Cutting Edge: ICOS-Deficient Regulatory T Cells Display Normal Induction of Il10 but Readily Downregulate Expression of Foxp3

The B7 family of ligands expressed by APC interact with the CD28 family of coreceptors expressed on T cells, delivering unique signals that were historically classified as either costimulatory or coinhibitory. More recently, it has become accepted that these interactions can have mechanistic effects beyond just modulation of T cell activation. For example, signaling via T cell coreceptors, including CD28, CTLA-4, herpes virus entry mediator, programmed cell death-1 (PD-1), and ICOS, figure prominently in the development and functions of regulatory T (Treg) cells, essential mediators of immune homeostasis (1–5).

Compared with wild-type mice, Icos^−/− mice harbor reduced Foxp3⁺ Treg cells in secondary lymphoid tissues (1, 6). Although dispensable for induction of Foxp3⁺, ICOS labels Treg cells with superior suppressive capacity (5) and promotes resistance of Treg cells to cell death (7). The ICOS pathway has also been implicated in the maintenance and/or function of CD4⁺ effector cells at homeostasis, following antigenic challenge, and during chronic inflammation (6, 8–11). Thus, despite often being used as a Treg cell marker, several of the functions of ICOS in T cell lineage maintenance also extend to effector and memory cells. Accordingly, a range of studies in animal models have positioned ICOS as a context-dependent negative or positive regulator or even a nonfactor in T cell–mediated diseases (12–14).

Loss-of-function polymorphisms in the gene encoding the ligand of ICOS (ICOSL) (15) confer susceptibility to inflammatory bowel disease, but neither Icos^−/− or Icosl^−/− mice develop overt intestinal inflammation. This is despite many reported links between ICOSL–ICOS and CD4 T cell production of IL-10 (7, 16–19), which is critical for intestinal immune homeostasis (20, 21). In this study, we show that murine ICOS deficiency results in increased induction of Il10 in CD4 T cells but reduced accumulation of large intestine (LI) Foxp3⁺ cells, including microbiota-dependent peripheral Treg (pTreg) cells. Icos^−/− Treg cells displayed increased methylation of Foxp3 conserved noncoding sequence 2 (CNS2) and preferentially downregulated Foxp3 relative to Icos^+/+ Treg cells. The extinction of Foxp3 rendered Icos^−/− Treg cells incapable of reversing gut inflammation. Our study identifies ICOS as an important mediator of Foxp3 stability that is dispensable for T cell production of IL-10 in the intestine.

CD45.1, Rag1^−/−, and Icos^−/−, and Icosl^−/−, mice were purchased from The Jackson Laboratory. Foxp3^−IRES-GFP, CBir1 TCR transgenic, and Myd88^−/−.Trif^−/− mice were gifts from Dr. V. Kuchroo, Dr. C. Elson, and Dr. S. Michalek, respectively. 10BiT mice have been previously described (22). All mice were bred and maintained at the University of Alabama at Birmingham in accordance with Institutional Animal Care and Use Committee guidelines.

The following mouse Abs were purchased from eBioscience: PE-anti–IL-17A, FITC–anti-Foxp3, allophycocyanin–anti-Helios, PE-Cy7–anti-CD4, anti-CD11c, and biotin–anti-ICOSL. The following were purchased from BD Biosciences: PE–anti-CD103, PerCP–anti-CD90.1, and PerCP-Cy5.5–anti-CD45.1. Samples were acquired on an BD LSR II instrument, and data were analyzed using FlowJo software.

The intestines were removed and stripped of mesenteric fat, and luminal contents were flushed using sterile HBSS. The epithelial layer was removed by incubating it in HBSS with 154 μg/l l-dithioerythritol and 2 μM EDTA. The remaining tissue was digested with 100 U/ml collagenase IV and 20 μg/ml DNase (Sigma-Aldrich) for 30 min at 37°C with gentle stirring. Total lamina propria cells were purified on a 40/75% Percoll gradient by room temperature centrifugation at 2000 rpm with no brake for 20 min.

