Systemic lupus erythematosus (SLE) is an autoimmune disease involving multiple tissues and organ systems, including kidneys, joints, skin, blood cells, and brain (1). Mechanisms of disease, however, can vary depending on the system affected. Manifestations of SLE in the nervous system are termed neuropsychiatric lupus (NPSLE), which can affect roughly 20% to 40% of patients (2). Common presentations of NSPLE include depression, cognitive dysfunction, and anxiety (3, 4), but unfortunately, most treatments in current use are non-specific for the neuropsychiatric disease itself or only treat the symptoms and not the underlying disease. One major difficulty with developing an effective and targeted treatment is that the pathogenesis of NPSLE has yet to be fully elucidated.

The blood-brain barrier has traditionally been believed to be the site of entry for immune cells and inflammatory mediators that initiate NPSLE (5). Nevertheless, leukocyte accumulation and immune complex deposition at another key brain barrier, the blood-cerebrospinal fluid barrier (i.e. the choroid plexus), has been described in human NPSLE (6, 7), suggesting possible involvement of lymphocyte aggregation at the choroid plexus in the pathogenesis of lupus-associated brain disease. Moreover, we found in lupus mice that choroid plexus lymphocytes organize into tertiary lymphoid structures (TLS), lymphoid aggregates that develop in response to chronic, unresolved inflammation (8).

TLS functionally and structurally resemble secondary lymphoid organs but differ in that they form after development in non-lymphoid tissues and resolve with the inflammation (9). Though the initial events involved in TLS formation are still being investigated, it is clear that chemokines such as CXCL13 are overexpressed to recruit and retain immune cells after initiation (10). TLS then organize into B and T cell zones and can eventually develop germinal centers, where they can produce autoantibodies specific for the local tissue that can worsen disease and perpetuate inflammation (11). Additionally, the complexity and size of the TLS affect the local production of homeostatic cytokines and chemokines (such as CXCL13), which can continue to damage surrounding tissue (9).

Chemokine CXC ligand 13 (CXCL13; also, B lymphocyte chemoattractant (BLC)) is a chemokine important in lymphocyte recruitment and retention and lymphoid formation (12). This key chemokine is primarily produced by stromal cells and follicular dendritic cells and acts to recruit B cells and T follicular helper (Tfh) cells via its receptor, CXCR5. Indeed, the lack of CXCL13 or CXCR5 alone is enough to disrupt normal lymph node formation, while overexpressing CXCL13 is sufficient to induce lymphoid formation (13, 14). CXCL13 has also been shown to be important in SLE; systemic levels of CXCL13 positively correlate with disease activity scores in SLE patients, and blocking CXCL13 in lupus mice attenuated renal disease (15–17). These findings strongly suggest CXCL13’s involvement in both TLS formation and disease progression, but its role has not been studied in the context of neuropsychiatric involvement.

We used a spontaneous SLE mouse model, MRL/MpJ-Fas^lpr (MRL/lpr), which presents with prominent neuropsychiatric manifestations. This strain has a mutation of the FAS gene that disrupts normal Fas-mediated apoptosis and results in ineffective clearing of autoreactive lymphocytes, leading to nuclear autoantibody production, hypergammaglobulinemia, lymphadenopathy, skin, and kidney disease (18). MRL/lpr mice also exhibit neurobehavioral abnormalities reminiscent of human NPSLE, including depression-like behavior and cognitive abnormalities in the form of memory impairment (18, 19). In this study, we neutralized CXCL13 to block lymphocyte recruitment and organization to understand the contribution of this process to the neuropsychiatric manifestations seen in MRL/lpr mice. MRL/lpr were treated with an anti-CXCL13 antibody delivered systemically or locally (in the brain), and assessed for changes in the behavioral phenotype to determine the potential of blocking CXCL13 as a therapeutic modality for NPSLE.

Currently, the treatment of NPSLE is not standardized as there is little evidence that any single medication or combination therapy works for all cases of NPSLE (32). Focusing specifically on the inflammatory presentation of NPSLE, current treatment recommendations differ based on disease severity; however, they still only have limited to moderate evidence of benefit or commonly lead to post-treatment relapse (4, 31). This unmet need in the treatment of NPSLE emphasizes the importance of studies that explore novel aspects of pathogenesis and investigate new therapeutic modalities.

