Latent gammaherpesvirus exacerbates arthritis through modification of age-associated B cells

Epstein-Barr virus (EBV) infection is associated with rheumatoid arthritis (RA) in adults, though the nature of the relationship remains unknown. Herein, we have examined the contribution of viral infection to the severity of arthritis in mice. We have provided the first evidence that latent gammaherpesvirus infection enhances clinical arthritis, modeling EBV’s role in RA. Mice latently infected with a murine analog of EBV, gammaherpesvirus 68 (γHV68), develop more severe collagen-induced arthritis and a Th1-skewed immune profile reminiscent of human disease. We demonstrate that disease enhancement requires viral latency and is not due to active virus stimulation of the immune response. Age-associated B cells (ABCs) are associated with several human autoimmune diseases, including arthritis, though their contribution to disease is not well understood. Using ABC knockout mice, we have provided the first evidence that ABCs are mechanistically required for viral enhancement of disease, thereby establishing that ABCs are impacted by latent gammaherpesvirus infection and provoke arthritis.


Introduction
Rheumatoid arthritis (RA) is one of the most common autoimmune diseases in adults, though the etiology and pathophysiology are not fully understood (Humphreys et al., 2013;Firestein and McInnes, 2017). RA, as well as other autoimmune diseases including multiple sclerosis (MS) and systemic lupus erythematosus (SLE), is associated with Epstein-Barr virus (EBV) infection (Balandraud et al., 2003;Ascherio and Munger, 2007;Gross et al., 2005). EBV infection typically takes place during childhood or adolescence, while RA generally becomes symptomatic during middle age, indicating that the latent EBV infection likely modulates the immune system over time in a manner that contributes to the development of RA (Humphreys et al., 2013;Dowd et al., 2013;Fourcade et al., 2017;Alamanos et al., 2006). The circulating EBV load is higher in individuals with RA than in otherwise healthy adults (Balandraud et al., 2003), and RA patients have increased levels of antibodies specific to multiple EBV-encoded proteins (Blaschke et al., 2000;Ferrell et al., 1981;Catalano et al., 1979;Alspaugh et al., 1981;Hazelton et al., 1987). Further, RA patients have increased EBV-specific CD8 + T cells (Lünemann et al., 2008), yet these cells have a reduced ability to kill EBV-infected B cells when compared to the same subset of EBV-specific CD8 + T cells from healthy controls (Takei et al., 2001). However, the precise role of EBV in RA pathogenesis remains unknown.
Evidence from in vivo models is scarce and previous studies have focused primarily on the direct relationship between EBV infection and damage to the joint capsule, with little attention given to systemic effects of EBV infection on immune modulation preceding and continuing throughout disease (Kuwana et al., 2011;LeBel et al., 2018). Mice with humanized immune systems, namely NOD/Shi-scid/IL-2Rg null mice reconstituted with CD34 + hematopoietic stem cells, that were infected with EBV went on to spontaneously develop erosive arthritis, suggesting a causative role of EBV in arthritis development (Kuwana et al., 2011). Related, a serum transfer-induced arthritis model was used to demonstrate that Ly6C high monocytes play a role in transporting murine gammaherpesvirus 68 (gHV68), an EBV homolog, to the synovium (LeBel et al., 2018). Our group has previously shown that latent gHV68 infection exacerbates experimental autoimmune encephalomyelitis (EAE) and leads to a disease that more closely resembles MS, with increased demyelination and infiltration of CD8 to cells of the central nervous system in gHV68-infected mice (Casiraghi et al., 2012). Critically, this enhancement was specific to gHV68; other viruses, including lymphocytic choriomeningitis virus (LCMV) and murine cytomegalovirus (MCMV), did not lead to enhancement of EAE. Additionally, enhancement took place without changes to autoantibody levels. An in vivo model that recapitulates the temporal and systemic immunological aspects of the relationship between EBV and RA is critical.
To examine the relationship between EBV and RA, we have adapted in vivo models of both.
gHV68 is a natural pathogen that is a well-established and widely-used murine model of EBV infection that shares an array of characteristics with human EBV infection, including latent persistence in B cells, viral reactivation from latency, a potent CD8 T cell response, and immune evasion tactics (Olivadoti et al., 2007;Wirtz et al., 2016). Type II collagen-induced arthritis (CIA) is a commonly used model of RA wherein mice are injected with type II collagen emulsified in an adjuvant.
