# CYTOKINE-MEDIATED ORGAN DYSFUNCTION AND TISSUE DAMAGE INDUCED BY VIRUSES

EDITED BY : Michael H. Lehmann, Piotr Religa and Juliet Spencer 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 978-2-88963-519-1 DOI 10.3389/978-2-88963-519-1

#### 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

# CYTOKINE-MEDIATED ORGAN DYSFUNCTION AND TISSUE DAMAGE INDUCED BY VIRUSES

Topic Editors:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany Piotr Religa, Karolinska Institutet, Sweden Juliet Spencer, Texas Woman's University, United States

Citation: Lehmann, M. H., Religa, P., Spencer, J., eds. (2020). Cytokine-Mediated Organ Dysfunction and Tissue Damage Induced by Viruses. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-519-1

# Table of Contents

*05 Editorial: Cytokine-Mediated Organ Dysfunction and Tissue Damage Induced by Viruses*

Juliet V. Spencer, Piotr Religa and Michael H. Lehmann

*08 Contribution of Cytokines to Tissue Damage During Human Respiratory Syncytial Virus Infection*

Karen Bohmwald, Nicolás M. S. Gálvez, Gisela Canedo-Marroquín, Magdalena S. Pizarro-Ortega, Catalina Andrade-Parra, Felipe Gómez-Santander and Alexis M. Kalergis


Tim Krischuns, Franziska Günl, Lea Henschel, Marco Binder, Joschka Willemsen, Sebastian Schloer, Ursula Rescher, Vanessa Gerlt, Gert Zimmer, Carolin Nordhoff, Stephan Ludwig and Linda Brunotte

*53 Hantavirus-Driven PD-L1/PD-L2 Upregulation: An Imperfect Viral Immune Evasion Mechanism*

Martin J. Raftery, Mohammed O. Abdelaziz, Jörg Hofmann and Günther Schönrich

*66 Cytokine-Mediated Induction and Regulation of Tissue Damage During Cytomegalovirus Infection*

Mathew Clement and Ian R. Humphreys

*75 Nef-induced CCL2 Expression Contributes to HIV/SIV Brain Invasion and Neuronal Dysfunction*

Michael H. Lehmann, Jonas M. Lehmann and Volker Erfle

*85 Fine Tuning the Cytokine Storm by IFN and IL-10 Following Neurotropic Coronavirus Encephalomyelitis*

Carine Savarin and Cornelia C. Bergmann

*93 TRIM21 Restricts Coxsackievirus B3 Replication, Cardiac and Pancreatic Injury via Interacting With MAVS and Positively Regulating IRF3-Mediated Type-I Interferon Production*

Hui Liu, Min Li, Yahui Song and Wei Xu

*106 Silencing the CSF-1 Axis Using Nanoparticle Encapsulated siRNA Mitigates Viral and Autoimmune Myocarditis*

Ingmar Sören Meyer, Carl Christoph Goetzke, Meike Kespohl, Martina Sauter, Arnd Heuser, Volker Eckstein, Hans-Peter Vornlocher, Daniel G. Anderson, Jan Haas, Benjamin Meder, Hugo Albert Katus, Karin Klingel, Antje Beling and Florian Leuschner

*124 Proteasomal Protein Degradation: Adaptation of Cellular Proteolysis With Impact on Virus—and Cytokine-Mediated Damage of Heart Tissue During Myocarditis*

Antje Beling and Meike Kespohl


Cordelia Manickam, Spandan V. Shah, Olivier Lucar, Daniel R. Ram and R. Keith Reeves

# Editorial: Cytokine-Mediated Organ Dysfunction and Tissue Damage Induced by Viruses

#### Juliet V. Spencer <sup>1</sup> , Piotr Religa<sup>2</sup> and Michael H. Lehmann<sup>3</sup> \*

<sup>1</sup> Department of Biology, Texas Woman's University, Denton, TX, United States, <sup>2</sup> Department of Medicine, Karolinska Institutet, Stockholm, Sweden, <sup>3</sup> Institute for Infectious Diseases and Zoonoses, Ludwig-Maximilians-Universität München, Munich, Germany

Keywords: coxsackievirus, hantavirus, HCMV, HIV, influenza, MHV, RSV, TMEV

#### **Editorial on the Research Topic**

#### **Cytokine-Mediated Organ Dysfunction and Tissue Damage Induced by Viruses**

Cytokines are small proteins, mostly secreted into the extracellular environment, that bind to specific cell surface receptors, which mediate cell differentiation, migration, growth, and death. Gene expression and cellular release of cytokines are strictly regulated to assure proper function of cells, tissues, and organs. Upon virus infection, a cell starts producing type I interferons (IFN) and inflammatory cytokines (ICs) to restrict spread and replication of the respective virus. Ideally, the virus is completely eliminated by the immune system and the antiviral mechanisms are turned off within a reasonable time frame. However, there are different scenarios where this process does not work efficiently or does not happen at all, leading to cytokine-mediated organ dysfunction and tissue damage.

#### First, if a virus inhibits type I IFN production and signaling but does not prevent expression of ICs, then this virus spreads further, and more viral components are in the system, which continuously amplifies ICs production. The lung is one organ that is especially vulnerable to such a "cytokine storm," triggered, for example, by infection with respiratory syncytial virus (RSV) or influenza virus. In a comprehensive review article, Bohmwald et al. describe in detail where and which cytokines are induced during human RSV infection and their potential contribution to damage of not only the lung but also the brain. The pathophysiological production of ICs is most probably also due to the ability of RSV to induce the expression of Toll-like receptor (TLR) 4 in human airway epithelial cells, which normally do not respond to endotoxin (1).

Aflatoxin B<sup>1</sup> (AFB1), a mycotoxin produced by Aspergillus flavus, increases TLR4 expression as well as TLR2, IL-1β, and IL-6 expression in human monocyte-derived dendritic cells (2). Sun et al. confirmed upregulation of TLR4 expression by AFB<sup>1</sup> in porcine alveolar macrophages (PAMs) and mice infected with swine influenza virus (SIV) subtype H1N1. Importantly, they show that AFB<sup>1</sup> exacerbates lung damage in mice during SIV infection, caused by a TLR4-dependent increase in viral replication and TNF levels. Consequently, uptake of AFB1, for example by contaminated food (3), could aggravate the course of flu.

Experimental evidence suggests that extrarespiratory induction of ICs such as TNF, IL-6, and IL-8 contributed to deadly infection with the 1918 H1N1 influenza A virus (IAV) strain (4), which hit mankind during World War I, a period when not sufficient food was available (5). Similarly, systemic high levels of IL-6 and IL-8 were detected in humans infected with IAV subtype H5N1, especially in those with fatal outcome, but not in humans infected with IAV subtype H3N2 or H1N1 (6). Hence, it appears that an IAV which just passed the animal-human barrier is much more harmful than a human adapted strain. In this respect, Krischuns et al. discovered that infection

#### Edited and reviewed by:

Vitaly V. Ganusov, The University of Tennessee, Knoxville, United States

#### \*Correspondence:

Michael H. Lehmann Michael.Lehmann@lmu.de; Orlataler@web.de

#### Specialty section:

This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology

Received: 26 November 2019 Accepted: 02 January 2020 Published: 22 January 2020

#### Citation:

Spencer JV, Religa P and Lehmann MH (2020) Editorial: Cytokine-Mediated Organ Dysfunction and Tissue Damage Induced by Viruses. Front. Immunol. 11:2. doi: 10.3389/fimmu.2020.00002

**5**

of human alveolar epithelial cells with highly pathogenic avian influenza virus (HPAIV) strains, but not with human adapted IAV strains, leads to constitutive phosphorylation of tripartite motif (TRIM) 28 at S473 and increased production of IFN-β, IL-6, and IL-8. TRIM28 negatively regulates transcription in a number of ways (7), and the inability of the non-human adapted influenza strains to prevent its deactivation could well explain the dysregulated cytokine expression observed in individuals infected with the IAV subtype H5N1 or the 1918 pandemic H1N1 IAV. In conclusion, severe organ damage can occur if a virus passes from its natural host to another species where it can or gains the ability to replicate without having an effective mechanism to block immunity of the other species. such a scenario is frequently observed in humans when they are infected with hantavirus (8). Possibly, activation of bystander CD8+ T cells after hantavirus infection, as shown in an original article by Raftery et al., could be a reason for the organ damages in humans. Indeed, viral-activated bystander memory CD8+ T cells can cause organ damage in a T cell receptor independent manner (9). In contrast, no apparent disease is observed in rodents, the natural reservoir hosts of all hantavirus species, which is similar to humans where a lifelong, persistent infection with human cytomegalovirus (HCMV) remains normally inconspicuous. However, there are several situations in which HCMV induces damage to organs and tissues, and Clement and Humphreys review the involvement of cytokines therein.

The cellular tropism of human and simian immunodeficiency virus is limited, but dysfunctions and damages also occur in the hearts, lungs, kidneys, and brains of infected individuals. Lehmann et al. provide a model to understand the complexity of these pathologies based on recent cell biology findings about systemic distribution of the viral Nef protein and improved understanding of C-C motif chemokine ligand 2 (CCL2) dependent transendothelial cell migration. The focus of this Mini Review is on the brain, where non-physiological induction of CCL2 expression not only drives encephalitis but also affects signaling and survival of neuronal cells. Indeed, antiviral and anti-inflammatory cytokine expression have to be well-coordinated during a viral infection in order to enable elimination of the pathogen on the one hand, and to avoid tissue and organ damage on the other hand. This problem is discussed by Savarin and Bergmann on the basis of a murine model of encephalomyelitis induced by a neurotropic strain of mouse hepatitis virus where the interplay of IFN and IL-10 dictates the extent of viral control and tissue pathology.

#### REFERENCES


Type I IFN is required to restrict coxsackievirus B3 (CVB3) replication (10), and Liu et al. found that functional TRIM21 is important for high level type I IFN expression in vitro and in vivo. Additionally, they show that TRIM21 deficiency leads to higher viral titers, stronger cardiac and pancreatic damages, and higher levels of ICs including CCL2 in mice infected with CVB3. Meyer et al. detected increased levels of colony stimulating factor 1 (CSF-1) in heart biopsies of patients with myocarditis, and by using nanoparticle-encapsulated siRNA directed against CSF-1 they could decrease CVB3-induced monocyte infiltration and heart damage in mice. Beling and Kespohl suggest that therapeutic targeting the proteasome could help to prevent immunopathology of the heart, which can be triggered by many different viruses. Theiler's murine encephalitis virus (TMEV) induces myocarditis in mice, but this virus can also induce demyelination of neurons depending on the mouse strain. A Hypothesis and Theory article by Omura et al. summarizes findings about the different nature of these diseases and provides evidence that TMEV induces cell-type specific innate immune responses and distinct organ-specific pathology. Thus, the choice of the right animal model to study virus induced immunopathology can be challenging. For that, Manickam et al. provide a thorough review of non-human primate models for understanding the extent of cytokine-mediated tissue damage during many different types of virus infection, including dengue virus, HCMV, hepatitis B and C virus, HIV, influenza virus, and Zika virus.

Collectively, this Research Topic introduces some of the complex virus-host interactions that can tip the scales toward immunopathology. The common themes that emerge from this collection include the potential for use of cytokines as markers of disease and the manipulation of certain cellular molecules as therapeutic options.

#### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

We extend our gratitude to the authors for their excellent contributions to the Research Topic, and we thank the reviewers for their time, effort, and helpful feedback.


**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 © 2020 Spencer, Religa and Lehmann. 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.

# Contribution of Cytokines to Tissue Damage During Human Respiratory Syncytial Virus Infection

Karen Bohmwald<sup>1</sup> , Nicolás M. S. Gálvez <sup>1</sup> , Gisela Canedo-Marroquín<sup>1</sup> , Magdalena S. Pizarro-Ortega<sup>1</sup> , Catalina Andrade-Parra<sup>1</sup> , Felipe Gómez-Santander <sup>1</sup> and Alexis M. Kalergis 1,2 \*

<sup>1</sup> Millenium Institute on Immunology and Immunotherapy, Departamento de Genética Molecular y Microbiología, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Santiago, Chile, <sup>2</sup> Departamento de Endocrinología, Facultad de Medicina, Pontificia Universidad Católica de Chile, Santiago, Chile

#### Edited by:

Michael H. Lehmann, Ludwig Maximilian University of Munich, Germany

#### Reviewed by:

Peter J. M. Openshaw, Imperial College London, United Kingdom Yashoda M. Hosakote, The University of Texas Medical Branch at Galveston, United States

> \*Correspondence: Alexis M. Kalergis akalergis@bio.puc.cl; akalergis@icloud.com

#### Specialty section:

This article was submitted to Cytokines and Soluble Mediators in Immunity, a section of the journal Frontiers in Immunology

> Received: 21 September 2018 Accepted: 19 February 2019 Published: 18 March 2019

#### Citation:

Bohmwald K, Gálvez NMS, Canedo-Marroquín G, Pizarro-Ortega MS, Andrade-Parra C, Gómez-Santander F and Kalergis AM (2019) Contribution of Cytokines to Tissue Damage During Human Respiratory Syncytial Virus Infection. Front. Immunol. 10:452. doi: 10.3389/fimmu.2019.00452 The human respiratory syncytial virus (hRSV) remains one of the leading pathogens causing acute respiratory tract infections (ARTIs) in children younger than 2 years old, worldwide. Hospitalizations during the winter season due to hRSV-induced bronchiolitis and pneumonia increase every year. Despite this, there are no available vaccines to mitigate the health and economic burden caused by hRSV infection. The pathology caused by hRSV induces significant damage to the pulmonary epithelium, due to an excessive inflammatory response at the airways. Cytokines are considered essential players for the establishment and modulation of the immune and inflammatory responses, which can either be beneficial or harmful for the host. The deleterious effect observed upon hRSV infection is mainly due to tissue damage caused by immune cells recruited to the site of infection. This cellular recruitment takes place due to an altered profile of cytokines secreted by epithelial cells. As a result of inflammatory cell recruitment, the amounts of cytokines, such as IL-1, IL-6, IL-10, and CCL5 are further increased, while IL-10 and IFN-γ are decreased. However, additional studies are required to elicit the mediators directly associated with hRSV damage entirely. In addition to the detrimental induction of inflammatory mediators in the respiratory tract caused by hRSV, reports indicating alterations in the central nervous system (CNS) have been published. Indeed, elevated levels of IL-6, IL-8 (CXCL8), CCL2, and CCL4 have been reported in cerebrospinal fluid from patients with severe bronchiolitis and hRSV-associated encephalopathy. In this review article, we provide an in-depth analysis of the role of cytokines secreted upon hRSV infection and their potentially harmful contribution to tissue damage of the respiratory tract and the CNS.

Keywords: human respiratory syncytial virus, cytokines, chemokines, tissue damage, inflammation

#### INTRODUCTION

#### Prevalence of hRSV Infection Worldwide

The human respiratory syncytial virus (hRSV) is one of the primary viral agents causing hospitalizations due to acute lower respiratory tract infection (ALRTI) in young children, immunocompromised and elderly individuals worldwide (1, 2). The epidemic period for hRSV infections usually takes place during the winter season in areas with temperate climates (3).

**8**

This pathogen causes pulmonary manifestations mainly in the upper and lower respiratory tract, promoting the development of bronchiolitis and pneumonia (**Figure 1**) (4, 5). Some of the risk factors associated with the development of hRSV-associated ALRTI are premature birth, low birth weight, maternal smoking, history of atopy and no history of breastfeeding in infancy, among others (6). A recent report estimated that -during 2015 hRSV-associated ALRTI episodes reached a global burden of 33.1 million, resulting in 3.2 millions hospital admissions and around 60,000 in-hospital deaths in children under the age of 5 (1), although, no global studies of other populations such as the elderly or patients with underlying medical conditions have been conducted (7). Reinfections during childhood and adulthood are very common, and the severity of hRSV infections in healthy adults is mild. This decrease in severity has been related to higher neutralizing antibody titers induced by constant challenges with the virus throughout life (8). Besides children, the elderly have been described as another high-risk population, probably because of their senescent immune system (9). In this population, hRSV is the leading viral pathogen, which causes morbidity and mortality, followed by influenza A (10).

Besides the airway pathologies caused by hRSV, neurologic complications have also been described after infection with this virus, although less frequently (11–13). The etiology of the neurological alterations remains unknown. However, it has been proposed that inflammatory mediators, such as cytokines could be playing an essential role in the development of neurologic alterations (14, 15).

The hRSV is a highly contagious virus, as it can live outside of the host for about 6 h on hard surfaces, and as much as 20 min on the skin (16). Also, people that are infected with this virus remain contagious up to 8 days starting from the day of infection (17). Studies have shown that at least a third of the children experiencing hRSV infection within their first year of life will get re-infected during their second or third year of life (18). Patients infected with this virus cannot promote an adequate immunological response and, therefore, can get infected again with the same virus in the same cohort (19). In this regard, it has been described that this virus can impair the assembly of a proper immunological synapse between the antigen-presenting cells (APC), such as the dendritic cells, and T cells (**Figure 1B**) (20). In this way, hRSV renders T cells unable to respond correctly, which may lead to a poor adaptive immune response against the virus and, consequently, the reinfections mentioned above (**Figure 1**).

Most studies, aimed to determine the economic burden associated with hRSV, measure its immediate impact on healthcare resources, such as hospitalizations, ambulatory care, and emergency department visits, focusing primarily on infant populations (21–23). It is noteworthy that hRSV has been associated with long-term illness such as asthma and recurrent wheezing (24, 25), which could represent a substantial increase in the economic burden related to this pathogen (26, 27).

Currently, there are no licensed vaccines available for preventing hRSV infection although several groups are working in the development of potentially effective vaccines and therapies. Nowadays, the only drug available on the market designed to ameliorate this disease is palivizumab, a humanized monoclonal antibody against the fusion protein (F-protein) of the virus. This product is used as a prophylactic option, along with ribavirin as a therapeutic option, although this strategy is only used in high-risk patients, such as children born after ≤29 weeks of gestation and preterm infants with chronic pulmonary disease (28, 29). Because this treatment fails to target most of the population susceptible to hRSV-caused disease, (i.e., healthy infants, children, and the elderly), the development of an effective vaccine is imperative (21, 22, 30). Several studies have concluded that the cost-effectiveness of palivizumab might not be enough to recommend the massive use of this antibody (22, 31–33). However, other studies have concluded that it does reduce the severity of infection and long-term effects on children, suggesting that it can diminish the spending of health-care resources (34, 35).

### hRSV: General Characteristics and Infective Cycle

The hRSV has been recently defined as a member of the Orthopneumovirus genus from the Pneumoviridae family being also recently renamed as human Orthopneumovirus and is an enveloped, negative-sense and single-stranded RNA virus with a genome of about 15.2 kb, possessing 10 genes that encode for 11 proteins (36–38). The viral particle displays 3 surface proteins, the F-protein, the glycoprotein (G) and the small hydrophobic (SH) protein (**Figure 1A**). Of all these, the Gprotein is responsible for the attachment with the membrane of the host cell (39), mainly by binding to the CX3CR1 receptor on ciliated epithelial cells (40, 41). The fpre F-protein is responsible for the fusion of the viral membrane with the host cell membrane and further entry of the viral genetic material into the cytosol, apparently by its interaction with the surface protein nucleolin, although other receptors have been described to play a role in this process (42).

This virus can be transmitted by aerosol particles person-toperson, or via direct contact of these aerosol particles with the exposed mucosa, such as conjunctival (43). After infection, the incubation period can vary between 2 and 8 days in healthy individuals (44). At the beginning of hRSV infection, the virus meets the first line of defense of the organism, consisting of epithelial cells from the nasal and upper respiratory tract (45, 46).

The airway epithelium presents the apical junctional complex (AJC), which seals the space between the layer of epithelial cells and acts as a barrier that prevents the entry of pathogens into the organism (47). It has been described that hRSV infection induces a dysfunction in the epithelial barrier in a protein kinase D (PKD)-dependent manner (48). After infection by hRSV, cells exhibit a disruption of the AJC, which can be prevented when PKD-inhibitors are added, as described previously (48). As mentioned above, once hRSV reaches the apical side of the ciliated epithelial cells, the G and F proteins allow the attachment and fusion of the virus to the host membrane, respectively (39, 42).

After the virus has fused with the membrane of the host cells, it then begins the mechanism of entering the cells. The entry of

infected epithelial cells, will induce the collapse of the lower respiratory tract. Finally, it has been described that hRSV infection can cause CNS pathologies, although the mechanism underlying this have not been described.

the virus is through an endocytosis-dependent mechanism and allows the entering of the whole virus, including its lipid envelope (49). Then, the virus is carried within endocytic vacuoles, and undergoes a second fusion, this time with the vacuole itself, that occurs when the F-protein is cleaved by a furin-like convertase, to render the virus able to infect the cell (49). Then the virus reaches the cytoplasmic inclusion bodies (IBs) of the cells, where it can replicate its RNA using the viral RNA-dependent RNA polymerase (RdRp) complex, which is composed of the large protein (L) and the phosphoprotein (P) of the virus (50). As transcription goes on, the viral protein M2-1 is added to the complex allowing the synthesis of the mRNA (50, 51). The virus starts the replication of its RNA in the nasal epithelial cells and then it moves toward the bronchioles, where the replication becomes more effective (44). The virus spreads via intercellular extensions between two cells or through the cell to cell transmission, and in both cases, the infected cell is the one who passes the virus to the target cell (52).

To study the pathology associated with hRSV and the immune response during the infection, the use of several animal models has shown to be extremely important (53, 54). Lately, mice have been the animal model of choice for most immunology studies on this virus (53), although it is important to emphasize that the immune response observed in mice is not necessarily identical to the one observed in human patients. Some of these differences in the immune response between mice and humans are remarkable, for instance, the fact that older mice are more susceptible to hRSV-infection as compared to younger mice (55). Some techniques and methods to determine hRSV disease severity used in murine models are also different from those used to evaluate these parameters in humans. For instance, recording the body weight changes as a parameter of disease severity (more weight loss implies a more severe disease) is frequently used in the murine model, but it is not used as a parameter in humans disease (53). Also, obtaining bronchoalveolar lavage fluid (BALF) samples from mice is a standard procedure to evaluate inflammatory parameters, and these results can vary significantly from those observed in humans (53, 56). Although some differences can be observed, the data relative to cytokines and chemokines in the lower respiratory tract of mice and humans varies little, and to our knowledge, no published studies are describing these molecules in the upper respiratory tract and central nervous system (CNS) of mice (**Table 1**).

Further, in this review, we will provide an in-depth analysis of the current information available regarding the inflammatory mediators that are induced upon hRSV infection, and which ones are produced, up-regulated and down-regulated in the different sections of the respiratory tract and the CNS (**Table 1**).



Additionally, we will discuss the contribution of cytokines to the immune response and immunopathology observed after hRSV infection.

#### CYTOKINES ELICITED BY hRSV INFECTION

Among the inflammatory mediators that have been described to play an essential role in the hRSV pathology are cytokines and chemokines. Cytokines are small secreted molecules that contribute significantly to the modulation of the immune response and T cells differentiation (79). Several cell types can produce and secrete cytokines including immune cells, epithelial cells, and endothelial cells, amongst others (80, 81). Depending on the effect that they generate over immune cells, they can be classified into two groups; pro-inflammatory and anti-inflammatory (79). Interleukin (IL)-1, tumor necrosis factor alpha (TNF-α), interferon-gamma (IFN-γ), and IL-6, among others (79, 82, 83) belong to the pro-inflammatory group, IL-10 is anti-inflammatory, and IL-12 can be pro- and anti-inflammatory cytokine (**Figure 1C**) (84, 85).

Among cytokines, chemokines are a group of proteins with chemoattractant properties and are characterized by three to four cysteine residues present in their structure (80, 84). These proteins can be classified -according to the position of the cysteines residues in their N-terminal portion- into four families. The first family is the C-C chemokines present the cysteine residues continuously. The second family is the C-X-C chemokines present one amino acid between the two cysteine residues. The third family is the X-C chemokine only present one cysteine residue in a conserved position (this family is composed of only one member; XCL1). Finally, the four family is the C-X-3-C chemokine, which presents two cysteine residues separated by three interchangeable amino acids (This family possesses only one member; CX3CL1) (84, 86). Another relevant characteristic of chemokines is that they are considered to be promiscuous proteins, as they can interact with more than one chemokine receptor and one receptor can bind more than one chemokine (86). Additionally to their chemoattractant function, chemokines play an essential role in maintaining the homeostasis during the development of the brain, heart, and hematopoietic system, among others (86). Besides, they are also critical players in the modulation of the immune response during infections, as they are responsible for the infiltration of immune cells into the site of injury (84, 86).

#### CYTOKINES INDUCED BY hRSV INFECTION IN THE UPPER RESPIRATORY TRACT

As mentioned above, the hRSV infection starts with the virus reaching the mucous membranes of the eyes, nose or mouth, allowing it to enter the organism (87). The first zone of infection is the upper respiratory tract, where it targets the ciliated epithelium of the nasopharynx, and then it moves toward the lungs, blocking the airways as the infection proceeds (**Figure 1C**) (88). This inflammation -known as bronchiolitis or pneumonia, accordingly to the degree of the disease- involves infiltration of polymorphonuclear cells (PMNs) such as neutrophils and eosinophils. Moreover, the rounding and shedding of the infected epithelial cells apparently caused by the NS2 protein, as described by Liesman et al. (74) inducing the collapse of the alveolar spaces and, therefore, impaired oxygen exchange (89). Remarkably, humans are born with at least a third of the alveoli that they will possess once the lungs are fully developed, with alveolar walls similar to the ones seen in an adult (90, 91). However, during childhood, these structures exhibit a lower area/volume ratio when compared with a fully developed lung, a rate that is increased until adolescence. Therefore, the useful space for gas exchange is reduced in early stages of human development. This phenomenon could explain for the exacerbated pathology observed in children, compared to teenagers and adults, with even further complications the younger they are (90, 91).

Several reports have described the changes in the ciliated epithelium upon infection with this virus. For instance, Wong et al. described that, upon infection, total loss of cilia is reported, mainly associated with microtubule damage (92). These could be in direct relation with reports indicating that this virus replicates in the apical cell surface of these cells (93). Remarkably, Smith et al. described that infection with hRSV could induce ciliary dyskinesia and ciliary loss of epithelial cells early during the infection, impairing in this way the clearance of the respiratory tract (94). Interestingly, Jumat et al. recently described the morphogenesis of hRSV in epithelial cells, using a primary culture of nasal epithelial cells as a model. They detected the presence of the F-protein of hRSV predominantly in cilia, but not the N-protein, observing this and several other proteins in the non-cilia locations of the cells, indicating that, probably the hRSV-F protein may be responsible for the damage to the cilia (46). Therefore, upon infection, hRSV seems to replicate and exit from non-cilia locations in the apical side of epithelial cells, somehow causing loss of ciliary function.

In response to all this damage, the airway epithelium generates cytokines and chemokines to recruit effector cells to the site of infection and restrict its propagation (95), causing an exacerbated immune response where infiltrating immune cells such as PMNs, T cells and inflammatory mediators cause damage to the tissues (63, 84). This exaggerated inflammatory response is increased as the infection progresses, with hRSV inducing a Th2-like immune response, promoting the inflammation (64). Notably, it has been described that primary infection with hRSV induces the transcription of nuclear factor kappa B (NF-κB) mainly through its M2-1 protein (96). This factor, in turn, produces the secretion of IL-8/CXCL8, TNF-α, CCL5, and CXCL10, among others. Accordingly, transcription factor AP-1 is also required for the expression of IL-8, as described by Dey et al. (97). Both NF-κB and AP-1 are regulated in their expression by the TGFß activation kinase 1 (TAK1), as deletion or inactivation of this kinase reduce gene expression of the transcription factors and decrease their nuclear translocation and DNA-binding activity (97), suggesting that the virus could be modulating these pathways. Finally, it has also been reported that STAT1 regulates the secretion of IL-4 by basophils upon infection with hRSV. In this line, Moore et al. described that KO mice for STAT1 showed higher levels of IL-4 in lungs, upon infection; a phenomenon that was reverted when mice were depleted from basophils. Remarkably, this increase in the expression of IL-4 correlated with more marked lung histopathology (98).

In light of all this, Das et al. reported that human nasal epithelial cells infected with hRSV exhibits increased levels of IL-6, CXCL8, and CCL5, as compared to non-infected cells (65). Remarkably, IL-2 levels in nasopharyngeal aspirates do not seem to correlate with hRSV infection, as Giugno et al. described that the concentration of this cytokine was heterogeneous among infected and non-infected children (99). The secretion of these pro-inflammatory cytokines may be adding to the exacerbated inflammation described in this disease (**Figures 1C**, **2**).

Since hRSV-infected children are only brought onto health centers once the disease has reached an advanced development stage, it is hard to determine the temporality of the secretion of cytokines and chemokines in humans, during this disease. In this line, Blanco et al. performed a study in cotton rats where they measured the transcription levels of several of these molecules during primary and secondary infection (100). Therein the authors show an increase in the transcription levels of all the cytokines measured except for IL-10 during the first day post-infection. A peak for IL-6, IFN-α, and TNF-α, was detected during day 1 post-primary infection, decreasing the first two by day 3, while the latter remained high up until day 10. Likewise, IL-1β, CCL5, CXCL1, and CXCL10 transcription levels peaked at day 2, remaining high up until days 5 or 6. The cytokines that peaked during day 3 were IL-10 and CCL4, recovering normal levels by day 7. The last molecule to reach its peak was IFN-γ, at day 4 post-primary infection which correlates with previous studies indicating that this virus inhibits the expression of this cytokine. Then, at day 14, the levels of IFN-γ were returned to normal levels (100). Remarkably, infectious virus was not detected in the lungs of cotton rats challenged in a secondary infection; however, changes in the lung structure were detected even earlier than in primarily infected cotton rats (100). Despite all this, and as indicated above, these data are all related to transcript expression level, and it is not a direct measure of proteins. Therefore, this information must be taken into account cautiously.

Another work performed by Legg et al. examined the cytokine response to the hRSV through nasal lavage fluid in infants (101). In this study when some respiratory symptoms the research team visited the infant to whom they performed a clinical examination and nasal lavage, considering this collection of samples days 1 and 2. The same procedure was performed at day 5 and 6 since the development of the symptoms. They found that IL-4/IFN-γ ratio was elevated at day 1–2 and 5–6, and during the first two days, the IL-10/IL-12 ratio reached its peak (101). The results obtained with IFN-γ in infants correlates with the results obtained in mice since during the first couple of days the secretion of this cytokine has a similar pattern, suggesting that the other cytokines should behave similarly in humans as it does in cotton rats.

Toll-like receptors (TLRs) are pathogen recognition receptors (PRRs) that are activated upon the recognition of pathogenassociated molecular patterns (PAMPs) (102). They are expressed in several cell types such as immune cells and epithelial cells. Moreover, they are significant players in the early response against pathogens, as they can regulate the secretion of several cytokines and chemokines (103, 104). In this line, the role of TLRs in the innate immune response against hRSV is significant, as TLR3, TLR4, TLR7, and TLR8 are upregulated upon infection (**Figure 1B**) (105–107). TLR4 has been described to interact with the F-hRSV protein, leading to the activation of NF-κB and the secretion of the cytokines mentioned above, such as CXCL8 and TNF-α (108, 109). In humans, mutations in TLR4 impair the activation of this pathway and, in mice this renders the organism unable to clear the virus, and the persistence of the virus has been described in TLR4 deficient mice (57, 58). TLR3, which recognizes viral double-stranded RNA, induces the secretion of type I IFN and the activation of the NF-κB pathway (105). Remarkably, TLR3 deficient mice have shown a Th2-like biased immune response, further exacerbating the

eosinophils infiltration and mucus secretion (60). TLR7, in turn, recognizes viral single-stranded RNA and induces the secretion of T cells-activator and mucus-secreting cytokines such as IL-12 and IL-23 (61).

Mucus secretion is also a significant factor associated with hRSV infection. The production of this thick layer that works as another defense mechanism of the organism is performed by goblet cells (**Figures 1C**, **2**) (62). These cells are activated by cytokines such as IL-13, IL-17, and IL-23 (67), TLRs such as TLR3, and TLR7 (60) and immune receptors such as CXCR2 (68). As described above, TLR3 upregulation, secretion of IL-13 by infiltrating eosinophils and activation of several immune receptors are hallmarks of hRSV infection. Therefore, higher production of mucus usually correlates with more severe disease. Remarkably, Mukherjee et al. described that upon blockade of IL-17 through neutralizing antibodies, the secretion of mucus by hRSV-infected mice was significantly reduced, leading to a less exacerbated obstruction of the airways (69).

As it can be seen, the immune response against hRSV may be redundant at some points, but this redundancy itself is in part aiding the exacerbated inflammation and the production of pro-inflammatory cytokines. Although the organism exhibits several mechanisms to impede the advance of hRSV throughout the upper respiratory tract, this virus can avoid and even take advantage of many of these, eventually reaching the lower respiratory tract, where it will continue to replicate and progress in its pathology.

### Cytokines Induced by hRSV Infection in the Lower Respiratory Tract

It has been described that hRSV is a mucosa-restricted virus, as in natural infections it initially replicates in the epithelium of the nasopharynx (110). In immunologically naïve infants, hRSV spreads through a cell to cell transfer and extracellular binding, producing discontinuous foci of infection in the tracheal epithelium (110). The lower respiratory tract is essential for the respiratory system and is composed of trachea, bronchi (primary and secondary) and alveoli (111). Under normal conditions, inhaled pathogens are cleared via the mucociliary escalator from ciliated epithelial cells. This defense mechanism is coordinated with the actions of the airway lining fluid, rich in antioxidants, defensins, and lysozyme secreted by Clara cells and submucosal glands, along with mucous glycoproteins secreted by goblet cells (**Figure 2**) (66).

An exacerbated hRSV infection is characterized by several symptoms including severe chesty cough, wheezing, apnea and cyanosis and all of these symptoms can be signs of a lower respiratory tract infection (LRTI) (112). In infants, the leading pathology caused by LRTI is bronchiolitis, which has been described to involve an acute inflammation—mainly associated with exacerbated infiltration of neutrophils- necrosis of epithelial airway cells and increased production of mucus, among others (113, 114). Additionally, it has been described that the damage observed in the respiratory tract is not only induced by the viral infection itself, but also by the local production of cytokines (88). In the bronchioles samples from post-mortem patients, hRSV was detected mainly in the ciliated cells (115). Moreover, the majority of inflammation observed was at the submucosa level (115).

To understand the nature of this inflammation, the majority of the studies that have been performed in patients are focused on the analysis of the production of cytokines that coordinate the infiltration of immune cells. Since it is difficult to obtain bronchoalveolar lavage fluids (BALFs) samples from patients, these studies have been performed on ventilated hRSV-infected infants (116, 117). McNamara et al. collected these samples from term and pre-term infants to determine the inflammatory mediator profile in these children. An increase of the transcript and protein levels of cytokines such as IL-6, TNF-α, CXCL8, CCL3, CCL4, CCL2, and CCL5 was observed, as compared to control groups (**Figure 2**) (116, 117).

It has been reported that cytokines associated with a Th2-like response, such as IL-4, IL-6, IL-9, IL-10, and IL-13, are elevated in nasal washes and lungs of children with hRSV-induced LRTI (**Figure 2**) (72). Among these, IL-6 is a pro-inflammatory cytokine which has been described to play an essential role in the host immune response against hRSV infection (116). McNamara et al. also showed in pre-term and term infants with hRSVinduced bronchiolitis that IL-6 levels were elevated at day 1 of intubation, in term infants as compared to pre-term and control group (116). According to this observation, it is possible that IL-6 plays a relevant role in the hRSV pathogenesis in the lung of infected infants. When comparing the concentration of chemokines on the first day of intubation and the extubating day, no differences were found between these critical days (57). During another study performed in children under the age of 2 with clinical manifestations of respiratory obstruction and distress due to viral infection, different cytokines related to hRSV-infection were evaluated at three-time points: admissionhospitalization, discharge and 1 month after release (70). At the beginning of the study, the children admitted exhibited an increase in Th2-like cytokines such as IL-4, IL-5, and IL-13 (70). The increase in these cytokines decayed progressively until 1 month after discharge. In another study in children with signs of severe LRTI and positive for hRSV infection (71), cytokines were evaluated at two points: discharge and 1 year after release. Th2-like cytokines, such as IL-4 and IL-6, decayed 1 year after the infection. Surprisingly, IL-13 levels remained higher in the initially infected group when compared with the control group 1 year after the viral infection, although the authors could not rule out the effects of other diseases or environmental factors (71). Furthermore, children admitted in hospital with bronchiolitis due to hRSV-infection exhibited higher concentrations of IL-6 in nasal swabs as compared with their older siblings (118).

IL-10 has also been described as a key cytokine in the response against this virus (119–121). The varying levels and the role of IL-10 during hRSV infection have not been entirely determined, as IL-10 fluctuates with the age of children (120). Importantly, a study found that lower levels of IL-10 correlate with the severity of the hRSV disease in infants (120). Additionally, it has been described that in infants older than 3 months of age with mild hRSV infection exhibit high IL-10 levels, which can be related to a protector effect. Nonetheless, in infants below 3 months of age, high IL-10 levels were reported in those with severe bronchiolitis, therefore being considered as a hallmark of disease (121).

Interestingly, it has been reported that infants younger than 3 months, hospitalized with hRSV-induced bronchiolitis, presented elevated amounts of Th2-related cytokines in BALF samples, such as IL-3, IL-4, IL-10, and IL-13 (75). Furthermore, an increase of pro-inflammatory cytokines such as IL-1β, IL-6, TNF-α, and also IL-12-p40 -a Th1-like related cytokinewas also reported (75). Importantly, IL-3 -which is involved in the infiltration of immune cells that are related to the asthma development- and IL-12p40 are necessary for the secretion of IFN-γ. Therefore, the increase of both cytokines correlates with recurrent episodes of wheezing in hRSV infection (75).

IFN-γ is a cytokine that stimulates viral clearance by promoting anti-viral immune effector responses. Therefore, low levels of this cytokine in patients have been associated with a higher severity index in the bronchiolitis caused by hRSV (**Figure 2**) (122). Semple et al. reported that in infants hospitalized due to hRSV-induced bronchiolitis who needed oxygenation or ventilated support, IFN-γ levels in BALF were low when compared with the infants that never required oxygenation (122). These low IFN-γ levels correlated with increased severity of the disease and its reduction is significant in the development of the bronchiolitis (122). Contrary to these findings, recent studies performed by Thwaites et al. shows high levels of IFNγ in patients from the pediatric intensive care unit (PICU) with hRSV-infection, along with high levels of IL-1 and IL-10 respect to the healthy controls (123). Also, reduced IFN-γ levels were detected in children with moderate bronchiolitis; however, in children with severe bronchiolitis, the levels of IL-17A and MUC5AC were increased (123). Considering the data mentioned above, the amount of IFN-γ in patients with hRSV-bronchiolitis is controversial.

Additionally, Semple et al. also analyzed the production of IL-9 (122). This cytokine is produced in Th9-like immune response and has been implicated in the severity of the hRSV pathology (124). The data obtained showed that IL-9 levels in BALF were increased in infants with severe bronchiolitis that required oxygenation. However, no differences were found when compared with infants that never needed oxygen supplementation (122). In another study performed in pre-term and term infants with hRSV bronchiolitis, the expression of the IL-9 mRNA in BALF was increased in both groups, as compared to control groups (124). Moreover, no significant differences were found in the levels of IL-9 transcript among pre-term and term infants. However, the protein secretion was increased in term infants when compared to both pre-term and control groups (124). Furthermore, the primary source of these cytokines in the lungs of hRSV-infected infants were neutrophils (124). Remarkably, it has been reported that IL-9 can upregulate genes involved in the mucus production in goblet cells, which could explain the elevated amounts of mucus in patients with hRSVinduced bronchiolitis (124). Additionally, it has been reported that IL-9 polymorphism has a different effect in the hRSV disease severity in boy and girls (125). The single nucleotide polymorphism (SNP) rs2069885 of the IL-9 gene was associated with higher susceptibility of severe disease caused by hRSV while in boys, is associated with a lower susceptibility (125).

Also, it has been described an association of the polymorphism of the IL-4 and IL-4Rα genes with hRSV disease severity (73). Hoebee et al. found that the−590T allele of the IL-4 gene was expressed more frequently in infants hospitalized by hRSV bronchiolitis compared to the control group (73). Moreover, the authors found an association of the hRSV disease severity and the IL-4 locus in children older than 6 months that were hospitalized by hRSV bronchiolitis (73). Additionally, this study found that 2 polymorphisms of the IL-4Rα gene, the I50V, and the Q551R. Only the Q551R SNP show an association with the children older than 6 months who were hospitalized by a severe hRSV bronchiolitis (73).

Another relevant cytokine reported upon hRSV infection is the thymic stromal lymphopoietin (TSLP), a cytokine associated with asthma development (126). Also, a strong association in Th2-like effector cytokines, such as IL-4 and IL-5, and IL-13 has been reported (127, 128). TSLP is secreted by epithelial cells associated with barriers (129) and bronchial smooth muscle cells (130, 131). In infants with hRSV-induced bronchiolitis, this cytokine was elevated as compared to healthy controls, suggesting that TSLP could play an essential role in the hRSV immunopathology (132).

Importantly, the response associated with IL-17 can be harmful to the patients, as mentioned above. Higher levels of IL-17 have been reported in patients with mechanic ventilation due to hRSV-induced bronchiolitis (76, 78). IFN-λ is a cytokine discovered in the year 2003 (77) and it has been reported to play a role in the establishment of the adaptive response to hRSV, with an increase in the secretion of IL-6, CXCL8, and IL-10 in peripheral blood mononuclear cells (PBMC) (133). Moreover, a deleterious effect of IFN-λ in hRSV infection has been seen (134), as a study of acute bronchiolitis-patients reported a significant increase in the transcription of IFN-λ in patients with increased respiration rate, a sign of acute bronchiolitis induced by hRSVinfection (134).

As we described earlier, chemokines are also involved in the inflammatory response elicited by hRSV-infection. One of these is CCL3, a small pleiotropic chemoattractant protein whose function is to attract or activate immune cells such as eosinophils, monocytes, basophils and lymphocyte subpopulations (135). This chemokine was increased in lower respiratory tract secretions from infants under 2 months old that were hospitalized with hRSV-induced bronchiolitis (135). Interestingly, this increase was correlated with the detection of eosinophil degranulation products, which suggests that CCL3 has an active role in this process during hRSV-induced bronchiolitis (135). In addition to this, it was also reported that CCL5 was increased in these infants (135). CCL5 is a chemoattractant cytokine that principally recruits monocytes, T cells, and eosinophils, acting via three chemokine receptors: CCR1, CCR3, and CCR5 (136). Evidence obtained from children with hRSV infection shows an increase of the CCL5 protein levels in both upper and lower airway secretions, and levels of CCL5 in upper airway secretions correlate positively with disease severity (137, 138). Recently a prospective study of 173 patients with bronchiolitis caused by hRSV was performed, holding 536 healthy controls whose samples of nasopharyngeal aspirate were taken (139). Therein, the authors found a single SNP in CCL5 (rs2107538∗CT), exhibiting an association with hRSVbronchiolitis and also with the need for mechanical ventilation (139). These data suggest that CCL5 contributes to bronchiolitis leading to airways damage in patients.

Furthermore, McNamara et al. also evaluated this chemokine in BALF from infants that required ventilation support and found an increase at the first day of the mechanical ventilation, but these levels decreased over time (117). This phenomenon was also observed for CXCL8 (117), which is a chemokine that attracts mainly neutrophils, one of the most frequent immune cells found in the airways of hRSV-infected infants (140). Subsequently to these results, another study performed in BALF samples from intubated infants reported elevated levels of CXCL8 transcript, which also correlates with the finding of this chemokine in nasopharyngeal aspirates (NPA) (141). Thus, the NPA samples might be an excellent alternative to study the implications of the infection caused by hRSV in the respiratory tract (**Figure 2**).

#### CYTOKINES SECRETED BY EPITHELIAL CELLS IN RESPONSE TO THE hRSV INFECTION IN VITRO

The majority of the knowledge available about the induction of pro-inflammatory cytokines and chemokines production upon hRSV infection has been described in vitro using airway epithelial cells (AECs) models such as A549, primary human small airway epithelial cells (SAECs), BEAS-2B and primary normal human bronchial epithelial cells (NHBE), among others (88, 142, 143). The data obtained using these models can vary depending on the cell line. According to this, experiments in the A549 cell line (human alveolar type II-like epithelial) with the Long strain of hRSV showed that infection with hRSV induces the secretion of IL-6, CCL3, and CCL5 at 48 h post-infection as compared to noninfected cells (144, 145). On the other hand, a study performed in this cell line but with a different strain and subgroup of hRSV obtained from clinical isolates showed that the induction of IL-6 and CCL5 could be variable and dependent on the virus strain used (146).

BEAS-2B is an SV40 transformed human normal bronchial epithelium cell line that exhibits a limited susceptibility to hRSVinfection and profile of virus resistance as compared to the A549 cell line (142). Infection of this cell line with the hRSV Long strain showed an increase in the transcript levels for CXCL8 at 4 h post-infection, which was observed up to 24 h post-infection (143). Regarding the upregulation of IL-6, it was observed only at 96 h post-infection (143). However, another study performed using the hRSV Long strain showed that CXCL8 levels were not changed upon infection with hRSV, while CCL3 and CCL5 levels were increased (64). Importantly, the authors observed that the amounts of CCL5 produced by the epithelial cells were enough to attract eosinophils (64). Furthermore, infected BEAS-2B cells with the hRSV strain A2 also exhibited an induction in the secretion of IL-6 and CXCL8 as compared to non-infected cells (145).

Currently, primary or normal epithelial cells are the most used model for hRSV infection as it is thought to be representative of the effects of hRSV infection in the respiratory tract. Accordingly, it has been reported that primary AECs obtained from hRSVinfected infants exhibited higher viral titers as compared to the BEAS-2B cell line when infected with the same virus (147). Besides, the amounts of IL-6 and CXCL8 were higher in the primary AECs as compared to BEAS-2B cells (147). Considering these data, AECs can be considered as an excellent model for understanding the effects of hRSV-infection and the production of cytokines and chemokines as may occur in infants.

Additionally, in a study in vitro using a WT hRSV A2 strain (6340WT) and a recombinant strain that lacks the Gprotein gene (63401G), infection of NHBEs cells induced the secretion of CCL2, CCL5, and CXCL8 by both viruses (148). However, only the recombinant virus was able to promote the secretion of CXCL10 in NHBEs (148). On the other hand, both F- and G-protein promoted the secretion of CXCL8 and CXCL10, whereas only G-protein induced the secretion CCL5 (148). In contrast to these findings, infection with the hRSV Long strain in NHBEs cells did not lead to the secretion of CCL2 and CCL3, but the levels of CCL5 were increased as compared to uninfected cells (149). Additionally, it has been described that hRSV infection in NHBEs cells induced the expression of TSLP transcript at 12 h post-infection and TSLP secretion exhibited a peak at 24 h post-infection as compared to ultraviolet (UV)-hRSV inactivated (150). This phenomenon was also seen in NHBEs cells obtained from asthmatic patients and infected with hRSV, as TSLP concentration were high when compared to healthy patients with hRSV-infection (150). Moreover, studies using A549 cells co-transfected with the human TSLP promoter with a reporter, and a dominantnegative form of RIG-I (DN-RIG-I), showed that hRSV-infection could induce activation of this pathway to increase TSLP expression (150).

### hRSV INDUCED CYTOKINE PRODUCTION AND TISSUE DAMAGE IN MICE

It has been described that hRSV-infected BALB/c mice can exhibit increased levels of IL-6 in BALFs at 12 h post-infection that remains elevated up until 14 days post-infection (59). Similar results were observed in lungs parenchyma and sera of hRSV-infected mice (59). The contribution of IL-6 to the hRSV immunopathology was evaluated by the depletion of this cytokine one day before hRSV infection, parameters of disease, such as weight loss were more severe (59). Furthermore, in these hRSV-infected mice the lung vascular permeability was evaluated by measurement of albumin in the airways, which was increased as compared to the isotype control at 7, 11, and 14 days post-infection (59). Further, in the absence of IL-6, hRSVinfected mice displayed an increase of lymphocyte recruitment at 7 days post-infection, while neutrophil infiltration was similar to the isotype control (59). These results suggest that the early production of IL-6 is essential to control the severity of the disease and to limit lung damage.

Furthermore, it has also been described that hRSV infection promotes an increase of IL-1β, TNF-α, IFN-γ IL-12, IL-6, CCL3, and CCL5 in BALF samples from mice (151). The elevated levels of IL-1β and TNF-α on the first day of hRSV infection correlate with the peak of weight loss, whereas increased levels of IL-12 were found before the induction of IFN-γ (151). Besides, histological analyses have shown that hRSV infection produces changes in the lung that are associated with airway and vascular cuffing and interstitial pneumonia (144). On the other hand, an effect of TNF-α alone over the hRSV-infection has not been demonstrated with knockout mice. However, in a study in BALB/cJ mice with pretreatment with antibody for TNF-α before the hRSV-infection, mice showed a significant increase of weight loss and slow recovery as compared to control mice (152). Therefore, these observations suggest that TNF-α can be established as a participant in the hRSV-infection, and in the absence of this cytokine the mice showed a delay in the viral clearance.

The role of IFN-γ during hRSV pathogenesis was evaluated using both an IFN-γ knockout mice model and the blockade of IFN-γ (153). The data obtained from this study shows that, both in IFN-γ knockout (IFN-γ <sup>−</sup>/−) mice and in the anti-IFN-γ treated mice, the immune cell infiltration (principally neutrophils and eosinophils) in BALF samples were higher than in control mice. However, when the respiratory rate was evaluated [the ratio between inspiration time and expiration time (Ti/Te)] the anti-IFN-γ treated hRSV-infected mice shows no difference in the ratio compared to control mice (153). Besides, in the absence of IFN-γ also increase the viral load of these mice compared to control mice. These results suggest that IFN-γ plays a dual role during hRSV infection, been necessary to control the viral replication and also prevents the obstruction of the airways (153).

Regarding to the role of chemokines, has been reported that elevated concentrations of CCL3 and CCL5 at day one postinfection are consistent with recruitment of monocytes and lymphocytes into the mice lungs (151, 154). Additionally, it has been described that CCL5 induction by hRSV infection contributes to a subsequent allergic pulmonary inflammation (155). Moreover, in mice, the secretion of CCL5 in the lungs was correlated with airway hyperreactivity (AHR). This association was evaluated by antibody neutralization of CCL5, showing that while viral loads were not affected by this treatment, a significant decrease for hRSV-induced AHR was observed, down to control mice levels (156). Furthermore, it was described that CCL5 exhibits a biphasic response during the hRSV infection, with an initial phase of innate immune response and a second phase consisting of lymphocyte-mediated responses (157). Besides, mice sensitized with recombinant vaccinia expressing G-hRSV protein (rVV-G) showed a significant increase of both mRNA and protein levels for CCL5 during the first 24 h post-infection (157). Then, CCL5 is also increased in the second phase of hRSV infection at 168 h post-infection (157). To understand the role of CCL5 when viral replication was eliminated, an inhibitory analog of CCL5, Met-RANTES, was used to treat hRSV-infected mice. These studies showed that mice treated with Met-RANTES exhibited a significant reduction of CD4<sup>+</sup> and CD8<sup>+</sup> T cell recruitment into the lungs after infection (157). Along these lines, blockage of CCL5 reduced both weight loss and eosinophilia, suggesting that this cytokine plays an essential role during lung inflammation (157). Accordingly, the induction of CCL5 by hRSV infection is involved in lung inflammation, although there is no evidence of a contribution or a direct role in airway damage.

Regarding the contribution of CCL3 to hRSV infection, it was shown that equivalent to CCL5, CCL3 displays a biphasic expression both for mRNA and protein, at day 1 and 7 postinfection (158). Moreover, blockage of CCL3 with a neutralizing antibody showed no change in the recruitment of NK cells and did not affect viral loads in the lungs of hRSV-infected mice after 4 days of infection (158). However, at 7 days post-infection, the number of CD4<sup>+</sup> and CD8<sup>+</sup> T cells was reduced in the lungs of infected mice (158). Accordingly, hRSV-infected BALB/c WT mice exhibited an infiltration of about 80% of mononuclear cells close to vessels and bronchioles, while CCL3−/<sup>−</sup> hRSVinfected mice exhibited a decrease of infiltrating cells in the lungs. Interestingly, in both mice, the viral loads were equivalent (154). Additionally, in CCL3−/<sup>−</sup> hRSV-infected mice the mRNA of CCL5, CCL2, and CXCL2 were decreased as compared to their wildtype littermates, suggesting that CCL3 is required for the development of the hRSV-induced immunopathology (154). Despite these data, there is no direct evidence of the pulmonary damage caused by CCL3, which be relevant to determine.

Related with the production of TSLP in hRSV-infected mice, it has been reported that at the peak of the immunopathology, high amounts of this cytokine are produced (150). The contribution of TSLP to the hRSV pulmonary immunopathology was analyzed using TSLP KO mice and results showed that expression of Gob5, IL-13, and mucus production decreased as compared to hRSV-infected WT mice (150). Moreover, Stier et al. showed that knockout mice for the TLSP receptor (TSLPR KO) infected with hRSV displayed moderate mucous cell metaplasia, as the WT hRSV-infected mice. However, the accumulation of intraluminal mucus was lower when compared to WT mice (159). The airways obstruction of both hRSV-infected WT and TSLPR KO mice, was evaluated by methacholine challenge. Consistently, hRSVinfected WT mice displayed an increase in the airway reactivity (increased amounts of methacholine) as compared to the hRSVinfected TSLPR KO mice. These later animals showed only minor symptoms of the disease, which were equivalent to the mocktreated mice (159). Accordingly, these results suggest that TSLP activity is relevant for the hRSV immunopathology and that also contributes to lung damage in murine models (150).

As described above, most of the work in this field suggests what cytokines are either up- or down-modulated during hRSV infection. However, little or nothing has been reported about the direct contribution of these mediators to the airway damage caused by hRSV. The development of new methodological approaches is still necessary to achieve a better understanding of the effects that this virus produces on the respiratory tract by inducing inflammatory mediators. However, it could be possible to suggest that, like what is seen for the upper respiratory tract, hRSV exhibits several redundant mechanisms that induce damage and inflammation in the lower respiratory tract.

### hRSV INFECTION IS ASSOCIATED WITH ELEVATED LEVELS OF CYTOKINES IN THE CNS

As described above, hRSV infection induces cytokines that damage the respiratory tract, but also these cytokines could affect the CNS. Years ago, a small number of hRSV-infected patients were reported to exhibit clinical signs associated with neurological complications, such as seizures (160, 161), apnea (12), encephalopathy (162) and encephalitis (163, 164). Nowadays, the cases of neurological abnormalities related to hRSV infection reported are increasing. However, our knowledge regarding the mechanisms involved in this phenomenon remains limited and controversial (**Figure 1**).

One of the first findings in patients with neurological manifestations associated with hRSV infection was the detection of virus-specific antibodies in cerebrospinal fluid (CSF) (165). Later, after many efforts to find viral genetic material in CSF, hRSV RNA belonging to the serogroup B was detected in the CSF of an infant with febrile convulsion and pneumonia (13). Researchers not only have focused on hRSV detection, but also on the possible production of cytokines that could be a consequence of viral infection and that could explain the symptoms affecting the CNS. Accordingly, an increase of IL-6 in CSF from an hRSVinfected patient was reported (14). The observation that serum IL-6 levels in these patients were normal (14) would suggest that this cytokine is produced locally in the CNS, most likely by CNS-resident cells, such as microglia and astrocytes. A report about 3 clinical cases where children infected by hRSV suffered from seizures, showed that the levels of IL-6 were increased and that the serogroup of hRSV found in the CFS belonged to the serogroup A (166). Additionally, the same authors also found viral RNA in the CSF of a different cohort of hRSVinfected patients, with increased levels of IL-6, IL-8, CCL2, and CCL4, suggesting that these inflammatory mediators may play a critical role during the hRSV-infection in the CNS pathogenesis (15). Importantly, the increased levels of IL-6 correlate with the severity of the CNS encephalitis mediated by a cytokine storm, which can be useful as a molecular marker of neurological prognosis (167). Based on all the data described above, it is possible that hRSV spreads from the lungs to the CNS and infects local cells, initiating an inflammatory immune response mediated by cytokines.

As we mentioned above, there is controversy in this field due to reports in which hRSV-derived genetic material was not found in CSF samples from patients with severe bronchiolitis (168). Analyses of blood and CSF samples from 10 patients with apneas showed that only 7 were positive for hRSV (168). This study showed that hRSV RNA was detected in PBMC of two patients, but was not found in their CSF (168). Possible explanations for this controversy could be due to differences in the clinical signs of the patients, to the hRSV serogroups found infecting them and also to technical differences used by the researchers.

Although there is clinical relevance in the CNS pathologies caused by hRSV infection, there is little research in this aspect that could provide conclusive evidence. A study using the mouse model described that hRSV could infect sensory neurons in the lungs through the interaction of the G-hRSV glycoprotein with the chemokine receptor CX3CR1 located at the surface of these cells (169). Experiments with mouse neuronal primary cultures showed that hRSV infected about 5% of these cells and this percentage decrease when CX3CR1 was blocked (169). Nevertheless, in this study, the authors did not evaluate the effect of hRSV infection in these cells or whether the neurons secreted inflammatory mediators. To approach these questions, neuronal N2a cells were infected with hRSV showing that these cells secrete IL-6 and TNF-α in vitro (170).

While these reports advance the knowledge in this field, there is still no evidence of neuronal infection by hRSV in vivo. In this regard, Espinoza et al. described that hRSV could be detected in several areas of the brain from infected mice, such as the cortex, ventromedial hypothalamic nucleus, and hippocampus (171). Interestingly, the finding of the virus in the hippocampus led to hypothesize that behavioral and learning processes may be altered. Marble burying (MB) and Morris Water Maze (MWM) test were performed 30 days after hRSV infection to test this hypothesis. In both trials, behavioral (MB) and spatial learning (MWM), performance was altered in hRSVinfected mice (171). The authors also evaluated the possible impairment in the functionality of the synaptic plasticity in the hippocampus. The data shows that the long-term potentiation (LTP) and the long-term depression (LTD) were altered in hRSVinfected mice, suggesting damage in the brain of these animals (171). It is possible to think that impairment in the behavior and learning is due to the neuronal infection by hRSV, which alters the normal function of these cells. Besides, it is also possible that hRSV infection promotes the secretion of several cytokines in the CNS, either by neurons or other resident cells, which could contribute to this neurological-associated phenomenon. However, more research is still necessary in this field to further advance our knowledge of the effects that this virus has on our CNS.

#### CONCLUDING REMARKS

HRSV remains one of the primary viral agents causing respiratory tract infections worldwide, for which there is no vaccine available. Once hRSV infection reaches the epithelium of the respiratory tract, it produces several symptoms such as wheezing, apnea, cyanosis, and bronchiolitis, related to acute

#### REFERENCES


lower tract infection. Most of the damage seen in patients with complications associated with hRSV infection is caused by an exacerbated immune response triggered mainly by the cytokines secreted by the infected cells of the respiratory tract epithelium.

In human studies, cytokines and chemokines have been detected in nasopharyngeal aspirates, tracheobronchial aspirates or bronchoalveolar lavage fluids, in children with mechanic ventilation due to bronchiolitis associated with hRSV infection. In these patients -usually children younger than 2 years of age- the cytokines that predominated were IL-4, IL-5, IL-6, IL-10, and IL-13. A low concentration of cytokines associated with a Th1-like response such as IFN-γ is also seen, which could be considered as a severity index. Also, chemokines such as CCL3, CCL5, and CXCL8 are increased in the lower respiratory tract of individuals infected with hRSV. These components contribute to generating a severe pathology in the patients, which is associated with an unbalance between Th1- and Th2-like cytokines, and an increase in chemokines that attract more inflammatory cells like granulocytes, which in turn generates a deleterious effect on the patient. Moreover, the secretion of many of the cytokines described above has also been seen in mice models, with associated tissue damage, although further studies are still required to fully elicit the specific role of each cytokine in this pathology.

Furthermore, infection by hRSV seems to reach CNS, which produces high levels of IL-6 in the zone. This infection might generate problems in the behavior and learning process of the children, but further studies are required to elucidate more information in this regard.

#### 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 FONDECYT grants N◦ 3180570, N◦ 1150862, N◦ 1190830 and the Millennium Institute on Immunology and Immunotherapy (P09/016-F). AMK is a Helen C. Levitt Visiting Professor, Department of Microbiology and Immunology, University of Iowa.

aspects. Mem Inst Oswaldo Cruz Rio Janeiro. (2003) 98:739–43. doi: 10.1590/S0074-02762003000600005


and meta–analysis. J Glob Health. (2015) 5:020416. doi: 10.7189/jogh.05. 020416


and allergy at age 13. Am J Respir Crit Care Med. (2005) 171:137–41. doi: 10.1164/rccm.200406-730OC


epithelial responses to it in human nasal epithelial cells. (2013) 8:e70225. doi: 10.1371/journal.pone.0070225


Int J Neurosci. (2017) 127:44–50. doi: 10.3109/00207454.2016. 1138951


**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 Bohmwald, Gálvez, Canedo-Marroquín, Pizarro-Ortega, Andrade-Parra, Gómez-Santander and Kalergis. 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.

# Aflatoxin B<sup>1</sup> Promotes Influenza Replication and Increases Virus Related Lung Damage via Activation of TLR4 Signaling

Yuhang Sun1,2, Jiarui Su1,2, Zixuan Liu1,2, Dandan Liu1,2, Fang Gan1,2, Xingxiang Chen1,2 and Kehe Huang1,2 \*

<sup>1</sup> Department of Animal Nutrition and Immunology, College of Veterinary Medicine, Nanjing Agricultural University, Nanjing, China, <sup>2</sup> Institute of Nutritional and Metabolic Disorders in Domestic Animals and Fowls, Nanjing Agricultural University, Nanjing, China

Aflatoxin B<sup>1</sup> (AFB1), which alters immune responses to mammals, is one of the most common mycotoxins in feeds and food. Swine influenza virus (SIV) is a major pathogen of both animals and humans. However, there have been few studies about the relationship between AFB<sup>1</sup> exposure and SIV replication. Here, for the first time, we investigated the involvement of AFB<sup>1</sup> in SIV replication in vitro and in vivo and explored the underlying mechanism using multiple cell lines and mouse models. In vitro studies demonstrated that low concentrations of AFB<sup>1</sup> (0.01–0.25µg/ml) markedly promoted SIV replication as revealed by increased viral titers and matrix protein (M) mRNA and nucleoprotein (NP) levels in MDCK cells, A549 cells and PAMs. In vivo studies showed that 10–40 µg/kg of AFB<sup>1</sup> exacerbated SIV infection in mice as illustrated by significantly higher lung virus titers, viral M mRNA levels, NP levels, lung indexes and more severe lung damage. Further study showed that AFB<sup>1</sup> upregulated TLR4, but not other TLRs, in SIV-infected PAMs. Moreover, AFB<sup>1</sup> activated TLR4 signaling as demonstrated by the increases of phosphorylated NFκB p65 and TNF-α release in PAMs and mice. In contrast, TLR4 knockdown or the use of BAY 11-7082, a specific inhibitor of NFκB, blocked the AFB1-promoted SIV replication and inflammatory responses in PAMs. Furthermore, a TLR4-specific antagonist, TAK242, and TLR4 knockout both attenuated the AFB1 promoted SIV replication, inflammation and lung damage in mice. We therefore conclude that AFB<sup>1</sup> exposure aggravates SIV replication, inflammation and lung damage by activating TLR4-NFκB signaling.

