Runx–CBFβ regulates the development of tolerogenic Thetis cells

Abstract

Peripherally induced regulatory T (pT_reg) cells play an essential role in establishing immune tolerance to gut microbiota and food antigens. A subset of RORγt-expressing antigen-presenting cells, Thetis cell subset IV (TC IV), plays an essential role in intestinal tolerance. However, the transcription factors governing the differentiation of this subset remain unclear. Here we show that Runx–CBFβ complexes regulate the development of specific TC subsets. Mice lacking CBFβ2 exhibited loss of TC II, III and IV, with loss of RORγt⁺ pT_reg cells. Transgenic CBFβ2 expression by Cd11c-Cre restored TC IV and RORγt⁺ pT_reg cell differentiation in CBFβ2-deficient mice. Conditional inactivation of Runx1 and Runx3 by Cd11c-Cre selectively impaired TC III and IV. Conversely Cd11c-Cre-driven transgenic Runx expression enhanced TC IV differentiation and thus RORγt⁺ pT_reg cell induction. These findings establish a critical pathway for TC IV differentiation and provide new insights into therapeutic interventions to promote RORγt⁺ pT_reg cell induction in autoimmune diseases.

Taniuchi and Brown and colleagues report that the Runx–CBFβ transcription factor complex plays a crucial role in the development of intestinal RORγt⁺ antigen-presenting cells, which are required to promote tolerance to food and gut commensal microorganisms.

Main

Establishing and maintaining tolerogenic immune responses to innocuous foreign antigens is vital for human health. Foxp3-expressing regulatory T (T_reg) cells are essential for immune tolerance and are first developed in the thymus. Medullary thymic epithelial cells that express the autoimmune regulator (Aire) present self-antigens, such as tissue-specific antigens, to guide some autoreactive T cells to differentiate into T_reg cells^1–4. Following encounter with antigens derived from the commensal microbiome and food, naive CD4⁺ T cells in the periphery can be induced to become Foxp3⁺ T_reg (pT_reg) cells, especially RORγt-expressing pT_reg cells, which are key to establishing immune tolerance in the gut^5–8. This process requires T cell priming by tolerogenic antigen-presenting cells (APCs)⁹.

Recent work has established a new paradigm for intestinal tolerance, demonstrating the existence of a novel lineage of RORγt⁺ APCs, Thetis cells (TCs), which encompass four distinct subsets¹⁰. Several studies have described transcriptionally overlapping RORγt⁺ APC subsets with variable nomenclature, including TC I–IV^10,11, RORγt⁺ dendritic cells (DCs) I–IV¹², Janus cells I–II^13,14, RORγt⁺ extrathymic Aire-expressing cells 1–3 (ref. ¹⁵) and PRDM16⁺ tolerizing DCs¹⁶, mainly based on single-cell RNA-sequencing (scRNA-seq) data. As a result, the same cellular subset has been given multiple names¹⁷. A recent integrated analysis of these scRNA-seq datasets revealed transcriptional concordance and resolved the nomenclature across studies. In addition, Smart-seq3 analysis of TC I–IV revealed unique expression patterns of several cell surface markers, allowing for analysis of individual TC subsets by flow cytometry^10,11. As previously shown^10,11, TCs (CXCR6⁻RORγt⁺MHCII⁺) can be divided into NCAM1⁺ TC I and EPCAM⁺NCAM1⁻ cells, which comprise CCR6⁺NRP1⁺ TC II, CCR6⁻CD11c⁺CD11b⁻ TC III and CCR6⁻CD11c⁺CD11b⁺ TC IV subsets, whereas major histocompatibility complex class II⁺ (MHCII⁺) group 3 innate lymphoid cells (ILC3s), also referred to as lymphoid tissue inducer (LTi) cells, are CXCR6⁺RORγt⁺MHCII⁺. Different genetic approaches have been used to decipher the TC subset(s) responsible for pT_reg cell differentiation. These studies have shown that tolerogenic APCs reside within α_vβ₈-expressing TCs that lack Aire expression, suggesting that Aire⁻α_vβ₈⁺ TC IV are the tolerogenic subset^10,12,16,18. Other studies have highlighted a role for the transcription factors PRDM16 and RORγt in the regulation of particular TC subsets^12,16. Although, the existence of distinct RORγt⁺ APC subsets has became clear, the hierarchical differentiation pathway and the transcription factors that regulate the development of individual TC subsets remain poorly characterized.

Intriguingly, Runx3 gene inactivation in hematopoietic cells results in spontaneous colitis development¹⁹, in line with the genetic association of Runx3 in human inflammatory bowel disease²⁰. Runx family transcription factors function as obligate heterodimers with CBFβ, a non-DNA-binding cofactor that stabilizes Runx proteins and is essential for their transcriptional activity. Although Runx–CBFβ transcriptional heterodimer complexes have pleiotropic effects on immune cell differentiation, including thymic-derived T_reg cells^21,22, effector CD4⁺ T cells^23,24, ILC3s/LTi cells^25,26 and classical DCs (cDCs)²⁷, the cellular mechanisms underlying Runx–CBFβ regulation of intestinal tolerance have not been established. Here, we sought to investigate the role of Runx–CBFβ in differentiation of RORγt⁺ APCs, encompassing TCs and LTi cells.

Results

CBFβ2-deficient mice lack RORγt⁺ pT_reg cells and develop colitis

The mouse Cbfb gene produces two functional variants, CBFβ1 and CBFβ2, by alternative RNA splicing using different splicing donor signals within exon 5. We previously generated a mouse line that specifically lacks the CBFβ2 variant due to mutations²⁸, referred to as Cbfb^2m/2m mice. We found that Cbfb^2m/2m mice spontaneously developed colitis with associated histological changes and ensuing weight loss between 8 and 12 weeks of age (Fig. 1a and Extended Data Fig. 1a). Given the critical role of RORγt⁺ pT_reg cells in gut microbiota tolerance^5,6,8, we examined T cell subsets in small and large intestine lamina propria (SI-LP and LI-LP, respectively) and found that RORγt⁺ pT_reg cells were reduced in both the SI-LP and LI-LP of Cbfb^2m/2m mice, whereas RORγt⁺Foxp3⁻ interleukin-17 (IL-17)-producing helper T (T_H17) cells were slightly increased in the SI-LP of Cbfb^2m/2m mice (Fig. 1b).

Fig. 1. Cbfb^2m/2m mice develop colitis due to dysfunction of APCs.

a, Left, representative images of the colon of Cbfb^+/+ (n = 6) and Cbfb^2m/2m mice (n = 6). Right, summary of colitis score. Data are shown as mean ± s.d. and were analyzed by an unpaired t-test (two-sided); ****P < 0.0001. b, Left, zebra plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of the SI-LP and LI-LP of Cbfb^+/+ (n = 7) and Cbfb^2m/2m (n = 7) mice. Right, statistical summary of the frequency of the indicated cell subsets in CD4⁺ T cells. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05; ***P < 0.001. c, Left, zebra plots showing RORγt and Foxp3 expression in CD4⁺ T cells of the SI-LP and LI-LP of mice with the indicated genotypes (Cbfb^+/+ (SI-LP: n = 10, LI-LP: n = 7), Cbfb^2m/2m (SI-LP: n = 9, LI-LP: n = 6), Vav1-iCre (SI-LP: n = 3, LI-LP: n = 3), Mb1-Cre (SI-LP: n = 4, LI-LP: n = 5), Cd4-Cre (SI-LP: n = 3, LI-LP: n = 3), Cd11c-Cre (SI-LP: n = 4, LI-LP: n = 5), Rorc-Cre (SI-LP: n = 8, LI-LP: n = 8) and Il7r-Cre (SI-LP: n = 3, SI-LP: n = 3)). Transgenic CBFβ2 protein is expressed from the Rosa26^lsl-Cbfb2 allele after excision of lsl sequences by Cre-mediated recombination. Right, statistical summary of the frequency of RORγt⁺Foxp3⁺ pT_reg cells in the SI-LP and LI-LP. Data are shown as mean ± s.d. and were analyzed by one-way analysis of variance (ANOVA); *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. d, Top, experimental schematic. Middle, pseudocolor plots showing representative RORγt and Foxp3 expression in CD45.2⁺OT-II⁺CD4⁺ T cells in the mLNs 1 week after transfer into 3-week-old CD45.1:Cbfb^+/+ (n = 6) and CD45.1:Cbfb^2m/2m (n = 5) recipient mice. Bottom, statistical summary of the cell number and frequency of Foxp3⁺RORγt⁻ and Foxp3⁺RORγt⁺ subsets within the CD45.2⁺OT-II⁺ population. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); ***P < 0.001; ****P < 0.0001; P21, postnatal day 21.

