Normal colonic mucosa was surgically resected from six colorectal cancer patients at approximately 10cm from the tumor. Samples were part of a tissue biocollection registered by the French Ministry for Higher Education and Research (DC-2014-2206), with approval from the ethic committee (CPP Ouest IV-Nantes) and the institutional board of Nantes University Hospital. Tissues were processed in accordance with Helsinki declaration. Each patient signed an informed consent form. The lamina propria was separated from the epithelium after incubation in 1mM EDTA PBS buffer (20min), minced into fragments and washed with RPMI containing 10μg/ml penicillin and 0.1mg/ml gentamycin (Sigma-Aldrich). Lamina propria fragments were digested with collagenase IV (1mg/ml; Sigma-Aldrich), with shaking at 37°C. Mucus and debris were removed by filtration through a 40μm-cell strainer (BD). Viable lamina propria-derived lymphocytes (LPLs) were obtained through Ficoll gradient centrifugation.

To isolate F. prausnitzii-reactive DP8α Treg clones from blood, purified VPD-stained CD4⁺ T cells, comprising DP8α Tregs, were co-cultured with purified autologous CD14⁺ monocytes loaded overnight with F. prausnitzii (1:1 ratio). Five days later, VPD^LOW CD3⁺/CD4⁺/CD8α^LOW cells were cloned from four healthy donors using the Aria III cell sorter ( Figure S1 ). Alongside, from the same donors, CD3⁺/CD4⁺/CD8^-/CD25^HIGH/CD127^LOW classical FoxP3⁺ Tregs were similarly cloned ( Figure S1 ). DP8α Tregs and FoxP3⁺ Tregs were also cloned from freshly-dissociated and sorted colonic LPLs ( Figure S1 ). Clones were then amplified on feeder cells (6). After 4 weeks, resting clones were screened for their response to autologous monocytes presenting F. prausnitzii and their FoxP3 expression.

Cells were cultured in RPMI-1640 supplemented with 5% human serum, 2mM L-glutamine and 10μg/ml penicillin-streptomycin (Gibco). rhIL-2 was used for T cell culture and expansion. Violet Proliferation Dye 450 (VPD) (1μM, BD Bioscience), Brefeldin A (10μg/ml, Sigma-Aldrich) and 4% paraformaldehyde (Sigma-Aldrich) were used. POM-1 and AB-680 (MedChemExpress) drugs, which inhibit CD39 and CD73, respectively, were used as indicated in the figure legends.

For surface staining, cells were washed and stained for 40min at 4°C in PBS/0.1% BSA with the following antibodies: anti-CD3 (clone UCHT1, Becton Dickinson), anti-CD4 (clone 13B8.2, Beckman Coulter), anti-CD8α (clone RPA-T8, Becton Dickinson), anti-CCR6 (clone G034E3, Biolegend), anti-CXCR6 (clone K041E5, Biolegend), anti-CXCR5 (clone 51505, R&D Systems), anti-CD39 (clone A1, Biolegend), anti-CD73 (clone AD2, Biolegend), anti-CD27 (clone O323, Biolegend), anti-CD40L (clone 24-31, Abcam), anti-CD19 (clone SJ25C1, Biolegend), anti-CD38 (clone HB-7, Biolegend), anti-CD25 (clone BC96, eBiosience), anti-CD127 (clone HIL-7R-M21, Becton Dickinson), anti-CRTAM (clone 210213, R&D Systems), and anti-CCR5/CD195 (clone 3A9, Becton Dickinson).

For intracellular staining, cells were fixed in 4% paraformaldehyde for 10min, washed and stained for 40min at RT in PBS/0.1% BSA/0.1% saponin with anti-IFNγ-APC (clone B27, Becton Dickinson), anti-CCL4/MIP-1β (clone D21-1351, Becton Dickinson), anti-CCL5/RANTES (clone 2D5, Becton Dickinson), anti-Granzyme A (clone CB9, Biolegend), anti-Granzyme B (clone GB11, Becton Dickinson), anti-Granzyme K (clone GM26E7, Biolegend), anti-IL-2 (clone MQ1-17H12, Becton Dickinson), anti-IL-4 (clone 8D4-8, Biolegend), anti-IL-5 (clone TRFK5, Biolegend), anti-IL-13 (clone JES10-5A2, Biolegend), anti-IL-21 (clone eBio3A3-N2, eBiosience), and anti-Perforin (clone δG9, Becton Dickinson). For transcription factors, cells were fixed and permeabilized using the Foxp3/Transcription Factor Buffer Set (eBioscience) and the following antibodies: anti-FoxP3 (clone 259D/C7, Becton Dickinson), anti-RORγt (clone AFKJS-9, eBioscience), anti-Eomes (clone WD1928, eBioscience) and anti-ID2 (clone ILCID2, eBioscience).

