The TNFRSF members CD27 and OX40 coordinately limit TH17 differentiation in regulatory T cells

See allHide authors and affiliations

Science Immunology  21 Dec 2018:
Vol. 3, Issue 30, eaau2042
DOI: 10.1126/sciimmunol.aau2042

Constraining skin Tregs

Aberrant expression of interleukin-17A (IL-17A) by regulatory T cells (Tregs) is one of the perturbations observed in chronic human inflammatory skin diseases such as psoriasis. Remedios et al. investigated the pathways that normally restrain Tregs from morphing into pathogenic cells secreting IL-17A. Absence of the CD27 and OX40 receptors on skin-resident Tregs in mice led to marked increases in their IL-17A expression and cutaneous inflammation. The subset of cutaneous Tregs making IL-17A in patients with psoriasis and hidradenitis suppurativa expressed low levels of CD27. These studies identify two receptor-ligand pathways that act in concert to maintain the normal homeostatic function of cutaneous Tregs.


Regulatory T cells (Tregs) are closely related to TH17 cells and use aspects of the TH17-differentiation program for optimal immune regulation. In several chronic inflammatory human diseases, Tregs express IL-17A, suggesting that dysregulation of TH17-associated pathways in Tregs may result in either loss of suppressive function and/or conversion into pathogenic cells. The pathways that regulate the TH17 program in Tregs are poorly understood. We have identified two TNF receptor superfamily (TNFRSF) members, CD27 and OX40, that are preferentially expressed by skin-resident Tregs. Both CD27 and OX40 signaling suppressed the expression of TH17-associated genes from Tregs in a cell-intrinsic manner in vitro and in vivo. However, only OX40 played a nonredundant role in promoting Treg accumulation. Tregs that lacked both CD27 and OX40 were defective in controlling skin inflammation and expressed high levels of IL-17A, as well as the master TH17 transcription factor, RORγt. Last, we found that CD27 expression was inversely correlated with Treg IL-17 production in skin of patients with psoriasis and hidradenitis suppurativa. Together, our results suggest that TNFRSF members play both redundant and distinct roles in regulating Treg plasticity in tissues.


Regulatory T cells (Tregs) are critical for maintaining immune homeostasis and mitigating tissue damage caused by excessive inflammation (1). It is becoming increasingly appreciated that Tregs adapt to the local inflammatory environment by acquiring specific programs that facilitate optimal immune regulation. In this capacity, Tregs use effector T cell (Teff)–associated transcription factors T-bet, interferon regulatory factor 4 (IRF4), and signal transducer and activator of transcription 3 (STAT3) to regulate T helper cell type 1 (TH1), TH2, and TH17 responses, respectively (24). Expression of these “master” TH cell lineage transcription factors imparts Tregs with the ability to colocalize with TH cells driving a given immune response through expression of TH-specific chemokine receptors (5). However, Tregs only acquire aspects of TH cell programs and generally fail to secrete cytokines characteristic of these cells within healthy tissue (6). Thus, Tregs maintain a delicate balance between expression of transcriptional programs that mediate immune regulation and expression of specific pathways associated with TH cell differentiation.

In multiple human diseases, Tregs within inflamed tissue produce appreciable amounts of inflammatory cytokines. In psoriasis, a cutaneous inflammatory disease driven by dysregulated TH17 responses, Tregs in lesional skin produce increased interleukin-17 (IL-17) compared with Tregs in nonlesional skin and skin from normal healthy individuals (7, 8). Similarly, both lamina propria and peripheral blood mononuclear cells isolated from patients with Crohn’s disease have increased IL-17–producing Tregs relative to healthy controls (9, 10). In support of the latter, dysregulation of the TH17 program within Tregs has been shown to contribute to disease pathogenesis. In murine models of severe asthma and chronic arthritis, Treg instability leads to acquisition of proinflammatory TH17 cytokines, loss of FOXP3 expression, and increased disease severity (11, 12). Thus, dysregulation of TH cell programs within Tregs can negatively affect their ability to control inflammation.

Members of the tumor necrosis factor (TNF) receptor superfamily (TNFRSF) of costimulatory receptors (including, but not limited to, CD27, OX40, GITR, and 4-1BB) play critical roles in both the development and regulation of productive immune responses (13). Although TNFRSF members mediate similar cellular processes, comprehensive studies comparing redundant and distinct effects of specific receptors are lacking. In addition, many of these receptors were originally studied in the context of Teff function; however, it is now evident that many TNFRSF members are also expressed by Tregs (14). The function of many of these receptors on Tregs is poorly understood.

We have found that CD27 is highly expressed on Tregs that stably reside in human skin (8). CD27 plays important roles in Teff priming, expansion, survival, and differentiation (15). The only known ligand for CD27 is CD70, which is transiently expressed on dendritic cells, B cells, T cells, and natural killer cells after activation (1618). CD27 engagement by CD70 induces the recruitment of the TNF receptor–associated factor 2 (TRAF2) and TRAF5 adaptor proteins, which activate the c-Jun N-terminal kinase and nuclear factor κB signaling pathways (19). In γδ T cells, CD27 signaling is a thymic determinant of interferon γ (IFN-γ)– versus IL-17–producing subsets (20). In CD4+ Teffs, CD27 promotes the development of productive TH1 responses (21, 22). More recently, CD27 signaling has been reported to attenuate TH17 responses via epigenetic silencing of IL-17A and CCR6 in TH17 cells (23). On Tregs, high levels of CD27 expression correlate with increased survival and suppressive function. CD27 also plays a minor role in thymic differentiation of Tregs (2427). However, the function of this TNFRSF member on Tregs within peripheral tissues is currently unknown.

OX40 is another member of the TNFRSF that is highly expressed on Tregs (28). Signaling through OX40 occurs after engagement with OX40L, which is broadly expressed by multiple immune lineages (29). Similar to CD27, OX40 signaling promotes Teff survival, differentiation, and thymic Treg development and can attenuate TH17 differentiation (3033). However, the effects of OX40 signaling on Treg biology are poorly understood.

