Abstract
Regulatory T cells (Treg) are critical for maintaining self-tolerance and immune homeostasis, but their suppressive function can impede effective antitumor immune responses. FOXP3 is a transcription factor expressed in Tregs that is required for their function. However, the pathways and microenvironmental cues governing FOXP3 expression and Treg function are not completely understood. Herein, we report that YAP, a coactivator of the Hippo pathway, is highly expressed in Tregs and bolsters FOXP3 expression and Treg function in vitro and in vivo. This potentiation stemmed from YAP-dependent upregulation of activin signaling, which amplifies TGFβ/SMAD activation in Tregs. YAP deficiency resulted in dysfunctional Tregs unable to suppress antitumor immunity or promote tumor growth in mice. Chemical YAP antagonism and knockout or blockade of the YAP-regulated activin receptor similarly improved antitumor immunity. Thus, we identify YAP as an unexpected amplifier of a Treg-reinforcing pathway with significant potential as an anticancer immunotherapeutic target.
Significance: Tregs suppress antitumor immunity, and pathways supporting their function can be novel immunotherapy targets. Here, the selective expression of YAP by Tregs, its importance for their function, and its unexpected enhancement of pro-Treg Activin/SMAD signaling are reported, as are validations of potential cancer-fighting antagonists of YAP and its regulatory targets. Cancer Discov; 8(8); 1026–43. ©2018 AACR.
This article is highlighted in the In This Issue feature, p. 899
Introduction
Regulatory T cells (Treg) play critical roles in promoting immunologic self-tolerance and immune homeostasis by suppressing aberrant or excessive immune responses that could give rise to autoimmune diseases (1). However, their ability to dampen the activation of other leukocytes can also pose a major barrier to effective antitumor immunity and the sterile cure of chronic infections (2). The signature forkhead family transcription factor FOXP3 anchors the gene expression profile that is responsible for the characteristic suppressive function of Tregs. Clearly demonstrating the importance of this factor, mutations to the gene encoding FOXP3 can lead to fatal autoimmune disorders in Scurfy mice and in human patients with IPEX alike (3, 4). Despite the undeniable importance of FOXP3 for Treg function and immune control, our grasp of the factors and mechanisms governing its expression remains incomplete.
The signaling pathways triggered in response to certain cytokines (e.g., IL2 and TGFβ) can be critical for induction and maintenance of FOXP3 expression in Tregs (5). TGFβ potently induces FOXP3 expression in vitro and in vivo through activation of SMAD signaling molecules, critical facilitators and regulators of TGFβ-initiated signaling events and downstream gene activation (6, 7). TGFβ signaling has also been reported to be critical for maintaining FOXP3 expression and Treg function (8, 9). Likewise, SMAD2 and SMAD3 are also apparently needed for the optimal phenotypic stability of Tregs (10). Importantly, mechanisms for the augmentation or amplification of TGFβ/SMAD signaling in Tregs can stabilize or enhance the suppressive function of these cells (11) and may be crucial determinants of Treg performance in a variety of microenvironmental niches.
YAP is a transcriptional coactivator that developmentally regulates organ size (12, 13). YAP is frequently elevated in a number of cancer types such as lung, colorectal, ovarian, liver, and prostate cancers, where it acts as a powerful tumor promoter, and its activation is a frequent event in tumor progression (14). The Hippo pathway is believed to be the major regulator of YAP nuclear localization, activity, and tumorigenic potential (15–17). However, the physiologic role of YAP in the immune system is unknown.
Unexpectedly, we found YAP to be highly expressed by Tregs. In this report, we characterize the role of YAP in these important cellular mediators of immune control. Our studies revealed that in the absence of YAP, Tregs failed to suppress immune activation in vitro as well as in vivo. We also found that YAP potentiates the signaling events triggered by dimeric members of the TGFβ cytokine superfamily known as activins by activating expression of a key signaling component of the activin receptor complex. Interestingly, we found that not only is this signaling axis active in Tregs, it could also effectively amplify TGFβ/SMAD signaling and the promotion of Treg differentiation and function. Moreover, disrupting this YAP/activin/SMAD axis dramatically slowed the growth of tumors in mice, including a highly aggressive melanoma model. This experimental treatment also enhanced the antitumor efficacy of an antitumor vaccine, suggesting that the targeting of this YAP/activin/SMAD axis can be used to improve anticancer immunotherapy efficacy.
Results
YAP Expression Is Induced by T Cell–Receptor Signaling, Is Highly Expressed by Tregs, and Supports Their Function
YAP is a transcriptional coactivator known for its role in the Hippo signaling pathway (13). As such, its importance in tumorigenesis and organ size determination is well recognized (14). However, little is known about the role of the Hippo pathway and YAP in immune cells. Reports of cross-talk between the Hippo and TGFβ signaling pathways (18, 19) led us to speculate that elements of the former may have a role in the mechanisms governing immune activation and tolerance.
We therefore screened YAP expression across different subsets of murine CD4+ T cells in order to assess the likelihood that Hippo signaling plays a role in these functionally distinct T-cell lineages. Little to no Yap mRNA was detected in naïve CD4+ T cells, but, notably, YAP expression was uniquely induced during the early stages of the in vitro induced (iTreg) differentiation. Meanwhile, other T effector subsets (Th0, Th1, Th2, and Th17 cells) failed to markedly upregulate Yap mRNA (Fig. 1A). Interestingly, transient Yap message accumulation was noted during Th17 skewing. However, 12 hours after stimulation, Yap transcript levels returned to baseline in these T cells (Fig. 1A; Supplementary Fig. S1A). Importantly, considerable levels of YAP protein were found in cells of the iTreg subset and not other Thelper lineages (Fig. 1B). Human Tregs isolated from the peripheral blood of healthy donors also displayed higher levels of Yap mRNA than their conventional CD4+ (non-Treg) counterparts (Fig. 1C). These results implicate YAP as a transcription factor preferentially expressed by developing and established Tregs of mice and humans.
Expression of YAP mRNA and protein by Thelper subsets. Naïve CD4+ T cells (CD4+ CD25−CD62L+) were isolated from the spleen and lymph nodes of WT C57BL/6 mice and activated under polarizing conditions to generate the indicated Thelper subset. The cells were harvested at different time points, and mRNA or protein levels of YAP were assessed by (A) qRT-PCR and (B) western blot. C, Human Tregs (CD3+/CD4+/CD8−/CD25hi/CD127lo/CD39+) and non-Treg CD4+ T cells were obtained from the peripheral blood of healthy donors by FACS after Ficoll–Paque PLUS gradient centrifugation and magnetic bead enrichment of CD4+ T cells. Yap mRNA was measured by qRT-PCR. For A and B, shown are representative findings from at least 3 independent experiments (mean ±SEM of triplicates for A). For C, mean expression of Yap mRNA is shown for 10 healthy human donor samples. The data were analyzed using the Student t test and considered significant if *, P < 0.05; **, P < 0.01; ****, P < 0.001.
Expression of YAP mRNA and protein by Thelper subsets. Naïve CD4+ T cells (CD4+ CD25−CD62L+) were isolated from the spleen and lymph nodes of WT C57BL/6 mice and activated under polarizing conditions to generate the indicated Thelper subset. The cells were harvested at different time points, and mRNA or protein levels of YAP were assessed by (A) qRT-PCR and (B) western blot. C, Human Tregs (CD3+/CD4+/CD8−/CD25hi/CD127lo/CD39+) and non-Treg CD4+ T cells were obtained from the peripheral blood of healthy donors by FACS after Ficoll–Paque PLUS gradient centrifugation and magnetic bead enrichment of CD4+ T cells. Yap mRNA was measured by qRT-PCR. For A and B, shown are representative findings from at least 3 independent experiments (mean ±SEM of triplicates for A). For C, mean expression of Yap mRNA is shown for 10 healthy human donor samples. The data were analyzed using the Student t test and considered significant if *, P < 0.05; **, P < 0.01; ****, P < 0.001.
Because YAP is a major component of the Hippo pathway, we assessed levels of several Hippo signaling factors known to be upstream of the transcription factor across T-cell subsets to determine if these are also expressed preferentially by Tregs. Interestingly, we found that LATS1/2 and MST1/2, unlike YAP, were not upregulated by iTreg-skewing conditions (Supplementary Fig. S1B). These findings suggest that unlike other Hippo pathway factors, YAP is uniquely upregulated in developing Tregs, and they imply a role for YAP in the biology of these cells outside of its traditional role.
In order to dissect the potential role of YAP in the biology of CD4+ T cells, including Tregs, we crossed Yapfl/fl mice to CD4-cre transgenic mice to generate animals with a T cell–specific deletion of Yap. These conditional knockout mice (Yap cKO) developed normally without apparent defects in T-cell development or peripheral immune cell populations (Supplementary Fig. S2). Additionally, no obvious spontaneous immune pathologies were noted in these mice. Likewise, the lung, kidney, liver, small intestine, and stomach of mice with Treg-specific YAP deficiency (generated by crossing Yapfl/fl mice to FOXP3Cre+ transgenic mice) appeared comparable to wild-type (WT) littermates (Supplementary Fig. S3). We used both strains to assess the impact of YAP deficiency on Thelper cytokine production and lineage commitment.