Source data

Extended Data Fig. 1. Restoration of RORγt⁺ pTreg cell by transgenic Cbfβ1 in Cbfb^2m/2m mice.

(a) Representative images of large intestine of four (n = 3), six (n = 3) and eight (n = 3) weeks old Cbfb^2m/2m mice. The right graph showing summary of colitis scores. Mean ± SD. **: p < 0.01, unpaired t-test (two-sided) (b) A scheme showing the structure of the Rosa26 allele with cDNA fragment inserted for the inducible expression of transgenic Cbfβ1 after removal of loxP-Neo/STOP-loxP (lsl) sequences by Cre-mediated site-specific recombination. (c) One representative image of two immunoblot experiments showing Cbfβ1 expression in thymocytes of mice with indicated genotype. The graph at right showing numbers of Peyer’s patches (PPs) of Cbfb^+/+ (n = 5), Cbfb^2m/2m (n = 5) and Cbfb^2m/2m:Rosa26^lsl-Cbfb1/+:Vav-iCre (n = 3) mice. Mean ± SD. ****: p < 0.0001, One-way Anova analyses. (d) Zebra plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of SI-LP of Cbfb^+/+ (n = 4), Cbfb^2m/2m (n = 4) and Cbfb^2m/2m:Rosa26^lsl-Cbfb1/+:Vav-iCre (n = 5) mice. Graph showing a statistical summary of the frequency of indicated cell subsets in CD4⁺ T cells. Mean ± SD. *: p < 0.05, **: p < 0.01, unpaired t-test (two-sided). (e) Contour plots showing representative RORγt and Foxp3 expression in CD45.2 OT-II⁺ CD4⁺ T cells in the SI-LP one week after their transfer into three-week-old CD45.1: Cbfb^+/+ (n = 6) and CD45.1: Cbfb^2m/2m (n = 5) recipients mice. Mean ± SD. ***: p < 0.001, unpaired t-test (two-sided). (f) Histograms showing RORγt and PRDM16 expression in indicated cell subsets. Mean fluorescent intensity (MFI) of RORγt expression was indicated. (g) Graph showing the statistical summary of the frequencies of the indicated cell subsets in mLNs of Cbfb^+/+ (n = 3) and Cbfb^2m/2m (n = 3) mice, after gating cell population as shown in Fig. 2c. Mean ± SD. *: p < 0.05, **: p < 0.01, unpaired t-test (two-sided).

Source data

To determine the cell type responsible for impaired RORγt⁺ pT_reg cell differentiation in mice with germline Cbfb^2m mutation, we took advantage of Rosa26^lsl-Cbfb2/+ mice to allow for lineage-specific Cre-driven removal of the ‘loxP-STOP-loxP’ (lsl) allele²⁹ to selectively rescue CBFβ2 expression across a range of immune cell types. As expected, differentiation of RORγt⁺ pT_reg cells was restored in Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Vav1-iCre mice (Fig. 1c). Restoration of RORγt⁺ pT_reg cell differentiation by transgenic CBFβ1 expression suggested that low dosage of CBFβ rather than loss of specific function of CBFβ2 is causal for loss of RORγt⁺ pT_reg cells in Cbfb^2m/2m mice (Extended Data Fig. 1b–d). By contrast, RORγt⁺ pT_reg cells were not rescued by transgenic CBFβ2 expression in B or T lymphocytes in Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Mb1-Cre or Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd4-Cre mice, respectively (Fig. 1c). These results indicate that impaired RORγt⁺ pT_reg cell differentiation in Cbfb^2m/2m mice occurs in hematopoietic cells but by T cell-extrinsic mechanisms.

To further address this point, we used the ovalbumin (OVA)-specific CD4⁺ OT-II T cell antigen receptor (TCR)-transgenic mouse model. We examined the differentiation of CD45.2⁺ naive CD4⁺ OT-II T cells following adoptive transfer into CD45.1 Cbfb^+/+ and Cbfb^2m/2m recipients and OVA feeding. In both mesenteric lymph nodes (mLNs) and the SI-LP, ~57% of OT-II cells differentiated into Foxp3⁺ pT_reg cells in Cbfb^+/+ recipients, whereas pT_reg cell differentiation was abolished in Cbfb^2m/2m recipients (Fig. 1d and Extended Data Fig. 1e). In addition, RORγt⁺Foxp3⁺ cells were detected in Cbfb^+/+, but not in Cbfb^2m/2m, recipients.

Given the role of RORγt⁺ APCs in pT_reg cell differentiation, we next generated Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Rorc-Cre and Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd11c-Cre mice. Although Rorc-Cre led to variable rescue in RORγt⁺ pT_reg cells, we observed complete restoration of RORγt⁺ pT_reg cell differentiation when transgenic CBFβ2 expression was induced by Cd11c-Cre (Fig. 1c), in line with previously reported loss of RORγt⁺ pT_reg cells by Cbfb inactivation by Cd11c-Cre²⁷. Such variegated rescue by Rorc-Cre led us to wonder whether CBFβ2 expression was required at an earlier stage of TC development before full acquisition of RORγt expression. Given previous reports of IL-7R fate mapping in TCs^10,11 and suggested descendancy from IL-7R⁺ lymphoid progenitors^11,12, we turned to Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Il7r-Cre mice. In these mice, we observed complete restoration of RORγt⁺ pT_reg cells (Fig. 1c). Together, these observations indicate that RORγt⁺ cells that express CD11c and IL-7R require Runx–CBFβ2 for development and/or pT_reg cell-inducing function. Overall, these findings suggest a role for Runx–CBFβ in RORγt⁺ APC development.

CBFβ2-deficient mice lack TC II–IV

To determine the role of Runx–CBFβ in TC differentiation, we examined RORγt⁺ APC subset composition in the mLNs of Cbfb^+/+ and Cbfb^2m/2m mice by flow cytometery using several cell surface markers, which enabled us to identify individual TC subsets (Extended Data Fig. 2). Analysis of these RORγt⁺ APC subsets in mLNs revealed almost complete absence of TC II, III and IV in Cbfb^2m/2m mice, but no change in TC I numbers (Fig. 2a). In addition, MHCII⁺ ILC3/LTi cell numbers were modestly impaired, in keeping with previous reports of CBFβ2 regulation of LTi cell development²⁵.

Extended Data Fig. 2. Schematic summary of nomenclature of RORγt⁺MHCII⁺ APCs.

RORγt⁺MHCII⁺ APCs are first divided into CXCR6⁺ ILC3/LTi cells and CXCR6⁻ Thetis cells (TCs).TC s are further divided into at least fours subsets, TC I, II, III and IV according to their unique gene expression profile and expression of indicated surface markers, which can be used as a reliable strategy to separate these fours TC subsets by flow cytometry. TC subsets have been referred to by different names across various studies. This scheme provides a cross‑reference of the alternative nomenclature used for each TC subset. JC, Janus cell; TolDC, tolerizing dendritic cell; R-eTAC, RORγt⁺ extra-thymic Aire-expressing cell.

Fig. 2. Severe reduction of TCs in Cbfb^2m/2m mice.

a, Left, representative flow cytometry plots showing gating of APC types in the RORγt⁺MHCII⁺ population in the mLNs of 2- to 3-week-old Cbfb^+/+ (n = 4) and Cbfb^2m/2m (n = 4) mice. Right, statistical summary of numbers of the indicated cell subsets. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05. b, Left, histogram showing RORγt expression in TC I from Cbfb^+/+ (n = 3), Cbfb^2m/2m (n = 3), Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Il7r-Cre (n = 3) and Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Vav1-iCre (n = 3) mice. Right, histograms showing PRDM16 expression in TC I of Cbfb^+/+ (n = 9), Cbfb^2m/2m (n = 7), Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Il7r-Cre (n = 3) and Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Vav1-iCre (n = 3) mice. Bottom, statistical summary of RORγt mean fluorescence intensity (MFI; left) and the frequency of PRDM16⁺ TC I (right). Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; *P < 0.05; ***P < 0.001; ****P < 0.0001. c, Top, representative flow cytometry plots showing gating of APC types in the Lin⁻RORγt⁻MHCII⁺ population in the mLNs of Cbfb^+/+ (n = 3) and Cbfb^2m/2m (n = 3) mice. Bottom, statistical summary of the numbers of the indicated cell subsets. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Source data

