CD4CD8αα Lymphocytes, A Novel Human Regulatory T Cell Subset Induced by Colonic Bacteria and Deficient in Patients with Inflammatory Bowel Disease

Gut bacterium Faecalibacterium prausnitzii activates a newly identified set of human IL-10-producing Treg cells (CD4CD8αα lymphocytes), revealing a mechanism by which commensal microbes contribute to host immunity.


Introduction
The gastrointestinal tract hosts a huge number of bacteria species. In mice, these bacteria play a major role in shaping local and systemic immune responses, notably by the induction of different effector and regulatory T cell subsets, whose adequate balance is required for the maintenance of gut homeostasis [1]. It may be postulated that this balance relies on the microbiota composition, as suggested by the observation that dysbiosis are frequent in chronic immune disorders, especially Inflammatory Bowel Disease (IBD) [2][3][4]. Nonetheless, data on how the microbiota composition may impact the disease process are still lacking and in particular the existence and potential role of microbiota-induced regulatory T cells (Treg) in humans remain to be addressed [5].
CD4 T cells that express the transcription factor fork head box p3 (Foxp3) are the best-known Treg. Some differentiate in the thymus in response to self-antigens and prevent self-reactive immune responses [6,7]. Others differentiate in the periphery under various conditions including chronic challenges by non-self antigens, such as commensal bacteria, and are strong contributors to tissue homeostasis [8]. In mice, a recent study highlighted that Clostridium bacteria are outstanding inducers of Foxp3 Treg in the colonic mucosa [9]. Additionally, some of these Treg expressing microbiota-specific T cell receptor (TCR) suggested that their induction involved the cognate recognition of bacterial antigens [10]. Foxp3 Treg are also present in the human gut mucosa, but their exact origin, distribution, and contribution to IBD prevention remain to be elucidated. Given that individuals with FOXP3 mutations do not always develop colitis and that intestinal inflammation is not associated with a decrease in the number of Foxp3 Treg, it has been postulated that non-Foxp3 Treg or other suppressive mechanisms may regulate colon immune homeostasis in humans [11,12].
Interleukin (IL)-10-secreting Foxp3 negative lymphocytes correspond to a heterogeneous population that remains poorly characterized, especially in humans [13,14]. In mice, several subsets of IL-10-secreting lymphocytes have been described in the small intestine [1], among which are double positive CD4CD8aa intraepithelial lymphocytes (IELs), which may prevent Th1induced intestinal inflammation, in an IL-10-dependent manner [15]. In contrast, these cells are absent in the lamina propria (LP) of the mouse colonic mucosa, where most IL-10-secreting lymphocytes express Foxp3 [16]. In humans, genetic studies pointed to an important role of IL-10 in the prevention of intestinal inflammation [17,18], but the abundance and distribution of IL-10-secreting non-Foxp3 and Foxp3 lymphocytes in the gut and the relative contribution of these cells to gut homeostasis remain poorly documented.
At least two subsets of double positive (DP) T lymphocytes, CD4CD8ab high and CD4CD8aa (CD8a low lacking CD8b) have been described in the human blood [19,20]. The latter subset has also been described in the intestinal mucosa of patients with and without IBD where it is relatively abundant compared with blood [21,22]. However, the functional significance of these DP subsets has remained unclear to date, and, in particular, a putative regulatory role of these cells has not been studied.
Here we showed that double positive CD4CD8aa (DP8a) T lymphocytes from the human colonic lamina propria (DP8a LPL) and blood (DP8a peripheral blood lymphocytes [PBL]) represent a yet undescribed subset of Foxp3-negative T lymphocytes that share all the regulatory functions of Foxp3 Treg and secrete IL-10. We also showed that these cells are decreased in the blood of patients with IBD, as compared with healthy donors, and in the colonic mucosa of patients with IBD compared with the healthy colonic mucosa of patients with colon cancer (CC). Importantly, we identified the induction of DP8a Treg by the gut commensal bacteria F. prausnitzii (F) as one of the mechanisms that may explain the impact of a balanced microbiota on the prevention or the control of IBD in particular, and on human health in general.

Double Positive CD4CD8aa T lymphocytes Are Frequently Found in the Human Colonic Lamina Propria
We analyzed the co-expression of CD4 and either CD8a or CD8b by T cells isolated from the epithelium or LP of healthy colonic mucosa from patients with CC. A significant fraction of CD3 LPL co-expressed CD4 and CD8a but not CD8b. The CD8a level expressed by these cells was variable and lower than that on CD8ab T cells (Figure 1a). CD4CD8aa LPL, hereafter referred to as DP8a LPL, were then quantified ( Figure 1b). They made up a mean of 8.5% (range 3.1-16.2) of CD3 LPL and 13.3% (range 5.9-24.8) of CD4 LPL. In the epithelium, smaller fractions of T cells co-expressed CD4 and CD8a (mean 2.4%, range 0. 7-5.6), and some of these cells expressed high levels of CD8a or CD8b ( Figure S2), likely corresponding to the CD4CD8ab colonic IEL subset that we previously described [23].
