Induction of stable human FOXP3+ Tregs by a parasite‐derived TGF‐β mimic

Abstract Immune homeostasis in the intestine is tightly controlled by FOXP3+ regulatory T cells (Tregs), defects of which are linked to the development of chronic conditions, such as inflammatory bowel disease (IBD). As a mechanism of immune evasion, several species of intestinal parasites boost Treg activity. The parasite Heligmosomoides polygyrus is known to secrete a molecule (Hp‐TGM) that mimics the ability of TGF‐β to induce FOXP3 expression in CD4+ T cells. The study aimed to investigate whether Hp‐TGM could induce human FOXP3+ Tregs as a potential therapeutic approach for inflammatory diseases. CD4+ T cells from healthy volunteers were expanded in the presence of Hp‐TGM or TGF‐β. Treg induction was measured by flow cytometric detection of FOXP3 and other Treg markers, such as CD25 and CTLA‐4. Epigenetic changes were detected using ChIP‐Seq and pyrosequencing of FOXP3. Treg phenotype stability was assessed following inflammatory cytokine challenge and Treg function was evaluated by cellular co‐culture suppression assays and cytometric bead arrays for secreted cytokines. Hp‐TGM efficiently induced FOXP3 expression (> 60%), in addition to CD25 and CTLA‐4, and caused epigenetic modification of the FOXP3 locus to a greater extent than TGF‐β. Hp‐TGM‐induced Tregs had superior suppressive function compared with TGF‐β‐induced Tregs, and retained their phenotype following exposure to inflammatory cytokines. Furthermore, Hp‐TGM induced a Treg‐like phenotype in in vivo differentiated Th1 and Th17 cells, indicating its potential to re‐program memory cells to enhance immune tolerance. These data indicate Hp‐TGM has potential to be used to generate stable human FOXP3+ Tregs to treat IBD and other inflammatory diseases.


INTRODUCTION
Heligmosomoides polygyrus is a parasitic nematode of mice that establishes long-term infections in the intestinal tract and releases a diverse array of excretory/secretory products (HES), which promotes parasite survival. Like most parasites, H. polygyrus has evolved several immune evasion strategies 1 including those that inhibit inflammatory pathways and promote immunosuppressive cell populations. 2,3 For example, HES contains molecules that block the cytokine IL-33 and its receptor ST2, 4,5 impair dendritic cell function 6 and suppress macrophages. 7 Recently, another HES protein was identified as a TGF-b mimic, termed Hp-TGM (H. polygyrus TGF-b mimic), which, despite the lack of sequence homology, triggered mammalian TGF-b signaling pathways through binding TGF-b receptors. 8 Importantly, Hp-TGM induced FOXP3 expression in both mouse and human CD4 + T cells, with mouse induced FOXP3 + regulatory T cells (Tregs) shown to have suppressive function both in vitro 8 and in vivo. 9 FOXP3 + Tregs have a key role in maintaining selftolerance and immune homeostasis and in vivo are generated in both the thymus and the periphery. 10 Peripheral induction of Tregs in the gut is important to ensure appropriate immune tolerance towards self, commensal and dietary antigens, a setting in which TGFb plays a critical role. 11,12 TGF-b signaling is tightly regulated, being first produced in a latent form that includes a latency-associated peptide (LAP), which prevents receptor binding. 13 One mechanism of generating active TGF-b is cleavage via the integrin avb8, which is expressed by dendritic cells in the gut; active TGF-b is then able to initiate Treg induction through binding to the heterodimeric surface receptors, TbRI/ TbRII, on CD4 + T cells. 14 Receptor binding initiates phosphorylation of SMAD2/3 proteins, which form a complex with SMAD4 that binds to regions in the promoter of FOXP3 and drive its expression. FOXP3 is the key master transcription factor of Tregs 14 as it orchestrates the expression of other Treg markers, resulting in the acquisition of immunosuppressive functions. 15 In contrast to mammalian TGF-b, Hp-TGM is secreted in an active form and does not require posttranslational processing to activate TGF-b receptormediated signaling. 16 The role of TGF-b in inducing Tregs has been known for over 15 years 17,18 and much work has been done to investigate whether it could be used to generate large numbers of Tregs in vitro for a cell therapy to induce immune tolerance. 15,19 However, thus far these approaches have been limited in their ability to generate Tregs with phenotypic stability. Previous studies that generated human Tregs in vitro (iTregs) with TGF-b, and with or without combinations of all-trans retinoic acid and/or rapamycin, showed that FOXP3 expression and suppressive function were unstable. Specifically, upon restimulation in the absence of polarizing conditions iTregs lost these features, likely due to the fact that the iTregs did not undergo the epigenetic re-modeling needed for stable FOXP3 expression. 20,21 Here we investigated whether Hp-TGM could induce human FOXP3 + Tregs and if the induced cells had superior function or stability compared with TGF-binduced Tregs. This knowledge is important to assess the therapeutic potential of this novel parasite-derived protein.

