Abstract
Chronic inflammation and African ancestry are implicated in prostate cancer aggressiveness, and inflammation-related genes are more highly expressed in prostate cancer in African American men. IL8 secretion is also implicated in prostate cancer progression and castration resistance. We used RNA in situ hybridization to localize IL1β, IL6, IL8, and IL10 mRNA in low- and high-grade prostate cancer from African American and European American men. IL8 was the most abundantly expressed and the only interleukin detected in tumor cells. We further interrogated IL8 expression in primary and metastatic prostate cancer tissue microarrays and both androgen-dependent and castration-resistant patient-derived xenografts (PDX). IL8 was significantly increased in both tumor and benign regions of higher grade cases (ISUP Grade Group 4–5), but there was no difference between races. We determined that IL8 expression in prostate cancer cell lines, distant metastases, and PDX lines was associated with androgen receptor (AR) loss, but not castration resistance. Reciprocal IL8 and AR expression was also observed in high IL8-expressing atrophy lesions with simultaneous AR downregulation. Finally, we show that IL8 is likely repressed by AR binding to the IL8 promoter and is inducible in prostate cancer cells stimulated with lipopolysaccharide only in cells with AR loss. Likewise, AR knockdown in androgen-dependent cells induced IL8 expression, further demonstrating that AR represses IL8 expression. In conclusion, IL8 expression in the tumor microenvironment is associated with aggressive prostate cancer and with AR loss in metastatic disease.
IL8 expression is repressed by AR and is associated with prostate cancer aggressiveness and AR loss in metastatic disease.
Introduction
Prostate cancer is the second leading cause of cancer-related deaths in American men, and African American (AA) men are two to three times more likely to die from prostate cancer than European American (EA) men (1, 2). Some studies indicate that disparities in access to quality care contribute to these differences in prostate cancer outcomes among races, whereas others identify genomic alterations and factors in the tumor microenvironment (TME), such as increased inflammation, in AA men that may promote prostate cancer aggressiveness (3–6).
Prostate cancer development may be mediated by environmental exposures that contribute to chronic inflammation—which is frequently observed in the adult prostate and is an enabling characteristic of cancer (7, 8). Inflammation-associated regions of prostate atrophy are frequently observed with aging, particularly in the peripheral zone where prostate cancer is prone to develop. These regions, referred to as proliferative inflammatory atrophy (PIA), contain atrophic epithelial cells with a markedly increased proliferative fraction compared with normal-appearing epithelium. Neoplastic transformation may commence in the atrophic cells in PIA (reviewed in ref. 7). Furthermore, PIA lesions are commonly observed in direct transition with both prostatic intraepithelial neoplasia (PIN) and, at times, adenocarcinoma and in a subset of cases share at least one somatic DNA alteration, hypermethylation of the CpG island in the GSTP1 gene (8). As such, chronic inflammation might contribute to prostate carcinogenesis; however, the molecular mechanism of this association is not yet well defined (9).
Inflammatory cytokines might mediate inflammation-associated prostate carcinogenesis, and SNPs in cytokine genes are associated with prostate cancer risk and aggressiveness (10, 11). African ancestry is associated with prostate cancer aggressiveness, and altered expression of genes involved in inflammatory pathways is consistently more prevalent in tumors from AA men versus EA men (1–6, 11, 12). Specifically, IL1β, IL6, IL8, and IL10 are reportedly overexpressed in the TME of AA men (6, 13, 14). However, there is little known about the cellular localization of interleukins in the prostate microenvironment.
IL8 is a proinflammatory chemokine produced by monocytes, neutrophils, endothelial, and epithelial cells, and is expressed in cancer cells (15). IL8 may be regulated by inflammatory signals such as IL1β, chemical and environmental stress, and steroid hormones such as androgens (16, 17). IL8 serves as a chemoattractant and activator for neutrophils and may promote cell proliferation, cell invasion, cell survival, and chemoresistance (16). IL8 is elevated in serum of men with prostate cancer and is associated with poorer outcomes (18, 19). Further, IL8 protein expression measured in the stroma surrounding prostate tumors is reportedly elevated compared with stroma surrounding normal-appearing prostate epithelium (20). The cellular localization of IL8 production and expression in tumor versus benign regions is not fully described.
Although IL1β, IL6, IL8, and IL10 have been assessed in serum or via meta-analysis of RNA, the spatial localization of these cytokines in prostate cancer tissues is largely unreported. Whether these cytokines are produced primarily by prostatic or immune cells is to be determined. Further, whether expression patterns within the prostate differ by cancer grade or between races is unknown. We aimed to address these specific questions using highly specific RNA in situ hybridization (RISH) assays and IHC to analyze formalin-fixed paraffin-embedded (FFPE) tissues of lower-grade (Gleason grades 6 and 7, ISUP Grade Group 1–3) and higher-grade (Gleason grade ≥8, ISUP Grade Group 4–5) prostate cancer from AA and EA men. We report differential expression of IL1β, IL6, IL8, and IL10 among men with prostate cancer and across regions within the prostate. IL8 expression was the most abundant cytokine assessed and, importantly, its expression was increased in both benign and tumor regions of cases with higher-grade tumors. There was no apparent difference in expression of IL8 between races. Our evaluation of IL8 expression in metastatic tissues and patient-derived xenografts (PDX) as well as a series of in vitro studies in prostate cancer cell lines suggests a reciprocal relationship between IL8 and androgen receptor (AR) expression, implying that regulation of this proinflammatory cytokine is related to the presence of the AR. Whereas inflammatory cytokines have previously been proposed to regulate the AR (21, 22), the converse has not yet been widely appreciated.
