Abstract
Tumors have evolved multiple mechanisms to evade immune destruction. One of these is expression of T cell inhibitory ligands such as programmed death-ligand 1 (PD-L1; B7-H1). In this study, we show that PD-L1 is highly expressed on mesothelioma tumor cells and within the tumor stroma. However, PD-L1 blockade only marginally affected tumor growth and was associated with the emergence of activated programmed death-1+ ICOS+ CD4 T cells in tumor-draining lymph nodes, whereas few activated CD8 T cells were present. Full activation of antitumor CD8 T cells, characterized as programmed death-1+ ICOS+ Ki-67+ and displaying CTL activity, was only observed when CD4 T cells were depleted, suggesting that a population of suppressive CD4 T cells exists. ICOS+ foxp3+ regulatory T cells were found to be regulated through PD-L1, identifying one potentially suppressive CD4 T cell population. Thus, PD-L1 blockade activates antitumor CD8 T cell most potently in the absence of CD4 T cells. These findings have implications for the development of PD-L1-based therapies.
Tumor growth is associated with chronic antigenic stimulation, immune evasion, and impaired T cell responses (1). The T cell inhibitory receptor programmed death-1 (PD-1,4 CD279) plays a key role in T cell inactivation after chronic exposure to Ag (2, 3, 4). PD-1 is a B7 family member that is expressed on activated T cells and that regulates T cell function through its interaction with programmed death-ligand 1 (PD-L1, also known as B7-H1, CD274) (5). PD-1 has a second ligand, programmed death-ligand 2 or B7-dendritic cell (DC; CD273), which is expressed on DCs and has a less clearly defined function (3). Persistent PD-1 ligation is responsible for functional inactivation (exhaustion) of antiviral CD8 T cells in mice chronically infected with lymphocytic choriomeningitis virus (6), and blockade of the PD-1/PD-L1 pathway rescues T cell functionality (6, 7, 8). In the context of persistent lymphocytic choriomeningitis virus infection, this means that rescued CD8 T cells restore their capacity to lyse target cells and to produce cytokines such as IFN-γ, IL-2, and TNF-α. Because PD-L1 is expressed on tumor cells (9) as well as on tumor-derived APCs (10), it has been proposed that the PD-1/PD-L1 pathway may play a central role in tumor immune evasion. Consistent with this, PD-L1 expression on tumor cells is associated with a poor prognosis in renal cell carcinoma patients (11). Thus, it was predicted that blockade of the PD-1/PD-L1 pathway would enhance anticancer T cell responses. Indeed, PD-L1 blockade enhances the efficacy of different immunotherapies in mouse tumor models (12, 13), suggesting that antitumor T cell responses that are induced by immunotherapy are restrained through PD-1/PD-L1 interactions. However, it has recently been shown that CD25+ regulatory T cells are also under the control of PD-1, i.e., PD-L1 blockade enhanced the suppressive function of regulatory T cells (14). Therefore, the objective of the present study was to investigate the role of PD-1 in tumor-induced immunosuppression in a mouse model of malignant mesothelioma (15), in which immunosuppression depends in part on regulatory T cells (16). To study this, we have used the AB1-hemagglutinin (HA) mesothelioma tumor cell line that was obtained by transfecting asbestos-induced AB1 mesothelioma cells (15) with the infuenza virus HA gene as model tumor Ag (17). Expression of the HA gene does not impact on the immunogenicity of the tumor (18). This tumor model has two important features. First, we have previously shown that tumor Ags are constitutively and efficiently cross-presented to CD8 T cells (19). Second, cross-presentation results in weak antitumor CTL responses, which are incapable of controlling tumor growth. Therefore, tumor growth in this model is associated with persistent antigenic stimulation, and it is tempting to speculate that inhibitory T cell receptors such as PD-1 are responsible for the weak antitumor T cell responses. In this study, we report that PD-L1 is ubiquitously expressed in the tumor stroma, that both CD8 and CD4 T cells express PD-1, and that PD-L1 blockade results in T cell activation. We found that PD-L1 blockade led to CD4 T cell activation, and that the full CD8-activating potential of PD-L1 blockade was only realized when CD4 T cells were depleted. These findings are consistent with the recent finding that CD25+ regulatory T cells are functionally controlled through PD-1 (14), which suggests that immunosuppressive responses can also be controlled by PD-1, and have implications for the design of PD-L1-based immunotherapies.
Materials and Methods
Reagents and Abs
d HA pentamers were purchased from ProImmune. CFSE was purchased from Molecular Probes. Flow cytometry was performed using a BD FACSCalibur or a FACSCanto II instrument and analyzed using Flowjo software (Tree Star).
