time, the functional significance of the lymphatic network to health and disease is still being unraveled. The lymphatic system is a network of vessels that drains protein-rich lymph from the extracellular fluid, transports it through a series of lymph nodes (LNs), and finally returns it to the bloodstream. Beyond their role in maintaining tissue fluid homeostasis, lymphatic vessels are an important part of the immune system: they allow transport of antigens from the periphery to LNs, where immune cells are primed, expanded, and eventually transported to the site of inflammation (2, 3). In addition to ferrying lymph and immune cells, the lymphatic system itself is directly involved in immune modulation and induction of tolerance to self-antigens (4, 5). Given the function of the lymphatic system in immunity, lymphatic dysfunction may also contribute to the pathophysiology of autoimmune diseases. Here, we review the current understanding of the lymphatic function within autoimmune disease. We begin with a brief overview of the lymphatic system, discuss what is known about lymphatic function in a number of rheumatic diseases, starting from the best studied to the least studied in terms of lymphatic function, and conclude with a consideration of manual therapies as potential approaches to improve lymphatic function in disease. Our goal is to bring more attention to this under-explored, yet promising, area of study.

### OVERVIEW OF THE LYMPHATIC SYSTEM

Lymphatic vessels form an extensive network throughout the body with the exception of only a few tissues including bone, heart myocardium and skeletal muscles, as well as the parenchyma of kidney, liver, and adrenal and thyroid glands. These exceptions either have little interstitial fluid or have an alternative drainage system, such as fenestrated blood vessels (6).

**71**

The lymphatic system is composed of initial lymphatic capillaries that merge to form collecting lymphatic vessels. The collecting vessels transport the lymph to and from a series of LNs and eventually drain into the thoracic duct that connects to the blood circulation by draining into the subclavian veins (6). Lymphatic capillaries are blind-ended and are composed of a thin layer of lymphatic endothelial cells (LECs) with discontinuous basement membrane and "button-like" cell junctions, allowing unidirectional flow of cells and fluid into the vessel (7). In contrast to the capillaries, collecting lymphatic vessels (LVs) have intraluminal valves that prevent backflow of lymphatic fluid (8) and several perivascular layers of lymphatic muscle cells, with characteristics of both smooth muscle cells and cardiac striated muscle cells, that provide vascular tone and rhythmic contractions of the vessels, enabling anti-gravity, active fluid transportation (9, 10). Previously, distinguishing blood endothelial cells from LECs was difficult. But recent research identified several LEC-specific markers, including Lyve-1, vascular endothelial growth factor receptor-3 (VEGFR-3), podoplanin (PDPN), and Prox-1, among others, that enabled tremendous advancements in the study of lymphatic vessel development and function (11, 12).

Previous studies have shown that the systemic vasculature can reabsorb up to 90% of extravasated water, while the remaining 10% is absorbed by the lymphatic vessels (13). However, congenital or acquired dysfunction of the lymphatic system result not only in lymphedema and its sequalae such as skin thickening, fibrosis, and adipose degeneration, but also in poor immune function, susceptibility to infections, and impaired wound healing, among other deleterious health effects (8). These observations point at the greater role lymphatic vessels have than simple fluid transportation. Literature has shown that LECs directly affect immune cell activity in many different ways, including secretion of transforming growth factor-β (TGFβ) leading to suppression of dendritic cell (DC) maturation (14); production of IL-7 to increase IL-2 sensitivity in regulatory T cells to increase their immune-regulatory function (15) and to sustain inflammation-induced lymphoid follicles in disease (16); secretion of colony stimulating factor-1 (CSF-1) promoting differentiation, proliferation and survival of macrophages that contribute to tumor growth (17). LECs also present peripheral tissue antigens together with programmed death-ligand-1 (PD-L1) leading to CD8+ T-cell response inhibition (18) and modulate CD4+ T-cell response via low level antigen presentation with MHC-II during inflammation (19). That lymphatic dysfunction exacerbates autoimmune disease is supported by the development of autoantibodies in mice lacking dermal lymphatics (5). Understanding the lymphatic system in the context of autoimmune diseases has the potential to provide insight into disease mechanisms and new approaches to treatment.

#### LYMPHATICS AND RHEUMATOID ARTHRITIS

Rheumatoid arthritis (RA) is one of the most studied autoimmune conditions, with regards to the role of lymphatics in the context of disease. RA is an autoimmune systemic disease, affecting 0.5–1% of the population, with its hallmark being symmetric polyarthritis, usually with small-joint distribution (20). Local lymph node enlargement was first described in RA in 1896 (21), but it wasn't until more specific markers of the lymphatic system were discovered that its role could be specifically investigated.

It is thought that together with the joint inflammation occurring in RA, the local lymphatics undergo two stages of alterations. As a response to the initial, pre-arthritic, synovial inflammation, the lymphatics undergo an "expansion phase," whereby they increase their capacity to remove excess cellular debris and inflammatory cells from the site of inflammation; whether by lymphangiogenesis (22), or by increased lymphatic vessel contraction frequency (23). This process is important to allow for the resolution of the inflammatory process; if the expansion process is stunted, by inhibition of lymphangiogenesis for example, the joint inflammation becomes more severe and clinical synovitis develops (24). Beyond the lymphatic vessel changes, the draining lymph nodes themselves increase in size during the expansion phase (23, 25), likely due to increased volume and pressure of fluid within the afferent vessels (25), intra-nodal lymphangiogenesis (23), and infiltration of a unique subtype of IgM+CD23+CD21hiCD1dhi B cells found in inflamed lymph nodes, known as Bin cells (26, 27). Nevertheless, while the removal of the excess debris is important to allow for inflammation resolution in the acute setting, the inflammatory cells, and catabolic factors that are being removed have been shown to directly damage the LECs and lymphatic muscle cells, both in the afferent lymphatic vessels and the draining lymph nodes (28). As a result of this ongoing stress on the lymphatic system, the lymphatics progress to the "collapsed phase," in which the local lymphatic conduit system breaks down, and the lymph node is no longer able to efficiently drain the fluid from the inflamed synovium (24, 25). The lymphatic vessels are damaged, with increased leakiness and reduced contractions, leading to poor lymphatic clearance, and stasis of the inflammatory fluid within the joint and the afferent lymphatic vessels (23, 28– 31). The process is thought to be mediated by several factors, including inflammatory cytokines in the vessels triggering LEC expression of inducible nitric oxide synthase (iNOS), as well as iNOS-producing activated myeloid cells, now static within the lymphatic vessels. The increased local NO production abrogates the constitutive endothelial NOS (eNOS) activity that is an important mediator of lymphatic vessel contraction (32, 33). Reduced vessel contraction is likely also due to increased fluid flow and pressure inside the vessels, beyond the vessels' ability to compensate (34, 35). At the same time, Bin cells in the draining lymph node migrate from the lymph node follicles to the sinuses, as extensively reviewed by Bouta et al. (31), leading to clogging of lymph node sinuses and blocking passive lymphatic drainage. The resulting impairment in lymphatic drainage contributes to increased joint inflammation and synovial hyperplasia, eventually leading to joint destruction (36). Importantly, known and effective RA treatments, such as tumor necrosis factor (TNF) inhibition and anti-CD20 therapy, have both shown to also have a beneficial effect on lymphatic flow. Inhibition of TNF has been shown to restore lymphatic vessel contractility (28), and anti-CD20 therapy, i.e., Rituximab, depletes Bin cells from the lymphatic sinuses, thereby promoting the restoration of lymphatic flow (37).

Preliminary work with indocyanine green near-infrared (ICG-NIR) fluorescence imaging has been promising in providing sensitive, real-time non-invasive means to evaluate the layout and function of the lymphatic vasculature (31, 38, 39), and the first study of lymphatic flow in RA patients with this modality is currently being performed (ClinicalTrials.gov NCT02680067). Meanwhile, the change in size of the local draining lymph nodes has been shown to reflect joint inflammatory activity, as well as response to therapy (40). Even prior to clinical lymphadenopathy, evaluation of draining lymph nodes of inflamed joints by power Doppler ultrasound (PDUS), demonstrated hypertrophy of the lymph node cortex, in addition to power Doppler signal amplification in cortical and hilar regions likely indicating increased flow. Importantly, these findings reversed with treatment (41). At the same time, low PDUS signal at baseline despite active arthritis, likely representing a collapsed lymph node, predicts poor clinical response to therapy (40). Similarly, in a pilot study using contrast enhanced MRI (CE-MRI) to monitor LN size before and after treatment with Certolizumab in RA patients, there was an inverse correlation between the extent of treatment-related pain relief and decrease in LN size. The LNs with the more notable reduction in size are the ones that are more likely to have undergone collapse, leading to inadequate inflammation resolution and reduced pain relief (42). Thus, in addition to providing an important, non-invasive means by which to monitor disease activity, response to therapy, and even predict prognosis; these findings also support the bi-phasic lymphatic response model, namely the "expansion phase" and "collapse phase," seen in murine models of inflammatory arthritis.

#### LYMPHATICS AND SYSTEMIC SCLEROSIS

Scleroderma is an autoimmune connective tissue disease characterized by abnormalities in vasculature, immune function, and extracellular matrix that ultimately manifest as fibrosis of the cutaneous, vascular, musculoskeletal, gastrointestinal, pulmonary, cardiac, and renal systems. Although the etiology of the disease is poorly understood, vascular injury and abnormal endothelial cell function are hypothesized to be among the primary defects responsible for disease pathogenesis (43, 44).

Studies into vascular abnormalities leading to fibrosis have generally been more focused on blood endothelial cells (BECs), while the roles of LECs and lymphatic dysfunction have been less well-studied. In 1999, Leu et al. first demonstrated using lymphangiography that scleroderma lesional skin had signs of lymphatic microangiopathy, with absence or fragmentation of visualizable lymphatic networks and evidence of vessel leakiness and backflow (45). Similar findings using immunohistochemical staining of scleroderma skin recently showed a decrease in lymphatic vessels and increased cross-sectional area of the remaining vessels, suggesting vessel dilatation and a block in lymphatic flow downstream (44). The changes were most significant in the reticular dermis, with a similar trend found in the papillary dermis. Lymphatic changes have been found in other fibrosing conditions as well (**Table 1**), further supporting the idea that lymphatic dysfunction may be a therapeutic target in scleroderma.

### LYMPHATICS AND SYSTEMIC LUPUS ERYTHEMATOSUS AND DERMATOMYOSITIS

Systemic lupus erythematosus (SLE) is the prototypical systemic autoimmune disease, affecting between in 6.5 and 187 per 100,000 people worldwide, with a 9:1 female predominance and mortality rate that is three times that of the general population (57, 58). There have been no systematic studies of lymphatic function in SLE to date. There are, however, hints that there might be lymphatic dysfunction in SLE, as there are case reports of chylous ascites or pleural effusions, lymph fluid found in the abdomen or thoracic cavity, respectively, that can result from lymphatic obstruction in the mesentery (59–62). Lymphedema from peripheral lymphatic obstruction has also been described (63). These occurrences are rare, however, and whether more subtle problems with lymphatic flow that could perhaps result from the lymphadenopathy commonly found in SLE (64, 65) are not known.

Similarly, lymphatic function in dermatomyositis, a group of autoimmune diseases primarily directed against the muscle and skin (66), has not been systematically studied. Dermatomyositis patients can rarely present with generalized edema, which may reflect poor lymphatic function and lymphedema (67). Gottron's papules are characteristic red, raised lesions on the knuckles of dermatomyositis patients, and one study examining the histopathology of these papules noted dilated PDPN<sup>+</sup> lymphatic vessels (68). Interestingly, benign lymphadenopathy is less frequent in dermatomyositis than in SLE or RA, and, because of the association of cancer with dermatomyositis in adults, lymphadenopathy in dermatomyositis has to be evaluated carefully for metastatic cancer or lymphomas (69).

#### POTENTIAL APPROACHES TO IMPROVING LYMPHATIC FUNCTION IN RHEUMATIC DISEASES: NEW AND ANCIENT

As we begin to understand the role of lymphatic dysfunction in autoimmune diseases, it is also time to consider how we might improve lymphatic function as part of disease treatment. Schwarz and colleagues have recently outlined molecularly targeted therapies that are currently being investigated (31). In contrast to pharmacologic approaches, manual therapies have been used since ancient times and are currently the mainstay in improving lymphatic flow in diseases. Below, we briefly discuss some of these approaches to consider their potential utility in improving lymphatic function in autoimmune diseases.

Lymphatic-directed massage techniques are used in the treatment of primary and secondary forms of lymphedema, such as that which occurs in the arms of 20% of breast cancer surgery patients when axillary lymph nodes have been removed (70).

#### TABLE 1 | Summary of lymphatic dysfunction in fibrosis of different organs.


*SSc, systemic sclerosis; LEC, lymphatic endothelial cells; IPF, idiopathic pulmonary fibrosis; VEGF, vascular endothelial growth factor; TGF-*β*, transforming growth factor-beta.*

Manual lymphatic drainage (MLD) is a specific light pressure massage technique that moves from the trunk to the distal portion of the affected limb to stimulate lymph flow away from the peripheral tissue (71). Indeed, hand edema is observed in systemic sclerosis patients in the early edematous phase, and MLD has been shown to significantly reduce the swelling and improve hand function in these patients (72). Similarly, a dry brushing massage technique used in Ayurvedic medicine that originated in India 5,000 years ago is meant to relieve lymphatic congestion that is thought to contribute to stress and disease. It has also been used to reduce lymphedema and inflammation from lymphatic filariasis (73). Interestingly, the sports industry, which has been interested in promoting post-training recovery and reducing edema, is investigating the utility of peristaltic pulse dynamic compression (PPDC) devices that simulate manual lymphatic therapies. Recently, PPDC was shown to increase the pressure-to-pain threshold in elite athletes (74) and also to induce expression of anti-inflammatory genes (75), although it is yet unclear whether these effects are attributable to improving lymphatic flow. Potentially, then, lymphatic massage techniques could be used to improve lymphatic function to help reduce tissue inflammation in autoimmune diseases.

Interestingly, acupuncture, a component of traditional Chinese medicine, may potentially have a lymphatic basis. Acupoints are specific points on the body that practitioners target in an attempt to mobilize stagnant qi, thought to be a form of energy that flows through the meridian system and enhance well-being (76). Recent analysis of acupoints demonstrates that they co-localize with tissue planes rich in nerves, blood and lymphatic vessels, and mast cells. Acupuncture involves insertion of metal needles into the skin and spinning the needle between the acupuncturist's fingers. It has been proposed that this process disturbs local tissues and transmits a biomechanical signal to surrounding cells and structures (77) that can stimulate lymphatic vessels. Additionally, activation of the nerves or mast cells in the area could result in release of vasoactive cytokines that can then stimulate lymphatic vessels to better mobilize fluid and inflammatory cells from the area (77–80). There are observational trials showing efficacy of acupuncture on breast cancer-associated lymphedema, supporting the idea that acupuncture can modulate lymphatic function (81–83). A recent randomized controlled trial examining the ability of acupuncture to further reduce lymphedema on top of current standard therapies such as lymphatic massage drainage and compression sleeves did not show additional benefits (84). However, whether acupuncture alone is at least as good as current standard therapies is not yet known. It should be of interest to better study whether acupuncture could modulate lymphatic function to aid in the treatment of autoimmune diseases.

While the objective of these manual therapies in lymphedema is to improve lymphatic flow and reduce inflammation and swelling in the affected tissues, improving lymphatic flow has the potential to also modulate immune cell activity in a number of ways. First, as mentioned in section Overview of the Lymphatic System, LECs can directly regulate immune cell function, and stimulation of lymphatic flow can modulate the ability of LECs to regulate immune cells (85). Second, lymphatic flow, by means of transporting antigen from the periphery, can impact the tolerance and activation of lymph node lymphocytes (5). Third, cytokines expressed in peripheral tissues can impact immune function in lymph nodes (86), potentially in part by lymphatic transport to the lymph nodes (87). Here, it is possible that cytokines transported to the draining nodes can both activate and regulate lymph node responses, suggesting that improving lymphatic flow can help reduce the duration and/or magnitude of ongoing autoimmune responses. Thus, in the study mentioned above examining the effects of MLD on hand edema in scleroderma patients (72), it would be interesting to understand whether MLD reduced autoantibody levels when edema was reduced. Manual therapies to improve lymphatic flow, then,

#### REFERENCES


may be a well-tolerated, relatively low-cost method to improve many facets of lymphatic function to reduce inflammation and autoimmunity in rheumatic diseases.

#### CONCLUSIONS AND FUTURE DIRECTIONS

The lymphatic system has not been well-studied in autoimmune diseases generally, but the existing evidence, especially in RA and, to a more limited extent, in systemic sclerosis suggests that there is at least dysfunction of lymphatic flow. Further studies focused on the consequences of dysfunctional flow as well as alterations in the direct effects of lymphatic vessels and LECs on innate and adaptive immune cells should provide insights into how best to target the lymphatics in autoimmune rheumatic diseases. Additionally, understanding the causes of lymphatic dysfunction in these diseases may help us better target upstream mediators and perhaps reveal that lymphatic targeting is a mechanism of action of some medications. Finally, as we consider new approaches to targeting lymphatics in autoimmune diseases, there may be value in better understanding older approaches in the context of Twenty-First century biomedical understanding of lymphatic and immune function to expand our therapeutic armamentarium for autoimmune diseases.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work was supported by NIH T32AR071302-01 to the Hospital for Special Surgery Research Institute Rheumatology Training Program (to NS and WS), NIH MSTP T32GM007739 to the Weill Cornell/Rockefeller/Sloan-KetteringTri-Institutional MD-PhD Program (to WS), NIH R01AI079178, Scleroderma Foundation, St. Giles Foundation, and the Lupus Research Alliance (all TL).


arthritis by accelerating local lymphatic drainage in mice. Arthritis Rheum. (2011) 63:2318–28. doi: 10.1002/art.30421


association with fingertip ulcers. Arthritis Rheum. (2010) 62:1513–22. doi: 10.1002/art.27406


cancer treatment. Cochrane Database Syst Rev. (2015) 2015:Cd003475. doi: 10.1002/14651858.CD003475.pub2


**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 © 2019 Schwartz, Chalasani, Li, Feng, Shipman and Lu. This is an openaccess 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(s) 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.

# Dendritic Cells and T Cells Interact Within Murine Afferent Lymphatic Capillaries

Morgan Campbell Hunter <sup>1</sup> , Alvaro Teijeira<sup>1</sup> , Riccardo Montecchi <sup>1</sup> , Erica Russo<sup>1</sup> , Peter Runge<sup>1</sup> , Friedemann Kiefer 2,3 and Cornelia Halin<sup>1</sup> \*

1 Institute of Pharmaceutical Sciences, ETH Zürich, Zurich, Switzerland, <sup>2</sup> Max Planck Institute for Molecular Biomedicine, Münster, Germany, <sup>3</sup> European Institute for Molecular Imaging – EIMI, Westfälische Wilhelms-Universität Münster, Münster, Germany

Afferent lymphatic vessels contribute to immunity by transporting antigen and leukocytes to draining lymph nodes (LNs) and are emerging as new players in the regulation of peripheral tolerance. Performing intravital microscopy in inflamed murine ear skin we found that migrating dendritic cells (DCs) and antigen-experienced effector T cells spend considerable time arresting or clustering within afferent lymphatic capillaries. We also observed that intralymphatic T cells frequently interacted with DCs. When imaging polyclonal T cells during an ongoing contact-hypersensitivity response, most intralymphatic DC-T cell interactions were short-lived. Conversely, during a delayed-type-hypersensitivity response, cognate antigen-bearing DCs engaged in long-lived MHCII-(I-A/I-E)-dependent interactions with antigen-specific T cells. Long-lived intralymphatic DC-T cell interactions reduced the speed of DC crawling but did not delay overall DC migration to draining LNs. While further consequences of these intralymphatic interactions still need to be explored, our findings suggest that lymphatic capillaries represent a unique compartment in which adaptive immune interaction and modulation occur.

#### Edited by:

Beth Ann Tamburini, University of Colorado Denver, United States

#### Reviewed by:

Pierre-Yves Von Der Weid, University of Calgary, Canada Veronique Angeli, National University of Singapore, Singapore

\*Correspondence:

Cornelia Halin cornelia.halin@pharma.ethz.ch

#### Specialty section:

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

Received: 13 December 2018 Accepted: 26 February 2019 Published: 22 March 2019

#### Citation:

Hunter MC, Teijeira A, Montecchi R, Russo E, Runge P, Kiefer F and Halin C (2019) Dendritic Cells and T Cells Interact Within Murine Afferent Lymphatic Capillaries. Front. Immunol. 10:520. doi: 10.3389/fimmu.2019.00520 Keywords: dendritic cells, T cells, immune interactions, lymphatic vessels, adaptive immunity, migration

### INTRODUCTION

Afferent lymphatic vessels are present within most vascularized tissues and functionally convey lymph toward and into a draining LN. By transporting soluble inflammatory mediators, antigens and leukocytes, afferent lymphatic vessels establish an immunological connection between peripheral tissues and LNs. In addition to these traditional transport functions, several emerging studies highlight the role of lymphatic endothelium itself as a key modulator of peripheral immune responses (1–3).

The main cell types migrating via afferent lymphatic vessels are antigen-experienced CD4<sup>+</sup> T cells and antigen-presenting dendritic cells (DCs) (3, 4). While T cell recirculation through afferent lymphatic vessels is thought to contribute to immunesurveillance, DC migration is important for maintenance of tolerance and for induction of protective immunity in draining LNs (3). In this regard, DCs within the tissue take up antigen and migrate via afferent lymphatic vessels to a draining LN. Within the LN, naïve T cells survey arriving DCs for presentation of antigen. In the case that a naïve T cell encounters a cognate antigen, the T cell undergoes clonal expansion and differentiation into effector and memory T cells. Intravital microscopy has revealed that

**78**

such adaptive DC-T cell interactions progress through distinct phases of contact that depend on factors such as antigen recognition, timing of activation, signal strength, and the inflammatory environment (5, 6). At the end of the proliferation and differentiation phase, antigen-experienced effector, and memory T cells exit the LN via efferent lymphatic vessels and migrate to inflamed peripheral tissues. There, effector/memory T cells may be re-stimulated by antigen to perform local effector functions, or exit the tissue via afferent lymphatic vessels. Performing intravital microscopy in the murine ear skin we and others have recently described that both DCs and CD4<sup>+</sup> effector/memory T cells spend several hours actively patrolling within initial capillaries and are only passively transported to the draining LN once they reach the larger downstream collecting vessels (7–10). Considering the long time spent in lymphatic capillaries and the emerging knowledge of the immune-modulatory functions of lymphatic vessels (1, 3), we here set out to further characterize the intralymphatic migratory behavior of DCs and T cells and to specifically investigate whether these cells might interact inside lymphatic capillaries.

### MATERIALS AND METHODS

### Mouse Strains

Wilde-type (WT) C57BL/6 mice, VE-cadherin-Cre×RFP (7), hCD2-DsRed×Prox1-GFP (10), Prox1-Orange×CD11c-YFP (11, 12), and hCD2-DsRedxOTII (13, 14) mice were crossed and/or bred in specific-pathogen-free (SPF) facilities in-house. I-A/I-E−/<sup>−</sup> mice (15) were acquired from an SPF facility at University of Zurich Laboratory Animal Services Center (LASC), Schlieren. All experiments were approved by the Cantonal Veterinary Office Zurich.

#### Bone Marrow Chimeras

Bone marrow chimeras were generated as described in (7).

### Generation of BM-DCs

WT or I-A/I-E−/<sup>−</sup> BM-DCs were generated as described in **Supplemental Experimental Procedures**.

#### Intravital Microscopy Specifications

Intravital microscopy of mouse ear pinna was performed as previously described (10). Exact imaging conditions and cell motility specifications are listed in **Supplemental Experimental Procedures**. For adoptive transfer experiments: five to six hours prior to imaging, mice were anesthetized using isoflurane (2–5%) and 500'000 to 750'000 WT or I-A/I-E−/<sup>−</sup> labeled BM-DCs, or CD11c-YFP BM-DCs, were adoptively transferred into the ear skin in 2-3 injections of up to 5 µl each.

### Analysis of DC and T Cell Contacts

For DC - T cell contact analysis, individual DCs were followed frame-by-frame and contact with T cells manually annotated. Direct contact for longer than 2 min was considered an active interaction. Contacts shorter than 2 min were excluded from the analysis. A gap size of 2 min between contacts with the same T cell was considered a single continuing contact. Consequently, gaps in contact for more than 2 min were considered independent contacts. Using these criteria, a contact plot for each DC was generated. The length of each contact, DC occupancy and number of T cell contacts per DC were analyzed. DC occupancy index = a measure of the percentage time that a DC is contacted by a T cell/s during an imaging period.

#### Flow Cytometry

Flow cytometry was performed on ear skin, LNs or BM-DCs as described in **Supplemental Experimental Procedures**.

### CHS-Induced or DTH-Induced Inflammation

See **Supplemental Experimental Procedures**.

### Statistical Analysis

All cell tracking data are presented as medians and all other results presented as mean plus standard deviation. Data sets were analyzed using Prism 7 (GraphPad). Kruskal-Wallis test followed by post hoc analysis was used for multiple comparisons and Mann-Whitney U-test for simple comparisons.

## RESULTS

### DCs Patrol and Arrest Within Lymphatic Capillaries

We previously reported that DCs actively entered lymphatic capillaries and migrated in a semi-directed manner toward the collecting vessels (7, 9). Intriguingly, in these studies we also frequently observed DCs that remained arrested for long time periods. To assess the relative proportion of time that DCs spend arrested or patrolling, we classified DCs into four different groups, depending on whether they had spent the entire imaging period of 60 min (a) actively patrolling, (b) mainly patrolling, (c) mainly arrested, or (d) completely arrested. Imaging was performed in the ear skin of VE-cadherin-Cre × RFP mice, which feature red-fluorescent blood and lymphatic vessels (7), either upon adoptive transfer of LPS-matured YFP<sup>+</sup> bone marrow-derived DCs (BM-DCs; **Figure 1A**), or upon reconstitution of these mice with bone marrow from CD11c-YFP mice (BM chimeras, **Figure 1B**). In chimeric animals, endogenous YFP<sup>+</sup> DCs were induced to migrate into lymphatic vessels by intradermal LPS injection and topical imiquimod treatment. In both imaging setups only a small proportion of DCs (≈ 15%) continuously migrated or "patrolled" intravascularly, whereas a large proportion of DCs (≈ 75%) exhibited an intermittent arresting and patrolling behavior, and ∼10% remained completely arrested throughout the imaging period (**Figure 1C** and **Movie S1**). Most notably, in both setups, around 30% of DCs were arrested for more than half of the imaging period (**Figure 1C**, orange and red sectors combined).

tracked cell. Median is shown as a red bar. Pooled data from 3 mice are shown.

### T Cells Patrol, Cluster, and Frequently Arrest Within Lymphatic Capillaries

Performing intravital microscopy in contact hypersensitivity (CHS)-inflamed ear skin of hCD2-DsRed×Prox1-GFP mice, which feature red T cells and green lymphatic vessels (**Figure 1D**), we recently reported that, similarly to DCs, CD4<sup>+</sup> effector/memory T cells entered into and actively patrolled within lymphatic capillaries in mouse ear skin (10). After Hunter et al. DC-T Cell Interactions in Lymphatic Capillaries

further imaging and closer inspection of our videos, we also observed that several T cells remained arrested or clustered inside lymphatic capillaries (**Figure 1E**, **Movie S2**). Singlecell tracking analysis revealed that T cells in clusters were notably less motile than patrolling T cells (**Figure 1F**). Since clustering and swarming of T cells are hallmark phases of T cell activation in the draining LN (5, 6), we speculated that clustering or arresting T cells might be interacting with as yet "invisible" arrested DCs. In support of this "interaction hypothesis", we also occasionally observed T cells interacting with motile GFP<sup>+</sup> cells inside lymphatic capillaries of hCD2- DsRed×Prox1-GFP mice (**Figures S1A,B**, **Movie S3**). In flow cytometry a small fraction of GFP<sup>+</sup> cells were found to be CD45<sup>+</sup> cells and to express the DC markers CD11c and MHCII (I-A/I-E), indicating that the interacting cells might be DCs that had phagocytosed dying Prox1-GFP<sup>+</sup> lymphatic endothelial cells (**Figures S1C,D**).

### Adoptively Transferred DCs Interact With T Cells Inside Lymphatic Capillaries During a CHS Response

To more definitively show that DCs interact with T cells inside lymphatic capillaries, we adoptively transferred YFP<sup>+</sup> LPS-matured BM-DCs (**Figure S2A**) into the ear skin of hCD2-DsRed×Prox1-GFP mice, which had been inflamed by induction of a CHS response to oxazolone. As previously reported (7, 9), YFP<sup>+</sup> DCs avidly entered and actively migrated within lymphatic capillaries. Most notably, YFP<sup>+</sup> DCs frequently interacted with T cells inside lymphatic capillaries (**Movie S4**).

#### The Majority of Intralymphatic Interactions Between T Cells and Transferred DCs Are Short-Lived and I-A/I-E−/−-independent During a CHS Response

In LNs, migratory DCs are known to present processed antigen on MHCII (I-A/I-E) to circulating naïve T cells. To further characterize DC-T cell interactions inside dermal lymphatic capillaries, and to investigate their requirement for I-A/I-E, we adoptively transferred DeepRed-labeled wild-type (WT) or I-A/I-E−/<sup>−</sup> BM-DCs into the CHS-inflamed ear skin of hCD2-DsRed×Prox1-GFP mice (**Figures 2A,B**). Activated WT and I-A/I-E−/<sup>−</sup> DCs expressed similar levels of co-stimulatory molecules CD80 and CD86 (**Figures S2B,C**) and crawled with equal migratory speeds on lymphatic endothelial cell monolayers in vitro (**Figure S2D**) and in lymphatic capillaries of CHSinflamed skin in vivo (**Figure 2C**). Both WT and I-A/I-E−/<sup>−</sup> DCs interacted with T cells inside lymphatic capillaries, and in most cases intralymphatic DC-T cell interactions were dynamic in nature: DCs interacted with several T cells during the imaging period and frequently interacted with more than one T cell simultaneously (**Figure 2B**, **Movie S5**). To quantify intralymphatic DC-T cell interactions, we generated contact plots whereby interacting DCs were analyzed frame by frame for contact with T cells (**Figure 2D**). This assessment revealed that the majority (≈ 80%) of contacts were short-lived (<10 min), with only a handful (≈ 5%) of contacts lasting longer than 30 min (**Figure 2E**). No long-lasting contacts were observed for I-A/I-E <sup>−</sup>/<sup>−</sup> DCs, but overall no major differences in T cell contact times between WT and I-A/I-E−/<sup>−</sup> DCs were observed (**Figure 2E**). However, WT DCs showed a tendency to be more occupied by T cells than I-A/I-E−/<sup>−</sup> DCs were (**Figure 2F**).

#### Adoptively Transferred Antigen-Presenting DCs Engage in Prolonged Interactions With Cognate Intralymphatic T Cells During a Delayed-Type Hypersensitivity (DTH) Response

Although not analyzed, most probably only a fraction of DsRed<sup>+</sup> T cells recruited into the skin was hapten-specific in our CHS model (**Figure 2**). Moreover, considering that we had not exposed DCs to the CHS-inducing agent oxazolone prior to adoptive transfer, cognate DC-T cell interactions were unlikely to have been observed by intravital microscopy in this model. To overcome this limitation, we switched to investigating DC-T cell interactions during a DTH response in which only T cell receptor (TCR) transgenic, cognate antigen-specific T cells were DsRed+. To do so, we crossed TCR transgenic OTII mice, in which T cells are specific to ovalbumin-derived peptide OVA323−<sup>339</sup> presented on I-A/I-E (14), with hCD2-DsRed mice. CD4<sup>+</sup> T cells from hCD2-DsRed×OTII mice were transferred intravenously into Prox1-GFP mice, and mice were immunized with ovalbumin 1 day later (**Figure 3A**). After 4–7 days, ovalbumin was injected into the ears in order to elicit a DTH response (**Figure 3A**). Two days after elicitation, mouse ears were visibly reddened and ear thickness had increased (**Figure S3A**). By intravital microscopy we observed many DsRed<sup>+</sup> T cells within the tissue and inside lymphatic capillaries (**Figure S3B**). Characterization of the T cell population in DTH-inflamed ears revealed that DsRed+OTII T cells constituted ≈ 5–20% of CD4<sup>+</sup> T cells in the ear skin (**Figures S3C,D,E**).

With a working DTH model, we adoptively transferred DeepRed-labeled, OVA323−339-pulsed WT or I-A/I-E−/<sup>−</sup> BM-DCs into the DTH-inflamed ear skin of our model mice (**Figures 3A,B**). In comparison to the CHS model (**Figure 2**), contacts between WT DCs and T cells were less dynamic in nature: although WT DCs occasionally contacted more than one T cell, the majority of WT DCs engaged in long-lived contacts with a single T cell (**Figures 3D,E**, **Movie S6**). Most notably, more than 60% of contacts between WT DCs and T cells lasted longer than 30 min (**Figure 3E**). Conversely, the majority of contacts between I-A/I-E−/<sup>−</sup> DCs and T cells were short-lived, with <20% of contacts lasting longer than 30 min (**Figure 3E**). Moreover, whereas more than 65% of WT DCs were occupied by T cells for more than 80% of their track duration, only around 25% of I-A/I-E−/<sup>−</sup> DCs were equally occupied by T cells (**Figure 3F**). Consequently, intralymphatic I-A/I-E−/<sup>−</sup> DCs migrated faster than their WT counterparts (**Figure 3C**). However, in a competitive homing experiment, adoptively transferred WT and I-A/I-E−/<sup>−</sup> DCs migrated equally well to the draining LN (**Figure S4**).

FIGURE 2 | DCs interact with T cells inside lymphatic capillaries and short interactions are I-A/I-E-independent in CHS-inflamed mouse ear skin. (A–F) Intravital microscopy was performed in CHS-inflamed ear skin of hCD2-DsRed×Prox1-GFP mice after adoptive transfer of DeepRed-labeled WT or I-A/I-E−/<sup>−</sup> BM-DCs. (A) Schematic diagram of the experimental setup. (B) Time-lapse images of a DeepRed<sup>+</sup> WT DC (DC, cyan) contacting DsRed<sup>+</sup> T cells (T1 and T2) inside a lymphatic capillary (scale bars: 30µm). Times are shown in min. (C) Speed of WT and I-A/I-E−/<sup>−</sup> DCs within lymphatic capillaries. (D) Plots of contact between WT and I-A/I-E−/<sup>−</sup> DCs and T cells inside lymphatic capillaries. Each line is a DC indicating contact (green) and no contact (gray) with T cells. WT = 69 DCs, 174 contacts; I-A/I-E−/<sup>−</sup> = 77 DCs, 196 contacts. (E) Quantitative analysis of contact times from (C) are shown individually and after classification into three contact time groups. Median is shown as a red bar. (F) The occupancy of DCs by T cells from (C) are shown individually and after classification into three groups. Each dot in (C,E,F) represents a tracked cell. Medians are shown as red bars. Pooled data from 6 mice per group are shown.

3–4 mice/group each are shown.

### Endogenous DCs Interact With T Cells Inside Lymphatic Capillaries During a DTH Response to Ovalbumin

To investigate whether also endogenous DCs interact with T cells inside lymphatic capillaries, we established our hCD2- DsRed×OTII DTH model in Prox1-Orange×CD11c-YFP mice, which feature orange lymphatic vessels and yellow DCs (**Figure 4A**). Two days after challenge with ovalbumin, both endogenous YFP<sup>+</sup> DCs and in vivo-expanded CD4<sup>+</sup> hCD2- DsRed×OTII cells could be observed actively migrating and interacting inside lymphatic capillaries (**Figure 4B**, **Movie S7**). Analysis of contact plot data (**Figure 4C**) revealed that the majority of DCs engaged in short-lived contacts (<10 min) with T cells, with only a small percentage (5%) of DCs engaging in contacts longer than 30 min (**Figure 4D**). Although several DCs contacted more than one T cell, the majority of DCs engaged in a single contact with a T cell during an imaging period (**Figure 4C**). Consequently, the majority of DCs were occupied by a T cell/s for <40% of their track duration (**Figure 4E**).

### T Cells, but Not DCs, Can Exit Lymphatic Capillaries in Murine Skin

Upon close inspection of videos generated from our CHS and DTH imaging setups, we occasionally observed T cells that exited lymphatic capillaries back into the surrounding tissue (**Figure 4F**, **Movie S8**). Egress across the endothelium was brief (≈ 2–5 min) and T cells visibly squeezed their cell body in order to exit the lymphatic lumen (**Figure 4F**, **Movie S8**). Although these events were not seen in every video, quantification in our endogenous DTH setup revealed that in total, ≈ 5–10% of intralymphatic T cells exited capillaries during an imaging period of 45 min (**Figure 4F**). Conversely, in all our imaging experience over the years, and quantitatively shown in our endogenous DTH setup (**Figure 4F**), we never observed a DC exit a lymphatic capillary.

### DISCUSSION

In this study we have used intravital microscopy to further detail the behavior of DCs and T cells within dermal lymphatic capillaries during an ongoing immune response. In agreement with previous reports by us and by others (7–10) DCs and T cells actively migrated and patrolled within lymphatic capillaries. Moreover, we found that both cell types frequently arrested or clustered within lymphatic capillaries and that T cells, but not DCs, occasionally exited from the vessel lumen back into the tissue. Most intriguingly, we observed the occurrence of intralymphatic DC-T cell interactions.

Interactions between DCs and T cells are crucial for mounting an adaptive immune response. While they have mostly been studied in the draining LN during priming of naïve T cells (5, 6), only a few studies have investigated interactions of DCs with antigen-experienced effector/memory T cells in peripheral tissues like the skin (16–18). Given the spatial confinements of lymphatic vessels and the enhanced recruitment of activated DCs and T cells and drainage of antigen into the vessels during inflammatory processes, afferent lymphatic vessels might provide an ideal local compartment for adaptive modulation of the ongoing immune response. Intriguingly, several previous studies analyzing leukocyte subsets in afferent human lymph already reported the presence of cell aggregates comprising DCs and IFNγ-secreting CD4<sup>+</sup> T cells (4, 19–21), indicating that also in humans DC-T cell interactions might be occurring in afferent lymphatic vessels.

When eliciting a DTH response toward ovalbumin in mice with DsRed<sup>+</sup> TCR-transgenic OT-II T cells the majority of adoptively transferred OVA323-<sup>339</sup> peptide-presenting DCs engaged in long-lived, MHCII (I-A/I-E)-dependent interactions with TCR transgenic OTII T cells (**Figure 3**). By contrast, when imaging polyclonal DsRed<sup>+</sup> T cells 24 h after elicitation of a CHS response—a setup where likely only few of the adoptively transferred (unpulsed) DCs were presenting a cognate, haptenated antigen—numerous short-lived but only few long-lived DC-T cell interactions were observed (**Figure 2**). Similarly, imaging in our endogenous model of an ovalbumininduced DTH response 48 h after ovalbumin injection (**Figure 4**), long-lived interactions only occurred in 5% of all cases. The reason why not more long-lived endogenous DC–T cell interactions occurred may be linked with the (unknown) level of OVA323−<sup>339</sup> peptide presentation: In contrast to the BM-DC transfer experiments, where imaging was carried out shortly after transfer of OVA323−<sup>339</sup> peptide pulsed DCs (**Figure 3**), likely much less OVA323−<sup>339</sup> was present on endogenous intralymphatic DCs when imaging 2 days after ovalbumin challenge (**Figure 4**). Overall antigen availability has been recognized as an important determinant of the duration of DC-T cell contacts in other studies (22–24). Moreover, somewhat in line with our findings, interactions within the tissue of DTHinflamed rat ear skin were shown to progress from long-lived contacts during onset to less frequent short-lived contacts during the peak of the response (17).

At this point we do not know the specific subset of T cells involved in the observed intralymphatic DC-T cell interactions, and we can only speculate about the potential immunological significance of these interactions. Given current knowledge of T cell trafficking through inflamed afferent lymphatic vessels (3, 25) it is likely that intralymphatic DC-T cell interactions either involve CD4<sup>+</sup> effector T cells or regulatory T cells (Tregs). During an ongoing immune response, both effector T cells and Tregs are recruited into peripheral tissues irrespective of their antigenic specificity (18, 26, 27). By contrast, exit from the inflamed tissues via lymphatic vessels appears to be at least in part dependent on whether or not the T cell encountered its cognate antigen while surveying the interstitial space (28, 29). Particularly in the initial phase of a developing immune response (e.g., an infection), when antigen is still scarcely distributed, cognate effector T cells might not encounter their antigen on antigen-presenting cells scanned in the tissue and hence exit into lymphatics. In the case that an effector T cell now encountered a cognate antigen-presenting DC within lymphatic capillaries, the effector T cell could be re-activated and instructed to exit the lymphatic vessel back into surrounding tissue and continue searching for antigen in order to exert its effector functions in the tissue. Considering this scenario, it is intriguing that we

occupancy of DCs by T cells from (C) are shown individually and after classification into three groups. Each dot in (D) and (E) represents a tracked cell. Medians are shown as red bars. (F) Intravital microscopy snapshot of a DsRed<sup>+</sup> T cell (T) exiting a lymphatic capillary (scale bar: 30µm) and quantification of the percentage of intralymphatic DCs or T cells that exited a lymphatic capillary during an imaging period. Each dot represents a movie analyzed. Mean and standard deviation are shown. Pooled data from 5 mice are shown.

found a substantial fraction of intralymphatic T cells exiting the vessel again (**Figure 4**). Overall, this exiting behavior could contribute to immunosurveillance, as these cells would take a "short-cut" back into tissue where their cognate antigen might be located, rather than recirculating through draining LNs, lymphatic vessels, and blood. In any case, our finding of T cells exiting back into the tissue asks for a revision of the current lymphatic trafficking paradigm, in which afferent lymphatic vessels have thus far exclusively been regarded as cellular tissue exit routes.

Interestingly, besides effector T cells, Tregs were found to constitute ∼50% of T cells emigrating from CHS-inflamed skin to draining LNs via afferent lymphatic vessels (30, 31). Moreover, Tregs arriving via afferent lymphatic vessels were shown to contribute to the suppression of immunity in draining LNs and to be important for preventing exacerbated CHSinduced inflammation in skin (30, 31). In addition to directly suppressing T cell priming (32), Tregs are capable of suppressing the maturation phenotype of antigen-presenting DCs (33). Considering that afferent capillaries accumulate both DCs and Tregs that are exiting from the tissue, it is perceivable that DC-Treg interactions that modulate the DC phenotype might already take place in this compartment. Given the availability of Treg-specific reporter mice, such as FoxP3-GFP mice (34), this hypothesis could be investigated in the near future.

We also observed that endogenous intralymphatic DCs or T cells frequently clustered amongst themselves. While we occasionally found several clustering T cells interacting with DCs, DC-T cell interactions do not seem to explain the formation of all or larger T cell clusters. Homotypic T cell-T cell clusters have previously been described within LNs and have been shown to augment T cell activation and differentiation via paracrine signaling of IL-2 and IFN-γ (35, 36). Although we cannot yet specifically determine the activation status of clustering T cells within lymphatic vessels, this might become possible in the future using new photoconvertible systems, such as Kaede mice (37).

Our simultaneous imaging of DCs and T cells also revealed that not all immotile DCs or T cells were necessarily engaging in interactions with other DCs or T cells, but that some cells simply remained arrested on the lymphatic endothelium for long time periods. It is tempting to speculate that DCs and T cells might be exchanging immune-modulatory signals with lymphatic endothelium during these lengthy arrest and interaction periods. Interestingly, autoantigen-presenting lymphatic endothelial cells in LNs have been identified as important players in the regulation of peripheral CD8<sup>+</sup> T cell tolerance (1, 38). Moreover, LN lymphatic endothelial cells have been shown to archive exogenous antigen derived from viral infections or vaccinations and pass it on to migratory DCs capable of antigen crosspresentation (39). Besides impacting CD8<sup>+</sup> T cell responses, emerging studies have also identified a role for LN lymphatic endothelial cells in modulating CD4<sup>+</sup> T cell responses, either by directly presenting antigen to T cells or by transferring selfantigen to DCs, leading to the induction of T cell anergy and tolerance (40, 41).

Of interest, we have observed that similarly to LN lymphatic endothelial cells, lymphatic endothelial cells in afferent lymphatic vessels upregulate PDL-1 and MHCII in response to inflammation [(42) and data not shown]. Together with our intravital microscopy observations of lengthy DC and T cell arresting, this could suggest that lymphatic endothelial cells in afferent lymphatics might exert similar immune-modulatory functions as in draining LNs. In fact, in the case of DCs, the production of prostaglandins (43) and ICAM-1 by lymphatic endothelial cells of afferent lymphatic vessels (44) were already suggested to impact maturation of migratory DCs. Moreover, dermal lymphatic vessels have recently been identified as

#### REFERENCES


important players in the regulation of peripheral CD8<sup>+</sup> T cell tolerance during tumor growth (2). Given the extensive time that DCs and T cells spend inside afferent lymphatics during their exit from the tissue, future studies should investigate the expanding role of afferent lymphatic endothelium in the immunomodulation of intralymphatic passengers.