To assess the stability and polyclonality of DP8a LPL, LPL populations freshly dissociated from four donors were expanded by a single polyclonal stimulation. The populations obtained contained similar fractions of DP8a cells as the original LPL (unpublished data). The CD4 (CD8a neg) and the CD4CD8a cells were then FACS-sorted, to derive pure DP8a and CD4 LPL lines by polyclonal expansion. At any time during culture the phenotype of the sorted cell lines was unchanged (Figure 1c and unpublished data). We then asked the degree of polyclonality of DP8a LPL. Two of these LPL lines (C101 and C114), obtained by a single stimulation of freshly isolated LPL expressed, respectively, 11 and 21 Vb chains (out of the 25 tested). Moreover, DP8a LPL sorted from freshly dissociated LPL from one donor (C150) expressed 22 out of the 25 Vb analyzed (Table S1). Therefore, DP8a T cells represent in vivo an abundant and polyclonal subset of colonic LPL, distinct from the CD4 subset by a stable expression of CD8a.
To further characterize these cells, we studied their cytokine profile. The DP8a LPL lines and freshly dissociated DP8a LPL

Author Summary
It has become evident that bacteria in our gut affect health and disease, but less is known about how they do this. Recent studies in mice showed that gut Clostridium bacteria and their metabolites can activate regulatory T cells (Treg) that in turn mediate tolerance to signals that would ordinarily cause inflammation. In this study we identify a subset of human T lymphocytes, designated CD4CD8aa T cells that are present in the surface lining of the colon and in the blood. We demonstrate Treg activity and show these cells to be activated by microbiota; we identify F. prausnitzii, a core Clostridium strain of the human gut microbiota, as a major inducer of these Treg cells. Interestingly, there are fewer F. prausnitzii in individuals suffering from inflammatory bowel disease (IBD), and accordingly the CD4CD8aa T cells are decreased in the blood and gut of patients with IBD. We argue that CD4CD8aa colonic Treg probably help control or prevent IBD. These data open the road to new diagnostic and therapeutic strategies for the management of IBD and provide new tools to address the impact of the intestinal microbiota on the human immune system. Moreover, upon polyclonal activation, the DP8a LPL lines secreted tumour necrosis factor-alpha (TNF-a) and interferon gamma (IFN-c) (approximately 50% cells), but little if any IL-2 ( Figure 2d) and no IL-4, IL-5, IL-13, IL-17, or IL-22 (unpublished data). Ex vivo, the DP8a LPL exhibited the same cytokine profile as the DP8a cell lines ( Figure S3c). Importantly, as shown by ELISA ( Figure 2e) and quantitative PCR ( Figure S3d), activated DP8a LPL lines, but not their CD4 counterparts, secreted IL-10. Therefore, DP8a LPL exhibit a phenotype and a cytokine profile of Treg but lack Foxp3. We then addressed the regulatory potential of these cells in vitro.

Regulatory Properties of Human DP8a LPL
Similarly to Foxp3 Treg [24], the DP8a LPL lines inhibited the maturation of dendritic cells (DCs), as revealed by the inhibition of CD86, CD83 (Figure 3a), and CD80 (unpublished data) upregulation, in a CTLA-4-and LFA-1-dependent manner ( Figure 3a). The DP8a LPL lines also inhibited CD4 T cell proliferation induced by anti-CD3 and anti-CD28 antibody at all effector-target ratios used (Figure 3b and 3c), and this inhibition was partially blocked by an anti-IL-10 but not by an antitransforming growth factor (TGF)-bR-antibody (Figure 3d). In contrast, the CD4 LPL lines induced DC maturation ( Figure 3a) and failed to inhibit the proliferation of CD4 lymphocytes (Figures 3b and S4a). Notably, contrasting with Foxp3 and Tr1 Treg, the suppressive DP8a LPL proliferated upon CD3 activation in the absence of IL-2 addition, during the inhibition of CD4 T cell proliferation ( Figure S4b).
DP8a LPL Specifically Reacted with a Gut Commensal Bacteria: F. prausnitzii Bacteria belonging to the Clostridium cluster IV and XIV have been shown to induce Foxp3 Treg in the mouse colonic LP [9]. We therefore investigated if this could be the same for DP8a colonic LPL Treg. We first assessed the microbiota reactivity of DP8a LPL lines using four bacteria strains: F. prausnitzii (F), a major human gut bacterium of the Clostridium IV group, that is present at a decreased level in the faeces of patients with IBD and was shown to induce IL-10 expression by peripheral blood mononuclear cell (PBMC) [2,25], Bacteroides thetaiotaomicron (B), and Lactobacillus casei (L), which may promote Foxp3 Treg differentiation/expansion in mice [1], and Escherichia coli (E), a potential pathobiont. The DP8a LPL lines and their CD4 counterparts were incubated with the bacteria alone or with a mix of allogeneic monocytes (as monocytes from the patients with cancer were usually not available) previously loaded overnight with each bacterium. The LPL lines (DP8a and CD4) did not proliferate or secrete cytokines when incubated with the bacteria alone (unpublished data). In contrast, the DP8a lines responded to monocytes loaded with F but not or much less to monocytes alone and to monocytes loaded with B, L, or E bacteria, while the CD4 LPL lines did not respond specifically to F nor to the other bacteria. This was shown both by a proliferation assay (Figure 4a and 4b) and by intracellular IFN-c and IL-10 labelling (Figures 4c and S4c). We also observed in these experiments that DP8a LPL responses to F were suppressed by an anti-MHC class-II antibody but not by an irrelevant antibody (Figure 4a and 4c). This finding suggested that F recognition by DP8a LPL could be TCR dependent. However it could also be due to F-restricted superantigens. To further assess this important point, we compared the capacity of antibodies to CD4, CD8, MHC-I, MHC-II, HLA-DP, HLA-DQ, and HLA-DR to block the response of DP8a LPL to F presentation by a mix of three allogeneic monocytes and we analyzed the response of HLA class  (Table S2). As expected from TCR-dependent responses, cytokine responses of the LPL lines to F-loaded monocytes were totally inhibited by the anti-MHC-II and -CD4 but not by the anti-MHC-I and -CD8 antibodies and were inhibited at different levels by the anti-HLA-DP, -DQ, and -DR antibodies (Table 1). In addition, the LPL lines reacted to F presentation by the monocytes that shared at least one HLA class II allele but did not react with those that did not (with the exception of the response of the LPL line C192 to monocyte 5,  which may be due to a cross-presentation between two similar alleles such as DQ*0301 and *0303) (Tables 2 and S2). Moreover, not supporting the superantigen hypothesis, we observed that most Vb expressed by F-reactive DP8a LPL (identified by the colabelling of intracellular cytokines and of Vb in F-stimulated DP8a LPL) were also expressed by the CD4 LPL lines and that these cells did not respond to F-loaded monocytes (unpublished data). These results indicated that DP8a LPL recognized F antigens in the context of autologous HLA DP, DQ, or DR alleles and therefore in a TCR-dependent manner.
