The Immune Checkpoint Regulator PD-L1 Is Highly Expressed in Aggressive Primary Prostate Cancer

Prostate cancer is the second most frequently diagnosed cancer in men, accounting for approximately 15% of all newly diagnosed male cancers worldwide (1). Overall, it is the fifth most common cause of death from cancer in men, with 307.000 estimated deaths (6.6% of all estimated deaths) in 2012 (1). First-line therapies for early-stage localized prostate cancer include surgery and radiotherapy, and the 5-year relative survival rate approaches 100% (2). For patients affected by metastatic prostate cancer, androgen deprivation therapy (ADT) still is the mainstay of treatment (3). Although surgical or chemical castration can initially be effective delaying disease progression, the majority of patients eventually develops castration-resistant prostate cancer (CRPC) and has an adverse prognosis (4, 5). Recent advances and the arrival of several new agents in the field of CRPC, however, have improved overall survival in this patient population. Enzalutamid, a novel androgen receptor (AR) inhibitor that reduces nuclear translocation of the AR complex and subsequent DNA binding, was shown to significantly prolong overall survival in two phase III randomized, placebo-controlled trials (6, 7). Furthermore, abiraterone, an inhibitor of cytochrome P450, family 17, subfamily A, polypeptide 1 (CYP17A1), suppresses androgen biosynthesis in combination with prednisone, increasing survival time and time to prostate-specific antigen (PSA) progression (8, 9). Above all, the early addition of docetaxel to ADT has recently been shown to extend survival for men with newly diagnosed hormone-sensitive prostate cancer by more than 13 months (10).

Among many concepts under investigation, augmenting immune responses to cancer has been proposed as a valid therapeutic option, providing an alternative approach to improve survival. In particular, checkpoint inhibitors have emerged as a complementary avenue of clinical research in prostate cancer (11, 12). These treatments target checkpoint molecules in the regulation of the immune system harnessing preexisting anticancer immune responses. The programmed death 1 (PD-1)/programmed death ligand 1 (PD-L1) pathway plays a pivotal role in the regulation of T-cell activity at the time of inflammatory response. The PD-1 receptor thereby acts as a negative checkpoint regulator to prevent off-target immune activation of T-lymphocytes (13). Binding of PD-1 to its ligand PD-L1, the predominant mediator of immunosuppression, inhibits proliferation of activated T cells leading to “T-cell exhaustion” (14). PD-L1 itself is an immunomodulatory cell-surface glycoprotein primarily expressed by antigen-presenting cells on myeloid dendritic cells, activated T cells, and some non-hematopoietic tissues (15).

Recent evidence strongly suggests that the activation of the PD-1/PD-L1 pathway represents a mechanism allowing tumors to elude the host's immune system. Therapies targeting this signaling pathway promote marked antitumor immunity and have shown promising results in a subset of solid tumors (16, 17). As reported by Topalian and colleagues, blockade of PD-L1 using an anti-PD-1 antibody induced objective response rates of 6% to 17% and prolonged stabilization of disease in patients with several advanced cancers, that is, malignant melanoma, non–small cell lung cancer, and renal cell cancer (17). Further, immunohistochemical assessment of PD-L1 in pretreatment cancer specimens from 42 patients revealed that response to treatment was seen exclusively in PD-L1–positive tumors (9/25, 36%), indicating that PD-L1 expression might be a predictive biomarker for anti-PD-1 therapy (17). However, several studies have demonstrated response to treatment in patients with low or absent PD-L1 expression, although these patients seem to be in the minority (18–23). So far, no data are available to support that PD-1/PD-L1–targeted therapy may be effective in prostate cancer. This prompted us to analyze PD-L1 expression in primary prostate cancer.

