T-bet deficiency and Hic1 induction override TGF-β-dependency in the formation of CD103⁺ intestine-resident memory CD8⁺ T cells

SUMMARY

Transforming growth factor β (TGF-β) represents a well-established signal required for tissue-resident memory T cell (T_RM) formation at intestinal surfaces, regulating the expression of a large collection of genes coordinately promoting intestinal T_RM differentiation. The functional contribution from each TGF-β-controlled transcription factor is not entirely known. Here, we find that TGF-β-induced T-bet downregulation and Hic1 induction represent two critical events during intestinal T_RM differentiation. Importantly, T-bet deficiency significantly rescues intestinal T_RM formation in the absence of the TGF-β receptor. Hic1 induction further strengthens T_RM maturation in the absence of TGF-β and T-bet. Our results reveal that provision of certain TGF-β-induced molecular events can partially replace TGF-β signaling to promote the establishment of intestinal T_RMs, which allows the functional dissection of TGF-β-induced transcriptional targets and molecular mechanisms for T_RM differentiation.

In brief

Wang et al. find that T-bet deficiency significantly rescues intestine T_RM differentiation in TGF-βR-KO cells. Suppressing the type 17 program or enforced expression of Hic1 further boosts T_RM formation in T-bet/TGF-βR DKO T cells. These results show the key molecular events downstream of TGF-β signaling during intestine T_RM differentiation.

Graphical abstract

INTRODUCTION

Tissue-resident memory T cells (T_RMs) are one of the key adaptive immune components and the first line of defense in mucosal tissues.^1–3 To form mucosal T_RMs, a subset of effector T cells migrate to the mucosal surface, receive local signals, and initiate a unique differentiation program. Mucosal T_RMs often express the surface markers CD69 and CD103. It is generally accepted that transforming growth factor β (TGF-β) is required for the establishment of the mucosal T_RM population, especially for CD103⁺ T_RMs, since CD103 (encoded by Itgae) is a direct target of TGF-β signaling in CD8⁺ T cells.^4–9 However, the function of individual molecular targets downstream of TGF-β signaling is less clear during T_RM establishment.

T-box transcription factors T-bet (encoded by Tbx21) and Eomes (encoded by Eomes) play essential roles in effector and memory CD8⁺ T cell differentiation.^10,11 It has been demonstrated that TGF-β signaling downregulates the expression of T-bet and Eomes during skin T_RM differentiation¹² and forced expression of either T-bet or Eomes significantly suppresses skin T_RM formation. Downregulation of T-box transcription factors leads to enhanced TGF-β receptor expression, which results in a feedforward loop to reinforce TGF-β-dependent T_RM differentiation.¹² Considering the fact that Eomes cannot replace T-bet in effector CD8⁺ T cells,¹³ it remains unknown whether T-bet and Eomes play equivalent roles in T_RM differentiation. Importantly, mature T_RMs almost completely turn off the expression of Eomes while carrying a low level of T-bet expression to maintain responsiveness to interleukin (IL)-15.¹² However, IL-15 dependency is a tissue-specific feature for T_RMs. In contrast to IL-15-dependent skin T_RMs, intestinal T_RMs are IL-15 independent.¹⁴ Whether TGF-β-dependent quenching of T-box transcription factors is the central hub of the TGF-β-controlled T_RM differentiation program and whether the same mechanisms are operating at the intestinal surface remain elusive. Recently, it was shown that the transcriptional repressor Hic1 acts as a specific regulator of intestinal T_RM differentiation.¹⁵ However, the functional interaction between Hic1 and other T_RM-promoting factors is not entirely clear.

Here, using genetic models, we show that T-bet deficiency, but not Eomes deficiency, significantly rescues the differentiation of CD103⁺ intestinal T_RMs in the absence of TGF-β receptor. T-bet deficiency allows the induction of the tissue-residency program at both transcriptional and epigenetic levels in TGF-β-receptor-deficient cells. T-bet deficiency induces a type 17 program in T_RMs. Suppressing the type 17 program further boosts T_RM differentiation. Interestingly, T-bet deficiency cannot fully restore Hic1-controlled intestinal T_RM differentiation. Forced expression of Hic1 further enhances the formation of CD103⁺ T_RM in the absence of both T-bet and TGF-β receptor. In contrast, forced induction of Hic1 in TGF-β-receptor-deficient cells only improves the differentiation of CD69⁺CD103⁻, but not CD103⁺, T_RMs at the intestinal surface. Together, our genetic models have revealed the function of essential events in the TGF-β-induced intestine T_RM differentiation program.

RESULTS

TGF-β signaling downregulates the expression of T-bet and Eomes during gut T_RM differentiation

To dissect the components of the TGF-β-induced T_RM differentiation program, we first focused on T-box transcription factors T-bet and Eomes. We employed lymphocytic choriomeningitis virus (LCMV) acute infection model and TGF-β receptor II conditional-knockout (KO) (Tgfbr2^f/f distal Lck-Cre,¹⁶ hereafter referred to as Tgfbr2^−/−) P14 TCR transgenic mice, which carried CD8⁺ T cells specific for LCMV epitope H-2D^b-GP_33–41. As illustrated in Figure S1A, naive P14 T cells carrying distinct congenic markers were isolated from wild-type (WT) control (CD45.1/1) and Tgfbr2^−/− mice (CD45.1/2), mixed at a 1:1 ratio and adoptively co-transferred into unmanipulated sex-matched C57BL/6 (B6, CD45.2/2) recipient mice followed by LCMV Armstrong infection. In this system, WT and Tgfbr2^−/− P14 T cells were compared side by side in the same WT environment. Consistent with published results for T_RM isolated from the skin, we observed TGF-β-dependent downregulation of both T-bet and Eomes during gut T_RM differentiation (Figures S1B–S1E). Importantly, the downregulation of both T-bet and Eomes occurred before the induction of CD103 expression (Figures S1B and S1C). Further, during in vitro T cell activation, TGF-β inhibited the expression of both T-bet and Eomes in purified WT CD8⁺ T cells (Figure S1F). Thus, TGF-β-controlled downregulation of T-box transcription factors represents an early event for intestinal T_RM differentiation.

T-bet and Eomes deficiency alters circulating effector and memory CD8⁺ T cells

To directly address the question of whether failed downregulation of T-bet and/or Eomes in Tgfbr2^−/− cells is responsible for defective T_RM formation, we generated double-conditional-KO mouse strains for TGF-βR II and T-bet (i.e., Tgfbr2^f/fTbx21^f/f dLck-Cre, hereafter referred to as Tgfbr2^−/−Tbx21^−/−) as well as TGF-βR II and Eomes (i.e., Tgfbr2^f/fEomes^f/f dLck-Cre, simplified as Tgfbr2^−/−Eomes^−/−). Furthermore, we bred all double-and single-KO strains with P14 TCR transgenic mice carrying congenic markers so that we could perform the same co-transfer experiments as in Figure S1A to carry out a side-by-side comparison of virus-specific CD8⁺ T cells with different genetic manipulation in a WT environment.

As illustrated in Figures S2A and S3A, naive P14 T cells isolated from WT plus one of the single- and double-KO mice were mixed at a 1:1 ratio and co-transferred into B6 recipients followed by LCMV arm infection. First, we examined circulating CD8⁺ T cells isolated from the spleen. Consistent with previous publications, Tgfbr2^−/− CD8⁺ T cells exhibited a slight reduction of initial expansion and increased KLRG1⁺ subset.⁶ Interestingly, Eomes deficiency largely corrected these defects as Tgfbr2^−/−Eomes^−/− P14 T cells exhibited expansion and KLRG1 expression comparable to co-transferred WT controls (Figures S2B, S2C, and S2E). In contrast, T-bet deficiency severely impaired the initial expansion of effector P14 T cells as seen in both T-bet single-KO (Figure S3B) and Tgfbr2^−/−Tbx21^−/− cells (Figure S2B). As expected, T-bet deficiency almost completely abolished KLRG1⁺ subset (Figures S2D, S2E, and S3C).^10,17,18 Using a different set of markers (CD62L and CD127) to define circulating memory T cells, T-bet-single and Tgfbr2^−/−Tbx21^−/− cells exhibited similarly enhanced T_EM population and greatly reduced terminal T_EM (T-T_EM) subset (Figure S2F).¹⁹ Further, expression of the well-established T-bet target gene, CXCR3, was severely defective for Tgfbr2^−/−Tbx21^−/− cells (Figure S2D), validating our genetic models. Together with certain expected defects, the double-KO P14 T cells differentiated into circulating effector and memory T cells.