Recent reports have indicated that IL-10-secreting Foxp3 Treg could not be induced in mice by a single Clostridium species but only by a pool of Clostridium [9,26]. Although these inductions had not been shown so far to be antigen dependent, this finding raised the question whether F recognition by DP8a Treg was species specific or cluster specific. We assessed this using nine additional Clostridium strains, Suboligranulum variabile, Roseburia intestinalis, C. coccoides, C. leptum, Ruminococcus gnavus, C. ramosum, C. symbosium, C, boltea, and Anaerotruncus colihominis, which are abundant in human faeces. While the LPL lines (n = 3) systematically responded to F, no response to these bacteria was observed (Table 3). These results indicated that F-reactive DP8a LPL do not respond to a widely shared Clostridium antigen and thus may be F-specific.
We next investigated which proportion of ex vivo colonic DP8a LPL could be F-specific. To assess this, three freshly dissociated LPL populations were stimulated by a mix of F-loaded allogeneic monocytes, and pure populations of freshly sorted DP8a and CD4 LPL from one patient with colorectal cancer (C150) were stimulated by F-loaded autologous monocytes. We measured the proportions of DP8a cells, and of their CD4 counterparts, which divided in response to F-loaded but not to E-loaded monocytes nor to monocytes alone. A mean of 45% (range 20-80) of DP8a LPL divided specifically to F but not to E, while the CD4 LPL did not proliferate ( Figure 4d). It has to be stressed that the fraction of Fspecific cells was likely underestimated in three of these LPL populations as they were evaluated with allogeneic antigenpresenting cells (APCs) (Figure 4d black symbols). Interestingly, when autologous monocytes were used to present F to a pure population of DP8a LPL (Figure 4d, grey circle and right panels), 80% of these cells responded to F. We then assessed, with short-term cultured LPL lines C101 and C114 and with the freshly sorted DP8a LPL population (C150), the Vb diversity of F-reactive DP8a lymphocytes. Vb labelling showed that the F-reactive DP8a LPL expressed the majority of the Vb of the total DP8a populations, respectively, for C101, C114, and C150 LPL: 9/11 Vb and 1/3 Va, 16/21 Vb and no Va, and 23/23 Vb (Table S1). In sum, these data demonstrated that DP8a colonic LPL are highly polyclonal and, considering that F antigen presentation by allogeneic monocytes was suboptimal, they suggested that the TCR repertoire of these cells is highly skewed towards the recognition of F. prausnitzii.

DP8a Regulatory T Cells Are Present amongst PBL
Variable proportions and distinct subsets of double positive CD4CD8 T cells have been described in the blood of patients and healthy donors among which CD4CD8ab lymphocytes expressing high levels of CD4, CD8a and CD8b, and CD4CD8aa lymphocytes that expressed low levels of CD8a and no CD8b. Nonetheless, very little is known about the repertoire and immunological functions of these cells [19,27]. We confirmed that CD4CD8a low PBL lacked CD8b and they will hereafter be referred to as DP8a PBL. DP8a PBL represented a mean of 1.5% (range 0.1%-5.7%) of CD3 PBL and of 2.3% (range 0.1%-7.8%) of CD4 PBL in healthy donors (Figures 5a and S5). It had been reported that some DP PBL represent monoclonal T cell expansions [27]. Therefore we first assessed ex vivo the polyclonality of DP8a PBL from two donors, using 25 Vb antibodies. DP8a PBL expressed 24 out of 25 Vb, suggesting that they were polyclonal (Table S3). To determine whether these cells could be Treg originating from the colonic mucosa, we next assessed their expression of regulatory markers (CTLA4, CD25, GITR, and LAG3), and their reactivity to F. Ex vivo, most DP8a PBLs lacked regulatory markers (unpublished data). Nonetheless, a significant percentage of these cells proliferated in an MHC class-II dependent manner among PBMC cultured for 5 d with F (mean divided cells 7.54% range 1%-22%), but not, or at much lower levels, with B, L, and E ( Figure 5b). Compared with DP8a PBL, CD4 PBL yielded much lower proliferative responses to F ( Figure 5c) and did not respond to E (unpublished data). Importantly, a majority of F-reactive DP8a PBL and of those that did not respond to F in the proliferation assay expressed the regulatory markers CTLA-4 and LAG-3, within 5 days of coculture with F while their CD4 counterparts did not (Figure 5d). The fraction of DP8a PBL specific for F was re-analyzed using pure DP8a and CD4CD25 high CD127 low (Foxp3 + ) PBL populations sorted from PBMC from healthy individuals stimulated by autologous monocytes loaded with F or E or not loaded. As observed previously with total PBMC, the DP8a, but not the CD4 PBL, proliferated in a specific manner to F-loaded monocytes (mean 14.4% range 7.3-26.9) (Figure 5e). Therefore the mean percent of F-specific cells among DP8a PBL may be estimated to be above 15%.
