Dendritic Cells Control Regulatory T Cell Function Required for Maintenance of Intestinal Tissue Homeostasis

Dendritic cells (DCs) can capture and process foreign- and self-derived Ags and, depending on the context, induce either immunity in presence of danger or tolerance to harmless substances and self-molecules. This balancing of immunity and tolerance makes DCs key players in maintaining immune homeostasis and preventing autoimmunity. Tolerance induction is especially important in mucosal tissues, where the immune system is constantly confronted with environmental Ags derived from food or commensal bacteria that would otherwise elicit ongoing immune responses. Several mechanisms are described by which DCs can induce tolerance in the periphery. Among them are the induction of anergy and clonal deletion of autoreactive T cells (1, 2). In addition, Foxp3⁺ regulatory T cells (Tregs) play a major role in prevention of spontaneous autoimmunity (3). Treg differentiation can be induced by tolerogenic DCs via cell–cell contact-dependent signaling and through secretory proteins like TGF-β, IL-10, IL-2, or IL-35 (4, 5). Tregs in turn are able to induce tolerogenic DCs via inhibitory surface molecules like CTLA-4 or Lag-3, secretion of IL-10, and inhibition of autophagy in DCs (6, 7). Impaired Treg or DC function can have fatal effects on immune homeostasis. Loss of Treg function may lead to severe autoimmune diseases like immunodysregulation polyendocrinopathy enteropathy X-linked (IPEX) syndrome and is also associated with autoimmune disorders such as type 1 diabetes, systemic lupus erythematosus, and multiple sclerosis (8, 9).

Our previous experiments showed that mice constitutively lacking DCs (∆DC) develop autoimmunity characterized by reduced body weight, hyperimmunoglobulinemia, autoantibodies, splenomegaly, and intestinal inflammation (10). Several other studies provide evidence for DC-controlled intestinal Treg function. In the lamina propria, CD11b⁺CD103⁺ DCs can induce Treg differentiation through TGF-β secretion and retinoic acid metabolism (11). It was further shown that integrin α(v)β8 on DCs is required for activation of TGF-β and prevention of autoimmunity and inflammatory bowel disease (12). Constitutive CD40 signaling in DCs resulted in dysbiosis and fatal colitis caused by abolished induction of RORγ⁺ Tregs (13). Disruption of the calcineurin–NFAT–IL-2 pathway in CD11c⁺ DCs leads to a reduction in Tregs, an expansion of Th1 and Th17 cells, and development of spontaneous colitis (14). MHC class II (MHC-II) deletion on DCs was reported to result in intestinal inflammation in a microbiota-dependent manner (15). So far, direct effects of impaired DC function or abundance on Treg-dependent immune homeostasis have not been investigated, although a correlation between Treg and DC numbers was described (16).

To investigate the role of DCs in regulation of Treg-controlled intestinal homeostasis, we used ∆DC mice generated by crossing CD11c-Cre BAC-transgenic mice with mice harboring the diphtheria toxin α-chain under the control of a loxP-flanked stop cassette, which leads to depletion of more than 95% of conventional DCs, the majority of plasmacytoid DCs, and Langerhans cells (10). In ΔDC mice, we observed an altered transcriptome of peripheral but not thymic Tregs, which expressed higher levels of inhibitory receptors and were less potent in suppressing T cell proliferation in vitro. However, Tregs from ΔDC mice were functional in an in vivo model of transfer colitis but only in the presence of DCs in recipient Rag1-ko mice. Alterations of the Treg compartment in ΔDC mice correlated with higher IgA production and more IgA-coated microbiota in the feces. Our data demonstrate a critical role for DCs to maintain intestinal Treg functionality. Missing this control by DCs leads to loss of intestinal tissue homeostasis resulting in intestinal inflammation.

Constitutively DC-ablated ΔDC mice have been described (10). For generating ∆DC_Foxp3-GFP mice and ∆DC_Rag1–knockout (ko) mice, ΔDC mice were crossed to Foxp3-GFP mice (17) and Rag1-ko mice (18). Ly5.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ) and Rag1-ko (B6.129S7-Rag1^tm1Mom/J) mice were originally obtained from The Jackson Laboratory (Bar harbor, ME) and maintained at our facility. MHC-II-ko [I-A^b-ko (19)] bone marrow was kindly provided by L. Nitschke (Division of Genetics, University of Erlangen). Wild-type (WT) C57BL/6 mice used as recipients for chimeric mice were purchased from Charles River Laboratories (Sulzfeld, Germany). All mice were maintained under specific pathogen-free conditions and used at 6–10 wk of age. All mice were on C57BL/6 background, and all animal experiments were approved by the Federal Government of Lower Franconia and performed in accordance with German animal protection law.