Fluorescence was measured on a FACS LSR II flow cytometer and analyzed using the FlowJo and Diva softwares (Becton Dickinson). Relative fluorescence intensity (RFI) corresponds to the mean fluorescence intensity (MFI) obtained with the specific antibody divided by the MFI obtained with the isotypic control.

T cells were stimulated using coated anti-CD3 (clone OKT3, 1μg/ml, eBioscience) for 24-48h at 37°C. Supernatants were tested for their IL-10 and IFNγ contents using the Ready-Set-Go ELISAs according to the manufacturer’s guidelines (eBioscience). XCL1, total hIgA and hIgG were measured using ELISA kits (Invitrogen) according to the manufacturer’s instructions.

Clones were stimulated or not with 0.5μM OKT3 for 4h, washed extensively and dry pellets were frozen at -80°C. mRNA was extracted from all samples at once (to avoid batch effects) using the NucleoSpin RNA kit (Macherey-Nagel) and quantified with a NanoDrop.

Freshly sorted CD4⁺ T cells (Miltenyi) were stained with 1μM VPD before being co-cultured with Treg clones with 20UI/ml rhIL-2, with or without CD3/CD28 microbeads (Miltenyi). In some cases, CD39 or CD73 inhibitors were added. VPD dilution was assessed 5 days later to determine proliferation.

Statistical analyses were performed using GraphPad Prism version 9.3.0. Most single comparisons were performed using 2-sided t-tests or Mann-Whitney tests, as indicated in figure legends. For multiple comparisons, one-way ANOVA (either Friedman tests for matched data or Kruskal-Wallis tests when there are no matching or pairing), followed by Dunn’s multiple comparison tests to obtained adjusted p values were used, as indicated in figure legends. p<.05 was considered statistically significant.

3’seq-RNA Profiling protocol was performed as described (8). Libraries were prepared from 10ng total RNA in 4µl. The mRNA poly(A) tails were tagged with universal adapters, well-specific barcodes and unique molecular identifiers (UMIs) during template-switching reverse transcriptase. Barcoded cDNAs from multiple samples were pooled, amplified and tagmented using a transposon-fragmentation approach, which enriches for cDNA 3′ ends. 100ng of full-length cDNA were used as input to the Nextera DNA Sample Prep kit (Illumina), which enriches for cDNA 3′ ends. Library’s size was controlled on 2200 Tape Station System (Agilent Technologies). A library of 350–800bp length was run on an Illumina HiSeq 2500, using a Hiseq Rapid SBS Kit v2-50 cycles and a Hiseq Rapid PE Cluster Kit v2 according to the manufacturer’s protocol (Denaturing and Diluting Libraries for the HiSeq^® and GAIIx, Part # 15050107 v03 protocol, Illumina).

Raw fastq pairs match the following criteria: the 16 bases of the first read correspond to six bases for a designed sample-specific barcode and 10 bases for a unique molecular identifier (UMI). The second read (58 bases) corresponds to the captured poly(A) RNA sequence. Bioinformatics steps are performed using a snakemake pipeline (https://bio.tools/3SRP). We performed demultiplexing of these fastq pairs to generate one single-end fastq for each sample. These fastq files were then aligned with bwa to the reference mRNA refseq sequences and the mitochondrial genomic sequence, both available from the UCSC download site.

Gene expression profiles were generated by parsing the alignment files (bam) and counting, for each sample, the number of UMIs associated with each gene. Reads aligned on multiple genes, containing more than three mismatches with the reference sequence or having a polyA pattern were discarded. Finally, a matrix containing the counts of all genes on all samples was produced. The expression values, corresponding to the absolute abundance of mRNAs in all samples, were then ready for further gene expression analysis.

