Capicua deficiency induces autoimmunity and promotes follicular helper T cell differentiation via derepression of ETV5

High-affinity antibody production through the germinal centre (GC) response is a pivotal process in adaptive immunity. Abnormal development of follicular helper T (TFH) cells can induce the GC response to self-antigens, subsequently leading to autoimmunity. Here we show the transcriptional repressor Capicua/CIC maintains peripheral immune tolerance by suppressing aberrant activation of adaptive immunity. CIC deficiency induces excessive development of TFH cells and GC responses in a T-cell-intrinsic manner. ETV5 expression is derepressed in Cic null TFH cells and knockdown of Etv5 suppresses the enhanced TFH cell differentiation in Cic-deficient CD4+ T cells, suggesting that Etv5 is a critical CIC target gene in TFH cell differentiation. Furthermore, we identify Maf as a downstream target of the CIC–ETV5 axis in this process. These data demonstrate that CIC maintains T-cell homeostasis and negatively regulates TFH cell development and autoimmunity.

T he germinal centre (GC) response is one of the most elegant processes in adaptive immunity and produce antibodies that have high affinity to antigens. In follicles, B cells that recognize antigens proliferate and form GCs. GCs are expanded by the proliferation of GC B cells and polarize into two microenvironments, the dark zone and the light zone [1][2][3] . GC B cells cycle between these two zones. In the dark zone, GC B cells rapidly proliferate and undergo somatic hypermutation, which enables antibody diversification and affinity maturation. In the light zone, GC B cells are selected on the basis of antigen affinity, undergo immunoglobulin class-switch recombination, and eventually give rise to antibody-secreting plasma cells or memory B cells [1][2][3] . During the GC response, several types of immune cell collaborate with B cells in the follicles, where follicular helper T (T FH ) cells have an instrumental function. T FH cells facilitate the selection and maturation of high-affinity GC B cells by multiple rounds of cognate interaction with B cells in the light zone; these interactions provide the selected B cells with crucial signals for survival and re-entry into the dark zone 3 . Because T FH cells have an important function in the generation of isotype-switched and affinity-maturated antibodies, dysregulation of T FH cell development and function is closely associated with immunodeficiency-related pathogenesis or antibody-mediated autoimmune diseases including systemic lupus erythematosus [4][5][6] .
Differentiation of T FH cells is initiated by the interaction of naïve T cells with dendritic cells (DCs), which, together with environmental factors, including cytokines, triggers expression of the chemokine receptor CXCR5 on DC-primed T cells 7,8 . The surface expression of CXCR5 enables T cells to migrate into B-cell follicles 7,8 . T cells that are targeted to enter B-cell follicles upregulate expression of the transcriptional repressor BCL6 and express an intermediate level of typical T FH molecules (for example, CXCR5, PD-1, ICOS and SAP) at the junction between T-cell and B-cell zone 9,10 . At this stage, developing T FH cells interact with cognate B cells and differentiate into GC T FH cells that express high levels of T FH molecules, such as PD-1 and CXCR5 (ref. 9). BCL6 as a master transcription factor for T FH cell differentiation [11][12][13] and BLIMP1 as an antagonist of BCL6 (ref. 11), plus several other transcription factors, help orchestrate T FH cell differentiation by exerting either a positive or negative effect, depending on the cellular context 8 . Among these factors, MAF (also known as c-MAF) was identified as a positive regulator of T FH cell differentiation in mice and humans. Maf deficiency decreases the frequency of CD4 þ CXCR5 þ T cells in mice 14 . MAF regulates expression of Il21 in mouse T FH cells and, in conjunction with BCL6, MAF induces expression of CXCR4, CXCR5, PD-1, ICOS and IL-21 in human T FH cells [14][15][16] . Moreover, MAF expression is induced in CD4 þ T cells by ICOS co-stimulation 14 or by IL-6, an important cytokine for initiating T FH cell differentiation 15 .
Capicua/CIC is a transcriptional repressor that is evolutionarily conserved from cnidarians to mammals 17 , and it exists in short (CIC-S) and long (CIC-L) isoforms 17 . In mammals, CIC interacts with Ataxin-1/ATXN1 (ref. 18), of which polyglutamine (polyQ)expanded form causes spinocerebellar ataxia type-1 (SCA1) neuropathogenesis, and its haploinsufficiency alleviates SCA1 progression 19 . Loss of the ATXN1-CIC complex results in hyperactivity, impaired learning and memory, and abnormal maturation and maintenance of upper-layer cortical neurons in mice 20 . CIC also suppresses the progression of several types of cancer [21][22][23] . CIC target genes that are critical for regulation of cancer progression include PEA3 group genes, ETV1/ER81, ETV4/PEA3 and ETV5/ERM, which are frequently overexpressed and promote tumourigenesis and metastasis in various types of cancer cell 21,[24][25][26] . In vivo functions of CIC have been reported in studies of Cic hypomorphic (Cic-L À / À ) mice 19 . These mice die within 4 weeks after birth and have defects in lung alveolarization 27 and bile acid homeostasis 28 . However, CIC is expressed in most mouse tissues 27,28 , so it may also have other important physiological functions. CIC levels are relatively high in the thymus 27,28 and in immune cells including T and NK cells (http://www.humanproteomemap.org) 29 . Moreover, our previous study shows that genes associated with immune activation and autoimmune diseases, such as SLE, asthma and thyroiditis, are significantly upregulated in the liver of Cic-L À / À mice 28 . These findings suggest that CIC may regulate immune responses and immune disorders.
Here we investigate CIC functions in immune system using various immune cell-specific Cic null mice. Our study finds previously unrecognised functions of CIC in regulation of T-cell activation and GC responses and suggests CIC as a key transcription factor in the suppression of autoimmunity.

