A Central Role for Induced Regulatory T Cells in Tolerance Induction in Experimental Colitis

Abstract

In addition to thymus-derived or natural T regulatory (nT_reg) cells, a second subset of induced T regulatory (iT_reg) cells arises de novo from conventional CD4⁺ T cells in the periphery. The function of iT_reg cells in tolerance was examined in a CD45RB^highCD4⁺ T cell transfer model of colitis. In situ-generated iT_reg cells were similar to nT_reg cells in their capacity to suppress T cell proliferation in vitro and their absence in vivo accelerated bowel disease. Treatment with nT_reg cells resolved the colitis, but only when iT_reg cells were also present. Although iT_reg cells required Foxp3 for suppressive activity and phenotypic stability, their gene expression profile was distinct from the established nT_reg “genetic signature,” indicative of developmental and possibly mechanistic differences. These results identified a functional role for iT_reg cells in vivo and demonstrated that both iT_reg and nT_reg cells can act in concert to maintain tolerance.

Mature CD4⁺ “conventional” T (T_conv)⁴ cells isolated from peripheral blood and lymphoid organs can be induced to express Foxp3 in vitro by T cell activation in the presence of TGF-β1 and IL-2 (1–4). This process is enhanced by retinoic acid and antagonized by IL-6 and IFN-γ. Foxp3⁺ “induced” regulatory T (iT_reg) cells have suppressive function, as measured by their capacity to inhibit T cell proliferation in vitro and to inhibit the induction of experimental autoimmune disease. TGF-β1 also acts with IL-6 to induce a T cell-specific isoform of the retinoic acid-related orphan receptor-γ (RORγt), which is necessary for the development of proinflammatory Th17 cells. Foxp3 and RORγt can be expressed in the same cell population and therefore lie at a critical crossroad between inflammation and tolerance.

In TGF-β1 and conditional T cell-specific TGF-β receptor II knockout mice, a lethal lymphoproliferative autoimmune disease, similar to that seen in Foxp3 knockout mice, develops within the neonatal period (5–9). Although “natural” regulatory T (nT_reg) cell development in the thymus appears normal, survival in the periphery is compromised (7–9). Furthermore, TGF-β deficiency may abrogate the conversion of T_conv cells into iT_reg cells. These observations demonstrate an important role for TGF-β in maintaining the regulatory T (T_reg) cell pool, but do not functionally discriminate between the two types of T_reg cells.

Recent studies have examined the in situ generation of iT_reg cells under a variety of different conditions. Chronic exposure to systemically delivered submitogenic amounts of agonistic peptides or T cell expansion in a lymphopenic environment results in Foxp3 expression in up to 15% of Ag-experienced T_conv cells. These cells have the capacity to suppress T cell proliferation (10–13). However, studies in diabetic NOD mice with a high frequency of self-reactive T cells showed little overlap in the T_conv and nT_reg cell TCR pools, suggesting peripheral conversion plays a minimal role in tolerance (14). Peripheral conversion was also not observed after an acute viral infection and only modestly following immunization (15, 16). Arguably, the published data support the interpretation that chronic Ag exposure creates conditions that favor iT_reg cell generation and survival, while during primary responses iT_reg cells seem to contribute marginally to T_reg cell control. Nevertheless, these studies left unresolved the genetic and functional relationships between iT_reg and nT_reg cells, as well as the biological relevance of the conversion process itself.

In this study, we used mice that express enhanced GFP (EGFP) under the control of the endogenous Foxp3 promoter with either a functional or a nonfunctional Foxp3 protein (Foxp3^EGFP or Foxp3^ΔEGFP mice, respectively) to examine in situ conversion and the contribution of iT_reg cells to tolerance. As a test of function, we compared the capacity of iT_reg and nT_reg cells to modify autoimmune inflammation in a lymphopenia-induced model of inflammatory bowel disease. We also compared the gene expression profiles of iT_reg and nT_reg cells. We found a synergistic role for iT_reg cells in the maintenance of dominant immunological tolerance that depended upon Foxp3 to stabilize differentiation induced by T cell activation and TGF-β1.