Given that CD27 and OX40 are highly expressed on Tregs and limit TH17 differentiation in Teffs, we hypothesized that signaling through these receptors would play a role in regulating the TH17 program in Tregs in inflamed tissues. In addition, we speculated that these pathways would have completely overlapping and redundant roles in Tregs. Consistent with our hypothesis, we found that both CD27 and OX40 limit expression of TH17-associated genes in Tregs; however, signaling through OX40 resulted in enhanced Treg accumulation when compared with the CD27 pathway. Our results elucidate distinct, partially redundant roles for members of TNFRSF in limiting TH17 differentiation in Tregs and promoting their accumulation in peripheral tissues.


Tregs in skin are poised to respond to TH17 inflammation

Within the gastrointestinal tract, IL-17–producing cells play a major role in maintaining barrier homeostasis, in part by promoting epithelial integrity (34, 35). A defined subset of Tregs within the gut expresses the TH17-associated transcription factor, RAR-related orphan receptor gamma 2 (RORγt), which allows them to regulate TH17 responses within this tissue (3638). In contrast, the skin contains relatively few IL-17–producing cells in the steady state, and increased numbers of these cells are observed during pathologic skin inflammation (39, 40). Because Tregs in the gut are constantly regulating TH17 responses and Tregs in the skin need to regulate these responses only in specific inflammatory contexts, we hypothesized that gut-resident Tregs would be more “TH17 skewed” than skin-resident Tregs. To test this hypothesis, we compared expression of the master TH17 transcription factor, RORγt, in Tregs within the colonic lamina propria, skin, and skin-draining lymph nodes (SDLNs) in healthy adult mice. Compared with the gut where >40% of Tregs express RORγt, only ~6% of Tregs within both murine skin and SDLNs expressed this transcription factor (Fig. 1A). RNA sequencing (RNA-seq) of Tregs and Teffs sorted from healthy murine and human skin confirmed that Tregs express very low Rorc transcript levels (Fig. 1B and fig. S1). Tissue-resident Tregs in both murine and human skin expressed high levels of other transcription factors involved in TH17 polarization (including Stat3, Rora, Irf4, and Batf), as well as receptors for TH17-polarizing cytokines (including Il6ra, Il6st, and Il21r) (Fig. 1B and fig. S1). However, skin-resident Tregs expressed little to no TH17-associated cytokines, including Il17a, Il17f, Il18, Il21, and Il22, as well as low levels of the proinflammatory cytokine receptor Il23r (Fig. 1B and fig. S1). These results suggest that Tregs in skin are not TH17 skewed in the steady state but instead are poised to be able to respond to TH17-inducing stimuli.

Fig. 1 Skin-resident Tregs differentiate toward IL-17–producing cells during TH17-mediated tissue inflammation.

(A) Percentage of RORγt-expressing Tregs (gated on Live CD45+CD4+FOXP3+ cells) in the colon, skin, and SDLNs of WT mice as quantified by flow cytometry. Results are from one experiment with n = 6 mice. (B) Tregs and Teffs were sort purified from normal healthy skin of FOXP3-GFP mice, and gene expression was quantified by whole-transcriptome RNA-seq. Heat maps of cytokines (top), transcription factors and cytokine receptors (middle), and Treg-specific genes (bottom) are shown. (C and D) The percentage of (C) RORγt- and (D) IL-17A–expressing Tregs in the skin of WT mice on days 0, 3, and 6 (d0, d3, and d6, respectively) after cutaneous C. albicans infection was quantified by flow cytometry. Results are from two replicate experiments with n = 2 to 6 mice per group. Data are means ± SEM. P values are determined using one-way ANOVA. **P < 0.01 and ****P < 0.0001.

To test whether Tregs can respond to TH17 inflammation, we infected wild-type (WT) mice cutaneously with Candida albicans, a pathogen known to drive potent TH17 responses (41). Tregs in the skin expressed increasing levels of RORγt with time after infection (Fig. 1C), which coincided with an increase in skin Treg production of IL-17A (Fig. 1D). These results suggest that Tregs in skin induce expression of RORγt and begin to secrete appreciable levels of IL-17A in response to TH17 inflammation in this tissue.

The TNFRSF members CD27 and OX40 are preferentially expressed on Tregs in skin

Tregs in tissues can express components of TH17 differentiation, and aberrant regulation of this process is thought to contribute to human disease pathology (8, 9, 11, 36, 38, 42). Mechanisms that regulate TH17 differentiation in Tregs are poorly understood. Recently, two members of the TNFRSF, CD27 and OX40, have been reported to attenuate Teff differentiation into the TH17 lineage (23, 33). TNFRSF members are often expressed by Tregs in addition to Teffs; however, it is currently unknown whether these receptors influence Treg function in tissues. We hypothesized that CD27 and OX40 limit TH17 differentiation in Tregs during tissue inflammation. To begin to test this hypothesis, we first quantified the expression of CD27 and OX40 on skin-resident Tregs. CD27 is expressed on both Tregs and Teffs in SDLNs (Fig. 2A). However, CD27 is preferentially expressed on skin-resident Tregs relative to skin-resident Teffs. In contrast, OX40 was preferentially expressed by Tregs in both the skin and SDLNs (Fig. 2B). We observed that CD27 and OX40 were coexpressed by Tregs within both the skin and SDLNs (fig. S2). High expression of both CD27 and OX40 on skin-resident Tregs suggests that these costimulatory receptors might preferentially influence Treg function in this tissue.

Fig. 2 The TNFR family members CD27 and OX40 are preferentially expressed by skin-resident Tregs.

Representative flow cytometry plots and quantification of percentages and MFI of (A) CD27 and (B) OX40 expression on Tregs and Teffs in healthy murine skin. Populations are pregated on Live CD45+CD4+ cells. Results are pooled from three independent experiments with n = 9 to 10 mice per experiment. Data are means ± SEM. P values are determined using paired Student’s t test. ****P < 0.0001.

