Epigenetic Suppression of SERPINB1 Promotes Inflammation-Mediated Prostate Cancer Progression

Prostate cancer is the most common noncutaneous cancer in men in the United States, with more than 160,000 new cases diagnosed each year. Although most localized prostate cancers have an indolent course, it is essential to accurately assess which tumors will behave more aggressively (1). The development of biomarkers has revolutionized oncology; yet few useful prostate cancer biomarkers have emerged in recent years. Biomarkers under investigation in prostate cancer include TMPRSS2:ERG rearrangement, PC antigen 3 (PCA3), and the androgen receptor splice variant-7 (ARv7; ref. 2). Moreover, several genomic, proteomic, and epigenetic tests such as the OncotypeDx, ProMark, and ConfirmMDx evaluate a panel of biomarkers and attempt to stratify patients by risk (2). At this time however, clinical stage, Gleason grade, and serum PSA remains the primary prognostic indicators used by clinicians to guide therapeutic decisions (2).

Growing evidence suggests that the tumor microenvironment contributes to cancer development and progression. Thus, understanding the complex interactions between tumor cells and surrounding reactive stroma may uncover novel biomarkers and therapeutic targets (3, 4). Although prostate cancer is defined as an adenocarcinoma, tumors in fact have a significant nonmalignant cellular compartment consisting of fibroblasts, smooth muscle cells, endothelium, and immune cells (3, 4). The tumor microenvironment is also rich in noncellular components such as extracellular matrix (ECM), cytokines, growth factors, and proteases, which form fertile ground for cancer cell growth, migration, and invasion (5). The cellular and noncellular components evolve over the natural course of cancer. Specifically, tumor and stromal cells undergo genotypic and phenotypic changes, inflammatory cells infiltrate tumors, and the ECM becomes remodeled (3, 4). Despite progress in other solid tumors, much remains to be learned about microenvironment changes in prostate cancer.

We recently showed that granulocytic myeloid cells [likely myeloid-derived suppressor cells (MDSCs)] accumulate peripherally in blood and locally in tumors using a xenograft mouse model of human prostate cancer. These granulocytic myeloid cells directly promote tumor growth in the absence of adaptive immunity, because tumor size is significantly reduced upon sustained myeloid cell depletion in athymic mice (6). Furthermore, we demonstrated the importance of myeloid-derived neutrophil elastase (NE) as a protumorigenic factor for human prostate cancer cells, as NE promoted human prostate cancer cell line growth, migration, and invasion, whereas NE inhibition with the drug sivelestat attenuated human xenograft progression (6).

Although accumulation of NE-producing myeloid cells may be sufficient to explain enhanced NE activity within the tumor microenvironment, loss of endogenous NE inhibitors in tumor cells may further shift the balance (5). There are several endogenous NE inhibitors that are expressed by numerous tissue types, including elafin (PI3), secretory leukocyte peptidase inhibitor (SLPI), and SERPINB1 (7–10). We chose to focus our investigation on the most recently identified NE-specific inhibitor, SERPINB1. SERPINB1 is a 42-kD member of the Clade B family of serpins (serine protease inhibitors) that potently inhibits NE and neutrophil extracellular trap (NET) formation similar in mechanism to sivelestat (10–12). SERPINB1 regulates neutrophil homeostasis within the bone marrow and promotes neutrophil survival via intracellular inhibition of neutrophil serine proteases (13–15). Interestingly, SERPINB1 is expressed by many cell types beyond immune cells and detected both intra- and extracellularly (12, 16, 17). Although SERPINB1 lacks a classical signal peptide, it is secreted via an unconventional caspase-1–dependent secretory pathway that also mediates IL1β and IL18 secretion (17, 18). The role of SERPINB1 in epithelial cells and cancer is not clear; however, limited screening studies demonstrate reduced expression in hepatocellular carcinoma (HCC), glioma, and melanoma compared with normal tissue counterparts, as well as possible roles in suppressing apoptosis and promoting tumor migration (19–21). Microarray analysis of laser captured micro-dissected human prostate cancer specimens suggest that SERPINB1 mRNA expression is reduced early in the development of prostatic intraepithelial neoplasia (PIN) and remains low in prostate cancer (22). 2D DIGE/MS proteomic analysis comparing prostate cancer to benign prostatic hyperplasia (BPH) also suggests lower SERPINB1 expression in tumors (23). Finally, meta-analysis of epithelial-to-mesenchymal transition (EMT) encompassing several cancers, including prostate, identifies SERPINB1 as a commonly downregulated gene (24). However, all of the aforementioned studies simply classify SERPINB1 as a gene on a long list of other dysregulated genes, and neither a functional role for SERPINB1 loss nor mechanisms for its repression have been elucidated in prostate cancer.