In our prior study, we identified a role for Runx–CBFβ in the regulation of Rorc expression²⁸. Given recent reports demonstrating dependence of RORγt⁺ APCs on a +7-kb cis-regulatory element in the Rorc locus^12,16 and a transcriptional co-regulator PRDM16 (refs. ^11,16,30), we wondered whether Runx–CBFβ was required for RORγt and PPRDM6 expression during TC development. Among TC subsets, we observed differential levels of RORγt protein expression, with the highest level in TC II and low levels in TC I (Extended Data Fig. 1f). In addition, analysis of PRDM16 revealed expression within TC II, III and IV (Extended Data Fig. 1f), but only low levels in a fraction of TC I, consistent with published scRNA-seq results^10,12,15. Intriguingly, although TC I numbers were not impacted in Cbfb^2m/2m mice, the level of RORγt expression was reduced (Fig. 2b), suggesting a role for Runx–CBFβ in regulation of Rorc expression in TCs. In contrast to the modest decrease in RORγt expression in TC I, we observed complete absence of PRDM16-expressing TC I in Cbfb^2m/2m mice (Fig. 2b). These observations suggested additional roles for Runx–CBFβ in the regulation of Prdm16 expression, presumably independent of RORγt. Both RORγt expression levels and the frequency of PRDM16⁺ TC I were restored by transgenic CBFβ expression, most notably when driven by the Vav-iCre transgene (Fig. 2b).

We also examined the RORγt⁻ APC compartment, comprising cDCs¹⁷, in the mLNs of 2- to 3-week-old Cbfb^+/+ and Cbfb^2m/2m mice. Total numbers of both CCR7⁺ and CCR7⁻RORγt⁻ APCs were higher in Cbfb^2m/2m mice than in Cbfb^+/+ mice (Fig. 2c). In the CCR7⁺ migratory compartment, both CD11b⁻CD103⁺ cDC1 and C11b⁺CD103⁺ cDC2 populations were increased in Cbfb^2m/2m mice, whereas CD103-expressing cells in the CCR7⁻ population were almost lost in Cbfb^2m/2m mice (Fig. 2c and Extended Data Fig. 1g).

Runx3 is required for TC development

CBFβ is a common partner for all three mouse Runx family proteins, Runx1, Runx2 and Runx3. Among these, dysfunction of Runx3 has been shown to result in spontaneous colitis development¹⁹. To specifically address the role of Runx3 in intestinal tolerance, we generated two mouse lines. In the Runx3^WRPW allele, the terminal tyrosine residue within the WRPY motif in endogenous Runx3 was replaced with tryptophan. The WRPY motif is known to serve as a platform for recruiting transducin-like enhancer of split (TLE) co-repressor family proteins^31,32. Because the WRPW motif in bHLH transcription factors, such as HES1, has higher affinity to TLE family proteins³³, the tyrosine-to-tryptophan replacement transforms Runx3 protein to a dominant repressor and interferes with Runx3-dependent gene activation. Consequently, the differentiation of several Runx3-dependent immune cell populations, including CD8⁺ T cells and dendritic epidermal T cells, was impaired in Runx3^+/WRPW mice³⁴. Another mutant Runx3 allele, which we refer to as Runx3^AID, was generated with the aim of enabling inducible Runx3 degradation through the auxin-inducible degron (AID) 2 system³⁵.

In the Runx3^AID allele, a DNA fragment encoding mini-Auxin-inducible degron (mAID) degron sequences was inserted upstream of the C-terminal end of Runx3, resulting in the production of a Runx3–mAID fusion protein (Extended Data Fig. 3a). Unfortunately, insertion of the mAID degron tag at this position led to impaired Runx3 function, as evidenced by the absence of Peyer’s patches (PPs; Extended Data Fig. 3a) and most peripheral LNs (pLNs) in Runx3^AID/AID mice. The frequency and number of B cells (B220⁺), γδ T cells (TCRγδ⁺), monocytes (CD64⁺) and natural killer (NK) cells (NK1.1⁺) were comparable between Runx3^+/+ and Runx3^AID/AID mice (Extended Data Fig. 3b). By contrast, within the splenic αβ T cell compartment, the frequency of CD8⁺ T cells was reduced, resulting in an increased CD4⁺:CD8⁺ T cell ratio. In addition, CD103 expression was nearly lost in splenic CD8⁺ T cells (Extended Data Fig. 3c). However, we did not observe any apparent CD4 derepression, a typical phenotype caused by loss of Runx3 function in CD8⁺ T cells³⁶. These observations indicated that the Runx3^AID allele functions as a hypomorphic Runx3 allele. Analysis of Runx3^+/WRPW mice revealed absence of RORγt⁺ pT_reg cells (Fig. 3a) with histological evidence of colitis by 3 months of age (Extended Data Fig. 3d). pT_reg cell generation was likewise impaired (Fig. 3b), with development of colitis by 3 months of age in Runx3^AID/AID mice (Extended Data Fig. 3e).

Extended Data Fig. 3. Colitis development by attenuated Runx3 function.

(a) A schematic structure of around exon 6 in murine Runx3 gene and structure of donor ssDNA for insertion of mAID sequence. Amino acid sequences at the C-terminal of Runx3^AID mutant protein are shown; GS linker (green), mAID (red) and Runx3 derived amino acid (black). One representative result of two immunoblots with α-Runx3 antibody showing Runx3 and Runx3^AID protein expression in CD8⁺ T cells. Gapdh immunoblot was used as a loading control. The graph at bottom right showing numbers of PPs of Runx3^+/+ (n = 6) and Runx3^AID/AID (n = 4) mice. Mean ± SD. ****: p < 0.0001, unpaired t-test (two-sided). (b) Graphs summarizing the frequencies and the numbers of B cells (B220⁺), γδT cells (TCRγδ⁺), monocytes (CD64⁺) and NK cells (NK1.1⁺) in the spleens of 3-weeks-old Runx3^+/+ (n = 3) and Runx3^AID/AID (n = 3) mice. (c) Representative pseudocolor plots showing CD4 and CD8α expression in spleen TCRβ⁺ population of Runx3^+/+ (n = 4) and Runx3^AID/AID (n = 3) mice. Histograms showing CD103 expression in CD8⁺ T cells. The graphs on the left and right showing CD4⁺/CD8α⁺ cell ratio within splenic TCRβ⁺ population and the frequency of CD103⁺ cells among CD8⁺ T cells, respectively. Mean ± SD. **: p < 0.01, ****: p < 0.0001, unpaired t-test (two-sided). (d) Representative images of colon of Runx3^+/+ (n = 4), Runx3^+/WRPW (n = 4) mice. Graph at right showing a summary of colitis score. Mean ± SD. **: p < 0.01, unpaired t-test (two-sided). Bars indicate 100 μm scale. (e) Representative images of colon of Runx3^+/+ (n = 5), Runx3^AID/AID (n = 7) mice. Graph at right showing a summary of colitis score. Mean ± SD. ****: p < 0.0001, unpaired t-test (two-sided). Bars indicate 100 μm scale. (f) Representative pseudocolor plots showing RORγt and MHCII expression in Lin⁻ population of mLNs of Runx3^+/+ (n = 6), Runx3^+/WRPW (n = 4) and Runx3^AID/AID (n = 4) mice. Graph showing absolute numbers of RORγt⁺MHCII⁺ cells. Mean ± SD. ***: p < 0.001, One-way Anova analyses. (g) Histogram showing PRDM16 expression in TC I subset of Runx3^+/+ (grey, n = 3) amd Runx3^AID/AID (green, n = 4) mice. The graph showing the summary of the frequency of PRDM16⁺ TC I cells. Mean ± SD. **: p < 0.01, unpaired t-test (two-sided).

Source data

Fig. 3. Severe reduction of TCs in Runx3-mutant mice.

a, Contour plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of the SI-LP and LI-LP of Runx3^+/+ (n = 3) and Runx3^+/WRPW (n = 3) mice. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05; **P < 0.01. b, Zebra plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of the SI-LP and LI-LP of Runx3^+/+ (SI-LP: n = 7, LI-LP: n = 3) and Runx3^AID/AID (SI-LP: n = 6, LI-LP: n = 4) mice. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); ***P < 0.001; ****P < 0.0001. c, Left, representative flow cytometry data showing gating of APC types in RORγt⁺MHCII⁺ populations of mLNs of Runx3^+/+ (n = 3) and Runx3^+/WRPW (n = 4) mice. Right, statistical summary of numbers of RORγt⁺MHCII⁺ cells. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05; **P < 0.01. d, Left, representative flow cytometry data showing gating of APC types in the RORγt⁺MHCII⁺ population of mLNs of Runx3^+/+ (n = 3) and Runx3^AID/AID (n = 4) mice. Right, statistical summary of numbers of RORγt⁺MHCII⁺ cells. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05; **P < 0.01; ***P < 0.001.