CD27 and OX40 signaling regulates TH17 differentiation in Tregs in vitro

To test whether signaling through CD27 and OX40 attenuates TH17 differentiation in Tregs, we developed an in vitro assay in which expression of TH17-associated genes could be induced in these cells. Tregs (CD4+CD25+GFP+) were sorted from FOXP3-GFP reporter mice (43) to >97% purity and cultured with anti-CD3/CD28 activator beads with the TH17-polarizing cytokines IL-6, IL-1β, and IL-23. When compared with Tregs cultured with IL-2 alone, Tregs cultured under TH17-promoting conditions expressed increased levels of the TH17-associated cytokines IL-17A and IL-17F after phorbol 12-myristate 13-acetate (PMA)–ionomcyin restimulation (Fig. 3, A and B). No difference was observed for the TH1 cytokine, IFN-γ (Fig. 3C). Treg culture under TH17-polarizing conditions also resulted in increased expression of the TH17-associated chemokine receptor CCR6 and RORγt (Fig. 3, D and E).

Fig. 3 CD27 and OX40 signaling attenuate TH17 differentiation in Tregs in vitro.

Tregs were sort purified from FOXP3-GFP reporter mice and cultured with anti-CD3/anti-CD28–coated Dynabeads under either TH0- or TH17-polarizing conditions in the presence or absence of FcCD70 and/or an agonistic anti-OX40 monoclonal antibody (OX86). On day 6, Tregs were restimulated with PMA and ionomycin. Representative flow cytometry plots of (A) IL-17A, (B) IL-17F, (C) IFN-γ, (D) CCR6, (E) RORγt, (F) FOXP3, and (G) CD25 after gating on Live+CD4+FOXP3+ cells are shown and either percentages (A to D) or MFIs (F and G) are quantified. Data are representative of two to three independent experiments with technical replicates and graphs depict means ± SEM. P values are determined using one-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001.

Using this culture system, we assessed whether stimulation through CD27 and/or OX40 could influence the expression of TH17- or TH1-associated proteins. The addition of either FcCD70, a recombinant Fc-tagged CD70 that can engage and agonize CD27 signaling (44), or OX86, an agonistic antibody against OX40 (45), was able to suppress Treg expression of IL-17A, IL-17F, CCR6, and RORγt (Fig. 3, A, B, D, and E). Neither of these agonists influenced IFN-γ expression (Fig. 3C). Addition of either FcCD70 or OX86 did not affect Treg FOXP3 or CD25 expression, as assessed by mean fluorescence intensity (MFI) (Fig. 3, F and G), suggesting that signaling through these receptors did not primarily influence TH17 differentiation in Tregs by increasing Treg stability.

We observed a trend toward an additive suppressive effect when both FcCD70 and OX86 were added in combination. OX86 treatment induced greater reduction in expression of the TH17-assocaited molecules compared with FcCD70. Antibody-mediated cross-linking of receptors can result in increased receptor oligomerization and activation compared with recombinant ligands (46, 47). Therefore, discrepancy between FcCD70 and OX86 in these cultures may be due to fundamental differences in activating TNF receptors using recombinant ligands compared with agonistic antibodies.

To validate our findings with CD27 and OX40 agonists, we cultured Tregs in the presence of expanded and matured bone marrow–derived dendritic cells (BMDCs) as a source of natural CD70 and OX40 ligands (fig. S3A). Compared with culturing Tregs under TH0 conditions, culturing Tregs under TH17 conditions induced the expression of IL-17A (fig. S3B) but had no effect on IFN-γ (fig. S3C). FACS (fluorescence-activated cell sorting)–sorted WT, CD27−/−, or OX40−/− Tregs were cultured in the presence of BMDCs under TH17-promoting conditions. Compared with WT Tregs, both CD27−/− and OX40−/− Tregs expressed higher levels of IL-17A when cultured with matured BMDCs (fig. S3, D and F). Minimal difference was observed in IFN-γ production between the groups (fig. S3, E and G).

To assess how CD27 engagement on Tregs affects TH17-associated gene expression, we performed whole-transcriptome RNA-seq. Tregs were cultured under TH17-polarizing conditions in the presence or absence of FcCD70. After 6 days of culture, RNA was isolated for RNA-seq. Over four replicate experiments, after this culture period, there were only 18 differentially expressed genes between FcCD70-treated and untreated groups that had a false discovery rate of less than 0.05 (fig. S4, A and B). Consistent with our results at the protein level, transcripts for both IL-17A and RORγt were significantly reduced after FcCD70 treatment. Gene set enrichment analysis (GSEA) revealed that TH17-associated genes (48) were significantly enriched in untreated compared with FcCD70-treated samples (fig. S4C). Because CD27 signaling resulted in a decrease in RORγt at both the transcript and protein levels, we also used GSEA to determine whether the RORγt-regulated signature was attenuated after FcCD70 treatment. RORγt-driven genes (49) were significantly enriched in untreated compared with FcCD70-treated samples (fig. S4D). At the RNA level, Tregs expressed low levels of other TH17-assicated cytokines such as Gmcsf, Il22, and Il21, and no difference was detected upon FcCD70 treatment (table S1). Consistent with our findings at the protein level (Fig. 3, F and G), we did not see major differences in Treg-associated genes, including Foxp3, CD25, and CTLA-4 (table S2). This again suggested that CD27 engagement did not attenuate the TH17 program through increasing the expression of FOXP3 or other Treg-associated genes. Together, these results indicate that signaling through CD27 and OX40 regulates the TH17 differentiation program in Tregs and that engagement of CD27 most markedly affects RORγt and IL-17 expression in this assay.