Here we report that SERPINB1 expression is reduced in mouse Pten-null prostate tumors and treatment with the NE inhibitor sivelestat (SERPINB1 pharmacomimetic) attenuates tumor growth. We confirm that SERPINB1 is expressed in normal human prostatic epithelium but downregulated in prostate cancer. We demonstrate that SERPINB1 is secreted by nonmalignant human prostate cells and its expression is ERK1/2 dependent. Functionally, we show that SERPINB1 loss in nonmalignant human prostate cells induces EMT and promotes a proliferative phenotype. In contrast, SERPINB1 overexpression in human prostate cancer cells suppresses xenograft progression. Finally, we find that SERPINB1 is epigenetically repressed in human prostate cancer cells by EZH2-mediated histone methylation (H3K27me) and DNMT-mediated DNA methylation. Overall, our data suggest a novel inhibitory role for SERPINB1 in prostate cancer progression, with potential application to both biomarker and therapeutic development.

RWPE-1 cells (ATCC) were cultured in keratinocyte serum-free media (K-SFM; Gibco) supplemented with 25 μg/mL bovine pituitary extract (BPE; Gibco), 5 ng/mL EGF (Gibco), and 1% penicillin–streptomycin (P–S; Gibco). C4-2 (Ganesh Raj, The University of Texas Southwestern, Dallas, TX), PC3 (ATCC), and LNCaP (ATCC) cells were cultured in RPMI1640 media (Gibco) supplemented with 10% FBS (Seradigm) and 1% P–S. BPH-1 cells (Donald DeFranco, University of Pittsburgh, Pittsburgh, PA) were cultured in 5% FBS and 1% P–S in RPMI1640. VCaP, DU145, and CWR22Rv1 cells (Kent Nastiuk, University at Buffalo, Buffalo, NY) were cultured in 10% FBS and 1% P–S in DMEM (Gibco). LAPC-4 cells (Kent Nastiuk) were cultured in 10% FBS and 1% P–S in RPMI1640, 10 mmol/L HEPES pH 7.4 (Gibco). Pten-Cap8 cells (Kent Nastiuk) were cultured in 10% FBS, 25 μg/mL BPE, 6 ng/mL EGF, 5 μg/mL human recombinant insulin (Sigma), and 1% P–S in DMEM. All cells were maintained below passage 30 at 37°C, 95% air, 5% CO₂.

Cells were serum starved for 24 hours and stimulated with 20 ng/mL EGF (Corning) in SFM for 4 or 24 hours before subsequent lysis and analysis by quantitative PCR or Western blot analysis, respectively. For inhibitor studies, cells were pre-incubated with 0.25 μmol/L MEK inhibitor PD0325901 (Selleckchem) or 0.25 μmol/L PI3K inhibitor LY294002 (Sigma) for 30 minutes before the addition of EGF. Cells were treated with 10 μmol/L DNMT inhibitor 5-aza-2-deoxycytidine (5-AZA; Sigma) and 10 μmol/L EZH2 inhibitor GSK343 (Selleckchem) in complete media for 72 hours.

RWPE-1 cells were transfected with 25 nmol/L Silencer Select Negative Control #1 siRNA (#4390843; Ambion) or 25 nmol/L Silencer Select SERPINB1 siRNA (#4392420; Ambion) using jetPRIME (Polyplus). Knockdown was carried out for 72 to 96 hours as specified.

Human SERPINB1 was cloned via nested PCR. SERPINB1 cDNA was inserted into pcDNA3.1/Hygro(+) vector (Invitrogen). C4-2 cells were transfected with SERPINB1-pcDNA3.1/Hygro(+) or empty vector using X-tremeGENE 9 DNA Transfection Reagent (Sigma) and selected with 100 μg/mL hygromycin (Sigma). Monoclonal cell lines were established and maintained in 10% FBS and 1% P–S in RPMI1640 containing 25 μg/mL hygromycin.

RWPE-1 and C4-2 cells were grown to 80% confluence in respective complete medias and then cultured without supplements for an additional 48 hours. Conditioned media were collected, centrifuged at 2000 rpm for 10 minutes at 4°C, and concentrated five-fold using Amicon Ultra 0.5 mL 3 kDa centrifugal filter units (Millipore).

Conditioned media samples were diluted 1:1 with 2× nonreducing, nondenaturing sample buffer and separated on a 10% gel containing 1 mg/mL gelatin (Sigma). Zymograms were performed and analyzed as previously described (25, 26).