Source data

Similar to Cbfb^2m/2m mice, we did not observe defects in the mLNs in Runx3^+/WRPW and Runx3^AID/AID mice, although pLN and PP development was also disrupted in Runx3^+/WRPW mice³⁴, as was observed in Runx3^AID/AID mice. Numbers of RORγt⁺MHCII⁺ cells in the mLNs of Runx3^+/WRPW and Runx3^AID/AID mice were severely reduced (Extended Data Fig. 3f). Accordingly, the development of all TC subsets was almost completely abrogated with severe reduction of LTi cell numbers in Runx3^+/WRPW mice (Fig. 3c). In Runx3^AID/AID mice, TC II, III and IV were absent, whereas ILC3s/LTi cells were only mildly reduced, and TC I numbers were not substantially altered (Fig. 3d). Because the Runx3^WRPW mutation would interfere with Runx1 function when a Runx3^WRPW–Runx1 heterodimer is formed³⁴, we cannot exclude the possibility of redundant functions of Runx1 and Runx3 in TC development, analogous to the role of Runx proteins in LTi cell development²⁵. However, the loss of TC II, III and IV in Runx3^AID/AID mice, which presumably possess weakened Runx3 function, demonstrates a critical requirement for the Runx3–CBFβ complex in the development of these TC subsets. Given the presence of TC I in Runx3^AID/AID mice, we investigated PRDM16 expression and found that the frequency of PRDM16⁺ TC I was also reduced in Runx3^AID/AID mice (Extended Data Fig. 3g), suggesting a role for Runx3 in the regulation of Prdm16 expression.

We also examined the RORγt⁻ APC compartment in the mLNs of 2- to 3-week-old Runx3^+/WRPW and Runx3^AID/AID mice. In contrast to the increase in both CCR7⁺ and CCR7⁻ RORγt⁻ APCs observed in Cbfb^2m/2m mice, no such increase was observed in either of the Runx3-mutant strains (Extended Data Fig. 4a,b). CD103⁺CCR7⁻RORγt⁻ APCs, including CCR7⁻C11b⁺CD103⁺ cDC2s, were modestly impaired in Runx3^+/WRPW mice. These findings indicate that a germline mutation that compromises Runx3–CBFβ function severely impairs TC II, III and IV development.

Extended Data Fig. 4. cDC development in Runx3 mutant and Rorc^{Δ+11E/Δ+11E} mice.

(a) Representative flow cytometry showing gating of antigen presenting cell types in Lin⁻RORγt⁻MHCII⁺ population in mLNs of two to three weeks-old Runx3^+/+ (n = 3) and Runx3^+/WRPW (n = 4) mice. Graphs showing the statistical summary of the frequency in indicated population and the numbers of indicated cell subsets. Mean ± SD. *: p < 0.05, **: p < 0.01, unpaired t-test (two-sided) (b) Representative flow cytometry showing gating of antigen presenting cell types in Lin⁻RORγt⁻MHCII⁺ population in mLNs of two to three weeks-old Runx3^+/+ (n = 3) and Runx3^AID/AID (n = 4) mice. Graphs showing the statistical summary of the frequency in indicated population and the numbers of indicated cell subsets. Mean ± SD. unpaired t-test (two-sided). (c) Representative pseudo-color plots showing Lin and CD45.2 expression, RORγt and MHCII expression in Lin⁻CD45.2⁺ population, Cxcr6 and MHC-II expression in Lin⁻CD45.2⁺RORγt⁺ cells, and CCR6 and MHC-II expression in Lin⁻CD45.2⁺ RORγt⁺ Cxcr6⁺ cells. Graphs showing the numbers of LTi cells and TCs that were reconstituted from fetal liver cells of the indicated genotypes. Mean ± SD. **: p < 0.01, ***: p < 0.001, ****: p < 0.0001, One-way Anova analyses. (d) Contour plots showing CD103 and CD11b expression in Lin⁻RORγt⁻MHCII⁺CD11c⁺ CCR7⁺ (left) and CCR7⁻ (right) population in samll intestine lamina propria (SI-LP) of Wt (n = 3) and Rorc^{Δ+11E/Δ+11E} (n = 3) mice. Graph showing a summary of frequencies of cDC1 (CD103⁺CD11b⁻) and cDC2 (CD103⁺CD11b⁺) cells in CCR7⁺ and CCR7⁻ population. Mean ± SD. *: p < 0.05, unpaired t-test (two-sided).

Source data

A +11-kb Rorc enhancer regulates TC development

We previously identified a +11-kb genomic region in the Rorc locus that was bound by Runx–CBFβ complexes in fetal liver and required for RORγt expression in fetal liver LTi cells²⁸. Analysis of previously published single-cell assay for transposase-accessible chromatin with sequencing (scATAC-seq) data for TCs and ILC3s¹⁰ revealed accessible chromatin at the +11-kb enhancer (+11E) in all TC subsets (Fig. 4a), as well as accessibility at the +7-kb region in TC II–IV. Additional TC II–IV-specific peaks around regions 94392000 and 94395000 did not contain Runx motifs and were not occupied by CBFβ2 (Fig. 4a). Given the essential role of PRDM16 in TC development and the possible regulation of Prdm16 expression by Runx3–CBFβ, we similarly analyzed ATAC-seq peaks and CBFβ2 binding in the Prdm16 locus and identified several TC‑specific ATAC‑seq peaks and numerous CBFβ2 binding sites (Fig. 4b). Given the complexity of Runx–CBFβ binding patterns at the Prdm16 locus, we focused on determining the role of the Rorc +11E in regulating TC differentiation using previously generated +11E-deficient Rorc^{Δ+11E/Δ+11E} mice²⁸. In the SIs and LIs of Rorc^{Δ+11E/Δ+11E} mice, both RORγt⁺ pT_reg cells and RORγt⁺ T_H17 cells were almost completely absent (Fig. 4c). In line with the role of the +11E in LTi cell development, there were no LNs or PPs in these mice, making it impossible to examine TCs, which predominantly reside in LNs. To circumvent this, we generated fetal liver chimeric mice by transferring fetal livers from embryonic day 14.5 (E14.5) Rorc^{Δ+11E/Δ+11E}, knock-in/knockout Rorc^GFP/GFP, Cbfb^+/+ or Cbfb^2m/2m embryos to sublethally irradiated recipient mice. Two weeks later, we observed reconstitution of TCs from wild-type Cbfb^+/+ fetal liver but substantially impaired TC numbers in Cbfb^2m/2m recipients and complete absence of TCs in Rorc^{Δ+11E/Δ+11E} and Rorc^GFP/GFP fetal liver recipients (Fig. 4d). Similarly, E14.5 Runx3^+/WRPW fetal livers failed to reconstitute both TC and LTi cell compartments, whereas both cell types were differentiated to some extent from Runx3^AID/AID fetal livers (Extended Data Fig. 4c). In the SI-LP of Rorc^{Δ+11E/Δ+11E} mice, we observed a slight decrease and increase in the frequency of CCR7⁻CD103⁺CD11b⁺RORγt⁻ APCs and an increase in CCR7⁻CD103⁺CD11b⁻RORγt⁻ APCs (Extended Data Fig. 4d). Together, these observations identify a critical role for Runx3–CBFβ in TC II–IV development as well as in regulation of RORγt expression at least in part through activation of the +11E in the Rorc locus. However, the changes observed in the RORγt⁻ APC compartment in Rorc^{Δ+11E/Δ+11E} mice preclude direct assessment of TC functions in these mice.