Small intestine and colon samples were digested as described (10). Peyer’s patches (PP) were digested for 10 min at 37°C with 0.7 mg/ml collagenase D (Roche, Basel, Switzerland). Single-cell suspensions were prepared by passing digested intestinal tissue or undigested tissue of mesenteric lymph node (mLN), spleen, and thymus through 70-μm cell strainers (Becton Dickinson, Franklin Lakes, NJ). Afterwards, samples were incubated with anti-CD16/CD32 blocking Ab (clone 2.4G2; Bio X Cell, West Lebanon, NH) and stained in FACS buffer (PBS containing 2% FCS [Life Technologies by Thermo Fisher Scientific, Waltham, MA]) at 4°C for 20 min. The following Abs were purchased from eBioscience at Thermo Fisher Scientific: PerCP-Cy5.5–labeled anti-CD11c (clone N418), FITC-labeled anti-Foxp3 (clone FJK-16S), eFluor 450–labeled anti–KLRG-1 (clone 2F1), PerCP-eFluor 710–labeled anti-TIGIT (clone GIGD7), eFluor 450–labeled anti-Helios, PerCP-Cy5.5–labeled anti-CD4 (clone RM4-5), PE-labeled anti-CD25 (clone PC61.5), PE-labeled anti-IgA (clone mA-6E1), and Alexa Fluor 488–labeled anti-GL-7 (clone GL-7). Allophycocyanin-Cy7–labeled anti–I-A/I-E (clone M5/11.15.2), PE-labeled anti-CD85K (clone H1.1), Alexa Fluor 647–labeled anti-CD103 (clone 2E7), PE-Cy7–labeled CD279 (clone RMP1-30), PerCP-Cy5.5–labeled CD45R (clone RA3-6B2), FITC-labeled anti-CD45RB (clone C363-16A), PE-Cy7–labeled anti-CD38 (clone 90) were purchased from BioLegend (San Diego, CA). PE-labeled anti-ST2 (clone DJ8) was purchased from MD Biosciences (Oakdale, MN) and allophycocyanin-labeled anti-CD138 (clone 281-2) were from BD Pharmingen at Thermo Fisher Scientific.

Intracellular staining for Foxp3 and Helios was performed with the Foxp3/Transcription Factor Staining Buffer Set from eBioscience according to the manufacturer’s protocol.

To exclude dead cells, fixable ViaDye eFluor 506 or eFluor 780 (eBioscience) or DAPI were used. Samples were acquired on FACSCanto II (BD Biosciences) or on LSRFortessa (BD Biosciences), and data were analyzed with FlowJo 10 software (Tree Star, Ashland, OR).

Naive CD4⁺CD45RB^highCD25⁻ T cells or CD4⁺Foxp3-GFP⁺ Tregs were sorted using the S3 sorter from Bio-Rad (Hercules, CA). To speed up the sorting process, T cells were pre-enriched with the CD4⁺ T cell isolation kit (STEMCELL Technologies, Vancouver, Canada).

Fecal samples were prepared as described previously (20). Briefly, samples were solved in 100 μl sterile PBS per 10 mg feces, centrifuged for 5 min at 700 × g, and supernatants were again centrifuged for 5 min at 12,000 × g. The remaining pellet was resuspended in PBS containing 1% BSA and stained with PE-labeled anti-mouse IgA and allophycocyanin-Cy7–labeled anti-mouse Igκ for 30 min on ice. After washing in PBS, samples were fixed in 4% paraformaldehyde overnight and measured by flow cytometry.

Naive CD4⁺CD62L⁺ responder T cells from WT mice and Tregs (CD4⁺CD25⁺) were sorted from spleen of naive WT and ΔDC mice. CD4⁺ responder T cells were stained with CellTrace Violet (Life Technologies by Thermo Fisher Scientific) and cocultured with Tregs from either WT or ∆DC mice at a ratio of 3:1 on anti-CD3–coated plates (2 μg/ml, clone 2C11; from eBioscience at Thermo Fisher Scientific). Proliferation of responder T cells was analyzed after 3 d by flow cytometry.