The R package DESeq2 (9) was then used for the differential analyses.

Data are available at GSE197906.

T cell clones were co-cultured with 100,000 indicated target cells at indicated ratios at 37°C, with or without CD3/CD28 activation (T-Cell TransAct polymeric nanomatrix inducing low to no death of activated T cells). Supernatants were harvested 18h later and cell death was measured using a bioluminescent, non-destructive cytolysis assay kit designed to measure the release of the adenylate kinase from damaged cells (ToxiLight™ BioAssay Kit, Lonza). The ToxiLight Lysis Control Set (Lonza) was used to determine 100% lysis. Lysis percentages were calculated as follows:

((Luminescence read - Luminescence of target cells alone lysed spontaneously x 100)/(Luminescence of totally lysed target cells – (Luminescence of target cells alone lysed spontaneously + Luminescence of corresponding T cells alone lysed spontaneously)

Freshly sorted CD3^-/CD19⁺/CD27^- naïve B cells (20x10³ cells/well) were incubated at a 1:1 ratio with autologous sorted DP8α Tregs (CD3⁺/CD4⁺/CD8α^LOW/CCR6⁺/CXCR6⁺), Tfh cells (CD3⁺/CD4⁺/CD8^-/CXCR5⁺) or FoxP3⁺ Treg clones in RPMI-1640 supplemented with 10% fetal bovine serum (to avoid any contamination with human Ig) in the presence of an agonistic anti-BCR (anti-IgA/IgG/IgM) antibody (Jackson ImmunoResearch), with or without CD3/CD28-stimulation. At day 7, frequency of CD3^-/CD19^INT/CD38⁺ plasmablasts among total CD19⁺ B cells was measured, and total IgG and IgA were measured in the supernatants.

We selected 6 DP8α and 4 FoxP3⁺ Treg clones derived from colonic lamina propria lymphocytes (LPLs) from five different donors, and matched-pairs of DP8α and FoxP3⁺ Treg clones from the blood from four healthy donors, as described in the Methods section.

All DP8α clones responded to F. prausnitzii ( Figures 1A-D ), and lacked FoxP3 expression (mean RFI=4.46 ± 1.27 and 6.31 ± 0.95 for blood- and LPL-derived clones, respectively), as compared to FoxP3⁺ Treg clones (mean RFI=21.19 ± 2.08 and 46.29 ± 9.80 for blood- and LPL-derived clones, respectively) ( Figures 1E, F ). Furthermore, DP8α clones secreted IFNγ ( Figures 1A, B ) and IL-10 ( Figures 1C, D ) upon stimulation. A fraction of FoxP3⁺ Treg clones also secreted IL-10, however, none of these clones secreted significant levels of IFNγ, which when co-produced with IL-10 represents a feature of Tr1-like Tregs (10). All DP8α clones expressed the purinergic receptors CD39 and CD73 ( Figures 1G, H ), and blocking these enzymes neutralized their ability to inhibit effector T cell proliferation ( Figures 1I, J ). In contrast, whereas some FoxP3⁺ Treg clones also expressed CD39, they expressed low to no CD73 ( Figures 1G, H ), and their inhibition of effector T cell proliferation was unaltered following CD39 and CD73 inhibition ( Figures 1I, J ). Therefore, DP8α regulatory function, unlike FoxP3⁺ Treg function, depends on the purinergic pathway. Similar to clone data, ex vivo analyzed circulating DP8α Tregs systematically expressed CD39 and CD73 in all donors, whereas CD39 was expressed by blood FoxP3⁺ Tregs in only two donors, and CD73 in none of them ( Figure 1K ).

As shown by Principal Component Analysis (PCA), DP8α clones, whether they were derived from PBLs or LPLs, clustered apart from FoxP3⁺ clones ( Figure 2A ). At variance, DP8α clones of blood and colonic origin, despite their dispersion, clustered in the same area at the resting state and, to some extent, also upon activation, Moreover, no differentially expressed gene (DEG) distinguished resting colon- and blood-derived DP8α clones and only 63 were identified upon activation, underlying the robust relationship between colon- and blood-derived DP8α Tregs (not shown).