Fig. 4. Lack of TC progenitors in Cbfb^2m/2m and Rorc^{Δ+11E1/Δ+11E} mice.

a,b, scATAC-seq tracks showing chromatin accessibility at the Rorc locus (a) and Prdm16 locus (b) in the indicated cell subsets. TC-specific scATAC-seq peaks and CBFβ2 peaks detected by chromatin immunoprecipitation with sequencing (ChIP–seq) are shown as blue and red boxes, respectively; Chr, chromosome. c, Left, contour plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of the SI-LP and LI-LP of Rorc^+/+ (n = 5) and Rorc^{Δ+11E1/Δ+11E1} (n = 5) mice. Right, statistical summary of the frequency of the indicated cell subsets in CD4⁺ T cells. Data are shown as mean ± s.d. and were analyzed by unpaired t-test (two-sided); *P < 0.05; **P < 0.01; ***P < 0.001. d, Reconstitution of RORγt⁺MHCII⁺ cells from fetal liver cells (Cbfb^+/+: n = 7, Cbfb^2m/2m: n = 7, Rorc^{Δ+11E/Δ+11E}: n = 6, Rorc^GFP/GFP: n = 6). Two weeks after injection of fetal liver cells from CD45.2 E14.5 embryos with the indicated genotype, mLNs of CD45.1 sublethally irradiated recipients were examined by flow cytometry. Left, representative pseudocolor plots showing Lin and CD45.2 expression, RORγt and MHCII expression in the Lin⁻CD45.2⁺ population, CXCR6 and MHCII expression in Lin⁻CD45.2⁺RORγt⁺ cells and CCR6 and MHCII expression in Lin⁻CD45.2⁺RORγt⁺CXCR6⁺ cells. Right, numbers of LTi cells and TCs that were reconstituted from fetal liver cells of the indicated genotypes. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; ****P < 0.0001.

Source data

Transgenic Runx expression enhances TC differentiation and pT_reg cell abundance

We next sought to determine whether Runx1 and Runx3 serve distinct versus redundant functions in regulating TC development and pT_reg cell differentiation. In contrast to the loss of RORγt⁺ pT_reg cells with germline hypomorphic Runx3 mutation, Runx3 inactivation alone by Cd11c-Cre did not lead to a substantial reduction of RORγt⁺ pT_reg cells (Extended Data Fig. 5a). Given redundant functions between Runx1 and Runx3 in other types of APCs, such as cDC2s²⁷, Runx1 may compensate Runx3 function for TC development from a certain developmental stage where Cd11c-Cre was expressed. Indeed, differentiation of both RORγt⁺ pT_reg cells and RORγt⁺ T_H17 cells was severely inhibited in Runx1^fl/fl:Runx3^fl/fl:Cd11c-Cre(Runx1/Runx3^fl/fl:Cd11c-Cre) mice (Fig. 5a). Although TC II–IV express CD11c, TC I exhibit variable levels of CD11c¹⁴. However, it remains unclear which RORγt⁺MHCII⁺ cell types have a history of Cd11c-Cre expression. We therefore conducted lineage tracing using a Rosa26^lsl-tdTomato reporter and found that almost all TC II, III and IV were marked by tdTomato expression by Cd11c-Cre, whereas less than 40% of TC I were marked with tdTomato (Fig. 5b). Consistent with the negligible levels of labeling in LTi cells, the formation of pLNs and PPs was not impaired in Runx1/Runx3^fl/fl:Cd11c-Cre mice. We then examined the TC compartment in mLNs and observed that TC III and TC IV numbers were markedly decreased in Runx1/Runx3^fl/fl:Cd11c-Cre mice, whereas TC II numbers were not substantially reduced (Fig. 5c). The TC II subset that emerged in Runx1/Runx3^fl/fl:Cd11c-Cre mice retained PRDM16 expression (Fig. 5d), suggesting that the differentiation of TC II and the maintenance of PRDM16 expression in this subset become independent of Runx–CBFβ at a certain point during development. Similarly, the frequency of PRDM16⁺ TC I was not reduced in Runx1/Runx3^fl/fl:Cd11c-Cre mice (Fig. 5d).

Extended Data Fig. 5. Rescue of RORγt⁺ MHCII⁺ development by transgenic Runx expression.

(a) Contour plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of SI-LP of Rx3^fl/fl (n = 3) and Rx3^fl/fl: Cd11c-Cre (n = 3) mice. Graph showing a summary of frequencies of RORγt⁺ pTreg cells. Mean ± SD. unpaired t-test (two-sided). (b) Contour plots showing CD103 and CD11b expression in Lin⁻MHCII⁺CD11c⁺ cells in mLNs of mice with indicated genotype (n = 4 for each indicated genotype). Graph showing a summary of the frequencies of cDC1 (CD103⁺CD11b⁻) and cDC2 (CD103⁺CD11b⁺) cells. Mean ± SD. *: p < 0.05, ****: p < 0.0001, One-way Anova analyses. (c) Representative flow cytometry plots showing gating of antigen presenting cell types in Lin⁻RORγt⁺ and Lin⁻RORγt⁻ populations in mLNs of Rx1/3^fl/fl (n = 3), Rx1/3^fl/fl: Cd11c-Cre (n = 3) and Rx1/3^fl/fl: Rosa26^lsl-Rx1/+: Cd11c-Cre (n = 3) mice. Graphs showing statistical summary of the numbers of indicated cell subsets. Mean ± SD. *: p < 0.05, **: p < 0.01, One-way Anova analyses.

Source data

Fig. 5. Runx3 regulates the development and function of TC III and IV.

a, Left, zebra plots showing representative RORγt and Foxp3 expression in CD4⁺ T cells of the SI-LP of mice with the indicated genotype (Runx1/Runx3^fl/fl: n = 8, Runx1/Runx3^fl/fl:Cd11c-Cre: n = 7, Runx1/Runx3^fl/fl:Rosa26^lsl-Runx1/+:Cd11c-Cre, n = 4, Runx1/Runx3^fl/fl:Rosa26^lsl-Runx3/+:Cd11c-Cre, n = 5). Right, statistical summary. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; **P < 0.01; ****P < 0.0001. b, Histograms showing tdTomato expression from the Rosa26^lsl-tdTomato locus in the indicated cell subsets of Rosa26^{lsl-tdTomato/+}:Runx1/Runx3^+/fl:Cd11c-Cre mice (n = 3). Data are shown as mean ± s.d. c, Left, flow cytometry plots showing the gating of APC types in the RORγt⁺MHCII⁺ population in mLNs of 2- to 3-week-old mice with the indicated genotype (Runx1/Runx3^fl/fl: n = 13, Runx1/Runx3^fl/fl:Cd11c-Cre: n = 8, Runx1/Runx3^fl/fl:Rosa26^lsl-Runx1/+:Cd11c-Cre, n = 5, Runx1/Runx3^fl/fl:Rosa26^lsl-Runx3/+:Cd11c-Cre, n = 4). Right, statistical summary of numbers of LTi cells/ILC3s and each TC subset. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. d, Top, representative histograms showing PRDM16 expression in TC I and II of Runx1/Runx3^fl/fl (n = 7) and Runx1/Runx3^fl/fl:Cd11c-Cre (n = 5) mice. Bottom, statistical summary of the frequency of the PRDM16⁺ population. Data are shown as mean ± s.d. e,f, Left, representative pseudocolor plots showing Texas Red fluorescence in cDC1s and cDC2s (e) and TC III and IV (f) from the mLNs 16 h after oral gavage of OVA–Texas Red into 2- to 3-week-old Runx1/Runx3^fl/fl (n = 4), Runx1/Runx3^fl/fl:Cd11c-Cre (n = 8), Runx1/Runx3^fl/fl:Rosa26^lsl-Runx1/+:Cd11c-Cre (n = 4) and Runx1/Runx3^fl/fl:Rosa26^lsl-Runx3/+:Cd11c-Cre (n = 4; e) and Rorc^+/GFP and Rorc^+/GFP:Runx1/Runx3^fl/fl:Rosa26^{lsl-Runx3-ΔIRES/+}:Cd11c-Cre mice (n = 3 for each indicated genotype; f). Right, statistical summary. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA (e) and unpaired t-test (two-sided; f); *P < 0.05; **P < 0.01; ****P < 0.0001.