In summary, our study for the first time reports the occurrence of adaptive DC-T cell interactions within lymphatic capillaries. Combined with the current literature, our findings provide several further pieces of evidence suggesting that afferent lymphatic vessels represent more than just a trafficking route to draining LNs but rather a new compartment for adaptive immune interactions and immune modulation.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

### AUTHOR CONTRIBUTIONS

MH designed research, performed research, analyzed data, and wrote the paper. AT designed research, performed research and analyzed, and discussed data. RM, ER, and PR performed research and analyzed and discussed data. FK provided mice and discussed data. CH designed research, analyzed data, and wrote the paper.

#### FUNDING

CH acknowledges financial support from the Swiss National Science Foundation (grants 310030\_182528 and 310030\_156269) and ETH Zurich.

#### ACKNOWLEDGMENTS

The authors acknowledge support of the Scientific Center for Optical and Electron Microscopy (ScopeM) of ETH Zurich. We thank Angela Vallone (ETH Zurich) for excellent technical assistance and the staff of the ETH Rodent Center HCI for animal husbandry.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00520/full#supplementary-material


interaction kinetics and memory fate decisions. Immunity. (2013) 39:496–507. doi: 10.1016/j.immuni.2013.08.034


cells and dendritic cell migration in a stimulus-dependent manner. Blood. (2011) 118:205–15. doi: 10.1182/blood-2010-12-326447


**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 © 2019 Hunter, Teijeira, Montecchi, Russo, Runge, Kiefer and Halin. 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(s) 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.

# Mesenteric Lymphatic Alterations Observed During DSS Induced Intestinal Inflammation Are Driven in a TLR4-PAMP/DAMP Discriminative Manner

#### Matthew Stephens <sup>1</sup> , Shan Liao<sup>2</sup> and Pierre-Yves von der Weid<sup>1</sup> \*

<sup>1</sup> Department of Physiology and Pharmacology, Inflammation Research Network, Snyder Institute for Chronic Diseases, University of Calgary, Calgary, AB, Canada, <sup>2</sup> Department of Microbiology, Immunology and Infectious Diseases, Inflammation Research Network, Cumming School of Medicine, Snyder Institute for Chronic Diseases, University of Calgary, Calgary, AB, Canada

#### Edited by:

Timothy Padera, Massachusetts General Hospital and Harvard Medical School, United States

#### Reviewed by:

Sanjukta Chakraborty, Texas A&M University Baylor College of Dentistry, United States Sylvie Lesage, Université de Montréal, Canada

\*Correspondence:

Pierre-Yves von der Weid vonderwe@ucalgary.ca

#### Specialty section:

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

Received: 10 December 2018 Accepted: 01 March 2019 Published: 26 March 2019

#### Citation:

Stephens M, Liao S and von der Weid P-Y (2019) Mesenteric Lymphatic Alterations Observed During DSS Induced Intestinal Inflammation Are Driven in a TLR4-PAMP/DAMP Discriminative Manner. Front. Immunol. 10:557. doi: 10.3389/fimmu.2019.00557 Background: Inflammatory bowel disease (IBD) is characterized by both acute and chronic phase inflammation of the gastro-intestinal (GI) tract that affect a large and growing number of people worldwide with little to no effective treatments. This is in part due to the lack of understanding of the disease pathogenesis and also the currently poorly described involvement of other systems such as the lymphatics. During DSS induced colitis, mice also develop a severe inflammation of terminal ileum with many features similar to IBD. As well as inflammation within the ileum we have previously demonstrated lymphatic remodeling within the mesentery and mesenteric lymph nodes of DSS-treated mice. The lymphatic remodeling includes lymphangiogenesis, lymphatic vessel dilation and leakiness, as well as cellular infiltration into the surrounding tissue and peripheral draining lymph nodes.

Methods: Intestinal inflammation was induced in C57BL/6 mice by administration of 2.5% DSS in drinking water for 7 days. Mice were treated with TLR4 blocker C34 or Polymyxin-B (PMXB) daily from days 3 to 7 of DSS treatment via I.P. injection, and their therapeutic effects on disease activity and lymphatic function were examined. TLR activity and subsequent effect on lymphangiogenesis, lymphadenopathy, and mesenteric lymph node cellular composition were assessed.

Results: DSS Mice treated with TLR4 inhibitor, C34, had a significantly improved disease phenotype characterized by reduced ileal and colonic insult. The change correlated with significant reduction in colonic and mesenteric inflammation, resolved mesenteric lymphangiectasia, and CD103<sup>+</sup> DC migration similar to that of healthy control. PMXB treatment however did not resolve inflammation within the colon or associated mesenteric lymphatic dysfunction but did however prevent lymphadenopathy within the MLN through alteration of CCL21 gradients and CD103<sup>+</sup> DC migration. Conclusions: TLR4 appears to mediate several changes within the mesenteric lymphatics, more specifically it is shown to have different outcomes whether stimulation occurs through pathogen derived factors such as LPS or tissue derived DAMPs, a novel phenomenon.

Keywords: lymphatics, mesentery, toll-like receptors, inflammation, lipopolysaccharides, colitis

#### INTRODUCTION

IBD constitutes of two major phenotypes of gastrointestinal diseases, Ulcerative colitis and Crohn's disease. Both diseases have an inflammatory component which results in impaired nutrient absorption, cell recruitment, and chronic inflammation. During the pathogenesis of IBD, a major alteration occurring within the hosts GI tract is focused upon the lymphatics.

Severe lymphatic remodeling has been observed within the intestinal wall of IBD patients, at the stage of the initial (lacteal) lymphatics, through to the collecting mesenteric lymphatics, and mesenteric lymph nodes (1–5). However, what effect these alterations are having upon disease progression is still not fully understood. The expansion of the lymphatic network, also known as lymphangiogenesis, is mediated through the binding of the lymphatic vascular endothelial selective growth factors VEGF-C and VEGF-D to VEGFR3, and is a common feature in Crohn's disease (6–8). Blockade of lymphangiogenesis through anti-VEGFR3 antibodies do not provide any therapeutic benefit but rather exacerbates submucosal oedema in animal models of IBD (9, 10), while stimulation of lymphatic functions with VEGF-C ameliorates experimental IBD (10). Therefore, it presents the idea that lymphangiogenesis may in fact be a reparative measure in response to inflammation and pro-lymphangiogenic factors, such as VEGF-C, may provide novel strategies for the treatment of chronic inflammatory diseases. Lymphangiectasia, the dilation of lymphatic vessels, is a common sign of collecting lymphatic vessel disruption. It has been shown to be associated in case of inflammation, and intestinal inflammation in particular, with increase permeability of the lymphatics (3, 4). This can result in oedema, hypoproteinaemia, lymphocytopenia, and immunologic anomalies. Another great concern associated with the leakage is lipid absorption issues resulting in weight loss and fat deposition within the mesothelium. To what extent the fat contributes to inflammation, and the effect it has on resident cells, is still not fully elucidated however has been suggested that lymphaticassociated fat can be a source of inflammatory material and may play a greater role in disease pathogenesis than first expected (11– 13).

Toll-like receptors (TLRs) play a key role in mucosal innate immunity and may be involved in the pathogenesis of IBD (14). An evolutionarily conserved family of transmembrane pattern recognition receptors, TLRs recognize pathogen-associated molecular patterns (PAMPs) conserved between microbes (15). Activation of TLRs results in the induction of cytokines, chemokines, and antimicrobial molecules, all important factors in the initial innate response aiding in priming the adaptive immune system (16, 17). TLR4 binds the gram-negative bacterial cell wall component lipopolysaccharide (LPS) and through coreceptor MD-2, interaction triggers both MyD88-dependent and independent pathway leading to the translocation of NFκB and subsequent production of inflammatory cytokines and proteins (18, 19). TLR4-mediated signaling is important for the recruitment of immune cells to the site of inflammation promoting reparative mechanisms, but can be described as a double-edged sword, as aberrant stimulation can induce chronic inflammation (20).

The lymphatic system is a complex network of specialized vessels involved in tissue fluid homeostasis. Lymphatic vessels drain fluid from tissues and associated organs, and propel it unidirectionally as nutrient- and cell-rich "lymph" back into peripheral blood circulation. Initial lymphatic vessels comprised of closed-end, lymphatic capillaries, which branch into tissue then amalgamate to form larger collecting vessels, which, through the presence of smooth muscle cells surrounding the endothelium wall, propel lymph via peristaltic-like contractions toward the draining lymph node. Formation of lymph is believed to occur through the swelling of the interstitium, respiration, arterial pulsations and skeletal movement. Increased interstitial pressure opens the initial lymphatic vessels through small anchoring filaments attaching endothelial cells to the extracellular matrix. During inflammatory diseases such as IBD, increased localized swelling within a tissue creates an increased burden upon the draining lymphatics. Within IBD, disruptions in the mesenteric lymphatic architecture has been correlated to worsened disease progression, putting the changes under scrutiny for their potential contribution to pathogenesis (21, 22).

We aimed to determine whether lymphatic disruption in the mouse model of DSS-induced ileitis/colitis, was in portion driven by TLR4. In order to block TLR4 activity directly and indirectly, two drugs were used. The first, Polymyxin-B, inhibits TLR4 recognition of LPS by binding the lipid-A component of LPS preventing recognition by the receptor. The other, C34, is a direct chemical inhibitor of TLR4 binding to the receptor in an antagonistic and competitive manner. Identified by Neal and colleagues, C34 was a potent TLR4 blocker in enterocytes and macrophages in vitro, and reduced systemic inflammation in mouse models of endotoxemia and necrotizing enterocolitis (23). These treatments allowed us to differentiate between the activation of TLR4 by LPS or by TLR4-directed agonists of another source.

#### MATERIALS AND METHODS

#### Mice

All mice used were housed at constant temperature (22◦C) on a 12:12-h light-dark cycle, with food and water ad libitum. The animal handling and experiments were approved by the University of Calgary Animal Care and Ethics Committee and conformed to the guidelines established by the Canadian Council on Animal Care.

#### Cell Culture

HEK293 (TLR4/MD2/CD14) dual reporter cells (Invivogen, USA Cat. No HKD-mTLR4ni) were maintained in supplemented DMEM (High Glucose) 10% FCS with Normocin, Hygromycin Gold, and Zeocin as per manufacturers instruction. Cells were passaged every 3–5 days at 80% confluency and maintained in a 37◦C incubator, 5% CO<sup>2</sup> atmosphere. All cells were used between passages 3–8 and experiments were performed on 3 or more distinct passages of cells.

#### Induction of Colitis and Administration of Treatments Acute DSS

Six-week-old C57BL/6 mice were obtained from Jackson Laboratories. Colitis was induced in these mice by administration of 2.5% (weight/vol) dextran sulfate sodium (DSS; Affymetrix, Cleveland, Ohio, USA) in drinking water for 7 days. Sham mice were given normal drinking water. I.P. injections of C34 (50 mg/kg) (Tocris, USA, Cat. No 5373) or Polymyxin-B (50 mg/kg) (Sigma Aldrich, USA, Cas. No 1405-20-5) were administered from days 3 to 7 diluted to 200 µl total in saline, control mice received saline only. Mice were euthanized by exposure to isoflurane and cervical dislocation.

#### Disease Evaluation

In order to assess the severity of DSS-induced inflammation a multi-parameter approach was used in order to quantify inflammation by region. Colon shortening, a common sign of inflammation-driven fibrosis, was measured as a marker of colonic inflammation. Differences in weight were calculated as the percentage weight loss pre- and post-treatment (SHAM/DSS/DSS + treatment). Additionally, fecal matter consistency and visual blood presence were assessed. All of these factors were evaluated, and using the scoring system detailed in **Table 3**, the disease activity score (DAI) was calculated (see **Table 1**) [adapted from (24)].

### Alterations in Lymphatics

#### Lymphangiectasia

Lymphatic vessels were identified as initials by positive staining with CCL21 (R&D systems, USA, Cat. No AF457) and as collectors by staining with αSMA (Sigma-Aldrich, USA, Cat. No C6198) in whole-mount mesenteric preparations. Vessel diameters were measured in 3–5 vessels per sample at 3 random sites along each vessel.

#### Lymphadenopathy

Mesenteric lymph nodes (MLNs) were isolated from the mouse, cleaned of fat and connective tissue and measured for both size (lengthways) and weight.


#### Lymphangiogenesis

Lymphangiogenesis was measured via CCL21 staining of mesenteric whole-mounts. Vessel branch points and numbers were determined in a fixed area of interest kept uniform between samples.

#### Whole Mount Immunofluorescence

Whole mount mesenteries were fixed on sylgard coated dishes and fixed with 4% PFA for 1 h at room temperature. Tissues were washed, permeabilized in PBST (PBS + 0.03% Triton X-100) and blocked with 2–3% BSA in PBST. Primary antibodies incubation occurred for 24 h at 4◦C. Samples were washed three times in PBST (PBS + 0.01% Triton X-100) for 10 minutes per wash and then incubated with secondary antibodies (in 2% BSA containing 0.01% Triton X-100) for 1–2 h. Samples were washed as previously described before preparation for optical clearing. Fat-clearing was obtained by serial ethanol dehydration followed by methyl salicylate (MeS) immersion for 15 min. Immediately after clearing, samples were mounted with DAPI containing mounting medium and installed with a coverslip for imaging. The imaging occurred within 2 h of clearing due to fluorescent diminishment. Vessel diameter and branching was quantified using the LASX software attached to a Leica SP8 confocal microscope.

### Quantitative Real-Time Polymerase Chain Reaction (qPCR)

The total RNA isolated from given samples was purified using the QIAGEN RNA total cleanup kits as per manufacturers instruction. One hundred nanogram of the RNA was converted using EvaGreen RT conversion kit in a gradient thermocycler as per manufacturers description. One nanogram per microliter of the converted cDNA samples was added to EvaGreen SYBR qPCR master-mix and qPCR analysis was performed in an ABI StepOne Plus PCR system. Annealing temperatures were kept at 60◦C, and


#### TABLE 2 | RT-PCR primer sequences.

qPCR primer sets used for the amplification of genes of interest.

TABLE 3 | TLR4 mRNA expression changes in murine mesentery during DSS treatment.


DSS induces expression of TLR4 within the mesentery. Mesenteric lymphatic samples from SHAM and DSS treated mice (2.5% 7d) were analyzed for TLR expression using qPCR. Unpaired Mann-Whitney test, two-tailed.

40 cycles of amplification were performed to produce a sufficient read. Sequences of primers used are detailed in **Table 2**.

#### Total LPS Isolation From Murine Feces

Samples were homogenized in PBS and subsequently filtered in order to remove non-soluble components. Protein content of the fecal homogenate was determined through the Precision Red Protein Quantification assay (Cytoskeleton, Inc., USA Cat. No ADV02). The concentration of samples was then equilibrated to 1 mg/ml through addition of supplemented DMEM, before being assayed for endotoxin content using the chromogenic-Limulus amebocyte lysate (LAL) assay (ThermoFisher, USA, Cat. No 88282). Equal volumes (3.5 ml) of quantified fecal homogenate was then passed through a high capacity endotoxin removal spin column (Pierce, USA. Cat No 88274) reducing LPS content from an average of 87.67 EU/mg/ml (±7.36) to <1.71 EU/mg/ml (±0.39). The bound LPS was removed for later studies.

#### HEK-TLR4 Cell Stimulation

Cells were maintained in Normocin, Hygromycin Gold, Zeocin DMEM-high glucose media as per manufacturers instruction. During stimulation Hygromycin Gold and Zeocin were not present in the media. "Endotoxin-low" samples isolated using the PMXB columns were added to HEK293 (TLR4/MD2/CD14) dual reporter cells (Invivogen, USA Cat. No HKD-mTLR4ni) in order to assess the stimulatory capacity of the material through TLR4. Samples were diluted 1:10 in supplemented DMEM to remove toxic effects of salt exchange which occurred during the LPS removal process reducing endotoxin levels to below threshold (<0.1EU/ml) a value at which comparable LPS concentrations do not activate the cells.

#### Statistics

Data are expressed as the mean ± one standard error of the mean (SEM). Sample size varies from 3 to 9 as indicated, performed as a minimum experimental triplicate. Statistical significance was assessed through the use of two-tailed unpaired Student's t-test for parametric data, while the Mann-Whitney test was performed for non-parametric data. Multiple analyses were performed using a one-way Anova with post-hoc Tukey test where indicated. <sup>∗</sup>P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001.

### RESULTS

#### DSS Alters TLR mRNA Expression Within the Mesentery

Previous findings have demonstrated DSS directly impacting on the expression of TLRs within the inflamed colon mimicking that found in the patient cohort (28–31). We wanted to ascertain whether there the same was true for the murine mesentery. In order to do so, qPCR screening of all known murine TLRs, was assessed. Total RNA from the mesentery of sham-control mice, DSS, and other treatments were isolated and converted to cDNA before analysis via SYBR-green qPCR. Of the 14 known murine TLRs only 7 were detectable within the murine mesentery samples (**Table 3**). Analysis of expression within sham controls provided a baseline for subsequent comparison with DSS which demonstrated a significant upregulation in TLR4 (P < 0.0001). We hypothesized that TLR4, which recognize the bacterial component LPS, could be involved heavily in lymphatic-driven inflammation and dysfunction during DSS induced colitis. Additionally, during the progression of DSS induced colitis, the epithelium of the gut is severely disrupted allowing a vast influx of lumenally-derived microbial content into the mesenteric lymphatics.

### DAMPs Created Within the Colon During DSS Treatment Activate Cells in a TLR4 Dependent Manner

In order to discern what, within the GI tract, could be activating upon TLR4 directly, samples of colonic fecal matter were collected from sham and DSS treated mice for analysis. Data shown in **Figure 1A** show protein normalized samples and their subsequent LAL-determined endotoxin levels per milligram of fecal matter from both sham and DSS treated mice. The levels of endotoxin present in the sample are denoted as: PRE (before) and POST (after) endotoxin removal via the Polymyxin-B column. **Figure 1B** demonstrates that the induction of the NFκBSEAP (Secreted Embryonic Alkaline Phosphatase) and IL-8 luciferin reporters found within the HEK-TLR4 reporter cells can be driven by substances within the murine fecal matter. A significant reduction in the induction of gene expression can be seen through the removal of LPS. However, a proportion of the (POST) DSS sample can still induce both NF-κB and IL-8 gene expression suggesting other molecules are being recognized by TLR4.

#### TLR4 Blockade Through C34 Treatment Ameliorates the Progression of DSS Induced Disease Activity

Being home to the majority of the microbiome, the gut must function effectively as a barrier in order to prevent the influx of microbial pathogens into the normally sterile sub-mucosa. The multi-layered composition of the intestinal tract aids in this function through the secretion of mucins, the epithelial barrier itself and the rapid response of immune cells a site of breach (32). During DSS induced colitis, the breakdown of the epithelial barrier leaves the potential for invasion of commensal bacterium, fungi, viruses and dietary substances to permeate the pseudosterile barrier (33, 34). Activation of the resident macrophages, dendritic cells and mast cells within the epithelial sub-lining promotes the recruitment of neutrophils, the induction of proreparative measures, and the clearance of antigens to the lymph node in an attempt to create and effective immune response to the infection (35, 36).

With high levels of bacterial LPS and DAMPs present within the intestinal luminal space, a potential to activate a TLR4 mediated innate immune response is rife. Inflammatory molecules induced by TLR4 activation are documented to negatively impact on lymphatic function, thus potentially reducing flow of antigens to the lymph node and subsequent immunosuppression (37). Therefore, we attempted to determine the effect of inhibition of TLR4 and subsequent effect on inflammation within the local drainage lymphatic system. Mice treated with 2.5% DSS for 7 days, received a daily I.P injection of either saline, C34 or PMXB from days 3 to 7 (See Methods). DSS colitis in mice is characterized by the development of diarrhea, colonic inflammation, and subsequent weight loss. Fecal consistency, blood presence in feces and the extent of the colon shortening was converted to a DAI and recorded as described in Methods (see **Table 3**). When compared to sham controls, treatment with C34 significantly reduced weight loss (**Figure 2A**), reduced disease activity score (**Figure 2B**) and reduced colon shortening (**Figure 2C**). However, treatment with PMXB did not aid significantly in the characteristic disease phenotype. We also tested an alternative drug delivery method via oral gavage of the treatments in the same dosage and time frame, however, neither treatment alleviated any tested condition (Data not shown).

### TLR4 Modulates Lymphatic Alterations Within the Mesentery in an LPS-Independent Manner

The promotion of lymphangiogenesis within the mesentery of DSS treated animals is a well-documented phenomenon (3, 10). Alongside lymphangiogenesis, vessel dilation mediated by iNOSdependent production of nitric oxide is a common feature associated with inflammation within the tissue surrounding the lymphatics (37). The dilation often correlates with increased vessel permeability and particulate exchange, although the effectors regulating this process are still unknown. This "leakiness" can disrupt the flow of antigen-bearing immune cells to the lymph node and cause lymph and its content to spill out into the surrounding tissue. We hypothesized that one of the functions of mesenteric lymphatic network expansion is to resorb this lost material.

Mesenteric sections were isolated from sham and DSS mice treated or not with C34 or PMXB, fixed as whole-mount and stained with lymphatic vessel markers. The initial-lymphatic endothelial marker CCL21, and collecting lymphatic marker αSMA, were used to highlight the border of the lymphatic endothelium and/or smooth muscle layer, allowing vessel size measurement (**Figure 3**). As illustrated in **Figure 3**, shamcontrol collecting lymphatic vessels average luminal diameter and branch points were assessed creating baseline values. DSS treated samples showed extensive expansion of the lymphatic network and a significant increase in lymphatic vessel diameter. Restoration of the normal phenotype was successfully achieved through administration with C34. However, PMXB had no significant effect. This finding was compounded through the analysis of CCL21 mRNA levels within the affected tissue, whereby C34 reduced expression to that of sham control whilst PMXB had no effect on transcript levels. Altogether, these data show mesenteric lymphangiogenesis and lymphangiectasia can be resolved through the I.P. blockade of TLR4 using C34.

### TLR4 Activation Modulates Lymphangiogenic and Inflammatory Molecules Within the Mesentery

The inflammation associated with DSS-induced intestinal inflammation promotes lymphangiogenesis within the mesentery as well as altering lymphatic structure through dilation of the collecting lymphatic vessels (**Figure 2**). Analysis of mRNA levels of poignant lymphatic markers (**Figure 4**) revealed

significant increases in LYVE-1 (P < 0.001) and CCL21 (P < 0.05) transcription during DSS treatment with the effect ameliorated by treatment with C34. Interestingly, the widely accepted universal lymphatic endothelial marker PROX-1, was not significantly induced through DSS treatment. Rather, inhibition of TLR4 signaling through C34 treatment induced PROX-1 transcription suggesting that TLR4 regulates the lymphangiogenic transcription factor in an activation dependent manner. Additionally, COX2 (P < 0.0001) and iNOS (P < 0.01) mRNA levels spiked during DSS treatment, indicating signs of mesenteric dysfunction, with both treatments ameliorating this induction suggesting an anti-inflammatory effect of TLR4 inhibition within the mesentery itself. Interestingly VEGFR3 expression was not significantly increased in the DSS group compared to sham but was rather significantly reduced through C34 (P < 0.01) and PMXB (P < 0.01). Furthermore, analysis of common inflammatory markers (TNFα, IL-1β and IL-6)

in paired mesenteric samples showed no significant induction within any group at day 7, inferring the passage of the acute inflammatory phase within that region (Data not shown).

### LPS Drives Lymph Node Expansion and Cellular Migration During DSS Treatment

Lymphadenopathy, or swelling on the lymph node, is a common occurrence during the response to infection presenting in either a localized (regional) or diffuse (generalized) phenotype (38). Immune cells within the mesentery can become activated and promote expression of chemotactic agents within the collecting lymphatics such as CCL21 which, though production of a gradient, attracts CCR7<sup>+</sup> cells, such as dendritic cells, from the lamina propria to the mesenteric lymph node for antigen presentation (39, 40). These cells accumulate within the lymph node after trafficking antigens from a peripheral site of inflammation, and subsequently produce a wide array

of proliferative and chemoattractant agents, which, result in structural remodeling of the node (41, 42). TLR4 activation is known to induce TNF-a production, IL-8 secretion, and matrix protease secretion from a wide variety of cells, including fibroblastic reticular cells and macrophages, key players in lymph node remodeling (43, 44).

Examination of the lymph node corroborated that DSS had a distinct effect on lymphadenopathy significantly increasing the MLN size (P < 0.0001), weight (P < 0.001), and cellular content (P < 0.0001) (**Figure 5A**). This effect was not abrogated through C34 blockade of TLR4, however, PMXB treatment reduced significantly the MLN size (P < 0.0001), weight (P < 0.0001), and cell count (P < 0.0001) suggesting the possible involvement of a non-TLR4 dependent LPS interaction in lymphadenopathy. Altogether, these results suggest that TLR4 does not directly influence lymphadenopathy in a DSS model of colonic inflammation but LPS does, and it does so in a TLR4-independent manner. DSS treatment also significantly upregulated CCL21 expression within the MLN (P < 0.05) (**Figure 5B**). C34 in combination with DSS treatment had no effect on CCL21 expression in the MLN, however PMXB treatment significantly reduced it (P < 0.01), impacting the recruitment of CCR7<sup>+</sup> CD103<sup>+</sup> DCs accumulation (P < 0.05) (**Figure 5C**).

### DISCUSSION

Within the intestine there is a delicate balance between innate immune activation and inflammation. Invasive pathogens can be recognized by a myriad of pattern recognition receptors and induce an inflammatory response, a feature critical to the successful clearance of aforementioned pathogen (45). However, in a system that has the potential to be exposed to an enormous volume of these microbes, a correct magnitude of response is paramount (46). Patients suffering from IBD have an exacerbated and possibly unregulated inflammation within the intestinal tract, a feature that does not resolve normally as expected, but rather perpetuates in a chronic fashion. During the progression of IBD, and perhaps its conception, the structure of the gastrointestinal-associated and mesenteric lymphatic vascular environment changes drastically (22, 47–49). Whether this phenomenon is a direct cause of IBD or a causal agent of IBD

2–3 separate experiments. One-way ANOVA with Tukey post-hoc test \*P < 0.05, \*\*P < 0.01, \*\*\*P < 0.001, and \*\*\*\*P < 0.0001.

is debated. However, from our own lympho-centric opinion, the lymphatic dysfunction itself could cause both. Increased intestinal permeability, combined with reduced lymphatic function could lead to a stagnation of material within the effected intestinal region leading to a hot-spot of inflammation. In the DSS model, intestinal permeability is increased through the chemical ablation of the intestinal epithelium via the formation of nano-lipocomplexes between medium-chain-length fatty acids and DSS, highly abundant in the colon, therefore greatly isolating its effector venue (50). Loss of this barrier allows a vast milieu of microbial and dietary content to enter the submucosa and be directly exposed to the lymphatic system via the initial lacteal vessels. Within this transport system the exposure of immune and stromal cells to bacterial products such as LPS cause a

large inflammatory response, tissue remodeling and cellular proliferation/recruitment, a phenomenon well-documented in the lung (51, 52). During the progression of IBD extra-cellular matrix remodeling is common, the role of this remodeling however, is poorly understood and it is thought that extracellular molecules produced during this destructive/reparative stage may in fact perpetuate inflammation (53). We have previously demonstrated that even after the removal of DSS, remodeled lymphatics persist, a phenomenon that is evident with the IBD patient population (3).

TLR4 has been implicated in the pathogenesis of many inflammatory diseases including IBD, as through the recognition of LPS and a wide array of previously mentioned PAMPs and DAMPs, a large inflammatory stimulus can be generated (54). Our data reveals not only the TLR4-driven lymphatic alterations during DSS-induced colitis, but the gross-mechanisms which they act. Through the use of a competitive TLR4 antagonist (C34) and Polymyxin we were able to selectively differentiate between total TLR4-driven lymphatic alterations/consequences during DSS induced colitis, and those driven by LPS. Data presented, indicates that substances separate from LPS, i.e., TLR4-DAMPs, and other PRR PAMPs, modulate inflammation and lymphatics to a much greater extent than previously estimated. This hypothesis was confirmed through the detection and analysis of TLR4 activating material within the colon of DSS mice separate from the LPS content. Currently, specific DAMPs have not been elucidated in this system, however it would be feasible to expect well-published TLR4 DAMPs such as Tenacin-C, HMGB1, HSP90, or S100 proteins to be candidates as levels are known to drastically rise during tissue damage (55–57).

A key marker of correct lymphatic response was the potent lymphadenopathy seen during DSS treatment. This effect was disrupted through the PMXB treatment, accredited to the lack of CD103<sup>+</sup> DC migration to the lymph node. However, through C34 inhibition of TLR4, CCL21 (a potent lymphatic chemokine) was downregulated in the mesentery but not the lymph node, creating a gradient for increased movement of CCR7<sup>+</sup> CD103<sup>+</sup> DCs. This gives partial explanation to the reduced cellular content of the lymph node and significant reduction in lymphadenopathy which was ameliorated through PMXB treatment where the CCL21 content in the lymph node is

downregulated. Therefore, any CCL21-driven chemotaxis within the mesentery has no directionality. This also suggests that the lymphadenopathy, whilst caused by LPS, is not driven through TLR4, a novel finding in this context.

We also present evidence supporting our hypothesis that restoration of lymphatic function to a "normal" phenotype, significantly aids in the reparation of DSSinduced disease activity. Achieved through TLR4 blockade via C34, mesenteric lymphatic disruption was significantly reduced, evidenced by reduced lymphangiogenesis and lymphangiectasia. We note that this finding, of reduced lymphangiogenesis improving DSS-induced phenotype, is in somewhat opposition to D'Alessio's, work whereby lymphangiogenesis was beneficial to their model of IBD (10). This discrepancy could be solely contributed to the timing of the treatments or the target itself. Our experimental method and timings were designed to modulate inflammation after its genesis rather than in a preventative capacity. We attempted to modulate TLR4-associated inflammation and therefore the subsequent lymphatic remodeling, whereas D'Alessio and colleagues focused intentionally on promoting lymphatic remodeling through the overexpression of VEGFC, a method that likely had many targets separate from VEGFR3 induced lymphangiogenesis.

We know with certainty that alterations occur within the lymphatics of patients with of IBD and we are able to mimic them in murine models of DSS-induced intestinal inflammation. However, what is not yet understood is whether TLR4 could be a potential target for IBD in humans. Targeting such an important receptor undoubtedly has its risks but data presented in this paper suggest the plasticity of the receptor in delineating pathogenic material from self, a phenomenon that could be utilized in the future for the development of novel treatment of IBD.

### DATA AVAILABILITY

The datasets generated for this study are available on request to the corresponding author.

## AUTHOR CONTRIBUTIONS

Project planning and experimentation was solely carried out by MS. Interpretation of data, technical advice, and manuscript proofreading was performed by P-YvdW and SL.

### FUNDING

This study was supported by the Lymphedema Research and Education Program, Snyder Institute for Chronic Diseases, Cumming School of Medicine, University of Calgary to P-YvdW.

### 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.

Copyright © 2019 Stephens, Liao and von der Weid. 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(s) 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.

# Halted Lymphocyte Egress via Efferent Lymph Contributes to Lymph Node Hypertrophy During Hypercholesterolemia

Meng Hwee Daniel Tay 1†, Swee Yeng Jason Lim1†, Yew Fai Ivan Leong<sup>1</sup> , Chung Hwee Thiam<sup>1</sup> , Kar Wai Tan<sup>1</sup> , Federico Tesio Torta<sup>2</sup> , Pradeep Narayanaswamy <sup>2</sup> , Markus Wenk <sup>2</sup> and Véronique Angeli <sup>2</sup> \*

1 Immunology Programme, Department of Microbiology and Immunology, Life Science Institute, Yoon Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore, <sup>2</sup> Department of Biochemistry, Life Science Institute, SLING-Singapore Lipidomics Incubator, Yoon Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore

#### Edited by:

Amanda W. Lund, Oregon Health and Science University, United States

#### Reviewed by:

Witold Kilarski, University of Chicago, United States Joseph Rutkowski, Texas A&M University, United States

#### \*Correspondence:

Véronique Angeli micva@nus.edu.sg

†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: 19 December 2018 Accepted: 04 March 2019 Published: 27 March 2019

#### Citation:

Tay MHD, Lim SYJ, Leong YFI, Thiam CH, Tan KW, Torta FT, Narayanaswamy P, Wenk M and Angeli V (2019) Halted Lymphocyte Egress via Efferent Lymph Contributes to Lymph Node Hypertrophy During Hypercholesterolemia. Front. Immunol. 10:575. doi: 10.3389/fimmu.2019.00575 Dyslipidemia is a central component of atherosclerosis and metabolic syndrome linked to chronic inflammation and immune dysfunction. Previously, we showed that hypercholesterolemic apolipoprotein E knock out (apoE−/−) mice exhibit systemic effects including skin inflammation and hypertrophic lymph nodes (LNs). However, the mechanisms accounting for LN hypertrophy in these mice remain unknown. Here, we show that hypercholesterolemia led to the accumulation of lymphocytes in LNs. We excluded that the increased number of lymphocytes in expanded LNs resulted from increased lymphocyte proliferation or entry into those LNs. Instead, we demonstrated that the egress of lymphocytes from the enlarged LN of apoE−/<sup>−</sup> mice was markedly decreased. Impairment in efferent lymphatic emigration of lymphocytes from LNs resulted from an aberrant expansion of cortical and medullary sinuses that became hyperplastic. Moreover, CCL21 was more abundant on these enlarged sinuses whereas lymph levels of sphingosine 1 phosphate (S1P) were decreased in apoE−/<sup>−</sup> mice. Normal LN size, lymphatic density and S1P levels were restored by reversing hypercholesterolemia. Thus, systemic changes in cholesterol can sequester lymphocytes in tissue draining LNs through the extensive remodeling of lymphatic sinuses and alteration of the balance between retention/egress signals leading to LN hypertrophy which subsequently may contribute to poor immunity. This study further illustrates the role of lymphatic vessels in immunity through the regulation of immune cell trafficking.

Keywords: lymph node, lymphatic vessel, lymphocyte egress, mouse model, dyslipidemia

### INTRODUCTION

Lymph nodes (LNs) are highly specialized organs that play an essential role in immune priming and function. LN comprises of three main regions namely the cortex, paracortex, and the medulla. Lymph enters the LN via the afferent LVs and reaches the subcapsular sinuses first before it drains through the cortical sinuses in the cortex region, and into the medullary sinuses in the medulla region, and then eventually exits the LN via efferent lymphatic vessels. Protein, lipid, antigens, microorganisms, and immune cells such as lymphocytes and dendritic transported via the lymph enter the subcapsular sinus of the LN. Subcapsular and medullary sinuses are directly connected and thus lymph can pass through LN without filtering through the cortex (1). In most cases, lymphocytes enter the LN via the high endothelial venules (2), except for a small proportion of memory lymphocytes that enter via the lymphatics (3, 4). Activated and naïve lymphocytes in the LN matrix would eventually need to enter the sinuses so that they can be transported by the efferent lymph to reach the effector sites or return back to the circulation for immune surveillance, respectively. Lymphatic endothelial cells (LECs) in the cortical sinuses control lymphocyte trafficking within the LN by accumulating lymphocytes for further transit to medullary sinuses (5). Furthermore, lymphocytes can migrate from the lymphatic sinuses back to the LN parenchyma (5). Lymph borne lymphocytes can also be passively collected into the peripheral medullary sinuses, and then enter the LN parenchyma in a CCR7-independent manner by moving into adjacent peripheral medullary cords (6). The medulla is formed of a threedimensional labyrinthine structure of sinus channels starting as cortical sinusoids and expand to become wider medullary sinuses that finally drain collectively into the efferent lymphatic vessel (7). In addition to cortical sinuses, medullary sinuses have been proposed as exit routes for the egress of lymphocytes from LNs (5, 8, 9). Lymphocyte egress are governed by mechanisms controlling the entry of lymphocytes into the efferent lymphatic vessels including the regulation of CCR7 and sphingosine-1 phosphate receptor-1 expression (10–12), or those regulating lymphatic endothelial barriers (13, 14). It is well-established that lymphocyte egress from LNs via cortical and medullary sinuses is dependent on signals generated by lymph-borne S1P (8, 9, 15) produced by LECs via S1P kinase 1 kinase (16). S1P levels are low in LN parenchyma but high in lymph fluid thus creating a gradient. This S1P gradient guides T cells exhibiting decreased CCR7/CCL21-retention signals from LN parenchyma into medullary and cortical sinuses and ultimately facilitates T cell egress (10). Thus, the structural integrity and functions of these lymphatic sinuses have to be maintained as they are not only just passive carrier of lymph, but active player in the regulation of lymphocyte egress.

Local immune responses and inflammation are associated with alterations in the trafficking of lymphocytes through activated LNs. Indeed, the entry of lymphocytes into the draining LNs is increased whereas their exit into the efferent lymph is transiently blocked from few hours to days depending on the nature of the antigen and inflammation (17–20). In contrast to our knowledge about the mechanisms controlling lymphocyte retention within the inflamed LN after the initiation of an immune response (8, 21, 22), little is known about those reestablishing lymphocyte egress to steadystate levels. We showed previously that the expansion of cortical and medullary sinuses in the stimulated LNs at later stages of inflammation contribute to the reestablishment of lymphocyte egress rates (23). Reestablishment of steady-state egress from inflamed LNs may help to content the enhanced lymphocyte entry into LNs when inflammation is prolonged. This may prevent these LNs from becoming a "sink" that may compromise efficient lymphocyte recirculation and the timely induction of a suitable immune response (24, 25). This raises the possibility that LN hypertrophy and hyperactivation occurring with autoimmune and chronic inflammatory diseases may arise when lymphocytes fail to emigrate from the inflamed LNs in a timely manner. Notably, we and others reported in mice that hypercholesterolemia associated with atherosclerosis disease leads to systemic effects including skin inflammation, hypertrophic LNs, and compromised immunity (26–33). However, how hypercholesterolemia lead to LN hypertrophy in still remains an open question. Here, we explored how lymphocyte trafficking and LN structure and function are affected by hypercholesterolemia that occurs in mice lacking apoE (apoE−/−). The study reveals an accumulation of lymphocytes in enlarged skin draining LNs which results from a severe blockade of their egress rather than increased lymphocyte proliferation or entry within the enlarged LNs. We further show that hypercholesterolemia induces an aberrant remodeling of the cortical and medullary sinuses and an imbalance between retention and egress signals that account for the impaired lymphocyte egress in dyslipidemic mice.

### METHODS

#### Animals

Male CD45.2 C57BL/6 mice deficient in apolipoprotein E (apoE−/<sup>−</sup> mice) and low-density lipoprotein receptor (Ldlr−/−), CD45.2 wild type C57BL/6 (WT), and CD45.1 C57BL/6 WT were obtained from The Jackson Laboratory (Bar Harbor, ME). All mice were given a chow diet (18% protein and >5% fat, Harlan Teklad, Madison, WI) until 6 weeks of age. At 6 weeks of age, the diet was switched to a high fat and cholesterol rich diet containing 21.2% fat and 0.2% cholesterol (Harlan Teklad) until sacrifice. In some experiments, 13– 16 weeks old male apoE−/<sup>−</sup> and WT mice fed a high fat diet were treated daily with ezetimibe (5 mg/kg; Kemprotec) or corn oil (vehicle) via oral gavage corn oil for 12 weeks prior sacrifice as previously described (34). All mice were housed under specific pathogen free conditions with unrestricted access to food and water in the animal housing unit of the National University of Singapore. All studies were performed under protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the National University of Singapore.

#### Adoptive Cell Transfers

For adoptive transfers, CD45.2 recipient mice were intravenously injected with 2 × 10<sup>7</sup> of CD45.1 spleen and LN cells. In short term homing experiments to assess lymphocyte entry into skin draining LNs (axillary and brachial), CD45.1, or CD45.2 cells were adoptively transferred into CD45.2 or CD45.1 recipient mice, respectively and allowed to equilibrate for 2.5 h before sacrifice. At this time point, entry from the peripheral blood into skin draining LNs directly determined the CD45.1 cell numbers within recipient LNs (23, 35). In experiments where donor lymphocytes from CD45.1 WT and CD45.2 apoE−/<sup>−</sup> mice were co-transferred, lymphocytes were labeled with carboxyfluorescein succinimidyl ester (CFSE). To assess lymphocyte egress from skin draining LN (axillary and brachial), CD45.1, or CD45.2 cells were adoptively transferred into CD45.2 or CD45.1 recipient mice, respectively and allowed to equilibrate for 24 h. Following equilibrium, half of the mice were sacrificed (T0) while the other half of the mice were treated with anti-CD62L antibody and sacrificed 20 h after administration (T20). Intraperitoneal administration of anti-CD62L antibody at a dose of 100 µg per mouse (clone Mel-14 hybridoma from American Type Culture Collection) blocked further entry of circulating lymphocytes into LNs without affecting lymphocyte egress from LNs (8, 10). Therefore, the population of transferred lymphocytes remaining within skin draining LNs at T<sup>20</sup> compared to T<sup>0</sup> will be a measure of lymphocyte egress from LNs (23).

### Spleen, LN, and Lymph Cell Suspensions

Cell suspensions from skin draining LNs and spleens were prepared by mechanical disruption. For quantification of LECs by flow cytometry, LNs were first digested with 4 mg/mL collagenase IV (Roche) in calcium and magnesium free HBSS at 37◦C for 45 min with gentle agitation. After 45 min, EDTA was added to the digestion mixture to a working concentration of 10 mM and the lymph nodes were further digested for 5 min at 37◦C in a final digestion step. For lymph cell collection, mice were fasted overnight prior sacrifice and lymph was collected as described by Matloubian et al. (36). The peritoneal cavity of anesthetized mice was exposed ventrally and the cysterna chili identified. Lymph was subsequently drawn from the cysterna chili with the use of extended length gel-loading pipette tips (Neptune Scientific).

### Flow Cytometry Analysis

Flow cytometric analysis of skin draining LN (brachial and axillary) cell suspensions stained for CD45, podoplanin and CD31 allowed differentiation and quantification of LECs (CD45−, CD31+, podoplanin+) (23). Antibodies used included the following: rat anti-mouse CD31 (Serotec) detected with anti-rat-APC, hamster anti-mouse podoplanin (clone 8.1.1; Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA) detected with anti-hamster PE and PerCPcy5.5 conjugated anti-mouse CD45.2 (BD Biosciences). FACS analysis was also employed to quantify congenic transferred lymphocytes T and B cell populations in LNs, and lymph. Antibodies used included the following: FITC or PerCPCy5.5—or Pacific Blue conjugated anti-B220, APC-conjugated anti-CD3e, PerCPCy5.5-conjugated anti-CD45.2, biotin–conjugated anti-CD45.1 revealed with Streptavidin-PE. Live and dead cells were discriminated during flow cytometry using a LIVE/DEAD <sup>R</sup> Fixable Aqua Dead Cell Stain Kit (Molecular Probes, Invitrogen). Cell counts were determined during flow cytometry using Count Bright <sup>R</sup> Absolute Counting Beads (Molecular Probes, Invitrogen). FACS analysis was performed using a CyAn ADP Analyzer (Beckman Coulter) and data were analyzed with Flowjo software (Treestar).

#### Immunohistochemistry

Skin draining LNs (axillary and brachial) were either freshly embedded in tissue freezing medium or fixed overnight in 2% paraformaldehyde/30% sucrose solution at 4◦C and embedded in tissue freezing medium. 6–8µm thick cryostat sections were cut for imaging by the fluorescence microscope. Primary antibodies used included biotinylated or purified anti-B220 (eBiosciences), anti-TCRβ (BD Biosciences), anti-LYVE-1 (Upstate), anti-CD31 (Serotec), anti-Ki67 (Dako), anti-CCL21 (R&D Systems), anti-collagen type IV (Cosmo Bio), FITC-conjugated anti-CD169 (Serotec), biotinylated anti-CD45.1 (eBiosciences) antibodies. Secondary antibodies used included Dylight647-conjugated streptavidin, Cy2 or Cy3 conjugated anti-rat IgG, Dylight647 or Dylight549-conjugated anti-armenian hamster IgG, Dylight 647-conjugated anti-goat IgG and Cy2, Cy3, or Dylight647-conjugated anti-rabbit IgG antibodies (Jackson Immunoresearch). Endogenous avidin and biotin were quenched using the Avidin/Biotin blocking kit (Vector Laboratories).

### Microscopy, Image Analysis, and Lumen Area Measurements

To anatomically locate the lymphatic sinuses in the paracortex and medulla of the skin draining LNs images of LN sections stained for LYVE-1 and B220 were captured with a fluorescence microscope (Axio imager.Z1, Axiocam HRM camera; Carl Zeiss Micro Imaging, Inc., Jena, Germany). Lymphatics were identified in cortical or medullary sinuses if they were found in the LN paracortex or medulla and to contain B cells within their lumen (23). To determine lumen area of lymphatic sinuses, images were acquired from 6 representative LN sections per mouse. Within each section, the area of 15 lymphatic vessels in the cortex region and 10 lymphatic vessels in the medulla region were quantified. Therefore, a total of 90 cortical and 60 medullary sinuses were sampled per mouse. Lumen area was obtained by using the "measure" tool of the Axiovision software (version 4.8; Carl Zeiss Micro Imaging, Inc.).