Cells were lysed in mammalian cell lysis buffer (Abcam) supplemented with 1× Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Lysates or conditioned media were diluted 1:1 with 2× sample buffer containing 2-mercaptoethanol (Sigma) and denatured. Samples were processed for gel electrophoresis and blotted with mouse anti-SERPINB1 (1:2,000, #TA800093; Origene), rabbit anti-GAPDH (1:2,000, #2118; Cell Signaling Technology), rabbit anti-phospho-ERK1/2 (1:1,000, #9101; Cell Signaling Technology), and rabbit anti-total-ERK1/2 (1:1,000, #9102; Cell Signaling Technology).

RNA was extracted using the E.Z.N.A. Kit (Omega). Quantitative PCR (qPCR) was performed using the qScript XLT 1-Step RT-qPCR ToughMix Kit (QuantaBio) and TaqMan primers (Applied Biosystems) for: human SERPINB1 (Hs00961948_m1), MMP9 (Hs00234579_m1), SNAI1 (Hs00195591_m1), TWIST1 (Hs01675818_s1), PI3 (Hs00160066_m1), SLPI (Hs00268204_m1), GAPDH (Hs03929097_g1). Expression was normalized to GAPDH using the ΔΔCt method.

Chromatin immunoprecipitation (ChIP) for SERPINB1 was performed using the MAGnify Chromatin Immunoprecipitation System (Invitrogen; ref. 27). Briefly, chromatin fragments were immunoprecipitated with mouse monoclonal anti-H3K27me3 ChIP-grade antibody (#6002; Abcam) or mouse IgG. Quantitative PCR was performed using PerfeCTa SYBR Green SuperMix Reagent (QuantaBio) with primers directed against two regions of the human SERPINB1 promoter (P1-forward-5′ CGT GCG ATT CTA GAG ACG ATT T 3′, P1-reverse-5′ CGA GGA CAG GCA AAG AAG AA 3′; P2-primer-5′ TCT GAG AGT GGA GAT CGA GAT G 3′, P2-primer-5′ GGT GTA GGA TGT GCC AGT TT 3′). Results were normalized by the fold enrichment method.

RWPE-1 cells were transfected with control and SERPINB1 siRNA for 72 hours. Cell pellets were collected with 0.05% trypsin-EDTA (Gibco), neutralized with 2% FBS in PBS, and washed with cold PBS. Apoptosis was assessed using the FITC Annexin V Apoptosis Detection Kit I (BD Pharmingen). Briefly, cells were resuspended in 1× Annexin Binding Buffer at 1 × 10⁶ cells/mL, stained with PI and FITC Annexin V, collected on a LSRII flow cytometer (BD Biosciences), and analyzed with FCS Express 6 software.

Cells were treated with 10 μmol/L 5-aza-2′-deoxycytidine for 72 hours and harvested for genomic DNA isolation with Blood and Cell Culture DNA Mini Kit (Qiagen). Two micrograms of genomic DNA from each sample was used for bisulfite conversion with EpiTect Fast DNA Bisulfite Kit (Qiagen). PyroMark PCR Kit (Qiagen) was used for subsequent PCR reactions. Three pairs of primers amplifying three regions, each containing 2 CpG sites within 1500bp upstream of TSS, were designed with Qiagen PyroMark Assay Design Software 2.0. PyroMark Q24 sequencer (Qiagen) was used to analyze amplified PCR products, and methylation of CpG sites was determined with PyroMark Q24 Advanced software.

RWPE-1 cells were transfected with control and SERPINB1 siRNA for 72 hours. Proliferation was assessed using the BrdUrd Cell Proliferation Assay Kit (Cell Signaling Technology; ref. 6).

Experiments were performed in accordance with the guidelines for the Care and Use of Laboratory Animals and approved by the University Committee on Animal Resources at the University of Rochester. Six- to eight-week-old male athymic nude mice were subcutaneously injected with 2 × 10⁶ C4-2 cells (expressing SERPINB1 or vector) in 0.1 mL of a 1:1 mixture of Matrigel (Corning) and PBS. Tumor growth was monitored over 12 weeks. Once largest tumors reached end point (∼2,000 mm³), all tumors were harvested. Work with prostate-specific PbCre4/Pten^fl/fl mice in C57/BL6 background (28) was approved by the Roswell Park IACUC. Tumor volume was monitored using 3D ultrasound (29). Mice bearing tumors of 300 to 500 mm³ were blindly randomized to vehicle (PBS) or sivelestat (Tocris; 5 mg/kg in PBS) treatment daily via intraperitoneal injection. Tumor volume was monitored weekly via ultrasound.

Blood was collected from retro-orbital sinuses at indicated times and processed for CD11b, Ly6G, and Ly6C staining as previously described (6).