Keywords: aflatoxin B1 , swine influenza virus, replication, inflammation, lung damage, TLR4, NFκB, TNF-α

#### INTRODUCTION

Swine influenza virus (SIV), a single-stranded negative-sense RNA virus, causes severe systemic effects, resulting in significant economic losses in the animal husbandry industry. SIV also causes human disease and can even give rise to human pandemics, including the pandemic caused by the H1N1/2009 virus (1). Increasing evidence indicates that viral infection is associated with several environmental, nutritional, and immune factors, such as mycotoxin contamination (2, 3), selenium deficiency (4, 5), and macrophage polarization (6). The involvement of these factors may partly

#### Edited by:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany

#### Reviewed by:

Florian Krammer, Icahn School of Medicine at Mount Sinai, United States Dunja Bruder, Universitätsklinikum Magdeburg, Germany

> \*Correspondence: Kehe Huang khhuang@njau.edu.cn

#### Specialty section:

This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology

Received: 19 May 2018 Accepted: 17 September 2018 Published: 04 October 2018

#### Citation:

Sun Y, Su J, Liu Z, Liu D, Gan F, Chen X and Huang K (2018) Aflatoxin B1 Promotes Influenza Replication and Increases Virus Related Lung Damage via Activation of TLR4 Signaling. Front. Immunol. 9:2297. doi: 10.3389/fimmu.2018.02297

**24**

explain the differences in morbidity and mortality in infected animals and humans all over the world.

Aflatoxin B<sup>1</sup> (AFB1), which is produced by Aspergillus flavus, is one of the most common mycotoxins in contaminated food and plant products from tropical and subtropical areas with high temperature and humidity (7, 8). It is well known that AFB<sup>1</sup> is harmful to the liver and kidney of mammals and is regarded as a representative orally ingested carcinogen (9, 10). However, increasing evidence indicates that AFB<sup>1</sup> can also affect immune responses in mammals (11, 12); these evidences show that low doses of AFB<sup>1</sup> (≤0.025 mg/kg) significantly increase the secretion of pro-inflammatory cytokines by T cells and NK cells in rats, but high doses of AFB<sup>1</sup> (0.4–0.8 mg/kg) markedly decrease macrophage migration and the lymphocyte response to mitogens in pigs. Specifically, some reports propose that mycotoxins can eventually decrease resistance to infectious diseases (13), and aflatoxins are thought to feature prominently in the progression of some viral diseases, such as HIV (3). However, so far, there have been no studies investigating whether influenza virus infection in mammals exposed to AFB<sup>1</sup> is more severe than infection in unexposed mammals.

Toll-like receptors (TLRs) compose a main family of pattern recognition receptors with a critical role in the activation of the innate immune response (14). To date, there are at least 13 members (TLR1-TLR13) of this family in mammals that recognize specific components of pathogenic microorganisms. TLR4 is a unique receptor for pathogen recognition that was initially found in various cell types, including porcine alveolar macrophages, and in mice. In the past, many studies have focused on TLR4 structure and function. On the one hand, TLR4 activation leads to nuclear factor kappa (NFκB) translocation and the expression of proinflammatory cytokines, including tumor necrosis factor (TNF-α), which is responsible for activating the innate immune system (15). On the other hand, the overexpression or continuous activation of TLR4 leads to excessive inflammatory responses and/or tissue injury in the body (16–19). Interestingly, viruses can evade the host immune response when TLR4 is inhibited, thereby enhancing viral replication, and one study has shown that a TLR4 antagonist can protect mice from lethal influenza infection (20). Nevertheless, few studies are available regarding the role played by TLR4 in AFB1−promoted SIV replication.

Thus, given the differences in morbidity and mortality following SIV infection, we hypothesized that AFB<sup>1</sup> promotes SIV replication. In this study, multiple cell lines and mouse models were established to assess the involvement of AFB<sup>1</sup> in SIV replication in vitro and in vivo and to elucidate the underlying mechanism of such involvement.

#### MATERIALS AND METHODS

#### Ethics Statement

This research protocol was approved by the Ethics Committee for Animal Experimentation of Nanjing Agricultural University (approval number: SYXK-SU-2011-0036). All animal care and use procedures were conducted in strict accordance with the Animal Research Committee guidelines of the College of Veterinary Medicine at Nanjing Agricultural University, and all efforts were made to minimize animal suffering and to reduce the number of animals used.

#### Reagents

AFB<sup>1</sup> (1 mg/mL; Sigma-Aldrich, USA), BAY 11-7082 (10 mM; MCE, USA) and TAK-242 (50 mM; ApexBio, USA) were dissolved in dimethyl sulfoxide (DMSO), packaged, and stored frozen at −20◦C until use. Forin vitro studies, the dissolved AFB<sup>1</sup> was diluted with serum-free medium, and equal concentrations of DMSO were used in the vehicle and in the control solution. For in vivo studies, the dissolved AFB<sup>1</sup> was diluted in fresh sterile endotoxin-free saline daily, and the solution was then injected intraperitoneally (i.p.) at concentrations of 10, 20, and 40 µg/kg b.w.; diluted TAK-242 was also prepared daily and then injected i.p. (3 mg/kg b.w.) 1 h prior to other treatments as previously described (16, 17, 21).

### Cell Culture

Madin-Darby canine kidney (MDCK, NBL-2) cells, human lung cancer cells (A549) and porcine alveolar macrophages (PAMs, 3D4/21) that were free of any respiratory or systemic diseases were purchased from the China Institute of Veterinary Drug Control (Beijing, China). MDCK and A549 cells were grown in Dulbecco's modified Eagle's medium (Gibco, USA) containing 10% fetal calf serum (FCS; Gibco, USA) and 1% penicillinstreptomycin (Solarbio, China) at 37◦C in 5% CO2. PAMs were cultured in Roswell Park Memorial Institute-1640 medium (Gibco, USA) supplemented with 10% FCS, 1% penicillinstreptomycin and 1% nonessential amino acids (Gibco, USA) at 37◦C in 5% CO2. Cells and serum and culture medium were tested for mycoplasma using MycoTestTM kit (Seebio, China). During viral infection, all cell lines were transferred to serum-free medium supplemented with 1µg/ml tolylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin (Sigma, USA).

### Cell Viability Determination by MTT and LDH Assays

MDCK cells, A549 cells and PAMs were cultured in 96-well plates for 24 h and were then exposed to various concentrations of AFB<sup>1</sup> or to 1µg/ml DMSO for an additional 24 h before being subjected to colorimetric 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) assay. Subsequently, the absorbance was measured at 490 nm with a reference wavelength of 655 nm, and all experiments were performed in triplicate.

Lactate dehydrogenase (LDH) release was also measured using commercially available kits to assess cell viability. Briefly, cells were seeded in 96-well plates and exposed to various concentrations of AFB<sup>1</sup> or to 1µg/ml DMSO. After 24 h of incubation, the supernatant was collected for the measurement of LDH release according to the manufacturer's protocol (Jiancheng, China). The absorbance was measured at a wavelength of 450 nm, and all samples were measured in triplicate.

#### Apoptosis Assay by DAPI Staining

4 ′ ,6-diamidino-2-phenylindole (DAPI) staining was performed as described previously (22) with a minor modification. Briefly, PAMs were seeded on coverslips (WHB, China) into 12-well culture plates and incubated with AFB<sup>1</sup> and DMSO for 24 h. Next, the PAMs were washed three times with PBS and fixed with 4% paraformaldehyde for 20 min at 4◦C. After three washes, cell nuclei were counterstained with DAPI (Beyotime, China) for 5 min in the dark. Finally, the stained PAMs were washed three times and examined by fluorescence microscopy (Nikon Ti-S, Japan).

## Viral Titration by TCID<sup>50</sup>

Influenza virus strain A/swine/Guangxi/18/2011 (H1N1) was kindly provided by Dr. Weiye Chen, Harbin Veterinary Research Institute, Chinese Academy of Agricultural Sciences (Harbin, China). The virus was propagated in MDCK cells, and the supernatant was harvested at 72 h post infection (hpi) to ensure that enough virus was obtained. The viral titers were determined by the 50% tissue culture infectious doses (TCID50) in MDCK cells, A549 cells and PAMs. Briefly, the MDCK cells, A549 cells and PAMs were seeded in a 96-well plate (Corning, USA) for 24 h, infected with 10-fold serial dilutions of virus in serum-free medium supplemented with 1µg/ml TPCK-treated trypsin and then exposed to various concentrations of AFB1. The cytopathic effect induced by the virus was observed and recorded after 24 hpi to calculate the virus titers by the method of Reed and Muench. A biosafety level 2 facility was used for all the experiments with the H1N1 virus.

### Animals and Study Design

Male TLR4 knockout (C57BL/10ScNJNju, TLR4−/−) and wildtype (C57BL/10JNju, WT) mice, 6–8 weeks old and weighing 18– 20 g, were purchased from Nanjing University (Nanjing, China). TLR4−/<sup>−</sup> mice do not express functional TLR4 or TLR4 mRNA because of the TLR4 lps-del mutation. All mice were housed in a specific pathogen-free environment (22 ± 2 ◦C) with a 12 h light/dark cycle. Water and food were available ad libitum throughout the whole study. All mice were acclimatized for 1 week before the onset of experiments. Body weight changes and illnesses were monitored daily.

For the first randomized trial, WT mice were randomly divided into 6 groups (each group included 3 replicates, with 4 mice per replicate): 4 groups were challenged intranasally with a nonlethal dose of H1N1 virus (1000 TCID50) (23) prior to treatment with AFB<sup>1</sup> on d 1, d 7, and d 14 as described previously (24, 25), and the other two groups were given equivalent amounts of PBS intranasally. Among the 4 infected groups, three groups were given 10, 20, and 40 µg/kg b.w. AFB<sup>1</sup> i.p. daily for 15 days, and the fourth group was given an equivalent amount of PBS i.p. Likewise, two uninfected groups were injected with equivalent amounts of PBS or AFB<sup>1</sup> (40 µg/kg).

For the second randomized trial, WT mice were randomly divided into 2 groups (each group included 3 replicates, with 3 mice per replicate): the first group was given 3 mg/kg of TAK242 i.p.1 h prior to the other treatments, and the other group was given an equivalent amounts of PBS i.p.

For the third randomized trial, TLR4−/<sup>−</sup> and WT mice were likewise divided into 2 groups (each group included 3 replicates, with 3 mice per replicate).

All mice from the second and third randomized trials were treated with equivalent amounts of H1N1 virus and 40 µg/kg of AFB<sup>1</sup> as described for the first randomized trial.

### Histopathological Examination and Immunohistochemical Staining

At the end of the experiments, mice were euthanized. Lung and spleen tissues were taken from each mouse. Approximately 75% of the lung tissue was stored at −80◦C for the subsequent experiments, and the other 25% of the lung tissue was fixed in 4% formaldehyde for hematoxylin-eosin staining (H&E) according to standard protocols as described previously (26, 27) with some modifications. Briefly, lung tissue was fixed in 10-fold volume of 4% formaldehyde for 48 h. Next, samples were embedded in paraffin and cut into 4-µm-thick sections. One section from each tissue sample was stained with H&E.

For immunohistochemical staining, spleen tissues were incubated with a monoclonal antibody for TLR4 (Abcam, UK) and then incubated with an appropriate horseradish peroxidase (HRP)-conjugated secondary antibody. Subsequently, the HRP conjugates were visualized using a diaminobenzidine solution. Images were captured with a Pannoramic viewer (Pannoramic MIDI, 3D HISTECH), and data were analyzed using DensitoQuant software (QuantCenter, 3DHISTECH). A histochemistry score (H-score) was calculated according to a previously reported equation (28, 29).

#### Fluorescent Quantitative Real-Time PCR (qRT-PCR) Analysis

Cells and lung and spleen tissues were collected to determine the relative mRNA expression levels of viral matrix (M) protein, TLRs, TNF-α and IL-10. Primers for the reference genes and target genes (**Table 1**) were designed and synthesized by Invitrogen based on known sequences. Briefly, total RNA was first extracted from tissues and cells using an RNAiso Plus kit (TaKaRa, Japan). First-strand cDNA was synthesized using a reverse transcription kit (TaKaRa, Japan). Subsequently, the samples of cDNA were subjected to qRT-PCR (TaKaRa, Japan) using specific primers with a no-cDNA template as a calibrator. The relative expression levels of the target genes were calculated by the 2−11CT method with 18S or GAPDH as an endogenous reference gene.

### Western Blot Analysis

Cells and lung and spleen tissues were collected for western blotting analysis to assess the relative expression levels of viral nucleoprotein (NP), phosphorylated NFκB p65 (pp65) and TLR4. Briefly, total protein was extracted, and the protein concentration was measured with a BCA kit (Beyotime, China). The proteins were denatured, subjected to 10–15% of sodium dodecyl sulfatepolyacrylamide gel electrophoresis and then transferred to polyvinylidene difluoride membranes (Bio-Rad, USA). Next, the membranes were blocked for 2 h at room temperature (RT) in 5% bovine serum albumin (BSA) in Tris-buffered saline containing



M, influenza A virus matrix protein; TLR, toll-like receptor; IL-10, interleukin 10; TNF-α, tumor necrosis factor-α; GAPDH, glyceraldehyde-phosphate dehydrogenase.

0.1% Tween 20 (TBST), incubated overnight at 4◦C with specific primary antibodies from (anti-NP, ab128193; anti-TLR4, ab13556; anti-pp65, ab76302 or anti-actin, ab14128; Abcam, UK), and then incubated for 1 h at RT with appropriate secondary antibodies (horseradish peroxidase-labeled anti-mouse or antirabbit secondary antibodies; Cell Signaling Technology, USA). Finally, the bound antibodies were visualized using an enhanced chemiluminescence kit (Beyotime, China).

#### Determinations of TNF-α and IL-10 by ELISA

Whole blood from mice was collected from the retro-orbital plexus in heparinized tubes by a trained individual and was allowed to clot at RT. Sera were separated by centrifugation and stored at −80◦C until analysis. The contents of TNF-α and IL-10 in sera were measured using ELISA kits (Jiancheng, China) according to the manufacturer's instructions.

#### Short Interfering RNA (siRNA) Transfection

A pig TLR4-specific siRNA sequence, 5′ - GGAUUUAUCCAGAUGUGAATT-3′ , and a control siRNA sequence were obtained from a paper published by our coauthor (18). qRT-PCR was performed to determine the interfering efficiency of siTLR4. The siRNA experiment was carried out as our coauthor described previously (4). Briefly, PAMs were seeded in 12-well plates and were transfected for 6 h with XtremeGENE siRNA transfection reagent (Roche, USA), siTLR4 and negative control siTLR4 diluted in medium according to the manufacturer's protocol, when the cells had reached approximately 70–80% confluence. Next, PAMs were infected with H1N1 virus and exposed to AFB<sup>1</sup> for an additional 24 h for further experiments.

#### Statistical Analysis

Statistical analysis was conducted using Prism 6 (GraphPad Software, La Jolla, CA). Data are presented as the means ± SEM. Unpaired two-tailed Student's t-tests were performed to evaluate statistical significance for two-group comparisons, and ordinary one-way (nonparametric) ANOVA with Tukey's posttests and two-way ANOVA with Dunnett's posttests were performed to evaluate statistical significance for multigroup comparisons. A value of P < 0.05 was considered significant, and P < 0.01 was considered strongly significant.

#### RESULTS

#### The Cytotoxic Effects of Various Concentrations of AFB<sup>1</sup> on MDCK Cells, A549 Cells and PAMs

To remove the effects of AFB1-induced cytotoxicity on viral replication, the effects of various concentrations of AFB<sup>1</sup> on cell viability were determined by MTT and LDH assays. As shown in **Figures S1A–C**, the viability of MDCK cells, A549 cells and PAMs decreased significantly when the AFB<sup>1</sup> concentrations were greater than 0.5, 0.5, and 0.1µg/ml, respectively. Correspondingly, LDH assay showed that LDH release increased markedly in MDCK cells, A549 cells, and PAMs when the AFB<sup>1</sup> concentrations were greater than 0.5, 0.5, and 0.1µg/ml, respectively (**Figures S1D–F**). Afterwards, DAPI staining was performed to determine the extent of apoptosis and thus to further assess the cytotoxicity of AFB<sup>1</sup> on PAMs. As shown in **Figure S1G**, apoptosis began to occur when the AFB<sup>1</sup> concentration reached 0.1µg/ml and was identified by the condensation and fragmentation of nuclei. In addition, given that AFB<sup>1</sup> was dissolved in DMSO, the effects of DMSO on MDCK cells, A549 cells and PAMs were also measured, and no significant differences were observed between the DMSO (1µg/ml) group and either of the control groups (no DMSO and no AFB1). Taken together, these results suggest that AFB<sup>1</sup> at concentrations between 0.01 and 0.25µg/ml, 0.01 and 0.25µg/ml, and 0.01 and 0.05µg/ml are not toxic to MDCK cells, A549 cells and PAMs, respectively. Thus, for subsequent experiments, AFB<sup>1</sup> was used at concentrations of 0.01, 0.05, and 0.25µg/ml in both MDCK and A549 cells and at concentrations of 0.01, 0.025, and 0.05µg/ml in PAMs.

### AFB<sup>1</sup> Promotes SIV Replication in MDCK Cells, A549 Cells and PAMs

To investigate the potential role AFB<sup>1</sup> plays in SIV replication, viral titers, viral M mRNA expression levels and NP expression levels were measured by TCID50, qRT-PCR and western blotting, respectively, as described previously (30). All cells were infected with SIV and then treated with various concentrations of AFB<sup>1</sup> for 24 h. As shown in **Figure 1**, viral titers, M mRNA expression levels and NP levels were significantly increased in SIV-infected MDCK (**Figures 1A,D,G**) and A549 cells (**Figures 1B,E,G**) treated with 0.01–0.25µg/ml AFB<sup>1</sup> compared with levels in cells without AFB<sup>1</sup> treatment. Correspondingly, viral titers, M mRNA expression levels and NP levels were also markedly increased in SIV-infected PAMs (**Figures 1C,F,G**) treated with 0.025– 0.05µg/ml AFB<sup>1</sup> compared with levels in non-AFB1-treated PAMs. To confirm that the increase in SIV replication induced by AFB<sup>1</sup> was not due to the presence of DMSO, we compared viral M mRNA expression of the three cell lines exposed to DMSO to that of the three cell lines exposed to medium and demonstrated that viral M mRNA expression in the three DMSO-exposed cell lines was identical to that in the cell lines exposed to medium alone (data not shown). Taken together, our results suggest that AFB<sup>1</sup> exposure promotes SIV replication in vitro.

#### AFB<sup>1</sup> Upregulates TLR4-NFκB Signaling and Promotes Inflammatory Responses in SIV-Infected PAMs

TLRs are a main family of pattern recognition receptors with a critical role in the activation of innate immune responses, but it has been proven that the overexpression or continuous activation of TLR4 can lead to excessive inflammatory responses or to injury in the body (16–18). To determine whether the promotion of SIV replication by AFB<sup>1</sup> is associated with TLRs-induced innate immune responses or injury, the expression levels of TLRs 1-10 in SIV-infected PAMs were investigated. As shown in (**Figure 2A**), the relative expression of TLR4 mRNA was significantly elevated following exposure to 0.025–0.05µg/ml AFB<sup>1</sup> compared with the expression in the control group. This finding was confirmed by the marked increases in TLR4 protein levels (**Figure 2B**). TLR4 induces NFκB activation (15), and a previous study indicated that NFκB signaling involves pathogen- or cytokineinduced immune and inflammatory responses (31). To further confirm whether TLR4-NFκB was activated, the levels of pp65 were also determined. The results showed that 0.025–0.05µg/ml AFB<sup>1</sup> significantly increased the relative protein levels of pp65 (**Figure 2C**). The inflammatory response was quantified based on the expressions of the TNF-α and IL-10 genes, and the results indicated that 0.025–0.05µg/ml AFB<sup>1</sup> significantly increased the relative TNF-α mRNA level but decreased the relative IL-10 mRNA level (**Figures 2D,E**). Taken together, our results

FIGURE 1 | AFB<sup>1</sup> promotes swine influenza virus (SIV) replication in multiple cell lines. Cells were infected with SIV (MOI = 1), and then incubated with various concentrations of AFB1 for 24 h. (A–C) Viral titers. Infectious virus particles were quantified by TCID50. (D–F) Viral M protein mRNA and (G) nucleoprotein (NP) levels were analyzed by qRT-PCR and western blotting, respectively. Data are presented as the means ± SEM of three independent experiments. Significance compared with the SIV control group, \*P < 0.05 and \*\*P < 0.01.

demonstrated that AFB<sup>1</sup> upregulated TLR4-NFκB signaling and promoted inflammatory responses in the SIV-infected PAMs.

### TLR4 Knockdown and BAY 11-7082 Administration Block the AFB1-Promoted SIV Replication and Inflammatory Responses in SIV-Infected PAMs

To further investigate the mechanism of SIV promotion by AFB1, a TLR4-specific siRNA sequence was used to remove the effects of TLR4, and a control siRNA sequence was used as a negative control. The interfering efficiency of siTLR4 was determined by qRT-PCR. As shown in (**Figure 3A**), TLR4 knockdown significantly decreased TLR4 mRNA expression by >70% compared with the expression in the blank; no significant difference in TLR4 mRNA expression was observed between the blank and siControl groups. In addition, our results demonstrated that 0.05µg/ml AFB<sup>1</sup> significantly elevated viral titers (**Figure 3B**), M mRNA expression (**Figure 3C**) and NP levels (**Figure 3D**) in SIV-infected PAMs compared to the corresponding parameters in control cells without AFB1. In contrast, TLR4 knockdown significantly reduced AFB1 promoted SIV replication, as indicated by lower viral titers, M mRNA expression and NP levels in the TLR4 knockdown group than in the siControl group; no significant difference in SIV replication was observed between the TLR4 knockdown and control groups (**Figures 3B–D**). These findings indicated that TLR4 knockdown blocked the promotion of SIV replication induced by AFB1. Likewise, TLR4 knockdown significantly reduced pp65 protein and TNF-α mRNA levels compared with the levels in the siControl group and even compared with the levels in the control group (**Figures 3E**,**F**), suggesting that TLR4 knockdown drastically counteracted the AFB1-promoted inflammatory responses in the SIV-infected PAMs.

Furthermore, our previous study indicated that BAY 11-7082 (10µM), a specific inhibitor of NFκB, significantly reduces pp65 in PAMs and does not have cytotoxicity in PAMs (32). In the present study, BAY 11-7082 was used to further confirm the mechanism of SIV promotion by AFB1. The results showed that compared with medium alone, BAY 11-7082 significantly reduced the elevations in viral titers (**Figure 3G**), M mRNA levels (**Figure 3H**), and TNF-α mRNA levels (**Figure 3I**) promoted by AFB<sup>1</sup> in SIV-infected PAMs, and no significant differences in the above parameters were observed between the BAY 11-7082 group and the control groups. Taken together, the results indicated that TLR4 knockdown and BAY 11-7082 blocked the AFB1-promoted SIV replication and inflammatory responses.

### AFB<sup>1</sup> Promotes SIV Replication and Lung Damage Induced by SIV in Mice

To further verify the in vitro results, lung tissues were taken from SIV-infected mice exposed to AFB<sup>1</sup> to assess viral replication

FIGURE 2 | AFB<sup>1</sup> upregulates TLR4-NFκB signaling and promotes inflammatory responses in SIV-infected PAMs. PAMs were incubated with or without SIV (MOI = 1), and then, the SIV-infected PAMs were incubated with various concentrations of AFB1 for 24 h. (A) Relative TLRs mRNA, (B) TLR4 protein, (C) phosphorylated NF-κB p65 (pp65), (D) TNF-α mRNA levels and (E) IL-10 mRNA levels. Data are presented as the means ± SEM of three independent experiments. Significance compared with the SIV control group, \*P < 0.05 and \*\*P < 0.01.

FIGURE 3 | TLR4 knockdown and BAY 11-7082 reduce the AFB1-promoted SIV replication and inflammatory responses. PAMs were infected with SIV (MOI = 1) and then treated with (AFB1 group) or without (control group) 0.05µg/ml AFB1. A TLR4-specific siRNA sequence was used to remove the effects of TLR4, and a control siRNA sequence was used as a negative control. (A) The knockdown efficiency of TLR4 siRNA in PAMs. (B) Viral titers and (C) relative viral M mRNA levels, (D) NP, (E) pp65 and (F) TNF-α mRNA levels. A specific inhibitor of NFκB, BAY 11-7082 (10µM), was added to remove the effects of NFκB, and medium was used as a negative control. (G) Viral titers, (H) relative viral M mRNA levels, and (I) TNF-α mRNA levels. Data are presented as the means ± SEM of three independent experiments. Significance compared with the control group, \*P < 0.05, \*\*P < 0.01; significance compared with the negative control group, #P < 0.05 and ##P < 0.01.

as indicated by viral titers (**Figure 4A**), M mRNA levels (**Figure 4B**) and NP levels (**Figure 4C**). As expected, AFB<sup>1</sup> at doses of 10–40 µg/kg markedly increased viral titers, M mRNA levels and NP levels in lungs of SIV-infected mice compared with the levels in lungs of mice without AFB1. To further assess the impact of AFB<sup>1</sup> on viral replication, weight gain (**Figure 4D**), the lung index (**Figure 4E**) and histological damage (**Figures 4F,G**) were determined. As expected, SIV-infected mice exhibited decreased weight gain, but enhanced the lung index and inflammatory cell infiltration compared with mice from the blank group, and these changes were aggravated following exposure to 10–40 µg/kg AFB1. In addition, 40 µg/kg AFB<sup>1</sup> had no effects on these parameters in mice from the blank group (**Figures 4D-G**). Taken together, our data suggest that AFB<sup>1</sup> promotes SIV replication and SIV-induced lung damage in mice.

### AFB<sup>1</sup> Promotes TLR4 Expression and the Inflammatory Response in SIV-Infected Mice

To further verify the in vitro results, spleen tissues were taken from SIV-infected mice exposed to AFB<sup>1</sup> to assess TLR4 expressions as indicated by TLR4 protein and mRNA levels. The immunohistochemical assay demonstrated that AFB<sup>1</sup> at doses of 10–40 µg/kg significantly increased TLR4 expression

FIGURE 4 | AFB1 promotes SIV replication and lung damage in mice. Anesthetized mice were infected intranasally with 1000 TCID50 of SIV or PBS on d 1, d 7, and d 14; injected intraperitoneally with various concentrations of AFB1 daily; and sacrificed at 15 days post infection (dpi). (A) Viral titers in the lung homogenates were determined by TCID50 on MDCK cells at 2 and 6 dpi. Data are shown as mean log10 TCID50 per gram of lung for three mice per group. The lung tissues were harvested at 15 dpi to assess viral replication as measured by (B) viral M mRNA and (C) NP levels. (D) Comparison of weight change expressed as a percentage of starting weight. (E) The lung index was calculated as the ratio of lung weight and body weight. (F) Representative images taken from nine mice in six groups as indicated. The areas of hemorrhage are denoted with the blue arrows. (G) Pathological changes in lungs. The mouse lungs were removed at 15 dpi, sectioned and stained with H&E for histological examination. Representative images from nine mice in each group were obtained at 200 × magnification. Data are presented as the means ± SEM of nine mice in each group; Different lowercase letters indicate significant differences (P < 0.05). \*,#P < 0.05, \*\*,##P < 0.01, ns, not significant.

(H-score) in the spleens of SIV-infected mice (**Figure 5A**). As expected, qRT-PCR and western blot assays supported the above results, demonstrating that AFB<sup>1</sup> at doses of 10–40 µg/kg markedly increased TLR4 mRNA (**Figure 5B**) and protein levels (**Figure 5C**) in the SIV-infected mice compared with the levels in mice without AFB1. In addition, the inflammatory response was quantified by the release of TNF-α and IL-10, and the results showed that AFB<sup>1</sup> at doses of 10 to 40 µg/kg markedly increased TNF-α release but significantly decreased IL-10 release in sera (**Figures 5D,E**). Taken together, our data suggest that AFB<sup>1</sup> promotes TLR4 expression and the inflammatory response in SIV-infected mice.

#### TAK242 and TLR4 Knockout Alleviates AFB1-Promoted SIV Replication, Inflammation and Lung Damage in SIV-Infected Mice

To determine the roles TLR4 plays in the promotion of SIV replication by AFB<sup>1</sup> in vivo, the TLR4 inhibitor, TAK242, was used to treat mice (16). The results showed that TLR4 mRNA (**Figure 6A**), viral M mRNA (**Figure 6B**) and NP levels (**Figure 6C**) were markedly reduced in the presence of TAK242 compared with the levels in the no-TAK242 group, suggesting that TLR4 activation is required for the promotion of SIV replication by AFB1. However, no significant differences in weight gain (**Figure 6D**) or the lung index (**Figure 6E**) were observed between the TAK242 and control groups. Histological examination of lungs demonstrated that lung damage was alleviated after TAK242 administration (**Figures 6F–G**). In addition, TAK242 significantly reduced TNF-α content in sera (**Figure 6H**). Taken together, our results indicated that TAK242 alleviated AFB1-promoted SIV replication, inflammation and lung damage in SIV-infected mice.

To further confirm the roles TLR4 plays in the promotion of SIV replication by AFB1, TLR4−/<sup>−</sup> mice were used in this study. The results showed that TLR4−/<sup>−</sup> mice exhibited decreased TLR4 mRNA (**Figure 6I**), viral M mRNA (**Figure 6J**) and NP levels (**Figure 6K**) compared with WT mice, suggesting that TLR4 activation is indeed required for the promotion of SIV replication by AFB1. Likewise, no significant differences in weight gain (**Figure 6L**) and the lung index (**Figure 6M**) were observed between WT and TLR4−/<sup>−</sup> mice. As expected, histological examination of lungs from TLR4−/<sup>−</sup> mice did not reveal obvious lung damage (**Figures 6N–O**). In addition, the TNF-α content in sera of TLR4−/<sup>−</sup> mice was lower than that in sera of WT mice (**Figure 6P**). Taken together, our results indicated that TLR4 knockout attenuated AFB1-promoted SIV replication, inflammation and lung damage in SIV-infected mice.

FIGURE 6 | TLR4 deficiencies alleviate AFB1-promoted SIV replication, inflammation and lung damage in mice. Anesthetized mice were infected intranasally with 1000 TCID50 of SIV on d 1, d 7, and d 14 and were injected intraperitoneally with 40 µg/kg AFB1 daily. Mice from the TAK242 group were injected with 3 mg/kg TAK242 daily, and the mice from control group were given PBS. Mice were sacrificed at 15 dpi. (A) Relative TLR4 mRNA levels, (B) viral M mRNA levels, (C) viral NP levels, (D) body weight and (E) the lung index. (F) Representative images taken from mice in the control and TAK242 groups. The areas of hemorrhage are denoted with the blue arrows. (G) Pathological changes in lungs and (H) serum TNF-<sup>α</sup> in mice. TLR4−/<sup>−</sup> and WT mice were infected intranasally with 1000 TCID<sup>50</sup> of SIV on d 1, d 7, and d 14; injected intraperitoneally with 40 µg/kg AFB1 daily; and sacrificed at 15 dpi. (I) Relative TLR4 mRNA levels, (J) viral M mRNA levels, (K) viral NP levels, (L) body weight, and (M) the lung index. (N) Representative images taken from mice in the TLR4 knockout (TLR4−/−) and wild-type (WT) groups. The areas of hemorrhage are denoted with the blue arrows. (O) Pathological changes in lungs and (P) serum TNF-α in TLR4−/<sup>−</sup> and WT mice. Data are presented as the means ± SEM. Significance compared with control/WT mice, \*\*P < 0.01; ns, not significant.

### DISCUSSION

Swine are one of the species most sensitive to AFB1, and the maximum tolerance level of AFB<sup>1</sup> for pigs is approximately 0.385 mg/kg of feed (33). On the contrary, mice are highly resistant to AFB<sup>1</sup> (TD<sup>50</sup> > 5,400 mg/kg b.w.) (34). According to the World Health Organization, in humans, AFB<sup>1</sup> at concentrations of 30 to 50, 50 to 100, and 100 to 1,000 µg/kg b.w. produces mild, moderate and severe toxicity, respectively. According to the guidelines of the US Food and Drug Administration and the National Food Safety Standard (GB2761-2017, China), the maximum allowable dietary AFB<sup>1</sup> concentrations for humans and animals are 20 and 300 µg/kg, respectively. However, it was previously unknown whether low-dose AFB<sup>1</sup> could cause or exacerbate secondary diseases. Therefore, concentrations of 10, 20, and 40 µg/kg b.w. were used in this study. Our findings confirmed that 40 µg/kg AFB<sup>1</sup> has no effects on the weight gain and lung function of mice, which is consistent with a previous study (34) and suggests that the promotion of SIV replication by AFB<sup>1</sup> is not due to AFB<sup>1</sup> toxicity. In addition, to remove the potential effects of AFB1-induced cytotoxicity on SIV replication, the safe concentrations of AFB<sup>1</sup> were also determined by MTT and LDH assays and DAPI staining for further in vitro experiments.

Since the initial report in 1979 that AFB<sup>1</sup> decreases interferon production by the influenza virus (35), few studies have been performed to determine its effects on SIV replication. Our study shows that AFB<sup>1</sup> promotes SIV replication in vivo and in vitro. First, enhanced viral replication was observed in the MDCK cells, A549 cells and PAMs. Correspondingly, the in vivo results supported the conclusion of the in vitro experiments that AFB<sup>1</sup> promotes SIV replication in mice. In addition, SIV-infected mice exposed to AFB<sup>1</sup> also exhibited decreased weight gain but increased the lung index and lung damage. Our findings are consistent with the outcomes of SIV infection (36–38), suggesting that SIV infection is aggravated by AFB1.

Toll-like receptors (TLRs), which exist in porcine alveolar macrophages and in mice, are associated with the innate immune response (14, 39). Interestingly, viruses can evade the host immune response, thereby enhancing viral replication, when TLR4 is inhibited, but TLR4 antagonists can protect mice from lethal influenza infection (20). Therefore, the role of TLRs in the AFB1-induced promotion of viral replication was examined in our present study. Our data showed that AFB<sup>1</sup> upregulated TLR4, but not other TLRs, in the SIV-infected PAMs. We investigated the underlying mechanism by using TLR4 knockdown and TLR4−/<sup>−</sup> mice. The results showed that TLR4 knockdown and the inhibition of NFκB significantly reduced the AFB1-promoted SIV replication and inflammatory responses in PAMs, and TLR4 deficiencies also attenuated the AFB1-promoted SIV replication, inflammation and lung damage in mice. This may appear counterintuitive at first because the TLR4 pathway is often required for protection against influenza infection (40). Generally, TLR4 plays a critical role in the activation of innate immune responses to defend the body against pathogens. However, an increasing number of studies have shown that the overexpression and/or continuous activation of TLR4 can lead to excessive inflammatory responses or tissue damage in the body (16–18, 41). Our results are the first to suggest that AFB<sup>1</sup> promotes SIV replication and SIVrelated lung damage by activating the TLR4-NFκB pathway. This finding is supported by previous studies demonstrating that TLR4 antagonists or TLR4 knockout can prevent lethal influenza infection (20, 42). Therefore, we infer that AFB<sup>1</sup> might promote TLR4 overexpression and excessive inflammatory responses and reduce tolerance (43), thereby promoting SIV replication.

Previous study indicated that the effects of proinflammatory cytokines were antagonized by anti-inflammatory cytokines such as IL-10 (43). In addition, a delicate balance between proand anti-inflammatory cytokine production is essential for the recovery from and defense against viral infection (44), which has roles in the maintenance of homeostasis and immunity. Accordingly, our data suggested that the inflammatory response was aggravated to defend against SIV infection, and IL-10 decreased and was not enough for the maintenance of homeostasis and immunity, thereby reducing the tolerance and increasing viral replication. On the contrary, excessive inflammatory responses can induce anti-inflammatory responses (19), and M2 macrophage polarization (anti-inflammatory macrophage phenotype) is TLR4 dependent (45). Therefore, it is likely that AFB<sup>1</sup> promotes SIV replication via the TLR4 dependent induction of M2 macrophage polarization, but this possibility needs to be further studied.

In conclusion, our data suggest that AFB<sup>1</sup> promotes SIV replication and SIV-induced lung damage by activating TLR4- NFκB signaling in vitro and in vivo or at least promotes these processes in a TLR4-dependent manner. This finding suggests a new risk of AFB<sup>1</sup> exposure and reveals the vital role of TLR4-induced inflammation in the promotion of SIV replication and lung damage by AFB1, pointing to TLR4 as a potential therapeutic target for preventing lethal influenza infection.

## AUTHOR CONTRIBUTIONS

YS, FG, XC, and KH designed this project. YS, JS, ZL, and DL conducted the experiments. YS, KH, and DL wrote and revised the manuscript. FG, XC, KH, and DL gave helpful advice regarding the project. All authors reviewed the manuscript.

### FUNDING

This work was also supported by Jiangsu Agricultural science and technology independent innovation foundation of China (CX (15) 1067).

### ACKNOWLEDGMENTS

We thank Weiye Chen from the Harbin Veterinary Research Institute, Chinese Academy of Agricultural Sciences (Harbin, China), for kindly providing the H1N1 virus. This work was supported by the National Natural Science Foundation of China (Nos. 31472253 and 31772811).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02297/full#supplementary-material

Figure S1 | Effects of various concentrations of AFB1 on cells. Cells were exposed to various concentrations of AFB1 for 24 h, and then subjected to (A–C) MTT, (D–F) LDH, and (G) DAPI staining assays for the detection of cell viability. A DMSO group was included to remove the effects of DMSO on cell viability, as the AFB1 was dissolved in DMSO. Cells without any AFB1 and DMSO were used as the control group. Cell nuclei were counterstained with DAPI to assess apoptosis, and the apoptotic cells were identified by the condensation and fragmentation of nuclei (yellow arrows). Data are presented as the means ± SEM of three independent experiments. Significance compared with the control group, <sup>∗</sup>P < 0.05 and ∗∗P < 0.01.

### REFERENCES


45. Li X, Wang Z, Zou Y, Lu E, Duan J, Yang H, et al. Pretreatment with lipopolysaccharide attenuates diethylnitrosamine-caused liver injury in mice via TLR4-dependent induction of Kupffer cell M2 polarization. Immunol Res. (2015) 62:137–45. doi: 10.1007/s12026-015-8644-2

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

Copyright © 2018 Sun, Su, Liu, Liu, Gan, Chen and Huang. 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.

# Phosphorylation of TRIM28 Enhances the Expression of IFN-β and Proinflammatory Cytokines During HPAIV Infection of Human Lung Epithelial Cells

Tim Krischuns 1,2, Franziska Günl 1,2, Lea Henschel 1,2, Marco Binder <sup>3</sup> , Joschka Willemsen<sup>3</sup> , Sebastian Schloer 2,4, Ursula Rescher 2,4, Vanessa Gerlt 1,2, Gert Zimmer 5,6 , Carolin Nordhoff 1,2, Stephan Ludwig1,2 and Linda Brunotte1,2 \*

#### Edited by:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany

#### Reviewed by:

David Kong Ann, City of Hope National Medical Center/Beckman Research Institute, United States Kenrie PY Hui, University of Hong Kong, Hong Kong

> \*Correspondence: Linda Brunotte brunotte@uni-muenster.de

#### Specialty section:

This article was submitted to Cytokines and Soluble Mediators in Immunity, a section of the journal Frontiers in Immunology

Received: 31 May 2018 Accepted: 07 September 2018 Published: 28 September 2018

#### Citation:

Krischuns T, Günl F, Henschel L, Binder M, Willemsen J, Schloer S, Rescher U, Gerlt V, Zimmer G, Nordhoff C, Ludwig S and Brunotte L (2018) Phosphorylation of TRIM28 Enhances the Expression of IFN-β and Proinflammatory Cytokines During HPAIV Infection of Human Lung Epithelial Cells. Front. Immunol. 9:2229. doi: 10.3389/fimmu.2018.02229 1 Institute of Virology Muenster, Westfaelische Wilhelms-University Muenster, Muenster, Germany, <sup>2</sup> Cluster of Excellence "Cells in Motion", Westfaelische Wilhelms-University Muenster, Muenster, Germany, <sup>3</sup> Research Group "Dynamics of Early Viral Infection and the Innate Antiviral Response", Division Virus-Associated Carcinogenesis (F170), German Cancer Research Center (DKFZ), Heidelberg, Germany, <sup>4</sup> Center for Molecular Biology of Inflammation, Institute of Medical Biochemistry, Westfaelische Wilhelms-University Muenster, Muenster, Germany, <sup>5</sup> Institute of Virology and Immunology (IVI), Bern, Switzerland, <sup>6</sup> Department of Infectious Diseases and Pathobiology (DIP), Vetsuisse Faculty, University of Bern, Bern, Switzerland

Human infection with highly pathogenic avian influenza viruses (HPAIV) is often associated with severe tissue damage due to hyperinduction of interferons and proinflammatory cytokines. The reasons for this excessive cytokine expression are still incompletely understood, which has hampered the development of efficient immunomodulatory treatment options. The host protein TRIM28 associates to the promoter regions of over 13,000 genes and is recognized as a genomic corepressor and negative immune regulator. TRIM28 corepressor activity is regulated by post-translational modifications, specifically phosphorylation of S473, which modulates binding of TRIM28 to the heterochromatin-binding protein HP1. Here, we identified TRIM28 as a key immune regulator leading to increased IFN-β and proinflammatory cytokine levels during infection with HPAIV. Using influenza A virus strains of the subtype H1N1 as well as HPAIV of subtypes H7N7, H7N9, and H5N1, we could demonstrate that strain-specific phosphorylation of TRIM28 S473 is induced by a signaling cascade constituted of PKR, p38 MAPK, and MSK1 in response to RIG-I independent sensing of viral RNA. Furthermore, using chemical inhibitors as well as knockout cell lines, our results suggest that phosphorylation of S473 facilitates a functional switch leading to increased levels of IFN-β, IL-6, and IL-8. In summary, we have identified TRIM28 as a critical factor controlling excessive expression of type I IFNs as well as proinflammatory cytokines during infection with H5N1, H7N7, and H7N9 HPAIV. In addition, our data indicate a novel mechanism of PKR-mediated IFN-β expression, which could lay the ground for novel treatment options aiming at rebalancing dysregulated immune responses during severe HPAIV infection.

Keywords: influenza, TRIM28, KAP1, TIF1-beta, innate immunity, IFN-β, RIG-I, PKR

## INTRODUCTION

Influenza A viruses (IAV) are the leading cause of annually recurring respiratory infections affecting millions of people worldwide. Infection by seasonal viruses is accompanied by mild to severe symptoms, such as fever, headache and dry cough but immunocompetent patients usually recover within 2– 3 weeks. In contrast, infections with highly pathogenic avian influenza viruses (HPAIV), such as H5N1 often cause severe viral pneumonia as well as multiple organ failure and can lead to death, as exemplified by the "bird flu" outbreak in Hong Kong in 1997 with an overall mortality rate of 33% (1–3). Uncontrolled expression of type I and type II interferons (IFNs) and high levels of proinflammatory cytokines, such as TNF-α, IL-1ß, IL-6, und IL-8 due to a hyperinduction of the innate immune and inflammatory responses are the suspected reasons for HPAIVinduced immunopathology (reviewed in (4)). The underlying molecular mechanisms and signaling pathways, which are responsible for the increased and sustained expression of IFNs and proinflammatory cytokines during HPAIV infection are still not fully understood. However, a virus-induced imbalance of stimulatory and inhibitory factors, which normally regulate the controlled onset and resolution of immune responses, is hypothesized (5).

The innate immune response to IAV is rapidly initiated by pathogen recognition receptors (PRRs), such as RIG-I, which recognize viral RNA in the cytoplasm of infected cells and activate a signal transduction cascade involving the adaptor protein MAVS and the transcription factors IRF3/7. Upon phosphorylation, IRF3/7 dimerize and translocate into the cell nucleus where they bind to the IFN-α/β promotor and facilitate gene transcription. Alternatively, membrane associated tolllike receptors (TLRs) can detect viral glycoproteins or sense viral RNA in endosomal compartments and signal via the adaptor protein MyD88 resulting in the activation of IRF3/5/7 and subsequently in IFN-α/β expression (6, 7). Secreted IFNα/β bind to the interferon-α/β receptor on neighboring cells resulting in STAT1/2 phosphorylation by the receptor-associated Jak/Tyk kinases (8). This mediates the nuclear translocation of STATs and upregulation of the expression of hundreds of interferon-stimulated genes (ISGs), among them antiviral proteins, chemokines and proinflammatory cytokines. This allows the recruitment and activation of immune cells at the site of infection. To resolve the ongoing immune reaction and prevent immunopathology, negative immune regulators, such as the recently identified death-associated protein kinase 1 (DAPK1) (9) and others interfere with further signal transduction and cytokine expression.

Here, we have identified the host factor and transcriptional corepressor Tripartite motif-containing 28 (TRIM28/KAP1/TIF1β) as a critical regulator of IFN-β, IFN-γ and cytokine expression during infection with HPAIV. TRIM28 belongs to the family of TRIM proteins (10) of which most members are involved in the regulation of the immune response to diverse viruses (11, 12). Like most of the TRIM family members, TRIM28 possesses E3 ubiquitin ligase activity located in its N-terminal RBCC-domain. Its C-terminus contains a rather unique arrangement of functional domains including a heterochromatin protein 1 binding domain (HP1 BD), a plant homeodomain (PHD) and a bromodomain (Bromo), which is only shared by the three other TRIM-family members TRIM24/TIF1α, TRIM33/TIF1γ and TRIM66/TIF1δ. All four proteins are known for their function as transcriptional regulators and constitute the TRIM subfamily VI (13–15).

Functionally, TRIM28 is described as a universal genome regulator involved in embryonic and stem cell development, cell cycle regulation, apoptosis, cancer, diverse stress responses and immunity (16–18). Mice lacking TRIM28 die at an early embryonic stage emphasizing its crucial role during embryonic development (19). In addition, TRIM28 facilitates silencing of endogenous retroviruses (20), restricts pro-viral gene activation and suppresses lytic gene expression of Kaposi's sarcomaassociated herpes virus, Murine leukemia virus and human T-cell lymphotropic virus-1 (21–23). It possesses E3 SUMO ligase activity and interacts with diverse transcription factors and other proteins to modulate their activity. These functions of TRIM28 are suspected to be regulated by post-translational modification (PTM) including SUMOylation, phosphorylation and others, which often occur at acceptor sites located in close proximity to the functional domains in the C-terminus (24–26). In contrast to the majority of TRIM proteins, which comprise immune enhancing activities, TRIM28 is associated with immunosuppression (27). The protein was reported to downregulate the activity of several immune-related transcription factors, such as IRF7, IRF5 and IRF1 as well as STAT3 by varying mechanisms (28–30). A role of TRIM28 during IAV replication has not been investigated until today. First evidence for a possible functional relevance derives from a global SUMO-screening demonstrating that TRIM28 is deSUMOylated during IAV infection (31). Nevertheless, this study did not address whether TRIM28 is involved in the immune response to IAV infection.

In the present study, we demonstrate that TRIM28 is phosphorylated at serine 473 (S473), a site known to regulate TRIM28 corepressor activity, during infection of human lung epithelial cells with HPAIV. Furthermore, we establish a link of S473 phosphorylation to elevated IFN-β expression and provide compelling evidence that TRIM28 is a key factor in the development of cytokine overexpression during HPAIV infection. These results could be the starting point for the development of new immunomodulatory strategies targeting TRIM28 post-translational modification to control the expression of type I IFNs as well as proinflammatory cytokines.

#### MATERIAL AND METHODS

#### Cells and Viruses

Human alveolar epithelial cells (A549), African green monkey kidney epithelial cell (Vero), HEK293T, HEK293T-Phoenix and Madin-Darby canine kidney type II cells (MDCK-II) were cultivated in Dulbecco's modified Eagle's Medium (DMEM) (Sigma, Germany) supplemented with 10% fetal bovine serum (Merck, Germany) and 1% Penicillin/Streptomycin (P/S) (Merck, Germany) at 37◦C and 5% CO2. Human Umbilical Vein Endothelial Cells (HUVECs) were isolated from umbilical cords by dispase treatment and cultured on CellBIND <sup>R</sup> dishes (Corning, USA) in HUVEC-medium [50% EGM2 and 50% M199 (Biochrom, Germany) supplemented with 10% fetal calf serum (Sigma, Germany), 30µg/mL gentamycin (Cytogen, Germany), 15 ng/mL amphotericin B (Biochrom, Germany), 100 IE Heparin (Ratiopharm, Germany), 2 mM L-glutamine (Lonza, Switzerland)] at 5% CO<sup>2</sup> and 37◦C. Upon infection HUVECs were cultured in M199 medium containing 1% BSA, 30µg/ml gentamicin and 15 ng/ml amphotericin B. All work with HUVECs was conducted with the formal approval of the Ethics Committee of North Rhine-Westphalia and the University of Muenster. A/Thailand/KAN-1/2004 (H5N1) (KAN-1) was kindly provided by P. Puthavathana (Bangkok, Thailand). A/FPV/Bratislava/79 (H7N7) (FPV) was obtained from the virus depository of the Institute of Virology in Giessen, Germany. A/Hamburg/04/2009 (H1N1pdm) was a kind gift of the German National Reference Centre for Influenza (Brunhilde Schweiger, Berlin, Germany). A/Vietnam/1203/2004 (H5N1) (VN) and A/Anhui/1/2013 (H7N9) (Anhui) were kindly offered by Thorsten Wolff (RKI, Berlin). Recombinant A/Puerto Rico/8/34 (H1N1) (PR8), A/seal/Mass/1-SC35M/80 (H7N7) (SC35M) and A/WSN/33 (H1N1) (WSN) were generated using the pHW2000 reverse genetics system (32). All influenza viruses were propagated on MDCK-II cells in infection medium [DMEM supplemented with 1% P/S, 0.25% bovine serum albumin (BSA, Sigma) and 0.01% MgCl<sup>2</sup> and CaCl<sup>2</sup> (Roth, Germany)]. Infections were carried out by incubating cells in infection PBS (PBS supplemented with 1% P/S, 0.25% BSA and 0.01% MgCl<sup>2</sup> and CaCl2) at the indicated multiplicity of infection (MOI) for 30 min. Experiments involving HPAIV were conducted in a biosafety level (BSL) 3 approved laboratory. Recombinant VSV (serotype Indiana) encoding firefly luciferase (VSV-luc) was generated by replacing the GFP gene in the previously described VSV-GFP vector by the firefly luciferase gene according to published procedures (33). VSV-luc was propagated on Vero cells and titrated by immunostaining with a rabbit polyclonal anti-VSV serum as described previously (34).

#### Plasmids

Guide RNAs (gRNA) targeting TRIM28 and mCherry were designed with BbsI overhang sequences. The gRNA oligonucleotides were ordered phosphorylated, annealed and ligated into BbsI digested pSpCas9(BB)-2A-GFP plasmid (Addgene #48138) (35). For MyD88 and PKR, gRNAs were annealed, phosphorylated by PNK and cloned into BsmBI digested lentiCRISPR v2 vector (Addgene #52961) (36). Oligonucleotide sequences are included in **Supplementary Table S1**. Full-length human TRIM28 was subcloned from pEGFP-TRIM28 (Addgene #45568) into NotI and XhoI (NEB, USA) digested pBluescript II SK(+/–) vector. In pBluescript II SK(+/–), TRIM28 mutants (S473A, S473E) were obtained by site-directed mutagenesis with non-overlapping primers. Subsequently, TRIM28 wildtype and the phosphomutants were cloned into NotI and EcoRI (NEB, USA) digested retroviral vector pQCXIP. PCR primer sequences are included in **Supplementary Table S2**.

## Generation of Knockout Cells

A549 TRIM28 CRISPR-Cas9 knockout (KO) and control cells (Ctrl) were generated by transient transfection. In brief, A549 cells were transfected with either pSpCas9(BB)-2A-GFP harboring a gRNA targeting TRIM28 or a control gRNA targeting mCherry (plasmids were kindly provided by Nicole Fischer, Hamburg, Germany). Positively transfected cells were selected by fluorescence-activated cell sorting (FACS) and clonal cell lines were analyzed for TRIM28 KO by western blot. A549 PKR, RIG-I KO, MAVS KO and MyD88 KO cells were generated by lentiviral transduction as described elsewhere (9, 37). In brief, lentiviral particles were produced on HEK293T cells by transfection with the following three plasmids at a 3:1:3 ratio; (i) pCMV-DR8.91 (ii) pMD2.G (iii) lenti-CRISPR-vector. Virus particle-containing supernatants were harvested 48, 56, and 72 h post-transfection (h p.t.) and used for transduction of target cells. Successfully transduced cells were selected with 1µg/ml puromycin (Sigma, Germany). Gene knockout in single cell clones was validated by western blot.

### Retroviral Gene Transfer

The empty retroviral vector pQCXIP or pQCXIP-TRIM28 expressing the different phospho-mutants were transfected into HEK293T-Phoenix packaging cells (Orbigen, USA). Retroviruscontaining supernatants were harvested 48 and 60 h p.t., supplemented with polybrene (Santa Cruz Biotechnology, USA) to a final concentration of 4µg/ml and used for transduction of A549 TRIM28 KO cells. Transduced cells were selected with 1µg/ml puromycin for 5 days to obtain stable cell lines and TRIM28 expression levels were analyzed by western blot. Stable TRIM28 mutant-expressing cells were subcloned to obtain single cell clones with equal expression of TRIM28 as measured by western blot.

### Cell Treatments

Cells were treated with inhibitors for 1 h prior to infection, RNA transfection or induction of genotoxic stress. After removal of the inoculum, transfection mix or chemicals, inhibitors were added to the infection medium. The following inhibitors were used: ATM (KU-60019, Selleckchem, Germany), Chk2 (Chk2 inhibitor II, Abcam, Germany), p38 MAPK (SB202190, Calbiochem, USA), PKR (2-Aminopurine, Sigma, Germany), MEK (U0126, Taros Chemicals, Germany), MK2 (PF-3644022, Sigma, Germany), MSK1 (SB747651A, Axon Medchem, Netherlands) and the ROS-scavenging agent N-Acetyl-L-cysteine (NAC) (Sigma, Germany). Cells were stimulated by exposure to 1 kJ/m<sup>2</sup> UVC-light using a Stratalinker 2400 UV Crosslinker (BioSurplus, USA), or incubation with H2O2, etoposide (Sigma, Germany) or IFN-β (R&D Systems, Germany) for the indicated times and concentrations. For RNA and HMW poly(I:C) (Invivogen, USA) stimulations, A549 cells were transfected using Lipofectamine 2000TM (Invitrogen, USA) according to the manufacturer's instructions. Therefore, total RNA from MDCK-II cells either infected with WSN at an MOI of 5 for 8 h or non-infected cells was isolated using the RNeasy KitTM according to the manufacturer's instructions (Qiagen, Germany).

#### MTT-Assay

For cytotoxicity measurements, MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] (Sigma, Germany) was added to the cells at a final concentration of 5 mg/ml for 4 h at 37◦C and 5% CO2. As a positive control, 2µM staurosporine (Sigma, Germany) was added for 10 h. Supernatants were aspirated and DMSO (Roth, Germany) was added for 5 min before the optical density (OD) was measured at a wavelength of 562 nm (MicroLumat Plus LB96V, Berthold Technologies, Germany).

#### Western Blot and Antibodies

Cells were lysed with ice cold radioimmunoprecipitation assay (RIPA) buffer (25 mM TRIS pH 7.5, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% Triton X-100) supplemented with the following protease and phosphatase inhibitors; 10µM leupeptin (Sigma, Germany), 200 nM aprotinin (Roth, Germany), 5 mM benzamidine (Sigma, Germany), 2.5 mM pefabloc (Sigma, Germany), 10 mM beta-glycerophosphate (Sigma, Germany), 1 mM sodium orthovanadate (Sigma, Germany), 10 mM sodium fluoride (Roth, Germany) and 2.5 mM sodium pyrophosphate (Sigma, Germany). Lysates were sonicated for 20 s (pulse 50%, amplitude 30%) and pelleted at 4◦C, 14.000 g for 15 min. Protein amounts were adjusted to 20 µg, mixed with 4× sample buffer (0.25 M TRIS pH 6.8, 40% glycerol, 8% SDS, 10% β-mercaptoethanol, 0.01% bromophenol blue) and separated by SDS-PAGE. Proteins were transferred to nitrocellulose membranes and detected by using primary antibodies targeting tubulin (Sigma, Germany), PB1 (GeneTex, USA), TRIM28, TRIM28 S473-P, TRIM28 S824-P (Abcam, UK), CREB S133-P, RIG-I, eIF2α S51-P, ERK1/2, ERK1/2 T202/Y204-P, HSP27, HSP27 S82- P (Cell signaling Technologies, USA) and anti-mouse or anti-rabbit IgG secondary antibodies either conjugated to fluorophores (Licor, Germany) or horseradish peroxidase (Cell Signaling Technology, USA). Selected bands were densitometrically quantified using Licor Image studio software (Licor, Germany).

#### Immunofluorescence

A549 cells were seeded on glass coverslips and fixed with 4% paraformaldehyde (Sigma, USA). Cells were permeabilized with 0.1% Triton X-100, and blocked for 30 min in 3% BSA. Slides were incubated overnight at 4◦C with primary antibodies against IAV nucleoprotein (NP) (GeneTex, USA) and TRIM28 S473-P. Secondary antibodies anti-rabbit Alexa Fluor 488 (Invitrogen, USA) and anti-mouse Alexa Fluor 568 (Invitrogen, USA) were incubated for 1 h at room temperature. Cell nuclei were stained for 20 min with DAPI (Thermo Fisher Scientific, USA). Coverslips were mounted on glass slides in Mounting Medium S3023 (Dako Omnis, USA) and examined using a LSM-800 Airyscan confocal microscope (Carl Zeiss, Germany).

#### Phosphoproteomic Screen

A549 cells were stably labeled with "light" lysine (12C6, <sup>14</sup>N2) and arginine (12C6, <sup>14</sup>N4), "medium" lysine (13C6, <sup>14</sup>N2) and arginine (13C6, <sup>14</sup>N2) or "heavy" lysine (13C6, <sup>15</sup>N2) and arginine ( <sup>13</sup>C6, <sup>15</sup>N4). Labeled cells were infected with either PR8, FPV or KAN-1 at an MOI of 5 for 2, 4, 6, and 8 h. "Light" labeled cells were used as non-infected control (0 h), whereas "medium"- and "heavy"-labeled cells were infected for 2 or 6 h and 4 or 8 h, respectively. Lysates from non-infected, 2, 4 h infected cells (Mix 1) and non-infected, 6, 8 h infected cells (Mix 2) were subjected to tryptic digestion. Phosphopeptides were purified by cation exchange chromatography and TiO2 enrichment followed by LC-MS/MS analysis on a Proxeon Easy-nLC coupled to an LTQ-Orbitrap XL mass spectrometer. Data analysis was performed using Mascot and MaxQuant (v1.2.2.9) as previously described (38–40). Phosphorylation intensities of TRIM28 residues were quantified in relation to the phosphorylation of TRIM28 in non-infected cells in both lysate mixtures.

### Cytokine Analysis

A549 TRIM28 KO and Ctrl cells were stimulated by transfection with 200 ng of viral or cellular RNA. The LEGENDplexTM Human Anti-Virus Response Panel (BioLegend Cat. No. 740350) was used for the simultaneous determination of the concentrations of IFN-α, -β, -γ, -λ1 and λ2/3 as well as IL-1β, IL-6, IL-8, IL-10, IL-12p70, TNF-α, IP-10, and GM-CSF in the supernatant. Cytokine capturing was performed according to the manufacturer's protocol in filter plates. Bead-bound cytokines were measured on a FACSCalibur Cytometer (Becton Dickinson) and concentrations were calculated using the LEGENDplexTM Data Analysis Software (BioLegend, USA).

#### RNA Isolation and Quantitative Real-Time PCR (qRT-PCR)

RNA was isolated using peqGOLD TriFastTM according to the manufacturer's instructions (VWR, USA). Total RNA was reverse transcribed with oligo(dT) primers and RevertAid H Minus Reverse Transcriptase (Thermo Fisher Scientific, USA). RT-PCR was carried out in duplicates using a LightCycler <sup>R</sup> 480 II (Roche, Germany). Primer sequences are provided in **Supplementary Table S3**. Commercially available primers were used for analysis of IFN-β mRNA (Qiagen, Germany). Expression data were normalized to the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GADPH) and analyzed using the 2 <sup>−</sup>11CT method as described elsewhere (41).

#### IFN-Bioassay

A549 TRIM28 KO and Ctrl cells were stimulated by transfection of 250 ng of viral or cellular RNA and at 6 h p. t. supernatants were harvested. The cell-free supernatants were diluted 1:10 and added to Vero cells for another 16 h. Subsequently, Vero cells were infected with VSV-luc at an MOI of 5 for 5 h. Supernatants were aspirated, cells were lysed in passive lysis buffer (Promega, USA) and luciferase assay substrate (Promega, USA) was added. VSV-luc reporter gene expression was determined by measuring luminescence using a MicroLumat Plus LB96V luminometer (Berthold Technologies, Germany).

