Epithelia Use Butyrophilin-like Molecules to Shape Organ-Specific γδ T Cell Compartments

1The specialized differentiation of body surface epithelia is most commonly viewed from the perspective of organ-specific physiological functions, such as nutrient absorption in the gut and prevention of trans-epidermal water loss. Likewise, differentiated epithelia provide physical and chemical barriers to pathogens and toxins. However, it is now appreciated that body surfaces at steady state comprise diverse cell types, including many immune cells. Among these, intraepithelial lymphocytes (IELs) expressing αβ T cell receptors (TCRs) can mount rapid recall responses to pathogens, while other IELs, commonly expressing γδ TCRs, contribute to the maintenance of body surface integrity that is key to metazoan viability. Thus, murine gut γδ T cells regulate enterocyte differentiation and turnover and limit inflammatory damage, while TCRγδ+ dendritic epidermal T cells (DETCs) limit inflammation, promote wound resolution, and increase cutaneous resistance to carcinogens. Moreover, the skin and gut of jawless fish harbor cells similar to TCRγδ+ IELs, arguing that such compartments fulfil critical, evolutionarily conserved roles.

2In mice, large IEL compartments are defined by γδ TCRs that match particular anatomical sites: Vγ5 in skin, Vγ7 in gut, and Vγ6 in the uterus. Conceivably, these alignments are determined by organ-specific products of epithelial differentiation, particularly since TCRγδ+ IELs are largely unaffected by major histocompatibility complex (MHC) antigens that shape αβ T cell repertoires. Likewise, whereas some γδ TCRs bind CD1 and MR1, genetic studies have largely excluded these MHC-related molecules from the selection of murine γδ cell repertoires.

3One insight into how epithelia might shape local IEL repertoires was provided for the skin by the discovery of Skint1, the founding member of a novel multi-gene family specifically expressed by thymic epithelial cells and keratinocytes. Skint1 drives the selective maturation of Vγ5+ DETC progenitors, and DETCs are >90% ablated in Skint1 mutant mice, while all other T cells are unaffected. However, the generality of this mechanism for IEL selection was questioned, since neither DETCs nor Skint1 is broadly conserved and because Skint genes are only expressed in skin and thymus.

4This notwithstanding, Skint genes sit within the Btnl family comprising six rodent and five human genes. Their poorly understood gene products are structurally similar to CD80 and PDL1 co-stimulatory and inhibitory molecules, which are themselves considered to be evolutionarily related to the MHC. By definition, Btnl/BTNL genes are structurally similar to butyrophilin (Btn/BTN) genes, of which mice have two and humans six. Butyrophilin genes derive their name (“butter-loving”) from the founding member, Btn1a1/BTN1A1, that encodes a milk fat micelle-associated protein. However, this function appears atypical in contrast to the recent implication, albeit largely imprecise, of several BTN/Btn/BTNL/Btnl gene products in immune regulation. Provocatively, human BTN3A1 facilitates peripheral blood γδ T cell responses to low-molecular-mass microbial and endogenous metabolites (so-called phosphoantigens), although it is not known whether this is mediated by direct TCR-BTN3A1 binding.

5To explore whether Btnl genes might mediate epithelial regulation of local γδ T cells, we considered the mouse gut, the major site of Btnl1, Btnl4, and Btnl6 expression. Here, we identify a time window early in the development of young mice in which Btnl1 expressed by post-mitotic, small intestinal villus epithelial cells critically and selectively promotes the maturation and expansion of Vγ7+ T cells, thereby shaping the IEL compartment. Requiring neither microbial nor food antigens, this process evokes Skint1-mediated DETC selection and αβ T cell selection by the MHC. Indeed, we show that intestinal epithelial cells expressing Btnl1 jointly with Btnl6 can induce TCR-dependent stimulation uniquely of intestinal Vγ7+ T cells.

6γδ cells have often been viewed as species specific with few features conserved between mouse and humans. However, it is increasingly clear that human tissues too harbor large γδ cell compartments with TCRs distinct from those in peripheral blood (,; our unpublished data). Furthermore, a large bioinformatics study of thousands of cancer patients presenting with a broad range of carcinomas established that the best correlate of overall survival was a tumor-associated γδ T cell gene signature. Hence, there is a pressing need to define how human epithelia interact with tissue-resident γδ T cells.

7Addressing this, we provide a refined description of human colonic γδ cells and show that a signature subset expressing TCRVγ4 is specifically regulated by human BTNL3 and BTNL8 expressed by human gut epithelium. Hence, the specialized differentiation of intestinal epithelial cells in mice and in humans includes the expression at steady state of site-specific regulators of local T cell compartments.

8By flow cytometry of cells recovered from epithelium, and by confocal visualization of epithelial whole mounts, we found that the signature murine small intestinal Vγ7+ IEL compartment largely took shape at 2–3 weeks of age and remained stable for at least 9 months thereafter (Figures 1A and 1B). At day 21, Vγ7+ cells mostly phenocopied mature Skint1-selected DETCs, expressing uniformly high levels of CD122 (the IL-2R/IL-15Rβ chain), TIGIT (an inhibitory co-receptor), and the TCR (detected with anti-CD3 antibodies) and low levels of RNA for Rorgc and Sox13, two transcription factors contributing to γδ T cell differentiation (Figure 1C). Vγ7− IELs (mostly Vγ1 or Vγ4) did not show this phenotype, and whereas both Vγ7+ and Vγ7− IEL subsets were mostly CD45RBhi, CD44+, and CCR9+, Vγ7+ IELs were distinct in being Lag3+, Thy1−, CD69+, CD5−, and CD8αα+ (Figures 1C and S1A).

9Prior to day 21, however, Vγ7+ IELs phenocopied Vγ7− IELs of adult mice. Thus, by sequential gating and radar plots of surface protein co-expression, one could clearly distinguish mature Vγ7+ IELs (CD122hi[MFI > 500], Thy1−, TIGIT+, Lag3+, CD8αα+, CD5−, CD24−, TCRhi) from putative Vγ7+ IEL progenitors (CD122lo[MFI < 200], Thy1+, TIGIT−, Lag3−, CD8αα−, CD5+, CD24+, TCRlo) (Figures 1D, 1E, and S1B), with the latter also phenocopying DETC progenitors prior to Skint1 selection.

10To further compare IELs with their putative progenitors, CD122hi Vγ7+ and CD122lo Vγ7+ IELs were purified from the same day 14–17 mice on four independent occasions and assessed by RNA sequencing (RNA-seq) (Figure S1C). Consistent with their distinct phenotypes, the cells showed significantly different expression of many genes for cell surface proteins (Figure S1C). Furthermore, many genes up- (e.g., Tnfrsf9 [4-1BB/CD137], Xcl1 [lymphotactin], Nasp) or downregulated (e.g., sox13, Bcl11b, Cx3cr1) in CD122hi versus CD122lo Vγ7+ cells were likewise regulated by Skint1 selection of DETC progenitors (Figures 1F and S1C).

11Additionally, CD122hi Vγ7+ cells were enriched in cell-cycle genes, consistent with which ∼100% of Vγ7+ IELs at day 21–24 were Ki67+ (i.e., outside of G0), compared to <40% of Vγ7− cells (p < 0.0001) (Figure 1G). Likewise, Vγ7+ IELs at day 28 phenocopied rapidly dividing thymocytes in that ∼10% incorporated ethynyldeoxyuridine (EdU) (a labeled nucleotide) during a 3-hr pulse, compared to only 4% of Vγ7− IELs (Figure S1D).

12In sum, these data are consistent with the gut supporting the selective maturation and expansion of CD122hi, Thy1−, TIGIT+, Lag3+, CD8αα+, CD5−, CD24−, TCRhi Vγ7+ cells that by weeks 3–4 dominate the γδ IEL compartment. After week 5, the fraction of cycling (Ki67+) Vγ7+ IELs at steady state declined to levels comparable to Vγ7− IELs (Figure 1G).