CD27 and OX40 attenuate TH17 differentiation in Tregs in vivo

To determine whether the CD27 and OX40 pathways attenuate TH17 differentiation in Tregs in a cell-intrinsic fashion in vivo, we developed a Treg adoptive transfer model of cutaneous C. albicans infection (Fig. 4A). WT CD45.1+ Tregs and either CD27−/− or OX40−/− CD45.2+ Tregs were sort purified and combined with sort-purified CD45.1/CD45.2 WT Teffs at a 1:1:2 ratio and adoptively transferred into RAG2-deficient (RAG2−/−) recipients. Both Tregs and Teffs homeostatically proliferate in RAG2−/− hosts, with defined populations readily detected in skin 2 weeks after transfer. At this time point, mice were cutaneously infected with C. albicans according to a well-established model known to elicit potent TH17 responses (41, 50). Skin and SDLNs were harvested 7 days after infection (i.e., 21 days after transfer), and Treg accumulation and cytokine production was quantified by flow cytometry. In animals that received WT and CD27−/− Tregs, both of these populations were present at equal percentages within SDLNs (Fig. 4B). To determine whether CD27 suppressed TH17 differentiation in Tregs, we quantified RORγt expression and intracellular IL-17 production. Compared with WT Tregs, CD27−/− Tregs had increased expression of RORγt (Fig. 4C). In addition, CD27−/− Tregs produced increased levels of IL-17A, but not of IFN-γ (Fig. 4, D and E). Together, these data suggest that CD27 expression on Tregs limits RORγt and IL-17 expression in a cell-intrinsic fashion in vivo, with no obvious effect on cell accumulation.

Fig. 4 CD27 and OX40 attenuate TH17 differentiation in Tregs in a cell-intrinsic manner in vivo.

(A) WT CD45.1 Tregs, CD27−/− or OX40−/− CD45.2 Tregs, and WT CD45.1/CD45.2 Teffs were sort purified and cotransferred at a 1:1:2 ratio into RAG−/− recipients. Fourteen days later, mice were infected with C. albicans, and skin and SDLNs were harvested 7 days after infection. (B) Representative flow cytometry plots and quantification of percentages of WT (CD45.1) and CD27−/− (CD45.2) Tregs in the SDLNs after gating on Live+CD4+TCRβ+FOXP3+ cells. (C to E) Representative plots and quantification of (C) RORγt, (D) IL-17A, and (E) IFN-γ expression by WT or CD27−/− Tregs in the SDLNs. (F to I) Representative plots and quantification of (F) percentages of WT (CD45.1) and OX40−/− (CD45.2) Tregs and expression of (G) RORγt, (H) IL-17A, and (I) IFN-γ expression in Live+CD4+TCRβ+FOXP3+ cells in the SDLNs. Data are compiled from two independent experiments with n = 11 to 13 mice per group. Graphs depict mean ± SEM. P values are determined using a paired Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, P > 0.05.

To assess the role of OX40 signaling on Tregs, we cotransferred WT CD45.1+ Tregs and OX40−/− CD45.2+ Tregs with WT CD45.1/CD45.2 Teffs into RAG2−/− recipients followed by C. albicans infection. In contrast to CD27−/− Tregs, OX40−/− Tregs exhibited a severe defect in their ability to accumulate in the SDLNs relative to WT Tregs (Fig. 4F). Similar to CD27−/− Tregs, the remaining OX40−/− Tregs expressed higher levels of RORγt and IL-17A when compared with WT Tregs, with no difference in IFN-γ expression (Fig. 4, G to I).

To determine whether OX40 influences Treg accumulation in a noninfectious context, we assessed the effect of OX40 on Treg accumulation in the absence of C. albicans infection. WT or OX40−/− Tregs were transferred into RAG2−/− hosts, and Tregs were quantified in the SDLNs at days 5 and 15 after transfer. In this experiment, OX40−/− Tregs accumulated within the SDLNs at lower frequencies compared with WT Tregs (fig. S5), suggesting that OX40-mediated Treg accumulation is not dependent on C. albicans infection. These findings are consistent with previous reports demonstrating that OX40 is required for Treg survival in a noninfectious context (51, 52).

Together, these results demonstrate that on Tregs, signaling through either CD27 or OX40 resulted in cell-intrinsic suppression of IL-17A and RORγt expression. However, in contrast to CD27, OX40 played a nonredundant role in promoting Treg accumulation.

The CD27 and OX40 pathways synergize to attenuate TH17 differentiation in Tregs in vivo

To determine whether the CD27 and OX40 pathways have an additive effect in attenuating TH17 differentiation in Tregs in vivo, we crossed CD27−/− and OX40−/− mice to generate CD27−/−/OX40−/− double knockout (DKO) animals. WT CD45.2+ or DKO CD45.2+ Tregs were sort purified, combined 1:1 with sort-purified WT CD45.1+ Teffs, and adoptively transferred into RAG−/− recipients, followed by cutaneous C. albicans infection as described above (Fig. 5A). Unlike results observed in mice receiving Tregs deficient in either CD27 or OX40 (fig. S6), mice that received Tregs deficient in both of these receptors (i.e., DKO Tregs) had significantly more skin inflammation, as measured clinically by increased scaling and erythema and histologically by more pronounced epidermal hyperplasia and mononucelar cell infiltrate (Fig. 5, B to D). Compared with mice that received WT Tregs, mice that received DKO Tregs had significantly increased expansion of lymphocytes in SDLNs (Fig. 5E), further suggesting that DKO Tregs had a reduced capacity to regulate inflammation.

Fig. 5 Signaling through CD27 and OX40 synergize to attenuate TH17 differentiation in Tregs in vivo.

