Introduction

Antibody-mediated blockade of the programmed death-1/programmed death ligand-1 (PD-1/PD-L1) interaction is a promising clinical strategy in several tumor types,1, 2, 3 and PD-1-blocking reagents have recently been approved for both melanoma and squamous lung cancer. Immunologically, these agents function by blocking the interaction between PD-1 on a partially activated or ‘exhausted’ CD8+ T-cell and its primary ligand (PD-L1), which is expressed on tumor cells, as well as myeloid cells in the tumor microenvironment. That interaction inhibits CD8+ T-cell function, thus blocking PD-1 or PD-L1 can lead to T-cell activation, proliferation and tumor cell lysis.4, 5 Unfortunately, phase I data from men with metastatic castrate-resistant prostate cancer (mCRPC) suggest that PD-1 blockade is less effective in prostate cancer compared with other tumor types, with 0 out of 17 patients showing objective responses in the initial phase I study.6, 7

The relative lack of objective responses to PD-1 blockade in prostate cancer could be because of the observation that PD-L1 expression appears to be rare in prostate cancer, at least in the data accumulated thus far.2 This is not the case for immune-sensitive diseases such as melanoma, where PD-L1 expression is common,8 and expression of PD-L1 correlates with response to PD-L1 blockade.9, 10 Several studies show that PD-L1 expression can be mediated by proinflammatory factors secreted by immune cells, a phenomena that has been termed ‘adaptive immune resistance’.1, 11, 12 This hypothesis is supported by the colocalization of immune cells and PD-L1 expressing tumor cells, suggesting that signals from infiltrating immune cells trigger tumor cells to upregulate PD-L1.8 Other data suggest that PD-L1 upregulation may be driven by oncogene expression, a process termed ‘innate immune resistance’. This hypothesis suggests that PD-L1 is constitutively expressed in response to aberrant signaling in other pathways. For example, Parsa et al.13 showed that loss of the tumor suppressor phosphatase and tensin homolog (PTEN), and associated phosphoinositide 3-kinase (PI3K) activation, was associated with PD-L1 expression in both glioblastoma and prostate cancer.14 To date, the relationship of ‘innate immune resistance’ whereby PD-L1 is constitutively upregulated when PTEN is lost has not been further explored in-depth in prostate cancer, especially in primary tumor specimens.

In prostate cancer, PTEN loss is a relatively common occurrence and has been established as a predictor of poor clinical outcomes and progression to mCRPC.15, 16, 17, 18 Based on the lack of clinical responses to anti-PD-1 therapy in patients with mCRPC, we hypothesized that PTEN loss might not be associated with PD-L1 expression in prostate cancer. In the current study, we demonstrate that prostate cancer occasionally expresses PD-L1, and that this expression has no clear relationship to PTEN loss or activation of the PI3K pathway in this disease.

Materials and methods

Cell culture

Human prostate cancer cell lines were grown in a monolayer under standard culture conditions with 5% CO2 in a 37 °C incubator. Tissue culture media were supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. CWR22Rv1,19 E006AA,20 DU14521, LNCAP21 and PC321 were grown in RPMI 1640 medium (Invitrogen, Grand Island, NY, USA); LAPC-421 was grown in Iscove’s modified Dulbecco’s medium (Invitrogen); and VCAP22 was grown in Dulbecco’s modified Eagle's medium (Invitrogen). Cell lines were a generous gift from Dr John Isaacs, PhD, Johns Hopkins University. Cell lines have been previously described. Short tandem repeat and mycoplasma testing was performed on all cell lines. For stimulation experiments, cells were plated at a density of ~5000 cm−2 into 100 cm2 tissue culture dishes and allowed to adhere for 24 h. After 24 h, cell cultures were treated with recombinant human interferon-γ (IFN-γ) (300-02; Pepro Tech, Rocky Hill, CT, USA) at a concentration of 100 U ml−1 for 48 h before harvest23 or with 10 μm bicalutamide (B9061; Sigma-Aldrich, St Louis, MO, USA) for 48 h.24 All stimulation experiments were repeated at least three times.