### RESULTS

### Phosphorylation of TRIM28 Is Induced by HPAIV Infection

Viruses activate diverse signaling pathways in infected cells. To elucidate whether human adapted and highly pathogenic avian-derived IAV strains differentially activate kinase-governed signaling pathways a quantitative phosphoproteomic screen was performed (40). Human lung epithelial cells (A549) were infected with the human IAV strain A/Puerto Rico/8/34 (PR8, H1N1), the HPAIV strain A/Thailand/KAN-1/2004 (KAN-1, H5N1), which was isolated from a fatal human case following direct avian-to-human transmission and the HPAIV avian isolate A/FPV/Bratislava/79 (FPV, H7N7). This revealed that the host factor TRIM28 was increasingly phosphorylated at S473 during infection with KAN-1 and FPV but not with PR8 (**Figure 1A**, upper panel). For the neighboring serine 471 (S471), increased phosphorylation was only detected during FPV infection (**Figure 1A**, lower panel). These results were confirmed by western blot analysis using an antibody specific for phosphorylated TRIM28 S473 (**Figure 1B**). Based on these data, we speculated that TRIM28 phosphorylation could be a straindependent mechanism. To support this hypothesis, additional IAV strains were tested. We observed that TRIM28 S473 was also phosphorylated upon infection with the mouse-adapted HPAIV variant A/seal/Mass/1-SC35M/80 (SC35M, H7N7) and the HPAIV strains A/Vietnam/1203/2004 (VN, H5N1), A/Anhui/1/2013 (Anhui, H7N9) but not with the humanadapted 2009 pandemic H1N1 strain A/Hamburg/04/2009 (H1N1pdm) (**Figure 1C** upper panels). Quantitative western blot analysis further demonstrated that SC35M, KAN-1, and FPV induced S473 phosphorylation to different degrees, suggesting that all three strains have individual capacities to induce S473 phosphorylation (**Figures 1B,C**, lower panels). Plotting the virus strains according to the intensity of the induced S473-P signals indeed suggests that the degree of human adaptation inversely correlates with the capacity to induce S473 phosphorylation (**Figure 1D**). Like H5N1 viruses, H7N7 viruses can cross the species barrier from birds to humans and may cause severe to lethal respiratory disease in humans (42–44). As we observed S473 phosphorylation during infection with the mouse-adapted HPAIV variant SC35M, we used this strain as a representative for HPAIV in many experiments. This had the advantage that we could perform the experiments under BSL2 conditions. Interestingly, phosphorylation at S473 and S471 could be detected at 6 h p.i in the phosphoproteomic screen as well as in western blot analysis, indicating that it is not induced at an early stage of viral infection like viral entry or nuclear replication but rather at a later step. S473 phosphorylation was also observed at a low MOI of 0.1 (**Supplementary Figure S1A**). In addition, strain-dependent phosphorylation was also observed in primary HUVECs (**Supplementary Figure S1B**). Immunofluorescence data showed that the occurrence of nuclear S473 phosphorylation correlates with the cytoplasmic localization of the viral nucleoprotein (NP) 10 h after infection. In contrast, in cells infected for 5 h, only background phosphorylation was observed in the nucleus (**Figures 1E,F**). In summary, these results demonstrate that HPAIV of the H5N1, H7N7, and H7N9 subtypes induce phosphorylation of TRIM28 S473 at a late time point during infection. Furthermore, our data indicate that the capacity of IAV strains to phosphorylate TRIM28 inversely correlates with the degree of human adaptation.

### HPAIV-Induced Phosphorylation of TRIM28 Is Mediated by a Signaling Pathway not Related to the DNA Damage Response (DDR)

Phosphorylation of TRIM28 at positions S473 and serine 824 (S824) is widely described to occur in response to DNA damage and can be experimentally induced by various genotoxic stresses including treatment with H2O2, UV-radiation and etoposide (45–47). During DNA damage, phosphorylation at these sites is mediated by the kinase ataxia-telangiectasia mutated (ATM) and the checkpoint kinases 1 und 2 (Chk1/2) (48, 49) (**Figure 2A**). Phosphorylation is associated with different biological outcomes. While S473 is located in close proximity to the HP1 BD, which mediates the interaction with HP1 and repression of Krüppel-associated box zinc finger protein (KRAB-ZNF) dependent genes, S824 lies next to the C-terminal bromodomain. Functionally, phosphorylation of S473 has been demonstrated to ablate binding of TRIM28 to HP1 and TRIM28-mediated repression of KRAB-ZNF-dependent genes. In contrast, S824 phosphorylation facilitates local chromatin relaxation (48) and, in combination with TRIM28 deSUMOylation, leads to the de-repression of DDR-responsive genes (25). Because infection with IAV has been reported to induce DDR (50, 51), we examined whether infection with SC35M induces the same phosphorylation pattern on TRIM28 compared to UV-radiation, H2O<sup>2</sup> or etoposide treatment. Remarkably, we found that the induced phosphorylation patterns during IAV infection and DNA damage are different. Infection with SC35M induced phosphorylation of S473 but not S824 while all three genotoxic agents readily induced phosphorylation at both sites (**Figure 2B**). We further investigated whether ATM and Chk2 are also the responsible kinases for TRIM28 phosphorylation during IAV infection. Treatment of A549 cells with non-toxic concentrations of the inhibitors for ATM and Chk2 prior to stimulation with H2O2, etoposide or infection with SC35M clearly demonstrated that these kinases are not involved in IAV-mediated TRIM28 S473 phosphorylation (**Figure 2C**; **Supplementary Figures S2A–D**). Using NAC to scavenge reactive oxygen species (ROS) we could also exclude ROS as cause of TRIM28 S473 phosphorylation during SC35M infection (**Figure 2D**). In summary, these results demonstrate that during IAV infection TRIM28 S473 is not phosphorylated by the DDR-related kinases ATM and Chk2, which suggests that

FIGURE 1 | Phosphorylation of TRIM28 during HPAIV infection. (A) SILAC-labeled human A549 cells were infected with FPV (H7N7), KAN-1 (H5N1), and PR8 (H1N1) for the indicated times at an MOI of 5. Phosphorylated peptides were enriched and analyzed by mass spectrometry. The relative phosphorylation of TRIM28 at serine 473 (S473) and serine 471 (S471) are depicted. Western blot of A549 cells infected with (B) FPV, KAN-1, PR8 and (C) VN (H5N1), Anhui (H7N9), SC35M (H7N7), H1N1pdm at an MOI of 5 for the indicated time points. Phosphorylation of TRIM28 S473 was detected using a phospho-specific antibody. Detection of total TRIM28 and tubulin served as loading controls. Densitometric quantifications of S473 phosphorylation were normalized to tubulin intensities and are depicted as mean fold change (±SEM). (D) Schematic representation of mean fold changes of TRIM28 S473 phosphorylation compared to non-infected cells at 6 h p.i. A549 cells were infected with SC35M (E) at an MOI of 5 for 5 h or (F) at an MOI of 0.5 for 10 h. Nuclei were stained with DAPI. Viral nucleoprotein (NP) and TRIM28 S473 phosphorylation were stained using specific antibodies. Cells were analyzed by confocal laser scanning microscopy.

FIGURE 2 | Phosphorylation of TRIM28 S473 occurs by non-DDR related kinases during HPAIV infection. (A) Schematic representation of the involved kinases leading to TRIM28 S473 and S824 phosphorylation during DNA damage response. (B) A549 cells were treated with 250µM H2O2 for 3 h, 100µM etoposide for 6 h, infected for 10 h with SC35M at an MOI of 20 or treated with 1 kJ/m<sup>2</sup> ultraviolet light 30 min prior to western blot analysis. TRIM28 phosphorylation was detected using antibodies specific for TRIM28 phosphorylated at S473 or S824, respectively. Total TRIM28 and tubulin served as loading controls. Infection was validated using an antibody targeting the viral protein PB1. (C) A549 cells were treated with inhibitors for ATM (KU-60019, 2.5µM), Chk2 (Chk2 Inhibitor II, 10µM) and the solvent control (DMSO, 10µM) 1 h prior to stimulation with H2O2, etoposide and SC35M infection in the presence of the corresponding inhibitors. (D) A549 cells were treated with increasing amounts of the reactive oxygen species scavenger NAC (5, 10, 15 mM) for 1 h following stimulation with 500µM H2O2 for 2 h and infection with SC35M for 10 h at an MOI of 20 in the presence of NAC. TRIM28 S473 phosphorylation was monitored by western blot.

TRIM28 has a yet non-described, non-DDR related function in IAV infected cells.

### TRIM28 Is a Negative Regulator of the Innate Immune Response to IAV

To gain insight into the general function of TRIM28 during viral infection, TRIM28 KO cells were generated using CRISPR-Cas9 (**Figure 3A**). Growth curve analyses demonstrated no pronounced effect on viral replication of SC35M and FPV in cells lacking TRIM28 compared to control cells (**Supplementary Figures S3A,B**). Because TRIM28 is described as a negative immune regulator, we analyzed the expression of IFN-β in these cells. Intriguingly, infection with PR8, SC35M or FPV resulted in elevated levels of IFN-β compared to infected control cells (**Figure 3B**). Elevated levels of IFN-β as well as the proinflammatory cytokines IL-6 and IL-8 were also observed during infection of TRIM28 KO cells with KAN-1 (**Figure 3C**). In addition, transfection of viral RNA (vRNA), as a trigger for the innate immune response, also resulted in higher mRNA levels of IFN-β, IL-6 and IL-8 in the absence of TRIM28 (**Figure 3D**). Importantly, we could also demonstrate that transcriptional upregulation correlated with significantly increased secretion of IFN-β, IL-6, IL-8 and IFN-γ in vRNA-treated TRIM28 KO cells at 8 and 24 h p.t. (**Figure 3E**; **Supplementary Figures S4A,B**). Because we did not observe an effect on SC35M and FPV replication, the biological function of increased IFN levels was manifested in an IFNbioassay using a luciferase-expressing vesicular stomatitis virus (VSV-luc), which is highly sensitive to the action of IFNs. To induce an antiviral state, Vero cells were pre-treated with the supernatants from vRNA-stimulated TRIM28 KO and control cells. Infection with VSV-luc revealed a pronounced inhibition of viral replication in Vero cells that have been treated with the supernatant from stimulated TRIM28 KO cells compared to Vero cells treated with control cell supernatant, indicating that the increased IFN levels induced a more potent antiviral state (**Figure 3F**). In summary, these results demonstrate that TRIM28 functions as an important negative regulator of the expression of IFN-β, IFN-γ, IL-6 and IL-8 during IAV infection.

### Phosphorylation of TRIM28 S473 Occurs in Response to Viral RNA but Is Independent of RIG-I

The previous results demonstrated that TRIM28 negatively regulates the expression of IFN-β, IFN-γ, IL-6 and IL-8 during IAV infection. However, the role and biological function of S473 phosphorylation and the source of activation remained elusive. During IAV infection IFN-β is majorly expressed in response to sensing of viral RNA by cytosolic RIG-I (52, 53).

FIGURE 3 | TRIM28 is a negative regulator of the innate immune response during influenza A virus infection. (A) Western blot of A549 TRIM28 KO cells. (B) Infection of TRIM28 KO cells with PR8, SC35M and FPV at an MOI of 5. Total RNA was isolated at the indicated time points and IFN-β mRNA levels were analyzed by qRT-PCR. IFN-β levels are depicted as mean n-fold change (±SEM) compared to non-infected cells. \*\*p ≤ 0.002; \*\*\*p ≤ 0.0002; two-way ANOVA; Sidak's multiple comparisons test. (C) TRIM28 KO cells were infected with KAN-1 (H5N1) at an MOI of 0.01 for 24 h. IFN-β, IL-6 and IL-8 mRNA levels were determined as described in (B). \*p ≤ 0.03; \*\*\*\*p ≤ 0.0001; two-tailed unpaired t-test. (D) TRIM28 KO cells were transfected with 200 ng viral or cellular RNA. Total RNA was isolated 24 h p.t. and IFN-β, IL-6 and IL-8 mRNA levels were determined by qRT-PCR. Results are shown as mean n-fold change (±SEM) over non-treated cells. \*\*p ≤ 0.0021; \*\*\*p ≤ 0.0002; two-way ANOVA; Sidak's multiple comparisons test. (E) TRIM28 KO cells were treated as in (D). Supernatants were analyzed using LEGENDplexTM bead immunoassay for the indicated cytokines at 24 h p.t. Results of six independent experiments are plotted as well as the mean (±SD). \*p ≤ 0.03; \*\*\*\*p ≤ 0.0001; two-way ANOVA; Tukey's multiple comparisons test. (F) A549 cells were transfected with 250 ng viral or cellular RNA for 6 h. Supernatants were harvested and transferred to Vero cells for 16 h. Stimulated Vero cells were infected with a luciferase-encoding vesicular stomatitis virus (VSV-luc) at an MOI of 5 for 5 h. Cells were harvested and virus replication was determined by luciferase assay. Results of four independent experiments are plotted as well as the mean (±SEM).

Because our results demonstrate that TRIM28 is also involved in the expression of IFN-β, we speculated that TRIM28 S473 phosphorylation could be induced by a similar mechanism. Therefore, we analyzed whether transfection of vRNA induces S473 phosphorylation. We observed that TRIM28 S473 was markedly phosphorylated following transfection of vRNA or poly(I:C) (**Figure 4A**; **Supplementary Figure S5A**). Importantly, using RIG-I knockout cells (RIG-I KO) (**Figure 4B**), we could demonstrate that S473 phosphorylation during vRNA transfection (**Figure 4C**) and SC35M infection (**Figures 4D,E**) is retained in the absence of RIG-I. This provides evidence that S473 phosphorylation occurs independent of the RIG-I signaling pathway. To support the idea that RIG-I independent mechanisms contribute to the expression of IFN-β during infection with HPAIV, we infected wildtype and RIG-I KO cells with PR8, SC35M as well as FPV and measured the induction of IFN-β. This revealed that IFN-β expression was rather low in PR8 infected wildtype cells and seems to primarily depend on RIG-I as 6.2-fold less IFN-β was induced in the absence of RIG-I. In contrast, IFN-β was upregulated by 25-fold and 75-fold in SC35M and FPV infected wildtype cells, respectively. However, lack of RIG-I reduced IFN-β induction only by 2-fold in SC35M infected cells and 1.5-fold in FPV infected cells (**Figure 4F**). This indicates that the expression of IFN-β during FPV infection is not exclusively dependent on RIG-I but involves other signaling pathways.

with PR8, SC35M or FPV at an MOI of 5. Total RNA was isolated 8 h p.i. and IFN-β mRNA levels were measured by qRT-PCR. IFN-β levels are depicted as mean

n-fold change (±SEM) over mock-infected cells.

These data suggest that alternative RNA sensing receptors are responsible for the induction of S473 phosphorylation and contribute to the high levels of IFN-β during infection with HPAIV.

#### Detection of Viral RNA by the Cytoplasmic RNA Sensor PKR Induces TRIM28 S473 Phosphorylation

To further specify which immune recognition pathway comes into consideration for S473 phosphorylation and modulation of IFN-β expression during HPAIV infection, A549 cells lacking the adaptor proteins MAVS and MyD88 were examined. Infection with SC35M clearly demonstrated that TRIM28 S473 was still phosphorylated in cells lacking the RIG-I downstream effector MAVS, which supported the previous results obtained in RIG-I KO cells (**Figure 5A**, lane 7). Of note, RIG-I could not be detected in this western blot due to low induction by SC35M infection. However, RIG-I knockout in these cells was demonstrated following IFN-β treatment in **Figure 4B**. S473 phosphorylation was also retained despite lack of MyD88, which rules out the majority of TLRs as candidate receptors for mediating TRIM28 S473 phosphorylation (**Figure 5A**, lane 8). Another protein that is described to have RNA sensing capacity is the doublestranded RNA sensing protein kinase R (PKR), which also binds to double-stranded RNAs in the cytosol (54). Interestingly, inhibition of PKR using 2-Aminopurine (2-AP) impeded S473 phosphorylation in response to viral infection in a concentration dependent manner (**Figure 5B**) and following vRNA transfection (**Figure 5D**). Furthermore, **Figure 5C** shows that PKR inhibition also resulted in decreased levels of IFN-β, IL-6, and IL-8 during infection. As a genetic approach, A549 cells lacking PKR (PKR KO) were generated. Intriguingly, in these cells S473 phosphorylation after vRNA transfection was strongly reduced (**Figure 5E**). Infecting PKR KO cells with PR8, SC35M, and FPV revealed that the induction of IFN-β is differentially dependent on PKR. Although all three viruses induce less IFN-β in PKR KO cells, we observed a clear tendency that IFN-β induction is more dependent on PKR during infection with SC35M and FPV compared to PR8 (**Figure 5F**). This fits to our hypothesis that IFN-β induction in HPAIV but not PR8 infected cells is potentiated by a PKR activated signaling cascade. In summary, these results demonstrate that viral RNA sensing by PKR leads to TRIM28 S473 phosphorylation during HPAIV infection and presumably contributes to the high IFN-β levels.

### p38 MAPK and MSK1 Phosphorylate TRIM28 S473 During HPAIV Infection

In order to elucidate the signaling cascade responsible for TRIM28 S473 phosphorylation during viral infection, we concentrated further on kinases which are reported to be involved in the expression of IFN-β and proinflammatory cytokines during HPAIV infection and are known to be activated by PKR (55–57). This led us to investigate the stress inducible mitogen-activated protein kinase (MAPK) p38. Treatment of A549 cells with the p38 inhibitor SB202190 at specific and non-toxic concentrations efficiently blocked TRIM28 S473 phosphorylation during SC35M infection (**Figure 6A**; **Supplementary Figure S6A**) demonstrating that p38 plays a major role in this process. In contrast, treating cells with an inhibitor of MEK, thus blocking the ERK MAPK pathway, did

qRT-PCR. Results are depicted as mean n-fold change (±SEM) over non-infected cells. \*p ≤ 0.03; \*\*p ≤ 0.002; two-tailed unpaired t-test. (D) A549 cells pre-treated with 10 mM 2-AP for 1 h were transfected with 500 ng viral or cellular RNA. Lysates were harvested 4 h p.t. and TRIM28 S473 phosphorylation was analyzed by western blot. PKR inhibition was controlled by detection of eIF2α S51 phosphorylation. (E) A549 PKR KO cells were transfected with 50, 250, and 500 ng viral RNA or 500 ng cellular RNA. Lysates were harvested 8 h p.t. and TRIM28 S473 phosphorylation was analyzed by western blot. (F) Infection of PKR KO cells with PR8, SC35M or FPV at an MOI of 5. Total RNA was isolated 8 h p.i. and IFN-β mRNA levels were measured by qRT-PCR. IFN-β levels are depicted as mean n-fold change (±SEM) over non-infected cells.

not reduce S473 phosphorylation, excluding crosstalk from the classical MEK1/2-ERK1/2 MAP kinase pathway (**Figure 6B**; **Supplementary Figure S6B**). Well-described downstream kinases of p38 MAPK are MSK1 and MK2, which are both reported to be involved in the transcriptional regulation of cytokine expression (58, 59). Chemical inhibition of MSK1 but not MK2 resulted in the loss of S473 phosphorylation (**Figures 6C,D**; **Supplementary Figures S6C,D**). Importantly, inhibition of p38 MAPK and MSK1 led to reduced TRIM28 S473 phosphorylation during infection with the HPAIV KAN-1 and Anhui in primary HUVECs (**Supplementary Figures S6F,G**). This led us to conclude that MSK1 is the responsible kinase for S473 phosphorylation during IAV infection. Induction of S473 phosphorylation by transfection of vRNA was similarly abolished by inhibition of p38 and MSK1 but not by inhibiting MEK and MK2 (**Figure 6E**). Most importantly, loss of TRIM28 S473 phosphorylation by inhibition of p38 and MSK1 also resulted in decreased levels of IFN-β, IL-6, and IL-8 during infection with SC35M (**Figure 7A**), which was not caused by an inhibition of viral replication (**Figure 7B**). In conclusion, these data provide compelling evidence that TRIM28 S473 phosphorylation in response to PKR-dependent sensing of vRNA is mediated by the p38/MSK1-cascade during infection with HPAIV. Furthermore, these results strongly indicate that TRIM28 S473 phosphorylation results in enhanced expression of IFN-β and proinflammatory cytokines.

#### Constitutive Phosphorylation of TRIM28 S473 Leads to Increased Induction of IFN-β, IL-6 and IL-8 During HPAIV Infection

To establish the functional link between TRIM28 S473 phosphorylation and IFN-β expression, we reconstituted TRIM28 KO cells with either wildtype TRIM28 or the phosphomutants S473A and S473E. Infection of TRIM28 KO cells with VSV-luc resulted in decreased viral replication. Most importantly, reconstitution of TRIM28 KO cells with the wildtype protein rescued VSV-luc replication (**Figure 8A**). Substitution of S473 with alanine (S473A) eliminates the phospho-acceptor site, while substitution with glutamic acid (S473E) mimics constitutive phosphorylation. As our previous data suggested that S473 phosphorylation regulates TRIM28 mediated repression of IFN-β, expression of these mutants should affect VSV-luc replication. Indeed, infection with VSVluc demonstrated that reconstitution with TRIM28 S473E resulted in significantly decreased viral replication compared to cell expressing wildtype TRIM28 and TRIM28 S473A (**Figure 8B**). To proof that expression levels of IFN-β, IL-6, and IL-8 are also increased in the TRIM28 S473E expressing cells we infected the reconstituted cells with KAN-1 and performed qRT-PCR analysis. As seen in **Figure 8C**, the infected S473E expressing cells express higher levels of IFN-β, IL-6, and IL-8 compared to cells reconstituted with wildtype TRIM28 and TRIM28 S473A. Of note, also the non-phosphorylated form of TRIM28 harboring S473A showed increased levels of IFN-β,

FIGURE 6 | Inhibition of p38 and MSK1 ablates TRIM28 S473 phosphorylation. A549 cells were pre-treated with inhibitors for (A) p38 (SB202190), (B) MEK (U0126), (C) MSK1 (SB747651), and (D) MK2 (PF-3644022) at 0.1 µM, 0.5µM, 1µM, 5µM and 10µM before infection with SC35M at an MOI of 20 for 10 h and further incubation with the inhibitors. Lysates were examined with a phospho-specific antibody for TRIM28 S473 phosphorylation. Infection was verified by detection of the viral PB1 protein. Tubulin and total TRIM28 served as loading controls. MEK and MK2 inhibition was controlled by detection of ERK T202/Y204, and HSP27 S82 phosphorylation, respectively. (E) A549 cells were pre-treated with p38 (SB202190), MEK (U0126), MSK1 (SB747651) and MK2 (PF-3644022) inhibitors at 1µM for 30 min. Pre-treated cells were transfected with 500 ng viral or cellular in the presence of the corresponding inhibitors. Lysates were harvested 4 h p.t. and TRIM28 S473 phosphorylation was analyzed by western blot.

IL-6, and IL-8 compared to the cells expressing the wildtype protein. The reason for this is unknown. We speculate, that other phosphorylation sites, such as S471 and/or others, compensate for the lack of S473 phosphorylation. The phosphorylation dynamics of other phosphorylation sites of TRIM28 are not well-understood and require further investigation. In summary, our data demonstrate that S473 phosphorylation is functionally linked to increased expression of IFN-β, IL-6, and IL-8 and support our hypothesis, that phosphorylation at S473 modulates the corepressor activity of TRIM28 during infection with HPAIV.

#### DISCUSSION

Infection of humans with HPAIV is often associated with severe tissue damage and multiple organ failure caused by excessive production of IFNs and proinflammatory cytokines. The involved pathways as well as the underlying mechanisms leading to cytokine overexpression are not yet fully resolved. This knowledge gap impairs the development of new immunomodulatory treatment options due to the lack of suitable targets for efficient immunomodulatory therapies.

the indicated time points by plaque assay.

FIGURE 8 | Constitutive TRIM28 S473 phosphorylation potentiates the innate immune response. (A) TRIM28 KO cells were stably reconstituted with wildtype TRIM28 by retroviral transduction. Reconstituted cells were infected with VSV-luc at an MOI of 0.01 for 15 h. Cells were harvested and luciferase activity was measured. (B) In addition to wildtype TRIM28, phospho-mimetic variants (S473A and S473E) were stably reconstituted in TRIM28 KO cells. VSV-luc infection was carried out as described in (A). Results are depicted as mean RLU (±SEM) normalized to TRIM28 expression. \*\*\*\*p ≤ 0.0001; one-way ANOVA; Dunnett's multiple comparisons test. (C) Stably reconstituted TRIM28 KO cells were infected with KAN-1 at an MOI of 0.01. Total RNA was isolated 24 h p.i. and IFN-β, IL-6 and IL-8 mRNA levels were measured by qRT-PCR. mRNA levels are depicted as mean n-fold change (±SEM) over non-infected cells. n.s. p > 0.03, \*p ≤ 0.03; \*\*p ≤ 0.002; \*\*\*\*p ≤ 0.0001; one-way ANOVA; Dunnett's multiple comparisons test.

during IAV infection, viral RNAs are sensed by RIG-I. This initiates a signaling cascade, which includes the adaptor protein MAVS and results in the dimerization and nuclear translocation of the transcription factors IRF3/7. During infection with human adapted strains, this leads to non-pathological levels of IFN-β and proinflammatory cytokines, which usually leads to the clearance of IAV infection (left). In contrast, during infection with the HPAIV of the subtypes H5N1, H7N7 and H7N9 (right), the expression of IFN-β, IL-6 and IL-8 is potentiated. This is facilitated by PKR-mediated sensing of viral RNA followed by signal transduction via p38 and MSK1 resulting in phosphorylation of the transcriptional corepressor TRIM28 at serine 473. This leads to the release of TRIM28 corepressor activity and finally results in elevated expression of IFN-β, IL-6 and IL-8, which is commonly associated with tissue damage und high mortality during HPAIV infections.

Here, we report for the first time, that the cellular corepressor and negative immune regulator TRIM28 is the direct target of a signaling cascade involving the kinases PKR/p38/MSK1 during infection of human alveolar epithelial cells with HPAIV and contributes to the high expression levels of IFN-β, IL-6 and

IL-8. Based on our results we hypothesize that TRIM28 is a key determinant for IFN-β overexpression and cytokine-mediated tissue damage and may represent a potential therapeutic target for the treatment of HPAIV-induced hypercytokinemia in humans.

TRIM28 is widely described as a genomic corepressor and negative immune regulator of cytokine expression in response to different immune stimuli. The described mechanisms of action involve its intrinsic E3 SUMO ligase activity as well as HP1-BD-mediated corepressor activity, which are assumed to be fine-tuned by SUMOylation and phosphorylation (24). The exact contribution of SUMOylation and phosphorylation to the regulation of TRIM28 activities has remained enigmatic. However, several reports have established an attractive regulatory model. While SUMOylation might control the general and genome wide repressor function of TRIM28, stimulus-dependent phosphorylation presumably regulates the de-repression of specific gene subsets (25) to allow stimulus- and stress-specific host responses. In line with this, Kubota et al. reported that tyrosine phosphorylation at positions Y449, Y458, and Y517 regulates HP1-binding and the controlled de-repression of genes required for stress tolerance and repair processes (60) and Li et al. demonstrated that phosphorylation of S824 regulates the expression of genes involved in cell cycle control and apoptosis in response to genotoxic stresses (25). Intriguingly, the authors of this report observed that the level of TRIM28 SUMOylation was decreased when S824 was mutated to aspartic acid to mimic constitutive phosphorylation, suggesting PTM crosstalk (25). Our own data suggest that phosphorylation of TRIM28 at S473 regulates the de-repression of IFN-β, IL-6 and IL-8. Nevertheless, we assume that additional sites could be involved, as we observed phosphorylation of the neighboring serine 471 in our phosphoproteomic screen (**Figure 1A**, lower panel). Supportive evidence for the biological relevance of phosphorylation of S473 and S471 comes from other proteomic studies in which these sites have been identified as phosphoacceptor sites (61, 62). In addition, both sites are also highly conserved in mice, rats and dogs, suggesting a biological important function. The p38/MSK1/TRIM28 signaling-axis was described previously to play a central role in myoblast differentiation. In these cells, TRIM28 phosphorylation controls the activity of the central transcriptional activator MyoD and thereby differentiation of myoblasts into myotubes (63). In addition, this report convincingly demonstrated that TRIM28 is a bona fide target of MSK1 in an in vitro kinase assay.

The detailed mechanism of TRIM28-mediated cytokine amplification during HPAIV infection remained unsolved. Based on available reports we assume that phosphorylation at S473 attenuates HP1- and chromatin-binding of TRIM28, which results in the loss of its corepressor function and leads to the de-repression of the described genes (49). However, other mechanism cannot be excluded. Because TRIM28 itself does not possess DNA binding activity, it is likely that cytokine repression occurs through the interaction with other transcription factors and chromatin remodeling enzymes. Indeed, TRIM28 was shown to interact and modulate the activity of diverse immune-related proteins, including NF-κB (64), STAT1 (28), STAT3 (29), IRF7 (30), and IRF5 (65). Nevertheless, a conjoint conclusion for the mode of action of TRIM28 is difficult to extract because diverse cell lines and immune stimuli were employed and the impact of S473 phosphorylation was not addressed. Thus, it needs to be investigated whether one of these factors facilitates TRIM28 mediated cytokine upregulation upon S473 phosphorylation. Recently, a novel model for TRIM28-mediated control of gene expression was proposed (66, 67). In this model, TRIM28 is involved in tethering of the 7SK snRNP complex to the promotor proximal regions of many rapid response genes that contain paused RNA Polymerase II (Pol II). Thereby, TRIM28 facilitates recruitment of the positive transcription elongation factor P-TEFb, which releases paused Pol II by phosphorylating serine 2 in the pol II C-terminal domain (CTD) and allows rapid elongation of transcription (62, 67, 68). Most intriguingly, TRIM28 was found to be associated with more than 13,000 promotor proximal regions, giving a rough estimation of how many genes might be regulated by TRIM28 (69). So far, the importance of S473 phosphorylation and SUMOylation for the control of immune-related genes has not yet been addressed in this model.

Phosphorylation of TRIM28 S473 was induced in a straindependent manner. This suggests that the degree of human adaptation as well as the reported characteristic to induce hypercytokinemia and tissue damage in humans might be determinants for TRIM28 phosphorylation during infection. To challenge this theory, we included the pandemic 2009 H1N1 virus in our analysis because it is a triple reassortant virus containing genes derived from humans, swine and birds and has acquired stepwise human adaptation in pigs prior to human transmission. In contrast to other pandemic IAV strains, H1N1pdm demonstrated weak virulence and low mortality rates (70) and human infections with this virus are not necessarily associated with hypercytokinemia and tissue damage. Thus, we expected that this strain would not trigger S473 phosphorylation. Indeed, we could not detect S473 phosphorylation with H1N1pdm, supporting our hypothesis (**Figure 1C**). The reasons for strain-dependent phosphorylation of TRIM28 on a molecular level are not known. It is tempting to speculate that it is mediated by virus intrinsic properties, such as avian specific protein signatures, differences in the NS1-mediated inhibition of PKR activation or other factors that underlie human adaptation. Alternatively, differences in replication speed or nuclear export of vRNPs, leading to accumulation of cytosolic vRNA cannot be excluded.

In addition to the novel role of TRIM28, our results suggest a new mechanism for PKR-mediated cytokine expression. Here, PKR senses viral RNA at a late time point during infection with HPAIV and provokes TRIM28 S473 phosphorylation via p38 and MSK1 with the consequence of excessive production of IFN-β, IL-6 and IL-8. PKR-mediated regulation of IFN-β expression in virus infected cells is described to be facilitated by activation of the translation elongation factor eIF2α as well as by compromised IFN-β mRNA stability (71, 72). Here we show, that in HPAIV-infected cells, PKR signals via p38/MSK1 to inactivate TRIM28 and potentiates the expression of IFN-β, IL-6 and IL-8 in human lung epithelial cells. At this moment, it remains unknown whether this pathway is also present in other IAV susceptible cell types, such as macrophages and dendritic cells, which could have severe immunopathological consequences as these cells are the main producers of IFNs and cytokines. The results from the phosphoproteomic screen as well as western blot analysis demonstrate that S473 phosphorylation occurs at a surprisingly late time point during infection. We assume that PKR activation requires the accumulation of viral RNA in the cytoplasm, possibly in the form of exported vRNPs, in order to boost IFN-β and cytokine expression through S473 phosphorylation. This mechanism of PKR activation has been previously suggested for Influenza B viruses (73) but is not described for IAV. Although our results convincingly show that TRIM28 phosphorylation is mediated by PKR, we can currently not exclude that other signaling pathways and receptors, such as TLR3, which signals independently of the adaptor proteins MyD88 and MAVS, are also involved.

In summary, we propose a model for the TRIM28-mediated potentiation of cytokine expression during HPAIV infection. During infection with human adapted IAV strains, viral RNA is detected early during infection by RIG-I, which leads to the expression of non-pathological levels of IFN-β and proinflammatory cytokines (**Figure 9**, left side). In contrast, during infection with HPAIV of the H5N1, H7N7, and H7N9 subtypes cytosolic viral RNA is recognized by PKR, in addition to the RIG-I-dependent antiviral response. This leads to the activation of p38 and MSK1 and subsequently to phosphorylation of TRIM28 at S473 with the consequence of exacerbation of the ongoing immune response by amplification of IFN-β, IL-6 and IL-8 expression, which may lead to excessive immune cell recruitment and tissue inflammation (**Figure 9**, right side). We therefore propose, that controlling phosphorylation of TRIM28 by therapeutic interventions could prevent uncontrolled cytokine expression during HPAIV infections in humans.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

TK and LB are responsible for the concept and designed the experiments. TK, FG, VG, LH, SS, and CN conducted the experiments. MB and JW generated knockout cell lines. GZ generated recombinant VSV-luc. LB and TK wrote the manuscript. LB, TK, SL, and UR discussed and edited the manuscript.

#### FUNDING

This project was funded by the Young Investigator fund Innovative Medizinische Forschung of the Medical Faculty Muenster (granted to LB), several grants of the Deutsche Forschungsgemeinschaft (DFG) (BR 5189/1-1 to LB and LU 47723/1 to SL) and SFB/CRC 1009/B2 and A6.

#### ACKNOWLEDGMENTS

We would like to thank René Riedke, Maria Bethke, Sebastian Kordes and Ludmilla Wixler for technical assistance and Martin Stehling for FACS-based isolation of single cell clones. We are grateful for discussion and critical comments to Martin Schwemmle and Yvonne Börgeling. We acknowledge support by Open Access Publication Fund of University of Muenster.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02229/full#supplementary-material


inflammatory macrophages. Immunobiology (2012) 217:1315–24. doi: 10.1016/j.imbio.2012.07.026


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

Copyright © 2018 Krischuns, Günl, Henschel, Binder, Willemsen, Schloer, Rescher, Gerlt, Zimmer, Nordhoff, Ludwig and Brunotte. 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.

# Hantavirus-Driven PD-L1/PD-L2 Upregulation: An Imperfect Viral Immune Evasion Mechanism

#### Martin J. Raftery, Mohammed O. Abdelaziz, Jörg Hofmann and Günther Schönrich\*

Berlin Institute of Health, Institute of Virology, Charité–Universitätsmedizin Berlin, Humboldt-Universität zu Berlin, Berlin, Germany

Viruses often subvert antiviral immune responses by taking advantage of inhibitory immune signaling. We investigated if hantaviruses use this strategy. Hantaviruses cause viral hemorrhagic fever (VHF) which is associated with strong immune activation resulting in vigorous CD8+ T cell responses. Surprisingly, we observed that hantaviruses strongly upregulate PD-L1 and PD-L2, the ligands of checkpoint inhibitor programmed death-1 (PD-1). We detected high amounts of soluble PD-L1 (sPD-L1) and soluble PD-L2 (sPD-L2) in sera from hantavirus-infected patients. In addition, we observed hantavirus-induced PD-L1 upregulation in mice with a humanized immune system. The two major target cells of hantaviruses, endothelial cells and monocyte-derived dendritic cells, strongly increased PD-L1 and PD-L2 surface expression upon hantavirus infection in vitro. As an underlying mechanism, we found increased transcript levels whereas membrane trafficking of PD-L1 was not affected. Further analysis revealed that hantavirus-associated inflammatory signals and hantaviral nucleocapsid (N) protein enhance PD-L1 and PD-L2 expression. Cell numbers were strongly reduced when hantavirus-infected endothelial cells were mixed with T cells in the presence of an exogenous proliferation signal compared to uninfected cells. This is compatible with the concept that virus-induced PD-L1 and PD-L2 upregulation contributes to viral immune escape. Intriguingly, however, we observed hantavirus-induced CD8+ T cell bystander activation despite strongly upregulated PD-L1 and PD-L2. This result indicates that hantavirus-induced CD8+ T cell bystander activation bypasses checkpoint inhibition allowing an early antiviral immune response upon virus infection.

#### Keywords: bystander activation, hantaviruses, viral immune evasion, PD-L1, PD-L2, PD-1, CD86

#### INTRODUCTION

The immune response to infection is regulated not only by signaling through antigen receptors but also by co-receptors (1). The principal stimulatory co-receptor CD28 is constitutively expressed on T cells and interacts with CD80 and CD86 expressed on activated professional antigen-presenting cells (APCs) such as dendritic cells (DCs) (2). In contrast, programmed death-1 (PD-1), a member of the CD28 family, is a key negative regulator of immune responses (3). PD-1 is expressed on activated T cells whereas the known PD-1 ligands, PD-L1 and PD-L2, are detected on professional APCs similar to CD80 and CD86 (4). In addition, PD-L1 is expressed by non-hematopoietic cells such as endothelial cells (5–8). PD-L1

#### Edited by:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany

#### Reviewed by:

Tony Schountz, Colorado State University, United States Dimitris Lagos, University of York, United Kingdom Jorg Goronzy, Stanford University, United States

\*Correspondence: Günther Schönrich guenther.schoenrich@charite.de

#### Specialty section:

This article was submitted to Cytokines and Soluble Mediators in Immunity, a section of the journal Frontiers in Immunology

> Received: 25 May 2018 Accepted: 17 October 2018 Published: 03 December 2018

#### Citation:

Raftery MJ, Abdelaziz MO, Hofmann J and Schönrich G (2018) Hantavirus-Driven PD-L1/PD-L2 Upregulation: An Imperfect Viral Immune Evasion Mechanism. Front. Immunol. 9:2560. doi: 10.3389/fimmu.2018.02560

**53**

is further upregulated by proinflammatory cytokines that are released during virus infections such as type I and type II interferon (9). These pro-inflammatory cytokines also enhance PD-L2 expression, which is usually expressed at only low levels by a restricted number of cell types such as dendritic cells (DCs) (9).

Viruses have evolved mechanisms to exploit host inhibitory receptor signaling for subversion of host immune responses (10). Persisting viruses such as human immunodeficiency virus type 1 (HIV-1), hepatitis B virus (HBV) and hepatitis C virus (HCV) drive virus-specific CD8+ T cells into a dysfunctional or "exhausted" phenotype that is characterized by increased PD-1 expression (11, 12). In accordance, blockade of PD-1 or its ligands in chronic viral infection can enhance virusspecific CD8<sup>+</sup> T-cell responses and reduce the viral load. The functional consequences of PD-L1 upregulation during acute viral infection are less clear (13). For example, CD8+ T cell responses are impaired and immunopathology is attenuated by the PD-1 pathway during acute virus infections of the lower respiratory tract (14). On the other hand, it has been reported that PD-L1 upregulation on DCs contribute to the antiviral defense during acute HSV-1 infection (15). Moreover, during acute Friend retrovirus infection CD8+ T cells expressing high levels of PD-1 were both cytotoxic and critical for virus control (16).

Viral hemorrhagic fever (VHF) is a term for a group of similar but distinct zoonotic human diseases that are caused by RNA viruses including hantaviruses. Humans are infected with hantaviruses after inhalation of aerosols that contain virions derived from the natural host reservoirs, mostly rodents (17). The hallmarks of VHF are increased vascular permeability and loss of platelets (18). Hantaviruses are known to replicate without causing obvious cytopathic effects. As with other VHFs dysregulated immune responses play a role in hantavirusassociated diseases (19, 20). Paradigmatic experiments with lymphocytic choriomeningitis virus (LCMV)-infected mice have shown that PD-L1 is critical for prevention of immunopathology and virus-induced dysfunction such as vascular leakage (21, 22). Thus, it is important to understand how hantavirus replication modulates PD-L1 and PD-L2. In this study, we investigated how hantavirus replication affects the key stimulatory and inhibitory checkpoints of immune responses and explored the functional consequences thereof.

#### MATERIALS AND METHODS

#### Ethics Statement

The analyses of human sera were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this retrospective study, formal consent is not required. Buffy coat preparations were purchased from German Red Cross (Dresden). Blood samples were taken with the approval of the ethics committee of the Charité–Universitätsmedizin Berlin. Written informed consent was obtained from all donors.

#### Cells

Vero E6 and RPE-1 cells were cultured in Dulbecco's MEM (Gibco) supplemented with 10% hiFCS (BioWhittaker), 2 mM L-glutamine, penicillin and streptomycin (PAA). HUVECs were generated and cultivated as described (23). Adherent cells were passaged by first washing with PBS (Biochrom), addition of trypsin until cells detached and finally addition of FCS-containing medium to stop trypsin. HEL cells, an erythroleukemia suspension cell line, were cultured in RPMI 1640 (Gibco) with 10% hiFCS, 2 mM L-glutamine, penicillin and streptomycin (PAA). Huh7.5 cells is a human hepatoma cell line, which expresses an endogenous RIG-I with a mutation (T55I) in the first caspase-recruiting domain. This mutated RIG-I acts as a dominant-negative inhibitor (24). Transduced Huh7.5 clones overexpressing constitutive active RIG-I have been generated previously and were cultured as described (25). Huh7.5 cells were cultured as previously described (26).

Density gradient centrifugation using Ficoll-Paque was used to isolate PBMCs from buffy coat units (DRK, Dresden). In short, blood diluted 1:1 with RPMI wash (RPMI 1640, 2% heatinactivated FCS and 0.2 mM EDTA) was layered onto Ficoll (PAA) and centrifuged at 800 g, 30 min RT. PBMC were isolated from the interface, washed twice and CD14+ cells isolated using Blood CD14 isolation kit (Miltenyi Biotec). CD14+ monocytes were used to generate immature DCs by cultivation in RPMI1640 with 10% hiFCS (Hyclone), 2 mM L-glutamine, penicillin and streptomycin (PAA) and further supplemented with 500 IU/ml GM-CSF (ImmunoTools) and 200 IU/ml IL-4 (ImmunoTools). Medium and cytokines were changed every 2–3 days, cells were used for experiments at day 6.

#### Cytokines And Pathogen-Associated Molecular Patterns (PAMPs)

IFN-α, IFN-β, and IFN-γ were provided by ImmunoTools. Further samples of IFN-β were supplied by R&D Systems. TLR3 agonist polyinosinic:polycytidylic acid [poly(I:C)] and polydeoxyadenylic:polydeoxythymidylic acid [poly(dA:dT)], which indirectly stimulates retinoic acid–inducible gene I (RIG-I), were obtained from InvivoGen. Poly(I:C) was used at 10µg/ml and poly(dA:dT) at 1µg/ml.

#### Serum Samples And ELISAs

Samples from patients infected with Puumala virus (PUUV) or Dobrava-Belgrade virus (DOBV) were collected for diagnostic purposes and were anonymized and stored before being tested retrospectively. Routine diagnostic testing included qPCR of the L segment of hantavirus from RNA isolated from the sera, positivity indicating the presence of active viral infection and thus an acute infection. All serum samples were stored at −80◦C before use. The histone/dsDNA complexes were determined using Cell Death Detection ELISAPLUS (Roche) for quantification of neutrophil extracellular traps (NETs) in the serum as previously described (27). Human sPD-L1 and sPD-L2 levels were determined by using ELISA kits from R&D Systems, whereas the ELISA for measuring soluble CD86 (sCD86) was provided by PromoKine.

#### Flow Cytometry of Surface Molecules

Cells were harvested and washed twice in ice-cold FACS washing solution. Cells were then resuspended in 50 µl FACS blocking solution, containing the primary antibody in appropriate dilution, and incubated for 1 h. Cells stained with directlycoupled antibodies were washed and analyzed. For uncoupled primary antibodies after incubation cells were again washed twice with FACS wash and secondary antibody, diluted in FACS block solution, was added. After 45 min the cells were washed with FACS wash solution and resuspended in FACS fixation solution. For quantifying fluorescence of labeled cells a FACSCalibur <sup>R</sup> (BD Biosciences) was used. Results were evaluated with the flow cytometry analysis software programs CellQuestPro <sup>R</sup> and FlowJo V10 (BD Biosciences).

### Transfection

Transfection was undertaken using plasmid pcDNA3.1 HTNV N or empty pcDNA3.1 as control (1 µg) using Lipofectamine 3000 transfection reagent according to the manufacturer's protocol, including Optimem medium (Thermo Fischer Scientfic).

#### Antibodies And Staining Reagents

For flow cytometry and functional studies, respectively, the following antibodies and staining reagents were used: anti-CD40 (clone 5C3), anti-CD54 (clone HA58), anti-CD80 (L307.4), anti-CD83 (clone HB15e), anti-CD107a (H4A3), and anti-B7- H2 (clone 2D3) were supplied by BD PharMingen; anti-PD-1 (clone J116), anti-PD-L1 (clone MIH1), and anti-PD-L2 (clone MIH18), anti- B7-H3 (clone H74), anti-B7-H4 (clone MIH35) were purchased from eBioscience; anti-CD86 (clone IT2.2) was supplied by ImmunoTools; anti-DC-SIGN (Clone MR-1) was purchased from Acris; anti-MHC class I (clone w6/32) and II (clone L243) were produced in-house; HCMV pp65 495-503 loaded NTA HLA-A2 tetramer reagents were obtained from TCMetrix. Secondary antibodies coupled to fluorochromes were supplied by Dianova. Blocking monoclonal antibodies directed against human IL-15 (clone 34559) and anti-human IFNR chain 2 (clone MMHAR-2) were supplied by R&D Systems. Cells were incubated with blocking antibodies or isotype-matched control antibodies for 1 h before exposure to virus. Isotypematched control antibodies were supplied by BD PharMingen. For immunohistochemistry human-specific FITC-coupled anti-MPO (clone 7.17; ImmunoTools) and polyclonal goat anti human PD-L1 (R&D Systems) were used with bovine anti goat Fab fragment Alexa 594-coupled (Dianova) as secondary antibody, all used at 1:300 dilution.

### PD-L1 Uptake Protocol

Cells were incubated with PE-coupled anti-PD-L1 antibody for 1 h at 4◦C or 37◦C for 4 h before being washed and analyzed by flow cytometry. Uptake was calculated by subtracting MFI at 37◦C from MFI at 4◦C. Uptake of HTNV infected cells was then compared to uninfected cells.

### T Cell Assay

CD4+ cells were isolated from PBMCs using CD4-coupled beads (Miltenyi) and frozen on liquid nitrogen until use. HUVECs infected with HTNV at a MOI of 1.5 were incubated in flatbottom 96-well plates for 4 days before being were mixed with allogenic CD4+ cells at a ratio of 1:4 and treated with PHA at 5ug/ml for 2 days. Proliferation was measured by MTT dye test (EZ4U-test).

### Viruses

Virus stocks of Hantaan virus (HTNV, strain 76-118) and Tula virus (TULV, strain Moravia) were propagated on VeroE6 cells in a biosafety level 3 (BSL3) laboratory as previously described (28). For virus titration, supernatants from hantavirus-infected cells were incubated with Vero E6 cells and subsequently focusforming units (FFU) were counted in a chemiluminescence detection assay (29). Virus stocks were regularly tested for mycoplasma by PCR and stored at −80◦C before use. In order to infect cells virions were allowed to adsorb to cells for 1 h. After infection cells were washed three times with medium before incubation in a humidified incubator at 37◦C. Uninfected cells treated with medium instead of virions were used as mock control. Herpes simplex virus type 1 (HSV-1) strain KOS and Vesicular stomatitis virus (VSV, strain Indiana) was propagated and titrated as previously described. Titres were determined by plaque assay on Vero E6 cells and expressed as PFU per milliliter (30). UV inactivation was performed for 5 min and the remaining titer was tested and found to be less than 1 FFU/ml.

### qPCR

RNA was isolated from cells using RNeasy Plus mini kit (Quiagen) and reverse transcribed using SuperScript III (Thermo Fisher Scientific). qPCR was performed on a qTOWER<sup>3</sup> (Analytik Jena) using PrimeTime gene expression master mix and PrimeTime primers (IDT). The input RNA was normalized using average expression of β-actin and cyclophilin B housekeeping genes.

### Humanized Mouse Model

The generation of mice with a humanized immune system has been described elsewhere (31). Briefly, NSG mice expressing HLA-A2, a human MHC class I molecule, were humanized by reconstitution with HLA-A2+ human CD34+ hematopoietic stem cells isolated from umbilical cord blood. Engraftment was evaluated at 11 weeks post inoculation by cytofluorimetric analysis of PBMCs. Successfully engrafted mice were infected i.p. with 10<sup>5</sup> focus-forming units (FFU) of HTNV (strain 76- 118). Infection was successful as determined by qPCR from sera. Twenty-Two days post infection mice were sacrificed and liver, kidney, lungs and spleen fixed and mounted in paraffin blocks. The infection experiments were approved by the governmental animal-welfare committee of the state Berlin, Germany (G 0013/12).

#### ImageJ Analysis

Six cell-rich areas of five to twelve cells each were analyzed on each slide. Cell density was determined blind using DAPI staining and subsequently the mean intensity of staining of human PD-L1 (Texas Red) was determined.

### Statistical Analysis

Student's t-test and 1 way ANOVA test with Bonferroni correction were used to determine statistical significance. Pvalues below 0.05 (95% confidence) were considered to be significant. Prism 6 software (GraphPad) was used for statistical analysis.

### RESULTS

#### Strong Upregulation of PD-1 Ligands in Hantavirus-Infected Patients and in an Animal Model of Hantavirus Infection

Initially we tested if hantaviruses modulate the expression of the ligands of checkpoint inhibitor PD-1 during clinical infection of humans. For this purpose, we measured the amount of soluble PD-L1 (sPD-L1) and soluble PD-L2 (sPD-L2) in sera from hantavirus-infected patients. The level of both sPD-L1 (**Figure 1A**) and sPD-L2 (**Figure 1B**) were strongly upregulated in sera from hantavirus-infected patients as compared to normal healthy individuals. Sequential samples from the same patients indicate that for both PUUV and DOBV sPD-L1 levels decrease with time indicating active regulation and that acute samples still with active virus replication (hantavirus RNA positive) have high sPD-L1 levels (**Figure 1C**). Similarly, PUUV samples early in convalescence (IgM > IgG) had significantly raised sPD-L1 compared to samples taken later (IgG > IgM) (**Figure 1D**). We also detected elevated levels of neutrophil extracellular traps (NETs), a marker for recent hantavirus infection, in these sera (**Figure 1E**) (27, 32). The level of sPD-L1 detected in culture supernatants and plasma of patients is known to correlate with the level of membranebound PD-L1 (33, 34). Taken together, PD-L1 and PD-L2 are strongly upregulated in hantavirus infected patients. Using a previously established animal model of hantavirusinduced immunopathology we analyzed the spleen of hantavirusinfected mice with a humanized immune system as previously published (31). We observed enhanced expression of human PD-L1 in the spleen (**Figure 1F**) in addition to high levels of human myeloperoxidase (MPO)-expressing cells, presumably neutrophils (data not shown). Taken together this data shows that PD-L1 and PD-L2 are strongly upregulated during hantavirus infection in vivo.

#### Hantavirus-Infected Human Dendritic Cells Upregulate Both Costimulatory Molecules as Well as PD-L1/PD-L2

Next we investigated the possible source of sPD-L1 and sPD-L2 seen in sera from hantavirus-infected patients. The production of sPD-L1 by proteolytic cleavage of membrane-bound PD-L1 is a feature of activated monocyte-derived DCs (35). This important immunoregulatory cell type is susceptible to hantavirus infection (36–39). As previously reported, immature DCs infected with Hantaan virus (HTNV), the most common cause of human hantavirus infections, upregulated adhesion molecules and MHC molecules (**Figure 2A**). In addition, HTNV increased expression of costimulatory molecules on the surface of immature DCs (**Figure 2B**). Intriguingly, HTNV infection resulted in enhanced expression of both PD-L1 and PD-L2 whereas PD-1 was barely detectable on the surface of uninfected and HTNV-infected immature DCs (**Figure 3A**). In contrast, HTNV-infected DCs did not upregulate other members of the B7 family such as B7-H2, B7-H3, and B7- H4. (**Figure 3B**) (40). In summary, hantavirus replication in DCs drives surface expression of both T cell costimulatory molecules such as CD86 as well as the T cell inhibitory molecules PD-L1/PD-L2.

### Hantavirus Regulates PDL1/PDL2 Expression on the Transcription Level

In further experiments we analyzed the mechanism upregulating PD-L1 and PD-L2 during hantavirus infection of DCs. PD-L1 expression can be regulated on the genetic, transcriptional, post-transcriptional and post-translational level (41). We first determined the number of PD-L1 and PD-L2 transcripts in HTNV-infected DCs and DCs exposed to IFN-α by qPCR. HTNV increased the number of transcripts encoding PD-L1 and PD-L2 (**Figure 4A**). IFN-α also upregulated PD-L1 and PD-L2 transcripts. We also tested whether HTNV modulates DCs trafficking of PD-L1. As shown in **Figure 4B** HTNV-infected DCs endocytosed PD-L1 as efficiently as uninfected control cells excluding altered endocytosis kinetics as a mechanism of PD-L1 upregulation. In conclusion, hantaviruses increase the number of PD-L1/PD-L2 transcripts but do not modulate endocytosis of the corresponding proteins.

#### Hantavirus-Associated Inflammatory Signals Including Hantaviral N Protein Drive PD-L1 Expression

Next we examined which hantavirus-associated inflammatory stimuli modulate PD-L1 expression on immature DCs. IFNγ and to a lesser extent IFN-α upregulated cell-surface PD-L1 (**Figure 5A**). Hantavirus replication triggers pattern recognition receptors (PRRs) such as toll-like receptor 3 (TLR3) and retinoic acid–inducible gene I (RIG-I) (30, 42, 43). Strikingly, TLR3 agonist poly(I:C) strongly increased PD-L1 expression on immature DCs (**Figure 5A**). Poly(I:C) similarly induced PD-L2 (data not shown). In contrast, immature DCs treated with RIG-I activating signals such as UV-inactivated VSV or poly(dA:dT) did not show increased PD-L1 expression (**Figure 5A**). The absence of PD-L1 upregulation after stimulation of the RIG-I pathway was confirmed by using Huh7.5 cells expressing a constitutive active RIG-I molecule (RIG-CA) (25). These cells did not express elevated PD-L1 levels compared to the untreated cells whereas Huh7.5 cells treated with IFN-γ upregulated PD-L1 compared to untreated Huh7.5 cells (**Figure 5B**). We also tested the effect of hantaviral nucleocapsid (N) protein, which has many diverse functional activities during the viral life cycle (44). As shown in **Figure 5C** expression of N protein in HEL cells, a human erythroleukemia cell line, resulted in PD-L1 upregulation. In summary, type I IFN, hantaviral N protein, and TLR3 signaling induced PD-L1 expression whereas RIG-I signaling had no effect.

FIGURE 1 | Levels of sPD-L1, sPD-L2, hantavirus-specific IgG and NETs in sera from hantavirus-infected patients. Sera from normal healthy individuals or convalescent hantavirus-infected patients (after the viremic phase) was tested by ELISA for levels of (A) sPD-L1 and (B) sPD-L2. Error bars represent the mean ± SD (\*\*\*\*p < 0.0001, \*\*\*p < 0.001, paired Student's t-test). (C) Sequential sera samples from patients with PUUV (black) or DOBV (blue) were tested for sPD-L1. Red samples also tested additionally positive for hantavirus RNA and are therefore acute infections. (D) Levels of sPD-L1 in patients with kidney failure or in convalescence were further analyzed. Convalescent sera were separated into early convalescent (IgM dominant) or late convalescent (IgG dominant). Error bars represent the mean ± SD (\*p < 0.05, paired Student's t-test). (E) The level of NETs in sera from normal healthy individuals or convalescent hantavirus-infected patients was determined as previously described (27). Error bars represent the mean ± SD (\*\*\*p < 0.001, paired Student's t-test). (F) Spleen sections from uninfected or HTNV-infected humanized mice were stained for human PD-L1 (red) and nuclei (blue). HTNV-infected spleen sections show large areas of human cells with enhanced PD-L1 expression in comparison to uninfected spleen sections (upper left and right panel; inserts show higher magnification of cells; bars represent 100µm). Slides from uninfected and HTNV-infected humanized and unreconstituted mice animals (N = 3 each group; 12 total) were analyzed using ImageJ to determine the intensity of human PD-L1 staining (Lower panel). Error bars represent the mean ± SEM (\*\*\*\*p < 0.0001, paired Student's t-test). The samples from unreconstituted mice were used to determine the background staining. No significant difference was found in background staining in HTNV-infected or uninfected unreconstituted mice.

## Subversion of T Cell Responses by Hantavirus-Induced Checkpoint Inhibitors

We next analyzed whether PD-L1 and PD-L2 is upregulated on hantavirus-infected endothelial cells, which play a pivotal role in hantavirus pathogenesis (45, 46). Upon hantavirus infection human umbilical vein endothelial cells (HUVECs) upregulated both PD-L1 and PD-L2 (**Figure 6A**). PD-L1 expression started to increase on HTNV-infected cells at 12 h post infection similar to MHC class I expression (**Figure 6B**). PD-L1 expression further increased at later time points post infection (**Figure 6B**). We also tested whether hantavirus-induced PD-L1 and PD-L2 modulate T cell responses. For this purpose, HTNV-infected

HUVECs were mixed with allogeneic CD4+ cells and stimulated with PHA. T cells strongly upregulate PD-1 upon stimulation with PHA (47). As shown in **Figure 6C** the numbers of surviving T cells and endothelial cells was strongly reduced in comparison to control T cells exposed to uninfected HUVECs, suggesting that T cell proliferation may be reduced. These results indicate that hantaviruses upregulate both PD-L1 and PD-L2 on endothelial cells which has a functional effect on T cells.

### Hantavirus-Induced Bystander Activation Despite Upregulation of Checkpoint Inhibitors

To test the functional consequences of PD-1 ligand upregulation we investigated the behavior of T cells when exposed to infected autologous myeloid cells. We infected PBMCs from healthy human donors with HTNV and subsequently stained T cells for expression of the C-type lectin CD69 as an early marker of T-cell activation (48). Recently, it has been shown that CD69 regulates the metabolism and migrationretention ratio of T cells as well as the acquisition of T cell effector or regulatory phenotypes (49). Surprisingly, we observed increased percentages of activated cells especially in the CD8+ T cell population early after infection of PBMCs with HTNV (**Figure 7A**). Bystander activation of T cells during viral infections is common and is initiated by stimulated professional APCs such as DCs (50). In order to identify the responding cells, we tested whether heterologous memory CD8+ T cells are activated in this experimental setting.

FIGURE 3 | Hantavirus-induced upregulation of PD-L1 and PD-L2 on immature DCs. (A) Immature DCs were infected with HTNV at a MOI of 1.5 and incubated for 4 days before staining for PD-1, PD-L1 or PD-L2. (B) Immature DCs infected as for (A) were stained for members of the B7 family other than PD-L1/PD-L2. The results shown are representative of three independent experiments using three different donors. Positive controls are given in the lower panel (B7-H2 and B7-H3 from HUVEC, B7-H4 from HEK293 cells transfected with a B7-H4 plasmid).

number of indicated transcripts quantified by qPCR according to the delta-delta-Ct (ddCt) method. (B) Immature DCs infected as for (A) were incubated with PE-coupled anti-PD-L1 antibody at 4◦C 1 h or at 37◦C for 4 h before being washed and analyzed by flow cytometry. Uptake was calculated by subtracting MFI at 37◦C from MFI at 4◦C. Uptake of HTNV infected cells was then compared to uninfected cells. Results are derived from three independent experiments, error bars represent the mean ± SD.

For this purpose we infected PBMCs derived from HLA-A2+ human healthy donors that were seropositive for human cytomegalovirus (HCMV), a member of the human herpesvirus family. A HLA-A2 tetramer loaded with a immunodominant

FIGURE 5 | Control of PD-L1 expression by inflammatory stimuli. (A) Immature DCs were exposed to the following inflammatory stimuli before staining for PD-L1: type I IFN (IFN-α at 1,000 U/ml), type II IFN (IFN-γ at 1,000 U/ml), poly(dA:dT), UV-inactivated VSV or poly(I:C) for 24 h. The results shown are representative of three independent experiments using three different donors. (B) Huh7.5 cells (control), Huh7.5 cells permanently expressing a constitutively active form of RIG-I (RIG-CA) or Huh7.5 cells stimulated with IFN-γ at 1,000 U/ml for 24 h were stained for PD-L1 and analyzed by flow cytometry. Results are derived from three independent experiments, error bars represent the mean ± SEM (\*p < 0.05, paired Student's t-test). (C) HEL cells were transfected with HTNV N-expressing plasmids or empty plasmids (Control). After 2 days cells were stained for PD-L1. Results are given as a percentage of control and are derived from three independent experiments, error bars represent the mean ± SD.

peptide derived from pp65 (CMVpp65TET) was used to detect HCMV-specific CD8+ memory T cells. After HTNV infection of PBMCs the percentage of CMVpp65TET+ CD8+ T cells that expressed CD107a (LAMP-1), a marker for degranulation

of activated CD8+ T cells (51), significantly increased in PBMCs as compared to uninfected PBMCs (**Figure 7B**). In contrast, in PBMCs infected with herpes simplex virus type 1 (HSV-1), another member of the human herpesvirus family, no significant increase in activated CMVpp65TET+ CD8+ T cells was observed (**Figure 7B**). In conclusion, heterologous T cells are activated at an early time point after hantavirus infection despite increased expression of PD-L1 on antigen-presenting cells.

(EZ4U-test). Results are derived from three independent experiments using

three different donors, error bars represent the mean ± SD.