TLS can form in a variety of contexts in autoimmune diseases, and accumulating evidence suggests that TLS develop in cases of SLE and may greatly impact disease progression (33). Evidence for TLS formation has been found in the skin, vasculature, and kidneys in SLE animal models, as well as in the kidneys of lupus patients (34–37). Chang et al. reported that half of the renal biopsies from LN patients had well-organized infiltrates in the tubulointerstitial space, demonstrating germinal center formation, as well as B cell clonal expansion and somatic hypermutation, suggesting antibody production (37). The kidney TLS were strongly associated with local immune complexes found in the tubular basement membrane, suggesting that TLS contributed to disease progression. Multiple sclerosis (MS) is a specific case in which TLS have been found in a neuroimmunological autoimmune disease, with antibodies proposed to recognize myelin and other neuronal components (11). Well-organized lymphocyte infiltrates in MS correlate with severe inflammation, as evident by increased meningeal and perivascular inflammation with B cells, plasma cells, and T cells found in the sulci of the frontal, temporal, and parietal lobes (38). Gatumu et al. demonstrated that blocking LTβR, an important cell recruitment factor in TLS formation, partially restored function of the salivary glands in NOD mice, a model for Sjogren’s syndrome, by reducing the size and number of follicles that formed and inhibiting CCL19 expression (39). Finally, we previously identified TLS in the brains of MRL/lpr mice, and previous studies reported the presence of CD45+ cells and immune complex deposition in the brains of NPSLE patients (6, 7, 40). These data collectively highlighted the potential importance of TLS in SLE and other autoimmune conditions (including those affecting the CNS) and suggested that interfering with TLS formation can be beneficial, which encouraged us to investigate the contribution of TLS to NPSLE pathogenesis and progression.

CXCL13 is a key component in lymphoid organization, including in TLS formation. Several studies show that mice deficient in CXCL13 or its receptor, CXCR5 (expressed on follicular dendritic cells and follicular T helper cells, respectively), lack peripheral lymph nodes and have reduced B-cell homing to follicles in existing lymph nodes and the spleen (12, 13, 41). Moreover, CXCL13 alone has been shown to be sufficient to initiate the formation of lymphoid structures in peripheral organs. Luther et al. reported that overexpression of CXCL13 under a pancreatic insulin promoter in diabetic rats induced pancreatic TLS formation (14). Additionally, local expression of CXCL13 positively correlates with TLS size and complexity (9), further implicating not just the importance of CXCL13 in TLS formation but also its influence on lymphoid tissue development.

Previous studies have looked at the association of CXCL13 and SLE specifically, reporting a positive correlation between CXCL13 and disease activity scores (15, 16, 42). Interestingly, separate from its chemotactic function, a direct role for CXCL13 in the pathogenesis of LN was suggested by Worthmann et al., who found that exposing cultured podocytes to CXCL13 increased secretion of CXCL12, MIF, ICAM-1, and VCAM-1 (43). Looking at CSF of patients with neurological diseases, CXCL13 was the cytokine most closely linked to B cell recruitment to the CNS and intrathecal Ig synthesis (44). The latter study emphasizes the importance of CXCL13 in the context of the CNS and the recruitment of B cells that may be contributing to disease.

In this study, we look at the effect of neutralizing CXCL13 in the context of NPSLE. Specifically, we assessed how treatment altered brain disease manifestations when delivered systemically or intracerebroventricularly. Treated mice showed less memory impairment and depression-like behavior, suggesting that CXCL13 contributes to the development or maintenance of these behavioral deficits and may likely play an important role in NPSLE pathogenesis. ICV administration of anti-CXCL13, which restricted the treatment to the CNS, allowed us to uniquely determine that the observed attenuation of the behavioral manifestations was due to changes in the brain disease progression and not simply secondary to any undetected changes in systemic disease.

The histological scores and immunofluorescence staining with cell counts indicated little change in terms of cellular infiltration extent and its overall composition, despite the expectation that neutralizing CXCL13 would either prevent recruitment of lymphocytes to the site or disrupt retention of the cells. The immune system, however, has proven redundancies, and it is likely that another similar chemokine, such as CCL20, could have recruited the cells in the absence of CXCL13. Cellularly, the one parameter that was different was the number of AID-positive cells between the two antibody groups and the control group not receiving active therapy. B cells positive for AID indicate germinal center development and increase of antibody diversity (45), suggesting the infiltrates in the antibody groups had less germinal center activity. As intravenous immunoglobulin (IVIg) is one of the established treatments for lupus (46), it could be that delivering any antibody alone, regardless of specificity, may decrease antibody diversification.

The PCR array assessed for RNA expression levels of genes involved in processes such as neuronal apoptosis, inflammatory response, and cell adhesion. CCL3 (or MIP-1α) is a chemokine secreted by macrophages that was significantly down-regulated in the hippocampus of the IP cohort. A previous study by Marciniak et al. demonstrated that injecting excess CCL3 into the brain significantly impaired performance in a Y-maze test (47). This study suggests that CCL3 is relevant to memory-based tasks and that decreased CCL3 in the treated group could have contributed to the improvement we saw in the OP task. CCL5 (or RANTES), a molecule found to be upregulated in CSF levels of NPSLE patients compare to non-NPSLE patients (48), was also down-regulated in the hippocampus of mice treated IP. In cases of HIV infection, CCL5 correlated with increased viral load and was one of the strongest predictors of neurocognitive impairment based on a global deficit score (49). Levels of CCL5 are elevated in the serum of patients with depression (50), suggesting, along with the connection with cognitive impairment and NPSLE, that the decrease of CCL5 in the treatment group may have contributed to the improvement in both behavioral indicators we tested for in this study.