13Because Skint1 selects for signature Vγ5+ DETC progenitors in the thymus, DETCs are absent from athymic NU/NU mice. By contrast, intestinal IELs were present in NU/NU, and although there was some decrease in numbers (average of ∼1.3 × 106 cells compared to >2.0 × 106 cells in euthymic mice; see below), the compartment was again dominated by CD122hi Vγ7+ IELs. Moreover, ∼25% of Vγ7+ IELs in NU/NU and in euthymic mice reacted with antibody GL2 that detects Vδ4 (TRDV2-2 encoded) chains. Consistent with this, TRDV2-2 sequences accounted for ∼25% of TCRδ chain RNAs expressed by purified Vγ7+ IELs (Figures 2A and S2A). In sum, the shaping of the gut Vγ7+ IEL compartment did not require a thymus.

14Consistent with this, Vγ7+ thymocytes were rare, comprising <10% of TCRγδ+ cells in fetal and post-natal thymi across the first 8 weeks of life, the peak period of thymus function in mice (Figure S2B). Furthermore, most Vγ7+ thymocytes were CD45RBlo, Thy1+, CD5hi, CD122lo, TCRlo, and CD8αα−, thus offering no evidence for intrathymic maturation (Figures S2C–S2E). Likewise, neither lymph nodes nor Peyer’s patches (PP) were required to shape the IEL compartment, since normal numbers of Vγ7+ and Vγ7GL2+ IELs with signature phenotypes were present in aly/aly (alymphoplasia) mice that following surgery were confirmed to lack PP and peripheral and mesenteric lymph nodes (MLNs) (Figure S2F).

15As intestinal driver(s) of IEL maturation in weanling mice, microbial and/or food antigens were logical candidates. However, C57Bl/6 mice bred into and maintained in a germ-free environment and/or on elemental, protein-antigen-free diet displayed Vγ7+ and Vγ7GL2+ IEL compartments comparable to conventionally housed counterparts (Figure 2B). The Vγ7+ IELs were uniformly TCRhi, CD122hi, and absolute numbers were somewhat increased, partially compensating for the decline of TCRαβ IELs in germ-free and protein-antigen-free mice (Figure 2B). Thus, the local T cell compartment is most likely shaped by an endogenous intestinal element(s).

16In seeking that element(s), we focused on three genes, Btnl1, Btnl4, and Btnl6, that are closely related to Skint1. Btnl4 was expressed at low levels in proximal small intestine, commencing in the fetus. Btnl1 and Btnl6 RNAs were detected at day 6 postpartum, and Btnl1 levels further increased at around day 14 before the expression of all three Btnl genes stabilized (Figure 2C). Expression was in post-mitotic villus enterocytes that are interspersed with IELs and was essentially absent from villus crypts that house replicating epithelial cell progenitors and lack IELs (Figures 2D, 2E, and S2G). Btnl1 expression peaked in proximal and medial small intestine (Figure S2H), where its expression was >107-fold higher than in the thymus (Figure S2I). These expression patterns could permit Btnl1, Btnl4, and/or Btnl6 to act locally upon Vγ7+ IELs in weanling mice. To investigate this possibility, we obtained three independent strains of Btnl1−/− mice and one of Btnl4−/− mice, each generated by targeted mutagenesis of embryonic stem cells (ESCs), and a strain with an internally deleted Btnl1 locus generated by clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 in mouse eggs (Figure S2J). The strains were confirmed as gene knockouts by DNA analysis and loss of respective Btnl RNAs (Figures 2E and S2I–S2K).

17The four Btnl1−/− strains each displayed major, highly selective losses of Vγ7+ IELs, assessed by flow cytometry or confocal microscopy (Figures 3A, 3B, and S3A). Vγ7+ IEL numbers were depleted by ∼90%, with Vγ7GL2+ cells almost ablated. Because Vγ7− IEL numbers barely increased, the percentage representation of Vγ7+ cells among γδ IELs was reduced only by ∼3-fold relative to wild-type (WT) mice, but this was set against a background of dramatically reduced γδ IEL numbers (Figures 3A and S3A). By contrast, TCRβ+CD8αα+ IEL numbers increased significantly in Btnl1−/− mice (Figures 3A and 3B).

18The specificity of Btnl1 for Vγ7+ IELs was emphasized by comprehensive immune phenotyping of Btnl1−/−, WT, and Btnl1+/− mice that showed comparable splenic or MLN immune cell subsets (including γδ cell repertoires) and comparable representation and phenotypes of Vγ7+ thymocytes from day 4 to week 8 (Figures S3B–S3E). Consistent with its expression pattern, Btnl1 acted extrathymically; thus, Btnl1 deficiency crossed onto NU/NU mice reduced the average number of Vγ7+ IELs by ∼90%, with almost total loss of Vγ7GL2+ IELs (Figure 3C). Since NU/NU mice lack most αβ T cells, this result excludes the formal albeit unlikely possibility that Vγ7+ IEL losses in euthymic Btnl1−/− mice indirectly reflected expanded TCR αβ IELs. The specificity of Vγ7+ IELs for Btnl1 was emphasized by the fact that Btnl4−/− mice displayed no overt defects in any major IEL subset (Figure 3D).

19To determine how and when Btnl1 impacts IELs, we examined residual Vγ7+ and Vγ7GL2+ IELs in Btnl1−/− mice. Relative to those in WT mice, significantly fewer Vγ7+ IELs incorporated EdU at D28 [p < 0.0001] or expressed Ki67, and this did not change until W7 when most Vγ7+ IELs in WT mice moved out of cycling (Figures 4A and S4A). Thus, there was no selective expansion of Vγ7+ IELs in Btnl1−/− mice. Similarly, although their TCR levels were high, many residual Vγ7+ and Vγ7GL2+ IELs in week 3–5 Btnl1−/− mice were CD122lo, Thy1+, TIGIT−, Lag3−, CD8αα−, CD5+, CD24+, thereby phenocopying immature Vγ7+ IELs of day 14–17 WT mice (Figures 4B, 4C, and S4B).

20To demonstrate that Btnl1 exerts its selective impact on Vγ7+ T cells in trans, we established conditions in which the Vγ7+ and Vγ7GL2+ IEL compartments of week 3–5 WT mice were reconstituted within ∼5 weeks of donor bone marrow (BM) transfer to irradiated 8- to 10-week, γδ T cell-deficient TCRδ−/− mice (Figure S4C). BM from either WT or Btnl1−/− mice proved equally effective at IEL reconstitution (Figure 4D). Not surprisingly, WT BM reconstitution of Vγ7+ IELs in irradiated, congenic T cell-sufficient CD45.2+ WT recipients was less effective than it was in TCRδ−/− hosts (compare plots, top left in Figures 4D and 4E), but nonetheless, reconstitution of Btnl1−/− hosts was much less effective, and the few Vγ7+ and Vγ7GL2+ IELs that developed in Btnl1−/− recipients phenocopied residual Vγ7+ cells in Btnl1−/− mice and immature IELs in day 14–17 WT mice (Figure 4E). Complementing these findings, purified IELs from 4-week-old mice could reconstitute Vγ7+ IELs in recipient TCRδ−/− mice, albeit very inefficiently, and this too was greatly impaired in Btnl1−/−TCRδ−/− hosts (Figure S4D). Thus, Btnl1 acts in non-hematopoietic cells to support the selective expansion and maturation of Vγ7+ IEL.