### S1P Measurement by Liquid Chromatography-Mass Spectrometry

The lymph samples were centrifuged at 14,000 g for 10 min at 4◦C and the supernatant was obtained for subsequent S1P extraction. Five microliter of lymph was mixed with 5 µl ISTD and 90 µl methanol. The samples were sonicated for 30 min at room temperature and centrifuged at 14,000 g for 10 min. The supernatant was recovered and 10 µl of TMS-Diazomethane (2M in hexane) was added. The sample incubated for 20 min at room temperature under gentle mixing at 750 rpm. The reaction was stopped by adding 1 µl of acetic acid. The derivatized samples were dried in speedvac and reconstituted in 100 µl of mobile phase before injecting 1 µl of sample into the LC-MS system. All solvents for LC-MS analysis were LC-MS grade and were purchased from Fisher Scientific and Merck Millipore. Lipid standards: isotope labeled standard D-erythro-Sphingosine-1-phosphate (13C2D2–S1P,) were purchased from Toronto Research Chemicals. All LC-MS/MS experiments were performed using Agilent 1200 series HPLC-Chip systems connected to the Agilent 6490 QQQ mass spectrometer as described in our previous report (37). A customized HILIC-chip containing Amide-80 stationary phase (Tosoh Bioscience, LLC. Montgomeryville, PA, 5 Nm particle size, 80 Å pore size) was used for the chromatographic separation, including a 160 nl trapping column and a 75 Nm × 150 mm analytical column (Agilent Technologies Corp., Santa Clara CA). Solvents used for HILIC HPLC: 50% acetonitrile in water containing 25 mM ammonium formate pH 4.6 (solvent A), 95% acetonitrile containing 25 mM ammonium formate pH 4.6. The pH value was adjusted with formic acid. Analytes were eluted with the following gradient: 100% B from 0 to 1.5 min, 40% B from 1.5 to 8.5 min, 30% B from 8.5 to 10.5 min, 0% B from 11.5 to 13.0 min, 100% B from 13.1 to 19 min. The chip cube was operated with back flush mode and samples were injected through the enrichment column at 4 µl/min. The valve was switched 1.5 min after injection to place the enrichment column in line with the analytical column at a flow rate of 400 nl/min. Murine lymph samples were spiked with known amounts of internal standard (S1P-13C2D2). The Agilent 6490 triple quadrupole (QQQ) mass spectrometer was operated in positive mode for MRM where the CID fragment at m/z 60 was used as a "quantifier" and m/z 113 was used as a "qualifier." These ions were present after fragmentation of all species. Quantification was performed according to the internal standard method, comparing peak areas of the endogenous S1P extracted with MassHunter Quant (Agilent) to the ISTD peak.

### CCL21 ELISA

Skin draining LNs were harvested and homogenized in lysis buffer (RIPA buffer, Sigma Chemicals) with a protease inhibitor cocktail (Roche Diagnostics). Homogenates were centrifuged for 10 min at 4◦C at 14,000 g, and supernatants were assayed using commercial CCL21 (R&D Systems) ELISA kits as per manufacturer's protocols.

### Quantitative RT-PCR

Total RNA from skin draining LNs was homogenized and extracted using TRIZOL <sup>R</sup> reagent (Invitrogen) and NucleoSpin <sup>R</sup> RNA II kit (Macherey-Nagel). First strand cDNA was synthesized using TaqMan <sup>R</sup> Reverse Transcription Reagents (Applied Biosystems). Real-time PCR were performed using iTaqTM SYBR <sup>R</sup> Green supermix with ROX (Biorad) on a 7500 Real-Time PCR System (Applied Biosystems). Expression of genes of interest expression was normalized to the expression of GAPDH. The following primers were used: Sphk1, forward 5′ -AACTTGACTGTCCATACCTGGTTC-3 ′ and reverse 5′ -CACATACCATCAGCTCTCCATCC-3′ ; Sgpl1, forward, 5′ -CCTGTTGGGCCGCCTTGATGC-3′ and reverse 5′ -AAATTCCACCCCTTAGC-3′ .

### Statistical Analysis

Statistical analysis was performed with Prism 5 (Graph-Pad Software, Inc.). All values were expressed as the mean of n samples ± SD. Statistical significances were determined using the unpaired two-tailed t-test. Whenever more than two groups were compared, the one-way ANOVA test with Bonferroni's post-test was applied. For all tests, a p < 0.05 was considered significant.

## RESULTS

### Lymphocytes Accumulate in Hypertrophic LN From apoE−/<sup>−</sup> Mice

Consistent with our previous report (33), the substantial increase in skin draining LN cellularity was evident in 22 to 28 weeks old apoE−/<sup>−</sup> mice fed a diet rich in fat and cholesterol compared to age-matched WT mice but not in 6 week old apoE−/<sup>−</sup> mice (**Figures 1A,B**). LN hypertrophy was also observed in Ldlr−/<sup>−</sup> mice, another hypercholesterolemic mouse model (**Supplemental Figure 1**). Flow cytometry analysis revealed T and B cells accumulation in the enlarged LNs of apoe−/<sup>−</sup> mice (**Figure 1C**) with a proportionally greater increase in B cells compared to T cells (**Figure 1D**). The number of CD4<sup>+</sup> and CD8<sup>+</sup> T cells increased in hypertrophic LN of Apoe−/<sup>−</sup> mice but the ratio of CD4+/CD8<sup>+</sup> T cells was similar to WT mice (data not shown).

#### Lymphocyte Proliferation and Entry Within Expanded LNs From apoE−/<sup>−</sup> Mice Are Not Altered

Accumulation of lymphocytes in an activated LN may result from increased proliferation within LN, increased entry of lymphocytes from the blood into the LN or, conversely, a decreased efferent lymphatic emigration from the enlarged LN. Therefore, we sought to determine which of these possibilities could account for LN hypertrophy in apoE−/<sup>−</sup> mice. We excluded the possibility of lymphocyte proliferation contributing to the LN hypertrophy in apoE−/<sup>−</sup> mice since we did not detect any obvious differences for proliferative marker Ki-67 in apoE−/<sup>−</sup> and WT LN sections co-stained with TCRβ or B220 to detect T cells and B cells, respectively (**Figure 2A**). To determine whether lymphocyte trafficking into the enlarged LNs of apoE−/−mice is affected, we quantitated lymphocyte entry into LNs in short-term homing assays (23). 2.5 hours after adoptive transfer of CD45.1 WT lymphocytes into CD45.2 apoE−/<sup>−</sup> or WT mice, CD45.1-transferred T cells accumulated similarly in the LN of WT and apoE−/<sup>−</sup> recipient mice (**Figure 2B**) whereas the number of transferred B cells was increased in apoE−/<sup>−</sup> LN compared to WT LN. Despite subtle differences in T and B cell trafficking, the overall entry of CD45.1-transferred lymphocytes into apoE−/<sup>−</sup> and WT LN was similar (**Figure 2B**), indicating that lymphocyte entry into the hypertrophic LN of Apoe−/<sup>−</sup> mice was not augmented. To substantiate these findings, we performed reverse adoptive transfer experiments. Lymphocytes, T or B cells isolated from CD45.1 WT and CD45.2 apoE−/<sup>−</sup> mice were labeled with CFSE and co-transferred into CD45.1 WT recipient mice. Unexpectedly, we found that the entry of apoE−/<sup>−</sup> lymphocytes into WT LN was decreased compared to WT lymphocytes (**Figure 2C**). Further analysis revealed that apoE−/<sup>−</sup> T cell entry was decreased whereas input of apoE−/<sup>−</sup> B cells into WT LN was comparable to WT B cells. Taken together, these data indicate that lymphocyte accumulation in enlarged LN

of apoE−/<sup>−</sup> mice is not due to increased lymphocyte entry into the LNs.

### The Efferent Lymphatic Emigration of Lymphocytes From LN Is Markedly Blocked in apoE−/<sup>−</sup> Mice

Since we rule out the involvement of lymphocyte entry and proliferation to impact lymphocyte accumulation in LNs of Apoe−/<sup>−</sup> mice, we hypothesized that the increased number of lymphocytes in skin draining LNs from apoE−/<sup>−</sup> mice could result from impaired egress of these cells from the enlarged LN via efferent lymphatic vessels. In support of this hypothesis, we found a marked decrease in the number of total lymphocytes, T and B cells in efferent lymph from apoE−/<sup>−</sup> mice collected at the cysterna chyli (36, 38) compared to WT efferent lymph (**Figure 3A**). This prompted us to assess the exit of lymphocytes from LN using long-term adoptive transfer assays (23, 36). We first evaluated the capacity of CD45.1 WT donor lymphocytes to egress from CD45.2 WT and apoE−/<sup>−</sup> LN mice. The data were expressed as mean fraction of egressed cells—defined by dividing the mean CD45.1 T cell that have exited (T20) by the mean population of CD45.1 T cells present at baseline (T0) (23). This experiment revealed that the egress of WT transferred lymphocytes from hypertrophic apoE−/<sup>−</sup> recipient LN was severely abrogated compared to WT recipient LN (**Figure 3B**). This phenomenon was observed for both T and B cell egress. Next, we compared the egress of CD45.2 apoE−/<sup>−</sup> lymphocytes from CD45.1 WT recipient LN with the egress capacity of CD45.2 WT lymphocytes. No significant difference was observed between the egress index of WT and apoE−/<sup>−</sup> lymphocyte, T or B cells (**Figure 3C**). Considering that no other alternative possibilities could account for the greatly increased numbers of lymphocyte in the hypertrophic LN from apoE−/<sup>−</sup> mice, we conclude that increased lymphocytes accumulation in expanded skin draining LN occurs because less lymphocytes emigrate from those LNs into the efferent lymph.

### Lymphangiogenesis Is Induced in Hypertrophic LNs

Our long-term adoptive transfer assays indicate that the impaired egress of lymphocytes from apoE−/<sup>−</sup> hypertrophic LN was not due to intrinsic defects in lymphocytes but rather to environmental changes in the enlarged LN. We hypothesized that the hypercholesterolemic environment in apoE−/<sup>−</sup> mice induces LN remodeling compromising the egress of lymphocytes, which in turn promotes subsequent LN hypertrophy. As the lymphatic vessels are critical routes for lymphocyte egress from LNs and inflammation has been shown to induce lymphangiogenesis in activated LN such as upon immunization (23, 38), we examined LN lymphangiogenesis in apoE−/<sup>−</sup> mice. We employed flow cytometry to quantify LECs in LNs (**Figure 4A**). Hypercholesterolemia in apoE−/<sup>−</sup> mice stimulated a significant increase in LEC numbers over WT LN (**Figure 4B**). This marked expansion of lymphatic network in hypertrophic skin draining LNs from apoE−/<sup>−</sup> mice was also apparent in LN sections immunostained for LYVE-1 which identifies lymphatic sinuses in the medulla and cortex (**Figure 4C**). Moreover, immunostaining of LN sections for proliferative marker Ki67 revealed that the lymphatic expansion resulted from proliferation of pre-existing lymphatics (**Figure 4D**).

### Expanded Cortical and Medullary Sinuses Are Overly Dilated in Hypertrophic LN

We next investigated whether the extensive expansion of cortical and medullary sinuses observed in apoE−/<sup>−</sup> LN was accompanied by any morphological changes in the sinuses. A close-up examination using LYVE-1 revealed that majority of the cortical and medullary sinuses in hypertrophic LN from apoE−/<sup>−</sup> mice were notably dilated with open lumens compared to those in WT LN with more partially collapsed state (**Figure 5A**). This was further supported by the marked increase in lumen area of cortical and medullary sinuses in apoE−/<sup>−</sup> mice LN compared to WT control LN (**Figure 5B**). Moreover, the dilated sinuses were more irregular and tortuous with occasional focal absence of LYVE-1 expression (**Figure 5C** ). These morphological changes resemble those described in certain tumor-associated vessels having regions lacking endothelial cell markers and termed "mosaic vessels" (39). Additional co-staining of LN sections with LYVE-1 and collagen type IV or CD31 that identify the basement membrane and endothelial cells of sinuses, respectively, confirmed the focal absence of LYVE-1 expression in the cortical and medullary sinuses of apoE−/<sup>−</sup> mice compared to WT mice although collagen type IV (**Figure 5D**) and CD31 (**Figure 5E**) remained intact in those regions. These data indicated extensive structural changes in the LN cortical and medullary sinuses from dyslipidemic mice.

### The Balance Between Retention and Egress Signals Is Altered in Hypetrophic LNs

We next investigated whether these structural alterations may compromise the entry of lymphocytes into the sinuses. To address this question, we stained LN sections from long-term adoptive transfer experiments for CD45.1 at T<sup>0</sup> and T20h to identify the WT donor lymphocytes in LNs from WT and apoE−/<sup>−</sup> mice. While we failed to capture transferred lymphocytes within collapsed sinuses in WT LN (data not shown), numerous transferred lymphocytes were detected within the dilated sinuses in apoE−/<sup>−</sup> LN (**Figure 6A**). Thus, although the lymphocytes were able to migrate into the exit structures of enlarged LN they failed to be captured by the lymph flow and transported to the efferent lymph (**Figures 3A,B**). This may suggest an imbalance between CCL21-mediated retention and S1P-mediated exit signals in the enlarged LN. Measurement of total CCL21 levels by ELISA that reflects the intracellular or secreted (inactive) CCL21 and the extracellular or gradientforming (active) CCL21 revealed a significant increase in this chemokine in LN from apoE−/<sup>−</sup> mice (**Figure 6B**). Microscopic examination of CCL21 distribution in LN from apoE−/<sup>−</sup> mice by immunostaining revealed lower levels of CCL21 associated with non-endothelial cells which might reflect its continuous secretion but higher levels of active CCL21 bound to the basement

membrane of the dilated cortical and medullary sinuses but also of LYVE-1<sup>−</sup> vessels (**Figure 6C**). Together, this finding suggest that CCL21 is more actively released in LN of apoE−/<sup>−</sup> mice compared to WT mice. Moreover, quantification of S1P in lymph from WT and apoE−/<sup>−</sup> mice by mass spectrometry revealed a significant decrease in lymph S1P levels in dyslipidemic mice (**Figure 6D**). This reduction in lymph S1P in apoE−/<sup>−</sup> mice likely resulted from decreased production in LN as supported by reduced expression of LN Sphk1 but also from increased S1P degradation by S1P lyase whose expression was markedly increased in apoE−/<sup>−</sup> mice LN (**Figure 6E**). Together these data indicate that the decreased egress of lymphocyte from hypertrophic LN observed in apoe−/<sup>−</sup> mice is due to reduced S1P exit signal and conversely, increased CCL21 retention signal.

### Reversing Dyslipidemia in apoE−/<sup>−</sup> Restores Lymphocyte Egress

Next, we tested whether reduction of circulating cholesterol levels in apoE−/<sup>−</sup> mice could restore lymphocyte egress. Consistent with our and others work (34, 40, 41) daily gavage of ezetimibe, a FDA approved cholesterol-lowering drug, reduced total plasma cholesterol levels in apoE−/<sup>−</sup> mice but did not affect plasma cholesterol levels in WT mice (**Table 1**). Following ezetimibe treatment, LN cellularity was significantly reduced in apoE−/<sup>−</sup> mice compared to untreated apoE−/<sup>−</sup> mice whereas LN cellularity in WT mice was not affected (**Figure 7A**). In ezetimibe treated apoE−/<sup>−</sup> mice, lymphangiogenesis was reduced (**Figure 7B**) and the number of LECs was reversed to WT LEC number (**Figure 7C**). Since ezetimibe did not have any effect in WT mice, vehicle-treated WT mice were used as control for comparison with vehicle-treated apoE−/<sup>−</sup> mice in the subsequent experiments. Ezetimibe treatment also affected the dilation of lymphatic sinuses in apoE−/<sup>−</sup> mice. Indeed, the lumen area of cortical and medullary sinuses was not equivalent to WT mice but was significantly improved over untreated apoE−/<sup>−</sup> mice (**Figure 7D**). Furthermore, the expression of Sgpl1 in LN from ezetimibe treated apoE−/<sup>−</sup> mice was improved to WT expression whereas the expression of Sphk1 was significantly increased compared to untreated apoE−/<sup>−</sup> mice (**Figure 7E**). Importantly, these changes in ezetimibe treated apoE−/<sup>−</sup> mice resulted in the restoration of lymph S1P levels comparable to WT controls (**Figure 7F**). In contrast, the expression of CCL21 surrounding the cortical and medullary in LN from ezetimibe treated apoE−/<sup>−</sup> mice was not significantly different compared to untreated apoE−/<sup>−</sup> mice (data not shown). Nevertheless, the amelioration of cortical and medullary sinuses dilatation and the reversal of LEC numbers and lymph S1P levels in ezetimibe treated apoE−/<sup>−</sup> mice were sufficient to override CCL21 retention signal since lymphocyte egress from LN was partially restored in these mice, increasing lymphocyte egress index by 50% (**Figure 7G**). This effect was mainly due to the improvement of T cell egress but not B cell.

#### DISCUSSION

Here, we elucidate the mechanisms whereby hypercholesterolemia in mouse models leads to the accumulation of lymphocytes in skin draining LNs and subsequent LN hypertrophy. We show that the increased number of lymphocytes does not result from increased proliferation or input of lymphocyte within the enlarged LN but from the severe impairment of lymphocytes to emigrate from those LN via efferent lymph. The adoptive transfer experiments used to determine lymphocyte egress also revealed that defects in extrinsic microenvironment surrounding lymphocytes rather than intrinsic defects in lymphocytes account for the impaired lymphocyte egress in apoE−/<sup>−</sup> mice compared to WT mice. Lymphatic vessels are essential for lymphocyte trafficking. We showed previously that skin lymphatic transport is severely compromised in hypercholesterolemic apoE−/<sup>−</sup> and Ldlr−/<sup>−</sup>

mice which also exhibit LN hypertrophy and impaired skin DC migration (34, 42). Therefore, we hypothesized that hypercholesterolemia may alter the lymphatic sinuses of the activated LN.

Consistent with this hypothesis, we found a pronounced expansion of the lymphatic network in LNs from apoE−/<sup>−</sup> mice as denoted by a >20-fold increase in LEC numbers and the expression of proliferative marker Ki67. This expansion affects predominantly the cortical and medullary sinuses. Further microscopic examination revealed structural abnormalities in this expanded sinuses network including tortuous morphology, severe dilation and local loss of LYVE-1 expression. Such "mosaic" vessels have also been observed in various tumors as a result of extensive intra-tumoral lymphangiogenesis (39, 43) and have been shown to be "leaky" and as a consequence poor lymph carrier (43). Thus, it is possible that these structural abnormalities in the overly expanded sinuses of apoE−/<sup>−</sup> LN may lead to

indicate loss of LYVE-1 expression.

FIGURE 6 | The balance between retention and exit signals is altered in apoE−/<sup>−</sup> mice LNs. (A) Immunoreactivity for LYVE-1, B220, and CD45.1 was examined on LN sections from long-term adoptive transfer at t = 0 and t = 20 h to determine the localization of CD45.1 transferred lymphocytes in enlarged LNs from apoE−/<sup>−</sup> mice. Images are representative of three to four independent experiments (n = 3–4 mice per group) (B) CCL21 protein content was analyzed in homogenates of 22 to 28 weeks-old WT and apoE−/<sup>−</sup> mice LNs by ELISA. (C) Immunoreactivity for CCL21 and LYVE-1 was examined in LN sections from 22 to 28 weeks-old WT and apoE−/<sup>−</sup> mice. Images are representative of three to four independent experiments (n = 3–5 mice per group) (D) S1P content was analyzed in efferent lymph of 22 to 28 weeks-old WT and apoE−/<sup>−</sup> mice by mass spectrometry. (E) Quantitative RT-PCR analysis was performed on whole LN RNA to examine expression of Sph1k and Spgl1 in WT and apoE−/<sup>−</sup> mice at 22 to 28 weeks of age. n = 6–10 mice per group; \*p < 0.05.


Values are mean ± SEM. \*\*\*p < 0.0001.

"leaky" sinuses and disturb efferent lymph flow which is critical for optimal lymphocyte transport from the LN (44). Altered efferent flow may also diminish the transport of lymphocytes within efferent lymph which is consistent with the poor number of lymphocyte collected in efferent lymph from apoE−/<sup>−</sup> mice and conversely, increase the opportunity for CCL21 to attract lymphocytes back to the LN parenchyma. Notably, numerous adoptively transferred lymphocytes in apoE−/<sup>−</sup> mice were found within and surrounding cortical and medullary sinuses where

(D) Lumen area of cortical and medullary sinuses was determined on LN sections NT WT, NT, and T apoE−/<sup>−</sup> mice. n = 4 mice per group; \*\*\*p < 0.0005. (E) Quantitative RT-PCR analysis was performed on whole LN RNA to examine expression of Sph1k and Spgl1 in NT WT, NT, and T apoE−/<sup>−</sup> mice. Data is pooled (Continued) FIGURE 7 | from three independent experiments with 4 mice per group in each experiment; \*p < 0.05; \*\*p < 0.005; \*\*\*p < 0.0005. (F) S1P content was analyzed in efferent lymph of NT WT, NT and T apoE−/<sup>−</sup> mice by mass spectrometry. Data is pooled from three independent experiments with 3–4 mice per group in each experiment (G) Egress of adoptively transferred CD45.1 lymphocytes, T and B cell from NT WT, NT and T apoE−/<sup>−</sup> mice was examined. n = 4–6 mice per group; \*p < 0.05; \*\*p < 0.005; \*\*\*p < 0.0005.

CCL21 was more abundant compared to WT mice. Although the measure of lymph flow will require further investigations which are beyond the scope of this study, our previous data on skin lymphatics in apoE−/<sup>−</sup> mice showing that they are leaky and exhibit abnormal valve (34) support the idea of altered efferent lymph flow in these hypercholesterolemic mice. The extensive remodeling of lymphatic sinuses observed in apoE−/<sup>−</sup> mice was also associated with a significant decrease in lymph S1P levels. This decrease may result from a reduced local production of S1P by LECs as supported by lower expression of S1PK but also from an increased degradation by S1P lyase whose expression was significantly higher in apoE−/<sup>−</sup> LN compared to WT LN. However, we cannot exclude the possible contribution of systemic abnormalities in sphingolipids as shown in a lipidomic study by Chen et al., reporting elevated plasma levels of sphingolipids and abnormal sphingolipid metabolism in apoE−/<sup>−</sup> mice (45).

Finally, we provide evidence that hypercholesterolemia accounts in part for the impaired lymphocyte egress, extensive remodeling of lymphatic sinuses and decrease in lymph S1P levels as these defects were restored by reversing hypercholesterolemia with the lowering-cholesterol drug ezetimibe. This is in line with our previous report demonstrating that ezetimibe improves lymphatic drainage and skin dendritic cell migration in both apoE−/<sup>−</sup> and Ldlr−/<sup>−</sup> mouse models (34). By analogy to arterial alterations induced by hypercholesterolemia, the accumulation of lipoprotein including LDL and oxidized LDL may compromise the structural integrity of lymphatic vessels by for example altering the expression of tight junction assembly (46) or may induce lymphatic dysfunction by affecting smooth muscle contraction of efferent lymphatic vessels through the modulation of nitric oxide levels (47). However, hypercholesterolemia in apoE−/<sup>−</sup> mouse is often associated with systemic inflammation including in skin (26, 27, 33). Therefore, inflammation that can affect lymphatic vessel structure and function may also account for lymphatic alterations and impaired lymphocyte trafficking. Although ezetimibe has been shown to reduce atherosclerosis progression mainly through its effect on plasma LDL cholesterol by inhibiting its absorption by the intestine (40, 41), it is possible that ezetimibe may indirectly affect inflammation associated with hypercholesterolemia in apoE−/<sup>−</sup> mice by reducing LDL cholesterol and subsequently inflammatory modified LDL. Therefore, it would be interesting to investigate whether anti-inflammatory strategies alone without affecting hypercholesterolemia would be sufficient to improve lymphocyte egress and lymphatic function in apoE−/<sup>−</sup> mice. That the improvement of lymphocyte egress in apoE−/<sup>−</sup> mice by ezetimibe treatment resulted from the improved T cell egress but not that of B cell suggests that the egress of B cells may rely on additional signals than S1P. In fact, most of the knowledge on lymphocyte egress is based on studies on T cell egress.

It is now apparent that the expansion of lymphatic vessels in LN can modulate the immune response during inflammation (38, 48, 49). All of these studies however focused on lymphangiogenesis occurring at early phases of inflammation. More recently, we reported a biphasic remodeling of lymphatic vessels during the course of inflammation, the subcapsular sinuses being expanded first followed by the cortical and medullary sinuses (23). Notably, this differential remodeling is biologically and functionally important. Indeed, the early expansion of subcapsular sinuses enhances DC migration from the periphery into the inflamed LN (38) whereas the expansion of cortical and medullary sinuses at later phases of inflammation reestablishes the steady-state egress of lymphocytes from those LNs (23). Therefore, the increased accumulation of lymphocytes into LNs during inflammation has to be accompanied by a proportional increase in lymphocyte exit in the efferent lymph to prevent LN hypertrophy. Here, we provide evidence that under certain circumstances such as chronic inflammation associated with hypercholesterolemia, the aberrant expansion of cortical and medullary sinuses can conversely inhibit the emigration of lymphocyte from the inflamed LNs leading to subsequent LN hypertrophy.

Activated antigen-specific lymphocytes must leave the LN to migrate into the effector sites in order to exert their appropriate immune responses (9). Naïve lymphocyte must also exit the LN via efferent lymph to reach the blood circulation for immune surveillance. Therefore, the impairment of lymphocyte egress is expected to affect the normal function of both naïve and effector lymphocytes and subsequently, compromised immune responses. This scenario is likely relevant to hypercholesterolemic mouse models. Indeed, apoE−/<sup>−</sup> and Ldlr−/<sup>−</sup> mice show impaired priming when immunologically challenged (33) and reduced capacity for clearance of bacteria (29, 30, 50), fungi (28), and virus (31). In addition, this increased susceptibility to infection does not occur in young apoE−/<sup>−</sup> mice, but increases with age (31, 51), consistent with the later onset of LN hypertrophy and decreased lymphocyte egress. Thus, the increased susceptibility to infection in hypercholesterolemic mice may not only result from impaired DC migration from peripheral tissue into the draining LN as we proposed previously (33) but also from the impairment in efferent lymphatic emigration of lymphocytes from the enlarged LN. This is supported by the study of Ludewig et al. (31) showing that apoE−/<sup>−</sup> mice infected with LCMV exhibit impaired migration of virus specific CD8 cytotoxic T cells into the blood and liver.

In conclusion, we provide evidence that dyslipidemia severely compromises the efferent lymphatic emigration of lymphocyte from LN which subsequently leads to LN hypertrophy by altering the structure and function of lymphatic vessels. Since LN hypertrophy is often associated with chronic inflammatory and autoimmune diseases such as systemic lupus erythematosus and rheumatoid arthritis, our findings may also be relevant to other chronic diseases than those related to dyslipidemia. This study further illustrates the importance of maintaining healthy lymphatic vessel for optimal immunity.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and the **Supplementary Files**.

#### AUTHOR CONTRIBUTIONS

SL, MT, YL, KT, FT, PN, MW, CT, and VA designed the study and developed the methodology. SL, MT, YL, KT, FT, PN, and CT performed and analyzed the experiments and VA wrote the manuscript.

#### REFERENCES


#### FUNDING

This work was supported by grants from National Medical Research Council, Ministry of education and by the National Research Foundation, Prime Minister's Office, Singapore under its Campus for Research Excellence and Technological Enterprise (CREATE) programme to VA.

#### ACKNOWLEDGMENTS

We thank Hwee Ying Lim for critical reading of the manuscript. We thank Paul Edward Hutchinson (Core Flow Cytometry Lab, Life Science Institute, National University of Singapore) and Esther Koh (Advanced Imaging lab, Life Science Institute, National University of Singapore) for sharing their expertise.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00575/full#supplementary-material


enhances dendritic cell mobilization. Immunity. (2006) 24:203–15. doi: 10.1016/j.immuni.2006.01.003


**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 © 2019 Tay, Lim, Leong, Thiam, Tan, Torta, Narayanaswamy, Wenk and Angeli. 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(s) 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.

# Tumor-Associated Lymphatic Vessel Features and Immunomodulatory Functions

#### Laure Garnier\*, Anastasia-Olga Gkountidi and Stephanie Hugues\*

*Department of Pathology and Immunology, Faculty of Medicine, University of Geneva, Geneva, Switzerland*

#### Edited by:

*Amanda W. Lund, Oregon Health & Science University, United States*

#### Reviewed by:

*Traci Renae Lyons, University of Colorado Anschutz Medical Campus, United States Mihaela Skobe, Icahn School of Medicine at Mount Sinai, United States*

\*Correspondence:

*Laure Garnier laure.garnier@unige.ch Stephanie Hugues stephanie.hugues@unige.ch*

#### Specialty section:

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

Received: *19 December 2018* Accepted: *18 March 2019* Published: *09 April 2019*

#### Citation:

*Garnier L, Gkountidi A-O and Hugues S (2019) Tumor-Associated Lymphatic Vessel Features and Immunomodulatory Functions. Front. Immunol. 10:720. doi: 10.3389/fimmu.2019.00720* The lymphatic system comprises a network of lymphoid tissues and vessels that drains the extracellular compartment of most tissues. During tumor development, lymphatic endothelial cells (LECs) substantially expand in response to VEGFR-3 engagement by VEGF-C produced in the tumor microenvironment, a process known as tumor-associated lymphangiogenesis. Lymphatic drainage from the tumor to the draining lymph nodes consequently increases, powering interstitial flow in the tumor stroma. The ability of a tumor to induce and activate lymphatic growth has been positively correlated with metastasis. Much effort has been made to identify genes responsible for tumor-associated lymphangiogenesis. Inhibition of lymphangiogenesis with soluble VEGFR-3 or with specific monoclonal antibodies decreases tumor spread to LNs in rodent models. Importantly, tumor-associated lymphatics do not only operate as tumor cell transporters but also play critical roles in anti-tumor immunity. Therefore, metastatic as well as primary tumor progression can be affected by manipulating tumor-associated lymphatic remodeling or function. Here, we review and discuss our current knowledge on the contribution of LECs immersed in the tumor microenvironment as immunoregulators, as well as a possible functional remodeling of LECs subsets depending on the organ microenvironment.

Keywords: lymphatic vessels, anti-tumor immune response, metastasis, lymphangiogenesis, tumor microenvironment

#### INTRODUCTION

Over the last few years, immunotherapy has evolved into a very promising new approach for fighting tumor progression. However, the proportion of cancer patients that positively respond to these treatments is still limited. Indeed, tumor cells foster mechanisms to escape immunosurveillance either by inducing poorly immunogenic tumors (immunoselection) or by setting up a tolerogenic environment that inhibits immune effector cells (immunosubversion) [reviewed in (1–3)]. Therefore, manipulations aiming at boosting anti-tumor immune cell responses and in particular tumor-specific T cell priming currently represent an extensive axis of investigations.

During tumor development, lymphatic endothelial cells (LECs), the principal components of lymphatic vessels (LVs), undergo active modifications that facilitate metastatic dissemination, and induce immunoregulation. LEC phenotype and functions are strongly altered by inflammation or infections, which may directly influence on-going immune responses (4). In particular, it has been suggested that LECs immersed in the tumor microenvironment (TME) can act as immunoregulators of the anti-tumor T cell response (5). In vitro studies have further shown that, tumor derived LECs exhibit altered gene expression profiles compared to dermal derived LECs (6) and upregulate PD-L1 to inhibit T cell activation (7, 8). On the other hand, a recent study has suggested that tumor-associated (TA) LVs might be beneficial for the efficacy of anti-PD-1 immunotherapy (9). Therefore, depending on the stage of tumor progression and on the immunological settings (immune evasion/immunosubversion or immunotherapy), LV might display positive and/or negative effects on tumor immunity. It is thus urgent to decipher precisely the roles for LVs in tumor cell dissemination and anti-tumor T cell immunity. In this review, we discuss the ability of LECs to shape tumor development through their contribution to tumor cell spreading and regulation of anti-tumoral T cell responses.

### LYMPHATIC VESSELS AS IMMUNOREGULATORS IN NON-TUMOR CONTEXT

LVs develop as a hierarchical vasculature facilitating a unidirectional drainage system of fluid and cells from tissues toward draining lymph nodes (LNs) (10). They interlace the blood vessel circulation and play a crucial role in lipid absorption, tissue fluid homeostasis and immunity (11). The lymphatic system is a linear and blind-ended circuit. Initial lymphatic capillaries are composed of a single layer of LECs with minimal basement membrane and are not covered by pericytes or smooth muscle cells. This particular organization of LECs is highly permeable for the uptake of cells, macromolecules and interstitial fluids (12). Lymphatic capillaries drain to collecting lymphatics defined by pericyte and smooth muscle cell coverage, continuous basement membrane with "zipper-like" junctions, and a system of valves preventing retrograde flow (12, 13). Our knowledge of multiple LV functions has quickly evolved, based on the identification of LEC markers such as the transcription factor Prox-1 and the surface protein LYVE-1, that are not expressed by others endothelial cells. Prox-1 is primordial for the development and the maintenance of LECs (14–16). LYVE-1 is enriched in lymphatic junctions, highly expressed in initial lymphatics, but mostly absent from LV collectors [reviewed in (12)]. This molecule is implicated in dendritic cells (DCs) trafficking within LVs (17). LECs also express GP38 (podoplanin) and platelet endothelial cell adhesion molecule (PECAM-1 or CD31) that are markers shared with fibroblastic reticular cells (FRCs) and blood endothelial cells (BECs), respectively. An important function of lymphatics is to transport immune cells from peripheral tissues to LNs and therefore to participate to immune response initiation (18–21). Transcriptomic analysis of ex vivo LN stromal cell (LNSC) subsets in distinct immunological situations established that FRCs, BECs, and LECs express a multitude of immune mediators and growth factors that may influence the immune system. LNSCs are strongly modulated by inflammation or infections, and may contribute as active participants of on-going immune responses. In addition, a more precise characterization of these cells within distinct conditions suggested that LNSCs are specialized for their unique microenvironment (4). This might reflect a functional specialization of LNSC subsets depending on the organ microenvironment. Apart from their effect on tissue drainage and immune cell migration, LECs regulate T cell responses through different mechanisms (22). First, different studies in mice showed that steady-state LN LECs participate to peripheral T cell tolerance by presenting endogenously expressed tissue-restricted antigens (17, 18) through MHC class I (MHCI) molecules and eliminating autoreactive CD8<sup>+</sup> T cells (23–25). LN LECs can also cross-present exogenous antigens onto MHCI molecules, and further drive the apoptosis of antigen-specific CD8<sup>+</sup> T cells (26). Whether LN LECs have an impact on peripheral CD4<sup>+</sup> T cell responses in different immunological settings remains largely unknown and controversial. On the one hand, Rouhani et al. showed that LECs were unable to load MHC class II (MHCII) molecules with antigenic peptides due to their lack of H2-M expression at steady-state (27). However, LECs express the promoter IV (pIV) of CIITA, the master regulator for MHCII molecule expression (28). CIITA pIV being inducible by IFN-γ (29), LECs might require exposure to IFN-γ to upregulate H-2M molecules and be capable of MHCII-restricted antigen presentation. On the other hand, we published that surface MHCII expression on LNSCs results from the combination of both endogenous and acquired molecules. In vitro and in vivo, LNSCs further present peptide-MHCII complexes acquired from DCs to CD4<sup>+</sup> T cells to induce their dysfunction. In particular, LECs specifically induce CD4<sup>+</sup> T cell death, whereas LECs, BECs and FRCs all induce T cell anergy (28). Moreover, our recent studies demonstrate that the loss of MHCII expression on LNSCs in murine LNs impairs peripheral CD4<sup>+</sup> T cell tolerance, and alters regulatory T cell populations, resulting in signs of spontaneous autoimmunity in elderly (30). Their lack of costimulatory molecules could explain LNSC implication in T cell tolerance. By releasing the sphingosine 1-phosphate (S1P), LECs play also an important role in the egress of activated T cells from LNs (31, 32). In addition, LEC-derived S1P is involved in naïve T cells survival, its signaling further providing sufficient energy to maintain their steady-state recirculation (33). LECs are also capable of preventing T cell activation and proliferation in a negative regulatory feedback process. Indeed, LECs from LNs produce nitric oxide in response to inflammatory signals (IFN-γ and TNF) produced by T cells, inhibiting back T cell activation (34). Finally, during inflammation, LECs present in collecting LV or in the skin suppress DC maturation via a Mac-1/ICAM-1 dependent mechanism (35), or through prostacyclin synthesis, respectively (36), leading to subsequent dampening of T cell activation.

#### TUMOR-ASSOCIATED LYMPHANGIOGENESIS AND METASTASIS

The mortality linked to solid tumors is mainly associated with their capacity to disseminate to distant organs in a process known as metastasis (37). LVs are essential in tumor cell spreading as they function as "highways" connecting primary tumors to secondary lymphoid organs. The process of lymphatic proliferation, sprouting and enlargement during tumor progression, known as tumor lymphangiogenesis, and its implication in the spread of the disease has been studied for many years. TA-lymphangiogenesis correlates with metastasis and poor prognosis in several cancer types [as depicted in (10)], illustrating the relevance of lymphatic vasculature to cancer biology. A retrospective analysis of melanoma patients with lung metastases showed that high LV density and lymphatic invasion in metastatic regions were associated with poor prognosis (38). Moreover, LVs and immune cell infiltrates positively correlate in human metastatic cutaneous melanoma and colorectal cancer (39, 40). Therefore, therapies aiming at blocking tumor lymphangiogenesis are being considered as promising approaches for the treatment of such malignancies [as discussed in (41)]. Importantly, several inhibitors targeting distinct actors of lymphangiogenesis have been developed in murine tumor models and could be translated into clinic to reduce metastasis. Accordingly, the inhibition of the prolymphangiogenic VEGFR-3 signaling by using VEGFR-3 blocking antibodies or VEGF-C/D trap reduces LN, and/or distant organ metastasis in different tumor mouse models (42–45). Conversely, overexpression of two lymphangiogenic factors VEGF-C and VEGF-D increases metastasis dissemination to sentinel LNs (46–49). In particular, molecules or antibodies blocking VEGF-C/VEGFR3 signaling have been tested in clinical trials, some have gone one to be approved for cancer treatment (10, 50) [as reviewed in (10)]. However, although blockade of VEGFR3 has no noticeable effect on established lymphatics, VEGF-C signaling has been described to promote homeostasis of intestinal and brain meningeal LVs (51, 52). Therefore, it would be crucial to develop treatment that specifically target pro-tumorigenic LEC functions in order to exclude any potential intestinal or neurological side effect.

In mice, LECs from tumors present a distinct molecular profile compared to dermal LECs. Altered pathways include chemokines, extracellular matrix, cell adhesion, and inflammatory responses (6). These observations reflect significant levels of LEC plasticity that is highly regulated by the tissue microenvironment.

### Pro-Lymphangiogenic Factors and Tumor Cell Spreading

The TME is composed of cancer cells, the extracellular matrix (ECM), stromal cells, and various immune cell types, impacting both tumor cell development and anti-tumor immunity. All these cells produce many factors that lead to the establishment of an intratumoral environment characterized by chronic inflammation, immunosuppression, angiogenesis, and lymphangiogenesis, the latter being the focus of this review **(Figure 1.1)**.

The proliferation, migration and survival of LECs depend mainly on VEGFR2/3 signaling axis, which is driven by VEGF-C and VEGF-D (Vascular Endothelial Growth Factors –C and –D) (53, 54) produced by many different cell types, including tumor cells and immune cells. VEGF-C and VEGF-D, considered to be major drivers of tumor lymphangiogenesis, are associated with LN and/or distant organ metastasis (38, 46–48, 55–57) **(Figure 1.1)**. Using orthotopic spontaneous metastasis models in nude mice, it has been shown that VEGF-C expression by tumor cells favors metastatic propagation in distal organs (57). Moreover, a recent study indicated that, in a transgenic mouse model with increased lymphangiogenesis in the lung, TA-LVs contribute to the dissemination of metastases to distant organs (38). In vitro studies have deciphered the molecular mechanisms implicated in the activation of VEGF-C/D signaling pathway. Following VEGFR-3 engagement, the protein kinase C is activated, leading to the phosphorylation of AKT, and subsequent LEC migration, survival and proliferation (53). Neuropilin 2 (Nrp-2), an additional receptor for VEGF-C, is also expressed by LECs, and contributes to lymphatic sprouting (58, 59).

In several human cancer, VEGF-C and COX-2 (cyclooxygenase 2, an enzyme implicated in prostaglandin pathway) expression are associated with LV density and LN metastasis (60– 63). Interestingly, preclinical and clinical trials using different Non-Steroidal Anti-Inflammatory Drugs (NSAIDs), blocking COX-2 and subsequent prostaglandin production, have reported a decrease in cancer incidence, tumor cell dissemination, and finally global cancer morbidity. These observations suggest that NSAIDs could be applied for the treatment of metastasis (64–66). In mice, beside a direct effect on LECs, VEGF-C/D increases the levels of prostaglandins in the TME, further promoting TA-lymphangiogenesis. VEGF-D indeed inhibits the enzyme 15-hydroxyprostaglandin dehydrogenase (15-PGDH), therefore enhancing LEC exposure to prostaglandins in collecting lymphatic vessels (67). The engagement of EP3 signaling (prostaglandin E2 receptor 3) on tumor-associated stromal cells promotes lymphangiogenesis (68). Moreover, increased amounts of prostaglandins amplify the production of VEGF-C by tumor and immune cells, contributing to lymphangiogenesis and tumor cell dissemination (62, 63).

TNF-α interaction with its TNF receptor 1 (TNFR-1) triggers VEGF-C secretion by tumor-associated macrophages (TAM), amplifying LV expansion and metastasis (69). On the other hand, TNF-α signaling in LECs directly favors their proliferation and their migration, without however being sufficient to constitute a fully competent lymphatic network (69). Indeed, TNF-α induced lymphangiogenesis completely depends on the VEGF-C/VEGFR3-induced LEC tip formation. Similarly, VEGF-C/VEGFR3-induced LEC tip formation is required to trigger fibroblast growth factor (FGF2) induced lymphangiogenesis and foster tumor metastasis in mice (70). In contrast, proangiogenic factors such as platelet derived growth factor B (PDGF-BB) (71) and angiopoietins (ANGPTs) (72) can act as direct lymphaniogenic factors by binding, respectively, PDGF-BB

(3). DCs further present tumor-antigens to naïve T cells, leading to initiation of adaptive anti-tumor immunity. In tumor-draining LNs, VEGF-C exposed LEC cross-present tumor-antigens (Ags) and induce the deletion of anti-tumor CD8<sup>+</sup> T cells (4). Intratumorally, naïve and activated T cells are weakly restimulated by local DCs, due to the immunosuppressive TME. The TME favors in particular the infiltration of Treg and naïve T cells through a CCL21-dependent pathway. TA-LECs also express high levels of PD-L1 in response to IFN-γ produced by effector T cells (5). Upon anti-PD-1 immunotherapy, TA-LEC mediated immunosuppression might be abrogated, contributing to enhanced T cell activation and tumor cell elimination (6). Drawing designed by Rémi Jeandenand.

receptors and receptors Tie1/2 expressed by LECs. Whether TGFβ enhances (73, 74) or inhibits (75) lymphangiogenesis depends on tumor models, rendering difficult the targeting of this cytokine for the regulation of LEC remodeling in tumors.

### Mechanism of Metastasis Dissemination

Emerging evidence suggest that lymphatic vessels undergo several changes in response to lymphangiogenic factors during the course of metastasis. In addition to promoting tumoral cell transportation, LVs deliver lymphangiogenic factors produced by the primary tumor to condition sentinel LNs prior to the arrival and seeding of cancer cells (56, 76–78). In the mouse B16F10 melanoma model, LVs from distant metastatic regions, such as LNs and lungs, attract chemoresistant CD133+CXCR4<sup>+</sup> melanoma cells by secreting CXCL12 (79).

TA-LVs were for long described as passive conduits for tumoral cell spread toward sentinel LNs and distant organs. However, several studies highlighted that the tumor microenvironment actively modifies LV features of primary tumor and draining LNs to further promote metastasis. VEGF-C acts in an autocrine manner to improve metastasis dissemination by favoring proteolytic activity and motility of tumor cells (80). Besides its direct effect on tumoral cells, VEGF-C modulates the expression of integrins and chemokines by LECs to facilitate tumor invasiveness. The integrin α4β1 (or VLA-4), which is considered as a marker of activated and proliferating LECs in human and murine tumors (81), is activated by the VEGF-C/PI3Kα pathway in LECs to promote lymphangiogenesis and tumor metastasis in LNs. Therefore, the blockade or the genetic deletion of this integrin on LVs prevents LEC migration and invasion, and inhibits VCAM-1 mediated adhesion of tumoral cell to LECs (82). The secretion of CCL21 by LECs, which drives CCR7-dependent tumor migration through LVs, is also enhanced in response to VEGF-C (80). Moreover, CCL21-dependent recruitment of innate lymphoid cells results in the production of CXCL13 by tumoral stromal cells, which in turn induces metastasis through RANK/RANKL signaling (83). Transmural flow modulates LEC function by promoting the expression of CCL21 and by downregulating VE-cadherin and PECAM-1, two adhesion molecules crucial for cellular junctions (84). Modification of interstitial flow influences CCR7 ligand secretion by tumoral cells, providing an autologous chemotactic gradient (85). In human, CCR7 expression by tumor cells is associated with LN metastasis in several cancers (86–88) **(Figure 1.2)**.

