HIF1α-dependent and independent pathways regulate the expression of PD-L1 in prostate cancer

PD-L1/PD-1 pathway is a major pathway exploited by human cancer types, which is a target for current immunotherapy. We investigated tumor microenvironmental factors involved in PD-L1 induction in prostate cancer (PC). We studied the expression of PD-L1 in a series of 66 PCs, in parallel with the expression of hypoxia- and acidity-related immunohistochemical markers (Hypoxia-inducible factor HIF1α, and lactate dehydrogenase LDHA) and tumor-infiltrating lymphocyte TIL density. Experiments with three PC cell lines, the 22Rv1, DU145, and PC3 were conducted focusing on the inducibility of PD-L1 by hypoxia, acidity, lymphocyte interactions, and radiation. In tissues, PD-L1 expression by cancer cells was directly related to PD-L1 expression by  TILs and macrophages (p < 0.05), and the overexpression of HIF1α and LDH5 (p < 0.05). TIL density was inversely related to ΗΙF1α (p = 0.02). Exposure of PC cell lines to hypoxia strongly induced PD-L1 and protein and mRNA levels, directly controlled by HIF1α function (p < 0.001). Irradiation with 20 Gy had no apparent effect on PD-L1 expression. Culturing PC cell lines with culture medium (CM) from PBMCs strongly induced PD-L1 at protein and mRNA levels, independently from HIF1α, which was also confirmed when cells were incubated with Interferon-γ (p < 0.001). It is concluded that the combination of anti-PD-L1/PD-1 immunotherapy with hypoxia/HIF-targeting may be important in the treatment of specific subgroups of PC patients.


Introduction
Blocking of immune-control checkpoint molecular pathways (ICMPs) has emerged as a promising strategy for cancer therapy. Recent studies have prompted a lot of understanding of the immunotherapy-related mechanisms targeting ICMPs [1]. The binding of the PD-1 (programmed cell death protein 1) T-cell (T-lymphocytes) receptor to the PD-L1 (programmed death ligand 1) ligand expressed by regulatory immune cells or cancer cells restricts normal T-cell function. The PD-1/PD-L1 mechanism is essential for normal immune physiology, as it suppresses immune over-activation allowing tolerance to various antigens and preventing autoimmune reactions [2]. By exploiting this very immunosuppressive pathway, cancer cells suppress the antitumor immune response [3].
Over the years, studies have also brought forward the critical role of the tumor microenvironment (TME) in the repression of the cytotoxic function of tumor-infiltrating immune cells. Poor tumor vascularization and impaired blood flow due to an immature vascular network produce a hypoxic TME. Cancer cells undergo a reprogramming of their metabolic profile characterized by intense glycolysis and anaerobic usage of pyruvate, even under aerobic environmental conditions [4]. Glycolytic metabolism involves lactate dehydrogenase A (LDHA) overexpression so that pyruvate is converted to lactic acid, one of the principal causes of intratumoral acidity [5]. Several recent studies suggest that a hypoxic and acidic tumor microenvironment prevents lymphocyte proliferation and the antitumor activity of T-cells and natural killer NK-cells, also promoting the accumulation of myeloid and M2-type macrophages with immunosuppressive activity [6][7][8][9][10]. Moreover, PD-L1 gene has been postulated to be under the direct control of the Hypoxia-Inducible Factor 1α (HIF1α) [11].
Tumor-infiltrating lymphocytes and monocytes secrete immunostimulatory or immunosuppressive cytokines, including interferons (IFNs), interleukins (ILs), tumor necrosis factors (TNFs), or transforming growth factor-β (TGFβ). The biological activity of such cytokines on cancer cells' expression of immune checkpoint molecules (ICMs) is under intense investigation. Up-regulation of the PD-L1 expression in cancer cells by such cytokines has been reported [12]. IFN-γ, for example, induces PD-L1 expression in a dose-dependent manner, which raises concerns, as cytokinesecreting antitumor immune cells may produce a cancer cell defensive response that blocks immunity [13,14].
In the current study, we investigated the role of intratumoral hypoxia and anaerobic metabolism in the expression of PD-L1 in prostate cancer (PC) at the tissue level and in vitro experiments. We provide direct evidence that HIF1α and IFNγ are involved in distinct molecular pathways controlling the up-regulation of PD-L1 expression in this common human malignancy.

