PRDM16-dependent antigen-presenting cells induce tolerance to gut antigens

Abstract

The gastrointestinal tract is continuously exposed to foreign antigens in food and commensal microorganisms with potential to induce adaptive immune responses. Peripherally induced T regulatory (pT_reg) cells are essential for mitigating inflammatory responses to these agents^1-4. Although RORγt⁺ antigen-presenting cells (APCs) have been shown to programme gut microbiota-specific pT_reg cells^5-7, their definition remains incomplete, and the APC responsible for food tolerance has remained unknown. Here we identify an APC subset that is required for differentiation of both food- and microbiota-specific pT_reg cells and for establishment of oral tolerance. Development and function of these APCs require expression of the transcription factors PRDM16 and RORγt, as well as a unique Rorc(t) cis-regulatory element. Gene expression, chromatin accessibility, and surface marker analysis establish the pT_reg-inducing APCs as myeloid in origin, distinct from type 3 innate lymphoid cells, and sharing epigenetic profiles with classical dendritic cells, and designate them PRDM16⁺RORγt⁺ tolerizing dendritic cells (tolDCs). Upon genetic perturbation of tolDCs, we observe a substantial increase in food antigen-specific T helper 2 cells in lieu of pT_reg cells, leading to compromised tolerance in mouse models of asthma and food allergy. Single-cell analyses of freshly resected mesenteric lymph nodes from a human organ donor, as well as multiple specimens of human intestine and tonsil, reveal candidate tolDCs with co-expression of PRDM16 and RORC and an extensive transcriptome shared with tolDCs from mice, highlighting an evolutionarily conserved role across species. Our findings suggest that a better understanding of how tolDCs develop and how they regulate T cell responses to food and microbial antigens could offer new insights into developing therapeutic strategies for autoimmune and allergic diseases as well as organ transplant tolerance.

Fig. 1. RORγt is required by tolerance-inducing APCs to promote microbiota-specific pT_reg cell differentiation.

a, Representative flow cytometry plots (top) and frequencies (bottom) of Hh-specific pT_reg (FOXP3⁺RORγt^+/−), T_H17 (FOXP3⁻RORγt⁺T-bet^+/−) and T_H1 (FOXP3⁻RORγt⁻T-bet⁺) cells in the LILP of Hh-colonized control (Rorc(t)^fl/gfp, Rorc(t)^wt/gfp and Cd11c^creRorc(t)^wt/gfp; n = 7) and Rorc(t)^ΔCD11c (Cd11c^creRorc(t)^fl/gfp; n = 5) mice at 14 days after adoptive transfer of naive Hh7-2tg CD4⁺ T cells. The top row of flow cytometry plots is gated on total Hh7-2tg cells (CD45⁺B220⁻TCRγδ⁻TCRβ⁺CD4⁺Vβ6⁺CD90.1⁺) and the bottom row is gated on FOXP3⁻ Hh7-2tg cells. b, Bulk ATAC–seq data showing accessible regions in the Rorc locus of several RORγt-expressing cell types, including double-positive (DP; CD4⁺CD8⁺) thymocytes, in vitro differentiated T_H17 cells and SILP-derived T_H17 (TCRβ⁺CD4⁺IL23R-GFP⁺) cells, Tγδ17 (TCRγδ⁺IL23R-GFP⁺) cells and presumptive ILC3s (Lin⁻IL7R⁺Klrb1b⁺NK1.1⁻). Prom, promoter. c, Phenotype of Hh-specific T cells in the LILP of Hh-colonized control (n = 6) and Δ+7 kb (n = 4) mice at 14 days after adoptive transfer of naive Hh7-2tg CD4⁺ T cells. The flow cytometry plots are gated on total (top) and FOXP3⁻ (bottom) Hh7-2tg cells. Data in a are pooled from two independent experiments. Data in c are representative of two independent experiments. Data are mean ± s.e.m. Unpaired two-sided t-test. *P < 0.05, **P < 0.01, ***P < 0.001; NS, not significant. Source Data

Fig. 2. RORγt regulates development of PRDM16-expressing tolDCs within mLNs.