Animals
BALB/c (H-2d) wild-type mice were purchased from the Animal Resources Centre and maintained under specific pathogen-free conditions. All experiments used female mice between 6 and 8 wk of age. Clone 4 TCR transgenic mice, which express a TCR specific for the H-2d-restricted peptide IYSTVASSL (residues 518–526) of A/PR8/8/34 (H1N1) influenza virus HA, were generated and screened, as previously described (18). Animal experiments were conducted according to University of Western Australia Animal Ethics Committee approvals.
Tumor cell culture and inoculation
Generation of the BALB/c-derived mouse mesothelioma cell line AB1 and transfection with the gene encoding influenza HA (AB1-HA) have been previously described (18). Cell lines were maintained in RPMI 1640 (Invitrogen Life Technologies) supplemented with 20 mM HEPES, 0.05 mM 2-ME, 60 μg/ml penicillin (CSL), 50 μg/ml gentamicin (West), and 10% FCS (Invitrogen Life Technologies). Selection of AB1-HA transfectants was achieved using culture medium containing the neomycin analog geneticin (Invitrogen Life Technologies) at a final concentration of 400 μg/ml. Tumor cells (1 × 106 in 100 μl of sterile PBS) were injected s.c. into the shaven right flank of recipient mice on day 0. Tumor growth was monitored by taking two perpendicular diameter measurements using microcalipers. Mice were euthanized when tumors reached 10 × 10 mm2
In vivo Ab blockade
PD-1 (29F.1A12) and PD-L1 (10F.9G2) blocking Abs were supplied by G. Freeman (Harvard Medical School, Boston, MA) and were administered every 3 days for a total of three doses of 200 μg. Rat IgG2b isotype control (administered as for anti-PD-L1 and anti-PD-1).
Cell depletion studies
CD4 T cell depletion was performed using purified GK1.5 Ab, prepared by Dr. K. Davern (Western Australia Institute for Medical Research, Perth, Western Australia, Australia) CD4-depleted mice received an initial dose of 100 μg i.v. on the first day of treatment, followed by two additional doses 3 days apart (three doses, every third day) or as indicated for each figure. CD4 depletion was verified during treatment by FACS analysis of peripheral blood using Abs specific for TCR-β and CD4.
Preparation and staining of tissues for flow cytometry
Tumors, spleens, and lymph nodes were collected into cold PBS containing 2% FCS (v/v). The axillary and inguinal lymph nodes were pooled for the tumor flank (draining lymph nodes) and for the contralateral flank (nondraining lymph nodes). Tissues were homogenized into cell suspensions using frosted glass slides, washed and resuspended in staining buffer (PBS with 2% BSA (w/v), 2% FCS (v/v), and 0.01% NaN3 (w/v)), and incubated for 30 min at 4°C with appropriate dilutions of Abs or isotype controls. Cells were washed twice and resuspended in stabilizing fixative (BD Biosciences) before FACS analysis. Pentamer staining was done as follows: cells were first blocked with 10% FCS (20 μl, 10 min at room temperature). After adding 5 μl of pentamer, cells were incubated for 30 min at 4°C, and then washed in FACS buffer, followed by incubation with anti-CD8 and anti-CD19 Abs, as well as other Abs, as required.
Collagenase-DNase digestion for DC analysis
For isolation and analysis of DCs in murine tissue, organs were disrupted by enzymatic digestion, as described previously (20, 21). Briefly, lymph nodes and tumors were finely minced and digested for 20 min at room temperature in RPMI 1640 containing 2% FCS, HEPES, 1 mg/ml collagenase type II (Worthington Biochemical), and 1 μg/ml grade II bovine pancreatic DNase I (Boehringer Mannheim). To disrupt DC-T cell rosettes, 0.1 M EDTA was added and the suspension was digested for an additional 5 min. Undigested fragments were allowed to settle, and the supernatant was transferred to a fresh tube. Cells were recovered by centrifugation and resuspended for subsequent procedures in ice-cold HBSS (Invitrogen) supplemented with 5 mM EDTA and 5% FCS (EDTA-balanced salt solution-FCS). For digestion of solid tumors, ∼100 μl of RPMI 1640 containing 2% FCS, HEPES, 1 mg/ml collagenase, and 100 μg/ml DNase was injected into the tumor tissue before mincing. Tumor fragments were resuspended in the same solution and digested with mixing for 1 h at room temperature. A single-cell suspension was then generated using the protocol described above.