T-bet deficiency, but not Eomes deficiency, corrects gut T_RM differentiation in the absence of TGF-βR

Next, we focused on P14 T cells isolated from the small intestine. When T-bet-single or Eomes-single conditional-KO cells were examined, enhanced induction of T_RM markers CD69 and CD103 was observed at early time points (e.g., day 5 and day 7) (Figures S3D and S3E). However, at memory phase (≥d30), the phenotypic difference between WT and Tbx21^−/− or WT and Eomes^−/− cells was either subtle or not significant (Figures S3D and S3E). Thus, we concluded that T-bet or Eomes deficiency accelerated gut T_RM differentiation, consistent with the observation in skin T_RM differentiation.¹² Interestingly, we often detected more dramatic changes in T-bet KO compared with Eomes KO (Figures S3B, S3D, and S3E).

Next, we examined whether T-bet or Eomes deficiency could overcome the blockade of intestinal T_RM differentiation in the absence of TGF-βR. When comparing Tgfbr2^−/− vs. Tgfbr2^−/−Eomes^−/− gut T_RM differentiation, we did not detect any significant difference in either the total population size or the induction of CD69 and CD103 in small intestine intraepithelial lymphocyte (SI-IEL) compartment (Figures 1A, 1B, 1D, and 1E). In contrast, when examining Tgfbr2^−/−Tbx21^−/− vs. Tgfbr2^−/− cells, there was a clear and significant rescue of the population size as well as phenotypic markers of gut T_RM (Figures 1A, 1C, 1D, and 1E). We could detect a significant rescue of gut T_RM differentiation as early as day 7 post infection and a gradual increase in the proportion of CD69⁺CD103⁺ subset in Tgfbr2^−/−Tbx21^−/− cells (Figures 1C–1E). Together, we conclude that T-bet deficiency, but not Eomes deficiency, partially rescues SI-IEL T_RM differentiation in the absence of TGF-β receptor.

Figure 1. T-bet deficiency, but not Eomes deficiency, rescues gut-resident memory T cell differentiation in the absence of TGF-β receptor.

Same experimental setup as in Figure S2A.

(A) The percentage of donor P14 T cells in the total CD45⁺ cell population isolated from SI-IEL is shown (n = 8–37 individual recipient mice for each time point). Day 30 results are presented as a bar graph. (B and C) Representative FACS profiles of pre-gated donor P14 T cells isolated from SI-IEL are shown. (D and E) (D) The percentage of CD69⁺ and (E) CD103⁺ cells in donor P14 T cells isolated from SI-IEL are shown (n = 5–30). Mean ± SEM is shown for each data point in (A), (D), and (E). Pooled results from three to six independent experiments are shown in (A), (D), and (E). N.S., not significant; ***p < 0.001; and ****p < 0.0001 by ordinary one-way ANOVA with multiple-comparison post test for the last time point.

Next, we performed a side-by-side comparison of WT, Tgfbr2^−/−, Tgfbr2^−/−Tbx21^−/−, and Tbx21^−/− P14 T cell in the same experiments and expanded the analysis to include lamina propria (LP) and Peyer’s patches (PPs). As expected, Tgfbr2^−/− cells exhibited severe defects in T_RM differentiation in all intestinal tissues (Figures 2D and 2G). Tbx21^−/− T_RMs showed similar or slightly increased frequency and number of CD69⁺CD103⁺ cells compared with WT controls at memory time points (Figures 2B–2H). Tgfbr2^−/−Tbx21^−/− exhibited a partial, but significant, rescue of both population size and surface markers of gut T_RMs, including CD69 and CD103, but not CD49a and CXCR3 (Figures 2B–2H). Together, T-bet deficiency, but not Eomes deficiency, significantly rescues intestinal T_RM differentiation in the absence of TGF-β receptor.

Figure 2. T-bet deficiency partially overcomes the differentiation block in Tgfbr2^−/− cells.

Similar experimental setup as in Figure S2A.

(A) Day 7 post infection, the expression of T-bet in SI-IEL P14 T cells was measured by flow cytometry (n = 5).

(B–H) (B–E) Day 25–30 post infection, (F) to (H) day 45–60 post infection. (B and F) Representative FACS profiles of pre-gated donor P14 T cells isolated from SI-IEL are shown. (C) MFI of CD49a and CXCR3 on pre-gated SI-IEL P14 T cells are shown (n = 4–14). (D and G) The percentage of donor P14 T cells in total CD8 is shown. (E and H) The percentages of CD103⁺ (left) and CD69⁺ (right) in donor P14 T cells are shown (n = 9–57 for D and E; n = 4–29 for G and H). Each symbol represents the results of an individual mouse. Mean ± SEM is shown. Pooled results from two to six independent experiments are shown. N.S., not significant; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001 by ordinary one-way ANOVA with multiple-comparison post test or Student t test.

T-bet deficiency allows TGF-β-independent differentiation of gut T_RMs

With this surprising rescue phenotype, we would like to validate the deletion efficiency first to rule out the possibility that some cells escaped Cre-mediated deletion in the intestinal T_RM compartment. First, we examined Tbx21 locus. When gated on IEL P14 T cells, the expression of T-bet was dramatically reduced in both Tbx21^−/− and Tgfbr2^−/−Tbx21^−/− cells at protein level (Figure 2A). Further, when sorted Tgfbr2^−/−Tbx21^−/− IEL P14 T cells were subjected to RNA sequencing (RNA-seq) analysis, the expression of Tbx21 was almost completely abolished (Figure S4E and housekeeping control in Figure S4G). Next, we focused on Tgfbr2 locus. To this end, we first compared the expression of CD103 on naive CD8⁺ T cells in uninfected animals. WT and Tbx21^−/− CD8⁺ T cells expressed comparable levels of CD103. In contrast, both Tgfbr2^−/− and Tgfbr2^−/−Tbx21^−/− CD8⁺ T cells were almost completely devoid of CD103 expression (Figure S4A). Further, we employed an in vitro culture system. Briefly, naive P14 T cells were isolated from WT, Tbx21^−/−, Tgfbr2^−/−, and Tgfbr2^−/−Tbx21^−/− mice, activated by TCR/CD28 stimulation and cultured in IL-2 with added TGF-β or TGF-β neutralizing antibody. Four days later, only WT and Tbx21^−/− P14 T cells expressed CD103 in a TGF-β-dependent manner, while Tgfbr2^−/− and Tgfbr2^−/−Tbx21^−/− cells did not induce CD103 expression (Figure S4B). To further validate the findings, we used the well-established in vivo priming + ex vivo culture system (Figure S4C).^20,21 As shown in Figure S4D, in vivo-primed WT P14 T cells responded to TGF-β and induced CD103 expression. Tbx21^−/− cells exhibited greatly enhanced response to TGF-β, which is consistent with previous findings.²² In contrast, Tgfbr2^−/− and Tgfbr2^−/−Tbx21^−/− cells failed to upregulate CD103 in this setting. Lastly, when sorted Tgfbr2^−/−Tbx21^−/− IEL P14 T cells were examined by RNA-seq analysis, a specific loss of transcripts starting from exon 5 was detected (Figure S4F), which was the floxed exon in our conditional-KO model. Based on this set of experiments, we concluded that there was no escaped deletion in either Tbx21 or Tgfbr2 loci. T-bet deficiency allows TGF-β-independent differentiation of gut T_RMs after arrival at local intestinal tissues.

T-bet deficiency cannot rescue Tgfbr2^−/− T_RM formation in the kidney and salivary glands

To determine whether the rescue phenotype is intestine specific or a general phenomenon, we examined kidney and salivary glands (SGs). As we and others have published,^23,24 CD8⁺ T_RM differentiation and maturation in both kidney and SG are TGF-β-dependent. Kidney T_RMs represent non-barrier tissue T_RMs and often lack the expression of CD103. In contrast, SG T_RMs reside in an IEL compartment and carry CD103 expression. Using the same LCMV infection system, we found that the population and phenotype of Tgfbr2^−/−Tbx21^−/− T_RMs were indistinguishable from Tgfbr2^−/− ones in both kidney and SG (Figures S2G and S2H). Thus, T-bet deficiency rescues Tgfbr2^−/− T_RM formation in an intestine-specific manner.