Small pieces of the colon were fixed in 4% paraformaldehyde overnight at 4°C and stored in PBS until paraffin embedding. Sections were cut at 8 μm thickness and stained with H&E according to the manufacturer’s protocol (Morphisto, Frankfurt am Main, Germany). Images were taken with ZEISS Axio Vert.A1 microscope (Zeiss, Jena, Germany).

The inflammation of the gut was analyzed by high-resolution mini-endoscopy of the colon (Karl Storz, Tuttlingen, Germany) using the murine endoscopic index of colitis severity system as previously described (21).

Total RNA from sorted CD4⁺CD25⁺ Tregs of thymus and spleen was extracted using RNeasy Mini kit (Miltenyi Biotec, Bergisch-Gladbach, Germany). RNA quality was assessed by the Agilent 2100 Bioanalyzer platform, and all samples revealed RNA integrity number values between 8.4 and 9.9, indicating a high RNA purity and integrity. Ten nanograms of each RNA sample was used to generate Cy3-labeled cRNA, which was then hybridized to 8 × 60K Agilent Whole Mouse Genome Oligo Microarrays. Labeling and hybridization was done by Miltenyi Biotec. The dataset has been submitted to the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/geo/) with accession number GSE120690.

CD4⁺ T cells from spleen and mLN of ΔDC_Foxp3-GFP reporter mice were isolated using the mouse CD4 Nanobeads (BioLegend). A total of 16–22 × 10⁶ cells were i.v. injected into WT mice. After 1 and 3 d, Tregs in mLN were analyzed by flow cytometry. Transferred Tregs were gated as Foxp3-GFP–positive. Endogenous Tregs were stained intracellularly for Foxp3.

Bone marrow cells were prepared from ΔDC, WT, and I-A^b ko mice. A total of 1 × 10⁶ cells of a mixture of 80% ΔDC bone marrow and 20% WT or I-A^b ko bone marrow were i.v. transferred into 8 Gy irradiated WT B6 mice. Mice were treated with antibiotics in drinking water for 4 wk (2 g/l neomycin sulfate, 100 mg/l polymyxin B; Sigma-Aldrich, St Louis, MO). Mice were weight twice a week and analyzed at week 11 after cell transfer.

A total of 4 × 10⁵ sorted naive T cells (CD4⁺CD25⁻CD45RB^high) were transferred i.p. into Rag1-ko or ∆DC_Rag1-ko mice to induce colitis. In some experiments, 1.3 × 10⁵ sorted Tregs of WT or ∆DC_Foxp3-GFP reporter mice were cotransferred. Weight was determined and fecal samples were collected weekly. Mice were sacrificed at indicated time points and colon samples were taken for histology.

To determine IgA levels in the serum, standard ELISA was performed using goat anti-mouse Ig for coating and goat anti-mouse IgA-AP for detection (Southern Biotech, Birmingham, AL). For measuring Lipocalin-2 levels, the mouse Lipocalin-2/NGAL DuoSet ELISA (R&D systems, Minneapolis, MN) was used according to the manufacturer’s protocol. Samples were prepared as described by Chassaing et al. (22). Briefly, feces were collected and homogenized with PBS containing 1% Tween20 (10 mg feces/ml; AppliChem, Darmstadt, Germany) in a shaker for 25 min at room temperature. After centrifugation, supernatants were diluted in PBS containing 1% BSA (Carl Roth, Karlsruhe, Germany) and used for ELISA.

Student t test was performed with SigmaPlot (Systat Software, San Jose, CA) software and p values < 0.05 were considered statistically significant.

All data are available from the authors upon request.

Tregs play an essential role in maintaining self-tolerance, and changes in Treg numbers or function can lead to autoimmunity (3). To investigate whether the constitutive lack of DCs affects the gene expression profile of thymic Tregs or peripheral Tregs, we performed microarray analysis with Tregs sorted from thymus and spleen of WT and ∆DC mice. Comparison of gene expression profiles revealed that there are only minor differences of the transcriptomes between thymic Tregs from WT and ∆DC mice. Only 11 genes with a log₂-fold change >2 and a combined expression level >10 were upregulated in thymic Tregs of ∆DC in comparison with WT mice (Fig. 1A). However, 288 genes were higher expressed in splenic Tregs of ΔDC compared with WT mice (Fig. 1A). This suggests that Tregs from WT and ∆DC mice are more distinct from each other in the periphery than in the thymus. Principal component analysis confirmed these findings as samples derived from thymus clustered close together, whereas spleen samples from WT and ∆DC mice did not (Fig. 1B). Furthermore, we performed unsupervised hierarchical cluster analysis of the 1000 topmost differentially expressed genes among WT and ∆DC Tregs from thymus and spleen and again Tregs from thymus appeared to be more similar to each other than Tregs derived from spleen (Fig. 1C).