A number of JAK/STAT pathway genes were altered with anti-CXCL13 treatment as well, with the most notable being STAT3, which was decreased in both cortex and hippocampus of the IP cohort. STAT3 is a key transcription factor in the development of Th17 cells, which have been hypothesized to be potential inducers of TLS formation, and also are involved in pathways regulating cell proliferation, differentiation, and inflammation (51, 52). Previous studies in MRL/lpr mice inhibiting STAT3 saw decreased anti-dsDNA antibodies and nephritis, as well as decreased renal tubulointerstitial lesions (53–55). Another study in rats following chronic mild stress correlated increased levels of STAT3 with increased immobility in the Porsolt swim test, implying STAT3’s potential role in affective deficits (56). Slight-Webb et al. suggested that the reduction of phosphorylated STAT3 is the primary mechanism responsible for disease suppression in lupus for mycophenolate mofetil, a current therapy for NPSLE (57). Surprisingly, all tissues (disregarding the cortex of the ICV cohort) showed up-regulation of JAK1 or JAK2. Multiple studies showed that inhibiting Jak1 or Jak2 attenuated lupus-like disease in multiple mouse models (53), yet we saw an increase in the brains of the MRL/lpr strain. While the exact mechanistic changes are important to investigate and should be addressed in future studies, it is likely that unexpected changes in gene expression could have been the result of compensation by the immune system as the CXCL13 pathway was disrupted.

Another potential explanation of the neurobehavioral changes we saw in the treatment mice is that CXCL13 may be acting directly on brain cells (rather than mainly contributing to the TLS). CXCL13 is upregulated in the CNS in neuroinflammation by perivascular mononuclear cells and endogenous microglial in active MS plaques as well as by stromal cells in meningeal B cell aggregates (58). Additionally, animal studies have also not only demonstrated that CXCL13 may be produced by microglia with neuroinflammation but also that CXCR5 signaling in the brain can then lead to astrocyte activation (59) as well as affect the hippocampus to induce cognitive dysfunction (60). Furthermore, we previously described significantly increased CXCL13 and CXCR5 in the choroid plexus of MRL/lpr mice (8) and, in preliminary single-cell RNA-sequencing studies, determined that the cellular sources of these molecules are macrophages/microglia and stromal fibroblasts (for CXCL13) and B cells and double-negative T cells (for CXCR5) (data not shown). Choroid plexus tissue was however not obtained in our current study. While we did not find any significant change with treatment in CXCL13 or CXCR5 expression in the cortex and hippocampus (data not shown), this might be occurring in the choroid plexus, and/or might be detectable in the cortex and hippocampus by single cell rather than bulk analysis. In any case, the upregulated CXCL13/CXCR5 pathway in brain cells of MRL/lpr mice and the previous descriptions of enhanced brain CXCL13/CXCR5 expression in animal models provide a likely explanation for the beneficial effect of CXCL13 blockade in this study regardless of CNS leukocyte accumulation or TLS formation.

The current study demonstrated that treatment with an anti-CXCL13 neutralizing antibody attenuated behavioral deficits in lupus-prone mice and could be doing so on a molecular level by altering expression levels of certain inflammatory mediators or direct effect on resident brain cells. At least for the ICV cohort (e.g. Supplementary Figure 3B ) we identified relatively few differentially expressed genes to fully explain the effects of treatment. However, the approach we chose was a selective array containing less than 100 genes, in particular pathways, dictated by the commercial product. A more comprehensive and unbiased characterization by bulk RNA sequencing may very well have disclosed many more gene differences. Nevertheless, the latter approach, while potentially helpful to further elucidate the basis for the therapeutic benefit of the antibody, was outside the scope of the present study and can be applied in future research to more definitively address the mechanism of benefit of anti-CXCL13 treatment in NPSLE. Moreover, we studied gene expression in the cortex and hippocampus, and as mentioned above the benefit we demonstrated was possibly derived in part from inhibition of the pathway in additional brain regions as well, e.g. the choroid plexus.

