Research ArticleCancer

Androgen receptor antagonists compromise T cell response against prostate cancer leading to early tumor relapse

See allHide authors and affiliations

Science Translational Medicine  06 Apr 2016:
Vol. 8, Issue 333, pp. 333ra47
DOI: 10.1126/scitranslmed.aad5659

A winning combination

Promising results for cancer immunotherapy in the clinic have led to attempts to combine immunotherapy with standard-of-care therapies, such as androgen deprivation therapy (ADT) for prostate cancer. Now, Pu et al. report that although surgical ADT synergizes with immunotherapy to treat cancer, medical ADT may actually suppress adaptive immune responses and block the efficacy of immunotherapy. However, surgical ADT is irreversible, with ethical, psychological, and surgical concerns. The authors therefore characterized the mechanism of the immunosuppression and found that medical ADT interfered with initial T cell priming rather than reactivation and expansion. These data suggest that careful regulation of timing and dose could improve the efficacy of this combination therapy.

Abstract

Surgical and medical androgen deprivation therapy (ADT) is a cornerstone for prostate cancer treatment, but relapse usually occurs. We herein show that orchiectomy synergizes with immunotherapy, whereas the more widely used treatment of medical ADT involving androgen receptor (AR) antagonists suppresses immunotherapy. Furthermore, we observed that the use of medical ADT could unexpectedly impair the adaptive immune responses through interference with initial T cell priming rather than in the reactivation or expansion phases. Mechanistically, we have revealed that inadvertent immunosuppression might be potentially mediated by a receptor shared with γ-aminobutyric acid. Our data demonstrate that the timing and dosing of antiandrogens are critical to maximizing the antitumor effects of combination therapy. This study highlights an underappreciated mechanism of AR antagonist–mediated immunosuppression and provides a new strategy to enhance immune response and prevent the relapse of advanced prostate cancer.

INTRODUCTION

Prostate cancer is the most prevalent cancer in men. Androgen deprivation therapy (ADT) is initially effective; however, resistance eventually develops in most cases despite aggressive treatment (1). Although medical ADT and orchiectomy are both clinical options, medical ADT with a luteinizing hormone–releasing hormone analog (LHRH-A) is preferred, because orchiectomy is an irreversible procedure involving ethical, psychological, and surgical concerns (24). Notably, in the past decade, LHRH-A therapy has been widely used as a part of the combined androgen blockade (CAB), which consists of LHRH-A and androgen receptor (AR) antagonists. Although there is some debate over the merits of CAB, it currently remains the most popular option for patients with advanced prostate cancer (5). Androgens are growth factors that activate the AR signaling pathway on prostate tissues in both normal and cancerous states (6). AR antagonists such as flutamide, bicalutamide, and enzalutamide are a diverse group of nonsteroidal drugs engineered to counteract the effects of androgens on various organs, including the prostate (7). Recent studies have revealed an immunological component to antiandrogens (8). In addition to stromal cells, AR is also expressed on various types of immune cells, exerting significant influence on both innate and adaptive immune regulations (911). Recent studies using conditional AR knockout mice, with knockout of AR on selective immune cells, revealed that androgen exerts suppressive effects on the development and activation of T and B cells. Removal of such suppression causes thymic enlargement and excessive export of immature B cells (12). Additionally, T cells from orchiectomized mice showed higher peripheral absolute T cell levels but also exhibited stronger antigen- and tissue-specific activation (13, 14). Consistently, after medical castration with LHRH-A, T cell infiltration into the prostate was observed in some patients as well (15). Together, these observations suggest that medical castration potentially plays a similar positive role in the host immune system as orchiectomy (16, 17). However, whether pharmacologic ADT with LHRH-A plus AR antagonists can trigger effective antitumor immunity still remains controversial (18). Additionally, nonsteroidal AR antagonists are identified as belonging to a class of antiandrogens that cause undesirable symptoms, such as seizures, likely mediated through antagonism of the central nervous system (CNS)–based γ-aminobutyric acid type A (GABA-A) receptor by an off-target mechanism (19). Whether these off-target effects impact immune regulation remains unclear.

A number of clinical trials evaluating immunotherapy and medical castration are currently ongoing, highlighting the importance and potential of this combination treatment. Two of the most promising immunotherapeutic approaches are immune checkpoint blockade and therapeutic vaccination, which harness the powerful capabilities of antitumor immunity while overcoming immunosuppression (20). Although anti–programmed cell death protein 1 (PD-1) or anti–programmed cell death ligand 1 (PD-L1) monoclonal antibody (mAb) have been widely used in various cancers and have achieved impressive responses, the lack of clinical response to anti–PD-L1 for prostate cancer patients is still poorly understood (21). Among numerous immunotherapy strategies, the most notable has been the first U.S. Food and Drug Administration–approved therapeutic vaccine, sipuleucel-T (PROVENGE, Dendreon Corp.), which is an autologous vaccine prepared using an individual patient’s peripheral blood mononuclear cells (22). However, the therapeutic efficacy is controversial when this vaccine is combined with antiandrogens in different doses and timing sequences. Given the controversial results, more trials are necessary before new protocol recommendations can be made as to where immunotherapy may fit with current ADT treatment standards.

Here, we used Myc-CaP, a well-known transplantable prostate tumor, to study the role of ADT on immune regulation (23). We showed that medical ADT with AR antagonists unexpectedly has a negative effect on the host immune response, compromising the potential synergistic effects of the combination of immunotherapy and conventional therapy in prostate cancer patients. We also developed a novel combination therapy superior to the current one that can control tumor relapse. This study provides insight into the rational design of combination therapies involving immunotherapy and ADT in prostate cancer.

