SMAD4 governs a feedforward regulation of the TGF-β-effects in CD8 T cells that contributes to preventing chronic intestinal inflammation

SMAD4, a key mediator of TGF-β signaling, plays a crucial role in T cells to prevent chronic gut inflammation. However, the molecular mechanisms underlying this control remain elusive. Using different genetic and epigenetic approaches, we unexpectedly reveal that SMAD4 in CD8 T cells prevents chronic intestinal inflammation by a feedforward mechanism that is TGF-β-independent. Prior to any TGF-β-receptor engagement, SMAD4 acts as an active and basal repressor of epigenetic, transcriptional and functional TGF-β imprinting in CD8 T cells. Thus, in sharp opposition to total TGF-β signaling deletion, SMAD4 deletion impairs naïve CD8 T cell effector predisposition but promotes CD8 T cell accumulation and epithelial retention by promoting their response to IL-7 and their expression of integrins such as Itgae. Besides, SMAD4 deletion unleashes the induction of a wide range of TGF-β-signaling-repressors such as Smad7, Ski, Skil, and Smurf2 and hampers TGF-β-mediated CD8 T cell immunosuppression. Mechanistically, prior to any TGF-β signal, SMAD4 binds to the loci of several TGF-β-target genes, and by regulating histone acetylation, represses their expression. The massive gut epithelial colonization, associated with their escape from the immunoregulatory TGF-β effects overtakes their poor effector preconditioning and elicits microbiota-driven chronic epithelial CD8 T cell activation. Hence, in an anticipatory manner, independently of TGF-β, SMAD4 governs a feedforward regulation of TGF-β effects in CD8 T cells, preventing chronic intestinal inflammation.


Introduction
established mice with a deletion of SMAD4 (SKO), TRIM33 (TKO), double deletion of 87 TRIM33 and SMAD4 (STKO) or double deletion of TGF-RII and SMAD4 (R2SKO) in T 88 cells (Fig. 1a, Supplementary information, Fig. 1a). Consistent with previous studies, 89 TKO, SKO, STKO and R2SKO mice did not display any signs of autoimmunity even at a 90 more advanced age [22][23][24] . However, strikingly, the weight of all mice lacking SMAD4 in T 91 cells (SKO, STKO and R2SKO) stopped increasing from 4 months of age onwards (Fig.   92 1b). Postmortem analysis revealed an important intestinal inflammation in these animals, 93 illustrated by an enlargement of the duodenum and a shortening of the colon (Fig. 1c-d). 94 Histological analysis revealed massive immune cell infiltrations in the mucosa and homeostasis. Given that TGF-RII-deficient (R2KO) mice die within 3-4 weeks 5,6 and 106 intestinal inflammation only develops at 5 months in SMAD4-deficient mice, we employed 107 a bone marrow (BM)-engrafted mouse model to compare age-matched adult mice. We 117 immunopathology observed in SMAD4 deficient mice 118 Next, we examined which effector T cell population mediates this intestinal 119 immunopathology by using specific anti-CD4 and anti-CD8 depleting antibodies. To 120 avoid undesirable long-term side effects of depleting antibody treatment, we used the BM-121 engrafted mouse models (Fig. 2a). Flow cytometry analysis confirmed the effective 122 ablation of conventional CD8αβ and CD4 T cells in secondary lymphoid organs (SLOs) 123 and in the gut without depleting the other populations, such as CD8 TCRand  136 We then assessed the mechanism by which SMAD4 in CD8 T cells prevents 137 intestinal immunopathology. Strikingly, we observed in all SMAD4-deficient mice (SKO,138 STKO and R2SKO) a substantial increase in the frequency and numbers of CD8αβ T cells 139 in secondary lymphoid organs, as well as in the lungs, skin, colon, and small intestine, 140 compared to WT or TKO mice ( Fig. 3a-b, Supplementary information, Fig. 3a-b). This 141 data revealed a systemic accumulation of CD8 T cells in the absence of SMAD4.