Currently, only few PD-L1 antibodies have been validated for formalin-fixed paraffin embedded (FFPE) material, only one of which being commercially available ((E1L3N) XP; refs. 14, 24–31). The utilization of this antibody for predicting response to anti-PD1 or anti-PD-L1 therapy, however, remains unknown. The majority of clinical trials targeting the PD-1/PD-L1 pathway are using proprietary antibodies, implying that validation data for these particular antibodies are undefined. So far, no immunohistochemical detection method for quantifying PD-L1 expression is uniformly accepted as standard. We here comprehensively validated a novel monoclonal rabbit antibody against PD-L1 (clone EPR1161(2)) amenable to FFPE tissue and analyzed two large, well-characterized cohorts of primary prostate cancer after radical prostatectomy (RP) for PD-L1 expression.

Two previously described RP cohorts were analyzed in the present study (32–34). Clinicopathologic details, including Gleason grade grouping according to the latest International Society of Urological Pathology (ISUP) consensus (35), are given in Table 1. The training cohort included 262 patients having undergone radical prostatectomy at the University Hospital of Bonn between 1998 and 2008. Immunohistochemical PD-L1 staining was available for 209 cases. The test cohort comprised 640 patients diagnosed at the Charité University Hospital, Berlin between 1999 and 2005, and a total of 611 cases were eligible for PD-L1 immunhistochemistry. All research conducted was approved by the Charité Ethics Committee (EA1/06/2004) and the University Hospital of Bonn, Germany.

Tissue microarrays (TMA) were constructed as previously described (32). Briefly, FFPE tissue specimens were selected according to tissue availability and retrieved from the archive of the Charité University Hospital Berlin and the University Hospital Bonn. Each case was represented by one to three tissue cores with a core diameter of 1.2 mm or 1.8 mm.

The prostate cancer cell line DU145 was obtained from the ATCC and maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS, 1% l-glutamine, and 1% antibiotics (Life Technologies). Cells were grown at 37°C in a humidified 5% CO₂ atmosphere. PD-L1 expression was analyzed using flow cytometric analysis. Briefly, 5 × 10⁵ cells were washed with 5 mL PBS and resuspended in 300 μL of PBS. FACS analysis was performed on a Navios Flow Cytometer (Beckman Coulter) following incubation with FITC-tagged anti-human PD-L1 or mouse IgG isotype control (BD Pharmingen) for 20 minutes at room temperature, a second wash, and resuspension in PBS.

Cells were lysed in RIPA buffer (Sigma-Aldrich) in the presence of protease inhibitors. Cleared lysates were separated on NuPAGE 4% to 12% Bis–Tris gels (1.0 mm, 15 well; Life Technologies) and transferred onto an Invitrolon PVDF membrane (Life Technologies). Membranes were stained using anti-PD-L1 rabbit monoclonal antibody (mAb) clone EPR1161(2) (Abcam; 1:500) in 5% nonfat dry milk, and protein concentrations were determined by a colorimetric assay (Bio-Rad Laboratories Inc.). Blots were stripped and reprobed with anti-β-actin mAb (Sigma-Aldrich; 1:500). Western blot analysis with the validated anti-PD-L1 rabbit mAb (E1L3N) XP (Cell Signalling Technology; 1:500) was performed for comparison.

DU145 cells were seeded with 2 × 10⁵ cells in 6-well plates and incubated for 72 hours in DMEM in the presence of 10 to 20 nmol/L siRNA directed against PD-L1 (FlexiTube Gene Solution GS29126, Qiagen) or non-targeting control siRNA (Qiagen) complexed with HiPerFect Transfection Reagent (Qiagen) according to the manufacturer's instructions. Cells treated with HiPerFect Transfection Reagent only served as controls. siRNA-induced knockdown was evaluated using Western blot analysis as described above.

Prior to standard proceedings, the anti-PD-L1 antibody clone EPR1161(2) was incubated with a 10-fold concentration in excess of blocking peptide (Abcam) that corresponds to the epitope recognized by the primary antibody. Western blot analysis was carried out with the neutralized antibody side-by-side with the antibody alone. In analogy to this experiment, immunohistochemical staining was performed on placental and tonsillar tissue, following primary antibody neutralization with a 10-fold concentration in excess of blocking peptide, and compared with standard staining results.