Materials and Methods

Mice

Foxp3^EGFP and Foxp3^ΔEGFP mice on the BALB/c background were generated and screened as previously described (16–18). Thy1.1⁺ Foxp3^ΔEGFP newborn mice were rescued by i.p. transfer of 60 × 10⁶ unfractionated Thy1.2⁺ BALB/c splenocytes to generate naive Thy1.1⁺CD4⁺ CD45RB^high T cells with the nonfunctional Foxp3^ΔEGFP allele. Rag1^−/− mice were obtained from The Jackson Laboratory. The Animal Resource Committees at the Medical College of Wisconsin and at the University of California, Los Angeles, approved all animal experiments, and moribund mice were sacrificed and analyzed per these protocols.

Cell purification and adoptive transfer

Pooled splenocytes and lymph node cells were stained with either anti-CD4-PE or anti-CD4-Pacific Blue and anti-CD45RB-allophycocyanin as appropriate and sorted on the basis of Ab and EGFP fluorescence. All sorting was done on a FACSAria (BD Biosciences). The average purity and viability of the sorted CD4⁺ populations was 98.8 ± 0.2% and 87.6 ± 1.6%, respectively (SEM, n = 23). Colitis was induced in 6-wk-old BALB/c mice by i.p. injection of 4 × 10⁵ CD4⁺EGFP⁻CD45RB^high cells suspended in 1 ml of PBS. Following this injection, mice were weighed twice weekly. In some experiments, when mice had lost 5–7% of their initial body weight, they were treated by i.p. injection of T_reg cells purified by cell sorting.

Analytical flow cytometry

Cells were stained as described previously (18) and analyzed using an LSRII (BD Biosciences) and FlowJo software (Tree Star).

TGF-β1-mediated in vitro conversion

Sorted EGFP⁻ splenocytes from Foxp3^EGFP or Foxp^ΔEGFP mice (1 × 10⁶/ml) were cultured in anti-CD3 mAb (clone 14-2C11 at 2.5 μg/ml)-coated dishes in the presence of soluble anti-CD28 mAb (clone 37.51) at 1 μg/ml with or without TGF-β1 (5 ng/ml; R&D Systems). After 72 h, cells were either analyzed by flow cytometry or resorted based upon EGFP fluorescence.

Suppression assay

BALB/c splenocytes containing 2.5 × 10⁵ responder CD4⁺ T cells were cultured in 96-well flat-bottom plates in the presence of 1 μg/ml anti-CD3 mAb. The number of responder CD4⁺ cells was kept constant, while the number of suppressor T_reg cells was titrated to achieve the indicated ratios in triplicate wells. Cultures were maintained for 48 h, then pulsed with 0.4 μCi/well [³H]TdR for an additional 18 h, harvested, and counted.

Histology

Complete colons were fixed in formalin, processed, and stained with H&E using a histology core facility. Blinded sections from the entire colon were examined by a pathologist (N.H.S.) and large intestine colitis scores were determined for the following inflammatory changes on a 4-point semiquantitative scale with 0 representing no change (19). The following features were considered: severity, depth and chronic nature of the inflammatory infiltrate, crypt abscess formation, granulomatous inflammation, epithelial cell hyperplasia, mucin depletion, ulceration, and crypt loss.

Statistics

To compare groups with respect to the colitis score, nonparametric methods were used because the data were skewed and ordinal. A Kruskal-Wallis test overall was used with a Mann-Whitney U test for pairwise comparison. Since there were ties, the results were checked for consistency using a Savage test or ordered data or a Cochran-Armitage Savage trend test, whichever was relevant. Lymphocyte numbers and lymphocyte percentages were compared by a two-tailed Student’s t test. To compare percent weight change over time, a random coefficient regression model was used. In the first 10 days, there was little difference between the groups. Analysis was done after day 10. Where the data were nonlinear, a polynomial was fit. The random coefficient model essentially fits each mouse separately and calculates an aggregate weighing with respect to fit of the data to the hypothesized model. Post hoc pairwise comparisons were done using a contrast. No adjustment was made for multiple comparisons. Kaplan-Meier survival curves were compared using the log-rank test.