(A) Model of cutaneous C. albicans infection. WT or CD27−/−OX40−/− DKO CD45.2 Tregs were sort purified and adoptively transferred with CD45.1 Teffs into RAG−/− recipients. On day 14 after transfer, mice were infected cutaneously with C. albicans. (B) Representative histology of dorsal skin 7 days after infection. Scale bars, 100 μm. (C) Representative clinical images of dorsal skin inflammation 7 days after infection. (D) Quantification of disease based on blinded histologic scoring of epidermal hyperplasia and mononuclear cell infiltrate. (E) Absolute numbers of SDLNs cells as measured by a hemocytometer. (F and G) Quantification of percentages and absolute numbers of CD45.2+ (Tregs) and CD45.1+ (Teff) cells (gated on Live CD4+TCRβ+) in (F) the skin and (G) SDLNs by flow cytometry. WT or DKO Treg expression of (H and I) RORγt, (J and K) IL-17A, and (L and M) IFN-γ (gated on CD4+CD45.2+FOXP3+ cells) in the skin and SDLNs as quantified by flow cytometry. Data are combined from three independent experiments with n = 13 to 14 mice per group, and graphs depict mean ± SEM. P values are determined using unpaired Student’s t test. *P < 0.05; ***P < 0.001; ****P < 0.0001; ns, P > 0.05.

Similar to our findings using OX40−/− Tregs, DKO Tregs were present at significantly lower frequencies and absolute numbers in the SDLNs with a similar trend observed in skin (Fig. 5, F and G). Of the transferred CD45.2+ Tregs that remained, DKO Tregs expressed more than double the amount of RORγt in both the skin and SDLNs when compared with WT Tregs (Fig. 5, H and I). DKO Tregs also produced more than double the amount of IL-17A with no difference in IFN-γ production (Fig. 5, J and M). These results suggest that the CD27 and OX40 pathways synergize to attenuate TH17 differentiation in Tregs and that Tregs that are deficient in both of these TNFRSF members are not capable of effectively controlling skin inflammation.

CD27 expression inversely correlates with Treg production of IL-17 in diseased human skin

Similar to Tregs in murine skin, CD27 and OX40 are preferentially and highly expressed on Tregs in healthy human skin relative to Teffs (Fig. 6A and fig. S7). In multiple human chronic inflammatory diseases, Tregs within inflamed tissue have been reported to express increased levels of IL-17A compared with healthy controls (710). Furthermore, single-nucleotide polymorphisms within CD27 have been identified as susceptibility loci in psoriasis (53). Thus, we explored the relationship between IL-17 production and CD27 expression on tissue Tregs in two inflammatory skin diseases: psoriasis and hidradenitis suppurativa (HS).

Fig. 6 CD27 expression inversely correlates with Treg production of IL-17 in diseased human skin.

(A) Representative flow cytometry plots of CD27 on Tregs and Teffs (gated on Live CD45+CD3+CD4+) within healthy human skin and blood. Percentages and MFI of CD27 on skin-resident cells are quantified, and results are pooled from n = 21 healthy human skin biopsies. (B) Treg IL-17 production in healthy or lesional psoriatic (PSO) skin was quantified by flow cytometry after PMA/ionomycin restimulation [n = 22 (healthy) and n = 26 (PSO)]. (C) Representative plot and quantification of CD27 expression on Tregs and Teffs in psoriatic skin are shown. (D) Correlation between Treg IL-17 production and CD27 MFI in lesional psoriatic skin was quantified by flow cytometry. Lines represent paired data from the same sample (n = 10). Results are pooled from 10 patients with psoriasis. (E) Treg IL-17 production in biopsies from paired nonlesional (NL-HS) and lesional (L-HS) skin of patients with HS was quantified by flow cytometry (gated on Live CD45+CD3+CD4+FOXP3+ cells). Lines represent paired data from a single patient (n = 8). (F) Representative plot and quantification of CD27 expression on Tregs and Teffs in lesional HS skin. (G) Correlation between Treg IL-17 production and CD27 MFI in lesional skin biopsies from patients with HS as quantified by flow cytometry. Lines represent paired data from the same sample (n = 20). P values are determined using either an unpaired (A and B) or a paired (D to G) Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Under inflammatory conditions, recently activated Teffs can transiently express FOXP3 (54), yet transient FOXP3 expression tends to be lower compared with bona fide Tregs (55, 56). We therefore compared percentages and MFI of FOXP3 in healthy and diseased skin (fig. S8). There was a greater percentage of FOXP3+ cells within psoriatic skin compared with that within healthy skin, and the MFI of FOXP3 was also increased (fig. S8A). In HS, both FOXP3 percentages and MFIs were unchanged between lesional and adjacent nonlesional tissue (fig. S8B). Together, these data are consistent with our previous findings (8) and suggest that our definition of Tregs most likely represent true regulatory cells rather than Teffs that transiently express FOXP3.

Consistent with previous studies (7, 8), Tregs within lesional psoriatic skin biopsies express higher levels of IL-17A compared with Tregs in healthy skin (Fig. 6B). Most of the Tregs within psoriatic skin expressed CD27 (Fig. 6C). Treg expression of IL-17 was inversely correlated with CD27 expression levels as measured by MFI of staining (Fig. 6D).

HS is a highly inflammatory skin disease characterized by painful intertriginous follicular abscesses and fibrous tracts. Recent studies have shown that lesional biopsies from patients with HS have increased frequencies of TH17 cells compared with healthy controls (57, 58). Whether Tregs within inflamed HS tissue produce inflammatory cytokines has not been assessed. Thus, we quantified IL-17 production from Tregs in lesional and adjacent nonlesional skin obtained from the same patient with HS. When compared with nonlesional skin, Tregs within lesional HS skin produced significantly more IL-17A (Fig. 6E) and expressed higher levels of CD27 compared with Teffs (Fig. 6F). Consistent with results observed in psoriatic skin, Treg expression of IL-17 was inversely correlated with CD27 expression in HS skin (Fig. 6G). Together, these results demonstrate that Tregs produce IL-17 in inflamed human skin and that expression of this cytokine inversely correlates with CD27 expression on these cells.