The effects of YAP deficiency on CD4+ T-cell subsets. A–C, Naïve CD4+ T cells (CD4+ CD25−CD62L+) were isolated from WT Yapfl/fl CD4Cre− or Yapfl/fl CD4Cre+ (Yap cKO) mice (n = 5/group/experiment) and were activated under the indicated polarizing conditions for 4 days. The cells were harvested and signature cytokines and transcription factors for each Th subset were assessed by flow cytometry and qRT-PCR. Shown in A and B are representative flow cytometry results and the mean percentage of cells ± SEM from at least 3 independent experiments. C, Relative expression level of Il17a transcript during the early stages of Th17 cell differentiation for WT and Yap cKO derive cells ± SEM. D, iTreg differentiation of WT (Yapfl/fl FOXP3Cre−), and FOXP3Cre-driven Yap knockout mice (Yapfl/fl FOXP3Cre+). Naïve CD4+ T cells were isolated from the indicated mice as above before activation in the presence of IL2 and varying concentrations of TGFβ. Treg differentiation was assessed by intracellular staining for FOXP3 and flow cytometry analysis. Shown are representative histograms (left) and the mean fluorescence intensity (MFI) of FOXP3 staining was found in at least 3 independent experiments; average MFI ± SEM are shown. E, The suppressive function of WT or Yap cKO-derived Tregs (CD4+ CD25hi T cells FACS isolated from lymph node and spleen cell suspensions) was determined using an in vitro suppression assay. Naïve CD4+ T cells (responders) and Tregs were isolated from the indicated mice (n = 5/group/experiment). WT responders were prestained with CFSE and cocultured with WT and Yap cKO-derived Tregs at the indicated ratios. The cultures were activated with anti-CD3/anti-CD28–conjugated beads at a cell-to-bead ratio of 1:1. The percentage of proliferating (CFSElo) responder cells in each culture was determined by flow cytometry. Shown are representative histograms (top) and the mean percentages of proliferating cells ± SEM over at least 3 independent experiments (bottom). For A-E, significant differences were determined by the Student t test (*, P < 0.05; **, P < 0.02; ***, P < 0.002; ns, not significant).
The effects of YAP deficiency on CD4+ T-cell subsets. A–C, Naïve CD4+ T cells (CD4+ CD25−CD62L+) were isolated from WT Yapfl/fl CD4Cre− or Yapfl/fl CD4Cre+ (Yap cKO) mice (n = 5/group/experiment) and were activated under the indicated polarizing conditions for 4 days. The cells were harvested and signature cytokines and transcription factors for each Th subset were assessed by flow cytometry and qRT-PCR. Shown in A and B are representative flow cytometry results and the mean percentage of cells ± SEM from at least 3 independent experiments. C, Relative expression level of Il17a transcript during the early stages of Th17 cell differentiation for WT and Yap cKO derive cells ± SEM. D, iTreg differentiation of WT (Yapfl/fl FOXP3Cre−), and FOXP3Cre-driven Yap knockout mice (Yapfl/fl FOXP3Cre+). Naïve CD4+ T cells were isolated from the indicated mice as above before activation in the presence of IL2 and varying concentrations of TGFβ. Treg differentiation was assessed by intracellular staining for FOXP3 and flow cytometry analysis. Shown are representative histograms (left) and the mean fluorescence intensity (MFI) of FOXP3 staining was found in at least 3 independent experiments; average MFI ± SEM are shown. E, The suppressive function of WT or Yap cKO-derived Tregs (CD4+ CD25hi T cells FACS isolated from lymph node and spleen cell suspensions) was determined using an in vitro suppression assay. Naïve CD4+ T cells (responders) and Tregs were isolated from the indicated mice (n = 5/group/experiment). WT responders were prestained with CFSE and cocultured with WT and Yap cKO-derived Tregs at the indicated ratios. The cultures were activated with anti-CD3/anti-CD28–conjugated beads at a cell-to-bead ratio of 1:1. The percentage of proliferating (CFSElo) responder cells in each culture was determined by flow cytometry. Shown are representative histograms (top) and the mean percentages of proliferating cells ± SEM over at least 3 independent experiments (bottom). For A-E, significant differences were determined by the Student t test (*, P < 0.05; **, P < 0.02; ***, P < 0.002; ns, not significant).
To this end, we isolated naïve CD4+ T cells from Yap cKO and WT mice for activation under different helper CD4+ T cell (Th) polarizing conditions for 72 hours. Yap cKO CD4+ T cells express moderately higher levels of IL2 and IFNγ upon unbiased activation (Th0 conditions; Fig. 2A). Yap cKO CD4+ T cells also express a greater amount of IL17A than WT CD4+ T cells under Th17 polarizing conditions (Fig. 2B), and, consistently, Yap cKO CD4+ T cells expressed higher levels of Il17a mRNA than WT cells (Fig. 2C). A modest decrease in FOXP3+ cells was also seen in Yap cKO-derived T cells cultured under Th17 conditions (Fig. 2B). These observations, coupled with our earlier discovery that YAP is upregulated in iTregs, led us to suspect that YAP positively affects the generation of iTregs in vitro over other CD4+ T-cell fates. In line with this, the percentages of FOXP3+ cells induced from naïve Yap cKO T cells activated under iTreg skewing conditions were modestly, yet significantly, lower than those seen in polarized WT CD4+ T cells (Supplementary Fig. S4). Naïve CD4+ T cells isolated from Yapfl/fl FOXP3Cre+ mice were also consistently less able to upregulate FOXP3 than WT controls in response to activation and various concentrations of the Treg-promoting cytokine TGFβ. Here, YAP deficiency specifically in T cells having already “turned on” FOXP3 expression reduced the intensity of signal for the Treg transcription factor (Fig. 2D). Taken together, these findings suggest that YAP likely plays an important role in the initiation or maintenance of Treg differentiation.
In addition to FOXP3 induction, we also hypothesized that YAP might contribute to the suppressive function of Tregs as well. Indeed, an in vitro suppression assay showed that whereas WT Tregs readily dampened the proliferation of naïve T cells, Yap cKO Tregs were much less effective suppressors (Fig. 2E). In all, these findings implicate YAP as a Treg-associated factor with a role in both the generation and function of these cells.
YAP-Deficiency Enhances Anti-Melanoma Immunity
Although Tregs are necessary to maintain immune homeostasis, they pose an obstacle in mounting effective antitumor immune responses, and their suppressive function dampens the efficacy of anticancer immunotherapies (20). For these reasons, therapies aimed at inhibiting Treg activity are promising additions to the cancer immunotherapy arsenal (21). We hypothesized that the apparent loss of Treg-suppressive function seen in the absence of YAP could enhance antitumor immune responses. To test this, WT and Yap cKO mice were challenged with B16 melanoma, an aggressive “nonimmunogenic” cancer model. Tumor growth was measured in these mice over time, and, strikingly, we found that Yap cKO mice controlled the subcutaneous growth of the implanted melanoma cells whereas tumors grew robustly in WT mice (Fig. 3A and B). In line with our in vitro findings, the activation of CD4+ and CD8+ tumor-infiltrating lymphocytes (TIL) from Yap cKO mice was apparently much less restrained than that of their WT counterparts. Intracellular cytokine staining revealed these cells produced significantly higher levels of IFNγ and TNFα (Fig. 3C) compared with those from WT tumors. These results suggest that in the absence of YAP in T cells, a more robust antitumor immune response is mounted.