Tissue sections were freshly cut (3 μm), mounted on Super Frost Plus slides (Thermo Fisher), and rehydrated in descending gradient alcohols. Immunohistochemical staining was carried out on the Ventana BenchMark Ultra automated staining system (Ventana) and visualized with the Ventana amplifier detection kit using the following antibodies: PD-L1, clone EPR1161(2) (Abcam; 1:75), Ki-67, clone MIB-1 (Dako; 1:500), and AR, clone AR441 (Dako; 1:400). Omission of the primary antibody was used as a negative control. Immunohistochemical stainings were evaluated independently by two pathologists (G. Kristiansen and H. Gevensleben) who were blinded to patients' clinical outcome. Specific membranous and cytoplasmic staining of epithelial tumor cells was considered positive. Because the staining was uniformly homogeneous, the intensity of PD-L1–positive cells was scored semi-quantitatively as negative (0), weak (1), moderate (2), or strong (3). Discrepant cases were reviewed at a multi-headed microscope and the consensus was reported.

Following dichotomization by median, associations between PD-L1 protein expression and clinicopathologic variables, including AR and Ki-67 expression, were analyzed using the χ² test. The correlation of Ki-67 and AR with PD-L1 expression as continuous variables was further tested using the Kendall τ rank correlation. Biochemical recurrence (BCR)-free survival was calculated for PD-L1 expression dichotomized by median using the Kaplan–Meier method, and survival time differences were compared using the log-rank test. PD-L1 expression as continuous variable was also examined within univariate and multivariate Cox proportional hazards regression models. P values refer to the Wald test. An interrater reliability analysis using the Kappa (κ) statistic was conducted to determine consistency among raters. All statistical tests were two-sided. P < 0.05 were considered to be statistically significant. All analyses were carried out using the SPSS 21 software package (IBM).

For immunohistochemical analysis of PD-L1 expression in FFPE tissue, we validated a novel mAb against PD-L1 (clone EPR1161(2)) and compared its performance with the previously validated PD-L1 mAb (E1L3N) XP (14, 24–31). The prostate cancer cell line DU145 served as a positive control. Flow cytometric analysis (FACS) of untreated DU145 cells compared with concentration matched mouse IgG isotype control demonstrated positive PD-L1 membrane expression in 40% of cells (Fig. 1A). For Western blot analysis, cell lysates from untreated DU145 and MCF7 cells were probed with PD-L1 mAbs. Both antibodies detected a band at the expected size of glycosylated PD-L1 (∼55 kDa) for the FACS-positive cell line DU145 (Fig. 1B). As reported previously (36), PD-L1 was not detected in MCF7 cells. Antibody specificity was confirmed by small interfering RNA (siRNA) knockdown of PD-L1 and Western blot analysis (Fig. 1C). Cross-reactivity was excluded using immunizing peptide blocking experiments. Primary antibody neutralization with a specific blocking peptide prior to Western blot and immunohistochemical proceedings abolished immunoreactivity and thus verified specificity of this PD-L1 mAb clone EPR1161(2) (Fig. 1D).

Detailed clinicopathologic characteristics of the patient cohorts included in the present study are shown in Table 1. Tumoral immunoreactivity of PD-L1 was uniformly homogenous, and scoring results were highly concordant (κ = 0.75, P < 0.001). PD-L1 expression was detected at the membrane or in the cytoplasm of the tumor cells (Fig. 2). Additionally, stromal and tumor infiltrating lymphocytes displayed a strong PD-L1 immunoreactivity.