a, Uniform manifold approximation and projection (UMAP) representation of 21,504 transcriptomes obtained from scRNA-seq of MHCII-expressing innate immune system cells (CD45⁺Ly6G⁻B220⁻TCRγδ⁻TCRβ⁻MHCII⁺) in the mLN, combining data from 3-week-old control and Δ+7 kb mice for joint clustering. Macs, macrophages; NK, natural killer cells. b, Dot plot of indicated clusters from a examining expression of genes that have been previously ascribed to proposed RORγt-APC subsets. c, Stacked bar plots comparing the proportion of each cluster in a, as derived from control and Δ+7 kb mice. The dashed line at 51.2% indicates total contribution from Δ+7 kb mutants. d, Stacked bar plots analogous to experiment a–c, but comparing proportion of cell clusters as derived from control and Rorc(t)^ΔCD11c mice. The dashed line at 43% indicates the total contribution from Rorc(t)^ΔCD11c mutants. e, Chromatin accessibility profiles for Rorc, Prdm16 and Clec9a loci across the indicated cell types. The green shaded region demarcates the Rorc(t) +7 kb CRE. f, Gating strategy for tolDC, MHCII⁺ ILC3 and cDC populations. The bottom left flow cytometry plot is gated on CD45⁺Ly6G⁻B220⁻TCRγδ⁻TCRβ⁻ and was generated by concatenating samples from four wild-type mice. g, Expression of the indicated proteins in tolDCs, MHCII⁺ ILC3s and cDCs, as gated in f. h, Numbers of tolDCs and MHCII⁺ ILC3s in the mLN of 3-week-old and 9-week-old control and Δ+7 kb mice. n = 4 per group. i, Numbers of tolDCs and MHCII⁺ ILC3s in the mLN of 3-week-old and 8-week-old control and Rorc(t)^ΔCD11c mice. n = 4 per group. Data in f–i are representative of two (h,i) or three (f,g) independent experiments. Data are mean ± s.e.m. Unpaired two-sided t-test. Source Data

Fig. 3. PRDM16-dependent tolDCs promote microbiota-specific and food antigen-specific pT_reg cell differentiation.

a, Phenotype of Hh-specific T cells in the LILP of Hh-colonized control (Prdm16^fl/fl and Prdm16^fl/wt; n = 4) and Prdm16^ΔRORγt (Rorc(t)^crePrdm16^fl/fl; n = 4) mice at 14 days after adoptive transfer of naive Hh7-2tg CD4⁺ T cells. b, Experimental design for the experiments in c–f. Created in BioRender; Fu, L. (2025) https://BioRender.com/o19q348. LP, lamina propria. c–f, Representative flow cytometry plots and frequencies of OT-II pT_reg (FOXP3⁺RORγt^+/−), T_H2 (FOXP3⁻GATA3⁺), T_H17 (FOXP3⁻GATA3⁻RORγt⁺), T_H1 (FOXP3⁻GATA3⁻RORγt⁻T-bet⁺) and T_FH (FOXP3⁻GATA3⁻RORγt⁻T-bet⁻BCL6⁺) cells in the mLN of OVA-treated control and Prdm16^ΔRORγt mice (c) and SILP of OVA-treated control and Prdm16^ΔRORγt (d), Δ+7 kb (e) or MHCII^ΔRORγt (Rorc(t)^creI-AB^fl/fl) (f) mice at 5 and 12 days post-adoptive transfer of naive OT-II CD4⁺ T cells. The flow cytometry plots are gated on total OT-II cells (CD45⁺B220⁻TCRγδ⁻TCRβ⁺CD4⁺Vα2⁺Vβ5.1/5.2⁺GFP⁺). c, mLN: control mice, n = 4; Prdm16^ΔRORγt mice, n = 3. d, SILP: control mice, n = 4; Prdm16^ΔRORγt mice, n = 4. e, SILP: control mice, n = 4; Δ+7 kb mice, n = 4. f, SILP: control (I-AB^fl/fl) mice, n = 6; MHCII^ΔRORγt mice, n = 6. Data in f are pooled from two independent experiments; data in a,c–e are representative of two (a,c,d) or three (e) independent experiments. Data are mean ± s.e.m. Unpaired two-sided t-test. Source Data

Fig. 4. tolDCs are required to develop oral tolerance.