Source data

A major question is whether the RORγt⁺ pT_reg cell-inducing ability of TCs can be harnessed for therapeutic approaches in intestinal inflammation such as inflammatory bowel disease or food allergy. Given the rescue of RORγt⁺ pT_reg cells by transgenic CBFβ2 expression, we wondered whether we could exploit Runx regulation to augment TC numbers and RORγt⁺ pT_reg cell differentiation. Taking advantage of having Rosa26^lsl-Runx1 and Rosa26^lsl-Runx3 mouse strains³⁴, we tested Runx1 and Runx3 function by generating Runx1/Runx3^fl/fl:Rosa26^lsl-Runx1/+:Cd11c-Cre and Runx1/Runx3^fl/fl:Rosa26^lsl-Runx3/+:Cd11c-Cre mice. Both transgenic Runx1 and Runx3 restored, or rather enhanced, RORγt⁺ pT_reg cell development with a higher number of RORγt⁺ pT_reg cells induced by transgenic Runx1 (Fig. 5a). This finding was associated with parallel changes in TCs, specifically within TC III and IV. Severe abrogation of T_H17 cell differentiation due to Runx1/Runx3 inactivation by Cd11c-Cre was also restored to levels above control mice with both transgenic Runx1 and Runx3 expression (Fig. 5a). Underlying this phenomenon, within the MHCII⁺CD11c⁺ population in mLNs, the severe reduction of CD103⁺CD11b⁺ cells, a critical cell type that has been implicated in intestinal T_H17 cell differentiation due to its IL-23 production capacity^37,38, was restored after rescue of Runx1 or Runx3 expression (Extended Data Fig. 5b). Further analysis, in which MHCII⁺CD11c⁺ cells were divided into RORγt⁺ and RORγt⁻ populations, with or without CCR7 expression, revealed that the restored CD103⁺CD11b⁺ subset stemmed primarily from enhanced differentiation of RORγt⁺CCR7⁻CD103⁺CD11b⁺ cells (Extended Data Fig. 5c), which shared characteristics with TC IV. However, the increase in RORγt⁺CCR7⁺CD103⁺CD11b⁺ cells by transgenic Runx1 expression also highlighted its broad effect across a range of APC types. Thus, further careful investigation is required to elucidate the mechanisms underlying the enhanced differentiation of RORγt⁺ pT_reg and T_H17 cells in these mice.

To further characterize the function of RORγt⁺MHCII⁺ cell subsets, we next examined in vivo antigen uptake capacity using oral gavage of Texas Red–OVA as well as a Rorc^GFP reporter allele³⁹. Sixteen hours after oral gavage in Rorc^+/GFP mice, incorporation of Texas Red–OVA was detected in cDC1s, cDC2s and TC IV, with highest incorporation efficiency in cDC1s and TC IV, but not in ILC3s/LTi cells nor in TC I, II and III (Extended Data Fig. 6a), as was recently reported¹⁸. Uptake of Texas Red–OVA by cDC1s and cDC2s was substantially reduced by loss of Runx1 and Runx3 and was restored by either transgenic Runx1 or Runx3 expression, with more efficient restoration by Runx3 than by Runx1 (Fig. 5e). After excision of lsl sequences, the Rosa26^lsl-Runx3 locus produced green fluorescent protein (GFP) signal that overlayed RORγt–GFP signals, preventing us from identifying RORγt⁺ cells. To overcome this experimental barrier, we generated the Rosa26^{lsl-Runx3-ΔIRES} allele by mutating IRES sequences in the Rosa26^lsl-Runx3 locus (Extended Data Fig. 6b) and confirmed that restored TC IV by transgenic Runx3 retained antigen uptake capacity to a similar extent as control TC IV (Fig. 5f). Interestingly, TC III acquired antigen uptake capacity following transgenic Runx3 expression (Fig. 5f). These observations suggest that Runx proteins regulate not only development but also function of cDCs and TCs.

Extended Data Fig. 6. Specific rescue of TC IV subset by Cd11c-Cre driven transgenic Cbfβ2.

(a) Representative pseudocolor plots showing Texas-Red fluorescence in indicated cell subset from mLNs 16 h after oral gavage of OVA-Texas-Red into three-week-old Rorc^+/gfp mice (n = 3). Mean ± SD. *: p < 0.05, One-way Anova analyses. (b) A scheme showing the structure of the Rosa26^lsl-Rx3 allele and strategy for targeting mutations into the ires sequences by CRISPR/Cas9 genome editing. Sequence of non-functional ires after genome editing is shown in which sequences deleted were highlighted with yellow color font. gRNA sequences are indicated with wavy underline. (c) Graphs showing summary of the numbers of LTi and TC subsets of Cbfb^+/+ (n = 8), Cbfb^2m/2m (n = 6), Cbfb^2m/2m: Rosa26^lsl-cbfb2/+: Cd11c-Cre (n = 5), Cbfb^2m/2m: Rosa26^lsl-cbfb2/+: Il7r-Cre (n = 3) and Cbfb^2m/2m: Rosa26^lsl-cbfb2/+: Cd11c-Cre (n = 3) mice. Mean ± SD. *: p < 0.05, **: p < 0.01, ***: p < 0.001, ****: p < 0.0001, One-way Anova analyses. (d) Representative flow cytometry showing gating of TC subset in CXCR6⁻RORγt⁺MHCII⁺ population in mLNs of two to three weeks-old Cbfb^+/+ (n = 3), Cbfb^2m/2m (n = 3) and Cbfb^2m/2m: Rosa26^lsl-Rx1/+: Cd11c-Cre (n = 4) mice. Graphs showing the statistical summary of cell numbers of indicated cell subsets. Mean ± SD. **: p < 0.01, ***: p < 0.001, ****: p < 0.0001, One-way Anova analyses.

Source data

Rescue of TC IV differentiation restores RORγt⁺ pT_reg cell differentiation

Having established the rescue of specific TC subsets by Cd11c-Cre-driven transgenic Runx protein, we also examined which TC subsets were restored by transgenic CBFβ2 expression by Cd11c-Cre in Cbfb^2m/2m mice. Only TC IV was restored in Cbfb^2m/2m:Rosa^lsl-Cbfb2/+:Cd11c-Cre mice (Fig. 6a), indicating that recovery of TC IV alone was sufficient for restoration of RORγt⁺ pT_reg cells in Cbfb^2m/2m mice (Fig. 1c). Interestingly, we noticed that restored TC IV by Cd11c-Cre-driven transgenic CBFβ2 expression lacked PRDM16 (Fig. 6b). Similarly, PRDM16 expression was not restored in most TC IV rescued by transgenic Runx1 expression in Runx1/Runx3^fl/fl:Cd11c-Cre mice (Fig. 6b). These findings suggest that TC IV do not require PRDM16 for their function. By contrast, the frequency of PRDM16-expressing TC I was restored, or even enhanced, by Cd11c-Cre-driven transgenic Runx1 expression, whereas transgenic CBFβ2 expression in Cbfb^2m/2m mice did not restore PRDM16⁺ TC I (Fig. 6c).We next tested whether other Cre transgenes, Vav-iCre and Il7r-Cre, which are expected to be expressed at early stages of TC development, could restore PRDM16 expression in TCs. Both transgenes rescued TC II, III and IV in CBFβ2-deficient mice, with Il7r‑Cre additionally enhancing the differentiation of TC I, II and IV (Extended Data Fig. 6c). Furthermore, PRDM16 expression in all TC subsets was restored by both transgenes (Fig. 6d), suggesting that the restoration of CBFβ dosage at early TC developmental stages is sufficient to recover both TC development and PRDM16 expression. As expected, Cd11c-Cre-driven transgenic Runx1 could not compensate for CBFβ2 deficiency, and TC II, III and IV remained absent in Cbfb^2m/2m:Rosa26^lsl-Runx1/+:Cd11c-Cre mice (Extended Data Fig. 6d). Collectively, these findings indicate that recovery of TC IV alone is sufficient to restore RORγt⁺ pT_reg cell differentiation and that TC IV function does not require PRDM16 expression.

Fig. 6. Rescued TC IV lacking PRDM16 expression support RORγt⁺ pT_reg cell differentiation.

a, Left, representative flow cytometry plots showing gating of APC types in the RORγt⁺MHCII⁺ population in mLNs of 2- to 3-week-old Cbfb^+/+ (n = 5), Cbfb^2m/2m (n = 5) and Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd11c-Cre (n = 5) mice. Right, statistical summary of numbers of the indicated cell subsets. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; *P < 0.05; ***P < 0.001; ****P < 0.0001. b, Top, histograms showing PRDM16 expression in TC IV of 2- to 3-week-old wild-type (n = 9), Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd11c-Cre (n = 4), Runx1/Runx3^fl/fl:Rosa26^lsl-Runx1/+:Cd11c-Cre (n = 3) and Runx1/Runx3^fl/fl:Rosa26^lsl-Runx3/+:Cd11c-Cre (n = 4) mice. Figures in histograms show the frequency of PRDM16-expressing cells in wild-type mice. Bottom, summary of the frequency of PRDM16⁺ TC IV. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; ****P < 0.0001. c, Top, histograms showing PRDM16 expression in TC I of 2- to 3-week-old wild-type (n = 4), Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd11c-Cre (n = 4) and Runx1/Runx3^fl/fl:Rosa26^lsl-Runx1/+:Cd11c-Cre (n = 3) mice. Figures in histograms show the frequency of PRDM16-expressing cells in wild-type mice. Bottom, summary of the frequency of PRDM16⁺ TC I. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; **P < 0.01; ****P < 0.0001. d, Summary of the frequency of PRDM16⁺ cells in the indicated TC subsets of Cbfb^+/+ (n = 8), Cbfb^2m/2m (n = 6), Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd11c-Cre (n = 5), Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Il7r-Cre (n = 3) and Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Cd11c-Cre (n = 3) mice; NA, not applicable. Data are shown as mean ± s.d. and were analyzed by one-way ANOVA; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Source data