Tissue immunohistochemistry
The immunohistochemical method used to detect PD-L1, HIF1α, and LDH5 in formalin-fixed paraffin-embedded PC tissues has been reported in previous studies of ours [15,16]. PD-L1 protein was detected with the rabbit monoclonal anti-PD-L1 antibody [clone CAL10; BiocareMedical, CA, USA, dilution 1:1000, 60 min incubation]. For the detection of LDH5 (composed of 4 LDHA subunits), we used the Ab9002 (Abcam, UK) sheep polyclonal antibody raised against the human placenta LDH5 from at dilution of 1:200, and overnight incubation. The HIF1α protein was detected using the mouse monoclonal ESEE122 antibody (gift by professor K.C. Gatter, Oxford, UK), raised against a recombinant fragment corresponding to human HIF1 alpha aa 300-550 (dilution 1:200, overnight incubation).
For PD-L1, the percentage of cancer cells with strong membrane (with or without cytoplasmic) expression was recorded in the entire tissue section in × 200 optical fields, and the mean value was used to score each case. Membrane staining in at least 1% of cancer cells was considered for PD-L1 positivity. The extent of PD-L1 expression in tumorinfiltrating lymphocytes and monocytes (TILMs) in the tumor stroma was assessed in all available optical fields. The percentage of stroma area with PD-L1 + TILMs was used to score cases. Moreover, the tumor-infiltrating lymphocyte TIL density was also assessed in hematoxylin and eosin sections by counting the number of TILs per optical field in all available optical fields covering the tumor and using the mean value to score each case. For HIF1α and LDH5 immunostaining, the proportion of tumor cells expressing cytoplasmic and/or nuclear reactivity was recorded after examining the entire tumor area at × 200 magnification. The median value was used to score each case. Expression in ≥ 50% of cells defined cases with high HIF1α or LDH5 expression.

Hypoxia and acidosis conditions
For the hypoxia conditions, PC cell lines 22Rv1, DU145, and PC3 were seeded in 6-well plates, in an approximate yield of 2 × 105 cells per well. After 24 h, cells were incubated in hypoxic conditions in FLUOstar® Omega (BMGLABTECH) for 24 h in mixed gas airflow, consisting of 5% CO2, 1% O2, and 74% N2. For the acidosis conditions, PC cell lines 22Rv1, DU145, and PC3 were seeded in 6-well plates, in an approximate yield of 2 × 105 cells per well. After 24 h, cells were incubated in culture medium DMEM (Dulbecco's Modified Eagle Medium) containing 18.12 mmol/L HCl [37%; CL00.0310.2500; CHEM-LABS] (pH = 6.5) for 24 h. After that, protein isolation for Western Blot Analysis and RNA isolation protocol for Real-Time PCR (Polymerase Chain Reaction) experiments was performed, following the experimental procedures mentioned below.

RNA interference procedure
Prostate cancer cell lines 22Rv1, DU145, and PC3 were stably transfected with a second-generation Lentiviral system via two plasmid vectors purchased from GenePharma [Shanghai GenePharma, China], the first one encoding unique sequences for the shRNA (Small hairpin RNAs) of HIF1α gene [LV10N(U6/mCherry&Puro); HIF1Ahomo-964; Sequence: 5′GCT GAT TTG TGA ACC CAT TCC3′] and the other one free of any coding sequence. Both plasmids included a single restriction site for the EcoRI, genes conferring resistance in Ampicillin and Puromycin, and a gene expressing the mCherry protein. Plasmid amplification was performed on DH5-Alpha competent bacteria cells. The selection procedure of the successfully transformed bacterial clones was performed via Ampicillin. The amplified-plasmid DNA extraction and purification were performed using the NucleoSpin Plasmid Miniprep kit (REF#740,588.10, Macherey-Nagel, Germany). Transient transfection of the host cell line HEK293T with the lentiviral vector [LV10N (U6/mCherry&Puro&Amp)] was performed in order to produce the lentiviral particles. The lentiviral vector was combined with the adequate plasmids, an envelope plasmid [pczVSV-G; cat#8454, Addgene] and a packaging plasmid [pCMVR8.74; cat#22,036, Addgene]. The aforementioned transfection was performed using Lipofectamine2000 transfection reagent [cat#11,668,019, Thermo-Fisher Scientifc Inc., USA]. Lentiviral particles, produced from HEK293T cells, were isolated from the supernatant, centrifuged at 1000 g for 10 min, filtrated and used to infect 22Rv1, DU145, and PC3 cells for 48 h. Successfully transfected cells were selected using Puromycin at a concentration of 1 μg/mL to 4 μg/mL.