Results
Autoimmunity in immune cell-specific Cic null mice. To investigate in vivo requirement of CIC in homeostasis and functions of immune cells, we generated haematopoietic lineage cell-specific Cic null mice (Cic f/f Vav1-Cre) by crossing mice carrying floxed Cic alleles (Cic f/f ) with Vav1-Cre mice ( Supplementary Fig. 1a), and characterized them at 9-12 weeks of age. Development of lymphocytes in primary lymphoid organs was largely normal ( Supplementary Fig. 1b,c), although an increase in proportion of thymic CD4 þ FOXP3 þ regulatory T (T reg ) cells was observed ( Supplementary Fig. 2).
Strikingly, all Cic f/f Vav1-Cre mice developed enlarged secondary lymphoid organs, including spleens (Fig. 1a), mainly due to proliferation of B220 þ B cells (Fig. 1b). In addition, the number of CD8 þ T cells was significantly decreased, whereas that of macrophages and DCs was increased in spleen of Cic f/f Vav1-Cre mice (Fig. 1b). The expression levels of several T-cell co-stimulatory ligands, such as CD80, CD86, CD40 and ICOSL, on the surface of DCs were comparable between Cic f/f (WT) and Cic f/f Vav1-Cre mice ( Supplementary Fig. 3).
Cic f/f Vav1-Cre mice had a higher frequency of effector/memoryphenotype (CD44 hi CD62L lo ) cells for both CD4 þ and CD8 þ subsets, compared with wild-type (WT) mice (Fig. 1c). Cic-deficient CD4 þ CD44 þ T cells showed elevated expression of cell surface co-stimulatory molecules such as ICOS, PD-1, OX40 and GITR ( Supplementary Fig. 4). Flow cytometry analysis on cytokine expression profiles revealed that the frequency of IFNg þ , TNF þ , IL-21 þ , IL-13 þ and IL-22 þ CD4 þ T cells was significantly increased in spleen of Cic f/f Vav1-Cre mice ( Supplementary Fig. 5a). Cic f/f Vav1-Cre mice also had a higher frequency of CCR4 À CXCR3 þ (T H 1-related) and CCR4 þ CCR6 À (T H 2-related) CD4 þ T cells 30 (Supplementary Fig. 5b). Thus, loss of CIC spontaneously activates T cells, resulting in accumulation of effector/memory cells with mixed T helper (T H ) phenotypes in mice. On the other hand, Cic null naïve CD4 þ T cells differentiated into each T H subset with similar efficiency to WT cells under in vitro polarizing conditions ( Supplementary  Fig. 6), suggesting that CIC might not be a crucial factor for each T H subset differentiation.
Cic f/f Vav1-Cre mice had a higher proportion of CD4 þ CD25 À FOXP3 þ T cells in spleen, but a comparable proportion of CD4 þ CD25 þ FOXP3 þ T cells, compared with WT mice (Supplementary Fig. 7a). Expression levels of surface molecules that are critical for controlling T reg cell homeostasis and function, such as GITR, CTLA-4, CD103 and GARP (refs 31,32), were comparable between WT and Cic null CD4 þ FOXP3 þ T cells regardless of CD25 expression ( Supplementary Fig. 7b). Consistent with this result, WT and Cic-deficient CD4 þ CD25 þ   (a,b) Spleen weights (a) and the numbers of total splenocytes, B220 þ B, CD3 þ T, CD4 þ T, CD8 þ T, macrophages (CD11b þ CD11c À F4/80 þ ) and dendritic cells (CD11b þ CD11c þ MHCII þ ) (b) in 10-week-old Cic f/f (WT) and Cic f/f Vav1-Cre (cKO) mice. (c) FACS analysis showing the proportion of CD44 lo CD62L hi naïve T cells and CD44 hi CD62L lo effector/memory T cells in spleen of Cic f/f and Cic f/f Vav1-Cre mice. All data are representative of two independent experiments. Data are presented as box-and-whisker plots. n ¼ 4-5 per each genotype. (d) Serum levels of total IgG, IgG1, IgG2a and IgM in 12-week-old Cic f/f and Cic f/f Vav1-Cre mice were measured by ELISA. (e) Images of dissected spleens (SPL) and draining lymph nodes (LN) from Cic f/f and Cic f/f Vav1-Cre mice at the age of 12 months. The graph shows average weights of spleens from the Cic f/f and Cic f/f Vav1-Cre mice. M, male; F, female. (f) Serum levels of anti-dsDNA antibody in the 12-month-old Cic f/f and Cic f/f Vav1-Cre mice were measured by ELISA. The graphs in a,b,d-f show data as mean ± s.e.m. and each dot in graphs represents an individual mouse. *Po0.05, **Po0.01 and ***Po0.001 (two-tailed two-sample unequal variance Student t-test). (g) Immune cell infiltration in liver and lung from the 12-month-old Cic f/f Vav1-Cre mice. Tissue sections were stained with haematoxylin and eosin (H&E). Arrows indicate immune cell infiltrates. (h) H&E-stained kidney sections showing glomerulonephritis in the 12-month-old Cic f/f Vav1-Cre mice. (i) Immunostaining for IgG deposition in kidney glomeruli of the 12-month-old Cic f/f Vav1-Cre mice. Representative images from two independent experiments are shown. (g-i) Scale bars, 100 mm.
T reg cells had a comparable suppressive activity on IL-2 secretion and proliferation of responding CD4 þ CD25 À T cells ( Supplementary Fig. 7c,d). The numbers of CD8 þ CD25 þ FOXP3 þ and CD8 þ ICOSL þ CXCR5 þ (Qa-1-restricted CD8 þ T reg cells) 33,34 T cells were also comparable between WT and Cic f/f Vav1-Cre mice ( Supplementary Fig. 7e,f). These results, together with the increased frequency of thymic T reg cells ( Supplementary Fig. 2), indicate that the T cell hyperactivation in Cic f/f Vav1-Cre mice was not due to defects in T reg cell compartment.