Among the most differentially regulated genes between WT and ∆DC Tregs in spleen are inhibitory receptors like ILT3 (Lilrb4, −4.13), TIGIT (Tigit, −2.29), TIM-3 (Havcr2, −4.57), or LAG3 (Lag3, −1.10) showing higher expression on Tregs from ∆DC as compared with WT mice (Supplemental Fig. 1). Moreover, members of the killer cell lectin like receptor family with inhibitory functions like KLRB1 (Klrb1a, −4.90; Klrb1c, −4.99), KLRC1 (Klrc1, −6.08), KLRD1 (Klrd1, −4.25), and KLRG1 (Klrg1, −2.98), are higher expressed on ∆DC Tregs in spleen in comparison with the expression on WT Tregs. Interestingly, no differences in expression of the Treg-associated master transcription factor Foxp3 (Foxp3, 0.00) and the coinhibitory molecule CTLA4 (Ctla4, 0.00) were detected between WT and ∆DC Tregs, neither in thymus nor in spleen.

We selected some candidate genes, which are known to be associated with Treg function and determined their expression on Tregs from thymus and spleen of WT and ∆DC mice by flow cytometry. We analyzed expression of TIGIT, PD-1, and ILT3, which have been shown to transduce inhibitory signals into T cells (23–25). We also included analysis of the transcription factor Helios, which can be used as a marker for naturally occurring, thymic-derived Tregs, although it may also be induced in some peripheral Tregs under certain conditions (26, 27). As expected, the expression of TIGIT, PD-1, ILT3, and Helios was comparable on thymic Tregs from WT and ΔDC mice (Fig. 1D, 1E). In contrast, we found a significant upregulation of TIGIT and ILT3 as well as a trend to higher expression of PD-1 on splenic Tregs from ∆DC mice (Fig. 1D, 1E).

Next, we asked whether Tregs that developed in a DC-deficient environment downregulate expression of inhibitory receptors when DCs are present. Therefore, we transferred Tregs from ΔDC_Foxp3-GFP reporter mice into WT mice. After 1 d, the frequencies of TIGIT-, PD1-, and ILT3-positive transferred Tregs started to decrease compared with frequencies detected before transfer and adapted the phenotype of the endogenous Treg population 3 d after transfer (Fig. 1F). This demonstrates that upregulation of inhibitory receptors on Tregs from DC-deficient mice is a reversible process.

The different gene expression profile of peripheral Tregs from ∆DC mice suggested that this may affect the functionality of the cells. To address this question, we performed in vitro suppression assays by coculturing CellTrace-labeled CD4⁺ T cells from WT mice together with Tregs sorted from the spleen of either WT or ∆DC mice and analyzed the capacity of the Tregs to suppress T cell proliferation. WT Tregs were able to suppress more than 60% of T cell proliferation, whereas Tregs derived from ∆DC mice showed only ∼15% suppression (Fig. 2A, 2B).

To analyze the functionality of Tregs from ∆DC mice in vivo, we used the model of T cell transfer colitis. Transfer of naive T cells into immunodeficient mice leads to severe colitis, but cotransfer of Tregs can prevent disease onset by suppressing proliferation and activation of colitogenic T cells (28). We transferred naive CD4⁺CD25⁻CD45RB^high T cells from WT mice together with Tregs from WT_Foxp-eGFP or ΔDC_Foxp3-eGFP mice into Rag1-ko recipients at a ratio of 3:1 and monitored the mice for more than 8 wk for clinical signs of disease development. Both groups of recipient mice did not show signs of colitis like hunched posture or diarrhea. Also, mice that received WT Tregs and mice that received ∆DC Tregs did not show weight loss even over a long period of time (Fig. 2C). After 62 d we analyzed the recipient mice by endoscopy and afterward performed histology sections of the colon. Again, both groups of mice did not show symptoms of colitis (Fig. 2D). This result indicates that ∆DC-derived Tregs are fully functional in vivo.