### CD86-Dependency of Hantavirus-Induced Bystander Activation

We next examined the mechanisms by which hantavirus-infected DCs cause bystander activation despite checkpoint inhibitors. First, we tested whether hantavirus-infected DCs express inflammatory cytokines that can cause bystander activation of memory CD8+ T cells in the absence of cognate antigen such as IL-15, IL-18, and IL-21 (52). For this purpose RNA from immature DCs infected with HTNV or exposed to IFN-α was isolated and subjected to qPCR. As shown in **Figure 8A** HTNV upregulated production of mRNA encoding IL-15, IL-18, and IL-21 in immature DCs. This finding is in line with cytokinedrive bystander activation of T cells during hantavirus infection of PBMCs. In order to further dissect the mechanism we used antibodies to block IL-15 as this cytokine has been implicated in hantavirus-induced natural killer (NK) cell activation (53). We also blocked type I IFN, which also can contribute to bystander activation of T cells (52). The IL-15 block had no significant effect whereas the type I IFN block significantly reduced T cell bystander activation (**Figure 8B**). In comparison depletion of CD14+ cells completely abrogated hantavirus-induced bystander activation (**Figure 8B**). CD14 serves as marker for monocytes which are detected in PBMCs at frequencies of 10–20% (54) and represent the major hantavirus-permissive cell type in PBMCs. In addition, by blocking the T cell costimulatory molecule B7-2 (CD86) during HTNV infection of PBMCs we could also prevent bystander activation of CD8+ T cells (**Figure 8C**). In contrast, blocking of MHC class I molecules had no effect (**Figure 8C**). These result suggested that CD86 expressed by CD14+ cells plays a major role in hantavirus-induced bystander activation whereas interaction of T cell receptors (TCRs) with MHC-bound peptides interactions is not required (**Figure 8C**). In accordance, CD14+ cells strongly upregulated CD86 during infection with HTNV (**Figure 8D**) and high levels of soluble CD86 were detected in hantavirus-infected patients (**Figure 8E**). Taken together, these results demonstrate that CD14+ monocytes are inducing hantavirus-driven bystander T cell activation in a CD86-dependent manner.

### DISCUSSION

In this study, we detected high amounts of sPD-L1 and sPD-L2 in sera of hantavirus-infected patients. Hantaviruses strongly upregulated PD-L1 and PD-L2 on endothelial cells, which play a pivotal role in hantavirus-induced pathogenesis. In line with an inhibitory role of PD-L1/PD-L2 hantavirusinfected endothelial cells did not induce T cell proliferation. Hantaviruses also strongly increased expression of PD-L1 and PD-L2 on monocyte-derived DCs. However, monocyte-derived inflammatory cells could still activate heterologous CD8+ T cells in a CD86-dependent fashion. This indicates that hantavirusinduced CD8+ T cell bystander activation bypasses inhibitory checkpoints.

Gene expression of PD-L1 and PD-L2 is controlled by inflammatory signals (9). Hantavirus-induced upregulation of PD-L1 and PD-L2 could be indirect due to release of IFNs. In line with this view, endothelial cells and DCs predominantly produce IFN-β upon infection with pathogenic hantaviruses (37, 38, 55, 56). PD-L2 is upregulated equally well by IFN-β and IFN-γ whereas PD-L1 is especially sensitive to IFN-γ (57). In hantavirus-infected patients vigorous responses of NK cells and CD8+ T cells resulting in increased levels of IFN-γ are observed (19, 58–61). In addition to this, we show that hantaviral N protein in HEL cells resulted in PD-L1 upregulation although the underlying mechanism is unclear. Thus, IFN-independent mechanisms may contribute to hantavirus-induced PD-L1/PD-L2 expression as recently shown for MHC class I molecules (62). In conclusion, PD-L1/PD-L2 upregulation in hantavirus-infected patients is due to both IFNs and additional IFN-independent mechanisms.

Hantavirus infection is detected by pattern recognition receptors, primarily TLR-3 (42) and RIG-I. (30, 43). We found that the TLR3 ligand poly(I:C) strongly increased PD-L1 levels on immature DCs. In accordance, poly(I:C) has been reported to upregulate PD-L1 on DCs (63, 64) as well as endothelial cells (65) and airway epithelial cells (66). In contrast, PD-L1 was not upregulated upon stimulating RIG-I. Taken together, our in vitro observations would fit with hantavirus infection strongly inducing PD-L1 and PD-L2 by triggering TLR-3, which transmits downstream signals through the TIR-domaincontaining adapter-inducing IFN-β (TRIF) pathway. Production of IFN-β by both TLR3 and RIG-I induced signaling would be expected to further increase expression of PD-1 ligands later in infection.

Other viruses have also been reported to modulate checkpoint inhibitors. Similar to hantaviruses the Japanese encephalitis virus nonlytically infects monocyte-derived DCs thereby inducing phenotypic maturation and a significant increase in PD-L1 expression (67). Replication competent but not inactivated KSHV induces PD-L1 expression in human monocytes in a dose-dependent manner although the precise mechanism has

convalescent hantavirus-infected patients were tested by ELISA for levels of sCD86. Error bars represent the mean ± SD (\*\*\*\*p < 0.0001, paired Student's t-test).

not been defined (68). Akhmetzyanova et al. observed a type I IFN-dependent increase in PD-L1 expression after infection of spleen cells with the murine Friend retrovirus (FV) (69). PD-1 and PD-L1 are also up-regulated in monocytic cells upon HIV-1 infection (70, 71). In accordance, the HIV-1 Tat protein has been observed to increase PD-L1 expression on DCs through TNF-α and TLR4 (72). The HCV core protein up-regulates PD-L1 expression on Kupffer cells, which binds PD-1 to promote T cell dysfunction and development of viral persistence (73). A subset of macrophages upregulated PD-L1 expression via type I IFN during infection with LCMV (74). In addition, influenza virus enhances PD-L1 expression of lung macrophages through type I IFN signaling (75). Taken together, it appears that PD-L1 upregulation is a relatively common consequence of viral infection which is driven by type I IFN and viral PRR triggering.

PD-L1 expression on professional APCs facilitates the induction of regulatory T cells (Tregs) and enhances expression of the key transcription factor forkhead box p3 (Foxp3) (76, 77). Tregs not only regulate effector T cell function but also humoral immunity (78). A recent report has shown that the severity of hantavirus-associated disease correlates with expression of Foxp3 (79). This strongly suggests that hantavirus-induced upregulation of PD-L1 on DCs induces Tregs. In accordance, other investigators have shown that virus-induced PD-L1 upregulation on monocyte-derived DCs leads to expansion of Tregs (67).

We observed that hantavirus-infected human endothelial cells upregulate surface expression of PD-L1 and PD-L2 and inhibit proliferation of PHA-stimulated T cells. Other investigators detected increased amounts of PD-L1 in hantavirusinfected cultures of rat endothelial cells (80). In HFRS patients, hantavirus-induced PD-L1 may be responsible for the contraction of a newly identified highly cytotoxic T cell subset that strongly upregulates PD-1 in the late phase of hantavirus infection (81). Hantavirus-induced expression of PD-L1 and PD-L2 may contribute to the recently described protection of hantavirus-infected endothelial cells from cytotoxic attack by CD8+ T cells and NK cells (82). In line with this notion, antibody blockade of PD-L1 and PD-L2 on IFN-γ treated endothelial cells enhanced cytolytic activity of antigen-specific CD8+ T lymphocytes (8). Similarly, failure of the inhibitory PD-1/PD-L1 axis during hantavirus infection of vascular tissue may lead to unbalanced immunostimulation and immunopathology as proposed for inflammatory blood vessel diseases (83).

Despite checkpoint inhibition we observed bystander activation in a subset of T cells. Bystander activation of T lymphocytes represents a first line of antiviral defense and may contribute to hantavirus-induced immunopathogenesis. In line with his view, bystander T cells responding to dengue virus, another VHF virus, secrete IFN-γ (84). It has been reported that virus-induced bystander T cell activation bypasses control checkpoints such as Tregs (85). In accordance, we observed that hantavirus-induced bystander T cell activation is not prevented by PD-Ll/PD-L2 upregulation on monocyte-derived inflammatory DCs. This can be explained by the fact that bystander CD8+ T cell activation does not result in TCRinduced PD-1 upregulation. In contrast, TCR signaling induced by cognate antigen upregulates PD-1 expression on CD8+ T cells within the first 24 h during infection (86). This may ensure that virus-specific T cells are excluded from innate responses and differentiate into effector T cells of the adaptive immunity.

Hantavirus-induced bystander activation was strictly dependent on CD14+ cells. This may be explained firstly by the fact that monocyte-derived cells are needed for hantavirus infection in PBMCs. Secondly, CD86 is expressed almost exclusively on monocyte-derived cell types and we could show that CD86 was required for hantavirus-induced bystander activation. Thus, CD86 on hantavirus-infected DCs may activate heterologous CD8+ T cells through CD28. The importance of CD28 for bystander activation of CD8+ T cells has been previously described (87). It is unlikely that hantaviruses directly activate T lymphocytes through PRRs. However, previous reports demonstrated that inflammatory cytokines such as type I contribute to innate T cell activation (88–90). In accordance, we observed that blocking of type I IFN reduced bystander activation of CD8+ T cells upon hantavirus infection.

Many acute viral infections are known to trigger bystander activation of heterologous CD8+ T cells (91–93). Often CD8+ T cells specific for human herpesviruses contribute to the heterologous antiviral immune response (92). In line with this view, we observed activation of HCMV-specific memory CD8+ T cells in PBMCs from HCMV-seropositive patients after hantavirus infection. In fact, bystander activation of CD8+ T cells represent an early line of antiviral defense (94). Bystander activated CD8+ T lymphocytes control early pathogen load in virus-infected tissue by a NKG2D-dependent mechanism (95). In accordance with this concept, cytotoxic CD8+ T cells strongly expressing NKG2D were detected in the lung of hantavirusinfected patients (96). NKG2D ligands are upregulated by PRRs that sense viral replication (97). These include RIG-I, which has been shown to detect hantaviruses (30). Interestingly, a strong plasmablast response with reactivity against virus-unrelated antigens has recently been detected in patients with acute hantavirus pulmonary syndrome (98). Whether this heterologous B cell response has a pathogenic or protective role is unclear.

In conclusion, hantavirus-infected patients suffer from immunopathology in the face of immunosuppressive PD-L1 upregulated by hantaviral N protein and most likely hantavirusinduced TLR3 signaling. This apparent discrepancy could be explained by rapid cleavage and removal of PD-L1 from the surface of hantavirus-infected cells in vivo. In accordance, we detected large quantities of sPD-L1 in the serum of patients with hantavirus-associated disease. Moreover, the lack of opportunistic infections in these patients implies that PD-L1 does not globally suppress the immune system. Finally, early activation of heterologous CD8+ T cells during acute virus infections bypasses or overwhelms the inhibitory PD-1/PD-L1 axis and represents a means of eluding viral immune subversion at least in the short term (99).

#### AUTHOR CONTRIBUTIONS

MR designed research, performed all experiments, analyzed data, contributed to writing, and prepared figures. MA analyzed data and contributed to manuscript revision. JH provided serum samples from hantavirus-infected patients, contributed to manuscript revision and provided intellectual input. GS was involved in experiment conception, wrote the paper, analyzed data, provided intellectual input, and contributed to figure preparation.

#### ACKNOWLEDGMENTS

We acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité–Universitätsmedizin Berlin. This work was also supported by a grant (GALHANT) from the Federal Ministry of Education and Research (BMBF) and the FAZIT foundation. We thank M. Binder for providing the Huh7.5 RIG-CA cells and C. Priemer for assistance in culturing cells.

#### REFERENCES


molecule on microvascular endothelial cells. Microcirculation (2002) 9:133– 45. doi: 10.1038/sj/mn/7800123


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

Copyright © 2018 Raftery, Abdelaziz, Hofmann and Schönrich. 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.

# Cytokine-Mediated Induction and Regulation of Tissue Damage During Cytomegalovirus Infection

Mathew Clement and Ian R. Humphreys\*

*Division of Infection and Immunity/Systems Immunity University Research Institute, Cardiff, United Kingdom*

Human cytomegalovirus (HCMV) is a β-herpesvirus with high sero-prevalence within the human population. Primary HCMV infection and life-long carriage are typically asymptomatic. However, HCMV is implicated in exacerbation of chronic conditions and associated damage in individuals with intact immune systems. Furthermore, HCMV is a significant cause of morbidity and mortality in the immunologically immature and immune-compromised where disease is associated with tissue damage. Infection-induced inflammation, including robust cytokine responses, is a key component of pathologies associated with many viruses. Despite encoding a large number of immune-evasion genes, HCMV also triggers the induction of inflammatory cytokine responses during infection. Thus, understanding how cytokines contribute to CMV-induced pathologies and the mechanisms through which they are regulated may inform clinical management of disease. Herein, we discuss our current understanding based on clinical observation and *in vivo* modeling of disease of the role that cytokines play in CMV pathogenesis. Specifically, in the context of the different tissues and organs in which CMV replicates, we give a broad overview of the beneficial and adverse effects that cytokines have during infection and describe how cytokine-mediated tissue damage is regulated. We discuss the implications of findings derived from mice and humans for therapeutic intervention strategies and our understanding of how host genetics may influence the outcome of CMV infections.

Keywords: cytokine, cytomegalovirus infection, immunopathologic process, virus, mcmv

#### INTRODUCTION

Human cytomegalovirus (HCMV) is a ubiquitous beta-herpesvirus that has co-evolved with its host for millions of years and acquired multiple immune evasion functions that manipulate and hide the virus from host immunity (1, 2). Primary HCMV infection and latency in immune-competent hosts is usually asymptomatic (3). Thus, HCMV is typically thought to establish lifelong infection without inducing overt pathology often triggered by other viruses. It is becomingly apparent, however, that chronic HCMV carriage in 'healthy individuals' may exacerbate conditions from general frailty (4) to cardiovascular disease (5).

HCMV causes morbidity and mortality in immune-compromised patients including transplant recipients and HIV co-infected individuals. Solid-state organ or human stem cell transplantation remains challenging as immune suppression can facilitate uncontrolled HCMV reactivation from host and/or donor tissue, resulting in organ pathology and systemic disease (6). HCMV co-infection is the leading cause of vision loss in untreated HIV/AIDS individuals (7, 8) and

#### Edited by:

*Juliet Spencer, Texas Woman's University, United States*

#### Reviewed by:

*Mary Hummel, Northwestern University, United States Sarah Rowland-Jones, University of Oxford, United Kingdom Laura Kay Hanson, Texas Woman's University, United States*

#### \*Correspondence:

*Ian R. Humphreys humphreysir@cardiff.ac.uk*

#### Specialty section:

*This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology*

Received: *14 September 2018* Accepted: *11 January 2019* Published: *29 January 2019*

#### Citation:

*Clement M and Humphreys IR (2019) Cytokine-Mediated Induction and Regulation of Tissue Damage During Cytomegalovirus Infection. Front. Immunol. 10:78. doi: 10.3389/fimmu.2019.00078*

**66**

remains an issue in patients receiving anti-retroviral therapy (9). HCMV causes gastrointestinal and neurological diseases during HIV co-infection (7, 10). Further examples of viralinduced morbidity include congenital infection where HCMV is the leading infectious cause of all congenital birth defects (11, 12). Life-long neurological defects ensue, including microcephaly, encephalitis, seizures, and blindness, and HCMV is the leading cause of congenital deafness (6, 12, 13).

The fact that HCMV preferentially causes disease in immune compromised individuals highlights the importance of immune control of virus replication. Indeed, many HCMVassociated disease manifestations correlate with viral replication and respond to antiviral drug treatment. However, certain syndromes, particularly chronic diseases, do not typically correlate with high HCMV load (14), suggesting that direct cellular destruction by virus is not the sole cause of tissue damage.

Cytokines participate in immune responses to viruses that activate innate immune responses and orchestrate the development of adaptive antiviral immunity. However, uncontrolled cytokine production can cause off-target effects, participating in various immune-driven pathological processes. Due to the limitations of what can be investigated in humans, the murine CMV (MCMV) model has been used for decades to study mechanisms influencing CMV pathogenesis in vivo, including how cytokines orchestrate antiviral immunity [summarized in detail elsewhere (15)]. Herein, we examine evidence from both clinical studies and experimental models of CMV infection showing that although cytokines are required to limit viral replication, they can cause host damage. We discuss these findings in the context of different tissues where damage during CMV infection can ensue and describe the mechanisms that restrict these harmful processes (see **Figure 1** for summary).

### PRO-INFLAMMATORY CYTOKINES, SYSTEMIC CYTOMEGALOVIRUS-INDUCED DISEASE, AND ORGAN DAMAGE

Cytokine responses during HCMV viremia have been mostly studied in the transplantation setting where time of virus exposure is known. Following initial replication, sustained type 1 cytokine signatures are observed that are characterized by production of IFNγ [in some but not all studies (16)], IL-18 and IL-6, and is further accompanied by acute phase protein and chemokine (IP-10) secretion (16, 17). T-cells are implicated as a significant source of type 1 cytokines (18, 19). Furthermore, numerous pro-inflammatory chemokines and cytokines, including IL-6, are secreted directly following HCMV infection (20). HCMV triggers cytokine production through the stimulation of pattern recognition receptors (PRRs), most notably Toll-like receptor 2 (21), the cytoplasmic DNA sensor STING (22) and IFI16 (23). Mice defective in PRRs mount reduced cytokine responses to MCMV in vivo (24–26). Although differences in the relative contributions of PRRs to the recognition of MCMV and HCMV may exist, these data suggest that innate immune recognition of viral infection by PRRs contributes to HCMV-induced cytokine profiles. Furthermore, in vitro, HCMV stimulation of peripheral blood-derived monocytes increases expression of TLRs, CD14, and adaptor molecules and transcription factors downstream of TLRs (27). Thus, active HCMV replication likely induces systemic pro-inflammatory cytokine responses both following via direct host recognition but also, potentially, by priming the host immune response to respond strongly to unrelated microbial signals.

Given the established role for type 1 cytokines in antiviral immunity, is such a response to CMV infection a bad thing for the host? Certainly, substantial evidence from clinical and experimental studies point toward a protective role for type 1 cytokine responses in cytomegalovirus infections (28–30). However, studies using MCMV show that T-cell responses, particularly CD8<sup>+</sup> T-cells, known to be induced by type 1 cytokines cause substantial tissue damage if insufficiently regulated (31, 32). Also, severe inflammatory cytokine responses or "cytokine storms" occur during MCMV hepatitis (33). Thus, these processes may drive acute HCMV-associated diseases. Furthermore, HCMV is implicated in organ rejection (34, 35) and, in cardiac transplants, graft atherosclerosis (36). Experimental studies using MCMV have recapitulated the observation that acute infection and viral reactivation can influence graft longevity (37, 38). MCMV reactivation induces expression within the graft of IFNα and IL-12 (37), implying that viral infection may elicit cytokine responses that activate cellular immunity capable of mediating graft rejection. Furthermore, HCMV induces IP-10 and fractalkine production during infection (17, 39), both of which are markers of allograft rejection (40).

HCMV establishes life-long infection within multiple host tissues (41) where some genomes are silent but others are transcriptionally active and express many genes (41–43). Immunological data highlights the likelihood that frequent reactivation events occur that re-stimulate the host immune system (44). Subsequently, HCMV may contribute to cytokine mediated inflammatory diseases in latently-infected immune competent individuals via continued gene transcription and reactivation, stimulating immune recognition and subsequent cytokine production. For example, HCMV is implicated in cardiac diseases (45) including atherosclerosis (46) where plaque formation and instability is an inflammatory-driven processes initiated by IFNγ (47). HCMV also induces accumulation of virus-specific cytotoxic CD4<sup>+</sup> T-cells expressing CX3CR1 (48). CX3CR1 binds fractalkine which is expressed by activated endothelium in response to TNFα and IFNγ produced by HCMV-specific T-cells (39). Interestingly, the HCMV chemokine receptor homolog US28 also binds fractalkine (49) and may contribute to localized inflammation. Thus, HCMV-induced cytokine and T-cell responses may mediate endothelial damage that in turn promotes vascular diseases and contributes to damage in multiple tissues and organs. Whether such processes underpin other harmful associations of HCMV, such as increased frailty in elderly individuals (4), is unclear.

Cytokines may also indirectly enhance tissue damage by promoting CMV reactivation and subsequent replication. IL-6 promotes HCMV reactivation in dendritic cells via ERK-MAPK

mediated transcriptional induction of major immediate early (IE) genes (50, 51). TNFα and IL-1β also induce IE gene transcription by latent HCMV (52–54) and are implicated in reactivation of HCMV and/or MCMV in vitro and in vivo (55–59). An additional role for IFNγ in initiating HCMV reactivation has been described (56, 58). Data from MCMV suggest that overt pro-inflammatory cytokine responses may also impinge on innate antiviral immunity. Inadequate proinflammatory cytokine regulation can promote activationinduced NK cell death (60, 61) in a process involving IL-6 (60). Thus, inflammatory cytokines may directly and indirectly promote virus replication, which in turn drives peripheral tissue damage.

#### CYTOKINES AND DAMAGE IN IMMUNE PRIVILEGED SITES

When HCMV accesses immune privileged organs, immunemediated pathology can ensue. HCMV-induced retinitis is a significant problem in AIDS/HIV patients (7–9). Interestingly, elevated expression of type 1 cytokines including IL-6 and IFNγ in aqueous and/or vitreous fluids from patients is detectable (62– 64). Systemic CMV infection in immune competent mice induces significant myeloid cell and T-cell infiltrations into ocular tissue including the neural retina (65). Although cytokines likely play a role in mediating these inflammatory processes in immune competent hosts, this has yet to be investigated.

A role for inflammation in HCMV-induced hearing loss in infants is suggested by autopsies showing inner ear inflammation (66, 67). In mice, systemic infection of newborns induces progressive hearing loss and decreased spiral ganglia neuron density that is indicative of congenital HCMV infection (68). In MCMV, hearing loss does not correlate with the presence of virus in the cochlea but rather associates with persistent expression of chemokines and pro-inflammatory cytokines including TNFα, IL-6, and IL-1β (68). Similarly, intracranial MCMV infection induces hearing loss and chronic inflammatory cytokine expression (69).

Murine neonatal infection models have also been used to recapitulate central nervous system pathology triggered by congenital HCMV infection. After systemic infection, MCMV induces widespread focal encephalitis accompanied by mononuclear inflammation and microglial activation (70, 71), including TNFα expression (72). This is accompanied by STAT1 activation and IFN (type I and II) expression, in addition to TNFα (73). Interestingly, glucocorticoid treatment of these mice reduced cytokine expression and associated morphogenic abnormalities and cellular inflammation without influencing virus load, suggesting that virus-induced inflammation could be safely targeted to improve CMV-induced CNS pathogenesis (73). Indeed, neutralization of TNFα reduced expression of cytokines and myeloid cell activation and accumulation in the brain, and corrected cerebellar abnormalities and developmental gene expression (74). These important studies provide proof-of-concept that anti-inflammatory approaches can be safely utilized to ameliorate CMV pathogenesis in vivo.

HCMV is implicated in esophagitis in HIV-infected individuals and associates with elevated TNFα production (75). Inflammatory bowel diseases are common during HIV co-infection (7) and HCMV maintains active replication in the gastrointestinal epithelium of individuals treated with antiretroviral therapy, where replication disrupts epithelial integrity in a manner partially dependent upon IL-6 (76). HCMV also associates with gastrointestinal inflammation in healthy individuals (77), where the virus may drive local production of cytokines such as TNFα (78) via induction of pattern recognition receptor expression and/or downstream, adaptor molecules (27, 79).

HCMV may also impact on neurological diseases in adults, with associations with HIV-associated neurological disorder (HAND) and impaired cognitive performance in HIV-infected individuals being reported [reviewed in (10)]. The link between HCMV and multiple sclerosis in immune competent hosts is controversial, with contradicting findings regarding the association between HCMV seropositivity and disease occurrence (80–82). In the murine experimental autoimmune encephalomyelitis (EAE) experimental model, MCMV worsens disease in genetically susceptible mice (83) and increases EAE occurrence in resistant (BALB/c) strains. Here, infection increases CD4 T-cell-dependent disease that is associated with IFNγ- and IL-17-expressing T-cells (84), further demonstrating that CMV can exacerbate tissue damage in the central nervous system.

Like many herpesviruses, HCMV is implicated as a risk factor in Alzheimer's Disease (AD) and cognitive decline (85). PBMCs from HCMV seropositive AD subjects produce more IFNγ following polyclonal and viral protein stimulation than non-AD subjects (86), and IFNγ is detectable only in cerebrospinal fluid of HCMV seropositive but not seronegative AD patients (87). Thus, although the role of HCMV in AD development is controversial, (88) it appears that HCMV-infected AD sufferers exhibit heightened cytokine responses which in turn could contribute to disease development and/or progression.

### REGULATION OF CYTOKINE-DRIVEN CMV-INDUCED PATHOGENESIS

Despite its inflammatory potential, HCMV rarely causes inflammatory conditions in healthy individuals. Furthermore, infection in immune compromised and immunologically immature hosts does not always cause overt tissue damage, suggesting that virus-induced inflammatory cytokine responses are tightly regulated.

#### Regulatory T-Cells

The association between inducible regulatory T-cell (iTregs) expansions and reduced vascular pathology in elderly HCMVinfected individuals suggests a protective function for Tregs in HCMV infection (89). In MCMV, Tregs (promoted by IL-33) restrict liver pathology following systemic MCMV infection (32) and chronic reactive gliosis triggered by MCMV encephalitis (90). Although hepatic Tregs are known to be dependent upon IL-33 (32), whether Treg-mediated control of pathogenic T-cell responses involves restriction of inflammatory cytokine secretion is currently unknown.

### Cytokines

Inflammatory cytokine responses during acute HCMV infection are accompanied by secretion of the immune modulatory cytokine IL-10 (16, 91). HCMV re-programmes human hematopoietic progenitor cells (HPCs) into immune-suppressive monocytes that express IL-10 in a process requiring US28 (92). In mice, genetic and pharmacological targeting of IL-10 demonstrates that IL-10 limits systemic inflammatory cytokine responses induced by CMV, including IL-6 and TNFα (61, 93, 94). This alleviates MCMV-induced disease, assessed using body weight (93, 94), and weight loss in IL-10−/<sup>−</sup> mice is alleviated by TNFα neutralization (93). IL-10 also restricts MCMV-induced hepatic inflammation and preserves liver function by limiting inflammatory effector cell infiltration, hepatocyte apoptosis and necrosis (95, 96). Experiments performed in perforin-deficient mice that are unable to control MCMV replication reveal that IL-10 restricts liver inflammation primarily by limiting pathogenic CD8<sup>+</sup> T-cell responses (31), a conclusion supported by data derived from immune competent Il-10−/<sup>−</sup> mice (95). Following injection of MCMV into the brain, IL-10 limits fatal immunopathology characterized by proinflammatory cytokine production and neutrophil infiltration (97, 98). Although the physiological relevance of some of these experiments in terms of HCMV pathogenesis is unclear, these data clearly highlight that IL-10R signaling can suppress CMV-induced immune pathology.

Importantly, genetic variation within the human IL-10 gene correlates with altered HCMV disease occurrence following allogeneic stem cell transplantation (99) and during HIV coinfection (100). This suggests that host genetic variation may influence tissue damage caused by HCMV-induced cytokines. Furthermore, HCMV encodes a functional IL-10 otholog (UL111A, vIL-10) that is expressed in lytic replication (101) and an alternate isoform in latency [LAcmvIL-10 (102)]. vIL-10 suppresses numerous innate and adaptive host immune responses including pro-inflammatory cytokine secretion (103, 104). Given that cellular IL-10 promotes MCMV carriage (93, 105–107), one may predict that HCMV vIL-10 facilitates virus persistence. However, using rhesus macaque CMV (rhCMV) that, like HCMV but not MCMV, expresses UL111A, it has been demonstrated that vIL-10 restricts acute inflammation at the initial site of infection, the skin. Interestingly, UL111A had no obvious impact on virus shedding in these experiments. This implies that virus persistence may not be influenced by UL111A in vivo (108) but instead that restriction of tissue pathology is an important function of viral IL-10 orthologs and perhaps other immune evasion gene products expressed by HCMV. Intriguingly, certain clinically-isolated HCMV strains have disrupted UL111A genes (109, 110). It will be interesting to investigate whether these HCMV strains preferentially associate with overt inflammatory responses.

IL-27 is an IL-12 family member that restricts numerous infection-induced pathologies (111). IL-27 facilitates MCMV persistence in the mucosa by suppressing IFNγ <sup>+</sup> (107) and/or cytotoxic (112) CD4<sup>+</sup> T-cells. Given that cytotoxic CD4<sup>+</sup> T-cells are implicated in tissue damage (113), IL-27-faciliated shedding of virions may be a necessary evil to restrict the development of these cells. Data regarding the function of IL-27 during HCMV infection is limited. Spector and colleagues identified that IL-27 limits IFNγ expression by virus-specific T-cells in HIV<sup>+</sup> and HIV<sup>−</sup> HCMV-infected individuals. This was accompanied by IL-27-mediated induction of IL-10 secreting CD4<sup>+</sup> T-cells (114). Whether IL-27 also alters the development of HCMV-specific cytotoxic T-cells is unknown. However, overall these data are consistent with the idea that IL-27 restricts chronic tissue damage by limiting HCMV-specific T-cell responses.

Data from HCMV and MCMV highlights that the cytokine TNF-related apoptosis-inducing ligand (TRAIL) contributes to control of virus replication (115–117). During persistent MCMV infection in the salivary glands, however, TRAIL expression by NK cells restricts pathogenic CD4<sup>+</sup> T-cell responses in this tissue. TRAIL-deficient mice exhibit hallmarks of Sjogren's syndrome (SS), an autoimmune disease of the salivary glands that is characterized by ectopic germinal centerlike structures in the glands, elevated autoantibody production and impaired saliva secretion (113). Thus, TRAIL can limit both viral replication and potentially harmful infection-induced inflammatory responses.

#### Antiviral Restriction Factors

Interferon induced transmembrane protein 3 (IFITM3) is an antiviral restriction factor that inhibits endocytosisdependent cell entry of numerous viruses (118). IFITM3 polymorphisms associated with reduced function are linked to increased risk of severe viral pathogenesis, most notably influenza-induced disease (119–121). Although IFITM3 does not directly impinge on either MCMV or HCMV replication (60, 122), Ifitm3−/<sup>−</sup> mice are dramatically more susceptible to MCMV-driven pathogenesis (60). Disease, which can be fatal, consists of extensive weight loss, transient pulmonary and hepatic mononuclear inflammation, and extensive and irreversible splenic damage. Blocking the action of IL-6 alleviates pathogenesis in MCMV-infected Ifitm3−/<sup>−</sup> mice and also inhibits activation-induced NK cell death and promotes NK cell immunity (60). Thus, it is unclear whether IL-6 drives CMV-induced pathology by promoting tissue damage and/or by impairing cellular antiviral innate immune responses and subsequent control of virus replication. Irrespective, these data again highlight the possible role for genetics in determining host cytokine responsiveness to HCMV and the subsequent disease outcome.

#### Glucocorticoids

Endogenous glucocorticoids are steroid hormones produced in the adrenal cortex following activation of the hypothalamicpituitary-adrenal (HPA) axis. Initial inflammatory cytokine responses during acute MCMV infection are accompanied by robust glucocorticoid production (123, 124), the maximal release of which is dependent upon virus-induced IL-6 (123). The importance of glucocorticoids in modulating CMV-induced pathogenesis is highlighted in studies where mice are rendered globally deficient in glucocorticoids by adrenalectomy and display increased production of pro-inflammatory cytokines and susceptibility to TNFα-mediated lethal disease (125). Furthermore, glucocorticoid receptor signaling in NK cells, via an axis involving the inhibitory PD-1 receptor, exerts tissue-specific regulation of IFNγ production. Here, unrestricted NK cell expression of IFNγ in spleens of mice lacking the glucocorticoid receptor in NCR1<sup>+</sup> cells results in necrotizing splenitis and destruction of the white pulp (124). Although pathology in medically important sites of CMV pathogenesis like the liver were unaffected by this process (124), these data suggest that neuro-immune pathways may be critical for control of cytokine-driven pathogenesis during CMV infection.

### CONCLUSIONS

Many associations exist between production of inflammatory cytokines and CMV-associated pathologies in humans and in experimental systems. Experimental models like MCMV have their limitations in terms of variations in virus genetics (including lacking key immune regulatory genes like vIL-10) and the imperfect recreation in mice of HCMV-induced pathologies. However, important predictions regarding roles that cytokines play in virus-induced tissue damage and how inflammatory cytokines are regulated can be derived from these studies. Moving forward, these models will be critical to examine whether targeting CMV-induced inflammation is an effective, safe and viable approach to alleviating pathogenesis. Understanding exactly how cytokines cause tissue damage and how production of these cytokines is regulated will hopefully lead to more refined and effective strategies to help alleviate the pathological consequences of HCMV infection. These studies may also help identify host genetic variations that influence cytokine responsiveness and susceptibility to HCMV disease. Finally, these studies may help form novel hypotheses regarding the possible influence of genetic variation in virus-encoded immune evasion genes on HCMV pathogenesis.

### AUTHOR CONTRIBUTIONS

IRH defined the manuscript focus and structure. MC and IRH wrote and edited the manuscript.

### FUNDING

This work was funded by a Welcome Trust Senior Research Fellowship to Ian Humphreys (207503/Z/17/Z).

## ACKNOWLEDGMENTS

The authors wish to thank Dr. Matthew Reeves for critical reading of this manuscript.

## REFERENCES


acceptance by latent cytomegalovirus. Am J Transl. (2009) 9:42–53. doi: 10.1111/j.1600-6143.2008.02457.x


immunosuppressive monocytes to achieve latency. Nat Microbiol. (2018) 3:503–13. doi: 10.1038/s41564-018-0131-9


genome diversity highlights the widespread occurrence of gene-disrupting mutations and pervasive recombination. J Virol. (2015). 89:7673–95. doi: 10.1128/JVI.00578-15


**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 Clement and Humphreys. 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.

# Nef-induced CCL2 Expression Contributes to HIV/SIV Brain Invasion and Neuronal Dysfunction

Michael H. Lehmann1,2 \*, Jonas M. Lehmann<sup>3</sup> and Volker Erfle<sup>1</sup>

1 Institute of Virology, Technische Universität München, Munich, Germany, <sup>2</sup> Institute for Infectious Diseases and Zoonoses, Ludwig-Maximilians-Universität München, Munich, Germany, <sup>3</sup> Department of Informatics, Technische Universität München, Munich, Germany

C-C motif chemokine ligand 2 (CCL2) is a chemoattractant for leukocytes including monocytes, T cells, and natural killer cells and it plays an important role in maintaining the integrity and function of the brain. However, there is accumulating evidence that many neurological diseases are attributable to a dysregulation of CCL2 expression. Acquired immune deficiency syndrome (AIDS) encephalopathy is a severe and frequent complication in individuals infected with the human immunodeficiency virus (HIV) or the simian immunodeficiency virus (SIV). The HIV and SIV Nef protein, a progression factor in AIDS pathology, can be transferred by microvesicles including exosomes and tunneling nanotubes (TNT) within the host even to uninfected cells, and Nef can induce CCL2 expression. This review focuses on findings which collectively add new insights on how Nef-induced CCL2 expression contributes to neurotropism and neurovirulence of HIV and SIV and elucidates why adjuvant targeting of CCL2 could be a therapeutic option for HIV-infected persons.

#### Keywords: AIDS, astrocyte, autophagy, chemokine, dementia, inflammation, neuron, virus

#### INTRODUCTION

Acquired immune deficiency syndrome (AIDS), caused by the human immunodeficiency virus (HIV) (1, 2), has to date resulted in the deaths of over 32 million people. According to the 2019 UNAIDS Global AIDS Update, 1.7 million people became newly infected with HIV in 2018 resulting in a total number of 37.9 million people living with HIV worldwide. To date, there is no effective protective vaccine against HIV or even a feasible cure available for HIV-infected patients (3, 4).

In the mid-1990s combined anti-retroviral therapy (ART) was introduced, which considerably reduced the mortality of HIV-infected patients. However, since then, the prevalence of HIV-associated diseases has increased. A major obstacle toward the development of therapies against these diseases that affect a number of organs such as the heart, lungs, kidneys, and the brain is due mainly to the fact that the disease pathogenesis is poorly understood (5–8). In the meantime, there exists at best a consensus that a systemic and persistent activation of the immune system plays a major role in the disease pathogenesis (9–11). Moreover, it is difficult to accurately differentiate between age-related neurodegeneration, other neurodegenerative diseases and HIV-associated neurocognitive disorders (HAND) (12). However, attempts have been made to identify biomarkers to diagnose neurocognitive impairment in HIV-infected persons and activated

#### Edited by:

Serge Benichou, Centre National de la Recherche Scientifique (CNRS), France

#### Reviewed by:

Laura Fantuzzi, National Institute of Health (ISS), Italy Tetsuo Tsukamoto, Kindai University, Japan Fan Wu, Fudan University, China Matthias Clauss, Indiana University Bloomington, United States

#### \*Correspondence:

Michael H. Lehmann Orlataler@web.de

#### Specialty section:

This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology

Received: 27 June 2019 Accepted: 01 October 2019 Published: 15 October 2019

#### Citation:

Lehmann MH, Lehmann JM and Erfle V (2019) Nef-induced CCL2 Expression Contributes to HIV/SIV Brain Invasion and Neuronal Dysfunction. Front. Immunol. 10:2447. doi: 10.3389/fimmu.2019.02447

**75**

monocytes/macrophages and C-C motif chemokine ligand 2 (CCL2) appear to be the most promising amongst them (13).

CCL2, also named monocyte chemoattractant protein 1, is a chemotactic cytokine for monocytes (14) and T cells (15), which are the main target cells of HIV-1. CCL2 decreases interferonalpha expression (16), and promotes HIV/SIV replication by up-regulation of surface C-X-C motif chemokine receptor 4 expression (17).

CCL2 binds to the C-C motif chemokine receptor 2 (CCR2), which is expressed by neurons (18), human fetal astrocytes (19) and brain microvascular endothelial cells (BMECs) (20). CCL2 also binds to the D6 chemokine decoy receptor, which is expressed on adult human astrocytes (21).

The CCL2-CCR2 axis has been shown to play a key role in multiple sclerosis and in experimental autoimmune encephalomyelitis (22), in addition to exacerbating neuronal damage after status epilepticus (23), eliciting itch- and pain-like behavior in allergic contact dermatitis (24), as well as mediating alcohol-induced neuroinflammation and neurotoxicity (25).

### HIV AND SIV ASSOCIATED DEMENTIA, ENCEPHALITIS AND NEURONAL DAMAGE

Without combined ART, HIV causes dementia which is characterized by deficiencies in cognition, motor disorders, and behavior abnormalities (26). Pathological manifestations of HIV-associated dementia (HAD) appear as meningitis, encephalitis and vacuolar myelopathy (27). A similar clinical picture has been observed in the SIV/macaque model (28, 29). Even after the introduction of combined ART, HAND remain (8), and, in fact SIV-infected macaques treated with suppressive ART also show ongoing neurodegeneration and inflammation (30). The reason for this phenomenon is unknown although several explanations have been proposed, e.g., that anti-retroviral drugs cannot access the central nervous system (CNS), are not effective in eliminating viral reservoirs, or themselves contribute to HAND (31, 32). However, since specific CNS-targeted ART failed to improve neurocognition in HIVinfected patients compared to non-CNS-targeted (33), it has been hypothesized that early events after primary infection with HIV/SIV are critical for initiating the development of HAND (8).

Indeed, SIV was detected in the brains of macaques within a few days after intravenous infection (34, 35). Further, HIV nucleic acid was detected in the brain of an HIV naïve patient who died 15 days after intravenous inoculation of indium-111 labeled white blood cells, which originated from an HIV-infected individual (36). Additionally, a more recent study showed that HIV RNA is present in the cerebrospinal fluid (CSF) of humans as early as 8 days after HIV infection (37). This suggests that HIV/SIV is capable of exploiting a distinct mechanism to enter the brain rapidly.

Entry of SIV into the brain and induction of neuropathology does not appear to depend on a sustained high viral load because the SIVmac32H(pC8) strain, whose replication is attenuated in vivo (38, 39), was detected in the brain 3 days after infection of macaques where it caused persisting neuroinflammation (40). The attenuated phenotype of SIVmac32H(pC8) is most probably due to a 12 base-pair deletion in its nef gene, which results in an in-frame deletion of the amino acids 143–146 of the translational product (38). Although this Nef variant was detected at lower levels in vitro compared to other variants (41), it was definitely detected in the brain of macaques infected with SIVmac32H(pC8) (42). However, SIV strains containing nucleotide deletions in the nef long-terminal repeat (nef/LTR) overlap region, analogous to the HIV strain of the Sydney blood bank cohort (SBBC), could not be detected in the brains of macaques despite viral replication in the periphery (43). Of note, members of the SBBC who had become infected with an HIV strain containing the nucleotide sequence deletions in the nef/LTR region that results in a truncated Nef protein of 24 amino acids (44), did not or only slowly progressed to AIDS including HAD (45).

### THE NEF PROTEIN OF HIV/SIV: IMPORTANCE FOR AIDS PROGRESSION AND ITS INTERCELLULAR TRANSFER

The importance of Nef for AIDS progression was confirmed in SIV-infected rhesus monkeys and HIV-transgenic mice (46, 47). Additionally, it was shown that Nef is required for high viral load in vivo (47). These findings have stimulated a series of studies aiming to identify the mechanistic background with the ultimate goal to exploit the knowledge for therapeutic intervention. Indeed, numerous cellular interaction partners and pathophysiological functions of Nef have been detected (48, 49), and several models of how Nef executes its role in HIV/SIV replication and immunopathogenesis have been proposed (50). In 2009, Kyei et al. showed that HIV Nef inhibits autophagic maturation in human macrophages and thereby provided a convincing explanation of how Nef acts at the molecular level to enable efficient replication of HIV (51). Inhibition of autophagy increases the production of proinflammatory cytokines (52, 53) including CCL2 (54, 55). Thus, Nef also seems to contribute to chronic inflammation, which occurs in HIVinfected persons (56).

HIV Nef was found in supernatants of nef-expressing BHK cells (57), yeast (58), and HEK293 cells (59), which was surprising at the time of these discoveries because nef does not code for an N-terminal signal sequence that would direct the protein to the cell secretory pathway leading to export. Thus, the mechanism by which Nef is released from infected cells was regarded as an open question. In the past, on analyzing the supernatants of BHK cells infected with recombinant vaccinia virus expressing HIV Nef, it was assumed that Nef could be released by vesicles (57). Today, it is recognized that not only proteins but also lipids and RNA can be released from a cell by extracellular vesicles (60).

In 2003, it was shown that HIV Nef induces an accumulation of multivesicular bodies (MVBs) and that Nef itself is present in MVBs (61). MVBs can fuse with the cell plasma membrane, leading to the release of 40–90 nm diameter vesicles, termed exosomes, into the extracellular environment (62). Consequently, it was tempting to speculate that Nef could be released from cells by exosomes. However, it was challenging to test this hypothesis in HIV-infected cells because Nef is incorporated in virions (63, 64).

The astrocytoma cell line TH4-7-5 is persistently infected with the HIV isolate TH4-7-5 which has a mutation in the nef gene (GenBank accession number: L31963.1), resulting in a myristoylation-deficient Nef (65). However, myristoylation of Nef is required for optimal HIV replication in vitro (66). Thus, a myristoylation-deficient Nef and a block in HIV Rev function most probably effected a very low production of infectious virus but a high production of Nef in astrocytoma TH4-7-5 cells (65, 67). We took advantage of the astrocytoma cell line TH4-7-5 and examined whether Nef is present in the supernatants of these cells. Application of a two-step centrifugation protocol, previously shown to enable the enrichment of microvesicles including exosomes from cellular supernatants (68), resulted in the detection of Nef in the pellets of centrifuged supernatants of these cells (69).

It was later confirmed that Nef is released from HIV-infected cells (70, 71) by microvesicles and it was even claimed that this occurs via exosomes (72). However, there is still an ongoing debate regarding the type of vesicle by which Nef leaves the cell (73, 74). Further, Nef was detected in microvesicles and exosomes isolated from the plasma of HIV-infected persons despite them receiving ART, and it has been shown that exosomes derived from HAD patients can transfer nef mRNA to cells, leading to Nef expression and subsequent induction of cellular genes (75, 76).

Nef was also found in uninfected human peripheral blood mononuclear cells (PBMCs), which can transfer Nef to human umbilical cord vein endothelial cells (77). A recent study not only reported that Nef is released by vesicles from HIVinfected cells, but also confirmed the result for SIV-infected cells and has additionally shown that extracellular vesicles containing Nef circulate in the blood of SIV-infected macaques (78). Meanwhile, the process of protein and mRNA transfer by exosomes and other extracellular vesicles even between different types of cells is well understood (79). In summary, irrespective of the type of extracellular vesicle from which Nef is released by HIV/SIV infected cells, Nef is present in the extracellular environment independently of virions and can enter uninfected cells where it affects cellular functions and gene expression.

Additionally, cells can exchange molecules and organelles directly via tunneling nanotubes (TNTs), which are about 50– 200 nm long thin actin rich membrane conduits, even between different types of cells (80). Nef can induce TNT formation (81, 82), and it can also be transferred to B cells via TNTs from HIV-infected macrophages (83), from macrophages to T cells (82), from nef-expressing T cells to hepatocytic cells (84) and also between macrophages (81). Importantly, Nef is transferred from T cells and monocytes to human coronary arterial endothelial cells via TNTs, leading to apoptosis and CCL2 expression (85).

### NEF-INDUCED CCL2 EXPRESSION AND THE FUNCTION OF THE BLOOD-BRAIN-BARRIER

CCL2 increases the blood-brain barrier (BBB) permeability (86, 87), and much progress has been made in revealing the molecular mechanism of how leukocytes, governed by CCL2, pass the BBB (88). Therein, astrocyte and BMEC-derived CCL2 play complementary roles (89).

A natural repair mechanism to restore damaged brain tissue after experimentally-induced ischemia starts with the recruitment of CCR2+Iba1<sup>+</sup> monocytes from the periphery, which then differentiate into brain Iba1+NG2<sup>+</sup> cells within the brain parenchyma (90, 91). Transmigration of CCR2+Iba1<sup>+</sup> monocytes through the BBB is enabled by a transient expression of CCL2 in astrocytes and endothelial cells that lasts for only 2 days (92). Indeed, under normal physiological conditions, the BBB is impermeable for circulating monocytes (93, 94), and therefore invasion of a healthy brain by HIV/SIV should not happen as fast as it has been observed. But a specific HIV/SIVtriggered mechanism leading to CCL2 expression in BMECs may enable HIV/SIV to get access to the brain either in the form of free virions or via infected CCR2<sup>+</sup> cells. In this respect, it was significant to observe that Nef (i) can be transferred from human PBMC to human endothelial cells (77), (ii) was detected in endothelial cells of nef-transgenic mice and macaques infected with SHIV-nefSF33, and (iii) induces CCL2 expression in endothelial cells (85).

Invasion of the brain by leukocytes would additionally require an upregulation of adhesion molecules on both endothelial and infected cells. Indeed, it has been shown that HIV Nef upregulates the intercellular adhesion molecule 1 (ICAM-1) in vascular endothelial cells (95). ICAM-1 interacts with the lymphocyte function-associated antigen 1 (LFA-1), and its subunits, CD11a and CD18, are upregulated in HIV-infected monocytes (96, 97). Endothelial-derived CCL2 activates CD11a, leading to a firm arrest of monocytes on endothelial cells (98, 99), and mediates the subsequent transendothelial migration (100).

In summary, the findings collectively result in a model in which Nef-containing PBMCs and extracellular vesicles carrying Nef attach to and transfer Nef into endothelial cells, leading to CCL2 production that can cause BBB leakiness and subsequent entry of HIV/SIV by infected cells into the brain (**Figure 1**). Of note, this provides a simple explanation of why SIV with a deleted nef gene cannot enter the brain (43).

### NEF-INDUCED CCL2 EXPRESSION AND NEURONAL DYSFUNCTIONS

Once in the brain, HIV/SIV cannot be eliminated by ART. The virus persists and triggers a chronic inflammation leading to sustained leukocyte infiltration, astrogliosis and neuronal degeneration (102, 103). In brain tissues of HIV-infected patients, HIV DNA was detected in the cells of the macrophage lineage and in astrocytes, the most abundant cell type in the brain. However, it was not found in neurons (104), which is in accordance with

the finding that perivascular macrophages and microglia, but not neurons, can be productively infected with HIV/SIV (105, 106). These findings indicate that an indirect mechanism causes neuronal dysfunction and damage, and microglia that release exosomes and microvesicles containing Nef (107) may play an important role therein. It has long been known that HIV and SIV antigens are present in astrocytes of primary infected tissues (106, 108). Recently, a hypothesis was proposed that explains this finding (109) and challenges the consensus that HIV/SIV can infect astrocytes (110).

Significantly, Nef is highly expressed in astrocytes (111), promotes replication of HIV (112), and is also released by exosomes (113) or any other extracellular vesicle (114). Human astrocytes infected with recombinant Sindbis virus vector encoding HIV nef produced elevated CCL2 mRNA levels, which was independent of the nef variant tested (115). Induction of CCL2 expression by HIV Nef was confirmed in U-251 MG astroglioma cells transfected stably with nef (116), in primary rat astrocytes in vivo (117), and in primary murine macrophages and microglia (118). Animal models have provided evidence that there is a direct link between Nef-induced CCL2 expression and neuronal dysfunction and damage. Macrophages expressing HIV Nef, which were implanted into the rat hippocampus, triggered immigration of monocytes/macrophages, tumor necrosis factor expression, and astrogliosis, a hallmark of HIV encephalitis (HIVE). In addition, the neurotoxicity triggered by Nef was associated with cognitive deficits (119). Cognitive deficits in particular spatial and recognition memory were observed in rat brains in which primary astrocytes were implanted that expressed HIV Nef. This was associated with Nef-induced CCL2 expression, which resulted in immigration of macrophages in the hippocampus and loss of hippocampal CA3 neurons in these animals (117). In transgenic mice, in which HIV Nef was expressed specifically in macrophages and microglia, CCL2 was increased in the brain, and the dopamine system was affected, leading to mania-like behavior, especially in males (118).

There are several studies demonstrating that increased CCL2 concentrations correlate with HAD/HAND. Elevated levels of CCL2 were detected in the CSF of HIV-infected individuals positively diagnosed with HAD (120, 121). Microglia and astrocytes of HIV-infected persons suffering from HIVE produce CCL2 (122), which was confirmed for SIV infected macaques (123). Additionally, a specific small nucleotide polymorphism in the CCL2 promoter, which leads to increased CCL2 expression and infiltration of mononuclear phagocytes into tissues correlates positively with the risk of HAD (124). Cocaine, known to exacerbate neurodegeneration in persons infected with HIV, induces CCL2 expression in microglia and leads to increased transmigration of monocytes into the brain (125).

It is now also known that CCL2 affects neurons directly in addition to enhancing the transmigration of infected leukocytes through the BBB (126). For example, over-induction of CCL2 in astrocytes causes dopaminergic neurodegeneration in 1 methyl-4-phenyl-1,2,3,6-tetrahydropyridine mice (127), and an inhibition of CCL2 expression protects neurons against amyloidbeta-induced toxicity (128). Indeed, CCL2 mediates cell death

in neurons of the hippocampal CA3 region after kainic acidinduced seizures in mice. Neuronal degeneration was associated with behavioral impairment, memory decline, and anxiety (129), all characteristics which have been observed early after infection of humans with HIV and even in HIV-infected persons receiving ART (130–132).

CCR2, the receptor of CCL2, is present on neurons (18), and its absence reduced brain damage as well as BBB permeability in an experimental stroke model in mice (133). Similar to the process in an HIV/SIV infection, CCR2 plays a key role in the accumulation of myeloid cells in the brain and the activation of hippocampal myeloid cells upon infection with Theiler's murine encephalitis virus (TMV). Notably, CCR2 deficient mice had almost no hippocampal damage during TMV infection (134). Thus, CCL2 represents a convincing candidate to explain neuronal dysfunction and damage (135) which occur in HIV/SIV infected humans and animals (**Figure 2**). Additionally, CCL2 is major mediator of pain (137), and chronic pain is a common burden in people living with HIV/AIDS (138).

#### SUMMARY

The findings summarized herein not only integrate well into the "Trojan horse" model that states that a cell infected with HIV/SIV enters the brain leading to a persistent infection and consequently HAND (139) but also add to this model the fact that the transfer of Nef by microvesicles into endothelial cells and the subsequent induction of CCL2, mimics a pathophysiological state of the brain to which monocytes are recruited normally. Nef, in combination with other HIV/SIV proteins and even antiretroviral drugs, possibly work together more efficiently to enable a rapid entry of HIV/SIV-infected cells into the brain (140). This interplay presumably plays a general role in HIV-associated diseases (141).

In the brain, HIV/SIV-infected cells such as astrocytes and microglia distribute Nef to uninfected cells via microvesicles and TNTs. Thereby, there is a steady increase in the number of Nefbearing, non-infected cells which produce CCL2. HIV Tat in astrocytes seems to contribute to an increase in the levels of

#### REFERENCES


CCL2 in the brain (142, 143). The persistent non-physiological expression of CCL2 leads to sustained cell infiltration into the brain and a disturbance of neuronal functions. If a person is infected with HIV subtype B then Tat could enhance CCR2 activation through its acidic region (144, 145). Moreover, when present in sufficiently high concentrations in the brain, Tat could definitely exacerbate neuronal dysfunctions through its basic region (146). Moreover, besides CCL2, the C-X-C motif chemokine 10 (CXCL10) has also been identified as a biomarker for HAND (13), especially in HIV-infected women (147) and this chemokine can also be induced by Nef (115).

#### CONCLUSION

The findings summarized here classify HIV/SIV Nef-induced CCL2 expression in the complex pathogenesis of HAND, and once again highlight the special role which the CCL2-CCR2 axis can play in a neurological disease. Consequently, drugs which have been developed to target this chemokine or its receptor could also be an option for an adjuvant therapy in HIV-infected persons.

#### AUTHOR CONTRIBUTIONS

ML wrote the initial draft. ML and VE discussed the manuscript. JL, ML, and VE edited the manuscript. JL and ML designed and drew the illustrations.

#### FUNDING

This work was supported by the German Research Foundation (DFG) and the Technical University of Munich (TUM) in the framework of the Open Access Publishing Program.

#### ACKNOWLEDGMENTS

The authors would like to thank Jacqueline Weber-Lehmann for critical reading of the manuscript.


MCP-1: implication for increased monocyte transmigration. Blood. (2010) 115:4951–62. doi: 10.1182/blood-2010-01-266221


vesicles in the spinal cord is a major mediator of local inflammation and pain after peripheral nerve injury. J Neurosci. (2011) 31:5865–75. doi: 10.1523/JNEUROSCI.5986-10.2011


**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 Lehmann, Lehmann and Erfle. 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.

# Fine Tuning the Cytokine Storm by IFN and IL-10 Following Neurotropic Coronavirus Encephalomyelitis

Carine Savarin and Cornelia C. Bergmann\*

*Department of Neuroscience, Cleveland Clinic Foundation, Lerner Research Institute, Cleveland, OH, United States*

The central nervous system (CNS) is vulnerable to several viral infections including herpes viruses, arboviruses and HIV to name a few. While a rapid and effective immune response is essential to limit viral spread and mortality, this anti-viral response needs to be tightly regulated in order to limit immune mediated tissue damage. This balance between effective virus control with limited pathology is especially important due to the highly specialized functions and limited regenerative capacity of neurons, which can be targets of direct virus cytolysis or bystander damage. CNS infection with the neurotropic strain of mouse hepatitis virus (MHV) induces an acute encephalomyelitis associated with focal areas of demyelination, which is sustained during viral persistence. Both innate and adaptive immune cells work in coordination to control virus replication. While type I interferons are essential to limit virus spread associated with early mortality, perforin, and interferon-γ promote further virus clearance in astrocytes/microglia and oligodendrocytes, respectively. Effective control of virus replication is nonetheless associated with tissue damage, characterized by demyelinating lesions. Interestingly, the anti-inflammatory cytokine IL-10 limits expansion of tissue lesions during chronic infection without affecting viral persistence. Thus, effective coordination of pro- and antiinflammatory cytokines is essential during MHV induced encephalomyelitis in order to protect the host against viral infection at a limited cost.

#### Keywords: central nervous system, viral infection, JHMV, IFNα/β, IFNγ, IL-10, demyelination

## INTRODUCTION

The central nervous system (CNS) is susceptible to various neurotropic viral infections associated with acute inflammation. Depending on the distinct anatomical regions infected, inflammation is referred to as meningitis (meninges), encephalitis (brain), myelitis (spinal cord), or meningoencephalitis and encephalomyelitis if multiple sites are afflicted (1). Viral meningitis is overall more clinically benign, whereas encephalitis is associated with clinical evidence of neurological dysfunctions, which can range from behavioral changes to seizures and paralysis. Many encephalitic viruses such as insect borne viruses, enteroviruses, and nonendogenous retroviruses can rapidly invade the CNS early following peripheral infection. However, encephalitis caused by members of the herpes viruses, e.g., Herpes Simplex Virus (HSV)-2, cytomegalovirus (CMV), or the polyomavirus John Cunningham virus (JC virus) are more commonly caused by immune suppression allowing re-activation of otherwise controlled chronic or latent peripheral infections and invasion of, or reactivation within the brain, resulting in severe disability and death (2). For example, premature death of multiple sclerosis patients treated with

#### Edited by:

*Michael H. Lehmann, Ludwig Maximilian University of Munich, Germany*

#### Reviewed by:

*Lauren Alene O'Donnell, Duquesne University, United States Tian Wang, The University of Texas Medical Branch at Galveston, United States*

\*Correspondence:

*Cornelia C. Bergmann bergmac@ccf.org*

#### Specialty section:

*This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology*

Received: *16 October 2018* Accepted: *06 December 2018* Published: *20 December 2018*

#### Citation:

*Savarin C and Bergmann CC (2018) Fine Tuning the Cytokine Storm by IFN and IL-10 Following Neurotropic Coronavirus Encephalomyelitis. Front. Immunol. 9:3022. doi: 10.3389/fimmu.2018.03022* Natalizumab due to JC-virus mediated progressive multifocal leukoencephalopathy emphasizes the importance of CNS immune surveillance to prevent viral recrudescence (3, 4).

As many neurotropic viruses predominantly target highly specialized and/or non-renewable cells controlling cognitive and vital physiological functions, an efficient anti-viral immune response is essential to limit viral CNS dissemination to prevent lethal outcomes. However, the anti-viral immune response needs to be tightly regulated to minimize bystander tissue damage and neurological dysfunction, which can be long term sequela even after virus control (2). Given the limitations in obtaining human CNS samples, several murine models of viral encephalitis provide complementary tools to unravel activation, effector function and regulation of protective immune responses within the CNS; these include Vesicular stomatitis virus (VSV), Sindbis virus, West Nile virus, Theiler's encephalomyelitis virus (TMEV) and mouse hepatitis virus (MHV). This review primarily focuses on encephalomyelitis induced by neurotropic MHV, namely the sublethal glia tropic variant of the John Howard Muller MHV strain, designated v2.2-1, and the non-lethal dual liver and neurotropic MHV-A59 strain (5). Both viruses are characterized by an acute encephalomyelitis which resolves into a persistent infection characterized by demyelination and sustained detection of viral RNA in the absence of infectious virus. As demyelination is immune-mediated and neuronal infection is sparse in the v2.2- 1 model, it provides a useful tool to study the dynamics and regulation of antiviral host immune responses associated with ongoing immune-mediated tissue damage balanced by repair during chronic infection.

#### MOUSE HEPATITIS VIRUS

Mouse hepatitis viruses (MHV), members of the positive-strand RNA enveloped Coronaviridae, are natural murine pathogens that infect the liver, gastrointestinal tract and CNS (6, 7). Virus tropism and pathogenesis depends upon virus strains and variants, as well as inoculation route (8). The attenuated MHV-JHM v2.2-1 referred as v2.2-1 from hereon is a monoclonal antibody derived variant of the lethal MHV-JHM strain (9), which has been extensively used to unravel immune correlates of protection and viral-induced demyelination. Upon intracranial infection the MHV-A59 strain is more neuronotropic than v2.2-1, but also infects glia and causes immune mediated demyelination, although clinical disease severity in immune competent adult infected mice is less severe (10). Unless otherwise stated, this review pertains to encephalomyelitis induced by v2.2-1. Following intracranial administration, v2.2-1 infects the ependymal cells lining the ventricles before spreading to microglia, astrocytes, and oligodendrocytes (OLG); neurons are largely spared. Peak virus replication around day (d) 5 post-infection (p.i.) correlates with activation of astrocytes and microglia, disruption of the blood brain barrier (BBB) and CNS recruitment of neutrophils, NK cells and predominantly bone marrow derived monocytes (6, 11). Monocytes and neutrophils enhance BBB disruption (12) and pave the way for infiltration of T and B cells. T cell recruitment is associated with signs of encephalitis observed around d7 p.i. Both CD8 and CD4 T cells are essential for reducing infectious virus below detectable levels 2 weeks p.i. (6, 13). T cell mediated antiviral function also correlates with onset of demyelination, which peaks 2–3 weeks after control of infectious virus. While virus replication is no longer detectable in chronically infected mice, persisting viral RNA remains present in spinal cords at slowly declining levels. Deprivation of local humoral immunity constitutes the only manipulation resulting in reemergence or lack of clearance of infectious v2.2-1 or A59 virus (14), suggesting virus persists in a replication competent form controlled by local Ab (15).

Induction of cytokines and chemokines, as well as CNS recruitment of innate and adaptive immune cells, is highly regulated during neurotropic MHV infection, emphasizing the orchestration of specific functions at times critical to efficiently control infectious various, while restraining subsequent tissue destruction. This review discusses findings from our colleagues and own laboratories on the role of signature cytokines associated with effective, yet dampened anti-viral responses and limited tissue damage with focus on Interferon (IFN)α/β, IFNγ and IL-10.

#### TYPE I IFN: CONDUCTOR OF THE EARLY ANTI-VIRAL RESPONSE

The induction of innate immune responses, including type I IFNs, provides the first critical line of immune defense in stemming viral spread throughout the CNS (16, 17). Although coronaviruses are known to be poor IFNα/β inducers, the importance of IFNα/β signaling following both MHV-A59 and v2.2-1 infection, became apparent following infection of IFNα/β receptor deficient (IFNAR−/−) mice. Uncontrolled viral replication, extensive viral dissemination throughout the CNS, and expanded tropism to neurons coincided with rapid mortality (18, 19). Early viral replication also induces cytokines and chemokines, some of which are IFNα/β dependent (20). Together, the early response regulates the adaptive immune response essential for reducing viral replication.

Since the naïve CNS is devoid of plasmacytoid dendritic cells, potent peripheral IFNα/β inducers, IFNα/β production relies on sensing of virus invasion by glial and neuronal cells. Although glia and neurons are known to express pattern recognition receptors (PRRs), which recognize diverse pathogen associated molecular patterns (PAMPs) and endogenous danger signals (DAMPs), the diversity and magnitude varies not only between CNS cell type, but also their regional anatomical localization within the CNS (2, 21–23). While all CNS cell types have been shown to be capable of producing IFNα/β in vitro, the ability to induce IFNα/β in vivo depends on the specific virus, its replication cycle, cellular tropism and respective repertoire of PRRs and associated signaling factors. The disparities between CNS cells in their ability to produce and respond to IFNα/β in vivo have recently been reviewed (20). Our own studies with v2.2-1 revealed that oligodendrocytes (OLG) are poor inducers of IFNα/β relative to microglia consistent with low basal levels and limited diversity of PRRs detecting viral RNAs (24). The low expression of IFNα/β receptor chains further coincides with reduced and delayed expression of interferon sensitive genes (ISG) encoding factors with anti-viral activity, including interferon-induced protein with tetratricopeptide repeats 1 and 2 (Ifit1 and Ifit2). Both their reduced ability to establish an antiviral state and upregulate IFNα/β-induced major histocompatibility complex (MHC) class I presentation components may enhance their propensity to become the predominantly infected glia cells and set the stage for establishment of persistent infection (24, 25).