Because the abnormal activation of T cells and the increased frequency of CD4 þ CD25 À FOXP3 þ T cells are often associated with autoimmunity 35,36 , we assessed whether Cic f/f Vav1-Cre mice display autoimmune phenotypes with age. The Cic f/f Vav1-Cre mice developed hyperglobulinemia at 12 weeks of age (Fig. 1d), and exhibited several typical autoimmune attributes at around 1 year of age, including enlarged secondary lymphoid organs (Fig. 1e), increased serum levels of anti-dsDNA antibody (Fig. 1f), infiltration of immune cells into tissues (Fig. 1g), glomerulonephritis (Fig. 1h) and IgG deposition in kidney glomeruli (Fig. 1i). Together, these findings demonstrate that CIC is indispensable for suppression of lymphoproliferative autoimmunity.

Spontaneous induction of GC responses by CIC deficiency.
Increased frequencies of T FH -like cells are often observed in patients with autoimmune diseases including SLE 6 , and studies of Roquin san mice show that overexpression of IFNg and ICOS, a co-stimulatory molecule highly expressed in T FH cells 37,38 , promotes T FH cell development, which in turn spontaneously induces a lupus-like autoimmune syndrome 39,40 . It is also known that IL-21 and IL-4 are key cytokines that T FH cells express during the GC response 41 . Because Cic f/f Vav1-Cre mice had the increased T-cell surface expression of ICOS, the increased proportions of CD4 þ IFNg þ , CD4 þ IL-21 þ and CD4 þ IL-4 þ T cells, and the autoimmune-like symptoms similar to those in Roquin san mice 42 , we set out to determine whether the T FH subset was elevated in Cic f/f Vav1-Cre mice. The frequencies of T FH (PD-1 int CXCR5 int ) and GC T FH (PD-1 hi CXCR5 hi ) cells were significantly increased in the spleen of 9-week-old Cic f/f Vav1-Cre mice at the expense of non-T FH cells (Fig. 2a). The increased T FH cell frequency in Cic f/f Vav1-Cre mice was confirmed by flow cytometry analysis of CD4 þ BCL6 þ CXCR5 þ T cells (Fig. 2b). Consistent with these results, the proportion of CD4 þ ICOS þ T cells was also markedly increased (Fig. 2c). On the other hand, levels of BCL6 in PD-1 þ CXCR5 þ T FH cells were comparable between WT and Cic f/f Vav1-Cre mice (Fig. 2d). The frequency of IFNg þ T FH cells was substantially increased in Cic f/f Vav1-Cre mice, whereas that of IL-21 þ and IL-4 þ T FH cells was comparable between WT and Cic f/f Vav1-Cre mice (Fig. 2e). Because the primary function of T FH cells is to help B cells form GC reactions, we further analysed GC B (B220 þ Fas þ GL-7 þ ) cells. The proportion of GC B cells was significantly higher in Cic f/f Vav1-Cre mice than in WT mice (Fig. 2f). Taken together, these data indicate that CIC is required for suppression of spontaneous induction of GC reactions.
T-cell-intrinsic functions of CIC. Because several abnormalities in T cells were observed in Cic f/f Vav1-Cre mice, we next examined CIC in T cells. CIC protein levels gradually increased along with the activation of CD4 þ T cells by anti-CD3 and anti-CD28 antibodies (Fig. 3a), and were higher in effector/memory CD4 þ T cells than in naïve CD4 þ T cells (Fig. 3b). Cic-deficient CD4 þ T cells secreted IL-2 and proliferated more efficiently than WT T cells when stimulated with anti-CD3 (Fig. 3c,d), suggesting that CIC negatively regulates an activation signal that is mediated by T-cell receptors (TCRs). However, this difference was largely abrogated by addition of anti-CD28 (Fig. 3c,d), indicating that a strong co-stimulatory signal through CD28 overrides the enhanced TCR response in Cic-deficient T cells.
To determine the in vivo consequences of CIC deficiency in T cells, we generated mice with Cic deletion specific to T cells (Cic f/f Cd4-Cre) (Fig. 4a). Development of T cells in thymus was largely normal in Cic f/f Cd4-Cre mice ( Supplementary Fig. 8a). Given that CIC deficiency induced autoimmunity and that Cic null CD4 þ T cells more robustly responded to TCR stimulation than WT cells in vitro (Fig. 3c,d), we assessed thymic negative selection, which is a critical process that removes autoreactive T cells in thymus 43 , in WT and Cic f/f Cd4-Cre mice. TCR-induced apoptosis comparably occurred in WT and Cic null double positive (DP) thymocytes ( Supplementary Fig. 8b), suggesting that TCR-induced negative selection might not be affected by CIC deficiency. Consistent with this finding, WT and Cic null CD4 þ thymocytes exhibited a comparable induction of Nur77, which has been implicated in negative selection through its ability to convert BCL2 into a proapoptotic molecule 44,45 , in response to TCR stimulation ( Supplementary Fig. 8c).
On the other hand, as observed in the Cic f/f Vav1-Cre mice, hyperglobulinemia, T-cell hyperactivation and increased populations of CD25 À FOXP3 þ À , T H 1-type and T H 2-type CD4 þ T cells in spleen occurred in Cic f/f Cd4-Cre mice at 12 weeks of age ( Fig. 4b, Supplementary Fig. 9). These mice also showed several systemic autoimmune phenotypes with age, including increased serum levels of anti-dsDNA antibody (Fig. 4c), infiltration of immune cells into tissues (Fig. 4d) and glomerulonephritis (Fig. 4e). Moreover, the proportions of T FH , GC T FH , CD4 þ ICOS þ T and GC B cells were significantly higher in Cic f/f Cd4-Cre mice than in WT mice ( Fig. 4f-h). The frequency of IFNg-expressing T FH cells was increased in the spleen of Cic f/f Cd4-Cre mice ( Supplementary Fig. 10a). Because IFNg drives IgG2a class-switching in B cells during GC reactions 40,46,47 , we further assessed IgG2a þ GC B cells. As expected, the frequency and the number of IgG2a þ GC B cells were also increased in Cic f/f Cd4-Cre mice ( Supplementary  Fig. 10b). Taken together, these findings demonstrate that the spontaneous inductions of T-cell activation, GC response and systemic autoimmunity were due to loss of CIC in T cells.