a, Experimental design for the airway allergy experiments in b–i. IG, intragastric; IN, intranasal; IP, intraperitoneal. b, H&E staining and inflammation score of lung sections from control and Δ+7 kb mice. Scale bars, 500 μm. Non-tolerized control mice, n = 3; tolerized control mice, n = 4; non-tolerized Δ+7 kb mice, n = 3; tolerized Δ+7 kb mice, n = 3. c–f, Eosinophil (CD45⁺CD11b⁺CD11c^low/−Siglec-F⁺) numbers in the lung (c), OVA-specific IgE and IgG1 levels in the serum (d) and phenotype (e) and quantification (f) of OVA:I-A^b tetramer⁺ CD4 T cells in the lung of control and Δ+7 kb mice. Flow cytometry plots in e were generated by concatenating the samples from each group. Non-tolerized mice, n = 4 per group; tolerized mice, n = 5 per group. g, Eosinophil numbers in the lung of control and Prdm16^ΔRORγt mice. n = 4 per group. h, Eosinophil numbers in the lung of control (n = 5 per group) and Rorc(t)^ΔCD11c (n = 4 per group) mice. i, Eosinophil numbers in the lung of control (n = 5 per group) and MHCII^ΔRORγt (n = 4 per group) mice. j, Experimental design for the food allergy experiments in k,l. CT, cholera toxin. k,l, Changes in rectal temperature (k) and OVA-specific IgE and IgG1 levels in the serum (l) of control and Δ+7 kb mice. n = 8 per group. Data in k,l are pooled from two independent experiments. Data in b–f are representative of two independent experiments. Data are mean ± s.e.m. Unpaired two-sided t-test (b–d,f–i,l) and two-stage step-up method of Benjamini, Krieger and Yekutieli (k). Schematics in a,j Created in BioRender; Fu, L. (2025) https://BioRender.com/o19q348. Source Data

Fig. 5. PRDM16-expressing tolDCs are conserved in humans.

a, Input scheme for human scRNA-seq, with four mLNs resected and enriched for APCs. Created in BioRender; Fu, L. (2025) https://BioRender.com/o19q348. b, UMAP of 12,928 resultant human mLN transcriptomes. c, Dot plot of indicated clusters from b examining the expression of genes previously ascribed to RORγt-APC subsets. d, UMAP derived from public datasets of human lamina propria (ileum and colon) resected from six donors, enriched for APCs. e, UMAP derived from public datasets of human tonsil from nine donors, enriched for APCs. Black arrows in b,d,e highlight populations of tolDCs. f, Chromatin accessibility profiles for RORC, PRDM16 and CLEC9A loci across the indicated cell types. Green shaded region demarcates the human orthologue of mouse Rorc(t) +7 kb CRE, part of conserved non-coding sequence 9 (CNS9).

Extended Data Fig. 1. RORγt expression in CD11c lineage APCs is necessary for directing the differentiation of gut microbiota-specific pTregs.

a, Numbers of Hh-specific pTreg, Th17 and Th1 cells in the LILP of Hh-colonized control (n = 7) and Rorc(t)^ΔCD11c (n = 5) mice at 14 days after adoptive transfer of naïve Hh7-2tg CD4⁺ T cells. b, Phenotype of Hh-specific T cells in the mLN of mice shown in a. c, Phenotype of host CD4 T cells in the LILP of mice shown in a, with representative flow cytometry profiles (left) and aggregate quantitative data (right). The flow cytometry plots are gated on total (upper) and FOXP3^– (lower) host CD4 T cells (CD45⁺B220^–TCRγδ^–TCRβ⁺CD4⁺CD90.1^–). Data are pooled from two independent experiments. Data are means ± s.e.m.; ns, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test. Source Data

Extended Data Fig. 2. Characterization of lineage-specific Rorc(t) cis-regulatory elements.

a,b, Comparison of SILP GFP⁺ and mCherry⁺ populations (a) as well as mCherry expression of in vitro differentiated Th17 cells (b) in BAC transgenic mice bred to RORγt-GFP knock-in/knockout reporter mice. a, Tg (Control Rorc(t)-mCherry);Rorc(t)^wt/gfp mice, n = 3; Tg (Control Rorc(t)-mCherry) mice, n = 2; Tg (Δ+3 kb Rorc(t)-mCherry);Rorc(t)^wt/gfp mice, n = 3. c, Frequencies of SILP ILC3 in control (n = 12), Δ+6 kb (n = 3) and Δ+7 kb (n = 19) mice. d, Frequencies of SILP Tγδ17 in control (n = 9), Δ+6 kb (n = 3) and Δ+7 kb (n = 19) mice. e, Frequencies of SILP Th17 in control (n = 13), Δ+6 kb (n = 3) and Δ+7 kb (n = 14) mice. f-i, RORγt expression in the SILP RORγt⁺ populations (f,g) and in vitro differentiated Th17 cells (h,i) from mice shown in c-e. j,k, Phenotype of ILC3 (CD45⁺Lin^–CD127⁺RORγt⁺) subsets in the SILP (j) and LILP (k) of control and Δ+7 kb mice. SILP: control mice, n = 3; Δ+7 kb mice, n = 5. LILP: control mice, n = 6; Δ+7 kb mice, n = 4. l, Numbers of Peyer’s patches in control (n = 5) and Δ+7 kb (n = 5) mice. m-o, Body weight changes (m), fecal C. rodentium counts (l) and frequencies of LILP IL-22⁺ ILC3 (o) (day 14, ex vivo stimulation with IL-23) in control (n = 7) and Δ+7 kb (n = 5) mice post infection. Data in m-o are pooled from two independent experiments. Data in j-l are representative of two (j,k) or three (l) independent experiments. Data are means ± s.d. (a,c-e) or means ± s.e.m. (j-o); ns, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test. Source Data