Altered effector program in Tgfbr2^−/−Tbx21^−/− T_RMs

Because T-box transcription factors have well-established roles in the CD8 effector program, we addressed whether Tgfbr2^−/−Tbx21^−/− gut T_RMs were functional. For this purpose, WT plus Tgfbr2^−/−Tbx21^−/− P14 T cells or WT plus single-KO controls were co-transferred into WT recipients followed by LCMV infection (Figure 3A). Eomes^−/− memory T cells exhibited similar effector functions to co-transferred WT controls in both spleen and SI-IELs (Figures 3B–3D). Compared to WT controls, both Tbx21^−/− and Tgfbr2^−/−Tbx21^−/− memory T cells produced reduced levels of interferon (IFN)-γ and granzyme A in the spleen (Figures 3B, 3F, 3H, and S5A). However, in the SI-IEL compartment, T-bet deficiency had minimal impact on IFN-γ and granzyme A production (Figures 3B, 3G, 3H, and S5B). Both Tbx21^−/− and Tgfbr2^−/−Tbx21^−/− memory T cells produced similar levels of tumor necrosis factor (TNF) and significantly increased amounts of IL-2 and IL-17 in both spleen and SI-IEL compartments (Figures 3C–3H). Together, Tbx21^−/− and Tgfbr2^−/−Tbx21^−/− memory CD8⁺ T cells exhibit enhanced type 17 response in both circulating memory T cells and T_RMs. In contrast, T-bet was only required for the optimal type 1 effector program in circulating memory T cells but not in small intestine T_RMs.

Figure 3. Tgfbr2^−/−Tbx21^−/− gut-resident memory T cells exhibit an altered effector program.

(A) Schematics. Naive P14 isolated from WT and one of the KOs (gray Tbx21^−/−, green Eomes^−/−, and blue Tgfbr2^−/−Tbx21^−/−) were co-transferred into B6 recipients followed by LCMV infection.

(B–H) Day 30–32 (B–E) or day 45 post infection (F–H), the percentage of cytokine-producing P14 T cells are shown in (B)–(E) and (H). Representative FACS profiles of pre-gated donor P14 T cells isolated from the spleen (F) and SI-IEL (G) are shown. Each symbol in(B)–(E) and (H) represents the results from an individual recipient mouse (n = 5–13). Mean ± SEM is shown. N.S., not significant; *p < 0.05; ***p < 0.001; and ****p < 0.0001 by ordinary one-way ANOVA with multiple-comparison post test.

Type 17 program inhibits gut T_RM differentiation in the absence of T-bet

It has been reported that T-bet deficiency promotes a RORγ-dependent type 17 response in CD8⁺ T_RMs.^12,18 Indeed, we could detect a significant portion of P14 T cells producing IL-17 in the absence of T-bet (Figure 3). This finding was not T_RM specific as IL-17⁺ cells were present in both spleen and IEL. To determine whether this type 17 program was involved in T_RM differentiation in the absence of T-bet, we first examined the expression of RORγ in gut T_RMs. Compared with WT controls, both Tbx21^−/− and Tgfbr2^−/−Tbx21^−/− T_RMs carried significantly increased expression of RORγ (Figures 4A and 4B). Interestingly, in Tgfbr2^−/−Tbx21^−/− cells, the expression of RORγ was greatly enhanced in CD103⁻ subset compared with CD103⁺ counterpart (Figures 4A and 4B), suggesting that RORγ may not be required for or may suppress CD103⁺ T_RMs in this setting. To directly address whether RORγ was involved in Tbx21^−/− gut T_RM differentiation, we generated T-bet and RORγ double-KO mice carrying P14 TCR transgene (i.e., Rorc^−/−Tbx21^−/−). As illustrated in Figure 4C, WT plus Tbx21^−/− or WT plus Rorc^−/−Tbx21^−/− P14 T cells were adoptively co-transferred into WT recipients followed by LCMV infection. As expected, Rorc^−/−Tbx21^−/− cells completely lost RORγ expression (Figure 4D). However, we could not detect major defects in gut T_RM differentiation for Rorc^−/−Tbx21^−/− cells (Figure 4E). Indeed, there was a slight but significant increase of CD103 expression in Rorc^−/−Tbx21^−/− cells (Figure 4E).

Figure 4. Type 17 program suppresses gut T_RM in the absence of T-bet.

Similar experimental setup as in Figure S2A.

(A) Representative FACS profiles of pre-gated donor P14 T cells isolated from SI-IEL are shown.

(B) The percentage of RORγ⁺ cells in each subset of SI-IEL P14 T cells at day 30 post infection (n = 9–10).

(C) Schematics for the experiments shown in (D) and (E).

(D) Representative FACS profiles of pre-gated SI-IEL P14 T cells are shown.

(E) The percentage of CD69⁺ (left, n = 5–10) and CD103⁺ (right, n = 8–18) cells in donor P14 T cells isolated from SI-IEL at day 30 post infection are shown.

(F) Schematics for the experiments shown in (G)–(I).

(G) The ratio of Cas9-mediated KO P14s over co-transferred control P14s is shown. Left, Rorc-sgRNA; right, Rora-sgRNA (n = 5).

(H) Representative FACS plots to show the deletion efficiency of Rorc in IEL P14 T cells.

(I) The percentage of CD103⁺ cells in gut P14 T cells are shown (n = 5). Each symbol in (B), (E), (G), and (I) represents the results from an individual mouse. Mean ± SEM is shown. Pooled results from two independent experiments are shown. N.S., not significant; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001 by ordinary one-way ANOVA with multiple-comparison post test.

Next, to directly test whether the type 17 program is required for Tgfbr2^−/−Tbx21^−/− T_RM formation, we employed a CRISPR-Cas9-based system.²⁵ As shown in Figure 4F, naive P14 T cells were isolated from two congenically distinct Tgfbr2^−/−Tbx21^−/− mice. A pre-made control single guide RNA (sgRNA)/Cas9 complex was delivered into CD45.1/1 P14, while Rorc- or Rora-targeting sgRNA/Cas9 was delivered into CD45.1/2 cells. After electroporation-mediated sgRNA/Cas9 delivery, control- and targeting-sgRNA/Cas9-treated naive P14 T cells were 1:1 mixed, adoptively transferred into B6 recipients, followed by LCMV infection. To be noted, in this system, no in vitro T cell activation is required for sgRNA/Cas9 delivery. P14 T cells are primed in vivo following LCMV infection. Further, we could achieve high targeting efficiency in intestinal T_RMs (Figure 4H). Four weeks post infection, we observed significant enrichment of Rorc^−/−Tgfbr2^−/−Tbx21^−/− and Rora^−/−Tgfbr2^−/−Tbx21^−/− P14s over the co-transferred control sgRNA-treated Tgfbr2^−/−Tbx21^−/− ones in the SI-IEL compartment (Figure 4G). Both Rorc and Rora deletion enhanced CD103 expression in Tgfbr2^−/−Tbx21^−/− T_RMs in the small intestine (Figure 4I). In summary, although T-bet deficiency leads to enhanced type 17 program, Rorc/Rora-controlled type 17 program suppresses CD103⁺ T_RM formation in the small intestine.

T-bet deficiency partially rescues gut Tgfbr2^−/− T_RM differentiation at the transcriptional level

To further characterize Tgfbr2^−/−Tbx21^−/− T_RM, we determined their transcriptional profiles. Briefly, using fluorescence-activated cell sorting (FACS), we sorted different subsets of P14 T cells isolated from SI-IEL compartment together with a WT splenic P14 T cell subset (i.e., KLRG1⁻) as a circulating memory T cell control. All sorted cells were subjected to bulk RNA-seq analysis. Using principal-component analysis, circulating memory T cells were separated from all SI-IEL subsets based on PC1, accounting for 57% of variance (Figure 5A). Along PC1, WT CD103⁺ and Tbx21^−/− CD103⁺ IEL cells were similarly positioned, while Tgfbr2^−/−Tbx21^−/− CD103⁺ ones were situated between T_RMs and circulating controls (Figure 5A). We did observe difference along PC2 (19% variance) between different CD103⁺ IEL subsets (Figure 5A). When focused on established circulating and resident gene signatures, Tgfbr2^−/−Tbx21^−/− IELs carried a gene set variation analysis (GSVA) score between T_RMs (including both WT and Tbx21^−/− IELs) and circulating T cells (Figure 5B). This finding was further validated by unsupervised clustering in heatmaps focusing on T_Cir and T_RM signature genes (Figure S6). When performing gene set enrichment analysis (GSEA) to compare different IEL subsets vs. WT splenic T cells for T_Cir and T_RM signatures, all CD103⁺ subsets (including WT, Tbx21^−/−, and Tgfbr2^−/−Tbx21^−/−) were positively enriched for T_RM signature and negatively enriched for the T_Cir one (Figure 5C top row). When comparing CD69⁺CD103⁻ IEL subsets and splenic T cells, we observed similar T_RM-like enrichment for WT and Tgfbr2^−/−Tbx21^−/− cells (Figure 5C bottom row). Interestingly, Tgfbr2^−/− CD69⁺CD103⁻ IEL subsets were also positively enriched for T_RM signature and trending negatively enriched for the T_Cir one (Figure 5C, bottom right). When the Tgfbr2^−/− CD69⁺CD103⁻ subset was compared with its WT counterpart, we detected significant enrichment of T_Cir in Tgfbr2^−/− cells, while the T_RM signature was comparable (Figure 5D left). When Tgfbr2^−/− CD69⁺CD103⁻ cells were compared with WT CD103⁺ T_RM, WT T_RMs were positively enriched for T_RM and negatively enriched for T_Cir signature (Figure 5D right). These results demonstrate that Tgfbr2^−/− cells can initiate but cannot complete the T_RM differentiation program. Importantly, both Tgfbr2^−/−Tbx21^−/− CD69⁺CD103⁺ and CD69⁺CD103⁻ cells exhibited a T_RM GSEA pattern (Figure 5C).