We speculated that the discrepancy between in vitro and in vivo functionality of ΔDC-derived Tregs might be due to the fact that DCs are present in the Rag1-ko mice in the colitis model. To figure out if endogenous DCs of the recipients are important for the outcome of transfer colitis, we generated Rag1-ko mice lacking DCs by breeding ∆DC mice on a Rag1-ko background. We first tested whether colitis can be induced in a system that lacks DCs as APC to prime colitogenic T cells. Therefore, we transferred only naive CD4⁺CD45RB^high T cells in ∆DC_Rag1-ko mice and monitored the mice for signs of colitis. The weight loss in ∆DC_Rag1-ko mice was slightly increased when compared with normal Rag1-ko mice (Fig. 3A). We also determined Lipocalin-2 levels from fecal samples as a noninvasive marker for intestinal inflammation as described previously (22). Here, we observed a strong increase in Lipocalin-2 levels in both Rag1-ko and ∆DC_Rag1-ko mice, indicating normal development of colitis in both mouse strains (Fig. 3B). Three weeks after cell transfer, we also noticed cellular infiltrates, loss of goblet cells and epithelial hyperplasia in both groups of recipient mice (Fig. 3C). These results show that DCs are not required for activation of colitogenic CD4⁺ T cells in this model of transfer colitis.

Next, we performed a cotransfer of naive T cells from WT mice together with Tregs from either WT_Foxp3-eGFP or ∆DC_Foxp3-eGFP mice into Rag1-ko or ∆DC_Rag1-ko mice and monitored weight and Lipocalin-2 levels for 6 wk. Both groups of T cell–transferred Rag1-ko mice gained weight, displayed only low levels of Lipocalin-2 in fecal samples, and showed no further signs of inflammation. In contrast, both groups of T cell–transferred ∆DC_Rag1-ko mice did not gain weight and developed mild colitis indicated by slightly elevated Lipocalin-2 levels and colonic cellular infiltrations (Fig. 3D–F).

These data demonstrate that DCs are indeed important for Treg function. Without DCs, Tregs lose their ability to efficiently suppress effector T cells in vivo, which may lead to intestinal inflammation.

We further determined whether lack of DCs also affects the phenotype of Tregs in intestinal tissues. Because some macrophages can also express low levels of CD11c in the intestinal mucosa and might therefore be deleted in ΔDC mice, we first addressed this issue by gating on all CD11c⁺ cells and determined the frequency of DCs (CD103^hiF4/80^lo) and monocytes/macrophages (CD103^loF4/80^hi) within this population. CD11c^hi DCs appeared to be efficiently deleted, whereas CD11c^lo monocytes/macrophages were not (Fig. 4A). In addition, we and others have previously shown that only a few CD11c⁺ activated T cells express the Cre recombinase in CD11cCre mice, and this population is not diminished in ΔDC mice (29, 30). We performed flow cytometric analysis of Tregs isolated from the lamina propria of small intestine and colon of ΔDC and WT mice for expression of Foxp3, ST2, KLRG1, CD103, ILT3, TIGIT, and Helios. The expression levels of Foxp3 in small intestine and colon were comparable between WT and ΔDC mice, but the frequency of Tregs within the CD4⁺ T cell population was reduced in small intestine and colon of ΔDC mice as compared with controls, which likely reflects the expansion of effector CD4⁺ T cells in these tissues, as we have shown before (10) (Fig. 4B).

In the small intestine of ΔDC mice, we observed more ST2⁺ and TIGIT⁺ Tregs as compared with controls (Fig. 4C–F). In contrast, the colonic Treg population of ΔDC mice contained an increased frequency of ILT3⁺ and CD103⁺ Tregs (Fig. 4C–F). The KLRG1 receptor marks terminally differentiated Tregs (31), and its expression was also slightly increased on Tregs from ∆DC mice in small intestine and colon (Fig. 4B–E). Helios was expressed in ∼50% of Tregs in the colon but in only 10% of Tregs in the small intestine of both mouse strains (Fig. 4C–F).

Overall, the flow cytometric analysis demonstrates that similar to the spleen, Tregs with expression of inhibitory receptors are also more frequent in small intestine and colon of ΔDC mice.