RESULTS

Surgical orchiectomy synergizes with immunotherapy, whereas medical ADT involving AR antagonists suppresses the immune response

Orchiectomy is reported to enhance dense infiltration of tumors by CD3+ T cells after 1 to 2 weeks following castration (24). Myc-CaP is a semi–hormone-dependent tumor cell line that undergoes massive apoptosis and antigen release after orchiectomy. However, despite initial tumor regression, these tumors frequently relapse (25). We hypothesized that orchiectomy rapidly induces massive apoptosis of tumor cells that release tumor antigens, providing a therapeutic window for immunotherapy. To test whether proper targeting of Toll-like receptors (TLRs) inside tumor tissues during the castration-induced apoptotic stage can activate dendritic cells (DCs) that timely uptake and process excess tumor antigens, we selected CpG, a TLR9 agonist that has been reported to activate DCs for cross-priming (26). Myc-CaP tumors were implanted into mice, and 14 days later, tumors were treated with orchiectomy and/or CpG. CpG alone slightly inhibited tumor growth, whereas orchiectomy alone initially reduced tumor burden due to its androgen-dependent potential. Treatment with a combination of surgical orchiectomy and CpG reduced tumor burden and more effectively delayed tumor relapse than either single treatment (Fig. 1A). We also tested an adenovirus expressing LIGHT (Ad-LIGHT), which not only attracts and activates innate immune cells but also costimulates recruited T cells (27). A similar major reduction in tumor growth was observed when Ad-LIGHT was combined with orchiectomy (fig. S1A). Together, the data suggest that orchiectomy can synergize with immunotherapy.

Fig. 1. Surgical orchiectomy synergizes with immunotherapy, whereas medical ADT involving AR antagonists suppresses the immune response.

(A) FVB male mice were inoculated subcutaneously with 3 × 106 Myc-CaP tumor cells on day −14. Mice received castration or shamed operations on day 0 and/or 30 μg of CpG intratumorally on days 7, 4, and −1. *P < 0.05 for the comparison of mice receiving combination treatment and CpG alone. (B) FVB male mice were inoculated subcutaneously with 3 × 106 Myc-CaP tumor cells on day −14. Mice were intraperitoneally treated with flutamide (60 mg/kg) and leuprolide (10 mg/kg) from day 0 to day 21 daily and/or 30 μg of CpG intratumorally on days 7, 4, and −1. *P < 0.05 for the comparison of mice receiving combination treatment and CpG alone. Flu, flutamide; Leu, leuprolide. (C) Relative efficacy of orchiectomy or flutamide combined with CpG to delay tumor relapse, calculated from (A) and (B). *P < 0.05 for the comparison of mice receiving CpG alone and combination treatment. (D) Myc-CaP tumor–bearing mice were treated as described in (A) and (B). The draining lymph nodes (DLNs) were removed 13 days after the first CpG injection. (E) DLN cells (2 × 105) were stimulated with irradiated tumor cells. The ratio of DLN cells to tumor cells was 10:1. IFN-γ–producing cells were enumerated by ELISPOT assay. Results are expressed as number of spots per 106 DLN cells (*P < 0.05) for the comparison of the CpG + orchiectomy group and the CpG + flutamide group. All of the in vivo representative data were conducted with five mice per group.

Medical ADT is known to be more widely used for patients than orchiectomy. A report from the National Cancer Institute demonstrated that the current most common CAB—LHRH-A (for example, leuprolide or goserelin) plus AR antagonist (for example, flutamide or enzalutamide)—significantly improved patient survival in comparison with orchiectomy (5). To evaluate whether leuprolide plus flutamide will be synergized with CpG, we administered clinical doses of leuprolide and flutamide 14 days after tumor implantation, whereas CpG was injected sequentially. The combination treatment of LHRH-A plus AR antagonists with CpG did not result in more rapid tumor regression compared to the CpG monotherapy group. Rather, it led to earlier relapse (Fig. 1B). These results raised concerns that unlike orchiectomy, CAB—leuprolide plus flutamide—suppresses the immune responses initiated by CpG treatment.

However, surgical and medical castrations (via LHRH-A only) are known to result in the complete regeneration of the male mouse thymus, restoration of peripheral T cell phenotype and function, and enhanced thymus regeneration in various well-established studies (13, 28, 29). Thus, we asserted that it is the AR antagonists, rather than the LHRH-A, that induced the unexpected immunosuppression resulting from the combination of CAB and immunotherapy. To evaluate this supposition, we administered clinical doses of flutamide daily 14 days after tumor implantation, whereas CpG was injected sequentially. Similarly, we observed that the efficacy of CpG was impaired by AR antagonists alone (Fig. 1, C and D). To further determine the impact of orchiectomy and AR antagonists on the immunotherapy-mediated antitumor T cell response, DLN cells from different treatment groups were used for an interferon-γ (IFN-γ) enzyme-linked immunosorbent spot (ELISPOT) assay. The number of tumor antigen–specific IFN-γ–producing T cells significantly increased in the orchiectomy plus CpG group, but not in the AR antagonists plus CpG group (Fig. 1E). Together, these findings indicate that the immunosuppression observed in the combination therapy of CAB and immunotherapy results from AR antagonists.

To test whether the AR antagonist–mediated immune dysregulation is prostate cancer–specific, we used MC38, an anti–PD-L1 responsive tumor model, and treated tumor-bearing hosts with anti–PD-L1 mAb and flutamide. Consistent with clinical data, the additional flutamide attenuated the efficacy of anti–PD-L1 mAb (fig. S1B). These results indicate that surgical orchiectomy plus immunotherapy improves tumor control by enhancing host tumor antigen–specific T cell response, whereas conversely, the common clinically used CAB—LHRH-A plus AR antagonists—may play a negative role during this process because of AR antagonist–induced immunosuppression.

AR antagonists inhibit the broader immune response independently of ARs

Because the combination treatment of immunotherapy and AR antagonists is deleterious to the antitumor-specific immune response, we explored whether AR antagonists themselves could broadly dysregulate the host immune response. Herpes simplex virus (HSV) relapse after drug-mediated immunosuppression is a common clinical occurrence affecting postoperative cancer patients, including prostate cancer patients (30).We used an HSV mouse model to investigate the effects of AR antagonists on infectious disease processes. Mice intramuscularly infected with 5 × 107 plaque-forming units (PFU) of HSV-1 were orchiectomized or treated daily with flutamide 1 day after infection. Progression and severity of infection were measured by weight loss and survival. We found that 60% of the daily flutamide-treated mice lost over 25% of their body weight, but not other groups (Fig. 2A). Fifty percent mortality by 12 days after infection was observed in flutamide-treated groups only. In contrast, no infected mice receiving vehicle or orchiectomy succumbed to infection (Fig. 2B). The loss of body weight and increase of mortality are not due to flutamide-induced liver toxicity because flutamide-treated HSV-1–infected mice or flutamide alone did not show significant increases in alanine transaminase (ALT) and aspartate transaminase (AST), in comparison with the HSV-1–infected control group at day 10 (fig. S2A). Although there is a slight elevation of serum levels of AST and ALT in the flutamide-treated HSV-1–infected moribund animals, we believe that the rise resulted from the systemic inflammatory response induced by lethal herpes simplex encephalitis (HSE) in immunocompromised individuals.

