IL-27 promotes T cell–dependent colitis through multiple mechanisms

To determine whether IL-27 responsiveness of T cells was required for the establishment of the colitis phenotype, we used the CD4⁺CD45Rb^hi transfer model of colitis (Powrie et al., 1993). First, we crossed the Il27ra^−/− allele (Chen et al., 2000) into the balb/c background for 12 generations. As previously described in C57BL/6 mice, Il27ra deficiency causes no overt abnormalities in the balb/c background (unpublished data). However, to our surprise, transfer of FACS-purified balb/c Il27ra^−/− CD45Rb^hi T cells into CB17-SCID recipient mice resulted in partial protection against weight loss compared with transfer of WT CD45Rb^hi T cells (Fig. 1, A and B), suggesting that IL-27 plays a proinflammatory role in this model. Upon analysis of the colons at the end of the study, we found that recipients of Il27ra^−/− CD45Rb^hi cells also displayed a reduced amount of colonic shortening (Fig. 1 C) and had significantly reduced histological scores (Fig. 1, D and E). Therefore, the presence of Il27ra on T cells is required in this model for the development of both fulminant colitis and maximal weight loss.

As expected, control animals transferred with total CD4⁺ cells from either genotype suffered from neither weight loss nor colitis, and this is because of the presence of natural T_regs (nT_regs) in the cell graft (Fig. 1). We therefore investigated whether a difference in the presence of forkhead box P3–positive (Foxp3⁺) cells might account for the observed phenotype. As shown in Fig. 2 A, we could detect elevated T_reg cells in the absence of Il27ra on peripheral blood T cells as early as 5 wk after transfer of CD45Rb^hi cells. Furthermore, when we sacrificed mice at the end of the study, we found that recipients of Il27ra^−/− CD45Rb^hi cells contained approximately two to three times the normal proportion of Foxp3⁺ T cells in blood, spleen, mesenteric LN (mLN), and lamina propria (Fig. 2, A–C). In agreement with enhanced suppressive activity, we also noted that Il27ra^−/− recipients contained fewer total CD4⁺ T cells, especially in the lamina propria (Fig. 2, D and E).

Previous studies and experiments done here revealed no reduction in the frequency of Foxp3⁺ cells in naive Il27ra^−/− mice (Fig. S1 A; Batten et al., 2006). Furthermore, the suppressive capacity of these cells in vitro is unaffected by the absence of Il27ra (Fig. S2 A; Batten et al., 2006). However, because FACS-sorted WT and Il27ra^−/− CD4⁺CD45Rb^hi cells contained ∼0.5% nT_regs (Fig. S2 B), it remained possible that the increased frequency of Foxp3⁺ cells in Il27ra^−/− CD45Rb^hi recipients resulted from preferential in vivo expansion of or enhanced in vivo suppressive capacity by Il27ra^−/− nT_regs. To address this concern, we transferred purified CD4⁺CD25⁺ cells from WT or Il27ra^−/− mice into Rag2-deficient C57BL/6 mice that had been transferred with WT CD4⁺CD45Rb^hi cells 7 wk before. nT_regs from either genotype were fully capable of rescuing their hosts from systemic wasting disease (Fig. S2, C and D). Furthermore, we observed similar frequencies of Foxp3⁺ cells of both genotypes in the blood, spleen, and mLNs of the rescued recipient animals, suggesting that the in vivo expansion rate of nT_regs is not affected by the Il27ra genotype (Fig. S2, E and F).

Inducible T_regs develop from naive CD4⁺ T cells upon stimulation in the presence of TGF-β. It has been demonstrated in the context of transfer colitis that this type of conversion occurs in vivo in a small fraction of the transferred cells (Sun et al., 2007); however, the resulting number of Foxp3⁺ cells is insufficient to afford the host full protection, and colitis ensues. Prior studies have suggested that IL-27 can suppress the TGF-β–driven induction of Foxp3⁺ cells in vitro (Neufert et al., 2007; Huber et al., 2008); therefore, we investigated whether IL-27 normally restrains T_reg conversion in vivo. To enable experiments that are not encumbered by nT_reg contamination, we bred the Il27ra^−/− allele into a DO11.10⁺ and Rag2^−/− background. Such mice contain a pristine population of naive, OVA-specific T cells that is completely devoid of nT_regs (Fig. 3 A), and hence represent an ideal system to study T_reg conversion. To validate this system, we first cultured DO11.10⁺Rag2^−/− T cells in the presence of lamina propria DCs (LPDCs), OVA peptide, and TGF-β. Indeed, lamina propria DCs allowed for Foxp3⁺ conversion, and IL-27 suppressed T_reg conversion (Fig. 3 B). However, in this experimental system IL-27 also affected CD25 expression and generally suppressed T cell activation as measured by CD69 surface expression (Fig. 3 C). IL-27 acted directly on T cells to down-regulate Foxp3, CD25, and CD69, because it had no effect when Il27ra^−/− T cells were used in the experiment.