We next investigated whether cognate interaction between DCs and Tregs is essential for instruction of proper Treg function. To address this question, we generated mixed bone marrow chimeric mice by mixing 80% ΔDC bone marrow with either 20% WT or 20% I-A^b-ko bone marrow, resulting in mice that have either WT DCs or DCs lacking MHC-II (Fig. 5A). Around 40 d after bone marrow reconstitution, mice lacking MHC-II on DCs started losing weight and by the end of the experiment developed diarrhea and intestinal inflammation, whereas mice with WT DCs did not lose weight (Fig. 5B). After 11 wk we analyzed spleen and mLN of the chimeric mice. We found no significant differences in the frequency of CD4⁺Foxp3⁺ T cells in spleen and mLN of mice containing WT DCs or MHC-II–deficient DCs (Fig. 5C). However, Tregs of mice lacking MHC-II on DCs showed a significant upregulation of the inhibitory receptor TIGIT in spleen and in mLN as well as a trend of upregulated ILT3 in mLN, whereas expression of PD-1 was not altered (Fig. 5D, 5E). We further observed a trend to higher levels of IgA in the serum of mice lacking MHC-II on DCs in comparison with WT controls (Fig. 5F). Taken together, these findings indicate that MHC-II on DCs may promote proper Treg function and maintenance of intestinal homeostasis.

Intestinal Tregs have been shown to regulate IgA selection and diversification in the gut, which may affect the composition of the commensal flora (20). Therefore, we further analyzed IgA-producing B cells in the gut. We determined IgA expression in germinal center (GC) B cells and plasma cells (PC) from mLN and PP of WT and ∆DC mice by flow cytometry. We observed a significant increase of total PCs, IgA⁺ PCs and IgA⁺ GC B cells in mLN of ∆DC mice (Fig. 6A, 6B). The frequency of GC B cells was significantly higher in PP of ∆DC mice, whereas the frequency of PC, IgA⁺ GC B cells, and IgA⁺ PC was comparable to WT mice (Fig. 6C). Serum IgA levels were also higher in ΔDC mice as compared with WT controls (Fig. 6D). Because one prominent function of soluble IgA in the intestine is the coating of bacteria in the lumen, we next assessed the amount of IgA-coated bacteria in fecal samples of WT and ∆DC mice by flow cytometry. This revealed that in ∆DC mice ∼25% of bacteria is coated with IgA, whereas only around 5% of bacteria are highly IgA-coated in WT mice (Fig. 6E, 6F).

These results show that despite the lack of DCs, ∆DC mice display a higher frequency of IgA producing cells in mLN as compared with WT mice and that commensal bacteria in ΔDC mice are highly coated with IgA.

The communication between DCs and Tregs plays a crucial role in keeping the immune system balanced and preventing autoimmunity. In this study, we investigated the impact of missing DCs on Treg phenotype and function and the resulting autoimmune manifestations in the intestine. Previous studies showed a correlation between DC and Treg numbers. Depleting Tregs led to an increase in DCs by a mechanism that requires Flt3-ligand and diphtheria toxin receptor–mediated depletion of DCs resulted in decreased Treg numbers (16, 32). Depletion of DCs also led to lower Foxp3 levels, increased Th1 and Th17 cells, and autoimmune inflammation, like we previously observed in constitutively DC-depleted ∆DC mice (10). Lower Foxp3 levels have been reported to result in impaired suppressive capacity of Tregs (33). However, Foxp3 levels were not reduced in Tregs of ∆DC mice. This discrepancy might be explained by the constitutive ablation of DCs in our system in which mice grow up in the absence of DCs, and this is not the case in models of transient DC depletion in adult mice.