Gene expression profiling

Purified cells sorted by flow cytometry from 4 to 10 mice were used to generate total RNA, which was amplified and used to create probes for Affymetrix 430 2.0 chips. The subset of probe sets whose expression increased or decreased by 2-fold or more was identified and used for further analysis. The data were normalized using the justRMA algorithm from the Bioconductor group (20). Results represent mean fold change values derived from three to five independent arrays for each cell type and scored p < 0.05 by Student’s unpaired two-tailed t test. Selected targets were verified by flow cytometry. It should be noted that because the gene array probes for Foxp3 lie distal to the poly(A) start site in the Foxp3^EGFP allele, a false-negative result is generated for Foxp3 in Foxp3^EGFP T cells. The following cell populations were analyzed: freshly isolated and in vitro-activated (anti-CD3 mAb plus IL-2) CD4⁺EGFP⁺ (nT_reg) cells and CD4⁺EGFP⁻ (T_conv) cells from Foxp3^EGFP mice, iT_reg cells from Foxp3^EGFP and Foxp3^EGFP/+ mice, and ΔiTreg cells from Foxp3^ΔEGFP/+ mice. The microarray data have been deposited in the Gene Expression Omnibus repository (http://www.ncbi.nlm.nih.gov/geo/), accession number GSE14415.

Results

In situ-generated iT_reg cells in experimental colitis

To investigate the relative potency of iT_reg cells, the capacity of iT_reg cells to suppress autoimmune inflammation was examined in a lymphopenic model of inflammatory bowel disease (21). Colitis was induced in Rag1^−/− recipients by the i.p. transfer of 4 × 10⁵ CD4⁺EGFP⁻ Thy1.2⁺CD45RB^high syngeneic T_conv cells isolated by cell sorting from Foxp3^EGFP or Foxp3^ΔEGFP male mice. In the recipients of naive T_conv cells from Foxp3^EGFP mice, diarrhea and weight loss began ~30 days after transfer. Disease progressed until mice became moribund and were sacrificed ~100 days after transfer (Fig. 1, A and B). Colitis was marked by extensive leukocytic infiltration of the mucosa, submucosa, muscularis and serosal layers, absence of mucin-secreting goblet cells, epithelial cell hyperplasia with increased crypt length, epithelial cell ulceration, and crypt abscess formation (Fig. 1C). The mean colitis score from 12 randomly selected moribund mice was 3.8 ± 0.1 (Fig. 1D). Analysis of lymphocytes in the mesenteric lymph nodes 30 days after transfer revealed that 2.4 ± 0.2% of CD4⁺ T_conv cells had become EGFP⁺, reflecting the in situ generation of iT_reg cells (n = 5; data not shown). This proportion increased to 9 ± 1.3% of CD4⁺ T cells at the conclusion of the experiment (Fig. 1E). A similar proportion of T_reg cells was seen in the mesenteric lymph nodes of normal Foxp3^EGFP mice, although the total number of EGFP⁺ cells was reduced in mice with colitis by ~50% (Fig. 1F). In situ-generated iT_reg cells and nT_reg cells from healthy mice had similar suppressive function in vitro, as measured by their capacity to reduce Ab-mediated T cell proliferation. A 50:50 mixture of the two cell types did not increase the suppressive effect over that mediated by iT_reg cells alone (Fig. 1G). In situ-generated iT_reg cells in the mesenteric lymph nodes of mice with colitis could be distinguished from mesenteric lymph node nT_reg cells from normal mice on the basis of several cell surface markers including CD25, CD62, and CD103 (Fig. 1H).

FIGURE 1.

Induction of colitis in the absence of functional Foxp3. A, Weight change following the adoptive transfer of 4 × 10⁵ CD4⁺EGFP⁻Thy1.2⁺CD45RB^high T cells isolated from Foxp3^EGFP (dashed red lines; n = 47) and Foxp3^ΔEGFP (dashed black lines; n = 20) mice into Rag^−/− recipients. Control Rag^−/− littermates (gray lines; n = 11) did not receive transferred cells. Color-matched linear regression lines are plotted for each data set. B, Kaplan-Meier survival curves for the mice in A. C, Representative histological sections from the colons of randomly selected mice in A stained with H&E. D, Scatter plot showing the colitis score for each mouse where histology was obtained. E, Representative flow cytometry analysis of the mesenteric lymph nodes from each group (n = 7, 14, and 17, left to right). Numbers denote means for the adjacent gate. F, The total number of T_reg cells found in the mesenteric lymph nodes of healthy Foxp3^EGFP mice (green) and in mice with colitis with either functional (red) or nonfunctional (black) in situ-derived iT_reg cells. G, Representative data from in vitro suppression assays (n = 3) using sorted in situ-derived iT_reg cells (red), sorted nT_reg cells from healthy mice (green), or a 50:50 mix of each population (blue) to inhibit the anti-TCR-induced proliferation of unfractionated splenocytes. The ratio of T_reg cells to CD4⁺ responder splenocytes is indicated. H, Representative cell surface marker analysis of CD4⁺EGFP⁺ mesenteric lymph node T_reg cells from E above stained as indicated.