CCL1 secretion by LECs located in the subcapsular sinuses of LNs is crucial to control tumor cell invasion into LNs. Indeed, blocking the CCL1 receptor (CCR8) inhibits metastasis by preventing tumor cell egress from collecting lymphatics into LNs without affecting their entry into intratumoral lymphatics (89). Recently, the screening of 810 mutant mouse strains allowed the identification of 23 genes that, when disrupted, alter the establishment of metastatic foci (90). Notably, they demonstrated that the deletion of the sphingosine-1-phosphate transporter SPNS2 in LECs decreases pulmonary metastasis and promotes effector T cell and natural killer cell infiltration in lungs (90).

Recently, Black et al., have shown that the prolymphangiogenic factor COX-2 enhances the expression of semaphorin 7a (sema7a) in breast tumoral cells. This leads in turn to the activation of β1-integrin receptors on adjacent tumoral cells and LECs, to finally increase lymphangiogenesis and cancer cell dissemination (91). Moreover, sema7a induces gp38 upregulation by tumor-infiltrating macrophages, therefore promoting their adhesion to LVs and consequently boosting lymphangiogenesis and metastasis in breast cancer (92). In agreement, Sema7a gene expression is observed in a high frequency in human breast cancer and correlates with metastasis and poor prognosis (91).

Apart from their implication in metastasis dissemination, accumulating studies indicate that LECs modulate anti-tumor immunity. The roles of LECs in tumor spreading and antitumoral immune responses are discussed below.

#### DUAL ROLE OF TUMOR-ASSOCIATED LYMPHATIC VESSELS IN ANTI-TUMOR IMMUNITY

Growing evidence highlight that, in addition to acting as drains for soluble factors and tumoral cell transport, TA-LVs further play important roles in shaping antitumor immunity. Therefore, the modulation of lymphangiogenesis could impact not only metastasis dissemination but also anti-tumor immunity and primary tumor growth. In the context of solid tumors, lymph flow from tumors is increased, driving intense interstitial flow in the tumor stroma, and enhancing lymphatic drainage to the draining LNs (93). TA-LVs are primarily required for the recruitment of immune cells and adaptive immune response initiation (39, 94). However, immunosuppressive features of LECs in TME will subsequently dampen ongoing anti-tumor immunity (5). Therefore, LVs play a dual role on tumor immunity that might be temporally regulated. Finally, immunotherapy approaches can be potentialized by TA-lymphangiogenesis in melanoma tumors (9), further highlighting the relevance of modulating LV functions during tumor development.

#### Lymphatic Vessels Are Necessary for the Initiation of Anti-Tumoral Responses

T cell activation and infiltration in tumors are key steps of antitumor immunity. Indeed, while Treg infiltration is associated with a poor outcome in patients, intratumoral cytotoxic T lymphocytes are beneficial for clinical outcome (95, 96). Although some studies have suggested that naïve T cell could infiltrate tumors and be locally activated (97–99), antigen transport by dendritic cells (DCs) through LVs toward draining LNs is nevertheless crucial for the initiation of tumorspecific T cell responses, at least in melanomas (39, 100). Indeed, tumor drainage, DC trafficking and subsequent induction of antitumor adaptive immune responses are drastically impaired in transgenic mice lacking or with disturbed local LVs (39, 94). Upon inflammation, LECs in afferent LVs produce CCL21 that is necessary to DC egress from the tissue toward lymphatics (101, 102). Moreover, the expression of CLEC-2 by DCs is essential for their migration into LNs. The activation of CLEC-2 by GP38, which is highly expressed by LECs and FRCs, induces actin polymerization and motility of DCs (103). In a tumoral context, CCR7 expression by DCs is primordial for their migration into tumor draining LNs and subsequent T cell activation (100) **(Figure 1.3)**.

In agreement with a role for lymphatic vasculature in the initiation of anti-tumor immunity, lymphatic vessel density (LVD) or lymphatic gene expression in primary tumors of colorectal or melanomas patients positively correlates with inflammation and immune cell infiltration (9, 39, 40, 104).

#### Lymphatic Vessels Suppress Effector T Cells During Tumor Progression

The lymphangiogenic factor VEGF-C produced in the tumor favors immunological tolerance in murine melanoma, including the induction of tumor-specific CD8<sup>+</sup> T cell deletion (5) **(Figure 1.4)** (5, 26). This is consistent with studies in human melanoma, where active CTLs can be found in the circulation, while they exhibit an exhausted phenotype when localized in tumors (105). In addition, LECs in tumor draining LNs cross-present tumor antigens through MHCI complexes, and further drove the apoptosis of tumor-specific CD8<sup>+</sup> T cells. The expression of the immunosuppressive molecule PD-L1 is enhanced at the surface of LECs after antigen specific interaction with CD8<sup>+</sup> T cells in vitro (7, 26). Moreover, blockade of PD-L1 on antigen pulsed immortalized LECs in vitro increases CD8<sup>+</sup> T cell activation (7). In vivo, in several tumor mouse models, TA-LECs express higher levels of PD-L1 compared to naïve skin LECs (7, 8), the highest PD-L1 expression being observed in immunogenic tumors (8). Recently, Lane et al. demonstrated that PD-L1 expression by non-hematopoietic cells prevents CD8<sup>+</sup> T Garnier et al. Tumor Lymphatic Vessels in Immunity

cells accumulation in melanoma. IFN-γ production by antigenspecific CD8<sup>+</sup> T cells is primarily necessary to induce PD-L1 expression on LECs. Using mice with LECs deficient for IFN-γ receptor, they established that the specific loss of IFNγ sensitivity in LVs improves CD8<sup>+</sup> T cell-dependent control of melanoma tumor growth and mouse survival (8). Thus, during tumor development, a negative feedback loop is set up between LECs and T cells. LECs up-regulate PD-L1 expression in response to IFN-γ produced by tumor specific CD8<sup>+</sup> T cells, and subsequently inhibit T cell accumulation in tumors **(Figure 1.5)**.

In human metastatic melanoma, VEGF-C expression positively correlates with T cell infiltration and CCL21 expression (9). CCL21 plays a crucial role in the establishment of a tolerogenic tumor microenvironment by recruiting CCR7<sup>+</sup> regulatory T cells in primary tumors and by promoting the formation of lymphoid like stromal structures with immunosuppressive features (106). CCL21 further attracts naïve T cells that can be locally activated in response to immune blockade or vaccination (9).

#### CONCLUDING REMARKS

Recent studies indicate that tumor-associated LECs significantly contribute to shaping the immunosuppressive TME, therefore helping tumors hijack the immune system from an efficient to an incompetent anti-tumor response. Altogether, several observations highlight a new role for lymphatics in promoting tumor development, suggesting that lymphatic endothelium in the local microenvironment may be a novel target for immunomodulation. In agreement with these hypotheses, a recent publication demonstrated that following exposure to tumor derived factors, FRCs of the tumor draining LNs undergo multiple changes to convert into a immunosuppressive phenotype, such as decreased production of IL-7 and CCL19/21 (107). Whether a similar profound reprogramming occurs to LECs in tumor draining LNs remains to be determined.

#### REFERENCES


Whereas, VEGFC driven TA-lymphangiogenesis correlates with increased intratumoral inflammation (39) and immune suppression in progressing tumors (5), it seems also to be necessary for the response of the tumor microenvironment to immunotherapeutic intervention, as demonstrated for PD-1 blocking antibodies (9) **(Figure 1.6)**. This suggests that TA-LECs potentiate immunotherapy by attracting naive T cells through a CCL21 dependent mechanism. Accordingly, LVs and immune cell infiltrates positively correlated in metastatic cutaneous melanoma and colorectal cancer patients (39, 40). Once in the tumor, naive T cells can be locally primed upon PD-1 blockade, which reverts the immunosuppressive T cell imprinting and induces long-lasting anti-tumor immunity. Therefore, it is tempting to speculate that LV density in tumors could be used as a predictor for positive response to immune checkpoint blockade. Additional research will determine how to selectively target LEC immunosuppressive functions in tumors, which could, combined to immunotherapeutic approaches, lead to the conversion of a "cold" into "hot" immunogenic TME and potentiate anti-tumor T cell responses.

### AUTHOR CONTRIBUTIONS

LG, A-OG, and SH wrote the manuscript. LG conceptualized **Figure 1**.

### FUNDING

SH's laboratory is supported by the Swiss National Science Foundation (310030\_166541), the Swiss Cancer League, the Leenaards Foundation, and Carigest SA.

### ACKNOWLEDGMENTS

We thank Rémi Jeandenand for the design of the drawing (**Figure 1**).


and metastasis. Proc Natl Acad Sci USA. (2012) 109:15894–9. doi: 10.1073/pnas.1208324109


**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 © 2019 Garnier, Gkountidi and Hugues. 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(s) 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 Interplay Between Lymphatic Vessels and Chemokines

Rae H. Farnsworth1,2 \*, Tara Karnezis <sup>3</sup> , Simon J. Maciburko<sup>3</sup> , Scott N. Mueller 4,5 and Steven A. Stacker 1,2,6 \*

*<sup>1</sup> Tumor Angiogenesis and Microenvironment Program, Peter MacCallum Cancer Centre, Melbourne, VIC, Australia, <sup>2</sup> Sir Peter MacCallum Department of Oncology, The University of Melbourne, Parkville, VIC, Australia, <sup>3</sup> Lymphatic and Regenerative Medicine Laboratory, O'Brien Institute Department, St. Vincent's Institute of Medical Research, Fitzroy, VIC, Australia, <sup>4</sup> Department of Microbiology and Immunology, Peter Doherty Institute for Infection and Immunity, The University of Melbourne, Melbourne, VIC, Australia, <sup>5</sup> The Australian Research Council Centre of Excellence in Advanced Molecular Imaging, Melbourne, VIC, Australia, <sup>6</sup> Department of Surgery, Royal Melbourne Hospital, The University of Melbourne, Parkville, VIC, Australia*

#### Edited by:

*Amanda W. Lund, Oregon Health & Science University, United States*

#### Reviewed by:

*Burkhard Ludewig, Kantonsspital St. Gallen, Switzerland Bernahrd Ryffel, Centre National de la Recherche Scientifique (CNRS), France*

#### \*Correspondence:

*Rae H. Farnsworth rae.farnsworth@petermac.org Steven A. Stacker steven.stacker@petermac.org*

#### Specialty section:

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

Received: *21 December 2018* Accepted: *26 February 2019* Published: *12 April 2019*

#### Citation:

*Farnsworth RH, Karnezis T, Maciburko SJ, Mueller SN and Stacker SA (2019) The Interplay Between Lymphatic Vessels and Chemokines. Front. Immunol. 10:518. doi: 10.3389/fimmu.2019.00518* Chemokines are a family of small protein cytokines that act as chemoattractants to migrating cells, in particular those of the immune system. They are categorized functionally as either homeostatic, constitutively produced by tissues for basal levels of cell migration, or inflammatory, where they are generated in association with a pathological inflammatory response. While the extravasation of leukocytes via blood vessels is a key step in cells entering the tissues, the lymphatic vessels also serve as a conduit for cells that are recruited and localized through chemoattractant gradients. Furthermore, the growth and remodeling of lymphatic vessels in pathologies is influenced by chemokines and their receptors expressed by lymphatic endothelial cells (LECs) in and around the pathological tissue. In this review we summarize the diverse role played by specific chemokines and their receptors in shaping the interaction of lymphatic vessels, immune cells, and other pathological cell types in physiology and disease.

Keywords: chemokine, lymphatics, endothelial, chemokine receptor, lymphangiogenesis, lymphatic remodeling

#### INTRODUCTION

Cells in complex vertebrates receive signals from extracellular environment which coordinate a raft of important cellular programs and functions (1). These signals can be through direct cell-to-cell contact or by the use of soluble molecules synthesized and secreted by neighboring or distant cells. Growth factors and cytokines are examples of soluble proteins that have potent cellular effects through designated cell surface receptors, such as growth and differentiation (2). A subset of the cytokine proteins that act to induce the movement of cells are the chemokines (-kinos from the Greek for movement) (3). Chemokines are small, highly conserved polypeptides of 70–100 amino acids. While having a conserved three-stranded β-sheet/α-helix tertiary structure they are divided into several subfamilies (CXC, CC, XC, and CX3C) based on variations in their quaternary structure and critical cysteine residues (4, 5). They exert their effects through cell surface G-protein coupled receptors on target cells (4) that can act as homo- or heterodimers depending on the context. This family has now expanded to include at least 51 chemokines and 20 receptors, plus (presently) four atypical or decoy receptors which typically dampen chemokine activity by binding and internalizing chemokines without initiating G-protein-dependent signaling (5–7).

Chemokines act by establishing gradients to direct random or directed migration of cells bearing cognate receptors from lower to higher concentrations of ligands. These gradients are often formed through the interaction with proteoglycans attached to the cell surface or extracellular matrix. Diversity within the chemokine system is generated both structurally and functionally through an array of different receptors and ligands with precise or promiscuous binding affinity, where splice variants, posttranslational modifications including nitrosylation, citullination, and many forms of proteolytic cleavage (8) can all diversify signaling leading to events that are either chemoattractive or chemorepulsive (5, 9, 10). The biological effects of the chemokine family are broad-ranging as they can be used to move individual cells, subsets of cells or large groups of cells in order to achieve the outcomes of significant processes such as immune cell development, embryogenesis, angiogenesis, phagocytosis and survival/apoptosis (5). Expand this to controlling these cell population during infection, immunity, inflammation, and other pathologies and the extensive roles of chemokines in the mammal is clear.

The movement of cells in normal and pathological situations is highly dependent on the circulatory system, which allows long and short range transport, and exit and entry from all tissues. Previous studies have shown the critical role of blood vessels in chemokine action, in particular directing key cellular effectors of the immune response (11). Blood and lymphatic vessels work together to control fluid and cells in the circulation and tissues, yet the blood vessels have often received the most attention. However, the important and independent roles the lymphatics play in cellular interactions in normal physiology, development, and pathology are becoming evident through studies in a number of areas highlighting the organ- and subtype- specific activity of lymphatic vessels (12–14).

Lymphatic vessels have gained a greater prominence in our thinking over the past two decades as molecular tools have facilitated clear discrimination from blood vessels (15– 17). Further, the characterization of factors required for growth and differentiation of lymphatic endothelial cells (LECs) in vitro has provided a more in-depth understanding of their unique biological function and differences to blood vascular endothelium (18). Extensive in vivo studies using promoters with specificity to the lymphatic compartment has also identified key functional roles for the lymphatics and LECs in development and disease (14, 19), and other functional screens have highlighted the unique features of LECs (20, 21). These unique responses of lymphatic vessels are often regulated through the interaction of cells and signaling molecules with the LECs lining the lumens of lymphatic vessels.

The paradigm of chemokine action involving the lymphatics is potentially complex. The lymphatics can both be the source of the chemokines, express the receptors, or both (**Figure 1**; **Table 1**). As a vessel for the passage of many circulating cells lymphatics also act as a conduit allowing the flow of chemokines or cells to other targets; for instance to lymph nodes (LNs). Akin to the action of chemokines in blood vessel function the lymphatics provide a surface for the attraction and interaction of immune cells in pathological contexts (3, 63, 64). This review aims to highlight the interplay between lymphatic vessels and chemokines in a range of biological contexts from embryonic development through to regulation of immunity and a range of human pathologies.

### DEVELOPMENT

#### Regulation of Lymph Node Organogenesis

In accordance with their function in providing immune surveillance for particular organs or regions of tissue, LNs develop at strategic locations along the vasculature, typically at the branch points of large veins (25, 65). Although the mechanisms controlling the precise location and subsequent assembly of LNs are incompletely understood, recent evidence suggests multiple roles for lymphatics as well as the venous vasculature.

Chemokines are already known to be critical for initiating the development of LNs and other secondary lymphoid organs such as the Peyer's patches in the gut (25). Mice lacking CXCL13 or its receptor CXCR5 fail to develop particular subsets of peripheral LNs and also exhibit impaired Peyer's patch formation (66, 67). Combined deficiency of CXCR5 and CCR7 completely ablated the formation of peripheral LNs (68), although interestingly deficiency of CCR7 alone had only a mild impact (68, 69). In the prevailing model of LN development, lymphoid tissue organizer (LTo) cells—cells of mesenchymal origin induced by neuronally-derived retinoic acid signaling (70) at putative sites of LN development—secrete the chemokines CXCL13, CCL19, and CCL21 which in turn recruit CXCR5- and CCR7-expressing haematopoietic lymphoid tissue initiating (LTi) cells into the growing LN anlage (25, 65, 68). LTi cells extravasate from veins at junctions where smooth muscle coverage is sparse (71) and make contact with LTo cells, whereupon a positive feedback loop ensues: activation of IL-7Rα on LTi cells by LTo-expressed IL-7 upregulates lymphotoxin expression by LTi cells, which in turn promotes further chemokine secretion by LTo cells (66, 72). These reciprocal interactions lead to the expansion of both cell populations, growth of the LN anlage and subsequent differentiation of lymphoid organ subcompartments (25, 65). However, the persistence of LN formation when LTβR signaling was specifically ablated in CXCL13 or CXCL19 expressing mesenchymal cells (73, 74) suggested that additional LTo cell types may exist.

The contribution of lymphatics to LN organogenesis was previously unclear—LN development is still initiated in mice that lack lymphatics due to global or Tie2 promoter-restricted Prox1 knockout, although the anlagen in these animals show defects in the differentiation and organization of mesenchymal LTo cells, and are often reduced in size (75). Recent studies utilizing cell type-specific gene knockout approaches have now elaborated multiple roles for lymphatics in LN formation. Lymphatic vessels contribute to early LN initiation by delivering recirculating LTi cells from peripheral tissue to the LN anlage

**Abbreviations:** BEC, Blood vascular endothelial cell; DC, Dendritic cell; LEC, Lymphatic endothelial cell; LN, Lymph node; LTi, Lymphoid tissue initiator; LTo, Lymphoid tissue organizer; SCS, Subcapsular sinus (of lymph node).

through CCL21/CCR7-mediated chemotaxis (71, 74). The same sphingosine-1-phosphate (S1P) signaling that regulates lymphocyte egress in adult LNs retains LTi cells at the LN anlage, potentiating the molecular crosstalk between LTo and LTi cells that results in further chemokine-mediated LTi recruitment and subsequent LN maturation (25, 74). Peripheral LN anlagen typically form near major venous junctions which run parallel with collecting lymphatic vessels. LECs within the collecting vessel adjacent to the accumulation of LTi and LTo cells subsequently proliferate in a VEGF-C/VEGFR-3 dependent manner to form a disc which eventually expands to envelop the growing LN (71, 76). Functional lymph flow also appears to be essential for complete LN formation as it generates interstitial fluid force which likely stimulates CXCL13 expression by fibroblastic LTo cells (71). Notably, many of the cellular mechanisms and signaling pathways (including chemokines) involved in LN organogenesis are recapitulated in the development of tertiary lymphoid organs in response to pathological insult (25). This suggests that enhanced lymphangiogenesis and lymph flow may contribute to the de novo development of lymphoid organs in order to strengthen local immune responses, and that this mechanism may be therapeutically manipulable (25).

### Regulation of Lymphatic Vascular Patterning

Where chemokines mediate interactions between the blood and lymphatic vasculature and other cell types, the endothelial cells lining these vessels are commonly characterized as the source of the chemokine ligand, or the surface to which it binds. However, endothelial cells themselves also express chemokine receptors and can respond to chemokine gradients generated by other cell types. As such, a growing list of chemokines and their receptors have been implicated in directing the growth and patterning of blood and lymphatic vasculature.

CXCR4 and its ligand CXCL12 have well-described roles in promoting angiogenesis and patterning the embryonic vasculature (77–81). Recently their role in patterning the lymphatic vasculature has also been described (45). In zebrafish, expression of cxcr4a and cxcr4b was detected in lymphatic progenitors sprouting from the posterior cardinal vein, as well as in the developing parachordal line, intersegmental lymphatic vessels, and other large trunk lymphatics such as the thoracic duct (45). Loss- and gain- of function experiments confirmed that these receptors were required for the development of the large trunk lymphatics. Accordingly, dynamically regulated expression of ligand-encoding genes cxcl12a and cxcl12b in the dorsal aorta and arterial intersomitic vessels directed the parallel migration of the growing lymphatic vessels along these paths (45). Interestingly, another group has shown that upregulation of cxcl12a was mediated by the microRNA miR-126, which also synergises with Flt4 (VEGFR-3) signaling (82).

Atypical or decoy chemokine receptors also play important roles in shaping developmental lymphangiogenesis. Mice deficient in ACKR3 (formerly known as CXCR7) exhibit defects in lymphatic development, typified by precocious development TABLE 1 | Chemokine-mediated cellular interactions with LECs.

#### CHEMOKINE LIGANDS EXPRESSED BY LECs



#### ATYPICAL CHEMOKINE RECEPTORS EXPRESSED BY LECs


*<sup>a</sup>Compiled from references (5) and (49).*

of lymph sacs and hyperplasia (60). This phenotype was found to be caused by excessive pro-proliferative signaling from adrenomedullin, a non-chemokine ligand for ACKR3 which is a positive regulator of lymphangiogenesis. Deficiency of ACKR2 (formerly D6; CCBP2, chemokine-binding protein 2) in mice results in hyper-branched lymphatics (40). ACKR2 scavenges LEC-expressed CCL2, which is chemotactic for monocytes via CCR2 signaling, thereby reducing the accumulation of macrophages in proximity to developing lymphatics (40). These macrophages deliver lymphangiogenic growth factors and play important roles in shaping developmental lymphangiogenesis (83, 84). These studies highlight the complex mechanisms by which chemokines orchestrate multiple cellular interactions within the developing embryo.

### LEUKOCYTE TRAFFICKING IN INFLAMMATION AND IMMUNITY

Arguably the best-characterized chemokine-mediated functions involving the lymphatics are those that regulate trafficking of leukocytes in physiological homeostasis and during inflammation and immune responses. Leukocytes in the peripheral interstitium typically enter initial lymphatics in the first instance, and subsequently migrate through the local plexus of pre-collecting lymphatics before entering the large collecting lymphatic vessels that pump lymph and cells over long distances to LNs, where encounters between antigen-presenting cells and cognate T and B lymphocytes are coordinated (49, 85, 86).

#### Entry of Leukocytes Into Peripheral Lymphatic Vessels

CCL21, constitutively expressed by peripheral LECs, has a prominent role in trafficking CCR7-expressing dendritic cells (DCs) through afferent lymphatic vessels to LNs along with other CCR7-expressing cells such as T cell subsets and neutrophils (22– 24). Notably, in the peripheral vasculature CCL21 expression is relatively specific to the endothelial cells of initial lymphatics; it is generally absent from blood vascular endothelial cells (BECs) with the exception of high endothelial venules in the LN (23). The elongated, positively-charged C terminus of CCL21 mediates strong binding affinity to diverse proteoglycans as well as collagen IV, allowing it to form chemotactic gradients on the surface of LECs and adjacent extracellular matrix (87). The requirement of CCL21-CCR7 interactions for trafficking immune cells through afferent lymphatics has been long recognized and has been reviewed extensively elsewhere (23, 49), but recent studies continue to shed light on the precise mechanisms and additional chemokines that are involved.

Although low levels of cellular trafficking occur under homeostatic conditions, this increases dramatically during immune responses (49, 86). Accordingly, constitutive expression of homeostatic chemokines in LECs is supplemented during inflammation by increased expression of these chemokines, along with additional "inflammatory" chemokines that shape the immune response. CCR7 is upregulated in DCs by inflammatory stimuli such as TNF-α, while the same stimuli increase CCL21 release by LECs by upregulating transcription and by releasing intracellular stores of the chemokine (88, 89). ACKR4 (previously known as CCRL1) expressed in dermal LECs and keratinocytes plays an essential role in properly directed egress of DCs from skin during inflammation by scavenging the more soluble CCR7 ligand CCL19, which would otherwise retain DCs in skin (61). LEC-expressed ACKR2 also regulates DC egress during inflammation by scavenging inflammatory chemokines to ensure preferential presentation of CCL21 on the cell surface of LECs. This in turn supports adhesion of mature CCR7+ DCs to LECs and their transport to LNs, in preference to immature DCs and inflammatory myeloid cells (52, 53). In mice lacking ACKR2, elevated presentation of inflammatory chemokines such as CCL2 on peripheral and LN LECs leads to congestion of lymphatics by myelomonocytic cells, with downstream impairment of lymphatic transport and consequently dampened antigen-specific immune responses (52). Similar roles for ACKR2 in orchestrating cell migration and resolving the inflammatory response have been described in a range of pathological contexts (50).

DC migration toward lymphatic vessels is also mediated by expression of CXCR4 in activated DCs and its ligand CXCL12 in LECs, although DCs seem to preferentially migrate toward CCL21 when both chemokines are present, indicating a coordinated rather than additive function (33). CX3CL1, an atypical chemokine possessing a transmembrane domain, is upregulated in LECs by TNF-α and mediates basolateral-toapical migration of CX3CR1-expressing DCs through lymphatic endothelium and DC trafficking to LNs during dermal contact hypersensitivity responses in vivo (41). Interestingly, this chemokine is predominately shed from the basolateral surface of LECs by ADAM10 and ADAM17 metalloproteases, in contrast to remaining membrane-bound and behaving as a leukocyte adhesion molecule in BECs (41).

DCs commonly enter lymphatics through binding to immobilized CCL21 puncta specifically localized between the button-like intercellular junctions characteristic of initial lymphatics (90, 91). Direct contact between DCs and LECs also dynamically triggers localized release of CCL21 from within the trans-Golgi network of LECs, further potentiating transendothelial migration (92). Once inside lymphatic vessels, DCs crawl in a semi-directed manner within the flattened lumen, moving in multiple directions but ultimately following an intralymphatic gradient of CCL21 that is generated by lymphatic flow (91, 93). In this context, CCL21 immobilized on the LEC surface mediates not only chemotaxis but also adhesion. Although most DCs and other leukocytes exiting the periphery are thought to enter the initial lymphatics, specific chemokines may regulate entry of particular cell types into other segments of the lymphatic vasculature. CCL27 was found to be specifically expressed in pre-collecting lymphatics, where it promoted the attraction of CCR10+ T lymphocytes in vitro and in vivo (27). Pre-collecting LECs were also found to overexpress CCL27, CXCL12, CXCL14, and the promiscuous CC chemokine decoy receptor ACKR1 (formerly DARC, Duffy antigen receptor for chemokines) compared to initial LECs, whereas CCL21 was more abundantly expressed in initial LECs (27). The same study found that pre-collecting and initial LECs in the adult human dermis could be discriminated by flow cytometry and immunofluorescence according to their expression levels of Podoplanin—pre-collecting LECs being designated as Podoplaninlow and initial LECs Podoplaninhigh (27). The correspondence between Podoplanin and CCL21 expression levels between LEC subtypes may relate to the ability of the glycoprotein Podoplanin to bind and present CCL21 on the LEC cell surface (94). More remains to be understood about the specific immune cell types which are selectively recruited to initial vs. pre-collecting lymphatics, and the functional importance of these differences.

#### Chemokine Signaling Within Lymph Nodes

Once cells pass into collecting lymphatics the lymph flow rate accelerates, and cells are transported passively to LNs where they are delivered into the subcapsular sinus (SCS) (91, 93). Here chemokine gradients also are important in regulating migration and localization of different cell types in the LN parenchyma (49, 85). It has recently been demonstrated that the LECs comprising the LN SCS and medullary lymphatic sinuses have distinct expression profiles, including differential expression of several chemokines and receptors (95). ACKR4 is specifically expressed in LECs of the SCS "ceiling" where it plays an important role in scavenging and internalizing CCL21 to create a gradient that directs DC migration toward and ultimately through the SCS floor into the LN parenchyma (62). CCR7+ T cells arriving through afferent lymphatics have been observed to enter the LN parenchyma preferentially through medullary sinuses, but will transmigrate through the SCS floor only in conjunction with local changes induced by DCs (96). CCL21 is also produced abundantly by fibroblastic reticular cells (FRCs), which constitute the majority of the LN stroma and guide interactions between DCs and naïve T cells both structurally and chemically (97). Notably, lymphatic flow upregulates CCL21 expression by FRCs (98), reiterating the importance of the lymphatics for maintaining proper immune function in the LN microenvironment. Some DCs require additional signals to CCL21/CCR7 to access the LN parenchyma: in cutaneous allergic responses, CD301b+ DCs were found to require CCR8 signaling in response to CCL8 from interfollicular CD169+SIGN-R1+ macrophages (99). Within the LN parenchyma, chemokines from a variety of cellular sources exquisitely regulate localization of specific leukocyte subsets to coordinate effective immune responses, reviewed in detail elsewhere (49, 85).

In response to infection and inflammation, LN LECs respond robustly and dynamically. Proliferation of LECs supports the expansion of LNs during immune responses and coincides with increased expression of chemokines including CXCL9, CXCL10, CCL2, CCL5, and CCL20 (31, 32, 100) (**Table 2**). In the latter stages of inflammatory remodeling of the LN, the cortical and medullary sinuses expand substantially (31, 103). During certain infections, such as persistent infection with the helminth Heligmosomoides polygyrus, lymphangiogenesis driven by VEGF-A and VEGF-C from B lymphocytes results in a sustained expansion of LN LECs (104). Such changes potentially support the egress of leukocytes from the inflamed LN and the restoration of homeostasis, however may also influence lymphoid tissue functions in response to subsequent infections.

Egress of leukocytes from LNs occurs predominately via the medullary lymphatic sinuses, which channel cells into efferent collecting lymphatics (49, 86). Leukocyte retention in or egress from a LN is regulated by a balance of directional signals and changes in receptor expression (49). Prolonged signaling through CCR7 in T cells or CXCR5 in B cells leads to reduced expression or responsiveness of these receptors to their ligands expressed by LN LECs or other stromal cells, and leukocytes instead upregulate S1P receptor 1 (S1PR1), the main receptor promoting leukocyte egress (105–107). Notably, LECs have been defined to be the key cellular source of S1P regulating lymphocyte egress, as determined by conditional gene deletion of the two enzymes responsible for S1P generation (Sphk1 and Sphk2) by LYVE-1-directed expression of Cre recombinase (108).

While many of the major LEC-expressed chemokines and receptors that regulate leukocyte trafficking have been defined, many more questions remain. It is evident that different pathological stimuli elicit expression of different suites of chemokines and receptors in LECs (31, 101). This indicates a role for LECs in trafficking context-specific subsets of leukocytes to LNs, as well as interacting with other cells within the tissue microenvironment, which remains to be explored.

### WOUND HEALING

Chemokine signaling plays an integral part in the healing process of various wounds including lacerations, surgical incisions, burns and skin grafts, as well as chronic diabetic and aging wounds. Wound healing is a dynamic and highly coordinated process with three distinct stages—inflammation, tissue formation, and tissue remodeling, with each stage having a distinct chemokine profile (109, 110). Inflammation is a crucial part of wound healing and despite the variety of tissue injuries that can occur, the subsequent events share a similar course (111). The aim of the initial inflammatory phase is to prevent further blood/fluid loss, protect against infection, and initiate the clearance of dead or dying cells and tissue debris (111). It begins immediately after tissue damage and is characterized by the formation of a platelet plug, the deposition of a fibrin matrix and the recruitment of neutrophils, which are the predominant inflammatory effector cells in the first 24–48 h (112). Monocytes enter the wound area 48–72 h after injury, differentiate into macrophages and play an important role in coordinating subsequent events of wound repair (111). The second stage, tissue formation, spans 2–10 days after injury, and aims to restore the barrier function of the epithelium (110, 111). Angiogenesis occurs from blood vessels at the wound edge and the newly formed capillaries, along with macrophages and fibroblasts, replace the fibrin matrix with granulation tissue, and allow for the proliferation and migration of keratinocytes across the wound surface (111, 113). In addition to angiogenesis, an adequate growth of lymphatic vessels in and around the wound zone is critical for the normal healing. Lymphangiogenesis follows angiogenesis via


*Results derived from microarray studies of LECs isolated from in vivo settings.* ↑ *relatively higher,* ↓ *relatively lower mRNA expression in first vs. second disease/tissue setting. ND, none detected/none described.*

sprouting from existing lymphatic vessels at the wound edge and is primarily stimulated by VEGF-C or VEGF-D secreted by macrophages located in the microenvironment (114–116). This facilitates the drainage of tissue edema and transport of DCs from the wound zone (114, 117–119). Macrophages also stimulate some fibroblasts to differentiate into myofibroblasts and working together with fibroblasts, a predominately type III collagen extracellular matrix is deposited and the edges of the wound are brought together over time (120, 121). The final stage of tissue remodeling begins 2–3 weeks after injury and can take over a year to complete (111). It is characterized by the progressive cessation of the inflammatory response and the remodeling of the type III collagen matrix to type I collagen (122).

Although the three phases of wound healing are distinct, the inflammatory reaction continues until tissue remodeling, albeit with changing cellular mediators of inflammation (113). Leukocytes have the dual role of acting as immunological effector cells as well as modulators of inflammation. In the acute phase, the production of proteases and reactive oxygen species aids with tissue degradation, while the secretion of growth factors in the later stages promotes tissue formation (113). Chemokines are integral in activating and recruiting leukocytes to specific microanatomical sites of the wound as well as stimulating angiogenesis (63, 109). Neutrophils, the initial responders of the acute inflammatory response, are recruited by CXCL1, CXCL5, CXCL7, and CXCL8 (formerly IL-8), secreted by activated platelets, BECs, pericytes and resident monocytes within the injured tissue (123–126). Monocyte and macrophage recruitment follows closely behind and is mediated by CCL2 secretion (63). CCL2 is also chemotactic for lymphocytes but after day 4 postinjury, CXCL9, CXCL10, and CCL22 secreted by monocytes and macrophages take over (127). The role of the CXC family of chemokines in angiogenesis is well-established (63) and the concentration of CXCL1, CXCL8, and CXCL12 in the healing wound is greatest during days 1–4 post-injury and correlates with an increasing number of blood vessels within the wound (113). The high levels of CXCL1 and CXCL8 within the wound also stimulate keratinocytes via CXCR2 to increase proliferation and migration, which enhances re-epithelialization (128, 129).

While the mechanisms of wound repair (111) and the chemokines involved (109, 113) have been extensively reviewed, the effect of these chemokines on the lymphatic vasculature and the role of chemokines secreted by the lymphatic endothelium on the healing wound are not well-established. The extent to which lymphatics will respond to chemokines secreted during the various stages of wound healing will largely depend on their expression of chemokine receptors. Like BECs, LECs express receptors CXCR1 and CXCR2, both of which are upregulated during inflammation (48, 130). As such, LECs have the potential to interact with the CXCL1, CXCL5, CXCL7, and CXCL8 chemokines that are expressed in the healing wound. Of these, CXCL1 and CXCL8 have been shown to promote lymphangiogenesis via increased LEC migration and tube formation, and additionally increased LEC proliferation in the case of CXCL8 (38, 48, 131). Interestingly, CXCR2 expressed on the surface of lymphatics acts as a scavenging receptor, capable of binding various inflammatory chemokines that can shape chemokine gradients. LEC-expressed CXCR2 is thus likely to influence the inflammatory response, and has also been shown to be important in lymphatic vessel remodeling, two key components of wound healing. For example, CXCR2 ligands, CXCL1 and CXCL2, have been shown to be elevated during the inflammation stage of wound healing and in skin graft wounds (109, 132). LECs also have the ability to differentially secrete CXCL1 and CXCL8 depending on the local environment, which is thought to act in an autocrine and/or paracrine manner to increase lymphangiogenesis (133). However, in the context of a healing wound, LEC secretion of these chemokines may also have an endocrine effect in creating a chemokine gradient to recruit distant neutrophils. Another lymphangiogenic chemokine present in the healing wound is CXCL12, which binds CXCR4 expressed by LECs to induce migration and tube formation in a novel pro-lymphangiogenic pathway that is distinct from the classical VEGFR-3 pathway (46). Furthermore, increased secretion of CCL21 by LECs in the inflammatory wound environment may increase migration of DCs and other antigen-presenting cells to help activate an immune response, which may assist in healing of infected wounds (88, 101, 134).

Complications in wound healing impair the ability of lymphatic vessels to regenerate and repair, leading to impaired lymphatic drainage which results in lymphedema, with the risk for recurrent infection. Therefore, there is a clinical need to better understand the regulation of lymphatic vessel function during wound healing. Chemokines have been a focus for therapeutic approaches to promote wound healing, in particular targeting the CXCL12/CXCR4 signaling axis, a key pathway regulating the recruitment of bone-marrow derived stem cells with regenerative capacity (135). Greater understanding of the specific chemokine pathways involved in wound healing will present additional therapeutic opportunities.

## CANCER

While cancer is a genetic disease initiated through the acquisition of specific mutations in key genes, the resulting changes to the cell biology within a host drives the important clinical manifestations of the disease. The progression of cancer, from its evasion of the immune system to its ultimate spread to critical organs systems in the body, has a reliance on altered chemokine signaling resulting from the presence of mutated tumor cells. The lymphatics play a key role in both controlling access to and interaction with the immune system, and also provide an initial means of escape for primary tumor cells, while chemokines also influence immune responses and the pattern of metastatic spread through directed migration of tumor cells in a tissue-specific manner (11, 17, 34, 136).

#### Leukocyte Recruitment and Egress

Multiple chemokines and receptors have been implicated in the recruitment of specific immune cell subsets to tumors and in influencing cancer immunotherapy responses (26, 137). However, the involvement of lymphatics in anti-tumor immune responses is only beginning to be understood. Powerful lymphangiogenic growth factors that drive the formation and remodeling of lymphatic vessels have been shown to upregulate chemokines in the context of cancer. VEGF-C has been shown to upregulate CCL21 expression by LECs, driving CCR7-dependent tumor chemoinvasion toward lymphatic vessels (138). CCL21 has also been shown to promote lymphoid-like stromal components and immune escape in melanoma tumors in mice, raising the concept that CCL21-secreting tumors can alter the host immune response from immunogenic to tolerogenic which then impacts on tumor progression (139). Recent extension of these observations has shown a role for VEGF-C-induced CCL21 in the tumor infiltration of naive T cells prior to immunotherapy via CCR7-dependent chemotaxis (140). The authors of this study propose that VEGF-C, through VEGFR-3 signaling, can potentiate immunotherapy by attracting abundant CCR7+ naive T cells, which are then locally activated by the immunotherapy. These studies point to a role for VEGF-C and potentially other lymphangiogenic factors as predictive biomarkers for immunotherapy, with a chemokine providing a key link in the signaling chain (140). Meanwhile, other studies are beginning to unravel the complex mechanisms by which lymphatics influence the tumor immune microenvironment (141, 142). Other groups have speculated that "key driver chemokines"—for example CXCL10, which is expressed by LECs in several pathological contexts and implicated in metastasis to LNs (28–30, 143)—may be valid targets in diseases including cancer because of their ability to enhance T-cell-dependent anti-cancer immunity (144).

### Tissue-Specific Patterns of Metastasis

A common manifestation of chemokine involvement in directing patterns of metastasis is that tumor cells express chemokine receptors that respond to chemokines secreted by cells of a given tissue or organ, often co-opting the chemokine signaling used for tissue-specific homing of leukocytes (11, 26, 34, 35). This is true for lymphatic vessels and LECs present in the primary tumor, regional LNs or distant organs that are targets of metastatic spread. Studies in a variety of tumor types have shown that CCR7 present on the tumor cells mediates their migration toward CCL21-expressing initial lymphatics (in preference to blood vessels) and/or LNs, thereby promoting spread via the lymphogenous route (145) [recently reviewed in (11, 26)]. A similar mechanism is mediated by tumor-expressed CXCR4 and CXCL12 expressed in LECs and LNs (11, 26). In particular, CXCL12 secreted by LECs in the LN SCS contributes to an attractive and supportive metastatic niche for CXCR4+ tumor cells (36, 37).

The list of chemokines that similarly promote LN metastasis is expanding. Expression of CXCR3 in colorectal cancer was linked to increased metastasis to LNs, likely in response to expression of ligands CXCL9, CXCL10, and CCL21 in the lymphatic sinuses and paracortex of LNs (29). As the rate of spread of CXCR3 expressing or -deficient cell lines was similar, potential growth effects were also considered. A similar effect was also seen in melanoma (30). In gastric cancer, LECs upregulated expression of CXCL1, which in turn increased tumor cell invasiveness via CXCR2 and stimulated lymphangiogenesis. Expression of both ligand and receptor in patient gastric cancer specimens was associated with LN metastasis and poor survival (38, 39). Soler-Cardona et al. characterized a mechanism in melanoma where neutrophils recruited to melanomas by CXCL5 appeared to facilitate transmigration of tumor cells through lymphatic endothelium (47). Expression of CCL1 in the LN SCS was also shown to regulate entry of CCR8+ melanoma cells into the

LN (43). It is noteworthy that recent studies in mouse models have shown that initial lymphogenous spread can transfer to the blood vascular system via high endothelial venules within regional LNs (146–148).

With the exception of their arrival to LNs via afferent lymphatics, tumor cells are presumed to home to specific organs via the blood vessels, guided by chemokines produced by BECs or transcytosed to the vessel lumen (11, 34). Nonetheless, LECs in distant organs can also participate in distant organ metastasis and metastatic niche formation. Lee et al. showed in mouse models that circulating IL-6 secreted by orthotopic breast cancer cells could influence LECs in LNs and lung. These distant LECs were induced to express CCL5, which was chemotactic for the tumor cells, and VEGF-A, which increased angiogenesis and vascular permeability at metastatic sites (42). Another study used a Vegfr3-luciferase reporter mouse and melanoma models to demonstrate pre-metastatic lymphangiogenesis in distant organs, and to identify midkine as a regulator of metastatic niche formation with prognostic significance (149). These studies open up new avenues of investigation into how lymphatics at distant metastatic sites can also influence metastasis.

#### Atypical Chemokine Receptors in Cancer

Atypical or decoy chemokine receptors have in a number of contexts been involved in modulating cancer progression through shaping the inflammatory response (7, 150). The atypical chemokine receptor ACKR2 has been shown to internalize and sequester an array of pro-inflammatory chemokines of the CC family (5, 49, 50). Mice deficient in ACKR2 had an increased susceptibility to the development of cutaneous tumors that was linked to the recruitment of immune cells (e.g., T cells and mast cells) to support their development (54). In this study ACKR2 was predominately expressed in LECs of human oral squamous cell carcinomas (OSCC) and not tumor cells or epithelial cells, and the levels in tumor LECs were upregulated compared to normal LECs (54). This same group had previously shown that ACKR2 is expressed by lymphatic endothelium and may influence the recirculation of leukocytes via a chemokine driven mechanism, as ACKR2 was expressed on the afferent lymphatics (51). Antigen-experienced T cell subsets express multiple CCR receptors, with CCR4 specifically implicated in cutaneous T cell homing (151). Mast cells also express CCL3 receptors, and these cells generally play a role in promoting tumor angiogenesis and recruiting other pro-tumorigenic leukocyte subsets. CCL3 further directly contributes to mast cell degranulation via CCR1 (152). Expression of ACKR2 therefore limited inflammation by restricting availability of chemokines that attracted these pro-inflammatory leukocytes (54). Other studies of ACKR2 have confirmed its role in the lymphatic system in other organs. ACKR2-deficient mice are more susceptible to inflammation-induced colon carcinogenesis, an effect that was attributed to lymphatic expression of ACKR2 using bone marrow transplantation experiments (55). ACKR2 was also found to be upregulated on lymphatics of inflamed and cancerous colon specimens (55). However, other groups have reported contrasting results (56, 57), potentially suggesting dynamic and context-specific roles for ACKR2 during inflammatory colon carcinogenesis. ACKR2 is also highly expressed in vascular tumors of lymphatic origin (51) and the spindle cells of Kaposi's sarcoma (58). In more aggressive tumors this receptor is downregulated through the KRAS/BRAF/ERK pathway, leading to chemokine-mediated macrophage recruitment and increased angiogenesis and tumor growth (58).

In breast cancer the absence of a number of members of the atypical chemokine receptor subset predict involvement of axillary LN metastasis, a key clinicopathological indictor of disease progression (153). These observations were further validated by the characterization of genetic variants of two chemokine decoy receptors, ACKR1 and ACKR2, that associated with the metastatic potential of breast cancer (154, 155). Yu et al. found that the expression of the atypical chemokine receptors also predicted relapse-free survival in breast cancer where coexpression and co-genotype (two major alleles of DARC-rs12075 and D6-rs2228468) of the chemokine decoy receptors ACKR1 (DARC) and ACKR2 (D6) had significant associations. This data shows that host factors such as polymorphisms of major chemokine receptor genes and the expression of the protein receptors in cancers, including in lymphatic or blood vessels, could help predict prognosis (155).