Sections were de-paraffinized and rehydrated (6, 25). Heat-mediated antigen retrieval was performed in 0.01 M Citrate pH 6 at 95°C. For human tissues, mouse antihuman SERPINB1 (#TA800093; Origene) was diluted 1:200 in antibody diluent (Thermo Scientific) and incubated overnight at 4°C. Biotinylated horse anti-mouse IgG (#BA-2000; Vector Laboratories) was diluted 1:200 in blocking serum (2.5% normal horse serum in PBS; Vector Laboratories). For mouse tissues, rabbit anti-mouse SERPINB1 (#TA340203; Origene) was diluted 1:200 in antibody diluent and incubated overnight at 4 °C. Biotinylated goat anti-rabbit IgG (#BA-1000; Vector Laboratories) was diluted 1:200 in blocking serum (2.5% normal goat serum in PBS; Vector Laboratories). Immunoreactivity was detected using the Vectastain Elite ABC and DAB peroxidase substrate kits (Vector Laboratories). Sections were counterstained with hematoxylin and mounted using Cytoseal 60 (Thermo Fisher Scientific).

Primary antibodies were: biotin rat anti-mouse Ly6G (1:50, #127604; BioLegend) and goat anti-mouse proliferating cell nuclear antigen (PCNA; 1:50, #sc-9857; Santa Cruz Biotechnology). Primary antibodies were detected using streptavidin-Alexa Fluor 488 (1:200, #S11223; Thermo Fisher Scientific), donkey anti-rat Alexa Fluor 488 (1:200, #A21208; Thermo Fisher Scientific), and donkey anti-goat Alexa Fluor 568 (1:200, #A11057; Thermo Fisher Scientific). Sections were counterstained and mounted using VECTASHIELD antifade mounting media with DAPI (Vector Laboratories).

Mice received 4 nmol of Neutrophil Elastase 680 FAST probe (PerkinElmer) in 0.1 mL PBS via tail-vein injection. Activity was measured on excised tumors using fluorescent microscopy (6, 25) and intensity was analyzed using ImageJ v1.48 software.

Microarray datasets of prostate adenocarcinoma (PRAD) and normal tissue were accessed through the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) and queried for SERPINB1 expression. Chandran and colleagues (GEO accession number: GSE6919; ref. 30), Varambally and colleagues (GEO accession number: GSE3325; ref. 31), Arredouani and colleagues (GEO accession number: GSE55945; ref. 32), and Taylor and colleagues (GEO accession number: GSE21036; ref. 33) expression datasets were analyzed. Kaplan–Meier analysis for recurrence-free survival was performed using the Taylor et al dataset through the open web interface Project Betastasis (http://www.betastasis.com). Analysis for DNA methylation at the SERPINB1 promoter and SERPINB1 expression in cancer and normal tissue was performed using The Cancer Genome Atlas (TCGA) dataset through MethHC (http://methhc.mbc.nctu.edu.tw/php/index.php; ref. 34).

Data are presented as mean ± SEM. Comparison between two groups was performed using two-tailed t test, unless otherwise indicated. Comparison between more than two groups was performed using one-way ANOVA with appropriate post hoc testing. Statistical analyses were performed using GraphPad Prism 7.0 software, and significance defined as P < 0.05.

We recently demonstrated that NE inhibition with sivelestat reduces human prostate cancer xenograft growth in athymic mice. Furthermore, as in our xenograft models, we showed that granulocytic myeloid cells (likely MDSCs) are elevated in peripheral blood and infiltrate prostate tumors of immunocompetent probasin-driven Pten-null mice. Accordingly, Pten-null prostate tumors exhibit enhanced NE activity (6).

To determine whether NE is a protumorigenic factor in Pten-null mice, we treated mice with sivelestat or vehicle and monitored tumor growth over eight weeks (Fig. 1A). Moreover, we assessed dynamics of peripheral blood MDSC accumulation via flow cytometry at 0, 4, and 8 weeks of treatment (Fig. 1A). We examined prostates by ultrasound to measure tumor volumes and initiated treatment at ∼400 mm³. Tumors were smaller in sivelestat versus vehicle treated mice throughout the study, reaching statistical significance by the final two time points (Fig. 1B), suggesting that NE indeed promotes tumor growth in a prostate cancer mouse model with an intact immune system. Interestingly, we observed time dependent accumulation of MDSCs (CD11b+/Ly6G+/Ly6C+) in the blood of both sivelestat and vehicle-treated animals (Fig. 1C), confirming that NE inhibition does not mitigate tumor induced MDSC production. Examination of MDSC infiltration within tumors via Ly6G immunofluorescence revealed that, whereas there was no difference between sivelestat and vehicle-treated tumors (not shown), there was a strong positive correlation between infiltrating MDSCs and expression of proliferating cell nuclear antigen (PCNA) in prostate cells (Fig. 1D and E), supporting the hypothesis that MDSCs contribute to tumor proliferation.