Cell types, which are not effective initial type I IFN inducers, may nevertheless be protected after inducing ISG, which also include PRRs, in response to IFNα/β produced by heterologous cells. Similar to OLG, lower constitutive PRR, and ISG levels were found in astrocytes relative to microglia. However, studies with MHV-A59 revealed delayed but substantial upregulation of IFNα/β pathway genes within astrocytes following infection (26). Some PRRs, ISGs and IFNα were even expressed at higher levels in astrocytes at d5 p.i. compared to microglia, indicating that astrocytes are critical to the innate antiviral activity through amplification of the IFNα/β response. The importance of IFNα/β signaling within astrocytes was confirmed by uncontrolled viral replication and premature death (1 week p.i.) of mice lacking IFNAR expression specifically on astrocytes (26). However, delayed mortality compared to total IFNAR deficiency indicated that other CNS cells, presumably microglia, contribute early to limiting virus dissemination. Analysis using the v2.2-1 virus will determine whether the astrocytic contribution to IFNAR mediated protection remains similar in a model with sparse astrocyte infection.

Altogether, these data shed light on the individual in vivo contribution of glial cells in overall IFNα/β mediated early protection against MHV CNS infection. More studies using conditional ablation of IFNAR and selected ISGs in various encephalitic virus models will be beneficial in unraveling the importance of autocrine and paracrine protective IFNα/β effects on subsequent adaptive responses and potential establishment of cell type specific persistence.

#### IFNγ AND PERFORIN: WHEN ADAPTIVE IMMUNITY TAKES THE RELAY

Although innate anti-viral immune responses are critical in containing initial CNS virus spread, virus-specific T cell effector functions are essential to eliminate or reduce infectious virus load during most acute infections (27–29). Importantly, CNS cells appear to shape the adaptive immune response to avert direct T cell cytolytic effector mechanisms, especially targeted to neurons, as recently reviewed by Miller at al. (2). While various mechanisms, including intrinsic deviation from cellular targets of lytic granules, T cell inhibitory molecules, as well as antiinflammatory factors have been demonstrated to dampen T cell effector functions, the same mechanisms also favor establishment of persistent infection.

The requirement for adaptive immune responses to control neurotropic MHV was evidenced by uncontrolled viral replication and mortality of v2.2-1 infected immunodeficient Rag2−/<sup>−</sup> or SCID mice (30, 31). However, the absence of adaptive immunity also revealed that virus itself does not cause demyelination (6, 9, 32), supporting T cell effector function in mediating pathology. T cell depletion studies subsequently revealed that v2.2-1 control required both CD4<sup>+</sup> and CD8<sup>+</sup> T cells, with CD4<sup>+</sup> T cells providing helper function for CD8<sup>+</sup> T cells, which are the primary effector T cells within the CNS (13, 33). Efforts to define prominent anti-viral effector function further demonstrated that mice deficient in perforin-mediated cytolysis could not control viral replication in microglia and astrocytes, while virus control in oligodendrocytes (OLG) was unaffected (34). In contrast, IFNγ <sup>−</sup>/<sup>−</sup> mice exhibited loss of viral control specifically in OLG (35). The requirement for IFNγ mediated control in OLG was further confirmed by specifically abrogating IFNγ receptor signaling in OLG (36). These data thus demonstrated that T cell mechanisms affecting viral control in vivo were clearly cell type dependent, although CD8<sup>+</sup> T cells isolated from the infected CNS exerted both potent cytolytic activity and produced IFNγ ex vivo. The distinct susceptibilities of glia cells to CD8<sup>+</sup> T cell effector functions was further confirmed by adoptive transfer of virus-specific CD8<sup>+</sup> T cells deficient in either IFNγ or perforin into infected T cell-deficient mice (13, 31). The overall higher dependency on IFNγ for MHV control may also reside in the differential dependence of glia on IFNγ to upregulate MHC class I and antigen processing components. Whereas, class I surface expression by microglia coincides with IFNα/β expression, OLG appear to require IFNγ to upregulate class I (25). This delayed class I expression coinciding with enhanced expression of the inhibitory receptor B7-H1 may protect OLG from CD8<sup>+</sup> T cell cytolysis (37).

Analysis of the relative contribution of CD8<sup>+</sup> vs. CD4<sup>+</sup> T cells to express IFNγ following v2.2-1 infection surprisingly revealed that CD4<sup>+</sup> T cell express higher levels of IFNγ mRNA at the population levels than CD8<sup>+</sup> T cells (38). However, the APC triggering IFNγ production by CD4<sup>+</sup> T cells have not been identified, but may be meningeal or perivascular DC. CD4<sup>+</sup> T cells can indeed mediate direct anti-viral activity in addition to enhancing CD8<sup>+</sup> T cell migration and survival within the CNS (39). However, adoptive transfer of perforin- or IFNγdeficient CD4<sup>+</sup> T cells into infected immunodeficient recipients revealed that viral control was independent of either anti-viral function (13, 17). Moreover, sparse MHC class II upregulation on microglia in the absence of IFNγ, and lack of MHC class II expression on astrocytes and OLG suggest that CD4<sup>+</sup> T cells contribute to viral control indirectly via a viral antigen cross presenting APC or via an MHC class II-independent mechanisms (17). Cell types presenting viral antigen to activate CD4<sup>+</sup> T or CD8<sup>+</sup> T cells in the CNS in vivo requires more extensive investigation not only in the MHV model, but also models of neuronotropic infection.

Although the anti-viral T cell response is vital to protect the host following neurotropic infection, it induces tissue damage characterized by demyelination and modest axonal damage. A role for cytolytic infection of OLG was discounted based on the lack of tissue damage in immunodeficient mice, as well as restored myelin loss by transfer of virus specific CD4<sup>+</sup> or CD8<sup>+</sup> T cells (7). Direct T cell-mediated cytolysis of OLG is also unlikely given the IFNγ dependent control of infectious virus and difficulties to detect apoptotic OLG (30). Delayed virus control in both perforin−/<sup>−</sup> as well as IFNγ <sup>−</sup>/<sup>−</sup> mice did not alter pathology compared to wt mice, indicating that these effector molecules did not play a role in demyelination (34, 35). Similarly, enhanced OLG infection in the absence of IFNγR signaling in OLG did not result in increased demyelination even in the presence of intact T cell function (36). These studies gave the first indication that IFNγ signaling in OLG, independent of their virus load, does not directly affect demyelination.

The role of IFNγ in demyelination nevertheless still remains unresolved. T cell transfer studies with select virus primed T cell populations further indicate that the source of IFNγ in CD4<sup>+</sup> or CD8<sup>+</sup> T cells influences pathogenesis. Less demyelination after transfer of IFNγ <sup>−</sup>/<sup>−</sup> CD8<sup>+</sup> T cells into RAG−/<sup>−</sup> mice correlated with decreased macrophage/microglia activation and recruitment into white matter areas (40). By contrast, transfer of IFNγ <sup>−</sup>/<sup>−</sup> CD4<sup>+</sup> T cells into RAG−/<sup>−</sup> mice correlated with increased demyelination and mortality (41). The dichotomy of enhanced demyelination in RAG−/<sup>−</sup> recipient of IFNγ −/− CD4<sup>+</sup> T cells, which also exhibit selectively increased OLG infection, is likely due to increased IFNγ-regulated neutrophil infiltration and induction of pathogenic Th17 cells (42–44), which had not been uncovered at the time. Distinct from the later studies, lack of IFNγ production by CD4<sup>+</sup> T cells partially protected SCID recipients from myelin loss, but led to premature mortality (17). Decreased demyelination in SCID recipients of IFNγ <sup>−</sup>/<sup>−</sup> CD4<sup>+</sup> T cells nevertheless also correlated with reduced macrophage infiltration and microglia activation. A direct toxic effect of CD4<sup>+</sup> T cells on OLG is unlikely due to their lack of MHC class II expression. Some inconsistencies between results in RAG−/<sup>−</sup> vs. SCID recipients remain to be resolved and may reside in different genetic backgrounds or activation state of transferred T cells (17, 41). Irrespectively, together these data indicate that while IFNγ is vital to reduce MHV virus load, the side effect of extensive macrophages/microglia activation promotes myelin destruction. On the other hand, the total absence of IFNγ not only enhanced virus load, but also maintained neutrophil function and activated Th17 cells (44), which normally do not play a role during a strongly Th1 skewed response during neurotropic MHV infection. More in depth analysis of the role of IFNγ, specifically its cellular targets, is expected to reveal a better understanding of IFNγ as a major regulator of inflammation by promoting MHC class II and iNOS expression and shaping the composition of CNS inflammatory response by regulating chemokine expression. Although iNOS upregulation and oxidative damage have been implicated as factors contributing to CNS tissue damage during demyelination (45), neither genetic ablation of iNOS or pharmacological inhibition of NO affected viral control, demyelination or mortality following infection with v2.2-1 or the neuro attenuated MHV-OBLV60 (46, 47). By contrast, compounds reducing reactive oxygen species (ROS) reduced neuronal loss and demyelination during MHV-A59 induced optic neuritis (48). The contribution of ROS to pathogenesis thus requires more in depth analysis.

### IL-10: THE GAMEKEEPER OF TISSUE DAMAGE DURING CHRONIC JHMV INFECTION

Incomplete control of neurotropic MHV results in persistent infection characterized by low levels of viral RNA in spinal cord, sustained detection of cytokine and chemokine expression, retention of CD4<sup>+</sup> and CD8<sup>+</sup> T cells and ongoing primary demyelination balanced by remyelination (6, 7, 11). The inability to completely eliminate virus suggested an important host response to dampen myelin loss at the expense of virus persistence. One checkpoint molecule was the T cell inhibitory molecule B7-H1, strongly upregulated on OLG. The severity of tissue destruction within lesions in the absence of B7-H1 coincided with increased mortality, although viral control was accelerated (37). Another molecule counteracting tissue damage is the anti-inflammatory cytokine IL-10, known to be a master regulator of immunity to infection (49) as well as balancing immune responses and neurodegeneration in the brain (50). IL-10 is upregulated during acute v2.2-1 infection, at which time it is mainly produced by CD4<sup>+</sup> and to a lesser extent CD8<sup>+</sup> T cells (51). While IL-10 expression by CD8<sup>+</sup> T cells wanes during persistence, it is maintained by CD4<sup>+</sup> T cells (52, 53). Both Foxp3 regulatory CD4<sup>+</sup> T cells (Tregs) and virus-specific IFNγ <sup>+</sup>IL-10<sup>+</sup> CD4<sup>+</sup> T cells (Tr1) are sources of IL-10 throughout the course of JHMV infection and their role have been recently reviewed by Perlman et al. (54). V2.2-1 infection of IL-10−/<sup>−</sup> mice resulted in faster control of virus replication during acute infection and reduced initial demyelination; surprisingly however, the severity of demyelination increased 2 weeks after viral control without altering viral persistence (55). IL-10 deficiency was also associated with sustained MHC class II expression on Iba1<sup>+</sup> myeloid cells and increased iNOS levels in lesions. These data suggested a critical role of IL-10 in limiting tissue damage, despite similar levels of persisting virus. Increased IL-10 production following CNS infection using an engineered IL-10 expressing v2.2-1 variant also resulted in decreased demyelination while virus clearance was slightly delayed (56).

The confirmation of IL-10 as a critical regulator of demyelination questioned whether Tr1 and Foxp3 Tregs played a distinct role. As IL-10 induction in Tr1 cells is IL-27-dependent, mice deficient in IL-27 signaling (IL-27R−/−) infected with v2.2-1 were analyzed for a role of Tr1 cells (57). Infected IL-27R−/<sup>−</sup> displayed drastically reduced Tr1 cells as anticipated, and significantly reduced IL-10 levels at d7 p.i. consistent with faster viral control, similar to IL-10−/<sup>−</sup> mice. However, impaired IL-27R signaling also correlated with decreased demyelination distinct from the IL-10−/<sup>−</sup> infected mice. While these findings implied that IL-10 mediated suppression of demyelination is Tr1-independent, it is noted that IL-27R−/<sup>−</sup> mice have several other dysregulated immune pathways (58, 59). Switching the focus on Foxp3 Tregs, transfer of naïve Foxp3 Tregs into wt or RAG1−/<sup>−</sup> recipients during acute infection ameliorated tissue damage without affecting virus control (52, 60). These results from a gain of function approach were supported by depletion of CD25<sup>+</sup> Tregs prior to infection, which resulted in increased

demyelination (57). While the effect of Foxp3 Tregs on tissue damage is manifested during chronic infection, their regulatory function may already be initiated during acute infection. Indeed, depletion of Foxp3 Tregs during chronic infection had no effect on the extent of myelin loss (61). Similarly, IL-10 neutralization coincident with CNS infection induced increased demyelination whereas delayed IL-10 inhibition did not affect tissue damage (56). Lastly, although Foxp3 Treg transfer during acute infection decreased CNS tissue damage, they were not detected within the CNS. They rather exerted their functions within CNS draining cervical lymph nodes (CLN) by dampening dendritic cell activation and T cell proliferation (60). These data are consistent with a critical regulatory role of Foxp3 Tregs at the time of initial T cell activation with remote consequences on tissue damage.

Irrespective of Treg effects on effector T cells, increased demyelination in IL-10−/<sup>−</sup> mice correlated with sustained microglia activation and impaired glial scar formation (55). These results supported a local regulatory role of IL-10 acting directly on CNS resident cells. The downregulation of IL-10Rα expression on microglia, yet upregulation on lesion associated astrocytes further highlights the complex dynamics of the CNS environment in responding to IL-10 (55). The identity of the Foxp3 Treg population limiting tissue damage also requires further investigation. A small population of virus-specific Foxp3 Tregs was detected in both CLN and CNS, where they effectively regulated the pro-inflammatory T cell response at both sites (62). Whether these virus-specific Foxp3 Tregs also play a role in directly regulating demyelination remains to be ascertained. Foxp3 Tregs may also prevent tissue damage during chronic

FIGURE 1 | Balance IFN and IL-10 responses determine viral control and pathology. IFNα/β limits viral spread throughout the CNS following MHV infection. The collaboration of microglia as early IFNα/β inducers, and astrocytes as amplifiers of IFNα/β, is crucial to protect from viral dissemination and expanded tropism. The innate response promotes virus-specific T cell recruitment and anti-viral activity critical to eliminate infectious virus below detection limits. CD4<sup>+</sup> T cells enhance CD8<sup>+</sup> T cell functions and survival and exhibit uncharacterized anti-viral activity. Virus-specific CD8<sup>+</sup> T cells eliminate virus using perforin-dependent mechanism in astrocyte/microglia and IFNγ in OLG. CNS T cell recruitment also correlates with initiation of demyelination. Both CD4<sup>+</sup> and CD8<sup>+</sup> T cells participate in tissue destruction by instructing myeloid cells to initiate tissue damage. The adverse effects mediated by the pro-inflammatory anti-viral response are balanced by IL-10, a master regulator of immunity to infection. While the role of IL-10 during acute infection remains unknown, it limits myelin loss during chronic infection without affecting viral persistence. Both Foxp3 Tregs and Tr1 cells produce IL-10, which restrain demyelination by regulating microglia activation and astroglial scar formation. A direct role of Foxp3 Treg on peripheral T cell activation, with remote temporal effects on tissue damage, has been suggested by T cell transfer studies.

MHV infection by limiting the autoimmune response (63). Global Foxp3 Treg depletion during acute infection correlated with increased proliferation of transferred self-reactive T cells within both CLN and CNS (64). A correlation with potential expansion of demyelinated lesions was however not evaluated. The interplay of various IL-10 secreting Tregs acting at specific sites and on selective target cells at critical time points emphasizes the complex role of IL-10 in dampening JHMV-induced tissue damage without affecting viral clearance and persistence.

Pronounced effects of IL-10 on pathogenesis and clinical outcome rather than viral control in the CNS are also clearly evident in other viral encephalitis models. In the TMEVmediated transient polioencephalitis model using SJL mice, peak virus load in the hippocampus coincides with peak expression of IL-10, IL-10ra, and relates genes. IL-10R neutralization resulted in increased loss of mature neurons and axonal damage, which correlated with enhanced inflammation, although virus load was not altered (65). Further, increased accumulation of Foxp3 Tregs and arginase-1 expressing microglia/macrophages suggested unsuccessful efforts of the host to compensate for the abrogated IL-10 signaling. IL-10 signaling also protects from CNS damage in mice infected with a virulent strain of the mosquito borne alphavirus Sindbis virus by mitigating detrimental Th17 cell functions (66). By contrast, using a more attenuated Sindbis virus, IL-10 deficiency led to longer morbidity, higher mortality, and delayed viral clearance without affecting Th17 cells. Morbidity was rather associated with increased Th1 and decreased Th2 T cells and delayed humoral immunity (67). Along with TNF-α and IL-2, IL-10 is also a key factor for disease remission from fatal encephalitis due to infection with Oshima strain of Tick born encephalitis virus (68). In a murine model of Japanese encephalitis virus infection, elevated IL-10 and reduced IFNγ also correlated with better survival (69). Lastly, IL-10 treatment has been shown to reduce levels of proinflammatory cytokines and infiltrate in murine HSV keratitis without impairing viral clearance (70). In vivo results further suggest that IL-10 has the ability to regulate microglial cell production of immune mediators and thereby dampen the proinflammatory response to HSV-1 (71).

#### CONCLUSION

Animal models of viral CNS infection have been crucial in revealing mechanisms of viral control, establishment of persistence and tissue damage. A common theme, not only applying to neurotropic MHV encephalomyelitis, are the protective activities of IFNα/β signaling in limiting initial viral dissemination and predominantly non-cytolytic T cell effector functions in reducing infectious virus load (1, 2). While some viruses are cytolytic to their target cells, the immune response also actively contributes to bystander damage manifested in glia and neuronal dysfunction or demyelination associated with axonal damage. The neurotropic MHV model specifically highlights the critical role of IFNα/β signaling in a single cell type in stemming overwhelming viral dissemination despite no evident defects in T cell function (**Figure 1**). It further demonstrates that maximal T cell anti-viral activity during acute infection coincides with maximal anti-inflammatory IL-10 expression, suggesting that an overaggressive adaptive immune response is already counterbalanced during the viral clearance phase, and does not necessarily emerge as a result of tissue damage (**Figure 1**). Most importantly, the lack of this antiinflammatory activity can manifest in exacerbated tissue damage remote from acute infection. An immune mediated imbalance early during encephalomyelitis may thus also explain distinct severities of neurological sequelae following human viral disease. For example, IL-6 and IFNγ levels in CSF may be associated with enterovirus (EV)71-induced neuropathology (72). Further, analysis of serum and CSF samples from patients with acute encephalitis syndrome, including with Japanese encephalitis virus supported that higher IL-10 levels in both serum and CSF correlates with protection (73). Similarly, a distinct study of encephalitis patients, including a subcohort with HSV-1, revealed that IL-10 levels were associated with a better coma score on admission in the overall cohort. Elevated IL-10 levels were also associated with a lesser degree of BBB permeability (74). IL-10 signaling also supports BBB integrity following traumatic CNS injury in rodent models (75). With respect to human virus induced encephalitis, it is also interesting to note IL-10 gene polymorphisms as potential susceptibility factors (76). Mutations in IL-10Ra have also been identified as a risk factor of severe influenza-associated encephalopathy (77).

The imprinting of the innate immune response on subsequent adaptive immunity and its effects on bystander cells such as microglia and infiltrating myeloid cells make it difficult to tease apart critical checkpoints determining disease progression or resolution. However, the availability of numerous conditional knockout mice blocking cytokine responses in distinct cell types and in a temporal fashion promise to shed more light on pathways ameliorating pathology while preserving viral control. Confirmation of similar pathways in multiple viral encephalomyelitis models will ultimately enhance targeted treatment options at early stages of disease manifestation. Accumulating literature in both rodent models and human encephalitis implicate that manipulation of IL-10 and IFNγ may have broad implications to treat encephalitis more broadly.

### AUTHOR CONTRIBUTIONS

CS and CB contributed to the writing, editing of the manuscript and approved the final version for publication.

#### FUNDING

This work was supported by the National Institutes of Health grant NS091183.

### REFERENCES


for autoimmune inflammation of the brain. Nature (2003) 421:744–8. doi: 10.1038/nature01355


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

Copyright © 2018 Savarin and Bergmann. 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.

# TRIM21 Restricts Coxsackievirus B3 Replication, Cardiac and Pancreatic Injury via Interacting With MAVS and Positively Regulating IRF3-Mediated Type-I Interferon Production

#### Edited by:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany

#### Reviewed by:

Antje Beling, Charité Universitätsmedizin Berlin, Germany Haizhen Zhu, Hunan University, China

> \*Correspondence: Wei Xu xuweifd828@126.com

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology

Received: 22 May 2018 Accepted: 08 October 2018 Published: 25 October 2018

#### Citation:

Liu H, Li M, Song Y and Xu W (2018) TRIM21 Restricts Coxsackievirus B3 Replication, Cardiac and Pancreatic Injury via Interacting With MAVS and Positively Regulating IRF3-Mediated Type-I Interferon Production. Front. Immunol. 9:2479. doi: 10.3389/fimmu.2018.02479

#### Hui Liu† , Min Li † , Yahui Song and Wei Xu\*

Jiangsu Provincial Key Laboratory of Infection and Immunity, Institute of Biology and Medical Sciences, Soochow University, Suzhou, China

Tripartite motif-containing 21 (TRIM21) is a regulator of tissue inflammation and pro-inflammatory cytokine production, and has been implicated in negative regulation of IRF3-dependent type I interferon signaling. However, the antiviral activity of TRIM21 varies among diverse viruses and its role on regulation of type I interferon remains inconsistent in different microbial infections. Here, we investigate the potential role for TRIM21 in controlling Coxsackievirus B3 (CVB3) replication and susceptible organ pathology. We found that CVB3 infection up-regulated the expression of TRIM21 in hearts of mice and cardiomyocytes at early phase of infection. Knock-down of TRIM21 resulted in increased viral replication, while overexpression led to increased phosphorylation and dimerization of IRF3, increased IFN-β transcription and reduced viral replication in vitro. We demonstrate that TRIM21 promotes the activation of IRF3 in CVB3-infected cells via interacting with MAVS and catalyzing the K27-linked polyubiquitination of MAVS, thereby enhancing type I interferon signaling. The RING domain of ubiquitin ligase activity and PRY-SPRY domain of TRIM21 are critical for its anti-viral effect. In vivo overexpression of TRIM21 significantly protected mice against viral myocarditis by suppressing CVB3 replication and reducing cardiac inflammatory cytokine production. While TRIM21 deficient mice exhibited a decreased IFN-β production, an increased cardiac and pancreatic CVB3 replication, and aggravated pancreatic injury as well as myocarditis during acute infection. Thus, our results demonstrate TRIM21 as a positive regulator of IFN-β signaling by targeting MAVS during CVB3 infection and suggest it as a potent host defense against CVB3 infection and viral-induced injury in hearts and pancreas.

Keywords: TRIM21, coxsackievirus B3 (CVB3), viral myocarditis, IFN-β, IRF3

## INTRODUCTION

Coxsackievirus is a single-Stranded RNA non-enveloped virus of the Enterovirus genus within Picornaviridae associated with several human and mammalian diseases, of which B3 type Coxsackievirus (CVB3) is well-identified as a major causative agent of viral myocarditis (VMC) (1, 2). CVB3 has been involved in 25–27% cases of acute myocarditis and dilated cardiomyopathy in children and young adults (3). VMC has been identified as an important etiology of heart failure and dilated cardiomyopathy, which contribute to nearly 50% of the indication for the heart transplantation (4). CVB3 infection also involves brain and pancreas, resulting in aseptic meningitis and pancreatitis (5, 6). Early direct virus-induced cytopathic effect and intense inflammatory injury followed by host immune responses are the main pathological processes of VMC and pancreatitis. Although excessive activation of immune response triggered by virus infection maybe a major factor contributing to tissue injuries, the virus itself is critical to the progression of VMC via direct attack on cardiomyocytes (7, 8). Despite considerable effort for decades, the fundamental mechanism responsible for the pathogenesis of viral myocarditis has not been well-understood and no effective therapies for VMC are currently available. During acute phase, CVB3 replication leads to myocardial and pancreatic injury directly through inducing apoptosis and necrosis of cardiomyocytes and pancreatic acinar cells. CVB3 RNA can be detected in the chronic stages in infected animals by 21 days post-infection, initiating the disease progression to more severe myocardial fibrosis and DCM (8, 9). In this sense, development of novel anti-viral compounds and early intervention represents an alternative way to treat CVB3 myocarditis and related cardiomyopathy.

The tripartite motif (TRIM) protein family contains over 70 members of TRIM protein family in human and is structurally characterized by a RING domain, one or two B-boxes, and a coiled-coil domain (10). TRIM proteins have been reported to be involved in multiple biological processes including the regulating innate immunity, carcinogenesis, cell differentiation and apoptosis, which are mainly dependent on the RING domain of ubiquitin ligase activity and B-box domain of interacting motif (11, 12). Recently, a growing body of evidence suggests that many TRIM proteins play important roles in direct antiviral activities and in the regulation of antiviral innate immunity. TRIM5α was found to inhibit HIV-1 replication by directly interacting with viral proteins (13). TRIM22 has been reported to exert antiviral activity against several viruses, such as hepatitis B virus (HBV), encephalomyocarditis virus (ECMV), and human immunodeficiency virus (HIV-1) (14–16).

TRIM21, initially known as an autoantigen Ro52/SS-A, is an ubiquitously expressed cytosolic E3 ubiquitin ligase and plays important roles in immune regulation and microbial restriction. TRIM21 has been well known as a regulator for type I interferon (IFN) production, however it may positively or negatively modulate the antiviral innate signaling according to the types of viruses. TRIM21 has been reported to be a positive regulator of IRF3 signaling by preventing its ubiquitination and degradation, thus enhancing IRF3 mediated antiviral responses (17).On the other hand, Higgs et al. claimed that TRIM21 catalyzed IRF3 ubiquitination and promoted its degradation leading to inhibition of interferon-β (IFN-β) production postpathogen recognition (18). TRIM21 also serves as a negative regulator of IFN-β during Japanese encephalitis virus (JEV) infection in human microglial cells (19).Recently, Xue et al. report that TRIM21 is upregulated upon RNA virus infection and promotes K27-linked polyubiquitination of MAVS to upregulate type-I interferon signaling, thereby inhibiting viral infection (20). Thus, the antiviral activity of TRIM21 varies among different viral infection. Up to the present, there is no report of TRMI21 on CVB3 infection; and almost no biological function of TRIM21 has been confirmed in animal models of viral infection. It is of great interest to explore the possible antiviral function of TRIM21 on CVB3 infection and its role in the disease progression of CVB3-induced VMC and pancreatitis.

Here, we investigate the antiviral activity of TRIM21 against CVB3 replication and its role in CVB3-induced acute vial myocarditis and pancreatic injury. Our results indicate that TRIM21 inhibits CVB3 replication via interacting with MAVS for promoting the K27 polyubiquitination of MAVS, thereby enhancing IRF3-mediated type I IFN signaling pathway and protecting mice against CVB3-induced myocarditis as well as pancreatic acinar cell necrosis.

### MATERIALS AND METHODS

#### Mice and Virus

Six–eight weeks old male BALB/c mice were purchased from the Shanghai Slac Animal Inc. TRIM21−/<sup>−</sup> mice were constructed from Cyagen Biotech (Guangzhou, China).CVB3 (Nancy strain) was a kind gift from Professor Yingzhen Yang (Key Laboratory of Viral Heart Diseases, Zhongshan Hospital, Shanghai Medical College of Fudan University).

#### Cell Culture

HeLa cells and HEK-293 cells were grown and maintained in DMEM medium supplemented with 10% FBS (Gibco) and 100 units/ml penicillin and streptomycin in a 5% CO<sup>2</sup> incubator at 37◦C.

#### Virus Titers Assays

The viral titer was determined by TCID<sup>50</sup> assay on HeLa cell monolayers with standard methodology (AdEasy Application Manual, version 1.4; Qbiogene, Carlsbad, CA, United States).Cell culture and tissue lysis supernatants were diluted serially using 10-fold dilutions and titered on HeLa cell monolayers by the TCID<sup>50</sup> assay.

#### Plasmids and Transfection

The Flag-MAVs or HA-K27Ub plasmids were the gifts from Prof. Hui Zheng (Institute of biological and medical sciences, Soochow University). Human TRIM21 cDNA was amplified from RNA of HeLa cells using primers: For: 5′ -GCCA CCATGGATTACAAGGATGACGACCGATAAGGCTTCAGC

ACGC-3′ and Rev: 5′ -AAAGCCATCAATAGTCAG-3′ . Mouse TRIM21 cDNA was amplified from RNA of cardiomyocytes using primers:5′ -ATGGATTACAAGGATGACGATAAGCAC CCTCTACAACCTCAAAA-3′ and 5′ -CCTGGCTCCTGACCA TCACA-3′ . cDNA of truncated forms of TRIM21 lacking Nterminal RING,B-box C-terminal and PRY-SPRY domain were amplified using primers: 1RING-For: 5′ -CCGGCAGCGCTT TATGCTGCTC-3′ and 1RING-Rev: 5′ -AAAGCCATCAAT AGTCAG-3′ ; 1B-box-For: 5′ -ATGCGGTGTGCAGTGCAT GGA-3′ and 1B-box-Rev: 5′ -AAAGCCATCAATAGTCAG-3′ ; 1PRY/SPRY-For: 5′ -GCCACCATGGCTTCAGCAGCACGC-3′ and 1PRY/SPRY-Rev: 5′ -TCACACATGGCACACACTC-3′ . For the transfection experiment, HeLa cells were seeded into 24-well plates and transient transfection was performed by Lipofectamine 2000 according to the manufacturer's instruction. Cells were cultured for 24 h before infection of CVB3.

### Quantitative Real-Time PCR (Q-PCR)

Total RNA was isolated from cells or tissues using RNAiso reagent (Takara, Cat. No. 9109), and cDNA was prepared using reverse transcriptase (Takara, Cat. No. DRR063A). Quantitative real-time RT-PCR (Q-PCR) was performed using SYBR green real-time PCR kits (TaKaRa, Cat. No. DRR041A)on a Bio-Rad iCycler using the following primers:


The 2−11CT method was used to normalize the transcription of the detected gene mRNA to that of the GAPDH mRNA and calculate the fold induction relative to the control.

### Short Interference RNA (siRNA)

Human siRNA oligonucleotides targeting sequences named as TRIM21 siRNA1 (UCAUUGUCAAGCGUGCUGC) and TRIM21 siRNA2 (UGGCAUGGAGGCACCUGAAGGUGG) were ordered from GenePharma.Inc (Shanghai, China). The siRNA was transfected into HeLa cells using INTERFERin in vitro siRNA transfection reagent (Polyplus, NewYork, United States).

### Western Blotting

HEK293 cells were transfected with plasmids containing human or murine TRIM21 (1 µg) using the Lipofectamine Plus reagent (Invitrogen) for 24 h, and then infected by CVB3 (MOI = 5) for 18 h. Samples were resuspended in sample lysis buffer (Bio-Rad). Lysates were resolved by SDS-polyacrylamide gel electrophoresis and transferred to PVDF membranes. The blots were probed primary antibodies for Flag (1:1,000, CST 8146S), IRF3 (1:1,000, CST, D8389), pIRF3 (1:1,000, CST, S396), actin (1:2,000, ABGEN, SG140609AB), GAPDH (1:10,000, Sigma, G9545), and VP-1 (1:2,000, Dako, M706401). HRP-conjugated anti-rabbit (1:4,000, CST, 7074) or anti-mouse IgG (Bioworld, AB54151) was used as a secondary antibody. Proteins were detected by chemiluminescence (Pierce). The intensities of the bands in the blots were quantified by densitometry using the Image Studio Lite program according to the developer's instructions.

### IP and Immunoblotting

HEK293 cells were transfected with an expression plasmid encoding full-length of Flag-tagged MAVS. Cell lysates were collected using radioimmunoprecipitation assay (RIPA) lysis buffer with protease inhibitors (1 mM phenylmethylsulfonyl fluoride, Roche complete protease inhibitor), followed by immunoprecipitation with anti-Flag beads. Proteins were eluted from the beads after washing six times with PBS. The protein binding to the beads was subjected to Western blot with anti-TRIM21 (1:2,000 Santa Cruz Biotechnology, SC25351) or anti-Flag (1:1,000, CST 8146S).

### Ubiquitination Assays and Native Page

For analysis of ubiquitination of MAVS in HeLa cells, cells were co-transfected with TRIM21, HA-K27ub or Flag-MAVS, followed by infection with CVB3. Cell lysates were immunoprecipitated with anti-Flag and analyzed by immunoblotting with the anti-HA antibody. Native page for the detection of IRF3 dimerization was performed on acrylamide gel without SDS. Cells were lysed with ice-cold lysis buffer including 50 mM of Tris-Hcl at PH = 7.5, 150 mM of NaCl and 0.5% NP-40 containing protease inhibitor cocktail. After centrifugation at 13,000 g for 15 min, proteins in the supernatant were quantified and diluted with 5x native PAGE sample buffer (312.5 mM Tris-HCl, pH = 6.8; 75% glycerol; 0.25% bromophenol blue). The gel was pre-run for 30 min at 40 mA on ice with 25 mM Tris-HCL (pH = 8.4), and 192 mM glycine with or without 1% of deoxycholate in the cathode chamber and anode chamber, respectively. The unboiled total protein was added into the gel for 80 min at 25 mA on ice.

### Luciferase Reporter Assay

HEK293 cells were co-transfected with 100 ng luciferase reporter plasmid, 10 ng thymidine kinase promoter-Renilla luciferase reporter plasmid, and the TRIM21-expression or control vector plasmid using the Lipofectamine 2000 transfection reagent (Invitrogen,Cat.No.116688-019). 48 hrs later, cell lysates were prepared and the luciferase activities were determined by the Dual-Luciferase Reporter Assay System (Promega,Cat.No.E10910) according to the manufacturer's instructions.

#### CVB3 Infection

Mice were infected intraperitoneally (i.p.) with 100 µl PBS containing 1000 TCID<sup>50</sup> dose of CVB3. Body weight and mortality of mice were recorded upon the termination of experiment. Individual experiments were conducted at least three times with 7 to 10 mice per group.

#### Histopathological Analysis

Three hearts and pancreas of each group of mice were collected 7 days post infection. The apical parts of the tissues were fixed in 10% phosphate-buffered formalin, embedded in paraffin wax, sectioned at 5µm and stained with hematoxylin–eosin (H&E). Stained sections were used for image analysis with a Nikon Eclipse TE2000-S microscope and five images were captured under high power fields randomly.

#### Immunohistochemistry

Hearts were fixed with 10% formalin in 0.1 M phosphate buffer, pH 7.4. Sections were deparaffinized and irradiated at 750 W in a microwave oven in 10 mM sodium citrate buffer, pH 6.0. Sections were then treated with 3% hydrogen peroxide to inhibit endogenous peroxidases. After washing in TBS with 0.025% Triton X-100, the sections were blocked with 10% BSA. Following blocking, sections were incubated with goat polyclonal antibody against TRIM21 (sc-21362; 1:500; Santa Cruz Biotechnology) diluted in TBS-1% BSA overnight at 4◦C. After washing, sections were incubated with a biotinylated anti-goat secondary antibody (Jackson immunoresearch) for 1 h and a peroxidase-labeled streptavidin for 5 min at room temperature. Peroxidase activity was detected with DAB (Mouse and Rabbit Specific HRP/DAB Detection IHC Kit, ab64264, Abcam), and sections were counter stained with hematoxylin. The level of protein accumulation was estimated as the percentage of the total counterstained area that was positively stained for the protein of interest, which was determined using Image Jsoftware (Nikon Eclipse TE2000-S microscope).

#### Primary Cardiomyocyte Culture

Neonatal cardiomyocytes were isolated from 1 to 3 days BALB/c mice. The ventricles obtained from 1 to 3 days BALB/c mice were removed rapidly into cold Hanks' balanced salt solution (Gibco). After washing and mincing, tissues were digested in 0.05% trypsin (Gibco) for 30 min at 4◦C with rotation before transfer into DMEM (Gibco) containing 20% FBS (Fetalbovine serum, Gibco) to terminate the digestion. After washing with HBSS, the tissues were incubated with Liberase TH (0.1 U/mL, Roche, Germany) at 37◦C for 5 min, and the dissociated cells were collected into 20% FBS DMEM. This procedure was repeated until most of the cells were released. The isolated cells were incubated with 5% CO<sup>2</sup> at 37◦C for 2 h. The unattached cardiomyocytes were seeded into fibronectin-coated 8-well Live Cell Imaging Culture Dish (Bestmagsystem Medical Co. Ltd., Suzhou, China) and experiments were performed when the cardiomyocytes formed a confluent monolayer and beat in synchrony at 72 h.

#### Immunofluorescence

Immunofluorescence was performed to assess the cellular expression and location of TRIM21 and viral RNA expression according to the manufacturer's instructions. Freshly cultured cardiomyocytes were infected with CVB3 (MOI = 5) for 0, 12, and 24 hrs. Heart tissues of infected mice were embedded in OCT and made into 5µm cryo section. Cells were fixed with 4% paraformaldehyde for 1 h before blocking (2% BSA) for 1 h. Cells were then incubated with primary antibodies against TRIM21 (1:200, Santa Cruz Biotechnology) and anti-dsRNA (1:300, J2 mAb, English and Scientific Consulting) at 4◦C overnight. Fluorochrome-conjugated secondary antibodies (1:200, goat anti-rabbit IgG, goat anti-mouse IgG, Southern Biotech) and DAPI were used for immunofluorescent staining. Images were captured and analyzed with Nikon A1 confocal microscope.

#### Enzyme-Linked Immunosorbent Assay (ELISA)

Levels of TNF-α, IL-6, IL-10, IFN-γ, MCP-1 were determined using sensitive mouse, IL-6, IL-10, and IFN-γ kits according to the manufacturers' instructions (eBioscience, San Diego, USA).

#### In vivo Overexpression of TRIM21

Mice were retro-orbitally injected with 1.0 ml reagent containing 50 µg of mouse TRIM21-expression plasmid or vector plasmid using in vivo-JetPEITM–Gal transfection agent according to the manufacturer's instructions (Polyplus-transfection Inc., USA). Mice received 2 doses of TRIM21 plasmids 2 days before and 1 day after CVB3 infection to sustain in vivo over-expression of TRIM21.

#### Statistical Analysis

Data were presented as the mean ± SEM and statistical analysis was analyzed by GraphPadPrism 5 software. For two-group comparisons, statistical significance was determined by Student's t-test. Survival curves were estimated from Kaplan-Meier procedure with the Lonrank test to compare survival among groups. P < 0.05 was considered to be statistically significant and are indicated as follows: ∗ , 0.05 ≥ P > 0.01; ∗∗, 0.01 ≥ P > 0.001; ∗∗∗ , P ≤ 0.01.

### RESULTS

#### TRIM21 Is Up-Regulated in Hearts of Mice and in the Murine Cardiomyocytes Upon CVB3 Infection

To explore the role of TRIM21 in CVB3 infection, first we investigate whether TRIM21 is induced in heart tissues of mice by CVB3 infection. After 1000 TCID<sup>50</sup> CVB3 i.p.

24 h. CVB3 dsRNA, endogenous TRIM21 protein and nucleus were stained with anti-dsRNA antibody (green), anti-TRIM21Ab (red) and DAPI dye (blue) and observed under confocal microscope. Photomicrographs were captured under high power fields (100 × magnifications) (E).TRIM21 mRNA level were analyzed by Q-PCR. Data were presented as mean ± SEM of three representative independent experiments. The GAPDH expression level of heart tissues was set as 1.0 (F). \*\*p < 0.01; \*\*\*p < 0.001.

infection, a massive inflammatory infiltration and cardiomyocye necrosis were observed in hearts at day 7 p.i. (**Figure 1A**). The viral load in myocardium increased and peaked at day 3 p.i., then declined at day 7 p.i. (**Figure 1B**). Then we detected the expression kinetics of TRIM21 by Q-PCR and immunohistochemistry. The protein and mRNA levels of TRIM21 in hearts of mice were significantly increased and peaked at day 3 p.i. (**Figures 1C,D**). To confirm the expression and localization of TRIM21 in cardiomyocytes, we cultured primary cardiomyocytes from newborn mice and infected cells with CVB3. The immunofluorescence assay showed that TRIM21 was localized in cytoplasm and CVB3 infection enhanced its expression at protein and RNA levels (**Figures 1E,F**).Therefore, our result demonstrate that cardiac TRIM21 expression is upregulated by CVB3 infection, which may be involved in the regulation of CVB3 infection and the progression of viral myocarditis.

## TRIM21 Suppresses CVB3 Replication in vitro

To investigate the role of TRIM21 on CVB3 replication, HeLa cells were transiently transfected with a plasmid expressing TRIM21 or vector alone and then infected with CVB3 at MOI of 5. The efficiency of overexpression of TRIM21 was confirmed by real-time PCR and Western blot (**Figure 2A**). The supernatant was subjected to TCID<sup>50</sup> assay to determine the role of TRIM21 on viral progeny release. As shown in **Figure 2B**, TRIM21 overexpression significantly reduced the virus particle release. Furthermore, the protein level of CVB3 capsid VP1 was significantly inhibited by TRIM21 overexpression (**Figure 2C**).

To further verify the antiviral effect of TRIM21, we designed and screened two specific siRNA targeting the open reading frame of TRIM21, which led to a 75–80% reductions in the overall levels of the TRIM21 mRNA and

protein (**Figure 2D**). Knockdown of TRIM21 increased CVB3 progeny production (**Figure 2E**) and viral capsid protein VP1 expression (**Figure 2F**) significantly, compared to the effect of NC siRNA. Collectively, these results confirm that TRIM21 significantly restricts CVB3 replication in vitro.

### TRIM21 Increases IFN-α/β Activation Pathway

Type I interferons (IFNs) play an important part in the resistance to viral infection. TRIM21 is reported to be involved in modulating host innate type I signaling against viral replication. Thus, we first examined the IFN-β mRNA production in HeLa cells overexpressing TRIM21 upon CVB3 infection by real-time PCR. Cells transfected with an empty vector were used as a control. In comparison to vector-transfect cells, a moderate promotion in IFN-β mRNA levels was observed in the TRIM21-overexpressed cells infected with CVB3. Additionally, IFN-α mRNA level was increased significantly by TRIM21 overexpression (**Figure 3A**). To confirm IFN-α/β promoting role of TRIM21, HeLa cells were co-transfected with TRIM21 vector and IFN-β promoter-luciferase plasmid. As demonstrated in **Figure 3B**, overexpression of TRIM21 enhanced the activity of IFN-β promoter in a dose-dependent manner after CVB3 infection. Furthermore, co-transfection of TRIM21 increased MAVS-activated and MAD5-activated IFN-β reporter gene expression. Next, we detected IFN-stimulated genes (ISGs) expression and found TRIM21 also up-regulated the expression of ISG15 and ISG54 upon CVB3 infection (**Figure 3C**). Thus, our data suggest that TRIM21 up-regulates the activation of IFN-β signaling pathway.

### TRIM21 Positively Regulates IRF3 Activation via K27-Linked Polyubiquitination of MAVS Upon CVB3 Infection

Since TRIM21 promotes IFN-β activation after CVB3 infection, we suggest that TRIM21 might positively modulate the upstream molecules of type I interferon signaling pathway. RIG-I and MDA-5 recognition of CVB3 RNA leads to the activation of IRF3 and transcription factors required for transcription activation of IFN-α/β. We next explored the effect of TRIM21 on IRF3 activation upon CVB3 infection. As shown in **Figure 3D**, overexpression of TRIM21 significantly enhanced the reporter activity of IRF3 at basal level and after CVB3 infection. Then, native page assay was performed and demonstrated that TRIM21 overexpression promoted the dimerization and phosphorylation of IRF3 (**Figures 3E,F**). Recently, Xue et al. reported that TRIM21 catalyzed the K27-linked polyubiquitination of MAVS to upregulate type-I interferons signaling upon RNA virus infection. So we examined the interaction between TRIM21 and MAVS. The CO-IP experiments revealed that TRIM21 interacted with MAVS while CVB3 infection reduced MAVS expression (**Figure 3G**). Degradation of MAVS by CVB3 pro2A

FIGURE 3 | Overexpression of TRIM21 activates the virus-induced type I IFN signaling via catalyzing the K27-linked polyubiquitination of MAVS and increases IRF3 phosphorylation. (A) HeLa cells were transfected with TRIM21 expression or vector plasmids. 24 h after transfection, cells were infected with CVB3 (MOI = 5) for the indicated time. The mRNA levels of endogenous IFN-β and IFN-α were detected by Q-PCR assay. Data were presented as mean ± SEM of three representative independent experiments. (B) HEK293 cells were co-transfected with the IFN-β promoter reporter plasmids and the indicated amounts of TRIM21 expression plasmid or 100 ng vector plasmid. 24 h later, the luciferase activity was determined. HEK293 cells were co-transfected with the IFN-β promoter reporter plasmids and TRIM21 expression plasmid or mock vector. 24 h later, cells were infected with CVB3 (MOI = 5) for 12 h and the luciferase activity was determined. HEK293 cells were co-transfected with the IFN-β promoter reporter plasmids and TRIM21 expression plasmid or mock vector, together with MAVS or MDA5 expression plasmids. 24 h later, cells were infected with CVB3 (MOI = 5) for 12 h and the luciferase activity was determined. Data were presented as mean ± SEM of three representative independent experiments. (C) HeLa cells were transfected with TRIM21 expression or vector plasmids for 24 h before infection with CVB3 (MOI = 5). The mRNA levels of endogenous ISG15 and ISG54 were detected by Q-PCR assay. (D) HEK293 cells were co-transfected with the IRF3 promoter reporter plasmids and TRIM21 expression plasmid or mock vector. 24 h later, cells were infected with CVB3 (MOI = 5) for 12 h and the luciferase activity was determined and data were presented as mean ± SEM of three representative independent experiments. (E,F) HeLa cells were transfected with TRIM21 expression or vector plasmids for 24 h before infection with CVB3 (MOI = 5). Cell lysates were subjected to probe withanti-IRF3, anti-pIRF3 and anti-GAPDH antibodies by Western blotting and Native page for the detection of IRF3 dimerization. (G) HeLa cells were transfected with Flag-MAVS plasmids as indicated, 24 h later cells were infected with CVB3 (MOI = 5). Cellular lysates were immunoprecipitated with anti-Flag. Immunoprecipitates were analyzed by WB with anti-Flag and anti-TRIM21. (H) HeLa cells were co-transfected with TRIM21, Flag-MAVS, and HA-K27Ub for 24 h and treated with CVB3 for additional 12 h. Ubiquitination and immunoblotting assays were performed with indicated antibodies. \*p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001.

(21) may counteract the effect of CVB3 on TRIM21 upregulation and TRIM21-MAVS interaction in vitro. Furthermore, we observed that TRIM21 catalyzed the formation of K27 linked polyubiquitin chains on MAVS (**Figure 3H**, lane 1–2). Importantly, CVB3 infection enhanced the formation of the K27-linked polyubiquitinon MAVS by TRIM21 (**Figure 3H**, lane 3–4).Our data suggest that TRIM21 positively regulates type I IFN pathway during CVB3 infection via interacting with and promoting the ubiquitination of MAVS, thereby enhancing IRF3 activation.

### The Ring and PRY-SPRY Domains Are Required to Facilitate the TRIM21-Mediated Anti-viral Activity

TRIM21 contains three classical motifs including a RING finger domain, a B-box domain and a B30.2 domain. We constructed various domain mutants of TRIM21 to define which part was involved in its antiviral role (**Figure 4A**). As compared with the full-length TRIM21, B-Box mutant showed similar anti-CVB3 effects, while RING and PRY-SPRY domain mutants abolished the antiviral effects as measured by western blot of viral VP-1 protein (**Figure 4B**). Furthermore, dysfunction of RING domain and PRY-SPRY domain obstructed the activating role of TRIM21 on the promoter of IFN-β and IRF3, while B-box mutant and B30.2 mutant had no effect (**Figure 4C**). Collectively, the RING domain with E3 ubiquitin ligase and the PRY-SPRY domain were required for the TRIM21-mediated type I IFN activation and anti-viral effect.

### In vivo Overexpression of TRIM21 Protects Mice Against CVB3-Induced Myocarditis

We next evaluate the antiviral effect of TRIM21 in vivo according to an in vivo-JetPEITM strategy (22) and one retroorbital injection of 50 µg TRIM21- plasmids led to an enhanced protein

expression of cardiac TRIM21 which sustained for 2–3 days confirmed by IHC analysis (**Figure 5A**). Therefore, groups of mice were retro-orbitally injected with 50 µg TRIM21-plasmids or vector-plasmids using in vivo-Jet PEI reagent 2 days before and 1 day after CVB3 infection (**Figure 5B**), and susceptibility to CVB3 myocarditis as well as viral replication were evaluated in a course of 7 days infection. The survival rate and bodyweight loss of mice were monitored by day 7 p.i. and a significantly improved disease condition and reduced mortality were observed in mice with in vivo TRIM21 overexpression. More than 60% mice injected with mock plasmids died by day 7 and lost their 28% bodyweight; while TRIM21-overexpressed mice underwent a gentle decline loss of bodyweight (∼17%) and ∼70% mice survived by day 7 p.i. (**Figures 5C,D**). Consistently, histological analysis revealed that mice with mock plasmids developed severe myocarditis with diffuse inflammation and necrotic lesions, whereas TRIM21 treatment attenuated myocarditis with restricted inflammation and necrosis (**Figure 5E**). By analyzing the levels of cardiac inflammatory cytokines we found that inflammatory cytokines, such as IL-1β, TNF-α, IL-6, IL-10, and MCP-1, were significantly reduced by TRIM21 overexpression while IFN-γ level was not affected (**Figure 5F**).

To test whether differences in CVB3 disease susceptibility were due to differences in viral titers, CVB3 burden in the hearts of mice were measured. As shown in **Figure 5G**, CVB3 titer was significantly reduced in hearts of mice with TRIM21 over-expression at day 3 p.i. Immunofluorescent staining of the heart sections also confirmed dramatically reduced viral RNA level in hearts (**Figure 5I**). In accordance with that, a significantly up-regulated mRNA expression of IFN-β in heart was confirmed in TRIM21 overexpressed mice at day 3 p.i.(**Figure 5H**).

#### TRIM21 Deficient Mice Exhibits Increased Cardiac and Pancreatic Viral Burden and Aggravated Myocarditis as Well as Pancreatic Necrosis

To further confirm the antiviral effect of TRIM21 in vivo, we constructed deficient mice by CRISPR-CAS9 strategy (**Figure 6A**) and confirmed the deletion of mRNA and protein level of TRIM21 in tissues of mice (**Figures 6B,C**). Next, WT and TRIM21-deficient mice were infected i.p. with CVB3. Throughout the 7 days infection, TRIM21-deficient mice exhibited greater signs of sickness at day3 p.i. and lost weight more promptly by day 7 p.i. (14.8 vs. 5.1%, TRIM21-deficient vs. WT, p < 0.05, **Figure 6D**). Histopathology analysis revealed that TRIM21-deficient mice exhibited a significantly aggravated coagulative necrosis and acinar cell necrosis in the pancreas

FIGURE 5 | injected retroordibitally with 2 doses of 50 µg PEI-packaged mock or TRIM21 plasmids on day −2 and 1 and subjected to 1000TCID<sup>50</sup> CVB3 on day 0(n = 6). The survival rate (C) and body weight change (D) were monitored daily until day 7 p.i. (E) Representative image of HE-staining hearts of CVB3-infected mice (day 7 p.i.) treated with PEI-TRIM21 or PEI-vector, showing intra-cardiac immune infiltrates (marked with arrows). Scale bar: 100µm. Pathological scores of the heart of mice are shown. Results are presented as mean ± SEM; \*p < 0.05. (F) Protein levels of inflammatory cytokines in the homogenates of heart were measured by ELISA. Data were presented as mean ± SEM of three representative independent. (G) The cardiac CVB3 titer at day 3 p.i. were determined by TCID<sup>50</sup> assay. Data represent mean values of CVB3 PFU per gram of the heart tissues. Results are presented as mean ± SEM; Data pooled from 3 independent experiments. \*p < 0.05; \*\*p < 0.01. (H) Relative mRNA level of IFN-β (day 3 and 7p.i.) was detected by Q-PCR. Data were normalized to GAPDH expression and presented as mean ± SEM of three representative independent. (I) Hearts of mice at day 3 and 7 p.i. were OCT-embedded and cyrosections (5µM) were subjected to fluorescent staining. Composite confocal represented images show dsRNA (red, anti-dsRNA Ab) and nuclear (blue, DAPI).Low magnification (magnification, × 100) and higher magnification of the boxed areas (magnification, × 200) are shown. The number of red-stained viral-infected cells in the heart sections of mice were numerated. Data are expressed as mean ± SEM from three repeated experiments (n = 3). \*\*\*p < 0.001.

at day3 p.i., and an increased cardiac immune infiltration at day 7 p.i. (**Figure 6E**) compared to WT mice. Consistent with enhanced tissue pathology, the levels of cardiac inflammatory cytokines were significantly increased in TRIM21 KO mice than in WT mice (**Figure 6F**).To test whether differences in CVB3 disease susceptibility were due to differences in viral replication, CVB3 burden in the hearts and pancreas of mice were measured. At 3 dpi, the peak of viral replication, TRIM21 deficient mice exhibited significantly increased viral titers in hearts and pancreas compared to WT mice (**Figure 6G**). To confirm the anti-viral effect of TRIM21 in vivo, the mRNA and protein level of IFN-β in hearts were measured and were found significantly decreased in TRIM21 deficiency mice at early infection stage compared to those in WT mice (**Figure 6H**). These data confirm that TRIM21 effectively suppresses CVB3 replication in vivo. We thus propose a model depicting the role of TRIM21 in CVB3 infection: CVB3 infection up-regulates the expression of TRIM21 in cardiomyocytes, which interacts with MAVS and promotes IRF3-mediated IFN-I signaling to suppress viral replication in vivo, thereby decreasing virus-induced inflammatory injury in both hearts and pancreas of mice (**Figure 6I**).

#### DISCUSSION

In this study, we try to explore the role of TRIM21 in the susceptibility of mice to CVB3 induced myocarditis. TRIM21 expression is significantly up-regulated in hearts of mice on day 3 post infection, and systemic TRIM21 effectively inhibits CVB3 replication in vivo. TRIM21 restricts CVB3 replication by positively regulating IRF3 activation and IFN-β production after CVB3 infection via interacting with and promoting K27-linked polyubiquitination of MAVS. Silencing of TRIM21 significantly enhances CVB3 replication in tissues and alleviates virus-induced cardiac and pancreatic injury. Treatment with CVB3-infected mice with TRIM21 significantly reduces CVB3 replication in hearts and the severity of viral myocarditis.

TRIM21, initially known as an autoantigen Ro52/SS-A, is an ubiquitously expressed cytosolic E3 ubiquitin ligase and plays important roles in immune regulation and microbial restriction (23, 24). It has been reported that TRIM21 is constitutively and broadly expressed in various organs and cell types, but with highly divergent levels of expression. Highest expression is seen in cells of the immune system with particularly high levels in T cells, macrophages and DCs, where the expression is further augmented by stimulation with IFNs and TLR ligation (25). Previous study finds that TRIM21 expression is substantially increased in human primary lymphocytes and monocyte-derived macrophages in response to interferons (IFNs, type I and II), suggesting TRIM21 as an interferon-induced gene (26). It has been reported that SeV, NDV or HCV infection could significantly induce the expression of TRIM21 through JAK/STAT signaling pathway (27). Thus, although CVB3 infection does not induce robust production of IFNβ, a significant induction of TRIM21 protein is observed in heart tissues upon CVB3 infection. More convincingly, TRIM21 expression is significantly enhanced in primary cardiomyocytes upon CVB3 infection (**Figures 1E,F**) which is localized in cytoplasm of cells.

The host cells activate a series of signaling events that lead to induction of type I interferons (IFNs), including IFN-β and IFN-α. Type I IFNs further induce the expression of downstream proteins, which mediate innate immune responses, such as suppression of viral replication, clearance of virus-infected cells (28, 29). The role of TRIM21 in regulating the type I interferon signaling has been controversial. Sunit K Singh demonstrates TRIM21 as a negative regulator of IFN-β production mediated by IRF-3 during JEV infection in human microglial cells (19). In 2015, another study finds that TRIM21 facilitates Nmi-mediated negative regulation of the innate antiviral response (32). And Liu Y's group reports that TRIM21 as an E3 ligase which induces the Lys48 (K48)-linked ubiquitination and degradation of DDX41 and negatively regulates the innate immune response to intracellular dsDNA in myeloid dendritic cells (30). All the above data indicate that TRIM21 is a negative regulator of IRF3 (or DD41) activation and IFN-β production. However, Chen Wang reports that TRIM21 is induced and interacts with IRF3, preventing IRF3 ubiquitination and degradation thus playing an anti-viral effect during SeV infection (17). Recently, Xue et.al have reported (20) that TRIM21 is upregulated upon RNA virus (SeV, VSV) infection, interacts with MAVS and catalyzes the K27-linked polyubiquitination of MAVS, thereby promoting the activation of IRF3 and inhibiting viral infection. In our study, we confirm the co-immunoprecipitation and polyubiquitination of TRIM21 with MAVS (**Figures 3G,H**), which is in consistency with Xue's report that TRIM21 interacts with MAVS and promotes K27-linked polyubiquitination of MAVS. It also supports our conclusion that the anti-viral effect of TRIM21 is RING domain dependent (**Figure 4C**). Finally we propose a model depicting the role of TRIM21 in CVB3 infection:

FIGURE 6 | TRIM21 deficient mice increases CVB3 replication in organs and aggravates pancreatic acinar cell necrosis as well as myocarditis. (A) Schematic diagram of deficient mice construction by CRISPR-CAS9 strategy. (B,C) Q-PCR and Western blot analysis of TRIM21 expression from tissues and BM cells of wild-type (WT) and TRIM21-/- mice, GAPDH was used as a loading control. (D-H) WT and TRIM21-deficient mice were infected i.p. with CVB3 (n = 6).The body weight change were monitored daily until day 7 p.i. (D). Representative image of HE-staining hearts and pancreas of CVB3-infected WT or TRIM21-/-mice (day 7 p.i.), showing intra-cardiac immune infiltrates or intactpancreatic acini (marked with arrows). Scale bar: 100µm. Pathological scores of the heart and pancreas of mice are shown. Results are presented as mean ± SEM; Data pooled from 3 independent experiments (E).The mRNA levels of inflammatory cytokines in the homogenates of heart (day 7 p.i.) were measured by Q-PCR. Data were presented as mean ± SEM of three representative independent (F).Viral loadin pancreas and hearts of mice(day 3p.i.) was assessed by TCID<sup>50</sup> assay. Results are presented as mean ± SEM;Data pooled from 3 independent experiments. \*p < 0.05; \*\*p < 0.01 (G).The mRNA and protein level of IFN-β (day 1–3 p.i.) in hearts of mice was detected by Q-PCR and ELISA. Data as mean ± SEM of three representative independent. \*p < 0.05; \*\*p < 0.01 (H). (I) Proposed model depicting the role of TRIM21 in positive regulation of IFN-I production during CVB3 infection. TRIM21 targets and promotes the activity of MAVS, leading to the increased phosphorylation and translocation of p-IRF3 into the nucleus, leading to enhanced transcription and production of IFNs and IFN-stimulated genes (ISGs) that limits CVB3 infection.

CVB3 infection up-regulates the expression of TRIM21 in mice, TRIM21 interacts with MAVS and promotes the activation of IRF3 resulting in an up-regulation of type I innate signaling during CVB3 infection (**Figure 6I**).Although MDA5, MAVS and RIG-I are cleaved by CVB3 2Apro and 3Cpro (21) indicating TRIM21, one of ISGs, might be hardly up-regulated during CVB3 infection and TRIM21-mediated IFN-I response enhancing effect might be counteracted, there is article suggesting that enhancing IFNs production might be an alternative prescription in CVB3 related syndromes (33). And our in vivo over-expression and deficiency experiment in mice confirm TRIM21 supplementation as a promising strategy to limit CVB3 infection and related cardiac and pancreatic pathology.

CVB3 has evolved many strategies to suppress host innate immunity therefore does not cause robust interferon release and ISG expression. As shown in **Figure 3A**, the mRNA expressions of IFN-α and IFN-β in cells at 24 h after CVB3 infection were quite low (as similar as the level seen before infection). 48 h after infection, CVB3 did not induce IFN-α expression while induced a very modest up-regulation of IFN-β. Only upon transfection with TRIM21 plasmid, the mRNA expressions of IFN-α/β were significantly up-regulated at both 24 and 48 h post infection. The induced up-regulation of ISG15 by CVB3 was also very moderate, only TRIM21 overexpression significantly promoted ISG15 expression. It seems that upregulation of IFN-α/β may be a TRIM21-mediated effect. However, our data demonstrate that CVB3 infection significantly increases expression of TRIM21 in hearts of mice (**Figures 1 A–D**) and in primary cardiomyocytes (**Figures 1E,F**). Therefore, the upregulation of IFN-α/β signaling by TRIM21 is at least partially dependent on CVB3 infection. And the interaction of TRIM21 with MAVS (**Figures 3G,H**) further supports our data that CVB3-induced TRIM21 could enhance the phosphorylation of IRF3 upon CVB3 infection leading to elevated IFN-β production.