RESULTS
Hp-TGM induces FOXP3 + cells with a Treg phenotype from human na€ ıve CD4 + T cells To test the ability of Hp-TGM vs. TGF-b to induce Tregs, na€ ıve CD4 + T cells were isolated from peripheral blood (Supplementary figure 1a) and stimulated with artificial antigen-presenting cells expressing CD80 and CD58 and loaded with anti-CD3 in the absence or presence of the indicated cytokine. The majority of published protocols for TGF-b-induction of human iTregs from na€ ıve CD4 + T cells use between 1 and 5 ng mL À1 of TGF-b. [21][22][23] We tested TGF-b at concentrations from 1 to 100 ng mL À1 and confirmed that maximal FOXP3 expression (as determined by both the percentage of positive cells and MFI) had occurred by 1 ng mL À1 (dose curve data in Figure 1). This equates to 0.04 nM, as activated TGF-b exists as a 25 kDa dimer, thus 1 ng mL À1 TGF-b was selected as our comparison condition for the study. We found that after 6 days, in the presence of Hp-TGM or TGF-b, > 50% of CD4 + T cells expressed FOXP3, compared with a mean of 21% expression in their absence (Figure 1a). At least 10 ng mL À1 (0.21 nM) of the 46.8 kDa Hp-TGM was required to induce levels of FOXP3 that were equivalent to 1 ng mL À1 (0.04 nM) TGF-b (i.e.~5.25 times more than TGF-b) (Figure 1b). Analysis of the dose titration results obtained from all donors, and also in experiments that used anti-CD3/CD28 bead stimulation (Supplementary figure 1 and data not shown), led us to select 100 ng mL À1 Hp-TGM as the concentration that most consistently induced maximal FOXP3 expression and this concentration was used for the remainder of the study.
Although there was no significant difference between TGF-b and Hp-TGM in terms of the induced proportion of FOXP3-expressing cells (Figure 1c), there was a small, but significant, increase in FOXP3 MFI in Hp-TGM cultures (Figure 1d). A similar effect was seen for the induction of two Treg markers CD25 (IL-2Ra) and the co-inhibitory receptor CTLA-4, for which MFIs were significantly higher in Hp-TGM cultures compared to those with TGF-b (Figure 1d). The proportion of induced FOXP3 + cells remained consistent upon repeated TCR stimulation and re-addition of Hp-TGM or TGF-b (Figure 1e), and in Hp-TGM cultures it could be further enhanced by the addition of rapamycin, the mammalian target of rapamycin (mTOR) inhibitor that suppresses conventional T-cell proliferation 24 (Figure 1f). Similar effects were seen when na€ ıve CD4 + T cells were stimulated with anti-CD3/CD28 beads, although under these conditions the overall proportions of induced  Hp-TGM signals via TbRI and SMAD2/3 pathways in na€ ıve CD4 + T cells We next asked whether, as seen in mice, 8 Hp-TGM signaling required the TGF-b receptor subunit I (TbRI, the serine/threonine kinase ALK5), and stimulated phosphorylation of SMAD2/3 in human cells. PBMCs were cultured in serum-free media (no exogenous TGFb) with either TGF-b or Hp-TGM and the levels of pSMAD2/3 in CD4 + T cells were measured by flow cytometry. TGF-b rapidly induced pSMAD2/3 after 15 min; stimulation with 100 ng mL À1 of Hp-TGM required 30 min to observe a similar effect. However, increasing the concentration of Hp-TGM from 100 to 500 ng mL À1 enabled a more rapid activation of pSMAD2/3 ( Figure 2a). We confirmed that both Hp-TGM and TGF-b-mediated FOXP3-induction in na€ ıve CD4 + T cells required TbRI signaling as inhibition of this receptor with the ALK5 kinase inhibitor SB-431542 reduced the proportions of FOXP3 + cells in both Hp-TGM and TGF-b conditions by approximately 60% (Figure 2c, d). TbRI inhibition also reduced CD25 and CTLA-4 expression, with both TGF-b and TGM cultures having reduced MFIs of approximately 30% (for CD25) and 20% (for CTLA-4) compared with cultures without inhibitor added (Figure 2c, d).