Here, we chose to use C57BL/6J mice due to the extensive past characterization of gHV68 infection in C57BL/6 mice and the numerous knockout (KO) strains available on this background. Multiple strains of mice are susceptible to CIA, including C57BL/6 mice that, despite displaying a less severe disease course than other strains, generate a robust T cell response (Inglis et al., 2007;Campbell et al., 2000). In C57BL/6 mice, CIA follows a chronic disease course with a sustained T cell response, presence of anti-collagen IgG, and infiltration of inflammatory lymphocytes into the joint capsule (Inglis et al., 2007). EBV primary infection generally takes place in childhood or adolescence (Dowd et al., 2013;Fourcade et al., 2017), and RA can occur at any age, though the mean incidence is in the sixth decade of life (Myasoedova et al., 2010). Accordingly, we have infected immunologically and sexually mature 6-to 8-week-old C57BL/6J mice with gHV68 and have induced CIA when the mice were adults at 11-13 weeks old. Here we have shown that C57BL/6J eLife digest Rheumatoid arthritis is one of the most common autoimmune diseases, leaving patients in pain as their immune system mistakenly attacks the lining of their joints. The precise cause is unknown, but research suggests a link to the Epstein-Barr virus, the agent responsible for mononucleosis (also known as glandular fever). After infection and recovery, the virus remains in the body, lying dormant inside immune 'B cells' which are often responsible for autoimmune diseases.
Of particular interest are a sub-group known as 'age-associated B-cells', which are mostly cells left over from fighting past infections such as mononucleosis. Yet, the link between Epstein-Barr virus and rheumatoid arthritis remains hard to investigate because of the long gap between the two diseases: the virus mostly affects children and young people, while rheumatoid arthritis tends to develop in middle age. To investigate how exactly the two conditions are connected, Mouat et al. created a new animal model: they infected young mice with the murine equivalent of the Epstein-Barr virus, and then used a collagen injection to trigger rheumatoid arthritis-like disease once the animals were older.
Next, Mouat et al. monitored the paws of the mice, revealing that viral infection early in life worsened arthritis later on. These animals also had more age-associated B cells than normal, and the cells showed signs of participating in inflammation. On the other hand, early viral infection did not make arthritis worse in mice unable to produce age-associated B cells. Taken together, these results suggest that the immune cells are required to enhance the effect of the viral infection on rheumatoid arthritis. This new insight may help to refine current treatments that work by reducing the overall number of B cells. Ultimately, the animal model developed by Mouat et al. could be useful to identify better ways to diagnose, monitor and treat this debilitating disease. mice infected with latent gHV68 and induced for CIA develop a more severe clinical course and an altered immunological profile compared to uninfected CIA controls, with expanded CD8 + T cells and Th1 skewing. We have utilized gHV68 infection and CIA induction to investigate mechanism(s) by which EBV contributes to RA, in particular through the modulation of age-associated B cells (ABCs).
The role of B cells in the relationship between EBV and RA is intriguing because B cells contribute pathogenically to RA, and EBV infects B cells and alters the B cell profile (Marston et al., 2010;Hatton et al., 2014). ABCs are a subset of B cells that are of particular interest as they have been implicated in both autoimmunity and viral infection. When compared to healthy adults, the relative proportion and/or absolute circulating counts of ABCs are elevated in RA patients, a subset of individuals with MS, individuals with SLE, and a subset of people with common variable immune deficiency that displays autoimmune complications (Adlowitz et al., 2015;Rubtsov et al., 2011;Thorarinsdottir et al., 2019;Wang et al., 2019;Claes et al., 2016;Wang et al., 2018;Zhang et al., 2019;Rakhmanov et al., 2009). ABCs are required for disease development in mouse models of SLE (Rubtsova et al., 2017). Also, ABCs are increased during viral infections in mice and/ or humans including LCMV, gHV68, vaccinia, hepatitis C virus, HIV, and influenza (Rubtsova et al., 2013;Chang et al., 2017;Knox et al., 2017;Johnson et al., 2020). ABCs display an array of functional capacities, including the secretion of anti-viral or autoantibodies, initiation of germinal centers, antigen presentation to T cells, and secretion of cytokines (Rubtsova et al., 2013;Chang et al., 2017;Knox et al., 2017;Johnson et al., 2020). It is yet to be examined whether ABCs play a role in viral contribution to autoimmunity. We found that ABC KO mice are unable to develop the gHV68-exacerbation of CIA and therefore act as a mediator between viral infection and autoimmunity.