While our study suggests a promising new treatment approach to treat NPSLE, there are several limitations to consider. For one, we did not see systemic differences among the groups, even though previous studies have demonstrated effectiveness in blocking CXCL13 on autoimmunity. In experimental autoimmune encephalomyelitis (EAE), a MS murine model, blocking CXCL13 with an antibody or knocking it out entirely still allowed onset of disease to occur normally, but severity was attenuated over time, as shown by lower mean clinical scores and less activated microglia (61). In murine collagen-induced arthritis, a model of rheumatoid arthritis, blocking CXCL13 reduced disease severity and decreased synovial germinal center formation (62). Differences with our study may stem from how CXCL13 was blocked. The EAE study used a knockout mouse model (61); since complete depletion of CXCL13 has been shown to affect normal lymph node development (13), these differences in development may be why we did not see a similar improvement in systemic disease. The study utilizing collagen-induced arthritis mice neutralized CXCL13 using an antibody similarly to our study, but they do not specify the dosage or frequency of treatment (62), making it challenging to use for comparison. However, in a study by Finch et al. that specifically looked at lymphoid follicle development, CXCL13 neutralization in adult mice had no obvious effects on germinal centers, follicular dendritic cell networks, or Peyer’s patch structure (63). Their results are consistent with those of our study, suggesting that CXCL13 may have a more important role in lymphoid tissue formation than in its maintenance.

Interestingly, we also saw no differences in renal function in either of the two cohorts with the different delivery methods. Though improving renal dysfunction was not the focus of the present study, the results in the IP cohort differ from those reported by Wu et al. who saw significantly decreased proteinuria with CXCL13 blockade in MRL/lpr mice (17). These differences may be explained by differences in study design. Wu’s group, though delivering less per injection IP than we did in the present study (50 µg vs. 100 µg, respectively), treated the mice every day while we only treated three times a week. Moreover, the Wu study was also limited to 6 mice per group, whereas we studied groups of 13-21 mice; perhaps, with a larger sample size, the significant differences in systemic and renal disease initially reported would not have been confirmed. Our results also differed from a study done by Moreth et al. that looked at renal disease in MRL/lpr mice (64). They found that blocking the proteoglycan biglycan improves both systemic and renal parameters, which they attributed to declines in CXCL13 levels due to biglycan’s role in its regulation. However, biglycan also influences a number of other important inflammatory proteins, including TLR2 and TLR4, making it difficult to definitively attribute the beneficial effects of treatment in this study to any effects on the CXCL13 pathway.

One caveat of the current study was in the PCR array experiment; while still providing much information on the gene expression differences, a panel specific to NPSLE ideally would have been used, but since it is still unclear exactly what processes are involved in its pathogenesis, one could not be created accurately. Additionally, the ICV anti-CXCL13 treated and IgG control groups’ expression levels in the cortex had very similar profiles; this suggests that the surgery to implant the cannula did impact inflammation in the brain, as related genes were significantly lower in the non-surgery group. The hippocampus of the ICV cohort did not mirror this likely because the cannula does not cross the hippocampus at the coordinates we had chosen for its insertion.

A necessary consideration of the translatability of an anti-CXCL13 treatment is the delivery method. We used two different delivery methods, IP and ICV, in this study with lupus-prone mice. Both methods were effective in improving behavioral manifestations, but each poses individual concerns when considering administration in the clinical setting. Delivering the treatment systemically in humans will certainly have many off-target effects in secondary lymphoid organs. Since many SLE patients are already on immunosuppressants, blocking CXCL13 would serve to further dampen homeostatic immune responses. Delivering the treatment in a way that isolates it to the CNS would require an invasive approach, and infections are of major concern, such that ICV administration has little translational application at this time. Nevertheless, our study is one of the few that demonstrate that a systemically administered antibody can improve neuropsychiatric manifestations in a lupus-prone animal model.

Since we saw less than a full reversal of the NPSLE phenotype with CXCL13 inhibition, for future studies it may be beneficial to explore different dosages and/or frequency of delivery and to also test the same treatment in other lupus-prone mouse models with neuropsychiatric presentations to see if the treatment is equally advantageous. Additionally, some evidence suggests many lupus therapies would be effective concomitantly and should be studied in combination rather than one therapy at a time. A study by Sharma et al. demonstrated the effectiveness of neutralizing CXCL13 in conjunction with the BAFF receptor in a Sjogren model, preventing loss of salivary flow while reducing IgM levels (65). This may be the best approach with consideration of potential redundancies in the immune system, as discussed above. Especially if the TLS are an important contributing factor to the neuropsychiatric manifestations, targeting them by inhibiting certain components, such as CXCL13, or blocking certain functions, is worth pursuing further to develop more effective therapies for lupus and NPSLE.

The raw data supporting the conclusions of this article will be made available upon request from qualified individuals, without undue reservation.

Animal studies were reviewed and approved by the Institutional Animal Care and Use Committee at Albert Einstein College of Medicine.

AS and CP conceived of the presented idea. MH designed the study with input from AS and CP. MH performed and analyzed all experiments. CP was in charge of overall direction and planning. MH and CP wrote the manuscript in consultation with all authors. All authors contributed to the article and approved the submitted version.

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: NIH Training Program in Cellular and Molecular Biology and Genetics, T32 GM007491 (to MH).

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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.