Next, we transferred purified DO11.10⁺Rag2^−/− T cells into naive balb/c recipients and exposed them to OVA in the drinking water. Exposure to antigen led to a significant increase in Foxp3⁺ cells in the spleens and mLNs (Fig. 4, A–C). Consistent with our data obtained from the colitis model, we observed that Il27ra deficiency significantly augmented peripheral T_reg development, indicating that IL-27 limits T_reg conversion even in a noninflammatory environment. This effect was further accentuated when we measured absolute numbers of Foxp3⁺ DO11.10⁺Rag2^−/− cells (Fig. 4 C). Because only naive Foxp3⁻ cells were transferred into recipients, this experiment also conclusively proves that IL-27 signaling limits T_reg conversion rather than expansion of nT_regs. Consistent with previous observations (Villarino et al., 2006), Il27ra^−/− DO11.10 cells produced more IL-2 (Fig. 4 D), whereas IFN-γ production was minimal irrespective of Il27ra expression in the noninflammatory environment of unchallenged balb/c mice (Fig. 4 E). However, increased production of IL-2 is not responsible for enhanced T_reg conversion because IL-2 does not override the suppressive effect of IL-27 on Foxp3 induction (Neufert et al., 2007 and unpublished data), which has been shown in vitro to be a direct, STAT3-mediated effect of IL-27 on T cells (Huber et al., 2008). Importantly, we still observed enhanced T_reg conversion when we repeated this experiment under inflammatory conditions in mice that had received WT CD45Rb^hi cells 4 wk earlier (Fig. 4 F), suggesting that the inflammatory milieu by itself is insufficient to restrict Foxp3 expression in the absence of IL-27 signaling. However, introduction of Il27ra^−/− CD45Rb^hi cells into mice with preestablished colitis failed to demonstrate a therapeutic effect (Fig. S3).

Despite the increased conversion rate of Il27ra^−/− CD45Rb^hi cells, the majority of the transferred cells remained Foxp3⁻. We therefore also examined cytokine production in the colon by RT-PCR and in restimulated lymphocytes isolated from the spleen, the mLN, and the colonic lamina propria of CD45Rb^hi recipient animals by intracellular staining. Although IL-12, IL-23, IL-2, IL-6, and IL-27 were all found to be induced in colitic mice, no statistically significant changes were noted between WT and Il27ra^−/− CD45Rb^hi recipients (Fig. S4). In accordance with prior observations (Chen et al., 2000; Artis et al., 2004; Batten et al., 2006; Stumhofer et al., 2006), T cells isolated from Il27ra^−/− CD45Rb^hi recipients produced significantly less IFN-γ as assessed by intracellular staining (Fig. 5, A and B), mimicking decreased production of IFN-γ in naive mice (Fig. S1 B). The overall decrease of CD4⁺ cells (Fig. 2 E) further accentuates the decrease in absolute numbers of IFN-γ–producing cells (not depicted). Therefore, the question of whether diminished IFN-γ production is partially responsible for the protective effect of Il27ra deficiency merits consideration. Using IFN-γ neutralizing antibodies, others have previously reported a pathogenic role for IFN-γ in this model (Powrie et al., 1994). In our hands, IFN-γ^−/− CD45Rb^hi cells elicited less severe colitis, but also caused highly aggressive wasting disease requiring early termination of the experiment (Fig. S5, A–C). Consistent with findings obtained by others (Wang et al., 2006), IFN-γ^−/− CD45Rb^hi cells had a diminished propensity to become Foxp3⁺ and expanded more aggressively (Fig. S5, D and E). Thus, IFN-γ deficiency phenocopies IL-27Ra deficiency in terms of reduced colitis, but not in terms of the improved wasting disease, increased T_reg conversion, or reduced expansion of effector cells. Therefore, a mere reduction in IFN-γ production cannot fully explain the Il27ra^−/− phenotype.