Because sivelestat is a pharmacomimetic of the endogenous NE inhibitor SERPINB1 (35), we next examined SERPINB1 expression via IHC in untreated Pten-null tumors and wild-type mouse prostates. SERPINB1 strongly stained glandular epithelium of normal prostates but was reduced in the overgrown epithelium of Pten-null tumors (Fig. 1F; Supplementary Fig. S1). We confirmed by Western blot analysis that SERPINB1 was reduced in Pten-null tumors and the Pten-null prostate cancer cell line Pten-CaP8 compared with wild-type prostate (Fig. 1G), suggesting that epithelial downregulation of SERPINB1, along with increased MDSC infiltration, may contribute to the observed enhanced NE activity in Pten-null prostate tumors (6).

To determine whether SERPINB1 downregulation occurs in human prostate cancer, we queried public datasets of PRAD and normal tissue. Interestingly, SERPINB1 mRNA expression was reduced in prostate cancer compared with normal tissue in all four datasets, although the extent and pattern of reduction varied (Fig. 2A–D). Furthermore, SERPINB1 mRNA expression was lowest in metastatic prostate cancer (Fig. 2A, B, and D). Accordingly, using the Taylor and colleagues dataset that contains clinical patient parameters, we found that low SERPINB1 mRNA expression significantly correlated with diminished recurrence free survival using both 50% and 25% cut-off thresholds (Fig. 2E).

To begin elucidating the functional significance of SERPINB1, we performed Western blots for SERPINB1 in human nonmalignant and malignant cell lines. The most commonly used nonmalignant prostatic epithelial cell line RWPE-1 exhibited strong SERPINB1 expression (Fig. 2F). The BPH cell line BPH-1 expressed lesser amount of SERPINB1. Of the seven prostate cancer cell lines tested, four did not express detectable SERPINB1 (LNCaP, C4-2, VCaP, and LAPC-4), two expressed low levels of SERPINB1 (DU-145 and PC3), and one expressed equal amount of SERPINB1 (22Rv1), relative to RWPE-1.

Because nonmalignant RWPE-1 cells express abundant SERPINB1, we knocked down both cellular and secreted SERPINB1 expression by siRNA to determine how gene expression and physiology would be altered (Fig. 3A). Similar to gene arrays in other cancers (19, 20, 24), we found that mRNA expression of EMT-related genes MMP9, TWIST1, and SNAI1 were induced by SERPINB1 knockdown (Fig. 3B). Gelatin zymography on conditioned media collected from SERPINB1 and NSP siRNA-treated RWPE-1 cells confirmed increased expression of pro-MMP9, but not pro-MMP2, activity (Fig. 3C, quantified in Fig. 3D). Because SERPINB1 has also been implicated in cell viability (12), we assessed effects of SERPINB1 loss on apoptosis via flow cytometry for PI and Annexin V. Apoptotic cells (Annexin V+/PI−) were significantly reduced by ∼36% in response to SERPINB1 knockdown (Fig. 3E, quantified in Fig. 3F). Furthermore, proliferation by BrdU was increased by ∼39% with SERPINB1 knockdown (Fig. 3G). Together, these data suggest that SERPINB1 loss contributes to EMT, enhanced proliferation, and reduced apoptosis in normal prostatic epithelial cells and may consequently drive malignant transformation.

To determine how SERPINB1 is repressed in prostate cancer, we first examined regulation of SERPINB1 expression in nonmalignant prostatic epithelial cells. We focused on EGF-induced pathways, hypothesizing that enhanced growth factor signaling as seen in cancers might suppress SERPINB1 expression. Unexpectedly, EGF significantly induced SERPINB1 as measured by qPCR and Western blot analysis in RWPE-1 cells (Fig. 4A and B). EGF induced expression of mRNAs encoding other endogenous NE inhibitors, PI3 (elafin) and SLPI (secretory leukocyte protease inhibitor), though less dramatically (Fig. 4A). In contrast, EGF did not increase SERPINB1 mRNA or SERPINB1 protein expression in PC3 (Fig. 4C and D) or DU145 prostate cancer cells (Supplementary Fig. S2A and S2B), which express low but detectable SERPINB1 at baseline (Fig. 2F). Moreover, EGF was unable to induce SERPINB1 mRNA in LNCaP prostate cancer cells that do not express endogenous SERPINB1 (Supplementary Fig. S2C). In RWPE-1 cells, MEK inhibition by PD0325901 abrogated EGF-induced SERPINB1 transcription, whereas the PI3K inhibitor had no effect (Fig. 4E), with similar but less dramatic results seen when examining PI3 and SLPI transcription (Fig. 4E). Thus, SERPINB1 expression in nonmalignant RWPE-1 cells may be regulated in part by ERK1/2 signaling.