Co-Cultures and irradiation protocol
PC cell lines 22Rv1, DU145, and PC3, cell lines containing a non-coding (nc) plasmid, and cell lines with permanent silencing of HIF1α gene, were exposed to 20 Gy of irradiation, using the 6-MV beam produced by an I Infinity™ linear accelerator (Elekta, Stockholm, Sweden). The medium of these cultures was used to incubate the respective cancer cell lines for two days (matched types). Moreover, cell lines were incubated with culture medium obtained from PBMCs from a healthy donor cultured for 2 days in DMEM cell culture medium. PD-L1 expression was evaluated after 48 h via western blot analysis.  for PD-L1 and LDH5, and 18 μg for HIF1a protein were loaded for analysis. Protein samples were separated on discontinuous SDS gels using 15%, 10%, and 8% separating gels for PD-L1, LDH5, and HIF1a proteins, respectively, and 6% stacking gel. Electroblotting was performed with PVDF membrane [Immobilon-P PVDF for WB analysis, 26.6 × 1.8, pore size 0.45um; Cat. no IPVH00005; Merck Millipore, Germany], and then membranes were blocked with 5% Non-fat dry milk in 150 mM NaCl, pH 7.

Statistical analysis
Graph presentation and statistical analysis were performed by using the GraphPad Prism 8 version. The paired or unpaired two-tailed t test was applied for analysis of groups of continuous variables, as appropriate. A p value of < 0.05 was considered for significance.
Cytoplasmic/nuclear expression of HIF1α was evident in 36/56 (64.3%) cases and ranged from 0-100% of total cancer cells (median 30%); Fig. 1C. High expression (> 50% of cancer cells) was noted in 21/56 PCs. Cytoplasmic/nuclear expression of LDH5 was recorded in 45/56 (80.3%) cases and ranged from 0 to 100% of total cancer cells (median 70%); Fig. 1D. High LDH5 expression (> 50% positive cancer cells) was noted in 32/56 PCs. PD-L1 expression by cancer cells was directly related to high HIF1α and high LDH5 expression in cancer cells; Fig. 1E; Table 1. The median percentage of HIF1α + cancer cells was 60% in cases with high PD-L1 cancer cell expression vs. 20% in cases with low (p < 0.01). The median percentage of LDH5 + cancer cells was 90% in tumors with high PD-L1 cancer cell expression vs. 30% in tumors with low (p < 0.05). Expression of PD-L1 by TILMs was not related to HIF1α or LDH5 expression.
The tumor-infiltrating lymphocyte TIL density ranged from 6 to 520 lymphocytes per × 200 optical field. The median value was 96, and this was used to split PCs in two groups of low and high TIL density. Out of 21 PCs with high HIF1α expression, 6 (28.5%) had high TIL density, while out of 45 cases with low HIF1α expression, 22 (48.9%) had high TIL density (p < 0.05). There was no association between LDH5 expression and TIL density.

Basal levels of PD-L1 in PC cell lines
Western blot analysis of PD-L1 expression in 22Rv1, DU145, and PC3 PC cell lines showed very low expression in the 22Rv1 hormone-sensitive cell line. In contrast, a robust expression was evident in the DU145 (2.6-fold compared to 22Rv1) and, especially, in the PC3 cell line (3.9-fold compared to 22Rv1) (Fig. 1F,G). The Western blot patterns of PD-L1 expression were also confirmed in immunocytochemistry, where the 22Rv1 was negative, while the DU145 and PC3 cell lines expressed PD-L1 in 20% and 30% of cancer cells, respectively (Fig. 1H). In RT-PCR, the DU14 and PC3 cell lines had 1.6-and 2.5-fold increased PD-L1 mRNA expression compared to 22Rv1 (Fig. 1I) (*p < 0.05, **p < 0.01, ***p < 0.001).

Effects of hypoxia/acidity on PD-L1 expression
PC cell lines were exposed to hypoxic and acidic conditions, as described in the methods. In western blot analysis, hypoxia strongly induced the expression of PD-L1 in all three cell lines (p < 0.001; Fig. 2A,B). This was also confirmed in RT-PCR (p < 0.01) (Fig. 2C). Acidity had a

Silenced HIF1α gene and response to hypoxia
The creation of stably transfected sh22Rv1, shDU145, shPC3 cell lines with shRNA of the HIF1α was confirmed with RT-PCR analysis (Fig. 2D). Gene expression of HIF1α was significantly reduced in all three PC cell lines (in 22Rv1 cell line 3.3-fold compared to control cell line, p < 0.001; in DU145 cell line 4.76-fold compared to control cell line, p < 0.001; in PC3 cell line 7.14-fold compared to control cell line, p < 0.001). Transfection with non-coding sequence (nc22Rv1, ncDU145, ncPC3) had no effect on HIF1α expression.