Conversely, and consistent with prior studies (Artis et al., 2004; Batten et al., 2006; Stumhofer et al., 2006; Yang et al., 2008), we observed mild elevations in TH17 (IL-17A and IL-22) and TH2 (IL-5 and IL-13) cytokines produced by Il27ra^−/− CD45Rb^hi cells (Fig. 5, C–E and not depicted). However, these changes were effectively neutralized in absolute terms by the lower total number of CD4⁺ cells (Fig. 2 E and not depicted). We did not observe increased neutrophil infiltration, which one would expect as a consequence of IL-17 overexpression (Fig. 6). Therefore, the minor relative changes in IL-17 and IL-22 production by Il27ra^−/− CD45Rb^hi cells are unlikely to contribute significantly to disease protection, although IL-17 (Izcue et al., 2008; Leppkes et al., 2009; O’Connor et al., 2009) has been shown to either be neutral or exert protective effects in the context of this model, and IL-22 is protective (Zenewicz et al., 2008).

We demonstrate conclusively that Il27ra^−/− CD45Rb^hi T cells are less colitogenic than their WT counterparts in the murine transfer colitis model. Our analysis of the T cell graft after 12 wk of incubation in lymphopenic hosts revealed that a much greater proportion of IL27ra^−/− CD45Rb^hi T cells assumed a Foxp3⁺ phenotype, whereas the remaining Foxp3⁻ cells proliferated less and produced much less of the TH1 cytokine IFN-γ. Although there were mild relative increases in Il27ra^−/− effector cells producing TH2 and TH17 cytokines, the lower abundance of Il27ra^−/− effector cells effectively neutralized this difference, which is consistent with the absence of an increase in neutrophil infiltration into the lamina propria. Therefore, the reduced colitis and weight loss in recipients of Il27ra^−/− CD45Rb^hi T cells is a consequence of increased T_reg conversion, lower IFN-γ production, decreased effector cell proliferation, or a combination of these effects.

Together, the changes in T cell phenotype and abundance form a chicken and egg conundrum: do effector cells proliferate less because there are more Foxp3⁺ cells present, or is there an increase in Foxp3⁺ cells because effector cells fail to proliferate and cause inflammation efficiently in the absence of IL-27Ra, which in turn de-represses T_reg conversion? Several of our data points favor the first of these two possibilities: First, IL-27 is known to antagonize TGF-β driven T_reg conversion in vitro (Fig. 3; Neufert et al., 2007; Huber et al., 2008), and thus conversion is expected to be de-restrained in Il27ra^−/− T cells in vivo. Second, we observed an increase in absolute numbers, not just relative percentages, of Il27ra^−/− Foxp3⁺ T cells in our OVA-dependent tolerization model (Fig. 4). This difference persisted even in the context of established colitis (Fig. 4 F), and thus represents a cell intrinsic effect that is independent of the inflammation status of the environment into which the cells are transferred. The OVA oral tolerance model is a much shorter term experiment than the 12-wk colitis study, and is therefore more amenable to mechanistic interpretation. And lastly, although a reduction in IFN-γ production could potentially explain the reduction in colitis, T_reg conversion of IFN-γ^−/− T cells was not enhanced, and they still caused significant wasting disease. Therefore, reduced T_reg conversion is not secondary to reduced production of IFN-γ in the absence of IL-27Ra. In summary, these observations provide strong suggestive evidence that the primary effect of IL-27 in this context is to restrain T_reg conversion.

We attempted to therapeutically rescue mice with preestablished transfer colitis by introducing Il27ra^−/− CD45Rb^hi cells, but this experiment was not successful (Fig. S3). Although a trend toward increased Foxp3 expression was still noticed 6 wk after transfer of the Il27ra^−/− CD45Rb^hi cells into the inflammatory environment, and although the transferred cells clearly produced less IFN-γ (Fig. S3, F and G), we did not detect any changes in weight loss or colon length. This result suggests that although de-repression of T_reg conversion through loss of IL-27Ra at the onset of inflammation significantly affects the ultimate outcome, it is not robust enough to overcome preestablished inflammation.