It is becoming increasingly appreciated that Tregs co-opt Teff transcriptional programs for their optimal suppressive function (24). Tregs and TH17 cells share a considerable amount of both phenotypic and functional plasticity, and excessive expression of TH17-associated genes in Tregs can affect disease pathogenesis (11, 12, 59). Here, we demonstrate that two TNFRSF members, CD27 and OX40, are highly expressed by skin-resident Tregs. Using multiple approaches, we demonstrate that both CD27 and OX40 synergize to limit TH17 differentiation in these cells without appearing to affect Treg stability. Furthermore, OX40, but not CD27, influenced Treg accumulation in vivo. This study reveals both overlapping and nonredundant roles for TNFRSF receptors in influencing Treg function.

TNFRSF receptors have been shown to play a variety of roles in T cell biology, influencing cell survival, expansion, and differentiation. More recently, both CD27 and OX40 have been reported to regulate immune responses by attenuating the pathogenicity of TH17 cells (13, 23, 33). Despite these studies, the roles that these receptors play in peripheral Treg function are poorly understood. Because TNFRSF receptors play such diverse roles in T cell biology and Tregs are functionally distinct from Teffs, it cannot be assumed that these receptors will affect Tregs in the same way that they affect other T cell subsets. Our results demonstrate that, in the context of cutaneous TH17 inflammation, OX40 and CD27 signaling in Tregs attenuates the expression of TH17-associated genes.

Skin-resident Teffs express significantly lower levels of CD27 and OX40 compared with Tregs in the steady state and in inflamed skin. The reason that Tregs maintain high CD27 and OX40 expression while Teffs lack and/or down-regulate expression in skin is unclear. It is interesting to speculate that the lack of CD27 and OX40 expression enables Teffs to be poised to rapidly mount a TH17 response upon barrier breach or infection. In contrast, high expression of these receptors on skin-resident Tregs inhibits these cells from converting into IL-17–producing “TH17-like” cells in the context of TH17-mediated tissue inflammation. The mechanisms that regulate the differential expression patterns of TNFRSF receptors on Tregs versus Teffs remain to be elucidated.

One of the major roles for TNFRSF receptors is promoting T cell survival. OX40 had a significant effect on Treg accumulation in vivo. Compared with WT Tregs, OX40-deficient Tregs accumulated at much lower frequencies in both the SDLNs and skin in the steady state and after C. albicans infection. These findings are consistent with previous studies demonstrating that OX40-deficient Tregs exhibit defective survival in vivo (51, 52, 60). Because Tregs use OX40 for thymic development (32, 52), it is possible that defects in Treg accumulation in peripheral tissues are secondary to a cell-intrinsic developmental defect, resulting in less “biologically fit” cells exiting the thymus of these mice. When both OX40 and CD27 were deleted from Tregs, these cells had lower accumulation in both the skin and SDLNs (Fig. 5, F and G). This is most likely secondary to lack of OX40 signaling. A much more pronounced Treg survival defect was observed when only OX40 was deleted in Tregs (Fig. 4G). However, mice that receive OX40-deficient Tregs failed to develop signs of heightened inflammation compared with mice given WT Tregs (fig. S8). Thus, our data suggest that the increase in skin inflammation observed when Tregs lacked both CD27 and OX40 cannot be solely attributed to reduced Treg survival and that this is most likely secondary to the pronounced TH17 skewing observed in these cells in addition to reduced numbers.

In contrast to OX40, we observed that CD27 had no role in Treg accumulation in this study. This finding suggests that there are fundamental differences between CD27 and OX40 signaling on Tregs. This was somewhat surprising given that there are studies demonstrating that CD27 promotes Teff survival (6163). Furthermore, similar to OX40 (32, 52), CD27 also plays a minor role in promoting survival during Treg development within the thymus (27). Rather than ruling out a role for CD27 in Treg survival, our findings further illustrate that the role of any given TNFRSF receptor is highly contextual and highlight the complexity of these pathways in T cell biology.

Previous studies have demonstrated that CD27 and OX40 attenuate TH17 differentiation in Teffs via epigenetic silencing of the IL-17A/F and CCR6 loci, independent of RORγt (23, 33). In contrast, we found that CD27 and OX40 signaling in Tregs resulted in decreased RORγt expression. GSEA of our RNA-seq dataset comparing Tregs under TH17-polarizing conditions revealed that RORγt response genes were significantly reduced upon signaling through CD27. Thus, our results suggest that engagement of the CD27 pathway in Tregs reduces both RORγt levels and the expression of genes driven by this transcription factor, revealing what may be a fundamental difference between CD27-mediated signaling in Tregs compared with Teffs.

We found that CD27 expression was inversely correlated with Treg IL-17 production in lesional skin biopsies from patients with psoriasis and HS. This inverse correlation has also been observed for Teffs and γδ T cells in mice (20, 64). It has also been reported that human Tregs that differentiate into IL-17–producing cells express lower levels of CD27 (65). In psoriasis, genome-wide association studies have identified CD27 as a susceptibility locus (53). While Teffs within psoriatic skin are major producers of IL-17 (8), we observed that CD27 is not expressed by most of the Teffs within diseased tissue. It is interesting to speculate that defective signaling through CD27 (and/or OX40) in Tregs plays a role in human disease by failing to control the TH17 differentiation pathway in these cells. While the exact mechanisms by which these pathways might influence Treg IL-17 production in human disease remain unclear, there are several possibilities. First, CD27 and/or OX40 might be down-regulated on a subset of Tregs during skin inflammation. We did not observe a decrease in CD27 expression on Tregs in either psoriasis or HS, so it seems unlikely that this is the explanation. Another scenario is that there is a preexisting population of Tregs that express low levels of CD27 and/or OX40, rendering this population more susceptible to differentiate into TH17-like cells in human disease. A final possibility is that, rather than differences in receptor expression, the ligands for these receptors, CD70 and OX40L, are differentially expressed during disease and that the lack of ligand-receptor engagement renders Tregs more susceptible to IL-17 production.