This result raised the question of whether the F-specific DP8a PBL might be a DP subset distinct from the F-non-responding cells. To assess this, we asked whether F-reactive DP8a PBL might be distinguished from the non-responding ones by their Vb. We investigated this question using PBL lines enriched in DP8a cells (DP8) by a single sorting. In these cells lines the DP8a cells expressed 20 to 23 of the Vb tested (Table S3). We stimulated these cell lines by monocytes loaded with F and determined the Vb expressed by the F-reactive cells. As shown in Table S2, the Freactive DP8a cells (VPV low ) expressed the majority of the Vb tested: 17/17 and 12/17. These data indicated that the repertoire of F-specific DP8a PBL is diverse and not clearly distinct from that of the total DP8a PBL population.
To investigate the regulatory potential of DP8a PBL, we derived pure DP8a-and, as controls, pure CD4-PBL lines, by FACS-sorting followed by polyclonal expansion, i.e., independently of any stimulation by the bacteria. We also derived two DP8a cell lines and one CD4 cell line with similar sort but starting loaded overnight by F (1:5) in the presence or absence of anti MHC class-II antibody or of an irrelevant antibody, or not loaded, or loaded with B, L, or E: representative response of the C139 DP8a LPL line and the percentages of cells secreting IL-10 or IFN-c or both in independent experiments (two to seven experiments performed with four DP8a LPL lines: C114, black circles; C139, white circles; C192, white squares; and C140, black squares; ***p,0.001 (paired t-test). (d) Flow cytometry analysis of the proliferative response of freshly dissociated LPL from three donors stimulated by allogeneic monocytes (black symbols) and of sorted DP8a and CD4 LPL from one donor (grey symbols) stimulated by autologous monocytes. Responses were analyzed after 5 d of co-culture with monocytes alone or monocytes loaded overnight with F or E: mean percentage of F specific divided cells (with deduction of the divided cells to monocytes alone) (calculated on FlowJo software) and representative cytometry data of the proliferative response of the freshly sorted DP8a LPL population.  (Figure 6d and 6e) in contrast to their CD4 homologues (unpublished data). Amongst DP8a PBL lines, between 9% and 31% of cells secreted IFN-c and/or IL-10 in response to a mix of allogeneic monocytes (or autologous for one cell line) loaded with F but not to monocytes loaded with the other bacteria, in a MHC class-II-dependent ( Figure 6f) and MHC class-II isotype-dependent (unpublished data) manner, but did not secrete IL-2 or IL-4 (unpublished data). Importantly, in some experiments all the DP8a cells that responded to F by secreting IFN-c also secreted IL-10 ( Figure 6g), suggesting that the cytokine profile of these cells is homogeneous. The CD4 PBL lines did not respond to F by proliferation or cytokine secretion (unpublished data). Therefore, DP8a PBL appear to be polyclonal and functionally homogeneous populations of circulating Treg, distinct from their CD4 counterparts as they acquired regulatory markers and functions in culture, and as a significant fraction of them reacted to F in a TCR specific manner.

Decreased Frequency of DP8a LPL and PBL and Decreased F-Reactivity of DP8a PBL in Patients with Inflammatory Bowel Diseases Compared with Healthy Donors
As F is present at reduced levels in the gut microbiota of patients with IBD [2][3][4]28,29], we asked whether the frequencies of DP8a LPL and PBL were altered in patients with IBD compared with non-IBD donors (Tables S4 and S5 for characteristics of patients and healthy individuals). The mean frequency of DP8a lymphocytes amongst CD3 LPL was lower in the inflamed colonic mucosa of patients with IBD than in the healthy colonic mucosa of patients with CC (respectively, 4.8%, range 0.9-9.9 and 8.5%, range 3.1-16) (Figure 7a). Moreover, the mean fraction of DP8a lymphocytes amongst CD3 PBL was lower for patients with IBD than for healthy donors (respectively, 0.5%, range 0.1-1.3 and 1.5%, range 0.1-5.7) (Figure 7b). Separate analysis of the frequency of DP8a PBL in patients with Crohn disease (n = 20) and ulcerative colitis (UC) (n = 14) also indicated that in both diseases the frequency of these cells was lower than in healthy donors, respectively, mean fraction 0.4 and 0.6 range 0.1- 1.2 and 0.2-1.3 (Figure 7b).