In vivo CTL assay
Detection of HA-specific in vivo CTL was conducted, as previously described (18, 22). Briefly, naive BALB/c spleens were homogenized into cell suspensions, and erythrocytes were lysed by incubation of cells in RBC lysis buffer for 3 min, followed by two washes in RPMI 1640 with 10% FCS. Cells were resuspended at 5 × 106 cells/ml, and then divided into two populations, one of which was pulsed with 1 μg/ml clone 4 peptide for 90 min at 37°C. Cells were washed and then labeled with CFSE for 10 min at room temperature in serum-free conditions using a final concentration of 2 μM for peptide-pulsed cells (CFSEhigh) and 0.2 μM for unpulsed cells (CFSElow). Both populations were washed in FCS three times, followed by sterile saline before being pooled in equal proportions, and 2 × 107 cells were injected i.v. into recipients. After 18 h, draining and nondraining lymph nodes, tumors, and spleens were harvested and processed, as described above, and analyzed by FACS for the recovery of CFSEhigh and CFSElow cells. Percentage of killing was calculated using the following formula: (1 − (CFSEhigh events/CFSElow events)) × 100.
Statistics
Data were statistically evaluated using Prism software (GraphPad), with growth curves compared using two-tailed paired Student’s t test, with pairs defined by time point and survival responses analyzed by Kaplan-Meier using a log-rank test. Significance was defined as p < 0.05.
Results
Expression of PD-L1 by tumor cells and in the tumor milieu
Several studies have demonstrated that the efficacy of antitumor immunotherapy or adoptive T cell transfer is limited by negative PD-1/PD-L1 interactions (12, 23, 24). We used a mouse model of mesothelioma (15, 17) to assess the role of PD-1-mediated T cell inhibition in natural tumor-induced immune responses (i.e., without additional therapy). First, we evaluated the expression patterns of PD-L1 on AB1-HA mesothelioma cells. Under in vitro culturing conditions, we found that AB1-HA cells expressed low levels of PD-L1 (Fig. 1⇓A). PD-L1 expression was rapidly induced after exposure to IFN-γ (Fig. 1⇓B). This is similar to findings in other tumor models (1), suggesting that IFN-γ-dependent PD-L1 expression may be a general feature of tumor cells. To analyze in vivo PD-L1 expression patterns, i.e., in the tumor milieu, we prepared single-cell suspensions from growing tumors and analyzed these for PD-L1 expression. Although the majority of cells from the tumor were PD-L1low, we observed subpopulations of PD-L1int and PD-L1high cells (Fig. 1⇓C). Ex vivo exposure of tumor cell suspensions to IFN-γ led to a rapid PD-L1 up-regulation (Fig. 1⇓D). This was observed mostly as a shift of the PD-L1int subpopulation (Fig. 1⇓, C and D). No changes were observed in the PD-L1high population (Fig. 1⇓, C and D). Because IFN-γ-dependent PD-L1 expression appears to be a feature of the tumor cells (Fig. 1⇓, A and B), we reasoned that the IFN-γ-sensitive PD-L1int population could include the tumor cells. PD-L1high cells could represent tumor stroma cells, such as DCs. To test this, tumor-resident CD11c+ DCs were analyzed for PD-L1 expression. Indeed, we observed high-level PD-L1 expression on CD11chigh cells in the tumor, whereas CD11clow cells expressed low levels of PD-L1 (Fig. 1⇓E). PD-L1 analysis on DCs from tumor-draining and nondraining lymph nodes revealed systemic, high-level PD-L1 expression (Fig. 1⇓F). Thus, PD-L1 is ubiquitously expressed both within the tumor milieu and in the lymphatic system in tumor-bearing mice. Importantly, total PD-L1 expression in the tumor is sensitive to IFN-γ, suggesting that potential IFN-γ-secreting antitumor T cells may induce a counterregulatory response in the tumor, which could block their efficacy.
PD-L1 is expressed in the AB1-HA tumor microenvironment. A and B, AB1-HA tumor cells growing in vitro were stained for PD-L1 expression before (A) and after (B) exposure to 20 ng/ml murine IFN-γ (R & D Systems) for 24 h and analyzed using FACS. C and D, Expression of PD-L1 in the ex vivo tumor milieu before (C) and after (D) exposure to 20 ng/ml murine IFN-γ for 24 h. Whole tumor was removed at day 15, homogenized, exposed to IFN-γ or PBS, and analyzed for PD-L1 expression. Percentages of tumor cells in the marked gates are indicated. E, PD-L1 expression on tumor-resident DCs. Homogenized tumor samples were gated on CD11chigh (left panel) or CD11clow (right panel) and analyzed for PD-L1 expression. F, PD-L1 expression on CD11chigh DCs from the tumor-draining lymph nodes (DLN, left panel) and nondraining lymph nodes (NDLN, right panel).