Latent gHV68 infection exacerbates the clinical course of CIA
The development of RA often occurs years after initial infection with EBV when the virus is latent. To mimic this temporal relationship, we infected mice with gHV68, waited 5 weeks for the lytic infection to clear and the virus to establish latency, and induced CIA. Clearance of the acute virus and establishment of latency have previously been shown by plaque assay on spleens collected 35 days postinfection (Casiraghi et al., 2012;Barton et al., 2014). Following CIA induction, mice were assessed three times per week for redness and swelling in the two hind paws (Figure 1-figure supplement 1A), which informed a clinical score for each mouse. We observed that CIA in latent gHV68-infected mice (herein referred to as gHV68-CIA mice) had a more severe clinical course than uninfected mice (herein referred to as CIA mice), as evidenced by consistently higher clinical scores and changes in paw heights throughout the clinical course ( Figure 1A-B, Figure 1-figure supplement 1B). gHV68-CIA mice also developed onset of disease symptoms an average of 7 days earlier than CIA mice, reached a higher score at endpoint, and displayed a higher cumulative score ( Figure 1C, Figure 1figure supplement 1C-D). In agreement with other research groups (Campbell et al., 2000), male and female mice displayed similar clinical scores during CIA, and we also did not observe a sex difference in gHV68-CIA mice (Figure 1-figure supplement 1E). As expected, latent gHV68-infected mice (without CIA) did not display any signs of disease ( Figure 1A-B). Titers of anti-type II collagen autoantibodies (total IgG, IgG1, and IgG2c) were elevated in sera from mice with CIA compared to naive mice without CIA, yet were similar in mice with CIA regardless of infection ( Figure 1-figure supplement 1F-H). Additionally, we found that inducing CIA in gHV68-infected mice did not impact viral load (Figure 1-figure supplement 1I), indicating that gHV68 is not reactivating. These findings are in line with our previous work showing that latent gHV68 infection enhances EAE without influencing autoantibody levels or reactivating gHV68 (Casiraghi et al., 2012). These data demonstrate that latent gHV68 infection leads to earlier onset and more severe CIA, though the exacerbation is not due to higher titers of autoantibodies against type II collagen or changes in abundance of particular immunoglobulin isotypes.
The profile of immune cells infiltrating the synovium is altered in gHV68-CIA To assess the types and relative proportions of immune cells infiltrating the joint synovium, synovial fluid cells were collected on day 56 post-CIA induction. Synovial cells were collected from the knee and ankle joints by flushing each joint with phosphate-buffered saline (PBS) and subsequently analyzing isolated cells by flow cytometry. Synovial cells were not collected from naive or gHV68-infected mice without CIA because we would not expect there to be sufficient infiltration of immune cells for analysis. The number of CD8 + T cells infiltrating the synovium during gHV68-CIA was increased compared to CIA (3.6-fold change), while there was no significant difference in the number of CD4 + T cells ( Figure 2A). Additionally, the CD8 + and CD4 + T cells in gHV68-CIA synovium displayed a significant increase in Tbet expression compared to those in CIA ( Figure 2B), indicating Th1 skewing.
As further evidence that infiltrated T cells are immunologically active, we used real-time quantitative PCR (RT-qPCR) to evaluate the expression of key T cell-derived cytokines Ifng and Il17. The relative expression of Ifng in synovium cells of gHV68-CIA mice compared to CIA mice was increased (129-fold change), while the relative expression of Il17a was trending down in infected mice ( Figure 2C; 3.8-fold change), though the sample size was low due to the difficulty of obtaining these samples. Together, these results indicate that IFNg-producing T cells were preferentially infiltrating the synovium in our model of gHV68-CIA, which is consistent with what was observed in the synovium of RA patients (Yamada et al., 2008). Our data also demonstrate a skewing toward cytotoxic CD8 + T cells in mice latently infected with gHV68 prior to CIA.