Flow cytometry

Cells were stained with phycoerythrin (PE)-labeled mouse anti-human PD-L1 (CD274, clone MIH 1, 12-5983-41; Ebiosciences, San Diego, CA, USA), PE-labeled mouse anti-human second major ligand for PD-1 (PD-L2) (CD273, clone MIH18 clone, 12-5888-41; Ebiosciences) or fluorescein-labeled mouse anti-human HLA DR, DP, DQ (clone Tu39, 555558; Becton Dickinson, Franklin Lakes, NJ, USA). Antibodies were diluted 1:200 and staining was performed in FACS buffer for 15 min at room temperature. The same procedure was performed using isotype control antibodies, PE-labeled mouse immunoglobulin G1κ (IgG1κ) or fluorescein-labeled mouse IgG2aκ. Cells were analyzed using a BD FACS Calibur (Becton Dickinson) and FlowJo software (Tree Star, Ashland, OR, USA). Flow cytometric analyses were repeated at least three times.

Western blotting

Cells were grown with or without IFN-γ as above and lysed directly in tissue culture flasks using ice-cold RIPA buffer (R0278; Sigma-Aldrich) supplemented with Sigmafast protease inhibitor tablets (S8820-2TAB; Sigma-Aldrich). Protein quantitation was performed using a Coomassie Blue (Fisher Scientific, Waltham, MA, USA) chromogenic assay. Eighty micrograms of protein per well was resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane. Primary antibodies included either p-AKT Ser473 (clone D9E, 9188; Cell Signaling, Danvers, MA, USA) or PTEN (clone D4.3 XP 9271; Cell Signaling) at a concentration of 1:1000. Protein was detected using horseradish-peroxidase linked anti-rabbit IgG as a secondary antibody, and was developed using the Amersham ECL Detection Kit (RPN2232; GE Healthcare UK, Little Chalfont, Buckinghamshire, UK). These blots were subsequently stripped and reincubated with anti-actin antibody at a concentration of 1:5000 (A2066; Sigma-Aldrich). Images were obtained using Biospectrum Imaging Center/VisionWorks LS software (UVP, Upland, CA, USA). Western blot analysis was performed two times.

Human prostate samples

Twenty whole-mount paraffin-embedded primary prostate cancer specimens were obtained from a prostatectomy database maintained by the Department of Pathology at the University of Colorado under Institution review board-approved protocol, 00-812. Informed consent was obtained for tissue acquisition of each sample. Eleven of these samples were treated with leuprolide before prostatectomy and nine received no pretreatment. This was carried out at the discretion of the treating physician and not part of a clinical trial.

Immunohistochemistry

Immunohistochemistry for PD-L1 was performed by the Johns Hopkins Pathology Core using the 5H1 clone of the mouse anti-human CD274 monoclonal antibody, with a mouse IgG1 isotype as a negative control.8 Immunohistochemistry for PTEN was also performed by the Core Laboratory using rabbit anti-human PTEN antibody clone D4.3 XP (9188; Cell Signaling) at a 1:50 dilution as described previously.16

Scoring

Scoring was performed by two independent pathologists (AMD, RAA). PTEN status was scored using a validated dichotomous scoring system.16 PD-L1 scoring has also been described previously.8 Briefly, samples in which >5% of malignant epithelial cells show membrane staining were designated as PD-L1 positive.

Results

IFN-γ upregulates PD-L1 (adaptive immune resistance)