T cells in tumor-draining lymph nodes express PD-1
PD-1 expression on CD8 T cells is usually associated with persistent antigenic stimulation and may predispose these cells to functional inactivation (4, 6). To test whether this was the case in AB1-HA tumor-bearing mice, CD8 T cells in tumor-draining and nondraining lymph nodes were analyzed for PD-1 expression. Because AB1-HA tumor Ags are exclusively cross-presented in tumor-draining lymph nodes (25), we reasoned that a comparison between PD-1 expression levels on T cells in tumor-draining and nondraining lymph nodes would provide us with a surrogate for PD-1 expression on the tumor-specific T cell pool. As shown in Fig. 2⇓A, PD-1 expression levels on CD8 T cells in the draining lymph nodes were variable, but overall significantly (p < 0.001) higher than in nondraining nodes. PD-1 expression patterns on tumor-specific CD8 T cells were further evaluated by analyzing T cell responses in TCR-transgenic clone 4 mice. TCR-transgenic CD8 T cells from these mice are specific for a Kd-restricted epitope in the influenza virus HA protein (HA518–526: IYSTVASSL), which is expressed as a neo-Ag in AB1-HA tumor cells (17). Thus, TCR-transgenic mice were inoculated with AB1-HA tumor cells, and PD-1 expression on tumor-specific CD8 T cells was measured in draining and nondraining lymph nodes, as well as in the tumor. Again, we found higher PD-1 expression levels in draining compared with nondraining lymph nodes. PD-1high TCR-transgenic CD8 T cells were LFA-1high and CD62Llow, consistent with an activated phenotype (Fig. 2⇓B and data not shown). High PD-1 expression levels were also observed on tumor-infiltrating transgenic CD8 T cells (Fig. 2⇓B). Note that PD-1 expression levels on CD8negative cells were higher than on CD8positive cells (Fig. 2⇓A). It seemed likely that these CD8negative PD-1high cells were CD4 T cells. Indeed, we found that CD4 T cells expressed PD-1 (Fig. 2⇓C) and that PD-1 expression was largely confined to the tumor-draining lymph nodes (Fig. 2⇓C).
PD-1 expression on CD4 and CD8 T cells in the tumor and the local lymph nodes. A, PD-1 expression on CD8+ T cells from the tumor-draining lymph nodes (DLN) was significantly increased compared with non-DLN (NDLN); unpaired Student’s t test. Representative dot plots gated on lymphocytes show PD-1 expression on CD8+ T cells from the non-DLN (left panel) and DLN (right panel). Percentage of PD-1high CD8+ T cells is indicated. B, TCR-transgenic HA-specific CD8+ T cells from the NDLN (left panel), DLN (center panel), and tumor (right panel) were analyzed for expression of PD-1 and LFA-1. Percentages of PD-1+CD8+ T cells are indicated. Cells shown are gated on CD8+ cells. C, PD-1 expression on CD4+ T cells from the tumor nondraining lymph nodes (left panel) and draining lymph nodes (right panel). Percentages of PD-1+CD4+ T cells are indicated.
Thus, the data show that PD-1 is expressed on both CD4 and CD8 T cells in the tumor-draining lymph nodes, and suggest that immunosuppression in this tumor model could be driven by PD-1/PD-L1 interactions.
PD-L1 blockade slows tumor growth and activates lymph node CD4 T cells
To test whether AB1-HA tumor growth depended on PD-1/PD-L1-mediated inhibition of antitumor T cell responses, we treated tumor-bearing mice with PD-L1 blocking Abs. Treatment was started when tumors were just palpable (day 8 after inoculation), using a four doses every third day schedule (200 μg of Ab, i.p.). PD-L1 blockade had a weak, but significant (p < 0.01) effect on tumor growth (Fig. 3⇓A). We then assessed the impact of PD-L1 blockade on T cells by analyzing expression of PD-1 and ICOS on CD4 and CD8 T cells. ICOS (CD278) is a B7 family member that is expressed on activated T cells after TCR ligation (26). Interestingly, we observed that PD-L1 blockade induced a population of activated PD-1high ICOShigh CD4 T cells in tumor-draining lymph nodes (Fig. 3⇓B, lower right panel). CD8 T cells in the draining lymph nodes were not activated to a similar extent, in particular in terms of the expression levels of PD-1 and ICOS (as jugded on the basis of mean fluorescence intensity; Fig. 3⇓C, lower right panel). Given the key role of CD8 T cells in AB1-HA tumor resolution (22), it is possible that there is a correlation between the relatively weak antitumor effects of PD-L1 blockade and the relatively stronger CD4 T cell activation (compared with CD8 T cell activation, in terms of frequency and mean fluorescence intensity).