Earlier studies by our group and others showed that IL-27 can induce IL-10 production in T cells and lead to the development of Foxp3⁻, IFN-γ⁺IL-10⁺ Tr1 cells (Awasthi et al., 2007; Fitzgerald et al., 2007; Stumhofer et al., 2007; Batten et al., 2008). Because Tr1 cells have potent immunoregulatory effects (Anderson et al., 2007; Jankovic et al., 2007; Trinchieri, 2007), it has been postulated that this is a mechanism by which IL-27 exerts immune suppression in infectious and autoimmune disease. These earlier observations are at odds with the proinflammatory role of IL-27 described here. However, in transfer colitis, IL-10 production by cells originating from the Foxp3⁻ CD45Rb^hi graft is minimal (Uhlig et al., 2006). We found no Il27ra-dependent difference in IL-10 production by transferred CD45Rb^hi cells, and IL-10 production in general was minimal in this model (unpublished data). Thus, the Foxp3-suppressing effects of IL-27 are not in contradiction with its Tr1-inducing effects; the dichotomy merely reflects the differences between the physiological contexts in which IL-27 stimulation occurs and may explain why the essential function of IL-27 in the regulation of T_reg differentiation has not been noted previously.

It is perhaps appropriate to point out that the majority of in vivo effects assigned to IL-27 have been inferred from the analysis of Il27ra^−/− mice, and only some have been confirmed by studies of Il27p28^−/− or Ebi3^−/− mice. This study is no different in that regard. To date, IL-27 is the only confirmed ligand for IL-27Ra. However, IL-35 has been described as an IL-27–related heterodimer with potent immunoregulatory effects (Collison et al., 2007, 2010). IL-35 consists of IL-12p35 and Ebi3, and although its receptor has not been identified, IL-27Ra remains a candidate. Another recently described IL-27–related heterodimer consists of IL-27p28 and cytokine-like factor and appears to bind to IL-27Ra, but conclusive proof that IL-27Ra is required for signaling is currently not available (Crabé et al., 2009). Thus, it remains possible that IL-27Ra has ligands other than IL-27, which might further contribute to the apparent complexity and dichotomous nature of IL-27 biology in vivo.

In summary, we demonstrate that IL-27 exerts proinflammatory effects in the T cell transfer colitis model. Our experiments demonstrate conclusively that IL-27 acts to suppress induced T_reg development in vivo, and thus reveal a hitherto unrecognized proinflammatory mechanism. Collectively, our results suggest that targeting of IL-27 in situations where pathology results from a breakdown in T_reg-mediated tolerance may result in significant therapeutic benefit.

All mice were maintained under pathogen-free conditions, and experiments were approved by the Institutional Animal Care and Use Committee of Genentech, Inc. The Il27ra^−/− allele (Chen et al., 2000) was backcrossed onto the balb/c background for 12 generations. This strain was crossed further to the DO11.10⁺Rag2^−/− background (Taconic). CB17-SCID mice were purchased from Charles River. balb/c and Rag2^−/−.C57BL6 mice were obtained from Taconic.

Unless otherwise indicated, all cytokines were purchased from R&D Systems and all antibodies used in flow cytometry or culture experiments were purchased from BD. Anti-IL-22 (clone 3F11, isotype mouse IgG2a; Genentech) was directly conjugated to Alexa Fluor 647 (Invitrogen).

Naive CD4⁺CD45Rb^hi cells were isolated from the spleens of female Il27ra^+/+ or Il27ra^−/− mice by FACS sorting. In brief, single-cell suspensions were depleted of red blood cells by hypotonic lysis and CD4⁺ cells were purified by MACS-positive L3T4 selection (Miltenyi Biotec). After staining with Pacific blue–conjugated anti-CD4, PE-conjugated anti-CD45Rb, and PECy5-conjugated anti-CD44, CD4⁺CD44^loCD45Rb^hi naive T cells were purified (>98%) by cell sorting (FACSAria; BD). Female CB17-SCID mice were injected with 3 × 10⁵ CD45Rb^hi cells or unsorted CD4⁺ cells i.v. Mice were monitored for weight loss and sacrificed by CO₂ asphyxiation 11–12 wk after initiation of the experiment. Blood was obtained by retroorbital bleeds under isoflurane anesthesia at the indicated time points.

At the time of sacrifice, mouse colons were removed and flushed, and the length was measured from rectum to cecum. Tissues were immediately fixed in 10% buffered formalin, and 4–5-µm paraffin-embedded sections were stained with hematoxylin and eosin. Colitis severity was scored in the proximal colon, medial colon, distal colon, and rectum on a scale of 0–5, with 0 and 5 representing a normal colon and severe colitis, respectively. The scores of four anatomical regions were totaled for each mouse to yield a total histological score.