As a member of tripartite motif (TRIM) family protein, TRIM21 contains a RNIG motif in the N-terminal domain, a B-box motif, a coiled-coil domain. And TRIM21 protein also contains a carboxy-terminal B30.2 (SPRY) domain (31). Previous study demonstrates that TRIM21 interacts with IRF3 directly via its C-terminal SPRY domain, resulting in the polyubiquitination and proteasomal degradation of IRF3 and reduced IFN-β promoter activity (18). Liu Y's group report that TRIM21 cannot interact with IRF3 in mDC but interact with DD41 for promoting the ubiquitination and degradation of DDX41 therefore negatively regulates IFN-I response to DNA virus (HSV) (30). In 2015, another study finds that during SeV and VSV infection, up-regulated TRIM21 interacts with both Nmi and IFI35 and activates K63-linked ubiquitination on K22 residue of Nmi (SPRY domain dependent) which facilitates the negative regulatory function of the Nmi-IFI35 complex on innate antiviral signaling (32). Yang et al. report in their study that upon RNA virus infection TRIM21 interacts with IRF3, interferes with the interaction between Pin1 and IRF3, thus preventing IRF3 ubiquitination and degradation via its B30.2 domain (17). Xue et al. reports that TRIM21 interacts with MAVS and catalyzes the K27-linked polyubiquitination of MAVS through its RING domain (20). In our study, we confirmed (**Figure 3G**) that TRIM21 interacts with and promotes the K27-ubiquitination of MAVS, and the anti-viral effect of TRIM21 is RING and PRY-SPRY domain dependent (**Figure 4**), which is in consistency with the recent report (20).

Currently, there is only limited report of antiviral effect of TRIM21 in mice model. Our study demonstrate the in vivo effect of TRIM21 on CVB3 replication and tissue pathology. By using in vivo-Jet PEI-transfection of TRIM21-plasmids and TRIM21 deficient mice, we demonstrated that the viral replication and CVB3-induced cardiac immune infiltration, cardiac proinflammatory cytokines production and injuries were significantly decreased upon in vivo over-expression of TRIM21 (**Figure 5**). In accordance with that, cardiac and pancreatic CVB3 replication as well as virus-induced pancreatic acinar cell necrosis and myocarditis were significantly aggravated in TRIM21 deficient mice (**Figure 6**). Our data identify TRIM21 as a potent viral inhibitory factor during CVB3 infection. Recently, TRIM21 is also identified as an intracellular Fc receptor linking cytosolic antibody recognition to the ubiquitin proteasome system (34–36).So we cannot rule out the antiviral effect of TRIM21 is partially dependent on antibody-dependent intracellular neutralization (ADIN) effect of TRIM21 in vivo. And our preliminary data show that TRIM21 has IgA-mediated ADIN effect on CVB3 replication in vitro (data not shown). So in further study we will focus on clarifying whether TRIM21 exerts IgA-mediated ADIN function on intestinal CVB3 replication considering CVB3 as an oral-fecal disseminating virus.

Overall, our study identifies cytosolic TRIM21 as a positive regulator of CVB3-triggered MAVS-mediated type I Interferon signaling pathway that restricts viral infection. TRIM21 expression is up-regulated by CVB3 infection at early phase of viral infection. TRIM21 inhibits CVB3 replication in vivo and in vitro through interacting with MAVS thereby promoting the activation of IRF3 and Type I Interferon production. The anti-viral effect of TRIM21 is dependent on RING and PRY-SPRY domain. We also demonstrate the antiviral effect of systemic TRIM21 in vivo which leads to the increased resistance to CVB3-induced myocarditis and pancreatic injury. Our data help to clarify the biological role of TRIM21 in severe tissue pathology caused by viral infection and indicating a therapeutic target potential for TRIM21.

### ETHICS STATEMENT

All animal experiments were performed in accordance with Soochow University institutional guidelines, and the study was approved by the Ethics Committee of Soochow University in written form (SYXK2015-0036).

## AUTHOR CONTRIBUTIONS

WX conceived and supervised the project. HL and YS performed the experiments. HL and ML interpreted data and wrote the manuscript. All authors approved the final version of the paper.

## FUNDING

This work was supported by NSFC grant (31870903, 31670930, 31470869, 31270973), the Natural Science Foundation of Jiangsu Province Higher education (14KJA310005), the Jiangsu Provincial Innovative Research Team, and Program for Changjiang Scholars and Innovative Research Team in University (grant PCSIRTIRT1075).

#### REFERENCES


Polyubiquitination of MAVS. J Virol. (2018) 92. doi: 10.1128/jvi.003 21-18


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

Copyright © 2018 Liu, Li, Song and Xu. 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.

# Silencing the CSF-1 Axis Using Nanoparticle Encapsulated siRNA Mitigates Viral and Autoimmune Myocarditis

Ingmar Sören Meyer 1,2†, Carl Christoph Goetzke3,4†, Meike Kespohl 3,4, Martina Sauter <sup>5</sup> , Arnd Heuser <sup>6</sup> , Volker Eckstein<sup>7</sup> , Hans-Peter Vornlocher <sup>8</sup> , Daniel G. Anderson9,10,11 , Jan Haas 1,2, Benjamin Meder 1,2, Hugo Albert Katus 1,2, Karin Klingel <sup>5</sup> , Antje Beling3,4 \* † and Florian Leuschner 1,2 \* †

#### Edited by:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany

#### Reviewed by:

Milena Botelho Pereira Soares, Instituto Gonçalo Moniz (IGM), Fiocruz Bahia, Brazil Yasushi Fujio, Osaka University, Japan

#### \*Correspondence:

Antje Beling antje.beling@charite.de Florian Leuschner florian.leuschner@ med.uni-heidelberg.de

†These authors have contributed equally to this work and are share first and senior authorship

#### Specialty section:

This article was submitted to Cytokines and Soluble Mediators in Immunity, a section of the journal Frontiers in Immunology

Received: 17 July 2018 Accepted: 17 September 2018 Published: 08 October 2018

#### Citation:

Meyer IS, Goetzke CC, Kespohl M, Sauter M, Heuser A, Eckstein V, Vornlocher H-P, Anderson DG, Haas J, Meder B, Katus HA, Klingel K, Beling A and Leuschner F (2018) Silencing the CSF-1 Axis Using Nanoparticle Encapsulated siRNA Mitigates Viral and Autoimmune Myocarditis. Front. Immunol. 9:2303. doi: 10.3389/fimmu.2018.02303 1 Internal Medicine III, University Hospital Heidelberg, Heidelberg, Germany, <sup>2</sup> DZHK (German Centre for Cardiovascular Research), Partner Site Heidelberg-Mannheim, Heidelberg, Germany, <sup>3</sup> Institute of Biochemistry, Charité - Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, <sup>4</sup> DZHK (German Centre for Cardiovascular Research), Partner Site Berlin, Berlin, Germany, <sup>5</sup> Cardiopathology, Institute for Pathology and Neuropathology, University Hospital Tuebingen, Tuebingen, Germany, <sup>6</sup> Max-Delbrueck-Center for Molecular Medicine Berlin, Berlin, Germany, <sup>7</sup> Internal Medicine V, University Hospital Heidelberg, Heidelberg, Germany, <sup>8</sup> Axolabs GmbH, Kulmbach, Germany, <sup>9</sup> David H. Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA, United States, <sup>10</sup> Cardiovascular Division, Department of Medicine, Brigham and Women's Hospital, Boston, MA, United States, <sup>11</sup> Department of Chemical Engineering, Massachusetts Institute of Technology (MIT), Cambridge, MA, United States

Myocarditis is an inflammatory disease of the heart muscle most commonly caused by viral infection and often maintained by autoimmunity. Virus-induced tissue damage triggers chemokine production and, subsequently, immune cell infiltration with pro-inflammatory and pro-fibrotic cytokine production follows. In patients, the overall inflammatory burden determines the disease outcome. Following the aim to define specific molecules that drive both immunopathology and/or autoimmunity in inflammatory heart disease, here we report on increased expression of colony stimulating factor 1 (CSF-1) in patients with myocarditis. CSF-1 controls monocytes originating from hematopoietic stem cells and subsequent progenitor stages. Both, monocytes and macrophages are centrally involved in mediating tissue damage and fibrotic scarring in the heart. CSF-1 influences monocytes via engagement of CSF-1 receptor, and it is also produced by cells of the mononuclear phagocyte system themselves. Based on this, we sought to modulate the virus-triggered inflammatory response in an experimental model of Coxsackievirus B3-induced myocarditis by silencing the CSF-1 axis in myeloid cells using nanoparticle-encapsulated siRNA. siCSF-1 inverted virus-mediated immunopathology as reflected by lower troponin T levels, a reduction of accumulating myeloid cells in heart tissue and improved cardiac function. Importantly, pathogen control was maintained and the virus was efficiently cleared from heart tissue. Since viral heart disease triggers heart-directed autoimmunity, in a second approach we investigated the influence of CSF-1 upon manifestation of heart tissue inflammation during experimental autoimmune myocarditis (EAM). EAM was induced in Balb/c mice by immunization with a myocarditogenic myosin-heavy chain-derived peptide dissolved in complete Freund's adjuvant. siCSF-1 treatment initiated upon established disease inhibited monocyte infiltration into heart tissue and this suppressed cardiac injury

**106**

as reflected by diminished cardiac fibrosis and improved cardiac function at later states. Mechanistically, we found that suppression of CSF-1 production arrested both differentiation and maturation of monocytes and their precursors in the bone marrow. In conclusion, during viral and autoimmune myocarditis silencing of the myeloid CSF-1 axis by nanoparticle-encapsulated siRNA is beneficial for preventing inflammatory tissue damage in the heart and preserving cardiac function without compromising innate immunity's critical defense mechanisms.

Keywords: inflammation and immunmodulation, innate immunity, cytokines, monocytes/macrophages, RNA interference, virus, infection-immunology, myocarditis

#### INTRODUCTION

Myocarditis and its sequela, dilated cardiomyopathy, are leading causes of heart failure and sudden death in young adults (1). While various agents may provoke cardiac inflammation, viral infections are the most common trigger of myocardial inflammation in the Western world. Although various viruses are putative invaders of heart tissue, most of our knowledge on disease pathology comes from infection with enteroviruses, in particular CoxsackievirusB3 (CVB3). CVB3 had been reported among the most prevalent pathogens causing viral myocarditis in North America and Europe in the past (2, 3). Mouse models using different strains with divergent susceptibility for cardiotropic CVB3 elegantly reflect human disease with highly diverse disease outcome (4, 5). The hereditary susceptibility involves a certain immune-anchored genetic phenotype leading either to altered virus control and/or to induction of deleterious immunopathology (6–8). Severe virus-induced inflammation can result in a subsequent loss of self-tolerance against cardiac proteins, which contributes to additive auto-destructive activity of infiltrating cells and exaggerates heart tissue damage (9, 10). Cardiac myosin is such a crucial autoantigen in both human and murine virus-induced myocarditis (9). Administration of cardiac myosin or its pathogenic epitope in combination with an adjuvant induces experimental autoimmune myocarditis (EAM) in mice, a model that mimics certain aspects of myocarditis and heart failure in humans (11).

Treatment options for patients with myocarditis are sparse and both conventional immunosuppressive as well as antiviral approaches have not yielded the desired results in clinical trials (12). Recent data suggest that it is not the presence and/or replicative activity of invading viruses in the myocardium that determines outcome, but the virus-triggered abundance of infiltrating leukocytes is an independent risk factor (13, 14). At the acute state of myocarditis in mice, the majority of accumulating leukocytes in inflamed heart tissue are CD11b<sup>+</sup> monocytes and macrophages (15, 16). Consistently, the presence of CD68<sup>+</sup> macrophages is a diagnostic hallmark for human myocarditis (3). Infiltration of immune cells is cytokine/chemokine-dependent. Consistent with previous findings (17), we have demonstrated that not CVB3-mediated cytotoxicity itself, but the overwhelming cytokine response initiated by viral PAMPs is responsible for disease severity. Lower pro-inflammatory cytokine/chemokine production during the early phase of infection paralleled in reduced inflammatory heart tissue damage and protected mice from cardiac failure (14). As monocytes and macrophages are key players that secrete proinflammatory and pro-fibrotic cytokines thereby exacerbating acute and chronic inflammatory injury during myocarditis (4, 18), effector molecules that modulate their differentiation, activity, and cytokine secretion might be putative drug targets for myocarditis. We have previously described the precise targeting of inflammatory monocytes and their precursors by optimized lipid nanoparticles which were encapsulated with siRNA directed against CCR2 or CD115 (19, 20). Injection of mice with these nanoparticles resulted in rapid blood clearance, accumulation in spleen and bone marrow, and localization to monocytes (19).

Here, we demonstrate RNA sequencing data obtained from endomyocardial biopsies of patients with myocarditis indicating a significantly increased production of Colony Stimulating Factor 1 (CSF-1). The development of monocytes depends on CSF-1 (21) and its receptor CSF-1R/CD115. CSF-1 can be expressed and produced by various cells including monocytes themselves (22). Local production of CSF-1 stimulates tissue-resident macrophage proliferation and reduces apoptosis, thereby influencing cellular survival (23). CSF-1R is expressed on monocytes, macrophages, dendritic cells and their precursors, including "granulocytemacrophage progenitors" (GMP), "monocyte-macrophage DC progenitors" (MDP) and "common monocyte progenitors" (cMoP) (24, 25). CSF-1 receptor signaling is a well-described mechanism that leads to monocyte production from progenitors and stimulates mature monocytes screwing them into a proinflammatory state (26). Based on this, we hypothesized that disruption of the CSF-1 axis in myeloid cells attenuates heart muscle inflammation and the resulting organ damage during myocarditis. Using mouse models of CVB3-induced myocarditis and experimental autoimmune myocarditis, we have found that silencing of CSF-1 upon treatment of mice with CSF-1 siRNA encapsulated nanoparticles substantially mitigated inflammatory heart muscle damage leading to less fibrosis formation and improved heart muscle function without the risk of exacerbating direct viral pathology.

#### MATERIALS AND METHODS

#### Study Approval

All subjects gave written informed consent in accordance with the Declaration of Helsinki. The study protocol was approved by the ethic committee of the Medical Faculty—University of Heidelberg—project 390/2011 "Central biobank of Department Internal Medicine III for research on molecular and genetic markers in patients with cardiovascular disease."

#### RNA-Seq Analysis, Read Processing and Mapping

Patient enrollment and biomaterial processing for RNA-seq analysis of heart biopsies was performed as previously described (27). In detail, biopsy specimens were obtained from the apical part of the free LV wall during cardiac catheterization using a standardized protocol. Biopsies of 1- to 2-mm diameter were immediately washed in ice-cold saline (0.9% NaCl), transferred and stored in liquid nitrogen until RNA extraction. After diagnostic workup of the biopsies (histopathology), the remaining material was used to isolate RNA with an Allprep Kit (Qiagen). RNA purity and concentration were determined using the Bioanalyzer 2100 (Agilent Technologies) with a Eukaryote Total RNA Pico assay for RNA from biopsies. Sequencing libraries were generated using the TruSeq Stranded Total RNA Sample Preparation Kit with Ribo-Zero Human/Mouse/Rat from Illumina, adhering to the standard protocol of the kit. Sequencing was performed using 2 × 75 bp paired end sequencing on an Illumina HiSeq2000 instrument. For transcriptome analysis, raw read files were mapped with STAR v2.4.1c5 using GRCh37/hg19 and the Gencode 19 gene model (http://www.gencodegenes. org/). Read counts were generated with help of subread's feature counts program 6 (subread version 1.4.6.p1), using uniquely mapped reads only (28). Normalization was performed with help of rlog-normalization (29). RNA seq data were deposited to the public repository Gene Expression Omnibus (GEO) - NCBI, accession number GSE120567. RNA seq data for DCM patients are partially demonstrated in (30).

#### Differential Gene Expression- and Gene Set Enrichment Analysis

Differential gene expression analysis of RNA-seq data was carried out within the RStudio framework using the edgeR package (31). Gene set enrichment analysis was performed with KEGG gene sets.

#### Histology and Immunohistochemistry

Human endomyocardial biopsy tissue and murine tissue was stained as described elsewhere (32). For AVM, paraffin embedded organ tissue sections were stained with hematoxylin/eosin (HE) or Masson's trichrome according to standard protocols. Immunohistochemical stains for CSF-1 (rabbit polyclonal, abcam), T lymphocytes (CD3 and CD4) and mononuclear phagocytes (Mac-3) was performed as previously described (32). For EAM, hearts were excised 30 days after primary immunization. Hearts were rinsed in PBS, fixed in 10% formaline for 24 h and embedded in paraffin. Serial 5µm sections were stained with Masson's trichrome staining to quantify fibrotic tissue formation. Severity of EAM was evaluated according to a 6-tier scoring system as previously described (19, 20). All slides were counterstained with hematoxylin. Sections were mounted with Pertex mounting media (Medite). Slides were viewed with a Zeiss Axioskop 40 microscope.

#### Candidate Identification of CSF-1 siRNA

Lysates of several murine cell lines were tested on CSF1 expression. NIH-3T3 cells showed high CSF1 expression, are readily to be transfected, and were therefore used in candidate identification experiments. siRNA loaded lipid-based nanoparticles were generated by Axolabs GmbH (Kulmbach, Germany) as previously described (19). siRNA targeting CSF-1 receptor (CSF-1R, CD115) is described elsewhere (33). To generate siRNAs that target the CSF-1 transcript, NIH 3T3 cells were transfected with siRNAs targeting CSF-1 or non-targeting control siRNA complexed with Lipofectamine2000 Transfection Reagent at 5 and 50 nM final concentration in quadruplicates. Values for CSF-1 were normalized to GAPDH and related to the mean value of three different control siRNAs (100% expression). Optimal siRNA concentration yielding most efficient knockdown of CSF-1 production was obtained with RNA transfections starting at 100 nM in 6-fold dilution steps down to 10 fM. CSF-1 siRNAs that showed the best knockdown in both the dual concentration screen and the concentration response curve screen were used for nanoparticle encapsulation and in vivo experiments.

#### Induction of Acute Viral Myocarditis (AVM) and Experimental Autoimmune Myocarditis (EAM)

For induction of AVM, 5–7 weeks old male A.BY/SnJ mice were infected intraperitoneally (i.p.) with 5 × 10<sup>5</sup> PFU CVB3 (cardiotropic Nancy strain) provided by Klingel (15) and Rahnefeld et al. (32). Original breeding stocks for A.BY/SnJ mice were purchased from the Jackson Laboratory. For EAM, male BALB/c were purchased from Janvier (Saint-Berthevin, France). Myocarditis was induced by subcutaneous injection of an emulsion containing 150 µg myosin peptide SLKLMATLFSTYASAD (PSL GmbH, Heidelberg, Germany) supplemented with complete Freund's adjuvant (CFA) (Sigma-Aldrich, Taufkirchen, Germany) and 5 mg/ml Mycobacterium tuberculosis H37Ra (Sigma-Aldrich, Taufkirchen, Germany). Directly after the initial immunization, mice were injected with 500 ng pertussis toxin (Sigma-Aldrich, Taufkirchen, Germany) i.p. Seven days after the primary immunization, mice received a second subcutaneous injection of 150 µg myosin peptide supplemented with CFA and complemented with Mycobacterium tuberculosis. All mice were housed under standard laboratory conditions with a 12-h light-dark cycle and access to water and food ad libitum. For AVM, the protocol was approved by the Committee on the Ethics of Animal Experiments of Berlin State authorities [G0034/16]. EAM experimental protocols were approved by the institutional review board of the University of Heidelberg, Germany, and the responsible government authority of Baden-Württemberg, Germany (project number 35-9185.81/G-209/12). All mouse studies were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the German animal welfare act, which is based on the directive of the European parliament and of the council on the protection of animals used for scientific purposes. All efforts were made to minimize suffering.

### In vivo Silencing of CSF-1 During AVM and EAM

The optimal CSF-1–targeted siRNA (siCSF-1) was scaled up for in vivo studies. For viral myocarditis, mice were intravenously treated with 0.5 mg/kg nanoparticle encapsulated siLUC or siCSF-1 immediately prior to CVB3 infection and 2, 4, and 6 days after infection. For EAM, nanoparticle treatment started 14 days after the primary immunization with myosin peptide. Animals received four i.v. injections of 0.5 mg/kg siCSF-1 or siLuciferase (LUC)-nanoparticles (control siCD115) per week.

#### Evaluation of Knockdown Efficacy of siCSF-1

Male BALB/c mice received single injections of 1.5 mg/kg lipidbased nanoparticle containing either siLUC or siCSF-1 on three consecutive days. Animals were sacrificed 24 h after the third injection. Bone marrow cells were isolated and prepared for flow cytometry-based sorting of monocytes, which were identified as Lin−(CD90;B220;CD49b;NK1.1;Ly6G,Ter119);F4/80−; CD11c−; CD11b+. Cell sorting was performed on a FACS ARIAII (BD Bioscience, Heidelberg, Germany). RNA from sorted cells was isolated using Trizol (Life Technologies, Darmstadt, Germany). Knockdown efficacy was evaluated using quantitative real-time PCR. Gene expression was normalized to HPRT. The following primers were used: CSF-1: TCCCAT ATGTCTCCTTCCATAAA (fwd), GGTGGAACTGCCAGT ATAGAAAG (rev); CD115: CGAGGGAGATCTCAGCTACA (fwd), GACTGGAGAAGCCACTGTCC (rev). HPRT: GTCAAC GGGGGACATAAAAG (fwd), TGCATTGTTTTACCAGTG TCAA (rev). For the AVM model, spleen tissue was isolated 8 days after virus inoculation and tissue homogenization was performed using a lysis buffer containing 20 mM HEPES, 1 % (v/v) Triton X-100, 4 mM EDTA, 1 mM EGTA, 5 mM TCEP, 50 mM NaF, 5 mM NaPP, 2 mM Na-o-vanadate and Complete <sup>R</sup> protease inhibitor cocktail (Roche). Western blot analysis was performed following standard procedures. After blocking with 5% milk/PBS-Tween at 4 ◦C overnight, membranes were probed with the primary antibody α-CD115 (ab32633, Cell Signalling) and α-actin (Merck Millipore). The bound primary antibodies were detected using IRDye800CW labeled goat anti-mouse secondary antibodies in conjunction with an Odyssey CLx infrared imaging system (Li-Cor Biosciences, Bad Homburg, Germany).

#### Echocardiography

Cardiac function and morphology of mice with AVM were assessed with a VisualSonics Vevo 770 High-Frequency Imaging System with the use of a high-resolution (RMV-707B; 15–45 MHz) transducer during anesthesia with 1.5–2% isoflurane. Temperature and ECG were continuously monitored. For the EAM model, echocardiography was performed in conscious animals on a VisualSonics Vevo 2100 30 days after the first immunization. Standard imaging planes, M-mode, and functional calculations were obtained. For AVM, the parasternal long-axis four-chamber view of the left ventricle (LV) was used to guide calculations of percentage fractional shortening, ventricular dimensions and volumes. M-mode echocardiographic images were recorded at the level of the papillary muscles from the parasternal short-axis view. An experienced reader blinded to treatment performed all measurements. Ejection fraction (EF) and fractional shortening (FS) were calculated based on M-mode measurements.

### Flow Cytometry

Flow cytometric analysis was performed 8 days after infection in AVM and 21 days after the first immunization in EAM. Single cell suspension of bone marrow, spleen and heart tissue were prepared as previously described (34). Hearts were flushed with PBS and homogenized in RPMI 1,640 medium (Biochrom) containing 10% (v/v) fetal calf serum (FCS) (Biochrom), 1% (v/v) penicillin/streptomycin (Pan Biotech), 30 mM HEPES, 0.1 % (w/v) collagenase type 2 (Worthington) and 0.015% (w/v) DNase I (Sigma-Aldrich) at 37◦C at 800 rpm for 30 min. Afterwards, 10 mM EDTA was added. Cells were washed with PBS and passed through a 70µm cell strainer as described in reference (35). For the identification of myeloid cells, cell suspensions were stained with a cocktail of PE-conjugated anti-mouse antibodies targeting hematopoietic lineage markers (B220 for B cells (RA3-6B2, BD Bioscience), CD90.2 for T cells (53-2.1, BD Bioscience), CD49b for NK cells (DX5, eBioscience), NK-T/NK Cell Antigen for NK cells (U5A2-13, BD Bioscience) and Ter-119 for erythroid cells (TER-119, BD Bioscience)) and fluorescentdye conjugated antibodies against the following cell surface markers: CD45.2 (104, Brilliant Violet 711TM, BioLegend), CD11b (M1/70, PE-CF594, BD Bioscience), Ly6G (1A8, PerCP/Cy5.5, BioLegend), Ly6C (HK1.4, Pacific BlueTM, BioLegend), CD11c (N418, Brilliant Violet 510TM, BioLegend), I-A[b] (AF6-120.1, FITC, BD Bioscience) and F4/80 (BM8, APC, BioLegend). Cells were stained in PBS containing 2% FCS, 2 mM EDTA for 20 min at 4◦C. For the identification of lymphoid cells, cell suspensions were stained with fluorescent-dye conjugated anti-mouse antibodies against CD45.2 (104, Brilliant Violet 711TM, BioLegend), CD3e (145-2C11, PerCP/Cy5.5, BioLegend), CD4 (RM4-5, V500, BD Bioscience), CD8a (53-6.7, Pacific BlueTM, BD Bioscience), B220 (RA3-6B2, FITC, BioLegend) and CD19 (6D5, APC, BioLegend). The antibody staining was followed by a cell viability stain (Fixable Viability Dye eFluor <sup>R</sup> 780, eBioscience) according to the manufacturer's protocol.

Monocytes were identified as Linlow (CD90;B220;CD49b;NK1.1;Ly6G,Ter119), F4/80low, CD11clow, CD11bhigh; or Fixable Viability Dyelow, CD45.2high, CD11bhigh , (B220, CD90.2, CD49, NK-T/NK Cell Antigen, Ter-119)low, Ly6Glow, SSClow, F4/80low and CD11clow and further differentiated according to Ly6C-expression. Inflammatory monocytes express high levels of Ly6C and patrolling monocytes express low levels of Ly6C. Macrophages were identified as Fixable Viability Dyelow, CD45.2high, CD11bhigh, Linlow, Ly6Glow, SSClow, F4/80high and CD11clow/high. Dendritic cells were identified as Fixable Viability Dyelow, CD45.2high , CD11bhigh, Linlow, CD11chigh and MHC IIhigh (compared to isotype control). Neutrophils were identified as Fixable Viability Dyelow, CD45.2high, CD11bhigh, Linlow, Ly6Ghigh and SSChigh. B cells were gated as Fixable Viability Dyelow, CD45.2high, CD3low, B220high and CD19high. T cells were gated as Fixable Viability Dyelow, CD45.2high, B220low, CD3high and either CD4high or CD8high. For identifying proliferating GMPs mice received two s.c. injections of 1 mg/kg Bromdesoxyuridin (BrdU) 12 and 24 h before the animals were sacrificed. BrdU was stained using BrdU flow kit (BD Biosciences). Proliferating GMPs were identified as (CD90;B220;CD49b;NK1.1;Ly6G;CD11b;CD11c;IL-7R;Sca-1)low and (CD117;CD34;CD16/32;BrdU)high .

For the assessment of quantitative data, 123 count eBeads (eBioscience) were used according to manufacturer's protocol. Data were acquired on a FACS Verse (BD Biosciences, Heidelberg, Germany) or on a LSR II (BD Bioscience) and analyzed with FlowJo v10.0 software (FLOWJO, Ashland, United States). Reported cell numbers were normalized to the weight of total hearts, yielding the number of respective cell fraction per mg tissue.

#### Determination of Viral Load in Heart Tissue

Plaque assays were performed in triplicates on sub-confluent green monkey kidney cell monolayers as described recently (32). In situ hybridization of CVB3 RNA was performed using probes generated with the DIGoxigenin (DIG) RNA labeling kit (Roche) and the pCVB3-R1 plasmid. Plasmid cDNA was linearized with SmaI (36); all other steps were conducted as previously described (37). DIG-labeled CVB3 RNA was detected using a horseradish-peroxidase-conjugated DIG antibody (Roche 1:100). HistoGreen (Linaris) was used as a substrate. All slides were counterstained with hematoxylin.

#### High-Sensitive (hs)-Troponint (TnT)

Blood was sampled by facial vein puncture and collected in a heparinized capillary. Thereby obtained plasma was diluted 1:15 in PBS. hs-TnT was determined by the electrochemiluminescence method (ECLIA; Elecsys 2010 analyzer) according to the method described in reference (38).

#### Statistics

Statistical analysis of the data was performed in GraphPad Prism v6.00/v.700 for Windows (GraphPad Software, La Jolla, California, United States). Logarithmic data (virus titer, semiquantitative RNA quantification) measured on a linear scale was transformed logarithmically prior to data plotting and data analysis. Data summary is indicated on plots as mean ± SD unless stated otherwise. Unpaired t-tests were used for two group comparisons. If samples had unequal variances (determined by an F-test), an unpaired t-test with Welch's correction was used. For multiple group comparison unequal variance versions of ANOVA (1-way or 2-way ANOVA) were performed followed by a Sidak-Holm's multiple comparison test. The significance threshold for all tests was set at the 0.05 level.

### RESULTS

## Increased Abundance of CSF-1 in Heart Muscle Tissue During

Myocarditis/Inflammatory Cardiomyopathy RNA-seq analyses revealed that myocarditis results in a diverse transcriptional response in the patient's heart tissue. We observed 1963 differentially expressed genes in biopsies taken from patients with acute myocarditis vs. patients with a non-inflammatory dilated cardiomyopathy (DCM) (**Figure 1A**). Gene set enrichment analysis (GSEA) revealed that a decent amount of differentially expressed genes participates in inflammatory processes (especially cytokines and cytokine receptors) and differentiation of hematopoietic cell lineages (**Table 1**). Myocarditis leads to a massive infiltration with immune cells into heart tissue. Monocytes represent the most prominent leukocyte population both during virusmediated and experimental autoimmune myocarditis (16, 32). Monocyte production and maturation is strongly dependent on CSF-1 and CSF-1R, and, both effector molecules were identified in two gene sets mentioned above. Our data indicate a pronounced up-regulation particularly of CSF-1 and CSF-1R in endomyocardial specimen from patients with myocarditis/inflammatory cardiomyopathy (**Figure 1B**). Immunohistochemical stain of heart tissue from patients with myocarditis revealed CSF-1 expressing cells only within inflammatory foci, with a strong focus on mononuclear immune cells (**Figure 2A**). Altogether, these data argued toward a significant contribution of monocytes/macrophages to the cardiac CSF-1 expression, which we found in patients with inflammatory heart disease.

Local production of CSF-1 stimulates tissue-resident macrophage proliferation and reduces apoptosis (23). In addition to this function and a direct role of CSF-1 during monocyte development, it might also influence pro-fibrotic processes under inflammatory conditions. Since inflammation and fibrosis are hallmarks of inflammatory heart disease, we aimed to investigate the pathophysiological influence of CSF-1 with respect to manifestation of inflammatory heart tissue injury. First, we determined CSF-1 production in a mouse model of CVB3-induced myocarditis. We performed immunohistochemical stains to evaluate CSF-1 abundance during viral myocarditis. Consistent with our findings in patients, CSF-1 production was increased within inflammatory foci at the acute state of myocarditis in mice (**Figure 2B**). Since monocytes/macrophages represent the major infiltrating cell population in acute myocarditis, it is very likely that these cells are also involved in CSF-1 production. By double labeling immunohistochemistry we found CSF-1 protein expression in a part of Mac-3 positive mononuclear phagocytes within the cardiac inflammatory lesions (**Figure 2C**).

#### Nanoparticle-Encapsulated siRNA Effectively Downregulates CSF-1 Production in Monocytes

CSF-1 can be expressed and produced by various cells including monocytes themselves (22). siRNA encapsulated in lipid-based nanoparticles has been shown to effectively downregulate target genes in monocytes and their lineage progenitors (19, 20). Furthermore, in vivo knockdown of CSF-1R and monocyte depletion with nanoparticle-encapsulated siRNA has recently been demonstrated for ischemic heart disease (33). Thus, in order to investigate the pathophysiological function of CSF-1 production by monocytes/macrophages on inflammatory tissue damage during myocarditis, we decided to use a nanoparticleencapsulated siRNA approach to target CSF-1 production in myeloid cells. To identify siRNAs leading to highly efficacious suppression of CSF-1 production, 24 different siRNAs targeting CSF-1 were investigated regarding their influence on CSF-1 mRNA levels in vitro (**Figure 3A**). Six different siRNAs, which yielded optimal in vitro suppression of CSF-1 mRNA production, were further investigated for their knockdown efficacy. Next, we screened the respective CSF-1 directed siRNA regarding to the concentration-dependent knockdown efficacy (**Figure 3B**) and selected the most efficacious siRNA for in vivo nanoparticle studies. Naive BALB/c mice were intravenously inoculated with


RNA-Sequencing data of endomyocardial biopsies from patients with clinically diagnosed myocarditis or dilated cardiomyopathy as a control were analyzed for relative expression of different gene sets. Indicated are the names, description and p-values (with FDR-adjustment) of the most relevant gene sets.

from patients with acute myocarditis were stained by immunohistochemistry. (A) Representative micrographs stained with an anti-CSF-1 antibody [left column] or with a secondary antibody only [right column] are depicted. top: scale bar = 120µm; bottom: scale bar = 60µm. (B) Heart tissue sections were obtained from CVB3-infected A.BY/SnJ mice on day 8 p.i. Representative micrographs of anti-CSF-1 stained heart tissue are shown [scale bar = 36µm]. (C) In addition, cardiac sections from mice were double-stained with an antibody directed against Mac-3 (red) [left: scale bar = 36µm; center: scale bar = 12µm] and against CSF-1 (green). As control, Mac-3 stained tissue sections were counterstained omitting the anti-CSF-1 directed antibody [right: scale bar = 12µm].

1 mg/kg of this nanoparticle-encapsulated siRNA (termed siCSF-1) on three consecutive days. Expression of CSF-1 was found to be effectively downregulated in monocytes that were sorted from spleen of siCSF-1 treated mice and further evaluated by quantitative PCR analysis (**Figure 3C**). Since pathogens are frequently involved in the pathogenesis of myocarditis (5), as a next step we set up an experimental approach to decipher the CSF-1 axis using nanoparticle-encapsulated siRNA in a mouse model of virus-mediated myocarditis. A.BY/SnJ mice with high hereditary susceptibility for the development of acute viral myocarditis (AVM) were treated with siCSF-1 or respective controls directly prior to infection with cardiotropic CVB3 (Nancy). siCSF-1 treatment was repeated every other day until mice were sacrificed 8 days after infection at the respective peak of infiltration in heart muscle (**Figure 3D**) (6–8). Following this protocol, we monitored the abundance of CSF-1 receptor levels in spleens of infected mice, which allowed us to conclude on the efficiency of siCSF-1 treatment during infection. Consistent with virus-mediated mobilization of monocytes/macrophages from bone-marrow sources, viral infection resulted in increased CSF-1R levels in the spleen. In siCSF-1-treated mice, we found reduced CSF-1R levels being indicative of suppressed myeloid cell mobilization upon siCSF-1 injection during AVM (**Figure 3E**).

### siRNA-Mediated Knockdown of CSF-1 Attenuates Virus-Mediated Pathology

Since we found reduced mobilization of monocytes/macrophages in siCSF-1-treated mice during CVB3 infection, this mouse model allowed us to delineate the pathophysiological role of CSF-1 production particularly by monocytes/macrophages during viral myocarditis. First, we questioned whether siCSF-1 treatment manipulated the viral load during AVM. The viral burden as reflected by the amount of infectious viral particles (**Figure 4A**) was not substantially influenced by siCSF-1 in heart tissue at the acute state of infection. Thus, targeting the CSF-1 axis represents a safe approach regarding to control of virus dissemination and replication in A.BY/SnJ mice. During virusmediated myocarditis, there is a strong spatial-temporal relation

homogenates (n = 6 mice per group) 8 day after virus inoculation. CSF-1R (high molecular weight band) and the cytoplasmic domain of CSF-1R (around 55 kDa) are depicted. Fluorescence was quantified by the Image Studio Lite Ver 5.2 software. Signal intensity was normalized to actin and is depicted as relative expression levels compared to siLUC-treated mice in the bar graph. Unpaired t-tests were used; p-values are indicated on the graph and significant differences (p < 0.05) are marked with blue color.

between virus-induced cellular injury and the emergence of inflammatory foci in heart tissue (39). Likewise, viral genome abundance as detected by CVB3 in situ hybridization was spatially connected with high-grade inflammation and most impressive in siLUC-treated mice (**Figure 4B**). Since CVB3 does not only target the heart, but also replicates in the pancreas, we also determined the magnitude of virus-induced pancreas destruction and found similar tissue injury in both siLUC- and siCSF-1-treated groups (**Figure S1**).

Next, CVB3-infected mice that received siCSF-1 or siLUC as a control were followed for global signs of acute infection. siCSF-1 treatment profoundly attenuated overall virus-induced pathology as represented by significantly less pronounced body weight reduction (**Figure 4C**) and only a minor loss of body temperature

FIGURE 4 | Depletion of CSF-1 attenuates virus-mediated pathology. Mice with AVM were subjected to CSF-1 siRNA treatment as indicated in Figure 3D. (A) Infectious virus particles were determined in heart tissue homogenates by plaque assay. Data summary is mean ± SEM. A student's t-test was conducted and the p-value is shown. (B) To localize viral RNA in infected heart tissue, in situ hybridization for the detection of the CVB3 genome was performed and slides were counterstained with hematoxylin/eosin. Representative micrographs are depicted (scale bar = 60µm). During viral infection, mice were monitored for body weight (C) and body temperature (D) at the indicated points in time. Dots represent mean ± SEM. Repeated measurements versions of two-way ANOVA were performed followed by a Sidak- Holm multiple comparison procedure. P values are indicated (blue color indicates p < 0.05; only significant results are depicted on the graph). (E) To assess injury of cardiomyocytes prior to peak of inflammation, blood was sampled 5 days after infection by facial vein puncture and high sensitive (hs) troponin T plasma levels were determined. Obtained results were normalized to the results obtained with blood samples from non-infected siLUC-treated mice and are depicted (Continued)

FIGURE 4 | as fold changes. Data summary is mean + SD. Repeated measurements two-way ANOVA was performed followed by a Sidak-Holm's multiple comparison test and the p value is depicted. (F) After sacrificing mice 8 days after infection, heart tissue sections were stained with hematoxylin/eosin. (G) To quantify cell infiltration, single cell suspensions of heart tissue obtained from naive mice (uninfected mice that did not receive siRNA treatment; white bars, n = 4) as well as AVM and siRNA-treated mice (siLUC: gray squares, n = 7; siCSF-1: green squares, n = 8) were stained with CD45 antibodies to quantify total leukocyte count in the heart. (H) Cardiac function was assessed by echocardiography prior to CVB3 infection in A.BY/SnJ mice (baseline) by an experienced and blinded investigator. Mice were allocated to respective groups: siLUC and siCSF-1. In all CVB3-infected mice, echocardiography was repeated 8 days after CVB3 infection (siLUC n = 16; siCSF-1 n = 17 mice). Data were analyzed regarding putative alteration during AVM in the respective treatment groups (day 8 after infection vs. baseline measurements of the same cohort). Relative changes of stroke volume, heart rate and cardiac output compared to baseline measurements were calculated for each group and these fold changes are depicted for siLUC and siCSF-1-treated groups. One-sample t-tests were performed to compare baseline measurements and values obtained 8 days after infection. All p values are depicted, p < 0.05 are in blue color.

during infection in comparison to controls that received siLUC (**Figure 4D**). Overall diminution of virus-mediated pathology under siCSF-1 influence was further corroborated by a significant reduction of cardiac troponin T serum levels as a heartspecific sign of tissue damage (**Figure 4E**). Cardiac troponin T at an early state of myocarditis might reflect both direct virus-induced cytotoxicity and tissue destruction by innate mediators of the immune response. In line with this, analysis of heart tissue obtained from siCSF-1 treated animals sacrificed 8 days p.i. revealed distinct differences. Histological staining of heart tissue (**Figure 4F**) demonstrated a profound myocarditis in siLUC-treated A.BY/SnJ mice and in contrast to that only moderate signs of myocarditis after siCSF-1 treatment. Since viral injury of cardiomyocytes provokes an inflammatory response that significantly contributes to tissue damage and functional impairment of the heart (32), next we quantified infiltration with CD45<sup>+</sup> immune cells into hearts from siCSF-1 and siLUC-treated mice by flow cytometry. We found a significant reduction of infiltrating leucocytes in siCSF-1 treated mice (**Figure 4G**), thus indicating reduced inflammatory organ damage under suppression of the CSF-1 axis. Following up on observed systemic and heart-tissue specific responses to siCSF-1 treatment, siCSF-1 effects on cardiac performance were assessed by echocardiography during the inflammatory peak of viral myocarditis. In siLUC-treated, infected A.BY/SnJ mice, both the stroke volume and cardiac output were significantly reduced in comparison to baseline measurements (**Figure 4H**) and **Table S1**. Consistent with its heart-directed effects, siCSF-1 treatment mitigated these detrimental changes and CVB3 infection in this group resulted only in minor, non-significant reduction of cardiac performance (**Table S1**).

#### siRNA-Mediated Knockdown of CSF-1 Diminishes Immune Cell Infiltration During Acute Viral Myocarditis

As a next step, we aimed to determine whether siCSF-1 specifically influenced infiltration with monocytes/macrophages or whether other immune cells were affected as well. Immunohistochemistry using antibodies directed against marker proteins for myeloid (Mac3) and T cells (CD3 and CD4) indicated reduced infiltration of these respective immune cell populations in siCSF-1-treated mice during AVM (**Figures 5A,B**). These findings were corroborated by the results obtained from a quantitative flow cytometry-based analysis of the different immune cell populations in infected mouse hearts. We detected 756 ± 63 CD11b+/lineage<sup>−</sup> cells in siLUC- vs. 273.3 ± 30.8 CD11b+/lineage<sup>−</sup> cells/mg heart tissue in siCSF-1 treated mice (p < 0.0001) and 198 ± 34 T cells in siLUC- vs. 95 ± 12 T cells/mg heart tissue in siCSF-1 treated mice (p = 0.02; **Figures 5C,E**). The vast majority of infiltrating myeloid cells belonged to the pool of inflammatory monocytes (**Figure 5D**). Consistent with the influence of CSF-1 on monocyte recruitment and differentiation, inflammatory monocytes were highly significantly reduced in infected mouse hearts upon siCSF-1 treatment. siCSF-1 also led to a significant reduction of patrolling monocytes, macrophages and dendritic cells (**Figure 5D**), which might all originate from inflammatory monocytes. In correspondence to previous reports (15), the pool of invading T cells during AVM was majorly comprised of CD4<sup>+</sup> T cells. Comparable to siCSF-1-induced effects on myeloid cell infiltration, we also found a significant reduction of the CD4<sup>+</sup> T cell count in heart tissue upon siCSF-1-treatment (**Figure 5E**).

#### siRNA-Mediated Knockdown of CSF-1 Mitigates Inflammatory Heart Tissue Damage in a Mouse Model of Experimental Autoimmune Myocarditis

siCSF-1 treatment impressively improved cardiac function upon mitigating the inflammatory damage response during acute viral myocarditis. A high inflammatory disease burden at the acute state might directly translate into the manifestation of a chronic functional impairment. Long-term sequela of acute myocarditis involve cardiac remodeling processes with substantial fibrosis formation and a reduction of systolic cardiac function (4). Experimental autoimmune myocarditis (EAM) represents an excellent model that enables researchers to follow inflammatory disease progression from acute to chronic states of myocardial dysfunction (40). Therefore, we investigated the pathophysiological role of CSF-1 also in EAM, and started siCSF-1 treatment upon manifestation of myosin-heavy chaindirected autoimmunity 14 days after the first immunization (**Figure 6**). This application strategy might also be considered as a therapeutic regime starting upon manifestation of acute disease. Similar to the AVM model, intravenous siCSF-1 nanoparticle application was repeated every 48 h for a 1-week course. First, we performed a quantitative flow cytometry-based analysis of inflammatory monocytes in the injured hearts directly after this siCSF-1 treatment period. As expected from our results obtained in AVM, we indeed found a reduced number of inflammatory Ly6Chigh monocytes in siCSF-1-treated animals compared to the siLUC-treated controls during EAM (**Figure 6B**). To investigate

FIGURE 5 | bars = 60µm). Further differentiation by flow cytometry (Figure S2) was performed. (C) Total infiltrating myeloid cells (identified as CD45+, CD11b+, lymphoid lineage<sup>−</sup> life single cells) were quantified. (D) Myeloid cells were further differentiated according to the gating strategy depicted in Figure S2A. (E) Equally, lymphoid cells were further analyzed. Representative flow cytometry dot blots of siLUC- and siCSF-1-treated groups are depicted. Unpaired t-tests were performed between siLUC- and siCSF-1-treated groups and p-values are shown. Significant differences (p < 0.05) are marked with blue color.

whether CSF-1 promotes monocyte development from bonemarrow sources during EAM, we next evaluated the influence of CSF-1 knockdown on granulocyte-monocyte progenitor (GMP) cell numbers. siRNA-mediated knockdown of CSF-1 led to markedly increased numbers of GMPs in the bone marrow during EAM (**Figure 6C**). However, proliferation rates of GMPs did not differ significantly between siLUC- and siCSF-1-treated groups (**Figure 6D**). To further validate that CSF-1 knockdown leads to an arrest of progenitor cells in the bone marrow, we measured myeloid cell numbers in the blood and found reduced numbers 21 days after EAM induction (**Figure 6E**).

Based on our finding of reduced infiltration with inflammatory monocytes upon siCSF-1 treatment, next we followed mice for a total of 30 days after EAM induction and determined the formation of fibrotic scars as an integral hallmark of cardiac remodeling at this chronic disease state (**Figure 7A**). Knockdown of CSF-1 production resulted in a significant reduction of fibrosis formation in the heart muscle as indicated by quantitative assessment of Masson's trichrome stains (**Figure 7B**). Consistent with this reduction of longterm, inflammation-mediated tissue damage in the siCSF-1 group, functional investigation of cardiac performance by echocardiography revealed an improved ejection fraction in siCSF-1-treated animals compared to siLUC-treated control animals (**Figure 7C**). Chronic disease in siLUC-treated mice was mirrored by a substantial reduction of cardiac contractility. As a proof of principle, we also tested whether knockdown of CSF-1R expression could exert similar effects during disease course. Therefore, upon establishment of autoimmune myocarditis mice were treated with a nanoparticle-encapsulated siRNA that specifically targets CSF-1R (33). Formation of cardiac fibrosis assessed 30 days after EAM induction was found to be diminished in comparison to siLUC-treated controls, and was reduced to similar levels as achieved by siCSF-1-treatment (**Figure 7B**). Likewise, echocardiographic imaging of siCSF-1R-treated animals demonstrated significant improvement of cardiac performance as indicated by higher left ventricular ejection fraction in comparison to siLUC nanoparticle-treated animals (**Figure 7C**). Altogether, our data demonstrate that silencing the CSF-1 axis hampers monocyte development and substantially mitigates inflammatory heart tissue damage both in a mouse model of viral and autoimmune-mediated myocarditis. Moreover, initiation of CSF-1 knockdown upon manifestation of inflammatory tissue damage attenuates the manifestation of debilitating long-term sequela of acute inflammatory injury, and this parallels in preserved systolic contractility of the heart muscle (**Figure 8**).

#### DISCUSSION

One of the major causes of heart failure particularly in young patients is myocarditis. Direct viral cytotoxicity stimulates infiltration and immune response activation leading to pathogen clearance and resolution of organ damage. Nevertheless, in immune-genetically predisposed individuals there is also an adverse scenario where pathogen-induced immune response activation subsequently induces overwhelming inflammatory cytokine response and detrimental immunopathology or autoimmune processes, both leading to cardiac remodeling and fibrotic scarring. It appears to be a slim line between the induction of inflammation to fight the virus and exaggerated immune responses that begin to be deleterious. We here downregulated an important branch of the innate immunity the development of monocytes/macrophages—by siCSF-1 treatment for approximately a week during the acute phase of a viral infection, yet found no impairment on pathogen control. In fact, silencing CSF-1 production impressively mitigated acute inflammatory heart tissue damage and attenuated the development of a debilitating long-term sequela of acute inflammatory injury to the heart. This nanoparticle-mediated immune-modulation improved the course of disease both in acute viral myocarditis and autoimmune inflammatory heart disease and importantly, did not adversely influence viral burden and clearance of infectious particles. As it might be hypothesized that attenuating inflammatory activity could allow for enhanced virus-induced cell death, this is not what we observed with siCSF-1 treatment. Our finding is in line with previous data: out of 46 studies that intervened the immune response, more than 90% found no adverse effect on viral load (4). In line with this, we could recently demonstrate that inhibition of cellular proteolysis in immune cells by the immunoproteasome inhibitor ONX 0914 is highly efficient to reverse high grade AVM and this protective effect was attributed to maintained immune cell homeostasis, but not to direct antiviral aspects (14).

A.BY/SnJ mice that were used in this study for AVM are characterized by a high hereditary susceptibility to CVB3 induced cardiomyopathy (6, 15, 41). Upon viral infection of cardiomyocytes, heart tissue injury during the first days is mainly attributed to direct virus-induced cytotoxicity, while the activation of local type I interferon responses (T1IFN) is substantially hampered in this strain (7, 8). This phase of ongoing viral replication is accompanied and succeeded by the recruitment of cells of the innate immune response such as NK cells and monocytes (42). Later on, infiltration with immune cells of the adaptive immune response such as T and B cells follows (15). Analysis of the composition of leukocytes in the heart revealed that 8 days after viral infection a general reduction in invading immune cells was observed which suggests that hampering early responders, such as monocytes e.g., by reduced production of lymphocyte-attracting chemokines might dampen the infiltration and activation of subsequent populations as well. Although we found an overall reduction of CD4<sup>+</sup> T cells in the heart in siCSF-1 treated mice, we cannot conclude on the influence of silencing the CSF-1 axis on CD4<sup>+</sup> T cell

(Continued)

FIGURE 6 | to its potential to manipulate EAM in comparison to siLUC (control). Therefore, mice were treated with 0.5 mg/kg siRNA intravenously every other day starting 14 days after EAM induction, and mice were sacrificed 21 days after the first immunization. (B) Representative dot plots (left) and enumeration (right) of inflammatory Ly6Chi monocytes in heart tissue of siLUC- and siCSF-1-treated mice (n = 8). (C) Representative dot plots (left) and quantification of granulocyte-monocyte progenitors (GMPs) (right) found in bone marrow of siLUC- and siCSF-1-treated mice during EAM (n = 12). (D) Representative FACS plots (left) and quantification (right) of BrdU incorporation in GMPs of the bone marrow in siLUC- and siCSF-1 treated mice (n = 8). (E) Representative FACS density plots (left) and quantification of monocytes/macrophages (right) in the blood of siLUC- and siCSF-1- treated mice (n = 7). Unpaired t-tests were used. p-values are indicated on the graph and significant differences (p < 0.05) are marked with blue color.

differentiation e.g., into regulatory T cells or Th1 and Th17 cells. From our data we cannot conclude on possible additional effects e.g., of regulatory CD4<sup>+</sup> T cells that are known to mitigate the inflammatory tissue damage in the heart (43, 44). CSF-1 facilitates myeloid cell differentiation, monocyte survival, and macrophage proliferation. It was recently shown that CSF-1R also plays a role in splenic monocytopoiesis (33). Monocytes are also important producers of CSF-1 themselves. Hume and MacDonald suggested that modulation of the CSF-1 axis may be beneficial under pathological conditions (45). We here show that siRNA-mediated knockdown of CSF-1 in monocytes and its imminent precursors leads to a significant reduction of inflammatory Ly6Chi monocyte numbers in the inflamed heart. We speculate that this observation results from arresting the CSF-1-dependent monocyte/myeloid cell development at precursor stages. This hypothesis is supported by detection of increased GMP cell numbers in the bone marrow in siCSF-1-treated animals during EAM. Although increased cell numbers may also arise from increased proliferation rates, this situation seems to be unlikely, since the percentages of proliferating GMPs in the bone marrow of siCSF-1- and siLUC-treated animals did not differ. Consistently, the numbers of myeloid cells were reduced

damage in the heart and preserving cardiac function at acute and chronic disease states without compromising innate immunity's critical defense mechanisms.

in the blood stream of siCSF-1-treated animals during EAM. In agreement with this and the fact that other than in siLUCtreated mice enhanced monocyte counts were not observed in the blood in siCSF-1-treated mice during AVM (data not shown), we found a reduction of CSF-1R abundance in the spleen during AVM in the siCSF-1 group. The spleen itself is a source for monocytes with local production and from which they are recruited to the site of inflammation (34). CSF-1R is expressed throughout the mononuclear phagocyte system, which is primarily composed of monocytes and macrophages (22, 46). A reduction of CSF-1R expression might also be indicative for reduced numbers of monocytes and macrophages. These findings further underscore our assumption of a halted monocyte/macrophage production due to CSF-1 knockdown during inflammation. Mice that carry a deleterious mutation in CSF-1 develop characteristic skeletal malformation, caused by defective osteoclasts. In the context of an inflammatory disease, such as atherosclerosis however, deletion of CSF-1 results in dramatically reduced atherosclerotic plaque size (47, 48). After myocardial infarction an upregulation of CSF-1 is observed in ischemic areas for more than 5 days after the injury (49). In this context, CSF-1 also appears to exhibit indirect effects by regulating chemokine production (49). In the absence of CSF-1, GM-CSF controls the differentiation of selected macrophage subsets and may lead to an enhancement of macrophage lineage numbers (50). Under certain conditions, GM-CSF induces an inflammatory program marked by increased IL-1, IL-6 and TNF-α secretion, whereas the presence of CSF-1 suppresses the production of pro-inflammatory signals (51). Others have reported that administration of CSF-1 in EAM between days 21 and 29 after disease induction ameliorated cardiac fibrosis and left ventricular dysfunction by preventing the accumulation of fibroblasts (52). Notably, no effects were observed in the above mentioned study when CSF-1 was administered at later stages. These conflicting results stress the importance of timing in CSF-1 modulation. The reduced numbers of inflammatory monocytes due to silencing myeloid CSF-1 from days 14 to 21 observed in this study may outweigh the beneficial effects of CSF-1 in the inflamed heart at these early stages. In addition, macrophages have been shown to release pro-fibrotic cytokines such as TGF-β, which causes the differentiation of fibroblasts to myofibroblasts and a massive deposition of extracellular matrix proteins such as collagens (53). Since we see a reduction of myeloid cells during myocarditis in response to siCSF-1 treatment and this includes macrophages, we propose that thereby achieved reduction of pro-fibrotic signals most likely contributes to lower scar formation as observed at advanced states of EAM.

It has been reported that other cell types beyond cells of the myeloid lineage are capable of CSF-1 expression under inflammatory conditions, including fibroblasts and endothelial cells (54, 55). These cells may also influence monocyte production/proliferation during myocarditis. Nevertheless, at the acute state of virus-mediated myocarditis, we observed that Mac-3 positive cells like monocytes/macrophages, which infiltrate the heart, may represent the main producers of CSF-1. Many studies using different pathogenic models of bacterial, viral and fungal infections have highlighted the importance and requirement of TNF- and NO-producing monocytes as facilitators of the resolution of an infection (56). Also, during inflammation bone-marrow derived monocytes can mature into macrophages (56) and macrophage-depletion results in increased virus titers in infected mouse hearts (17). Therefore, it was somewhat surprising at the first glance that depletion of monocytes upon siCSF-1 treatment actually improved virusmediated pathology. We found no experimental evidence that virus dissemination, control and elimination had been adversely affected by such immune-modulating intervention. Since virus titers were not significantly influenced by siCSF-1-treatment, the impressively reduced inflammatory injury and preserved cardiac function in this group is not influenced by direct virus-mediated effects. These data suggest that virus-induced inflammation can be segregated from pathways that promote and limit virus infection and CSF-1-induced monocyte maturation. Monocytes can be directly activated by CVB3 infection resulting in the production of pro-inflammatory cytokines (57). Thereby, these cells can contribute to a strong inflammatory response and eventual tissue damage in the myocardium. Further experimental evidence for an adverse function of myeloid immune cells comes from macrophage depletion studies, where myocardial injury and formation of cardiac fibrosis were substantially diminished despite and in clear contrast to increased viral burden during AVM (17). Interestingly and in line with our findings in siCSF-1 treated A.BY/SnJ mice, macrophage depletion did not adversely affect clearance of infectious virus particles (17). We conclude that attenuated innate immune cell mobilization and hampered CSF-1-driven differentiation of innate myeloid cells in siCSF-1-treated mice directly suppresses cytotoxicity induced by cytokine production and/or infiltration with lymphocytes in viral myocarditis. Although there is a clear causal relationship between T1IFN-mediated suppression of viral load in infected cardiac cells and attenuation of inflammation and chronic tissue damage (8, 32), several publications including work of our group support the concept that—independent of direct virus-induced cell injury—particularly monocytes and macrophages are important players in inflammation and chronic organ damage in response to Coxsackievirus infection (4, 14, 58, 59).

Taken together, modulation of the CSF-1 axis in the myeloid cell lineage with siRNAs at early stages has beneficial acute and long-term effects in both viral and autoimmune myocarditis. Our data support the notion that particularly infiltrating myeloid cells contribute to acute and chronic functional impairment in inflammatory heart disease. Since pathogen control was not influenced upon suppression of the CSF-1 axis in myeloid cells, this study yields important insights for translating the pathophysiological role of CSF-1 from animal models to putative novel therapeutic targets for patients with inflammatory heart disease.

### ETHICS STATEMENT

The requested information is provided in the material and method section of the manuscript.

## AUTHOR CONTRIBUTIONS

AB and FL conceptualization. AB, FL, IM, CG, MK, MS, BM, JH, and KK methodology. IM, CG, KK, MS, AH, VE, H-PV, DA, JH, and MK investigation. IM and CG formal analysis. CG and IM visualization. FL, AB, IM, and CG writing-original draft. HK drafting/revising critically. AB, FL, and HK funding acquisition. AB and FL supervision.

### FUNDING

AB is supported by the Foundation for Experimental Biomedicine Zurich, Switzerland. CG received a MD scholarship from the Berlin Institute of Health. This project was additionally supported by DFG VO 1602/4-1 to AB, DFG KL 595/2-3 to KK, and CRC 1292/project 2 to AB. FL is supported by the DZHK (German Centre for Cardiovascular Research), the German Research Foundation (DFG, LE 2530/2-1), the German Heart Research Foundation, and the BMBF (German Ministry of Education and Research; Project DeCaRe).

### ACKNOWLEDGMENTS

We appreciate the excellent technical support of Anika Lindner, Karolin Voß and Sandra Bundschuh, Stefanie Schelenz und Martin Taube. We acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité–Universitätsmedizin Berlin.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02303/full#supplementary-material

#### REFERENCES


coxsackieviral clearance and accelerates myocardial disease. J Virol. (2014) 88:5087–99. doi: 10.1128/JVI.00184-14


macrophage phenotypes display differences in cytokine profiles and transcription factor activities: implications for CSF blockade in inflammation. J Immunol. (2007) 178:5245–52. doi: 10.4049/jimmunol.178.8.5245


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

Copyright © 2018 Meyer, Goetzke, Kespohl, Sauter, Heuser, Eckstein, Vornlocher, Anderson, Haas, Meder, Katus, Klingel, Beling and Leuschner. 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.

# Proteasomal Protein Degradation: Adaptation of Cellular Proteolysis With Impact on Virus—and Cytokine-Mediated Damage of Heart Tissue During Myocarditis

#### Antje Beling1,2 \* and Meike Kespohl 1,2

<sup>1</sup> Charité—Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health (BIH), Institute of Biochemistry, Berlin, Germany, <sup>2</sup> Deutsches Zentrum für Herz-Kreislauf-Forschung (DZHK), Partner Site Berlin, Berlin, Germany

#### Edited by:

Michael H. Lehmann, Ludwig-Maximilians-Universität München, Germany

#### Reviewed by:

George W. Booz, University of Mississippi Medical Center School of Dentistry, United States Jason Brice Weinberg, University of Michigan, United States

> \*Correspondence: Antje Beling antje.beling@charite.de

#### Specialty section:

This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology

Received: 19 September 2018 Accepted: 24 October 2018 Published: 28 November 2018

#### Citation:

Beling A and Kespohl M (2018) Proteasomal Protein Degradation: Adaptation of Cellular Proteolysis With Impact on Virus—and Cytokine-Mediated Damage of Heart Tissue During Myocarditis. Front. Immunol. 9:2620. doi: 10.3389/fimmu.2018.02620 Viral myocarditis is an inflammation of the heart muscle triggered by direct virus-induced cytolysis and immune response mechanisms with most severe consequences during early childhood. Acute and long-term manifestation of damaged heart tissue and disturbances of cardiac performance involve virus-triggered adverse activation of the immune response and both immunopathology, as well as, autoimmunity account for such immune-destructive processes. It is a matter of ongoing debate to what extent subclinical virus infection contributes to the debilitating sequela of the acute disease. In this review, we conceptualize the many functions of the proteasome in viral myocarditis and discuss the adaptation of this multi-catalytic protease complex together with its implications on the course of disease. Inhibition of proteasome function is already highly relevant as a strategy in treating various malignancies. However, cardiotoxicity and immune-related adverse effects have proven significant hurdles, representative of the target's wide-ranging functions. Thus, we further discuss the molecular details of proteasome-mediated activity of the immune response for virus-mediated inflammatory heart disease. We summarize how the spatiotemporal flexibility of the proteasome might be tackled for therapeutic purposes aiming to mitigate virus-mediated adverse activation of the immune response in the heart.

Keywords: virus, myocarditis, proteasome, cytokine, immunopathology, heart failure

## INTRODUCTION

Myocarditis and its debilitating sequela, inflammatory cardiomyopathy, are leading causes of heart failure and sudden cardiac death particularly in infants, children, and young adults (1) with viral infections being the most common trigger of non-ischemic myocardial inflammation in the Western world (2). Acute injury of the heart muscle upon viral infection stimulates infiltration of immune cells, aiming to support pathogen clearance and alleviate organ damage. However, this pathogen-induced immune response can result subsequently in overwhelming immunopathology or the development of auto-aggressive immunity against cardiac self-antigens. These processes comprise fibrotic scarring and cardiac remodeling (3, 4). Both the high mortality of acute viral myocarditis in childhood and the putative progression of acute myocarditis to chronic disease support the need to define precisely the underlying mechanisms.

Most of our knowledge on the pathology of viral myocarditis comes from infection with enteroviruses, in particular coxsackievirus B3 (CVB3) in mice. CVB belong to the picornavirus family and have a non-enveloped, icosahedral capsid surrounding a positive-strand RNA genome. CVB3 used to be among the most prevalent pathogens known to cause viral myocarditis in North America and Europe (3, 5). Infection of laboratory mouse strains mirrors the variable manifestation of the disease in man (6, 7) by causing susceptibility for cardiac pathogenesis to a highly varying degree. A certain genetic background determines both control of viral pathogens and the activation of deleterious immune response pathways (6, 8). Recent observational studies suggest that it is not primarily the presence and/or replicative activity of invading viruses in the myocardium that determines outcome. In fact, the virustriggered abundance of infiltrating leukocytes is an independent risk factor for adverse outcome (9). Although it is indisputable that primary encounter of virus in the heart triggers death of cardiomyocytes, the pathogenic role of persisting viral genomes was poorly defined in the past. Recently, experimental mouse data demonstrated that persisting enteroviral RNAs do not actively contribute to ongoing myocardial disease after viral myocarditis (10).