Bisulfite conversion, pyrosequencing, and analysis were performed by EpigenDx (Hopkinton, MA). Assays ADS568-FS1 and ADS568-FS2 were used to analyze nine CpGs of the mouse Foxp3 CNS2 (−2369 to −2207 from the Foxp3 transcriptional start site).

CD45RB^hi T cells were FACS purified from B6.CD45.1 splenocytes and a total of 4 × 10⁵ cells were injected into each Rag1^−/− recipient. CD4⁺GFP⁺ cells were FACS sorted from Icos^+/+.Foxp3^gfp or Icos^−/−.Foxp3^gfp mice, both on the CD45.2 background. Each recipient received either PBS, a total of 2 × 10⁵ Icos^+/+ GFP⁺, or a total of 2 × 10⁵ Icos^−/− GFP⁺ cells on either day 0 (prevention) or day 28 (reversal). At necropsy, representative sections of the proximal, middle, and distal colon were fixed in formalin and embedded in paraffin, and 5-μm sections were cut and stained with H&E. Histological scoring was performed in a blinded fashion. Remaining tissue was processed to isolate lamina propria cells.

Statistical significance was calculated by an unpaired Student t test, Mann–Whitney U test, or ANOVA, as appropriate, using Prism software (GraphPad, San Diego, CA). All p values ≤0.05 are considered significant and are referred to as such in the text.

Our preliminary analysis of splenic and intestinal CD4 T cells suggested that coexpression of ICOS is not an essential feature of IL-10–competent cells, particularly in the LI (Supplemental Fig. 1). To further examine this, we compared the impact of ICOS deficiency on gut Treg cells with that of an extraintestinal tissue (spleen) and an inductive site (thymus). Consistent with previous findings, there was no difference in Foxp3⁺ cell frequencies in the thymus of Icos^+/+ and Icos^−/− mice (Fig. 1A, 1B), but there were significant reductions in splenic and LI Foxp3⁺ Treg cells in absence of ICOS. Importantly, in the LI, where we observed the greatest difference in frequency, we detected similar numbers of total CD4 T cells but significantly reduced numbers of Foxp3⁺ cells (Fig. 1C).

The transcription factor Helios is expressed by the majority of thymic Treg (tTreg) cells but only a minor fraction of colonic lamina propria Treg cells, although the utility of Helios as a marker of tTreg cells remains controversial. As expected (23), in Icos^+/+ mice, the majority of Foxp3⁺ cells in the thymus and spleen coexpressed high levels of Helios, and importantly, the same was true of Icos^−/− Foxp3⁺ cells in these two compartments (Fig. 1D, 1E). However, in the colonic lamina propria of Icos^−/− mice, we detected a significantly increased representation of Helios^Hi and a concomitant decrease in Helios^Lo/− Foxp3⁺ cells (Fig. 1D, 1E). Analysis of TCR excision circles (24) indicated that there is a statistically significant increase in thymic output of Treg cells in Icos^−/− mice relative to wild-type mice (Fig. 1F), which may help explain the increase in Helios^Hi cells in the lamina propria.

To confirm the impact of ICOS deficiency on definitively colonic pTreg cells and avoid the controversy surrounding the use of Helios as a marker, we employed the CBir1 TCR transgenic system (25). This strain expresses a TCR specific for CBir1 flagellin, a microbiota-derived Ag detectable in healthy mice and humans. As with most TCR transgenics, these mice can rearrange a nontransgenic (endogenous) TCR α-chain, meaning they can generate self-antigen reactive Treg cells in the thymus (Fig. 1G, left panel). However, by rendering these mice deficient for the RAG-1 (Rag1), we eliminated tTreg cell development (Fig. 1G, center and right panels). Therefore, any Treg cell detected in the periphery of these mice is a bona fide pTreg cell. In healthy, 6- to 8-wk-old CBir1.Rag1^−/− mice, ICOS deficiency resulted in reduced numbers and frequencies of Foxp3⁺ LI Treg cells, despite an elevated number of total CD4⁺ T cells (Fig. 1H, 1I).

Importantly, developing pTreg cells likely receive an ICOSL signal because lamina propria CD11c⁺CD103⁺ dendritic cells express ICOSL. This expression occurs independent of the microbiota or TLR signaling, suggesting that it is developmentally regulated (Supplemental Fig. 2A, 2E). The same signal is likely available to developing tTreg cells because ICOSL is also expressed by MHC class II–expressing thymocytes (Supplemental Fig. 2F, 2G).