Extended Data Fig. 3. Rorc(t) +7 kb regulates RORγt⁺ Treg in a cell-extrinsic manner.

a, Numbers of Hh-specific T cells in the LILP of Hh-colonized control (n = 6) and Δ+7 kb (n = 4) mice at 14 days after adoptive transfer of naïve Hh7-2tg CD4⁺ T cells. b, Phenotype of Hh-specific T cells in the mLN of mice shown in a. c, Phenotype of host CD4 T cells in the LILP of mice shown in a. d, Experimental design for the bone marrow (BM) chimeric experiments in e-i. Created in BioRender; Fu, L. (2025) https://BioRender.com/o19q348. e,f, Relative CD45.2⁺CD45.1^– leukocyte chimerism normalized to CD45.2⁺CD45.1^– splenic B cells. n = 5 per group. g-i, RORγt mean fluorescence intensity (MFI) of CD45.2⁺CD45.1⁺ and CD45.2⁺CD45.1^– RORγt⁺ cells in the SILP and LILP. n = 5 per group. Data are representative of two independent experiments. Data are means ± s.e.m.; ns, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test (a-c,e,f) and paired two-sided t-test (g-i). Source Data

Extended Data Fig. 4. Sequencing annotation and analysis of Δ+7 kb mouse model mLN and multiome datasets.

a, Dot plot of all 16 cell types from Δ+7 kb mouse model mLN (mutant and control mice combined), demonstrating canonical genes used to annotate each cluster. b, Violin plot of Rorc expression across all clusters. c, Dot plot of select APC clusters, examining genes described for Nrg1_Pos as well as FRC/mTEC cell types. d, Violin plot of Prdm16 expression across all clusters. e, Violin plots of Rorc and Ccr6 expression within the ILC3 cluster, comparing Δ+7 kb mutant versus control biological conditions. f, Annotated UMAP with datasets combined from all murine experiments (all mutant and control animals, as well subsequent multi-ome experiment). g, Re-analysis of raw data from Akagbosu et al., which was computationally integrated alongside data in f.

Extended Data Fig. 5. Comparing the epigenetic landscape across APC populations.

a-h, Chromatin accessibility profiles for Itgax, Itgam, Itgae, Sirpa, Cxcr6, Csf1r, Flt3, and Clec10a loci across the indicated mouse APC populations.

Extended Data Fig. 6. Selective loss of tolDC in Rorc(t)^ΔCD11c mice.

a,b, Expression of the indicated proteins in tolDC, MHCII⁺ ILC3 and cDC, as gated in Fig. 2f. c, Numbers and frequencies of tolDC (CD45⁺Ly6G^–B220^–TCRγδ^–TCRβ^–MHCII⁺RORγt⁺CXCR6^–Prdm16^high) in mLN from week 1 to week 12. n = 6 (week 1); n = 5 (all other timepoints). d, Representative flow cytometry plots (top), frequencies (bottom left) and RORγt MFI (bottom right) of tolDC and MHCII⁺ ILC3 in the mLN of 3-week-old and 9-week-old control and Δ+7 kb mice. n = 4 per group. e, Representative flow cytometry plots (left) and frequencies (right) of tolDC and MHCII⁺ ILC3 in the mLN of 3-week-old and 8-week-old control and Rorc(t)^ΔCD11c mice. n = 4 per group. The top flow cytometry plots are gated on CD45⁺Ly6G^–B220^–TCRγδ^–TCRβ^–. Data in a,b,d,e are representative of two (d,e) or three (a,b) independent experiments. Data are means ± s.e.m.; ns, not significant; **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test. Source Data

Extended Data Fig. 7. tolDC deficiency leads to type 2 gastrointestinal pathology.