Follicular regulatory T (T FR ) cells, defined as CD4 þ FOXP3 þ PD-1 þ CXCR5 þ cells, are a specialized subset of regulatory T cells that inhibit antibody production [48][49][50] . Because the proportion of CD4 þ FOXP3 þ T cells increased in Cic-deficient mice (Supplementary Figs 7a and 9b), we determined whether cells were also significantly overrepresented in the spleen of Cic f/f Cd4-Cre mice compared with WT mice (Fig. 4i). Importantly, the ratio T FH /T FR , which is a critical factor that dictates the magnitude of antibody production 50 , was also increased in the Cic f/f Cd4-Cre mice (Fig. 4j), consistent with the observation that GC B cells and serum immunoglobulin levels were elevated in Cic f/f Cd4-Cre mice (Fig. 4b,h).
Normal T-cell homeostasis in T reg -specific Cic null mice. Although the suppressive activity toward conventional T cells and expression profiles of T reg cell-associated surface molecules were comparable between WT and Cic-deficient CD4 þ T reg cells ( Supplementary Fig. 7b-d), the frequency of CD4 þ CD25 À FOXP3 þ T cells was substantially increased in Cic-deficient mice ( Supplementary Figs 7a and 9b). To directly address the function of CIC in development and function of FOXP3 þ T reg cells, we generated FOXP3 þ T reg cell-specific Cic null (Cic f/f Foxp3-YFP-Cre) mice by crossing Cic f/f mice with Foxp3-YFP-Cre mice 51 . The ablation of Cic alleles in Foxp3-expressing (YFP þ ) cells did not affect formation of CD25 þ FOXP3 þ and CD25 À FOXP3 þ CD4 þ T cells in both thymus and spleen (Supplementary Fig. 11a-c). The proportion of naïve and effector/memory cells for both CD4 þ and CD8 þ subsets, T FH , T FR and GC B cells was also comparable between Cic þ / þ Foxp3-YFP-Cre (WT) and Cic f/f Foxp3-YFP-Cre mice (Supplementary Fig. 11d-f). These results suggest that CIC deficiency in FOXP3 þ T reg cells does not cause defects in development and function of FOXP3 þ T reg cells.
Since the FOXP3 þ T reg cell-specific Cic deletion did not promote formation of CD4 þ CD25 À FOXP3 þ T-cell population in spleen ( Supplementary Fig. 11c), we examined expression of FOXP3 in conventional CD4 þ T cells from WT and Cic f/f Cd4-Cre mice. Interestingly, the proportion of FOXP3-expressing naïve (CD45RB hi CD44 lo CD62L hi ) CD4 þ T cells was markedly increased in the spleen of Cic f/f Cd4-Cre mice, compared with WT mice ( Supplementary Fig. 12a). Consistent with this result, CD4 þ CD25 À FOXP3 þ T cells in Cic f/f Cd4-Cre mice were composed of both CD44 lo CD62L hi and CD44 hi CD62L lo populations, while those in WT mice were mainly CD44 hi CD62L lo cells ( Supplementary Fig. 12b). We also determined the origin of CD4 þ CD25 À FOXP3 þ T cells in Cic f/f Cd4-Cre mice by flow cytometry analysis for expression of Helios and Neurophilin-1/ NRP1, both of which are markers for thymus-derived T reg (tT reg ) cells [52][53][54] . The frequency of Helios À NRP1 À cells for both CD25 þ FOXP3 þ and CD25 À FOXP3 þ T-cell subsets was significantly increased in the spleen of Cic f/f Cd4-Cre mice ( Supplementary Fig. 12c), demonstrating that the increased frequency of CD4 þ CD25 À FOXP3 þ T cells in Cic f/f Cd4-Cre mice was due to the expansion of peripherally-induced T reg (pT reg ) cells. Taken together, our data suggest that CIC deficiency induces FOXP3 expression in naïve CD4 þ T cells, thereby increasing the population of CD4 þ CD25 À FOXP3 þ T cells in the periphery.

Regulation of T FH cell differentiation by the CIC-ETV5 axis.
To understand the molecular mechanism mediating the CIC deficiency-induced increase in T FH cells, we set out to identify CIC target genes in CD4 þ T cells. We analysed gene expression profiles in naïve and anti-CD3/CD28 antibody-activated CD4 þ T cells from WT and Cic f/f Cd4-Cre mice by RNA sequencing. Among the genes significantly upregulated (log 2 (fold change) 41.5; Po0.05) in Cic-deficient CD4 þ T cells, only Etv4 and Etv5 were upregulated irrespective of the status of T-cell activation (Supplementary Data 1). Both Etv4 and Etv5 have been wellcharacterized as CIC target genes in various tissues 19,22,24,27 . qRT-PCR analysis for mRNA levels of Pea3 group genes (Etv1, Etv4 and Etv5) (Fig. 5a) and quantitative PCR (qPCR) analysis for Pea3 group gene promoter regions in total CIC-associated DNA fragments (Fig. 5b) verified that Etv4 and Etv5 are direct targets of CIC in CD4 þ T cells.