#### Tumor Lymphangiogenesis

Lymphangiogenesis and lymphatic remodeling in tumors, commonly driven by VEGF-C and VEGF-D, are strongly associated with metastasis to LNs and distant organs (17). We recently identified a cooperative role for CCL27, CCL28, and their receptor CCR10 in VEGF-D driven tumor lymphangiogenesis (44). Here, CCR10 was expressed by LECs and upregulated by VEGF-D and the pro-inflammatory cytokine TNF-α. LECs were attracted to both CCL27 and CCL28 in a CCR10-dependent fashion. Further examination of CCR10 deficient mice confirmed a role for this receptor in lymphatic patterning. While CCL27 alone was not sufficient to drive metastasis, both chemokines enhanced LEC migration and worked in combination with VEGF-D to recruit LECs and form coherent vessels (44). The study suggests a cooperative action of chemokines, inflammatory mediators, and lymphangiogenic growth factors during cancer progression. Interestingly, VEGF-D was also shown to upregulate expression of ACKR2 in LECs in vitro (53). Other studies have shown a link between chemokine signaling and VEGFR-3-driven lymphangiogenesis in cancer where VEGF-C can upregulate CXCR4 and thereby cooperate with CXCL12 in driving lymphangiogenesis and metastasis (46). Notably CXCR4 has a well-established role in tumor angiogenesis as well, and is being actively pursued as a therapeutic target (64, 156). Tumor-expressed CXCL5 in melanoma has also been found to drive tumor lymphangiogenesis and lymphogenous metastasis through CXCR2 expressed on LECs (47). Gastric cancer cells induce expression of CXCL1 in LECs, which subseqently drives tumor lymphangiogenesis and lymphogenous metastasis (38). These studies illustrate that as well as being a source of chemokine ligands in cancer, lymphatic vessels can also be guided by chemokine receptor signaling.

Lymphangiogenesis is also coupled with chemokine signaling by fluid mechanics (157). Lymphatic flow is important in stimulating chemokine secretion by LECs and other cells, as well as for generating gradients of chemokines that can be followed by migrating tumor cells (98, 157, 158). Under conditions of interstitial flow, tumor cells co-expressing a chemokine and its receptor can thereby exhibit "autologous chemotaxis," following a self-generated chemokine gradient toward lymphatics (159).

### OTHER PATHOLOGIES

Lymphatic vessels have been observed to intersect with chemokine-mediated movement of important effector cells in a variety of diverse human pathologies. In type 2 diabetes patients a range of chemokines and related genes were differentially expressed in dermal LECs compared to non-diabetic patient LECs (28) (**Table 2**). Enhanced lymphatic density was observed in skin, along with upregulation of CXCL10 and downregulation of CCL27 and CXCL14 in response to pro-inflammatory conditions. TNF-α upregulated CXCL10 in LECs, and LECderived CXCL10 was able to mediate macrophage adhesion to LEC monolayers and invasion into agarose plugs (28). The study identified paracrine cross-talk allowing macrophage recruitment toward LECs via a chemokine-mediated mechanism.

Studies of the mechanisms of human kidney transplant rejection show that inflammatory infiltrates rich in lymphocytes attack both cortical tubules and endothelial cells. This is accompanied by significant increases in local lymphatic vessel density due to "lymphatic neoangiogenesis" (94). LECs from these vessels express and secrete CCL21 which attracts CCR7+ cells (94). A later study showed that ACKR3 (CXCR7) was also expressed by LECs during kidney rejection with nearly 1/3 of adult dermal lymphatics expressing ACKR3, and both ACKR3+ blood and lymphatic vessels increasing in number during allograft rejection (59).

A role for ACKR2 on lymphatic endothelium in autoimmune disease is implied from studies of the ACKR2-deficient mice during experimental autoimmune encephalomyelitis (EAE) (160) where encephalitogenic responses, including DC migration and T cell priming, were impaired (160). Interestingly other studies have shown that ACKR2-deficient mice develop enhanced symptoms of EAE (as well as collagen-induced arthritis) due to

#### REFERENCES


enhanced Th17 responses (161). These differences could be due to control of IL-17 production by ACKR2 (50).

#### FUTURE DIRECTIONS AND CONCLUSIONS

Lymphatic vessels, like blood vessels, are a highly interactive surface for cells of the immune system, and through the use of chemokines and their receptors can coordinate key interactions. These pathways can control the entry and function of particular immune subsets in a number of pathological conditions. Nonetheless LECs have distinct patterns of chemokine secretion and expression of chemokine receptors that distinguish them from the blood vessel system and mediate distinct roles and responses. The abundance and diversity of the chemokine family point to the likelihood that a plethora of novel chemokine functions and interactions remain to be discovered. Of note, several recent studies have undertaken differential expression profiling of LECs by microarray in a range of different pathologies, revealing multiple chemokines with as-yet undefined roles in disease (**Table 2**). These studies are complemented by in vitro analyses examining chemokines and receptors upregulated in LECs by specific stimuli (46, 60, 88). The emerging data suggests that chemokines and their receptors play a complex role in helping coordinate the movement of LECs and interactive circulatory cells in both normal development and a range of pathological conditions.

### AUTHOR CONTRIBUTIONS

RF and SS conceived the review. RF, TK, SiM, ScM, and SS wrote and edited the manuscript.

### FUNDING

The authors are funded by a Program grant from the Australian NHMRC (1053535), which supports the overall work of the Tumor Angiogenesis Program. SS is funded by an NHMRC Fellowship (487901).


by the collecting lymphatic endothelium. Cancer Cell. (2012) 21:181–95. doi: 10.1016/j.ccr.2011.12.026


induced by retinoic acid and neuronal stimulation. Nat Immunol. (2009) 10:1193–9. doi: 10.1038/ni.1789


lymphocyte egress and lymphatic patterning. J Exp Med. (2010) 207:17–27. doi: 10.1084/jem.20091619


**Conflict of Interest Statement:** SS has ownership interest in Opthea Ltd. that develops therapeutics in vascular biology.

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.

Copyright © 2019 Farnsworth, Karnezis, Maciburko, Mueller and Stacker. 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(s) 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.

# Comparative Transcriptomic Analysis Identifies a Range of Immunologically Related Functional Elaborations of Lymph Node Associated Lymphatic and Blood Endothelial Cells

Stella J. Berendam<sup>1</sup> , Alexander F. Koeppel <sup>2</sup> , Nicole R. Godfrey <sup>1</sup> , Sherin J. Rouhani <sup>1</sup> Amber N. Woods <sup>1</sup> , Anthony B. Rodriguez <sup>1</sup> , J. David Peske<sup>1</sup>

#### *United States* Reviewed by:

Edited by: *Beth Ann Tamburini, University of Colorado Denver,*

*Jason G. Cyster, University of California, San Francisco, United States Theresa T. Lu, Hospital for Special Surgery, United States Sirpa Jalkanen, University of Turku, Finland*

> \*Correspondence: *Victor H. Engelhard vhe@virginia.edu*

#### Specialty section:

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

Received: *15 December 2018* Accepted: *27 March 2019* Published: *16 April 2019*

#### Citation:

*Berendam SJ, Koeppel AF, Godfrey NR, Rouhani SJ, Woods AN, Rodriguez AB, Peske JD, Cummings KL, Turner SD and Engelhard VH (2019) Comparative Transcriptomic Analysis Identifies a Range of Immunologically Related Functional Elaborations of Lymph Node Associated Lymphatic and Blood Endothelial Cells. Front. Immunol. 10:816. doi: 10.3389/fimmu.2019.00816*

, , Kara L. Cummings <sup>1</sup> , Stephen D. Turner <sup>2</sup> and Victor H. Engelhard<sup>1</sup> \*

*<sup>1</sup> Department of Microbiology, Immunology, and Cancer Biology, Carter Immunology Center, University of Virginia School of Medicine, Charlottesville, VA, United States, <sup>2</sup> Department of Public Health Sciences and Bioinformatics Core, University of Virginia School of Medicine, Charlottesville, VA, United States*

Lymphatic and blood vessels are formed by specialized lymphatic endothelial cells (LEC) and blood endothelial cells (BEC), respectively. These endothelial populations not only form peripheral tissue vessels, but also critical supporting structures in secondary lymphoid organs, particularly the lymph node (LN). Lymph node LEC (LN-LEC) also have been shown to have important immunological functions that are not observed in LEC from tissue lymphatics. LN-LEC can maintain peripheral tolerance through direct presentation of self-antigen via MHC-I, leading to CD8 T cell deletion; and through transfer of self-antigen to dendritic cells for presentation via MHC-II, resulting in CD4 T cell anergy. LN-LEC also can capture and archive foreign antigens, transferring them to dendritic cells for maintenance of memory CD8 T cells. The molecular basis for these functional elaborations in LN-LEC remain largely unexplored, and it is also unclear whether blood endothelial cells in LN (LN-BEC) might express similar enhanced immunologic functionality. Here, we used RNA-Seq to compare the transcriptomic profiles of freshly isolated murine LEC and BEC from LN with one another and with freshly isolated LEC from the periphery (diaphragm). We show that LN-LEC, LN-BEC, and diaphragm LEC (D-LEC) are transcriptionally distinct from one another, demonstrating both lineage and tissue-specific functional specializations. Surprisingly, tissue microenvironment differences in gene expression profiles were more numerous than those determined by endothelial cell lineage specification. In this regard, both LN-localized endothelial cell populations show a variety of functional elaborations that suggest how they may function as antigen presenting cells, and also point to as yet unexplored roles in both positive and negative regulation of innate and adaptive immune responses. The present work has defined in depth gene expression differences that point to functional specializations of endothelial cell populations in different anatomical locations, but especially the LN. Beyond the analyses provided here, these data are a resource for future work to uncover mechanisms of endothelial cell functionality.

Keywords: endothelial cell, lymph node, lymphatic, RNA-Seq, antigen presentation, scavenger receptors, chemokines, cytokines and receptors

#### INTRODUCTION

Lymphatic and blood vessels are formed by specialized endothelial cells that are closely related but distinct (1). These endothelial populations form vessels in peripheral tissue, but also supporting structures in secondary lymphoid organs, particularly lymph node (LN). Blood endothelial cells (BEC) form high endothelial venules, which control the entry of lymphocytes from the bloodstream, while lymphatic endothelial cells (LEC) form lymphatic sinuses that control entry of tissue-localized immune cells, and organization and exit of all immune cells, in addition to the flow of lymph. To determine the basis for these functional attributes, several studies have evaluated transcriptomes of LEC and BEC, primarily from peripheral tissue vessels. Most of these have used microarray approaches and often relied on endothelial cells cultured in vitro (1–11), (see also EndoDB (12) for a comprehensive listing of prior studies, associated databases, and analysis tools). While they have revealed differences in LEC and BEC in genes implicated in vascular tube formation, transport of solutes, and immune cell trafficking, microarray hybridizationbased approaches posed several limitations, including high background levels and limited range of detection. Furthermore, these studies also concluded that even short-term primary cultures of LEC and BEC ex vivo resulted in some level of de-differentiation. Additionally, these studies used cells isolated from the skin and did not compare LEC and BEC from different anatomical sites. Analysis of transcriptional programs to understand the functionality and diversity of LEC and BEC in different anatomical locations remains to be done.

Recent studies have demonstrated that LN-associated LEC (LN-LEC) also actively participate in controlling innate and adaptive immune responses. We previously demonstrated that LN-LEC, but not LEC in tissue lymphatics, adventitiously expressed transcripts for proteins otherwise restricted to a small number of peripheral tissues. We showed that a peptide epitope from one of these, the melanocyte protein tyrosinase (Tyr), was presented on LN-LEC associated MHC-I molecules to Tyr-specific CD8 T cells (13–15). Although this induced activation and proliferation, LN-LEC also expressed high levels of PD-L1 that resulted in deletion of Tyr-specific CD8 T cells (15). LEC from tissue lymphatics express negligible levels of PD-L1 (14). In a separate study, we established that LN-LEC could induce Lag3 dependent CD8 T cell deletion via expression of MHC-II molecules, and that LEC from tissue lymphatics express negligible levels of MHC-II (16). While LN-LEC were incapable of presenting acquired Ag via these MHC-II molecules, they nonetheless transferred endogenous antigens to dendritic cells (DC) for presentation to CD4 T cells, resulting in anergy (16). These results point to an important role for LN-LEC in establishing systemic peripheral T cell tolerance. Conversely, others have shown that LN-LEC capture and archive exogenous antigens that induce antigen-specific memory CD8 T cell persistence (17). This occurs via transfer of LEC-archived antigens to migratory DC as a result of LEC apoptosis during LN contraction and also via direct exchange of archived antigens by the two cell types (18). The molecular mechanisms involved in these different processes of antigen acquisition, expression, and transfer by LN-LEC remain unclear, and the specific microenvironmental influences that control the phenotypic as well as functional distinctions between LEC in the LN and in the periphery remain to be fully understood.

In this study, we address these issues, as well as the technical limitations of previous studies, by using RNA-Seq analysis to compare the transcriptomes of freshly isolated murine LNassociated LEC and BEC (LN-BEC) as well as freshly isolated LEC from the diaphragm (D-LEC) as representative of peripheral tissue lymphatics. RNA-Seq has greatly improved the analysis of whole transcriptomes with higher sensitivity and dynamic range coupled to lower technical variations compared to microarrays and quantitative PCR (19, 20). Our work provides an important resource for further exploration of endothelial cell functionality in different anatomical locations.

### RESULTS AND DISCUSSION LN-LEC, LN-BEC, and D-LEC Are Transcriptionally Distinct

LEC and BEC populations were purified from relevant tissues using magnetic bead enrichment and electronic cell sorting from 10 to 33 C57BL/6 mice for each replicate sample, and subjected to RNA-Seq (**Figures S1A,B**). This yielded 48–98 million reads per replicate, with an average length of 180 nucleotides, and an average of 85.7% uniquely mapped reads. These reads mapped a total of 23,284 genes. One previous study estimated that one transcript copy per liver cell corresponds to 3 FPKM (21), while another estimated that genes expressed at FPKM > 1 were reproducibly and accurately detected in bulk RNA-Seq experiments (22). Based on this, we identified

**Abbreviations:** BEC, blood endothelial cells; LEC, lymphatic endothelial cells; LN, lymph node; LN-BEC, lymph node-associated blood endothelial cells; LN-LEC, lymph node-associated lymphatic endothelial cells; D-LEC, lymphatic endothelial cells from diaphragm; DC, dendritic cells; DEG, genes differentially expressed between any two cell types; 5X-DEG, genes whose differential expression in pairwise comparisons was greater than 5-fold; GO, Gene Ontology; ECM, extracellular matrix.

genes with FPKM ≥ 1 and p-adjusted < 0.05 in all replicate comparisons, which ensures that low level FPKM values are consistent. This gave a total of 15,331 genes considered to be expressed in at least one cell population (**Table S1**). Similar gene numbers were expressed in LN-LEC, LN-BEC, and D-LEC, respectively (**Figure S1C**). Principal component analysis revealed that the transcriptional profiles of replicates clustered tightly, and LN-LEC, LN-BEC, and D-LEC differed from each other (**Figure S2A**). The pan-endothelial marker (CD31) was strongly expressed in all endothelial cell populations (**Table 1**). Established markers of LEC (LYVE-1, PDPN, PROX-1, and RELN) were strongly expressed in LN-LEC and D-LEC with minimal (1.0–3.4%) cross expression by LN-BEC (**Table 1**). Established markers of BEC (NRP-1, VEGFR-1, VWF, and NOTCH-4) were strongly expressed in LN-BEC with minimal cross expression in LN-LEC and D-LEC (0.2–4.7%). The low levels of cross expression of these genes are consistent with very low cross-contamination or genuine low-level expression. Known markers of fibroblast reticular cells and hematopoietic subpopulations were evident only at very low to negligible levels (**Table 1**). Consistent with our previous findings (13, 14), Tyr was expressed by LN-LEC but not LN-BEC and D-LEC, while (PD-L1) was expressed at high levels in LN-LEC and LN-BEC but not D-LEC. These data established a high level of confidence in further analyzing gene expression patterns of that differ among LN-LEC, LN-BEC, and D-LEC.

#### Differential Gene Analyses Reveal Subsets of Genes Specific Only to LN-LEC, LN-BEC, or D-LEC, and Subsets of Genes Shared by at Least Two Cell Populations

Genes that were differentially expressed between any two cell types (DEG) were identified based on an adjusted p < 0.05. Comparisons of LN-LEC vs. D-LEC, LN-LEC vs. LN-BEC, and D-LEC vs. LN-BEC identified 7210, 6109, and 6994 DEG, respectively (**Figure 1A**). We next identified genes whose differential expression in these pairwise comparisons was greater than 5-fold (5X-DEG). A total of 1512, 1634, and 937 5X-DEG were overexpressed in LN-LEC, LN-BEC, and D-LEC respectively, accounting for a total of 3137 unique 5X-DEG (**Figures 1A,B**). Since the total expressed genes in these populations were similar (**Figure S1C**), the substantially higher numbers of 5X-DEG in LN-BEC and LN-LEC relative to D-LEC suggests that the two LN populations have more elaborated functionalities.

Hierarchical clustering identified subsets of 5X-DEG distinct to only LN-LEC, LN-BEC, and D-LEC, and subsets shared by two cell types: LN-LEC+LN-BEC, LN-LEC+D-LEC, and D-LEC+LN-BEC (**Figure 1C**; **Table S2**). There were relatively few 5X-DEG in the D-LEC+LN-BEC shared subset, consistent with the distinct developmental origins and anatomical locations of these two populations. Intriguingly, the LN-LEC+LN-BEC shared subset contained 3.4 times more 5X-DEG than the LN-LEC+D-LEC shared subset. Thus, despite their different developmental origins, the two LN-localized endothelial populations are more transcriptionally related to TABLE 1 | RNA-seq validation of stromal cell-specific markers and hematopoietic cell-lineage markers based on normalized gene expression levels (FPKM).


*Genes and FPKM values highlighted in bold represent previously identified lineage and phenotypic markers associated with the cell types.*

one another than the two LEC populations that occupy distinct anatomical niches.

We used GOrilla software to identify biological process and molecular function Gene Ontology (GO) terms in each 5X-DEG subset that were highly ranked based on enrichment score (see Methods), which emphasizes co-expression of multiple genes associated with a term, rather than overexpression of individual genes. We identified GO terms with significant (p-adjusted < 0.001) enrichment scores in all 5X-DEG subsets except D-LEC+LN-BEC (**Figure S2B**), and this subset was thus excluded from the analyses below. These GO terms were often interrelated and were further grouped into clusters based on visual inspection (**Table S3**; **Figure S2C**). These clusters, and the overexpressed genes that they contained, are discussed in more details in sections below.

### Differential Expression of Extracellular Matrix Components and Cell Adhesion Molecules Suggest Specialized Structural and Functional Attributes of LEC and BEC in Distinct Tissue Microenvironments

GO terms related to extracellular matrix (ECM) were highly ranked in all 5 5X-DEG subsets and identified overexpressed

subsets by at least two populations. (A) Venn diagrams showing pairwise comparisons of LN-LEC, LN-BEC, and D-LEC. (B) Total number of unique and cell type specific 5X-DEG. (C) Hierarchical clustering of distinct and shared subsets of 5X-DEG. Complete lists of 5X-DEGs in each subset are listed in Table S2.

genes in several different processes and functions. Each subset overexpressed different collagen molecules while 3 laminin family members were overexpressed in one of the two LN endothelial cell populations, and 4 fibronectin family members were overexpressed in one of the two LEC populations (**Figure 2**). Members of the tenascin, thrombospondin, elastin, and proteoglycan families were overexpressed in one or both of the two LEC populations, but none were overexpressed in LN-BEC. Similarly, ECM remodeling enzymes of the MMP, ADAM, and LOXL families were widely overexpressed in the LEC subpopulations, but minimally in LN-BEC. These data point to an elaboration of ECM components in LEC compared to BEC, and also suggest that LEC in different anatomical locations create distinct ECM microenvironments through both synthesis and remodeling activities. These may contribute to distinct structural and functional attributes of adjacent lumenal and ablumenal compartments.

GO terms related to cell adhesion were also highly ranked in all 5 5X-DEG subsets. Each of the 5 subsets overexpressed different integrin molecules (**Figure 2**). The LN-LEC, LN-BEC, D-LEC, and LN-LEC+LN-BEC also overexpressed different cadherin and cadherin-like family members and cell adhesion molecules (CAMs), known to mediate homophilic adhesion of endothelial population of common lineage and origin. No integrins or cadherin family members were overexpressed in the LN-LEC+D-LEC subset, suggesting a significant distinction in the patterns of cell engagement by these two LEC subpopulations. The leukocyte CAMs, all of which play well-established roles in mediating the extravasation of cells into lymphoid and peripheral tissues, were overexpressed in the LN-BEC only, but also LN-LEC+LN-BEC 5X-DEG subsets. Their function in LN-LEC remains to be established. Claudin and catenin family members were also overexpressed almost exclusively in the LN endothelial populations, with only a single catenin gene overexpressed in D-LEC associated subset. Taken together, these data suggest that LN endothelial subpopulations are endowed with an enriched capacity for interactions with a diversity of other cells relative to D-LEC.

Conversely, GO terms related to cytoskeleton were highly ranked in the D-LEC only 5X-DEG subset, and to a lesser extent, the LN-LEC+D-LEC subset. These subsets contained a variety of overexpressed genes encoding cytoskeletal proteins and binding molecules (**Figure 3**). The enhanced expression of these cytoskeletal proteins and binding molecules are likely attributes of LEC in peripheral tissue lymphatics that allow them to maintain shape in the face of ECM-mediated displacement during body movement.

### Chemokine Expression Patterns Suggest a Collaborative Division of Labor Between LEC and BEC in Maintaining Spatial Organization and Compartmentalization of Cells in LN

GO terms related to chemokines were highly ranked in the LNassociated 5X-DEG subsets, but not those associated with D-LEC. Nonetheless, the patterns of chemokine expression in the LNassociated subsets revealed a surprising degree of complexity. The homeostatic chemokines CCL19 and CCL21 are two CCR7 ligands that have been implicated in homing of multiple immune cell subsets to LN via blood and lymph and organizing the T-cell zone of secondary lymphoid organs. CCL19 was overexpressed in the LN-LEC+LN-BEC subset (**Figure 4**), while CCL21 was not expressed at all in any analyzed EC population (<1 FPKM).

While this is at odds with previous studies (10), we also confirmed using Q-PCR that CCL21 expression in LN-LEC was negligible compared to expression in bulk LN (**Figure S3**). Previous studies demonstrated the preferential ability of CCL19 to recruit CCR7<sup>+</sup> cells compared to CCL21 (23–26) and that CCL19 signaling blocks directed migration of CCR7<sup>+</sup> cells toward weak CCL21 signal (27). Our results suggest that autocrine secretion of CCL19 by LEC and BEC may play a role in organizing CCR7<sup>+</sup> cells in the face of distinct gradients of CCL21.

The homeostatic chemokine CXCL13, a ligand for CXCR5 that organizes the B-cell zone of secondary lymphoid organs, was

FIGURE 4 | 5X-DEG revealed immunomodulatory role of LEC and BEC via cytokines and innate effector molecules production. Heatmap analysis based on Z-score values of average log2 FPKM for replicates in each cell type (details in Methods).

also overexpressed in the LN-LEC+LN-BEC subset (**Figure 4**). A previous study reported CXCL13 expression in LEC but not BEC isolated from peripheral LN (28), while another reported that it is highly expressed on HEV from Peyer's Patches (29). Because our LN samples were pooled from both peripheral and mesenteric LN, it is possible that LN-BEC in mesenteric LN may be similar to those in Peyer's Patches due to their close anatomical proximity. Nonetheless, insofar as CXCL13 appears to be essential for organization of LN but not B-cell entry (30, 31), our results suggest that both LN-LEC and LN-BEC have the potential to influence this process.

CXCL12, a chemokine that plays multiple roles but is particularly involved in homing of both T- and B-cells into LN (29, 32–34), was overexpressed only in LN-BEC (**Figure 4**). This is in keeping with its principle function in homing, as opposed to organization, of LN. LN-BEC also selectively expressed CX3CL1, a membrane associated chemokine that binds to CX3CR1<sup>+</sup> cells. These include memory CD8 T-cells that reside in the LNsubcapsular cortex (35), and efferocytic T-zone macrophages (TZM) (36). There is no current evidence demonstrating that CX3CL1 mediates recruitment or localization of these cells in LN. To the contrary, CX3CR1 deficient TZM localize in normal numbers, but are deficient in clearance of apoptotic cells (36) because CX3CL1 also functions as an "eat me" signal (37) Nonetheless, the selective expression of CX3CL1 in LN-BEC is not entirely consistent with this role, and suggests that it may play a role in recruitment of cells to LN.

In contrast, LN-LEC overexpressed another membranebound chemokine CXCL16, which has activities as a chemoattractant and a scavenger receptor (**Figure 4**). As a chemoattractant it binds to CXCR6, which is expressed on activated CD8 and CD4 T cells (38–42). CXCL16 expressed in LN fibroblastic reticular cells was shown to mediate migration and mild adhesion of CXCR6<sup>+</sup> CD8 and CD4 T cells (43). CXCL16 expression by LN-LEC could function similarly, and could mediate immune regulation of activated T cells along with other molecules such as PD-L1 and Lag3 (15, 16).

Perhaps surprisingly, the LN-LEC+LN-BEC, LN-LEC, and LN-BEC subsets each overexpressed several proinflammatory chemokines, which collectively support recruitment of a variety of immune cells, many of which are not resident in resting SLO (**Figure 4**). CXCL1 is essential for neutrophil migration and neutrophil extracellular trap formation (44), while CCL5 mediates recruitment of leukocytes expressing its cognate receptor, CCR5 (45).However, D-LEC do not overexpress any similar chemokines. Given the pervasiveness of blood and lymphatic vessels in the LN, the shared expression of these chemoattractant molecules seems consistent with a role in recruitment of T-cells and DC on the one hand and Bcells on the other. However, it is also conceivable that only well-localized subpopulations of each endothelial cell type express either chemokine, enabling them to participate in organizational processes.

CXCR3 is a chemokine receptor that is widely expressed on activated and memory type I CD4 and CD8 T cells. Interestingly, LN-LEC and LN-BEC both overexpress one CXCR3 ligand, CXCL9, while the two others, CXCL10 and CXCL11, are overexpressed only in LN-LEC or LN-BEC, respectively, although CXCL11 is a pseudogene in C57Bl/6 mice (**Figure 4**). CCR5 is a chemokine receptor with a similar expression pattern on T cells, and additionally on macrophages and dendritic cells, and two of its ligands, CCL5 and CCL3, are overexpressed by LN-LEC+LN-BEC and LN-LEC subsets, respectively (**Figure 4**). This suggests a subtle interplay between LN-LEC and LN-BEC in fine-tuning organization and movement of different antigen experienced and antigen presenting cells in the LN.

In keeping with this idea, we also found overexpression of atypical chemokine receptors in LN-LEC, LN-BEC, D-LEC, and LN-LEC+LN-BEC subsets, but not the LN-LEC+D-LEC subset (**Figure 5**). These molecules typically function as decoy receptors to create chemoattractant gradients through chemokine sequestration. The LN-LEC+LN-BEC subset overexpressed ACKR5, which binds to CCL19. Since this same subset overexpresses CCL19, this may suggest that expression of the chemokine and the decoy receptor differs based on precise location within the LN. LN-LEC and D-LEC respectively overexpressed the atypical chemokine receptors, ACKR4 and ACKR3, which bind to CCL21, and CXCL11 and CXCL12, respectively. LN-BEC overexpressed ACKR1, which binds to CCL2, CCL5, CXCL1, CXCL4, CXCL5, CXCL7, and CXCL8. These data suggest that the expression of these chemokine decoy receptors may further augment cooperative interplay of LEC and BEC in controlling chemokine gradients in LN to promote directional sensing, migration, and activation of immune cells.

#### TNF and TNFR Superfamily Expression Patterns Suggest Overlapping but Distinct Involvement of LEC and BEC in Maintenance of LN Microarchitecture and Involvement in Autocrine and Paracrine Signaling Mechanisms

GO terms related to TNF and TNFR superfamily members were also highly ranked in the LN-associated 5X-DEG subsets, but not those associated with D-LEC. The LN-BEC 5X-DEG subset overexpressed TNFR2 (TNFRSF1B) (**Figure 5**). TNFR1 (TNFRSF1A) and LTβR were overexpressed at similar levels in LN-LEC and LN-BEC relative to D-LEC, but the fold change was <5 (GSE119499). However, LN-LEC overexpressed TNFα and LTβ, while LN-BEC overexpressed LTα (**Figure 4**). Since LTα can form either a homotrimer that binds TNFRs or a heterotrimer with LTβ that binds LTβR, this suggests that these two populations differ in expression of these alternative forms, possibly leading to differences in autocrine or paracrine signaling that could influence lymphoid tissue microarchitecture. It is well-established that LTβR signaling is required form homeostatic maintenance of HEV phenotype, including expression of PNAd and MadCAM-1 (46) and one major source of LTα1β2 is dendritic cells (47). Conversely, LTβR signaling plays a role in expression of homeostatic chemokines (48) and dendritic cell maturation (49, 50). Our data suggest that dendritic cells influence LN-LEC and LN-BEC phenotypes via the LTβR signaling pathway, and vice versa. Additionally,

the LN-LEC+LN-BEC 5X-DEG subset overexpressed RankL (**Figure 4**). The roles of these molecules in LN development and maintenance are well-established (51–56), but the specific involvement of endothelial cells as either initiators or recipients of TNF-related signaling has not been well-described.

Other overexpressed TNF and TNFR superfamily members have been associated with induction of cell death. The LN-LEC+LN-BEC subset overexpressed TRAIL (TNSFSF10), a TNF-superfamily ligand known to induce death of activated cells that express TRAIL receptors (57–60) (**Figure 4**). However, we did not detect pro-apoptotic TNFRSF10B (TRAIL-R2) expression in either LN endothelial subset. Instead, they expressed a decoy receptor for the ligand, TNFRSF23 (mDCTRAILR1) (61) (**Figure 5**). These data suggest that LEC and BEC in the LN induce TRAIL-mediated apoptosis of other cell types while protecting themselves. Conversely, the LN-LEC+LN-BEC 5X-DEG subset overexpressed TNFRSF19L (RELT). RELT has been shown to induce cellular cell death in multiple cell types via a mechanism distinct from TNFR1 (62). A ligand for RELT has not been identified (63). It is intriguing to consider whether RELT represents a mechanism by which inflammation driven LN angiogenesis and lymphangiogenesis might be downregulated after resolution.

Several overexpressed TNF and TNFR superfamily members have costimulatory or survival promoting functions. LN-LEC+LN-BEC overexpress TNFSF15 (VEGI, TL1A) (**Figure 4**), which acts as a T cell co-stimulator to induce a variety of distinct T cell subsets and immunopathologies (64–69) and promotes DC maturation (70). It also inhibits expression of VEGFR1 and induces endothelial apoptosis to inhibit vasculogenesis, but promotes lymphangiogenesis (71–74). LN-BEC selectively overexpress TNFSF18 (GITRL) (**Figure 4**), which co-stimulates both effector and regulatory T cells (75–77). Interestingly, LN-BEC also selectively overexpress TNFRSF9 (4-1BB), which could render them susceptible to signals delivered by 4-1BBL<sup>+</sup> cells such as DC, and TNFRSF13C (BAFFR), which promotes Bcell survival and isotype switching (78, 79) (**Figure 5**). Another receptorfor BAFF, TNFRSF17 (BCMA), which promotes survival of long-lived plasma cells (80) was overexpressed in LN-LEC (**Figure 5**). The impact of signals delivered by these receptors on endothelial function is unknown.

#### Selective Expression of Multiple TGFβ-Superfamily Members Suggests That LN-LEC Contribute to Immunosuppressive Functions in Homeostasis

GO terms related to the TGFβ-superfamily were highly ranked in the LN-LEC+D-LEC and LN-LEC 5X-DEG subsets, but not in the D-LEC, LN-BEC, and LN-LEC+LN-BEC subsets. LN-LEC+D-LEC overexpressed TGFβ3 and BMP7, while LN-LEC overexpressed BMP2, BMP3, BMP3B, and BMP8A (**Figure 4**). TGFβ3 is highly homologous to TGFβ1 and TGFβ2, which were comparably expressed in all three endothelial populations. However, TGFβ3 binds more potently to TGFβ receptors I (TGFBR1/ALK-5) and II (TGFBR2) (81–83). TGFβ3 plays similar roles in immunosuppression and stimulation as TGFβ1 (84, 85), TGFβ3, BMP7, and BMP2 have been shown to suppress survival, proliferation, differentiation of in vitro grown human B cells into antibody-secreting cells (86–88). BMP2 inhibits T cell proliferation (89) and promotes Foxp3<sup>+</sup> Treg generation in the context of TGFβ treatment (90). Little is known about the immunological activity of the other BMPs remaining molecules expressed in LN-LEC. Nonetheless, expression of

these molecules may be associated with previously described tolerogenic properties of LN-LEC (13, 15, 16, 91).

### Expression of Cytokines and Innate Effector Molecules Suggests Additional Immunomodulatory Roles of LN-Localized LEC and BEC

GO terms for several additional cytokines and growth factors were highly ranked, almost exclusively in LN-associated 5X-DEG subsets. These included the common γ-chain cytokines, IL-7 and IL-15, which were overexpressed LN-LEC+D-LEC and LN-BEC, respectively (**Figure 4**). These observations corroborate earlier work (7), and suggest a division of labor between LEC and BEC in maintaining IL-7Rα <sup>+</sup> and IL-15Rα <sup>+</sup> cells in LN. They also point to D-LEC as a source of IL-7 for homeostatic T cell maintenance. LN-LEC+D-LEC also overexpressed IL-18, while LN-LEC expressed IL-33, both members of the IL-1 family. IL-18 synergizes with IL-7 in activation and priming of naïve CD8 T cells (92), in that IL-7 upregulates IL-18R. Finally, LN-LEC+LN-BEC and LN-LEC subsets overexpressed IL-12a (p35) and EBI3, respectively. IL-12a (p35) pairs with EBI3 to form IL-35 (93, 94), which has been demonstrated to induce inducible regulatory T cells (iTregs) (95), and suppress T cell proliferation (96). The LN-LEC+LN-BEC subset overexpressed KITL, a ligand for the cell surface tyrosine kinase KIT found on lymphocytes and hematopoietic stem cells, while LN-LEC overexpressed CSF1 and CSF2, ligands for CSFR1 and CSFR2 receptors expressed on macrophages. Collectively, these patterns of expression reinforce the expansive roles played by both LN endothelial populations, and particularly LN-LEC, in promoting the survival of different immune subpopulations, and in providing a context for their differentiation.

Consistent with earlier work (7), LN-LEC expressed the proangiogenic factor, VEGFA, while LN-BEC expressed the lymphangiogenic factor, VEGFC (**Figure 4**). This creates the possibility that these two cell types could cross-regulate one another. LEC and BEC also expressed several members of the PDGF family (**Figure 4**), which could support fibroblastic reticular cells. These data point to a dynamic co-dependence among different stromal cell types that may regulate the balance between cell populations under steady state and inflammatory conditions.

LN-LEC and D-LEC subsets, respectively overexpressed IGF1 and IGF2, distinct members of the insulin-like growth factor (IGF) family (**Figure 4**), while transcripts for IGF-family receptors (IGFR1 and IGFR2) were detected but not differentially expressed in all 3 endothelial populations. These 3 populations collectively overexpressed most of the IGF binding proteins (IGFBP), which bind to IGFs and modulate their activity in distinct ways (97–100). Together, these data demonstrated LEC in different anatomical niches control the local tissue milieus to support cellular growth, differentiation, and function via intricate networks of cellular IGF1- and IGF2-signaling and counter balance mechanisms to maintain tissue homeostasis.

GO terms associated with innate host defense mechanisms were also highly ranked in the LN-LEC+LN-BEC 5X-DEG subset, corresponding to a range of genes with antimicrobial activities to viral, bacterial, and fungal organisms (**Figure 4**). This suggests an as yet unappreciated role for LN endothelial populations to prevent pathogen dissemination. In addition, the LN-LEC 5X-DEG subset overexpressed a number of molecules associated with classical and non-classical complement cascades. These included C2, C3, and C5, which participate directly in the proteolytic cascade. The resulting products include the anaphylatoxins C3a and C5a, which serve as chemoattractant for neutrophils, monocytes, and macrophages (101–103), and modulate the functions of APCs and T cells (104–106), and C3b, which binds to pathogens, immune complexes, and apoptotic cells to promote phagocytosis (107). These data suggest that LN-LEC may collaborate with subcapsular sinus macrophages, follicular dendritic cells, and B cells to promote both innate and adaptive immune responses through complement component secretion. LN-LEC also overexpressed of CD55 (DAF), DAF2, and CD59a, all of which prevent formation of the membrane attack complex and enable LN-LEC to protect themselves from the actions of the products they secrete.

#### Expression of Cytokine Receptors and Pathogen Sensing Molecules Suggests Additional Immunosensory Roles of LN-Localized LEC and BEC

GO terms for several cytokine receptors and pathogen sensing molecules were highly ranked, again almost exclusively in LNassociated 5X-DEG subsets. LN-LEC+LN-BEC, LN-BEC, and most prominently, LN-LEC, overexpressed members of the type I cytokine receptor superfamily, including components of the IL-1, IL12, IL18, IL27 receptors, and antagonists and decoys (**Figure 5**). LN-LEC and LN-BEC also overexpressed receptors for several other immune relevant molecules, including IL10, sphingosine-1-phosphate, and prostaglandins. These data indicate that the LN endothelial cells are poised to sense and respond to a variety of cytokine cues in their local milieu, although the consequences of signaling by any given receptor remain to be established.

LN-LEC, LN-BEC, and LN-LEC+LN-BEC overexpressed several toll-like receptors (TLR) and NOD-like receptors (NLR). LN-LEC+LN-BEC overexpressed TLR4, while several other TLR were selectively overexpressed by LN-BEC. The patterns of NLR overexpression were more complex. As with the immune receptors above the consequences of signaling by any given TLR or NLR remain to be established. Interestingly, however, LN-BEC overexpressed NLRP6, which inhibits inflammasome formation (108, 109). The LN-LEC+LN-BEC 5X-DEG subset overexpressed several ligands for the NKG2D receptor that is expressed on NK, NKT, γδ T cells, and activated CD8 T cells (**Figure 5**). While these ligands are generally associated with promoting effector activity via NKG2D signaling, RAET1E expressed on endothelial cells was demonstrated to inhibit NK cell activation by inducing NKG2D internalization (110). This raises the question of whether the responses of LN-LEC and LN-BEC to prototypical pro-immune receptor signaling may be counter-regulatory.

### LN-LEC and LN-BEC Overexpress Molecules Involved in MHC-I and MHC-II Antigen Processing and Presentation

GO terms for MHC-I and MHC-II antigen processing and presentation pathways were highly ranked exclusively in the LN-LEC+LN-BEC 5X-DEG subset. Overexpressed MHC-I pathway components included H-2K<sup>b</sup> , β2 microglobulin, several Qa and Cd1d molecules, TAP1 and 2 components of the immunoproteasome (**Figure 6**). Overexpressed MHC-II pathway components included H-2Aβ, invariant chain, peptide editors (H-2Oα, H-2DMβ2, H-2DMβ1), and cathepsins S and G. While not achieving 5X differential expression, the MHC-I H-2D<sup>b</sup> molecule, and the MHC-II components H-2Aα and H-2DMα were also overexpressed in the two LN cell populations relative to D-LEC (GSE119499). Cathepsin L was also overexpressed by 5X in LN-LEC relative to D-LEC, but this did not give rise to an enriched GO term score (**Table S1**). These data reinforce previous studies from our lab demonstrating that LN-LEC efficiently present endogenous antigens via H-2K<sup>b</sup> to CD8 T cells (14–16), and suggest that LN-LEC have elevated capacity to present antigens via additional classical and non-classical MHC-I molecules. They also suggest that LN-BEC have a similar capacity. We have also reported that LN-LEC are unable to present endogenous or exogenous antigens to CD4 T cells despite expressing MHC-II molecules, and have suggested that this is due to a deficiency in H-2DM expression (16). While both H-2DMα and H-2DMβ were overexpressed in LN-LEC, their level of expression was still low (<100 FPKM), compared with FPKM values >1000 for H-2Aα, H-2Aβ, invariant chain, and cathepsins S and L, consistent with this earlier conclusion. We have concluded that the MHC-II molecules expressed on LN-LEC are primarily involved in engaging LAG-3 on T cells to induce peripheral tolerance (16). Given the similar expression of MHC-II components in LN-LEC and LN-BEC, we suggest that the latter cells serve a similar function.

### LN-LEC Express Elevated Number of Molecules Involved in Exogenous Material Acquisition That Potentially Contribute to Their Functions in Antigen Archival and Peripheral Tolerance

GO terms for receptor mediated endocytosis were highly ranked almost exclusively in the LN-LEC 5X-DEG subset. Overexpressed molecules included C-type lectin receptors, scavenger receptors, and Fc receptors (**Figure 7A**). C-type lectin receptors have been categorized as binding to either carbohydrate, noncarbohydrate structures, or both. LN-LEC overexpressed some C-type lectins that bind carbohydrates (CLEC4A3, CLEC4D, CLEC4E, CLEC4G, CD209D) and others that bind noncarbohydrates (CLEC1A, CLEC1B, CLEC9A). We found a single C-type lectin (CLEC4G) in the LN-LEC+LN-BEC, although this subset was not enriched for GO terms associated with receptor mediated endocytosis. C-type lectin receptors also can signal via immune tyrosine activation or inhibitory motifs, or through noncanonical structural features that mediate positive or negative

immune stimuli. C-type lectin receptors in the LN-LEC include all 3 of these signaling categories.

The scavenger receptors MSR-1 (SR-A1), MARCO (SR-A6), STAB-2 (SR-H2), and CXCL16 (SR-G1) were also overexpressed in LN-LEC (**Figure 7A**). Previous studies demonstrated that the class A scavenger receptors, MSR-1 and MARCO, are expressed primarily on macrophage subpopulations, and are associated with recognition of surface molecules of Grampositive and -negative bacteria (111, 112), modified and oxidized LDL (113, 114), hepatitis C virus (115), β-amyloid (116), and heat shock proteins (117). The expression of MSR-1 and MARCO under steady state condition was reported to be restricted to macrophages in the LN and the marginal zones of the spleen (118). Our data extends this. In addition, we demonstrated by immunofluorescence and flow cytometry that subpopulations of LN-LEC express MARCO and MSR-1, while LN-BEC, fibroblastic reticular cells (FRC), and other CD45neg LN stromal cell populations do not (**Figures 7B,C**). Membrane-bound CXCL16 and STAB-2 bind phosphatidylserine and oxidized lipids (119–123), and membrane-bound CXCL16 also mediates phagocytosis of bacteria (124). LN-LEC also overexpressed the Fc receptor, FCGR2B, which has been shown to be essential for internalization of immune complexes by DC (125–127).

Again, while the subset did not have high GO enrichment scores, LN-LEC+LN-BEC also expressed two scavenger receptors, CD36 (SR-B2) and SCARF-1 (SR-F1). SCARF-1 and CD36 have been previously reported as pattern recognition molecules for fungal pathogens in innate immunity (128). More recently SCARF-1 expressed by splenic DC, macrophages, and

activity. (A) Heatmap analysis based on Z-score values of average log2 FPKM for replicates in each cell type (details in Methods). (B) Flow cytometry analyses of LN stromal cell populations for expression of Marco and Msr-1. (C) Immunofluorescent (IF) staining in adjacent tissue sections of LN for detection of Marco and Msr-1 co-expression with Lyve-1 (LEC marker).

endothelial cell was shown to bind the complement molecule, C1q to mediate apoptotic cell clearance, thus preventing generation of autoantibodies to DNA-containing antigens that lead to lupus-like disease and autoimmunity (129). CD36 was reported to facilitate transfer of surface antigen between CD8α + DC and mTEC to promote tolerance to self-antigens during T cell development (130).

Collectively, these data point to a previously undescribed but comprehensive capability of LN-LEC to internalize a broad array of extracellular materials using a variety of pattern recognition elements, either alone or in conjunction with other immune recognition molecules. We suggest that many of these material may be delivered into antigen processing and presentation pathways, at least for MHC-I molecules (13, 16). Our previous work indicates that the MHC-II pathway is non-functional in LN-LEC in the steady state (16), but it remains possible that this changes under conditions of pathogen exposure and inflammation. An important question is whether this is primarily a means of generating tolerance to exogenous self-antigens, or whether LN-LEC may sometimes also serve as accessory antigen presenting cells during an active immune response. In addition, this internalization capability is likely to be important in the antigen acquisition and archiving functions of LN-LEC (17, 18).

#### Attempting to Identify Peripheral Tissue Antigens Expressed by LN-LEC

We previously demonstrated that LN-LEC adventitiously expressed transcripts for proteins otherwise restricted to a small number of non-hematopoietic, non-endothelial peripheral tissues, typified by Tyr (13, 14). Because expression of these molecules is not dependent on Aire, as is the case for medullary thymic epithelial cells (131), it has not been possible to characterize the full range of peripheral tissue antigens potentially displayed by LN-LEC. With the expansion of available data on tissue specific expression, it has also become more difficult to unambiguously identify genes whose expression is rigorously limited to only a small number of tissues. Using the characteristics of Tyr expression (FPKM<343), we hypothesized that there would be an elevated number of transcripts at or below this expression level in LN-LEC compared to D-LEC. However, we found that these two subsets contained equal number of 5X-DEG that met this criterion (**Table S4**). Thus, although the cut-off criteria based on Tyr expression is a good first step, a more comprehensive approach is needed to identify the candidate peripheral tissues antigens expressed by LN-LEC.