Impact of in vivo PD-L1 blockade on tumor growth and T cell activation. A, Tumor growth in AB1-HA tumor-bearing BALB/c mice treated with 200 μg of anti-PD-L1 Ab (days 8, 11, and 14 posttumor inoculation). Data shown are mean ± SEM (n = 3) from one experiment. ∗∗, p < 0.05 when untreated is compared with anti-PD-L1. B, PD-1 and ICOS expression on CD4+ T cells from the nondraining lymph nodes (NDLN; left panels) and draining lymph nodes (DLN; right panels) from untreated (top panels) or anti-PD-L1-treated (bottom panels) mice. Tumor-bearing mice were left untreated or treated with anti-PD-L1 (days 11 and 14), and lymph nodes were harvested at day 16. C, PD-1 and ICOS expression on CD8+ cells, as per B.
CD8 T cells are activated by PD-L1 blockade after CD4 T cell depletion
To further investigate the effect of PD-L1 blockade on the activation of CD8 T cells, we performed PD-L1 blockade experiments in the absence of CD4 T cells. CD4 T cells were depleted using GK1.5 mAbs (three doses every third day), and depletion was verified by staining PBMC 3 days after Ab injection. Ab treatment routinely led to >99% CD4+ T cell depletion at 3 days after injection (Fig. 4⇓A), and high levels of depletion were maintained throughout the experiment (day 16 lymph node data shown; Fig. 4⇓B). All treatments were started when tumors were just palpable (day 7–8 after tumor cell inoculation). To study T cell activation in response to PD-L1 blockade, we analyzed the phenotypes of CD4 and CD8 T cells in tumor-draining lymph nodes using PD-1 and ICOS as activation markers. These experiments revealed that PD-L1 blockade in CD4-depleted mice induced high frequencies of activated PD-1+ ICOS+ CD8 T cells (Fig. 4⇓Civ). PD-L1 blockade alone did not activate CD8 T cells (Fig. 4⇓Cii), but did induce a population of activated CD4 T cells (Fig. 4⇓Cvi). CD4 depletion without PD-L1 blockade resulted in modest levels of activated CD8 T cells, as judged by ICOS and PD-1 expression (Fig. 4⇓Ciii). By analyzing CD8 T cell numbers using BD Biosciences TruCount tubes, we found that CD8 T cells did not have a homeostatic response to CD4 depletion (data not shown), excluding the possibility that PD-1 was up-regulated on CD8 T cells as a response to homeostatic proliferation (27). Thus, PD-L1 blockade leads to CD8 T cell activation only when CD4 T cells are absent. Interestingly, low frequencies of remaining or recovering CD4 T cells (∼5%; Fig. 4⇓B) can be detected in CD4-depleted mice (Fig. 4⇓C, vii and viii), and these CD4 T cells display an activated phenotype. PD-L1 blockade in CD4-depleted mice induced further activation of the recovering CD4 T cells (Fig. 4⇓Cviii).
Increased activation of CD8 T cells by PD-L1 blockade after CD4 depletion. Representative depletion data are shown for PBL samples at day 3 after GK1.5 Ab injection (A) and for lymph nodes at day 16 after GK1.5 Ab injection (B). C, Effect of CD4 depletion on the activation of T cells in the tumor-draining lymph nodes. Tumor-bearing mice were left untreated or depleted of CD4+ T cells (days 8, 11, and 14) with or without PD-L1 blockade (days 8, 11, and 14). Draining lymph node cells were harvested on day 16 and were gated on CD8+ (upper row) or CD4+ (lower row) cells and analyzed for PD-1 and ICOS expression. D, As per C, but showing expression of PD-1 and Ki-67.
We then assessed the proliferative status of anti-PD-L1-activated CD4 and CD8 T cells by intracellular Ki-67 staining (Fig. 4⇑D). Ki-67 is a nuclear protein that is only expressed in cycling cells (28) and that can be used to identify activated T cells (29). Activated PD-1+ ICOS+ CD4 T cells, induced by PD-L1 blockade, were Ki-67low, indicating that they did not proliferate (Fig. 4⇑Dvi). In contrast, we did observe Ki-67high proliferating cells among the PD-1+ ICOS+ CD8 T cells that were induced by PD-L1 blockade in CD4-depleted mice (Fig. 4⇑Div). The small, but activated CD4 T cell populations in GK1.5-treated mice harbored Ki-67high cells (Fig. 4⇑Dvii). Importantly, PD-L1 blockade in CD4-depleted mice resulted in even stronger activation and Ki-67 expression of the remaining CD4 T cells (Fig. 4⇑Dviii).