Cytokine-driven upregulation of PD-L1 has been implicated in the ability of tumors to evade detection and destruction by tumor-specific cytotoxic CD8+ T-cells. To assess the potential of human prostate cancer to demonstrate ‘adaptive immune resistance’,1, 11, 12 PD-L1 expression was assayed in seven human prostate cancer cell lines: CWR22Rv1, DU145, E006AA, LAPC-4, LNCaP, PC3 and VCaP by flow cytometry. We also quantified the expression of PD-L2, as well as class II major histocompatibility complex, which may contribute to immune evasion by binding to the immune checkpoint molecule LAG-3 on exhausted CD8+ T-cells.25 To simulate broadly a TH1-driven immune response, cells were cultured in the presence of IFN-γ for 48 h; IFN-γ is a known mediator of PD-L1 expression in human cancer cell lines and has also been shown to induce major histocompatibility complex.1, 23, 26 The majority of prostate cancer cell lines tested upregulated PD-L1 in response to IFN-γ, suggesting that prostate cancer cells are indeed capable of ‘adaptive immune resistance’. Three representative lines are shown in Figure 1, additional cell lines are shown in Supplementary Figure 1. Interestingly, two of the cell lines, Du145 and PC3, demonstrated PD-L1 expression at baseline, suggesting the possibility of ‘innate immune resistance’. All of the cell lines examined expressed PD-L1 to some degree, except for the two lines derived from lymph node metastases (LAPC-4 and LNCaP; Table 1 and Supplementary Figure 1). Prostate cancer is a hormone-sensitive tumor and alterations in the PD-1 pathway have been linked to steroid hormones.24, 27 To determine whether activity of the androgen receptor contributes to the expression of PD-L1 or other immune-related molecules, cells were treated with the androgen receptor antagonist, bicalutamide. As expected, the two cell lines that do not express androgen receptor showed no effects of bicalutamide treatment (Figure 2). The other cell lines tested also evinced no changes in immune-related cell surface markers either (Tables 2 and 3 and Supplementary Figure 2).

Figure 1
figure 1

Human prostate cancer cell lines display varying expression of PD-L1 in response to interferon-γ (IFN-γ). Histograms representing PD-L1 surface expression as detected by flow cytometry in human prostate cancer cell lines with and without exposure to IFN-γ. Open histograms represent cells stained with an isotype immunoglobulin G1 (IgG1) antibody tagged with a matching phycoerythrin (PE) fluorochrome. Shaded histograms represent PD-L1. Quantitative measurements reflect mean fluorescence index (MFI). LNCaP does not express PD-L1 at rest or in response to IFN-γ. DU145 expresses PD-L1 at rest and has little response to IFN-γ, indicating an innate immune resistance phenotype. PC3 expresses PD-L1 at rest but robustly upregulates this expression in response to IFN-γ displaying a more classic adaptive immune resistance phenotype. PD-L1, programmed death ligand-1.

Full size image
Table 1 Response to IFN-γ
Full size table
Figure 2
figure 2

Human prostate cancer cell lines display no change in PD-L1 expression in response to bicalutamide. Histograms representing PD-L1 surface expression as detected by flow cytometry in human prostate cancer cell lines with and without exposure to bicalutamide. Open histograms represent cells stained with an isotype immunoglobulin G1 (IgG1) antibody tagged with a matching phycoerythrin (PE) fluorochrome. Shaded histograms represent PD-L1. Quantitative measurements reflect mean fluorescence index (MFI). LNCaP, DU145 and PC3 displayed no changes in PD-L1 expression upon exposure to bicalutamide. PD-L1, programmed death ligand-1.

Full size image
Table 2 Response to bicalutamide
Full size table
Table 3 PD-L1 expression is independent of PTEN status in human prostate cancer cell lines
Full size table

Expression of immunologic markers is independent of PTEN status

Loss of the tumor suppressor PTEN has been linked to increased expression of PD-L1 in glioblastoma13 and prostate cancer.14 In patients, PTEN loss occurs in up to 60–70% of primary prostate cancer cases and portends a less favorable outcome.15, 16, 17, 18 Mechanistically, loss of PTEN derepresses the PI3K pathway leading to transcriptional activation of several downstream targets and is associated with mTORC2 signaling activation and the phosphorylation of AKT to its active form, phospho-AKT (p-AKT S473).28 To determine whether PTEN loss and PD-L1 upregulation are linked in prostate cancer cell lines, PTEN and p-AKT levels were evaluated by western blotting, and correlated with PD-L1 protein expression. Consistent with previously published reports, PTEN expression was intact in CWR22Rv1, DU145 and LAPC-4, and absent in E006AA and PC328 (Figure 3). All lines exhibiting PTEN loss had coordinate upregulation of p-AKT. A paired evaluation with and without IFN-γ was carried out on each sample. As expected, PTEN status was not affected by IFN-γ as this is a genetic loss-of-function mutation. Importantly, p-AKT expression was not altered by exposure of cell lines to IFN-γ, indicating that IFN-γ was not activating PI3K by another mechanism. Perhaps, most significantly, there was no correlation between PTEN status and PD-L1 expression in these cell lines, arguing against PTEN loss driving innate immune resistance in vitro.