In contrast, LPL-derived and PBL-derived FoxP3⁺ clones clustered apart by PCA ( Figure 2A ), indicating significant differences between these cells according to their location. Nonetheless, the number of DEGs between these clones was low both at the resting state (n=98, up: 47, down: 51) and after activation (n=230, up: 99, down: 131). This may result from a significant heterogeneity of colonic FoxP3⁺ Tregs likely due, in part, to the co-existence in this compartment, of tissue-resident Tregs and central memory Tregs circulating through colonic lymphoid structures. Analysis of individual FoxP3⁺ Treg clone transcriptomes and cytokine profiles indeed revealed a duality within colon-derived FoxP3⁺ clone profiles ( Figure S2 ), particularly for CD62L transcripts (SELL) expression, a lymphoid tissue related gene.

We then investigated the activation-induced responses of DP8α Treg clones through DESeq2 analysis of 3’SRP transcriptomic data obtained in activated versus resting colon- and blood-derived DP8α clones. Applying the criteria for significance (FDR < 0.05 and absolute value of log2 fold-change > 0.5), 323 and 379 genes were significantly up-regulated and 93 and 208 genes down-regulated upon activation, in colon- and blood-derived DP8α clones, respectively ( Figures 2B, C ). We first focused on the 102 genes up-regulated in activated colon-derived DP8α clones, for which roles in T cells had already been documented.

Strikingly, ten genes clearly conveyed cytotoxic functions. The class-I-restricted cytotoxic associated molecule, CRTAM, known as an inducer of cytotoxic activity and IFNγ production in CD4⁺ T cells (11), transcription factors (TFs) involved in cytotoxic cell differentiation, EOMES, PRDM1 and RUNX3, genes encoding cytotoxic cytokines or receptors: FASLG, TNFSF10 (or TRAIL), GZMB and LTA, as well as STX11 which facilitates cytolytic granule exocytosis (12) and SERPINB9 ( Figure 2D and Figure S3 ), known to protect GZMB-producing Tregs from lysis (13).

Six genes encoded chemokines: CCL1, which attracts Th2 lymphocytes via CCR8; CCL3 and CCL4, which recruit cells expressing CCR1 and/or CCR5, essentially T cells (14); XCL1, which acts specifically on antigen cross-presenting dendritic cells (15); CXCL8 (IL-8) ( Figure 2E ), a chemokine which can recruit neutrophils (16) and CKLF ( Figure S4 ), which attracts CCR4-expressing targets such as Th2 cells (17).

CD3-mediated activation also up-regulated numerous genes encoding molecules known to provide help to B cells and/or drive antibody class-switch recombination or secretion, such as CD40LG, IL10, IL4, IL21 (18), CD70, IL2, IL13, IL5, LTA, MAF, TNFSF14 (or LIGHT) ( Figure 2F and Figure S5 ), as well as genes encoding additional cytokines or growth factors ( Figure 2G ).

Particular TF combinations were reported to characterize homogeneous cell subsets more accurately than surface markers. Interestingly, 23 TFs, including those mentioned above, were induced or up-regulated by DP8α Tregs upon activation ( Figure 2H and Figure S6 ). Strikingly, all these TFs have been reported to play roles in the differentiation, homeostasis or suppressive functions of Tregs. This is the case for EOMES, which together with PRDM1 (or BLIMP1), induces IL-10 expression and Tr1-like cell differentiation (19, 20), PRDM1 also being a crucial regulator for microbiota-induced murine FoxP3⁺/RORγt⁺ Tregs (21, 22); AHR, that marks and promotes gut-derived Tregs and positively regulates IL-10 and Tr1 differentiation (23); CBFB, a master regulator of genes involved in the induction of FoxP3⁺ Tregs and their suppressive functions (24); EGR2, a Tr1-associated TF (25); ID2, IRF4, BATF and PPARG, which orchestrate the development and maintenance of adipose tissue Tregs (26); IRF1, which together with BATF is required for the development of IL-27-induced Tr1 cells (27); and MAF, a key regulator of intestinal immune homeostasis through sustaining RORγt expression in mouse microbiota-induced gut Tregs and positively regulating IL-10 production (21, 22).