To confirm that PD-L1 blockade activated a specific antitumor CD8 T cell response and that this response was augmented by CD4 T cell depletion, we performed an in vivo CTL assay and we used a MHC I pentamer specific for the Kd-restricted HA epitope. In vivo CTL activity was measured by injecting CFSE-labeled HA peptide-coated and uncoated targets into tumor-bearing mice and measuring their recovery 18 h later. We compared the effects of PD-L1 blockade in CD4-depleted and control mice. The results (Fig. 5⇓A) revealed that there was little to no in vivo CTL activity in untreated tumor-bearing mice, consistent with a strong suppression of antitumor immune responses. The CTL response was somewhat increased in mice receiving PD-L1 blocking Abs, consistent with the modest antitumor effect of this treatment. CD4 depletion resulted in higher in vivo CTL responses, but the variability in this treatment group was very high, indicating that not all animals responded (Fig. 5⇓A). Finally, PD-L1 blockade in CD4-depleted mice resulted in high in vivo CTL responses in all mice (Fig. 5⇓A). Responses were detected in draining and nondraining lymph nodes, as well as in the spleen and in the tumor. These results were confirmed by pentamer staining (Fig. 5⇓B). Few pentamer+ CD8 T cells were detected in untreated tumor-bearing mice, and these cells were ICOSlow and CD43low (Fig. 5⇓B and data not shown). PD-L1 blockade or CD4 depletion alone led to a modest increase in the frequency of pentamer+ CD8 T cells (Fig. 5⇓B). However, PD-L1 blockade in CD4-depleted mice triggered a strong increase in pentamer+ cells (Fig. 5⇓B). These cells were CD8low (Fig. 5⇓B) and expressed high levels of ICOS (Fig. 5⇓C), consistent with an activated phenotype.
A, CD4 depletion with PD-L1 blockade promotes CTL activity. Mice were either left untreated or treated with anti-PD-L1, CD4 depletion, or a combination of the two (days 7, 10, and 13). Two days after the last dose of treatment (day 15), 2 × 107 HA-coated CFSE-labeled target splenocytes were injected i.v., and their recovery in the draining lymph nodes (DLN), nondraining lymph nodes (NDLN), spleen, and tumor was measured 18 h later. Responses of individual mice are shown with group mean (horizontal scale bar; n = 3 per group). ∗, p < 0.05; ∗∗∗, p < 0.005 (unpaired Student’s t test). B, MHC I pentamer staining on tumor-draining lymph nodes in CD4-depleted and/or PD-L1-blocked mice. Tumor-bearing mice were treated with Abs from day 7 onward (anti-CD4 on day 7; anti-PD-L1 on days 8 and 11), and lymph nodes were harvested on day 14 and used for pentamer analysis. C, Expression of ICOS on pentamer+ CD8 T cells, as shown in B.
Impact of PD-L1 blockade on tumor growth with or without CD4 T cells
PD-L1 blockade potently activated CD8 T cells and tumor-specific CD8 T cell responses when CD4 T cells were depleted. Thus, we examined the effect of PD-L1 blockade in CD4-depleted mice. PD-L1 blockade in the absence of CD4 T cells resulted in a complex response pattern (Fig. 6⇓). Clearly, PD-L1 blockade was much more effective when CD4 T cells were depleted (Fig. 6⇓), identifying CD4 T cells as a limiting factor. CD4 depletion alone also had a strong antitumor effect, suggesting that CD4+ cells are overall immunosuppressive (Fig. 6⇓). Detailed analysis of the effects of PD-L1 blockade in CD4-depleted mice revealed a biphasic response pattern. Comparison of CD4-depleted mice with CD4-depleted and PD-L1-blocked mice revealed a highly significant (p < 0.005) benefit of PD-L1 blockade from day 13–28 after tumor cell inoculation (i.e., day 6–21 after Ab treatment; Fig. 6⇓). Importantly, tumor regression was only observed with the PD-L1 blockade/CD4 depletion combination, whereas the effects of CD4 T cell depletion only were limited to tumor growth arrest. From day 28 onward, tumor growth recurred and tumors in PD-L1-blocked and CD4-depleted mice grew faster than tumors in mice that had only received CD4-depleting Ab (p < 0.05, day 28–41; Fig. 6⇓). Thus, PD-L1 blockade in CD4-depleted mice initially enhanced antitumor responses, leading to tumor regression, but then resulted in accelerated tumor growth that coincided with CD4 recovery (last GK1.5 injection at day 14 after inoculation).
