Endophilin A2 deficiency protects rodents from autoimmune arthritis by modulating T cell activation

The introduction of the CTLA-4 recombinant fusion protein has demonstrated therapeutic effects by selectively modulating T-cell activation in rheumatoid arthritis. Here we show, using a forward genetic approach, that a mutation in the SH3gl1 gene encoding the endocytic protein Endophilin A2 is associated with the development of arthritis in rodents. Defective expression of SH3gl1 affects T cell effector functions and alters the activation threshold of autoreactive T cells, thereby leading to complete protection from chronic autoimmune inflammatory disease in both mice and rats. We further show that SH3GL1regulates human T cell signaling and T cell receptor internalization, and its expression is upregulated in rheumatoid arthritis patients. Collectively our data identify SH3GL1 as a key regulator of T cell activation, and as a potential target for treatment of autoimmune diseases.

Aberrant activation of autoreactive T cells has been suggested to initiate and drive autoimmune diseases, such as rheumatoid arthritis (RA) and multiple sclerosis (MS). It is believed that RA is initiated years before the clinical onset through an activation of major histocompatibility complex (MHC) class II-restricted autoreactive T cells that give specific help to B cells, leading to production of autoantibodies1. In RA, the chronic inflammatory attack of the joints, initiated by the adaptive immune system, is followed by an increased production of inflammatory cytokines such as TNF and IL-6 and during the last decades effective pharmaceutical treatments neutralizing this late inflammatory phase have been developed. Although early treatment has been shown to be more effective, as it improves the possibility to interfere with the underlying cause of autoimmune diseases, patients receive treatment well after disease is established2. Thus, a better understanding of the role of the adaptive immune system that drives the early stages of the disease i.e. before the disease is clinically overt, is needed to improve treatment of autoimmune diseases.

Genetic predisposition of RA is known to be strongly associated with MHC class II genes implicating a major contribution of T cells. Additionally, a recent genome wide meta-analysis revealed over 100 non-MHC risk loci where many were suggested to be related to T cell immune functions3. T cells can be found in abundance in the synovium of RA patients4and the successful use of T cell targeting therapy, such as the CTLA-4-Ig fusion antibody abatacept, confirm the importance of T cells in the chronic stage of the disease. These observations are corroborated in animal models for RA, where depletion of T cells by use of antibodies or gene deletion protects against disease induction5,6. A specific role of the T cell receptor (TCR) in regulating development of arthritis has been further illustrated in the SKG mouse7, where a mutation in TCR signaling molecule Zap70 leads to altered TCR signaling. This in turn skews T cell development in the thymus increasing the number of autoreactive T cells leading to subsequent development of arthritis.

Herein, we have identified Endophilin A2 (EA2), encoded by the SH3gl1 gene, as a regulator of TCR internalization, signaling and downstream T cell effector functions. So far, EA2 has predominantly been studied in synaptic transmission in the central nervous system and in cancer8,9,10. We demonstrate that deficient expression of SH3gl1, caused by either a spontaneous mutation in the DA rat, or in genetically modified mice, leads to protection against autoimmunity. We hereby report that EA2 has a fundamental role in autoimmunity and limits the induction of autoreactive T cells. The discovery of the EA2’s impact on T cell activation opens up to explore new pathways and treatment possibilities for not only RA but for all T cell dependent inflammatory diseases.

A spontaneous mutation in the SH3g11 gene protects against autoimmune arthritis

The DA rat is commonly used in autoimmunity research due to its high susceptibility to a number of chronic inflammatory disorders such as arthritis and experimental autoimmune encephalomyelitis11. We noticed an increased variability in susceptibility to arthritis in our inbred DA rat colony. Arthritis-resistant rats were selected and bred to establish a new line (denoted DA mutated (DAMut)), which was found to be completely protected against pristane-induced arthritis (PIA; Fig. 1a, b). We suspected the underlying cause to be of a genetic origin and not environmental. To test this hypothesis, the DAMut colony was re-established by cesarean sectioning into a SPF (FELASA II)- controlled facility. We injected the rats with pristane and followed arthritis development. The resulting clinical scores were similar to the conventional facility ruling out obvious environmental factors (Supplementary Fig. 1a, b). To genetically position the mutation, we crossed DAMut rats to the genetically different, arthritis susceptible, E3.DA-Pia457 rat strain12 and F2 offspring from this cross were immunized to induce PIA. Subsequent linkage analysis disclosed a significant association with arthritis incidence for a polymorphic marker on chromosome 9. Typing with additional markers lead to the identification of a quantitative trait locus (Pia43), at the telomeric end of chromosome 9 (Fig. 1c, d). The Pia43 locus had not been identified in previous E3xDA crosses13,14, confirming that this locus was unique to the DAMut line. The fragment was introgressed into a congenic strain DAMut.E3-Pia43 and minimized to 2 Mb by new recombinations (Fig. 1e, f).