Fig. 2. AR antagonists inhibit the broader immune response independently of ARs.

(A) B6 mice divided into four groups were infected intramuscularly with 5 × 107 PFU of HSV-1 and were treated with flutamide (60 mg/kg) or vehicle from day 0 to day 12 daily, castration, or sham castration before infection, respectively. Mice were weighed every 4 days. **P < 0.01 represents the mean percent change in body weight between the flutamide-treated group and the vehicle-treated group on day 30. (B) B6 mice divided into four groups were infected intramuscularly with 5 × 107 PFU of HSV-1 and were treated with flutamide (60 mg/kg), leuprolide (10 mg/kg), or vehicle from day 0 to day 12 daily, castration, or sham castration before infection, respectively. The survival rate was calculated using Kaplan-Meier analysis. LHRH-A, leuprolide. (C) Mice were vaccinated subcutaneously with OVA and alum adjuvant on day 0. Castration or sham operation was performed 5 days before vaccination. Flutamide (60 mg/kg) or vehicle was administrated from day 0 to day 12 daily. Twelve days later, the serum was collected and subjected to enzyme-linked immunosorbent assay (ELISA) to measure the anti-OVA IgG concentrations of the mice. *P < 0.05. (D) Mice were vaccinated subcutaneously with 1 × 107 irradiated B16-OVA tumor cells on day 0. Castration was performed 5 days before vaccination. Flutamide (60 mg/kg) was administrated from day 0 to day 6 daily. Nine days after vaccination, DLNs were removed and subjected to ELISPOT assays. *P < 0.05; ns, not significant. Representative data are shown from two (A and B) or three (C and D) experiments conducted with five mice per group.

In addition, hindlimb paralysis, a sign of HSE, indicative of the virus invading the nervous system, was observed before death. The log-rank (Mantel-Cox) test indicated that mice subjected to flutamide treatment died at a significantly faster rate than mice subjected to either vehicle or orchiectomy (P = 0.0045). These studies imply that AR antagonists may down-regulate the virus-specific T cell response leading to lethal HSE.

We next examined the impact of AR antagonists on the generation of humoral immunity to thymus-dependent antigens in vivo. We considered whether the flutamide-induced immunosuppression depends on the AR pathway. Mice were subcutaneously immunized with ovalbumin (OVA) protein emulsified with alum adjuvant on day 0. Clinical doses of flutamide were intraperitoneally administered daily over a period of 14 days commencing on day 0, and orchiectomy was performed on day 0. We divided the flutamide-treated group into two subsets, five mice in each subset (see legend Fig. 2C). To determine whether the effect of flutamide on the humoral immune response is androgen-dependent, we treated one set of mice with flutamide alone, whereas the second set received orchiectomy before vaccination, followed by flutamide treatment. Anti-OVA immunoglobulin G (IgG) antibodies were measured on day 14 as an indicator of the humoral immune response. As predicted, flutamide-treated mice failed to produce vigorous humoral responses to OVA, whereas the other groups generated significantly higher anti-OVA IgG antibodies (Fig. 2C). After creating androgen-null mice by precastration, flutamide treatment still showed inhibition of anti-OVA IgG production (Fig. 2C). This suggests that the immunosuppression effect of flutamide is androgen-independent.

We then sought to determine whether the loss of antibodies against OVA antigens is accompanied by loss of cellular immune responses. To explore this possibility, we used an experimental tumor vaccine model. We introduced a model antigen, OVA, into B16 melanoma to more easily track the T cell response. Mice vaccinated with irradiated B16-OVA tumor cells received flutamide or vehicle control treatments. DLN cells were collected on day 14 after vaccination, and IFN-γ ELISPOT assays were performed with exogenous OVA protein stimulation. We observed a marked decrease in IFN-γ–producing T cells after flutamide treatment (Fig. 2D). Similarly, we performed castration before flutamide treatment to exclude the influence of androgen. In the androgen-null mice, we still observed a marked decrease in the T cell response after the flutamide treatment (Fig. 2D). To further prove the androgen-independent regulation of the drug, we used an androgen-independent melanoma B16-humanEGFRhigh tumor model, which is a regressive tumor cell line that cannot grow in wild-type mice because of high immunogenicity. Wild-type mice support the aggressive growth of this tumor when treated with flutamide (fig. S2B). These results further solidify the concept that AR antagonists can broadly dysregulate the host immune response independently of androgen-related immune interplay.

AR antagonists inhibit the immune responses by suppressing T cells

There are studies that mentioned that the infiltration of intratumoral leukocyte and T cell functions have been enhanced in prostate cancer patients after medical ADT (1517). To verify these results, we used an in vitro syngeneic culture system to examine a structurally diverse set of drugs spanning AR antagonists (flutamide and enzalutamide), an upstream gonadotroph inhibitor (LHRH-A), and an androgen synthesis inhibitor (abiraterone). Splenocytes from B6 wild-type mice were stimulated with anti-CD3 and anti-CD28 in vitro and treated with different antiandrogen drugs simultaneously. Different from the enhanced immune readouts in previous studies, we observed that nonsteroidal AR antagonists significantly inhibited IFN-γ production, whereas LHRH-A, the upstream gonadotroph inhibitor, and the androgen synthesis enzyme inhibitor did not suppress IFN-γ production (Fig. 3A). These results are consistent with our in vivo data (Fig. 1). Because IFN-γ can be produced by either T cells or antigen-presenting cells (APCs), these responses in unfractionated splenocytes could reflect a direct effect on either T cells, APCs, or both. To determine which cells are directly responsible for this phenotype, we separately tested T cells and APCs. Purified T cells were directly stimulated with anti-CD3 and anti-CD28 in vitro, producing polyclonal expansion. Similar to unfractionated splenocytes, purified T cells exhibited diminished production of IFN-γ in response to nonsteroidal AR antagonists. In addition, splenocytes taken from Rag1−/− mice, which contain only innate immune cells and no adaptive lymphocytes, were directly stimulated with LPS. However, in contrast to T cells, the AR antagonists showed no significant direct effects on the production of IFN-γ and TNF-α (tumor necrosis factor–α) in Rag1−/− mice splenocytes, indicating that nonsteroidal AR antagonists may not severely affect APCs function (Fig. 3A).