21To attempt to restore IEL selection, we rendered Btnl1−/− mice transgenic for Btnl1 expressed from a doxycycline (Dox)-inducible promoter (Figures S5A and S5B) and crossed them onto Btnl1−/− mice, into which was introduced a Dox-responsive trans-activator (rtTA) regulated by the ubiquitously expressed Rosa26 promoter. As reported, adding low-dose Dox to drinking water had little overt impact on the gut over a several-week period. Therefore, this system offered a means to globally induce Btnl1 de novo in Btnl1−/− bitransgenic (BiTg) mice that inherited both rtTA and the inducible Btnl1 transgene. Conversely, BiTg mice administered sugar water and Dox-treated single-transgenic (SiTg) Btnl1−/− mice (that inherited only the Btnl1 transgene) served as controls. Btnl1 induction in mice of appropriate genotypes was validated by qPCR (Figure S5C).

22When several W11 Btnl1−/− BiTg mice were treated in this way, the representation of Vγ7+ and Vγ7+GL2+ IELs was unchanged relative to littermate controls, but the percentage of Vγ7+ IELs that were CD122hi increased greatly, and most expressed Ki67. This was not true for Vγ7− or TCRβ+ IELs (Figures S5D–S5F). Thus, Btnl1 induction de novo in adult mice partially phenocopied Vγ7+ IEL-specific maturation in early life but did not reconstitute Vγ7+ IEL numbers. Global Btnl1 induction also did not drive ectopic Vγ7+ cell maturation in any other tissues (unpublished data).

23In further experiments, we restricted Btnl1 induction to mature enterocytes by generating Btnl1−/− mice transgenic for rtTA expressed from the villin promoter. Within 1–2 weeks of Dox-treatment of several W11 BiTg Btnl1−/− mice, most Vγ7+ IELs had become Lag3hi, Thy1−, and CD122hi, of which the majority were also Ki67+ (Figures 5A and 5B). Again, this phenotypic transition was Vγ7+ IEL specific, albeit there were sometimes Btnl1-independent increases in Ki67+ TCRβ+ IELs in sugar-water-treated mice. (Figure 5B). Once more, the numbers and representation of Vγ7+ IELs were unchanged, even in mice retained on Dox for 3–4 weeks (Figure 5C). This establishes that Btnl1 can effect phenotypic conversion of immature Vγ7+ IELs rather than merely promote a selective outgrowth of mature Vγ7+ cells.

24However, there was significantly increased representation of Vγ7+ and Vγ7GL2+ IEL when Btnl1 expression was induced in Btnl1−/− mice in early-life by commencing Dox-treatment of nursing females at D7 or of weanlings at D21 and then maintaining treatment for 2-5 weeks (Figure 5D). Moreover, the expanded Vγ7+ cells phenocopied IEL of W4-5 WT mice, and were significantly different from the IEL of Dox-treated SiTg mice and Btnl1−/− mice (Figure 5E). Thus, acute expression of Btnl1 purely in the gut epithelium induces the selective phenotypic maturation and expansion of Vγ7+ IEL, but these effects are separable, with selective expansion mostly confined to a developmental window within the first five weeks of life.

25Given that acute Btnl1 expression drives the selective maturation of Vγ7+ IELs in vivo, we tested whether Btnl1 might show specificity for Vγ7+ IEL ex vivo. Since primary intestinal epithelial cells reportedly harbor Btnl1 in a complex with Btnl6, we sought evidence for heterotypic interactions of Btnl proteins. Indeed, cell surface expression of Btnl1 on Btnl1-transfected MODE-K cells (an established intestinal epithelial cell line in which endogenous Btnl genes are negligibly expressed) was greatly enhanced by co-transfection with Btnl4 or Btnl6 (Figure S6A). Likewise, surface Btnl6 expression was greatly enhanced by Btnl1, but there was no evidence for collaboration between Btnl6 and Btnl4 (Figure S6A). Given the specificity of Btnl6 for Btnl1, and given that Btnl4−/− mice showed no IEL phenotype, we focused on Btnl1 and Btnl6.

26MODE-K cells stably transduced with Btnl1 (L1), Btnl6 (L6), Btnl1 plus Btnl6 (L1+6), or empty vector (EV) were co-cultured with freshly explanted IELs that were then assayed for CD25 (IL-2Rα chain) upregulation, which is among the most robust readouts of TCR stimulation for systemic T cells. Whereas mixed TCRαβ+ and TCRγδ+ IELs showed minor CD25 upregulation upon co-culture with EV, L1, or L6 cells, IELs exposed to L1+6 cells showed highly significant CD25 upregulation, wholly attributable to ∼20% of Vγ7+ cells (both Vγ7GL2+ and Vγ7GL2− cells) (Figure 6A). Significant CD25 upregulation was first evident within 4–6 hr, as is true for systemic TCR stimulation (Figure S6B).

27When L1+6 cells were co-cultured with primary IELs from Nur77.gfp mice in which GFP is upregulated by nuclear factor of activated T cells (NFAT) activation downstream of TCR signaling, essentially all IELs that upregulated CD25 were GFP+ (Figure 6B). This phenotype was not shown by Vγ7− or TCRαβ+ IELs in the same co-cultures (Figure 6B). IEL upregulating CD25 also downregulated CD122, an additional symptom of systemic TCR stimulation (Figure S6C). Thus, CD25+GFP+CD122− cells arose uniquely among Vγ7+ IEL co-cultured with L1+6 cells (Figure 6C). Moreover, CD25 upregulation was accompanied by slight but significant TCR downregulation, another rapid response to TCR stimulation (Figure 6D). Btnl1+Btnl6 acted directly on Vγ7+ IEL, since CD25 was upregulated on cells that had been purified by flow cytometry prior to L1+6 cell co-culture (Figure S6D). (Note that background CD25 expression was increased by TCR-dependent sorting, but this did not obscure the result.)

28When IEL were separated from L1+6 cells by transwells, CD25 upregulation was abrogated and could not be secondarily trans-activated by IELs that were in contact with L1+6 cells, e.g., via secreted cytokines (Figure 6E; compare blue with yellow bars in bottom panel). CD25 upregulation by IELs in contact with L1+6 cells showed dose-dependent inhibition by PP2, which inhibits signaling by src-family kinases, such as Lck and Fyn, but was not inhibited by PP3, an established control for PP2 specificity (Figure S6E).

29Just as residual Vγ7+ IEL in Btnl1−/− mice responded to acute transgenic Btnl1 induction in vivo, they were comparable to WT Vγ7+ IELs in responding to Btnl1 plus Btnl6 ex vivo (Figure 6F). Interestingly in the same experiments, WT Vγ7+ IELs showed relatively poor responses to anti-CD3, phenocopying the attenuated responsiveness imposed on Vγ5+ DETC progenitors by Skint1, whereas Vγ7+ IELs from Btnl1−/− mice, and TCRαβ+ and Vγ7− IELs from WT mice, none of which subsets had experienced prior Btnl1 selection in vivo, all showed strong responses to anti-CD3 (Figures 6F and 6G).

30Finally, the supernatants of IEL co-cultures with L1+6 cells showed small but significant increases in interferon-γ (IFN-γ), CCL4, and granulocyte macrophage colony-stimulating factor (GM-CSF) among 36 cytokines tested (Figure 6H). These are typical IEL effectors, and although it was technically challenging to attribute production to Vγ7+ IELs, such increases were not seen in supernatants of L1+6 cells cultured with IELs from TCRδ−/− mice. (Note the higher background cytokine expression in TCRδ−/− mice may reflect spontaneous inflammation often associated with γδ deficiency;.) In sum, a range of metrics attested to a highly specific and direct interaction ex vivo of Vγ7+ IELs with Btnl1 and Btnl6 co-expressed on gut epithelial cells.