If EGF was not promoting SERPINB1 expression in tumor cells, then we hypothesized that some signal was preventing the EGF effects. Tumor suppressors are frequently turned off via epigenetic alterations such as histone and DNA methylation in cancers. Because EZH2 is upregulated in and considered a driver of prostate cancer, we performed ChIP for histone 3 lysine 27 tri-methylation (H3K27me3) on the SERPINB1 promoter in nonmalignant RWPE-1 versus castration resistant prostate cancer C4-2 cells. We observed marked enrichment of H3K27me3 in C4-2 compared with RWPE-1 cells (Fig. 5A). Seventy-two-hour treatment of C4-2 cells with an EZH2 inhibitor (GSK343) modestly induced expression of SERPINB1 mRNA, with minimal induction of PI3 and SLPI mRNA (Fig. 5B). Because this effect was relatively small, we examined the effects of DNA methyltransferase (DNMT) inhibition on SERPINB1 expression. Seventy-two-hour treatment of C4-2 cells with 5-AZA enhanced SERPINB1 mRNA expression by over 2,000-fold, comparable to transcript levels detected in RWPE-1 cells (Fig. 5C). Moreover, 5-AZA, but not GSK343, enhanced SERPINB1 protein expression in C4-2 cells (Fig. 5D). Similar effects on mRNA levels were observed in LNCaP prostate cancer cells (Supplementary Fig. S3A and S3B).

Accordingly, pyrosequencing analysis on the SERPINB1 promoters revealed that CpG1, CpG2, and CpG3, the islands closest to the transcription start site, had markedly elevated DNA methylation in C4-2 cells compared with nonmalignant RWPE-1 cells (Fig. 5E). 22Rv1 cancer cells that expressed comparable level of SERPINB1 as RWPE-1 also exhibited a similar pattern of CpG methylation as RWPE-1 (Fig. 5E). Conversely, LNCaP cancer cells that lacked endogenous SERPINB1 had a methylation pattern that closely resembled C4-2 cells (Fig. 5E). Furthermore, in assessing DNA methylation at the SERPINB1 promoter in various cancers using TCGA data available through the open web interface MethHC, only PRAD exhibited increased methylation relative to adjacent normal tissue (Fig. 5G). Similar to individual prostate cancer expression datasets (Fig. 2A–D), SERPINB1 mRNA expression was significantly reduced in tumor compared with normal samples using the PRAD TCGA (Fig. 5F), with SERPINB1 promoter methylation negatively correlating with SERPINB1 mRNA expression in tumors (Fig. 5H). These data suggest that DNA methylation may silence SERPINB1 expression specifically in human prostate cancer versus other cancers.

We next determined whether rescuing SERPINB1 expression in normally low-expressing prostate cancer cells would diminish tumor burden. We therefore generated two stable monoclonal cell lines overexpressing SERPINB1 or vector control in C4-2 cells (Fig. 6A). Notably, these cells expressed and secreted SERPINB1 just like RWPE-1 (Fig. 6A). Xenografts expressing SERPINB1 were markedly smaller than vector control xenografts after 12 weeks of growth in athymic nude mice (Fig. 6B and D). Stable SERPINB1 overexpression was verified in tumors of both clonal cell lines collected at the end of the experiment (Fig. 6C and E). Using a NE activity probe (6), we observed reduced NE activity in SERPINB1 overexpressing tumors compared with vector controls (Fig. 6F and G, quantified), suggesting that SERPINB1 inhibits local NE activity to attenuate prostate cancer xenograft growth.

Finally, we confirmed loss of SERPINB1 expression in prostate cancer by performing IHC on prostatectomy and core biopsy specimens that contain regions of both normal and malignant tissue. As expected, SERPINB1 was localized to cells lining normal prostate glands and exhibited diffuse stromal staining, both of which were largely absent in cancer (Fig. 7). Scattered SERPINB1 positive cells present in cancer regions were likely infiltrating immune cells expressing SERPINB1. Therefore, epithelial and/or stromal downregulation of SERPINB1 may serve as a novel biomarker for prostate cancer progression.