Latent gHV68 infection skews the T cell response toward a pathogenic profile during CIA To examine how latent gHV68 might contribute to CIA, we specifically examined the systemic T cell profile. It is known that latent gHV68 infection expands cytotoxic T cells and reduces Tregs (Casiraghi et al., 2015). Both cell types play a role in CIA with cytotoxic T cells being crucial  mediators of CIA while Tregs play a protective role (Tada et al., 1996;Morgan et al., 2003). We examined T cells in the spleen and inguinal lymph nodes (ILNs), a draining lymph node in which we observed a significant increase in overall abundance of immune cells during CIA (Figure 3-figure supplement 1A). gHV68-CIA mice displayed a decrease in relative proportion of FoxP3 + Tregs and an increase in relative proportion of CD8 + T cells in the spleen compared to control CIA mice ( Figure 3A-B, Figure 3-figure supplement 1A, B-C). This is similar to what was observed in people with RA, as activated CD8 + T cells were increased and Tregs were decreased in the circulation of RA patients compared to otherwise healthy people (Morita et al., 2016;Ramwadhdoebe et al., 2016). In the ILNs of gHV68-CIA mice, we observed a nonsignificant trend of decreased CD8 + and CD4 + T cell relative proportions, indicating potential T cell egress from the ILNs during disease, and found that the proportion of regulatory T cells was unchanged between CIA and gHV68-CIA mice ( Figure 3A-B, Figure 3-figure supplement 1E-G). We also observed a significant increase in relative proportion of CD11c + CD8 + dendritic cells (DCs) in gHV68-CIA mice compared to CIA (Figure 3figure supplement 1H-J). These data show that the T cell profile of gHV68-CIA mice is skewed pathogenically, with decreased Tregs and increased cytotoxic T cells.
T cell polarization is modulated in gHV68-CIA mice Although IL17 has been highly studied due to its predominance in animal models of arthritis, both IL17 and IFNg are involved in RA (Yamada et al., 2008;Ramwadhdoebe et al., 2016;   2009; Nistala et al., 2010). As expected from our previous work with gHV68-EAE, we found that in gHV68-CIA, greater numbers of splenic CD8 + and CD4 + T cells express IFNg compared to CIA alone ( Figure 3C-D). There is a maintenance of Th17 cells in the spleen, with a similar proportion of CD4 + T cells expressing IL17A in CIA and gHV68-CIA ( Figure 3D). In the ILNs, we observed a significant increase in Il17a by RT-qPCR ( Figure 3E) and a corresponding trend toward more IL17A-expressing CD4 + T cells. We propose that the combined Th1 and Th17 profile observed in gHV68-CIA is more reminiscent of what is observed in people with RA than in CIA without gHV68 infection.   Latency is required for the clinical and immunological gHV68exacerbation of CIA To examine the requirement of gHV68 latency, as opposed to residual effects from acute infection, for exacerbating CIA, we used a recombinant gHV68 strain that does not develop latency, ACRTA-g HV68. In ACRTA-gHV68, the genes responsible for latency were deleted and a lytic gene, RTA, was constitutively expressed, resulting in clearance of the acute virus by day 14 postinfection (Rickabaugh et al., 2004). We infected mice with ACRTA-gHV68, waited 35 days for clearance of the acute infection, and induced CIA. We found that ACRTA-gHV68-infected mice did not develop the CIA clinical enhancement that we observed in latently gHV68-infected mice, with the clinical course and day of onset resembling that of uninfected CIA mice ( Figure 4A-B). Furthermore, the immunological changes observed in gHV68-CIA mice, when compared to CIA mice, were absent in ACRTA-gHV68 CIA mice. The increase in relative proportion of CD8 + T cells in  Source data 1. Disease progression and immune profile of latency-free ACRTA-gHV68 CIA mice compared togHV68-CIA and CIA mice.
the spleen was less pronounced in ACRTA-gHV68 CIA compared to gHV68-CIA, while there was no change in relative proportion of CD4 + T cells ( Figure 4C-D). In ACRTA-gHV68 CIA mice, there was abolishment of the gHV68-induced upregulation of IFNg in CD8 + and CD4 + T cells, reflecting altered functional capacity and possibly specificity, and no change in IL17A expression by CD4 + T cells ( Figure 4E-G). The decrease in relative proportion of splenic Tregs and CD8 + infiltration into the synovial fluid observed in gHV68 CIA mice was not present in ACRTA-gHV68 CIA mice ( Figure 4H-I). Together, these data show that ACRTA-gHV68 CIA mice displayed a similar clinical and immunological profile to uninfected CIA mice. This demonstrates that the enhancement is not due to innate immune stimulation during the acute infection, but, rather, the latency phase of gHV68 infection is critical for the clinical and immunological exacerbation of CIA. The requirement of gHV68 latency mirrors the RA patient clinical course, wherein patients are infected with EBV years before the onset of disease.