In the training cohort, 52.2% of cases (109/209) expressed moderate to high levels of PD-L1, which positively correlated with proliferation (Ki-67, P < 0.001, τ = 0.26) and AR expression (P < 0.001, τ = 0.21). These results were confirmed by performing the χ² test (Table 1). Furthermore, an association of PD-L1 with Gleason score was found (P = 0.004; Table 1). The introduction of a cutoff for patients' stratification is required for a Kaplan–Meier analysis. However, data dichotomization based on a cutoff leads to a loss of information on the one hand and a multiple testing problem due to the multitude of possible cutoffs on the other hand. Therefore, PD-L1 expression was initially analyzed as a continuous variable. In univariate proportional hazards model analysis, semi-quantitative PD-L1 expression was strongly prognostic for BCR [P = 0.004; HR, 2.37; 95% confidence interval (CI), 1.32–4.25; Table 2]. Kaplan–Meier survival analysis of PD-L1 expression dichotomized by median confirmed that high PD-L1 expression was associated with significantly reduced BCR-free survival (P = 0.022; Fig. 3A).

In the independent test cohort, moderate to high PD-L1 expression was detected in 61.7% of cases (377/611). The strong association of PD-L1 expression with AR (P < 0.001, τ = 0.16) and Ki-67 (P < 0.001, τ = 0.22) was substantiated in this cohort, whereas no association with Gleason score was verified. Along with known prognostic factors (pT status, Gleason score, surgical margins, and preoperative PSA levels), PD-L1 expression was shown to be significantly prognostic for BCR in univariate Cox proportional hazards model analysis (P = 0.011; HR = 1.49; 95% CI, 1.10–2.02; Table 2). After dichotomization, PD-L1 expression conferred a significantly shorter PSA relapse-free survival in Kaplan–Meier survival analysis (P = 0.009; Fig. 3B). An association of age and presurgical PSA with PD-L1 could only be shown in either the training or test cohort, respectively (Table 1).

In multivariate Cox proportional hazards model analysis of all patients, PD-L1 furthermore remained an independent prognostic factor of BCR (P = 0.007; HR, 1.46; 95% CI, 1.11–1.92; Table 2) when tested together with pT status, Gleason score, preoperative PSA, and surgical margins.

This study is the first to demonstrate that PD-L1 expression is not only highly prevalent in primary prostate cancer but is also an independent prognostic factor of disease progression in cohorts of primary tumors following RP. Consistent with previous studies, our results suggest an association of PD-L1 with aggressive tumor behavior in primary prostate cancer, indicating that PD-1/PD-L1 pathway activation assists the evasion of antitumor immune response, driving tumor proliferation and progression. Further studies in watchful waiting cohorts need to ascertain whether PD-L1 might also contribute to the identification of lethal prostate cancer.

We here comprehensively validated a novel monoclonal rabbit antibody against PD-L1 (clone EPR1161(2)) amenable to FFPE tissue and analyzed two large, well-characterized cohorts of primary prostate cancer after RP for PD-L1 expression. Our study shows that the immunoreactivity of clone EPR1161(2) for the detection of PD-L1 on FFPE tissue is specific and robust. In preliminary experiments (not shown), which we routinely conduct to establish a new antibody in our laboratory, we used different immunohistochemical platforms and found this antibody's performance to be independent of the detection system. It necessitates conventional antigen retrieval, but yields a crisp and clean signal with no background issues at all.

Increased expression of PD-L1 on tumor cells has previously been described for several malignancies, including glioblastoma, pancreatic, ovarian, breast, renal, head and neck, esophageal, and non–small cell lung cancer (37–42). Moreover, PD-L1 expression has been associated with poor prognosis and adverse clinicopathological characteristics (28, 31, 43–46). In line with these findings, we have found moderate to high PD-L1 expression in 52.2% to 61.7% of primary prostate cancers after RP. Multivariate analysis further revealed that high PD-L1 expression was significantly associated with reduced BCR-free survival independently of other clinicopathological factors, suggesting that expression of PD-L1 on tumor cells promotes tumor recurrence by interrupting antitumor immunity (29). In addition, we found a strong correlation of PD-L1 with tumor cell proliferation as estimated by Ki-67, which has been associated with a highly adverse prognosis and primary therapy failure in prostate cancer (47–50).