Interestingly, consistent with the Tr1-like properties of DP8α Treg clones detailed above, i.e., mainly the lack of FoxP3 expression together with the secretion of both IL-10 and IFNγ, DP8α Tregs significantly shared features with either human or mouse Tr1 and Tr1-like signatures, namely GZMB, CSF1, HAVCR2 (or TIM3), EGR2, PRDM1, IRF4, MAF, EOMES, AHR, IRF1, BATF, ID2, BHLHE40, TNFRSF9 (or 41BB), TNFRSF4 (or OX40), IL-5, IL-2 ^LOW (19, 20, 28–32) ( Figure 2I ), as well as GZMK (31, 32), identified in DP8α Treg clones as down-regulated upon activation ( Figure S7 ). Moreover, all activated DP8α Treg clones but one also expressed CRTAM ( Figure 2J ), which together with EOMES, FASLG, GZMB and PRDM1 represent a cytotoxic CD4⁺ effector cell signature (11).

Further supporting the chemotactic and lytic potentials of DP8α Tregs, CCL5, CCR5, GZMA and GZMK were expressed by resting DP8α clones and down-regulated but still highly expressed upon activation ( Figure S7 ).

Regarding blood-derived F. prausnitzii-reactive DP8α Treg clones, most genes up-regulated by activation overlapped those up-regulated in colon-derived DP8α Treg clones ( Figure 2C ), although some of these alterations remained statistically unsignificant, likely due to the low number of studied clones. This result, together with the lack of DEG between blood- and colon-derived clones (not shown), strengthens the notion that DP8α cells found in both compartments are identical and originate from the gut.

We then compared DP8α to canonical FoxP3⁺ Treg clones derived from the same compartments, i.e., blood or colonic lamina propria, though DESeq2 analysis. Importantly, despite the limits mentioned above regarding the low number and heterogeneity of FoxP3⁺ clones, colon-derived DP8α and FoxP3⁺ Treg clones clearly diverged by PCA ( Figure 2A ) as well as by 286 (up: 108, down: 178) and 506 (up: 199, down: 307) DEGs at the resting and activated states, respectively ( Figures 3A, B ).

In contrast, blood-derived DP8α and FoxP3⁺ clones differed by a limited number of DEGs ( Figure S8 ). However, examination of unfiltered results revealed that most up-regulated DEGs identified in colon-derived DP8α versus FoxP3⁺ clones were also expressed at higher levels in DP8α, as compared to FoxP3⁺ blood-derived clones, as shown by log2 fold-changes > 1. Nonetheless, most of these differences remained statistically non-significant ( Figure S8 ), likely due to the low number and heterogeneity of both PBL-derived Treg clone subsets.

We next focused on DEGs identified between colon-derived DP8α and FoxP3⁺ Tregs. Top up-regulated DEGs identified in resting and activated DP8α versus FoxP3⁺ Treg clones (log2 fold-change ranging from 3 to 6.6) are represented ( Figures 3C, D ), respectively. As mentioned above, the expression of EOMES together with GZMK, GZMA, GZMB and CRTAM are hallmarks of cytotoxic CD4⁺ T cells (11, 33), while the co-expression of EOMES, GZMK and ITGA2 targets Tr1-like cells (31, 32). Other salient features of top DP8α Treg DEGs were a substantial prevalence of genes that 1/define a cytotoxic profile, such as FASLG, GZMA, GZMB, GZMK, GNLY, NKG7, SLAMF7, and SYTL2, 2/encode chemokines, including CCL1, CCL4, CCL4L1, CCL5 and XCL1, and 3/encode TFs reported to inhibit FoxP3⁺ expression or FoxP3⁺ Treg differentiation: EOMES (34), EPAS1 (35) and NFIL3 (36).

Altogether, these data showed that DP8α Tregs have a highly differentiated transcriptomic signature evocative of major regulatory functions via cytotoxicity as well as chemotaxis through both chemokine secretion facilitating their suppressive interaction with effector T cells and DCs, and robust expression of chemokine receptors, such as CCR1, and CCR5 (at steady state), favoring migration towards inflamed tissues.