a,b, Representative flow cytometry plots and numbers of tolDC and MHCII⁺ ILC3 in the mLN (a), SILP and LILP (b) of control (n = 4) and Prdm16^ΔRORγt (n = 4) mice. The flow cytometry plots are gated on CD45⁺Ly6G^–B220^–TCRγδ^–TCRβ^–MHCII⁺RORγt⁺. c, Numbers of Hh-specific T cells in the LILP of Hh-colonized control (n = 4) and Prdm16^ΔRORγt (n = 4) mice at 14 days after adoptive transfer of naïve Hh7-2tg CD4⁺ T cells. d, Phenotype of Hh-specific T cells in the mLN of mice shown in c. e,f, Representative flow cytometry plots and frequencies of RORγt⁺ Treg, Th17 and Th2 cells in the SILP and LILP of control and Δ+7 kb mice. e: n = 6 per group (SILP), n = 4 per group (LILP); f: n = 6 per group (SILP), n = 5 per group (LILP), per group. g,h, Frequencies of RORγt⁺ Treg, Th2 and Th17 cells in the SILP of control and mutant mice. g, control mice, n = 6; Rorc(t)^ΔCD11c mice, n = 4. h, control mice, n = 4; Prdm16^ΔRORγt mice, n = 4. i,j, H&E staining and inflammation score (i), and average muscularis propria thickness (j) of the distal small intestine sections in 8-week-old and 40-week-old control (n = 4) and Δ+7 kb (n = 4) mice. Scale bars, 100 μm. k,l, Small intestine length (k) and serum total IgE levels (l) of 8-week-old and 40-week-old control (n = 4) and Δ+7 kb (n = 4) mice. Data are representative of two (a-d,g,h) or three (e,f) independent experiments. Data are means ± s.e.m.; ns, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test. Source Data

Extended Data Fig. 8. tolDC are essential for the differentiation of food antigen-specific pTregs.

a-f, Phenotype of OT-II pTreg, Th2, Th17, Th1 and Tfh cells in the mLN, SILP and LILP of OVA-fed control and mutant mice at 5 and 12 days post-adoptive transfer of naïve OT-II CD4⁺ T cells. a, mLN: control mice, n = 8; Δ+7 kb mice, n = 8. b, mLN: control mice, n = 4; Rorc(t)^ΔCD11c mice, n = 4. c, mLN: control mice, n = 4; Prdm16^ΔRORγt mice, n = 3. d, SILP: control mice, n = 4; Prdm16^ΔRORγt mice, n = 4. e, SILP: control mice, n = 4; Δ+7 kb mice, n = 4. f, LILP: control mice, n = 4; Δ+7 kb mice, n = 4. Data in a are pooled from two independent experiments. Data in b-f are representative of two (b-d) or three (e,f) independent experiments. Data are means ± s.e.m.; ns, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test. Source Data

Extended Data Fig. 9. tolDC are required for establishing oral tolerance against allergic airway responses.

a, Magnified images of the lung sections in Fig. 4b. Scale bars, 100 μm. b-h, Eosinophil numbers in the BALF (bronchoalveolar lavage fluid) (b), phenotype of total CD4 T cells (c-f) and OVA:I-A^b tetramer⁺ CD4 T cells (g,h) in the lung of control and Δ+7 kb mice shown in Fig. 4c–f. Flow cytometry plots in g,h were generated by concatenating the samples from each group. Non-tolerized mice, n = 4 per group; tolerized mice, n = 5 per group. i, Phenotype of total Th2 cells in the lung of control and Prdm16^ΔRORγt shown in Fig. 4g. n = 4 per group. j, Phenotype of total Th2 cells in the lung of control (n = 5 per group) and Rorc(t)^ΔCD11c (n = 4 per group) shown in Fig. 4h. k, Phenotype of total Th2 cells in the lung of control (n = 5 per group) and MHCII^ΔRORγt (n = 4 per group) shown in Fig. 4i. Data in b-h are representative of two independent experiments. Data are means ± s.e.m.; ns, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001; statistics were calculated by unpaired two-sided t-test. Source Data

Extended Data Fig. 10. Analysis of differentially upregulated genes shared by human and mouse tolDC.

a, Dot plot of all 15 cell types from human mLN sc-RNA-seq experiment, demonstrating canonical genes used to annotate each cluster. b, Enumeration of differentially upregulated genes above a threshold of Log₂(Fold Change) = 3.2 after integrating all available data, from all tissues, for indicated murine and human APC populations. Overlaps within the Venn diagrams demonstrate conserved genes. c, Volcano plot of all 141 genes differentially upregulated in human tolDC, with explicit annotation of the 8 genes shared with mouse tolDC, as seen in b. P values (computed via FindMarkers Seurat algorithm) are two-sided and adjusted by Bonferroni correction using all genes in the dataset.