The reduced level of F. prausnitzii in the gut microbiota of Crohn disease [2][3][4]28,29] and UC patients might affect the production of F-specific Treg. To assess this hypothesis, we compared the proliferative responses to F of DP8a PBL from healthy individuals and patients with IBD. DP8a PBL from healthy donors proliferated systematically upon PBMC co-culture with F ( Figure 7c), but not when cultured alone or with E (unpublished data). A similar proliferation was observed among the PBMC from nine out of 21 patients with IBD ( Figure 7c). However, in the remaining 12 patients, DP8a PBL did not proliferate or proliferated only non-specifically, i.e., as much to E and/or in the absence of bacteria as to F. This lack of specific response was observed in particular with the PBMC from the five patients with UC included in this assay (unpublished data). As a result, the mean specific response to F of DP8a PBL from patients with IBD was lower than in healthy donors: respectively, mean percent divided cells among DP8a PBL, 7.5 and 2.4, ranges 1-22 and 0-14 ( Figure 7c). Because patients with Crohn disease with ileal involvement (L1 and L3 stages of the Montreal classification) exhibit the clearest deficit in F in the gut microbiota, we assessed whether remission in these patients was associated with a restored frequency and/or response of DP8a PBL to F. DP8a PBL from L1 and L3 patients in remission (n = 4) proliferated specifically to F (mean percent divided cells: 6.4 range 2-12.7). Such a response was significantly less frequent in patients with active disease (two out of ten), mean percent F-specific divided cells 1.3 range 0-7 ( Figure 7d). However, no significant difference in the frequency of DP8a circulating T cells was observed between the patients in remission and those in flair (unpublished data).
Finally, we could assess the proliferative response to F and E of LPL freshly isolated from the inflamed mucosa of two patients with UC and the IL-10/IFN-c response of an LPL line obtained from the inflamed mucosa of another patient with UC. Similarly to what we had observed with the PBL from many patients, freshly isolated DP8a LPL (Figure 7e) and the LPL line (unpublished data) failed to respond to F in contrast to freshly sorted LPL populations ( Figure 7e) and to LPL lines from healthy controls (Figure 4a).

Discussion
Herein we demonstrated that DP8a T cells, present in the human colonic LP and blood, represent a novel subset of T cells lacking Foxp3 but sharing with Foxp3 Treg most other regulatory markers and in vitro regulatory functions, and with Tr1 Treg the capacity to secrete IL-10. We also showed that a great part of these cells specifically recognized F. prausnitzii, a Clostridium cluster IV bacterium, that is a major component of the indigenous gut microbiota frequently decreased in patients with IBD [25]. In addition, we showed that DP8a cells are decreased in the blood and colonic mucosa of patients with IBD compared with healthy individuals and the healthy colonic mucosa of patients with CC, respectively. Together with the recent findings that bacteria of the Clostridium IV and XIV clusters induced the development of IL-10-secreting Foxp3 Treg in the colonic mucosa of germ-free mice and that some of these cells were specific for antigens of the inducing clostridia [9] [26], our results argue that DP8a T cells are Treg similarly induced in the human colonic mucosa through cognate activation by F. prausnitzii antigens. Thus, our data highlight a homology between mice and humans regarding colonic Treg induction by Clostridium bacteria. We also observed the presence of low fractions of DP8a lymphocytes in the epithelium of the colonic mucosa. However, as the phenotype and functions of these IEL were not characterized further, it remains unclear whether they correspond to the same Treg subset as DP8a LPL or to a colonic mucosa-specific IEL subset.
In mice, two subsets of CD4CD8aa T lymphocytes have been described in the epithelium of the small intestine [15,30,31]. One of these shared with DP8a LPL the capacity to secrete IL-10 and could prevent Th1-induced intestinal inflammation in an IL-10 dependent manner [15]. Interestingly studies with IL-10 reporter mice indirectly showed that this subset is absent from the mouse colonic LP, as far as all the IL-10-secreting lymphocytes of this compartment expressed Foxp3 [16]. This suggests that human DP8a and mouse Foxp3 LPL, both induced by Clostridium bacteria in the colonic mucosa, may be functional and developmental equivalents, in which case the distinct phenotype of these peripherally induced Treg populations represents a species-related divergence.
Whether Foxp3 Treg, induced by the microbiota, coexist with DP8a Treg in the human colonic mucosa is unknown. We observed that Foxp3 + CD25 high CD127 low T cells represented a much lower fraction of colonic LPL than DP8a Treg and did not react with F (unpublished data).
Another subset of CD4CD8a IEL of the mouse small intestine has been characterized recently as cytotoxic T lymphocytes. Interestingly the expression of the co-receptor CD8aa and of the cytotoxic function by CD4 IEL was shown to be associated with down-regulation of ThPOK and up-regulation of Runx 3 transcription factors [30,31]. In contrast, the human DP8a LPL expressed ThPOK and Runx 3 at same levels as their CD4 counterparts and lacked perforin and CD107a expression (unpublished data).
Since, the induction of mouse colonic Foxp3 Treg required a mix of at least three Clostridia strains [9,26], we asked whether nine Clostridia strains that are common in human faeces might be involved in the induction of DP8a Treg. Interestingly, four of these belonged to a mix of the 17 human Clostridial strains that enhanced Foxp3 Treg abundance in the mouse colonic mucosa [26]. None of these Clostridium strains was recognized by DP8a Treg. This showed that the target antigens of colonic DP8a Treg are not shared Clostridium antigens but rather are species specific. This further suggested that F is a major inducer of DP8a colonic Treg. However, it does not exclude that other bacteria including the Clostridium that induced Foxp3 Treg development in GF mice [26], could play a role in this induction, independently of TCR triggering. Besides cognate activation, DP8a Treg differentiation likely requires additional signals. TGF-b, vitamin A-derived retinoic acid, and tolerogenic DCs are Treg inducers constitutively present within the gut mucosa [11]. Interestingly, the production of these factors, within the mouse colon, has been shown to be stimulated by Clostridi [9].
Further supporting a prevalent role of F in the induction of colonic DP8a LPL, we observed that 80% of the freshly purified DP8a LPL from one donor reacted to F antigens (Figure 4d).