We then examined the levels of Etv4 and Etv5 in CD4 þ PD-1 þ CXCR5 þ T FH cells from WT and Cic f/f Cd4-Cre mice by    qRT-PCR analysis. Consistent with the previous result (Fig. 5a), levels of both genes were markedly increased in Cic-deficient T FH cells (Fig. 5c), indicating that their expression was derepressed in T FH cells in the absence of CIC. Interestingly, however, only ETV5 expression, but not ETV4 expression, was substantially increased in Cic-deficient T FH cells at the protein level (Fig. 5d), suggesting that ETV4 expression is tightly controlled in T FH cells at the posttranscriptional level. We also found that ETV5 levels were higher in T FH cells than in non-T FH cells (Fig. 5c,e), whereas CIC levels in the two T-cell populations were comparable (Fig. 5e), implying that ETV5 might be critical for differentiation or function, or both, of T FH cells. To further explore this possibility, we performed adoptive transfer experiments using OT-II cells, which express ovalbumin (OVA)-specific TCRs 55 . Thy1.1 þ OT-II cells were infected with control or ETV5expressing retrovirus ( Supplementary Fig. 13a) and adoptively transferred to Thy1.2 þ recipient mice. Immunization of the host mice with 4-hydroxyl-3-nitrophenyl (NP)-OVA in alum resulted in generation of a higher percentage of T FH cells (Fig. 5f), demonstrating that ETV5 promotes T FH cell differentiation. Consistent with this result, knockdown of ETV5 in OT-II cells by shRNA against Etv5 (shETV5) suppressed T FH cell differentiation (Fig. 5g). Lastly, we conducted the adoptive transfer experiment to examine whether the increased T FH cells in Cic mutant mice was due to derepression of ETV5. WT and Cic-deficient OT-II cells infected with control or shETV5-expressing retrovirus were adoptively transferred to Thy1.2 þ recipient mice. Seven days after immunization with NP-OVA in alum, the donor OT-II cells were analysed for T FH cell differentiation by flow cytometry. As expected from previous results (Figs 2a,b and 4f), T FH cell differentiation was more prominent with Cic-deficient donor OT-II cells than with WT OT-II cells (Figs 5h and 6f). Strikingly, such enhanced T FH cell differentiation of Cic-deficient OT-II cells was blunted when ETV5 expression was knocked-down with shETV5 ( Fig. 5h, Supplementary Fig. 13b). Altogether, these data demonstrate that the CIC-ETV5 axis is crucial for regulation of T FH cell differentiation.
Maf as a critical ETV5 target in T FH cell differentiation. ETV5 function in T FH cell development has not been investigated. To comprehensively understand how the CIC-ETV5 axis regulates differentiation of T FH cells, we sought to identify ETV5 target genes that could be involved in this process. We initially analysed expression profiles of several T FH -related genes in non-T FH and T FH cells from WT and Cic f/f Cd4-Cre mice. Bcl6, Batf, Maf, Cxcr5 and Ifng were significantly increased, whereas Lef1 and Foxp1 were decreased in WT T FH cells compared with WT non-T FH cells (Fig. 6a). Importantly, Batf, Maf, Icos and Ifng were significantly upregulated in Cic-deficient T FH cells compared with WT T FH cells; the increase was greatest in Maf, which promotes T FH cell differentiation and GC reactions 14,16 (Fig. 6a). We confirmed the overexpression of MAF in Cic-deficient T FH cells at the protein level (Fig. 6b).
We also examined which T FH -related genes are regulated by ETV5 in CD4 þ T cells. Anti-CD3/CD28 antibody-activated CD4 þ T cells were infected with control or ETV5-expressing retrovirus and cultured in the presence of IL-6 and IL-21 (T FH -like condition) 56 . Among the genes tested, only Maf levels were substantially increased by ETV5 overexpression (Fig. 6c). Consistent with this, it is known that Maf is a direct target gene of ETV5 in ocular lens cells 57 . We confirmed that ETV5 directly bound to Maf promoter in IL-6-treated CD4 þ T cells by chromatin immunoprecipitation (ChIP)-qPCR analysis ( Supplementary Fig. 14). In contrast, in the absence of IL-6 and IL-21, ETV5 overexpression did not significantly increase levels of Maf (Fig. 6d), suggesting that ETV5 might induce expression of Maf during T FH cell development. These results were also confirmed by western blot analysis (Fig. 6e). Because IL-6 and IL-21 can activate STAT3 pathway, we examined whether activation of STAT3 is required for ETV5-mediated induction of Maf expression. Treatment with AG490 (inhibitor of JAK2-STAT3 pathway) 58 and Stattic (inhibitor for STAT3 dimerization) 59 completely suppressed the induction of Maf expression by ETV5 in CD4 þ T cells incubated under T FH -like condition ( Supplementary Fig. 15a,b). However, treatment   with Stattic did not affect ETV5 promoter occupancy of Maf ( Supplementary Fig. 15c). These results indicate that STAT3 activity is required for ETV5-mediated induction of Maf expression, but dispensable for ETV5 binding to Maf promoter. Finally, we examined whether the upregulation of MAF expression contributed to the enhanced T FH cell differentiation in Cic-deficient T cells by the adoptive transfer experiment. WT and Cic null OT-II cells transduced with control or Maf shRNA (shMAF)-expressing retrovirus were adoptively transferred to Thy1.2 þ recipient mice. Seven days after immunization with NP-OVA in alum, the donor OT-II cells were analysed for T FH cell differentiation by flow cytometry. The enhanced T FH cell differentiation of Cic-deficient OT-II cells was indeed significantly alleviated by knockdown of Maf (Fig. 6f, Supplementary Fig. 13b). Taken together, these data suggest that the derepression of Etv5 increased Maf expression in Cic-deficient T FH cells, and subsequently promoted T FH cell differentiation and GC responses.

Discussion
In this study, we examined a previously unrecognised function of CIC in the immune system. Our findings demonstrate that CIC is a key transcriptional repressor that maintains peripheral immune homeostasis and restrains T FH cell differentiation, thereby suppressing autoimmunity (Fig. 6g). CIC deficiency causes lymphoproliferative autoimmunity in mice, which is accompanied by augmented T-cell responses and increased frequencies of T FH and GC B cells in secondary lymphoid organs. Given that excessive formation of T FH cells is often observed in patients with autoimmune diseases 5 and that autoimmunity in Roquin san mice is reduced when T FH cell development is suppressed either by removing one allele of Bcl6, by deleting SAP (Sh2d1a), or by dampening IFNg signalling 40,60 , spontaneous induction of the GC response might be a primary cause of the onset of autoimmune-like symptoms in Cic mutant mice.