142
Besides this important accumulation, CD8 T cells from SKO,STKO and R2SKO mice,143 expressed large amounts of cytotoxic molecules including granzymes A and B (GZMA 144 and GZMB) or pro-inflammatory cytokines and chemokines such as IFN-, TNF and CCL3 in the intestinal epithelium compared to WT and TKO mice (Fig. 3c, 146 Supplementary information, Fig. 3c-d). Importantly, the strong co-expression of the 147 epithelial retention marker CD103 and GZMB, suggests that the activated CD8T cells 148 were likely bona fide intra-epithelial lymphocytes (IELs) (Supplementary information, 149 Fig. 3e). Remarkably, CD8T cells from SMAD4-deficient mice were barely or not 150 activated in the spleen, the lung, skin, lymph nodes, and lamina propria of the intestine 151 (Fig. 3c, Supplementary information, Fig. 3f-g). This indicated a spatial-restricted-152 activation of SMAD4-deficient CD8 T cells within the intestine epithelium. 153 Next, we investigated the mechanisms triggering intestinal epithelial activation of 154 CD8 T cells in SMAD4-deficient mice. Given the importance of the microbiota in 155 shaping intestinal immunity and promoting IBDs 1,2 , we hypothesized that commensal 156 bacteria could be responsible for CD8 T cell intestinal epithelial accumulation and 157 exacerbated epithelial activation. In order to confirm this scenario, SKO mice were treated 158 with antibiotics (ATB). Strikingly, ATB treatment of SKO mice completely abrogated 159 CD8 T cell accumulation in the gut epithelium (Fig. 3d). In addition, the enhanced 160 production of IFN-γ and granzymes in CD8 IELs was also abolished in ATB-treated 161 SKO mice (Fig. 3e). Hence, these data reveal that the TGF--independent SMAD4 162 function prevents the spontaneous microbiota-driven activation of CD8T cells within 163 the epithelial layer of the intestine. To go deeper in the molecular processes governing SMAD4-mediated control of 168 intestinal homeostasis in CD8 T cells, we next performed a global gene expression profile 169 of CD8 T cells from WT, SKO and R2KO mice. In order to rule out any potential side 170 effects of the inflammatory environment, we used F5 TCR transgenic CD8 T cells in a 171 RAG2KO background, since these mice did not develop any inflammation 172 (Supplementary information, Fig. 4a). Unexpectedly, the comparison between SKO 173 CD8 T cells and R2KO CD8 T cells resulted in a larger set of significantly differentially expressed genes (DEGs) (1573 genes) (FDR < 0.05) than the comparison between SKO 175 and WT mice (408 DEGs) (Fig. 4a), highlighting a wider molecular gap between SKO and 176 R2KO naïve CD8 T cells. An unsupervised hierarchical clustering of all DEGs revealed 177 five distinct clusters. Strikingly, DEGs in which the SMAD4 deletion and the TGF-βRII 178 deletion show a distinct expression pattern (clusters II, III and V) represent more than 92 179 % of all DEGs, definitively, indicating a wide transcriptional disparity between SKO and 180 R2KO CD8 T cells. (Fig. 4b, Supplementary information, Fig. 4b). More importantly, 181 the large majority of the divergent DEGs are genes where SMAD4 deletion affects 182 negatively (cluster II) or positively (cluster III) their expression compared to WT and 183 oppositely to TGF- signaling depletion (Fig. 4b-c) belonging to the T cell effector program. In SMAD4-deficient naïve CD8 T cells, genes 199 encoding effector molecules such as Ifn, IL12r2,Gzmb,Gzma,Gzmk,and Cd244 were 200 repressed. Accordingly, the expression of transcription factors known to direct CD8 T cell 201 effector differentiation (Tbx21, Irf4, Zeb2, and Eomes) was also attenuated (Fig. 4f). 202 Conversely, genes associated with quiescence/naiveness of CD8 T cells (eg. Lef1,itgae,203 Il7r, Ets2) were slightly enhanced or not significantly affected in SMAD4-deficient CD8 204 T cells. Single TGF-RII deletion, in contrast, drastically promoted the expression of 205 effector genes (Fig. 4f). A gene set enrichment analysis (GSEA) of all DEGs and the 206 expression of 43 selected genes associated with T cell activation indicated that similarly 207 to SKO CD8 T cells, the effector gene predisposition was also repressed in naïve R2SKO 208 CD8 T cells (Fig. 4g, Supplementary information, Fig. 4c-d). Functionally, CD8 T cells 209 lacking SMAD4 displayed less activation judged by the weaker GZMB and TBET 210 expression compared to WT and R2KO cells after in vitro activation (Fig. 4h). Overall, 211 our data reveal that in the absence of TGF-, SMAD4 restricts transcriptional and 212 functional TGF- signature in CD8 T cells and endows naïve CD8 T cells with an effector 213 program.

215
Since SMAD4 deletion limits T cell activation, a compensatory mechanism must 216 allow microbiota-driven activation of CD8 T cells in the gut. Intriguingly, genes encoding 217 potent TGF- signaling repressors (e.g. Smad7, Ski, Skil and Smurf2) were enhanced in 218 SKO and R2SKO compared to R2KO CD8 T cells (Fig. 5a). We validated the 219 overexpression of those genes by real-time quantitative RT-PCR on naïve F5 CD8 T cells, 220 as well as on polyclonal CD8 T cells (Fig. 5b, Supplementary information, Fig.5a), 221 attesting that this overexpression was not restricted to a specific T cell receptor (TCR) 222 repertoire. The expression defect of TGF- repressors in R2KO CD8 T cells confirmed 223 that they are TGF- target genes 8,9,26 . Since the double deletion of TGF-RII and SMAD4 224 (R2SKO) restored the gene expression of TGF- repressors (Fig. 5b, Supplementary   225 information, Fig.5a), this demonstrated definitively that SMAD4 inhibits the expression 226 of TGF- repressors in a TGF--independent manner.