The impact of T cell– and Treg-restricted YAP deficiency and YAP inhibition on the antitumor response. WT (n = 5) or Yapfl/fl CD4Cre+ (Yap cKO; n = 5) mice were challenged with 5 × 105 B16 melanoma cells (s.c.), the tumor dimensions were measured every 2 days, and tumor volume was calculated (A, B). On day 21, the mice were euthanized and TILs were isolated from the excised tumors. C, TILs were gated on CD4+ and CD8+ T cells, and effector cytokines IFNγ and TNFα levels were measured by flow cytometry. D, Tumor challenge of Yapfl/fl FOXP3Cre+ mice (n = 4) and WT controls (n = 6) was carried out as above, and the frequencies of IFNγ- and IL17-producing leukocytes within the B16 TILs of these mice were determined by flow cytometry (E, left). Mean frequencies of IFNγ+ TILs as well as the number of IFNγ+/CD4+ and IFNγ+/CD8+ per gram of tumor tissue (± SEM) are also shown (E, middle and right, respectively) from at least 3 independent experiments. F, Proportions of FOXP3+ Tregs within the TILs of WT and Yapfl/fl FOXP3Cre+ mice (left) were also found by intracellular staining followed by flow cytometry analysis. Average Treg frequencies amongst CD4+ TILs were found and the ratio of tumor CD8+ T cell to FOXP3+ Treg numbers are also shown (center and right, respectively). G, Targeting YAP improves the antitumor effects of immunotherapies. C57BL/6 mice were challenged with B16 melanoma cells and tumor progression was monitored as mentioned above. Cohorts of mice were treated with i.p. injected VP, GM-Vac, anti–PD-1 antibody, VP and anti–PD-1, or VP and GM-Vac beginning day 7 after tumor injection. Control mice were left untreated (n = 5/group). Shown are the mean tumor volumes for the groups ± SEM. H, Characterization of TILs from treated and control mice were also determined by flow cytometry. The frequencies of IFNγ-producing CD4+ and CD8+ T cells as well as the ratio of tumor CD4+ and CD8+ T-cell numbersto FOXP3+ Treg numbers were also shown. For A, D, and G, the mean tumor volumes for the groups are shown over time ± SEM. Bar graphs in C, E, F, and H depict the mean frequency (%), ratio, or absolute number/gram tumor of the indicated immune cell subset ± SEM in 3 independent experiments. All other findings are representative of at least two independent experiments. Statistically significant differences were determined by t test for all panels except for G, where a two-way ANOVA was used. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
The impact of T cell– and Treg-restricted YAP deficiency and YAP inhibition on the antitumor response. WT (n = 5) or Yapfl/fl CD4Cre+ (Yap cKO; n = 5) mice were challenged with 5 × 105 B16 melanoma cells (s.c.), the tumor dimensions were measured every 2 days, and tumor volume was calculated (A, B). On day 21, the mice were euthanized and TILs were isolated from the excised tumors. C, TILs were gated on CD4+ and CD8+ T cells, and effector cytokines IFNγ and TNFα levels were measured by flow cytometry. D, Tumor challenge of Yapfl/fl FOXP3Cre+ mice (n = 4) and WT controls (n = 6) was carried out as above, and the frequencies of IFNγ- and IL17-producing leukocytes within the B16 TILs of these mice were determined by flow cytometry (E, left). Mean frequencies of IFNγ+ TILs as well as the number of IFNγ+/CD4+ and IFNγ+/CD8+ per gram of tumor tissue (± SEM) are also shown (E, middle and right, respectively) from at least 3 independent experiments. F, Proportions of FOXP3+ Tregs within the TILs of WT and Yapfl/fl FOXP3Cre+ mice (left) were also found by intracellular staining followed by flow cytometry analysis. Average Treg frequencies amongst CD4+ TILs were found and the ratio of tumor CD8+ T cell to FOXP3+ Treg numbers are also shown (center and right, respectively). G, Targeting YAP improves the antitumor effects of immunotherapies. C57BL/6 mice were challenged with B16 melanoma cells and tumor progression was monitored as mentioned above. Cohorts of mice were treated with i.p. injected VP, GM-Vac, anti–PD-1 antibody, VP and anti–PD-1, or VP and GM-Vac beginning day 7 after tumor injection. Control mice were left untreated (n = 5/group). Shown are the mean tumor volumes for the groups ± SEM. H, Characterization of TILs from treated and control mice were also determined by flow cytometry. The frequencies of IFNγ-producing CD4+ and CD8+ T cells as well as the ratio of tumor CD4+ and CD8+ T-cell numbersto FOXP3+ Treg numbers were also shown. For A, D, and G, the mean tumor volumes for the groups are shown over time ± SEM. Bar graphs in C, E, F, and H depict the mean frequency (%), ratio, or absolute number/gram tumor of the indicated immune cell subset ± SEM in 3 independent experiments. All other findings are representative of at least two independent experiments. Statistically significant differences were determined by t test for all panels except for G, where a two-way ANOVA was used. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
Tumor challenge of mice with Treg-restricted YAP deficiency yielded similar results. Although WT controls expectedly permitted rapid tumor development, Yapfl/fl FOXP3Cre+ mice maintained small tumors infiltrated by elevated populations of inflammatory cytokine-producing leukocytes. Specifically, producers of the tumoricidal Th1 cytokine IFNγ were found at higher frequencies and in greater numbers in the tumors of Yapfl/fl FOXP3Cre+ mice than those of WT controls (Fig. 3D and E). Analysis of FOXP3 expression by CD4+ TILs revealed that deletion of YAP in Tregs reduces the frequency of suppressive FOXP3+ Tregs in the tumor microenvironment (Fig. 3F, left and middle). The relative balance (i.e., the ratio) of Tregs and potential effector CD8+ T cells was similarly shifted in the tumors of mice with Treg-specific YAP deficiency compared with those of WT controls (Fig. 3F, right). Treg-specific YAP deficiency also slowed the growth of tumors caused by implanted MC38 adenocarcinoma cells (Supplementary Fig. S5A–S5B). Not only were MC38 tumors much smaller in Yapfl/fl FOXP3Cre+ mice 21 days after injection, the relative proportions of FOXP3+ Tregs among tumor-infiltrating T cells were reduced compared with WT tumors. In contrast, the frequencies of intratumoral producers of IFNγ and TNFα were elevated in the absence of Treg-specific YAP expression (Supplementary Fig. S5C–S5D). Corroborating results were seen in the injectable EL4 thymoma model in which Treg-restricted Yap knockout resulted in dramatically stunted tumor growth relative to WT mice. As with other tumor models, this derailed tumor progression was concurrent with reduced Treg proportions and an elevated presence of proinflammatory cytokine-producing T cells in the tumor microenvironment (Supplementary Fig. S6A–S6D). These experiments make a strong case for YAP's role as both a facilitator of Treg presence in the tumor niche and a potent and broadly active driver of Treg-enforced inhibition of endogenous antitumor immunity.
Some of the most promising immunotherapeutic agents (i.e., PD-1 and CTLA4 antagonist antibodies) show even greater antitumor effect when administered in concert (22–24) or alongside tumor vaccine strategies (25–28). We therefore tested the therapeutic potential of YAP targeting as an immunotherapeutic approach to combat cancer. Administration of a known YAP inhibitor, verteporfin (VP; ref. 29), to melanoma-bearing mice resulted in modest reduction in tumor size (Fig. 3G). Treatment of melanoma-bearing Yapfl/fl FOXP3Cre+ mice with VP, on the other hand, failed to alter the already stunted progression of tumors in these mice (Supplementary Fig. S7A), suggesting that potential off-target effects of this drug or any direct effects on tumor cells are not likely contributing to these in vivo observations. We also tested the effects of combining VP with the proven immunotherapeutic agents anti–PD-1 antibody and GM-Vac (irradiated GM-CSF–producing B16 cells). Both anti–PD-1 and GM-Vac treatments were able to slow tumor growth somewhat as monotherapies. Notably, combinatorial treatment with VP and anti–PD-1 neutralizing antibody suppressed tumor progression to a greater extent than any monotherapy tested. Even more dramatic were the synergistic effects of VP and GM-Vac, which prevented the development of tumors beyond a barely detectable size (Fig. 3G). The decidedly improved antitumor efficacy seen upon combination of either anti–PD-1 or GM-Vac treatment with VP was associated with enhanced proportions of IFNγ-producing CD8+ and CD4+ T cells and a compromised Treg presence in the tumor microenvironment (Fig. 3H; Supplementary Fig. S7B–S7E). These findings strongly suggest a major role for Treg-derived YAP in crafting the immunosuppressive nature of the tumor microenvironment. They also suggest the potential of immunotherapeutic approaches that include YAP-targeting agents.
YAP Potentiates Expression of Genes Involved in TGFβ/SMAD and Activin Signaling
To gain insight into the mechanism by which YAP contributes to Tregs and their enforcement of immune suppression, we isolated Tregs from mice lacking YAP in these cells (Yap cKO) and subjected them to RNA-sequencing (RNA-seq) analysis along with WT Tregs and naïve CD4+ T cells from both mice. The results of this analysis revealed that YAP-deficient Tregs display reduced expression of several genes known to be important in the signaling pathway triggered by the anti-inflammatory cytokine TGFβ. Interestingly, one of the genes most downregulated in the absence of YAP was that encoding the signaling component of the activin receptor complex known as Acvr1c (Fig. 4A; Supplementary Fig. S8A). Confirming a role for YAP in potentiating Acvr1c expression, we found that WT CD4+ T cells display considerable upregulation of the transcript for this receptor subunit during in vitro Treg differentiation, whereas their YAP-deficient counterparts did not. Interestingly, freshly isolated nTregs expressed modest levels of Acvr1c. However, upon activation, these Tregs dramatically activated activin receptor expression in a YAP-dependent manner (Supplementary Fig. S8B). Indeed, neither nTreg nor differentiating iTregs from Yapfl/flFOXP3Cre+ mice expressed considerable Acvr1c mRNA levels. Thus YAP-mediated activin responsiveness may have considerable influence over the biology of multiple Treg populations.