Hp-TGM induces epigenetic changes at the FOXP3 locus in na€ ıve CD4 + T cells As we had observed that Hp-TGM-induced FOXP3 expression was maintained in culture, we next asked if Hp-TGM caused epigenetic changes in the Treg-specific demethylated region (TSDR) of the FOXP3 locus. Reduced methylation in this intronic region is associated with stable FOXP3 expression. 25 After 28-32 days in culture, Hp-TGM-but not TGF-b-induced Tregs had significantly less methylation in the FOXP3 TSDR compared with na€ ıve T cells expanded with only IL-2 ( Figure 3a). We also calculated the percentage loss of methylation for each sample from matched control cultures (100 À (TSDR methylation in Hp-TGM culture/ TSDR methylation in control 9 100)), which was greater in Hp-TGM cultures than TGF-b cultures (P = 0.0625; Figure 3a).
To further explore possible epigenetic effects of Hp-TGM, we carried out ChIP-Seq analysis to measure changes in H3K27-acetylation (ac) on histones, with higher levels being indicative of more active/accessible genes. 26 Overall, the chromatin modifications detected in TGF-b-vs. Hp-TGM-induced Tregs were similar, although we did identify genes with different H3K27ac patterns (Figure 3b Hp-TGM-induced Tregs are functionally suppressive Next, we investigated whether Hp-TGM-induced Tregs acquired suppressive activity. We found that, compared with TGF-b-induced Tregs, Hp-TGM-induced Tregs were significantly better at suppressing allogeneic, polyclonally activated CD4 + and CD8 + T-cell proliferation over a range of cell ratios (Figure 4a, b). Another defining feature of FOXP3 + Tregs is their low expression of effector T-cell cytokines. 27 We confirmed that, similar to TGF-b-induced Tregs, Hp-TGM-induced Tregs had significantly reduced secretion of IFN-c and IL-2, compared with control culture cells as detected by intracellular cytokine staining ( Figure 4c) and cytokine secretion ( Figure 4d). Analysis of the cytokines secreted into culture supernatants also identified both Hp-TGMand TGF-b-induced Tregs had significantly reduced secretion of IL-4, IL-13 and TNF compared with control cultures (Figure 4d).

Hp-TGM-induced FOXP3 + Tregs are stable in the presence of inflammatory cytokines
A concern regarding the use of in vitro-induced Treg cells for therapy is reversion to an effector T-cell phenotype in vivo. 19 To investigate the stability of Hp-TGM-induced Tregs, we re-cultured induced Tregs in the absence of TGF-b or Hp-TGM and, in some cultures, in the presence of inflammatory cytokines (Figure 5a). Changes in cell phenotype were compared with populations that were expanded in the continual presence of TGF-b or Hp-TGM (baseline condition), which were defined as having 100% stability and used to calculate the percentage of stable cells in matched comparison cultures. The proportion of Hp-TGM-induced Tregs that remained FOXP3 + after Hp-TGM removal was significantly greater than for TGF-b-induced Tregs when TGF-b was removed, with a similar result for CTLA-4 expression   ( Figure 5b). Moreover, when the induced Tregs were recultured without Hp-TGM/TGF-b, but with the inflammatory cytokines IL-6, TNF and IL-1b, Hp-TGMinduced Tregs showed significantly greater stability of FOXP3 and CTLA-4 expression than did TGF-b-induced Tregs (Figure 5c). Importantly, following exposure to inflammatory cytokines, TGF-b-but not Hp-TGMinduced Tregs increased production of IFN-c and IL-2 compared with cultures with continual presence of TGFb/Hp-TGM and no inflammatory cytokines (Figure 5d).