### Conclusions

This study provides comprehensive comparative transcriptomic analyses of LN-LEC, LN-BEC, and D-LEC and has defined in detail gene expression differences that point to functional specializations of EC in different anatomical locations. Our goal was to provide a broad compendium of gene expression differences based on anatomic location and endothelial lineage, with a focus on genes of immunological interest. We believe that this information will be a significant and extremely useful resource for many workers in this field. Our data identify significantly expanded cohorts of immunologically significant genes either shared by LN-localized ECs, or expressed distinctly by one or the other LN-localized subset. These genes extend the understanding of both populations as regulators, not only of hematopoetic cell trafficking, but also cellular differentiation. They also point to an emerging understanding that LNlocalized EC express a variety of receptors that enable them to sense immunologically relevant changes in their environment. Further exploration of the consequences of this sensing on EC proliferation, differentiation, and function in immunoregulation is an important area for further work. This study clarifies that both LN-localized EC function as antigen presenting cells, and this issue explored in somewhat greater detail elsewhere (Santambrogio et al., Manuscript Submitted<sup>1</sup> ). Finally, this study highlights the surprising number of molecules involved in uptake of exogenous materials expressed distinctly by LN-LEC. The involvement of these molecules in enabling antigen archiving and peripheral tolerance to exogenous self- antigens is another rich area for further exploration.

While our comparative analysis identifies profound differences in bulk populations that are based on anatomical location, it does not address the almost certain heterogeneity that exists at the single cell level in each location, some of which has been pointed to in our own previous work (14). However, given the limited depth of single cell coverage, many of the differences we have identified might not have been immediately evident with that approach. Nonetheless, the results presented here provide a springboard for further work to establish the existence of heterogeneity in expression within LN anatomical niches using a variety of technical approaches. Taken together, the comparative gene expression profiles provided here would be useful resources for future work to uncover novel mechanisms of endothelial functionality and specialization in peripheral tissues and LN.

## MATERIALS AND METHODS

#### Mice

C57BL/6 mice (6–8 weeks of age) were purchased from NCI and were housed at pathogen-free facilities at the University of Virginia. All procedures were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and were approved by the University of Virginia Animal Care and Use Committee.

### LEC and BEC Isolation and Cell Sorting

Inguinal, axillary, brachial, cervical, and mesenteric LN were harvested, pooled, mechanically disrupted, and enzymatically digested for 15 min, followed by MACS bead depletion of CD45<sup>+</sup> cells as previously described (16). Diaphragm tissues were treated in the same way. CD45neg cells were electronically sorted based on absence of expression of CD45 (eBioscience, clone 30-F11), expression of pan-endothelial marker, CD31 (eBioscience, clone 390), and presence or absence of PDPN (Biolegend, clone 8.1.1) to distinguish LEC from BEC. Cells were collected in RNA Protect (Qiagen).

### RNA Extraction, cDNA Library Construction, and Sequencing

Total RNA was purified using RNAeasy mini kit (Qiagen) per manufacturer's instructions. cDNA library preparation and sequencing were performed by the Genomic Services Laboratory at Hudson Alpha, USA. Briefly, purified total RNAs (RIN score of 7.0 or higher) were prepared for sequencing using the Ovation RNA-Seq System V2 kit (Nugen) followed by RNA-Seq of 100 paired-end reads using the Illumina HiSeq 2500 v4 platform.

### Mapping, Quantification, and Differential Gene Analysis

Raw RNA-Seq read quality was assessed using FastQC (132) and low-quality regions were trimmed using Fastx-trimmer (http://hannonlab.cshl.edu/fastx\_toolkit/index.html). Cleaned reads were aligned to the mouse reference genome (build mm9) using STAR (133) and read counts on known mouse genes were calculated using featureCounts, part of the Subread package (134). Next, uniquely aligned reads were analyzed using the DEseq2 package in the R statistical computing environment (R Development Core Team, 2011, http://www.R-project. org/) to obtain normalized counts, estimate dispersion, and determine a negative binomial model for each gene. Principal

<sup>1</sup> Santambrogio L, Berendam SJ, Engelhard VH. The antigen processing and presentation machinery in lymphatic endothelial cells. Front Immunol. In Review.

Component Analysis was performed on rlog-transformed counts for quality assessment. Differentially expressed genes (DEG) were determined using DESeq2 and the Benjamini-Hochberg False Discovery Rate procedure was used to re-estimate the adjusted p-values. In our analyses, DEG were identified as those with an average FPKM of 1 or greater and replicate comparisons of p-adjusted < 0.05 in all cell types. 5X-DEG were identified as those with fold-change of 5 or greater. Hierarchical gene clustering analysis was performed using complete linkage and Euclidean distance as measure of similarity to display the DEG expression patterns.

#### Gene Ontology Analysis of 5X-DEG Subsets

Gene ontology analysis was performed with GOrilla software using the two ranked lists method (135). We used the 5X-DEG subsets as the target set and the all annotated genes from mouse reference genome (build mm9) as the background set. Briefly, GOrilla generates an enrichment score, which is the number of genes in the intersection of genes in the GO term (designated as B) and the number of genes in the target set (designated as b) for each associated GO term. We used the list of 5X-DEG for each subset as target. Enrichment of GO terms is then tested for statistical significance using a hypergeometric test and p-adjusted < 0.001 was considered as significant. Analyses were performed against gene ontologies: biological process and molecular function. We then identified the relationship between GO terms using hierarchical directed acyclic graph generated by GOrilla.

#### Immunofluorescence Staining and Flow Cytometry Analyses of LN-LEC

Immunofluorescence staining of adjacent LN tissue sections of C57BL/6 mice (purchased from NCI) were performed using rat anti-mouse LYVE-1 (R&D Systems, MAB2125), goat antimouse MARCO (R&D Systems, AF2956), goat anti-mouse MSR1 (R&D Systems, AF1797), normal goat IgG (R&D Systems, AB-108-C), and rat IgG2a (R&D Systems, MAB006) antibodies at final concentrations of 5 µg, respectively. Subsequent detection were performed using donkey anti-goat IgG-FITC conjugated (R&D Systems, F0109) and mouse anti-rat IgG2a (eBioscience, cat# 11-4817-82) at manufacturers' recommended concentrations. Flow cytometry detection of MARCO and MSR1 were performed using cells gated on CD31 and gp38 expressions as described above using anti-mouse MARCO-APC conjugated (R&D Systems, FAB 2956A), anti-mouse MSR1-APC conjugated

#### REFERENCES


(R&D Systems, FAB1797A), and isotype control antibodies (R&D Systems, IC005A and IC006A).

### DATA AVAILABILITY

The RNA-Seq datasets generated from this study have been deposited in GEO archives under the accession number GSE119499.

#### ETHICS STATEMENT

This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Procedures were approved by the University of Virginia Animal Care and Use Committee.

### AUTHOR CONTRIBUTIONS

SR, AW, JP, and KC purified the cells and extracted the RNA. AR performed the qRT-PCR. SB performed immunofluorescence staining and flow cytometry analyses. AK and ST performed bioinformatics and statistical analyses. SB and NG analyzed the data. SB, AW, and VE designed the figures. SB and VE wrote the manuscript.

### FUNDING

Support was provided by United States Public Health Services Research Grant R21 AI109250 (VE), Training Grants in Immunology (T32 AI007496; SR, NG, AW, and AR) and Medical Scientist (T32 GM007267; SR and JP), and Cancer Center Support Grant P30 CA44579 (University of Virginia Flow Cytometry Core Facility and Bioinformatics Core).

### ACKNOWLEDGMENTS

We would like to thank the University of Virginia Flow Cytometry Facility staff for their technical assistance in performing the cell sorting experiments.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00816/full#supplementary-material


the lymphoid tissue microenvironment. Blood. (2004) 103:4164–72. doi: 10.1182/blood-2003-10-3537


node fibroblastic reticular cells has the potential to regulate T cell migration and adhesion. Int Immunol. (2006) 18:301–311. doi: 10.1093/intimm/dxh369


acute and adoptive transfer experimental autoimmune encephalomyelitis. J Immunol. (2004) 173:1620–1627. doi: 10.4049/jimmunol.173.3.1620


**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 © 2019 Berendam, Koeppel, Godfrey, Rouhani, Woods, Rodriguez, Peske, Cummings, Turner and Engelhard. 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(s) 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 Antigen Processing and Presentation Machinery in Lymphatic Endothelial Cells

#### Laura Santambrogio<sup>1</sup> \*, Stella J. Berendam<sup>2</sup> and Victor H. Engelhard<sup>2</sup>

<sup>1</sup> Department of Pathology, Microbiology and Immunology, Albert Einstein College of Medicine, New York, NY, United States, <sup>2</sup> Department of Microbiology, Immunology and Cancer Biology, Carter Immunology Center, University of Virginia School of Medicine, Charlottesville, VA, United States

Until a few years ago, lymphatic vessels and lymphatic endothelial cells (LEC) were viewed as part of a passive conduit for lymph and immune cells to reach lymph nodes (LN). However, recent work has shown that LEC are active immunological players whose interaction with dendritic cells and T cells is of important immunomodulatory relevance. While the immunological interaction between LEC and other immune cells has taken a center stage, molecular analysis of LEC antigen processing and presentation machinery is still lagging. Herein we review the current knowledge of LEC MHC I and MHC II antigen processing and presentation pathways, Including the role of LEC in antigen phagocytosis, classical, and non-classical MHC II presentation, proteasome processing and MHC I presentation, and cross-presentation. The ultimate goal is to provide an overview of the LEC antigen processing and presentation machinery that constitutes the molecular basis for their role in MHC I and MHC II-restricted immune responses.

Edited by:

Amanda W. Lund, Oregon Health & Science University, United States

#### Reviewed by:

Stephanie Hugues, Université de Genève, Switzerland Ross M. Kedl, University of Colorado Denver, United States

#### \*Correspondence:

Laura Santambrogio laura.santambrogio@einstein.yu.edu

#### Specialty section:

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

Received: 26 February 2019 Accepted: 23 April 2019 Published: 07 May 2019

#### Citation:

Santambrogio L, Berendam SJ and Engelhard VH (2019) The Antigen Processing and Presentation Machinery in Lymphatic Endothelial Cells. Front. Immunol. 10:1033. doi: 10.3389/fimmu.2019.01033 Keywords: lymphatic endothelial cells, MHC class I, MHC class II, antigen processing and presentation, lymph

## MHC I AND MHC II ANTIGEN PROCESSING MACHINERY

#### MHC I and MHC II Molecules

Under basal physiological conditions both human and murine lymphatic endothelial cells (LEC) express both MHC class I and MHC class II molecules (1). However, as previously reported for blood endothelial cells (BEC) (2) the level of MHC II expression differs according to the anatomical location from which the cells are isolated (1, 3). LEC from LN (LN-LEC) express a high number of MHC II molecules while LEC from diaphragm express a much lower number (1). The MHC II surface expression in LN-LEC is similar to what observed in BEC but less than fibroblastic reticular cells from LN (1). LEC MHC II molecules are both endogenously synthesized or acquired from hematopoietic cells, as determined by chimera experiments in MHC II−/<sup>−</sup> mice (1, 4, 5). At the transcription level, MHC II expression is regulated by CIITA, which is not a DNA binding factor but instead a transactivator that regulates quantitative aspects of MHC-II expression by binding the MHC-II enhanceosome (6). CIITA expression is under the control of 4 different promoters (I, II, III, IV) and, in non-professional APC, MHC II expression is mostly regulated by CIITA IV (6). This promoter is responsive to IFNγ and other pro-inflammatory cytokines, which induce MHC II expression/up-regulation in fibroblasts and BEC (6). Similarly, in LEC it has been shown that endogenous MHC II expression is controlled by CIITA IV (4, 5). However, it is interesting to notice that, in contrast to other non-professional APC where pro-inflammatory stimuli greatly

**155**

up-regulate surface MHC II molecules, pro-inflammatory stimuli induce less robust MHC II up-regulation in LEC (3, 5, 7). In the future, it would be of interest to analyze why, even though LEC express the type IV INFγ-inducible CIITA, they do not strongly up-regulate MHCII during pro-inflammatory conditions (5).

### The Proteasome and TAP

Every cell expresses the constitutive 26S proteasome (8). This large barrel-shaped protein complex is formed in part by the catalytic 20S core, which consists of two pairs of outer α rings organized in seven α (α1–α7) subunits and two pairs of inner β rings organized in seven β subunits (β1–β7). The outer α subunits function as docking domains that regulate access of substrates to the catalytic chamber. Three of the β subunits (β1, β2, and β5) have proteolytic activities, including caspase-like activity (β1), trypsin-like activity (β2), and chymotrypsin-like activity (β5) (9). In the 26S proteasome, this 20S core is capped at both ends by the 19S regulatory complex (9). Ubiquitinated proteins are recognized by the 19S regulatory elements, which transfer them to the 20S for proteolysis (10). Peptides will then be transported in the ER by the transporter associated with antigen processing (TAP) and trimmed by the ER aminopeptidase I (ERAPI). In the ER the MHC class I heavy chain and β2m will transiently associates with TAP to load the peptides into the binding groove (11).

Following IFNγ or TNFα stimulation, new proteasome subunits are incorporated to generate the immunoproteasome (4, 12). β1 is exchanged with the large multifunctional peptidase 2 (LMP2) (also known as iβ1 or psmb9). β2 is exchanged with the multicatalytic endopeptidase complex-like-1 (MECL-1) (also known as iβ2 or psmb10). β5 is exchanged with the large multifunctional peptidase 7 (LMP7) (also known as iβ5 or psmb8). The 19S regulatory complex is exchanged with the Proteasome Activator α (PA28α) and PA28β, known as 11S regulator (2). The proteolytic functions of the immunoproteasome are greatly enhanced compared to those of the constitutive proteasome, as the immunoproteasome is more efficient in degrading ubiquitinated proteins and viral proteins, and in generating peptides for MHC-I presentation (13).

Until a few years ago, the presence of the proteasome in LEC was only indirectly analyzed by determining that TAP deficient mice were much less efficient in presenting MHC-I restricted OVA-derived peptides (7). However, a recent paper reported proteasome transcripts in LEC and BEC from different anatomical locations [**Table 1** and (14)]. All of these endothelial populations expressed comparable transcript levels for constitutive proteasome subunits and 19S regulatory subunits. However, LEC and BEC from LN expressed 5-8-fold higher levels of transcripts for psmb8, psmb9, and psmb10, and twice as much PA28α and β. This suggests that LN-localized LEC preferentially express immunoproteasomes. Similarly, LNlocalized LEC and BEC express 2–6 fold higher levels of TAP1 and TAP2, and twice as much ERAP1 and tapasin. Although these cells were isolated from resting LN, this suggests that their MHC-I processing and presentation capability is elevated.

Other non-proteasomal proteases have been implicated in MHC-I presentation (15). These additional peptidases can trim the proteasome-generated N-extended precursors or even destroy epitopes, by trimming below the size needed for presentation. Among these, LEC from both LN and lymphatic vessels express significant and comparable transcript levels of tripeptidyl peptidases I and II and nardilysin, but negligible levels of thimet [**Table 1** and (14)]. The functional implications of these additional LEC proteases, in generating the LEC MHC immunopeptidome, are currently unknown.

#### Endosomes and Lysosomes

Late endosomes (LE) and lysosomes (Lyso) are sub-cellular compartments, present in all cell types, specialized for the degradation of endogenous and exogenous materials for maintenance of cellular proteostasis and, in immune cells, for immunosurveillance (16). These organelles characteristically exhibit a low acidic pH, high concentrations of proteases, and expression of lysosome-associated membrane protein (Lamp) protein family members (16). In professional antigen presenting cells, LE and Lyso are also enriched in MHC class II proteins and molecules that regulate peptide loading (Invariant Chain, DM and DO) (17–21) and are referred as MHC class II compartments (MIIC) (22). Ultrastructurally these compartments can appear with different morphologies: multivesicular, multilamellar, or a combination of both (16).

Multivesicular bodies (MVB) are late endosomal compartments with a diameter of between 400 and 500 nm and a limiting membrane that encloses several internal vesicles with diameters of between 40 and 90 nm (16). MVB receive bio-synthetic cargo from the trans-Golgi, cytosolic cargo through autophagy, and exogenous proteins through phagocytosis. MVB are ubiquitously distributed and ultrastructural analysis has shown their presence in LEC (LS, unpublished observation) (23). However, it is currently unknown whether all/or a fraction of these compartments are MHC-II positive and whether there are differences in MHC-II expression in MVBs under steady state and inflammatory conditions. On the other hand the multilamellar bodies (MLB), which are lysosomal-like compartment formed by concentric lamellae and particularly enriched in MHC class II molecules (16) are more specifically expressed in professional APCs, such as DCs, B cells and macrophages, and they have not been found in LEC (LS unpublished observation).

#### Invariant Chain, DM, and DO

The MHC II molecules in association with their chaperone Invariant Chain, traffic from the trans-Golgi network to the plasma membrane before internalization to the endosomal MIIC. Sorting signals on the cytosolic tail of the Invariant chain are recognized by the clathrin-coated vesicle machinery for transport to LE/Lyso, where the Invariant chain will be processed

**Abbreviations:** BEC, blood endothelial cells; LEC, lymphatic endothelial cells; LN, lymph node; LN-BEC, lymph node-associated blood endothelial cells; LN-LEC, lymph node-associated lymphatic endothelial cells; DC, dendritic cells; LE, late endosomes; Lyso, lysosomes; MIIC, MHC class II compartments; MVB, multivesicular bodies; MLB, multilamellar bodies; DM, HLA-DM; Ii, invariant chain; DO, HLA-DO.

TABLE 1 | Comparative transcriptome profiling of antigen processing and presentation pathway genes from mouse lymphatic and blood endothelial cells.


(Continued)

#### TABLE 1 | Continued


(Continued)

#### TABLE 1 | Continued


<sup>a</sup>Data are reported as normalized gene expression levels as Fragments Per Kilobase of transcript per Million mapped reads (FPKM). Data is from Berendam. (14).

by Cathepsins to generate class II-associated invariant chain peptides (CLIP), which occupy the MHC II binding groove and will be exchanged with peptides derived from endosomal processing (24). MHC II/Invariant Chain complexes are present at high levels in LEC and confocal microscopy, performed on primary LEC indicates that MHC II is correctly targeted both at the cell surface and in endosomal compartments (1).

HLA-DM (H-2M in mice) is part of the endosomal antigen processing and presentation machinery and aids peptide loading onto MHC II molecules. HLA-DM (DM) was originally discovered following the analysis of B-cell lines that were inefficient at presenting peptides derived from the processing of phagocytosed proteins but easily presented peptides supplied exogenously (17–19, 25, 26). It was later determined that these cells were defective in the expression of either the HLA-DMA or HLA-DMB genes. Subsequent in vitro and in vivo experiments determined that the role of DM is to catalyze CLIP removal, stabilize empty MHC II molecules for peptide loading and skew the immunopeptidome repertoire toward high affinity peptides (17, 25, 26). Mice lacking H-2M expressed similar I-A<sup>b</sup> MHC II cell surface levels as wild type mice. However, the I-A<sup>b</sup> MHC II molecules were less compact/SDS-resistant and were predominantly associated with CLIP (27). In contrast, lack of DM led to decreased peptide capture by I-A<sup>d</sup> molecules, but enhanced peptide loading by I-E<sup>d</sup> MHC II molecules (28). Finally, lack of DM generated a substantial pool of empty or loosely occupied I-A<sup>k</sup> MHC II conformers with increase peptide binding activity. Mass spectrometry profiles confirmed the presence of an MHC II-peptidome in absence of DM (28, 29). Additionally DM requirements are different for CLIP binding in different registers (30). These results demonstrate that DM has distinct roles depending on its specific class II partners.

Subsequently, an additional protein, DO, was discovered, whose role is to inhibit DM function (18, 31). Importantly, while DM expression is not greatly increased following proinflammatory stimuli (TLR activation) that induces dendritic cell maturation, DO is down-regulated (32). As such it was hypothesized that high DO expression in immature dendritic cells would inhibit DM activity and skew the MHC II peptidome toward a broader and less stably bound repertoire. Upon DC maturation, reduced DO expression would lead to high DM activity, shaping the peptide repertoire toward long-lived surface class II MHC complexes, thus promoting productive immune responses (18, 33, 34).

Transcript analysis has shown that Invariant chain, I-A alpha, and I-A beta are expressed significantly in LEC and BEC from LN, but not LEC from lymphatic vessels, but DM and DO expression is very low to negligible, albeit DM is up-regulated following inflammatory stimuli (1, 4) [**Table 1** and (14)]. Because removal of CLIP from I-A<sup>b</sup> molecules is strongly DM dependent, this could explain the inefficient processing and presentation of I-A<sup>b</sup> restricted antigens by LEC (1). However, the haplotype variation data described above indicate that general conclusions about the ability of LEC to present MHC II restricted antigens should await analysis of other mouse haplotypes.

#### Cathepsins

Cathepsins are a large family of serine, cysteine or aspartyl proteases that are present in endo-lysosomal compartments, and may be secreted at steady state or during pathological conditions (35). Cathepsins are most active at acid pH, can still function at neutral pH but are inactive at alkaline pH (36). Although these enzymes are present in most cells, certain cathepsins are enriched in particular antigen presenting cells. For example, Cathepsin S is highly expressed in dendritic cells and B cells, Cathepsin F in macrophages, and Cathepsin L in thymocytes (36–41).

Transcriptome analysis indicated that LEC from vessels express relatively low levels of cathepsins L and F, and negligible levels of Cathepsin S, while the levels of Cathepsin S and L were significantly elevated in LEC from LN [**Table 1** and (14)]. However, measured Cathepsin L activity was variable among LN LEC and not evident in LEC from diaphragm (1). The activity of Cathepsin L indicates that at least some LEC could potentially cleave the Invariant Chain and generate CLIP peptides (1). Additionally, LEC could not efficiently process HA (an influenza membrane protein) and the IE-α protein as determined by either CD4 T cell recognition of the MHC II presented HA epitope or FACS analysis using the Y-Ae Ab that recognize I-Ab molecules loaded with the IE-α epitope, either under basal conditions or upon IFNγ stimulation (1). Furthermore, new evidence indicates that LEC express high levels of Cystatin C, B and 10 [**Table 1** and (14)], which function as natural inhibitors of cathepsins (42). Altogether, the data point to the possibility that CatL and S activity in LEC is diminished, which could affect the generation of LIP10 and CLIP, and might also diminish the processing of other endogenous antigens.

### Exogenous Peptides Binding and Antigen Exchange

The MHC I and MHC II presented immunopeptidome not only derives from endosomally processed proteins but also from pre-processed peptides that can be directly acquired from the extracellular milieu. Recent proteomic analyses have indicated that processed peptides are present in every biological fluid, among which lymph and blood, have been best characterized (43–50). The Eisen and Raghavan groups demonstrated binding of extracellular peptides to MHC I molecules and their regulation of CD8 T cell function (51, 52). Our group, among others, characterized extracellular peptide binding to MHC II surface molecules (44, 53–57). We determined that peptides carried in lymph were present in the HLA-DR1 immunopeptidome of immature dendritic cells and some of these peptides were not generated by endosomal processing, pinpointing the physiological relevance of MHC II surface/early endosomes loading (44). As such, the peptides present in the lymph, which derive from the metabolic and catabolic process of different parenchymal organs could contribute to the LEC MHC II immunopeptidome, since it has already been shown that LEC can readily bind and present pre-processed peptides (1).

## PHAGOCYTOSIS AND AUTOPHAGY

Only very recently LEC have been analyzed for their ability to capture exogenous and endogenous antigens through phagocytosis. In vivo experiments using fluorescently labeled OVA indicated that within 90 min the subcutaneously injected protein was identified in LYVE-1<sup>+</sup> cells, present in LN subcapsular sinuses (7). Additionally, genes encoding several scavenger receptors, known to be involved in receptor-mediated endocytosis, are upregulated in LEC from lymph node (14). LEC efficiency in processing phagocytosed proteins through the MHC II pathway in steady state condition is low (1); nevertheless LEC can transfer Ags to dendritic cells, which are known to be present in close proximity with LEC in the lymphatic capillary and collectors, to induce CD4 T-cell anergy (1, 58). In addition, LEC efficiently present MHC-I peptides, and it has been reported that phagocytosis in early endosomes can route exogenous antigens (both self and non self) for cross-presentation on MHC class I in a proteasome and TAP-1-dependent manner (1, 3, 7, 58– 61). It is interesting to consider that the acquisition of crosspresented material is mediated by these scavenger receptors. A second mechanism that can transfer endogenous proteins in the endosomes is autophagy. Although autophagy has been extensively characterized in BEC (62), there are no reports on the role of autophagy in antigen processing and presentation in LEC.

## LEC AND PATHOGEN IMMUNITY

A growing body of evidences indicates that LEC are involved in immune response to pathogens. It has been recently reported that in extrapulmonary tuberculosis, the lymphatic system is the most common site of infection and LEC function as a niche for Mycobacterium tuberculosis (59). Indeed M. tuberculosis can replicate in the LEC cytosol and within autophagosomes suggesting that LEC are a previously unrecognized site for infection persistence. Similarly, Hantaviruses have been shown to have a tropism for lymphatic vessels and LEC infection with either Andes virus and Hantaan virus induces LEC hyperpermeabilization and pulmonary edema (63). The edema can be inhibited by αvβ<sup>3</sup> integrin as well as VEGFR3 antibodies (63). A LEC role in HIV infection was also reported in promoting infection and latency formation in resting CD4+ T cells (64, 65). Recently an interesting role of LEC in antigen persistence, after resolution of the infection, has been shown (66). After viral challenge and vaccination, the antigen was captured by LEC under proliferative conditions and stored for extended periods of time. This "antigen archiving" mechanism positively influenced the degree of protective immunity provided by circulating memory CD8<sup>+</sup> T-cells (66, 67).

### COSTIMULATORY AND CO-INHIBITORY MOLECULES

Effective activation of T-cells requires the display of MHC-I and MHC-II-peptide complexes as well as an antigen-independent signals provided by co-stimulatory molecules, among which CD40, CD80 (B7.1), and CD86 (B7.2) have been extensively analyzed in their requirements for naïve and memory T-cells activation (68, 69). LN LEC were shown to express very low levels of CD40 and negligible levels of CD80 and CD86 (3, 60). More recent transcriptome analysis has validated these observations, and extended them to include additional costimulatory molecules [**Table 1** and (14)]. Importantly these costimulatory molecules did not up-regulate following stimulation with an MHC-I cognate ligand as well as inflammatory signals (TLRs binders or IFNγ) (3, 60).

In contrast, LEC in LN, but not in peripheral tissue lymphatics, express multiple inhibitory receptors that engage counter-receptors on activated T-cells to dampen the immune response (69). These include PD-L1 (CD274) and PD-L2 (CD273), which are present on resting LEC and greatly upregulated by inflammatory stimuli (1, 70). Interestingly, the ligand for LAG-3, another inhibitory receptor on T-cells, is MHC-II, and induction of CD8 T-cell tolerance by LEC depends on engagement of LAG-3 as well as PD-1 (1, 60). Consequently, it has been suggested that in the absence of functional Ag presentation, the expression of MHC-II molecules on LEC is concerned with inducing Lag-3 mediated tolerance. While the low expression of costimulatory molecules would suggest that LEC would be unable to activate T-cells, they stimulate profound proliferation of CD8 T-cells in vivo and in vitro, and after peptide pulsing and CD4 T-cell proliferation in vitro (58, 60). However, the expression of the co-inhibitory molecules leads to deletional tolerance of CD8 T-cells due to a failure to sustain upregulation of the IL-2 receptor. Thus, LEC represent an important mechanism for mediation of systemic peripheral tolerance (58, 60, 61).

#### EXOSOMES AND OTHER VESICLES

Most cells in the human body release vesicles of different sizes and content which can be classified as apoptotic bodies, micro and macrovesicles and exosomes (71). Exosomes are small (30–120 mm) vesicles generated from the multivesicular late endosomes upon fusion with the plasma membrane and release in the extracellular milieu. Exosomes from different sources have been shown to transport a protein cargo as well as mRNAs and microRNAs. Their physiological and pathological relevance has been established in several immune and cancerrelated models (72). Although very little is known about LEC-released exosomes, recently it has been shown that LEC release a vesicular fraction, which includes exosomes, following an inflammatory signal (73). The LEC-derived exosomes are reportedly enriched with a motility-promoting protein signature, which act as a cue for the dendritic cells migratory response (73).

#### REFERENCES


In particular LEC released vesicles accumulate in the perivascular stroma of small lymphatic vessels, mostly in the presence of inflammatory cytokines and promote directional migration of CX3CR1-expressing cells (73).

#### CONCLUDING REMARKS

LEC cells are anatomically placed between parenchymal organs and draining lymph nodes, functioning as a conduit for the lymphatic fluid and are known to control DC and T cell migration in and out of the lymph node (74, 75). During the last few years their functionality in antigen processing and presentation and T cell immune responses has emerged. Under steady-state conditions LEC can present self-antigens to induce T cell tolerance either through expression of peripheral tissue antigens (76) or acquisition of extracellular antigens through phagocytosis or by acquisition of pre-loaded MHC II molecules from DC. Under inflammatory conditions LEC also play an immunosuppressive role by decreasing DC maturation (77) and by up-regulating surface PDL1 (76).

However, the advances in understanding the cross-talk between LEC and T cells has not been paralleled by a detailed mechanistic analysis of their antigen processing and presentation machinery. Characterization of LEC immunoproteasomes, endosomal processing compartments, and antigen acquisition from the lymphatic fluid still needs to be investigated. Nevertheless, the work to date points to an emerging picture of the role played by LEC in maintenance of self-tolerance.

### AUTHOR CONTRIBUTIONS

LS and VE wrote the review. SB contributed the primary data presented in the table.

### FUNDING

Support was provided by United States Public Health Services Research Grant R21 AI109250 (VE) and RO1 AG045223 (LS).


or viral infection. Nat Commun. (2014) 5:3989. doi: 10.1038/ ncomms4989


**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 © 2019 Santambrogio, Berendam and Engelhard. 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(s) 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.

# Chronic Liver Disease in Humans Causes Expansion and Differentiation of Liver Lymphatic Endothelial Cells

Beth A. Jiron Tamburini 1,2,3 \* † , Jeffrey M. Finlon<sup>1</sup> , Austin E. Gillen<sup>3</sup> , Michael S. Kriss <sup>1</sup> , Kent A. Riemondy <sup>3</sup> , Rui Fu<sup>3</sup> , Ronald P. Schuyler <sup>2</sup> , Jay R. Hesselberth3,4, Hugo R. Rosen<sup>5</sup> and Matthew A. Burchill 1,3 \* †

*<sup>1</sup> Division of Gastroenterology and Hepatology, Department of Medicine, School of Medicine, University of Colorado, Aurora, CO, United States, <sup>2</sup> Department of Immunology and Microbiology, School of Medicine, University of Colorado, Aurora, CO, United States, <sup>3</sup> RNA Biosciences Initiative, School of Medicine, University of Colorado, Aurora, CO, United States, <sup>4</sup> Department of Biochemistry and Molecular Genetics, School of Medicine, University of Colorado, Aurora, CO, United States, <sup>5</sup> Keck School of Medicine, University of Southern California, Los Angeles, CA, United States*

#### Edited by:

*Anne Fletcher, Monash University, Australia*

#### Reviewed by:

*Veronika Lukacs-Kornek, Saarland University, Germany Gou Young Koh, Korea Advanced Institute of Science and Technology (KAIST), South Korea*

#### \*Correspondence:

*Beth A. Jiron Tamburini beth.tamburini@ucdenver.edu Matthew A. Burchill matthew.burchill@ucdenver.edu*

*†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: *27 November 2018* Accepted: *23 April 2019* Published: *15 May 2019*

#### Citation:

*Tamburini BAJ, Finlon JM, Gillen AE, Kriss MS, Riemondy KA, Fu R, Schuyler RP, Hesselberth JR, Rosen HR and Burchill MA (2019) Chronic Liver Disease in Humans Causes Expansion and Differentiation of Liver Lymphatic Endothelial Cells. Front. Immunol. 10:1036. doi: 10.3389/fimmu.2019.01036* Liver lymphatic vessels support liver function by draining interstitial fluid, cholesterol, fat, and immune cells for surveillance in the liver draining lymph node. Chronic liver disease is associated with increased inflammation and immune cell infiltrate. However, it is currently unknown if or how lymphatic vessels respond to increased inflammation and immune cell infiltrate in the liver during chronic disease. Here we demonstrate that lymphatic vessel abundance increases in patients with chronic liver disease and is associated with areas of fibrosis and immune cell infiltration. Using single-cell mRNA sequencing and multi-spectral immunofluorescence analysis we identified liver lymphatic endothelial cells and found that chronic liver disease results in lymphatic endothelial cells (LECs) that are in active cell cycle with increased expression of CCL21. Additionally, we found that LECs from patients with NASH adopt a transcriptional program associated with increased IL13 signaling. Moreover, we found that oxidized low density lipoprotein, associated with NASH pathogenesis, induced the transcription and protein production of IL13 in LECs both *in vitro* and in a mouse model. Finally, we show that oxidized low density lipoprotein reduced the transcription of *PROX1* and decreased lymphatic stability. Together these data indicate that LECs are active participants in the liver, expanding in an attempt to maintain tissue homeostasis. However, when inflammatory signals, such as oxidized low density lipoprotein are increased, as in NASH, lymphatic function declines and liver homeostasis is impeded.

Keywords: lymphatic endothelial cells, cirrhosis, fibrosis, non-alcoholic steatohepatitis, hepatitis C virus, alcoholic liver disease, interleukin-13, oxidized low density lipoprotein

## INTRODUCTION

Deaths from chronic liver disease (CLD) have increased by 31% between the years 2000 and 2015 (1). CLD arises due to chronic inflammation in the liver as a result of a number of environmental insults including viral infection (hepatitis C or B virus-HCV/HBV), alcohol consumption (Alcohol associated liver disease-ALD) and diet-induced obesity (Non-alcoholic steatohepatitis-NASH).

**164**

Due to the regenerative capacity of the liver, the liver participates in a dynamic process that can result in several rounds of injury and repair. However, chronic injury eventually leads to severe fibrosis, cirrhosis, and the decline of liver function. While removal of the insult causing injury can be effective at reversing liver fibrosis (2), many patients with advanced disease do not improve or ultimately progress to cirrhosis (3, 4). As a result, these people remain at an elevated risk for development of hepatocellular carcinoma despite the removal of chronic insult (5). Limited therapeutic options exist for these patients causing the rates of morbidity and mortality to continue to climb (6).

The lymphatic system transports interstitial fluid (lymph) from the tissue to the circulatory system for removal from the body (7, 8). In addition, lymphatics participate in the acquisition of fat and the formation of chylomicrons in the gut (9), reverse cholesterol transport (10), and trafficking of dendritic cells (DCs) from the tissue to the lymph node (LN) (11, 12). During CLD, increased lymphatic permeability has been implicated in the formation of ascites, or fluid accumulation, in the peritoneal cavity (13). Furthermore, increased lymphatic vessel permeability has been demonstrated to increase inflammation and immune dysfunction in other tissues (14–17). Obesity and hypercholesterolemia are also associated with lymphatic permeability, hyperplasia, and inflammation at peripheral sites in humans (18, 19) and in animals (20).

Lymphatic vessels are comprised of lymphatic endothelial cells (LECs). LEC interactions with immune cells can guide trafficking of immune cells as well as promote self-tolerance and enhance protective immunity (21–25). Despite the multi-faceted role of LECs in programming immune responses in the lymph node and skin, the role of lymphatics in coordinating the immune response in the liver has not been addressed. Furthermore, with the advent of single cell sequencing, several reports have addressed different cell populations within the liver, including specific interrogation of liver endothelial cell populations (26–28). However, in none of these reports have lymphatic endothelial cells been identified. Thus, the transcriptional profile and function of liver lymphatic endothelial cells in homeostasis or disease is yet unknown. However, previous case reports from almost 20 years ago, using common endothelial markers and histology, did demonstrate that lymphatic vessels increase in diameter and abundance during chronic viral hepatitis (29, 30). Despite these observations, little to nothing is known about lymphatics in non-viral cases of CLD, or the cause and/or consequence of lymphatic expansion in the liver. As diet-induced CLD has surpassed chronic viral infection as the leading cause of liver transplantation (31), understanding the role of lymphatics in the liver in people with diet-induced CLD is an important functional process that needs to be addressed.

Here we demonstrate a significant increase in lymphatic vessel density in patients with CLD. We show that in addition to virally induced CLD, that non-virally induced CLD also results in a significant increase in lymphatic vessel density in the liver. To identify specific differences in LECs during disease we performed transcriptional profiling of LECs, using a single-cell platform, from non-diseased and diseased human livers. While other endothelial cell populations have been identified by single cell RNA sequencing in the liver, this is the first demonstration of isolation of LECs and subsequent transcriptional profiling of this rare cell population in the liver. We find that LECs from NASH or HCV infected livers engage a transcriptional program that results in more LECs in active cell cycle and more CCL21 expression. However, when comparing the transcriptional profile of LECs from patients with NASH to patients with HCV we find that NASH specifically induces the activation of the IL13 pathway. Furthermore, we demonstrate that not only is the IL13 pathway increased in patients with NASH, but also that oxidized LDL, commonly associated with inflammation in NASH, can induce the upregulation of IL13 transcript and protein in LECs. Finally, we provide evidence that IL13 expression by LECs occurs specifically in the liver and that oxidized LDL results in the downregulation of the LEC transcription factor, PROX1, and reduced lymphatic stability.

### MATERIALS AND METHODS

#### Patient Samples

For immunohistochemistry archived patient specimens were obtained from the University of Colorado Anschutz Medical Campus biorepository core facility **(Supplementary Table 1)**. For single cell sequencing, patients were selected from a biorepository of patients who had undergone liver transplantation and collected under the IRB protocol of HRR and MSK. Transplanted livers were harvested and nonparenchymal cells (NPCs) were isolated and frozen in a single cell suspension as described (32). Additionally, non-diseased NPCs were purchased from Triangle research laboratories (Lonza, Triangle research park, NC). For Non-diseased patients (n = 6) the age range was 37 to 58 with a mean age of 49 where five patients were male and one was female. For HCV samples (n = 3) the age range was 47–55 with a mean age of 51 and all patients were male. For NASH samples (n = 2) the age range was 49–55 with a mean age of 52 and all patients were females. For Ki67 analysis: the age range was 35–61 for diseased patients and included NASH, ALD, HCV, AIH, PBC, and AIH with NASH. The average age of all diseased patients was 49.4 with 62.5% of the patients being female. The same non-diseased patients from Lonza were used as described above. All patients provided written and informed consent and the study was approved by the institutional review boards at the University of Colorado—Anschutz.

#### Flow Sorting and Flow Cytometric Analysis

To enrich LECs from hepatic NPCs we thawed frozen samples in RPMI containing 10% Human Serum AB (Gemini Bio-products, West Sacramento, CA) and 1% DNASE (MP Biomedicals, Santa Ana, CA). Cells were washed 2x with PBS containing 2% FBS

**Abbreviations:** CLD, chronic liver disease; LEC, lymphatic endothelial cell; PEC, portal endothelial cell; HBV, hepatitis B virus; HCV, hepatitis C virus; PBC, primary biliary cirrhosis; NASH, non-alcoholic steatohepatitis; ALD, alcoholic liver disease; AIH, autoimmune hepatitis; PSC, primary sclerosing cholangitis; PDPN, podoplanin; IL13, Interleukin 13; MELD, model for end-stage liver disease.

(Atlas Biologicals, Fort Collins, CO) and stained with antibodies against CD45 (clone HI30), CD31 (PECAM1, clone WM59), Podoplanin (clone NC-08) and CD146 (clone P1H12) from Biolegend (San Diego, CA), and CD68 (clone KP1) from abcam (San Francisco, CA). Cells from either Non-diseased, NASH or HCV were sorted using an aria Fusion sorter (BD Biosciences, Franklin lakes, NJ) and enriched LECs were sorted into RPMI containing 50% human serum AB. Enriched LECs from each condition were pooled for single cell sequencing as described below. Flow cytometric analysis of human liver LECs: isolated liver NPCs were stained with Fixable viability dye 510 (BD Biosciences) and stained with the above surface antibodies. Following surface staining cells were fixed and permeabilized (Thermo Fischer, Waltham, MA) and stained for Ki67 (clone 11F6) (Biolegend). Flow cytometry was performed using a BD FACSCanto II instrument and was acquired with BD FACSDiva software (BD Biosciences). Analysis ws performed using FlowJo 10 (Treestar, Woodburn, OR).

### Single Cell RNA Sequencing

Approximately 10,000 LEC-enriched hepatic NPCs were loaded onto a 10x genomics (San Francisco, CA) controller per manufacturer instructions to generate barcoded single cell GEMs using the 10x genomics 3′ kit. mRNA was converted to cDNA within each barcoded single cell GEM and libraries were generated as previously described. 10x libraries were sequenced using a NovaSeq 6000 (Illumina, San Diego, CA) to a predicted depth of 100,000 reads/cell. All cell preparation was performed at the University of Colorado Genomics Shared Resource Core. To control for sequencing batch effects, a minimum of three nondiseased samples were included with each diseased capture and sequencing run.

### Quantification of Single Cell RNA Sequencing

scRNA-seq data was processed with the 10x Genomics Cell Ranger Suite for demultiplexing, alignment, assignment of reads to genes, and unique molecular identifier (UMI) deduplication to remove PCR duplicates. Further analysis, including cell clustering, cell type identification, marker gene identification, and differential expression analyses was performed using the R packages Seurat (33) and scran (34). For cell cycle analysis Seurat assigns each cell a score based on its expression of G2/M and S phase markers. These marker sets are anti-correlated in their expression levels, and cells expressing neither are assigned to G1 phase. Cells with fewer than 250 detectable genes or >20% of UMIs derived from mitochondrial genes were excluded from the analysis to eliminate cells with insufficient expression data for clustering and dead cells, respectively **(Supplementary Figure 1)**. Ingenuity pathway analysis (IPA) was performed using IPA software (Qiagen, Venlo, Netherlands). Raw data, count matrices, metadata including cluster assignment are available at the Gene Expression Omnibus under accession GSE129933 (https://www. ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE129933).

### Multispectral Fluorescence Immunohistochemistry and Vectra Analysis

Five micron thick tissue sections were sequentially stained for human PDPN, CD3, CD19, CCL21, and CD68. Slides were dewaxed with xylene, heat treated in pH9 antigen retrieval buffer for 15 min in a pressure cooker, blocked in Antibody (Ab) Diluent (Perkin Elmer, Waltham, MA), incubated for 30 min with the primary Ab, 10 min with horseradish peroxidase (HRP)-conjugated secondary polymer (anti-mouse/anti-rabbit, Perkin Elmer, Waltham, MA), and 10 min with HRP-reactive OPAL fluorescent reagents (Perkin Elmer). Slides were washed between staining steps with PBS 0.01% tween 20 and stripped between each round of staining with heat treatment in antigen retrieval buffer. After the final staining round the slides were stained with spectral DAPI (Perkin Elmer), and coverslipped with Prolong Diamond mounting media (Thermo Fisher, Waltham, MA). Multispectral imaging was preformed using the Vectra 3.0 Automated Quantitative Pathology Imaging System (Perkin Elmer). Whole slide scans were collected using the 10x objective and 10 to 20 regions were selected for multispectral imaging with the 20x objective. The multispectral images were analyzed with inForm software (Perkin Elmer) to unmix adjacent fluorochromes, subtract autofluorescence, segment the tissue into lymphatic vessels and non-lymphatic vessels, segment the cells into nuclear, and membrane compartments, and to phenotype the cells according to morphology and cell marker expression. Lymphatic vessel density was quantified using the Nikon AR software where LVD = vessel area/total area × 100%. Lymphatic vessels with CCL21 staining less 0.036 were classified as CCL21lo or negative while lymphatic vessels with a value >0.036 were classified as CCL21hi or positive using inForm software. For CCL21 analysis three Non-diseased, 5 HCV, and three NASH patient samples were interrogated. For T, B, and macrophage cell analysis three or four non-diseased, 3–5 NASH, and 4 HCV patient samples were quantified. A student's t-test was performed where two asterisks represents a p-value less than 0.001. Patient data is included in **Supplementary Table 1**.

## Branch Forming Assay

Branch forming assay was performed as previously described (35, 36). Briefly, a 4.2 mg/ml matrigel (Corning, Tewksbury, MA) pad with either 100µg/ml oxidized LDL (Alfa Aeser, Ward Hill, MA), 0.25 mM Palmitic Acid (Caymen Chemicals, Ann Arbor, MI, or equivalent amount of DMSO was allowed to solidify for one and a half hours. Fifteen thousand Human Lymphatic Endothelial Cells (PromoCell, Heidelberg, Germany) mixed with 100µg/ml oxidized LDL, 0.25 mM Palmitic Acid, or equivalent amount of DMSO in endothelial cell growth medium (PromoCell) were placed on top of the matrigel pad. Cells were incubated at 37◦C for 21 h, then imaged using a Zeiss microscope, and Axiom camera. For RT-PCR the matrigel was dissolved using ice cold 5 mM EDTA rocking on ice for 1 h. HLECs were pelleted and lysed with RLT buffer, mRNA was extracted using the RNeasy micro kit per manufacturer instructions and cDNA was synthesized using the QuantiTect RT Kit (Qiagen, Venlo, Netherlands) following standard protocols. Quantitative PCR was performed on an Applied Biosystems 7,300 Real-time PCR machine and fold changes in mRNA levels were calculated using the delta-delta CT method. For each gene, all samples were normalized to the average fold change of the vehicle treatment group (DMSO). The following Qiagen QuantiTect primers were used: GUSB (QT00046046), IL13 (QT00000511), PROX1 (QT01006670), and VEGFR3/FLT4 (QT00063637).