Discussion

Here, we addressed the role of Runx–CBFβ complexes in regulating RORγt⁺ APCs using different mouse lines. Our results not only provided definitive evidence that TC IV is the critical tolerogenic APC subset for RORγt⁺ pT_reg cell induction but also shed new light on possible manipulation of TC IV development by engineering Runx expression. Our data suggest that attenuation of Runx3–CBFβ complexes by germline mutation results in reduced activation of cis-regulatory elements, such as +7Es and +11Es, in the Rorc locus, leading to impaired initiation of TC developmental programs. As such, Rorc-Cre was not always activated in Cbfb^2m/2m mice during TC development, causing a variegated rescue of RORγt⁺ pT_reg cells in Cbfb^2m/2m:Rosa26^lsl-Cbfb2/+:Rorc-Cre mice. Inactivation of Prdm16 by Rorc-Cre was shown to result in lack of PRDM16⁺ tolerizing DCs^11,16,30, including TC II, III and IV. Loss of PRDM16⁺ TC I in Cbfb^2m/2m mice suggests that Runx–CBFβ could be involved in Prdm16 regulation. Emergence of TC IV lacking PRDM16 expression by transgenic CBFβ2 or Runx suggest that PRDM16 is required for TC IV development only at the early stage but becomes dispensable for maintenance. Given the expected time lag of transgenic Runx protein induction after loss of endogenous Runx by Cd11c-Cre, the lack of PRDM16 expression in transgenic Runx-induced TC IV suggests a narrow time window for recovery of Prdm16 activation by Runx re-expression. Rescue of PRDM16 expression by Il7r-Cre or Vav1-iCre transgene-driven CBFβ2 expression indicates that the onset of Cd11c-Cre activity occurs later than these transgenes and that both differentiation of TC II and the maintenance of PRDM16 expression in this subset become independent of Runx–CBFβ at that stage. This distinct dependency on Runx–CBFβ for the differentiation of each TC subset highlights the heterogeneity of TCs and suggests that each subset, which emerges during TC differentiation from common TC progenitors¹¹, has a differential requirement for Runx–CBFβ. Future studies addressing the developmental pathways for individual TC subsets are required to further elucidate the role of Runx–CBFβ in TC subset specification. Recent studies have shown that PU.1 (refs. ^11,30) and REV-ERBα and REV-ERBβ³⁰ are essential transcriptional regulators for TC development. Runx has been reported to activate the Spi1 gene, which encodes PU.1, during early hematopoiesis through its activation of the −14-kb enhancer⁴⁰. Future investigation is required to further unravel the transcription factor network that governs TC development and differentiation into individual subsets.

Considering the essential role of TC IV in establishing tolerance to microbiota- and food-derived antigens^16,18, understanding the transcriptional regulation of TC IV is crucial for insights into therapeutic possibilities in inflammatory bowel disease and food allergies. Our results showed that the developmental program of TC III and IV can be manipulated by exogenous Runx expression. In addition, Runx proteins play a role in regulating the antigen uptake function of cDCs and TCs. It remains unclear how transgenic Runx1 induced by Cd11c-Cre enhances RORγt⁺ pT_reg cell differentiation. Besides the increase in the number of TC IV, enhanced function of TC IV and other uncharacterized APCs could be involved in this mechanism. Testing whether other Cre transgenes expressed during TC development, such as Il7r-Cre or Rorc-Cre, similarly enhance RORγt⁺ pT_reg cell differentiation would be informative in determining whether transgenic Runx1 expression in TC subsets, particularly TC IV, is sufficient for enhanced RORγt⁺ pT_reg cell differentiation.

Overall, these studies reveal a crucial role for Runx in the transcriptional regulation of TC differentiation and thus intestinal tolerance. Elucidating the mechanisms by which Runx influences APC differentiation may lead to the development of novel therapeutic approaches for immune-related diseases through increase of RORγt⁺ pT_reg cells.

Methods

Mice

Cbfb^2m28, Runx1^fl41, Runx3^fl41, Mb1-Cre⁴², Cd4-Cre⁴³, Cd11c-Cre⁴⁴, Il7r-Cre⁴⁵, Rosa26^lsl-Cbfb2 (ref. ²⁸), Rorc^Δ+11E28, OT-II⁴⁶, Rorc^GFP39 and Runx3^WRPW34 mice were previously described. Vav1-iCre (018968), Rorc-Cre (022791) and Rosa26^tdTomato (007909) mice were purchased from The Jackson Laboratory. CD45.1 mice were from Sankyo Labo Service. All mice were maintained under specific pathogen-free conditions at the RIKEN Center for Integrative Medical Sciences (IMS) and the Sloan Kettering Institute, with controlled temperature and humidity and a 12-h light/12-h dark cycle. All mouse strains were bred and housed in the animal facilities at RIKEN IMS and Sloan Kettering Institute. All animal procedures were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committees of the RIKEN Yokohama Branch (AEY2026-023(2)) and Sloan Kettering Institute (21-05-007). All mice were killed by CO₂ overdose following anesthesia with isoflurane. Data from both sexes were combined for analysis, and mice aged 2–14 weeks were used unless otherwise specified.

Runx3^AID mice and Rosa26^{lsl-Runx3-ΔIRES} mice were generated by genome editing conducted at RIKEN IMS. In brief, zygotes generated from C57BL/6NJcl mice purchased from CLEA Japan were electroporated with a recombinant Cas9 protein (100 ng μl¹; Alt-R S.P. Cas9 Nuclease V3, IDT), synthetic CRISPR RNA (crRNA; Alt-R CRISPR–Cas9 crRNA; IDT), trans-activating crRNA (Alt-R CRISPR–Cas9 tracrRNA; IDT) and synthetic single-strand donor DNA (IDT) at the animal facility at RIKEN IMS. Sequences for the crRNA and donor DNA used for genome editing are listed in Extended Data Table 1.

Extended Data Table 1.

The sequence information of the oligonucleotides used in the study

To construct the target vector for R26^lsl-Cbfb1 mice, cDNA encoding CBFβ1 was amplified by PCR to add AscI sites at both ends. This cDNA fragment was ligated into an AscI-cleaved pCTV vector (15912, Addgene). Thirty milligrams of the target vector was linearized by AsiSI enzyme and transfected into the M1 embryonic stem (ES) cell line by electroporation, as previously described²⁸. After G418 selection, G418-resistant ES clones were screened for homologous recombination by PCR. Appropriate ES clones harboring the Rosa26^lsl-Cbfb1 allele were aggregated with blastocysts to generate chimeric mice, through which the Rosa26^lsl-Cbfb1 allele was germline transmitted to offspring. Rosa26^lsl-Runx1 and Rosa26^lsl-Runx3 mice were also generated in a similar way as Rosa26^lsl-Cbfb1 mice.

Histology

Mouse intestinal tissues were collected, fixed in 10% neutral buffered formalin and paraffin embedded according to standard procedures. Four-micron sections were cut using a paraffin microtome with steel blades. Sections were stained with hematoxylin and eosin. Evaluation of inflammation in mouse colitis was performed in a blinded manner according to the previously described colitis score⁴⁷.

Tissue processing

LP lymphocytes from the SI and LI were isolated as previously described²⁷. SIs were cut into 5-mm pieces and incubated in 20 ml of RPMI containing 2% FBS and 5 mM EDTA, with shaking at 200 rpm (Taitec, BR-43FL) and 37 °C for 20 min. After incubation, the tissues were washed with PBS twice by vortexing for 20 s to remove epithelial cells. The rest of the tissues were filtered and digested with RPMI including 2% FBS, 0.5 mg ml¹ collagenase IV (Sigma, C-5138) and 50 μg ml¹ DNase (043-26773, Fujifilm) at 200 rpm (Taitec, BR-43FL) and 37 °C for 30 min. Digested tissues were passed through a 100-µm strainer and subjected to Percoll gradient centrifugation at 700g for 20 min using 40% and 80% Percoll in RPMI with 2% FBS. Cells located between the 40% and 80% Percoll layers were collected as LP lymphocytes.