Preliminary data from a phase I study on an anti-PD1 mAb treatment (nivolumab) also indicated that immunohistochemically detected PD-L1 is a potential predictive biomarker for therapeutic blockade of the PD1/PD-L1 pathway (17). In the landmark study by Topalian and colleagues immunohistochemical assessment of PD-L1 in pretreatment cancer specimens from 42 patients revealed that objective response to treatment was seen exclusively in PD-L1–positive tumors (9/25, 36%; P = 0.006). More recent data suggest that PD-L1–positive tumors have higher response rates to agents targeting the PD1/PD-L1 pathway, although response to treatment was also recorded for PD-L1–negative cases (18–23). Notably, none of the 17 patients with metastatic CRPC included in the original study by Topalian and colleagues responded to anti-PD-1 treatment. For two cancers that were eligible for immunohistochemical analysis, no PD-L1 expression was detected (17), all of which suggesting that PD-1/PD-L1–targeted treatment was not particularly promising for prostate cancer patients. Although the sample size with only two prostate cancer specimens eligible for immunohistochemical analysis is clearly limited, the absence of PD-L1 expression might as well be due to the fact that only heavily pretreated CRPC were enrolled in this study. So far, it is speculative if the strong correlation of AR and PD-L1 observed in our study provides a viable explanation for low expression rates of PD-L1 and lack of response to anti-PD-1 treatment in CRPC (17). However, the high rate of PD-L1 positivity found in primary, and hence hormone-naïve prostate cancers, indicates that PD-1/PD-L1 pathway–targeted therapy might potentially be a novel treatment option, and immunohistochemical assessment of PD-L1 might accordingly represent a biomarker for the identification of patients eligible for this therapy.

Strikingly, Bishop and colleagues recently reported that CRPC patients resistant to enzalutamide showed elevated levels of PD-L1–expressing dendritic cells in blood (51). Additional cell line and xenograft experiments suggested that enzalutamide-resistant tumors might suppress immune response not only through intrinsic PD-L1 expression, but also via induction of PD-L1 expression on circulating dendritic cells. The authors concluded that PD-L1 expression on tumor cells might be a mechanism of non-AR–driven resistance to enzalutamide. In addition to patients with hormone-sensitive prostate cancer, these patients might therefore also potentially respond to anti-PD-L1 checkpoint blockade immunotherapy.

In order to elucidate the potential of PD-1/PD-L1–targeted therapy in prostate cancer, further studies are needed to clarify (i) if PD-L1 expression can contribute to identifying insignificant prostate cancer cases that may be spared immediate definitive therapy, (ii) if PD-L1 expression in primary tumors is predictive for response to anti-PD-L1 therapy, (iii) if castration-sensitive prostate cancers have low PD-L1 expression levels due to androgen suppression and subsequently, (iv) how much ADT itself downregulates PD-L1, and (v) if anti-PD-L1 therapy is effective at all in prostate cancer.

No potential conflicts of interest were disclosed.

Conception and design: H. Gevensleben, D. Dietrich, P. Brossart, G. Kristiansen

Development of methodology: H. Gevensleben, C. Golletz, M. Jung, M. Majores, B. Uhl, P. Brossart, G. Kristiansen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Gevensleben, C. Golletz, S. Steiner, M. Jung, M. Majores, J. Stein, B. Uhl, S. Müller, C. Stephan, K. Jung

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Gevensleben, D. Dietrich, T. Thiesler, K. Jung, P. Brossart, G. Kristiansen

Writing, review, and/or revision of the manuscript: H. Gevensleben, D. Dietrich, J. Stein, J. Ellinger, P. Brossart, G. Kristiansen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Gevensleben, C. Golletz, S. Steiner, M. Majores, B. Uhl, S. Müller, J. Ellinger, C. Stephan, K. Jung, G. Kristiansen

Study supervision: D. Dietrich, P. Brossart

Other (tissue microarray—construction and quality assurance): M. Majores

Other (collection and database organization for TMA-related follow-up analyses): M. Majores

Other [providing data for clinicopathologic correlations (SPSS-database construction and collecting follow-up data): J. Stein

Other (construction of the tissue microarrays): J. Stein

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.