Figure 5. T-bet deficiency partially overcomes the transcriptional and epigenetic blocks in IEL T_RM differentiation in the absence of TGF-β signaling.

Various subsets of FACS-sorted P14 T cells were subjected to bulk RNA-seq and ATAC-seq analysis.

(A) Principal-component analysis (PCA) of RNA-seq results is shown.

(B) GSVA scores for T_RM signature (left) and T_Cir signature (right) are calculated based on RNA-seq results. Each symbol represents a biologically independent replicate.

(C and D) GSEA for circulating T cell signature genes and SI-IEL T_RM signature genes.

(E) ATAC-seq results for representative T_RM-related loci are shown.

When comparing differentially expressed genes (DEGs) between various T_RM subsets and circulating controls, we identified seven clusters of genes (Figure S7A). The common cluster shared by WT, Tbx21^−/−, and Tgfbr2^−/−Tbx21^−/− T_RMs was highly enriched for the biological processes related to the digestion system and actin-based cell projections (Figures S7B, C5). This finding supports the tissue specificity of this rescue phenotype demonstrated in Figures S2G and S2H. Together, consistent with phenotypic characterization, T-bet deficiency partially rescues the differentiation of intestinal T_RMs in the absence of TGF-β receptor at the transcriptional level.

T-bet-controlled Tcf-1 expression is not involved in intestinal T_RM formation in Tgfbr2^−/−Tbx21^−/− cells

In addition to the type 17 program, we were interested in other transcriptional programs induced by T-bet deficiency. Interestingly, we detected significantly increased expression of Tcf7. Using flow cytometry, we validated that, for both spleen memory T cells and IEL T_RMs, the expression of Tcf-1 (encoded by Tcf7) was significantly enhanced in the absence of T-bet (Figure S8A). To directly test whether the induction of Tcf-1 was responsible for T-bet deficiency-mediated T_RM rescue, we employed a similar CRISPR-Cas9 system to Figure 4F. Briefly, naive P14 T cells were isolated from two congenically distinct Tgfbr2^−/−Tbx21^−/− mice. One population of P14 T cells received control sgRNA/Cas9 and the other received Tcf7-targeting sgRNA/Cas9. Treated P14 T cells were mixed at a 1:1 ratio and adoptively co-transferred into B6 recipients followed by LCMV infection (Figure S8B). As shown in Figure S8C, sgRNA/Cas9-mediated deletion almost completely abolished Tcf-1 induction in the IEL compartment. However, no major differences were detected in gut T_RM differentiation except for a slight increase of CD103 expression at an early time point (Figures S8D and S8E). Together, the induction of Tcf-1 is not apparently required for T_RM formation in Tgfbr2^−/−Tbx21^−/− cells.

T-bet deficiency partially rescues gut Tgfbr2^−/− T_RM differentiation at the epigenetic level

Next, we examined whether T-bet deficiency had any impacts at the epigenetic level. To this end, P14 subsets were FACS sorted from SI-IEL compartment and subjected to assay for transposase-accessible chromatin with sequencing (ATAC-seq) analysis. When focusing on the transcription start site (TSS) region of all T_RM signature genes, we could detect significant defects in Tgfbr2^−/− cells, which were largely corrected in Tgfbr2^−/−Tbx21^−/− cells (Figure S9A dark red arrow). Further, motif enrichment analysis was largely consistent with our previous analysis, i.e., reduced T-bet, enhanced Tcf-1/Lef1 motif, and increased ROR-γ motif enrichment in the cells lacking T-bet, and decreased Smad4 motif enrichment in the cells lacking TGF-β receptor (Figure S9B). Narrowing down to individual genes, we identified several categories of gene loci. First, in multiple genomic regions harboring residency-related genes, Tgfbr2^−/−Tbx21^−/− CD69⁺CD103⁺ cells exhibited an intermediate phenotype between Tgfbr2^−/− CD69⁻ non-T_RMs and T_RMs (including both WT and Tbx21^−/− CD69⁺CD103⁺). This category included Itgae, Rgs1, Runx3, Litaf, Cdh1, and Xcl1 (Figures 5E and S9F). Second, for some genomic regions, Tgfbr2^−/−Tbx21^−/− CD69⁺CD103⁺ cells exhibited a similar phenotype to T_RMs (including both WT and Tbx21^−/− CD69⁺CD103⁺) and distinct from Tgfbr2^−/− CD69⁻ non-T_RMs. This category included S1pr5 (5′ region of the promoter), Eomes, Klrg1 (promoter region), S1pr1, Sell, and Cd69 (Figures 5E, S9C, S9E, and S9F). Finally, there were some regions exhibiting a T-bet-dependent pattern; i.e., Tgfbr2^−/− CD69⁻ were similar to WT CD69⁺CD103⁺ cells, while Tbx21^−/− CD69⁺CD103⁺ were similar to Tgfbr2^−/−Tbx21^−/− ones. The last category included the promoter region of S1pr5, Tbx21, Zeb2, distal region of Klrg1, and type 17-related genes Il17f and Ccr6 (Figures 5E, S9C, S9D, and S9E). Together, our ATAC-seq results largely support the conclusion that T-bet deficiency partially rescues Tgfbr2^−/− intestinal T_RM differentiation at an epigenetic level.

Hic1 further boosts CD103⁺ T_RM formation in the absence of T-bet

Since T-bet deficiency only partially overcame TGF-β dependency in gut T_RM differentiation, we wanted to identify key transcription regulator(s) missing in Tgfbr2^−/−Tbx21^−/− cells. Transcription repressor Hic1 has recently been demonstrated to be a key regulator for T_RM differentiation in an intestine-specific manner.¹⁵ We were curious whether Hic1 induction represented another downstream event of TGF-β signaling and a missing factor for Tgfbr2^−/−Tbx21^−/− T_RM formation.

First, we confirmed that Hic1 was induced by TGF-β during CD8⁺ T cell activation in vitro (Figure 6A). Next, we measured the expression of Hic1 at the protein level. Circulating T cells expressed minimal levels of Hic1, while IEL T_RMs exhibited a dramatic induction of Hic1 in a TGF-β-dependent manner (Figure 6B). T-bet deficiency partially rescued the defective expression of Hic1 in Tgfbr2^−/− cells. Interestingly, compared with WT controls, Tbx21^−/− T_RMs expressed slightly but significantly increased levels of Hic1 (Figure 6B). Together, optimal Hic1 expression requires both TGF-β signal and T-bet downregulation.

Figure 6. Hic1 OE and T-bet downregulation cooperate to mediate TGF-β-induced T_RM differentiation program.

(A) Purified WT naive CD8⁺ T cells were activated in vitro with TGF-β-neutralizing antibody or added TGF-β. Hic1 expression was measured by bulk RNA-seq (n = 4 independent replicates).

(B) Similar setup as in Figure S2A. Day 20 post infection, Hic1 expression in donor P14 T cells was measured by FACS (n = 5).

(C–K) (C and D) d13 and (I) d14 post infection, representative FACS of pre-gated IEL P14 subsets are shown. Numbers in (B) and (D) represent MFI. (E and J) The percentage of CD69⁺ and (F and K) the percentage of CD103⁺ cells in each IEL P14 subset are shown (n = 6–9 for E and F; n = 4–6 for J and K). D27–30 post infection, the relative population size of each IEL P14 subset (G) and the percentage of each donor subset in total CD45⁺ cells (H) are shown (n = 3–6). Each symbol and each pair of symbols represent the results from an individual recipient mouse. Mean ± SEM is shown. Pool results from two or three independent experiments are shown for each setting. N.S., not significant; *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001 by ordinary one-way ANOVA with multiple-comparison post test or Student t test.