Discussion
The PD-L1/PD-1 immune checkpoint inhibitory pathway is the best studied and targeted for the development of immunotherapy. Monoclonal antibodies (MoAbs) targeting PD-L1 or PD-1 have revolutionized the treatment of advanced cancer [17]. Unlike other solid tumors, anti-PD-L1/PD-1 immunotherapy is still investigational for PC and is mainly approved for mismatch-deficient tumors refractory to hormonal therapy and chemotherapy [18]. The addition of anti-PD-L1/PD-1 immunotherapy to hormonal therapy or chemotherapy has not provided clear evidence of benefit [19,20]. Although studies with RT combination with anti-PD-L1/PD-1 MoAbs are not available, radiotherapy increased by 2-3-fold the overall survival rates of patients with metastatic PC undergoing ipilimumab (anti-CTLA4 MoAb) immunotherapy [21]. The radio-vaccination effect during RT may significantly enhance the results of immunotherapy. The identification of subgroups of PC patients expected to benefit from anti-PD-L1/PD-1 immunotherapy would facilitate the conduct of clinical trials in patient subpopulations. Although extensively studied in other malignancies, PD-L1 expression is less studied in PC. It has been reported that PD-L1 is expressed by PC cells and stroma infiltrating immune cells in 10-50% of tumors, pending upon histological subtype and antibodies used for the immunohistochemical detection of the protein [22]. In our study, PD-L1 was expressed by cancer cells in 25% of PCs, while PD-L1 expression by tumor-infiltrating lymphocytes and macrophages was noted in 30% of PCs. As PD-L1 is an inducible gene [23], studying the microenvironmental conditions that are linked with PD-L1 expression is essential to identify subgroups of tumors with distinct biology that drives the PD-L1 expression status.
In the current study, we investigated the hypoxia and anaerobic metabolism/acidity-related gene expression, namely HIF1α and LDHA. A strong direct association of PD-L1 expression by cancer cells with these tumor markers was noted. Cell line experiments that followed confirmed that PD-L1 expression is strongly induced by hypoxia, although not by acidity, even in the hormonedependent cell line 22Rv1 which had very low steadystate levels of PD-L1. Hypoxia stimulated PD-L1 expression at both transcription and translation levels. This is in accordance with a previously reported study by Xu et al [24]. Similar results have been reported in other types of tumors, like bladder cancer and glioblastoma [25,26]. An important finding was also the association of HIF1α with low infiltration of PCs with TILs in accordance with previous studies in breast and head-neck cancer [27,28]. A hypoxic and acidic TME provides a functional barrier against cytotoxic lymphocytes, blocking their proliferation and cytotoxic activity, and furthermore, promotes regulatory T-cell and monocyte prevalence [6][7][8][9][10]. It seems that hypoxia has multiple immunosuppressive effects in the TME, by enhancing PD-L1 expression by cancer cells and repressing cytotoxic T-cell density and activity.
As HIF1α a key transcription factor involved in the adaptation of normal and cancer cells to hypoxia [29], we investigated the hypothesis that hypoxic PD-L1 regulation is HIF dependent. We created three PC cell lines with stably silenced HIF1α gene. Unlike the parental cells, these cells did not up-regulate PD-L1 mRNA and protein under hypoxic conditions, suggesting that PD-L1 is a HIF1α inducible gene. In fact, HIF-silenced cells showed a reduction of the PD-L1 expression, which may suggest that constitutively expressed levels of HIF1α in cancer cells drive PD-L1 expression that can be suppressed by HIFblockage. In 2014, Noman et al. showed the direct binding of HIF1α (but not ΗΙF2α) to a hypoxia-response element in the promoter of the PD-L1 gene [11]. Targeting HIF1α suppresses PD-L1 expression in experimental studies [30,31]. For this purpose, we exposed two PC cell lines with high PD-L1 protein and mRNA levels, namely DU145 and PC3, on 50 μmol/L of chrysin, a well-known inhibitor of HIF1α stability [32]. Exposure of the cell lines to chrysin resulted in down-regulation of both HIF1α and PD-L1, in both cell lines indicating the direct relation of a hypoxiainduced pathway on PD-L1 activity in PC. Similar studies [32] reported the down-regulation of PD-L1 via chrysin through the STAT3 (Signal transducer and activator of transcription factor 3) and NF-κB (Nuclear factor kappa B) pathways in hepatocellular carcinoma. In a subsequent step, we examined the role of the eventual immunogenic radiation-induced death