Hp-TGM can induce stable FOXP3 + Tregs from pre-committed memory Th cells As a potential therapeutic application of Hp-TGM is its direct administration in vivo, we investigated whether it may be able to convert pre-committed memory Th subsets to a regulatory T-cell phenotype. We sorted ex vivo Th1, Th2 and Th17 cells using the gating strategy in Supplementary figure 3a and expanded these cells by stimulating with anti-CD3/CD28 beads in the absence or presence of TGF-b or Hp-TGM (Figure 6a). Hp-TGMinduced significantly more Th1 and Th17 cells to express FOXP3 after 7 days in culture compared with TGF-b (Figure 6b), although both were similarly unable to induce FOXP3 expression in Th2 cells (Supplementary  figure 3b). However, in contrast to what was found with na€ ıve CD4 + T-cell cultures, neither TGF-b nor Hp-TGMconditioned Th1 or Th17 cells had substantial upregulation of CD25 or CTLA-4 compared to control cultures (Supplementary figure 3c). Interestingly, Hp-TGM, but not TGF-b, resulted in a small, yet significant, reduction in the percentage of IFN-c + Th1 cells and the amount of secreted IFN-c. Nevertheless, neither Hp-TGM nor TGF-b affected IL-17A expression by Th17 cells (Figure 6c and Supplementary figure 3d). Both TGF-b and Hp-TGM-conditioned Th1 and Th17 cells acquired in vitro suppressive function, with Hp-TGM-conditioned Th1 cells being slightly more suppressive than their TGFb counterparts over a range of cell ratios (P = 0.0065; Figure 6d). Finally, we assessed the stability of the induced Tregs from Th1 and Th17 cell cultures. Intriguingly, we again noted that the Hp-TGM-induced FOXP3 + cells had enhanced stability compared with TGFb-induced FOXP3 + cells in the presence of inflammatory cytokines in vitro, and this was significant for the Th17 cells (P = 0.0073, Figure 6e).

DISCUSSION
We show here that a novel parasite-derived TGF-b mimic protein (Hp-TGM) is able to induce FOXP3 + Tregs from both na€ ıve and memory human CD4 + T cells in vitro. The FOXP3 + Tregs induced from na€ ıve CD4 + T cells also had high expression of the Treg markers CD25 and CTLA-4, did not secrete effector cytokines and were able to suppress T-cell proliferation. Most striking was our finding that, in comparison with TGF-b, Hp-TGMinduced Tregs had superior stability in vitro in the presence of inflammatory cytokines. These data suggest that Hp-TGM not only phenocopies the ability of TGF-b to induce Tregs but that these Hp-TGM-induced Tregs may be a more stable cell product with potential therapeutic benefits.
Hp-TGM was similar to TGF-b in its ability to induce FOXP3 expression in na€ ıve CD4 + T cells; however, comparable levels of FOXP3 induction required at least 5 times higher molar concentrations of Hp-TGM compared with TGF-b, the opposite of what was previously shown for murine T cells. 7 The source of Hp-TGM, H. polygyrus, is a mouse parasite and these data may reflect hostparasite co-evolution and a specialized adaptation of Hp-TGM to the mouse immune system. As murine and human TGF-b receptors share 96% amino acid identity, the discordant concentration of Hp-TGM required for its effects on mouse and human T cells is unlikely to be due to differing affinities for the TGF-b receptors. However, we have shown that a truncated form of Hp-TGM, which retains the receptor binding domains but lacks the Cterminal portion of the protein, also requires higher concentrations to be effective on mouse cells, 16 implying that there may be an additional, mouse-specific, interaction that enhances the efficacy of Hp-TGM.