Single-cell suspensions were obtained from spleens and mLNs. Lamina propria leukocytes were isolated as described previously (Zheng et al., 2008). Cell counts were measured by ViaCount analysis on the Guava PCA-96 (Millipore). Anti-CD4 and anti-CD25 were used for surface staining of lymphocytes. For intracellular cytokine staining, single-cell suspensions from spleens, mLNs, and lamina propria were restimulated for 4 h in RPMI containing 10% FBS with 50 ng/ml of phorbol 12-myristate 13-acetate and 500 ng/ml of ionomycin in the presence of 5 µg/ml of brefeldin A (Sigma-Aldrich). Cells were then fixed in 1% paraformaldehyde in PBS, permeabilized with 0.5% saponin in flow cytometry buffer (0.5% BSA in PBS), and stained intracellularly with PE-conjugated anti–IL-17, PECy7-conjugated anti–mouse IFN-γ, Alexa Fluor 647–conjugated anti–mouse IL-22 (Genentech), and PE-conjugated anti–mouse IL-13. Alternatively, Foxp3 staining was performed according to manufacturer’s protocol (eBioscience).

Lamina propria DCs were purified after collagenase treatment by sorting of CD11c⁺MHCII⁺ cells. DO11.10⁺CD4⁺ cells were MACS purified and incubated (5 × 10⁴ cells/well) with LPDCs (5 × 10³ cells/well) in the presence of OVA_323-339 (0.3 µM), TGF-β (3 ng/ml), and IL-27 (20 ng/ml) as indicated. Cells were stimulated for 5 d in round-bottom 96-well plates, and intracellular Foxp3 was analyzed as described in the previous paragraph.

CD4⁺ T lymphocytes were enriched (∼85% purity) from the spleens and mLNs of female Il27ra^+/+ or Il27ra^−/− DO11.10⁺Rag2^−/− mice by negative MACS selection (Miltenyi Biotec). Female balb/c recipient mice (Taconic) were injected with 1.8 × 10⁶ cells I.V. on day 0. On day 1, mice were given 1.5% OVA (Grade V; Sigma-Aldrich) dissolved in drinking water for 5 d, as previously described (Sun et al., 2007). OVA water was replaced every 48 h, and control mice received normal drinking water. On day 6, mice were sacrificed and spleens, mLNs, Peyer’s patches, and lamina propria lymphocytes were assessed for Foxp3 expression. Cells were surface stained with PE-conjugated anti–KJ1-26 for the DO11.10 TCR, and then stained for intracellular IL-2, IFN-γ, or Foxp3. In some experiments, mice were injected with 3 × 10⁵ WT CD45Rb^hi cells to initiate colitis and, 4 wk later, were transferred with 2 × 10⁶ CD4⁺DO11.10⁺ cells from WT or Il27ra^−/− mice. Mice were either given normal water or 1.5% OVA for 5 d, as indicated.

Fig. S1 shows that naive Il27ra^−/− mice have unaltered levels of Foxp3⁺ T_regs and decreased levels of IFN-γ–producing CD4⁺ cells. Fig. S2 shows that Il27ra^−/− natural T_reg have normal suppressive capacity. Fig. S3 shows that Il27ra^−/− CD45Rb^hi cells fail to control previously established colitis. Fig. S4 shows cytokine expression during transfer colitis in recipients of WT or Il27ra^−/− cells, assessed by qPCR. Fig. S5 details transfer colitis elicited by transfer of IFN-γ^−/− CD45Rb^hi cells into rag2-deficient recipients. Onlen supplemental material is available.

We would like to thank Jeong Kim for critical discussion of the manuscript and William Forrest for assistance with statistical analysis. We also thank Jim Cupp and members of the Genentech flow cytometry group for contributions in cell sorting, the Genentech histology laboratory for specimen processing and staining, and Ben Torres for animal husbandry.

Nico Ghilardi, Nandhini Ramamoorthi, Noelyn M. Kljavin, Jennifer H. Cox, and Lauri Diehl are full time employees at Genentech, a member of the Roche group. Marcel Batten is supported by a fellowship from The Cancer Institute of New South Wales.

The authors declare no additional conflicting financial interests.