Alterations in the composition of the gut microbiota and especially reduced levels of F are characteristics of patients with IBD [2][3][4]28]. Additionally, in ileal Crohn disease, F reduction correlates with an increased risk of post-operative recurrence, suggesting that disruption of the F level is involved in the pathogenesis [2]. Importantly, the frequency of DP8a PBL and, among these, the frequency of the F-specific cells were decreased  in patients with IBD compared with healthy donors. In addition, DP8a lymphocytes were less frequent in the inflamed colonic mucosa of patients with IBD compared with healthy colonic LP of patients with CC. These observations may support a role of DP8a Treg in the control of IBD. However, the decreased frequency of DP8a LPL in the colonic mucosa of patients with IBD remains to be established by comparison with healthy donors. Indeed the use of individuals with CC as a control group has two main caveats. One is the great age disparity between the patients with IBD and with CC (Table S5). The other is the potential alteration of DP8a colonic LPL frequencies, which may be associated in these patients with gut microbiota dysbiosis especially F decrease [32]. The decreased frequency of DP8a PBL in patients with IBD together with the reported decrease of F in the faeces of these patients suggests that there is a connection between reduced levels of F and decreased F-specific regulatory activity, potentially resulting in increased inflammation. However, a formal assessment needs to be done of the correlation between the levels of F and the levels of DP8a PBL and LPL in patients with IBD. Moreover, the nature of the potential connection between F and DP8a Treg levels remains unclear. F decrease might limit the differentiation or survival of DP8a Treg. Conversely, a defect in DP8a Treg differentiation or function might contribute to dysbiosis, as suggested in mice when the differentiation of Foxp3 Treg was blocked [33].
DP8a LPL and PBL, at variance with their CD4 counterparts, express high levels of the regulatory markers CD25, CTLA-4, GITR, and LAG-3, secrete IL-10, and mediate robust inhibition of CD4 T cell proliferation and of DC maturation. Therefore, these cells appear adequately fitted to inhibit effector T cell responses under both inflammatory and non-inflammatory conditions [6,24]. These cells also expressed IFN-c and TNF-a and Th-associated transcription factors. Similar expressions have recently been observed in mouse and human Foxp3 Treg and appear to represent phenotypic and functional specializations of these cells, paralleling those of Th lymphocytes during inflammatory responses [34][35][36][37]. Importantly, a number of studies have established that IFN-c produced by Foxp3 Treg has essential immune regulatory functions [38].
Although DP8a Treg and Tr1 cells secrete IL-10, several phenotypic differences between them indicate that they represent distinct Treg populations. Among these is the expression by DP8a Treg of CD8a, high levels of CD25 and Gata-3, ex vivo and as cell lines, and the lack of expression by these cells of PD1 (unpublished data), considered to be a canonical marker of Tr1 cells by Roncarolo and colleagues [39]. In addition, both Treg seem to use different suppressive mechanisms as Tr1 suppression strongly depends on IL-10 and TGF-b activity, while these molecules, respectively, had a limited and no contribution to the suppressive activity of DP8a Treg (Figure 3d). Nonetheless, IL-10, the main inducer of Tr1 [13,14], might also contribute to DP8a Treg induction as F is strong inducer of IL-10 secretion by peripheral mononuclear cells [2].
No clear function could be ascribed so far to CD4CD8 PBL. We showed that among these the DP8a subset lacked Treg markers ex vivo, but acquired such markers and Treg functions in culture and that a fraction of them reacted specifically to F. The fraction of Freactive DP8a PBL seems to be lower than that of DP8a LPL. Indeed, at most 26.9% (mean 14.4%) of DP8a PBL reacted to F in an autologous antigen presentation context. As between 20% and 40% of T cells are usually non-responsive to their antigen in vitro (as observed with human clones), it may be estimated that no more than 40% of ex vivo DP8a PBL may be specific for F. Therefore, the remaining DP8a PBL should have distinct antigen specificities. Whether these cells are specific for other gut bacteria or for commensal strains present in other tissues is an interesting question. Alternatively, as DP8a PBL lacked Treg markers ex vivo, an important question was whether the fraction of these cells not-reactive to F could be effector lymphocytes expanded in response to infections, as suggested by others [20]. The observation that the majority of DP8a PBL expressed CTLA-4 and LAG-3 independently of their specificity for F, after a few days of culture with F, while very few of their CD4 counterparts expressed these markers (Figure 5d) argues that most DP8a PBL are Treg. It is likely that the circulating DP8a lymphocytes specific for F had been induced in the gut mucosa. The capacity of pTreg induced in the gut mucosa to recirculate in the blood was documented in mice by the increased frequencies of IL-10secreting Treg outside the gut following Clostridium reconstitution of GF mice [9].
The frequency of DP8a PBL was significantly reduced in the blood of patients with Crohn disease and UC. In addition, in a majority of these patients DP8a PBL failed to react to F, suggesting that the decrease in circulating DP8a PBL may specifically concern the F-specific ones. Alternatively, the defective response to F might be due to immunomodulatory treatments. However, this explanation appears unlikely because this defect was not observed in some of the patients who underwent the same treatments. It will be important to investigate if the frequency of circulating F-specific DP8aTreg is correlated with the level of F and to address its prognostic and diagnostic value.