RNA-seq analysis of WT and YAP-deficient Treg transcriptomes. Genes that had their expression level significantly changed by Yap knockout in naïve CD4+ T cells, unstimulated (unstim) Tregs, or stimulated (stim) Tregs were determined by RNA-seq. Results are presented as a heat map. Genes are arranged based on the fold change in expression between WT and YAP-deficient Tregs (genes with a fold change of more than 3 are shown). The color representation from green to red denotes log2-transformed FPKM from −2 to 2.
RNA-seq analysis of WT and YAP-deficient Treg transcriptomes. Genes that had their expression level significantly changed by Yap knockout in naïve CD4+ T cells, unstimulated (unstim) Tregs, or stimulated (stim) Tregs were determined by RNA-seq. Results are presented as a heat map. Genes are arranged based on the fold change in expression between WT and YAP-deficient Tregs (genes with a fold change of more than 3 are shown). The color representation from green to red denotes log2-transformed FPKM from −2 to 2.
It has been suggested that activin can promote TGFβ signaling. Pathway analysis of our RNA-seq results showed that the gene expression patterns most affected by YAP deficiency in Tregs were highly relevant to immune control and the diverse autoimmune pathologies resulting from the breakdown of such control. Among these, the genes associated with the TGFβ signaling cascade were markedly altered (Supplementary Fig. S8C). Furthermore, RT-PCR analysis also showed reduced transcript levels for several known TGFβ-responsive genes in Tregs from YAPfl/fl FOXP3Cre+ mice (Supplementary Fig. S8D). In light of these findings, we suspected that YAP contributes to Treg-mediated immune control at least in part by bolstering TGFβ/SMAD signaling through the activin/AcVR1C axis in these suppressor cells.
Although activin mRNA levels were low in naïve CD4+ T cells, in vitro differentiating Tregs (naïve CD4+ T cells activated with anti-CD3/CD28 in the presence of IL2 and TGFβ) upregulated activin expression over time (Supplementary Fig. S9A). The kinetics of this upregulation paralleled the appearance of FOXP3 expression in these cells (Supplementary Fig. S9B). qRT-PCR analysis also showed that expression of the activin receptor (AcVR1C) was similarly low in naïve CD4+ T cells, but was robustly upregulated under in vitro culture conditions that generate iTreg (Supplementary Figs. S8B and S9C). We went on to dissect which Treg-inducing stimuli were chiefly responsible for inducing expression of YAP and elements of activin/ACVR1C signaling. To this end, naïve CD4+ T cells were activated in vitro with anti-CD3/CD28 antibodies, either alone or in the presence of IL2, TGFβ, or IL2 and TGFβ. As expected, activation alone failed to induce upregulation of these genes or the canonical Treg transcription factor FOXP3. The cytokine TGFβ did trigger significant expression of FOXP3, as expected, but YAP as well. Exposure to IL2 along with TGFβ (but not IL2 alone) greatly augmented expression of YAP and FOXP3. Of the conditions tested, those upregulating robust YAP also brought about expression of activin and ACVR1C (Supplementary Fig. S9D–S9G). These findings further align the upregulation of YAP expression and activin signaling with the Treg lineage and shed some light on the largely unknown cast of molecular characters regulating these processes in T cells.
To gain further insight into the mechanism of YAP-mediated ACVR1C upregulation, we explored the potential involvement of a known YAP-collaborating factor. Mature YAP protein is known to contain a TEAD-binding domain, and prior studies (largely conducted in non-T cells) have identified numerous target genes controlled by the cooperation of these factors. Suggesting that transcription at the AcVR1c locus is activated through YAP–TEAD interaction, the promoter sequence of this gene was found to contain two TEAD consensus binding sites (Fig. 5A). To test the importance of TEAD binding for YAP-dependent AcVR1c expression, we prepared luciferase-based reporter constructs under the control of WT murine AcVR1c promoter sequence. Mutant constructs having either or both of the TEAD sites ablated were also designed (Fig. 5B). Each AcVR1c-luciferase reporter construct was delivered into Jurkat T cells along with an expression vector encoding YAP (“YAP1wt”) or a mutant version of this transcription factor unable to interact with TEAD (“YAP1mut”) owing to an S-to-A mutation at residue 94 (“S94A”). In this system, expression of TEAD1 or YAP1WT alone induced only modest activation of AcVR1c expression. In contrast, robust luciferase signal was detected when WT YAP and TEAD were coexpressed. Mutation of YAP's TEAD interaction site, however, resulted in far less reporter activity (Fig. 5C), supporting the notion that YAP–TEAD cooperation is necessary for optimal AcVR1c expression. Similarly, loss of a single TEAD binding site in the promoter sequence reduced YAP-induced transcription whereas mutation of both sites resulted in a significant and near-complete loss of reporter signal (Fig. 5D). These results clearly implicate a molecular partnership between YAP and TEAD in the potentiation of activin signaling through AcVR1c expression. This point was further supported by chromatin immunoprecipitation (ChIP) assays showing both YAP1 and TEAD1 are enriched at the AcVR1c locus in WT iTregs (Fig. 5E). Notably, in Yap cKO-derived iTregs, TEAD1 was still found interacting with the AcVR1c locus despite the absence of YAP (Fig. 5E). These findings illuminate the mechanism behind YAP's activation of activin signaling.
YAP drives AcVR1c transcription through molecular cooperation with TEAD. A, Shown is a portion (1.2 kb) of the murine AcVR1c promoter with TEAD binding sites determined by transcription factor prediction software. B, Design schema of luciferase-based AcVR1c reporter constructs;C, AcVR1c reporter assays. The mAcVR1c-WT reporter construct was cotransfected into Jurkat T cells along with the indicated YAP and TEAD expression constructs or an empty vector control. Cells were cultured with or without PMA/ionomycin (iono) activation for 8 hours prior to harvest and cell lysis. Luciferase activity was determined as previously described (47). D, As in C, YAP and TEAD expression were delivered to Jurkat T cells, except cells received variants of the mAcVR1c reporter possessing one, both, or none of the identified TEAD binding sites. Reporter activity was determined as in C. For both C and D, the mean relative luciferase values ± SEM are shown for the results of 3 independent experiments. E, A chromatin immunoprecipitation assay was carried out in iTregs generated from WT- and Yap cKO-derived naïve CD4+ T cells. The ability of antibodies against TEAD1 and YAP to pull down the indicated factors along with the AcVR1c promoter region was calculated based upon qPCR relative to a control IgG. The relative enrichment for each factor over 3 experiments is shown (±SEM). Significant differences for all experiments shown were determined by the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
YAP drives AcVR1c transcription through molecular cooperation with TEAD. A, Shown is a portion (1.2 kb) of the murine AcVR1c promoter with TEAD binding sites determined by transcription factor prediction software. B, Design schema of luciferase-based AcVR1c reporter constructs;C, AcVR1c reporter assays. The mAcVR1c-WT reporter construct was cotransfected into Jurkat T cells along with the indicated YAP and TEAD expression constructs or an empty vector control. Cells were cultured with or without PMA/ionomycin (iono) activation for 8 hours prior to harvest and cell lysis. Luciferase activity was determined as previously described (47). D, As in C, YAP and TEAD expression were delivered to Jurkat T cells, except cells received variants of the mAcVR1c reporter possessing one, both, or none of the identified TEAD binding sites. Reporter activity was determined as in C. For both C and D, the mean relative luciferase values ± SEM are shown for the results of 3 independent experiments. E, A chromatin immunoprecipitation assay was carried out in iTregs generated from WT- and Yap cKO-derived naïve CD4+ T cells. The ability of antibodies against TEAD1 and YAP to pull down the indicated factors along with the AcVR1c promoter region was calculated based upon qPCR relative to a control IgG. The relative enrichment for each factor over 3 experiments is shown (±SEM). Significant differences for all experiments shown were determined by the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
Activin Enhances SMAD/TGFa Signaling and Treg Differentiation
Because activin has been reported to promote SMAD signaling in non-T cells (30), we tested whether activin signaling in T cells could have a similar effect. SMAD activity was assessed by western blot analysis of SMAD phosphorylation. Indeed, we found that whereas untreated CD4+ T cells did not contain discernible levels of active (phosphorylated) SMAD molecules, treatment with 5 or 10 ng/mL of activin A resulted in elevation of phospho-SMAD levels. As expected, TGFβ treatment (0.5 or 2 ng/mL) also induced SMAD phosphorylation. Importantly, combined activin and TGFβ treatment resulted in even further activation of the SMAD signaling pathway (Fig. 6A). These findings suggest that activin signaling can augment signaling along the TGFβ/SMAD axis—a signaling pathway crucial for multiple aspects of Treg biology and immune tolerance (7).