Granulocytic myeloid cells or MDSCs are increased in the circulation, peripheral organs, and primary tumors in mouse models of numerous malignancies, including prostate cancer (36, 37). MDSCs contribute to cancer progression and metastasis indirectly via T-cell immunosuppression and directly via release of proinflammatory cytokines, growth factors, and proteases such as matrix metalloproteinases (MMP) and NE. Consequently, targeting MDSC production and recruitment leads to decelerated tumor growth and reduced metastases (5, 36). We reported that myeloid-derived NE directly promotes human prostate cancer growth in athymic mice, and that sivelestat (NE inhibitor) reduces tumor burden to the same extent as anti-Gr-1 MDSC depletion (6). We now demonstrate that targeting NE activity with sivelestat in the immunocompetent Pten-null mouse model of prostate cancer similarly attenuates tumor progression without altering levels of circulating or infiltrating MDSCs. Therefore, tumors continue to recruit protumorigenic MDSCs despite reduced NE activity, which likely explains why the tumor inhibitory effect of sivelestat is somewhat diminished by the later time points when the MDSC burden is too great. MDSC depletion via antibodies against CXCL5 and Gr-1 or inhibitors targeting CXCR2 similarly result in modestly decelerated prostate tumor growth in Pten-null mouse models (38–40). Unfortunately, MDSC levels rebound after prolonged antibody-mediated depletion, thus the translational potential of such therapy is uncertain (36). Therefore, we postulate that perhaps dual targeting of MDSCs and NE activity will lead to synergistic growth inhibitory effects.

Local MDSC recruitment likely drives the preponderance of immune-derived proteases in prostate and other cancers, and in fact we find that, in Pten-null mice, prostate cancer cell proliferation directly correlates with the density of surrounding MDSCs. Alternatively, although inflammatory protease levels are increasing in the tumor microenvironment, expression of endogenous protease inhibitors by epithelial cells is often suppressed, favoring enhanced proteolytic activity and resultant cancer cell proliferation, migration, and invasion (41). For instance, loss of Timp3, an endogenous MMP inhibitor, accelerates tumor growth and invasion of Pten-null prostate tumors via upregulated MMP activity (42). Here, we are first to demonstrate that SERPINB1, an endogenous NE inhibitor, is reduced in Pten-null mouse tumors, which might further explain their heightened intratumoral NE activity (6).

SERPINB1 is a potent regulator of neutrophil serine proteases and NET formation (13). Consequently, SERPINB1 contributes to the resolution of both acute infections and chronic inflammatory diseases, similar to the therapeutic effects of its pharmacomimetic sivelestat (12). SERPINB1 is widely expressed in various cell types, and several studies suggest a role in cancer progression separate from its NE inhibitory function. In HCC, glioma, and melanoma, SERPINB1 downregulation in tumors is associated with poor patient prognosis (19–21). Intriguingly, two exploratory studies identify SERPINB1 as a downregulated gene, among many others, in prostate cancer compared with nonmalignant prostate (22, 23). We now demonstrate that SERPINB1 is in fact reduced in human prostate cancer compared with normal prostatic epithelium using a variety of transcriptional and translational approaches. SERPINB1 is further reduced in metastatic disease, and low expression predicts diminished recurrence-free survival in patients. IHC staining reveals SERPINB1 is strongly expressed by the basal layer in normal prostate glands, which is largely absent in regions of cancer. Interestingly, there is strong evidence for the basal progenitor as the cell type of origin in both mouse models and human prostate cancer (43, 44). Moreover, loss of tumor suppressor Pten in basal cells promotes basal-to-luminal differentiation and development of invasive prostate cancer in mice (45). Accordingly, we observe decreased SERPINB1 expression in mouse Pten-null prostates and a Pten-null prostate cell line. It is therefore possible that SERPINB1 loss also plays a critical role during prostate tumorigenesis, although this remains to be carefully dissected.

The functional role of SERPINB1 in cancer at this time is unclear, although limited studies suggest an involvement in regulation of cell motility. For instance, SERPINB1 reduction in HCC and glioma cells enhances migration and invasion in vitro, potentially through increased MMP-2 expression (19, 20). Conversely, SERPINB1 overexpression in lung and breast cancer cells decreases migratory and invasive capacity (46). Here we demonstrate that SERPINB1 loss in nonmalignant prostatic epithelial cells induces expression of EMT markers, including MMP-9, TWIST1, and SNAI1. EMT is the process by which polarized epithelial cells acquire a mesenchymal, promigratory, and proinvasive phenotype, and activation of an EMT program is considered a critical step in malignant transformation (47, 48). Therefore, as mentioned above, loss of SERPINB1 may be an early event during tumorigenesis. Accordingly, a meta-analysis of 18 independent gene expression datasets of EMT identified SERPINB1 as a commonly downregulated gene in various cancers (24).