The physiological significance of circulating microbiota-reactive Treg will be important to address also in non-IBD contexts. In mice, increased resistance to experimental colitis and allergy models was observed in Clostridium-abundant animals. This suggests that the proportion of gut Clostridium may affect both colonic homeostasis and systemic immune responses and lead to conclude that Clostridium-induced Treg mediate these effects [9]. Likewise, the circulating F-reactive Treg identified in our study may play a role in systemic immunity. The quantification of DP8a PBL in diseases associated with microbiota imbalance should provide an answer to this question. Figure 5. Presence of F-reactive DP8a T lymphocytes in the blood of healthy individuals. (a) Flow cytometry analysis of the frequencies of DP8a T lymphocytes in the blood of 18 donors: representative dot plot of PBMC co-labelled with anti-CD3, anti-CD4, and anti-CD8a antibodies, representative histogram of CD8b expression by gated CD8, CD4, and DP8a PBMC, and the percentages of DP8a lymphocytes among the CD3 and the CD3CD4 PBMC. (b) Proliferative responses of DP8a peripheral blood T cells to F in the presence or absence of an anti-MHC class-II antibody or an irrelevant antibody, and to B, L, and E. PBMC were cultured for 5 d with the antibody and/or indicated bacteria at a bacterium:PBMC ratio 1:1: representative FlowJo analysis of the VPD dilution and percentage of F-specific divided DP8a T cells in the PBMC from several donors (n = 18); ***p, 0.001 and **p,0.01 (paired t-test). (c) Percent VPD dilution in paired DP8a and CD4 T cells among PBMC co-cultured with F for 5 d; ***p,0.001 (paired t-test). (d) CTLA4 and LAG3 expression in proliferative DP8a and CD4 PBL after 5 d of co-culture with allogeneic monocytes loaded overnight with F (n = 2). (e) Flow cytometry analysis of the proliferative response of freshly isolated DP8a or CD4 PBL after 5 d of co-culture with autologous monocytes alone or loaded overnight with F or E: representative cytometry data of a freshly sorted DP8a PBL population and mean percentage of Fspecific divided cells (calculated on FlowJo software) (experiments performed with five freshly isolated DP8a PBL (black circles) and two CD4 PBL (white circles). doi:10.1371/journal.pbio.1001833.g005  Importantly, the high in vitro proliferative capacity of DP8a cell lines, together with the high stability of their regulatory functions, are critical properties towards the use of these cells for adoptive transfer purposes.
In conclusion, we have identified an unanticipated new subset of Treg induced by F. prausnitzii that may contribute to the suppression of deleterious inflammation in humans. Importantly, these Treg exhibit a stable phenotype that allows their quantification. These results may provide new diagnostic and therapeutic strategies for IBD and systemic diseases that are associated with abnormal induction and/or function of DP8a Treg. These results should also lead to a much broader understanding of the impact of the microbiota on the human immune system than the one we currently have.

Cell Isolation and Cell Line Generation
PBMC were obtained from patients with IBD and healthy individuals (see Table S4 for characteristics of patients with IBD and healthy donors). Normal and inflamed colonic mucosa were obtained, respectively, from patients undergoing surgery for CC who did not undergo radiotherapy or chemotherapy, and from patients undergoing surgery for IBD (Crohn disease or UC) (see Table S5 for characteristics of patients with IBD and CC). Monocytes and PBL were obtained from healthy donor blood by elutriation (DTC platform, CHU, Nantes, France). Normal colonic mucosa was obtained from surgically resected tissue, taken approximately 10 cm downstream of the tumour. For normal mucosa, the LP was separated from the epithelium after incubation in EDTA buffer (30 min) and then minced into 1mm 2 fragments and washed with RPMI containing penicillin (10%) and gentamycin (0.1 mg/ml; Sigma-Aldrich). Tissue fragments were digested with collagenase and DNAse (2 mg/ml each; Sigma-Aldrich), with shaking at 37uC. Mucus and large debris were removed by filtration through a 40-mm-cell strainer (BD). Viable cells were obtained by Ficoll gradient centrifugation.
This study was approved by the ethics committee of the Comité de Protection des Personnes Ile-de-France IV (Suivithèque). All the patients signed informed consent forms. For cell line generation, CD3CD4CD8a neg and CD3CD4CD8a low LPL and PBL were isolated, by sorting on a FACS-Aria (Becton Dickinson). T cell lines were generated by stimulations with PHA, irradiated feeder cells, and IL-2, as described [40]. For suppression assays of T cell proliferation, CD4 T cells were isolated from PBMC using magnetic beads (130-045-101; Miltenyi). Immature DC were obtained from monocytes cultured for 5 d at 2610 6 cells/ml with 80 ng/ml IL-4 and 90 ng/ml GM-CSF (AbCys) in RPMI 1640 supplemented with L-glutamine, penicillin-streptomycin (10 mg/ ml) (GIBCO) and 10% SVF (PAA). DC maturation was induced by co-culture with activated CD4 lymphocyte or CD4 cell lines expressing the CD40L.