The YAP/activin/ACVR1C pathway enhances SMAD activation, Treg generation and function, and tumor progression. A, Freshly isolated CD4+CD25− T cells were isolated from the lymphoid tissues of WT mice (n = 6/experiment), cultured with plate bound anti-CD3 (2 μg/mL) and soluble anti-CD28 (2 μg/mL) for 24 hours, followed by treatment with different concentrations of activin A and TGFβ as indicated for an additional 12 hours. Cells were harvested and subjected to SDS-PAGE and western blot with the indicated antibodies (left). Band densities indicating protein amount were quantified by using ImageJ software, normalized to β-actin loading controls, and the mean density ± SEM across 3 independent experiments were found (right). B, Naïve CD4+ T cells from Yapfl/fl, FOXP3Cre+, and YapWT/WT, FOXP3Cre+ (WT) mice (n = 6/group/experiment) were stimulated with anti-CD3/CD28 antibodies (1 and 4 μg/mL, respectively) for 3 days in the presence of IL2 (100 U/mL) and the indicated doses of TGFβ and exogenous activin A. Activin was dosed at 50 ng/mL on days 0 and 2. Treg induction was assessed by flow cytometric detection of intracellular FOXP3. Shown are representative FOXP3 stainings (left) and the mean results of 3 independent experiments ± SEM (right). C, Effect of ectopic AcVR1c expression on YAP-deficient Tregs. As before, WT responder T cells and Tregs were isolated from the indicated mice (n = 6/group/experiment). Following lentiviral delivery of an ACVR1C overexpression construct or an empty vector control construct to Yapfl/fl FOXP3Cre+ Tregs (activated ex vivo overnight with anti-CD3/CD28 antibodies and IL2), the functional capacity of these cells was assessed in vitro. The transduced Tregs were cocultured with CFSE-stained CD45.1+ naïve CD4+ T cells (responders) at the indicated ratio and antigen-presenting cells (T cell–depleted splenocytes). After 5 days of activation, responder cell proliferation was assessed by flow cytometry. Shown at left are representative plots of responder cell gated (CD45.1+/CD4+) events from 1 of 2 independent experiments with like results. The immunoblot at right confirms expression levels of AcVR1c in transduced Tregs, and the bar graph (lower right) depicts the mean fraction of proliferating responder cells over all experiments ± SEM. Where indicated by asterisks, significant differences were found by the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
The YAP/activin/ACVR1C pathway enhances SMAD activation, Treg generation and function, and tumor progression. A, Freshly isolated CD4+CD25− T cells were isolated from the lymphoid tissues of WT mice (n = 6/experiment), cultured with plate bound anti-CD3 (2 μg/mL) and soluble anti-CD28 (2 μg/mL) for 24 hours, followed by treatment with different concentrations of activin A and TGFβ as indicated for an additional 12 hours. Cells were harvested and subjected to SDS-PAGE and western blot with the indicated antibodies (left). Band densities indicating protein amount were quantified by using ImageJ software, normalized to β-actin loading controls, and the mean density ± SEM across 3 independent experiments were found (right). B, Naïve CD4+ T cells from Yapfl/fl, FOXP3Cre+, and YapWT/WT, FOXP3Cre+ (WT) mice (n = 6/group/experiment) were stimulated with anti-CD3/CD28 antibodies (1 and 4 μg/mL, respectively) for 3 days in the presence of IL2 (100 U/mL) and the indicated doses of TGFβ and exogenous activin A. Activin was dosed at 50 ng/mL on days 0 and 2. Treg induction was assessed by flow cytometric detection of intracellular FOXP3. Shown are representative FOXP3 stainings (left) and the mean results of 3 independent experiments ± SEM (right). C, Effect of ectopic AcVR1c expression on YAP-deficient Tregs. As before, WT responder T cells and Tregs were isolated from the indicated mice (n = 6/group/experiment). Following lentiviral delivery of an ACVR1C overexpression construct or an empty vector control construct to Yapfl/fl FOXP3Cre+ Tregs (activated ex vivo overnight with anti-CD3/CD28 antibodies and IL2), the functional capacity of these cells was assessed in vitro. The transduced Tregs were cocultured with CFSE-stained CD45.1+ naïve CD4+ T cells (responders) at the indicated ratio and antigen-presenting cells (T cell–depleted splenocytes). After 5 days of activation, responder cell proliferation was assessed by flow cytometry. Shown at left are representative plots of responder cell gated (CD45.1+/CD4+) events from 1 of 2 independent experiments with like results. The immunoblot at right confirms expression levels of AcVR1c in transduced Tregs, and the bar graph (lower right) depicts the mean fraction of proliferating responder cells over all experiments ± SEM. Where indicated by asterisks, significant differences were found by the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
TGFβ/SMAD-mediated events are important during the upregulation of FOXP3 and the generation of Tregs from naïve CD4+ T-cell precursors. We next investigated whether YAP-dependent activin signaling can participate in the driving of this process. Having shown that YAP plays an important role in promoting or maintaining FOXP3 expression induced in the presence of various TGFβ concentrations (Fig. 2D), and having implicated the transcription factor in the regulation of TGFβ-sensitive genes (Fig. 4; Supplementary Fig. S8), we therefore postulated that YAP-mediated upregulation of AcVR1C and SMAD signaling might provide a crucial amplification of this important Treg-supporting signaling pathway that allows for more robust or sustained FOXP3 expression.
To explore the involvement of activin/AcVR1C signaling in the enhancement of Treg differentiation by YAP, the effect of supplemental activin A on in vitro Treg commitment was also investigated. As expected, activation of naïve CD4+ T cells without TGFβ yielded little to no FOXP3 induction regardless of YAP expression. Strikingly, activation of WT cells with exogenous activin A, even in the absence of TGFβ, generated a population of FOXP3+ cells. Although a suboptimal concentration of TGFβ resulted in modest upregulation of FOXP3 (mirroring the effects on SMAD activation), combined treatment of WT naïve CD4+ T cells with low doses of TGFβ plus activin A resulted in synergistic promotion of FOXP3+ T-cell induction. This induction of Tregs by activin A treatment alone was largely not seen upon FOXP3-driven knockout of Yap, and although dual treatment of differentiating Yapfl/fl FOXP3Cre+ iTregs did enhance the generation of FOXP3+ cells, it was to an extent far less than that seen in their WT counterparts (Fig. 6B). These results suggest that activin signaling via YAP-dependent AcVR1C expression on Treg not only augments TGFβ signaling but can also drive the process of FOXP3 upregulation. Supporting this notion, naïve T cells lacking AcVR1c were found to be less sensitive to TGFβ-induced iTreg differentiation than their WT counterparts, particularly when TGFβ concentrations were low (Supplementary Fig. S10A–S10B). Interestingly, exogenous activin supplementation could do little to rescue the deficient FOXP3 induction seen in naïve CD4+ T cells lacking either SMAD2 or SMAD3, or, for that matter, the pronounced defect in iTreg generation seen in T cells genetically lacking both SMAD molecules (Supplementary Fig. S11). This observation confirms that functional SMAD signaling is required for activin-mediated enhancement of Treg generation, in agreement with prior studies (31). In all, these findings are very much in line with a role for YAP-driven activin signaling in the augmentation of signaling down the SMAD/TGFβ axis in T cells.
Activin-Mediated Support of Treg Function Is YAP/AcVR1C Dependent
YAP deficiency leads to improved antitumor immunity and a Treg pool that is insensitive to an activator of the TGFβ/SMAD signaling pathway (i.e., activin). We therefore hypothesized that YAP facilitates robust Treg function in vivo through the induction of AcVR1C, which in turn amplifies the pro-Treg signaling cascade. In order to determine if the Treg-promoting effects of YAP were due to the upregulation of AcVR1C, we set out to test whether the defective Treg function seen in Yap knockouts could be restored by ectopic expression of AcVR1C. In an in vitro suppression assay, as expected, Yapfl/fl FOXP3Cre+-derived Tregs transduced with an empty vector control expressed reduced levels of ACVR1C protein and were much less efficient suppressors of naïve CD4+ T-cell proliferation than their WT counterparts. However, lentiviral-based delivery of an ACVR1C-encoding expression construct into Yapfl/fl FOXP3Cre+-derived Tregs more than rescued receptor expression, which greatly enhanced their suppressive potency beyond even that of WT Tregs (Fig. 6C). These results support the conclusion that activin signaling through AcVR1C (upregulated by YAP) can amplify the suppressive potency of established Tregs as well as the TGFβ-driven differentiation of iTregs and potentially other facets of this cytokine's broadly immunosuppressive action. Importantly, they also suggest that targeting either YAP or activin signaling is likely to undermine the tolerance-promoting attributes of TGFβ and both subsets of FOXP3+ Tregs in the cancer setting. These approaches may provide avenues to enhance antitumor immunity either as novel treatments on their own or as potent enhancers of other promising immunotherapeutic agents.