Although the percentages of FOXP3 + cells induced by the two ligands were similar, Hp-TGM-induced Tregs expressed higher amounts of FOXP3 protein, as indicated by MFI. In a parallel study, 9 we studied the kinetics of Hp-TGM signaling in mouse T cells and found it had a more sustained effect compared with TGF-b. This may be due to TGF-b being subjected to intricate regulation, which Hp-TGM may be able to circumvent, resulting in more prolonged signals and increased amounts of FOXP3. Importantly, we also showed that, compared to TGF-b, Hp-TGM could induce significantly more expression of CTLA-4, a co-inhibitory receptor with an important functional role in Tregs. 28 The in vitro effects of Hp-TGM could also be achieved with anti-CD3/CD28 bead-based stimulation, with FOXP3 + cell purity further enhanced by rapamycin, suggesting that Hp-TGM could be adapted for induction of a human Treg cell therapy product. Since not all cells were converted to Tregs, more work is needed to understand the biological differences between na€ ıve CD4 + T cells that are, or are not, susceptible to the effects of Hp-TGM.  It is remarkable that, despite sharing no sequence homology with mammalian TGF-b, Hp-TGM is able to signal through the human TGF-b receptor complex. Of note, one key difference is that Hp-TGM is able to directly bind to TbRI, while TGF-b must first form a complex with TbRII, 8 although the functional consequences of this difference are not yet known. We and others have shown that H. polygyrus infection induces host TGF-b production, 29,30 which may augment the effects of Hp-TGM in vivo. However, as Hp-TGM does not require processing for biological activity and is not subject to feedback controls that might limit the effects of TGF-b it is likely to have a more prolonged and potent effect in vivo.
We confirmed that Hp-TGM is able to drive SMAD2/3 phosphorylation in primary human cells, although it required either a longer incubation time or a higher concentration to achieve the same induction of FOXP3 as did TGF-b. While the mature form of TGF-b consists of a single domain, Hp-TGM consists of five domains and, in mice, only the first three domains were required for TGFb receptor binding and FOXP3 induction. 16 A recent study of Hp-TGM binding to human receptors identified that domain 3 competes for the same binding sites as TGF-b on TBRII, while domains 1 and 2 are required for optimal binding to TBRI, with the role of domains 4 and 5 in full-length Hp-TGM remaining unclear. 31 It is important for future studies to understand which domains are required for Hp-TGM signaling in human cells, to solve the crystal structure of Hp-TGM and TbRI/ II, and determine if Hp-TGM is able to bind other receptors. These data will provide clues as to how TGF-b signaling pathways could be targeted therapeutically.
Supporting the finding that Hp-TGM induced higher FOXP3 MFI was the fact that Hp-TGM induced a greater loss of methylation in the FOXP3 TSDR compared with TGF-b-induced Tregs, although neither TGF-b nor Hp-TGM achieved more than a 50% loss of methylation in na€ ıve CD4 + T cells. While it is accepted that most thymically derived Tregs have a fully demethylated TSDR, it is still unclear whether this is the case for peripherally derived Tregs and evidence to date suggests that Tregs induced in vitro with TGF-b do not acquire this signature. 20,21 Interestingly, a recent study in mice reported that TGF-b and IL-2 in vitro-induced Tregs acquired a demethylated TSDR in the absence of CD28 signaling. 32 Future studies should investigate the epigenetic and functional effects on Tregs induced with Hp-TGM without CD28 co-stimulation. However, as recent studies of iTregs generated by TGF-b and all-trans retinoic acid and rapamycin 22 ; or TGF-b and small molecule inhibitors 23 ; or lentiviral transfer of FOXP3 33 ; have found that the iTregs had robust and stable suppressive function without acquiring a methylated TSDR, this feature may not be essential for suppressive function of iTregs. Nevertheless, our data showing increased H3K27ac marks in the FOXP3 locus in Hp-TGM-induced Tregs suggest that this molecule may drive epigenetic changes that promote Treg stability 34 and warrant further investigation.