31Because of limited tissue access, human gut T cells are understudied. Nonetheless, there are gut-associated γδ cells whose TCR usage differs markedly from Vγ9+Vδ2+ cells that dominate the peripheral blood. To better characterize such cells, we submitted biopsy specimens from healthy ascending colon to a modified version of a protocol used to isolate human skin T cells. For 16 of 17 donors, the γδ cells were enriched in Vδ1+ cells, although Vδ1−Vδ2− cells were also present; hence, the term “Vδ2−” is used to distinguish tissue-associated γδ cells from Vδ2+ cells that could also be recovered from most gut samples, albeit in highly variable numbers (Figure 7A).

32Of six functional human Vγ chain genes (Vγ2, 3, 4, 5, 8, and 9), Vγ4 was reported to be the signature chain of intestinal Vδ2− cells. Indeed, for up to ten donors examined, most intestinal Vδ2− cells reacted with a Vγ2/3/4-specific antibody, but not a Vγ5/3-specific antibody (blue and red bars, Figure 7B; Table S1A), and TCR deep sequencing showed that Vγ4 sequences far outnumbered Vγ2 sequences (Figure S7A). Thus, despite individual variation, most gut γδ T cell compartments included a substantial Vγ4+Vδ2− subset, while some donors also displayed relatively high representation of Vγ8+Vδ1+ cells (Figure 7B; Table S1A).

33There is no human equivalent of the Btnl2-proximal amplicon on mouse chromosome 17 that encodes Btnl1, Btnl4, and Btnl6. However, adjacent to human BTNL9 is an amplicon that encodes BTNL3 and BTNL8 whose expression is highly enriched in gut, particularly EpCAM+ epithelial cells (Figures S7B–S7D). Interestingly, akin to the behavior of Btnl1 and Btnl6 described above, neither BTNL3 nor BTNL8 protein was efficiently expressed on cells transfected with their respective genes, unless both were co-expressed (Figure S7E, green line, top right histogram; blue line, bottom right histogram). Conversely, BTNL8S (a splice variant of BTNL8) failed to rescue surface BTNL3 expression (Figure S7E, red line, top right histogram).

34Whereas we could not test for a developmental dependence of human gut γδ cells on BTNL3 and BTNL8, we could assess whether BTNL3 and BTNL8 phenocopied Btnl1 and Btnl6 by specifically activating signature gut γδ cells in a TCR-dependent fashion. Thus, we established short-term co-cultures of primary-gut-derived lymphocytes with HEK293T cells transduced with BTNL3 (L3), BTNL8 (L8), BTNL3 and BTNL8 (L3+8), or EV (Figure S7F). For the representative donor shown, some of the discrete subsets of Vδ1− and Vδ1+ γδ cells that were apparent in T cell co-cultures with control (EV, L3, and L8) cells showed marked TCR downregulation when co-cultured with L3+8 cells (red arrows Figure 7C). Emphasizing specificity, TCR downregulation occurred in response to L3+8 cells in 21 of 23 donors but was never seen in co-cultures with L3 or L8 cells, and was never shown by intestinal Vδ2+ or TCRαβ+ cells, even in the same cultures as responding Vδ2− cells (Figure 7D). Although higher baseline CD25 expression reduced the sensitivity of this assay for human versus mouse gut T cell activation ex vivo, L3+8 cells induced significant CD25 upregulation vis-a-vis gut T cells co-cultured with control cells (Figure 7E), and CD25 upregulation was most evident on cells with downregulated TCRs (Figure S7G).

35Not all Vδ2− cells responded to L3+8 (Figure 7C). Thus, we considered that TCRγ chains might determine BTNL responsiveness, as is true in mice. Indeed, human Vδ2− populations that downregulated TCRs in co-cultures with L3+8 cells were detected by the Vγ2/3/4-specific antibody, but not by antibodies to Vγ8, Vγ5/3, or Vγ9 (Figures 7F and 7G). Moreover, productively rearranged Vγ4 genes were prevalent when L3+8-responsive cells with downregulated TCRs were flow cytometry sorted from one donor (Figure S7H) and their Vγ chains amplified without bias and sequenced (red notation, Table S1B). By contrast, TCRγ transcripts from skin-derived TCRγδ+ cells (G234SK01) were biased toward Vγ3 (purple notation Table S1B). Interestingly, of two donors showing no substantial response to BTNL3+8, one proved a posteriori to have an atypical intestinal γδ T cell repertoire dominated by Vγ8+ cells (Figure S7I). Likewise, L3+8 cells induced no significant TCR down-modulation by primary γδ+ T cells from skin or blood among which Vγ4+ cells are rare (Figure 7H). Thus, epithelial BTNL genes regulate human-tissue-resident γδ T cells in an organ-specific, TCRγ-chain-specific manner.

36This study shows that the unique composition of the murine intestinal intraepithelial T cell compartment arises from a selective maturation and expansion of Vγ7+ T cells driven by Btnl1, and most likely Btnl6, expressed by differentiated enterocytes. Likewise, BTNL3 and BTNL8 co-expressed by human intestinal epithelial cells selectively regulate gut Vγ4+ T cells. Given that Skint1 expressed by thymic epithelial cells and keratinocytes selectively regulates intra-epidermal Vγ5+ T cells, tissue-specific Btnl genes may offer a generalizable means by which epithelia shape and regulate local γδ T cell compartments. This may reflect an even broader utilization of BTNL/BTN genes in γδ biology, since BTN3A1 is critical to human peripheral blood γδ cell activation.

37Btnl1 and Skint1 effect many of the same changes in Vγ7+ and Vγ5+ IEL progenitors, respectively. These include upregulation of the receptor for IL15, a growth factor expressed by epithelial cells essential for γδ IEL maintenance; suppression of sox13, rorgc, and bcl11b that are associated with γδ cells producing IL-17, which is not a property of IELs; and attenuation of TCR responsiveness, consistent with IELs adopting an innate-like, rapidly responsive, tissue surveillance role.

38Being members of the B7 superfamily, Btnl/BTNL/Skint gene products may act as co-stimulators for IEL receptors yet to be identified. In this case, they will be the first co-stimulators specific for cells with particular TCRs; e.g., Vγ7+, but not Vγ7−, IELs from mouse gut; and Vγ4+, but not Vγ8+, T cells from human gut. Alternatively, their exquisite specificities may reflect interactions of Btnl/BTNL/Skint with cognate TCRs, possibly via unique Vγ-CDR1/2 regions. Likewise, human BTN3A1 appears to mediate its effects via the Vγ9Vδ2 TCR, although there are no clear direct binding data. This may reflect a highly complex interaction that includes critical co-factors. Of note, BTNL3, BTNL8, Btnl1, and Btnl6 each contain B30.2 domains related to the phosphoantigen-binding domain of BTN3A1, raising the possibility that low-molecular-weight metabolite(s) might have a broad role in γδ cell regulation by Btnl/BTNL genes. A requirement for co-factors might also explain the developmental time window during which Btnl1 could drive the maturation and expansion of Vγ7+ IELs.

39Peptide-MHC complexes have different impacts on αβ T-lineage cells, including positive and negative selection of thymocytes, T-regulatory cell differentiation, and activation or anergy of mature peripheral T cells. These outcomes are dictated by the state of the T cell and/or the biological context. In so far as parallels may be drawn, our study offers genetic evidence that Btnl1 drives Vγ7+ IEL selection and cell biological evidence that Btnl1 and Btnl6 can promote weak activation of Vγ7+ IELs, which was likewise true for human BTNL3 plus BTNL8 interactions with mature gut γδ cells. These different contexts might explain a seeming paradox that CD122 is upregulated by Btnl1-mediated selection in vivo, in common with agonist selection of thymocytes, and yet is downregulated by Btnl1 plus Btnl6-mediated cell activation ex vivo, in common with conventional T cell activation.