To determine the extent to which in situ-generated iT_reg cells might influence disease progression, we induced colitis with 4 × 10⁵ EGFP⁻Thy1.2⁺CD45RB^high syngeneic T_conv cells isolated by cell sorting from Foxp3^ΔEGFP mice. For these experiments, Foxp3^ΔEGFP mice were rescued at birth with 60 × 10⁶ unfractionated Thy1.1⁺ splenocytes to allow normal development of the host Thy1.2⁺CD4⁺ T_conv cells (our unpublished data). In the absence of a functional Foxp3 locus, weight loss began ~25 days after transfer and was accelerated relative to recipients of CD45RB^high cells from Foxp3^EGFP mice (Fig. 1A, p < 0.001). All mice became moribund and were sacrificed 50 days after initiation of the experiment, demonstrating a significant decrease in survival (Fig. 1B, p < 0.001). Colitis was severe, with a mean colitis score of 3.9 ± 0.1 (Fig. 1, C and D). Flow cytometry revealed a significant decrease in the percent and number of EGFP⁺ cells (Fig. 1, E and F). In situ-derived ΔiT_reg cells generated from Foxp3^ΔEGFP T_conv cells had no suppressive function in vitro (data not shown), although their cell surface phenotype was similar to the in situ-derived iT_reg cells generated from Foxp3^EGFP T_conv cells (Fig. 1H). Thus, Foxp3 is required to sustain iT_reg cell numbers and to develop iT_reg suppressive function.

Treating established colitis by T_reg cell adoptive transfer immunotherapy

In the preceding experiments, the iT_reg cells generated in situ from a starting pool of 4 × 10⁵ CD45RB^high T_conv cells delayed disease progression and prolonged survival, but were not sufficient to prevent severe colitis. Cotransfer of either iT_reg or nT_reg cells with the CD45RB^high T_conv cells can prevent disease from developing (22–24). In a setting that must include in situ-generated iT_reg cells (Fig. 1), established disease can also be treated by the adoptive transfer of nT_reg cells (25). Together, these observations suggest that iT_reg cells contribute to tolerance. In the treatment of established disease, it has not been determined whether iT_reg and nT_reg cells are interchangeable and function in an additive fashion or whether they act synergistically, perhaps by nonredundant mechanisms. To investigate the respective role of iT_reg and nT_reg cells in therapy, we treated mice at the point where they lost 5–7% of their initial body weight (approximately day 30) with 1 × 10⁶ CD4⁺Thy1.1⁺ iT_reg cells or 1 × 10⁶ CD4⁺Thy1.1⁺ nT_reg cells. EGFP⁺ iT_reg cells were generated in vitro by TCR cross-linking in the presence of TGF-β1 and purified by cell sorting. EGFP⁺ nT_reg cells were obtained from the spleens and lymph nodes of Foxp3^EGFP mice and also purified by cell sorting. The congenic Thy1.1 marker was used to discriminate between the adoptively transferred T_reg cells used to treat the mice and those Thy1.2⁺ iT_reg cells generated in situ during the induction of colitis.

When the number of T_reg cells was increased by treating with 1 × 10⁶ CD4⁺ Thy1.1⁺ iT_reg cells, the mice failed to gain weight but survived to complete the experiment at 125 days (Fig. 2, A and B). Histology showed moderate hypertrophy of the colonic mucosa, distortion of the crypts, and a patchy inflammatory cell infiltrate in the lamina propria (Fig. 2C). Colitis scores averaged 3.0 ± 0.5, consistent with less severe disease than in mice containing only in situ-generated iT_reg cells (Figs. 1D and 2D, respectively). In the mesenteric lymph nodes, 9 ± 1.2% of CD4⁺ T cells were Thy1.1⁺EGFP⁺ (in vitro-derived iT_reg cells) and 3 ± 0.5% were Thy1.1⁻EGFP⁺ (in situ-derived iT_reg cells) (Fig. 2E). On average, a total of 3.3 ± 0.7 × 10⁶ T_reg cells were recovered from the mesenteric lymph nodes and spleen of treated mice (Fig. 2F).