Subsequently, we would like to test whether reduced Hic1 expression in Tgfbr2^−/−Tbx21^−/− cells was responsible for their suboptimal differentiation compared with WT controls. To this end, WT P14 T cells were transduced with an empty retroviral vector (RV) and Tgfbr2^−/−Tbx21^−/− P14 T cells were transduced with a retroviral vector carrying Hic1 cDNA (Hic1 overexpression [OE]). After spin transduction, WT and Tgfbr2^−/−Tbx21^−/− P14 T cells were mixed 1:1 and adoptively transferred into B6 recipients immediately followed by LCMV infection. In this system, we were able to compare four subsets of P14 T cells isolated from the same tissue (i.e., WT, WT + empty RV, Tgfbr2^−/−Tbx21^−/−, and Tgfbr2^−/−Tbx21^−/− + Hic1 OE). Indeed, Hic1 OE significantly boosted both CD69 and CD103 expression in intestinal Tgfbr2^−/−Tbx21^−/− cells (Figures 6C, 6E, and 6F). We have demonstrated that IL-18 receptor downregulation is associated with the establishment of tissue residency.²³ In addition to CD69 and CD103, Hic1 OE facilitated IL-18R downregulation in Tgfbr2^−/−Tbx21^−/− cells (Figure 6D). When using splenic P14 subsets as an internal reference to calculate the relative abundance of IEL P14 population, we found that Hic1 OE significantly enhanced the total population of Tgfbr2^−/−Tbx21^−/− IEL T_RMs (Figure 6G left). When directly examining the total population size of each subset, Hic1 OE cells were highly enriched in the SI-IEL compartment compared with the spleen (Figure 6H left). Together, forced expression of Hic1 markedly enhances the differentiation of intestinal Tgfbr2^−/−Tbx21^−/− T_RMs.

Considering the impressive impacts of Hic1 OE in Tgfbr2^−/− Tbx21^−/− T_RMs, we wondered whether Hic1 could directly boost intestinal T_RM formation in Tgfbr2^−/− cells. Interestingly, using a similar retrovirus system, Hic1 OE was able to boost CD69 expression as well as the total population size of Tgfbr2^−/− cells (Figure 6G right, 6H right, 6I, and 6J). Remarkably, the expression of CD69 in Hic1 OE Tgfbr2^−/− IEL cells was comparable to that of WT controls (Figures 6I and 6J). In stark contrast, Hic1 OE led to no detectable improvement of CD103 expression in Tgfbr2^−/− cells (Figures 6I and 6K). At day 28 post infection, Hic1 OE even reduced CD103 expression (Figure 6K), likely due to the strong selection for retention in IEL compartment favors undeleted and therefore CD103⁺ “Tgfbr2^−/−” cells while Hic1 OE boosts CD69 levels and thus alleviates this selection pressure.

To rule out the possibility that Hic1 OE only affected a few T_RM-associated surface markers, we performed bulk RNA-seq analysis on FACS-sorted SI-IEL P14 T cells. Unsupervised principal-component analysis (PCA) plot showed that along PC1 (42% variance), Hic1 OE Tgfbr2^−/−Tbx21^−/− IEL cells almost overlapped with WT IEL controls (Figure 7A). Along PC2 (23% variance), we did observe a separation between WT, Tgfbr2^−/− Tbx21^−/−, and Tgfbr2^−/−Tbx21^−/− Hic1 OE subsets (Figure 7A). When narrowed down to the established T_RM and T_Cir signatures, an interesting pattern emerged. Hic1 OE did not significantly enhance T_RM signature in Tgfbr2^−/−Tbx21^−/− cells (Figure 7B blue line and 7E). Instead, Hic1 OE significantly reduced T_Cir signature in Tgfbr2^−/−Tbx21^−/− cells (Figure 7B red and 7D). Tgfbr2^−/−Tbx21^−/− Hic1 OE IEL cells carried decreased T_Cir signature even when compared with WT IEL (Figure 7C red). Bio-logical process Gene Ontology analysis revealed that Hic1 OE controlled multiple pathways, including leukocyte adhesion, immune response, and regulation of DNA-binding transcription factor activity (Figure S7C). Together, Hic1 OE boosts intestinal T_RM formation in Tgfbr2^−/−Tbx21^−/− cells mainly via suppressing T_Cir gene expression.

Figure 7. Hic1 overexpression suppresses the expression of circulating genes in Tgfbr2^−/−Tbx21^−/− T cells.

WT and Tgfbr2^−/−Tbx21^−/− P14 T cells were transduced by retrovirus before being co-transferred into LCMV-infected recipients. D22 post transfer, different subsets of IEL P14 T cells were FACS sorted and subjected to bulk RNA-seq.

(A) PCA plot is shown. GSEA for T_Cir and T_RM signatures between Tgfbr2^−/−Tbx21^−/− and Tgfbr2^−/−Tbx21^−/− Hic1 OE (B) and Tgfbr2^−/−Tbx21^−/− Hic1 OE and WT (C). Heatmap focused on T_Cir signature genes (D) and T_RM signature genes (E). Each column represents a biologically independent replicate.

These results demonstrate a hierarchy of interactions among TGF-β-induced T_RM differentiation events. TGF-β-induced T-bet downregulation and Hic1 induction exert synergistic efforts leading to the formation of intestinal T_RMs.

DISCUSSION

TGF-β has been established as one of the key signals required for T_RM differentiation. The TGF-β-induced gene signature has been widely used in the T_RM field²⁶; however, the key down-stream events mediated by TGF-β signaling required for T_RM differentiation are not entirely clear. Here, via a reductionist’s approach, we sought to determine which TGF-β downstream events are critically involved in intestinal T_RM differentiation and can replace TGF-β signaling. We found that T-bet deficiency, but not Eomes deficiency, partially rescues intestinal T_RM differentiation in the absence of TGF-β signaling. This finding is surprising as previous evidence supports a model that suppression of T-bet/Eomes sensitizes CD8⁺ T cells to TGF-β signaling.¹² Our results demonstrate that T-bet deficiency can partially bypass TGF-β signaling. T-bet deficiency supports the formation of CD103⁺ T_RMs in the absence of TGF-β receptor in vivo. However, T-bet deficiency cannot override the requirement of TGF-β signaling for CD103 induction in vitro or ex vivo. This finding suggests that a TGF-β-independent mechanism exists in vivo to support CD103 expression and intestinal T_RM differentiation, which is normally suppressed by T-bet. It will be interesting to define the molecular nature of this mechanism in the future.

Hic1 has been established as a key factor in promoting intestinal T_RM formation.¹⁵ Here, we find that, during IEL T_RM differentiation, TGF-β induces and T-bet suppresses Hic1 expression. Based on a previous publication, Hic1 OE slightly reduces T-bet expression in CD8⁺ T cells.¹⁵ These findings suggest that T-bet downregulation and Hic1 induction are not entirely independent events. Importantly, Hic1 expression alone is sufficient to boost CD69⁺ T_RM differentiation in the absence of TGF-β signaling. Interestingly, for the efficient formation of CD69⁺CD103⁺ mature T_RMs, both T-bet downregulation and Hic1 induction are required. Thus, it is highly possible that T-bet downregulation and Hic1 induction represent two essential events playing synergistic roles in TGF-β-mediated intestinal T_RM differentiation. Hic1 is also downstream of retinoic acid (RA) signaling. A recent publication has provided strong evidence that mesenteric lymph nodes provide essential RA signaling to license intestinal CD103⁺ CD8⁺ T_RM differentiation.²⁷ Whether enhanced RA signaling occurs in Tgfbr2^−/−Tbx21^−/− cells remains to be determined in the future.

How Hic1 OE enhances Tgfbr2^−/−Tbx21^−/− T_RM formation is not entirely clear. Our RNA-seq results suggest that Hic1 OE inhibits the expression of circulation-related genes, consistent with its established role as a transcription repressor. However, this effect cannot fully explain the diverged overall gene expression pattern seen in the PCA plot (Figure 7A), which requires future investigation.

Even though T-bet deficiency supports the formation of CD69⁺CD103⁺ T_RMs in the absence of TGF-β receptor, the resulting Tgfbr2^−/−Tbx21^−/− T_RMs do carry important distinctions from WT T_RMs, such as enrichment of the type 17 effector program and high levels of Tcf-1 expression. It is interesting to note that both features are associated with T-bet deficiency and not unique to Tgfbr2^−/−Tbx21^−/− cells. Further, both features are not T_RM-specific as they are present in splenic memory T cells. We have shown that type 17 differentiation suppresses intestinal T_RM differentiation. Although the downregulation of Tcf-1 is required for the efficient formation of the T_RM population,²⁸ our results indicate that the high level of Tcf-1 expression in Tgfbr2^−/−Tbx21^−/− cells is not apparently involved in intestinal T_RM formation.