40Weak activation of mature IEL may reflect the use by epithelia of site-specific Btnl/BTNL molecules to sustain cognate IELs, orienting them as to their correct anatomical location and to the status of the tissue. Consistent with this, Skint1 constitutively engages Vγ5+ DETCs (our unpublished data); Btnl1 can directly appose IEL at steady state, and the same may be true for BTNL3 and BTNL8 in humans, given RNAscope evidence that human colonic γδ cells are largely intraepithelial (our unpublished data). Such findings emphasize intimate and specific relationships between resting epithelial cells and neighboring T cells that provide a basis for tissue surveillance.

41Although γδ IELs make essential contributions to body surface integrity, our identification of Btnl/BTNL proteins as key IEL regulators offers to clarify the contexts in which those contributions are most important. For example, when during gut development, dysregulation, and/or repair might epithelial Btnl/BTNL proteins and/or co-factors communicate a need for IEL activation? Such insight can inform genome-wide association study (GWAS) data implicating BTNL/BTN polymorphisms in numerous immunopathologies and may reveal why epithelia use organ-specific, as opposed to pan-epithelial, Btnl gene products to regulate local T cells. In fact, different γδ TCRs offer IELs a means by which to discriminate organ-specific epithelia that is not obviously available to innate lymphoid cells.

42Our and others’ studies have by now implicated Btnl1, Btnl6, Skint1, BTNL3, BTNL8, BTN3A1, and BTN3A2 in γδ cell regulation. Nonetheless, they may be pleiotropic. For example, the long cytoplasmic tails of BTNL3, BTNL8, Btnl1, and Btnl6 may signal back to the cells that display them, consistent with our finding that Btnl1 attenuates epithelial cell sensitivity to IEL-derived cytokines. Yet, other BTNL/BTN protein functions may not relate to γδ cell biology, consistent with which we show that Btnl4 is not critical for gut γδ T cell development. A similar diversity of immunological functions may describe avian Btnl-like B-G genes, a subset of which may shape and/or regulate large γδ cell compartments in birds.

43A clear exception to the many parallels of Btnl1-mediated IEL selection and Skint1-mediated DETC selection is Skint1 expression by thymic epithelial cells. Conversely, the restriction of Btnl1 to the intestine, both naturally and in villin-driven BiTg mice, clearly establishes that the signature gut IEL compartment is shaped extrathymically, even if progenitors are thymus derived, thus resolving a long-standing controversy. This thymic independence may reflect a need to replenish gut IELs so as to maintain gut integrity post-thymic involution.

44For additional information about reagents and resources, contact the Lead Contact, Adrian Hayday at adrian.hayday@kcl.ac.uk.

45Wild-type (WT) C57Bl/6 mice were obtained from Charles River and Harlan. 3 independently derived embryonic stem (e.s.) cells for Btnl1−/− (Btnl1tm1(KOMP)Mbp) and e.s. cells for Btnl4−/− (Btnl4tm1(KOMP)Mbp) mice were obtained from the international mouse phenotyping consortium (IMPC) (project IDs: CSD67994 and CSD81524). Btnl1indel/indel mice were generated using CrisprCas Technology. Briefly: Two independent short guide RNAs, targeting the intronic region: between exon1 and 2 and between exon 5 and 6 were identified using the online tool: http://crispr.mit.edu/.

46Intron1/2: CCAGCTCCAAGATCCCCCTTGGG Intron5/6: TCCATAGCACCTTATCCGGTTGG

47The sg RNAs & PAM sequences were cloned into the g-RNA basic vector, translated in vitro, purified and co-injected with Cas9 into day 1 zygotes and transferred into pseudopregnant foster mice.

48WT and Btnl-knockout lines were generated and maintained at The Francis Crick Institute’s Biological resource facilities. B6.Cg-Foxn1 < Nu > /J (NU/NU) mice were obtained from the Jackson Laboratory. Nur77.gfp mice were kindly provided by K. Hogquist. For timed pregnancies, mice were mated overnight and E0 was considered as the day a vaginal plug is observed. Both male and female mice aged between 1 and 35 weeks (as indicated) were used in this study. No gender-specific differences were observed.

49C57Bl/6 mice maintained on germ-free or on solid-food antigen-free diets were bred at the Institute for Medical Microbiology and Hospital Epidemiology, University of Marburg, Germany where all experiments were conducted according to the German animal protection law.

50Germ-free (GF) mice were kept in plastic isolators (Metall and Plastik, Germany) with autoclaved food, bedding and water. Sterility of animals was checked bi-weekly by culturing faeces in thioglycollate medium under aerobic and anaerobic conditions for at least ten days. All handling procedures for GF mice were conducted in a laminar flow hood under sterile conditions. All experiments were conducted according to the German animal protection law.

51Food antigen-free (FAF) mice were raised on an amino acid-containing diet for up to five generations. Pellets of FAF diet (ssniff, S7242-E014/-E714) contained all essential vitamins, minerals, trace elements, fat, dextrin, sucrose and free amino acids equimolar to the protein content of normal rodent chow (LASQCdietRod16, LASvendi).

52Doxycycline (Dox)-inducible Btnl1-Tg mice were generated by injection of Btnl1−/− blastocysts with a linearized cassette containing a TRE/CMV-promoter upstream of the Btnl1-ORF. The TRE/CMV cassette has been previously described. R26-rtTA2-M2 or Villin-rtTA2-M2 mice were bred to homozygosity for Btnl1-deficiency and backcrossed onto Btnl-Tg mice for 3 generations to facilitate global (R26) or local (Villin) induction of Btnl1 transgene expression by doxycycline administered to drinking water (1mg/ml Dox, 2% suchrose). Animal experiments were undertaken in full compliance with UK Home Office regulations and under a project license to A.H. (80/2480).

53Flow cytometry was performed using the following antibodies, coupled to the indicated fluorochromes (see key resources table). Antibodies for mouse: CD3 APC Cy7 (17A2); CD3 PerCPCy5.5 (145-2C11); TCRβ Brilliant Violet 421 (H57-597); TCRβ APC (H57-597); CD122 PE (TMβ1); CD122 Brilliant Violet 421 (TMβ1); CD122 APC (TMβ1); TIGIT PE (GIGD7); CD45RB APC Cy7 (C363-16A); Thy1.2 Brilliant Violet 510 (53-2.1); Lag3 PerCP-efluor 710 (C9B7W); CD5 PE (53-7.3); CD24 FITC (M1/69); CD24 PECy7 (M1/69); CD8α PECy7 (53-6.7); CD8α PECy7 (53-6.7); TCR Vδ4 FITC (GL-2); TCR Vδ4 PE (GL-2); CD8β PerCpCy5.5 (YTS156.7.7); CD25 PerCpCy5.5 (PC61); CD69 PECy7 (H1.2F3); CCR9 PECy7 (CW-1.2); CD44 PECy7 (IM7); TCRVγ7 (F2.67) was provided by Pablo Pereira (Institut Pasteur, Paris, France); TCRVγ1 APC (2.11); TCRVγ4 APC (UC3-10A6); TCRδ BV421 (GL3); Ki67 FITC (B56/MOPC-21); CD45 Qdot 605 (30-F11); CD5 Brilliant Violet 510 (53-7.3); TCRδ PeCy7 (GL3); CD161/NK1.1 Brilliant Violet 650 (PK136); CD4 Brilliant Violet 786 (GK1.5); CD8α AlexaFluor 700 (53-6.7); CD25 APC (PC61); GITR PE (DTA-1); CD44 FITC (IM7); CD62L PerCP-Cy5.5 (MEL-14); KLRG1 BV421 (2F1); CD11c BV786 (HL3); CD11b BV510 (M1/70); F4/80 PerCPCy5.5 (BM8); Ly6G APC (1A8); Ly6C AlexaFluor 700 (AL-21); CD103 PE (M290); CD317 Brilliant Violet 650 (927); MHCII/IA/IE FITC (2G9); CD86 Pe-Cy7 (GL1); CD3 Brilliant Violet 421 (145-2C11); CD19 Brilliant Violet 421 (1D3); CD161/NK1.1 (lin) Brilliant Violet 421 (PK136); IgG1 PE (A85-1); B220 (CD45R) AlexaFluor 700 (RA3-6B2); IgM Brilliant Violet 786 (R6-60.2); IgD PerCPCy5.5 (11-26c.2a); GL-7 AlexaFluor 647 (GL7); CD95 PECy7 (Jo2); CD138 Brilliant Violet 650 (281-2); CD21/35 FITC (7G6); CD23 Brilliant Violet 421 (B-ly6). Antibodies for human: CD25 Brilliant Violet 421 (BC96); CD25 PE (BC96); CD3 Brilliant Violet 510 (OKT3); CD3 BUV (UCHT 1); EpCAM eFlour 660 (1B7); Streptavidin APC-Cy7; Streptavidin Brilliant Violet 421; TCRγδ PeCy7 (IMMU510); Vγ9 PC5 (IMMU360); Vγ9 PE (B3); Vδ1 APC (REA173); Vδ2 PerCP (B6); Vγ2/3/4 biotin (23D12), Vγ3/5 biotin (56.3) and Vγ8 biotin (R4.5.1) were provided by D. Kabelitz and D. Wesch (University of Kiel). Other antibodies: DYKDDDDK-PE (Flag); DYKDDDDK-APC (Flag); HA-DyLight 650; 6x-Histidine-PE.