Since clones were used to characterize the transcriptome of DP8α Tregs, the expression of molecules of interest as well as their in vitro function could be assessed using the exact same cells. First, based on the unique prevalence of cytotoxicity-associated genes in the transcriptome of DP8α Tregs and the significant overlap of the DP8α Treg signature with the Tr1 one, we measured the expression of such proteins by Treg clones, and their ability to kill cells of myeloid origin, which are typical targets of Tr1 cells (19, 37). In accordance with DEG data, DP8α Treg clones expressed EOMES protein, in contrast with FoxP3⁺ Treg clones ( Figure 4A ), as well as GZMA ( Figure 4B ), GZMB ( Figure 4C ), GZMK ( Figure 4D ) and Perforin ( Figure 4E ) at higher levels than FoxP3⁺ Treg clones did. Moreover, colon-derived DP8α Treg clones significantly expressed CRTAM shortly after activation (6h), which further increased at 16h post-activation, whereas colon-derived FoxP3⁺ Treg clones did not, regardless of the duration. At variance, CRTAM induction appeared delayed and was detected only at the 16h time-point in blood-derived DP8α clones. Moreover, CRTAM was also expressed by two blood-derived FoxP3⁺ Treg clones ( Figure 4F ), which, to our knowledge, has not been described so far. Similarly, in PBMCs, these molecules were all over-expressed by CCR6⁺/CXCR6⁺ DP8α Tregs, as compared to total CD4⁺ T cells or CD4⁺/CD25^HIGH/CD127^LOW classical FoxP3⁺ Tregs ( Figures 4G, H ).

To assess the ability of DP8α Tregs to kill myeloid cells, two cell lines, namely K562, of the erythroleukemia type, and U937, of the myeloid lineage, as well as two types of primary ex vivo sorted cells, CD19⁺ B cells and CD14⁺ monocytes, were co-cultured with Treg clones. No cell lysis was observed at 4h, timing usually required by conventional CD8⁺ T cells to exert cytotoxicity (not shown). In contrast, at 16h, potent lysis activity against cells of myeloid origin (U937 and monocytes) was induced by activated DP8α Tregs and, to a lesser extent, even non-stimulated DP8α Tregs, whether they were derived from blood or colon ( Figures 4I, J ). DP8α Treg clones neither significantly killed K562 nor sorted B cells ( Figures 4K, L ). No FoxP3⁺ Treg clone exerted any significant cytotoxicity ( Figures 4I-L ).

Therefore, in accordance with their expression of numerous molecules associated with a cytotoxic activity, DP8α Tregs unambiguously exert actual potent cytotoxic functions against myeloid cells, as reported for Tr1 cells (19, 37).

In mice, colonic IL-10-secreting Foxp3⁺ Tregs are thought to play a role in the maintenance of a healthy gut microbiota by regulating the production of microbiota-reactive secretory IgA (21).

Since the present study identified an array of genes involved in B cell help expressed by DP8α Tregs upon activation ( Figure 5A ), we asked whether blood-derived DP8α Tregs (ex vivo sorted and clones) could indeed provide B cell help and induce IgA production in vitro, as compared to both FoxP3⁺ Treg clones and sorted follicular helper T (Tfh) cells.

First, the expression of key proteins known to affect B cell activation and differentiation was analyzed. CD40L was strongly expressed by DP8α Treg clones, even at the steady state, and, to a lower extent, by ex vivo sorted DP8α Tregs, similar to Tfh cells used as a positive control, while Foxp3⁺ Treg clones barely did ( Figure 5B ). Production of several cytokines known to also influence B cell development and functions by both blood-derived clones and PBMCs was tested. Significant fractions of DP8α and FoxP3⁺ Treg clones did produce IL-2, IL-4, IL-5, IL-13 and IL-21 ( Figure 5C ) in addition to IL-10 ( Figures 1C, D ) and IFNγ ( Figures 1A, B ). Moreover, OKT3-stimulated PBMCs showed that DP8α Tregs produced higher levels of these cytokines than classical Tregs did ( Figure 5D ), demonstrating that DP8α Tregs are well-equipped to provide B cell help.