A recent publication using a commensal bacterial infection model has demonstrated that both T-bet-dependent type 1 T_RMs and c-Maf-dependent type 17 T_RMs are present in the skin.²⁹ Similar to Tbx21^−/− CD8⁺ T cells, Tgfbr2^−/−Tbx21^−/− CD8⁺ T cells carry a clear type 17 signature. However, obvious distinctions exist between Tgfbr2^−/−Tbx21^−/− T_RMs in the gut and T_RM17 cells in the skin, including TGF-β dependency and the role of Tcf-1. These distinctions may be due to the differences in infection models, tissues, or genetic models. Thus, the true molecular relationship between Tgfbr2^−/−Tbx21^−/− T_RMs in the small intestine and T_RM1/T_RM17 in the skin remains to be defined. It is conceivable that T_RMs isolated from different tissues require distinct transcriptional regulatory networks. For example, although the vast majority of T_RM subsets are TGF-β dependent, T_RMs isolated from the liver²⁶ and upper respiratory tract³⁰ are not. In addition, a CD69⁺CD103⁻ T_RM subset in SI-LP is TGF-β independent and occupies a different microscopic location in an oral bacterial infection model⁸. It will be interesting to compare the location of Tgfbr2^−/−Tbx21^−/− T_RMs and WT controls in different infection models and further characterize their microenvironmental niches.

Together, using genetic models, we have identified T-bet downregulation and Hic1 induction as two distinct, yet critical, events downstream of TGF-β signaling during intestinal T_RM formation. Enforcing these two events allows intestinal T_RM differentiation in the absence of TGF-β signaling.

Limitation of the study

Our study is limited to one systemic viral infection model and P14 TCR transgenic cells. Different infection systems may yield different local environmental signals that affect T_RM differentiation. The findings presented in our study need to be validated in an oral-infection- or bacterial-infection-induced intestinal T_RM population. Our conditional-KO models are all mediated by distal Lck-Cre, which is active after thymocyte positive selection. This is not a T_RM stage-specific KO system. Indeed, we observed significant alterations in circulating memory T cells in the spleen. Further, we relied on a retrovirus-based delivery system to overexpress Hic1 in CD8⁺ T cells. There are at least two caveats with this system. First, CD8⁺ T cells were activated in vitro before retrovirus transduction. Thus, the T cells were not primed in vivo in a physiological setting. Second, it is an OE system. The high level of enforced Hic1 expression from an early-stage post-T cell priming may introduce unexpected confounding factors.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Nu Zhang (zhangn3@uthscsa.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• The bulk RNA-seq data for splenic and SI-IEL P14 T cells are available for download on GEO data repository with accession number GSE184629. The bulk RNA-seq data for Hic OE P14 T cells available for download on GEO data repository with accession number GSE260630. The ATAC-seq results are available for download on GEO data repository with accession number GSE184628.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice and virus

C57BL/6J (B6) mice were obtained from the Jackson Laboratory and a colony of D^b-GP_33–41 TCR transgenic (P14) mice was maintained at our specific pathogen-free animal facilities at the University of Texas Health Science Center at San Antonio (San Antonio, Texas). B6.CD45.1 mice were originally obtained from the Jackson Laboratory and bred with P14 mice to generate congenically marked P14 mice. All recipient mice were used at 6 to 10 wk of age. Tgfbr2^f/f and dLck-Cre mice were described before^32,33 and available from Jax. Tbx21^f/f (Jax#022741,¹⁸), Eomes^f/f (Jax#017293,³⁴) and Rorc^−/− (Jax#007572,^35;36) were purchased from Jackson Laboratory. Both male and female mice are used in the experiments. No sex-dependent difference was observed. All mice were housed at our specific pathogen-free animal facilities at the University of Texas Health at San Antonio (San Antonio, TX). All experiments were done in accordance with the University of Texas Health Science Center at San Antonio Institutional Animal Care and Use Committee guidelines. Mice were infected i.p. by 2 × 10⁵ pfu LCMV Arm. Viruses were grown and quantified as described.³¹

Cell lines

293T cells (ATCC) were maintained in complete DMEM and used for retrovirus production.

METHOD DETAILS

Flow cytometry

Anti-CD16/32 (2.4G2) was produced in the lab and used in all FACS staining as FcR blocker. For intracellular cytokine staining, freshly isolated splenocytes were cultured with 0.1μM GP_33–41 peptide (AnaSpec) in the presence of Brefeldin A (BioLegend) for 4–5 h at 37°C. After surface staining, IFN—γ, TNF, IL-17 and IL-2 was performed using permeabilization buffer (BioLegend) following fixation. Ghost Dye Violet 510 (Tonbo Bioscience) was used to identify live cells. For granzyme staining, freshly isolated cells were surface stained, fixed and permeabilized using permeabilization buffer (BioLegend) before incubating with anti-granzyme antibodies. For transcription factor staining, surface-stained cells were treated by Foxp3/Transcription Factor Staining Buffer Kit (Tonbo). Washed and fixed samples were analyzed by BD LSRII or BD FACSCelesta, and analyzed by FlowJo (TreeStar) software.

Naive T cell isolation and adoptive transfer

Naive CD8⁺ T cells were isolated from pooled spleen and lymph nodes using a MojoSort mouse CD8 T cell isolation kit (BioLegend) following the manufacturer’s instruction. During the first step of biotin antibody cocktail incubation, biotin-αCD44 (IM7, BD) was added to label and deplete effector and memory T cells. Isolated naive CD8⁺ T cells were enumerated, 1:1 mixed (WT P14 plus one of the KO/DKO P14 mice), 10⁴ cells adoptively transferred into each sex-matched unmanipulated B6 recipient via an i.v. route before LCMV infection.

Lymphocyte isolation from the SI-IEL and SI-LP

Lymphocyte isolation procedures have been described before.^6,23 Briefly, small pieces of the small intestine were stirred at 800 rpm for 20 min in HBSS buffer containing 1mM dithiothreitol and 10% FCS at 37°C to release IEL. The remaining pieces of the small intestine were first treated by Ca²⁺/Mg²⁺-free HBSS containing 5mM EDTA to remove epithelia. EDTA-treated tissue was further digested by 0.08U/ml Liberase TL (Sigma, 5401020001) + 200U/ml DNase I (Sigma, D5025) + 1.33 mg/ml Dispase II (Sigma, D4693) with stirring for 45 min at 37°C. Both digested LP and released IEL were further purified by density gradient centrifugation with PBS-balanced 44% and 67% Percoll (Cytiva).

In vitro T cell activation

Naive P14 T cells were stimulated with 10nM GP_33–41 peptide (AnaSpec) plus soluble 1 μg/ml αCD28 (37.51, Bio X Cell) in the presence of 5 ng/ml IL-2 (BioLegend) with 2.5 ng/ml added hTGF-β1 (Biolegend) or 10 μg/ml anti-TGF-β (1D11, BioXcell). 4 days after culture, the expression of CD103 was determined on live CD8 T cells by FACS.

Ex vivo effector T cell culture

Day 5 post-LCMV Arm infection, total splenocytes containing P14 T cells were cultured in complete RPMI with 5 ng/ml IL-2 (BioLegend) in the presence or absence of added 20 ng/ml hTGF-β1 (Biolegend). 48 h later, the expression of CD103 on live P14 T cells was determined by FACS.

Retrovirus production and CD8 transduction

Retrovirus transduction was performed as described before.³⁷ Briefly, 293T cells were transfected with pCL-Eco and the plasmid of interest using FuGENE 6 (Promega). pCL-Eco was a gift from Inder Verma (Addgene).³⁸ MSCV-IRES-Thy1.1 DEST vector (Addgene)³⁹ was used to construct Hic1 OE vector. Retroviral supernatant was collected 48 h later. Purified naive P14 T cells were activated by 5 μg/ml αCD3+2 μg/ml αCD28 + 10 ng/ml IL-2 overnight. Live activated P14 T cells were purified by density gradient centrifugation with PBS-balanced 30% and 65% Percoll (Cytiva). Then, activated P14 T cells were spin infected by freshly collected retroviral supernatant in the presence of 4 μg/ml polybrene (Tocris) at 2,000g for 60 min at 30°C followed by 4-hour-incubation at 37°C. After extensive wash, 1×10⁵ retrovirus transduced P14 T cells were adoptively transferred into each recipient mouse, which had been infected by LCMV one day prior.

CRISPR/Cas9-mediated gene KO in naive T cells

We followed a published protocol using a Lonza 4D-Nucleofector and P3 primary cell 4D-Nucleofector X kit.²⁵ Pre-made Cas9 protein (IDT, Cat#1081059) and sgRNA (Synthego, CRISPRevolution sgRNA EZ kit) complex were prepared. Naive P14 T cells were resuspended in freshly prepared P3 buffer from P3 primary cell 4D-Nucleofector X kit. Resuspended cells were added to the preformed Cas9/sgRNA complex and were electroporated using a pre-configured program (Pulse DN100, for unstimulated mouse T cells). After electroporation, warm complete RPMI was added, and the cells were rested for 10 min in a cell culture incubator before live cell count and adoptive transfer.