#### Animal Studies

Six to eight week old C56BL6/J mice were IV injected with 85 µg of unlabeled or DIL or DIO labeled human highly oxidized LDL (Kalen Biochemicals, Germantown, MD) or 100 µg PolyI:C (InvivoGen, San Diego, CA). Following indicated incubation time mice were administered a second dose of stimulant. For in vivo BFA experiments, 18 h following the second injection mice were administered 250 µg of Brefeldin A as previously described for analysis of in vivo cytokine production by T cells (25, 37). Ninety minutes after Brefeldin A injection mice were euthanized and livers and lymph nodes were processed as described (38) with Brefeldin A in each buffer. For flow cytometric analysis of LEC production of IL13, single cell suspensions of liver and lymph node cells were stained with CD45 (clone 30-F11), CD31 (clone 390), PDPN (clone 8.1.1) from biolegend and CD146 (clone P1H12) and IL13 (clone ebio13A) from ebioscience. Flow cytometry was performed using a BD FACSCanto II instrument and data were acquired with BD FACSDiva software (BD Biosciences) or Cyan ADP and acquired with summit software. Analysis was performed using FlowJo 10 (Treestar). All procedures were approved by the University

anti-CD68 (macrophages-green), and dapi (nuclei-blue) and imaged using the Perkin Elmer Vectra 3.0 imaging system and linear unmixed with inFORM software. (C) Lymphatic vessels density was determined using inFORM software and normalized to area in each disease listed and designated by color of dot. Statistical analysis was performed using a student's *t*-test. \*\**P* < 0.01.

of Colorado School of Medicine Institutional Animal Care and Use Committee.

#### Statistical Analysis

For all graphs with statistical analysis an unpaired student's t-test was used evaluate statistical significance between two points using Prism software (GraphPad, San Diego, CA). One asterisk denotes a p-value of <0.05, two asterisks denotes a p-value <0.01 and three asterisks denotes a p-value <0.001. Statistical analysis for single cell RNA sequencing (differential expression testing) was performed using the Wilcoxon Rank-Sum Test implemented in Seurat.

### RESULTS

### Lymphatic Vessel Density Increases in End Stage Liver Disease

Liver tissue was obtained from 28 patients with end stage liver disease at the time of transplantation (Non-Alcoholic Steatohepatitis (NASH), Alcoholic liver disease (ALD), chronic Hepatitis C viral infection (HCV), Autoimmune hepatitis (AIH), Wilson's disease, Primary sclerosing cholangitis (PSC), and four patients with non-diseased livers **(Supplementary Table 1)**. Lymphatic vessels in the liver were assessed via immunofluorescence staining with the anti-podoplanin antibody (red) to mark lymphatic vessels in addition to anti-CD3 to mark T cells (white), anti-CD68 to mark macrophages (green), and dapi to label nuclei (blue). Shown are representative images from both a non-diseased liver **(Figure 1A)** and an ALD patient explant **(Figure 1B)**. In cirrhotic livers, the density of lymphatic vessels is significantly increased, regardless of disease etiology, when compared to liver sections from non-diseased controls (**Figure 1C)**. Changes in LVD were independent of Model for End Stage Liver Disease (MELD) score at time of transplant, age, body mass index (BMI), race, or disease etiology (**Supplementary Figure 2**). This increased lymphatic vessel density is confined to areas of active inflammation. As such, the lymphatic vessels are in close proximity to regions of inflammation as determined by the frequency of T cells, macrophages, and fibrotic areas. Alternatively, neither lymphatic vessels, nor T cells were found in the regenerative nodules **(Figures 1A,B** and **Supplementary Figure 3)**. This supports lymphatic expansion as a universal marker and potential critical mechanism for chronic liver disease progression.

### Increased CCL21 Expression and Immune Cell Infiltration Occurs in Cirrhotic Livers

As we determined lymphatic vessel density was increased during disease we asked if CCL21, T cell, B cell or macrophage frequency was changed. Others have reported expression of CCL21 by LECs in other tissues and CCL21 is a chemokine that can recruit CCR7+ dendritic cells and T cells (39). We first validated that liver LECs express CCL21 protein using a CCL21 specific antibody. In the imaging of non-diseased and diseased livers we observed that there were populations of LECs that express no or low levels of CCL21 **(Figure 2A)** while others express high levels of CCL21 protein (**Figure 2B**). To stratify lymphatic vessels based on CCL21 expression we calculated negative expression




*Shown is the average log2 gene expression fold change between LEC and PEC in non-diseased livers, the percent of cells in each subset that expresses a given gene, and the raw and adjusted P-values. Data was filtered to include only genes in which the p-value between LEC and PEC was* <*0.0006.*

to be <0.01 counts, based on a no-CCL21 antibody control, CCL21lo expression to be between 0.01 and 0.036 counts and CCL21hi expression to be between 0.036 and 0.252 counts, as assessed by InFORM software **(Supplementary Figure 4)**. Using PDPN to label lymphatic vessels we found that the number of lymphatic vessels that had CCL21hi/<sup>+</sup> expression was about 4 vessels per mm<sup>2</sup> in non-diseased livers, while the frequency of lymphatic vessels with high expression of CCL21 was between 6 and 10 vessels per mm<sup>2</sup> in patients with HCV and NASH **(Figure 2C)**. These vessels were also often associated with infiltrating immune cells and thus we quantified the accumulation of B cells, T cells and macrophages in the liver. Similar to previous results, we found that end stage liver disease resulted in the significant accumulation of T cells and macrophages in the liver of patients with HCV and NASH while B cells were less frequent **(Figure 2D)**. These studies led us to ask if chronic liver disease induces LECs differentiation in the liver that results in modulation of the inflammatory state of the liver microenvironment.

### Isolation and Single Cell Sequencing of Lymphatic Endothelial Cells in the Liver

To understand how LECs were transcriptionally regulated in the liver we isolated LECs and subjected them to single cell mRNA sequencing. While other endothelial cell populations within the liver have been evaluated transcriptionally, the transcriptome of liver lymphatic vessels has yet to be reported. This could be due to the LECs in the liver being a fragile, rare and difficult to identify population. Therefore, we used our expertise in lymphatic endothelial cell flow cytometry to isolate and acquire the transcriptional signature of lymphatic endothelial cells in the liver. Once LECs from explanted human livers were isolated by flow sorting, using our published liver LEC marker set (38) **(Figure 3A)**, we subjected the sorted cells to single cell mRNA sequencing using the 10x genomics 3′ platform. Individual groups of cells were clustered using TSNE clustering based on transcriptional profile **(Figure 3B)**. Interestingly, while we sorted our population of cells based on known phenotypic markers for LECs we still were able to visualize a number of contaminating cells based on their transcriptional profile. However, using this analysis we were also able to discern that in the non-diseased human liver there are two distinct populations of endothelial cells that are transcriptionally distinct from liver sinusoidal endothelial cells (LSECS) (26).

LECs were identified using the expression of prospero homeobox protein 1 (PROX1), lymphatic vessel endothelial hyaluronan receptor 1 (LYVE-1), podoplanin (PDPN), vascular endothelial growth factor receptor 3 (FLT4/VEGFR3), and CCL21 **(Figure 3C)**. Based on the expression of these LEC-associated markers we were able to divide these two populations into a fully differentiated LEC population and an endothelial cell population that resembles the recently reported portal endothelial cells (PEC) (27). Specifically, these two populations are distinguished by their expression of several markers such as PROX1, PDPN, CCL21, Neurotensin, and trefoil factor 3 (TFF3) by the LEC cluster; bone marrow stromal antigen two precursor (BST2), interferon alpha inducible protein 27 (IFI27) and ribonuclease 1 (RNASE1) by the other endothelial cluster similar to PECs **(Table 1).** These factors have been previously reported to be associated with either LECs or blood endothelial cells (BECs) in other model systems and in primary endothelial cell cultures confirming that these subsets are of lymphatic or blood origin, respectively (40, 41). As seen in **Figure 2**, liver LECs express CCL21 protein while LECs from diseased livers had more CCL21hi/<sup>+</sup> vessels. This was confirmed by our transcriptional analysis and similar to other reports demonstrating expression of the chemokine CCL21 by LECs **(Figure 3D)** (42, 43). We also discovered that TFF3—a gene upregulated in hypoxia that induces expression of VEGF and protects barrier function is upregulated by the LEC population (44, 45). The liver PEClike population has increased expression of genes such as IFI27 which is also expressed by LECs in the lymph node (46–48); and HSPG2 encoding Perlecan which is predominantly expressed by BECs, but whose expression can be increased during the final maturation of lymphatic vessels in the skin (49). These data suggest that the PEC-like population may be a progenitor cell for LECs with transcripts found in both cells from blood and lymphatic lineages. Finally, we confirm that the structures we visualized in the livers of cirrhotic patients **(Figure 1)** are the same LEC population we are evaluating transcriptionally based on the expression of PDPN in this specific endothelial population **(Figure 3C)**. Thus, based on transcriptional profiling we were able to distinguish LECs from other cells in the liver in order to evaluate changes in these cells during chronic liver disease.

#### Liver Disease Results in the Proliferation of Lymphatic Endothelial Cells

As stated above we identified lymphatic endothelial cells in the liver both by flow cytometry and transcriptional profiling.


*Shown is the average log*<sup>2</sup> *gene expression fold change differences between LECs isolated from NASH and HCV, the percent of cells in each subset that expresses a given gene, and the raw and adjusted P-values. Data was filtered to include only genes in which the p-value between HCV and NASH were* <*0.0006.*

Being that we saw a substantial increase in the frequency of lymphatic vessels in disease **(Figure 1)** we evaluated if chronic liver disease induced the specific expansion and differentiation of LECs. We first compared liver LECs from non-diseased to diseased livers (both HCV and NASH) and found LECs from diseased livers downregulated pathways involved in apoptosis while upregulating pathways involved in free radical scavenging **(Supplementary Table 2)**. Similarly, upstream pathways activated in LECs from diseased livers included Jun N-terminal kinase (JNK), mitogen-activated protein kinase (MAPK), as well as tumor necrosis factor (TNF) **(Supplementary Table 3)**. We also observed a higher proportion of LECs from diseased patients that are in active cell cycle compared to LECs from non-diseased livers or PECs **(Figure 4A)**. To confirm this transcriptional data we utilized flow cytometry to measure Ki67 expression by LECs from non-diseased and diseased livers. Using this approach, we were able to confirm our transcriptional data demonstrating that a higher frequency of LECs from diseased livers have Ki67 expression compared to LECs from non-diseased livers **(Figure 4B)**. This difference was consistent across patients, suggesting the expansion of LECs is a common event during chronic liver disease **(Figure 4C)**. Finally, when comparing genes between liver LECs from non-diseased, HCV or NASH we found that each group clustered differently **(Figure 4D)**. Many of the same genes were differentially expressed between non-diseased and both NASH and HCV **(Figure 4D)** suggesting different inflammatory stimuli induce some of the same transcriptional programs. However, while there were many similarities in the transcriptional programs of LECs from HCV or NASH explanted livers there were also differences **(Figure 4D**). Taken together, these data demonstrate that LECs and PECs, are differentially regulated during chronic liver disease and that chronic liver disease results in the preferential expansion of LECs. These data also suggest that different inflammatory insults may regulate different gene programs in LECs.

### CLD Alters Signaling Pathways in Liver LECs

To examine differences in gene expression in LECs dependent on disease etiology we compared the gene expression of LECs from NASH patients to HCV patients **(Table 2)**. We also used IPA software to evaluate transcriptional pathways **(Supplementary Table 4)** induced by the different disease etiologies. Intriguingly, the IL13 signaling pathway was upregulated in LECs isolated from patients with NASH **(Supplementary Table 4)**. IL13 has been shown to be involved both in maintaining lymphatic vessel structure and permeability (35, 50) and in the conversion of hepatic stellate cells (HSC) to myofibroblasts (51, 52). While IL13 expression was not detected in any of the cell types we evaluated, suggesting a signal strength issue; CD36, FABP4 and TFF3, genes that are targets of IL13 signaling, were upregulated (53–55), and ATF3, which inhibits IL13 transcription (56), was downregulated in LECs from NASH livers (**Figure 5A**). These data not only gave us potential leads to follow, but also led us to the conclusion that while CLD uniformly induces the expansion of lymphatic vessels in the liver, NASH-associated liver disease elicited a unique transcriptional profile in liver LECs that involves IL13 signaling.

We next asked if these transcriptional differences were a result of direct stimulation of LECs by factors associated with the disease. Therefore, we asked if oxidized LDL (oxLDL) or palmitic acid (PA) affected the lymphatic branching and transcriptional profile of human LECs (hLECs) in vitro (**Figure 5B**). We used oxLDL as it accumulates in response to free radicals generated by inflammation, has been shown to affect other cell types such as macrophages and endothelial cells in atherosclerosis and is elevated in NASH (57–62). Further, oxLDL has a dramatic effect on macrophage activation, while LDL alone does not (63). We used PA as it is an important dietary fatty acid that is largely consumed in foods [reviewed in (64)]. We observed that when LECs were treated with vehicle (DMSO), PA or oxLDL and

24 h with the indicated stimulus. (F) Quantification of (E). \**P* < 0.05. \*\*\**P* < 0.001.

visualized at 4 and 24 h post-treatment that only the vehicle and PA treated LECs were able to maintain their branched structures over the 24-h period (**Figures 5B,C**). OxLDL treated LECs were able to form the branch structures, but by 24 h the structures had collapsed (**Figures 5B,C**). To determine if these dietary constituents were inducing transcriptional changes in the LECs we performed qRT-PCR on the hLECs and evaluated the transcript abundance of IL13 (**Figure 5D**) because the IL13 pathway was upregulated in LECs from people with end stage NASH (**Figure 5A**) and IL13 has been shown to cause defects in lymphatic branching and inhibit PROX1 expression (35). Interestingly, we found that IL13 was upregulated only in LECs that were treated with oxLDL, but not PA (**Figure 5D**). Consistent with LEC structure collapse, oxLDL treatment resulted in the decreased expression of PROX1 and the PROX1 dependent gene, FLT4 (VEGFR3) (**Figure 5D**). This is in contrast to PA which did not affect PROX1 expression. We next asked if IL13 protein levels were increased after oxLDL, PA or interferon alpha. Interferon signaling is significantly increased in HCV while IL13 signaling is not. We found using flow cytometry that there is a significant increase in protein expression of IL13 in hLECs treated with oxLDL, but not PA or IFN **(Figures 5E,F)**.

These data demonstrate that oxLDL can induce transcriptional and functional changes in LECs in vitro that are similar to the signaling pathways we observed in patients with NASH. Thus, IL13 signaling in LECs from patients with NASH could be caused by increased levels of oxLDL in the liver.

### OxLDL Uniquely Induces IL13 Production by Liver LECs in vivo

Above we demonstrated that the IL13 signaling pathway is activated in LECs of people with NASH (**Figure 3**) and increased IL13 gene and protein expression in vitro when LECs were treated with oxLDL (**Figures 5D–F**). We next asked if LECs in the liver and LN of mice treated with oxLDL were able to produce IL13 protein in vivo. To answer this question, we intravenously injected C57BL/6 mice with fluorescently labeled, oxLDL or the interferon inducing toll like receptor agonist, polyI:C (as an inflammatory control that should not induce IL13). To determine if these stimuli resulted in the production of IL13 by liver and/or lymph node (LN) LECs we directly measured production of cytokines in vivo 6 days post-injection **(Figure 6A)**. Using an in vivo Brefeldin A assay (25, 37, 65), we found that acute stimulation with oxLDL, but not polyI:C elicited the production of IL13 by liver LECs **(Figure 6B)**. Interestingly, while the LECs in the skin draining LN were able to take up oxLDL, they did not produce IL13 (**Figure 6**). These findings were confirmed using Balb/c mice that have been engineered to express YFP under the control of the IL13 promoter (IL13-YFP) **(Supplementary Figure 5)**. These findings support our conclusions that liver LECs directly respond to oxLDL by producing IL13 and that liver LECs have a unique functional response to dietary stimulation compared to lymph node LECs.

#### DISCUSSION

The role of lymphatic vessels in normal and disrupted liver homeostasis has largely been ignored. Previous studies have demonstrated an increase in lymphatic vessel-like structures in the liver during chronic viral infection and in the setting of portal hypertension (29, 66). Others have proposed that the ascites associated with chronic liver disease may be a consequence of lymphatic dysfunction or increased lymphatic permeability (13). Furthermore, altered lymphatic function in sites peripheral to the liver has been documented in humans and animal models of obesity, infection, and hypercholesterolemia (18–20, 67). These findings seem to link liver function and lymphatic function, however even recent studies utilizing single cell RNA sequencing to evaluate liver cell populations or even specifically liver endothelial cell populations have failed to identify lymphatic endothelial cells (26, 27). This is likely due to the low frequency of lymphatic endothelial cells in normal human livers and the inability to maintain LEC viability or distinguish these populations for downstream transcriptional profiling. Thus, the precise identification of and transcriptional profile of liver lymphatic vessels in steady state and during chronic liver disease had yet to be achieved. In this study we aimed to understand the lymphatic system in the non-diseased human liver and in the setting of chronic liver disease.

We have previously developed methodology to evaluate liver LECs by flow cytometry (38). In this manuscript we demonstrate a strategy that utilized both flow cytometric sorting and singlecell mRNA sequencing to directly analyze the transcriptional profile of LECs from the liver during steady state and disease. As evident from our data **(Figure 2B)** our flow cytometric sorting strategy does not result in a pure population of LECs but rather an enrichment of these cells. However we do demonstrate that transcriptionally the markers PDPN, PROX1, VEGFR3, CCL21, and LYVE-1, when combined together, adequately label LECs and no other cell type in the liver expresses all of these markers. We were intrigued to find that the expansion of the LEC population in the liver was a result of active cell cycle and cell division suggesting that these cells are actively responding to accommodate the inflammation associated with disease. This expansion of LECs was a direct consequence of the increased expression of pro-proliferative and anti-apoptotic gene expression in LECs in the setting of chronic liver disease. Furthermore, LECs in diseased livers maintained high expression of the chemokine CCL21 suggesting that lymphatic vessels and lymphatic endothelial cells in particular may play an active role in immune cell recruitment, trafficking or programing during chronic liver disease.

When comparing different disease etiologies we found significant differences in the LEC transcriptional profile and pathways between NASH and HCV. Of the pathways that were differentially regulated we identified IL13 signaling as uniquely upregulated in LECs from individuals with end-stage NASH **(Figure 5** and **Supplementary Table 2)**. This pathway was interesting as IL13 has been identified as a pro-fibrogenic factor in the liver (52) as well as a factor involved in regulating lymphatic stability (35, 50). Indeed, our data points to a role for LECs in IL13 signaling through the direct stimulation of LECs with oxLDL. Our findings that IL13 is not produced by LECs either in livers from patients with HCV or in response to IFNα or polyI:C suggests that IL13 is not generally produced during inflammation. Instead, these findings suggest that IL13 production by LECs is a result of increased cholesterol, specifically oxLDL, found in the liver of patients with NASH. From these findings it is difficult to determine if release of IL13 from LECs results in either autocrine or paracrine signaling. However, our in vitro findings suggest that when IL13 is present, either by adding exogenous IL13 (35) or by inducing IL13 production by oxLDL, that LECs receive a signal to reduce PROX1 expression. Loss of PROX1 expression likely results in the decreased ability of LECs to maintain branched structures in vitro. Thus, we predict that while LECs receive signals to divide in order to accommodate the increased inflammation in the liver during disease, that in NASH the vessels become unstable and perhaps more permeable. These findings are intriguing as ascites is associated with chronic liver disease and may be due to increased permeability of the liver lymphatics (13). This is an important consideration when evaluating treatment options for patients with NASH compared to patients with HCV especially as NASH associated CLD is on the rise. Future studies will address whether IL13 production by LECs impacts liver specific cells such as hepatic stellate cells or is important for autocrine signaling within the LECs.

### ETHICS STATEMENT

All patients provided written and informed consent and the study was approved by the institutional review boards at the University of Colorado—Anschutz.

### REFERENCES


### AUTHOR CONTRIBUTIONS

BT and MB designed and executed experiments, analyzed results, and drafted the manuscript. JF performed experiments, analyzed the data, and critically reviewed the manuscript. AG, KR, RF, and JH designed experiments, analyzed single cell mRNA sequencing data and critically reviewed the manuscript. MK provided samples, designed experiments, and critically reviewed the manuscript. RS analyzed immunofluorescence data and critically reviewed the manuscript. HR provided samples and insightful discussion.

## FUNDING

This study was supported by funds from the GALIIP program, a grant from the National Institutes of Health (R01 AI121209) and the University of Colorado Department of Medicine Early Career Scholar/RBI Clinical Scholar Program to BT and a Gilead Sciences North American Liver Disease Research Scholars award to MB.

### ACKNOWLEDGMENTS

The authors thank members of the Consortium for Fibrosis and Translation (CFReT), GI and Liver Innate Immune Program (GALIIP), the Human Immunotherapy Initiative (HI3), and the RNA Bioscience Initiative (RBI) for support and tools used in this manuscript. Further, we would like to thank Kim Jordan and Angela Minic for acquisition and optimization of liver staining on the Vectra microscope and Rachael Dran for processing of explanted human livers at the time of transplant for downstream analysis. Finally, we would like to thank the Andrea Osypuk at the University of Colorado Biorepository Core facility for obtaining pathological samples used in this study.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01036/full#supplementary-material

biomarkers after virological cure in patients with chronic hepatitis. J Hepatol C. (2013) 59:675–83. doi: 10.1016/j.jhep.2013.05.015


**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 © 2019 Tamburini, Finlon, Gillen, Kriss, Riemondy, Fu, Schuyler, Hesselberth, Rosen and Burchill. 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(s) 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.

# Lymphatic Migration of Immune Cells

Henry R. Hampton<sup>1</sup> and Tatyana Chtanova2,3 \*

*1 Institute for Systems Biology, Seattle, WA, United States, <sup>2</sup> Immunology Division, Garvan Institute of Medical Research, Sydney, NSW, Australia, <sup>3</sup> Faculty of Medicine, St. Vincent's Clinical School, University of New South Wales Sydney, Kensington, NSW, Australia*

Lymphatic vessels collect interstitial fluid that has extravasated from blood vessels and return it to the circulatory system. Another important function of the lymphatic network is to facilitate immune cell migration and antigen transport from the periphery to draining lymph nodes. This migration plays a crucial role in immune surveillance, initiation of immune responses and tolerance. Here we discuss the significance and mechanisms of lymphatic migration of innate and adaptive immune cells in homeostasis, inflammation and cancer.

Keywords: lymphatic, migration, neutrophils, T cells, dendritic cells

#### Edited by:

*Amanda W. Lund, Oregon Health & Science University, United States*

#### Reviewed by:

*Alanna Ruddell, University of Washington, United States Maria M. Steele, Oregon Health & Science University, United States*

#### \*Correspondence:

*Tatyana Chtanova t.chtanova@garvan.org.au*

#### Specialty section:

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

Received: *31 January 2019* Accepted: *08 May 2019* Published: *28 May 2019*

#### Citation:

*Hampton HR and Chtanova T (2019) Lymphatic Migration of Immune Cells. Front. Immunol. 10:1168. doi: 10.3389/fimmu.2019.01168*

#### INTRODUCTION

The lymphatic system transports fluids from the periphery back into the circulatory system (1) using a series of open-ended capillaries known as lymphatic vessels (2). This network can be exploited by pathogens to facilitate rapid spread throughout the host (3). To prevent pathogen dissemination and enable a fast targeted immune response, the lymphatic system possesses filterlike structures termed lymph nodes (LNs) (3), where innate immune cells, such as macrophages, neutrophils and dendritic cells (DCs) trap and kill pathogens (3) and activate the adaptive immune response (4).

There are two routes by which immune cells can enter LNs: leukocytes can arrive from the bloodstream by crossing high endothelial venules (HEVs) (5). Alternatively, tissue-resident immune cells can enter afferent lymphatic vessels and migrate to draining LNs (dLNs) (5–8). Cells of the innate immune system including DCs, neutrophils, monocytes as well as adaptive immune leukocytes, such as T and B cells use lymphatic vessels to migrate from tissues into LNs (6– 11). Lymphocytes exit LNs via efferent lymphatic vessels, and eventually return to the circulatory system by the thoracic duct (12), however, in this review we will focus on the mechanisms and consequences of immune cell migration via the afferent lymphatic system.

In vitro and ex vivo models including adhesion and transmigration assays and analysis of immune cell migration in explanted skin provided important mechanistic insight into leukocyte entry and migration within lymphatic vessels (13–17), while in vivo approaches allowed to examine this complex biological process in situ (**Table 1**). Historically, in vivo analysis of immune cell migration in afferent lymphatics involved direct transfer of immune cells into the skin, lymphatic cannulation, as well application of fluorescent sensitizers to the skin to label cells and induce inflammation (18). More recently photoconvertible transgenic mice have been utilized to track immune cell migration from the skin and tumors (6, 9, 19–23), while intravital imaging approaches, such as in vivo two-photon microscopy enabled direct visualization of immune cell migration in lymphatic vessels (6, 16, 24, 25).

TABLE 1 | Methods for investigating immune cell migration *via* afferent lymphatic vessels.


### LYMPHATIC MIGRATION OF INNATE IMMUNE CELLS

#### Dendritic Cells

There are two distinct DC populations: plasmacytoid, which produce high amounts of type 1 interferon, and conventional DCs (cDCs) (26). Upon sensing inflammatory stimuli, cDCs enter lymphatic vessels and migrate to LNs (26, 27). They carry antigens (8) and pathogens including viruses (28, 29), spores (30) and bacteria (31–33) from the site of infection to LNs, while DC-mediated transport of innocuous antigens regulates tolerance (34). In dLNs DCs present antigen to CD4<sup>+</sup> T cells, or cross-present to CD8<sup>+</sup> T cells, thereby regulating adaptive immune responses (26). DC migration from the periphery has been discussed extensively in several recent reviews (27, 35). Here we provide a brief overview of the mechanisms of DC migration via lymphatic vessels (**Figure 1**).

The most important regulator of DC migration is the chemokine receptor CCR7. Consistent with this, CCR7-deficient DCs show a 90 percent reduction in migration from the periphery in response to inflammatory stimuli (10, 36). An elegant intravital microscopy study demonstrated that CCR7 is required for the LPS-induced directed migration of DCs toward lymphatic vessels and subsequent transmigration (25). DCs also use CCR7 for trafficking to dLNs from the lamina propria (37), lung (34), and skin (23) under homeostatic conditions. Interestingly, CCR7-dependent DC migration decreases lymphatic permeability (38), indicating bidirectional communication between the lymphatic network and immune cells.

Lymphatic vessels express CCR7 ligand, CCL21 (25, 39– 41), which is required for DC trafficking from skin to LNs under homeostatic (41) and inflammatory (42) conditions. Imaging studies have provided important insight into the role of CCL21 in DC migration. Firstly, the size of the CCL21 gradient, and distribution of lymphatic vessels, indicates that most skin DCs are able to sense CCL21 gradients (41). Secondly, DCs enter lymphatic vessels at sites of high CCL21 expression (25), suggesting that CCL21 directly regulates entry into lymphatics. Finally, intravital microscopy has revealed that CCL21 also enhances DC migration within lymphatic vessels (40). Collectively, these observations suggest that CCL21 regulates multiple steps in the lymphatic migration of DCs. In contrast, the other CCR7 ligand, CCL19, appears to be dispensable for DC lymphatic trafficking (42).

Inflammatory mediators regulate DC lymphatic migration (26). The cytokines Interleukin-1β (IL-1β) and Tumor Necrosis Factor-α (TNF-α) promote inflammation-induced DC migration to LNs (43–45). Furthermore, the lipid prostaglandin E2 increased CCR7 expression on DCs, augmenting migration towards CCL19 and CCL21 in vitro (46). Additional regulators of CCR7-mediated migration include the cell surface molecules

CD37, CD38, and CD47 which enhance DC movement toward CCR7 ligands and migration to dLNs (15, 47–50). In contrast, immunosuppressive molecules including IL-10 (51), TGF-β (52, 53) and the anti-inflammatory lipid Resolvin E1 (54) can inhibit DC trafficking.

In addition to CCR7, a number of other chemokine receptor/ligand pairs have been implicated in lymphatic DC migration. CXCR4/CXCL12 and CX3CR1/CX3CL1 enhance DC trafficking from inflamed skin (55, 56). The role of CCR8 is less clear with CCR8-deficient mice displaying reduced lymphatic migration of DCs following an injection of latex beads (57), but enhanced migration of DCs following FITC painting (58), suggesting that CCR8 plays a limited, or stimulus-specific, role in this process.

Intravital imaging and FITC-painting experiments have demonstrated that integrins, and integrin signaling are required for inflammation-induced DC migration to dLNs (14, 59, 60). However, DCs from mice lacking all integrins were able to migrate to LNs when injected into resting skin (61), indicating that integrins are important for DC migration in response to inflammation but dispensable for steady state DC egress. Accordingly, inflammation increases the expression of integrin ligands on lymphatic endothelial cells (LECs) (14). DC-expressed L1 cell adhesion molecule guides transendothelial migration of DCs thereby promoting trafficking to dLNs (17). A recent study demonstrated that interactions between LEC-expressed LYVE-1 and hyaluronan on the DC plasma membrane mediated DC adhesion and transmigration across LECs and subsequent migration to dLNs (13).

Sphingosine-1-phosphate (S1P), a lipid mediator of leukocyte egress from lymphoid organs (62), has been implicated in DC trafficking from the skin and lung (33, 63–65). However, in mice that lack S1P in lymphatic fluid, but not blood, the migration of adoptively transferred DCs to dLNs was comparable to that seen in wild-type mice (66). These results, and the fact that there are five S1P receptors (63), suggest that further experiments are required to uncover the precise role of S1P signaling in DC migration.

In contrast to cDCs, the lymphatic migration of pDCs is poorly understood. While one study reported that adoptively transferred pDCs migrated to dLNs from ovine skin (67), another showed that pDCs were not detected in the lymph of rats (68). However, pDCs transported harmless inhaled antigen from murine lungs to the mediastinal LN where they suppressed T cell activation, suggesting that pDC migration may play a role in preventing inflammation (69).

#### Neutrophils

Neutrophils are the first immune cells recruited to sites of inflammation, where they kill pathogens and release mediators that recruit other leukocytes (70, 71). Until recently neutrophils were thought to die at inflammatory foci. However, several groups, including ours, have shown that neutrophils can enter tissue lymphatic vessels and migrate to dLNs from the site of inflammation (6, 72–74).

Intravital imaging of inflamed mouse skin has enabled direct visualization of neutrophil migration within lymphatic vasculature (**Figure 2A**) (6, 72). However, in comparison to DCs, the significance and extent of neutrophil lymphatic migration are incompletely understood. Cannulation experiments have demonstrated that inflammation leads to a dramatic increase in neutrophils in ovine afferent lymph (8, 75, 76). Furthermore, neutrophils can transport antigens and microorganisms (**Figure 2B**) from the site of infection to LNs (6, 8, 77). Accordingly, inhibiting neutrophil lymphatic migration reduced early lymphocyte proliferation (6). Notably, a recent study did not detect substantial lymphatic migration of neutrophils in response to Staphylococcus aureus (S. aureus) (78). This likely highlights the fact that most neutrophils arrive in dLNs from the circulation via HEVs in response to bacteria already in dLNs, while a smaller population of neutrophils migrates directly from the site of inflammation to dLNs via afferent lymphatics. However, since neutrophils are the first innate immune cell subset to arrive in the LN from inflamed tissues, and often carry microbes, neutrophil lymphatic migration can exert considerable influence on the subsequent adaptive immune response (6, 77, 79).

Lymphatic migration of neutrophils could potentially be exploited by pathogens to enhance dissemination, since some microorganisms including the bacterium S. aureus can survive inside neutrophils (80). Consistent with this, injection of Leishmania major-containing neutrophils was sufficient to establish infection in mice, while depleting neutrophils reduced Leishmania burden when the pathogen was injected into the skin (81). Neutrophils also transported live Mycobacterium bovis bacille Calmette-Guérin from the skin to dLNs (77). In Toxoplasma gondii infection neutrophils removed the macrophages that line the subscapular sinus of the LN (82), however, it is not clear whether this favors pathogen control or spread.

CCR7 appears to be less important for neutrophil lymphatic migration than for that of DCs. Although it was required for neutrophil entry into lymphatic vessels in response to TNF-α and Complete Freund's Adjuvant (CFA) (72), and for CFA-driven migration from skin to LNs (83), neutrophil migration from the skin to dLNs in response to S. aureus was CCR7-independent (6). This suggests that the requirements for neutrophil trafficking vary depending on the stimulus and additional molecules may play key roles in guiding this migration.

The chemokine receptor CXCR4 regulates neutrophil migration from the bone marrow into the circulation (84) and may also play a role in their lymphatic migration. Inhibiting CXCR4 decreased neutrophil trafficking in response to immune complexes and S. aureus (6, 73). However, CXCR4 was not required for neutrophil entry into lymphatic vessels in response to CFA, as revealed by confocal imaging (72), again highlighting differences in neutrophil trafficking in response to distinct stimuli.

The cell surface receptor CD11b, which is involved in neutrophil recruitment from the vasculature into tissues (85), is emerging as a major regulator of neutrophil migration via lymphatics since neutrophil migration from inflammatory foci to LNs is substantially reduced when CD11b is inhibited (6, 73, 74). Intravital imaging demonstrated that blocking CD11b or its ligand ICAM-1 impaired neutrophil intraluminal crawling within lymphatic vasculature following CFA injection by reducing neutrophil speed and directionality (72). Likewise, inhibiting CD11b and ICAM-1 reduced neutrophil entry into lymphatic vessels and diminished egress from the skin in response to Mycobacterium bovis (74). Lymphocyte Functionassociated Antigen (LFA-1), which binds to ICAMs and is involved in neutrophil entry into tissues from the circulation, was required for neutrophil migration via afferent lymphatics in response to immune complexes (73) but not S. aureus (6).

Inflammatory cytokines may enhance neutrophil entry into lymphatics. TNF-α promoted neutrophil entry and crawling within lymphatic vessels in mouse cremaster muscle (72). However, since inflammatory cytokines also control neutrophil recruitment to sites of inflammation and lifespan, identifying a distinct role for these molecules in neutrophil lymphatic migration requires further investigation.

#### Monocytes and Macrophages

Monocytes are circulating leukocytes that phagocytose and kill bacteria and fungi and regulate the activity of other immune cells via cytokine release (86–88). They can also differentiate into DC and macrophage subsets (89). Several studies have demonstrated that monocytes egress tissues via afferent lymphatic vessels and transport antigen to dLNs (7, 8, 90–92). Once there, monocytes may present and cross-present antigens since a subcutaneous injection of antigen-pulsed monocytes induced the proliferation of antigen-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cells (91). The molecular mechanisms controlling monocyte migration via lymphatic vessels are yet to be identified. However, CCR7 may be important, since CCR7-deficient LPS-primed monocytes failed to migrate from the footpad to the popliteal LN (91). Accumulating evidence suggests that macrophages can also migrate from inflammatory lesions to dLNs (93–95) and that α1β1 integrin may limit macrophage egress via afferent lymphatics (93).

### LYMPHATIC MIGRATION OF ADAPTIVE IMMUNE CELLS

#### T Cells

T cells possess a rearranged T cell receptor which includes either αβ or γδ polypeptides (96). While αβ T cells are more abundant, γδ T cells are enriched in epithelial and mucosal tissues where they act as the first line of defense against pathogens. One of the main functions of CD8<sup>+</sup> T cells is to kill infected cells (97), while CD4<sup>+</sup> helper T cells secrete cytokines and regulate the function of other immune subsets (98). Photoconversion experiments (22), along with lymphatic cannulation (99), have demonstrated that effector, rather than naïve, T cells comprise the majority of lymph migrating T cells under homeostatic (22, 99) and inflammatory conditions (22, 100). This migration plays an important role in immune surveillance and in resolution of inflammation (18, 101–103).

Like DCs, T cells use CCR7 to migrate to LNs under homeostatic and inflammatory conditions. Following antigen challenge, T cells overexpressing CCR7 were preferentially lost from the lung and accumulated in the mediastinal LNs (104), while CCR7-deficient T cells failed to migrate from the footpad to the popliteal LN (11). However, the need for CCR7 in T cell migration might be context dependent, as T cells used CCR7 for trafficking from acute, but not chronically, inflamed skin (100). Tumor-infiltrating T cells could also emigrate to dLNs independently of CCR7 (9).

The lipid S1P, which promotes αβ T cell exit from LNs (62), can also mediate their migration via afferent lymphatics. Consequently, antagonizing S1P receptors led to T cell accumulation near skin lymphatic vessels and reduced migration to dLNs (100, 105). LEC-expressed macrophage mannose receptor (MR) and the cell surface molecule CD44, which interacts with MR, promoted T cell lymphatic migration by increasing T cell adhesion to lymphatic vessels (106, 107). Another LEC-expressed protein, CLEVER-1, was also demonstrated to be important for T cell migration via afferent lymphatics (108). Additionally, intravital imaging has shown that LFA-1 and its ligand ICAM-1, increased T cell velocity within afferent lymphatic vessels thereby promoting T cell migration to dLNs (16).

T cell egress from tissues to dLNs can either promote inflammatory responses or suppress them. For example, lymphatic migration of Regulatory T (Treg) cells may suppress allograft rejection (103). Likewise, CCR7-deficient Treg cells failed to migrate to the draining LN and accumulated in the skin, reducing skin inflammation during a delayed-type hypersensitivity reaction (101). Conversely, overexpression of CCR7 on antigen-specific Th1-cells enhanced their egress to dLNs and led to faster resolution of the inflammatory response in the skin (102).

Like αβ T cells, γδ T cells can migrate from the skin (23, 95, 109) and tumors (9) via lymphatic vessels to dLNs and comprise a large proportion of the cells in bovine lymph (110). Photoconversion experiments have demonstrated that murine γδ T cells can migrate from the skin to dLNs in the absence of CCR7 (23). Similarly, bovine γδ T cells can egress tissues independently of this receptor (111). The consequences of γδ T cell lymphatic migration are poorly understood, though it may enhance CD8<sup>+</sup> T cell proliferation (23).

## B CELLS

Cannulation experiments in sheep (112) and photoconversion experiments in mice (95), suggest that B cells use afferent lymphatic vessels to migrate from tissues to dLNs. The mechanisms of this migration are not yet known, however, at least in chronic inflammation, B cell egress may be independent of S1P and requires CCR7 (100). Similarly to T cells, blocking CLEVER-1 reduced B cell migration to dLNs (108).

### LYMPHATIC MIGRATION OF IMMUNE CELLS IN DISEASE

The importance of immune cell migration via lymphatics in host defense is illustrated by the observations that mice lacking CCR7 are susceptible to microbial and viral infections (113– 115). On the other hand, lymphatic migration of immune cells may also augment autoimmunity since preventing immune cell trafficking from the meninges to the cervical LNs reduced the severity of EAE (116). Furthermore, higher densities of lymphatic vessels in transplanted corneas (117) and kidneys (118) were associated with rejection, while preventing DC migration to the dLN by blocking Vascular Endothelial Growth Factor C (VEGF-C) improved corneal transplantation outcomes (119). Interestingly, in mice, obesity was associated with decreased lymphatic function and reduced immune cell migration to dLNs (120), suggesting that obesity may be linked to decreased immunity. CCR7 as well as its two ligands, CCL19 and CCL21, have been identified in mouse and human atherosclerotic lesions (121), consistent with accumulating evidence of a role for immune cell lymphatic migration in heart disease (122–125).

Lymphatic vasculature also plays a crucial role in tumor immunity by enabling transport of antigens from tumors to dLNs and egress of immune cells (9, 126). Lymphatic vessels can also serve as conduits for tumor cell spread (1). Their dual role in cancer is highlighted by the findings that many tumors overexpress VEGF-C, which promotes the growth and survival of LECs (127, 128) leading to increased LN metastasis (128–130). Furthermore, a recent study demonstrated that IFNγinduced PD-L1 expression by LECs may dampen anti-tumor immunity by limiting cytotoxic CD8<sup>+</sup> T cell accumulation in melanoma (131). On the other hand, overexpression of VEGF-C in a mouse melanoma model increased DC migration to dLNs (132). Consistent with an increase in lymphatic migration in cancer, enhanced trafficking of adoptively transferred B and T lymphocytes from footpads to dLNs was observed in melanomabearing mice (133).

Similarly to DC migration from the periphery, the DCdependent transfer of antigen from B16F10 melanoma to the dLN required CCR7. This correlated with an increase in CD8<sup>+</sup> T cell priming and a reduction in tumor growth (126). Overexpression of TGF-β1 in a model of squamous cell carcinoma reduced DC trafficking to dLNs, which led to

#### REFERENCES


an increase in LN metastasis (53). However, since CCR7 and the TGF-β1 receptor are not restricted to DCs (134), other immune cells may also contribute to these effects on tumor growth and metastasis. In contrast to DCs, photoconversion experiments demonstrated that tumor-infiltrating T cells do not require CCR7 to migrate to dLNs via afferent lymphatics (9).

#### CONCLUDING REMARKS

Lymphatic migration of immune cells presents opportunities for control of immune responses in infection and homeostasis. However, with the exception of DCs and T cells, the mechanisms controlling lymphatic migration of immune cells remain poorly understood. New tools, such as photoconversion and intravital imaging are poised to provide novel insight into the migration of previously overlooked immune subsets. A better understanding of the distinct mechanisms guiding lymphatic migration of specific immune subsets may suggest new approaches for treatment of cancer, autoimmunity and excessive inflammation.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work was funded by the National Health and Medical Research Council (NHMRC) project grant GNT1106043 (TC) and a Postdoctoral Fellowship from the Global Lyme Alliance (HH).


Salmonella enterica serovar typhimurium and limit systemic disease in mice. Infect Immun. (2009) 77:3170–80. doi: 10.1128/IAI.00272-09


CD14<sup>+</sup> CD16<sup>+</sup> monocytes, promote Th17 responses to Candida albicans. Eur J Immunol. (2011) 41:2915–24. doi: 10.1002/eji.201141418


**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 © 2019 Hampton and Chtanova. 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(s) 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.

# PD-1 Blockade During Post-partum Involution Reactivates the Anti-tumor Response and Reduces Lymphatic Vessel Density

Beth A. Jirón Tamburini 1,2 \*, Alan M. Elder 3,4, Jeffrey M. Finlon<sup>1</sup> , Andrew B. Winter <sup>1</sup> , Veronica M. Wessells 3,4, Virginia F. Borges 3,4 and Traci R. Lyons 3,4 \*

<sup>1</sup> Division of Gastroenterology and Hepatology, Department of Medicine, School of Medicine, University of Colorado Anschutz Medical Campus, Denver, CO, United States, <sup>2</sup> Department of Immunology and Microbiology, Department of Medicine, School of Medicine, University of Colorado Anschutz Medical Campus, Denver, CO, United States, <sup>3</sup> Division of Medical Oncology, Department of Medicine, School of Medicine, University of Colorado Anschutz Medical Campus, Denver, CO, United States, <sup>4</sup> Young Women's' Breast Cancer Translational Program and University of Colorado Cancer Center, Aurora, CO, United States

#### Edited by:

Ursula Grohmann, University of Perugia, Italy

#### Reviewed by:

Lianjun Zhang, Suzhou Institute of Systems Medicine (ISM), China Cornelia Halin, University of Zurich, Switzerland

#### \*Correspondence:

Beth A. Jirón Tamburini Beth.Tamburini@ucdenver.edu Traci R. Lyons Traci.Lyons@ucdenver.edu

#### Specialty section:

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

Received: 07 December 2018 Accepted: 23 May 2019 Published: 11 June 2019

#### Citation:

Tamburini BAJ, Elder AM, Finlon JM, Winter AB, Wessells VM, Borges VF and Lyons TR (2019) PD-1 Blockade During Post-partum Involution Reactivates the Anti-tumor Response and Reduces Lymphatic Vessel Density. Front. Immunol. 10:1313. doi: 10.3389/fimmu.2019.01313 Post-partum breast cancer patients, or breast cancer patients diagnosed within 10 years of last childbirth, are ∼3–5 times more likely to develop metastasis in comparison to non-post-partum, or nulliparous, patients. Additionally, post-partum patients have increased tumor-associated lymphatic vessels and LN involvement, including when controlled for size of the primary tumor. In pre-clinical, immune-competent, mouse mammary tumor models of post-partum breast cancer (PPBC), tumor growth and lymphogenous tumor cell spread occur more rapidly in post-partum hosts. Here we report on PD-L1 expression by lymphatic endothelial cells and CD11b+ cells in the microenvironment of post-partum tumors, which is accompanied by an increase in PD-1 expression by T cells. Additionally, we observed increases in PD-L1 and PD-1 in whole mammary tissues during post-partum mammary gland involution; a known driver of post-partum tumor growth, invasion, and metastasis in pre-clinical models. Importantly, implantation of murine mammary tumor cells during post-partum mammary gland involution elicits a CD8+ T cell population that expresses both the co-inhibitory receptors PD-1 and Lag-3. However, upon anti-PD-1 treatment, during post-partum mammary gland involution, the involution-initiated promotional effects on tumor growth are reversed and the PD-1, Lag-3 double positive population disappears. Consequently, we observed an expansion of poly-functional CD8+ T cells that produced both IFNγ and TNFα. Finally, lymphatic vessel frequency decreased significantly following anti-PD-1 suggesting that anti-PD-1/PD-L1 targeted therapies may have efficacy in reducing tumor growth and dissemination in post-partum breast cancer patients.