To directly determine whether DP8α Tregs can promote B cell activation and differentiation, clonal or ex vivo sorted DP8α cells were co-cultured with ex vivo sorted naïve B cells in the presence of an agonistic anti-BCR antibody with or without CD3/CD28 beads. As controls, ex vivo sorted Tfh cells and Foxp3⁺ Treg clones were identically run alongside. The gating strategy to sort these various T and B cell subsets is represented ( Figure 5E ). Ten days later, plasmablasts (CD3^-/CD4^-/CD19^LOW/CD38⁺) were detected when naïve B cells had been co-cultured with DP8α Tregs ( Figure 5F ). Plasmablasts represented a mean of 19.1% ± 0.21 and 14.1% ± 4.21 (ranging from 0.8 to 56.9%) of total CD3^-/CD4^-/CD19⁺ B cells after co-culture with activated sorted autologous DP8α Tregs and DP8α Treg clones, respectively ( Figure 5F ). It is noteworthy that DP8α Treg clones are heterogenous, i.e., while some induced high percentages of plasmablasts, other induced only low to no B cell differentiation. This plasmablast induction was comparable to that obtained in the presence of activated Tfh cells, while it was much lower (5.2% ± 1.39 of total B cells) when Foxp3⁺ Treg clones were used ( Figure 5F ).

In parallel, supernatants were tested for total IgA and IgG contents. Blood-derived DP8α Tregs, both clonal and sorted polyclonal cells, induced high levels of IgA production (1177ng/ml ± 219.3 and 990ng/ml ± 433.4, respectively), seemingly above levels induced by autologous Tfh cells (211ng/ml ± 210) ( Figure 5G ). DP8α Tregs also induced IgG production (90ng/ml ± 28.5 and 220ng/ml ± 71.8 for sorted and clonal cells, respectively) at similar levels than Tfh cells (164ng/ml ± 114.3) did ( Figure 5H ). Of note, Foxp3⁺ Tregs did not induce IgA nor IgG production ( Figures 5G, H ).

Therefore, DP8α Tregs do provide B cell help and seem prone to promote IgA class switch recombination.

Next, to formally establish the chemotactic potential of DP8α Tregs, we measured the protein expression of genes involved in this function, namely CCL4, CCL5, XCL1 and CCR5. In accordance with their differential transcription, all three chemokines were expressed at significantly higher levels by DP8α Treg clones, both colon- and blood-derived, as compared to FoxP3⁺ Treg clones derived from same compartments ( Figures 6A-C ), suggesting a significant ability of both circulating and colon-resident DP8α Tregs to recruit target cells. Strikingly, intracellular expression of CCL4 and CCL5 proteins by circulating DP8α Tregs was also detected ex vivo, clearly discriminating these cells from circulating FoxP3⁺ Tregs, at both resting and activated states ( Figure 6D ).

Also relevant in term of DP8α Treg regulatory potential, through recruitment to inflammatory sites (38), CCR5 expression levels were higher in resting colon- and blood-derived DP8α clones, as compared to matched FoxP3⁺ clones ( Figures 6E, F ), a difference confirmed ex vivo ( Figure 6G ).

Therefore, DP8α Tregs appear well-equipped to both recruit target cells and be recruited towards appropriate sites, especially when compared to classical FoxP3⁺ Tregs.

Finally, we assessed the protein expression by DP8α and FoxP3⁺ Treg cells of the RORγt TF, known to characterize microbiota-induced mouse Tregs (21, 22). In a previous study (6), we had reported the expression of RORγt by a few blood-derived DP8α Treg clones not studied here. Here, since this gene was not detected through the 3’seq RNA profiling method used, we assessed its expression by all studied Treg clones and their circulating counterparts. RORγt was expressed at the resting and activated states, by both blood- and colon-derived DP8α Treg clones, while it was not by FoxP3⁺ Treg clones ( Figure 7A ). Moreover, strikingly, this TF was also expressed by circulating DP8α Tregs among PBMCs, but not by other T cell subsets, such as total CD4⁺, classical CD25^HIGH/CD127^LOW Tregs, CD8⁺, as well as double positive CD4⁺/CD8⁺ T cells lacking CCR6 and CXCR6 co-expression, which do not respond to F. prausnitzii ( Figure 7B and not shown).

Therefore, RORγt represents yet another marker, which expression is preferably restricted to F. prausnitzii-induced DP8α Tregs.