Activin Blockade or AcVR1c Knockout Inhibits Tumor Growth
As an instigator of an apparent feed-forward loop capable of amplifying TGFβ/SMAD activity, YAP presents a tempting target for those aiming to break tolerance in the cancer setting. However, the targeting of YAP in patients with cancer may prove problematic owing to the molecule's intracellular location and the chemical drawbacks of known inhibitors (e.g., VP has noted solubility issues; ref. 29). Therefore, the activin/AcVR1C interaction is likely to serve as a desirable alternative strategy. Having demonstrated the positive effects of activin signaling on the TGFβ/SMAD signaling pathway and the processes of Treg generation and function, which can oppose immune-mediated tumor cell killing, we suspected that disrupting activin function should enhance antitumor immunity. We therefore tested the potential of activin targeting as an immunotherapeutic approach to combat cancer. Administration of anti-activin monoclonal antibody to mice injected subcutaneously with B16 melanoma markedly stunted the development of tumors relative to an inert isotype control (Fig. 7A). We also tested the value of combining anti-activin blocking antibody treatment with the anticancer vaccine GM-Vac. Treatment with GM-Vac alone was able to partially slow the growth of tumors to an extent similar to anti-activin monotherapy. However, combining anti-activin treatment with GM-Vac was able to arrest tumor growth at a barely detectable size (Fig. 7A). Anti-activin treatment also successfully reduced the frequency of FOXP3+ Tregs among TILs, and although GM-Vac–receiving mice displayed some reduced Treg presence in their tumors, combined GM-Vac and activin blockade resulted in dramatic loss of these suppressor T cells from the tumor microenvironment (Fig. 7B). The effect of blocking activin on Tregs coincided with increased frequencies of IFNγ-producing CD8+ and CD4+ T cells, an observation even more prominent upon combination of GM-Vac and anti-activin treatments (Fig. 7C). These results demonstrate the susceptibility of the YAP/AcVR1c/activin axis to therapeutic targeting at multiple points.
Activin blockade and AcVR1c deficiency slows B16 tumor growth and enhances the antitumor immune response. A, B16 melanoma cells were injected into individual female C57BL/6 mice (8–12 weeks of age). Tumor-bearing mice were randomly assigned into treatment groups once tumors were palpable ∼7 days after injection. Anti-activin A antibodies (R&D Systems) were administered (100 μg/mouse/injection) intraperitoneally twice a week once to one group. Another group received like doses of control IgG1. Other cohorts of tumor-bearing mice received GM-vaccine [100 μL of 1 × 106 lethally irradiated (150 Gy) B16 GM-vaccine cells or combined anti-activin/GM-vaccine treatment; n = 10 mice per group]. B, Treg frequencies among the TILs of treated mice. Intracellular staining of FOXP3 in CD4+ TILs from the indicated treatment groups were determined by flow cytometry. C, IFNγ-producing CD4+ and CD8+ T cells recovered from tumor cell suspensions were similarly assessed. D, The right flank of 8-week WT and AcVR1c KO female mice (C57BL/6 background; n = 8/group) were injected with 4 × 105 B16 cells in 100 μL PBS. E, The proportions of IFNγ- and TNFα-expressing T cells (CD3+) with the TILs of these mice were determined by flow cytometry (F) as were the frequencies of FOXP3 and IL17+ CD4+ T cells. For A and D, tumor development and changes in tumor volume were recorded for all groups, and the mean volume ± SEM for each are displayed. For B, C, E, and F, representative flow plots from a single mouse from each group are depicted (left) alongside the mean cell frequencies across 3 independent experiments (right). All experiments were repeated at least three times. Significant differences were determined by a Student t test for all panels, except A, where a two-way ANOVA was used. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
Activin blockade and AcVR1c deficiency slows B16 tumor growth and enhances the antitumor immune response. A, B16 melanoma cells were injected into individual female C57BL/6 mice (8–12 weeks of age). Tumor-bearing mice were randomly assigned into treatment groups once tumors were palpable ∼7 days after injection. Anti-activin A antibodies (R&D Systems) were administered (100 μg/mouse/injection) intraperitoneally twice a week once to one group. Another group received like doses of control IgG1. Other cohorts of tumor-bearing mice received GM-vaccine [100 μL of 1 × 106 lethally irradiated (150 Gy) B16 GM-vaccine cells or combined anti-activin/GM-vaccine treatment; n = 10 mice per group]. B, Treg frequencies among the TILs of treated mice. Intracellular staining of FOXP3 in CD4+ TILs from the indicated treatment groups were determined by flow cytometry. C, IFNγ-producing CD4+ and CD8+ T cells recovered from tumor cell suspensions were similarly assessed. D, The right flank of 8-week WT and AcVR1c KO female mice (C57BL/6 background; n = 8/group) were injected with 4 × 105 B16 cells in 100 μL PBS. E, The proportions of IFNγ- and TNFα-expressing T cells (CD3+) with the TILs of these mice were determined by flow cytometry (F) as were the frequencies of FOXP3 and IL17+ CD4+ T cells. For A and D, tumor development and changes in tumor volume were recorded for all groups, and the mean volume ± SEM for each are displayed. For B, C, E, and F, representative flow plots from a single mouse from each group are depicted (left) alongside the mean cell frequencies across 3 independent experiments (right). All experiments were repeated at least three times. Significant differences were determined by a Student t test for all panels, except A, where a two-way ANOVA was used. *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001.
Along this line, B16 tumor growth was also markedly slower in AcVR1c knockout mice than in WT controls (Fig. 7D). Correspondingly, the TILs from AcVR1c-deficient mice contained fewer FOXP3+ Tregs than their WT counterparts and displayed a selective elevation of IFNγ-producing T cells (Fig. 7E and F). As with chemical YAP inhibition and antibody-mediated activin blockade, administering GM-Vac to AcVR1c knockout mice enhanced the already considerable antitumor effect of genetic AcVR1c ablation (Supplementary Fig. S12). From these results, it is clear that disrupting any of the several elements of the YAP/activin/SMAD axis can undermine immune suppression and oppose tumor progression in mice.
In all, our findings support the conclusion that signaling along the YAP-regulated activin/ACVR1C axis can support Treg generation and function and potentially other broadly immune-suppressing effects of the TGFβ/SMAD pathway. Importantly, they also suggest that targeting this axis is likely to undermine the immune suppressive attributes of TGFβ and FOXP3+ Tregs in the cancer setting—either alone or in combination with other promising immunotherapeutic agents (e.g., immune checkpoint–blocking antibodies and anticancer vaccines).
Discussion
Tregs are indispensable for restraining potentially lethal self-directed (autoimmune) responses or overexuberant ones mounted against normally harmless commensal microbes (inflammatory bowel disease; ref. 1). However, in patients with cancer, Tregs can be greatly enriched within tumors, sometimes systemically (32). The suppressive function of these cells in this setting dampens the effectiveness of tumor-directed immunity and is a major obstacle for developing effective anticancer immunotherapies (21).
As part of an ongoing effort to identify precise mechanisms of Treg generation, maintenance, and function in the context of cancer, we have made the surprising discovery that YAP, a transcription factor critical in developmental regulation of organ size, is in fact an important factor in the generation and function of Tregs. Deletion of Yap1 in T cells somewhat enhances both Th1 and Th17 development but most impressively diminishes generation of iTregs under conditions of limited TGFβ. YAP deficiency also negatively affects the suppressive function of Tregs. The inability of Tregs to suppress immunity in vivo in the absence of YAP was dramatically illustrated by our B16 melanoma tumor model experiments (Fig. 3). The poorly immunogenic tumor failed to grow in mice with Treg-specific Yap deletion, which displayed markedly enhanced indicators of proinflammatory antitumor immunity compared with WT controls. This improved deployment of antitumor immunity was seen alongside a markedly diminished Treg presence in the tumor microenvironment (Fig. 3E and F)—observations also seen upon Treg-specific YAP deficiency across other, distinct tumor models as well. These findings strongly suggest that YAP is important for the accumulation and suppressive function of Tregs in the tumor microenvironment. Furthermore, they imply that targeting YAP should be a potent means of overcoming immune suppression in the cancer setting and improving the efficacy of endogenous and therapeutically induced tumor killing by leukocyte. Further characterization of YAP expression by Treg subsets found in different healthy and diseased tissues (including tumors) should more clearly define this factor's role in immune control in specific physiologic contexts.
Here, we present a body of data strongly suggesting a Treg-specific role for YAP in promoting the immune suppression capable of allowing the persistence and progression of tumors in the cancer setting. Indeed, YAP-expression patterns and the dramatically stunted tumor growth seen in Yapfl/fl FOXP3Cre+ mice support this. However, comparing the degree of antitumor effect resulting from T cell– and Treg-driven YAP deficiency, it appears that a slightly more dramatic effect is seen in the former case. Although the bulk of the effect seen in Fig. 3A is phenocopied by the more restrictive deletion of YAP in only FOXP3+ cells (Fig. 3D), it is possible that YAP may play a tumor-abetting role in some other T-cell population capable of inducing the factor in the cancer setting. Although such YAP expression appears to have relatively minor consequences next to Treg-derived YAP, at least in the tumor models used in our study, future work may bring to light additional layers of YAP's protumor effects involving cells beyond Foxp3+ Tregs (such as anergized or exhausted T cells, non–Foxp3-expressing TR1 Tregs, etc.). These too may be susceptible to YAP-targeting strategies, which, based on our results, clearly should have potent antitumor effects.