Figure 3
figure 3

Activation of the PI3K/MAPK pathway does not correlate with PD-L1 expression in human prostate cancer cell lines. Western blot quantitating PTEN and p-AKT protein in CWR22Rv1, LAPC-4, PC3, DU145 and E006AA with and without exposure to IFN-γ. PD-L1, programmed death ligand-1; PTEN, phosphatase and tensin homolog.

Full size image

PD-L1 expression in human prostate cancer samples is independent of PTEN status

To determine whether there was an association between PTEN status and PD-L1 expression in patients, 20 paraffin-embedded whole-mount primary prostate cancer samples were stained for both PTEN and PD-L1, using previously validated immunohistochemical (IHC) protocols.8, 16 Broadly consistent with previously reported results, 25% (5/20) of samples demonstrated loss of PTEN (Table 4). For these studies, PD-L1 ‘positivity’ was defined as 5% membrane staining, based on previous work showing a correlation between marker positivity and clinical response to blocking antibody.8, 10 Using this cutoff, 3/20 samples (15%) had focal areas of PD-L1 positivity, although in only two of the three positive samples was plasma membrane staining clearly observed on malignant epithelial cells (Supplementary Table 1). Strikingly, none of these PD-L1-positive samples demonstrated loss of PTEN, that is, consistent with the cell line data, none of the five samples with PTEN loss were PD-L1 positive (Figure 4 and Table 4). Of note, 11/20 patients in this series had received anti-androgen hormonal therapy before surgery. Again similar to our cell line results, none of the hormonally treated samples were positive for PD-L1. Interestingly, several of the samples in this series had PD-L1 expression in the 1% or lower range, the clinical significance of those low levels of staining is less clear10 (Supplementary Table 1). Some of this low-level staining appeared to occur in areas of proliferative inflammatory atrophy, a proposed precancerous inflammatory state of the prostate.29 Interestingly, some of this low-level staining occurred on infiltrating immune cells (Figure 5), and this finding has been correlated with response to PD-1 blockade in some clinical studies.30 Taken together, these results support previous data showing that PD-L1 expression in prostate cancers is rare, and that when expression does occur, it is not associated with regions of PTEN loss.

Table 4 PTEN vs PD-L1 expression in human prostate cancer samples
Full size table
Figure 4
figure 4

Loss of PTEN does not correlate with PD-L1 expression in resected primary prostate carcinomas. PD-L1 and PTEN staining in tumor lesions from two different patients. (a) Case number 8 showing focally positive PD-L1 area. Arrow indicates positive tumor cell staining, which is predominantly on the plasma membrane. (b) Similar region from adjacent section of case number 8 showing intact PTEN staining in all tumor cells (arrow shows a group of tumor cells with intact PTEN staining). (c) Case number 9 showing negative staining for PD-L1 in all tumor cells (arrow shows tumor cells without PD-L1 staining). (d) Adjacent region showing PTEN loss in nearly all of the tumor cells (arrow shows tumor cells with negative PTEN staining; arrowhead shows stromal element with strongly positive PTEN staining). All images taken at × 100 original magnification. PD-L1, programmed death ligand-1; PTEN, phosphatase and tensin homolog.

Full size image
Figure 5
figure 5

PD-L1 is expressed on infiltrating inflammatory cells. Surface expression of PD-L1 demonstrated by macrophages (arrows) infiltrating the lumen of a benign gland from prostatectomy Case number 1. Image taken at × 200 original magnification. PD-L1, programmed death ligand-1.

Full size image

Discussion

Immune checkpoint blockade with anti-PD-1 or anti-PD-L1 is emerging as a promising treatment modality in several tumor types,1, 2, 3 but to date blocking this interaction has been relatively disappointing in prostate cancer. The primary reason for this is likely that prostate cancer patients have little or no PD-L1 expression in their tumors, as demonstrated by several previous IHC studies and as confirmed here.