For staining of RORγt⁺MHCII⁺ cells from mLNs, all mLNs were collected in RPMI containing 10% FBS, 0.5 mg ml¹ collagenase IV and 50 μg ml¹ DNase I and incubated at 37 °C for 30 min with shaking. After incubation, the digested samples were passed through a 100-μm cell strainer and centrifuged at 310g for 5 min at 4 °C.

Flow cytometry

Surface molecules were stained with specific antibodies by incubating the cells for 30 min at 23 °C for mLN cells or for 10 min at 23 °C for other cell types. The following surface marker antibodies were purchased from BD Biosciences, BioLegend or Thermo Fisher Scientific: B220 (RA3-6B2), CCR6 (29-2L17), CCR7 (4B12), CD4 (RM4-5), CD8α (53-6.7), CD11b (M1/70), CD11c (N418), CD19 (6D5), CD24 (M1/69), CD25 (PC61.5), CD44 (IM7), CD62L (MEL-14), CD45 (30-F11), CD45.1 (A20), CD45.2 (104), CD56 (809220), CD64 (X54-5/7.1), CD88 (20/70), CD103 (M290), CD304 (3E12), CD326 (G8.8), CXCR6 (SA051D1), GR-1 (RB6-8C5), Ly6C (AL-21), NK1.1 (PK136), MHCII (M5/114.15.2), Siglec-F (S17007L), TCRβ (H57-597), TCRVα2 (B20.1) and γδTCR (GL3). For intracellular staining, cells were fixed and permeabilized using a Transcription Factor Buffer Set (562574, BD Biosciences) after surface staining. The following intracellular antibodies were purchased from BD Biosciences or Thermo Fisher Scientific: Foxp3 (FJK-16s) and RORγt (B2D and Q31-378). For intracellular PRDM16 staining, a rabbit monoclonal antibody to PRDM16 (ab303534, Abcam) was used as the primary antibody, followed by goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody, Alexa Fluor 647 (A21244, Invitrogen). Dead cells were excluded using Fixable Viability Dye eFluor 506 (65-0866, Invitrogen). Multicolor flow cytometric analysis was performed using a BD FACSCanto II and BD FACSAria Fusion (both from BD Biosciences), and data were analyzed with FlowJo software (BD Biosciences). Cell sorting was performed using a BD FACSAria III and BD FACSAria Fusion (BD Biosciences).

Immunoblotting

CD8⁺ T cells were isolated from the spleen using an EasySep Mouse CD8⁺ T Cell Isolation kit (STEMCELL Technologies). Cell lysates from total thymocytes or CD8⁺ T cells were resolved by SDS–PAGE and transferred to Immobilon-P Transfer Membranes (Millipore). Membranes were probed with an appropriate primary antibody, including rabbit anti-pan-CBFβ, anti-CBFβ1 and anti-CBFβ2 (ref. ²⁸); rabbit anti-Runx3 (D6E2, 9647, Cell Signaling) and anti-GAPDH (6C5, sc-32233, Santa Cruz) and horseradish peroxidase-conjugated secondary antibody, and immune complexes were detected using ECL Prime (GE Healthcare) with an Amersham Imager 600 (GE Healthcare).

Fetal liver transfer

Fetal livers collected from E14.5 CD45.2 embryos were used to prepare single-cell suspensions. Two to 3 weeks after injection, the recipient mice were analyzed by flow cytometry for reconstitution of RORγt⁺MHCII⁺ cells.

Adoptive OT-II T cell transfer

CD4⁺ T cells were pre-enriched from the spleens of CD45.2 OT-II TCR transgenic mice by negative selection using an EasySep Mouse CD4⁺ T Cell Isolation kit (STEMCELL Technologies). Live CD25⁻CD44^loCD62L⁺TCRβ⁺TCRVα2⁺ naive OT-II T cells were sorted by FACS (purity of >99%). A total of 5 × 10⁴ naive OT-II cells were transferred by retro-orbital injection into postnatal day 21 CD45.1:Cbfb^+/+ or CD45.1:Cbfb^2m/2m recipient mice. Recipient mice received 25 mg of OVA (grade III; Sigma, A5378) by oral gavage on days 1 and 3 after transfer and were simultaneously provided OVA-supplemented drinking water (0.75% (wt/vol)) ad libitum. One week after transfer, donor OT-II cells were recovered from the SI and mLNs and examined by flow cytometry.

Oral gavage and antigen uptake

To assess OVA uptake, postnatal day 14–21 Rorc^+/GFP and Rorc^+/GFP:Runx1/Runx3^fl/fl:Cd11c-Cre mice were gavaged with 200 μl of 20 mg ml¹ Texas Red-conjugated OVA (O23021, Invitrogen). Sixteen hours later, cells in the mLNs were analyzed by flow cytometry.

CBFβ2 ChIP–seq analysis

FASTQ files of CBFβ2 ChIP–seq data were obtained from GSE90794 (ref. ²⁸). Reads were trimmed using TrimGalore v0.6.19, aligned to the mouse reference genome (GENCODE GRCm38 release M22) using bowtie2 v2.5.1 and deduplicated. Peaks were identified using MACS2 v2.2.7.1 (-f BAM -g mm --nomodel --pvalue 0.1 --shift 0 --extsize 165 --call-summits).

scATAC-seq analysis

scATAC-seq data were obtained from GSE205065 (ref. ¹⁰) and processed using ArchR v1.0.1 as described previously. Briefly, latent semantic indexing was performed using the IterativeLSI function (iterations = 10, varFeatures = 100,000) of ArchR, and cells were clustered (method = Seurat, k.param = 30, resolution = 1.2) using 30 latent semantic indexing components. For peak calling, clusters for similar cells were grouped: C1 (TC IV), C2–C4 (TC I, II and III), C5 and C6 (NCR + ILC3) and C7–C13 (LTi). Peaks were called on each group using MACS2 v2.2.7.1 (--gsize mm --qval 0.01 --nomodel --ext 200 --shift -100 --call-summits). Peak summits were extended by 100 bp in each direction. Regions extending outside of mm10 chromosomes, arising from chromosome Y or the mitochondrial chromosome, overlapping with blacklist regions precompiled by ArchR (merged from the ENCODE mm10 v2 blacklist regions from https://github.com/Boyle-Lab/Blacklist/blob/master/lists/mm10-blacklist.v2.bed.gz and mitochondrial regions that are highly mappable to the mm10 nuclear genome from https://github.com/caleblareau/mitoblacklist/blob/master/peaks/mm10_peaks.narrowPeak) or containing ‘N’ nucleotides (>0.001 of the sequence) were filtered. Regions from groups of TC clusters (C1 and C2–C4) were compiled, and overlapping regions were merged to their union, resulting in a nonoverlapping set of 101,264 peaks for the TC clusters. Genome tracks were visualized using the plotBrowserTrack function in ArchR with peaks from TC clusters and CBFβ2 ChIP–seq data.

Statistics and reproducibility

Statistical analyses were performed using a one-way ANOVA and unpaired t-tests with GraphPad Prism (v8.4.3; GraphPad Software). All experiments were repeated at least twice as successful, independent experiments.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41590-026-02566-8.

Author contributions

C.O., C.Z., Y.A.P.I., J.H. and N.S.-T. analyzed phenotypes of gene-modified mice. M.Y. performed TC reconstitution from fetal liver. T.P. analyzed scATAC-seq data. M.K. performed histological analyses. C.C.B. provided technical advice and wrote the manuscript. I.T. designed the study, generated mice and wrote the manuscript.

Peer review

Peer review information

Nature Immunology thanks Daniel Mucida and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: L. A. Dempsey, in collaboration with the rest of the Nature Immunology team. Peer reviewer reports are available.

Data availability

The scATAC-seq data and CBFβ2 ChIP–seq data supporting the findings of this study were obtained from the NCBI Gene Expression Omnibus under accession numbers GSE205065 and GSE90794, respectively. For the scATAC-seq analysis, the reference mouse genome mm10 was used for mapping, whereas the GENCODE GRCm38 release M22 mouse reference genome was used for alignment in the CBFβ2 ChIP–seq analysis. All other data are available from the corresponding authors upon reasonable request. Source data are provided with this paper.

Competing interests

Although C.Z. is currently affiliated with the Discovery Research Department, RIBOMIC. The other authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41590-026-02566-8.

Supplementary information

The online version contains supplementary material available at 10.1038/s41590-026-02566-8.