Age-associated B cells are increased and display an inflammatory phenotype in gHV68-CIA . We found that CIA induction increased the proportion and total number of ABCs (CD19 + -CD11c + Tbet + ) in the spleen, and gHV68-CIA mice had further increased proportions of ABCs in the spleen compared to CIA ( Figure 5A-B). The proportion of ABCs in the ILNs was not significantly different between gHV68-CIA and CIA mice ( Figure 5-figure supplement 1C). The number of ABCs was substantially lower in the ILNs than in the spleen, concurring with other studies that found that ABCs primarily reside in the spleen (Johnson et al., 2020). During CIA and gHV68-CIA, we did not observe differences in the proportions of ABCs between male and female mice ( Figure 5-figure supplement 1B). We next examined the phenotypic characteristics and found that ABCs in the spleen were phenotypically distinct in gHV68-CIA compared to CIA. We examined a series of markers previously shown to be expressed by ABCs, including cytokines IL10, IFNg, and TNFa (Rubtsov et al., 2011;Hao et al., 2011;Russell Knode et al., 2017;Ratliff et al., 2013), an array of inhibitory receptors (Rubtsov et al., 2011;Wang et al., 2018;Knox et al., 2017), maturity and memory markers IgD, IgM, and CD27 (Rubtsov et al., 2011;Hao et al., 2011), and MHCII (Rubtsov et al., 2011;Knox et al., 2017;Aranburu et al., 2018). We found that fewer ABCs in the spleens of gHV68-CIA mice expressed IL10, while an increased proportion expressed IFNg ( Figure 5C), indicating that they are skewed toward a pathogenic Th1 phenotype. Further, fewer splenic ABCs in gHV68-CIA mice expressed inhibitory receptors CTLA4, PDL1, and PD1 ( Figure 5D-F), and thus ABCs in CIA displayed a more regulatory phenotype than those in gHV68-CIA mice. Additionally, the ABCs in gHV68-CIA mice displayed a more mature phenotype, with fewer IgD + IgM + naive B cells and increased MHCII expression, though the expression of memory marker CD27 was unchanged ( Figure 5G-I). These results indicate that ABCs in gHV68-CIA mice are more mature and may have increased antigen presentation capacities but are not primarily a memory subset. There were no differences in the expression of CD20, TNFa, CD95 (Fas), nor IDO expression ( Figure 5-figure supplement 1D-G). Collectively, these results indicate that ABCs in gHV68-CIA mice display a more pathogenic phenotype than those in CIA, with decreased expression of regulatory cytokine IL10 and inhibitory markers, and increased expression of IFNg.

Age-associated B cells are required for gHV68-exacerbation of CIA
To determine whether ABCs are a subset mediating the viral enhancement of CIA, we utilized ABC KO mice that harbor a B cell-specific Tbet deletion. The clinical course and immune profile of CIA and gHV68-CIA mice were compared in littermate controls of Tbx21 fl/fl Cd19 cre/+ (KO) and Tbx21 fl/fl Cd19 +/+ (Ctrl) mice ( Figure 6A). We observed that the clinical course was unchanged in CIA between Ctrl and KO mice, indicating that ABCs are not contributing to the disease course in CIA ( Figure 6B). Alternatively, when induced with CIA, gHV68-infected KO mice did not display the gHV68-exacerbated clinical course compared to gHV68-CIA Ctrl mice ( Figure 6C), indicating that ABCs are a pathogenic subset in gHV68-CIA. Without ABCs, gHV68-CIA mice did not display clinical exacerbation, but rather appeared similar to uninfected CIA mice in terms of disease severity and day of onset ( Figure 6B-D). We observed that the ablation of ABCs does not significantly alter the proportion of CD8, CD4, or Treg populations in the spleen during CIA or gHV68-CIA, nor the    expression of IFNg or IL17A (Figure 6-figure supplement 1). These results indicate that ABCs are a critical pathogenic population in gHV68-CIA though more work is needed to fully elucidate the mechanism by which ABCs are contributing to disease.