Importantly, Hp-TGM-induced FOXP3 + Tregs acquired robust in vitro suppressive function, which was statistically significantly superior to that of TGF-binduced Tregs. These cells also adopted the typical Treg characteristic of minimal production of effector cytokines, particularly IL-2, which is a key distinguishing feature from non-Tregs. Therefore, the Hp-TGM-induced Tregs fit the criteria proposed by Yamaguchi et al. for bona fide Tregs, being high, constitutive expression of FOXP3, CD25 and CTLA-4 and no IL-2 expression. 28 To more rigorously assess the stability of the Hp-TGMinduced Tregs we challenged these cells by removing Hp-TGM and adding the inflammatory cytokines IL-6, TNF and IL-1b. In this setting of inflammatory challenge, we saw a marked difference between Hp-TGM-and TGF-binduced Tregs, with the former having significantly greater stability of FOXP3 and CTLA-4 expression without secretion of IFN-c or IL-2.
We also showed that Hp-TGM is able to similarly induce stable and suppressive FOXP3 + cells from memory Th1 and Th17 cells. Interestingly, Hp-TGM-treated Th1 cells had reduced IFN-c secretion, whereas in parallel assays Hp-TGM-treated Th17 cells did not have altered IL-17A expression, indicating that Hp-TGM may be best suited for modulating the activity of Th1 cells. There is much interest in therapies that can convert pathogenic effector T cells into Tregs, and these data suggest that Hp-TGM may be a suitable candidate for such an in vivo approach.
In conclusion, we extend upon earlier work describing this novel parasite-derived TGF-b mimic, Hp-TGM, showing that it is also able to induce populations of suppressive FOXP3 + Tregs in vivo from both na€ ıve and memory human CD4 + T cells via TbRI and pSMAD2/3 signalling. These cells had superior stability compared to TGF-b-induced Tregs in an inflammatory milieu. Our data identify Hp-TGM as a potentially useful therapeutic molecule for the treatment of inflammatory disease, such as inflammatory bowel disease.   (Supplementary figure 3a).

Reagents
Recombinant full length Hp-TGM was produced as described previously 8 and used at 100 ng mL À1 unless otherwise stated. Recombinant mammalian TGF-b1 (R&D Systems Inc. Minneapolis, MN, USA; catalogue number 240-B) was used at 1 ng mL À1 unless otherwise stated. Recombinant human IL-2 (Proleukin; Prometheus Laboratories, San Diego, CA, USA) was added to Treg-induction cultures at 100 U mL À1 , and in cultures of ex vivo Tregs at 1000 U mL À1 . The TbRI (ALK5) inhibitor SB-431542, which also inhibits ALK4 and ALK7, was used at 5 lM (Tocris Bioscience, Bristol, UK).

In vitro induction of FOXP3 + Tregs
Sorted cells were expanded with anti-CD3/CD28 Dynabeads at specified ratios (Invitrogen, Carlsbad, CA, USA) or irradiated, modified L-cells at a 1:1 ratio. The modified L cells used were a mouse fibroblast line virally transfected with the human adhesion molecule CD58 (to bind CD2 on target cells and to stabilize the cell-cell interaction), Fc receptor CD32 (to bind the Fc receptor of the CD3 monoclonal antibody in culture to provide T-cell receptor stimulation), and the costimulatory ligand CD80. 35 L-cells were irradiated 50 Gy and 20 000 cells were plated in flat-bottom 96-well plates and 20 000 sorted CD4 + T cells added. Cytokines were added at specified concentrations and the cells incubated at 37°C (5% CO 2 ). T cells were restimulated every 14 days. The culture media for all experiments was X-VIVO 15 (Lonza, Basel, Switzerland) supplemented with 5% heat-inactivated human serum (NorthBio Inc, Toronto, ON, Canada), 1% Glutamax and 1% Penicillin-Streptomycin (Invitrogen). Cell proliferation was measured by staining with 5 lM cell proliferation dye eF450 (Invitrogen). The inflammatory cytokine challenge was the addition of 10 ng mL À1 each of recombinant IL-1b, IL-6 (STEMCELL Technologies Inc., #78034; #78148) and TNF (eBioscience, #14-8329-63).