RNA-seq analysis

Day 27 after infection, pooled P14 T cells from 5 to 10 recipient mice were isolated from SI-IEL compartment and FACS sorted into indicated subsets based on congenic markers (CD45.1 and CD45.2) and T_RM markers (CD69 and CD103). Total RNA was extracted from sorted cells using a Quick-RNA Miniprep kit from Zymo Research. Sequencing library was constructed according to Illumina TruSeq Total RNA Sample Preparation Guide (RS-122–2201). Each library was barcoded and then pooled for cluster generation and sequencing run with 50bp single-end sequencing protocol on an Illumina HiSeq 3000 platform by UT Health San Antonio Genomic Sequencing Core Facility. An independent set of samples were sequenced by Novogene. Original RNA-seq results can be accessed by GSE184629. For Hic1 OE RNA-seq, retrovirus-transduced WT and Tgfbr2^−/−Tbx21^−/− P14 T cells were adoptively transferred into B6 recipients followed by LCMV infection. Twenty-two days later, SI-IEL lymphocytes were FACS sorted into 3 subsets, i.e., WT (with and without empty control retrovirus), Tgfbr2^−/−Tbx21^−/− (CD90.1⁻, no retrovirus) and CD90.1⁺Tgfbr2^−/−Tbx21^−/− (Hic1 OE DKO). Total RNA was extracted from sorted cells and subjected to bulk RNA-seq analysis by Novogene. The results can be accessed by GSE260630.

For bioinformatic analysis, raw FASTQ files from RNA-Seq paired-end sequencing were trimmed and filtered by Fastp (version 0.19.5), and then aligned to the GRCm39/mm39 reference genome using Bowtie2 (version 2.4.1), the reads were counted by FeatureCounts (version 2.0.6). Genes with differential expression across samples (DEGs) were assessed using the DESeq2 (version 1.42.0) package of R. An FDR of 0.05 and Log2 fold change cut-off of 1 were imposed. PCA and heatmap plots were built using normalized and filtered log2 count. Gene set variation analysis (GSVA) and Rotation Gene Set Tests (Roast) were performed by the GSVA (version 1.50.0) and Limma (version 3.58.1) package in R, respectively. Gene Set Enrichment Analysis (GSEA) was performed using the Broad Institute software (https://www.broadinstitute.org/gsea/index.jsp). Multiple comparative analysis for T_RM and T_Cir were performed using published gene signatures.^4,40

ATAC-seq analysis

ATAC-seq was performed as described before.¹⁵ Briefly, 5 × 10⁴ P14 T cells were FACS sorted from pooled samples. The nuclei pellet was treated with Tn5 transposase from Nextera DNA Sample Prep Kit (Illumina). The transposase-associated DNA was purified, amplified and then size selected before deep sequencing. Original ATAC-seq results can be accessed by GSE184628.