54Commercial antibodies were purchased from Biolegend, eBioscience, BD-Bioscience, Thermo Fisher Scientific or Miltenyi (see key resource Information). Viability dyes (near IR or Blue) were from Invitrogen. Anti TCRVγ7 (F2.67) was purified from hybridoma supernatant using the mouse TCS purification system (Abcam) and conjugated to biotin or AF647 (see key resources table).

55Ki-67 staining was performed on cells fixed and permeabilised using the Foxp3 staining buffer set (eBioscience). BrdU (Sigma-Aldrich) and EdU incorporation was assessed 3h post-intraperitoneal injection (50mg/kg) by immunohistochemistry or by flow cytometry (Click-iT EdU AF647 Assay Kit, Invitrogen), respectively.

56Anti-TCRVγ7 (F2.67, provided by Pablo Pereira [Insitut Pasteur, Paris, France]) was purified from hybridoma supernatant using the mouse TCS purification system (abcam-ab128749). Purified anti-TCRVγ7 was conjugated to biotin (EZ-Link Sulfo-NHS-LC Biotinylation Kit, Thermo Fisher Scientific) or to AF647 (labeling kit, Thermo Fisher Scientific). Anti-human Vγ2/3/4 (23D12, biotinylated), Vγ5/3 (56.3, biotinylated) and Vγ8 (R4.5.1, biotinylated) were provided by D. Kabelitz & D.Wesch (Institute of Immunology, Kiel, Germany).

57Flow cytometry data analysis was performed on FlowJo (Version 9.9).

58The self-inactivating lentiviral vector pCSIGPW (SFFV promoter – Multiple Cloning Site [MCS] – IRES-GFP – CMV promoter – PuromycinR) was constructed by replacing the PuromycinR/mIR cassette from the pAPM vector by a custom EcoRI-XhoI-PmeI-NotI-BamHI-XbaI-MluI MCS. The IRES-GFP cassette was cloned by PCR from the pIRES2-eGFP vector (Clonetech) using the BamHI/XbaI sites. The CMV promoter was cloned by PCR from the pCDNA3.1+ vector (Thermo Fischer Scientific) using the MluI/ClaI sites. The Puromycin resistance gene was cloned by PCR from the pGIPZ vector (Dharmacon) using the ClaI/AgeI sites. The pCSIGHW variant was generated by exchanging the puromycin resistance gene with a hygromycin B resistance gene, which was cloned by PCR from the pLHCX vector (Clontech).

59cDNAs were (sub-)cloned into pCSIGPW or variant vectors (see Supplemental Information). Btnl1, Btnl4 and Btnl6 were previously described. BTNL3 (GenBank:NM_197975.2), BTNL8S (GenBank:NM_024850), and BTNL8 (GenBank:NM_001040462) were cloned from Caco-2 cells by conventional RT-PCR, using the following primers (See Table S2):BTNL3 For 5′-GAATATCCATGGCTTTTGTGC-3′BTNL3 Rev 5′-GTCTTCTCTGTCTCATCCCC-3′BTNL8 For 5′-CCATTCACAGAACACATCCATG-3′BTNL8S Rev 5′-TATGGGTTACAGTTTTCAGATCAG-3′BTNL8 Rev 5′-GTGGGATGTGATTCATCCTAC-3′

60BTNL3 For 5′-GAATATCCATGGCTTTTGTGC-3′

61BTNL3 Rev 5′-GTCTTCTCTGTCTCATCCCC-3′

62BTNL8 For 5′-CCATTCACAGAACACATCCATG-3′

63BTNL8S Rev 5′-TATGGGTTACAGTTTTCAGATCAG-3′

64BTNL8 Rev 5′-GTGGGATGTGATTCATCCTAC-3′

65FLAG, HA and HIS tags were added downstream of the putative leader peptides by overlapping PCR. Human full-length TCR γ and δ chains were cloned (XhoI / NotI, pCSIGPW) using the following primers (See Table S2):Vγ2/3/4 For 5′-ATGCAGTGGGCCCTAGCG-3′Vγ8 For 5′-ATGCTGTTGGCTCTAGCTCTGCTTC-3′Vγ9 For 5′-ATGCTGTCACTGCTCCACACATC-3′Cγ1/2 Rev 5′-TTATGATTTCTCTCCATTGCAGCA G-3′Vδ1 For 5′-ATGCTGTTCTCCAGCCTGCTG-3′Vδ2 For 5′-ATGCAGAGGATCTCCTCCCTCAT-3′Vδ3 For 5′-ATGATTCTTACTGTGGGCTTTAGCTTTTTG-3′Cδ Rev 5′-TTACAAGAAAAATAACTTGGCAGTCAAGAG-3′

66Vγ2/3/4 For 5′-ATGCAGTGGGCCCTAGCG-3′

67Vγ8 For 5′-ATGCTGTTGGCTCTAGCTCTGCTTC-3′

68Vγ9 For 5′-ATGCTGTCACTGCTCCACACATC-3′

69Cγ1/2 Rev 5′-TTATGATTTCTCTCCATTGCAGCA G-3′

70Vδ1 For 5′-ATGCTGTTCTCCAGCCTGCTG-3′

71Vδ2 For 5′-ATGCAGAGGATCTCCTCCCTCAT-3′

72Vδ3 For 5′-ATGATTCTTACTGTGGGCTTTAGCTTTTTG-3′

73Cδ Rev 5′-TTACAAGAAAAATAACTTGGCAGTCAAGAG-3′

74Expression of BTNL3 and BTNL8 was checked by conventional RT-PCR using the primers indicated above. BTN3A1, BTN3A2, EPCAM and GAPDH were used as control genes (See Table S2):BTN3A1 For 5′-AGTATCTCCTGATATGCAGCATG-3′BTN3A1 Rev 5′-GGAGGAACTCTCTTCTTCTTTTCAC-3′BTN3A2 For 5′-TGGTATCTCTTGATATGCAGCATAG-3′BTN3A2 Rev 5′-AGAGCATCAGGCTGACTTATTGG-3′EPCAM For 5′-GCCGCCACCATGGCGCCCCCGCAG-3′EPCAM Rev 5′-TTATGCATTGAGTTCCCTATGCA-3′GAPDH For 5′-GAAGGTGAAGGTCGGAGTC-3′GAPDH Rev 5′-GAAGATGGTGATGGGATTTC-3′