FIGURE 2.

Treatment of colitis in the presence of in situ-derived iT_reg cells. A, Weight change following colitis induced by adoptive transfer of 4 × 10⁵ CD4⁺EGFP⁻Thy1.2⁺CD45RB^high T cells isolated from Foxp3^EGFP mice. After losing 5–7% of their body weight, recipients were treated by i.p. injection of 1 × 10⁶ sorted in vitro-derived Thy1.1⁺EGFP⁺ iT_reg cells (red; n = 9) or by 1 × 10⁶ sorted Thy1.1⁺EGFP⁺ nT_reg cells (green; n = 6). Color-matched solid lines represent the polynomial fit based on a random coefficient regression model. B, Kaplan-Meier survival curves for the mice in A. C, Representative histological sections from the colons of the mice in A stained with H&E. D, Scatter plots showing colitis scores for each group. E, Representative flow cytometry analysis of the mesenteric lymph nodes from mice treated with nT_reg cells (left panel) or iT_reg cells (right panel). Numbers denote mean values for the quadrant. F, Bar graphs depicting the total number of T_reg cells found in the mesenteric lymph nodes and spleens of treated mice. The proportion of in situ-derived iT_reg cells (red), nT_reg cells (green), and in vitro-derived iT_reg cells (pink) is indicated.

When mice were treated with 1 × 10⁶ CD4⁺Thy1.1⁺ nT_reg cells,, they exhibited weight gain consistent with clinical recovery, although survival was not significantly different than in mice treated with iT_reg cells (Fig. 2, A and B). Histological examination at 125 days showed nearly normal colons, with an average colitis score of 1.0 ± 0.6 (Fig. 2, C and D). In the mesenteric lymph nodes, 11 ± 1.1% of CD4⁺ T cells were Thy1.1⁺EGFP⁺ (nT_reg cells) and 3 ± 0.4% were Thy1.1⁻EGFP⁺ (in situ-derived iT_reg cells) (Fig. 2E). The total number of T_reg cells recovered from the mesenteric lymph nodes was less than half that seen in the iT_reg cell-treated mice (1.3 ± 0.3 × 10⁶; Fig. 2F). Thus, despite a >50% reduction in T_reg cell numbers, the combination of in situ-derived iT_reg cells and nT_reg cells suppressed bowel inflammation and restored tolerance, whereas mice with a T_reg cell compartment containing only iT_reg cells had chronic disease. These data suggested more effective disease suppression when the T_reg compartment was comprised of both iT_reg and nT_reg cells.

Functional synergy between nT_reg and iT_reg cells

To dissect the individual contributions of iT_reg and nT_reg cells and to examine the possibility of synergy between the two cell types, we induced colitis by the adoptive transfer of Thy1.1⁺ CD4⁺EGFP⁻CD45RB^high T_conv cells isolated from Foxp3^ΔEGFP mice. In this experiment, functional iT_reg cells are not produced in situ (Fig. 1), making it possible to control the T_reg cell dose and track each T_reg cell population in subsequently treated mice. When the transfer recipients had lost 5–7% of their initial body weight, they were treated with increasing numbers of nT_reg cells or with a combination containing iT_reg and nT_reg cells.