and lymphocytic interactions in the release of soluble factors that could induce PD-L1 expression in viable cancer cells. Direct irradiation and culturing cancer cells with medium from irradiated cancer cells did not up-regulate PD-L1. On the contrary, culturing the cancer cell lines with medium from untreated PBMC cultures from a healthy donor resulted in PD-L1 up-regulation in all cell lines, including the ones with silenced HIF1α gene. It was, therefore, postulated that soluble factors released by lymphocytes induce PD-L1 independently of hypoxia and the HIF1α pathway. As IFNγ has been recognized to potently induce PD-L1 expression [33], we further tested whether this hypoxia-independent pathway is triggered by this cytokine. IFNγ strongly induced the expression of PD-L1 in all three PC cell lines, including the ones with silenced HIF1α gene. The up-regulation of PD-L1 by soluble factors released by lymphocytes should be sought in the activity of other cytokines and chemokines [34][35][36]. Although RT did not directly up-regulate PD-L1, post-irradiation inflammatory response may induce PD-L1 in PC patients treated with radiotherapy.
It is concluded that hypoxia induces PD-L1 expression in PC cell lines through a HIF1α-dependent mechanism. Soluble factors released by PBMCs and IFNγ can strongly induce this phenotype regardless of the activity of HIF1α, which may be intensified during radiotherapy of PC patients. We suggest that PD-L1 expression is induced at least by two distinct molecular pathways (Fig. 5). The first is dependent on the hypoxic tumor microenvironment, while the second is cytokine related and eventually driven by the tumor-infiltrating lymphocytes or monocytes in the tumor microenvironment. This finding demands further investigation to assess the cellular components of PBMCs that trigger PD-L1 expression. It is worrying that an intense infiltration of the tumor by TILs, a desirable antitumor immune response, can, at the same time, be the cause of the induction of a cancer defensive mechanism through PD-L1 up-regulation. Hypoxic tumors, also linked with poor TIL density, can escape anti-PD-1/PD-L1 immunotherapy as cytotoxic T-cell proliferation and function are blocked in such TME. Targeting the hypoxic conditions and HIF1α function in combination with anti-PD-L1 immunotherapy can overpass the adverse effect PD-L1, whether attributed to HIF1α or IFNγ in the TME, by enhancing the recognition of cancer cells and allowing effective cancer cell killing by the immune system. Author contribution ETX: Conception, Design, in vitro experiments, interpretation of data, writing of the paper, final approval, CK: in vitro experiments, writing of the paper, final approval, CN: Irradiation experiments, writing of the paper, final approval, AGG: Immunohistochemistry, writing of the paper, final approval, CK: Conception, Design, interpretation of data, writing of the paper, final approval, AG: Conception, Design, Immunohistochemistry analysis, interpretation of data, writing of the paper, final approval, MIK: Conception, Design, Study supervision, statistical analysis, interpretation of data, writing of the paper, final approval.
Funding Open access funding provided by HEAL-Link Greece. The study was supported by the Tumour and Angiogenesis Research Group and the Democritus University of Thrace Special Account, Project No 81006. Tumour and Angiogenesis Research Group, 81006, Michael I. Koukourakis.

Data availability
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.

Fig. 5
Two distinct cellular pathways regulating PD-L1 up-regulation. Pathway 1: HIF1α overexpression, whether hypoxia induced or inherently up-regulated, triggers PD-L1 overexpression on cancer cell membranes, producing a PD-L1-rich tumor (red structure), characterized by a hypoxic and/or acidic TME (gray). Anti-PD-1/ PD-L1 immunotherapy neutralizes the anti-immune PD-L1 enrichment (light blue). However, due to hypoxic/acidic conditions, cytotoxic T-cells (yellow/green structures) cannot proliferate and act in this adverse TME. Blockage of HIF1α activity will have a dual effect by repressing PD-L1 expression and reducing anaerobic metabolism and acidity. This may assist cytotoxic T-cells in thriving in the tumor and exerting antitumor cytotoxicity. Pathway 2: Constitutive or Radiotherapy-induced IFNγ produced by TILMs up-regulates PD-L1 expression by cancer cells, blocking cytotoxic T-cell activity. Anti-PD-1/PD-L1 immunotherapy blocks PD-L1 and facilitates cytotoxic T-cell activity, provided that this pathway does not co-exist with the hypoxia-driven one