Effect of CD4 depletion on effect of PD-L1 blockade. Tumor-bearing mice were left untreated or depleted of CD4+ T cells using GK1.5-depleting Ab (days 6, 9, and 12) with or without PD-L1 blockade (days 10, 13, and 16). Data shown are mean (n = 5) from one experiment. ∗∗, p < 0.005, anti-PD-L1 vs anti-PD-L1 + anti-CD4, day 13–28; ∗, p < 0.05, anti-PD-L1 vs anti-PD-L1 + anti-CD4, day 28–41.
PD-L1 blockade increases regulatory T cell numbers
Our data indicate that a pool of immunosuppressive CD4 T cells exists. CD25+ foxp3+ regulatory T cells constitute a well-described population of immunosuppressive CD4 T cells (30). We recently found that CD25 depletion alone did not affect tumor growth, but that it did enhance the efficacy of antitumor chemotherapy, identifying regulatory CD4 T cells as candidates for suppressive cells (16). Thus, we measured foxp3+ CD4 T cell numbers in tumor-draining lymph nodes from mice receiving different treatments. Interestingly, PD-L1 blockade resulted in an increase in the frequency of foxp3+ CD4 T cells (p < 0.05; Fig. 7⇓A). To further evaluate regulatory T cell responses, we analyzed ICOS expression on these cells. ICOS was selected because it may identify cells that respond to PD-L1 blockade and because it has been linked with a more suppressive regulatory T cell phenotype (31). Interestingly, foxp3+ CD4 T cells in anti-PD-L1-treated mice expressed higher levels of ICOS (Fig. 7⇓B). A potential functional role of ICOS on foxp3+ regulatory T cells is supported by the expression of ICOS-L on CD11c+ DCs in the tumor (Fig. 7⇓C) (32). Combined, these data provide a possible link between PD-L1 blockade and activation of suppressive immune responses, i.e., foxp3+ regulatory T cells.
PD-L1 blockade increases CD4+ regulatory T cell numbers. A, Percentage of foxp3+CD4+ T cells present in the draining lymph nodes of untreated mice or mice treated with poly(I:C), anti-PD-L1, or both (days 7, 10, and 13; top panel). Lymph nodes were harvested at day 14. ∗, p < 0.05; ∗∗, p < 0.01; unpaired Student’s t test. Representative plots of the same data (bottom panel). Numbers indicated represent the percentage of CD4+ T cells positive for foxp3. B, ICOS expression on CD4+ foxp3+ T regulatory cells from the draining lymph nodes of mice treated, as for A. C, ICOS-L expression on tumor-infiltrating CD11c+ DCs. CD11c+ DCs were gated (left panel) and analyzed for ICOS-L expression (green graph). Isotype control staining is shown in red.
Discussion
Inhibition of CD8 T cell responses via PD-1 and PD-L1 interactions has received considerable attention during recent years (3, 6, 9). In several studies, it has been shown that blockade of PD-1/PD-L1 interactions rescues T cell functionality, translating into improved antiviral (6) or antitumor responses (24). In the current study, we have investigated the effect of PD-L1 blockade on T cell activation and tumor growth. The data show that PD-L1 blockade indeed has the capacity to activate antitumor CD8 T cell responses, and that this effect was much stronger in the absence of CD4 T cells. These CD8 T cell responses are substantial: up to 15% of all CD8 T cells in the tumor-draining lymph nodes is activated in the absence of CD4 T cell-mediated suppression. Importantly, when CD4 T cells are not depleted, PD-L1 blockade activates CD4 T cells instead of CD8 T cells. This suggests that the CD4 T cells, possibly including those activated through PD-L1 blockade, may prevent CD8 T cell activation. The effect of PD-L1 blockade on tumor growth was weak, but, again, stronger when combined with CD4 depletion. Importantly, CD4 depletion alone already led to tumor growth arrest, pointing toward a role for immunosuppressive CD4 T cells. This phenotype of immunosuppressive CD4 T cells is similar to recently reported work by Melief and coworkers (33). Our data do not rule out the possibility that such suppressive CD4 T cells are also regulated through PD-L1 blockade. Indeed, our observation that PD-L1 blockade induces expansion of and ICOS up-regulation in foxp3+ regulatory T cells could support this view.