Fig. 3. AR antagonists inhibit the immune response by suppressing T cells.

(A) Splenocytes from B6 wild-type (WT) mice and Rag1−/− knockout mice were stimulated with soluble anti-CD3 (5 μg/ml), anti-CD28 (2.5 μg/ml), and lipopolysaccharide (LPS) (1 μg/ml), respectively, in the presence of 10 μM flutamide, 10 μM enzalutamide, 10 μM abiraterone, and 10 μM leuprolide. After 48 hours, the concentration of IFN-γ in supernatants was measured by BD cytometric bead array (CBA) kit. T cells were purified from B6 WT mice splenocytes by Mouse T Cell Isolation Kit from STEMCELL Technologies and stimulated with soluble anti-CD3 (5 μg/ml) and anti-CD28 (2.5 μg/ml), with antiandrogens added as indicated. (B and C) Splenocytes from OTI and OTII transgenic (Tg) mice were stimulated with OTI and OTII peptide (5 μg/ml), respectively, and treated with 10 μM flutamide, 10 μM enzalutamide, 10 μM abiraterone, and 10 μM leuprolide. After 48 hours, the concentration of IFN-γ in supernatants was measured by BD CBA kit. IL-2, interleukin-2. (D) B6 WT mice were inoculated subcutaneously with 1 × 106 B16-OVA tumor cells on day 0. The DLNs were removed 13 days after tumor inoculation. DLN cells (2 × 105) were stimulated with irradiated tumor cells in the presence of 10 μM flutamide, 10 μM enzalutamide, 10 μM abiraterone, and 10 μM leuprolide. The ratio of DLN cells to tumor cells was 10:1. IFN-γ–producing cells were enumerated by BD CBA kit. MFI, mean fluorescence intensity. (E) Myc-CaP tumor–bearing mice were treated as described in Fig. 1. DLNs were removed 13 days after the first CpG injection, and CD8+ T cells and DCs in DLN were isolated from DLN for ELISPOT analysis. T cells (2.5 × 105) from mice treated with CpG or a combination of CpG and flutamide were cocultured with naïve DC stimulated with irradiated Myc-CaP tumor cells. Tumor-bearing DCs loaded with irradiated Myc-CaP tumor cells were cocultured with T cells from irradiated Myc-CaP–vaccinated mice. The ratio of T cells to APCs was 20:1. IFN-γ–producing cells were enumerated by ELISPOT assay. Results are expressed as number of spots per 106 DLN T cells. Data are reported as a mean spots number ± SEM. *P < 0.05; **P <0.01(unpaired Student’s t test). One of three experiments is shown.

To further test which subset of T cells is involved, we used naïve splenocytes from OTI transgenic (CD8+ T cells for OVA) or OTII transgenic (CD4+ T cells for OVA) mice challenged with OTI and OTII peptide in vitro, respectively. Similar to the suppression effect observed in total T cells from B6 wild-type mice, only the nonsteroidal AR antagonists inhibited OTI and OTII T cell activation in vitro, as indicated by reduced cytokine production (Fig. 3, B and C). To rule out the potential contribution of B cells to the observed decrease in CD4 T cell cytokines and impaired vaccine efficacy, we analyzed peripheral CD19+CD11C B cell proliferations and found no substantial alteration in B cell proliferation when cocultured with AR antagonists (fig. S3). Furthermore, there was no change in cell viability of splenocytes with the treatment, suggesting the impaired cytokine production was not due to the general cytotoxicity of the drugs.

To determine whether these drugs affect antigen-experienced tumor-specific cytotoxic T lymphocytes besides naïve T cells, we next treated tumor-DLNs with these agents in vitro and measured T cell activation. We established a B16-OVA tumor cell line in which the OVA protein can serve as a surrogate marker that can be specifically recognized by endogenous or OTI CD8+ T cells from transgenic mice. DLNs were isolated from tumor-bearing mice, treated with different antiandrogen drugs, and stimulated with B16-OVA tumor cells in vitro. IFN-γ release was used to assess T cell recognition of tumor cells. A significant reduction in IFN-γ mean fluorescence intensity was observed when B16-OVA tumor cells were cocultured with flutamide- or enzalutamide-treated T cells, but not with other drugs (Fig. 3D).

Next, we investigated whether the effect of AR antagonists on T cells or DCs is sufficient for the change in the adaptive immune response observed in vivo. An antigen-specific system to track the priming and activation of tumor antigen–specific T cells was used. First, the direct impact on T cells was tested. DCs from naïve mice were pulsed with irradiated Myc-CaP cells to be used as tumor antigen–loaded APCs. T cells from DLNs of CpG- and flutamide-treated tumor-bearing mice were cocultured with the above-mentioned APCs. The number of antigen-specific IFN-γ–producing T cells of this system was significantly lower compared with those that were treated with CpG alone, suggesting that the direct impact of T cells may be a major contributor to the suppressive effect of flutamide. Second, we compared the function of DCs from DLNs of mice receiving CpG and flutamide combination treatment. DCs stimulated with irradiated Myc-CaP tumor cells were cocultured with the purified T cells from irradiated Myc-CaP–vaccinated mice. The specific IFN-γ–producing T cells were not significantly decreased compared to those treated with CpG alone, indicating that APC function was not significantly affected by antiandrogens (Fig. 3E). These data collectively suggest that nonsteroidal AR antagonists mediate immunosuppression in tumor-experienced T cell response.

AR antagonists inhibit the initial T cell activation

Because our data reveal that AR antagonists have a broader role in suppressing T cell responses, they might regulate the early stages of T cell priming and/or the later stages of T cell stimulation. To distinguish these two possibilities, we first examined whether AR antagonists could silence T cell effector functions or whether it has direct apoptosis-inducing effects, because induction of apoptosis is a key mechanism during T cell development and immune response termination (31, 32). We determined apoptosis of splenocytes after flutamide treatment, and we observed that flutamide showed only minimal apoptosis-inducing effects even at high physiological concentrations (fig. S4A). Thus, we hypothesized that AR antagonists create an anergic state in the early phase of T cell activation. To test this hypothesis, we divided the experiment into two groups. One group of splenocytes was stimulated with OTI peptide 6 hours before flutamide; the second group was stimulated with OTI peptide 24 hours before flutamide administration. After 48 hours, a CBA assay was performed to measure effector cytokine. Our results demonstrated that flutamide no longer inhibited the cytokine production of splenocytes when the addition of the drug was delayed 24 hours in vitro (Fig. 4, A and B). In addition, neither flutamide nor enzalutamide demonstrated significant changes to the T cell response to phorbol 12-myristate 13-acetate (PMA) and ionomycin, indicating that the inhibition of T cell activation may be downstream of the TCR (T cell receptor)–CD3 complex–gated signal pathway, but PMA/ionomycin did not induce the activation axis (fig. S4B). Together, these data indicate that AR antagonists–mediated T cell defects are determined in the priming phases of T cell activation, which suggests that the synergistic combination of immunotherapy and medical castration requires accurate optimization of the treatment schedule to avoid inhibitory effects.