The immune cell-specific Cic null mice shared many phenotypes including hyperglobulinemia, T-cell hyperactivation, accumulation of effector/memory cells with T H 1 and T H 2 phenotypes, systemic autoimmunity and increased proportions of CD4 þ FOXP3 þ CD25 À T, T FH and GC B cells with the T-cell-specific Cic null mice, demonstrating that T-cell-intrinsic functions of CIC are crucial for maintenance of T-cell homeostasis and suppression of spontaneous induction of the GC response and autoimmunity. A specific loss of CIC in B cells, from which antigen presentation and co-stimulatory signals are ARTICLE required for differentiation and maintenance of T FH cells 8 , did not promote T FH cell differentiation in mice ( Supplementary  Fig. 16), accentuating the T-cell-intrinsic requirement of CIC in regulation of T FH cell development. However, most phenotypes were more prominent in the immune cell-specific Cic null mice than in the T-cell-specific Cic-deficient mice; this difference implies that T-cell-extrinsic mechanisms also contributed in part to those phenotypes. To comprehensively understand the function of CIC in maintenance of peripheral immune tolerance, functions of CIC in other types of immune cell, such as DCs and B cells, need to be further determined.
Our study provided a molecular basis of how CIC deficiency promotes T FH cell differentiation. We determined that ETV5 is a critical CIC target that promotes T FH cell differentiation. Among Pea3 group genes, Etv4 and Etv5 are major CIC target genes in CD4 þ T cells. Although mRNA levels of both genes increased greatly in the Cic null T FH cells, only ETV5 expression increased significantly at the protein level, indicating that ETV4 levels are tightly controlled in T FH cells by posttranscriptional regulation. ETV5 expression also seems to be regulated at the posttranscriptional level in the Cic null T FH cells, because the increase in ETV5 expression by CIC deficiency was greater at the mRNA level (B5.5-fold) than at the protein level (B3.4 fold). One possible mechanism of posttranscriptional control of PEA3 group transcription factor levels in T FH cells is proteasomal degradation mediated by E3 ubiquitin ligase COP1 (also known as RFWD2). COP1-mediated degradation of PEA3 group transcription factors is involved in regulation of prostate cancer progression 61 , b-cell insulin secretion 62 and lung-branching morphogenesis 63 , but the function of COP1 in T FH cell development has not been determined. Because Rfwd2 is expressed in T lymphocytes 64 , one interesting follow-up would be to examine whether COP1 regulates T FH cell differentiation through degradation of ETV5 or other PEA3 group transcription factors in T cells.
We identified Maf as an important downstream target gene of the CIC-ETV5 axis in the process of T FH cell differentiation. To identify genes that may be regulated by ETV5 during T FH cell differentiation, we examined 14 genes that encode transcription factors, cytokines or cell surface molecules, all of which are critically involved in regulation of T FH cell development. Strikingly, ETV5 overexpression induced only Maf expression in CD4 þ T cells. The finding that T FH cell differentiation of the Cic null OT-II cells was suppressed by knockdown of Maf further demonstrates that the CIC-ETV5 axis regulates T FH cell differentiation via MAF. However, in the Cic-deficient OT-II cells, Maf RNAi suppressed T FH cell differentiation less efficiently than did Etv5 RNAi; this comparison suggests that, in addition to Maf, other ETV5 target genes may contribute to promotion of T FH cell differentiation. Moreover, Cic-deficient T FH cells exhibited the enhanced expression of IFNg, which is critical for T FH cell differentiation 40 , but not regulated by MAF 14 . This result also implies that Maf may not be the only ETV5 target gene that contributes to promotion of T FH cell differentiation. To improve our understanding of how ETV5 promotes T FH cell development, studies of the function of ETV5 in T FH cells, such as a genome-wide identification of ETV5 targets and their function in T FH cells, should be conducted.
Given that ETV5 upregulated Maf expression only in the presence of IL-6 and IL-21, which are cytokines that induce T FH cell development 56 , and that Maf levels were increased in Cic null T FH cells, but in neither naïve nor anti-CD3/CD28-activated Cic-deficient CD4 þ T cells (Supplementary Data 1), ETV5 might specifically regulate Maf expression during T FH cell development. Consistent with this inference, Maf levels were comparable between WT and Etv5 deficient T H 17 cells 65 .
However, ETV5 facilitates differentiation of T H 17 cells 65 and of IL-17-producing gd effector cells 66 by activating Il17 expression. ETV5 also promotes IL-9 production in T H 9 cells by binding and recruiting histone acetyltransferases to the Il9 locus 67 . These findings, together with our data, suggest that ETV5 participates in differentiation processes of several T H subtypes by regulating a different set of genes in each T H subset.
Our study is the first demonstration that the CIC-ETV5 axis is critical for regulation of T FH cell development. This axis also contributes to cancer progression [22][23][24] ; therefore, dysregulation of this regulatory axis could contribute to pathogenesis of both cancer and antibody-mediated autoimmune diseases. In this regard, CIC and ETV5 themselves or their target genes that commonly mediate pathogenesis of both diseases could be evaluated as molecular targets for treatment of both autoimmune diseases and cancers. Mice were fed standard rodent chow and water ad libitum. Three-to-thirteen mice that were used in individual experiments were assigned randomly to the experimental groups. Male and female mice were also randomly assigned to the experimental groups. Blinding was not possible in most animal experiments. All procedures were approved by the Pohang University of Science and Technology Institutional Animal Care and Use Committee.