So far there are no studies describing the impact on the transcriptome of Tregs that developed and persist in the absence of DCs. Transcriptome analysis revealed high expression of inhibitory receptors on peripheral Tregs isolated from ∆DC mice, and some of them are associated with Treg suppressive function but also with inhibition of Treg activity or competitive fitness. Among those inhibitory molecules, we found TIGIT highly upregulated on Tregs from ∆DC mice, especially in the spleen and small intestine. TIGIT⁺ Tregs have been shown to be highly suppressive and to induce the expression of the suppressive mediators IL-10 and fibrinogen-like protein 2 (Fgl2) upon ligation and selectively suppress Th1 and Th17 responses (34, 35). However, TIGIT can also mediate inhibitory signals into T cells, suggesting that the activity of Tregs can be suppressed by this molecule (23). The inhibitory receptor PD-1 was also expressed at higher levels in splenic Tregs from ΔDC as compared with control mice. ILT3 is another inhibitory receptor that we found upregulated in Tregs from colon and spleen of ΔDC mice. ILT3 was described to inhibit the suppressive activity of Tregs (25). The expression of the inhibitory receptor KLRG1 was slightly increased on intestinal Tregs of ΔDC mice as compared with controls and has been recently shown to inhibit Treg competitive fitness in the gut (36). High expression of KLRG1 on Tregs was also shown to have a negative impact on proliferation as well as stability of Foxp3 expression, and KLRG1⁺ Tregs were more prone to apoptosis than KLRG1⁻ Treg subsets in a NOD mouse model of type 1 diabetes (37). Similar effects were observed for the inhibitory receptor LAG3. LAG3 expression on Tregs leads to a downregulation of genes important for Treg maintenance and function such as transcription factor Eos and CD25, which negatively affects proliferation and survival of LAG3⁺ Tregs (38).

The higher expression level of inhibitory receptors on Tregs from ∆DC mice might thereby lead to an impairment of Treg function consistent with the reduced suppressive capacity of isolated Tregs from ∆DC mice during the in vitro proliferation assay. Utilizing the transfer colitis model we could show that once those Tregs can encounter endogenous DCs of the recipient Rag1-ko mice, they are fully functional again. This indicates that Tregs from ∆DC mice are per se functionally impaired but can regain full activity upon contact with DCs, which was also shown by the Treg transfer experiments in which the frequencies of Tregs from ΔDC mice positive for various inhibitory receptors were reduced after transfer into a DC-sufficient environment. The importance of this signal elicited by DCs also becomes obvious in the transfer colitis model with ∆DC_Rag1-ko mice as recipients. In this inflammatory environment without DCs, Tregs from WT mice also fail to efficiently suppress pathology.

A recent study by the Nussenzweig group has shown that conditional deletion of MHC-II in conventional DCs leads to development of an autoimmune phenotype with weight loss, splenomegaly, prolapses, and intestinal inflammation, features we can also observe in ∆DC mice (10, 15) and in part also in the ΔDC+ I-A^b ko chimeras (Fig. 5B). In contrast to our findings in ∆DC mice, they reported reduced GC B cells in mLN and PP, lower serum IgA levels, and lower IgA coating of intestinal bacteria as a result of insufficient CD4⁺ T cell help (15). Further studies should therefore investigate the role of different DC subsets and other CD11c-expressing cell types with regard to regulation of IgA responses in the gut. It was shown that in mice lacking T cells (Tcrb^−/−d^−/− mice) and in mice lacking Tfh cells (CD4-Cre Bcl-6^fl/fl), IgA coating of bacteria was identical to WT controls questioning an absolute requirement of T cells for generation of anti-commensal IgA in the intestine (39).

Reduced Treg frequency in the intestine and its associated tissues and the altered Treg transcriptome, as we describe in this article, can affect the mutual relationship between adaptive immunity and the microbiota. Tregs can modify microbiome composition via influencing IgA production in the GC by giving rise to T follicular Tregs. IgA in turn coats commensals and leads to diversification of the microbiome including enrichment of short chain fatty acid–producing bacteria, thereby facilitating the expansion of Tregs (20, 40). Moreover, highly IgA-coated bacteria have also been reported to be potentially colitogenic, and transfer of those IgA^hi bacteria in germ-free mice made them more susceptible to develop colitis (41). Indeed, we observed higher IgA coating in bacteria isolated from ΔDC mice compared with WT mice, suggesting a dysbiotic microbiota that often is associated with development of intestinal inflammation (42). Whether the enhanced antimicrobial IgA response is caused by the impaired function of Tregs or by the lack of DCs remains to be investigated in future experiments.

From our results, we conclude that Tregs need the presence of DCs to act as potent suppressors of aberrant immune activation in the gut and to support intestinal homeostasis.

We thank the Immunomonitoring and Cell Sorting Core Unit at the University Hospital Erlangen for cell sorting, Christoph Daniel for help with histology, Christoph Koch, Kirstin Castiglione, Daniela Döhler, and Laura Handl for technical assistance, and members of the Voehringer laboratory for helpful discussions.

The authors have no financial conflicts of interest.