Materials and Methods
Patient population and clinical samples
All specimens were acquired using protocols approved by the Johns Hopkins University's Institutional Review Board. FFPE whole tissue slides (e.g., standard tissue slides) from tumor-containing and matched benign region blocks from previously untreated patients with clinically localized prostate cancer were obtained from radical prostatectomy specimens from 19 self-identified AA patients and 19 self-identified EA patients (n = 38 patients with whole tissue slides) with lower grade (Gleason Grade 6 and 7, ISUP Grade Group 1–3, n = 22) and higher-grade (Gleason grade ≥ 8, ISUP Grade Group 4–5, n = 16) prostate cancer. One block containing the highest-grade index tumor and adjacent benign was chosen for RISH analysis from each prostatectomy case. Tissue microarray (TMA) sets were constructed from (i) lower grade and higher grade primary prostate cancer tissues obtained from prostatectomy specimens from 60 AA and 60 EA men matched for patient age (±3 years), tumor grade, and stage and contained four tissue spots of the index tumor from each case as well as four benign tissue spots taken from a block containing only benign tissue, with deliberate avoidance of areas with overt inflammation, and (ii) 26 autopsy metastatic tissue samples from 2 AA and 4 EA men. TMAs were used in RISH and IHC experiments. Further details of the patients and TMAs are provided in Supplementary Table S1.
PDX TMA
A TMA containing 42 of the LuCaP PDX lines (ref. 23; 3 cores from 3 xenografts from each line) was obtained from the Prostate Cancer Biorepository Network (PCBN; http://prostatebiorepository.org).
Cell lines, lipopolysaccharide treatment, and short hairpin RNA knockdown
LNCaP, CWR22Rv1, VCaP, and HEK293T cells were obtained from the American Type Culture Collection. PC3, DU145, MCF7, C4-2b, RWPE-1, and SR cells were obtained from the NCI-Frederick. LAPC4, and MDA PCa 2b, cells were obtained from J.T. Isaacs (Johns Hopkins University, Baltimore, MD). All cell lines used were authenticated via short tandem repeat profiling of 9 genomic loci with the Powerplex 1.2 system (Promega) before use. Mycoplasma testing was not performed. LNCaP, CWR22Rv1, and PC3 were maintained in RPMI1640 media containing l-glutamine and 10% heat-inactivated FBS at 37°C and 5% CO2. LAPC4 cells were maintained in Iscove's modified Dulbecco's medium media containing l-glutamine, 25 mmol/L HEPES, synthetic steroid R1881, and 10% FBS. Cells were maintained in serum and steroid-free media 2 hours prior to treatment. Cells were treated with 10 ng/mL lipopolysaccharide (LPS, L4391, Sigma-Aldrich) or 10 μmol/L enzalutamide (catalog no. HY-70002, MedChemExpress) for 48 hours. Cells treated with both were treated with enzalutamide 30 minutes prior to LPS treatment. AR-targeted short hairpin RNA (shRNA) plasmids were obtained from the Johns Hopkins University ChemCORE. AR shRNAs were incorporated into LNCaP cells by lentiviral transduction and puromycin (1 μg/mL) selection done for at least 5 days. Cells were transduced with shRNA within 20 passages of frozen stocks being in culture. Cells were collected into FFPE plugs as described previously (24).
Western blotting
Total protein extracts were obtained by homogenizing cells in total lysis buffer (25 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1.0% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, and cOmplete mini protease inhibitor cocktail; catalog no. 11836153001, Sigma-Aldrich) and centrifuging at 12,000 rpm for 20 minutes (4°C). Equal amounts of total proteins as determined by BCA Protein Assay were analyzed by SDS-PAGE with antibodies specific to IL8 (catalog no. 18672, Abcam) and prostate-specific antigen (catalog no. 2475, Cell Signaling Technology). The protein bands were detected by enhanced chemiluminescent reaction according to the manufacturer's instructions (Thermo Fisher Scientific). Blots were probed with antibody specific for β-actin (catalog no. A5441, Sigma-Aldrich) or TATA-binding protein (catalog no. 51841, Abcam) to ensure equal loading of proteins in each lane.