To determine whether the impact of ICOS on colonic Foxp3⁺ cells extends to IL-10–producing cells, we used IL-10 BAC-In transgenic (10BiT) mice in which induction of Il10 results in surface expression of Thy-1.1 (CD90.1) (22). In Icos^+/+ mice, the vast majority of colonic IL-10–competent cells coexpress Foxp3, with limited numbers of Foxp3⁻Thy1.1⁺ cells detected. In contrast, in Icos^−/− mice, we found that the majority of IL-10–competent cells were actually Foxp3⁻ cells (Fig. 2A). Overall, ICOS deficiency led to a significant reduction in the proportion of Foxp3⁺Thy1.1⁺ cells among CD4 T cells as a direct result of reduced Foxp3. (Fig. 2A, 2B). However, this was counterbalanced by significant increases in Foxp3⁻IL-10⁺ cells. The net effect was that the total expression of IL-10 by CD4 T cells as determined by cell frequencies or fluorescence intensity of Thy1.1 expression was not diminished (Fig. 2A, 2C).

Consistent with previous results (26), the majority of colonic Foxp3⁺IL-10⁺ cells in wild-type mice were Helios^Lo/−. However, Icos^−/− mice displayed a substantial population of Helios-expressing Foxp3⁺IL-10⁺ cells (Fig. 2D). Thus, despite the reduced frequency of Foxp3⁺ cells in Icos^−/− mice independent of Helios expression, the proportion that expressed IL-10 and the levels of IL-10 expression remained largely unchanged relative to Icos^+/+ mice (Fig. 2D, 2F). To eliminate any possible host-intrinsic effects on the foregoing results, we performed adoptive transfer of CD4 single-positive Foxp3⁻ thymocytes from congenically marked 10BiT.Foxp3^gfp.Icos^+/+ (CD45.1) and 10BiT.Foxp3^gfp.Icos^−/− (CD45.2) mice into the same T cell–deficient (Tcrβδ^−/−) recipients (Fig. 2G). After 3 wk, despite similar frequencies of colonic CD4 T cells derived from each source, the frequency of Icos^+/+ Foxp3⁺ cells was ∼5 times that of Icos^−/−, and conversely, IL-17 frequencies were elevated in the latter. The frequency of IL-10–competent CD4 T cells was similar in both Icos^+/+ and Icos^−/− cells (Fig. 2H, 2I), and there was a significant increase of Foxp3⁻IL-10⁺ cells among Icos^−/− colonic CD4 T cells, collectively mimicking intact mice (Fig. 2J, 2K).

Interestingly, among lamina propria CD4 T cells, the frequencies of IL-10–competent cells actually increased with aging (Supplemental Fig. 3A, 3B), suggesting a potentially compensatory role for IL-10 in the face of the colonic Treg cell deficits in Icos^−/− mice. Collectively, these results argue that ICOS is dispensable for induction of Il10 in intestinal Treg cells.

The intensity of Thy1.1 expression by Foxp3⁻ cells in the LI lamina propria of Icos^−/− mice strongly resembled that of LI Foxp3⁺Thy1.1⁺ (Fig. 2A), raising the possibility that some of the Icos^−/− Foxp3⁻Thy1.1+ cells were “ex-Foxp3” cells. To determine whether ICOS impacts the stability of Foxp3, we conducted pyrosequencing analysis of CNS2 of the Foxp3 locus, which is demethylated in Treg cells that stably express Foxp3 (27). In Icos^−/− Treg cells, there was significant methylation of Foxp3 CNS2 relative to Icos^+/+ cells. In fact, the methylation levels closely resembled that of naive T cells (Fig. 3A). To confirm the instability of Icos^−/−Foxp3⁺ cells, we FACS-sorted CD4⁺GFP⁺ cells from wild-type (CD45.1) and Icos^−/− (CD45.2) Foxp3^gfp reporter mice (Fig. 3B, left panel) and cocultured equal numbers in the presence of IL-2, anti-CD3, and anti-CD28. On day 3, the majority of CD45.2⁺ cells were Foxp3⁻ in contrast to cells of wild-type origin that were still mostly Foxp3⁺ (Fig. 3B, right panel). Stimulation (Stim) in the presence of proinflammatory cytokines IL-1β, IL-6, IL-12, and IL-23 (Stim plus cytokines) or blocking Abs targeting IL-6R, IL-12/23p40, and IL-21R (Stim plus blockade) had no major impact on the Foxp3 loss by Icos^−/− T cells (Fig. 3B). Altogether, these data argue that ICOS helps to imprint stable expression of Foxp3 mainly by promoting demethylation of Foxp3 CNS2. The highly methylated CNS2 leading to the rapid downregulation of Foxp3 may help to explain why the elevated thymic output of Foxp3⁺ cells (Fig. 1F) is incapable of restoring to wild-type levels the numbers of Foxp3⁺ cells in the spleen and especially the LI.