To identify the specific genetic alteration, we sequenced the DAMut rat genome and aligned it to the BN rat genome reference (Rno5). All variants in the DAMut.E3-Pia43 congenic region were manually compared to two previously sequenced DA genomes15,16. No single nucleotide polymorphism (SNP) or short insertion nor deletion (indel) was detected that could distinguish DAMut from the other DA strains in the congenic region, e.g., all 337 SNPs detected between DAMut and the reference sequence were also found in the other DA genomes. However, a structural variant was revealed, which was unique to the DAMut rat, and appeared to be the result of an insertion of a long terminal repeat (LTR) element of the ERV class I (ERV1) in intron 1 of the SH3gl1 gene (Fig. 2a).

Retrotransposons such as LTR elements have been shown to regulate gene expression in both mice and man17. To investigate if this was also the case for the ERV1 insertion in the SH3gl1 gene in DAMut rats, we determined the levels of lysine 4 methylation in histone 3 (H3K4me3) and the acetylation of histone 4 (H4Ac) using chromatin-immunoprecipitation (ChIP) and qPCR, as these modifications have been associated with active gene transcription18,19. We observed that H3K4me3 and H4ac levels upstream of the inserted LTR were similar in DA and DAMut rats. In contrast, levels in regions situated downstream of the insertion were down-regulated (Fig. 2B, C), indicating that SH3gl1 gene transcription was not active in DAMut rats. To determine if this had any impact on gene expression we analysed peripheral blood monocytic cells (PBMCs) from DA and DAMut rats and quantified gene expression with qPCR. Strikingly, the DAMut rats had almost no expression of the SH3gl1gene compared to DA rats (Fig. 2d). This vastly reduced gene expression also translated to a reduced level of the SH3gl1 encoded protein EA2 as determined by western blotting (Fig. 2e). To confirm that the arthritis resistance observed in the DAMut rats was due to a deficiency in EA2 expression we introgressed a deletion of the SH3gl1 gene into arthritis-susceptible B6N.Q mice through backcrossing and evaluated them with collagen-induced arthritis (CIA). Similar to the DAMut rats, the SH3gl1 deficient mice were also protected from arthritis, in contrast to their SH3gl1 sufficient wild-type littermates (Fig. 2f). To investigate whether the effect was restricted to control only arthritis development or also other autoimmune diseases, we tested experimental autoimmune encephalomyelitis (EAE), a T-cell-dependent model of MS, which confirmed the EA2 mediated protective effect seen in arthritis (Fig. 2g). Additionally, we tested the T-cell independent collagen-antibody induced arthritis model20and observed no difference in disease severity or induction (Fig. 2h, i). We conclude that EA2 is a major regulator of T cell dependent autoimmune disease.

EA2 deficiency alters the induction threshold of autoreactive T cells

Since EA2 deficiency had such a great impact on T cell dependent autoimmune disorders we investigated if the expression of SH3gl1 increased in T cells after arthritis induction. Indeed 8 days after in vivo activation the expression of SH3gl1 had more than doubled (Fig. 3a) suggesting an important function of SH3gl1 during T cell priming and activation. To further elucidate if the protection in SH3gl1 deficient rats is mediated by T cells we used the CD4+αβ+ T cell dependent pristane-induced adoptive transfer model21. Pristane-primed lymph node cells from DA or DAMut rats were expanded for T cells ex vivo before transfer into naïve DA recipients. Only cells from DA rat donors induced severe arthritis (Fig. 3b), indicating that the DAMut rats were not able to generate arthritogenic T cells. To confirm that the arthritis resistance was intrinsic to T cells, we transferred thymocytes from SH3gl1 deficient or wild-type littermate mice into TCRβ knockout mice before the induction of glucose-6-phosphate isomerase (GPI) induced arthritis. Mice receiving SH3gl1 deficient thymocytes developed nearly no arthritis (Fig. 3c) and had fewer antigen-specific cells compared to the wild-type recipients as determined by an in vitro recall assay using the GPI peptide while no significant difference could be seen in mitogen activated cells (Fig. 3d). To determine if this reduction in antigen-specific cells was due to a defect in the thymic T cell development, we assessed thymic T cell populations by flow cytometry. No difference in T cell populations could be observed between the SH3gl1 deficient mice compared to their wild-type littermates indicating that the reduced numbers of antigen-specific cells were not due to a defect in thymic T cell development (Fig. 3e) but were instead induced in the periphery. To better understand why the SH3gl1 deficient T cells do not become arthritogenic we looked at activation markers and induction of anergy. While no significant differences in activation markers could be seen ten days after GPI-peptide immunization, the SH3gl1 deficient T cells significantly upregulated anergy markers like PD-1 and CD73/FR4 (Fig. 3f). In a subsequent experiment comparing responses to GPI protein and ovalbumin, the same trend was observed for GPI but not for the exogenous antigen ovalbumin (Supplementary Fig. 2). We next investigated if the SH3gl1 deficient T cells had a regulatory or suppressive function in the pristane-induced T cell transfer model. Naïve DA rats were given either pristane-primed DA cells alone or together with an equal number of DAMut cells, with DAMut cells alone serving as a negative control. No reduction in arthritis severity was evident in recipients receiving pristane- primed DA cells together with DAMut cells, indicating that DAMut T cells do not exert their regulatory function via active suppression (Fig. 3g, h) but rather via inhibited activation.