Indeed, using a known YAP antagonist with modest inhibitory activity (29), we confirmed the potential of YAP as a target for Treg-undermining immunotherapies. Although inhibiting YAP alone slightly decreased tumor growth, we observed strong synergy in antitumor activity and immunity-boosting effects when the drug was combined with a tumor vaccine and checkpoint inhibitor treatment that alone possess much less potent effects. These findings suggest that YAP-targeting approaches should increase the efficacy of current immunotherapies, potentially by enhancing the presence of activated effector leukocytes in the tumor microenvironment.
Analysis of the downstream targets of YAP activity in Treg identified ACVR1C led to the finding that the activin–activin receptor signaling axis plays a major role in the augmentation of TGFβ/SMAD signaling and Treg generation and function (summarized in Supplementary Fig. S13). This pathway is highly important for the induction of extrathymic FOXP3+ T cells from naïve CD4+ precursors, as SMADs bind critical enhancer regions for the FOXP3 gene (6, 33). It is also important for sustaining FOXP3 expression and suppressive function in Tregs (7), and TGFβ has been implicated as a promoter of survival and phenotypic stabilization of thymic Tregs (34, 35). With such reliance on TGFβ and SMAD signaling, it stands to reason that Tregs use mechanisms to optimize or amplify the downstream signaling events and resultant gene regulation triggered by this pathway. Such amplification mechanisms can be important for maintaining the gene expression and phenotype traits underlying the suppressive function of Tregs. Documented examples include the enzymatic conversion of latent TGFβ to its active form (36) and the triggering of SMAD activation by galectin and CD44 (11). The upregulation of YAP and subsequently ACVR1C—the receptor for a known enhancer of SMAD signaling (i.e., activin)—may serve as an additional mechanism for amplifying this decidedly pro-Treg cascade. Herein activin/ACVR1C signaling can enhance the downstream signaling events triggered by TGFβ. Reports of activin expression in several tumor types (37, 38) support the notion that tumor-accumulating Tregs benefit particularly from activin/ACVR1C signaling facilitated by YAP induction.
Our proof-of-concept experiments demonstrate that this pro-Treg amplification mechanism is susceptible to therapeutic disruption. Particularly, our findings suggest that antibody-mediated activin blockade may prove a most effective means for the disruption of Treg- and tumor-abetting TGFβ activation in patients with cancer. Additionally, the development and vetting of therapeutic antibodies capable of neutralizing activin, AcVR1C, or blocking its association with AcVR1C in patients with cancer may lead to new and potent immunotherapeutic regimens capable of releasing antitumor immunity from stifling Treg-enforced tolerance. On the other hand, our findings suggest that supplementation of activin or other therapeutic enhancements of the activin/ACVR1C axis could have considerable potential as a strategy to correct inadequate immune regulation in settings of autoimmunity (e.g., multiple sclerosis) or inflammatory disease (e.g., inflammatory bowel disease). Future application of YAP inhibitors or activin/ACVR1C ablation in mouse models relevant to these and other pathologies of immune dysregulation will shed light on whether this pro-Treg loop is generally important for immune control or if it is principally operative in the tumor setting.
Our findings are, to our knowledge, the first to implicate YAP as a transcriptional facilitator of Treg differentiation and function. Although this molecule has been previously studied for its regulation of development, organ size, regeneration, and tumorigenesis (39), and its role as a transcriptional effector of gene expression downstream of the Hippo pathway is well established, the importance of the Hippo pathway and its associated cofactors in Tregs and immune control is only beginning to be understood. A recent study showed that the Hippo pathway kinase known as MST1 plays an important role in stabilizing FOXP3 protein levels and supporting Treg function (40). Our present findings reveal that YAP potentiates Treg-supporting SMAD activity in T cells through activin signaling. Notably, though, this unexpected role appears to be independent of other Hippo factors (i.e., MST1/2 and LATS1/2), as these, unlike YAP, were not highly upregulated in developing Tregs. Interestingly, another Hippo effector known as TAZ (regarded to be a YAP paralog) was recently identified as a promoter of Th17 differentiation in naïve CD4+ T cells and a negative regulator of FOXP3 function and expression in these cells (41). This role for TAZ in the generation of proinflammatory T cells was also apparently beyond its traditional Hippo-dependent role. Taken together, these newly uncovered immunologic roles played by YAP and TAZ suggest that different molecular players in the Hippo pathway can have functionally opposite and mechanistically distinct roles in determining the balance between inflammation and tolerance. Further dissection of this pathway in T cells should add considerably to our understanding of this balance and, based on our current study's findings, may lead to potent new immunotherapy approaches.
Methods
Mice
C57/BL6 Yapfl/fl mice were generous gifts of Dr. Duojia Pan. C57/BL6 AcVR1c knockout mice were gifts from Dr. Ning Lu. C57/BL6 CD4-cre and FOXP3-YFP-Cre transgenic mice were purchased from The Jackson Laboratory. Smad2−/−, Smad3−/−, and Smad2/3 double knockout mice on a C57BL/6 background were originally obtained from Dr. Se-Jin Lee's laboratory and were previously described (42). All animal experiments performed were approved by the Johns Hopkins University Institutional Animal Care and Use Committee.
In Vitro T-cell Differentiation
Naïve CD4+ T cells (CD4+ CD25− CD62Lhi) were sorted on a FACSAria II high-speed sorter. The sorted cells were activated with plate-bound anti-CD3 (1 μg/mL) and soluble anti-CD28 (2 or 4 μg/mL) in a 24-well plate with the following polarizing conditions: Th1 [IL12 (10 ng/mL), anti-IL4 (10 μg/mL)], Th2 [IL4 (10 ng/mL), anti-IFNγ (10 μg/mL), anti-IL12 (10 μg/mL)], Th17 [IL6 (10 ng/mL), TGFβ1 (1.25 ng/mL), IL23 (10 ng/mL), IL1β (10 ng/mL), anti-IFNγ (10 μg/mL), anti-IL4 (10 μg/mL)], Treg [TGFβ1 (5 ng/mL, or as indicated), IL2 (100 IU/mL)] typically for 4 days, unless otherwise indicated.
Human T-cell Isolation from Peripheral Blood
Deidentified human peripheral blood was obtained from blood bank in strict accordance with the Johns Hopkins University School of Medicine's Institutional Review Board guidelines. Samples were obtained from a total of 10 healthy adult volunteers (age range, 30–46 years). Peripheral blood mononuclear cells were extracted from whole blood through a gradient of Ficoll–Paque PLUS (GE Healthcare). CD4+ T cells were enriched using a Dynabeads Untouched CD4 T-cell isolation kit (Invitrogen). Tregs were identified and flow sorted via the following staining profile: CD3+/CD4+/CD8−/CD25hi/CD127lo/CD39+. Non-Treg CD4+ T cells were sorted as previously described (43).
In Vitro Suppression Assay
WT naïve CD4+ T cells (0.1 × 106) were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE) and cultured in a 96-well bottom plate with anti-CD3/CD28-conjugated beads at a cell-to-bead ratio of 1:1. Serially diluted Tregs (CD4+ CD25hi) were cocultured for 72 hours, and cellular proliferation by CFSE was measured by flow cytometry.
Lentivirus Production and Transduction
HEK293T cells were purchased from the ATCC in 2015 and were kept as a frozen stock. This cell line has not been authenticated by the laboratory. Recombinant lentiviruses were generated using a three-plasmid system as described previously (44). The AcVR1c cDNA was cloned into the modified pLV lentiviral vector carrying cytomegalovirus-driven Thy1.1 as a transduction efficiency marker. Virus was harvested at 48 and 72 hours after transfection, and titer was determined based on percentages of Thy1.1-positive Jurkat T cells after transduction with serially diluted viral supernatant. The titer, calculated as transducing units (TU)/mL of supernatant, was from 2 × 106 to 8 × 106 TU/mL. The virus-containing supernatant was concentrated using an Amicon Ultra Concentrator (Millipore) and stored at −80°C. Gene transduction into CD4+CD25− conventional T cells and CD4+CD25+ Tregs was performed by stimulating cells with plate-bound anti-CD3 (10 μg/mL) and soluble anti-CD28 (1 μg/mL) with 60 U/mL human recombinant IL2 for 16 hours. Activated T cells were transduced with viral supernatants supplemented with 60 U/mL IL2 and 8 μg mL-1 polybrene, followed by centrifugation for 1 hour at 2,500 rpm. After transduction, 20 U/mL human recombinant IL2 (eBioscience) was added to the culture. At 40 hours after transduction, Thy1.1+ Tregs were sorted for western blot and/or suppression assay as indicated.