Whereas CD27 and OX40 are expressed constitutively by skin-resident Tregs, signaling through these receptors is entirely dependent on engagement by their ligands. Both CD70 and OX40L can be expressed by a variety of immune cell types (29, 66). In the gut, there is a population of antigen-presenting cells (APCs) that constitutively express CD70 (67), but it is unclear whether an analogous population of APCs exists in skin. Langerhans cells that reside in the epidermis can promote Treg function and also play a critical role in TH17 responses to C. albicans (41, 68, 69). Langerhans cells have been shown to express CD70 after viral infection and OX40L in response to ultra violet blue (UVB) irradiation (70, 71). It is currently unknown whether CD70 and OX40L are expressed on Langerhans cells, or other APC populations, during TH17-mediated skin inflammation. Future studies are necessary to elucidate where and when these ligands are expressed in both healthy and diseased skin.

We studied the functional role of CD27 and OX40 signaling in attenuating TH17 differentiation in Tregs in a cutaneous C. albicans inflammation model. One of the limitations of our study is that we have not assessed the role of these receptors in Treg function in other models of inflammation (cutaneous and extracutaneous) or cancer. In addition, it is unclear whether other TNFRSF members synergize with CD27 and/or OX40 to suppress TH17 differentiation in Tregs.

A major outstanding question is whether cytokine-producing Tregs in inflamed tissues contribute to the pathogenesis of human disease. Do these cells retain suppressive capacity and are simply unable to control robust tissue inflammation? Are they unable to regulate because of cell-intrinsic defects in their suppressive function or do they actively contribute to disease by secreting cytokines known to mediate pathology? The answers to these questions are currently unknown and are difficult to discern in humans. Nevertheless, the results presented herein demonstrate that the TNFRSF receptors CD27 and OX40 can synergize to limit Treg differentiation into IL-17–producing cells. Future studies will be needed to definitively elucidate the roles of these cells in disease pathogenesis and evaluate whether targeting TNFRSF receptor pathways therapeutically can enhance Treg function in the setting of chronic tissue inflammation.


Study design

This study aimed to define the roles of the TNFRSF receptors CD27 and OX40 in Treg function during TH17 inflammation. To do this, we used a combination of in vitro and in vivo approaches. When hematoxylin and eosin images were scored, scoring was performed by blinded observers on randomized samples. All other experiments were quantified without blinding or randomization. The number of replicates for each experiment is indicated in the figure caption.


All mice were bred on a C57BL/6 background and maintained at the University of California, San Francisco (UCSF) in compliance with institutional guidelines. CD27−/− mice were generously provided by S. Schoenberger (with permission from J. Borst). C57BL/6 (WT), FOXP3-GFP (FOXP3tm2Tch), CD45.1, RAG2−/−, and OX40−/− strains were purchased from the Jackson Laboratory. DKO (CD27−/− /OX40−/− DKO) mice were generated by crossing CD27−/− mice to OX40−/− mice. All animals were socially housed in a 12-hour light/12-hour dark cycle. Animals 6 to 12 weeks of age were used in experiments, and all experiments were performed with sex- and age-matched controls.

Adoptive transfer model of C. albicans infection

C. albicans (derived from strain SC5314) was donated by D. Kaplan (University of Pittsburgh) (41). Protocols were approved by the UCSF Institutional Animal Care and Use Committee. In some experiments, 14 days before infection, 8- to 12-week-old RAG−/− mice were reconstituted with sort-purified Tregs (3 × 105) and Teffs (3 × 105). Cutaneous infection was performed as previously described (41). The day before infection, mice were anesthetized, dorsal skin was shaved with an electric razor, and depilatory cream was applied to the shaved area for 30 s before wiping clean. C. albicans was grown in YPAD medium at 30°C in a shaking incubator until the optical density at 600 nm was 1.5 to 2.0. C. albicans was washed in sterile phosphate-buffered saline (PBS) and resuspended at 4 × 109/ml. Upon infection, the stratum corneum was removed with 15 strokes of 220-grit sandpaper (3M), and 50 μl (2 × 108 C. albicans) was evenly applied to the skin with a sterile, prewetted cotton swab. Mice were harvested 7 days after infection. Infected mice were housed in a UCSF BSL2 facility according to National Institutes of Health (NIH) guidelines.

Human skin specimens

Studies using human tissue were approved by the UCSF Committee on Human Research and by the Institutional Review Board of UCSF. Normal human skin was obtained from patients at UCSF undergoing elective surgeries in which skin was discarded as a routine part of the procedure. Psoriatic skin biopsies were obtained from patients with active clinical disease (study number 10-02830). HS skin biopsies were obtained from patients with active clinical disease, and matching nonlesional skin was taken 10 cm away from active lesions (study number 16-19770). Skin samples were digested and processed as described previously (8). All patients provided written informed consent before biopsies.

Fluorescence-activated cell sorting

Spleen and lymph nodes (mesenteric and SDLNs) were harvested from 6- to 10-week-old mice. CD4+ T cells were enriched by negative selection using the EasySep CD4+ T Cell Isolation kit (STEMCELL Technologies, catalog no. 19852). Enriched cells were then stained for surface antigens. Tregs (live CD4+CD25+) and Teffs (live CD4+CD25) were sorted (>97% purity) on a FACSAria II (BD Biosciences). In some experiments, Tregs were sorted from FOXP3-GFP reporter mice (live CD4+CD25+GFP+) (43).

Tissue processing for flow cytometry

Skin and SDLNs were harvested for flow cytometry analysis. SDLNs (harvested from the axillary, brachial, and inguinal lymph nodes) were isolated and mashed over a sterile wire mesh to generate a single-cell suspension. Skin was harvested, lightly defatted, minced with scissors, and resuspended in a digestion mix with collagenase XI (2 mg/ml; Sigma-Aldrich, catalog no. C9407), hyaluronidase (0.5 mg/ml; Sigma-Aldrich, catalog no. H3506), and deoxyribonuclease (0.1 mg/ml; Sigma-Aldrich, catalog no. DN25) in RPMI 1640 with 2% fetal calf serum and 1% penicillin-streptomycin. The skin was digested for 45 min in a 37°C shaking incubator at 225 rpm. Digested skin was washed and vortexed for 15 s before filtering through a 100-μm strainer. Cell counts were performed using a Nucleocounter NC-200 (ChemoMetec). Cells were restimulated ex vivo with Cell Stimulation Cocktail (Tonbo Biosciences, catalog no. TNB-4975) for 4 hours before staining for flow cytometry analysis.