Discussion
In this report, we demonstrate that latent gHV68 exacerbates CIA clinically and immunologically, and Tbet + B cells, known as ABCs, are critical for this exacerbation. Investigation of the mechanism by which EBV contributes to RA has previously been challenging due to the lack of a murine model to  examine the systemic immune modulation caused by latent gammaherpesvirus infection and resulting influence on arthritis. Here, we show that infecting mice with latent gHV68 prior to CIA induction results in an immune course more similar to that of RA patients than CIA alone and is a suitable model for examining the contribution of EBV to RA. Elucidating how EBV infection contributes to the development of RA is critical to understanding the underlying pathophysiology of the disease. As EBV is associated with several autoimmune diseases, it is important to examine whether there are conserved mechanisms of contribution. The overlap in etiology and pathophysiology between these autoimmune diseases may help to explain the cross-efficacy of immunotherapies between MS and RA, including B cell-depletion therapies. Our lab has previously demonstrated that latent gHV68 infection enhances EAE, a common model of MS, clinically and immunologically (Casiraghi et al., 2012). In both the gHV68-CIA and gHV68-EAE models, we observed an increase in CD8 + T cells at the site of disease and increased expression of IFNg by cytotoxic and helper T cells. Latent gammaherpesvirus infection of mice clearly alters autoimmune disease onset and severity reminiscent of the strong association of latent EBV infection in RA patients. As such, these investigative models will serve to identify common mechanisms in which EBV contributes to multiple autoimmune diseases.
Due to EBV infection often taking place years before the onset of arthritis, we posit that latent EBV infection modulates the peripheral immune response in a manner that contributes to the development of RA. We suggest that latently EBV-infected B cells alter, either directly or indirectly, lymphocytes that go on to contribute to disease onset, likely through expanding and activating CD8 + T cells and skewing toward a Th1 response. CD11c + CD8 + DCs may play a role in priming the pathogenic CD8 + T cell response, as they have been shown to cross-present antigen (den Haan et al., 2000;Schulz and Reis e Sousa, 2002). By acting as a mediator between infected cells and pathogenic T cells, ABCs are likely critical moderators in driving the heightened Th1 immune response to latent viral infection.
Accumulating evidence shows that ABCs are expanded in multiple autoimmune diseases and function pathogenically in mouse models of lupus (Rubtsov et al., 2011;Thorarinsdottir et al., 2019;Wang et al., 2019;Claes et al., 2016;Wang et al., 2018;Zhang et al., 2019;Rakhmanov et al., 2009;Rubtsova et al., 2017). Precisely how ABCs are contributing to pathogenicity is unclear, and ABCs are known to display multiple functional capacities that could contribute to disease. In models of SLE, ABCs have been shown to secrete autoantibodies and compromise germinal center responses (Zhang et al., 2019). Additionally, ABCs function as excellent antigenpresenting cells (Rubtsov et al., 2015). In a model of SLE, the ablation of ABCs decreases activated CD4 + T cells and IFNg-CD8 + T cells (Rubtsova et al., 2017). How precisely ABCs alter the CD8 + T cell population, whether they are cross-presenting antigen or impacting the CD8 + T cells indirectly, warrants further investigation. Alternatively, ABCs have been shown to secrete regulatory IL10 (Rubtsov et al., 2011;Hao et al., 2011), suggesting that a portion of the ABC population, in some individuals or contexts, could function in a protective manner. Further characterization of the phenotype and functional capacities of ABCs in autoimmune patients may help to elucidate their functional role. RA patients who experience a relapse following B cell-depletion therapy are more likely to display a reconstitution profile with increased numbers of memory B cells (Leandro et al., 2006). Whether existing therapeutics, such as B cell-depletion therapies or other approved drugs for RA, such as Abatacept (CTLA4 Ig), impact the ABC repertoire remains unknown.
Further evaluation of the influence of viral infection on ABC pathogenicity is needed. It is intriguing that ABCs are pathogenic in a genetic model of SLE without the presence of a virus (Rubtsova et al., 2017), though we observe that latent gHV68 is necessary for the pathogenicity of ABCs in CIA. This discrepancy indicates that ABCs may be contributing to disease through various mechanisms or that different contexts can prime ABCs for pathogenicity. The role of ABCs in controlling viral infections is an ongoing topic of study, with multiple papers recently providing compelling evidence that ABCs are critical for an effective anti-influenza response (Johnson et al., 2020) and are required to control LCMV infection (Barnett et al., 2016), in part through their secretion of antiviral IgG2a. Additionally, the influence of aging on ABC population and on autoimmunity development and progression warrants further study.
In summary, we have developed an in vivo model of EBV's contribution to RA that recapitulates aspects of human disease. Further, we have examined the role of ABCs and found that they are critical mediators of the viral enhancement of arthritis.