75BTN3A1 For 5′-AGTATCTCCTGATATGCAGCATG-3′

76BTN3A1 Rev 5′-GGAGGAACTCTCTTCTTCTTTTCAC-3′

77BTN3A2 For 5′-TGGTATCTCTTGATATGCAGCATAG-3′

78BTN3A2 Rev 5′-AGAGCATCAGGCTGACTTATTGG-3′

79EPCAM For 5′-GCCGCCACCATGGCGCCCCCGCAG-3′

80EPCAM Rev 5′-TTATGCATTGAGTTCCCTATGCA-3′

81GAPDH For 5′-GAAGGTGAAGGTCGGAGTC-3′

82GAPDH Rev 5′-GAAGATGGTGATGGGATTTC-3′

83Transfections were carried out in HEK293T cells using PEI (3:1 PEI:DNA ratio, Polysciences). Btnl/BTNL expression was checked 48h post-transfection. Lentiviral particles were produced in HEK293T cells by co-transfection of pCSIGPW or pCSIGHW either empty or containing Btnl/BTNL cDNAs, pCMVΔR8.91 (HIV-1 tat/rev/gag/pol), and pHIT/G (MLV env). Transduced cells were treated with puromycin and hygromycin 48h post-transduction for 7 days, sorted on the basis of GFP expression and used for functional assays.

84Samples were stored in RNAlater (Ambion) or directly frozen in RLT buffer prior to RNA purification (QIAGEN RNeasy kit). cDNA was generated using Superscript-II (Invitrogen) and analyzed using Sybr-green assay (Invitrogen) using a ViiA7 Real-time PCR machine (Applied Biosystems) (See Table S2).

85Btnl1 For: 5′-TGACCAGGAGAAATCGAAGG-3′Btnl1 Rev: 5′-CACCGAGCAGGACCAATAGT-3′Btnl4 For: 5′-CATTCTCCTCAGAGACCCACACTA-3′Btnl4 Rev: 5′-GAGAGGCCTGAGGGAAGAA-3′Bntl6 For: 5′-GCACCTCTCTGGTGAAGGAG-3′Btnl6 Rev: 5′-ACCGTCTTCTGGACCTTTGA-3′β-Actin For: 5′-CAGCTTCTTTGCAGCTCCTT-3′β-Actin Rev: 5′-CACGATGGAGGGGAATACAG-3′Sox-13 For: 5′-CTCCAGGCCTTCCCAGAC-3′Sox-13 Rev: 5′-CATGGACTTCCAGCGAGAAC-3′Rorγc For: 5′-GGTGACCAGCTACCAGAGGA-3′Rorγc Rev: 5′-CCACATACTGAATGGCCTCA-3′Tbp For: 5′-GGGGAGCTGTGATGTGAAGT-3′Tbp Rev: 5′-CCAGGAAATAATTCTGGCTCA-3′CycloFor: 5′-CAAATGCTGGACCAAACACAA-3′Cyclo Rev: 5′-CCATCCAGCCATTCAGTCTTG-3′

86Btnl1 For: 5′-TGACCAGGAGAAATCGAAGG-3′

87Btnl1 Rev: 5′-CACCGAGCAGGACCAATAGT-3′

88Btnl4 For: 5′-CATTCTCCTCAGAGACCCACACTA-3′

89Btnl4 Rev: 5′-GAGAGGCCTGAGGGAAGAA-3′

90Bntl6 For: 5′-GCACCTCTCTGGTGAAGGAG-3′

91Btnl6 Rev: 5′-ACCGTCTTCTGGACCTTTGA-3′

92β-Actin For: 5′-CAGCTTCTTTGCAGCTCCTT-3′

93β-Actin Rev: 5′-CACGATGGAGGGGAATACAG-3′

94Sox-13 For: 5′-CTCCAGGCCTTCCCAGAC-3′

95Sox-13 Rev: 5′-CATGGACTTCCAGCGAGAAC-3′

96Rorγc For: 5′-GGTGACCAGCTACCAGAGGA-3′

97Rorγc Rev: 5′-CCACATACTGAATGGCCTCA-3′

98Tbp For: 5′-GGGGAGCTGTGATGTGAAGT-3′

99Tbp Rev: 5′-CCAGGAAATAATTCTGGCTCA-3′

100CycloFor: 5′-CAAATGCTGGACCAAACACAA-3′

101Cyclo Rev: 5′-CCATCCAGCCATTCAGTCTTG-3′

102Southern blots were performed with probes generated using a Dig-Probe labeling kit; blots were hybridized in DIG-Easy-hyb buffer overnight, and developed using the DIG-Luminescence Detection Kit (Sigma-Aldrich). For probe sequences see Supplementary Information. DIG labeled probes for Southern blotting were generated using the following primers (See Table S2):Btnl1 For: 5′-ACTGGCTTCCTCAGAGTCAT-3′Btnl1 Rev: 5′-CAGTAGTGAATGGCCCCTGA-3′Btnl4 For: 5′-GACCAACGCTTCCCTACCTC-3′Btnl4 Rev: 5′-GCCTTGGGTCCAACAAGACA-3′Btnl1-Tg-Ex3-For: 5′-GGTTTTCTGTGAAGGGACCA-3′Btnl1-Tg-Ex4-Rev: 5′-GGTCTGCAACTCAGAGGAGG-3′

103Btnl1 For: 5′-ACTGGCTTCCTCAGAGTCAT-3′

104Btnl1 Rev: 5′-CAGTAGTGAATGGCCCCTGA-3′

105Btnl4 For: 5′-GACCAACGCTTCCCTACCTC-3′

106Btnl4 Rev: 5′-GCCTTGGGTCCAACAAGACA-3′

107Btnl1-Tg-Ex3-For: 5′-GGTTTTCTGTGAAGGGACCA-3′

108Btnl1-Tg-Ex4-Rev: 5′-GGTCTGCAACTCAGAGGAGG-3′

109RNAscope was performed on paraffin embedded sections using probes and kits obtained from Advanced Cell Diagnostics using the RNAscope 2.0 HD Reagent Kit-BROWN. Reference sequences are as follows: Btnl1, GenBank:NM_001111094.1 (576-1723); Btnl4, GenBank:NM030746.1 (560-968); Btnl6, GenBank:NM_030747.1 (245-1552) (See Table S2).

110IEL were isolated from mouse small intestine as previously described. Small intestine was opened and washed in PBS, cut into 1cm pieces and incubated for 20min in RPMI 1640 supplemented with 1% penicillin/streptomycin (pen/strep), 10% fetal calf serum (FCS) and 1mM dithiothreitol on a turning wheel. Tissues were washed and vortexed in RPMI, then passed through a 70 μm nylon cell strainer twice, and centrifuged on a 20/40/80% Percoll density gradient at 700 g for 30min. IEL were harvested from the 40 to 80% Percoll interface.

111Comprehensive immunophenotyping of Btnl1-/- mice was performed using a platform developed by the Wellcome Trust Infection and Immunity Immunophenotyping (3i) consortium (www.immunophenotyping.org). In brief, Spleen and MLN were digested with collagenase (1mg/ml)/DNAse (0.1 mg/ml) in 2% FCS PBS (+ Ca/Mg) for 20 minutes at 37°C and filtered through 30μm cell strainers. Cells were plated on 96 well V-bottom plates, washed in PBS and stained with Zombie Near-IR (Biolegend) for live/dead discrimination. Antibody stains were performed at 4°C for 20mins. Full details regarding phenotyping panels are included in Table S3. Samples were acquired on a BD LSR Fortessa X-20 equipped with 405nm (40mW), 488nm (50mW), 561nm (50mW), and 640nm (100mW) lasers.