Fig. 4. AR antagonists inhibit initial T cell activation.

(A) Splenocytes from OTI transgenic mice were stimulated with OTI peptide (10 μg/ml) in the presence of 10 and 25 μM flutamide or vehicle. After 6 hours, the concentration of indicated cytokines in supernatants was measured by BD CBA kit. *P < 0.05 for the comparison of mice receiving vehicle and 10 μM flutamide. (B) Splenocytes from OTI transgenic mice were preactivated with OTI peptide (10 μg/ml) for 24 hours. After preactivation, cells were further stimulated with 10 and 25 μM flutamide or vehicle for an additional 24 hours. The concentration of indicated cytokines in supernatants was measured by BD CBA kit. (C) FVB male mice were inoculated subcutaneously with 3 × 106 Myc-CaP tumor cells on day −14 and treated with 30 μg of CpG or control intratumorally on days 7, 4, and −1. Flutamide was administrated intraperitoneally before (day −5 to day −2), after (day 9 to day 21), or during (day 0 to day 21) CpG treatment. **P < 0.01. One of two representative experiments is shown (A to C).

We speculated that there might be a window of time in vivo when AR antagonists can effectively reduce tumor burden without interfering with the host immune system. To determine whether the synergy of AR antagonists with combination therapy is dependent on sequential delivery, we conducted in vivo experiments with two treatment protocols: one completed treatment with CpG on days −1, 4, and 7 before the start of flutamide; the second one began CpG treatment simultaneously with flutamide. In the first protocol, the flutamide-induced immunosuppression was abrogated, whereas in contrast, flutamide still played an antagonistic role in the second protocol (Fig. 4C). This result is also supported by phase 2 clinical trials featuring patients with nonmetastatic castration-resistant prostate cancer (CRPC) who receive vaccination before AR antagonists. Results from these clinical trials show that this strategy may potentially result in improved survival compared with patients who received vaccination after hormone therapy (33). Therefore, we believe that the inhibiting role of AR antagonists in combination therapy for prostate cancer can be corrected by sequential administration after the host immune system has been strongly activated by immunotherapy.

The impaired immune response might be correlated with an off-target effect of GABA-A inhibition

It is reported that nonsteroidal AR antagonists, including flutamide and enzalutamide, have an off-target mechanism through inhibiting GABA-A currents (19). GABA is known as the principal inhibitory neurotransmitter in CNS (34). In addition to its well-known roles in CNS, engagement of GABA also modulates inflammation, and GABA receptor transcripts are reported to present in immune cells (35). To investigate whether AR antagonists could suppress T cell activation through off-targeting the GABA-A signal pathway, splenocytes from B6 wild-type mice and OTI and OTII transgenic mice were stimulated with anti-CD3 and anti-CD28, OTI and OTII peptide in the presence or absence of GABA and/or flutamide, respectively. Less IFN-γ was produced in groups treated with GABA or flutamide alone than the vehicle-treated group. However, additional flutamide added to the GABA-treated group did not further impair cytokine production, implying that GABA and flutamide are likely able to activate the same inhibitory pathway on T cells (Fig. 5A). Because there are two GABA receptors (GABA-A and GABA-B), we next dissected which GABA receptor antagonist is possible to rescue flutamide-mediated suppression. We found that the GABA-A competitive receptor antagonist bicuculline rather than the GABA-B competitive receptor antagonist CGP greatly neutralized the flutamide-induced immune inhibition. Because GABA-A receptors are ligand-gated ion channels that respond to GABA by opening their integral Cl channel, we additionally used picrotoxin, a GABA-A receptor Cl channel blocker, to further confirm it. The inhibitory effect of GABA was greatly reduced by the GABA-A competitive receptor antagonist bicuculline and almost completely abolished by the Cl channel blocker picrotoxin (Fig. 5, A to C). Similarly, human T cells stimulated by anti-CD3 are inhibited by AR antagonists in a dose-dependent manner in vitro, and immune inhibition could be corrected when T cells were cocultured with AR antagonists and GABA-A receptor antagonists together (fig. S5).

Fig. 5. The impaired immune response might be correlated with an off-target effect of GABA-A inhibition.

(A) Splenocytes from B6 WT mice were stimulated with soluble anti-CD3 (5 μg/ml) and anti-CD28 (2.5 μg/ml). Flutamide (10 μM), 1 mM GABA, 100 μM bicuculline, and 30 μM picrotoxin were added to the culture as indicated. After 48 hours, the concentration of indicated cytokines in supernatants was measured by BD CBA kit. *P < 0.05. (B and C) Splenocytes from OTI and OTII transgenic mice were stimulated with OTI and OTII peptide (5 μg/ml), respectively. Flutamide (10 μM), 1 mM GABA, 100 μM bicuculline, and 30 μM picrotoxin were added to the culture as indicated. After 48 hours, the concentration of indicated cytokines in supernatants was measured by BD CBA kit. *P < 0.05. (D) B6 WT mice were inoculated subcutaneously with 1 × 106 B16-OVA tumor cells on day 0. The DLNs were removed 13 days after tumor inoculation. DLN cells (2 × 105) were stimulated with irradiated tumor cells in the presence of 10 μM flutamide, 1 mM GABA, 100 μM bicuculline, and 30 μM picrotoxin. After 48 hours, the concentration of indicated cytokines in supernatants was measured by BD CBA kit. *P < 0.05. (E) B6 WT mice were injected subcutaneously with 2 × 106 B16–EGFR (enhanced green fluorescent protein) tumor cells. Flutamide (60 mg/kg), GABA-A receptor antagonist bicuculline (5 mg/kg), and GABA-B receptor antagonist CGP 55845 (10 mg/kg) were administrated intraperitoneally from day 0 to day 15 daily as indicated. Representative data are shown from experiments with five mice per group. *P < 0.05. Tumor volume was measured and compared twice a week. One of three representative experiments is shown (A to E).