Keywords: lymphatic endothelial cells, post-partum breast cancer, metastasis, immunotherapy, PD-L1, PD-1, T cells

**188**

## INTRODUCTION

Post-partum breast cancer patients in our cohort, or breast cancer patients diagnosed within 10 years of last childbirth, are ∼3– 5 times more likely to develop metastasis in comparison to non-post-partum, or nulliparous, patients (1, 2). Additionally, post-partum patients have increased LN involvement and peritumor lymphatic vessel density (LVD) (2–4). In 2010, Asztalos et al. identified alterations to gene expression patterns in normal mammary tissues from post-pregnant women that persist up to 10 years post-partum in comparison to nulliparous women (5). Specifically, they observed changes to genes involved in inflammation, angiogenesis, extracellular matrix (ECM), and breast cancer; suggesting that the mammary microenvironment after pregnancy may be conducive to malignancy. Consistent with this hypothesis, recent pregnancy can increase a woman's risk for developing breast cancer for more than 20 years following childbirth (6–8). Normal mammary gland development associated with pregnancy consists of a period of expansion of the mammary epithelium, to prepare the gland for lactation, followed by full differentiation of the mammary epithelium into milk secreting cells. Following lactation, or pregnancy in the absence of lactation, post-partum mammary gland involution occurs to return the mammary epithelium to the pre-pregnant state. Previous studies of normal post-partum mammary gland involution in rodents and women have revealed that attributes of this normal developmental process are similar to those observed in breast tumors (9–12). These attributes include establishment of a tissue microenvironment that is characterized by ECM remodeling, increased LVD, immune infiltration and evidence of immune suppression (3, 13–15). Additionally, non-metastatic tumor cells implanted into this tissue microenvironment in pre-clinical models grow and invade more rapidly, seed micro-metastases, and are durably altered to a more invasive and metastatic state; suggesting that post-partum involution can drive intrinsic, prometastatic changes, in tumor cells (3, 16). These findings predict that the process of normal involution may drive post-partum breast cancer (PPBC) metastasis.

Post-partum mammary gland involution, induced by weaning, has been extensively studied in rodents where it is characterized by two phases of tissue remodeling. The first, known as the reversible phase, is triggered by milk stasis and results in death of the secretory mammary epithelium (17, 18). The second phase, known as the irreversible phase, consists of stromal remodeling and repopulation of the gland with adipocytes. Insight into molecular programs that govern this developmental process in mice has been gained through gene expression profiling studies on whole mammary tissues where roles for death receptors and immune mediators were revealed (10, 11, 19). Additionally, influx of immunosuppressive Foxp3+ regulatory T cells and IL-10+ macrophages occurs during involution resulting in effector T cell suppression (13). Furthermore, M2-like or tissue repair type macrophages and macrophages with pro-lymphatic phenotypes are evident (3, 20, 21). As epithelial cell apoptosis during involution likely results in the increased presentation of self-antigens (22) it is not surprising that numerous cell types initiate an immune-tolerant microenvironment and, consequently, an environment that could be primed for post-partum tumor growth. Importantly, LVD also increases during mammary gland involution, presumably to promote the clearance of increased fluid—generated by milk stasis, apoptotic cell debris, and immune cell infiltrates. However, we have recently shown that the mammary lymphatics that arise during post-partum involution are also capable of transporting tumor cells to distant lymph nodes during the active phase of tissue remodeling and that post-partum patients are enriched for lymph node involvement (2, 21). These studies suggest that metastatic seeding via lymphatics may be an early event in post-partum patients, which may account for the increased metastasis observed (21).

Tumor-associated lymphatic vessels not only promote dissemination (23–27), but have recently been shown to reduce anti-tumor immune responses in tumor models (28, 29). Lymphatic endothelial cells (LECs) normally promote peripheral immune tolerance in the lymph node during homeostasis. Specifically, lymph node LECs express PD-L1 to inhibit autoreactive T cells via engagement of the inhibitory receptor Programmed Death-1 (PD-1) (30–34). In addition, inhibitory receptor expression of PD-1 is also an important marker of T cell effector function. However, upregulation of multiple inhibitory receptors, such as PD-1, Lag-3, and TIGIT can occur as a result of chronic antigen stimulation and lead to non-responsive T cells that fail to successfully clear the pathogen (35–40). In the cancer setting, a similar phenomenon occurs as tumor-infiltrating T cells upregulate multiple co-inhibitory receptors and are limited in their ability to produce multiple effector cytokines (such as IFNγ and TNFα) (41), making them less polyfunctional and unable to clear the tumor (42). Recently, several studies have pointed to a role for tumorassociated lymphatic and/or macrophage expression of PD-L1 in contributing to T cell inhibition (28, 29, 43, 44). In addition, we and others have published that PD-L1 expression by LECs promotes their survival during an immune response and a role for PD-L1 expression in promoting tumor cell survival has been demonstrated (45–48). Thus, PD-L1 clearly plays a role in promoting cell survival and immunosuppression in multiple cell types present in the tumor microenvironment (TME).

Since T-cell infiltration, immune suppression, macrophage infiltration, and lymphangiogenesis have all been described during post-partum involution (3, 13, 21), we sought to determine whether T cell expression of co-inhibitory receptors allows for immune evasion by tumor cells during post-partum involution and whether this mechanism could be reversed with anti-PD-1 treatment. In this manuscript, we investigate the immune regulatory state of the mammary tissue during postpartum involution and in tumors implanted during involution (post-partum tumors). We demonstrate that PD-L1 and PD-1

**Abbreviations:** LEC, Lymphatic endothelial cell; PDPN, Podoplanin; PD-1, Programmed Death-1; PD-L1, Programmed Death-Ligand 1; Lag-3, Lymphocyteactivation gene-3; LN, Lymph node; LVD, Lymphatic Vessel Density; TME, Tumor Micro Environment.

are a part of the involution program and show increased expression of PD-L1 on lymphatic endothelial cells, and cells of myeloid lineage, as well as PD-1 on T cells during mammary gland involution and in post-partum tumors. Importantly, we demonstrate that this mechanism is an integral part of the tumor promotional program effects of involution. Administration of an anti-PD-1 antibody to mice during involution, when the tumors are established, reduced growth of the post-partum tumors to levels observed in nulliparous hosts. Upon evaluation of the tumor infiltrating CD8+ T cells we discovered co-expression of two inhibitory receptors, PD-1 and Lag-3, that were specific to tumors established during involution. Following treatment with anti-PD-1, the PD-1+Lag-3+ inhibitory CD8+ T cell population disappeared and the frequency of total CD8+ T cells and polyfunctional CD8+ T cells increased significantly suggesting a reversal of at least some of the immunosuppressive effects of involution. Surprisingly, anti-PD-1 treatment also reduced tumor/involution associated LVD suggesting this treatment may have implications for stopping lymphatic-mediated metastasis. Our results lay the ground work for additional studies aimed at uncovering the potential role of the mammary LECs during involution, and in the tumor microenvironment, in promoting immunosuppression and suggest a potential treatment option for PPBC patients.

### MATERIALS AND METHODS

#### Animal Studies

All animal procedures were approved by the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee. BALB/c and C57Bl/6 (age 6–8 weeks) were obtained from Charles River Laboratories. Mice were crossbred for involution studies; C57Bl/6 female mice were bred in triad with BALB/c males and BALB/c females were bred with C57Bl/6 males. Age-matched nulliparous females were used as controls. At 10–14 days post-parturition, post-partum mammary gland involution was instigated in the bred females by force-weaning the pups. For normal involution studies, mammary glands, and draining lymph nodes (inguinal lymph nodes) were harvested from nulliparous and involution day 6 mice. Tumor studies for the 66cl4 and E0771 mouse mammary carcinoma models were performed as previously described (21) with the tumors, mammary glands, and lymph nodes being taken for flow cytometry, immunohistochemistry, and downstream biochemical analyses. For the E0771 PD-1 intervention studies, 250,000 tumor cells were injected into the number 4 mammary glands of either nulliparous or involution day 1 C57Bl/6 dams. Tumor sites were palpated daily (E0771) or twice weekly (66cl4) for tumors. Calipers were used to take measurements and the tumor volumes were calculated using length x width x width × 0.5. Additionally, 66cl4 tumor cells were luciferase and GFP tagged allowing for tumors to be detected using the Xenogen 200. Once tumors became measurable, mice were randomized into control or treatment groups and injected with 250 µg of either isotype control (anti-IgG2a [clone 2A3; Bio X Cell cat. #BP0089]) or anti-PD-1 (clone RMP1-14; Bio X Cell cat. #BP0146) antibody, respectively. Injections were administered intra-peritoneally every third day. Tumor studies were ended based on primary tumor cell growth or ulceration at 3–4 weeks post injection (66cl4) or 1–2 weeks post injection (E0771). In vivo studies were performed in triplicate with pooled or representative data shown.

### Mammary Gland Processing and Staining (IHC)

Mammary glands were harvested and placed into 10% neutral buffered formalin for 48 h. After 48 h, tissues were moved to 70% EtOH, processed, and stained for LYVE-1 as previously described (3, 14, 16, 21, 49).

### Lymphatic Vessel Density Quantification

Lymphatic vessel density (LVD) was performed as previously described (3, 21). Briefly, slides stained for PDPN (D2-40) or LYVE-1 were scanned into the Aperio ImageScope software. Lymphatic vessels were counted in the tumor-adjacent tissue (peri-tumor region) and LVD was quantified as the number of lymphatic vessels per area of tissue.

#### Human Tissue Acquisition

Research using de-identified human breast tissue (**Supplemental Table 1**) was conducted under a protocol deemed exempt from subject consent as approved by the Colorado Multiple Institution Review Board (COMIRB) and tissues were acquired by Virginia Borges as previously reported (1). Dr. Borges obtained written informed consent from the patients, the studies were conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule), and the studies were approved by an institutional review board.

#### Staining of Human Tissue Using Vectra

Four-micron thick sections were taken from Formalin Fixed Paraffin Embedded tissue, dewaxed in xylenes and rehydrated. Slides were placed in 10% NBF for 20 min for extra fixation, rinsed with DI water, then submerged in Target Retrieval Solution pH6 (Dako cat# S1699) and placed in a pressure cooker for 20 min. Slides were rinsed with Dako wash buffer (Dako cat# K8000), blocked for 10 min with Perkin Elmer Diluent/Block (Perkin Elmer cat# ARD1001EA), then sequentially stained for the following markers: PD-L1 (clone E1L3N), PD-1 (clone NAT105;), PDPN (clone D2-40), and CD68 (clone KP1). Incubation time for all primary antibodies was 1 h at room temperature. Slides were rinsed and stripped in Target Retrieval Solution in between every primary. Slides were then incubated in Perkin Elmer Opal Polymer HRP Mouse+Rabbit secondary (cat# ARH1001EA) for 30 min at room temperature, followed by a 10 min incubation in Opal Fluorophore reagents (Perkin Elmer). After the final stain, Spectral DAPI (Perkin Elmer cat# FP1490) was applied to slides for 5 min, then slides were rinsed and coverslipped with ProLong Diamond Antifade Mountant (Thermo cat# P36970). Multispectral imaging was then performed using the Vectra 3 Automated Quantitative Pathology Imaging System (Perkin Elmer). Whole slide scans were collected using the

10x objective and 5–10 regions were selected for multispectral imaging with the 20x objective. The multispectral images were analyzed with inForm software (Perkin Elmer) to unmix adjacent fluorochromes, subtract autofluorescence, segment the tissue into lymphatic vessels and non lymphatic vessels, segment the cells into nuclear, and membrane compartments, and to phenotype the cells according to morphology and cell marker expression. Cells with a PD-L1 threshold <0.95 were classified as PD-L1 negative while cells with a value >0.95 threshold were classified as PD-L1 positive using inForm software. To quantitate PD-L1 in lymphatics a blinded observer imaged 5–10 representative fields from the peritumor region that were positive for PDPN vessels by only the PDPN channel. For PDPN we also counted PD-L1+ lymphatic vessels by adding the PD-L1 channel and counted PDPN+PD-L1+ vessels as well as PDPN+PD-L1 vessels and calculated the percent positive per case, which was then normalized to area. PD-1+ cells were also counted in the same manner as PDPN+ vessels.

#### Flow Cytometry

Tumors were separated from the mammary gland. Both tumors and mammary glands were placed in six-well plates with 2 mL of Click's media without mercaptoethanol or L-glutamine (Irvine Scientific, Santa Ana, CA), where they were minced with scalpels, digested with 500 units/ml collagenase type II and IV and 20µg/ml DNase (Worthington Biochemical Corporation, Lakewood, NJ) and incubated for 1 h at 37◦C. The tissue suspension was then filtered through a 100µm strainer and washed with Click's. The filtered cells were centrifuged at 1,400 RPM for 5 min, the supernatant was removed, and the pellet was resuspended in 1 mL FACS buffer (500 mL 1x HBSS pH 7.4, 0.1% BSA, 0.02% sodium azide, up to 1L ddH2O). The tumor cells were stained with BD viability 510 dye prior to staining with CD45 (clone30-F11), CD8a APC/Cy7 (clone 53-6.7) (1:400), CD4 APC or PerCp-Cy5.5 (clone RM4-5) (1:300), PD-1 FITC or BV421 (clone 29F.1A12) (1:100), Lag-3 PerCP/Cy5.5 (clone C9B7W) (1:100), and/or CD11a FITC (clone M17/4) (1:200). The mammary glands were stained with BD viability dye 510 followed by CD45 APC-Cy7 or Pacific Blue (clone 30-F11) (1:300), CD31 Pacific Blue or PerCp-Cy5.5 (clone 390) (1:200), PDPN APC or PE-Cy7 (clone 8.1.1) (1:200), CD11b Pacific Blue or PerCp or FITC (clone M1/70)(1:400), F4/80 APC, APC-Cy7, PerCP-Cy5.5 or FITC (clone BM8) (1:100), PD-L1 PE, FITC, or BV421 (clone RMP1-30 or 29F.1A12) (1:200), and EpCAM PE-Cy7 or APC-Cy7 (clone G8.8) (1:100). Flow cytometry antibodies were purchased from Biolegend (San Diego, CA). CD8 T cells were identified from live, CD3+/CD8+, where they were further characterized by their expression of PD-1 and Lag-3. Lymphatic endothelial cells were identified from live, CD45-/EpCAM-, and CD31+PDPN+. Cells were run on the DakoCytomation CyAn ADP flow cytometer (Fort Collins, CO) or FACs Canto II, acquired using Summit software or Diva Software, and analyzed with FlowJo software (Tree Star, Ashland, OR). Geometric mean fluorescence intensity (gMFI) was calculated with FlowJo software.

### Intracellular Cytokine Staining

Cells were isolated from the tissue and treated with or without (unstimulated controls) phorbol 12-myristate 13-acetate (PMA) (20 ng/ml) (Sigma, St. Louis, MO) plus ionomycin (1 ug/ml) (Sigma, St. Louis, MO) for 4–6 h at 37 degrees in the presence of 2 ug/ml of brefeldin A (Adipogen, San Diego, CO) in RPMI+2.5% FBS. Cells were then stained with CD8, CD45, CD4, CD44, PD-1, and Lag-3 (as above) and incubated at 37◦C for 30 min. Following surface marker staining cells were fixed with 1% paraformaldehyde and 4% sucrose for 10 min in the dark at room temperature. Following fixation, cells were permeabilized with BD Perm Wash (BD Biosciences, San Jose CA) and stained for cytokines IFNγ (1:200) (APC; Biolegend clone XMG1.2) and TNFα (1:200) (FITC; Biolegend clone MP6-XT22). After washing cells were resuspended in FACs buffer (0.5% Bovine Serum Albumin and 0.1% Sodium Azide in PBS) and were run on an ADP Cyan. Gating was determined based on unstimulated controls. All antibodies were purchased from Biolegend (San Diego, CA).

#### TCGA RNASeq Analysis

Analysis was performed on cbioportal.org TCGA Breast provisional dataset using the co-expression tool and the RNA-Seq data.

#### Statistics

One-way ANOVA, unpaired t-test, and linear regression were run in the GraphPad Prism software, assuming normal distributions among independent samples. For **Figure 6**, Pearson and Spearman analysis was performed on cbioportal.org. P-values of <0.05 were deemed significant.

## RESULTS

### PD-L1 Expression and PD-1 T Cells Are Observed in Patients With PPBC and in Pre-clinical Models

To determine if the increased lymphatics (PDPN) that we observe in our patients with PPBC exhibit upregulation of PD-L1 and/or whether PD-1+ T cell infiltration occurs in the tumor microenvironment (TME) of PPBCs, we utilized multispectral imaging of the peri-tumor region in tissues from three different patients with PPBC, who were within 1-year (PPBC1), 3 years (PPBC2), and 4 years post-partum (PPBC3) (**Supplemental Table 1**). In comparison to three non-PPBC patients, who were all nulliparous, we observed increased PDPN+ LVD in the peri-tumor region from our PPBCs. We also observed frequent lymphatic vessel expression of PD-L1 (arrow) and that lymphatic vessels were frequently infiltrated with PD-L1 expressing tumor cells (asterisk), identified by their altered nuclear morphology. Furthermore, we also observed the presence of CD68+ macrophages that appear to express PD-L1 (+ symbol) as well as cells expressing PD-1 (arrowhead) in the surrounding areas (**Figure 1A** and **Supplemental Figure 1**). We quantitated PD-L1+ cells in the peritumor region and observed that total numbers of PD-L1+ cells per area did not differ between groups (**Figure 1B**), but that total PDPN+ vessel

FIGURE 1 | from A. (C) Number of PDPN+ vessels in the peritumoral region of the patients from A. (D) Number of PDPN+ vessels that are also PD-L1+ in the peritumoral region of the patients from A. (E) Number of PD-1+ cells in the peritumoral region of the patients from A. (F) Histogram of PD-L1 expression by LECs (CD45-EpCAM-CD31+PDPN+PD-L1+) from 66CL4 tumors implanted during mammary gland involution. Gray dotted line indicates a PD-L1 negative population (CD45+CD11b-F4/80-) that does not change. (G) gMFI of PD-L1 of LECs in tumor. (H) Histogram of PD-L1 expression by CD45+ CD11b+ cells from tumors implanted during mammary gland involution. Gray dotted line indicates a PD-L1 negative population (CD45+CD11b-F4/80-) that does not change. (I) gMFI calculated for PD-L1 from D. Data shown from animal experiments are from 1 representative experiment of 2 replicates with at least 4 tumors per group. Unpaired t-test: \*p < 0.05; \*\*p < 0.01. Scale bars are 100 microns in length.

density, as well as PDPN+PD-L1+ vessel density, were increased in our patients with PPBC compared to nulliparous controls (**Figures 1C,D**) suggesting that PD-L1 lymphatics are a part of the TME in patients with PPBC. Finally, we observe a significant increase in PD-1+ cells in the TME of our PPBC patients (**Figure 1E**).

Since we observed PD-L1 expression in cells of the TME from patients with PPBC, we next asked if PD-L1 expression was a characteristic of the TME of post-partum tumors in our murine model of PPBC. Using the 66cl4 isograft model where we have shown increased lymph vessel density (LVD) and lymph node metastasis in post-partum hosts (3), we examined post-partum tumor-associated LEC expression of PD-L1 in Balb/c mice. As previously observed, tumor cells implanted on day 1 of involution (involution group tumors) exhibited decreased latency and increased growth compared to nulliparous (**Supplemental Figure 2A**) (3, 13, 16). At study endpoint, 4 weeks post-injection, tumors were harvested, and populations were analyzed by flow cytometry (**Supplemental Figures 2B,C**). We observed an increase in the fluorescence intensity of PD-L1 on CD45-Epcam-CD31+PDPN + tumor associated LECs (**Figures 1F,G**), which we also observed with implantation of E0771 tumors into BL6 mice (**Supplemental Figure 3A**). We found that while there were similar frequencies of the monocyte (CD11b+) populations in both tumor models (**Supplemental Figure 3B** and not shown), CD11b+ cells in both 66cl4 and E0771 tumors implanted during mammary gland involution had higher average levels of PD-L1 (**Figures 1H,I** and **Supplemental Figure 3C**). The number of PD-L1+ cells was also increased in the involution group, but this increase was lost when normalized to tumor size (**Supplemental Table 2**). We also analyzed additional cell populations for PD-L1 including monocytes, fibroblasts, blood endothelial cells (BECs), and EpCAM+ tumor cells based on described markers. We did not see significant staining differences or staining above background in the fibroblasts, BECs or EpCAM+ cells (**Supplemental Figure 3D**). These results complement our recently published data showing that CD11b+ monocytes in the TME of involution group tumors contribute to lymphangiogenesis and extend our observations to describe their expression of PD-L1 (21).

As PD-L1 is the inhibitory ligand, we next examined expression of the inhibitory receptor, PD-1, by CD4+ and CD8+ tumor-associated T cells (gating-**Supplemental Figure 2D**) all of which are also positive for CD11a, a molecule that has been shown to render them unable to control tumor growth (50). We found an increase in the expression of PD-1 on CD4+ T cells (**Figure 2A**) and after quantification we found that an average of 30% of CD4+ T cells in the involution group tumors expressed PD-1 compared to 10% in the nulliparous controls (**Figure 2B**). As cytotoxic CD8 T cells are typically thought to be important in controlling tumors we next asked about expression of PD-1 on CD8+ T cells. We observed a striking difference in the expression profile of CD8+ T cells in tumors implanted during mammary gland involution compared to tumors implanted in nulliparous hosts (**Figure 2C**). When we quantified this difference, we found a >4-fold increase in expression of PD-1 by CD8+ T-cells from involution group tumors (**Figure 2D**). Additionally, we observed similar phenotypes in the mammary draining lymph nodes of tumor bearing animals (**Figure 2E**) suggesting this mechanism of tumor-associated immune suppression may extend beyond the local tumor microenvironment to drive the increased LN metastasis that we observe in post-partum patients. Finally, we observed similar phenotypes in the E0771 tumors, but not lymph nodes (**Supplemental Figures 4A,B**).

#### PD-L1 and PD-1 Expression Are Observed in Mouse Mammary Tissues During Normal Post-Partum Mammary Gland Involution

We next asked if the LECs of C57/BL6 (B6) or Balb/c female mice exhibit expression of PD-L1 during normal post-partum mammary gland involution, similar to what is observed in LN LECs to induce peripheral tolerance during tissue homeostasis (31). We evaluated mouse mammary LECs at involution day 6, the peak of the remodeling phase of involution, for expression of PD-L1. PD-L1+ LECs were evaluated based on the markers described above and validated by staining with isotype and fluorescence minus one (FMO) controls (**Supplemental Figure 2B**). We observed that expression of PD-L1 was increased in mammary LECs isolated from involution group mice compared to nulliparous (**Figure 3A**). We also found that the percentage and number of LECs expressing PD-L1, as well as the geometric mean fluorescence intensity (gMFI) of PD-L1, was increased on LECs during involution (**Figures 3B–D**), which is similar to that observed in our pre-clinical model of PPBC (**Figures 1F,G**). We found these increases in both the B6 mice as well as the Balb/C mice (**Supplemental Figures 5A–C**). Since we observed increased PD-L1 expression in the CD11b+ population in our model of PPBC, we first confirmed that the CD11b+ population increased (gating-**Supplemental Figure 2C**) during involution in both B6 (**Figure 3E**) and Balb/C (**Supplemental Figure 5D**) mouse mammary glands. Then, we also observed an increase in expression of PD-L1 by this

FIGURE 2 | Tumor-associated PD-1<sup>+</sup> T cells are increased in tumors implanted during involution. (A) Representative flow plots for cells from nulliparous or involution group tumors of CD11a and PD-1 expression by CD4<sup>+</sup> T cells. (B) Quantification of CD4<sup>+</sup> T cells that express PD-1 from A. (C) Representative flow plots of CD8<sup>+</sup> T cell expression of PD-1 and CD11a from nulliparous and involution tumors and (D) quantification of PD-1<sup>+</sup> CD8<sup>+</sup> T cells. (E) Representative flow plots of CD11a and PD-1<sup>+</sup> CD4<sup>+</sup> T cells or CD8<sup>+</sup> T cells in the tumor draining lymph nodes and quantification of PD-1+ frequency of each cell type. Data shown are from 1 representative experiment of 2 replicates with at least 4 tumors per group. Unpaired t-test: \*p < 0.05; \*\*\*p < 0.001.

FIGURE 3 | PD-L1<sup>+</sup> lymphatic endothelial cells and PD-L1<sup>+</sup> myeloid cells increase in the mammary gland during post-partum involution. (A) Representative histogram of PD-L1 by lymphatic endothelial cells (CD45−EpCAM−CD31+PDPN+PD-L1+) acquired by flow cytometry from B6 mammary glands of nulliparous (n = 6) or recently weaned, involution day 6 (n = 6), mice. Gray dotted line indicates a PD-L1 negative population (CD45+CD11b-F4/80-) that does not change. (B) Quantification of the frequency of PD-L1+ LECs of total LECs per mammary gland. (C) Number of PD-L1<sup>+</sup> LECs per mammary gland. (D) Quantification of gMFI of PD-L1 expression by LECs from A. (E) Frequency of CD11b<sup>+</sup> cells of the CD45<sup>+</sup> cells found in the mammary gland in nulliparous or involution group mice. (F) Histogram of PD-L1 expression by CD11b<sup>+</sup> cells in mammary gland of nulliparous or involution group mice. Gray dotted line indicates a PD-L1 negative population (CD45+CD11b−F4/80−) that does not change. (G) Quantification of PD-L1 expression by gMFI on CD11b+ cells as shown in F. (H) Percent of CD4<sup>+</sup> T cells that are PD-1<sup>+</sup> from the mammary glands of B6 mice and (I) as in H except from Balb/c mice. (J) Percent of PD-1<sup>+</sup> CD4<sup>+</sup> T cells in the axillary lymph nodes from nulliparous and involution group tissues. Data shown are from 1 representative experiment of 2 replicates with at least 6 mammary glands per group. Unpaired t-test: \*\*p < 0.01; \*\*\*p < 0.001; \*\*\*\*p < 0.0001.

#### population (**Figure 3F**) that was significant (**Figure 3G**) in both models (**Supplemental Figure 5E**).

To better understand if PD-1 expression on T cells is similarly increased during post-partum mammary gland involution, we evaluated the T-cell compartment at involution day 6 by assessing the frequency of PD-1 expression by CD4+ and CD8+ T-cells in B6 and Balb/c mice compared to isotype controls (gating-**Supplemental Figure 2D**). Similar to previous results in Balb/c mice (51), we observed a significant increase in the percent of PD-1+ CD4+ T cells in mammary glands from B6 (**Figure 3H**) and confirmed this in our Balb/c mice (**Figure 3I**). This increase in PD-1+ CD4+ T cells also extended to the LN (**Figure 3J**). Further, we found an increase in CD8+ T cells expressing PD-1 in the B6 mice (**Supplemental Figure 5F**), which was not significant in the Balb/c mice (**Supplementary Figures 5G,H**). These results suggest that a mechanism of LEC and/or monocyte/macrophage mediated T cell inhibition could be driving the decreased latency and increased growth rate that we observe in our preclinical models when tumor cells are implanted during involution (16, 21, 52).

### PD-1 Targeted Therapy Reduces Tumor Growth in Post-partum Hosts by Reactivating T-Cells

A prediction of our results is that inhibition of PD-L1/PD-1 signaling during involution would dampen the increased tumor growth observed when tumors are implanted during involution by reversing this involution-driven mechanism of immune suppression. To test our hypothesis, we utilized the E0771 mouse mammary tumor model since tumors implanted in this model become palpable during active involution; when the immune suppressive mechanism is most activated. Thus, we orthotopically implanted E0771 tumor cells into the intact mammary glands of C57BL/6 mice at involution day 1 or into nulliparous hosts. Then, we blocked PD-L1-mediated inhibition of T cells by administering an anti-PD-1 monoclonal antibody every third day after tumors were palpable and size-matched in both nulliparous and involution groups. Following two and three treatments, in involution and nulliparous mice, respectively, mice were euthanized and flow cytometry performed on tumors (**Figure 4A**). Similar to previous results, involution group tumors exhibited decreased latency and increased growth compared to tumors in nulliparous hosts (**Figure 4B**) (3, 13, 16). Importantly, the anti-PD-1 treatment significantly reduced the growth rate in involution group tumors, to levels more similar to those observed in the nulliparous hosts, and did not significantly affect tumor growth in the nulliparous group (**Figure 4B**). We then evaluated whether the anti-PD-1 treatment affected immune cell infiltration into the tumors in the involution group. Immunohistochemistry (IHC) on a single tumor from each involution group revealed increased intratumoral staining for CD45, which was validated by our flow cytometry where we observed a >2-fold increase in the number of CD45+ cells in the tumor (**Figures 4C,D**). We also observed a significant increase in the number of intra-tumoral CD8+ T cells following treatment with anti-PD-1 when the tumors were implanted into involution hosts (**Figure 4E**). Additionally, we found a significant decrease in LEC frequency in the adjacent mammary tissues of involution group mice treated with anti-PD-1 by both flow cytometry (**Figure 4F**) and by tissue staining with PDPN to assess LVD (**Figure 4G**). Conversely, we found no significant difference in PD-L1 expression by LECs after anti-PD-1 treatment (**Supplemental Figure 6A**). These findings suggest that blockade of PD-1 could have potential for blocking both tumor growth and lymphogenous tumor cell spread during involution.

To evaluate whether anti-PD-1 treatment was affecting the phenotype and functionality of the tumor-associated T cells, we used flow cytometry to assess the co-expression of co-inhibitory markers, PD-1 and Lag-3, by the CD8+ T cells (**Figure 5A**) as well as PD-1 single positive expression by the CD8+ and CD4+ T cells (**Supplemental Figures 6B,C**). With treatment, we observed a decrease in the frequency of PD-1 and Lag-3 double positive cells in the involution group tumors, but not in the nulliparous tumors (**Figure 5B**). Importantly, the Lag-3 gMFI was also significantly decreased following anti-PD-1 treatment in the involution group, but not the nulliparous, (**Figure 5C**) and this effect was not due to the treatment antibody (clone RMP1-14) blocking the staining antibodies (clone RMP1-30 or 29F1A12) (**Supplemental Figure 6D**). We also observed that the frequency of the CD45+ cells that were also positive for CD8 was increased in involution group tumors, and the nulliparous, that were treated with anti-PD-1 (**Figure 5D**). However, the number of CD8+ cells was not increased (see **Figure 4E**) nor was tumor growth significantly affected by anti-PD-1 treatment in the nulliparous group (**Figure 5D**). While PD-1 and Lag-3 are markers of T cells exhaustion, these markers do not evaluate the functionality of the T cells. Therefore, we next measured the production of the effector cytokines IFNγ and TNFα by tumor-associated CD8+ and CD4+ T cells in our involution group tumors. We evaluated CD8+PD-1+ T cells ex vivo in the presence or absence of stimulation with PMA/Ionomycin (**Figure 5E**). After anti-PD-1 treatment we found significant increases in the production of both IFNγ and TNFα by CD8+PD1+ T cells, from the involution group treated with anti-PD-1, suggesting that they are poly-functional (**Figure 5F**). The frequency of CD4+ T cells and the production of IFNγ by CD4+ T cells after anti-PD-1 treatment was not significantly different with treatment (**Supplemental Figures 6E,F**).

#### Co-expression of Immune Inhibitory Programs and PD-L1, PDPN, and CD68 Is Observed in Patients With Breast Cancer

Having shown that PD-L1+ LEC and monocytic cell populations likely contribute to the immune inhibitory microenvironment in the mammary gland during involution and in breast cancer, we examined whether breast cancer patient samples frequently exhibit co-expression of immune-inhibitory programs and of the LEC marker PDPN and the macrophage marker CD68. To accomplish this, we examined co-expression by RNASeq in breast cancers using The Cancer Genome Atlas cBioPortal for Cancer Genomics. As expected we observed co-expression of PD-L1 (gene name CD274) with CD8 (gene name CD8A), PD-1 (gene name PDCD1), and LAG3 (**Figures 6A–C**). We also observed significant co-expression of CD274 with PDPN and CD68 as well as between PDPN and CD68 (**Figures 6D–F**). Correlation coefficients and p-values for each relationship are reported in **Table 1**. Furthermore, consistent with our results during in mouse mammary tissue during involution we did not observe significant co-expression of CD274 with PDGFA, an established fibroblast marker, or with PDGFB the heterodimeric

FIGURE 4 | Anti-PD-1 treatment reduces tumor growth, enhances immune infiltration, and reduces lymphatic vessels in a model of post-partum breast cancer (A) C57/Bl6 mice were bred and allowed to lactate for 10 days before pups were removed to initiate mammary gland involution (parous/involution, n = 6); age-matched nulliparous controls were used (n = 6). Parous animals were injected with 250,000 E0771 tumor cells at Inv D1 (involution group) or into nulliparous animals (nulliparous group). Once tumors became measurable (involution = 4 days post injection (DPI); nulliparous=DPI D5), anti-PD-1 intervention was administered and continued every third day. E0771 mammary tumor growth curves from nulliparous and involution group C57Bl/6 mice treated with vehicle or anti-PD-1 are shown in (Continued) FIGURE 4 | (B). Results are representative from two independent studies. Dotted lines represent the slope of the tumor growth. (C) Representative images of H&E analysis and immunohistochemistry for CD45 (brown) in fixed tumor tissue to identify tumor infiltrating lymphocytes after anti-PD-1 treatment compared to vehicle controls, T = tumor and N = normal. CD45<sup>+</sup> area was 9.99% of tumor area for involution with vehicle and 22.62% of tumor area for involution with anti-PD-1 treatment. Scale bars are 50 microns. Number of (D) CD45<sup>+</sup> and (E) CD8<sup>+</sup> cells per area in tumors from B quantified by flow cytometric analysis. (F) % LECs of total in tumors from B quantified by flow cytometry. (G) Representative images of Lyve-1 stained fixed tumor adjacent tissues and quantitation of Lyve-1<sup>+</sup> vessels per area in involution group tumors +/- PD-1 treatment. Scale bars are 100 microns. \*p < 0.05; \*\*p < 0.01, \*\*\*p < 0.001, \*\*\*\*p < 0.0001.

staining with IFNγ and TNFα. (F) Quantification of double positive IFNγ and TNFα cells from upper right quadrant in G in involution group tumors with or without treatment. Data shown are from 1 representative experiment of 2 replicates with at least 4 tumors per group. Two-way ANOVA (A). Unpaired t-test (C–G): \*p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001.

partner of PDGFA. Therefore, we predict that this mechanism of immune suppression is mediated, in part, by cells of the TME may be at play in breast cancer patients and that analysis of LEC and monocyte/macrophage content could be utilized to predict whether a breast cancer patient is likely to respond to PD-1-targeted therapy.

### DISCUSSION

Our results suggest that an immune inhibitory microenvironment involving LEC and monocyte/macrophage expression of PD-L1 and T-cell expression of PD-1 is induced during involution and may be co-opted by post-partum tumors.

mRNA (by RNASeq) levels of multiple markers reveals significant positive correlations between CD274 (the gene encoding for PD-L1) and (A) CD8A (B) PDCD1 (the gene encoding for PD-1) (C) LAG3 and (D) PDPN and (E) CD68. As well as between PDPN and CD68 (F). No correlation was observed with (G) PDGFA, a marker of fibroblasts, or its heterodimeric partner PDGFB (H). Statistical analyses are provided in Table 1.

These inhibitory receptor-ligand interactions appear to promote increased tumor growth and immune evasion by increasing PD-1 and Lag-3 expression by tumor infiltrating T cells. In 2017, Dieterich et al. published that LECs in the tumor microenvironment could promote T cell exhaustion through expression of PD-L1 (28) and more recently Lane et al. identified TABLE 1 | Statistics for mRNASeq correlation analyses.


a mechanism by which T cell infiltration can manipulate PD-L1 expression via IFNg (29). These results are similar to normal mechanisms of T-cell inactivation by LECs in the lymph node where LEC expression of peripheral tissue antigens and PD-L1 act to prevent autoimmunity and promote peripheral tolerance (30, 31, 33). Here, we suggest that a similar mechanism is induced during post-partum mammary gland involution by showing that PD-L1+ LECs and monocyte/macrophage populations and PD-1+ T-cells are abundant. Additionally, we believe that both increased lymphatics and their expression of PD-L1, along with increased PD-1 and Lag-3 expression by T cells in the tumor microenvironment could lead to increased tumor metastasis in tumors established during involution. We also show that specific targeting of the PD-L1/PD-1 interaction in our model of PPBC decreases mammary tumor growth during involution. Identification of the mechanisms that underlie the increased expression of co-inhibitory ligands and receptors during involution, and in post-partum tumors, is important for proposing rational clinical trials for post-partum patients.

We predict that the immune suppressive environment, which occurs as a consequence of mammary gland involution and is mediated, in part, by lymphatics and monocyte/macrophage populations, aids in prevention of autoimmunity in a tissue healing environment (3, 10, 13, 20, 53, 54). While our studies do not identify the specific pathways activated in the mammary gland during involution that control PD-L1 expression by LECs or monocytes/macrophages, and other cells, several possibilities are evident from the existing literature. First, pro-inflammatory enzyme cyclooxygenase-2 (COX-2) is expressed and active during involution, and correlations between COX-2 expression and PD-L1 expression have been reported in lung cancer, melanoma, a mouse model of mammary cancer, and in tumorassociated macrophages and myeloid derived suppressor cells (55–59). Additionally, blocking COX-2 during involution also decreases some of the tumor-promotional effects of involution including increased growth (16). Second, STAT-3 is activated during involution and is a known activator of PD-L1 expression (60–65). Third, high expression of PD-L1 in the subcapsular sinus LECs in the lymph node (30, 45) occurs through lymphotoxin beta receptor (LTBR or TNFRSF3) signaling and is mediated by the presence of B cells (30). Interestingly, LTBR signaling is also a known regulator of epithelial cell apoptosis during involution (19) and B cells are increased in the mammary gland during involution (13) suggesting an additional mechanism that may drive LECs to upregulate PD-L1 expression. We propose that a number of mechanisms could be at play to result in the upregulation of PD-L1 on lymphatic endothelial cells, as well as other cells, in the mammary gland during mammary gland involution and in post-partum tumors. As we have previously published that macrophages contribute to lymphatic remodeling, and lymphatic mimicry, during involution it is notable that the monocyte/macrophage population expressing PD-L1 is also higher during involution and in tumors implanted during involution. We suggest that this increase in myeloid cells could either be to remodel the lymphatic vasculature or to drive expression of PD-L1 through as of yet undefined mechanisms. It is also possible that the increased PD-L1 myeloid population is contributing to the expression of PD-L1 by incorporating into the vasculature and expressing lymphatic markers. Dissecting these molecular mechanisms that drive PD-L1 expression is under active investigation.

In addition to increased PD-L1 expression, we show that tumor infiltrating T cells express more co-inhibitory receptors in mammary tumors implanted during post-partum involution. We also found that treatment with anti-PD-1 in these tumors resulted in decreased frequencies of PD-1 and Lag-3 doublepositive CD8+ T cells and decreased Lag-3 expression. The frequency and expression of Lag-3+ cells in anti-PD-1-treated involution group tumors was more similar to nulliparous group mice, suggesting that anti-PD-1 treatment could reverse the immune suppression observed in involution group tumors by decreasing or inhibiting the expansion of Lag-3 and PD-1 double positive T cells. An alternative explanation is that the anti-PD-1 depletes PD-1+ T cells and this possibility is being explored through complementary anti-PD-L1 studies by our group. Further, while IFNγ expression is relatively low in bulk tissue during involution, we do see IFNγ expression by T cells in mammary tumors implanted into both nulliparous and involution group hosts demonstrating that IFNγ may at least partially contribute to PD-L1 expression by the cells in the tumor (29). These findings indicate that while the production of IFNγ within the tumor is not increased, that instead the frequency and poly-functionality (IFNγ and TNFα double positive cells) of CD8+ T cells is increased following treatment with anti-PD-1. These findings predict increased tumor killing consistent with what is seen in patients (41) and is consistent with the decreased tumor volume we observed. Our findings that LVD is also decreased after anti-PD-1 treatment could either be a cause or consequence of tumor regression, but is likely due to the treatment induced pro-inflammatory environment in contrast to the anti-inflammatory environment observed in post-partum tumors. Loss of LVD could significantly impact the ability of tumor cells to migrate to the tumor draining lymph node, which could also reduce metastasis. A limitation of our current studies is the lack of metastatic data in our preclinical models as well as the lack of analysis of PD-L1 expression on post-partum tumors; these important studies are being actively pursued using additional models.

Our findings that an involution-targeted therapy, anti-PD-1, can mitigate the decrease in tumor growth and LVD that we observe after implantation during involution are consistent with our previous results showing that inhibition of COX-2 can decrease growth, invasion, LVD, and metastasis (3, 16). Whether anti-PD-1 therapy could also mitigate the increased metastasis observed in our models is unanswered by our current data. However, if such mechanisms are maintained long term, as is suggested by our 66cl4 model where tumors are isolated long after completion of involution, post-partum patients with metastasis may also benefit from anti-PD-1 therapy, which has shown efficacy in the metastatic setting (see below). Our findings may also lend insight into the increased aggressiveness and metastatic potential of breast cancer diagnosed within 5– 10 years of recent childbirth. We propose that a plausible mechanism is immune suppression in both the mammary tissue and the draining lymph node, which could account for the increased lymph node positivity observed in post-partum patients (2). This, along with the increased lymphangiogenesis and lymphatic vessel invasion observed in the adjacent mammary tissue of post-partum patients, could provide a favorable route for metastatic spread. While further studies addressing mechanisms of lymphatic expression of PD-L1 and lymphatic growth in tumors must be performed, we believe these studies provide substantial evidence that current immunotherapies could benefit patients with PPBC. Immunotherapy, specifically anti-PD-1/PD-L1 based, has been investigated in breast cancer with positive results. First, pembrolizumab—the highly selective monoclonalantibody-based therapy against PD-1—was the first shown to be successful as a monotherapy for metastatic triple-negative cases of breast cancer (TNBC), with some long-lasting responses reported, and has also shown benefit for advanced ER+/Her2 in the KEYNOTE trials (66–70). It is also currently in clinical trial with several chemotherapy partners, including in the neoadjuvant setting and as a single drug in the post-neoadjuvant setting for patients with residual cancer (71, 72). Anti-PD-L1 based therapy, specifically atezolizumab, is now a standard of care option in combination with the chemotherapy drug nabpaclitaxel for TNBC that have at least 1% PD-L1 expressing tumor infiltrating immune cells, based on the results of the Impassion 130 clinical trial, which demonstrated a 10-month improvement in overall survival for the combination (73). While the early immunotherapy trials with check-point block inhibitors have met with success, many patients do not benefit or do not achieve long-term benefit and death due to progressive cancer remains the norm. This highlights that additional markers, such as evaluation of LVD, may be better or synergistic with existing markers for predicting patient response (74). Importantly, several trials are ongoing to investigate the use of anti-PD-1 and anti-PD-L1 in the (neo) adjuvant setting and the treatment has proven safe and well-tolerated with preliminary results demonstrating promising efficacy (71, 72, 74). Here we have identified a specific population of patients who may benefit from (neo) adjuvant PD-1/PD-L1 blockade—women diagnosed within 10 years postpartum who are at high risk for metastasis. Ongoing research is investigating this possibility as well as whether a combined approach with an anti-lymphangiogenesis-based therapy could improve survival for PPBC patients.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule) of the Colorado Multiple Institution Review Board (COMIRB) committee with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the COMIRB. This study was carried out in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals, Animal Welfare Act and PHS Policy by the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee. The protocol was approved the institutional animal care and use committee.

### AUTHOR CONTRIBUTIONS

BT and TL designed and executed experiments, analyzed results, and drafted the manuscript. AE, JF, and AW performed experiments, analyzed data, and critically reviewed the

### REFERENCES


manuscript. VW performed imaging experiments. VB provided samples and critically reviewed the manuscript.

### FUNDING

All funds utilized to support this project were generated locally. Specifically, from the Cancer League of Colorado AWD #173586- TL and the University of Colorado Anschutz Medical Campus Department of Medicine Outstanding Early Career Scholars Program to TL and BT. The Linnea J. Basey endowment to TL and Carpenter Family Gift Fund to TL and VB.

### ACKNOWLEDGMENTS

We acknowledge Kim Jordan and the HIMSR for the Vectra imaging, and the Tissue Bio banking and Biorepository and the Shared Resource of Colorado's NIH/NCI Cancer Center Support Grant P20CA046934. We thank M. Burchill and A. Goldberg for critical review of the manuscript and D. Matthews and R. Torres for advice on dose and interval of PD-1 treatment. We gratefully acknowledge the patients from the University of Colorado Young Women's cohort for their contribution to this research.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.01313/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.

Copyright © 2019 Tamburini, Elder, Finlon, Winter, Wessells, Borges and Lyons. 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(s) 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.