Antibodies and flow cytometry

Cells were stained for surface antigens and a live/dead marker (Ghost Dye Violet 510, Tonbo Biosciences) in FACS buffer (PBS with 2% fetal calf serum and 1% penicillin-streptomycin) for 30 min at 4°C. To stain for intracellular markers, cells were fixed and permeabilized using the FOXP3-staining buffer kit (eBioscience). Antibodies used are listed in table S3. Samples were run on a Fortessa (BD Biosciences) in the UCSF Flow Cytometry Core. FlowJo software (FlowJo LLC) was used to analyze flow cytometry data. Gating strategies used for murine and human Tregs and Teffs are shown in fig. S9.

In vitro TH17 culture

FACS-sorted Tregs (1 × 105) were cultured in 96-well flat bottom plates in the presence of 1 × 105 Mouse T-Activator CD3/CD28 Dynabeads (Thermo Fisher Scientific, catalog no. 11456D). Tregs were cultured under TH0 conditions in the presence of IL-2 (100 U/ml). Polarization under TH17 conditions occurred with IL-6 (25 ng/ml), transforming growth factor–β (2 ng/ml), anti-INF-γ (5 μg/ml), and anti-IL-4(5 μg/ml) continually in culture with IL-1β (10 ng/ml) and IL-23 (10 ng/ml) added at day 3. Cells were cultured in the presence and/or absence of FcCD70 (0.5 mg/ml; Sinobiological, catalog no. 51129-M04H) and/or OX86 (5 μg/ml; Bio X Cell, catalog no. BE0031). Cells were restimulated on day 6 for cytokine analysis by flow cytometry. In some experiments, Tregs were sorted (CD4+CD25+FOXP3-GFP+) on day 6, and cell pellets were flash frozen in liquid nitrogen for RNA-seq.

RNA sequencing

Frozen pellets were sent to Expression Analysis, Quintiles (Morrisville, NC). All sample preparation and processing for sequencing was performed by expression analysis. RNA was isolated using QIAGEN RNeasy Spin Columns, and quality was checked using an Agilent Bioanalyzer Pico Chip. Total RNA samples were converted to complementary DNA (cDNA) libraries using the Illumina TruSeq Stranded mRNA sample preparation kit. Using an Illumina RNA-seq, cDNA was sequenced to a 25 M read depth. Reads were aligned to Ensembl mg GRCm38.p4 reference genome using TopHat (version 2.0.12). After alignment, SAM files were generated using SAMtools. Htseq-count (0.6.1p1, with union option) was used to obtain read counts. Differentially expressed genes between paired samples were determined using the R/Bioconductor package DESeq2.


Significance was determined using either two-tailed unpaired Student’s t test (for measuring differences between separate groups), paired Student’s t test (for measuring significance between populations within the same patient/mouse), or one-way analysis of variance (ANOVA) (for multiple comparisons) in GraphPad Prism Software. All in vivo experiments were performed with at least two to three independent experimental cohorts. The number of mice per group is annotated in the figure legends, and mean values are visually depicted with error bars representing the SEM. P values correlate with symbols as follows: ns, not significant (P > 0.05); *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. No animals were excluded from statistical analysis.


Fig. S1. Human skin-resident Treg expression of TH17-associated genes.

Fig. S2. CD27 and OX40 coexpression on Tregs.

Fig. S3. Deletion of CD27 or OX40 on Tregs results in increased IL-17 expression in vitro.

Fig. S4. CD27 signaling suppresses Treg expression of TH17-associated genes.

Fig. S5. OX40 is important for Treg accumulation in adoptive transfer model before C. albicans infection.

Fig. S6. Deletion of either CD27 or OX40 in Tregs has minimal effect on skin inflammation.

Fig. S7. OX40 is preferentially expressed on Tregs in healthy human skin.

Fig. S8. Percentages and MFI of FOXP3 in Tregs in human disease.

Fig. S9. Gating strategy for murine and human Tregs and Teffs.

Table S1. Effects of CD27 engagement on TH17 cytokines.

Table S2. Effects of CD27 engagement on signature Treg genes.

Table S3. Antibodies.

Table S4. Raw data.


Acknowledgments: We thank C. Benetiz and Y. Hu for assistance with animal husbandry. Histology was processed with assistance from the UCSF Mouse Pathology Core. Funding: This work was primarily funded by the following grants to M.D.R.: NIH K08-AR062064, Burroughs Wellcome Fund CAMS-1010934, NIH DP2-AR068130, NIH R21-AR066821, a National Psoriasis Foundation Translational Grant, and a Dermatology Foundation Stiefel Scholar Award in Autoimmune and Connective Tissue Diseases. Flow cytometry data were generated and supported by a Diabetes Research Center grant (NIH DK063720) to the UCSF Parnassus Flow Cytometry Core. HS biopsy analysis was supported in part by AbbVie’s funding contribution to the study. Author contributions: M.D.R. and K.A.R. conceived this project, designed the study, and wrote the manuscript. K.A.R. performed experiments and performed statistical analysis of data. B.Z., M.M.L., P.M.S., D.B., E.H., T.C.S., and S.B. assisted with experiments and data generation. M.M.L. and H.B.N. oversaw the HS study. W.L. oversaw the psoriasis study. Competing interests: M.D.R. is the founder of TRex Bio, a co-founder of Sitryx, and a consultant with Celgene. K.A.R. is an employee at TRex Bio. Data and materials availability: RNA-seq data have been deposited in the National Center for Biotechnology Information and are available from the Gene Expression Omnibus under accession numbers GSE121794 and GSE121795.

Stay Connected to Science Immunology

Navigate This Article