Mice with high susceptibility to severe virus-induced inflammation are pre-disposed also to a loss of self-tolerance against cardiac proteins (11). Additionally, viral infection of cardiomyocytes can trigger auto-destructive activity of infiltrating cells, as well as, the formation of autoantibodies directed against antigens of cardiac origin (12, 13) further exaggerating heart tissue damage. Establishment of autoimmune myocarditis in mice by priming with cardiac antigens revealed that the same strains of inbred mice, who develop post-viral inflammatory heart tissue injury are also prone to autoimmunetriggered heart pathology. This indicates that the background genetics and involved immune response pathways for both diseases might be overlapping (13, 14). Others have reviewed in detail how type B coxsackievirus interacts with the innate and adaptive immune system and inflammatory responses (7, 15). Our primary interest herein is to discuss how cellular proteolysis by the proteasome affects the innate and adaptive immune response during CVB3-induced inflammatory damage of heart tissue, and our focus will broaden to the adaptation of this multi-catalytic protease in different cells during infection and inflammation. We will specifically discuss recent findings regarding the functional importance of a specific proteasome subtype expressed in hematopoietic cells and its possible implications for cytokine-mediated pathogenesis and therapeutic interference during viral myocarditis.

## THE PROTEASOME: A DRUGGABLE MULTI-CATALYTIC PROTEASE

Several avenues of research have implicated the ubiquitinproteasome system (UPS) as a major regulator of cell signaling and transcription. It controls also antigen processing, apoptosis and cellular proliferation. The ubiquitination machinery tags degradation-prone proteins in a highly regulated system for processing by the proteasome. As an integral part of cellular proteostasis, proteasome-mediated protein degradation is the primary route for intracellular removal of misfolded, damaged, or short-lived proteins (16). Proteasomes are multi-subunit enzymes with a barrel-shaped structure and internal active sites are accessible through a gated pore (17, 18). Proteasomedestined cargoes are recognized by regulatory particles (19S regulator) associated with the proteasome core complex (20S proteasome). The recognition, de-ubiquitination, and unfolding of substrates in direct proximity to the gated entry channel made up of the outer α ring of the 20S proteasome is required for degradation (19). Peptide hydrolysis is restricted to three β subunits, β1, β2, and β5, within the interior of the 2-fold symmetric core 20S proteasome. In addition to the aforementioned functions of the proteasome, the UPS is also of particular importance under conditions of cellular stress, where a rapid elimination of unfolded and potentially toxic proteins is required to prevent formation of cytotoxic aggregates (16, 20). Restrained function of the UPS might lead to accumulation of harmful proteins to toxic levels, causing disease (21). Cells have several ways to meet such increased demand for protein turnover. In response to Interferon (IFN)-γ (22, 23), tumor necrosis factor (TNF)-α (24), doxorubicin (25), or H2O<sup>2</sup> exposure (26), others and we demonstrated an increased abundance of the immunoproteasome (iproteasome), a specific proteasome isoform that contains alternative catalytic subunits (β1i/low molecular weight protein (LMP) 2; β2i/multicatalytic endopeptidase complex (Mecl)-1, β5i/LMP7) (27). I-proteasomes at least partially replace their constitutively expressed standard proteasome counterpart in different tissues upon infection (28, 29). During viral myocarditis, the i-proteasome is upregulated strongly in heart tissue and its induction involves IFN-γ (30, 31), as well as, type 1 interferon (T1IFN)-mediated signaling (8). In the heart, i-proteasome formation results in increased peptide hydrolysis capacity (8). This adaptation within the proteolytic core of the 20S proteasome complex is advantageous since it contributes to maintenance of protein homeostasis during inflammation (23, 32, 33). I-proteasome assembly is very similar to the formation of the standard proteasome [reviewed recently by (34)]. Additional proteasome subtypes like the thymoproteasome with tissue-specific β5t subunit expression (35) and mixed proteasomes that contain only one (β5i) or two (β1i and β5i) of the three inducible catalytic subunits of the i-proteasome (36) contribute to the variety of proteasome-mediated proteolysis.

Both facts—the close vicinity of genes encoding β1i/LMP2, β5i/LMP7, and the transporter associated with antigen presentation (TAP) within the major histocompatibility (MHC) II region, as well as, the regulatory function of IFNγ for these molecules—were indicative for a specialized function of the i-proteasome during MHC class I antigen presentation (37). The finding that β1i/LMP2 and β5i/LMP7 enhance substrate cleavage after basic and hydrophobic amino acid residues further strengthened the notion for a specific role of the i-proteasome in the generation of antigenic peptides (22, 38). In fact, there are numerous examples for viral, bacterial, and parasitic pathogens for which in vitro peptide processing studies revealed facilitated MHC class I epitope liberation by the i-proteasome complex in comparison to lower epitope abundance upon processing of model polypeptides with the standard proteasome (39). This altered prevalence of antigenic peptide generation by the i-proteasome is attributed to different peptide cleavage site usage (40), and can elicit to altered CD8<sup>+</sup> T cell-mediated immune surveillance also (41–46). Nevertheless, these findings appear to be restricted to a defined pool of immunodominant epitopes with no effect of the i-proteasome on other epitopes (28, 47, 48).

During the last three decades, the experimental landscape investigating i-proteasome biology substantially broadened with the availability of knockout mice lacking either single immunosubunits (47, 49) or a combinatory deletion of the three genes encoding β5i/LMP7, β1i/LMP2, and β2i/MECL-1 (45). Because deletion of a single i-proteasome subunit might be outweighed by increased formation of standard proteasome complexes (50), research on the i-proteasome improved further with the availability of i-proteasome subunitselective inhibitors. Kisselev and Groettrup provided a detailed overview on inhibitors of the respective subunits of the immunoproteasome (51). Structure-guided optimization of such inhibitory compounds with subunit selectivity is actually an ongoing objective. Initially, development of i-proteasomeselective inhibitors was pursued with regard to the profound benefit in patients with multiple myeloma (MM) upon the implementation of non-selective proteasome inhibitors like bortezomib or carfilzomib (52–55). Despite their high efficacy for MM cells, targeting the proteasome in other organs like the heart constitutes a risk for heart failure (56). In comparison to heart tissue (57), MM cells are unique regarding the preferential expression of the i-proteasome in these cancer cells. Therefore, compounds with selective i-proteasome subunit specificity represent an alternative strategy for more selective tumor-directed targeting (54, 58). ONX 0914 initially known as PR957 is a potent i-proteasome-selective inhibitor that predominantly targets the β5i/LMP7 and to a lower degree the β1i/LMP2 i-proteasome subunit as well (29, 59). Beyond the tumor-suppressive potential of ONX 0914 (60, 61), pre-clinical research utilizing this compound and other i-proteasome-selective inhibitors revealed additional putative clinical scenarios, where such drugs might improve current medical treatment. Pioneering work by the Groettrup group and others highlighted the therapeutic potential of i-proteasome inhibitors for mitigation of autoimmune-driven inflammatory tissue damage (50, 59, 62–64). KZR-616—an ortholog of ONX 0914 with high selectivity for the human i-proteasome—passed successfully phase I trials and is now in phase II trials for patients with systemic lupus erythematosus. Since i-proteasome activity controls alloantibody production by B cells and influences processes resulting in T cell exhaustion, i-proteasome-selective compounds could be used to prevent allograft rejection upon organ transplantation as well (65, 66). All these recent reports shed light onto several previously unappreciated biological functions of the i-proteasome and support the requirement for a detailed overview on the pathological function of the proteasome during virus-induced inflammatory heart tissue injury.

#### VIRAL ENTRY, REPLICATION, AND RELEASE: CONTROL MECHANISMS BY THE PROTEASOME

Viruses subvert cellular processes to favor viral propagation. Given its central role in a wide range of cellular functions by maintaining a critical level of essential regulatory proteins, it is expected that the proteasome is involved in viral replication, and numerous examples have indeed been reported. Several viral proteins direct host-cell proteins to proteolytic degradation by the proteasome (67). Viruses have evolved e.g., by encoding specific ubiquitin ligase activity to employ the proteasome for degradation of host proteins that would impede viral growth. Since this review mainly focuses on the immunomodulatory function of the proteasome complex itself during manifestation of virus-mediated inflammatory damage of heart tissue, the reader is encouraged to refer to an excellent review recently provided by Honglin Luo on interactions between ubiquitin/ubiquitin family proteins and viral growth (68). Here, we will summarize examples of viruses with known cardiac tropism, where the proteasome complex is exploited for virus progeny formation and/or where inhibitors of proteasome activity affect viral replication (**Table 1**).

Approximately 20 viruses have been implicated in human myocarditis and some of them interfere directly with the UPS. Among them, parvovirus B19 is detected often in endomyocardial biopsies obtained from patients with clinically suspected myocarditis (9). Parvoviruses follow multiple strategies for nuclear transport, some of them requiring active proteasomes. Replication of minute virus of mice—a murine parvovirus is disrupted in the presence of proteasome inhibitors (81). In addition to parvoviruses, members of the herpesviridae family like human herpesvirus 6 (HHV6) are commonly detected pathogens in cardiac biopsies (9). HHV6 causes accumulation of p53 in the cytoplasm (86), and among many mechanisms regulating p53 activity, the cellular abundance of p53 is controlled by UPS-dependent turnover (87). In herpes simplex virus (HSV) infection, proteasome activity directly affects virus progeny formation. Since inhibitors of the proteasome block HSV entry at a step occurring after capsid penetration into the cytosol but prior to capsid arrival at the nuclear periphery, it was concluded that cellular proteasome activity facilitates virus entry at this early stage (74). The human cytomegalovirus (HCMV) pp71 protein stimulates quiescent cells to enter the cell cycle by targeting

#### TABLE 1 | Effect of the proteasome on the propagation of viral particles.


The table summarizes viruses with known cardiac tropism and the impact of different proteasome inhibitors (bortezomib, MG132, lactacystin, MLN353, MG115, as well as, the immunoproteasome-selective inhibitor ONX 0914 (59)), of the proteasome activator PA28 (85), as well as, of the i-proteasome (cell culture and mouse studies using LMP7−/<sup>−</sup> mice or cell lines obtained from these mice (47) on viral replication. CHO, chinese hamster ovary; MDCK, madin-darby canine kidney; Gag, group-specific antigen; n.r., not reported; MLN353, Millennium353 (proteasome inhibitor); ONX 0914, immunoproteasome-specific inhibitor; PA28α/β, proteasome activator α/β of 28 kDa. <sup>1</sup>Murine pathogens.

proteins of the retinoblastoma (Rb) family for proteasomedependent degradation (88) and proteasome inhibitors block viral DNA replication, as well as, assembly of HCMV (76). The annual influenza virus (IV) season also calls upon some cases of IV-induced myocarditis in man. Proteasome inhibitors attenuate virus progeny formation at a post-fusion step upon influenza A virus (IAV) infection, and UPS activity is required for RNA synthesis of the virus (80). A similar function of the proteasome machinery at a post-entry step during viral replication applies to DNA replication and expression of intermediate and late genes of the vaccinia virus (83). Work is still in progress to unravel the role of the proteasome in the replication of human immunodeficiency virus (HIV). Thus far, it was shown that proteasome inhibition interferes with gag polyprotein processing, release and maturation of HIV-1 and HIV-2 (78, 79).

Although the frequency of adenovirus and coxsackie B virus detection in human myocarditis has gradually declined in adults in Western Europe during the last two decades, they are still a common cause of myocarditis in children or reported in small regional outbreaks. The adenovirus (Ad) E4 protein requires active proteasomes to promote late gene expression (69). Moreover, the Ad E1A protein regulates proteasomal activity, but is also a substrate for proteasome-mediated degradation (89). Recently, the Weinberg group established a mouse model of pediatric Ad-mediated myocarditis following intranasal infection of neonatal C57BL/6 mice with mouse adenovirus 1 (MAV-1) (90). MAV-1-myocarditis induces IFN-γ-mediated i-proteasome formation in infected heart tissue, but the catalytic activity of the β5i/LMP7 i-proteasome subunit had no effect on viral genome copy numbers in heart tissue (31). Therefore, it is unlikely that MAV-1 replication is affected by i-proteasome activity. In addition to in vivo models for the investigation of viral heart disease, in vitro studies have substantial advantages to provide information on the function of the proteasome regarding virus progeny formation. Most detailed information on the proteasome during the replicative phase of a human cardiotropic virus is available for CVB3. The McManus/Luo group was first to report a substantial suppression of CVB3 replication in HL-1 cells upon treatment with pan-specific proteasome inhibitors. This inhibitory effect was independent of the blockade of viral entry into host cells and rather attributed to reduced genome replication (70). The Luo group followed proteasome inhibition also during CVB3-induced myocarditis using A/J mice, which are known to be highly susceptible for CVB3 induced pathogenesis. In their study, MLN353 was introduced as a novel proteasome inhibitor for in vivo application. In contrast to the robust suppression of viral replication upon MG132 treatment in the HL-1 myxoma cell line (70), MLN353 treatment of mice did not influence virus titers (71). These somewhat controversial findings indicate that other essential pathways for CVB3 control might possibly be adversely influenced by MLN353, and this could outweigh the suppressive effect of proteasome inhibitors in cells targeted by virus infection. Our group investigated the contribution of specific proteasome subunits on the replication cycle of CVB3 in cellulo under one-step conditions using both HeLa cells and murine primary embryonic cardiomyocytes. PR825, as well as, ONX 0914 were applied at non-toxic concentrations to specifically block the catalytic activity of either β5 or β5i/LMP7, respectively. The CVB3 replication cycle involving the adsorption, penetration, replication of the parent virus, and release of progeny virus was not altered by the selective inhibition of these proteasome subunits (72). In addition to diverging peptidase activities of the six catalytic subunits, proteasome activity can be regulated upon binding to regulatory particles like the proteasome activator of 28 kDa (PA28). PA28-capped proteasome complexes are equipped with increased peptide hydrolysis capacity (91), and by as yet unknown mechanisms PA28 suppresses the CVB3 replication machinery (73). Altogether, a broad spectrum of various viral pathogens exploits the proteasome machinery in cells of the host organism.

#### INNATE IMMUNITY: HOW THE PROTEASOME AFFECTS THE FIRST DEFENSE WAVE

### Type I Interferons During Viral Myocarditis: Control by Proteasome Activity

During viral infection, viral RNAs and replication intermediates bind to their respective intracellular pattern recognition receptors, including Toll-like receptors (TLRs) and retinoic acid-inducible gene I (RIG-I), and, mediated by several distinct signaling pathways, this increases the production of T1IFNs [refer to (92) for a detailed review on T1IFNs in infectious disease]. T1IFNs are an effective first line of defense against viral infections and as such, a robust T1IFN response is highly beneficial to counteract early CVB3 infection in mice (93–95). Results from a pilot trial indicated a putative beneficial therapeutic influence of T1IFN substitution in patients with coxsackieviral myocarditis (96, 97). Following activation of the IFNα/β receptor (IFNAR), a diverse repertoire of antiviral proteins is expressed including protein kinase R (PKR), 2,5 oligoadenylate synthetase-like protein 2 (OASL-2), IFN-induced proteins with tetratricopeptide repeats (IFITs), as well as, IFN-stimulated genes like ISG15. The latter is an ubiquitin family protein, which is strongly induced by T1IFNs and NF-κB signaling in cardiomyocytes (98, 99), suppresses coxsackieviral replication, mitigates profoundly viral myocarditis and blocks the progression to its debilitating sequela (99).

Plasmacytoid dendritic cells (pDCs) are a major source for T1IFNs during viral myocarditis (100) and unique regarding their TLR7 or TLR9-dependent activation of IFN regulatory factor 7 (IRF7)-mediated IFNα/β production (101). Whereas, molecular accounts on the influence of ubiquitin modifications on pattern recognition receptor (PRR)-mediated signaling are available (102), less is known about the role of the different peptidase activities of the proteasome during the process from engagement of PRR to T1IFN production. Pan-specific inhibitors of the proteasome like bortezomib or carfilzomib, which target both the standard proteasome and i-proteasome, are potent suppressors of TLR9 activation in murine bone marrow cells, as well as, human peripheral blood mononuclear cells (PBMCs), but other TLR-mediated pathways like Toll/interleukin-1 receptor-domain-containing adapter-inducing IFNβ (TRIF) mediated IRF3 activation are affected as well (63). Selective i-proteasome inhibitors assigned specifically the control of IFNα/β production in pDCs to i-proteasome peptidase activity (59, 63). Correspondingly, i-proteasome inhibition in CVB3-infected C57BL/6 (B6) mice substantially reduces T1IFN production. Thereby, i-proteasome inhibition aggravates disease parameters like viral load in B6 mice (72). On the other hand, ISGs in germline LMP7−/<sup>−</sup> mouse models are as active as in wild-type controls during viral myocarditis (23). Indisputably, numerous studies indicate that the effects of T1IFN on the host response to infection are not limited to the acute, cell-intrinsic antiviral response described above. IFNα/β are also involved at various stages in the activation of adaptive immune cell responses e.g., by evolving antigen presenting DCs into a mature state (92). Similar to this, in hosts exhibiting high susceptibility for development of severe acute and chronic heart pathology like A.BY/SnJ mice, a shifted and overall significantly impaired T1IFN response (9, 100) leads to reduced DC activation and lower cross-presentation (100, 103). Genetic defects of i-proteasome subunits in mice that lead to impaired i-proteasome formation or proteasome inhibitor treatment decrease DC activation, thus, influencing the immune-stimulatory capacity of DCs as reflected by altered co-stimulatory molecule and C-C chemokine receptor 7 (CCR7) expression, as well as, cytokine production, respectively (104, 105). Thereby, i-proteasome-mediated proteolysis might directly control the antigen presentation capacity of DCs.

In contrast to the classical antiviral function of T1IFNs, there is increasing appreciation that IFNα/β can also be harmful, e.g., by triggering excessive inflammation and tissue damage (106). Likewise, IFNα/β is a classical disease-trigger of autoimmunity and auto-inflammation, and a reduced IFNα/β production as achieved upon administration of i-proteasomeselective inhibitors attenuates disease manifestation in models of lupus erythematosus (63). Defects in the DNA three prime repair exonuclease 1 (Trex1), which result in high cyclic guanosine monophosphate–adenosine monophosphate synthase (cGAS) induced IFNα/β production, lead to spontaneous inflammatory myocarditis in mice and Aicardi-Goutières syndrome in man (107, 108). Similarly, mutations in different genes encoding protein subunits of the human proteasome restrain T1IFN production, and this commences to a syndrome involving chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE) (109–111).

### Effect of the Proteasome for Humoral Innate Immunity

In addition to the cellular branch of innate immunity that comprises cell-associated pattern recognition receptors, its humoral branch includes molecules such as the classic short pentraxin C-reactive protein (CRP), the long pentraxin PTX3, and complement recognition molecules (112). During viral myocarditis, PTX3 is produced mainly by monocytes and macrophages (113, 114). PTX3 promotes the engulfment of cellular debris by immune cells (115), and acts as a safeguard mechanism dampening myocardial injury induced upon pattern-associated molecular pattern (PAMP)/damageassociated molecular pattern (DAMP) signaling (112). Although, the detailed molecular aspects are unresolved, the peptidase activity of the i-proteasome controls PTX3 expression in TLR4 activated macrophages during viral myocarditis (114) and pneumococcal pneumonia (116), a function of the i-proteasome which cannot be compensated by enhanced formation of standard proteasome in LMP7−/<sup>−</sup> mice (114).

### The Proteasome Balances Protein Homeostasis

Myocarditis in CVB3 (Nancy)-infected LMP7−/<sup>−</sup> mice on a B6 background lacking intact i-proteasomes is not only mirrored by reduced PTX3 production (114), but it also comprises highgrade inflammation and increased cell death (23). In cells with high rate of protein synthesis e.g., in response to cytokine signaling, a reduction of translational fidelity often occurs, generating defective ribosomal products (16). Cells in general and cardiomyocytes in particular that produce higher amounts of i-proteasomes are equipped with increased proteolytic activity and can efficiently degrade defective proteins (32, 33, 117). Thereby, the i-proteasome diminishes tissue damage in mouse hearts of CVB3-infected wild-type B6 mice (23). Nevertheless, this finding in B6 mice is in clear contrast to findings made in A/J mice, which exhibit high susceptibility for virus-mediated inflammation of heart tissue (118, 119) and generally present with increased viral burden in the heart. Here, i-proteasome activity constitutes severe cytokine-mediated inflammatory heart tissue injury. I-proteasome inhibition blocks chemokine and cytokine production, and consequently reduces the appearance of misfolded proteins (72). The use of selective inhibitors targeting i-proteasome activity does not necessarily reflect the findings obtained in respective germ-line gene deficient mouse models (23, 50, 72). As an example, contrary to what was reported in LMP7−/<sup>−</sup> B6 mice, inhibition of i-proteasome activity by ONX 0914 in CVB3-infected wild-type B6 mice disrupts the T1IFN defense against the invading pathogen, facilitates virusmediated tissue damage and exacerbates PAMP/DAMP-signaling in the heart. Thereby, the production of chemokines, infiltration with immune cells, as well as, cytokine release increase (72). Such discrepancies between specific inhibitors for proteasome subunits and their knockout models might be due to a compensatory formation of standard proteasomes in LMP7−/<sup>−</sup> mice (50, 59), which is not observed at a similar level in ONX 0914-treated mice.

#### Innate Myeloid Cells: Proteasome Activity Regulates Chemokine and Cytokine Production

Neutrophils are the first and most abundant cell population of the host's innate immune response with well-known function in the defense against bacterial and fungal pathogens. Moreover, neutrophil recruitment in virus infection can be part of a protective strategy leading to prevention of viral disease (120). The i-proteasome influences the abundance of these cells in blood and spleen, but it controls the activation status of neutrophils as well (72, 121). Nevertheless, neutrophils have no disease modifying impact on CVB3-induced myocarditis (72, 122, 123). During myocarditis, particularly monocytes/macrophages—that emigrate the bone marrow, then sequester and differentiate in the spleen—infiltrate the infected mouse heart (23, 99). Chemokines attract these cells to the injured heart, where they are indispensable for waste removal and healing (7, 124). On the other hand, many studies have highlighted the requirement of monocytes/macrophages for the manifestation of the detrimental consequences of viral myocarditis—inflammatory injury and formation of fibrotic scar (72, 125–128). Similar to monocytes, macrophages also exacerbate inflammatory injury in infected mouse hearts (127). Monocytes and macrophages secrete pro-inflammatory and profibrotic cytokines (7, 126). Therefore, molecules involved in innate immune cell mobilization and differentiation or in the control of cytokine/chemokine production by these cells present putative drug targets for future investigation. Resembling their effects on neutrophils, i-proteasome inhibitors stimulate also monocyte/macrophage emigration from the bone marrow and increase the abundance particularly of Ly6Chigh monocytes in the spleen (72, 121), where they differentiate to macrophages under inflammatory conditions (129). These findings might be indicative for a pro-inflammatory function of the i-proteasome.

In contrast, there is considerable experimental evidence from various in vitro and in vivo approaches that argues substantially against this notion and rather advocates i-proteasome-selective inhibitors as anti-inflammatory drugs e.g., for autoimmunity or to prevent transplant rejection (50, 59, 62, 65, 66). As summarized in **Figure 1**, selective inhibitors of the i-proteasome suppress the production of pro-inflammatory cytokines such as TNF-α and IL-6 in TLR4 and TLR7 activated immune cells. Similar results were obtained in IFNγ and TLR4 activated mouse macrophages (137), TLR4 stimulated splenocytes (59, 138), TLR4 activated PBMCs from healthy donors and patients with rheumatoid arthritis (59), as well as, TLR7 engaged macrophages (72). Consistently, in CVB3 infected A/J mice, the i-proteasome affects cytokine production also (72). Nevertheless, it needs to be recalled that under conditions where i-proteasome activity is needed for pathogen control like during Candida albicans or CVB3 infection of B6 mice, this influence of iproteasome proteolysis on cytokine production seems to be outweighed by a higher PAMP burden (72, 121). In this case, the pathogen load is presumably a much stronger effector of cytokine production than the cellular content of the i-proteasome.

### Influence of Proteasome Peptidase Activity on TLR Signaling

In A/J mice, CVB3 replicates to about 10-fold increased titers in the heart in comparison to B6 mice (72). One might speculate that the overall increase in viral RNA ultimately stimulates PRR signaling in mouse hearts, thereby facilitating cytokine/chemokine production. In fact, the inflammatory response in infected heart tissue is higher in A/J mice if directly compared to B6 mice. It remains an enigma how i-proteasome catalyzed proteolysis controls PRR signaling at a molecular level. In addition, it is unclear why the i-proteasome affects differently the cardiac phenotype during MAV-1 and CVB3 induced myocarditis in B6 mice (31, 72). CVB3 as a singlestranded RNA virus is a bona fide activator of TLR7 and TLR8 (139) [in mice only TLR7 is active (140)]. Viral DNA from Ad however triggers the TLR9 pathway. Alternatively, Ad escaping the endosome reveals viral DNA complexes to the cytosolic compartment and sensors like cGAS, which acts by the stimulator of interferon genes (STING)-controlled immune pathway (141). Thereby induced signaling stimulates transcription factors like IRF7 (TLR7, TLR9) activator protein 1 (AP-1) (TLR7, TLR9), NF-κB (TLR7, TLR9, STING), and IRF3 (STING) leading to the induction of target genes that in addition to IFNs and other ISGs - also encode proinflammatory cytokines and chemokines (101, 142). Therefore, we have summarized the current understanding on how the i-proteasome influences e.g., TLR mediated cellular signaling in **Figure 1**.

The NF-κB family of transcription factors, which acts downstream of TLR7, TLR9, and STING, plays a central role in regulation of inflammation. In the canonical pathway of NF-κB activation, the proteasome degrades IκBα, releasing the active NF-κB dimer (usually p65/p50) and allowing translocation to the nucleus (**Figure 2**). The impact of the different proteasome isoforms on NF-κB signaling is reported controversially (summarized in **Table 2**). A defective NF-κB activation as a response to reduced LMP2 expression in non-obese mice was attributed to reduced processing of the NF-κB precursor p105 (143, 145), but two different laboratories rebutted these findings (146, 150). Other data confirmed the initial findings and suggested an altered stimulation of canonical NF-κB activation by the i-proteasome in comparison to the standard proteasome. 20S i-proteasomes accelerate IκBα degradation (144), p65 nuclear translocation is lower in IFN-γ activated murine embryonic fibroblasts from LMP7−/<sup>−</sup> mice (149), and LPS-activated B cells from LMP2−/<sup>−</sup> degrade IκBα less efficiently than controls do (147). However, different groups revisited these aspects and novel data reported on contradictory findings arguing that the i-proteasome plays no obligatory role in the degradation of IκBα and activation of the canonical NF-κB pathway (59, 114, 137, 148). Different model systems and heterogeneous read outs for the activation of canonical NF-κB activation might attribute to these controversial findings. As illustrated in **Table 2**, more recent reports utilized advanced models such as primary cells obtained from different i-proteasome deficient mouse strains (LMP7−/−, LMP7−/−/Mecl-1−/−, LMP2−/−), and, more importantly, applied selective proteasome inhibitors in diverse immune and non-immune cells. Moreover, the majority of these reports focused on transcriptional activity of the canonical NF-κB pathway, whereas earlier reports indicated effects on signaling primarily at the level of p105 processing and Iκ-Bα degradation.

Similar to TLR4-stimulated cells, cytokine/chemokine production in TLR 7 activated cells also involves MyD88 signaling, which in addition to NF-κB activates mitogenactivated protein kinase kinases (MAPKK) resulting in phosphorylation of p38, c-Jun N-terminal kinases (JNKs), and extracellular signal-regulated kinases 1/2 (ERK1/2), culminating in activation of AP-1 (101, 151). Pan-specific proteasome inhibition influences this MAPKK pathway in lipopolysaccharide (LPS)-stimulated DCs (104). Since the pool of proteasomes in DCs is mostly comprised of the i-proteasome (136), such findings are indicative for a specific effect of the i-proteasome. And indeed, data from more recent work showed that the i-proteasome controls specifically the abundance and/or activity of certain kinases, phosphatases and/or regulatory proteins involved in the complex MAPK signaling network, resulting in increased MAPK phosphorylation upon engagement of TLR4 and TLR7 (72, 114). A comprehensive system biology-based approach might be most appropriate to dissect the involved effectors that rely on functional i-proteasome activity. If and how i-proteasome activity influences mRNA transcription of genes that are under the control of IRF3, IRF8, and IRF7 is still a matter of ongoing investigation. TLR4 activated DCs from LMP7−/−/Mecl-1−/<sup>−</sup> mice show unaltered phosphorylation of IRF3 (105). The pan-specific inhibitor of the proteasome bortezomib interferes with IRF-3 and IRF-8 activation in response to LPS in human DCs (104), suggesting a selective effect of proteasome inhibition on the IRF-3 pathway as well.

FIGURE 1 | Impact of i-proteasome subunits on innate immune signaling in myeloid cells. Among many different pattern recognition receptors, TLRs are sensors of microbial antigens on monocytes/macrophages and dendritic cells. These membrane-bound receptors are located both on the cellular surface (TLR4—colored in red) and in endosomes (TLR3—green, TLR7—blue, TLR9—purple) (101). Signaling pathways down-stream of TLR4, TLR7, and TLR9 involve the common adaptor molecule MyD88 (130, 131). Upon TLR stimulation, the ubiquitin E3 ligase TRAF6 engages with the TLR/MyD88 complex and generates poly-ubiquitin scaffolds (132), thereby recruiting the TAK1 complex (133). TAK1 then activates the IKK complex, which in turn phosphorylates IκBα. Ubiquitination of IκBα marks it for degradation by the proteasome. Thereafter, NF-κB translocates into the nucleus. Simultaneously, TAK1 induces MAP kinase signaling (134), which results in the phosphorylation of ERK1/2, p38, and JNK and thereby activates the transcription factor AP-1. Both NF-κB and AP-1 induce the expression of co-stimulatory molecules (CD80, CD86, CD40) and migration signals (CCR7) on DCs, the secretion of pro-inflammatory cytokines (e.g., TNF-α, IL-6, IL-23, IL-1β), chemokines (e.g., Mip2α, MCP-1, IP-10, RANTES), and of Pentraxin3 by monocytes/macrophages (cytokines partially also by DCs). MyD88-dependent TLR7/9 signaling induces the phosphorylation of IRF7, which is a key regulator of T1IFN (IFNα, IFNβ) expression in pDCs (135). Signals from TLR3 and TLR4 are transmitted by a MyD88-independent, TRIF-dependent pathway involving activating kinases (131). Phosphorylation of IRF-3 induces translocation into the nucleus. Results obtained from in vitro studies, in which the impact of the different peptidase activities of the proteasome isoforms regarding to TLR signaling or the expression of effector molecules were investigated by different (Continued)

FIGURE 1 | approaches, are summarized. Colors indicate the type of TLR stimulated to activate innate immune cells of different origin including human PBMCs, murine splenocytes, bone marrow cells and peritoneal macrophages. Each box illustrates both the model used to alter a specific peptidase activity of the proteasome—innate myeloid cells isolated from knock out mice or proteasome inhibitors with different specificity studied in innate myeloid cells, as well as, the observed effect either on the respective signaling pathway or on the production of respective effector molecules. (↓): reduced phosphorylation of a key molecule in the indicated signaling pathway or lower production of the effector molecule, = no alteration of signaling or production of the effector molecule. AP-1, activator protein 1; BTZ, bortezomib—a pan-specific proteasome inhibitor included because the i-proteasome is highly abundant in DCs (136), CCR7, C-C chemokine receptor type 7; DC, dendritic cell; ERK, extracellular signal–regulated kinases; IκBs, inhibitors of κB; IKK, IκB kinase; IP-10, interferon-gamma induced protein 10; IRF3, interferon regulatory factor 3; JNK, c-Jun N-terminal kinase; MAPK, mitogen-activated protein kinase; MCP-1, monocyte chemoattractant protein 1; Mip2α, macrophage inflammatory protein 2α; MKK, mitogen-activated protein kinase kinase; MyD88, myeloid differentiation primary response 88; NF-κB, nuclear factor-κB; ONX 0914, immunoproteasome inhibitor (59); RANTES, regulated on activation; normal T cell expressed and secreted; T1IFN, type I interferon; TAK1, transforming growth factor-β activated kinase 1; TLR, Toll-like receptor; TNF-α, tumor necrosis factor α; TRAF, TNF receptor associated factor; TRIF, TIR-domain-containing adapter inducing IFNβ. (1) (59) (2) (63) (3) (104) (4) (105) (5) (114) (6) (72) (7) (137) (8) (138) (9) (116).

FIGURE 2 | Regulation of NF-κB signaling by the proteasome. Multiple inflammatory signals result in the activation of the transcription factor NF-κB through a variety of adapter proteins and kinases. The most abundant form of the NF-κB dimer is the p50/p65 heterodimer. (A) The p105 precursor is processed by the proteasome, thereby liberating the NF-kB p50 subunit for dimerization with p65. IκB retains the NF-κB heterodimer in the cytoplasm. (B) Ligand binding to cellular receptors like TLRs activates the IKK complex, which catalyzes the phosphorylation of IκB, inducing its poly-ubiquitination and degradation by the proteasome. (C) Activated NF-κB translocates into the nucleus, where it (D) activates target gene expression. Table 2 summarizes all reported effects of i-proteasome activity on the different steps in this canonical NF-κB signaling pathway. NF-kB, nuclear factor kappa B; IκBs, inhibitors of κB; IKK, IκB kinase.

### Natural Killer Cells

Natural killer (NK) cells as lymphoid effectors of the rapidly acting antiviral immune response are among the first cells to sense pro-inflammatory cytokines. More than two decades ago, the importance of NK cells for CVB3 clearance and disease progression was highlighted in mice (152, 153). More recently, this pathobiological significance could be extended by providing firm evidence for a protective role of the NK cell receptor NKG2D, which upon activation triggers effective virus clearance in myocarditis (154). Knowledge regarding the impact of proteasome activity on NK cell function is incomplete and data are mainly available from tumor models. Immune surveillance of tumor cells involves a tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-mediated cytotoxic pathway used by NK cells leading to tumor cell lysis. Proteasome inhibitors like bortezomib can sensitize tumor cells to TRAIL-mediated lysis (155). If findings in tumor models might be transferable to viral myocarditis, is unknown. There is no evidence for a specific influence of the different proteasome isoforms on NK cell abundance within the inflamed heart of mice after CVB3 infection (23). Nevertheless, our current comprehension of the role of proteasome activity on NK cell function during viral myocarditis remains incomplete.

#### INFLUENCE OF THE PROTEASOME ON ESTABLISHMENT OF ADAPTIVE IMMUNITY

The migration of NK cells and myeloid cells to the site of injury in conjunction with a considerable increase in proinflammatory cytokines is followed by a second wave of infiltration with CD4<sup>+</sup> and to a lesser extent B and CD8<sup>+</sup> T lymphocytes as well. Similar to innate immunity, virtually all knowledge about the biological function of the adaptive immune response with regard to the manifestation of viral myocarditis is based on the mouse model of CVB3-induced myocarditis. Experiments with immune-deficient mice revealed that both humoral and cellular immune responses are required to control CVB3 infection. Accordingly, mice with severe combined immunodeficiency, which lack mature B and T cell function, develop extensive myocarditis with high mortality rates (156). In this review, we expand upon established knowledge about the function of the proteasome in adaptive immunity and attempt to illuminate the implication of the different isoforms in virus control. For further details on interactions of coxsackievirus and adaptive immune system, we refer the reader to an excellent review by (15).



Known effects of the i-proteasome are summarized for each step of the NF-κB signaling pathway. These steps involve: (A) processing of the NF-κB p105 precursor protein, (B) IκBα degradation by the proteasome, (C) NF-κB nuclear translocation and DNA binding, and (D) NF-κB promoter activity, respectively (Figure 2). Results that indicate a specific role of the i-proteasome for canonical NF-κB signaling are colored in light blue. Controversial findings arguing against the notion that the i-proteasome has a specific effect on canonical NF-κB signaling are colored in dark blue. NF-kB: nuclear factor kappa B, IκBs, inhibitors of κB; T2 cells, human lymphoblast cell line defective in LMP2 and LMP7; UK-101, LMP2-specific inhibitor; LSK-01, LMP7-specific inhibitor; ONX 0914, immunoprotesome-specific inhibitor. WB, Western blotting; IF, immuno-fluorescence; EMSA, electrophoretic mobility shift assay; IVP, p105 in vitro processing assay, perit. macrophages: peritoneal macrophages; BM macrophages, bone marrow-derived macrophages; MEFs, mouse embryonic fibroblasts; LPS, lipopolysaccharide.

### Influence of CD8<sup>+</sup> T Cells on Viral Myocarditis and Role of the Proteasome

In contrast to the preponderant significance of B cell and CD4<sup>+</sup> T cell responses for CVB3 clearance (157, 158), the pathophysiological significance of CD8<sup>+</sup> T cells for CVB3 clearance and inflammatory injury is less clear. The protective function of CD8<sup>+</sup> T cells (159) involves the production of cytokines like IFNγ, yet is clearly separated from the direct cytolytic effect mediated by perforin, a classic hallmark of virusspecific CD8<sup>+</sup> T cells (160). CD8<sup>+</sup> T cells acting by perforin cause extensive destruction of myocardial tissue (160, 161). Evidence arguing in favor of a protective function of CD8<sup>+</sup> T cells during myocarditis was obtained from CD8<sup>+</sup> T cell-deficient β2-microglobulin−/<sup>−</sup> mice, in which injury of cardiac tissue exacerbates due to insufficient confinement of the initial viral load in the heart muscle (160). One needs to keep in mind that constitutive knockout models for perforin and β2-microglobulin do not only mirror the function of these molecules in CD8<sup>+</sup> T cells. Both perforin and β2-microglobulin affect also NK cell activation and function. Nevertheless, the finding that CD8<sup>+</sup> T cells restrain CVB3 in mice indicates that the virus could induce detectable CD8<sup>+</sup> T cell responses. However, the Whitton group provided data that coxsackieviruses do not elicit strong CD8<sup>+</sup> T cell responses. Investigation of mice infected with a recombinant CVB3 encoding known lymphocytic choriomeningitis virus (LCMV) derived CD8<sup>+</sup> T cell epitopes failed to trigger a marked expansion of CD8<sup>+</sup> T effector cells (162, 163). This is mainly due to the inhibition of antigen presentation by virus-induced disruption of host protein trafficking in infected cells (164, 165). The virus almost completely blocks antigen presentation via the MHC class I pathway, thereby evading CD8<sup>+</sup> T cell immunity (163). Our group followed a complementary approach employing prediction tools for proteasomal cleavage sites, MHC binding studies and in vitro peptide processing assays with the proteasome to identify MHC class I epitopes originating from CVB3 proteins (8, 166). Concordant with the findings by the Whitton group, expansion of respective CD8<sup>+</sup> T effector cells was weak in mice (8). Similarly, adoptive transfer of CD8<sup>+</sup> T cells isolated from mice with CVB3 myocarditis did not affect the manifestation of viral myocarditis in recipient mice (23).

Based on these virus-specific aspects, the role for the iproteasome with regard to induction of CD8<sup>+</sup> T cell responses needs to be revisited for viral myocarditis. Following up on robust i-proteasome formation in hearts of both MAV-1 and CVB3 infected mice (30, 31), the Weinberg lab and our workgroup investigated the role of the i-proteasome concerning virus clearance in myocarditis. The i-proteasome facilitates the release of peptides harboring hydrophobic or basic C-terminal amino acids typical for MHC class I epitopes (22, 27). By facilitating such specific peptide cleavages, the i-proteasome augments the pool of antigenic peptides (40). Nevertheless, we found uniformly that the i-proteasome can be adequately compensated by its standard proteasome counterpart during viral myocarditis (23, 31). Although the i-proteasome provides an increased capacity to liberate CVB3 epitopes for MHC class I antigen presentation (40, 166), it cannot compensate for the disruption of MHC class I presentation by the virus. If detectable at all, CD8<sup>+</sup> effector T cell responses remain weak during CVB3 infection (163).

### CD4<sup>+</sup> T Cells and Antibody Responses in CVB3 Myocarditis: Impact of the Proteasome

Infections with CVB3 trigger a rapid and effective antibody response. Neutralizing antibodies appear 4 days after CVB3 infection (167) and are essential for controlling virus dissemination and clearance in the heart (158). CD4<sup>+</sup> T cells activate B cells for production of protective antibodies. In contrast to MHC class I, MHC class II epitopes are presented efficiently upon infection with CVB3 and CD4<sup>+</sup> T cells mature quickly into effector and later on into memory T cells (163). The proteasome is involved in multiple cellular processes needed for antibody production. As outlined above, it controls the maturation and activation of DCs (104), but the proteasome regulates also B cell function (147). The canonical pathway for MHC class II antigen presentation is located within the endolysosomal compartment and thereby spatially separated from the proteasome. However, there is also a non-canonical cytosolic pathway of MHC class II-restricted antigen processing involving proteasome-dependent peptide processing. In addition to DCs exposed to exogenous influenza and vaccinia virus (168), cancer cells present peptides on MHC class II by such non-classical antigen-processing pathways (169). It is unknown whether the cleavage site preference of the different proteasome isoforms determine a specific CD4<sup>+</sup> T cell repertoire as reported for CD8<sup>+</sup> T cells. To dissect the function of the i-proteasome in CVB3 myocarditis, our group applied the i-proteasome-specific inhibitor ONX 0914, and alternatively utilized LMP7−/<sup>−</sup> mice. We found a strong induction of CVB3 directed immunoglobulins and neutralizing antibodies in mice lacking intact i-proteasome function (23). In fact, neutralizing antibody titers were higher in mice with ONX 0914 treatment, an observation that might be attributed to maintained survival of CD4<sup>+</sup> T cells during infection in response to i-proteasome inhibition (72). The latter finding during CVB3 infection was specific for A/J mice and did not occur in B6 mice. In B6 mice, i-proteasome inhibition resulted in a reduction of lymphocyte abundance in blood and spleen at the acute phase of the disease. In fact, other groups demonstrated also a pro-survival function of the i-proteasome in T cells during viral infection with IV and LCMV (147, 170).

The fact that re-infection of B6 mice with CVB3 4 weeks after primary virus inoculation completely revokes disease manifestation emphasizes the importance of memory immune status, as well as, antibody formation during CVB3 infection (72). Upon encountering CVB3, memory T and B cells initiate cell division much more rapidly than their naive counterparts do. These data suggest that the level of MHC/peptide complex upon initial infection is sufficient to trigger memory T cells (163). In CVB3-infected B6 mice, displaying impaired i-proteasome function, adequate immune memory develops unhindered as well (23, 72). Similarly, protective immunity to MAV-1 is preserved in LMP7−/<sup>−</sup> mice (31). Conclusively, the specific peptidase activities of the i-proteasome are not essential for establishment of an adaptive immune response in mouse models of viral myocarditis.

CVB-specific CD4<sup>+</sup> T cells show an effector phenotype with a Th1 cytokine profile (163). In addition, A/J mice induce an autoreactive CD4<sup>+</sup> T cell repertoire that contains IL-17 producing cells (11). The availability of i-proteasome selective inhibitors shed new light onto the role of the i-proteasome during CD4<sup>+</sup> T cell differentiation. Under Th17 skewing conditions, inhibition of the LMP7 subunit downregulates RORγt activity leading to reduced Th17 counts, whereby lower STAT1 phosphorylation reduces IFN-γ production under Th1 skewing conditions indicative for lower Th1 counts (171). Whether or not these in vitro findings are relevant during viral myocarditis needs further investigation—a challenging task given the relatively weak IL-17 signal obtained from CD4<sup>+</sup> T cells during acute myocarditis (11).

#### FUTURE PERSPECTIVES

Several mechanisms have been proposed for CVB3-mediated myocarditis in mice, including direct virus-mediated cell damage and destruction of heart tissue in response to the action of immune effector cells (7). Being the major cellular mechanism for protein degradation, the proteasomal system adapts to augmented protein turnover by increased formation of iproteasomes (32, 33). Based on structural information (17, 18, 29), site-specific inhibitors targeting particular subunits of the major proteasome isoforms have become available [reviewed in (51)] and our understanding about the pathophysiological role of the proteasome during CVB3-mediated myocarditis has thereby improved. In our concluding remarks, we discuss whether subunit-selective inhibitors might be applicable to suppress manifestation or progression of virus-induced cardiac injury.

Inactivation of the highly abundant β5 standard proteasome subunit in murine cardiomyocytes augments apoptosis in myocardial ischemia/reperfusion injury (172) or due to doxorubicin treatment. In contrast, even under conditions with cytokine-induced i-proteasome expression, selective i-proteasome inhibitors are advantageous in reducing cardiomyocyte death in comparison to compounds targeting either the standard or both the standard and the i-proteasome with similar efficacy (25). During viral myocarditis, i-proteasome formation and to a minor extent induction of PA28β also enhance cellular protein turnover reducing the accumulation of oxidant-damaged proteins (23, 73). The notion of a minor influence of the i-proteasome regarding the control of pathogens was supported by elimination of virus despite a reduction of T1IFN (63, 72) upon i-proteasome inhibitor treatment and induction of immune memory in CVB3 heart disease (72). This is consistent with findings for other pathogens as well (48, 59) and in addition, i-proteasome inhibitors are well tolerated in other viral infection models (31, 173). In none of these models, i-proteasome inhibition alters significantly the abundance of toxic aggregates. Most strikingly, in mice susceptible for CVB3 myocarditis, i-proteasome inhibition is highly beneficial. ONX 0914 treatment improves cardiac function and mortality by efficient suppression of cardiac and systemic chemokine and cytokine production (72).

In addition to myocarditis, experimental infection of susceptible mice with CVB3 results in severe systemic disease as well, with the pancreas being the primary and most affected organ (174). Early upon infection, mice become hypoglycemic, most likely due to pancreatitis and digestive dysfunction (175). With the release of cytokines, such systemic pathology alters the vascular tone and impairs diastolic filling as well. Systemic disease in A/J mice is reminiscent of a distributive shock in sepsis (118). Importantly, given the high abundance of i-proteasome in immune cells, i-proteasome specific inhibitors affect systemic pathology as well and this has immediate impact on the cardiac output and immunemediated damage of heart tissue (72). Other than in the experimental mouse model, myocarditis in man usually follows a benign respiratory, gastrointestinal or urogenital infection, and pancreatitis is reported only occasionally (3). Therefore, our current understanding of i-proteasome biology during myocarditis needs further clarification. Additional research ought to elucidate the contribution of the i-proteasome once virus-mediated injury of the heart muscle has developed. In addition, we need detailed knowledge on molecular and cellular aspects of i-proteasome biology and the underlying mechanisms that contribute to the protective outcome if the i-proteasome is blocked prior to the occurrence of viral heart disease. As the i-proteasome has wide-ranging functions, toxicity and immune-related adverse effects may represent significant hurdles regarding the application of i-proteasome inhibitors. A detailed comprehension of i-proteasome function at an advanced stage of myocarditis is particularly important, because the resolution of acute CVB3 myocarditis is followed by the onset of chronic inflammation, which has been attributed to autoimmunity, as shown in genetically susceptible mice (176). Whether the i-proteasome affects also manifestation of autoimmune heart disease is unknown. Nonetheless, our current understanding of i-proteasome biology encourages a continued look at this context to define novel treatment options for viral heart disease.

### AUTHOR CONTRIBUTIONS

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

### FUNDING

Funding was funded by the Foundation for Experimental Biomedicine (Zurich, Switzerland), by the German Research Foundation DFG BE 6335/4-1 and DFG BE 6335/6-1, CRC 1292 project 02, as well as, by the Federal Ministry of Education and Research (BacVirISG15 031L0123B) to AB. We acknowledge Martin Voss and Hannah Louise Neumaier for critical discussion of the manuscript. Publication was supported by the German Research Foundation (DFG) and the Open Access Publication Fund of Charité—Universitätsmedizin Berlin.

## REFERENCES


(Immuno)proteasome activities. Angew Chem Int Ed Engl. (2015) 55:4199– 203. doi: 10.1002/anie.201509092


infarction and infectious myocarditis. Cardiovasc Pathol. (2011) 20:e27–35. doi: 10.1016/j.carpath.2010.02.005


biosynthesis and stabilization of p38 MAPK. Mol Cell Biol. (2007) 27:170–81. doi: 10.1128/MCB.01456-06


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

Copyright © 2018 Beling and Kespohl. 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.

# Theiler's Virus-Mediated Immunopathology in the CNS and Heart: Roles of Organ-Specific Cytokine and Lymphatic Responses

Seiichi Omura1,2, Eiichiro Kawai <sup>2</sup> , Fumitaka Sato1,2, Nicholas E. Martinez <sup>2</sup> , Alireza Minagar <sup>3</sup> , Mahmoud Al-Kofahi <sup>4</sup> , J. Winny Yun<sup>4</sup> , Urska Cvek <sup>5</sup> , Marjan Trutschl <sup>5</sup> , J. Steven Alexander 3,4 and Ikuo Tsunoda1,2,3 \*

*<sup>1</sup> Department of Microbiology, Kindai University Faculty of Medicine, Osaka, Japan, <sup>2</sup> Department of Microbiology and Immunology, Center for Molecular and Tumor Virology, Center for Cardiovascular Diseases and Sciences, Louisiana State University Health Sciences Center-Shreveport, Shreveport, LA, United States, <sup>3</sup> Department of Neurology, Louisiana State University Health Sciences Center-Shreveport, Shreveport, LA, United States, <sup>4</sup> Department of Molecular and Cellular Physiology, Louisiana State University Health Sciences Center-Shreveport, Shreveport, LA, United States, <sup>5</sup> Department of Computer Science, Louisiana State University Shreveport, Shreveport, LA, United States*

#### Edited by:

*Michael H. Lehmann, Ludwig Maximilian University of Munich, Germany*

#### Reviewed by:

*Jun-ichi Kira, Kyushu University, Japan Andreas Beineke, University of Veterinary Medicine Hannover, Germany*

> \*Correspondence: *Ikuo Tsunoda itsunoda@med.kindai.ac.jp*

#### Specialty section:

*This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology*

Received: *28 September 2018* Accepted: *21 November 2018* Published: *10 December 2018*

#### Citation:

*Omura S, Kawai E, Sato F, Martinez NE, Minagar A, Al-Kofahi M, Yun JW, Cvek U, Trutschl M, Alexander JS and Tsunoda I (2018) Theiler's Virus-Mediated Immunopathology in the CNS and Heart: Roles of Organ-Specific Cytokine and Lymphatic Responses. Front. Immunol. 9:2870. doi: 10.3389/fimmu.2018.02870* Theiler's murine encephalomyelitis virus (TMEV) induces different diseases in the central nervous system (CNS) and heart, depending on the mouse strains and time course, with cytokines playing key roles for viral clearance and immune-mediated pathology (immunopathology). In SJL/J mice, TMEV infection causes chronic TMEV-induced demyelinating disease (TMEV-IDD) in the spinal cord about 1 month post-inoculation (p.i.). Unlike other immunopathology models, both pro- and anti-inflammatory cytokines can play dual roles in TMEV-IDD. Pro-inflammatory cytokines play beneficial roles in viral clearance while they are also detrimental in immune-mediated demyelination. Anti-inflammatory cytokines suppress not only protective anti-viral immune responses but also detrimental autoreactive immune responses. Conversely, in C3H mice, TMEV infection induces a non-CNS disease, myocarditis, with three distinctive phases: phase I, viral pathology with interferon and chemokine responses; phase II, immunopathology mediated by acquired immune responses; and phase III, cardiac fibrosis. Although the exact mechanism(s) by which a single virus, TMEV, induces these different diseases in different organs is unclear, our bioinformatics approaches, especially principal component analysis (PCA) of transcriptome data, allow us to identify the key factors contributing to organ-specific immunopathology. The PCA demonstrated that *in vitro* infection of a cardiomyocyte cell line reproduced the transcriptome profile of phase I in TMEV-induced myocarditis; distinct interferon/chemokine-related responses were induced *in vitro* in TMEV-infected cardiomyocytes, but not in infected neuronal cells. In addition, the PCA of the *in vivo* CNS transcriptome data showed that decreased lymphatic marker expressions were weakly associated with inflammation in TMEV infection. Here, dysfunction of lymphatic vessels is shown to potentially contribute to immunopathology by delaying the clearance of cytokines and immune cells from the inflammatory site, although this can also confine the virus at these sites, preventing virus spread via lymphatic vessels. On the other hand, in the heart, dysfunction of lymphatics was associated with reduced lymphatic muscle contractility provoked by

**141**

pro-inflammatory cytokines. Therefore, TMEV infection may induce different patterns of cytokine expressions as well as lymphatic vessel dysfunction by rather different mechanisms between the CNS and heart, which might explain observed patterns of organ-specific immunopathology.

Keywords: adhesion molecules, animal models, blood-brain barrier, computational analysis, GLYCAM1, LYVE1, Picornaviridae infection, unsupervised analysis

#### INTRODUCTION

### Theiler's Murine Encephalomyelitis Virus (TMEV) Induces Distinct Organ-Specific Diseases

Theiler's murine encephalomyelitis virus (TMEV) is a nonenveloped, single-stranded positive-sense RNA virus that belongs to the order Picornavirales, family Picornaviridae, genus Cardiovirus. Historically, Max Theiler discovered the Theiler's original (TO) strain of TMEV as an agent that induces acute polioencephalomyelitis in the central nervous system (CNS) of mice in 1934 (1–3). Since TMEV infects the gastrointestinal tract and induces an acute CNS disease similar to poliovirus (family Picornaviridae, genus Enterovirus), TMEV was originally classified into the genus Enterovirus and used as an animal model for poliomyelitis. In 1952, Joan Daniels reported that the Daniels (DA) strain of TMEV causes myositis in the skeletal muscle and a chronic inflammatory demyelinating disease in the spinal cord (4), the latter of which has been called TMEV-induced demyelinating disease (TMEV-IDD) and used as a viral model for multiple sclerosis (MS) (5–7), first by Howard Lipton in 1972. In 1996, Gómez et al. demonstrated that TMEV causes inflammation not only in the skeletal muscle (i.e., myositis) but also in the heart muscle (i.e., myocarditis) (8). Since 2014, TMEV-induced myocarditis has been applied as a viral model for myocarditis (9) (**Figure 1**). The resistance/susceptibility to TMEV-induced organ-specific pathology has been known to differ among mouse strains. The resistance to persistent CNS infection maps genetically to major histocompatibility complex (MHC) class I, H-2D region (3). The H-2 background also appears to influence myositis and myocarditis, although studies using congenic mice are necessary to determine the precise role of MHC molecules (8).

In general, viruses infect limited species and induce diseases in an isolated group of organs. The determination of the mechanism(s) of such organ-specific tropism/pathogenesis of virus infections could powerfully inform the development of treatments and methods of prevention for viral infections: currently the precise mechanisms of many types of viral pathogenesis still remain unknown. TMEV is a natural enteric pathogen of mice (11) and has been isolated from trapped wild mice (12), while no TMEV-induced disease has been reported in the wild. TMEV has been shown to infect only mice, and not other species in vivo (with a few exceptions) and causes distinct maladies that mimic human diseases (3). In experimental mice, intracerebral inoculation of TMEV results in CNS viral infection as well as viremia and induces diseases in the CNS and the heart (13). On the other hand, peripheral inoculation, such as intraperitoneal or intravenous injection, causes myocarditis more efficiently (9), but rarely causes CNS infection. Thus, TMEV has high neurotropism and high neurovirulence, but low neuroinvasiveness, despite the fact that TMEV can use at least three routes to gain access to the CNS: neural spread, hematogenous spread, and olfactory route (14). Low neuroinvasiveness of peripherally inoculated TMEV can be explained by the fact that vascular endothelial cells are not permissive for TMEV infection in vivo (15). Here, although TMEV can still invade the CNS hematogenously, using infected macrophages as Trojan horse (3), this is not an efficient way to achieve fast and successful viral invasion into the CNS. TMEV infects only certain cell types in restricted organsin vivo, although TMEV can infect most cell lines derived from various organs and different host species, even insect cells (with the exception of T cells in vitro) (15).

#### TMEV-Induced CNS Disease

TMEV is divided into two subgroups: the TO and GDVII, based on its neurovirulence following intracerebral inoculation. The GDVII subgroup, including GDVII and FA strains, causes acute fatal polioencephalomyelitis and kills all mice following intracerebral infection. One plaque forming unit (PFU) of GDVII virus is enough to kill mice by induction of neuronal apoptosis and axonal injury without inducing acquired immune responses (16). The TO subgroup, including DA and BeAn strains, induces a biphasic disease in susceptible mouse strains (highly susceptible, SJL/J mice; and intermediate susceptible, C3H mice), following intracerebral injection (17). During the acute phase, about 1 week post-inoculation (p.i.), TMEV infects neurons and induces neuronal apoptosis, neuronophagia, and inflammation, mainly in the gray matter of the brain, including the hippocampus and cerebral cortex (polioencephalitis), while induction of TMEV-specific cellular and humoral immune responses is accompanied by the clearance of the virus from the brain. Thereafter, a low level of TMEV can persistently infect oligodendrocytes and microglia/macrophages in the white matter of the spinal cord of susceptible mice, and recruit anti-viral immune cells into the infected regions, particularly ventrolateral funiculus of the thoracic segments, leading to inflammatory demyelination during the chronic phase, about 1 month p.i. (3).

During the acute phase of TMEV infection, CD4<sup>+</sup> and CD8<sup>+</sup> T cells and anti-viral antibodies enter the CNS, contributing to viral clearance from the gray matter without causing overt immune-mediated tissue damage (immunopathology). During the chronic phase, however, these same immune effector components are detected in the white matter, and play key roles

in immunopathology (18). Overall, gain-of-function and lossof-function approaches to clarify the roles of immune effector cells, antibodies, and cytokines, have demonstrated that antiviral pro-inflammatory effector molecules/cells, including CD4<sup>+</sup> T helper (Th)1, cells and CD8<sup>+</sup> cytotoxic T lymphocytes (CTLs), and antibodies play protective roles during the acute phase (7). For example, in GDVII virus infection, lack of CNS infiltrating immune cells results in acute fatal polioencephalitis (19), which has been associated with altered mRNA expression levels of cytokines, but not chemokines (20), as well as induction of transforming growth factor (TGF)-β1 protein in the neurons (21). On the other hand, during the chronic phase, these immune effector molecules/cells could play detrimental roles causing immunopathology in a bystander fashion and/or determinant (epitope) spreading to myelin antigens (22, 23), although the precise mechanisms of immunopathology remain unknown. While these immune effector molecules/cells (Th1, CTL, and antibody) seem to play both protective anti-viral and detrimental immunopathogenic roles in TMEV-IDD, anti-inflammatory cells including regulatory T cells (Tregs) can also be beneficial and detrimental depending on the disease phases in TMEV infection (24).

#### TMEV-Induced Myocarditis

Myocarditis is an inflammatory disease in the heart caused by microbial infections or autoimmunity and affects about 2 million people in the United States (25, 26). Among viruses, the picornavirus family, especially coxsackievirus B, is a well-known pathogen of myocarditis (27). In general, regardless of viral families or species, viral myocarditis has been proposed to be divided into three phases (28). In phase I (around 4 days p.i., experimentally), the virus infects and replicates in the heart, damaging cardiomyocytes, while innate immune responses against the virus are induced. In phase II, anti-viral T-cell and antibody responses are induced (after 5 or more days p.i.), with the penetration of these effector components into the heart. Under pathologic conditions, anti-viral immune responses not only clear the virus but also damage infected and uninfected cardiomyocytes by anti-viral CTLs and in a bystander fashion, respectively. This can be followed by induction of autoimmune responses to the heart reflecting determinant spreading and/or molecular mimicry between the virus and heart antigens. When the tissue damage caused in phase I and/or II is severe, cardiac remodeling and fibrosis with or without low-grade viral persistence occur, which can lead to dilated cardiomyopathy (phase III). Ideally, each patient with myocarditis should be treated depending on the phase (29): phase I, antiviral; phase II, immunomodulation; and phase III, standard heart failure therapy (e.g., immunosuppression may be appropriate for phase II, but will enhance virus replication in phase I). However, finding effective therapies has remained challenging because the phase-specific biomarkers and pathogenesis of myocarditis have not been conclusively identified (30), while serum cardiac troponin and creatine kinase, electrocardiogram and echocardiography, and the endomyocardial biopsy have been helpful, to some extent, for diagnosing myocarditis (31, 32).

To clarify the pathogenesis and discover the phase-specific biomarkers, we established a murine model for viral myocarditis using TMEV (9, 28, 33), which has unique characteristics not seen in most other animal models. For example, (1) most animal models don't have three phases; (2) fail to reproduce clinical and immunological findings in human viral myocarditis; and (3) fail to use a "natural" pathogen of the host, thus the TMEV model is possibly more relevant to human natural infections. Generally, peripheral injection (e.g., intraperitoneal) of TMEV in mice can efficiently cause inflammation in the heart, but not in the CNS (8, 13), while intracerebral injection of TMEV also causes myocarditis due to acute viremia. Susceptibilities to TMEV-induced myocarditis differ among mouse strains: the highly susceptible C3H strain, the intermediate susceptible C57BL/6 strain, and the highly resistant SJL/J strain. C3H mice develop all three phases, while SJL/J mice develop only phase I and C57BL/6 mice develop phases I and II; the different genetic susceptibilities to viral myocarditis has also been demonstrated in humans (34). TMEV-induced myocarditis can be divided into three phases as in human myocarditis. In phase I, innate immune molecules [interferon (IFN)-induced genes [e.g., interferon regulatory factor 7 (Irf7), interferon-induced protein with tetratricopeptide repeats 1 (Ifit1), and Ifit3] and chemokine genes [e.g., chemokine (C-X-C motif) ligand 9 (Cxcl9), Cxcl10, and chemokine (C-C motif) ligand 5 (Ccl5)] that can recruit Th1 and natural killer T (NKT) cells] were upregulated prior to immune cell infiltration in the heart. In phase II, T-cell infiltrates were observed with upregulation of proinflammatory IFN-γ pathway genes, followed by upregulation of cardiac remodeling genes (e.g., Mmp12 and Gpnmb) in phase III. Among transgenic and knockout (KO) mice infected with TMEV, NKT KO mice developed more severe myocarditis with lower ejection fraction in echocardiography than wild-type mice (10).