To assess the functional impairment of tumor-experienced T cells, we isolated splenocytes from B16-OVA tumor–bearing mice and treated them with flutamide, GABA, and its antagonists in vitro, and used B16-OVA tumor cells to stimulate IFN-γ production. Data revealed that bicuculline and picrotoxin are able to rescue the diminished T cell recognition induced by AR antagonists in cytotoxic T cell responses against tumors (Fig. 5D). These results suggest that the off-target immunosuppression induced by flutamide might involve the GABA-A pathway.

To determine whether GABA-A receptor antagonists can abrogate flutamide-mediated T cell suppression in vivo, mice were implanted with immunogenic B16-humanEGFRhigh tumor cells and treated with flutamide alone, bicuculline alone, CGP alone, flutamide plus bicuculline, or flutamide plus CGP. In contrast to rejection in wild-type mice, only tumors in mice treated with flutamide alone grew progressively, whereas other agents in single treatment–treated mice showed almost no impact on tumor growth. Treatment with a combination of flutamide and bicuculline effectively controlled tumor growth, whereas tumors treated with a combination of flutamide and CGP were not rejected in wild-type mice (Fig. 5E). These results collectively suggest that AR antagonist–mediated T cell impairment may in part engage through a GABA-A receptor pathway on T cells, which can be manipulated pharmacologically in a manner similar to that of neuronal GABA-A receptors.

The combination of immunotherapy and inhibitors of androgen biosynthesis achieves synergy

To thoughtfully consider an alternative to medical ADT that does not suppress T cell response, we explored whether abiraterone, a steroidal compound that interferes with androgen synthesis, could synergize with immunotherapy by avoiding off-target of GABA-A receptors. To test whether abiraterone could potentiate superior tumor rejection compared to classic medical castration (leuprolide + flutamide), we combined abiraterone with CpG treatment and observed that abiraterone further enhanced the long-term efficacy of CpG. Mice administered with a low clinical dose of abiraterone combined with CpG effectively reversed the growth of established tumors for at least 30 days after treatment (Fig. 6A). Emerging data show that a high dose of abiraterone might also inhibit 3βHSD (3β-hydroxysteroid dehydrogenase) isoenzymes and antagonize AR, suggesting that reverse abiraterone resistance is caused by sustained steroidogenesis (36). Accordingly, we tried combining a high dose of abiraterone with CpG. Notably, abiraterone itself worked moderately better than CpG alone at high doses, and the combination treatment significantly increased the survival of tumor-bearing mice, with near-completed tumor regression and all mice surviving 50 days after tumor implantation (Fig. 6, B and C). Collectively, these data suggest that this combination-based strategy could increase the overall response and cure rates of CRPC patients, even in hosts that initially fail to respond to medical castration.

DISCUSSION

Over the past decade, several studies have attempted to integrate immunotherapy into the existing standard treatments, such as ADT, radiotherapy, or chemotherapy, to improve overall survival in tumor-bearing hosts (3740). The rationale for a reciprocal potentiating action of ADT and immunotherapy is promising, but the results have not been convincing so far (33, 41, 42). Here, we have investigated the immune protection of combination treatment since having observed a promising synergistic effect of immunotherapy combined with orchiectomy. In contrast, combining some medical ADT with immunotherapy often fails to have long-lasting protection. We try to unravel this problem from three perspectives: (i) the functional significance of AR antagonist–induced immunosuppression in turning the regressive tumor into a progressive tumor and abolishing the synergistic effect of the combined immunotherapy, (ii) the molecular mechanism of AR antagonist–mediated immunogenic modulation involving T cell interference at its priming phase, and (iii) the AR-independent immunogenic modulation properties of AR antagonists, in part related to an unexpected off-target mechanism through the GABA-A receptor. In this context, the overall immune response may be complicated because the immunosuppressive effect of AR antagonists through targeting the GABA-A receptor could potentially be neutralized by its direct immune enhancement through blocking the AR. This could explain improved overall survival through the combination of AR antagonists and immunotherapy in some previous studies.

The mechanisms of the AR antagonist–mediated immunosuppression are complicated. First, in previous studies, researchers who observed an improved survival curve in enzalutamide and Twist vaccine combinational therapy also discovered an impairment of tumor-experienced CD4+ T cell proliferation and reduction of the antigen-specific T cell response in the same study model (18). These immune-related findings are particularly similar to our studies. Second, we have proven that the synergistic effect is independent of the unexpected off-target binding to GABA-A receptors. As shown in our study, only through binding to GABA-A receptors can the drug impair its synergistic effect with immunotherapy. Third, the treatment sequence plays a role in the synergistic effect because these drugs only affect the T cell priming phase. Previous studies and ours all have observed improved survival in proper combination and sequential administration. Accordingly, our findings are complementary to previous studies rather than in conflict with them. The AR antagonists, such as flutamide and enzalutamide, act as a double-edged sword in the immune system. AR antagonists activate the immune system through inducing tumor cell apoptosis, causing thymic enlargement, and increasing leukocytes and B cell migration (42). On the other hand, AR antagonists could suppress the immune system by an off-target mechanism, independent of the AR pathway.

As shown with several antiandrogen agents, some AR antagonists inhibited the T cell proliferative response to anti-CD3, as well as antigen-specific stimulation, in a dose-dependent manner in vitro. The inhibitory effect of these agents on T cell proliferation was associated with greatly reduced IL-2 and IFN-γ production by antigen-primed T cells. Furthermore, AR antagonists inhibited the ability of mice to mount both humoral and cellular responses to the antigens. Thus, these agents down-regulated T cell functions both in vitro and in vivo. The binding of TCR-CD3 co-receptors to the peptide/MHC (major histocompatibility complex) complex activated T cells through a cascade of intracellular messengers that triggered Ca2+ influx, Ca2+-dependent phosphorylation, and IL-2 gene expression. AR antagonists inhibited both antigen- and anti-CD3–primed T cell proliferation, suggesting interference with the signal transduction triggered by the TCR-CD3 complex. PMA and ionomycin overcame this inhibition by directly opening Ca2+ channels and activating downstream signal transduction, indicating that the action of antiandrogens was signaling events upstream. Our data suggest that these defects might be associated with GABA-A receptor off-target binding, although Bhat et al. claimed that GABA-A receptors played a more direct role in regulating APCs than T cells, based on an in vitro peritoneal macrophages activation assay. Additionally, in other studies, researchers also reported that GABA-A receptors could interfere with T cell function at its activation stage and showed that purified T cells would be negatively regulated in the presence of GABA agents (43). There are a few possibilities leading to the difference: (i) most of their data were from in vitro peritoneal macrophages stimulated with LPS stimulation, whereas DCs were measured only for the secretion of IL-1β and IL-6. However, we measured the function of DCs via an ex vivo tumor-experienced and antigen-specific assay that is closer to the clinical scenario. (ii) Our observations indicated that the mechanism may involve the GABA-A pathway, but other potential pathways cannot be excluded, such that we cannot consider that the effect of GABA-A on immune cells is entirely equivalent to the effect of AR antagonists on lymphocytes. Together, these results and ours demonstrate the importance of AR antagonists and their possible association with GABA-A receptors in the functional modulation of immune cells, with the detailed mechanism remaining to be clarified.