Mice treated with 0.25 × 10⁶ or 0.5 × 10⁶ Thy1.2⁺ nT_reg cells failed to gain weight and had reduced survival (Fig. 3, A and B). Mice treated with 1 × 10⁶ Thy1.2⁺ nT_reg cells initially gained weight for a short period of time, but then progressively lost weight. Only one animal lived to complete the experiment at 125 days. Histological analysis and colitis scores were consistent with severe disease in all groups (Fig. 3, C and D, average colitis score 3.7 ± 0.2). In mesenteric lymph nodes, the percentage of EGFP⁺ cells within the CD4⁺ gate increased with the nT_reg cell dose, reaching 13 ± 1.7% in mice that received 1 × 10⁶ Thy1.2⁺ nT_reg cells (Fig. 3E). Importantly, 80% of mice that received a total dose of 1 × 10⁶ T_reg cells comprised of a 1:1 mix of iT_reg and nT_reg cells gained weight and recovered completely. They had normal colons by histology (average colitis score, 1.2 ± 0.5). In the mesenteric lymph nodes of these mice, 2 ± 0.1% of the CD4⁺ T cells were Thy1.1⁺EGFP⁺ iT_reg cells and 10 ± 0.6% were Thy1.2⁺EGFP⁺ nT_reg cells (Fig. 3). The mesenteric lymph nodes and spleens of animals that survived for >50 days contained a similar number of T_reg cells, suggesting that homeostatic regulation of the T_reg cell compartment size is not tightly linked to the developmental origin of the T_reg cells (Fig. 3F). Taken together, the data in Figs. 2 and 3 show that the clinical effect of treating with a mixture of both T_reg cell types is greater than treating with either iT_reg or nT_reg cells alone. These data are consistent with the interpretation that iT_reg and nT_reg cells can act synergistically to establish and maintain tolerance. It seems likely that in situ-derived iT_reg cells and nT_reg cells also act in concert in a similar fashion.

FIGURE 3.

Treatment of colitis in the absence of in situ-derived iT_reg cells. A, Weight change following colitis induced by adoptive transfer of 4 × 10⁵ CD4⁺EGFP⁻ Thy1.1⁺CD45RB^high T cells isolated from Foxp3^ΔEGFP mice. After 5–7% weight loss, recipients were treated by i.p. injection of 0.25 × 10⁶ Thy1.2⁺EGFP⁺ nT_reg cells (red; n = 6), by 0.5 × 10⁶ Thy1.2⁺EGFP⁺ nT_reg cells (orange; n = 6), by 1 × 10⁶ Thy1.2⁺EGFP⁺ nT_reg cells (green; n = 5), or by a mixture of 0.5 × 10⁶ Thy1.1⁺ iT_reg and 0.5 × 10⁶ Thy1.2⁺nT_reg cells (blue; n = 5). Color-matched solid lines represent the polynomial fit based on a random coefficient regression model. B, Kaplan-Meier survival curves for the mice in A. C, Representative histological sections from the colons of the mice in A stained with H&E. D, Scatter plots depicting the colitis scores for each group. E, Representative flow cytometry analysis of the mesenteric lymph nodes from mice treated with nT_reg cells (left panels) or a mixture of both cell types (right panel). Numbers denote mean values for the quadrant or gate. F, Bar graphs depicting the total number of T_reg cells found in the mesenteric lymph nodes and spleens of treated mice. The proportion of nonfunctional in situ-derived ΔiT_reg cells (black), nT_reg cells (green), and in vitro-derived iT_reg cells (pink) is indicated. ns, Not significant).

Gene expression profiles from nT_reg and iT_reg cells

The genetic basis for the synergistic function of iT_reg and nT_reg cells was investigated by comparing the gene expression profile of in vitro-generated iT_reg cells with that of nT_reg cells isolated from Foxp3^EGFP mice. Results were also compared with other published gene expression profiles of T_reg cells and T_conv cells (15, 17, 26, 27). We first identified genes common to nT_reg and iT_reg cells as compared with T_conv cells. A total of 155 unique genes were found common to both nT_reg and iT_reg cells, of which 90 were overexpressed and 65 were underexpressed (Fig. 4A and supplemental Table I⁵). Comparing iT_reg and in vitro-activated nT_reg cells with activated T cells filtered out the contribution of T cell activation, revealing a smaller subset of common genes that included the prototypical T_reg cell genes Socs2, Gpr83, and Nrp1 (neuropilin) (supplemental Table II). Together, these findings confirmed the overlap of the genetic signatures of nT_reg and iT_reg cells.

FIGURE 4.

Overlap among the gene expression profiles of iT_reg cells and nT_reg cells. A comparison of the expression profiles of iT_reg and nT_reg cells with that of T_conv cells identified commonly (A) and selectively (B) expressed genes. All cells were derived from Foxp3^EGFP mice. The Venn diagrams illustrate the comparison groups and the number of distinct and shared genes. The bar graphs show a few of the genes that are differentially expressed in iT_reg and nT_reg cells as mean fold change values derived from three to five independent experiments, each representing mRNA pooled from 4 to 10 mice and scoring p < 0.05 by Student’s unpaired two-tailed t test. C, Scatter plots of mean gene expression values of iT_reg cells vs nT_reg cells (top panel) or activated nT_reg cells (bottom panel). D, Scatter plots of mean gene expression values of iT_reg cells vs ΔiT_reg cells (top panel) or iT_reg cells vs ΔiT_reg cells cultured for 7 days in 50 IU/ml IL-2 (bottom panel). Gene comparisons that are found significant in C and D are highlighted in red.