RISH
RISH was performed using the RNAscope 2.5 FFPE Brown Reagent Kit (catalog no. 322310, Advanced Cell Diagnostics) as per the manufacturer's instructions. Briefly, FFPE tissues were baked at 60°C for 30 minutes followed by deparaffinization in three changes of 100% xylene for 10 minutes each and two changes of 100% alcohol for 1 minute each. Next, the slides were treated with hydrogen peroxide for 10 minutes at room temperature. The slides were then added to boiling buffer for 15 minutes in a steamer and then treated with protease digestion buffer for 30 minutes at 40°C. The slides were incubated with a custom RNAscope target probe designed against IL1β mRNA [probe region 2-1319, NCBI reference sequence Accession (NCBI seq.) #NM_000576], IL6 mRNA (probe region 27-1112, NCBI seq. #NM_000600.3), IL8 mRNA (probe region 2-1082, NCBI seq. #NM_000584.3), IL10 mRNA (probe region 122-1163, NCBI seq. #NM_000572.2), CD68 mRNA (probe region 367-1149, NCBI seq. #NM_001040059.1), AR mRNA (probe region 2322-3578, GenBank FJ235916.1), or peptidyl prolyl isomerase B, also known as cyclophilin B, as a positive control mRNA (probe region 139-989, NCBI seq. #NM_000942.4) for 2 hours at 40°C, followed by signal amplification. 3,3′-Diaminobenzidine (DAB) was used for colorimetric detection for 10 minutes at room temperature. Slides were counterstained with Gill's hematoxylin. All FFPE blocks and sectioned slides were maintained at −20°C soon after collection, to avoid effects of block age on RISH analyses (25). Slides were scanned using an Aperio ScanScope (CS model) and then viewed and analyzed using Aperio ImageScope Software (version 12.2.2.8013). TMA-scanned images were segmented and then analyzed using TMAJ and FrIDA software (version 3.15.0) as described previously (25).
IHC
IHC was performed using the Power Vision+ Poly-HRP IHC Kit (catalog no. PV6119, Leica Biosystems). Slides were steamed for 25 minutes in antigen retrieval solution (catalog no., H-3300, Vector Laboratories, Inc.) and incubated with mouse monoclonal anti-CD66ce antibody (catalog no. AM325, BioGenex), mouse monoclonal anti-IL8 antibody (catalog no. 18672, Abcam), or rabbit monoclonal anti-AR antibody (catalog no. 5153, Cell Signaling Technology) for 45 minutes at room temperature. Poly–horseradish peroxidase-conjugated anti-mouse IgG antibody or anti-rabbit IgG was used as secondary antibody. Staining was visualized using DAB (catalog no. D4168, Sigma-Aldrich) and slides were counterstained with hematoxylin.
RISH and IHC dual stain
RISH staining was performed as described above. After DAB, slides were incubated with mouse monoclonal anti-CD66ce antibody (1 hour), mouse monoclonal anti-cytokeratin 8 (catalog no. MU142, BioGenex) for 45 minutes, or mouse monoclonal anti-CD68 (catalog no. MO814, Dako) for 45 minutes at room temperature. Power Vision+ Poly-AP IHC kit was used, followed by Vector red alkaline phosphatase substrate kit (catalog no. SK-5105, Vector Laboratories, Inc.), and then slides were counterstained with hematoxylin.
Controls for RISH and IHC
Cell lines expressing low or undetectable levels of IL1β mRNA (DU145), IL6 mRNA (MCF7), IL8 mRNA (MDA PCa 2b), and IL10 (PC3) were used as negative controls for RISH analysis (26). MCF7 (breast cancer cell line) cells transfected with the IL6 cDNA clone expression vector (catalog no. SC125236, OriGene) using Lipofectamine (catalog no. 11668-030, Thermo Scientific) and cells expressing high levels of IL1β mRNA (PC3), IL8 mRNA (PC3), and IL10 (SR) were used as positive controls for RISH analysis (26). All positive and negative RISH controls are shown in Supplementary Fig. S1A. FFPE Tonsil tissues were used as positive controls for anti-CD66ce and anti-CD68 IHC (data not shown).
In silico chromatin immunoprecipitation sequencing analysis
Chromatin immunoprecipitation sequencing (ChIP-seq) data were obtained from the Gene Expression Omnibus dataset GSE61268 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE61268) that contains ChIP-Seq data acquired with AR-specific antibody (Millipore; catalog no.: 06–680; lot number: 2328016) on cell lysates from LNCaP (GSM1501185) or C4-2b (GSM1501189) cells cultured with R1881 (27). ChIP-Seq visualization was performed using the Integrative Genomics Viewer from the Broad Institute (http://software.broadinstitute.org/software/igv/) using the hg38 reference version of the human genome.
Statistical analysis
Data were compared among groups by one-way ANOVA or between groups by two-tailed Mann–Whitney U or Wilcoxon matched-pairs signed rank test using GraphPad Prism Software (version 7.01; GraphPad Software, Inc.). Spearman rank-order correlation was used to compare trends of expression between groups. Values were considered significantly correlated at P < 0.05.