Enzyme-linked immunosorbent assay. Ninety six-well ELISA plates were pre-coated with 2 mg ml À 1 anti-mouse Ig (1010-01, Southern Biotechnology) at 4°C overnight. On the next day, the plates were washed with washing buffer (phosphate-buffered saline (PBS) with 0.05% Tween 20) and blocked with blocking buffer (washing buffer with 2% bovine serum albumin (BSA)) for 1-2 h at room temperature (RT); then 50 ml of diluted serum and standards were added and incubated for 2 h at RT. The plates were washed and incubated for 1 h at RT with secondary antibodies conjugated with horseradish peroxidase. For total IgG, goat anti-mouse IgG (H þ L) (pooled antisera from goats hyperimmunized with mouse IgG, 1:3,000 diluted, 1031-05) was used; for IgG1, goat anti-mouse IgG1 (pooled antisera from goats hyperimmunized with mouse IgG1, 1:3,000 diluted, 1070-05) was used; for IgG2a, goat anti-mouse IgG2a (pooled antisera from goats hyperimmunized with mouse IgG2a paraproteins, 1:1,000 diluted, 1080-05) was used; and for IgM, goat anti-mouse IgM (pooled antisera from goats hyperimmunized with mouse IgM, 1:2,000 diluted, 1020-05) was used. The plates were washed at least five times, then 50 ml of TMB substrate (SurModics, TMBM-1000-01) was added and incubated for 20 min in the dark. Finally, 50 ml of stop solution (1 M H 2 SO 4 ) was added and the plate was read at 450 nm. The serum concentration was calculated according to the standard curve generated. Serum anti-dsDNA concentration was measured using mouse anti-dsDNA ELISA Kit (Shibayagi, AKRDD-061) according to the manufacturer's instruction. For IL-2 ELISA, naïve CD4 þ T cells were activated with plate-bound anti-CD3 (1.0 mg ml À 1 , 145-2C11, BD) in the presence or absence of plate-bound anti-CD28 (2.0 mg ml À 1 , 37.51, BD), then cell supernatants were collected 48 and 72 h after stimulation. Briefly, plates pre-coated with 2 mg ml À 1 anti-IL-2 (14-7022, eBioscience) were washed and blocked as described above, then 50 ml of diluted samples and standards were added and incubated for 2 h at RT. The plates were washed, then incubated sequentially with biotin-conjugated anti-IL-2 (13-7021) for 1.5 h, avidin-HRP (18-4100) for 30 min, TMB substrate for 15 min, and 1 M H 2 SO 4 as a stop solution.
Tissue histology. Liver, lung and kidney tissues obtained from 12-month-old Cic f/f and Cic f/f Vav1-Cre mice or 14.5-month-old Cic f/f and Cic f/f Cd4-Cre mice were fixed in 10% formalin, then embedded in paraffin before sectioning. The tissues were cut into 5-mm sections (Leica RM2245), then sections were deparaffinized and dehydrated by using xylene, 100% ethanol and 95% ethanol sequentially. They were washed in distilled water, then stained with haematoxylin (Sigma, HHS32) and eosin (Sigma, HT110132). After H&E staining, Â 200 images of specimens were obtained using an Olympus CX31 light microscope.
Immunofluorescence staining of IgG deposition. Kidney tissues from 12-month-old Cic f/f and Cic f/f Vav1-Cre mice were snap-frozen in OCT medium (Sakura Finetek, 4583), then cut into 10-mm sections (Leica CM3050S). The kidney sections were air-dried for 24 h, then washed in PBS (pH 7.4) to remove fixation compound. Diluted anti-mouse CD16/CD32 antibodies (eBioscience, 14-0161) were treated for Fc blocking and incubated for 30 min at RT. After washing, the kidney sections were stained with anti-IgG antibody conjugated to fluorescein isothiocyanate (FITC) (1:500 diluted, Sigma, F0257) and 4 0 ,6-diamidino-2phenylindole to reveal the IgG complexes and the nuclei, respectively. The specimens were washed in PBS, then covered in mounting medium (Dako, C0563), and Â 400 images of specimens were obtained using an Olympus IX82-ZDC2 fluorescence microscope.
Plasmids and retrovirus production. The amino-terminal FLAG-tagged coding sequence (CDS) of mouse Etv5 was amplified by PCR, initially cloned into T-blunt vector (SolGent), and then verified by sequencing. A confirmed Etv5 CDS fragment digested by XhoI/HpaI was sub-cloned into MigR1 retroviral vector (MigR1-ETV5-GFP). The shRNA expression vectors for knockdown of mouse Etv5 and Maf were generated using MSCV-LTRmiR30-PIG (LMP) vector (Open Biosystems) according to the manufacturer's instruction. The target sequences were as follows. pure. In FACS plots, isotype controls for each of antibodies were used for separating negative and positive populations and all gates were based on this method. All FACS sorting/gating strategies are described in Supplementary  Fig. 17.
In vitro T-cell proliferation assay. Naïve CD4 þ T cells were purified to 495% purity from pooled spleens and lymph nodes of Cic f/f and Cic f/f Vav1-Cre mice by using a CD4 þ negative selection method (Stem Cell Technologies) for CD4 þ CD25 À CD44 lo CD62L hi cells. The cells were labelled with 5 mM Cell Trace Violet (CTV, Invitrogen, C34557) in pre-warmed PBS. Labelling was performed by incubating in a 37°C incubator for 20 min and stopped by adding five times the original staining volume of culture medium (containing Z1% protein). The cells were incubated for at least 10 min to allow the CTV reagent to undergo acetate hydrolysis. The cells were stimulated with plate-bound anti-CD3 (1.0 mg ml À 1 , 145-2C11, BioXcell) and anti-CD28 (2.0 mg ml À 1 , 37.51, BioXcell), then collected 72 h after stimulation. The samples were analysed using the LSRFortessa flow cytometer. Data were analysed using FlowJo software (Tree Star).