Materials and methods
Mice Tbx21 fl/fl Cd19 cre/+ mice were generated by crossing Tbx21 fl/fl Cd19 cre/+ and Tbx21 fl/fl Cd19 +/+ mice. Tbx21 fl/fl and Cd19 cre/+ mice were provided by Dr. Pippa Marrack (Rubtsova et al., 2017). C57BL/ 6J mice were originally purchased from The Jackson Laboratory. All animals were bred and maintained in the animal facility at the University of British Columbia. All animal work was performed per regulations of the Canadian Council for Animal Care (Protocols A17-0105, A17-0184).

Induction of CIA
On day 35 post-gHV68 or -ACRTA-gHV68 infection, CIA was induced by injection of immunizationgrade, chick type II collagen emulsified in complete Freund's adjuvant (CFA; Chondrex, Inc) intradermally at the base of the tail, followed by a booster injection of the same emulsion on day 14, as adapted from Inglis et al., 2008. Each mouse received 0.1 mg chick type II collagen and 0.25 mg CFA at days 0 and 14.

Evaluation of CIA severity
Clinical signs of CIA were assessed and scored three times per week beginning at the day of CIA induction: 0 = no symptoms; 1 = slight swelling and/or erythema; 2 = pronounced swelling and erythema; and 3 = severe swelling, erythema, and ankylosis, as adapted from Brand et al., 2007. Hind paws were scored individually by a blinded scorer and added for a single score. Day of onset considered two consecutive scoring days of a score of at least 1. The thickness of each hind paw was measured using a digital caliper and the size was expressed as the average thickness of the two paws.

Tissue harvesting and processing
Mice were anesthetized with isoflurane and euthanized by cardiac puncture. Blood was collected by cardiac puncture into empty sterile tubes and placed on ice until processing, and mice were perfused with 20 ml sterile PBS to allow for synovial fluid harvesting without blood contamination. ILNs and spleen were extracted and placed into 2 ml sterile PBS and stored temporarily on ice until processing. Synovial fluid was collected by exposing the knee and ankle joints, removing the patellar ligament, and flushing each flexed ankle and knee joint with sterile RNase/DNase-free PBS (Invitrogen) using an 18-gauge needle, adapted from Barton et al., 2007;Futami et al., 2012. Using a 70-mm cell strainer and a 3-ml syringe insert, spleens and ILNs were each mashed through the cell strainer mesh and a single-cell suspension was prepared for each sample. Splenocytes were incubated in ACK lysing buffer for 10 min on ice to lyse red blood cells, and remaining cells were kept on ice until further use.

Anti-type II collagen antibody ELISA
The sera were isolated by centrifugation 2000 Â g for 10 min, aliquoted, and stored for up to 14 months at À80˚C prior to running the enzyme-linked immunosorbent assay (ELISA). Anti-type II collagen antibodies were quantified by standard indirect ELISA. Briefly, ELISA plates (NUNC; Thermo Fisher) were coated with 5 mg/ml ELISA-grade type II collagen (Chondrex, Inc) overnight at 4˚C, washed 4x with wash buffer (PBS, 0.05% Tween-20), blocked with 5% newborn calf serum (NBCS; Sigma-Aldrich) for 1 hr at 37˚C, incubated with serial dilutions (1:100 to 1:12800) of test sera diluted in blocking buffer for 2 hr at 37˚C, and washed 4x with wash buffer. Bound (anti-collagen II) antibody was incubated with HRP-conjugated goat anti-mouse IgG (Thermo Fisher), rat anti-mouse IgG1 (BD Biosciences), or goat anti-mouse IgG2c (Thermo Fisher), all diluted 1:500 in blocking buffer, for 1 hr at 37˚C, washed 4x with wash buffer, and detected by TMB substrate (BD Biosciences). Absorbance was read at 450 nm on a VarioSkan Plate Reader (Thermo Fisher).

Statistics
Data and statistical analyses were performed using GraphPad Prism software 8.4.2 (GraphPad Software Inc). Results are presented as mean ± SEM. Statistical tests, significance (p-value), sample size (n, number of mice per group), and number of experimental replicates are stated in figure legends. Statistical analyses included two-way ANOVA with Geisser-Greenhouse's correction, Mann-Whitney test, or one-way ANOVA. P-values are indicated by asterisks as follows: ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

Study approval
All work was approved by the Animal Care Committee (ACC) of the University of British Columbia (Protocols A17-0105, A17-0184).