In some experiments, 100 ng mL À1 rapamycin was added (Sigma-Aldrich, St Louis, CA, USA). . The cells were incubated in this media and either left unstimulated or 5 ng mL À1 TGF-b or 100-500 ng mL À1 Hp-TGM were added for 15 or 30 min at 37°C and all cells were mixed by pipetting (1 9 10 5 cells in 500 lL). An equal volume of pre-warmed (to 37°C) Cytofix (BD) was added to the cells and incubated at 37°C for 10 min before adding 100 lL of ice-cold phosphate buffered saline (Thermo Fisher Scientific Inc) and then stained with pSMAD2/3(O72-670)-AF647 (BD).

Suppression assays
Suppression assays were performed as described previously 36 using heterologous PBMCs as responder cells and measuring the suppression of anti-CD3/CD28 bead stimulated cell proliferation.

Cytokine analysis
Cell supernatants were collected following 5 h stimulation with 10 ng mL À1 phorbol 12-myristate 13-acetate (PMA) and 500 ng mL À1 ionomycin in the presence of 10 lg mL À1 Brefeldin A (Sigma-Aldrich). The concentrations of secreted cytokines in the cell supernatants were measured using the LEGENDplex TM 13-plex Th cytokine bead array kit according to manufacturer's directions (BioLegend). CBA data were acquired on a 3-laser Cytoflex cytometer (Beckman Coulter) and analyzed using FCAP array v3 software (BD).

Treg-specific demethylation region
Methylation levels within the Treg-Specific Demethylation Region (TDSR) of FOXP3 was assessed by pyrosequencing as described previously. 37

ChIP-Seq
Na€ ıve CD4 + T cells were expanded with modified L-cells (as described above) for 22 days with either TGF-b, or Hp-TGM (controls had neither added), then viable CD4 + T cells were isolated from cultures by cell sorting for ChIP-seq analysis. The cells were fixed with 1% paraformaldehyde (ThermoFisher Scientific Inc), incubated for 10 min, then a 1/10 volume of 1.5 M glycine (Sigma-Aldrich) was added for 5 min while rotating. The cells were centrifuged then washed twice with 1 mL ice cold PBS and the cell pellets flash frozen in LN 2 and stored at À80°C. ChIP-Seq was performed as previously, 38 anti-H3K27Ac polyclonal antibody ab4729 was from Abcam (Cambridge, UK). The obtained sequences were mapped to hg19 using Bowtie 2 (v2.1.1). 39 Uniquely mapped reads were selected and then duplicated reads were discarded. Peak calls were performed using SICER (v1.1.1) 40 with the threshold of FDR less than 0.000000001. For further analyses and visualization, ChIPPeakAnno (v3.5) 41 and Integrative Genome Viewer (v2.3.91) 42 were used. These data have been deposited in NCBI's Gene Expression Omnibus 43 and are accessible through GEO Series accession number GSE164548. (https:// www.ncbi.nlm.nih.gov/geo/query/acc.cgi).

Statistics
Statistical analyses between two groups used the Mann-Whitney U-test or, for paired samples, the Wilcoxon signed rank test. Analysis of ≥ 3 groups used Kruskal-Wallis one-way ANOVA or, for paired samples, a Friedman one-way ANOVA with Dunn's multiple comparison post-test. Correlation analyses were calculated with Spearman's rho (r). P-values were considered significant when < 0.05. Prism v8 (GraphPad Software Inc., San Diego, CA, USA) was used for all statistical analyses. Error bars represent median AE interquartile range; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ***P ≤ 0.0001 and ns = not significant.