EA2 deficient rodents develop normally and mount a normal response to cancer and infection

EA2 is a ubiquitous protein with high expression in the central nervous system22. Thus, EA2 deficiency could potentially affect many cell types and impact development. However, we did not observe any major physiological disturbances caused by EA2 deficiency in DAMut rats by general measurements such as blood cell counts, body weight, breeding capacity and life span (up to 1.5 years) in which the DAMut rats were identical to their normal DA littermates (Supplementary Fig. 3a–e). Similarly to the DAMut rats, the SH3gl1 deficient mice develop normally as compared to their wild-type littermates8. Given that the EA2 deficient rodents exhibited a profound protection against autoimmune diseases, we assumed that they could have a severely defective immune system. Our animal facility is specific pathogen free but had an outbreak of S. aureus caused infections which provoked septic arthritis in immunodeficient mice23. However, no SH3gl1 deficient animals developed a clinical S. aureus infection during this outbreak. To further challenge the SH3gl1 deficient mice towards bacterial infections, an arthritogenic S. aureus LS-1 strain24 was inoculated intravenously. This septic arthritis model using S. aureus LS-1 strain has been shown to be dependent on T cells via activation of T cells by the superantigen, toxic shock syndrome toxin-1 (TSST-1)25. In this model of septic arthritis, we did not observe any difference in bacterial clearance or development of arthritis in SH3gl1 deficient mice compared to wild-type littermates (Supplementary Fig. 3f, g). To further challenge the immune system in EA2 deficient mice, we injected SH3gl1 deficient and wild-type littermate mice with B16F10 melanoma cells and followed tumor growth. No excessive tumor growth was observed in EA2 deficient mice compared to wild-type littermates (Supplementary Fig. 3h). Thus, EA2 deficiency seems to affect regulation of autoimmunity but does not affect the general health of the rodents nor their susceptibility to a bacterial infection caused by S. aureus nor an excess growth of cancer cells.

EA2 deficiency leads to reduced T cell activation via the TCR

Considering the endocytic function of EA2 and the profound effect that loss of EA2 expression had on T cells specifically, we reasoned that EA2 might be important for the internalization of the TCR. Thus, we checked whether EA2 co-localized with the TCR upon activation using a proximity-ligase assay. While no co-localization could be found in un-stimulated cells, we could see that EA2 and the TCR co-localized after only three minutes of anti-CD3/CD28 stimulation (Fig. 4a and Supplementary Fig. 4). We next studied the kinetics of the TCR internalization and indeed, SH3gl1 deficient T cells internalized their TCRs at a slower rate compared to SH3gl1 sufficient T cells (Fig. 4b) with a reduced internalization seen already at 15 min after stimulation. The observed decreased rate of TCR internalization was not due to an increased recycling of TCR as differences in TCR internalization between the SH3gl1 deficient and sufficient T cells sustain even after Brefeldin A treatment (Fig. 4c). To investigate if this had any impact on the responsiveness of the T cells we investigated downstream TCR signaling molecules in SH3gl1 deficient and wild-type T cells. A reduced TCR signaling cascade was observed in SH3gl1 deficient T cells compared to wild-type T cells, with reduced levels of phosphorylated Zap70 and activation of ERK1/2 (Fig. 4d–f). This reduction in responsiveness in the SH3gl1 deficient T cells translated to reduced T cell proliferation following in vitro stimulation via CD3/CD28 (Fig. 4g, h). The observed reduced proliferation was not the result of reduced internalization of the IL-2 receptor as previously described26 but must stem from a reduced activation via the TCR (Supplementary Fig. 5).