Despite sharing a portion of the T_reg genetic signature, nT_reg and iT_reg cells were genetically distinct, with each subset marked by a large number of differentially expressed genes (Fig. 4, B and C). A subset of canonical T_reg transcripts was selectively expressed in freshly isolated nT_reg cells but not in iT_reg cells, including the Ikaros family member Helios (Zfpn1a2), the calcium-binding protein S100a4 and S100a6, Klrg1, Itgb8, Swap70, Pim1, Ecm1, and Sell. In turn, a large number of transcripts were differentially enriched in iT_reg cells, including the serine protease inhibitors Ctla2a and Ctla2b, Tnfrsf5 (CD40), Acvr1 (activin receptor A type I), Lta (lymphotoxin α), Ahr (aryl hydrocarbon receptor), and Zbtb32 (repressor of gata) (supplemental Table III). The aryl hydrocarbon receptor is of particular interest given the recent report that activation of the aryl hydrocarbon receptor can generate iT_reg cells by a TGF-β1-dependent mechanism and that these iT_reg cells protect mice from experimental allergic encephalomyelitis (28). Although a subset of the genes selectively enriched in the freshly isolated nT_reg cells was down-regulated upon in vitro cell activation, the differential expression of several others, including Zfpn1a2, Sell, and Swap70 was retained (supplemental Table IV). Comparison with in vitro-activated nT_reg cells also revealed the iT_reg cells to be enriched in components of TGF-β-related pathways, including Acvr1, Tgfbr1, Tgfbr3, Smurf1, Id1, and Id2. Thus, at least some of differences in the gene expression profiles of nT_reg and iT_reg cells reflected a more fundamental divergence in their genetic programs that could not be attributed to their activation status.

To determine the role of Foxp3 in the induction and maintenance of the iT_reg cell genotype, we compared the gene expression profile of successfully derived iT_reg cells that expressed high levels of Foxp3 and exhibited potent in vitro suppressive activity with that of CD4⁺ T cells from the same culture that failed to express Foxp3 (supplemental Table V). We also compared the expression profiles of iT_reg cells with ΔiT_reg cells derived in vitro from Foxp3^ΔEGFP/+ mice (Fig. 4D and supplemental Table VI). Surprisingly, both comparisons revealed that the genetic signatures of the respective populations were virtually identical. However, when sorted ΔiT_reg cells were maintained in culture for 7 days in the presence of IL-2, the expression of a number of genes associated with the T_reg signature such as Gpr83, Nrp1, and Itgae was down-regulated (Fig. 4D and supplemental Table VII). These findings indicated that, within the set parameters of the analysis, Foxp3 contributed to the subsequent stability but not the initial establishment of the iT_reg cell genetic signature, which seems to be primarily shaped by the actions of TGF-β1 and TCR signaling (29).

Discussion

The central finding of this study is that iT_reg cells form a subdivision within the peripheral T_reg cell pool that can act synergistically with nT_reg cells to maintain tolerance. Although Foxp3 is required for iT_reg cell function and stability, the genetic signature of iT_reg cells only partially overlapped that of nT_reg cells, consistent with their different ontogeny and possibly nonredundant functional role. The concordant gene expression profiles seen at early time points in iT_reg and ΔiT_reg cells indicated that Foxp3 played little role in initiating iT_reg conversion. However, similar to nT_reg cells, the suppressive function of iT_reg cells is absolutely dependent upon Foxp3 (17, 30). The eventual loss of EGFP fluorescence in ΔiT_reg cells also indicated that stability of the iT_reg phenotype requires continued expression of a functional Foxp3 protein. This would explain previous observations that T_conv cells from Foxp3^ΔEGFP mice that were transferred into SCID recipients failed to give rise to iT_reg cells in vivo, even while inducing an aggressive colitis (17). A similar role has been suggested for Foxp3 in nT_reg cell phenotype stability, supporting the idea that a principle function of Foxp3 is to differentially regulate a subset of genes activated by TCR-, IL-2-, and TGF-β1-mediated signaling and to sustain its own expression (30).