In addition to EMT, two hallmarks of cancer are the ability of tumor cells to resist cell death and sustain proliferative signaling (48). Here, we report that SERPINB1 loss impedes apoptosis and stimulates proliferation in normal prostatic epithelial cells. SERPINB1 is proposed to control cell viability via interaction with several proteins involved in apoptosis, including PARP-1, BCL-2, and apoptosis inducing factor (AIF), among others (12). Moreover, SERPINB1 loss induces proliferation in keratinocytes by promoting G₁–S cell-cycle transition (49). Nonetheless, the effect of ectopic SERPINB1 overexpression on tumor growth in vivo has not been previously examined. Here, we demonstrate that stable ectopic expression of SERPINB1 in prostate cancer cells with low endogenous SERPINB1 expression reduces xenograft growth and NE activity in vivo. We also report SERPINB1 secretion by both normal prostatic epithelial cells and ectopically expressing SERPINB1 prostate cancer cells. Together, these data suggest that extracellular SERPINB1 may act as a functional NE inhibitor in vivo. Consequently, downregulation of SERPINB1 by cancer cells may permit NE activation within the tumor microenvironment to enhance tumor progression.

Still, how SERPINB1 is regulated in normal tissues and cancer is not well understood. In this study, we demonstrate that SERPINB1 is strongly induced by EGF signaling through the ERK1/2 pathway in nonmalignant but not malignant prostatic epithelial cells. Furthermore, we demonstrate a novel mechanism of SERPINB1 repression in prostate cancer cells via EZH2-mediated histone methylation (H3K27me) and DNA methyltransferase (DNMT)-mediated DNA methylation. Although EZH2 inhibition modestly induces SERPINB1 expression, DNMT inhibition induces SERPINB1 expression by over 2,000-fold and rescues protein expression from previously undetectable levels. EZH2 may facilitate gene silencing directly through histone methylation and indirectly by serving as a platform for DNMT complex recruitment (50). Thus, the two epigenetic repression systems may be functionally linked, and both are likely responsible for SERPINB1 silencing during the complex process of tumor initiation and progression. Accordingly, EZH2 and various DNMTs (i.e., DNMT1, DNMT3A, DNMT3B) are overexpressed in mouse and human prostate cancers (51, 52).

Remarkably, SERPINB1 promoter methylation appears to be specific to prostate cancer, as no other human cancers examined through the TCGA exhibit such a stark difference compared with their normal tissue counterparts. Not surprisingly, we observe that promoter methylation and SERPINB1 expression are tightly correlated in prostate cancer. Therefore, promoter methylation may serve as a surrogate marker of SERPINB1 expression and have diagnostic and prognostic potential. Furthermore, SERPINB1 downregulation, combined with elevated markers of inflammation such as neutrophil-to-lymphocyte ratio (NLR), may identify patients with prostate cancer who will best respond to MDSC or NE-targeting therapy. Better mechanistic understanding of the methyltransferases responsible for SERPINB1 repression will be advantageous for future drug and diagnostic test development. Analogously, GSTP1 is a tumor suppressor silenced via promoter methylation in prostate cancer, and this epigenetic alteration may be evaluated clinically through the ConfirmMDx assay (53).

In summary, our findings demonstrate that SERPINB1 is a novel tumor suppressor that is epigenetically silenced in prostate cancer. While SERPINB1 loss may permit NE-driven prostate cancer progression, it also exhibits significant NE-independent tumor promoting effects. Importantly, SERPINB1 expression and promoter methylation status may serve as a potential biomarker to guide therapeutic decisions.

No potential conflicts of interest were disclosed.

Conception and design: I. Lerman, K.L. Nastiuk, S.R. Hammes

Development of methodology: I. Lerman, M. de la Luz Garcia-Hernandez, K.L. Nastiuk, M. Susiarjo, S.R. Hammes

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Lerman, X. Ma, C. Seger, A. Maolake, M. de la Luz Garcia-Hernandez, J. Rangel-Moreno, J. Ackerman, K.L. Nastiuk, M. Susiarjo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Lerman, X. Ma, A. Maolake, J. Rangel-Moreno, J. Ackerman, K.L. Nastiuk, M. Susiarjo, S.R. Hammes

Writing, review, and/or revision of the manuscript: I. Lerman, X. Ma, M. de la Luz Garcia-Hernandez, J. Rangel-Moreno, J. Ackerman, K.L. Nastiuk, S.R. Hammes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Rangel-Moreno, J. Ackerman, M. Susiarjo, S.R. Hammes

Study supervision: M. Susiarjo, S.R. Hammes

S.R. Hammes was supported by the NIH grants R01GM101709 and R01CA193583-01A1. I. Lerman was supported by NIH grant F30CA203517.

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.