Results
IL1β, IL6, IL8, and IL10 mRNA is differentially expressed in prostate cancer tissues
We conducted RISH on whole tissue sections from a representative tumor-containing block and a matched benign tissue block from 19 EA and 19 AA patients (n = 38 patients, Supplementary Table S1) to determine the cellular localization and distribution of IL1β, IL6, IL8, and IL10. We opted for RISH because these cytokines are all secreted proteins, and even at optimal conditions, IHC may not represent accurately how much cytokine is being produced (Supplementary Fig. S1B). Our analysis revealed differential expression of IL1β, IL6, IL8, and IL10 mRNA in prostatectomy specimens. As recently described (28), we observed modest IL1β mRNA expression in stromal immune infiltrates and in epithelial cells in some inflamed atrophic glands (Fig. 1). Consistent with our previous report that included predominantly EA men (29), IL6 mRNA was not expressed in tumor cells and was largely confined to the stromal compartment in endothelial cell lining vessels and to a lesser extent in infiltrating immune cells and atrophic epithelial cells (Fig. 1). IL8 was the most abundantly expressed cytokine observed in every case in our study, and this is further detailed in the following sections. Limited IL10 mRNA expression was observed in atrophic glands and immune cells (Fig. 1). Each cytokine assessed was dramatically upregulated in regions of marked acute or chronic inflammation (Fig. 1). There was no apparent difference in the expression of any of the cytokines analyzed between races.
RISH analysis revealed differential mRNA expression of IL1β, IL6, IL8, and IL10 in prostatectomy tissues. Modest IL1β expression was observed primarily in stromal compartment cells consistent with immune/inflammatory cells (arrows), or at times in atrophic epithelial cells (arrowhead). IL6 expression was largely confined to the stromal compartment in endothelial cell lining vessels (arrowheads) and on immune cells (arrows). IL8 was the most highly expressed cytokine assessed, observed both in benign and tumor regions. IL8 expression was predominant in atrophic glands (red arrowheads) but was also seen on immune cells (red arrows) and potentially tumor cells (blue arrowhead). Limited IL10 expression was observed on immune cells (arrows). Cytokine expression was dramatically increased in highly inflamed regions.
RISH analysis revealed differential mRNA expression of IL1β, IL6, IL8, and IL10 in prostatectomy tissues. Modest IL1β expression was observed primarily in stromal compartment cells consistent with immune/inflammatory cells (arrows), or at times in atrophic epithelial cells (arrowhead). IL6 expression was largely confined to the stromal compartment in endothelial cell lining vessels (arrowheads) and on immune cells (arrows). IL8 was the most highly expressed cytokine assessed, observed both in benign and tumor regions. IL8 expression was predominant in atrophic glands (red arrowheads) but was also seen on immune cells (red arrows) and potentially tumor cells (blue arrowhead). Limited IL10 expression was observed on immune cells (arrows). Cytokine expression was dramatically increased in highly inflamed regions.
IL8 mRNA expression is markedly upregulated in inflammatory atrophy and in select tumor regions
We observed IL8 mRNA expression predominantly in epithelial cells in atrophic glands throughout the prostate (Figs. 1 and 2A). IL8 expression in epithelial cells in atrophic glands was most intense in areas with prominent stromal inflammatory infiltrates (Fig. 2A). Further, IL8 expression was often observed in atrophic glands containing basal cell–proliferative regions (Fig. 2B). Small (e.g., usually, but not always, microscopic) laminated bodies called corpora amylacea are common in benign regions of radical prostatectomy specimens and are believed to be deposits generated by acute inflammation as they are comprised largely of lactoferrin and other components of neutrophil granules (30). Atrophic epithelial cells in glands/acini encapsulating corpora amylacea often expressed considerable amounts of IL8 (Fig. 2C). Interestingly, there was dramatic IL8 mRNA expression observed in prostatic urothelial cells located in or near the prostatic urethra in each case containing urothelial tissue (Fig. 2D).
RISH analysis demonstrated IL8 mRNA expression in prostate tissues from men with prostate cancer. IL8 expression was predominant in atrophic glands particularly (A) in regions of intense inflammation, (B) atrophic glands involving basal cell proliferation, and (C) atrophic glands encapsulating corpora amylacea. IL8 expression was markedly increased in (D) urethral tissue in most cases. There was considerable IL8 expression observed in (E and F) tumor regions on immune infiltrates and in tumor cells (arrowheads). G and H, Concurrent IL8 RISH (brown) and CK8 IHC (red) analysis confirms that IL8 mRNA is expressed within prostate epithelial cells (arrows) in tumor regions; (I) CD68 (red)-positive macrophages do not express IL8, but some (J) CD66 (red)-positive neutrophils do (black arrows). There were also some IL8− neutrophils (red arrow).