PD-L1 expression can be driven by two major mechanisms. In the most common, ‘adaptive immune resistance’, PD-L1 expression on tumor cells is driven by immune cell production of proinflammatory cytokines such as IFN-γ. PD-L1 upregulation, in turn, protects the tumor cells from CD8+ T-cell mediated attack by binding to PD-1 on tumor-specific cytotoxic T-cells. Our data confirm that prostate cancer cells can express PD-L1 in vitro, in response to proinflammatory signals. In human samples, although, PD-L1 expression was relatively rare, confirming previous data2 and suggesting that the paucity of PD-L1 expression in patients may be because of a locally immunosuppressive environment that very effectively dampens CD8+ T-cell production of IFN-γ, as has been clearly demonstrated in several animal models.31, 32

The second major mechanism underlying PD-L1 expression is known as ‘innate immune resistance’, in which tumor cells autonomously upregulate PD-L1, potentially under the influence of oncogenic pathways.12 In the case of prostate cancer, previous data suggested that loss of PTEN, a common event in prostate cancer, is potentially associated with PD-L1 expression.14 If that were the case, then PD-L1 expression in prostate cancer might be assumed to be fairly widespread, as PTEN loss is a common event in prostate cancer. We tested this hypothesis in both cell lines and a small case series, and could not confirm the notion that PD-L1 expression is associated with PTEN loss, despite using well-validated staining protocols for both markers. Although there were only five samples of PTEN-negative prostate cancer in this series, if ‘innate immune resistance’ were a major mechanism driving PD-L1 expression in prostate cancer, it stands to reason that a few PD-L1-positive cells would have been seen. The reasons for this discrepancy are not immediately obvious, and a larger case series or tissue microarray study could be considered to explore this association further. Nevertheless, these data do support the conclusion that ‘innate immune resistance’ triggered by PTEN loss and subsequent PI3K pathway activation is not likely to be a major underlying mechanism driving PD-L1 expression in prostate cancer. It should be noted that a role for ‘innate immune resistance’ has recently been challenged in melanoma as well, where prior suggestions that common mutations in melanoma (i.e. BRAF V600E) were associated with PD-L1 expression were not supported in studies using patient samples.33

The ability of androgen ablation to modulate a proinflammatory tumor microenvironment in prostate cancer has been demonstrated by several previous studies.34, 35 For this reason, it has been postulated that androgen ablation might increase PD-L1 expression on tumor cells by attracting infiltrating immune cells and triggering an adaptive immune response.34, 36, 37 Therefore, it is surprising that in this study PD-L1 expression on prostate cancer cells remained rare and was not increased in those which were pretreated with androgen ablation. There was also no evidence of increased immune infiltration in androgen-ablated samples (Supplementary Figure 1). This suggests that if inflammation occurred in response to androgen ablation, it may have been transient and was not captured by this study. The elegant work by Mercader et al.34 indicates that immune infiltration peaks within 2 weeks of anti-androgen administration,34 and patients included in this study had been ablated for several months before surgery. These findings suggest that androgen ablation alone may not be sufficient to overcome the suppressive immune microenvironment of prostate cancer outside of a narrow therapeutic window. It is also important to note that from a translational standpoint, anti-PD-1 therapy has never been administered to newly diagnosed patients with prostate cancer. All of the phase I and II studies have been carried out in patients with mCRPC.7, 9 Therefore, it is possible that the lack of efficacy observed in clinical studies reflects the intrinsic biological differences between primary and metastatic prostate tumors.

In summary, these data suggest that treatments directed at the PD-1/PD-L1 interaction are unlikely to be successful as monotherapies in prostate cancer. Nevertheless, it is possible that the acute inflammation driven by androgen ablation34, 35 could transiently increase PD-L1 expression, suggesting a strategy in which blockade is combined with androgen ablation. Similar effects could be mediated by prostate cancer-specific vaccination, although the PD-L1 staining data above suggest that if PD-L1 expression does occur, it may be transient in nature. It is also possible that one of the many other immune checkpoint/ligand pairs are of greater importance compared with the PD-1/PD-L1 axis in prostate cancer; ongoing studies in our lab and many others are actively investigating that hypothesis.