Our observations have been supported by various clinical trials as well. Although some clinical trials showed survival benefits, others showed even higher rates of relapse when combining medical ADT with immunotherapy (18, 33). These varied outcomes are thought to be related to the different initial tumor microenvironments, especially in cases of anti–CTLA-4 (cytotoxic T lymphocyte antigen-4) or anti–PD-L1 mAb therapy (44). Our data uncovered that nonsteroidal AR antagonists directly inhibit T cell priming. Therefore, administering immunotherapy before AR antagonists could have a synergistic impact by temporally inhibiting the suppression of T cell priming. In our murine model, the AR antagonist–induced suppression in the combination therapy could only be alleviated, but not completely reversed, when CpG was administered before AR antagonists. In a recent clinical study, patients with nonmetastatic CRPC were randomized to receive either a poxvirus-based vaccine or the antiandrogen nilutamide, with the possibility of crossover at the time of disease progression. Patients who received the vaccine and subsequently nilutamide showed a significantly improved survival rate, compared with patients who received nilutamide first (33). Another randomized phase 2 trial evaluated the efficacy of sipuleucel-T by administering it either before or after AR antagonists. Cytokine responses and CD8+ T cell activation were higher when sipuleucel-T was administered after flutamide, suggesting an enhanced immune response (18, 45). Disparity in outcomes may result from different doses of AR antagonists administered and various immunotherapies that affect different stages of the immune response.

The rationale of medical ADT is to exert a positive effect on the activation and functional modulation of immune cells. Therefore, combining immunotherapy with medical ADT without off-target immunosuppression may optimize the efficacy of combination therapy. Our current study demonstrates that a selective combination of CpG and the androgen biosynthesis inhibitor abiraterone can have a synergistic effect on tumor regression.

In conclusion, our study revealed an unintended suppressive role of AR antagonists in the host immune system, which could impair the efficacy of combined immunotherapy. These results emphasized that the proper use of combination therapies could promote the immunotherapy-initiated antitumor immune response and suggest a new clinically relevant solution for better tumor control.

MATERIALS AND METHODS

Study design

The primary research objectives were to characterize the impact of ADT on the antitumor immune response and to evaluate combinatorial strategies of ADT with immunotherapy. The overall study design was a series of controlled laboratory experiments in mice, as described in the sections below. In all experiments, animals were randomly assigned to various experimental groups. The experiments were replicated two to three times as noted, and the final analysis included either pooled data or representative experiments where indicated. For in vivo experiments, five mice per group were used for each experiment, with at least two replicates. All outliers were included in the data analysis.

Mice

FVB male mice and C57BL/6 OTI and OTII mice were purchased from The Jackson Laboratory, and C57BL/6 Rag1−/− knockout mice were purchased from Harlan at 6 to 7 weeks of age. All mice were maintained under specific pathogen–free conditions. Animal care and use conditions were followed in accordance with institutional and National Institutes of Health protocols and guidelines, and all studies were approved by the Animal Care and Use Committee of the University of Chicago.

Cell lines and reagents

Myc-CaP is a murine prostate tumor cell line derived from the spontaneous prostate tumor of Myc transgenic mice (46). MC38 is a murine colon adenocarcinoma cell line [American Type Culture Collection (ATCC)]. B16-EGFR was selected from a single clone after being transduced by lentivirus expressing human EGFR (L858R). Myc-CaP, MC38, and B16-EGFR were cultured in 5% CO2 and maintained in vitro in Dulbecco’s modified Eagle’s medium supplemented with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich), 2 mM l-glutamine, 0.1 mM minimum essential medium (MEM) nonessential amino acids, penicillin (100 U/ml), and streptomycin (100 μg/ml). Myc-CaP, MC38, and B16-EGFR were cultured in 5% CO2 and maintained in vitro in Dulbecco’s modified Eagle’s medium (Corning Corp.) supplemented with 10% heat-inactivated fetal bovine serum (Sigma), 2 mM l-glutamine, 0.1 mM MEM nonessential amino acids, penicillin (100 U/ml), and streptomycin (100 μg/ml). T cells were cultured in modified RPMI-1640 medium (Corning) supplemented with 10% heat-inactivated fetal bovine serum (Sigma), 2 mM l-glutamine, 0.1 mM MEM nonessential amino acids, penicillin (100 U/ml), and streptomycin (100 μg/ml). Chemotherapeutic agents flutamide, bicalutamide (CDX), bicuculline, CGP-13501, leuprolide, picrotoxin, and GABA were purchased from Sigma and prepared according to the manufacturer’s recommendations. Anti–PD-L1 (10F.9G2) was purchased from Bio X Cell. Enzalutamide and abiraterone acetate were purchased from Toronto Chemicals Inc.

HSV-1 infection and analysis

The HSV-1F strain was provided by T. Kristie, Laboratory of Viral Diseases/National Institute of Allergy and Infectious Diseases/National Institutes of Health, amplified in Vero cells (ATCC), collected from cell supernatant, and purified through a sucrose-dextran gradient centrifuge with 5 × 107 PFU of HSV-1 in 40 μl. Phosphate-buffered saline (PBS) was subcutaneously injected into the footpads of the mice.