RISH analysis demonstrated IL8 mRNA expression in prostate tissues from men with prostate cancer. IL8 expression was predominant in atrophic glands particularly (A) in regions of intense inflammation, (B) atrophic glands involving basal cell proliferation, and (C) atrophic glands encapsulating corpora amylacea. IL8 expression was markedly increased in (D) urethral tissue in most cases. There was considerable IL8 expression observed in (E and F) tumor regions on immune infiltrates and in tumor cells (arrowheads). G and H, Concurrent IL8 RISH (brown) and CK8 IHC (red) analysis confirms that IL8 mRNA is expressed within prostate epithelial cells (arrows) in tumor regions; (I) CD68 (red)-positive macrophages do not express IL8, but some (J) CD66 (red)-positive neutrophils do (black arrows). There were also some IL8− neutrophils (red arrow).
In a subset of cases, there was elevated IL8 expression observed within tumor regions (Fig. 2E and F). A dual stain with IHC analysis of epithelial cell marker cytokeratin-8 and RISH analysis of IL8 confirmed that IL8 is expressed at times in tumor cells (Fig. 2G and H). Similar multiplex RISH analysis ruled out CD68+ macrophages as the major immune cells expressing IL8, but suggested that CD66ce+ myeloid cells, considered in most cases to be neutrophils, are among the IL8-expressing immune cells (Fig. 2I and J). Not all neutrophils expressed IL8, but neutrophils were generally present in regions with high IL8 expression. Because IL8 was the most abundantly expressed cytokine examined in this study, and the only cytokine found to be expressed in tumor cells, we assessed IL8 expression further in an expanded cohort of patients.
Higher IL8 expression in both tumor and adjacent benign regions of cases with higher-grade cancer
Quantification of the IL8 RISH signal in the whole tissue sections, including areas containing atrophy with and without inflammatory infiltrates, revealed no difference in IL8 mRNA expression in benign compared with tumor regions (P = 0.6475, Mann–Whitney U test; Supplementary Fig. S2A). We detected the highest IL8 RISH signal per unit area in the epithelial compartment within urethral tissue (P < 0.0001, Supplementary Fig. S2A). There was a positive correlation (r = 0.5098, P = 0.0056) between IL8 mRNA expression detected in urethral tissue and that detected in prostatic tissue (Supplementary Fig. S2B). However, there was no difference in urethral expression between grades or between races (Supplementary Fig. S2C and S2D).
IL8 mRNA expression was significantly elevated in tumors from men with higher-grade prostate cancer compared with those with lower-grade cancer (P = 0.005, Supplementary Fig. S2E). The same trend was observed in tumor–adjacent benign regions (e.g., benign regions assessed on the tumor-containing block). There was no difference in IL8 mRNA expression when assessed across the entire whole tissue sections between AA and EA men (Supplementary Fig. S2F). Among AA men, IL8 expression was elevated in sections with higher-grade cancer versus lower-grade cancer (P = 0.0203). The same trend was observed among EA men, but was not statistically significant, perhaps due to sample size (Supplementary Fig. S2G). Therefore, to perform similar analyses in a larger cohort of samples, we next performed IL8 RISH in TMAs.
Further demonstration of higher IL8 mRNA expression in tumor and benign regions of higher-grade prostate cancers using TMAs
We next quantified IL8 mRNA expression in a 120-case TMA set comprised of tissues from radical prostatectomies from 60 AA and 60 EA men (Supplementary Table S1). This TMA set contains 4 tissue spots of the largest/highest-grade index tumor from each case as well as 4 benign tissue spots taken most often from a block that did not contain tumor. Importantly, benign regions on this TMA set were sampled with deliberate avoidance of areas with overt inflammation. We observed significantly elevated IL8 mRNA expression in tumor regions compared with benign regions (Wilcoxon matched-pairs signed rank analysis, P = 0.0003, Fig. 3A). Similar to our results in whole tissue sections, IL8 mRNA expression was significantly increased in tissues from men with higher-grade compared with lower-grade prostate cancer (P = 0.0001, Supplementary Fig. S2H), in both tumor and benign tissues (Fig. 3B). There was no statistical difference in IL8 expression when all spots (tumor and benign) were assessed between AA and EA men (Fig. 3C). IL8 expression was elevated in higher-grade versus lower-grade tumors in both AA and EA men, but only statistically significant for EA men (P = 0.0275, Fig. 3D). During our analysis, we noted whether the benign spots on the TMA contained atrophy or PIN. IL8 mRNA expression was significantly elevated in spots that contained atrophy compared with normal, PIN, or tumor (Fig. 3E). Tumor IL8 mRNA expression was significantly higher than in normal tissue that did not contain atrophy and, interestingly, IL8 expression within PIN regions was comparable with that of normal glands (Fig. 3E). IHC analysis revealed equivalent CD66+ neutrophils in benign and tumor regions in the TMA set, suggesting that IL8-expressing neutrophils could not fully account for the increased IL8 mRNA expression measured in tumor regions (Fig. 3F). There was also no difference in CD66+ neutrophils between low-grade and high-grade tumors or between races (Supplementary Fig. S3).