The apparent dual requirement for iT_reg and nT_reg cells to enforce peripheral tolerance may reflect the distinct ontogeny and Ag specificity of the respective T_reg population. Early in life, iT_reg cells may be particularly vital due to the delay in nT_reg development relative to the production of T_conv cells by the newborn mouse thymus (31). Indeed, early thymectomy removes a large component of the developing nT_reg repertoire but leaves intact the capacity of peripheralized T_conv cells to express Foxp3, resulting in a milder disease phenotype than seen in Foxp3 deficiency (32). Other evidence theoretically supports the requirement for iT_reg cells. T_conv cells, from which iT_reg cells are derived, are by-and-large specific for foreign Ags. Conversely, several studies report that the nT_reg cell TCR repertoire may be enriched with self-specific receptors (33, 34), with some overlap with that of T_conv cells (33–36). Our data suggest that when comparing the nT_reg and T_conv TCR repertoires, overlap may be largely due to iT_reg cells. Expanding clones that produce both T_conv and iT_reg cell effectors would be predicted to generate the most frequently shared TCR isolates. Even if the molecular mechanisms for iT_reg- and nT_reg cell-mediated suppression are functionally equivalent, differences in their TCR repertoires could control the type and extent of T_reg cell activation. Additionally, local factors including the type of APC and the availability of cofactors like retinoic acid are likely to influence the size and Ag specificity of the iT_reg compartment (37–39). The composition of the antigenic environment may therefore be a primary determinant of the proportional contribution of iT_reg and nT_reg cells to the maintenance of tolerance.

Differences in the genetic signatures of iT_reg and nT_reg cells provide clues to their respective attributes. For example, Helios, a member of the Ikaros family of zinc finger transcription factors, was enriched in nT_reg relative to iT_reg cells. Helios is primarily expressed in T-lineage cells and early multipotential precursors (40, 41). It associates with the nucleosome remodeling deacetylase complex, which has been implicated in the activation of CD4 gene expression in developing thymocytes, suggesting a role for Helios-nucleosome remodeling deacetylase complexes in nT_reg development and perhaps phenotype stability (42, 43). In contrast, iT_reg cells were enriched in components of TGF-β-related pathways, such as the aryl hydrocarbon receptor, which may promote proliferation in the face of intense TGF-β signaling (44), and lymphotoxin-α, whose role in T_reg cell function remains unexplored. Both iT_reg and nT_reg cells differentially expressed a number of homing receptors such as Sell, suggesting distinct tissue-homing characteristics. These differences may individually or in combination provide differential effector mechanisms, which, together with their distinct TCR repertoires, enable the two populations to synergistically interact to promote in vivo suppression.

Recent reports demonstrate that nT_reg cells have epigenetic changes that are consistent with a stable, differentiated cell lineage (45, 46). Unlike nT_reg cells, the iT_reg cell compartment is not fixed and is maintained in dynamic equilibrium with T_conv cells. For example, in the colitis experiments, functional iT_reg cells are clearly generated in situ, while ~60% of the surviving in vitro-derived iT_reg cells lost Foxp3 expression. Our experiments are therefore most consistent with a model of T_reg cell function that involves a terminally differentiated nT_reg cell population augmented by iT_reg cells, with the latter entering and leaving the peripheral pool proportional to the needs of the host to control autoinflammatory responses.

Finally, the observation that both iT_reg and nT_reg cells are necessary to treat active colitis suggests that they may also have non-redundant roles in maintaining tolerance once established. In that regard, nT_reg or iT_reg depletion experiments in the absence of in situ-generated iT_reg cells may help to functionally discriminate between the two cell types. It will also be important to determine whether iT_reg cells function at sites other than the mucosal interface, where the microenvironment is conducive to their generation. These investigations will bear on the potential therapeutic use of T_reg cells in treating autoimmune diseases and graft-versus-host disease after bone marrow transplantation (47). Specifically, the use of iT_reg cells derived in vitro from T_conv cells may not prove sufficient to enforce tolerance, particularly in those situations where the indigenous nT_reg cells are either depleted or rendered ineffective. Strategies that promote both nT_reg and iT_reg responses may have a better chance of success.