Tumor growth and treatments

About 3 × 106 Myc-CaP or 1 × 106 MC38 or 2 × 106 B16-EGFR tumor cells were subcutaneously injected into the flank of the mice. Tumor volumes were measured by length (a), width (b), and height (c) and calculated as tumor volume = abc/2. For the antiandrogen agent combination experiment, tumors were allowed to grow for 9 to 14 days, and mice were treated with CpG (30 μg) or PBS (50 μl) intratumorally, and flutamide (60 mg/kg), leuprolide (10 mg/kg) intraperitoneally, or orchiectomy. For the B16-EGFR tumor model, mice were treated with flutamide (60 mg/kg) and/or with GABA-A receptor antagonist bicuculline (5 mg/kg) or GABA-B receptor antagonist CGP 55845 (10 mg/kg) daily since tumor inoculated. For the abiraterone treatment experiment, mice were treated with low-dose abiraterone (10 mg/kg) or high-dose abiraterone (50 mg/kg) intraperitoneally.

Surgeries

Mice were anesthetized using ketamine and xylazine. Surgical preparation consisted of cleansing the surgical site with three alternating wipes of 70% alcohol and betadine solution and applying a surgical drape. Maintenance of a surgical plane consisted of monitoring the respiratory rate and character, and response to toe pinch or lack thereof. A midline scrotal incision was made, followed by the ligation of the right and left spermatic cords using sterile monofilament absorbable nylon sutures. Testes were excised distal to the ligature, and the incision was closed with sterile dissolvable sutures. The surgery was performed in a biosafety cabinet that was disinfected with Virkon S broad spectrum disinfectant solution before use.

Detection of antibodies by ELISA

Serum samples were collected and examined for presence of antibodies to antigens by ELISA. Experimental values from separate experiments were expressed as nanogram per milliliter.

T cell isolation

OTI naïve CD8+ T cells were isolated from the lymph nodes (LNs) and spleens of 6- to 12-week-old OTI transgenic mice. Purification was carried out with a negative CD8 isolation kit (STEMCELL Technologies) following the manufacturer’s instructions.

Measurement of IFN-γ–secreting T cells by ELISPOT assay or CBA assay

Antitumor-specific T cells were measured by ELISPOT assay. Spleen or LN cells were resuspended in an RPMI-1640 medium supplemented with 10% fetal calf serum, 2 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 μg/ml). A total of 1 × 105 to 2 × 105 spleen or LN cells were used for the assay. Irradiated tumor cells were added at a ratio of 1:10 to spleen or LN cells. After 48 hours of incubation, IFN-γ was determined with an IFN-γ ELISPOT assay kit according to the manufacturer’s manual (BD Biosciences). The visualized cytokine spots were enumerated with the ImmunoSpot Analyzer (CTL).

In vitro OTI T cell activation assay

Purified OTI T cells were stimulated with OTI peptide (10 μg/ml). Forty-eight hours later, the supernatants were collected, and IL-2, TNF, and IFN-γ were measured by CBA assay (BD Biosciences).

Flow cytometric analysis

Single-cell suspensions of cells were incubated with anti-CD16/32 (anti-FcγIII/II receptor, clone 2.4G2) for 10 min and then stained with conjugated antibodies. All fluorescently labeled mAbs were purchased from BioLegend or eBioscience. Samples were analyzed on a FACS (fluorescence-activated cell sorting) Fortessa flow cytometer (BD Biosciences), and data were analyzed using FlowJo software (Tree Star).

Statistical analysis

No statistical method was used to predetermine sample size. Mice were assigned at random to treatment groups for all mice studies and, where possible, mixed among cages. There were no inclusion/exclusion criteria. Whenever possible, the investigators were blinded to group allocation during the experiments and when assessing outcomes. Experiments were repeated two to three times. Data were analyzed using Prism 5.0 software (GraphPad) and presented as means ± SEM. The P values were assessed using a two-tailed unpaired Student’s t test or a two-way analysis of variance (ANOVA), with P values considered significant as follows: *P < 0.05; **P < 0.01; and ***P < 0.001. For tumor-free mice frequency, statistics were done with log-rank (Mantel-Cox) test.

Fig. 6. The combination of immunotherapy and inhibitors of androgen biosynthesis achieves synergy.

(A) FVB male mice were inoculated subcutaneously with 3 × 106 Myc-CaP tumor cells on day −14. Mice were treated with low-dose abiraterone (10 mg/kg) from day 0 to day 21 daily and/or 30 μg of CpG intratumorally on day 7, day 4, and day −1. Tumor volume was recorded twice weekly. **P < 0.01 for the comparison of mice receiving the combination treatment and CpG alone; one of two representative experiments is shown, with five mice in each group (means ± SEM). (B) FVB male mice were inoculated subcutaneously with 3 × 106 Myc-CaP tumor cells on day −14. Mice were treated with a high dose of abiraterone (50 mg/kg) from day 0 to day 21 daily and/or 30 μg of CpG intratumorally on day 7, day 4, and day −1. *P < 0.05 for the comparison of mice receiving combination treatment and abiraterone alone. (C) Mice were monitored for overall survival. One of two representative experiments is shown, with five mice in each group (means ± SEM). The survival rate was calculated using a Kaplan-Meier analysis. CTR, control. **P < 0.01 for the comparison of mice receiving combination treatment and abiraterone alone.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/8/333/333ra47/DC1

Fig. S1. AR antagonists suppress various tumor immunotherapies in prostate cancer.

Fig. S2. The immunosuppression mediated by flutamide is not mediated by liver toxicity and is independent of AR.

Fig. S3. AR antagonists inhibit the immune response by suppressing T cells.

Fig. S4. High doses of antiandrogens fail to induce apoptosis in activated T cells.

Fig. S5. Antiandrogens inhibit human T cell activation through off-targeting GABA-A receptor as well.

REFERENCES AND NOTES

  1. Acknowledgments: We thank S. Fu, D. Harmon, and T. Tu for their editing and L. Zheng and R. R. Weichselbaum for their suggestions and comments. We thank M. Karin (UCSD) for his generous gift of the Myc-CaP cell line. Funding: This work was supported by NIH grant CA134563 to Y.-X.F. and the National 12.5 Major Project of China (no. 2012ZX10001006002004) to Y.-X.F. X.Y. was supported by the Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning (TP2015013) and Shanghai Pujiang Program. Author contributions: Y.P., X.Y., and Y.-X.F. designed the experiments, analyzed the data, and wrote the manuscript. Y.P., X.Y., M.X., Y.L., and K.Y. performed the experiments. M.X. and Y.G. contributed to the manuscript preparation. Y.-X.F. and X.Y. supervised the project. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article