#### Lymphatics and Viral Infections

The afferent lymphatic vessels transport interstitial fluid and antigens from tissues to lymph nodes and have specialized capillaries with an open structure; antigen transport to the draining lymph nodes is required to generate antigen-specific immune responses (35). Cancer cells and pathogens often "hijack" this transport system to achieve systemic spread (36), while dissemination to the blood circulation is first blocked at regional lymph nodes. In viral infections, while the mechanisms that limit systemic viral spread have not been studied extensively, several mechanisms have been proposed recently. Kastenmüller et al. (37) showed that vaccinia virus injected subcutaneously in mice was acquired by CD169<sup>+</sup> subcapsular sinus macrophages in the regional lymph nodes, but not in the spleen, 4 hours (h) p.i. Since local depletion of macrophages by clodronate-loaded liposomes resulted in viral spreading to the spleen, these results suggest that systemic viral spread ensues in the absence of effective viral capture by macrophages. On the other hand, Loo et al. (38) demonstrated that vaccinia virus infection by scarification, which did not spread the virus to draining lymph nodes, induced remodeling of the pre-existing cutaneous lymphatic vasculature, but not lymphangiogenesis. The remodeling was coincident with a rapid reduction in fluid transport, suggesting that lymphatic vessels negatively modulate fluid transport following viral infection in the skin, to limit the spread of viral particles into lymph nodes. Lymphatic vessel remodeling can result in not only compartmentalization of infectious virus, but also an accumulation of inflammatory mediators in the skin, which affect anti-viral immunity and immunopathology.

In the following sections, we will introduce our bioinformatics analyses of both supervised (such as heat map and k-means clustering) and unsupervised [particularly principal component analysis (PCA)] approaches to identify factors that contribute to organ-specific viral pathology. Previously, using these computational analyses, we were able to identify and rank key molecules involved in MS (39), stroke (40), and myocarditis (33). Here, we focus on two potential candidate factors contributing to organ-specific viral pathology: (1) innate immune responses by the major cell type of each organ, i.e., cardiomyocyte in the heart vs. neuron in the CNS; and (2) lymphatic vessel dysfunction induced by cytokines in the heart vs. downregulation of neurolymphatic molecules in the CNS.

### CELL-TYPE SPECIFIC INNATE IMMUNE RESPONSES IN TMEV INFECTION

### TMEV Infects and Damages Cardiomyocytes in vitro

The TMEV-induced myocarditis model in vivo is complemented by the in vitro model using a mouse cardiomyocyte cell line, HL-1, which was established by Dr. William C. Claycomb (Louisiana State University Health Sciences Center, New Orleans, LA) from an AT-1 subcutaneous tumor of a C57BL/6J mouse. HL-1 cells retain a differentiated cardiomyocyte phenotype and show contractile activity in vitro (41). To see the effects (innate immune responses and viral pathology) of direct virus infection without the involvement of immune cells (phase I mimic), we infected HL-1 cells, at a multiplicity of infection (MOI) = 1 or 10. TMEV infection induced cytopathic effects (CPE) on HL-1 cells, which became obvious 12 h p.i. (**Figure 2A**), while the cell viability started to decrease 8 h p.i., with most cells dying 36 h p.i. (**Figure 2B**). CPE was accompanied by the detection of cardiac troponin in the culture supernatants of HL-1 cells, which was measured by an enzyme-linked immunosorbent assay (ELISA) using the Ultra Sensitive Mouse Cardiac Troponin-I ELISA Kit (Life Diagnostics, West Chester, PA) (**Figure 2C**) (33). We also determined virus replication by plaque assays, using supernatants for cell-free virus and cell lysates for cell-associated virus (**Figure 2D**). Cell-free virus titers increased substantially 12 h p.i., which reflected a loss of plasma cell membrane integrity and showed similar kinetics with supernatant troponin concentrations. Cell-associated viral titers increased 8 h p.i. and peaked 12 h p.i., which was associated with the cell viability. In these assays, we also used a murine neuroblastoma cell line, Neuro-2a (43), since TMEV is known to infect neurons in vitro as well as during the acute phase following intracerebral infection in vivo. TMEV-infected Neuro-2a cells had similar kinetics of cell viability and viral replication to those of HL-1 cells, while cardiac troponin was not detectable in Neuro-2a cells regardless of infection, as expected (**Figures 2B–D**).

### Innate Immunity-Related Genes Are Upregulated Only in Cardiomyocytes Infected With TMEV

To characterize gene expression patterns in cardiomyocytes infected with TMEV, we conducted a supervised analysis using the 2-way comparison of microarray data between TMEVinfected and control mock-infected HL-1 cell culture samples (**Supplementary Materials and Methods**). We visualized the numbers of up- or downregulated genes of infected HL-1 cells compared with controls, using a volcano plot (**Figures 3A–D**) (44–46). We identified substantial numbers of genes whose expressions changed 4 h p.i. (185 upregulated and 413 downregulated genes, >2-fold compared with controls), and their numbers were increased 8 h p.i. (251 upregulated and 1,211 downregulated genes).

To compare these gene expression patterns among samples, we generated the heat map for highly up- or downregulated genes (13), using top 20 of up- or downregulated genes of HL-1 samples 8 h p.i. (**Figure 3E**). At 8 h p.i., TMEV infection upregulated genes associated with innate immunity: IFN-induced genes, including Ifit1, and Cxcl10 and Ccl5. TMEV-infected HL-1 samples 4 h p.i. showed a similar gene expression pattern to that of 8 h p.i. We categorized the genes up- or downregulated in TMEV-infected HL-1 cells, using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8 (Laboratory of Human Retrovirology and Immunoinformatics, Leidos Biomedical Research, Inc., Frederick, MD). Among the upregulated genes, DAVID identified 18 pathways whose P values were <0.05 (**Supplementary Table 1**), including "chemokine-mediated signaling pathway," "cellular response to IFN-α and IFN-β," and "positive regulation of T cell migration." Among the downregulated genes, DAVID identified 28 pathways, including "cell division" and "heart morphogenesis."

To determine the requirement of live virus for the gene expression changes, we incubated HL-1 cells with ultraviolet (UV)-irradiated (replication inactive) TMEV (UV-TMEV) (47). Following 8 h incubation, UV-TMEV upregulated 41 genes, among which only one gene Mir690 was identified, while none of the 41 genes significantly upregulated in live TMEV-infected HL-1 cells (**Figure 3C**; **Supplementary Table 2**). UV-TMEV also downregulated 10 genes whose immunological functions are unknown, while one gene [slingshot protein phosphatase 2 (Ssh2)] among the 10 genes was also downregulated in live TMEV-infected HL-1 cells. To identify cell-type specific gene expression, we conducted microarray analyses using TMEV- and mock-infected Neuro-2a cells (43). In Neuro-2a cells, TMEV infection did not upregulate any genes significantly, while two genes with unknown functions were downregulated (**Figure 3D**; **Supplementary Table 2**). No innate immunity-related genes were induced in HL-1 cells incubated with UV-TMEV or TMEVinfected Neuro-2a cells (**Figure 3E**; **Supplementary Table 2**). Thus, induction of innate immunity-related genes by TMEV requires live virus and is cell-type specific.

To identify sets of genes whose expression patterns were unique under the experimental conditions, we conducted kmeans clustering (**Figure 3F**) (33). Among 10 clusters, three clusters (clusters 3, 4, and 6) showed differentially expressed patterns, which were visualized by radar chart showing the different expression patterns of cluster centers in each cluster. Most upregulated genes in TMEV-infected HL-1 cells 4 and 8 h p.i., including Ifit1 and Cxcl10, were categorized in clusters 3, while the downregulated genes only 8 h p.i. or 4 and 8 h p.i. were categorized in cluster 4 or 6, respectively. Lists of genes in each cluster were shown in **Supplementary Table 3**.

### PCA of Microarray Data Separates Between the TMEV-Infected HL-1 and Control Groups

To compare overall gene expression patterns among samples, we conducted unsupervised PCA by entering microarray data from each sample without labeling of grouping (33, 42). In PCA, each principal component (PC) is determined automatically, and PC values for each sample data are plotted, for example,

was detectable in TMEV-infected HL-1 cells (open column), but not detectable (N.D.) in mock-infected HL-1 cells (closed column) or infected Neuro-2a cell culture (data not shown). (D) Viral titers of cell-free (•, N) and cell-associated virus (◦, 1) in HL-1 or Neuro-2a cell culture were determined by plaque assays with baby hamster kidney (BHK)-21 cells (24). In both HL-1 and Neuro-2a cells, cell-free virus titers increased substantially at 12 h p.i., while cell-associated viral titers increased at 8 h p.i. and peaked at 12 h p.i.

PC1 as the x-axis and PC2 as the y-axis. When the data of all HL-1 samples from mock-infection, TMEV-infection, and UV-TMEV incubation were entered, we found that the samples were separated into two distinct populations: live TMEV-infected samples vs. uninfected samples (mock-infection and UV-TMEV) (**Figure 4A**). According to the proportion of variance, PC1 explained 46% of the variation among samples (**Figure 4B**). Factor loading for PC1 showed that innate immunity-related genes, including Cxcl10, Ccl5, and Ifit1, contributed to PC1 positively, while a group of genes, including Ssh2 (48), listerin E3 ubiquitin protein ligase 1 (Ltn1), and MINDY lysine 48 deubiquienase 2 (Mindy2) (49), contributed negatively (**Figure 4C**). Thus, both supervised and unsupervised analyses suggested that innate-immunity-related genes, including Cxcl10, Ccl5, and Ifit1, could be biomarkers for the differences between the TMEV-infected and control groups in vitro.

The gene expression changes in TMEV-infected HL-1 cells appeared to be similar to those found in the heart during phase I of in vivo TMEV infection. Thus, we conducted PCA by entering microarray data from TMEV-infected HL-1 cells

FIGURE 3 | Supervised bioinformatics analysis of transcriptome data from cardiomyocyte HL-1 cells infected with TMEV (33, 42). (A–D) Volcano plots of significantly up-regulated (upper right) or down-regulated genes (upper left) in TMEV-infected cells by the OriginPro 8.1 (OriginLab Corporation, Northampton, MA), to assess significance together with log ratio of transcriptome data (Supplementary Materials and Methods) (33). Log ratios of gene expression in the TMEV-infected cell culture compared with mock-infected cell culture were used as the x-axis and the logarithms of *P* values to base 10 were used as the y-axis. (A) TMEV-infected HL-1 cells at MOI = 10 at 4 h p.i. (B) TMEV-infected HL-1 cells at MOI = 10 at 8 h p.i. (C) HL-1 cells incubated with ultraviolet (UV)-irradiated TMEV for 8 h. (D) TMEV-infected Neuro-2a cells s at MOI = 10 at 8 h p.i. (E) Heat map of 20 up- or down-regulated genes in TMEV-infected HL-1 cells at MOI = 10 at 8 h p.i. by R *(Continued)*

FIGURE 3 | version 3.4.3 and the R packages "gplots" and "genefilter." Red, blue, and white indicate up-regulation, down-regulation, and no change, respectively. Interferon-inducible genes (*Ifit1* and *Ifit3*) and chemokines (*Cxcl10* and *Ccl5*) were significantly up-regulated. While TMEV-infected HL-1 cells at MOI = 10 for 4 h or at MOI = 1 for 8 h showed a similar expression pattern in several genes, HL-1 cells incubated with UV-TMEV for 8 h did not up- or downregulated these genes. (F) Radar chart based on the values of cluster centers from *k*-means clustering. The number at each vertex is the cluster number (clusters 1 to 10), whereas the numbers along the axis (−2 to 1) are log ratios compared with mock-infected controls. Up-regulated genes in TMEV-infected HL-1 cells 4 and 8 h p.i., including *Cxcl10* and *Ifit1*, were categorized mostly in cluster 3, while the downregulated genes only 8 h p.i. or 4 and 8 h p.i. were categorized in cluster 4 or 6, respectively. In UV-TMEV-incubated HL-1 cells, most genes showed no change; the values of most cluster centers were log ratios = 0. List of genes in each cluster was shown in Supplementary Table 3.

and those from heart samples of all three phases in TMEVinfection in vivo (**Figure 4D**) (33, 42) to see whether the overall gene expression pattern of TMEV-infected HL-1 cells could be similar to those of TMEV-induced myocarditis in vivo. PCA clearly separated in vivo samples from three phases into three distinct groups by PC1 values; the PC1 values reflected distinct pathophysiology of three phases of myocarditis. Here, the PC1 values of in vitro TMEV-infected HL-1 cells 4 and 8 h p.i. were similar to that of heart samples of phase I in TMEVinduced myocarditis. On the other hand, PC2 values of in vitro TMEV-infected HL-1 cells were lower than those of in vivo samples. Thus, PC2 values could reflect the differences between in vitro and in vivo conditions, rather than phase-specific pathophysiology.

#### CYTOKINES AND LYMPHATICS IN TMEV INFECTION

#### Cytokines and Lymphatics in TMEV-Induced Myocarditis

Although several cytokines have been shown to influence lymphangiogenesis, the pro-lymphangiogenic cytokine, vascular endothelial growth factor (VEGF)-C or D (50), binds to VEGF receptor (VEGFR) 3 on lymphatic vessel endothelial cells to induce lymphangiogenesis during inflammation ("inflammationassociated lymphangiogenesis," IAL) (35), where macrophagesecreted VEGF induces sprouting of lymphatic vessels at the preexisting lymphatic vessels (51). The VEGF-A/VEGFR2, which is typically associated with angiogenesis (52), also induces lymphangiogenesis in a context-dependent manner, such as corneal lymphangiogenesis (50).

Cardiac lymphatic networks exist in all three layers of the heart, forming subendocardial, myocardial, and subepicardial plexuses, while these lymphatics share anatomical and physiological characteristics with those in other organs (53). Disturbed cardiac lymphatic drainage can contribute to many forms of cardiac pathology, such as dilated cardiomyopathy and heart failure. Myocarditis provokes myocardial edema and inflammatory infiltration of lymphocytes and macrophages; both can drive underlying lymphatic pumping disturbances. Lymphatic contraction is often impaired by inflammatory mediators, including cytokines, prostaglandins (PGs), and nitric oxide (54); inflammatory mediators produced during myocarditis could depress lymphatic pumping and drainage. In TMEV-induced myocarditis, we previously showed that pro-inflammatory cytokine interleukin (IL)-1β and tumor necrosis factor (TNF)-α upregulation was associated with myocarditis in vivo without induction of lymphatic markers, including lymphatic vessel endothelial hyaluronan receptor (LYVE)-1, or VEGFR3 (55). In addition, IL-1β reduced contractility of cardiac lymphatic muscle cells via cyclooxygenase (COX)-2/PGE<sup>2</sup> signaling with synergistic cooperation by TNF-α in vitro. These results suggest that a loss of cardiac lymphatic tonic contractility induced by IL-1β could exacerbate myocardial edema, leading to accumulation of inflammatory cytokines/chemokines and immune cells within the heart, while this may prevent viral spread to the systemic circulation.

### Lymphatics and Virus Infection in the CNS

The CNS has been regarded as an immunologically privileged site due to several characteristics that isolate it from systemic immune responses under physiological conditions: lack of MHC molecules on most resident cells, the presence of the blood-brain barrier (BBB) with low adhesion molecule expression on blood vessels, and no conventional lymphatic system (56). Recently, meningeal lymphatic vessels have been identified in the CNS that may be used for clearance of not only soluble molecules (57) but also immune cells (58) from the CNS and drainage to the deep cervical lymph node. Although there have been many experimental reports showing the transport of soluble molecules, the cellular transport from the CNS to cervical lymph nodes is still controversial. For example, even highly malignant cancer cells in the CNS do not metastasize to any peripheral lymph nodes; cellular transport using the lymphatics from the CNS seems to be regulated with unknown mechanisms. Although the soluble antigens transported from the CNS to cervical lymph nodes can be used for antigen presentation, it is unclear whether this pathway is a major priming site for presentation of CNS antigens since cervical lymph node swelling is not seen in CNS microbial infections or CNS inflammatory diseases.

Using experimental intravenous injection of simian immunodeficiency virus (SIV) in rhesus macaques, Dave et al. (59) demonstrated the presence of SIV in the CNS and cervical lymph nodes with lower levels of virus in plasma, suggesting SIV spread from the CNS to draining cervical lymph nodes. Although the exit of SIV from the CNS via lymphatic vessels should be confirmed by future studies, including the comparison of viral genotypes between the CNS and lymph nodes, this study showed the possibility that lymphatics might be used for virus clearance and/or exit from the CNS to the periphery.

(A) PCA separated samples into two groups: TMEV-infected samples vs. uninfected samples (mock infection and UV-TMEV), where principal component (PC) 1 reflected live virus infection. (B) The proportion of variance showed that PC1 explained 46% of the variance among the samples. (C) Factor loading for PC1 showed that chemokines (*Cxcl10* and *Ccl5*) and interferon-inducible genes (*Ifit1*) were correlated with PC1 positively, while several genes including slingshot protein phosphatase 2 (Ssh2) and listerin E3 ubiquitin protein ligase 1 (*Ltn1*) were correlated with PC1 negatively. (D) PCA of transcriptome data of TMEV-infected HL-1 cells 4 and 8 h p.i. and heart samples from phases I (4 days p.i.), II (7 days p.i.), and III (60 days p.i.) in TMEV-induced myocarditis *in vivo*. PCA showed that phase I samples and *in vitro* samples had similar PC1 values, compared with phases II and III samples. PCA was conducted using R version 3.4.3 (13). Microarray data were converted into tab-delimited text format and calculated using an R program "prcomp".

### Lymphocyte Entry/Exit and Lymphatics in CNS TMEV Infection

In MS and its animal models, the presence of immune cell infiltrates, particularly lymphocytes, in the CNS has been correlated with disease activity and neuropathology. Lymphocyte entry into the CNS is accompanied by upregulation of adhesion molecules on lymphocytes and blood vessels as well as a breakdown of the BBB (60) (**Figure 5**). Among the adhesion molecules, the interactions between very late antigen (VLA)-4 (CD49d/CD29) and vascular cell adhesion molecule (VCAM)- 1 (CD106) (63) as well as leukocyte function-associated antigen (LFA)-1 (CD11a/CD18) and intercellular adhesion molecule (ICAM)-1 (CD54) (64) have been shown to play a key role for lymphocyte extravasation into the CNS parenchyma (63). The BBB is composed of tight junctions of endothelial cells, the basement membrane, and astrocyte foot processes. Downregulation of tight junction proteins, including occludin and claudin, has been associated with BBB breakdown and disease activities in MS and its animal models (62). On the other hand, the pathophysiology of lymphocyte exit from the CNS is unclear, although newly identified CNS lymphatic vessels (58) might contribute to clearance of lymphocytes (and microbes) from the CNS, in theory.

In TMEV infection, we determined the extent of which expressions of the adhesion molecules, BBB and lymphatic molecules could be associated with CNS disease activity (53,

FIGURE 5 | Three components of lymphocyte entry into and exit from the CNS (61, 62). To initiate inflammation in the CNS, lymphocytes interact with endothelial cells of blood vessels via up-regulated adhesion molecules, particularly very late antigen (VLA)-4 and lymphocyte function-associated antigen (LFA)-1 on lymphocytes with vascular cell adhesion molecule (VCAM)-1 and intercellular adhesion molecule (ICAM)-1 on endothelia, respectively. Downregulation of molecules composed of the blood-brain barrier (BBB) also help in lymphocyte entry into the CNS parenchyma. While the precise mechanism of lymphocyte exit from the CNS is unknown, one hypothesis is the presence of vessels similar to peripheral lymphatic vessels, whose markers include prospero homeobox (PROX) 1 and lymphatic vessel endothelial hyaluronan receptor (LYVE) 1, may help in lymphocyte exit from the CNS to deep cervical lymph nodes. Increased lymphocyte entry together with decreased lymphocyte exit could lead to enhancement of CNS inflammation.

(acute polioencephalomyelitis) and 35 (TMEV-IDD) days p.i. (61). (A) We drew a heat map, using mRNA data of 20 adhesion molecules, 14 BBB molecules, and 12 lymphatic molecules listed in Figure 5 (total 46 genes). Most adhesion molecule genes were upregulated 7 and 35 days p.i., while only a few adhesion molecules were upregulated 4 days p.i. BBB and lymphatic molecules showed no change or slight downregulation. (B) Radar chart based on the values of cluster centers from *k*-means clustering (Supplementary Table 4). The number of each vertex is the cluster number (clusters 1 to 8), whereas the number along the axis (−2 to 8) are log ratios compared with mock-infected controls. Radar chart showed that the expression patterns of sets of genes were similar between days 7 and 35 p.i. Upregulated genes were categorized mostly in clusters 1, 3, and 7. (C) PCA of the 46 genes listed in Figure 5 separated controls/day 4 p.i. samples vs. days 7 and 35 p.i. samples based on PC1 values (proportion of variance was 85.9%), which reflect CNS cell infiltration. (D) Factor loading for PC1 showed that upregulation of adhesion molecules was associated with CNS inflammation, while downregulation of BBB and lymphatic molecules may play a minor role.

61). Using the RNA sequencing transcriptome data from the spinal cord of TMEV-infected mice harvested 4, 7, and 35 days p.i. (**Supplementary Materials and Methods**), we compared mRNA levels of representative 20 lymphocyte and vascular adhesion molecules, 14 BBB molecules, and 12 lymphatic molecules among samples (**Figure 5**). Both 7 and 35 days p.i., heat map showed that most adhesion molecules were upregulated, while lymphatic and BBB molecules showed no change or slight downregulation (**Figure 6A**). Since samples 7 days p.i. contain gray matter inflammatory lesions due to acute polioencephalomyelitis and those 35 days p.i. contain inflammatory demyelination in the white matter, we expected substantial difference in gene expression patterns between the two sample groups. Unexpectedly, however, the levels of most adhesion molecules 7 days p.i. were similar or slightly higher, compared with 35 days p.i. Only glycosylation-dependent cell adhesion molecule (GLYCAM) 1 was significantly upregulated from 7 to 35 days p.i. (65). Thus, GLYCAM1 may have a role in chronic demyelination. Most genes 4 days p.i. showed no or few changes, which is consistent with the histological finding that immune cell infiltrates become obvious 5 days p.i. in CNS TMEV infection. In radar chart that visualized gene expression patterns by k-means clustering, clusters 1, 3, and 7 were composed of highly upregulated genes 7 and 35 days p.i. (cluster 1, LFA-1 and 2, E- and L-selectin; cluster 3, ICAM-1, and other molecules; and cluster 7, VLA-4, VCAM-1, and GLYCAM1) (**Figure 6B**; **Supplementary Table 4**). Cluster 4 was composed of downregulated genes 4, 7, and 35 days p.i., including VEGF-C, LYVE1, and claudin 22.

We also conducted PCA using the same 46 gene expression data and found that expression patterns of molecules associated with CNS lymphocyte entry and exit could distinguish samples without CNS cell infiltration (control and 4 day p.i. samples) vs. with CNS cell infiltration (7 and 35 days p.i. samples) by PC1 values (**Figure 6C**). Factor loading for PC1 showed that upregulation of adhesion molecules (66) was correlated with PC1 values that reflect CNS inflammation 7 and 35 days p.i. (**Figure 6D**). Downregulation of several BBB molecules, including claudin 10 (67) and reelin, was weakly correlated with PC1 values. Downregulation of BBB may play a minor role in CNS inflammation induced with TMEV, although downregulation of BBB molecules has been reported not only in MS and autoimmune model for MS but also in another experimental CNS viral model induced with mouse hepatitis virus (68).

Inflammation has been reported to induce lymphangiogenesis in several organs and tissues. Following intracerebral TMEV infection in the CNS, however, most lymphatic markers were not upregulated at any time points, although the constitutive expression in control uninfected CNS tissues supports the presence of lymphatic-like structure in the CNS. This is consistent with our previous findings on the protein levels of lymphatic biomarkers, in which there was no increase in lymphatic markers, LYVE1 or prospero homeobox protein (PROX)1 in the CNS of TMEV-IDD (39). Most lymphatic molecules were actually downregulated slightly on 7 and 35 days p.i., while factor loading for PC1 showed that downregulation of



lymphatic molecules was weakly correlated with PC1 values. This suggests that dysfunction of lymphatic-like structure might delay exit of inflammatory cytokines/chemokines and/or cells from the CNS, enhancing inflammation, only to some extent. On 14 days p.i. when inflammation had subsided in the CNS, the levels of most lymphatic molecules of the TMEV-infected spinal cord were similar to those of uninfected control spinal cord (data not shown); this may reflect that recovery of lymphatic flow from the CNS contributes to exit of inflammatory cytokines/chemokines and/or cells from the CNS around 2 weeks p.i.

In TMEV-IDD, the balance between lymphocyte entry and exit could play a key role in inflammation in the CNS; upregulation of adhesion molecules rather than downregulation of BBB molecules could contribute to lymphocyte entry, while downregulation of lymphatic molecules may play a minor role in prolonged inflammation. In theory, dysfunction of the lymphatics results in the persistence of lymphocytes and cytokines/chemokines in the CNS (69). This would lead to chronic inflammation and immune-mediated demyelination by immunopathology, whereas such lymphostasis might confine TMEV to the CNS, limiting systemic viral spreading. Here, virusspecific lymphocytes among chronic cellular infiltrates in the CNS may also minimize virus replication in the CNS.

In summary, in TMEV infection, innate immune cytokines may play distinctive and diverse roles in lymphatic networks during inflammatory disease depending on the organs, which contribute to the levels of inflammation and to virus persistence (**Table 1**). Although TMEV can infect major cell types of the CNS (neurons) and the heart (cardiomyocytes), only infected cardiomyocytes expressed innate immunity-related molecules. In addition, lymphatic vessels in infected organs may also be differentially affected between the CNS and the heart. In the heart of TMEV-induced acute myocarditis, IL-1β with TNF-α could functionally alter lymphatics, while downregulation of lymphatic molecules might contribute to persistent virus infection and inflammation in the CNS of TMEV-IDD. These potential factors may contribute to organ-specific viral immunopathology in TMEV infection.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the criteria outlined by the National Institutes of Health (NIH). The protocol was approved by the Institutional Animal Care and Use Committee of LSUHSC-S and Kindai University.

### AUTHOR CONTRIBUTIONS

IT and SO for substantial contributions to the conception or design of the work. SO, EK, FS for the acquisition of data. SO, UC, and MT for analysis of data. IT, AM, MA-K, and JA for interpretation of data for the work. IT, SO, NM, JY, and JA for drafting the work or revising it critically for important intellectual content.

#### FUNDING

This work was supported by the fellowships (FS and SO) from the Malcolm Feist Cardiovascular Research Endowment, LSU Health Sciences Center, Shreveport, and grants from the National Institute of General Medical Sciences COBRE Grant (8P20 GM 103433, IT), and the Science Research Promotion Fund from the Promotion and Mutual Aid Corporation for Private Schools of Japan (FS), the Faculty Assistance and Development Research Grants from the Kindai University

#### REFERENCES


Research Enhancement Grant (FS and SO), the KAKENHI from the Japan Society for the Promotion of Science [Grant-in-Aid for Young Scientists (B), JP17K15628 (FS), Grants-in-Aid for Research Activity Start-up, JP16H07356 (IT), Grant-in-Aid for Scientific Research on Innovative Areas Frontier Research on Chemical Communications (No 17H06400 and 17H06404, IT)] and Novartis Pharma Research Grants (IT).

#### ACKNOWLEDGMENTS

We thank John A. Vanchiere, M.D., Ph.D., Melody Cunningham Baddoo, Rona S. Scott, Ph.D., Mitsugu Fujita, M.D., Ph.D., and Ah-Mee Park, Ph.D. for helpful discussions, and Gloria B. McClure, Paula Polk, Elaine A. Cliburn, Sadie Faith Pearson and Lesya Ekshyyan for excellent technical assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2018.02870/full#supplementary-material


encephalomyelitis by VLA-2 blockade. Brain Pathol. (2007) 17:45–55. doi: 10.1111/j.1750-3639.2006.00042.x


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

Copyright © 2018 Omura, Kawai, Sato, Martinez, Minagar, Al-Kofahi, Yun, Cvek, Trutschl, Alexander and Tsunoda. 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.

# Cytokine-Mediated Tissue Injury in Non-human Primate Models of Viral Infections

#### Cordelia Manickam<sup>1</sup> , Spandan V. Shah<sup>1</sup> , Olivier Lucar <sup>1</sup> , Daniel R. Ram<sup>1</sup> and R. Keith Reeves 1,2 \*

*<sup>1</sup> Center for Virology and Vaccine Research, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, United States, <sup>2</sup> Ragon Institute of Massachusetts General Hospital, MIT and Harvard, Cambridge, MA, United States*

Viral infections trigger robust secretion of interferons and other antiviral cytokines by infected and bystander cells, which in turn can tune the immune response and may lead to viral clearance or immune suppression. However, aberrant or unrestricted cytokine responses can damage host tissues, leading to organ dysfunction, and even death. To understand the cytokine milieu and immune responses in infected host tissues, non-human primate (NHP) models have emerged as important tools. NHP have been used for decades to study human infections and have played significant roles in the development of vaccines, drug therapies and other immune treatment modalities, aided by an ability to control disease parameters, and unrestricted tissue access. In addition to the genetic and physiological similarities with humans, NHP have conserved immunologic properties with over 90% amino acid similarity for most cytokines. For example, human-like symptomology and acute respiratory syndrome is found in cynomolgus macaques infected with highly pathogenic avian influenza virus, antibody enhanced dengue disease is common in neotropical primates, and in NHP models of viral hepatitis cytokine-induced inflammation induces severe liver damage, fibrosis, and hepatocellular carcinoma recapitulates human disease. To regulate inflammation, anti-cytokine therapy studies in NHP are underway and will provide important insights for future human interventions. This review will provide a comprehensive outline of the cytokine-mediated exacerbation of disease and tissue damage in NHP models of viral infections and therapeutic strategies that can aid in prevention/treatment of the disease syndromes.

#### Keywords: viral infections, cytokines, animal model, tissue damage, non-human primates

#### INTRODUCTION

Microbial pathogens are constantly evolving to evade the host's immune system, and even with several decades of research and modern therapeutics, chronic diseases such as those caused by human immunodeficiency virus (HIV-1) and hepatitis C virus (HCV) are still globally prevalent. Viruses use multiple evasive strategies such as avoiding detection by pattern recognition receptors, T cell receptors and antibodies, mimicking or blocking cytokines, chemokines and other host proteins, and/or directly depleting immune cell subsets [reviewed in (1)]. Disruption of the cytokine milieu is also an important and commonly used strategy by viruses (2–4), since cytokines play important roles in shaping both innate and adaptive immunity. Cytokines are soluble proteins

#### Edited by:

*Juliet Spencer, Texas Woman's University, United States*

#### Reviewed by:

*Vijayakumar Velu, Emory University, United States Cristian Apetrei, University of Pittsburgh, United States*

> \*Correspondence: *R. Keith Reeves rreeves@bidmc.harvard.edu*

#### Specialty section:

*This article was submitted to Viral Immunology, a section of the journal Frontiers in Immunology*

Received: *22 September 2018* Accepted: *20 November 2018* Published: *04 December 2018*

#### Citation:

*Manickam C, Shah SV, Lucar O, Ram DR and Reeves RK (2018) Cytokine-Mediated Tissue Injury in Non-human Primate Models of Viral Infections. Front. Immunol. 9:2862. doi: 10.3389/fimmu.2018.02862* secreted by cells during inflammation that act as key mediators of immune cell recruitment and modulators of the immune response via a complex network of cellular interactions and signaling pathways. So far, more than 300 cytokines including chemokines, interferons (IFN), and lymphokines have been described (5). While cytokines can be broadly classified based on the nature of the immune response as pro-inflammatory cytokines such as interleukin (IL)-1, IL-6, type 1 IFN, tumor necrosis factor (TNF)-α, and anti-inflammatory cytokines such as IL-4, IL-10, and transforming growth factor (TGF)-β, they have pleiotropic functions whereby individual cytokines can have either pro- or anti-inflammatory properties according to the cell system involved.

In viral infections, cytokines play central roles in the development of protective anti-viral responses, but also potential immunopathology associated with chronic viral diseases. Viral interactions with host cellular receptors triggers proinflammatory cytokine secretion which are essential for viral clearance. However, dysregulations in the cytokine type and quantitative levels can lead to overactivation of immune cells, which in turn cause tissue damage leading to fatal complications. For instance, extensive characterization of IFN-α and its direct antiviral activity since its discovery in 1957 (6), has led to successful treatment of "non-A, non-B (NANB) hepatitis" even before the actual identification of HCV as the causative agent (7). Combination therapy of pegylated IFN-α with ribavirin was the standard therapeutic regime for chronic HCV-infected patients until the recent introduction of directly acting antivirals. However, IFN-α therapy can induce side effects such as fever and headache to severe life threatening conditions including thyroid, visual, auditory, renal and cardiac impairments, and pulmonary interstitial fibrosis (8). The therapeutic use of cytokines for infectious diseases, autoimmune diseases and malignancies, may also come at a steep price, since prolonged use of cytokines present severe side-effects due to the pleiotropic nature of these molecules (9–14). While, it is necessary to understand cytokine dysregulations in viral diseases to anticipate potential tissue injury and deterioration, their pleotropic, rapid, and in some cases local and long term tissue effects make the study of cytokines in humans challenging with potential development of fatal complications. These challenges can be met by the use of animal models. Animal models have been used since more than 2400 years and currently are employed in all areas of biomedical research including basic biology, infections, immunology, cancer, metabolic diseases, and behavioral studies (15). This review is primarily focused on the virus mediated cytokine dysfunctions in animal models specifically non-human primates (NHP), which are already fundamental in the validation of human data.

#### NEED FOR ANIMAL MODELS IN STUDIES OF VIRAL IMMUNITY

Much of what is known regarding antiviral immunity and tissue inflammation comes from studies conducted in animal models of human diseases. Animal models act as preclinical and translational gatekeepers since they allow the study of cellular interactions in vivo and elucidation of disease pathogenesis in tissues that may be difficult to access in humans. While mouse models have provided tremendous benefits to immunologists in understanding immune responses in humans, 65 million years of divergent evolution has contributed to significant differences in cytokines and cytokine receptors for the two species. Studies have shown poor correlation in genomic responses to acute inflammatory stress between humans and mice (16), and engagement of different chemokine/cytokine pathways in response to oxygen and glucose deprivation by human neurons compared to murine neurons (17). IL-13 seems to induce B cell class switching for IgE production specifically in humans whereas mice require IL-4 (18, 19). Similarly, IL-7 receptor deficiency inhibits development of all T and B lymphocytes in mice (20), but only T cells in humans (21). Furthermore, a number of pathogens like influenza, HIV, or dengue are highly tropic to their respective hosts and do not mimic human pathologies in mice, potentially restricting the use of mice as models for some infectious diseases [reviewed in (22)].

NHP are perhaps the most commonly utilized models to study and understand immune responses against human infectious agents and for preclinical evaluation of therapeutics and vaccines (**Figure 1**). NHP have proven essential for research breakthroughs in maladies such as cancer, Parkinson's disease, heart diseases, and various infectious diseases such as HIV, Zika, Ebola, influenza, and others (23, 24). Even though NHP research accounts for <1% of the all the biomedical laboratories working in animal models (24), the advantages offered by NHP due to the genetic and physiological homology to humans are manifold. Indeed, human and NHP cytokines are relatively conserved with 95% amino acid identity of most cytokines such as IL-2 and IFN-γ for Old World NHP and up to 90% amino acid identity for New World NHP (25). In addition, many cross reactive reagents and monoclonal antibodies for the detection of cytokines have been evaluated and validated for NHP species (NIH Non-human Primate Reagents Resource; http://www.nhpreagents.org) (25–28), making NHP attractive animal models to study viral pathogenesis and disease progression.

### NHP MODELS COMMONLY USED FOR VIRAL DISEASES

#### Great Apes

The great apes used previously as animal models include chimpanzees (Pan troglodytes), and to a lesser extent orangutans (Pongo pygmaeus) and gorillas (Gorilla beringei) (29). Chimpanzees share >98% DNA sequence homology to humans; and yet surprisingly, have immune systems that respond much more robustly to infections like HIV and hepatitis B virus (HBV). HBV and HCV can only pathogenically infect humans and chimpanzees, thus making chimpanzees, at one time, the primary animal model for therapeutics and vaccine research (30–32). However, the use of great apes in biomedical research has become increasingly restricted for ethical and cost reasons and therefore other NHP models are being increasingly utilized.

### Old World Monkeys

The Old World monkeys are primarily found in the continents of Africa, Asia, and Europe with rhesus macaques (Macaca mulatta), cynomolgus macaques (Macaca fascicularis), sooty mangabeys (Cercocebus atys), African green monkeys (Chlorocebus aethiops), and baboons (Papio spp.) being the predominant species used in biomedical research. Rhesus/cynomolgus macaques are perhaps the most widely utilized NHP animal models to study human infectious diseases. Besides HIV (33), macaque models have been used for infectious diseases such as influenza (34, 35), HBV (36, 37), HCV (38–40), measles (Morbillivirus) (41–43), cytomegalovirus (CMV) (44– 46), among many others (47). Sooty mangabeys and African green monkeys are also used to study HIV and African green monkeys are used as a model for influenza (48). Less commonly used tree shrews (Tupaia belangeri) have also been explored as a model for HCV infection (49, 50).

#### New World Monkeys

New World monkeys or neotropical primates include cottontop tamarins(Saguinus Oedipus), common marmosets (Callithrix jacchus), owl monkeys (Aotus lemurimus), and squirrel monkeys (Saimiri boliviensis), which are commonly located in Central and South America. Although, the New World monkeys are more divergent than Old World NHP from humans, they provide a distinct advantage in biomedical research due to their relatively smaller size and lower cost compared to other NHP. Marmosets and tamarins have been used to study many flaviviruses such as HCV, Dengue, and Zika (51–56). Owl monkeys can be infected with Hepatitis E Virus (57) and at least some individual animals might have HIV-1 compatible CD4 alleles (58) making them potentially useful for HIV research. Squirrel monkeys have been utilized as animal models for HTLV-1 pathogenesis and vaccine development (59, 60) and as an experimental model for Nipah Virus (61).

#### CYTOKINE DYSREGULATION IN VIRAL INFECTION MODELS

#### HIV/Acquired Immunodeficiency Syndrome (AIDS)

The emergence of HIV (Genus: Lentivirus, Family: Retroviridae) is the result of the combination of at least four simian immunodeficiency virus (SIV) transmission events from chimpanzees or gorillas to humans (62, 63). Therefore, SIV and simian/human immunodeficiency virus (SHIV) infections in NHP are commonly used to model HIV pathogenesis and development of vaccines and therapeutics. Specifically, rhesus macaques and sooty mangabeys have been critical in understanding the early phase of the infection (33, 64). Several studies (discussed below) have shown the principal involvement of an unusually vigorous immune activation leading to the progression and establishment of AIDS.

Based on plasma parameters from HIV-infected patients, the virus-mediated cytokine storm starts early in infection even before peak viremia is reached (65, 66). It rapidly initiates a cascade of events characterized by the production of the early pro-inflammatory cytokines, IL-15, and IFN-α, quickly followed by the more sustained TNF-α and monocyte chemoattractant protein (MCP)-1 during infection. Other pro-inflammatory cytokines like IL-6, IL-8, IL-18, and IFN-γ are elevated 2 days post the first wave of proinflammatory cytokines. At the same time, the secretion of IL-10, an immunoregulatory cytokine exponentially increases until it peaks at 5 days of infection (65). While, the IL-10/IL10-R pathway has a major role in preventing tissue damage observed during HIV infection by inhibiting Th1 responses and the production of anti-viral cytokines (IFN-α, IFN-γ, IL-2), it also contributes to viral persistence. Furthermore, the expression of the PD-1/PDL-1 pathway drives the inhibition of T cell function (67) and indirectly up-regulates expression of IL-10 (68). Indeed, blockade of PD-1 by anti-PD-1 antibody in infected rhesus macaques augmented SIV specific IFN-γ responses in CD8+ T cells in the blood, and could be synergized with vaccination and anti-retroviral therapies (69, 70). However, more NHP studies are necessary to establish the importance of PD-1 blockade particularly in mucosal tissues.

The magnitude of the cytokine storm is broadly associated with the clinical outcome in infected rhesus macaques and sooty mangabeys (66, 71). Indeed, the progressive infection in rhesus macaques is associated with production of IL-15, IL-18, IFN-γ, granulocyte-colony stimulating factor (G-CSF), MCP-1 and macrophage inflammatory protein (MIP)-1β but not in nonprogressive sooty mangabeys (66). Similar cytokine dysregulation evidenced as elevated IL-12 has also been reported in HIV seroconverts (72) and South African women who are high risk population for acquisition of HIV infection (73). Furthermore, the cytokine storm leads to immune activation with global damage in mucosal tissues, specifically the gut and gut-associated lymphoid tissue (GALT) which are the early and major sites of virus replication (74). Specifically, the virus targets the IL-17/Th-17 pathway that is essential for preservation of the gut barrier, maintenance of the gut microbial environment, and prevention of translocation of microbial products into the circulation that could otherwise cause immune activation (75, 76). However, it is shown that cART can partially restore effective CD4+ T cells (more than 50% compared to non-treated) in the gut and enhance the Th17 subset which is associated with a better clinical outcome (77). This further illustrates the importance of NHP to study gut immunity in HIV infection and evaluate therapeutic modalities at mucosal tissues (78).

SIV infection in sooty mangabeys leads to a long nonprogressive infection as observed in some HIV-infected individuals (79). Sooty mangabeys do not develop disease symptoms due to a low level of immune activation despite high level of viral replication (80). Instead of an inflammatory immune response, elevated regulatory T cells (Treg) and associated cytokines, TGF-β and IL-10 limit the level of immune activation (80). Similarly in infected African green monkeys, an anti-inflammatory environment is rapidly established due to increases in Treg frequency, TGF-β, and IL-10 levels in the plasma (81). Interestingly, a comparison of acute infection in African green monkeys and rhesus macaques revealed that a rapid and elevated IFN-α is triggered in both models but return to baseline levels after 28 days of infection was observed only in African green monkeys (82). Further, no changes in the levels of pro-inflammatory cytokines such as IL-6, IL-18, and TNF-α were reported in infected African green monkeys compared to uninfected controls (83). It was also shown that sooty mangabeys have a unique genome that protects them from developing AIDS (84). Of importance, these animals possess a different TLR-4 gene compared to NHP that develop AIDS. TLR-4 is a pattern recognition receptor that senses lipopolysaccharides on bacteria and initiates pro-inflammatory cytokine induction. HIV can induce microbial translocation that elicited exacerbated TLR-4 stimulation and lead to chronic immune activation (85, 86). Therefore the differential cytokine response and an overall lower immune activation, in part confers immune protection, less tissue damage and maintenance of gut barrier in non-pathogenic SIV infection of sooty mangabeys as well as African green monkeys (87, 88).

Rhesus macaques are not natural hosts of SIV infection and therefore, some SIV strains can induce strong viral load and the development of AIDS similar to HIV-infected individual (89). In a rhesus macaque cohort infected with pathogenic or nonpathogenic strains of SIV/SHIV, the progressor cohort exhibited low IFN-γ induced by CD4+ T cells compared to CD8+ T cells whereas, the non-progressor monkeys did not develop a similar immunomodulation (90). Furthermore, infection with virulent SIVmac251 strain directly upregulated the cytokine production (IFN-α/β, IL-12, IL-18) and led to the activation of natural killer (NK) cells which are one of the major antiviral innate immune cells and also act as a bridge to the adaptive system. Interestingly, the production of antiviral cytokines (IFN-α, IFN-γ, IL-2) was also associated with viral establishment (91). An over production of IL-7 in the gut during the early days of acute SIV infection in rhesus macaques could contribute to the cytokine storm by inducing elevated chemokine expression triggering immune cell recruitment (92). Overall, the cytokine storm induces a vicious cycle by spreading the infection and causing tissue damage due to an extensive inflammation in SIV progressive NHP models. To overcome this cytokine mediated disease exacerbation, several therapeutic formulations that use cytokines including IL-12, IL-15, and IL-2 or block cytokine receptors are increasingly being tested in SIV infection models (discussed in later section).

#### Hepatitis B and C

Hepatitis B and C infections together are the leading causes of chronic liver disease worldwide (93). HBV (Genus: Orthohepadnavirus; Family: Hepadnaviridae) and HCV (Genus: Hepacivirus; Family: Flaviviridae) are hepatotropic viruses and cause both acute and chronic liver infections, which can progress to fibrosis and hepatocellular carcinoma. Interestingly, both viruses have a narrow host range (humans and chimpanzees) and have similar pathogenesis for progressive liver damage and persistence of infection. Studies in chimpanzees showed that HBV and HCV are not directly cytopathic (94–97) but instead cause liver injury due to chronic immune activation. Adaptive T cell and NK cell immunity are important in the control of viral hepatitis, but they can also prove detrimental in persistent infection. In cases of uncontrolled replication, infected hepatocytes secrete cytokines IL-8, CXCL-9, and CXCL-10, which recruit T cells to the infected liver, all correlating with histological damage (98–100). Further, innate immune NK cells are activated and recruited by high levels of IFN-α and IL-8 in the liver and induction of cytotoxic TRAIL pathway leads to killing of hepatocytes and liver injury (101). HCVmediated liver inflammation is promoted by IL-1β and the TNF superfamily cytokines such as TNF-α, TNF-β, TWEAK, and LIGHT through the activation of NF-kB and MLCK-signaling pathways to reduce hepatocellular tight junction integrity (102, 103). In HBV infection, TNF-α secretion was associated with significant fibrosis, and IL-10 and IFN-γ were associated with necroinflammation (104). Additionally, as a result of viral overload, induction of interferon stimulated genes and elevated IL-8 and chemokines such as CCL2, CXCL1, and CXCL5 results in cholestatic HCV, which is associated with metabolic dysregulation (105, 106).

Due to the narrow host range, chimpanzees were critical for initially understanding the natural history and pathogenesis of HCV and HBV (32, 107). However, because of the limited use of chimpanzees currently, other surrogate animal models are being employed. To model HBV, cynomolgus macaques have been used but with an indirect infection approach: ex-vivo baculovirusmediated HBV genome transfer in hepatocytes to cross the species barrier (108). Recently, a new virus called the capuchin monkey hepatitis B virus (CMHBV) has been discovered in Brazilian capuchin monkeys, a neotropical primate and has potential implications in the development of the much needed animals model for hepatitis B (109). The more commonly used NHP models for HCV are infections of neotropical primates, marmosets and tamarins, with the surrogate hepacivirus GBV-B of the same family Flaviviridae (51, 110, 111). Several studies showed that activated T cell immune responses and IFN-γ secretion are important for clearance of GBV-B (112, 113). However, similar to HCV-infected liver, immune activation correlated with liver damage in primary infections and reinfections in marmosets (114, 115). Activated NK cells expressing IFN-γ and perforin were accumulated in the liver and in addition elevated plasma IFN-γ and RANTES were associated with acute hepatitis in infected animals (114). Further, infected marmosets developed metabolic dysfunctions associated with GBV-B infection even after clearance of viremia indicating that viral hepatitis induces a cascade of events toward hepatic and systemic inflammation. Particularly, imbalance in levels of proinflammatory adipocytokines such as resistin and plasminogen activator inhibitor-1 secreted by dysfunctional adipose tissues contribute to local, systemic, and metabolic malfunctions (116). Given the importance of liver immune responses in progression of viral hepatitis, limited access to liver tissues has severely impeded development of HCV vaccine and HBV therapeutics.

#### Zika

Infections with Zika virus (ZIKV; Genus: Flavivirus; Family: Flaviviridae) have recently caused a pandemic due to abortions, stillbirths, congenital birth defects, and neonate deaths called the congenital Zika syndrome (CZS) (117). ZIKV induced neuronal necrosis in the cortical layer of the brain is mediated by a complex array of cytokines and immune factors (118–120). While studies in brain tissue are limited, in-situ immunostaining of infected fetal brain samples showed that the predominant immune response was characterized by IL-4, IL-10, IL-33, iNOS, and arginase and therefore was generally skewed toward a Th2 response (118). IL-33, in particular is directly involved in pyroptosis, activation of inflammasomes, endoplasmic reticulum stress potentially leading to cellular damage (119). However, other cytokine responses indicative of Th1, Th17, Treg, Th9, and Th22 response were also involved to a lesser extent. Immune cells including microglia, CD4+ and CD8+ T cells, Treg, NK cells, M1/ M2 macrophages, and antigen-presenting cells contribute to the pathogenesis of the ZIKV induced inflammation (118). Thus, a complex relationship between different immune factors, cell damage, and direct viral action leads to ZIKV meningitis and encephalitis.

While ZIKV induced pathology and pathogenesis studies in humans are limited to samples obtained from autopsy of severe fatal cases, NHP have been tremendously helpful in elucidating pathogenesis and fast tracked development of several vaccine candidates (121–123). Indeed, fetal neuropathology, microcephaly, and other CZS symptoms were evidenced in several NHP models including rhesus, pigtail, and cynomolgus macaques, common marmosets, and squirrel monkeys infected during early pregnancy (55, 56, 124–128). Infection studies in common marmoset dams identified immune pathways in maternal viral responses. Interestingly, an increase in IFNγ and pro-inflammatory cytokines as early as day 2 postinfection was reported. The pro-inflammatory response was maintained as elevated induction of type I/II IFN associated genes and pro-inflammatory cytokines even at day 7 postinfection and spontaneous abortion after 16–18 days of infection was reported with extensive viral infection in placenta and fetal tissues (56, 125). In infected rhesus macaques, viral persistence in the central nervous system and lymph nodes correlated with robust and early induction of proinflammatory responses and mTOR signaling pathways as evidenced by IFN-α induction at day 2, 4, and 6 postinfection and upregulation of transcript components of IFNα and IFN-stimulated genes (ISGs) (OAS2, IFT1/2/3, ISG15, IRF7, IFI44, MX1, and MX2), pro-inflammatory cytokines and chemokines (TNF- α, IL-1, IL18, CCR7, CCL2, and CCL20), immunomodulatory pathways (IL-10, TGF-β, and T regulatory cells), and inflammasome pathways (NOD2, NLRP3, CXCL10, BTG2, BST2, OSM) at day 6 post-infection (129). As a result of these activated pathways, ZIKV persistence could contribute to the characteristic neuropathology associated with ZIKV. Further several experiments in NHP are currently underway for preclinical testing of vaccine candidates and Zika is an excellent example to illustrate the importance of NHP in developing vaccines within a short span of time.

#### Dengue

Dengue virus (DENV; Genus: Flavivirus; Family: Flavivirdae), is a major vector borne disease in tropical and subtropical countries affecting approximately 100 million people worldwide, which can progress from the typical Dengue fever to fatal conditions such as Dengue hemorrhagic fever (DHF) and Dengue shock syndrome (DSS). Damage to vascular endothelium and uncontrolled activation of blood coagulation pathways in DHF can result in critical hypovolemic shock in DSS. Increased levels of cytokines, such as IFNs, IL-2, IL-8, TNF-α, and vascular endothelial growth factor A (VEGF-A) have all been reported to be associated with vascular leakage (130). Increased T cell activation and cytokine production in patients during both primary and secondary Dengue virus infections showed greater clinical severity of illness associated with cytokine storm characterized by elevated plasma pro-inflammatory cytokines such as IFN-γ, IL-6, IL-8, IL-10, CXCL9, CXCL10, CXCL11, MIF, TNF-α, and VEGF (130, 131).

Several NHP species are permissive to Dengue infection including chimpanzees, rhesus and cynomolgus macaques, sooty mangabeys, common marmosets, and owl monkeys, however the DENV induced hemorrhagic disease pattern is less common in NHP [reviewed in (132)]. In addition to elevated TGF-α and IFN-γ, increases in MCP-1, which drives immune cell recruitment, and potential cause of vascular damage was found elevated in rhesus macaques infected with DENV (133). A high dose intravenous inoculation of DENV induced classic dengue hemorrhage in infected rhesus macaques 3–5 days post-infection, with altered serum biochemical parameters indicative of coagulopathy (134). Similarly cytokine storm associated with enhanced dengue disease was detected in DENV infected marmosets, which showed a significant increase in plasma TNF-α as early as 3 days post-infection and significantly increased IFN-γ at 3, 6, and 20 days post-infection (52, 135). Indeed, antibody enhanced dengue disease in marmosets lead to CNS injury and was associated with intense TNF-α immunostaining in brain samples (135). Further, based on biomarker network analysis, two relevant strong axes during early stages of dengue fever were identified—a protective axis composed of TNF-α/lymphocytes/platelets, and a pathological axis IL-2/IL-6/monocyte/prothrombin time/viremia. Later time points post-infection showed the interaction of IFNγ/platelets/DENV-3/prothrombin time, and the involvement of type-2 cytokines (IL-4, IL-5) (136). Overall, these studies indicate that elevated proinflammatory cytokines in dengue-infected NHP have a pathogenic role associated with disease severity.

#### Influenza

Influenza A virus (Genus: Influenzavirus A; Family: Orthomyxoviridae) causes acute and severe respiratory illness in more than 1 billion people worldwide. The severity of influenza infection derives from the interplay between the virus and the host's ability to control viral infection and spread. In severe cases the host's response is hyperactivated and the resulting inflammatory response produces a cytokine storm (137–139) that is responsible for tissue injury and potentially death. This was seen during the 1918 H1N1 pandemic and more recently via the spread of H5N1. Endothelial cells from the lung have been implicated as key players in propagating the cytokine storm, in part from having elevated levels of CCL2, CCL5, and CXCL10 (140). Further inhibiting S1P1 receptor signaling on pulmonary endothelial cells, which leads to downregulation of cytokine/chemokine signaling, has been shown to decrease the development of cytokine storm following infection with influenza (140, 141).

One of the major issues in NHP modeling of influenza is the result of low animal mortality as compared to what happens in humans. While NHP can be infected with seasonal influenza strains they do not always display symptoms akin to those seen in humans (142). Influenza infection in NHP may lead to a biphasic subclinical fever early during the infection (143, 144), but this seems to be dependent on the mode of infection and dosage utilized (145, 146). Aerosol delivery using the full head chamber (145) results in a more lethal outcome, whereas the facemask leads to less severe symptoms. Infection with highly pathogenic influenza strains can induce clinical symptoms such as fever, cough and lethargy, and even showing signs of acute respiratory distress syndrome (124), bronchointerstitial pneumonia, peribronchiolar alveolitis, edema, and hemorrhaging (147–150). Further, in this model and others, increased levels of IP-10 (CXCL10), MCP-1 (CCL2), and IL-6 have been observed, which have been characterized as hallmarks of H5N1 human infection (138, 139, 151–153). Gene expression analyses have also shown that CXCL10 and CXCL11 are highly upregulated early during infection with highly pathogenic H1N1 and H5N1 and associated with elevated tissue damage (151, 152, 154). Using the full head chamber allows for the macaques to develop fulminant pneumonia that rapidly progressed to acute respiratory distress syndrome, which is the result of widespread alveolar epithelial cell death as well as depletion of alveolar macrophages.

#### CMV

CMV (Genus: Cytomegalovirus; Family: Herpesviridae) can infect and persist lifelong in multiple cell types such as macrophages, neutrophils, fibroblasts, neuronal cells, hepatocytes and others (155–159). Human CMV (HCMV) infections are often reported in patients with suppressed immune system, including the elderly, AIDS patients, cancer patients, and transplant recipients. After infection, CMV hijacks cellular machinery, induces significant alterations in gene expression including IFN signaling genes, followed by a complex cascade of signaling events (160, 161) leading to upregulation of transcription factors like NF-κB and altered cytokine production, and thus successfully evades the host immune surveillance and disseminates to all organs (162–167). While the pathogenesis is not completely clear, elevated levels of MCP-1 and MIP-1α recruiting monocyte and macrophages to the site of infection could mediate tissue damage with uncontrolled viral replication in immunocompetent patients (168, 169). In congenital CMV infections, which cause severe birth defects in newborn babies, elevated MCP-1 and TNF-α in placenta could lead to adverse pregnancy outcomes or even death in utero (170, 171). Another group reported severe CNS abnormalities and brain vasculature damage in newborn babies due to proinflammatory cytokines IL-8, IL-6, TNF-α, and IL-1β upregulated by CMV infection of pericytes (172).

HCMV does not infect animals due to the species specificity of beta herpesviruses and interestingly the virus has co-evolved with its host species (173). Therefore, the study of specific CMV in their respective species of animal models has been helpful in elucidating CMV specific immunity. Indeed, simian CMV seroprevalence was reported in baboons, African green monkeys, and rhesus macaques as early as 1971 (174) and currently, rhesus CMV (RhCMV) infections in rhesus macaques is more commonly used as a NHP model (175). Since the global prevalence of HCMV ranges from 60 to 100%, animal models offer a unique advantage of being specific pathogen free, in this case CMV-free, in order to understand CMV immunity in comparison to uninfected population. RhCMV is particularly useful to model congenital infections (176) and co-infections such as CMV and HIV infections in the same host (177). Intrauterine inoculation of pregnant dams and intraamniotic/intracranial inoculations of the fetuses with RhCMV led to severe neurological defects and CNS lesion similar to HCMV (45, 176, 178). Further, RhCMV studies helped identify that the primate CMV encodes and expresses IL-10 homolog genes in vivo (179). Interestingly, the viral homolog had evolved functions that are beneficial to viral replication, primarily through immunosuppressive and antiproliferative effects on host immune cells (179). The CMV IL-10 could also play a role in CMV's ability to subvert NK cell reactivity, thus avoiding NK cell lysis (179). Further, exploration of RhCMV infections in CMV free animals can identify immunopathogenesis pathways and therapeutic targets.

#### IMMUNOTHERAPEUTIC APPROACHES

Recombinant cytokines and anti-cytokine antibodies have recently gained traction in the pharmaceutical arena as a novel class of drugs for therapeutic purposes especially in autoimmune disorders and cancer (180, 181). There are few cytokine therapies that are already in use for therapy against viral infections such as IFN-α for HBV and HCV therapy. To overcome the severe side effects of IFN-α therapy, recently type III IFNs namely IFN-λ which have similar biological functions as IFN-α, have been tested preclinically in rhesus macaques (182). IFN-λ demonstrated antiviral effects similar to IFN-α without hematologic toxicity and thus could be used as an alternative therapy in chronic hepatitis patients. IL-12 administration has been previously studied in chimpanzees and rhesus macaques for understanding IL-12 mediated pathways and antiviral protection in SIV infections respectively (183, 184). IL-15 agonist, which has immunomodulatory functions, activates innate and adaptive immunity, and has been well characterized in NHP (185–188). Recently, a novel IL-15 superagonist ALT 803 potentiated T cell and NK cell responses leading to transient viral suppression in ART naïve SIV infected rhesus macaques (189). While the viral suppression was transient, this study illustrates IL-15 as a potential therapeutic agent particularly in combination therapy and ALT 803 is already in clinical trials for cancer therapy (190, 191). Even in DNA vaccine studies, IL-2 administration augmented vaccine elicited HIV-1, and SIV-1 specific immune responses in SHIV challenged rhesus macaques (192) thus showing that cytokine co-administrations can potentiate both vaccines and therapeutics.

Blocking of cytokine receptors or administration of cytokine antagonists can also be helpful in control of viral replication. Antagonists of CCR5 (maraviroc and vicriviroc) and CXCR4 inhibitor (Plerixafor) are relevant as they block HIV entry in cells and therefore can be used for HIV treatment (193). In addition to these small molecule CCR5 inhibitors, CCR5 blocking antibodies have also been characterized in preclinical rhesus macaques model of SIV infection (194–196). Further, maraviroc prevented cardiac dysfunction and cardiomyopathy associated with AIDS by blocking CCL5 and its recruitment of inflammatory macrophages in the heart tissue of SIV infected rhesus macaques (197).

Cytokine-based therapeutics are increasingly tested for other non-viral disease models of NHP. IL-13 neutralization for prevention of IgE mediated allergic responses in airway inflammation model of cynomolgus macaques (198), IL-6 receptor blocking and anti-TNF agent, infliximab for treatment of rheumatoid arthritis in cynomolgus macaques and rhesus macaques, IFN-α treatment effects in rhesus macaques model of cytokine induced depression (199, 200) are some of the few examples and could have potential applications in viral immunity and therapy. While cytokine therapy is advantageous in controlling viral replication or preventing tissue damage, systemic administration of cytokine, or cytokine blocking can result in altered hematopoiesis and immune activation, and

severe complications due to the pleiotropic nature of cytokines in long-term therapy. Even in co-inhibitor receptors/checkpoint blockade therapy such as anti-PD-1 or CTLA-4 therapy commonly used for reversion of exhausted T cells in cancer and chronic diseases, undue immune activation or autoimmune responses is a primary risk leading to systemic or organ toxicities associated with uncontrolled inflammatory cytokine secretion and cytotoxicity by activated immune cells, which in turn require additional or follow-up immunosuppressive treatment [reviewed in (201, 202)]. Therefore, development of site directed biologics or cytokine therapy targeting viral infected tissues would be more beneficial than systemic administration.

#### CONCLUSION

Within the last few years, cytokines have been identified as key diagnostic, prognostic, and therapeutic agents in human diseases. Their multifaceted roles in immunity, tissue protection, and remodeling, maintenance of systemic and metabolic homeostasis make them important biomarkers for understanding and treating infectious diseases, cancer, auto-immune diseases, metabolic dysfunctions and other inflammatory processes.

#### REFERENCES


However, it is very important that their use in conjunction with other therapeutic and preventative strategies needs to be tested in pre-clinical models due to their propensity to cause immunopathology and tissue injury leading to serious complications in certain conditions (**Figure 2**). The usage of NHP models will be helpful for early prevention of tissue injury and associated autoimmune and metabolic syndromes that arise in diseases caused by viral and non-viral causes.

#### AUTHOR CONTRIBUTIONS

CM and SVS performed most of the writing. OL and DRR contributed to writing of specific sections. RKR oversaw overall preparation of the manuscript, contributed to writing, and edited the final version of the manuscript.

#### FUNDING

This work was supported by National Institutes of Health (NIH) grants R01 DE026327, R01 DE026014, and R01 AI120828 to RKR. DRR was supported, in part, by T32 AI007387. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


for the detection of Interferon-gamma and Interleukin-2 in old and new world non-human primates. J Immunol. Methods (2017) 441:39–48. doi: 10.1016/j.jim.2016.11.011


in African green monkeys treated or not with interferon-alpha during primary SIVagm infection. PLoS Pathog. (2014) 10:e1004241. doi: 10.1371/journal.ppat.1004241


damage to the central nervous system in fatal Zika virus microcephaly cases. Cytokine (2018) 111:255–64. doi: 10.1016/j.cyto.2018.08.008


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

Copyright © 2018 Manickam, Shah, Lucar, Ram and Reeves. 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.