Quantification of RISH IL8 expression in TMAs showed (A) significantly increased expression in tumor regions compared with benign regions. B, There was elevated IL8 expression in both benign and tumor regions in higher-grade cases. C, There was no apparent difference in expression between races, (D) but a modest increase was observed in higher-grade tumors compared with lower-grade tumors from AA and EA men, but only significant for EA men. E, There was significantly elevated IL8 expression in atrophic glands compared with the remaining regions of the prostate. F, IHC analysis of CD66 showed comparable neutrophils in benign and tumor regions. Each data point represents the median of up to 4 TMA spots except for E where all TMA spots are shown. Center line shows the median for each group (*, P < 0.05; **, P < 0.005; ***, P < 0.0005; and ****, P < 0.0001).
Quantification of RISH IL8 expression in TMAs showed (A) significantly increased expression in tumor regions compared with benign regions. B, There was elevated IL8 expression in both benign and tumor regions in higher-grade cases. C, There was no apparent difference in expression between races, (D) but a modest increase was observed in higher-grade tumors compared with lower-grade tumors from AA and EA men, but only significant for EA men. E, There was significantly elevated IL8 expression in atrophic glands compared with the remaining regions of the prostate. F, IHC analysis of CD66 showed comparable neutrophils in benign and tumor regions. Each data point represents the median of up to 4 TMA spots except for E where all TMA spots are shown. Center line shows the median for each group (*, P < 0.05; **, P < 0.005; ***, P < 0.0005; and ****, P < 0.0001).
IL8 mRNA is expressed in prostate cancer metastases and is reciprocal to AR expression
IL8 RISH analysis of an autopsy TMA set containing prostate cancer that was in the prostate at time of autopsy (e.g., the men had not undergone radical prostatectomy), seminal vesicle invasion, and distant metastatic tissues revealed positive IL8 expression in all eight sites assessed (Fig. 4A). Similar to primary tumors, we observed IL8 expression in tumor, stromal, and immune cells. The majority of TMA spots (>80%) from bone, liver, lung, and lymph node metastases contained IL8-expressing cells. Most metastatic tissues from liver and lung were strongly positive for IL8 expression and significantly higher compared with primary tumor sites (7.8-fold, P < 0.0001; 5.3-fold, P = 0.0021, respectively, Fig. 4B). In the liver, the majority of IL8-expressing cells were likely immune cells suggestive of neutrophils (Supplementary Fig. S4A). Although there was decreased IL8 mRNA expression in EA compared with AA men that did not reach statistical significance (P = 0.06, Supplementary Fig. S4B), given small sample size, we could not accurately assess whether there were differences in IL8 expression by race in metastatic cases. Interestingly, AR mRNA expression as assessed by AR RISH was significantly higher in metastases from EA men than AA men (Supplementary Fig. S4C).
RISH analysis of a TMA set constructed from metastatic tumors obtained via autopsy showed that (A) IL8 mRNA is differentially expressed in every site assessed. B, Metastatic tissues from liver had the most intense signal compared with prostate (P < 0.0001). Each data point represents a TMA spot, and centerline shows the median for each group. The number of patients from which each site was sampled is noted (n = x). C, IL8 RISH analysis revealed IL8 mRNA expression in AR-null cell lines compared with AR-expressing cell lines (**, P < 0.005 and ****, P < 0.0001).
RISH analysis of a TMA set constructed from metastatic tumors obtained via autopsy showed that (A) IL8 mRNA is differentially expressed in every site assessed. B, Metastatic tissues from liver had the most intense signal compared with prostate (P < 0.0001). Each data point represents a TMA spot, and centerline shows the median for each group. The number of patients from which each site was sampled is noted (n = x). C, IL8 RISH analysis revealed IL8 mRNA expression in AR-null cell lines compared with AR-expressing cell lines (**, P < 0.005 and ****, P < 0.0001).
We next performed IL8 RISH analysis on benign prostate (RWPE-1) and prostate cancer cell lines (PC3, DU145, LNCaP, C4-2b, CWR22Rv1, LAPC4, MDA PCa 2b, and VCaP). We noted with interest that IL8 was expressed in AR-negative/nonfunctional (RWPE-1, PC3, and DU145) compared with androgen-dependent cell lines (LNCaP, MDA PCa 2b, CWR22Rv1, LAPC4, and VCaP, Fig. 4C). Interestingly, the androgen-independent C4-2b cell line lacked IL8 expression, but had intact nuclear AR expression (Fig. 4C). Further, reciprocal mRNA expression of IL8 and AR (increased IL8 with decreased AR) was observed in multiple prostate cell lines via the Broad Institute Cancer Cell Line Encyclopedia (Supplementary Fig. S5A). Due to this potential relationship between IL8 mRNA expression and AR loss, we next performed IHC analysis of AR expression in the metastatic TMA set. Although subtle, we found that AR-negative (AR−) metastases were more likely to express IL8 mRNA specifically in tumor cells, compared with AR+ tumors that were largely devoid of IL8 expression within tumor cells (Supplementary Fig. S5B). Analysis of whole tissue sections from metastatic tissues collected at autopsy further supported this observation. In a unique case, an autopsy patient (Patient 9) displayed positive AR expression in tumors at some metastatic sites but AR loss in others. In this patient, the AR+ metastases lacked IL8 mRNA expression, whereas there was IL8 mRNA expression observed in the AR− metastases (Fig. 5A). This same pattern was observed for additional AR+ versus AR− metastases in this autopsy series (Fig. 5A).