In vitro T reg cell suppression assay. CD4 þ CD25 À CD44 lo T cells were prepared from CD45.1 C57BL/6 mice. The cells were labelled with CTV and used as responder cells (Tresp). CD4 þ CD25 þ T reg cells were isolated from Cic f/f and Cic f/f Cd4-Cre mice by the MoFlo-XDP (Beckman Coulter). Tresp cells were mixed with an equal number of irradiated splenocytes (3000 Rad) from Rag1 À / À mice and different numbers of CD4 þ CD25 þ T reg cells and incubated in round bottom 96-well tissue culture plates with anti-CD3 (1 mg ml À 1 ). Seventy-two hours later, CD45.1 þ CD4 þ T cells were gated and proliferation was analysed based on CTV dilution. For detection of IL-2 secretion, CD4 þ CD25 þ T reg cells from Cic f/f and Cic f/f Cd4-Cre mice were used at a fixed concentration of 1 Â 10 5 cells per well and co-cultured with CD4 þ CD25 À Tresp cells in 96-well plates. The cells were stimulated with plate-bound anti-CD3 (1 mg ml À 1 ) and anti-CD28 (1 mg ml À 1 ), then supernatants were collected 72 h after stimulation. The samples were subjected to ELISA for IL-2 concentration.
Thymic negative selection assay. In vitro thymic negative selection assay was performed as described previously 45 . Briefly, DP thymocytes were prepared from Cic f/f and Cic f/f Cd4-Cre mice using the MoFlo-XDP and stimulated with anti-CD3 (20 mg ml À 1 ) and anti-CD28 (50 mg ml À 1 ) for 24 h to evaluate TCR-induced apoptosis. Cell death was assessed by staining for Annexin V according to the manufacturer's protocol (BD, 556419). To investigate the induction of Nur77 expression after TCR cross-linking, thymocytes from Cic f/f and Cic f/f Cd4-Cre mice were stimulated with anti-CD3 (5 mg ml À 1 ) and anti-CD28 (10 mg ml À 1 ) for 6 h. The samples were analysed using the LSRFortessa flow cytometer, and data were analysed using FlowJo software.
Retroviral transduction and cell transfers. OT-II Thy1.1 CD4 þ T cells were purified from whole splenocytes by negative selection using EasySep (StemCell), then re-suspended in T-cell medium (RPMI-1640 þ 10% FBS, supplemented with 2.05 mM L-Glutamine, 100 U ml À 1 Penicillin/Streptomycin and 55 mM BME). The purified OT-II T cells were stimulated in 24-well plates pre-coated with 2 mg ml À 1 anti-CD3 (145-2C11, BD) and anti-CD28 (37.51, BD) in the presence of rmIL-2 (50 U ml À 1 ). The OT-II T cells were transduced with control retrovirus (MigR1-GFP) or ETV5-expressing retrovirus (MigR1-ETV5-GFP) at 24 and 48 h after in vitro stimulation. After a total of 72 h stimulation, the cells were transferred into new six-well plates with rmIL-2 for 24 h, and then transferred again into new six-well plates in the presence of rmIL-7 for 24 h prior to cell sorting. For shRNA experiments, after in vitro stimulation and double transduction with shNC, shETV5 or shMAF expressing retrovirus for 72 h, the cells were transferred into six-well plates with rmIL-2 for 12 h and then transferred into new six-well plates with rmIL-7 for 12 h.
For adoptive transfer experiments, 0.2-0.5 Â 10 6 OT-II T cells infected with retroviruses were transferred into C57BL/6 Thy1.2 recipient mice by tail-vein injection. Three days later, mice were immunized intraperitoneally with 200 mg of NP-OVA (9006-59-1, Sigma) in 200 ml of alum. Seven or eight days after immunization, the donor cells were analysed for T FH differentiation by flow cytometry.
For STAT3 inhibitor treatment experiments, naïve CD4 þ T cells were polarized under T FH -like condition as described above. The cells were transduced with retrovirus (MigR1-GFP or MigR1-ETV5-GFP) at 24 h after in vitro stimulation, then treated with AG490 (Tokyo Chemical, 50 mM) for 16 h or Stattic (Santa Cruz Biotechnology, 20 mM) for 12 h. Two days after culture, the cells were rested for 24 h. The cells were collected for RNA samples.
For ETV5 ChIP experiment, CD4 þ T cells were activated with plate bound anti-CD3 (145-2C11, BD) and anti-CD28 (37.51, BD) in the presence of rmIL-6. Sixteen hours after in vitro stimulation, 1 Â 10 7 cells and 2 mg of anti-ETV5 antibody (sc-22807, Santa Cruz Biotechnology) were used for each ChIP experiment. To inhibit STAT3 activity, CD4 þ T cells were activated with anti-CD3/CD28 in the presence of rmIL-6 for 12 h, and then Stattic (20 mM) was treated for 12 h. The primer sequences used are listed in Supplementary Data 2.
qRT-PCR. Total RNA was extracted using Trizol reagent (301-001, GeneAll), then 1-2 mg of the total RNA was subjected to cDNA synthesis using a GoScript Reverse Transcription System (A5000, Promega). Each gene expression level was normalized to Hprt levels and presented as relative to WT. The primers used in qRT-PCR analyses are listed in Supplementary Data 3.
Statistical analysis. For statistical analysis, all experiments were performed more than three times independently. Statistical analyses were performed using the Student t-test (two-tailed, two-sample unequal variance). Po0.05 was considered significant. In bar graphs, bars indicate means and error bars indicate SEM. In box-and-whisker plots, boxes represent median with upper and lower quantiles and whiskers represent values of maximum and minimum.
Data availability. The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information Files, or from the corresponding authors on reasonable request. The RNA-seq data were deposited in the Gene Expression Omnibus (NCBI) data repository under accession number GSE84125.