SH3GL1 regulates TCR responses in human T cells and its expression is upregulated in RA patients

Since EA2 is highly conserved and had such a profound effect on T-cell-mediated arthritogenicity in both mice and rats, we next investigated to what extent SH3GL1 is of importance in human T cells. To demonstrate the relevance of EA2 in human T cells we made SH3GL1 CRISPR knock-out Jurkat T cells (Supplementary Fig. 6) and stimulated them in vitro with anti-CD3/CD28. Both TCR signaling and TCR internalization were reduced in SH3GL1deficient Jurkat cells compared to wild-type control and normal Jurkat T cells (Fig. 5a, b). We addressed the possibility that SH3GL1 could be overexpressed in T cells from RA patients and determined the expression of both SH3GL1 and the TCR molecule CD3e from whole blood of RA patients and healthy controls and found that the SH3GL1 expression correlated with expression of the CD3E molecule (Fig. 5c). Similar to the findings in animal models, the levels of SH3GL1 was higher in RA patients when normalized for CD3E (Fig. 5d) corroborating the data from the NCBI GEO database27 where SH3gl1 gene expression was found to be upregulated in sorted CD4 + T cells from RA patients compared to healthy control (Fig. 5e). Thus, SH3GL1 has a conserved mechanistic function in T cells across species and is relevant in a human disease setting.

Aberrant activation of autoreactive T cells is the key for the self-perpetuating vicious circle of activation of both the innate and adaptive immune response that leads to chronicity and lack of resolution in autoimmune diseases. Using a rat animal model of RA, we discovered SH3GL1 as a major regulator of T cell effector function and autoimmune diseases. We show that deficiency of SH3gl1 leads to a complete protection against autoimmune diseases in both mutated DA rats and in SH3gl1 knockout mice and that this results from loss of T cell effector functions. Furthermore, we show that SH3gl1 expression was increased in T cells during autoimmune arthritis in both rodents and RA patients.

TCR internalization and surface recycling following peptide-MHC recognition on the antigen presenting cell plays a pivotal role in establishing a stable immunological synapse and subsequent T cell activation28. To limit potential induction of autoreactive responses, T cell activation is tightly regulated via the TCR and co-stimulatory molecules. Both the surface expression of the TCR as well as the adapters in the TCR signaling cascade are of importance in dictating the effector functions of the T cells. For example, CD4+ T cells deficient in the WASH protein show reduced trafficking of the TCR and proliferation of T cells. Similar to SH3gl1 deficient animals, CD4+ conditional WASH knockout mice are also protected from EAE29. Several earlier studies demonstrate that SH3GL1 is necessary for internalization and trafficking of a number of receptors26,30,31. Our data demonstrate that SH3GL1 also regulates the internalization of the TCR. Because of the importance of TCR internalization and formation of the immunological synapse for efficient TCR signaling, a lack of SH3GL1 leads to reduced T cell signaling and would subsequently result in reduced numbers of autoreactive T cells, in turn limiting the arthritis development. The importance of adapters in the TCR signaling cascade was recently demonstrated by targeting the adapter protein NCK32. Like in SH3gl1 deficient animals, inhibition of NCK reduce proliferation of the T cells and leads to induction of immunosuppression and protection against EAE. Although the regulation of the TCR is different between SH3GL1 and NCK the resulting outcome seems to be similar. Development of effective tissue restricted immunotherapies in the field of autoimmune diseases has proven difficult, with a variety of antigen-specific vaccination strategies developed in the 1990s with limited success33,34. Consequently, instead of attempting to address antigen specificity, the field moved more into the depletion of molecules (e.g. TNF blockade35) or whole cell populations (e.g. B cells using anti-CD20 antibody36) involved in the pathogenesis. Although these therapies have been beneficial they are not curative and there is still a substantial number of patients who are non-responders or develop severe side-effects such as increased susceptibility to infections37,38. Our results with SH3gl1 in rodents demonstrate an interesting pathway by which activation of autoreactive T cells is modulated via the TCR while retaining the ability to respond to pathogens. Thus, SH3GL1 targeting could be a promising therapeutic treatment strategy for RA as well as in other T cell-mediated diseases.