A, AR-null metastatic tissues have modestly elevated IL8 mRNA in epithelial cells compared with in AR-positive metastases (IL8 CISH left, AR IHC right). B, Regions of prostatic atrophy with dramatically elevated IL8 expression (red arrowheads) were more likely to have AR downregulation in radical prostatectomy tissues.
A, AR-null metastatic tissues have modestly elevated IL8 mRNA in epithelial cells compared with in AR-positive metastases (IL8 CISH left, AR IHC right). B, Regions of prostatic atrophy with dramatically elevated IL8 expression (red arrowheads) were more likely to have AR downregulation in radical prostatectomy tissues.
To determine whether reciprocal IL8 and AR expression is unique to metastatic tissues, we performed AR IHC analysis on select radical prostatectomy whole tissue sections with significantly elevated IL8 expression in epithelial cells in prostatic atrophy, as AR protein levels are known to be heterogeneously decreased in prostatic atrophy (31). Intriguingly, we observed that AR was downregulated in regions with dramatically high IL8 expression (Fig. 5B).
We observed a negative correlation between IL8 and AR mRNA expression in primary tumors (n = 157) in the Memorial Sloan Kettering Cancer Center (MSKCC) dataset via the cBioPortal for cancer genomics (Supplementary Fig. S6A; ref. 32). Similarly, in metastatic tumors from the SUTC/PCF dream team dataset and a neuroendocrine tumor dataset, the samples with the highest IL8 mRNA expression had the least AR mRNA and the samples with the least IL8 had the highest AR mRNA expression (Supplementary Fig. S6A; refs. 33, 34). Similar patterns were observed when comparing IL8 receptors CXCR1 and CXCR2 with AR expression in primary tumors (Supplementary Fig. S6B).
The reciprocal relationship between AR and IL8 mRNA expression is associated with AR deficiency, not castration resistance
To further assess the relationship between IL8 expression and AR, we performed IL8 RISH analysis on a TMA set constructed from 42 LuCaP PDX cell lines (23). The TMA contained both androgen-sensitive and castration-resistant (LuCaP-CR) derivations of many of the LuCaP lines. There was no difference in IL8 mRNA expression between LuCaP and LuCaP-CR spots (Fig. 6A and C). Instead, the mean IL8 mRNA expression was significantly elevated in AR-low or -negative lines compared with AR+ lines (P < 0.0001; Fig. 6B and C). These data suggest that IL8 mRNA expression correlates to the absence of AR as opposed to the androgen dependence of the tumor.
IL8 RISH analysis of 42 LuCaP PDX TMA set revealed (A and C) comparable IL8 mRNA expression between PDX and castration-resistant PDX lines, but (B and C) elevated IL8 expression in AR− PDX compared with AR+ PDX lines. Each data point in A and B represents the mean of 3 TMA spots, and the centerline shows the mean of each group. ***, P < 0.0001.
IL8 RISH analysis of 42 LuCaP PDX TMA set revealed (A and C) comparable IL8 mRNA expression between PDX and castration-resistant PDX lines, but (B and C) elevated IL8 expression in AR− PDX compared with AR+ PDX lines. Each data point in A and B represents the mean of 3 TMA spots, and the centerline shows the mean of each group. ***, P < 0.0001.
AR represses IL8 mRNA expression, likely via binding to the IL8 promoter
IL8 has previously been characterized as an AR-repressed gene due to AR binding to the IL8 promoter in a series of ChIP together with microarray experiments in the HPr-1AR prostate epithelial cell line (35). An in silico analysis of AR ChIP-seq performed on LNCaP and C4-2b cell lines revealed AR binding at the IL8 promoter region (Fig. 7A), and this has also been recently described (36). AR binding at the IL8 promoter of C4-2b cells is of particular interest given our finding that IL8 is not expressed in the AR-positive but androgen-independent C4-2b cell line (Fig. 4C). To further demonstrate that IL8 mRNA expression is repressed by AR, we assessed whether IL8 may be induced in prostate cancer cell lines with functional AR. LPS treatment induced IL8 protein in AR− PC3 and DU145 cells (Fig 7B; Supplementary Fig. S7A), but not in AR+ LNCaP, LAPC4, or CWR22Rv1 cells (Fig. 7B). Pretreatment with AR antagonist enzalutamide (10 μmol/L) did not induce IL8, neither independently nor via LPS treatment (Fig. 7B; Supplementary Fig. S7B). Alternatively, AR knockdown in LNCaP cells resulted in increased IL8 mRNA expression as assessed by RISH (Fig. 7C). These data further support that the presence of AR represses IL8 expression and that AR loss is necessary (albeit not sufficient because not all AR-negative cells are IL8 positive) for increased IL8 expression.