RNA-seq Analysis
Spleen and peripheral lymph nodes were harvested from YapWT/WT;CD4-Cre-WT and Yap flox/flox (fl/fl); CD4-Cre+ mice (n = 5/group). CD4+ T cells were magnetically enriched, and naïve (CD4+ CD62L+ CD25−) T cells and natural Tregs (nTregs, CD4+ CD62L+/− CD25hi) were flow sorted from each group. For activation condition, sorted nTregs were further activated with 2 μg/mL of plate-coated αCD3 and 2 μg/mL of soluble αCD28 with TGFβ1 (5 ng/mL) and IL2 (100 U/mL) for 24 hours. nTregs (2 × 106; no stimulation or stimulation) from WT and Yap cKO mice were harvested and washed with 1× PBS twice and immediately snap-frozen until further RNA-seq analysis.
Construction of RNA-seq Libraries
Total RNA was isolated by TRIzol from naïve CD4+ T cells or natural Tregs with or without the stimulation anti-CD3/CD28 for 48 hours from WT or Yap cKO mice. RNA quality was monitored on Bioanalyzer. Strand-specific RNA-seq libraries were prepared using the TruSeq Stranded Total RNA LT Sample Prep Kit (with Ribo-Zero Gold, RS-122-2301, Illumina) from 322 ng of total RNA by following the manufacturer's protocols. Briefly, rRNAs were depleted using biotinylated, target-specific oligos combined with Ribo-Zero rRNA removal beads. After purification, RNA was fragmented using divalent cations under elevated temperature, and transcribed into first-strand cDNA using reverse transcriptase and random primers, followed by second-strand cDNA synthesis using DNA Polymerase I and RNase H. A single “A” base was added to these cDNA fragments that were subsequently ligated with the adapter. The products were enriched with 12-cycle PCR. The concentration of final cDNA libraries in 30 μL ddH2O reached 24 to 27 ng/μL as determined on Qubit 2.0.
Analysis of RNA-seq Data
Sequencing was performed on Illumina Hiseq2000 at Beijing Genomics Institute with the type of paired-end, 100 bp. Data quality was assessed by FastQC software (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Mapping to a mouse reference genome (mm10) was conducted by TopHat. Differentially expressed genes were called by Cuffdiff (45). The genes with P value < 0.05 and absolute values of log2-transformed fold changes larger than 1.5 between WT and Yap cKO T cells were considered differentially expressed. A heat map was generated in R statistical software using the geom_tile function under ggplot2 package. Clustering was done with the complete linkage and Euclidean distance using the hclust function in R statistical software. Pathway analysis (Ingenuity) was carried out as described previously (46).
Flow Cytometry
For extracellular staining, harvested cells were washed and incubated in PBS containing 1% FBS containing the below fluorochrome-conjugated antibodies in a U-bottom 96-well plate. For intracellular cytokine staining, harvested cells were restimulated in PMA and ionomycin in the presence of Golgi-Plug (BD Biosciences). After 5-hour incubation, the cells were fixed/permeabilized (eBioscience) and incubated with antibodies (see Supplementary Table S1A for a comprehensive list). For cellular proliferation, cell Trace CFSE cell proliferation kit (Invitrogen) was used per the manufacturer's manual.
Quantitative Real-Time PCR
RNA was extracted using TRIzol (Invitrogen) followed by cDNA synthesis reaction using SuperScript III (Invitrogen) in a 20 μL reaction/well. The same amount of RNA was used in each cDNA synthesis reaction measured by NanoDrop Spectrophotometer (ThermoScientific). The same volume of cDNA per sample was prepared for real-time PCR analysis using SYBR Green (Pierce) and the indicated primers to assess transcript levels of each gene.
Tumor Growth Experiments
Murine B16 melanoma, MC38 colon cancer, and EL4 thymoma cell lines were purchased from the ATCC and kept as frozen stock in 2015. These cell lines have not been authenticated by the laboratory. Cells were cultured in vitro in DMEM plus 10% heat-inactivated FBS and were detached by trypsinization and washed prior to s.c. injection into the shaved side flank of the indicated strains of female mice between the ages of 6 and 8 weeks on a C57BL/6 background (1 × 105 cells). In some experiments, 1 × 104 to 5 × 104 B16 melanoma cells were injected into each mouse in the footpad. Where indicated, once tumors were palpable (7–10 days after injection), 100 mL of 1 × 106 lethally irradiated (150 Gy) B16 GM-vaccine cells (GM-VAX) were injected s.c. into the contralateral limb. A hybridoma cell line expressing a blocking anti–PD-1 antibody (clone G4) was obtained from Dr. Charles Drake. One hundred microgram/mouse/injection of anti–PD-1 (G4) was injected intraperitoneally twice a week once tumors were palpable (7–10 days) in conjunction with vaccine and verteporfin (USP, USP-1711461) treatments. Verteporfin was dosed at 2 mg/mouse diluted to 200 μL with PBS and injected intraperitoneally every two days. Activin neutralization antibodies and isotype control IgG were purchased from R&D Systems. One hundred microgram/mouse/injection of activin-neutralizing antibodies was given intraperitoneally twice a week. For all these experiments, 5 to 10 mice were used per group. Tumor progression was assessed by measuring changes in tumor length (L) and width (W) and tumor volume (V) over time. Tumor volume was calculated using the formula (L × W2)/2.
Molecular Cloning and Site-Directed Mutagenesis
Mouse AcVR1c promoter (1.2 kb) was cloned from the genomic DNA of isolated CD4+ T cells, and the sequence was confirmed. The amplified clones were ligated to SacI/XhoI-digested pGL4.1-Basic Vector (Promega) using the In-Fusion Cloning Kit (Clontech). Site-directed mutagenesis was carried out using the QuikChange Lightning Kit (Agilent Technologies).
Transient Transfection and Luciferase Assay
Jurkat T cells (clone E6-1) were purchased from the ATCC in 2016 and were kept as a frozen stock. This cell line has not been authenticated by the laboratory. Jurkat T cells (1.5 × 107) were transfected with 5 μg pGL4.1-AcVR1c, 1 μg of pRL-TK Vector (Promega), and other indicated plasmids by electroporation using Nucleofector II (Amaxa/Lonza). The cells were rested overnight and stimulated with mock or PMA/ionomycin for 8 hours before being harvested and lysed followed by luminescence measurement using a Dual-Luciferase Assay (Promega) as per the manufacturer's instructions.
ChIP Assay
ChIP assay was performed according to the manufacturer's guidance (Invitrogen MAGnify ChIP system). Briefly, sorted CD4+ iTregs were activated with αCD3/αCD28–conjugated beads overnight and fixed with 2% formaldehyde. Sonicated DNA was immunoprecipitated with anti-YAP1 (Cell Signaling Technology), and anti-TEAD1 (Santa Cruz Biotechnology). The immunoprecipitated chromatin was analyzed on Roche LightCycler 480 by SYBR Green using the following primers for AcVR1c promoter: 5′-CATTGACGTCTCTATGGAAG-3′ (forward), 5′-CAAGCACCATTGCCTTCAGAC-3′ (reverse).
Statistical Analyses
Values are presented as means ± SEM where appropriate. Statistical differences among multiple groups were determined using a two-way analysis of variance (ANOVA) with a Newmane–Keuls multiple comparison test, unless otherwise indicated. Unpaired, two-tailed Student t tests were used for single comparisons. In general, P values <0.05 were considered statistically significant and are indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.002; ****, P < 0.001; ns, not significant. GraphPad Prism 7 was used to calculate P values.
Data Availability
RNA-seq dataset has been uploaded to an appropriate online repository. The GEO accession number is GSE112593.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: D. Pardoll, L. Lu, D. Pan, F. Pan
Development of methodology: J. Tao, Q. Chen, P. Wei, D. Pardoll, L. Lu, F. Pan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Ni, J. Tao, J. Barbi, Q. Chen, B.V. Park, N. Zhang, A. Lebid, A. Ramaswamy, P. Wei, Y. Zheng, X. Wu, P. Vignali, C.-P. Yang, L. Lu, D. Pan
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Ni, J. Tao, J. Barbi, B.V. Park, Z. Li, Y. Zheng, X. Zhang, P. Vignali, D. Pardoll, L. Lu, F. Pan
Writing, review, and/or revision of the manuscript: X. Ni, J. Barbi, B.V. Park, A. Ramaswamy, H. Li, D. Pardoll, L. Lu, D. Pan, F. Pan
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Barbi, B.V. Park, L. Lu
Study supervision: H. Li, D. Pardoll, L. Lu, F. Pan
Acknowledgments
F. Pan's research is supported by the Bloomberg-Kimmel Institute (Immunometabolism Program & Immune Modulation Program), the Melanoma Research Alliance, the NIH (RO1AI099300, RO1AI089830, and R01AI137046), and The DoD (PC130767). J. Barbi's research is supported by the Melanoma Research Foundation, Phi Beta Psi, the Roswell Park Alliance Foundation, and NCI grant P30CA016056. The Li Lab was supported by the National Natural Science Committee of China (No. 81725004) and Shanghai Science and Technology Committee (No. 16410723600). L. Lu's research is supported by the National Natural Science Fund of China (grants 81571564, 1521004, and 81522020) and the Foundation of Jiangsu Collaborative Innovation Center of Biomedical Functional Materials. D. Pan is an investigator of the Howard Hughes Medical Institute.