227
Since those genes are potent repressors of TGF-β signaling and have been 228 associated with a defect of T cell response to TGF- in IBDs 27 , we examined SMAD4-229 deficient CD8 T cell TGF--response. While TGF- inhibited impressively GZMB and 230 TBET expression in activated WT CD8 T cells even at low doses, impressively, their 231 expression was maintained even at high concentrations of TGF-in SKO CD8 T cells 232 ( Fig. 5c-d). Thus, these observations strongly reveal that SMAD4 ablation totally limits the immune-regulatory effects of TGF- on CD8 T cells. Importantly, this demonstrates 234 that SMAD4 is crucial for TGF--mediated immunosuppression and is not redundant.

235
Because TGF- is highly enriched in the gut 3 and represses T cell activation 28 , this 236 impaired response to TGF- could contribute to the chronic microbiota-driven CD8 T cell 237 activation. In order to confirm this assumption in vivo, we forced the activation of SMAD4 238 independent pathways of TGF- signaling by crossing SKO mice with mice bearing a 239 conditionally-expressed, constitutively-active form of the TGF-βR1 (LSL-TGFRICA 240 mouse strain) 29 . In the resulting SKO-RCA mice animals, CD8 T cells were as 241 abundant and activated in the gut epithelium as in SKO mice ( Fig. 5f-g), and more 242 importantly, SKO-RCA mice developed IBDs (Fig. 5h-i). Hence, the remaining TGF- 243 signaling pathways are unable to compensate for SMAD4 loss. Collectively, these data 244 suggest that the TGF--independent function of SMAD4 facilitates the response of CD8 245 T cells to TGF-, by restraining the expression of a wide range of TGF- repressors in a 246 feedforward mechanism (prior to any TGF- signal) and this is crucial and non-redundant 247 to mediate immune-regulatory-effect of TGF-. STAT5 phosphorylation, which is induced upon IL-7 stimulation, was slightly enhanced in SKO and R2SKO CD8 T cells, and impaired in R2KO CD8 T cells (Fig. 6b). A time-course 264 analysis of survival demonstrated that IL-7 did not prevent R2KO CD8 T cells from dying 265 in vitro compared to SKO and R2SKO CD8T cells that survived largely better (Fig. 6c). 266 Accordingly, we observed a substantial increase in the absolute number and the 267 proportion of CD8 T cells in secondary lymphoid organs from SKO and R2SKO F5 268 transgenic mice, unlike R2KO mice ( Fig. 6d and data not shown). These findings reveal 269 a critical role for the TGF--independent SMAD4 function in restraining CD8T cell 270 accumulation by repressing the IL-7 response, in sharp contrast to TGF- signaling. with a blocking antibody that specifically recognizes CD103 (Fig. 6f). The CD103 blockade 283 led to a decrease in CD8 T cell numbers within the intestinal epithelium of SKO mice, 284 without altering their accumulation in secondary lymphoid organs such as the spleen and 285 mesenteric lymph nodes (Fig. 6g). Although this treatment did not fully restore body 286 weight in SKO, the colon length and immune-histology analysis highlighted clear 287 improvement (Fig. 6h-j). The colon length reduction and the mucosal damage due to 288 immune infiltration were alleviated, indicating a beneficial effect of CD103 blockade in 289 SKO mice. Globally, in addition to the impaired response to TGF-β immune-regulatory Focusing on genes that were differentially expressed between WT and SKO cells and 320 between R2KO and SKO cells, we found 103 genes. Among those 103 genes, we found 321 genes implicated in CD8 T cell differentiation such as Tcf4 and Lef1 but also many well-322 characterized TGF--target genes. Importantly, we found TGF-β-repressors (Smad7, 7e, Supplementary information, Fig. 7c). Thus, SMAD4, directly by acting at the 325 chromatin level could restrict TGF- target gene expression. To identify putative partners 326 of SMAD4 in WT and R2KO CD8 T cells, we conducted an enrichment motif analysis and 327 uncovered similar motifs in the top 3 enriched motifs, notably ETS and RUNX family 328 domains, indicating a potential interaction between SMAD4 and ETS or RUNX 329 transcription factor families (Fig. 7f, 7g). Thus, revealing that SMAD4 interacts likely with 330 different partners to mediate its wide transcriptional impact in CD8 T cells. Our study uncovers an uncharacterized critical feedforward regulation of the TGF-347  effect governed by SMAD4 in CD8 T cells in a TGF- independent manner, crucial to 348 prevent chronic intestinal inflammation. Indeed, we reveal that the TGF-b-independent 349 function of SMAD4 acts as a basal and active repressor of a myriad of TGF- target genes, 350 restraining the TGF- signature in CD8 T cells in the absence of any TGF- signaling.