112Cells were co-cultured in RPMI 1640 supplemented with 10% FCS, Pen/Strep, 2.5% HEPES, 1% Glutamine, 1% non-essential amino acids, 1% sodium pyruvate, 0.2% β-mercapto-ethanol (GIBCO) and cytokines including IL-2 (10U/ml), IL-15 (10 ng/ml) (Immunotools), IL-3 (100U/ml), IL-4 (200U/ml) (R&D). 105 MODE-K were seeded in 48-well plates 24h prior to the addition of 105 unsorted or (where indicated) positively FACS-sorted (CD45+Vγ7+) IEL and incubated for 16-18h in 10% CO2 unless indicated otherwise. For transwell assays, 2x105 MODE-K cells were seeded onto 24-well transwell plates (3 μm pore size - Corning) 24h prior to the addition of 3x105 IEL, either in direct contact (below), sequestered from (above), or split 50:50 with MODE-K cells (above and below the transwell).

11396-well U bottom plates were coated overnight with 10 μg/ml LEAF-Purified anti-mouse CD3-ε or Hamster IgG Isotype control (Biolegend) at 4°C and washed once with PBS 1x before seeding IEL. 100,000 IEL were seeded per well. Cells were incubated at 37°C for 16-18h in 10% CO2 prior to analysis.

114Proximal small intestine (SI) samples were fixed in Zamboni’s fixative, blocked with normal goat serum and stained with antibodies against TCRβ, TCRδ, TCRVδ4 (encoded by TRDV2-2) (GL2), CD3 and Vγ7. Z-Sections were acquired on a confocal-LSM-710 microscope (Zeiss) and processed and analyzed using Imaris Software (Bitplane Scientific Solutions).

11510-12 week old recipient mice were irradiated with 950Rads 24h, injected (IV) with 5-10x106 donor bone marrow cells and analyzed 4-12 weeks later.

116IEL harvested from 4 week-old WT mice were column-purified using CD45 microbeads (MACS Miltenyi biotec) and IV-injected into 6 week-old TCRδ−/− and TCRδ−/−Btnl1−/− recipients. Analysis was performed 2-3 weeks later.

117Vγ7+CD122hi and Vγ7+CD122lo IEL were sorted from from pooled D14-17 pups directly into RLT buffer. RNA was prepared using the RNA-Micro-plus kit (QIAGEN). RNA libraries were generated using the KAPA Stranded RNA-seq Kit with RiboErase (HMR) (KAPA BIOSYSTEMS). Paired-end sequencing on HiSeq 2500 (illumina) using rapid run chemistry (read length: 100bp).

118Endoscopic biopsies were obtained from the ascending colon of adult donors undergoing routine diagnostic colonoscopy after informed consent and in compliance with local ethical approval (REC number 07/H0803/237). Excess resected skin discarded at the time of cutaneous or reconstructive surgery was obtained from adult donors after informed consent and in compliance with local ethical approval (REC number 06/Q0704/18). This study was conducted adhering to the principles of the Declaration of Helsinki.

119Primary gut lymphocytes were obtained using an adaptation of the method of Kupper and Clarke (Figure S7F). Skin lymphocytes were isolated using the method as originally described. 9mm x 9mm x 1.5mm Cellfoam matrices (Cytomatrix PTY Ltd), were autoclaved and incubated in 100mg/mL rat tail collagen I (BD Biosciences) in PBS for 30min at 37°C, and washed twice in PBS. In compliance with local ethical approval, 12 endoscopic biopsies were taken from the ascending colon of donors. Biopsies were washed for 20min in 5mL wash medium (RPMI 1640 10% FCS, β-mercaptoethanol, penicillin [500U/ml], streptomycin [500 μg/ml], metronidazole [5 μg/ml, Pharmacy department, Guy’s Hospital], gentamicin [100 μg/ml, Sigma-Aldrich] and amphotericin 12.5 μg/ml [Thermo Fisher Scientific]). One endoscopic biopsy was placed on top of each matrix, which was inverted, and pressure applied, to crush the biopsy into the matrix. The matrices were placed into a 24-well plate (1 per well) and covered with 2mL RPMI 1640 (supplemented with 10% FCS, β-mercaptoethanol, penicillin [100U/ml], streptomycin [100 μg/ml], metronidazole [1 μg/ml], gentamicin [20 μg/ml], amphotericin [2.5 μg/ml]), IL-2 (100U/mL, Novartis Pharmaceutical UK) and IL-15 (10ng/mL, Biolegend). 1 ml of medium was aspirated every second day and replaced with complete medium containing 2x concentrated cytokines. Cells were harvested and residual biopsy and empty wells were washed with PBS 0.02mM HEPES. The cell suspension was passed through a 70 μm nylon cell strainer, centrifuged at 400 g for 5min and resuspended in complete medium without additional cytokine and placed into co-culture immediately. Lymphocytes were used after 5-7 days of culture.

120PBMC were isolated by Ficoll gradient from blood obtained from the blood donation service.

121Colonic samples were incubated with 5 mM 1,4-dithiothreitol (Sigma), followed by enzymatic digestion with 1.5mg/ml collagenase VIII (Sigma) and 0.05 mg/mL DNase I (Sigma). EpCAM+ cells were sorted by flow cytometry directly into RLT lysis buffer. RNA and cDNA were prepared as described above.

1225x105 HEK293T cells, transduced with either empty vector (EV), BTNL3, BTNL8 or BTNL3+8 and 2x105 freshly harvested primary human lymphocytes were co-cultured in 96-well plates with complete medium (see Supplementary Information) without supplementary cytokine and incubated at 37°C at 5% CO2 for 16hrs (Figure S7F).

123Mouse TRDV gene: Amplification and sequencing of TCRδ CDR3 from RNA purified from sorted Vγ7+ IEL was performed using the Amp2Seq Platform (iRepertoire).Human TCRG Vγ gene: Amplification and sequencing of TCRγ CDR3 was performed using the immunoSEQ Platform (Adaptive Biotechnologies).

124Unless stated otherwise, bar/spider charts display mean ± SD and p values were derived from unpaired two tailed t tests, assuming equivalent SD (ns > 0.05).

125Confocal microscopy was performed using a LSM710 laser scanning confocal microscope (Zeiss) with a 40x oil objective (numerical aperture 1.3). 3D image analysis on z-stacks was carried out using Imaris (Bitplane). The surfaces tool was used to identify CD3+ cells. Voxels outside of these structures were set to zero in each of the channels to create masks.

126101 base-pair paired-end reads were aligned and quantified using RSEM (v1.2.11) with Bowtie2. Reads were aligned to a transcriptome constructed from the mm10 mouse genome and a UCSC knownGene gtf file. A mean alignment rate of 57.4 million fragments per sample was observed. Using the gene level quantification, only detected genes (mean TPM value across all samples > 1; 13,313 genes) were selected. Differential expression between the CD122hi and CD122lo Vγ7+ IEL groups using DESeq2 was identified by taking into account the paired structure within the replicate groups. Using an FDR of 0.01 2664 phenotype dependent gene expression effects were identified.

127The accession number for the RNA sequencing data reported in this paper is GEO: GSE85422.

128Conceptualization, A.H.; Methodology, A.H., A.J., L.D., R.D.M.B., P.V., R.J.D., O.N., U.S., and R.H.; Resources, D.G., P.M.I., P.P., and U.S.; Investigation, R.D.M.B., N.A.R., R.J.D., P.V., A.J., O.N., L.D., S.C., R.H., M.L.I., A.L., B.S.-D., P.E., and U.S.; Validation, R.J.D., L.D., and A.J.; Data Curation, R.D.M.B., B.S.-D., P.E., P.V., and A.H.; Writing – Original Draft, A.H.; Writing – Review & Editing, A.H., U.S., R.D.M.B., N.A.R., R.J.D., P.V., A.J., L.D., P.P., and S.C.; Visualization, R.D.M.B., N.A.R., R.J.D., A.J., R.H., B.S-D., P.V., and A.L.; Supervision, A.H.; Funding Acquisition, U.S. and A.H.