Suppressive Assays: Inhibition of CD4 Proliferation and DC Maturation
Freshly sorted CD4 PBL (5610 4 ) were incubated with 5 mM CFSE (Invitrogen) in PBS containing 0.1% BSA for 15 min, washed, and then stimulated with anti-CD3/anti-CD28 activation beads (Miltenyi) at a 1:1 ratio, in the presence or absence of DP8a or CD4 LPL lines at the indicated effector:target (E:T) ratios and in the presence or absence of anti-IL-10 or anti-TGF-b antibodies. The proliferation of target CD4 T cells was assessed by flow cytometry analysis of CFSE dilution among CD8negative T cells, on day 5. CD40L expressing CD4 LPL lines were cultured with immature DC (1:1 ratio) for 2 to 3 d (to induce DC maturation) in the presence or absence of DP8a cell lines and of anti-CTLA-4 or anti-LFA-1 antibodies. The cells were stained with APC-conjugated anti-CD3 (555335) and PEconjugated anti-CD80 (IM2729U) or PE-conjugated anti-CD83 (IM2218U), or PE-conjugated anti-CD86 (IM1976U) or isotype control antibodies. CD3 negative cells were analyzed by flow cytometry to determine the level of expression of CD80, CD83, and CD86.

Intracellular Cytokine Assay
Lymphocytes were incubated 6 h with 0.1 mg/ml plate-bound CD3 antibody (OKT3 eBioscience) or, at a 1:1 ratio, with bacteria or monocytes (when autologous monocytes were not available a mixture of monocytes from four donors were used to permit the sharing of HLA class-II alleles between the monocytes and the responding cells. Indeed, monocytes from a single donor that lacked any MHC overlap failed to induce a response to F, as shown on Table 2), previously incubated overnight with the different bacteria at a 5:1 ratio. To prevent cytokine secretion, 10 mg/ml brefeldin A (Sigma-Aldrich) was added for the last 6 h of stimulation. Stimulated cells were stained with PerCP-conjugated anti-CD3, FITC-conjugated anti-CD4, and APC-conjugated anti-CD8a. The cells were then fixed for 10 min in PBS/4% paraformaldehyde (Sigma-Aldrich) and washed. Cytokine-specific antibodies were then added for 30 min at room temperature. Reagent dilutions and washes were performed with PBS containing 0.1% BSA and 0.1% saponin (Sigma-Aldrich). Cytokine secretion was assessed by flow cytometry in DP8a or CD4-positive LPL and cell lines. In some experiments TCR Vb and intracellular labelling were analyzed among cytokine-labelled DP8a cell lines. The following antibodies were used: PE-conjugated anti-IL-2 (559334), anti-IL-4 (554486), anti-IL-5 (554395), anti-IL-10 (562400), anti-TNF-a (554418), and anti-IL-22 (515303) and APC-conjugated anti-IL-13 (554571), anti-IFN-c (554551), and anti-IL-17 (51717871). We also used in some experiments anti-CD3 Brilliant Violet (562426), anti-IFN-c APC (554702) combined with the anti-IL-10 PE. For blocking experiments, we used anti HLA class II ascites (clone 206 produced in our laboratory) and an irrelevant mouse IgG.
The levels of IL-10 in the supernatants were measured by ELISA (R&D Systems).

T Lymphocyte Proliferation Assays to Bacteria
Lymphocytes (PBMC, LPL, PBL, or cell lines) were labelled for 15 min incubation at 37uC in the dark with 1 mM VPD (BD Bioscience) in PBS containing 0.1% BSA. The cells were washed twice in medium containing 10% FBS. F, B, L, and E were sonicated for 15 min at high speed and then co-cultured with VPD labelled PBMC at 1:1 ratio, or with monocytes overnight (at a ratio 5:1) in presence of gentamycin (0.1 mg/ml). When autologous monocytes were not available a mix of allogeneic monocytes from three to four donors was used to present the bacteria. Monocytes, loaded with bacteria or left unloaded were washed and mixed with VPD-labelled cells (1 to 1.5610 5 ) at a 1:5 ratio. After 3 to 5 d, the proliferation of T cells was assessed by flow cytometry analysis of the VPD dilution in CD3CD4CD8a low and CD3CD4CD8a neg PBMC or in CD4-or in CD8a positive cells for T cell lines and the fraction of divided cells was determined by FlowJo analysis of the VPD dilution. The HLA class II dependency was analyzed by adding a specific antibody every 48 h or an irrelevant mouse IgG.

Quantitative RT-PCR Analysis
Briefly, after cell lysis using the Trizol reagent (Life technologies), total RNA were extracted using the RNeasy Micro kit (Qiagen) and reverse transcribed using the Superscript II reverse transcriptase (Life technologies). PCR amplification was performed with an amount of cDNA corresponding to 50 ng of total RNA. For qPCR, amplification was done using iQ SYBR Green Supermix (Bio-Rad) and specific gene expression was calculated using the 2DDCT method (using GAPDH as calibrator). The primer sequences used are available on request.

Statistical Analysis
Statistical analysis was performed with the GraphPad Prism version 5.0 (GraphPad software). Paired and unpaired t-tests and the Mann-Whitney test were used, as indicated in the Figures 1-7. Differences were considered significant at p,0.05. Table S1 Vb diversity of total DP8a and F-reactive DP8a-cells among LPL lines and ex vivo LPL. Cytometry analysis of Vb expressed by DP8a cells among LPL lines (n = 2 C101, C114) stimulated# or not* by F-loaded monocytes and among freshly purified LPL (C150) stimulated## or not** by Floaded monocytes. A gate was done on DP8a cells. (XLSX) Table S3 Vb diversity of total DP8a and F-reactive DP8acells among PBMC and PBL lines. Cytometry analysis of Vb expressed by DP8a cells among PBMC* (n = 2 DTC4, DTC28) and among VPD-labelled cell lines derived from these PBMC (by a single sorting and polyclonal expansion of DP8a cells) stimulated*** or not** by F-loaded monocytes. A gate was done on DP8a cells. (XLSX)