We then decided to definitively examine the functional consequence of this instability in vivo, first under homeostatic conditions. We employed the T cell cotransfer model of colitis and injected equal numbers of CD45.2⁺ wild-type or Icos^−/− Foxp3⁺ cells into Rag1^−/− recipients that simultaneously received CD45.1⁺ naive CD45RB^hi T cells. As expected, naive cells alone induced severe weight loss and colonic inflammation (Fig. 3C, 3E). However, as previously shown (28), both Icos^+/+ and Icos^−/− Treg cells prevented the development of colitis. Despite similarly sized donor CD45.2 fractions of Icos^+/+ and Icos^−/− origin, we detected significantly reduced frequencies of the latter that still expressed Foxp3 (Fig. 3F, 3H). Thus, Icos^−/− Treg cells were able to inhibit the development of colitis, despite the enhanced loss of Foxp3. In longer term analyses, under similarly homeostatic conditions, we examined the ability of Icos^−/− Treg cells to prevent the spontaneous autoimmunity and premature death experienced by Foxp3^−/− mice. Our results showed that Icos^−/− Treg cells could temporarily rescue Foxp3^−/− mice but preferentially lost expression of Foxp3 and ultimately failed to promote long-term survival in contrast to Icos^+/+ Treg cells (Supplemental Fig. 4).

To determine the potential consequences of Icos^−/− Treg cell instability during inflammation, we first induced colitis in recipient mice then transferred equal numbers of wild-type or Icos^−/− Treg cells. In this setting, Icos^−/− Treg cells failed to prevent the wasting disease characteristic of mice that did not receive a secondary Treg cell transfer (Fig. 4A). In contrast, recipients of Icos^+/+ Treg cells were rescued from disease, as further confirmed by histological analysis (Fig. 4B, 4C). Furthermore, the failure of Icos^−/− Treg cells to mitigate the ongoing inflammation in recipient mice correlated with their almost complete loss of Foxp3 expression, reflected in reduced frequencies and numbers of Icos^−/− Foxp3⁺ cells (Fig. 4D, 4E) despite similar numbers of cells of CD45.2 origin (Fig. 4F). These results provide compelling evidence that ICOS imprints Treg stability that is particularly important for Treg cell function during inflammation.

In this study we identified a novel role for ICOS signaling in imprinting the epigenetic stability of Foxp3⁺ Treg cells with no impairment in induction of Il10 in gut CD4 T cells. Despite this novel role, Icos^−/− mice retain a sizeable pool of Treg cells, and under specific pathogen–free housing conditions, do not succumb to the spontaneous autoimmunity characteristic Foxp3-deficient mice. This may be explained by 1) the thymic output of Foxp3⁺ cells even in aged mice and 2) the increase in IL-10–producing cells throughout life. However, during inflammation, this instability of Foxp3 produces detrimental consequences for the host. Our data, together with recent findings of a role for PD-1 signaling in stabilizing induced Treg cells (4), identify yet another contribution of T cell coreceptors in imprinting the long-term fate of Treg cells. Ultimately, these discoveries will present unique opportunities to target these pathways in cell-based treatment of chronic inflammatory diseases.

The authors have no financial conflicts of interest.