The concept that the effects of PD-L1 blockade may not be limited to IFN-γ-producing T cells has first emerged from work published by Freeman et al. (34). These authors showed that, depending on the polarization of the cells, PD-L1 blockade could rescue either IFN-γ or IL-10 production in CD4 T cells. These experiments were done in vitro, but suggested that PD-L1 blockade could also rescue different types of immune responses in vivo. The data presented in our current study could be consistent with this. We show in this study that PD-L1 blockade has the potential to expand and activate foxp3 regulatory CD4 T cells. Control of regulatory T cell responses through PD-1 is supported by a recent study showing that regulatory T cell functionality in chronically infected HCV patients is controlled through PD-1, in a counterregulatory fashion, i.e., PD-1 ligation shuts down the suppressors (14). Furthermore, PD-1 was up-regulated on CD4+ CD25+ foxp3+ regulatory T cells from nonresponders after hepatitis B surface Ag vaccination, consistent with increased suppressive activity (35). However, as we recently reported, CD25 depletion alone had no effect on tumor growth (16), suggesting that the CD4+ suppressive T cell pool that we identified in the current study may be composed of other non-T regulatory suppressive cell types as well. Indeed, the existence of PD-1+ CD4 T cells acting as regulatory T cells has recently been reported (36). Our preliminary data show that ICOShigh CD4 T cells, purified from tumor-draining lymph nodes in anti-PD-L1-treated mice, up-regulate the Th2 cytokines IL-5 and IL-13 (21- and 120-fold, compared with the ICOSlow population; A. J. Currie, A. Prosser, A. L. Cleaver, B. W. S. Robinson, G. J. Freeman, R. G. van der Most, manuscript in preparation). IL-13 could have an important role. For example, type-II CD4+ NKT cells exert their immunosuppressive functions through IL-13 (37, 38). IL-13 activates myeloid suppressor cells, which, in turn, produce the immunosuppressive cytokine TGF-β. Indeed, we have preliminary data showing that a combination of TFG-β neutralization and PD-L1 blockade slows tumor growth (R. G. van der Most, unpublished data). Furthermore, IL-2 produced by activated (ICOS+) CD4 T cells could enhance T regulatory proliferation (39), thereby enhancing immunosuppression.
The role of ICOS is interesting. ICOS is expressed on activated T cells, including CD4 and CD8 T cells (26), as well as regulatory T cells (31). ICOS on CD4 T cells has been linked to a Th2 phenotype (3, 40, 41, 42), but also with IL-10 production (31, 32, 40, 41, 43) and with more suppressive regulatory T cells (31, 44). Moreover, ICOS+ CD4 T cells were recently shown to control autoimmune responses (31, 40), and ICOS is expressed on tumor-infiltrating regulatory T cells (31, 44). Thus, there is evidence that is consistent with our conclusion that ICOS+ PD-1+ CD4 T cells may be suppressors, and that such responses can be controlled via PD-1/PD-L1 interactions.
We propose the following model (Fig. 8⇓). When the tumor establishes itself, it induces a CD4 T cell response that is composed of CD25+ regulatory cells and possibly IL-13-producing CD4+ T cells or NKT cells. Combined, these responses inhibit antitumor CD8 responses. The entire response may be under the control of PD-1, but the IFN-γ-dependent up-regulation of PD-L1 on tumor cells guarantees that the balance is tilted toward protumor T cell responses. PD-L1 blockade under these conditions will rescue all PD-1+ cells, including both CD4 and CD8 T cells, but reactivated CD4 T cells will suppress antitumor CD8 T cell responses, resulting in a weak antitumor effect. When CD4 T cells are depleted, PD-L1 blockade will activate CD8 T cells, leading to increased antitumor responses, but this response will be limited by the recovery time for CD4 T cells.
Schematic representation of the impact of PD-L1 blockade in the antiviral vs the antitumor setting. T cells are indicated in red as antiviral (A) or antitumor (B) effector cells, and in blue as immunosuppressive protumor T cells (B). Impact of PD-L1 blockade is indicated.
In conclusion, our data show that PD-L1 blockade has the capacity to unleash powerful antitumor CD8 T cell responses, but only when suppressive CD4 T cells are absent. The implication is that PD-L1 blockade may not always optimally activate antitumor CD8 T cell responses.
Acknowledgments
We thank Dr. Rafi Ahmed for stimulating discussions.
Disclosures
G.J.F. has patents and receives patent royalties on the PD-1 pathway.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was supported by a priming grant from the Raine Medical Research Foundation, Perth, Australia (to R.G.V.M.).
↵2 A.J.C. and A.P. contributed equally to this manuscript.
↵3 Address correspondence and reprint requests to Dr. Robbert G. van der Most at the current address: GSK Biologicals, Rue de l’Institut 89, B1330 Rixensart, Belgium. E-mail address: robbertvdm{at}gmail.com
↵4 Abbreviations used in this paper: PD-1, programmed death-1; DC, dendritic cell; HA, hemagglutinin; ICOS-L, ICOS ligand; int, intermediate; PD-L1, programmed death-ligand 1.
- Received April 2, 2009.
- Accepted October 19, 2009.
- Copyright © 2009 by The American Association of Immunologists, Inc.