For bioinformatic analysis, raw ATAC-seq FASTQ files were trimmed and filtered by Fastp (version 0.19.5), and then aligned to the GRCm38/mm10 reference genome using Bowtie2 (version 2.4.1), the Samtools (version 1.3.1) were used to remove unmapped, unpaired, mitochondrial reads. PCR duplicates were removed using Picard (version 2.25.0). Peak calling was performed using Macs2 (version 2.2.7.1). For each experiment, we combined peaks of all samples to create a union peak list and merged overlapping peaks with BedTools (version 2.30.0) merge. The peaks were visualized in Integrative Genomics Viewer (IGV, version 2.9.4). The functional genomic regions of samples were visualized by ngsplot (https://github.com/shenlab-sinai/ngsplot). The motif analysis was performed using Homer (version 4.11).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistic details can be found in the figure legends. Mean ± SEM is shown in all figures. p value was calculated by two-tail paired or unpaired Student t-test or One-way ANOVA using Prism 10 software. p values of <0.05 were considered significant.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PE anti-CD127 (A7R34)	BioLegend	Cat#135009; RRID: AB_1937252
PerCP/Cyanine 5.5 anti-CD127 (A7R34)	BioLegend	Cat#135022; RRID: AB_1937273
APC anti-CD127 (A7R34)	BioLegend	Cat#135012; RRID: AB_1937216
Brilliant Violet 785™ anti-CD127 (A7R34)	BioLegend	Cat# 135037; RRID: AB_2565269
Ultra-LEAF™ Purified anti-CD3e (145–2C11)	BioLegend	Cat# 100359; RRID: AB_2616673
InVivoMAb anti-CD28 (37.51)	Bio X Cell	Cat#BE0015–1; RRID: AB_1107624
APC anti-KLRG1 (2F1)	Cytek (Tonbo)	Cat#20–5893-U100; RRID: AB_2621607
PE/Cyanine7 anti-KLRG1 (2F1)	BioLegend	Cat#138416; RRID: AB_2561736
Brilliant Violet 605™ anti-KLRG1 (2F1)	BioLegend	Cat#138419; RRID: AB_2563357
FITC anti-CD8b (H35–17.2)	Thermo Fisher Scientific	Cat#11–0083-85; RRID: AB_657764
APC-eFluor™ 780 anti-CD8b (H35–17.2)	Thermo Fisher Scientific	Cat#47–0083-82; RRID: AB_2573943
eFluor™ 450 anti-CD8b (H35–17.2)	Thermo Fisher Scientific	Cat#48–0083-82; RRID: AB_11218504
PE anti-CD8b (YTS156.7.7)	BioLegend	Cat#126608; RRID: AB_961298
PerCP/Cyanine5.5 anti-CD45.1 (A20)	Cytek (Tonbo)	Cat#65–0453-U100; RRID: AB_2621893
APC/Cyanine7 anti-CD45.1 (A20)	Cytek (Tonbo)	Cat#25–0453-U100; RRID: AB_2621629
APC anti-CD45.1 (A20)	BioLegend	Cat#110714; RRID: AB_313503
Super Bright™ 600 anti-CD45.1 (A20)	Thermo Fisher Scientific	Cat#63–0453-82; RRID: AB_2717041
APC anti-CD45.2 (104)	BioLegend	Cat#109814; RRID: AB_389211
FITC anti-CD45.2 (104)	BioLegend	Cat#109806; RRID: AB_313443
PE/Cyanine7 anti-CD45.2 (104)	Cytek (Tonbo)	Cat#60–0454-U100; RRID: AB_2621851
APC/Cyanine7 anti-CD45.2 (104)	Cytek (Tonbo)	Cat#25–0454-U100; RRID: AB_2621630
Super Bright™ 780 anti-CD45.2 (104)	Thermo Fisher Scientific	Cat#78–0454-82; RRID: AB_2802469
FITC anti-CD62L (MEL-14)	Cytek (Tonbo)	Cat#35–0621-U500; RRID: AB_2621697
APC/Fire™ 750 anti-CD62L (MEL-14)	BioLegend	Cat#104450; RRID: AB_2629772
PE anti-CD69 (H1.2F3)	BioLegend	Cat#104508; RRID: AB_313111
APC anti-CD69 (H1.2F3)	BioLegend	Cat#104514; RRID: AB_492843
PE/Cyanine7 anti-CD69 (H1.2F3)	Cytek (Tonbo)	Cat#60–0691-U100; RRID: AB_2621856
Super Bright™ 600 anti-CD69 (H1.2F3)	Thermo Fisher Scientific	Cat#63–0691-82; RRID: AB_2688097
APC anti-CD183 (CXCR3–173)	BioLegend	Cat#126512; RRID: AB_1088993
Alexa Fluor® 647 anti-CD103 (2E7)	BioLegend	Cat#121410; RRID: AB_535952
PerCP/Cyanine5.5 anti-CD103 (2E7)	BioLegend	Cat#121416; RRID: AB_2128621
PE anti-CD103 (2E7)	Thermo Fisher Scientific	Cat#12–1031-83; RRID: AB_465799
Super Bright™ 600 anti-CD103 (2E7)	Thermo Fisher Scientific	Cat#63–1031-82; RRID: AB_2802433
PE anti-CD218a (P3TUNYA)	Thermo Fisher Scientific	Cat#12–5183-82; RRID: AB_2572617
eFluor™ 450 anti-CD218a (P3TUNYA)	Thermo Fisher Scientific	Cat#48–5183-82; RRID: AB_2574069
APC/Cyanine7 anti-CD90.1 (OX-7)	BioLegend	Cat#202519; RRID: AB_2201418
PE anti-CD49a (HMa1)	BioLegend	Cat#142603; RRID: AB_10945160
PE/Cyanine7 anti-IFN-g (XMG1.2)	BioLegend	Cat#505826; RRID: AB_2295770
Brilliant Violet 605™ anti-IFN-g (XMG1.2)	BioLegend	Cat#505840; RRID: AB_2734493
PE anti-IL-2 (JES6–5H4)	BioLegend	Cat#503808; RRID: AB_315302
Pacific Blue™ anti-TNF-a (MP6-XT22)	BioLegend	Cat#506318; RRID: AB_893639
FITC anti-TNF-a (MP6-XT22)	BioLegend	Cat#506304; RRID: AB_315425
Alexa Fluor® 488 anti-IL-17A (TC11–18H10.1)	BioLegend	Cat#506909; RRID: AB_536011
PE/Cyanine7 anti-Granzyme A (GzA-3G8.5)	Thermo Fisher Scientific	Cat#25–5831-82; RRID: AB_2573476
PE anti-Granzyme A (GzA-3G8.5)	Thermo Fisher Scientific	Cat#12–5831-82; RRID: AB_2572631
PE anti-RORgt (AFKJS-9)	Thermo Fisher Scientific	Cat#12–6988-82; RRID: AB_1834470
Anti-CD16/32 (2.4G2)	Produced in-house	N/A
FITC anti-T-bet (4B10)	BioLegend	Cat#644811; RRID: AB_2287097
PE anti-T-bet (4B10)	Thermo Fisher Scientific	Cat#12–5825-82; RRID: AB_925761
PE anti-Eomes (Dan11mag)	Thermo Fisher Scientific	Cat#12–4875-82; RRID: AB_1603275
Alexa Fluor™ 488 anti-Eomes (Dan11mag)	Thermo Fisher Scientific	Cat#53–4875-82; RRID: AB_10854265
Alexa Fluor® 488 anti-TCF1/TCF7 (C63D9)	Cell Signaling Tech	Cat#6444S; RRID: AB_2797627
Pacific Blue™ anti-TCF1/TCF7 (C63D9)	Cell Signaling Tech	Cat#9066S; RRID: AB_2797696
Alexa Fluor®647 anti-TCF1/TCF7 (C63D9)	Cell Signaling Tech	Cat#6709S; RRID: AB_2797631
Biotin anti-CD44 (IM7)	BD Biosciences	Cat#553132; RRID: AB_394647
InVivoMAb Anti-TGF-β (1D11.16.8)	Bio X Cell	Cat#BE0057; RRID: AB_1107757
Alexa Fluor®647 anti-Hic1 (H-6)	Santa Cruz	Cat#sc-271499; RRID: AB_10650134
Bacterial and virus strains
Lymphocytic choriomeningitis virus-Armstrong strain	Ma et al.31	N/A
Chemicals, peptides, and recombinant proteins
Liberase TL	Sigma	Cat#5401020001
EDTA	Sigma	Cat#EDS-100G
DNase I	Sigma	Cat#D5025
Dithiothreitol	Sigma	Cat#10197777001
Dispase II	Sigma	Cat#D4693
Percoll	Cytiva	Cat#17089101
H-2Db-restricted GP33–41 peptide	AnaSpec	Cat#AS-61296
RPMI 1640	Cytiva	Cat#SH30096.01
DMEM	Corning	Cat#10–013-CV
2-Mercaptoethanol	Gibco	Cat#21985–023
HBSS (w/o calcium/magnesium)	Corning	Cat#20–021-CV
HBSS	Gibco	Cat#14065–056
FuGENE 6	Promega	Cat#E2691
Polybrene	Tocris	Cat#7711
Alt-R™ S.p. Cas9 Nuclease V3	IDT	Cat#1081059
Brefeldin A Solution (1000X)	BioLegend	Cat#420601
Ghost Dye™ Violet 510	Cytek (Tonbo)	Cat#13–0870-T500
Recombinant human TGF-b1	BioLegend	Cat#781802
Recombinant mouse IL-2	BioLegend	Cat#575408
Critical commercial assays
MojoSort™ mouse CD8 T cell isolation kit	BioLegend	Cat#480035
Intracellular Staining Permeabilization Wash Buffer (10X)	BioLegend	Cat#421002
Foxp3/Transcription factor staining buffer kit	Cytek (Tonbo)	Cat#TNB-0607
Nextera DNA Sample Prep Kit	Illumina	Cat# FC-121–1030
P3 primary cell 4D-Nucleofector® X kit	Lonza	Cat#V4XP-3032
Quick-RNA Miniprep Kit	Zymo Research	Cat#11–328
Deposited data
Spleen and SI-IEL P14 bulk RNA-seq at d27 post LCMV	This paper	GEO: GSE184629
SI-IEL P14 with Hic1 OE at d22 post LCMV	This paper	GEO: GSE260630
Spleen and SI-IEL P14 ATAC-seq	This paper	GEO: GSE184628
Experimental models: Cell lines
293T	ATCC	Cat#CRL-3216; RRID: CVCL_0063
Experimental models: Organisms/strains
P14 (B6.Cg-Tcratm1Mom Tg(TcrLCMV)327Sdz/TacMmjax)	The Jackson Laboratory	Cat#037394-JAX; RRID: MMRRC_037394-JAX
C57BL/6J	The Jackson Laboratory	Cat#000664; RRID: IMSR_JAX:000664
CD45.1 (B6.SJL-Ptprca Pepcb/BoyJ)	The Jackson Laboratory	Cat#002014; RRID: IMSR_JAX:002014
CD45.1/2	Bred in-house	N/A
dLck-Cre (B6.Cg-Tg(Lck-icre)3779Nik/J)	The Jackson Laboratory	Cat#012837; RRID: IMSR_JAX:012837
Tgfbr2fl (B6; 129-Tgfbr2tm1Karl/J)	The Jackson Laboratory	Cat#012603; RRID: IMSR_JAX:012603
Tbx21fl (B6.129-Tbx21tm2Srnr/J)	The Jackson Laboratory	Cat#022741; RRID: IMSR_JAX:022741
Eomesfl (B6.129S1 (Cg)-Eomestm1Bflu/J)	The Jackson Laboratory	Cat#017293; RRID: IMSR_JAX:017293
Rorc−/− (B6.129P2(Cg)-Rorctm2Lit7J)	The Jackson Laboratory	Cat#007572; RRID: IMSR_JAX:007572
Oligonucleotides
CRISPRevolution sgRNA EZ kit for Rora (with 3 pre-mixed sgRNA targeting Rora)	Synthego	N/A
CRISPRevolution sgRNA EZ kit for Rorc (with 3 pre-mixed sgRNA targeting Rorc)	Synthego	N/A
CRISPRevolution sgRNA EZ kit for Tcf7 (with 3 pre-mixed sgRNA targeting Tcf7)	Synthego	N/A
Recombinant DNA
pCL-Eco	Addgene	Addgene plasmid #12371; RRID: Addgene_12371
MSCV-IRES-Thy1.1 DEST	Addgene	Addgene plasmid #17442; RRID: Addgene_17442
MSCV-Hic1-IRES-Thy1.1	This paper	N/A
Software and algorithms
FlowJo v10	Treestar Inc	RRID:SCR_008520
Prism 10	Graphpad Inc	RRID:SCR_002798
Fastp (version 0.19.5)	Anaconda	RRID:SCR_016962
Bowtie 2 (version 2.4.1)	Anaconda	RRID:SCR_016368
Subread (version 2.0.6)	Anaconda	RRID:SCR_009803
Macs2 (version 2.2.7.1)	Anaconda	RRID:SCR_013291
Samtools (version 1.3.1)	Anaconda	RRID:SCR_002105
Picard (version 2.25.0)	Anaconda	RRID:SCR_006525
BedTools (version 2.30.0)	Anaconda	RRID:SCR_006646
Homer (version 4.11)	Anaconda	RRID:SCR_010881
R (version 4.3.1)	R Foundation	RRID:SCR_001905
DESeq2 (version 1.42.0)	Bioconductor	RRID:SCR_015687
GSVA (version 1.50.0)	Bioconductor	RRID:SCR_021058
Limma (version 3.58.1)	Bioconductor	RRID:SCR_010943
Gene Set Enrichment Analysis (version 4.2.2)	Broad Institute	RRID:SCR_003199
ngsplot	shenlab	RRID:SCR_011795
Integrative Genomics Viewer (version 2.9.4)	Broad Institute	RRID:SCR_011793

Highlights.

• T-bet deficiency partially rescues TGF-βR-KO T_RM formation in the small intestine

• T-bet-deficiency-induced type 17 program inhibits gut T_RM formation

• Hic1 further boosts gut T_RM differentiation in the absence of TGF-β receptor