A, ChIP-seq analysis shows AR binding at the IL8 promoter (blue arrow) in LNCaP and C4-2b cell lines. B, LPS treatment induced IL8 protein expression in AR-negative PC3 cells but not in AR-positive LNCaP, LAPC4, or CWR22rv1 cells, even with AR antagonist (enzalutamide) treatment. C, AR-targeted shRNA (AR3717, AR3718, and AR26211) knockdown in LNCaP cells resulted in modest IL8 mRNA expression.
A, ChIP-seq analysis shows AR binding at the IL8 promoter (blue arrow) in LNCaP and C4-2b cell lines. B, LPS treatment induced IL8 protein expression in AR-negative PC3 cells but not in AR-positive LNCaP, LAPC4, or CWR22rv1 cells, even with AR antagonist (enzalutamide) treatment. C, AR-targeted shRNA (AR3717, AR3718, and AR26211) knockdown in LNCaP cells resulted in modest IL8 mRNA expression.
Discussion
Prior studies have not examined cell type–specific expression of cytokines in prostatic carcinomas from men of different races. Therefore, we performed RISH analysis to visualize cytokine expression in primary tumors and matched benign tissue from prostatectomies or castration-resistant metastatic tumors from autopsies from AA and EA men. Each of the cytokines, IL1β, IL6, IL8, and IL10, was differentially expressed in prostate tissues, but had similar expression patterns between prostate tissues from AA and EA men. Our results indicated that IL8 mRNA was not expressed differently between AA and EA men when matched on patient age, tumor grade, and stage. However, AA men are more likely to be diagnosed with higher-grade tumors, which we did observe to be significantly associated with higher IL8 expression levels. IL8 expression was significantly higher in tumors from AA compared with EA men in a previous microarray-based study that included significantly more higher-grade cancers among the AA men (13). In another study case matched for pathologic stage and Gleason score, genes associated with cytokine signaling were highly significantly different between EA and AA men, yet IL8 was not among the differently expressed genes (14). Similarly, in a cohort limited to Gleason 6 cases, gene ontology term analysis implicated ILs as more prevalent in AA men, yet IL8 was not reported as upregulated (6).
In our study, IL8 mRNA expression was observed predominantly in prostatic urothelium and prostate epithelial cells, specifically in atrophic glands. The initial cause of prostatic inflammation observed in a particular case is typically unclear but may include infectious agents, urine reflux, dietary factors including heterocyclic amines, and hormonal changes (7, 8). Phagocytes responding to the immune stimulus release reactive oxygen and nitrogen species that may cause DNA damage, cell injury, and cell death (7, 8). Chronic inflammation may therefore contribute to tissue damage. This proposed inflammation-injury cycle corresponds with the excessive IL8 mRNA expression observed within atrophic glands of the prostate. It is further proposed that inflammation-induced cell death and cell injury may promote cell regeneration, prompting otherwise quiescent prostate epithelial cells into proliferation (7, 8). IL8, specifically, may contribute to cell proliferation as its mitogenic properties have been described in multiple cell types (15, 37). IL8 induces proliferation in PC3 and DU145 prostate cancer cells via Erk-MAPK signaling or Akt phosphorylation (38, 39). Also, IL8 cytoplasmic expression in prostate biopsies was correlated with Ki-67 staining (39). Conversely, abrogation of IL8 signaling in PC3 cells with IL8 siRNA decreased proliferation and inhibited their invasive activity (40). IL8 is further described as protumorigenic based on its ability to promote cell invasion, cell migration, and angiogenesis. In addition, IL8 may stimulate tumor-associated macrophages to secrete growth factors that increase the proliferation of tumor cells (15). Altogether, it is likely that IL8 expression in benign regions helps create a microenvironment conducive to tumorigenesis, whereas IL8 expressed within the TME near and in tumors potentially promotes prostate tumor progression.
IL8 RISH quantification of whole tissue sections demonstrated significantly elevated IL8 expression in both tumor and adjacent benign tissues from higher-grade versus lower-grade prostate cancer. In this instance, there was no distinction made among benign regions between normal glands versus atrophy or PIN. In contrast to our analysis in whole tissue sections, quantification of IL8 RISH in the 120-case TMA showed elevated IL8 expression in tumor compared with benign regions. This is likely because for construction of this TMA, benign TMA spots were selected from regions with little or no inflammation, and also from a benign block that was distant to the tumor block. The TMA spots were further characterized as containing normal, atrophy, PIN, or tumor. Tissue spots with atrophy were much more frequently positive for relatively high levels of IL8 than normal, PIN, or tumor tissues. Atrophic lesions were also more likely to have a larger number of individual cells expressing IL8. Alternatively, in tumor regions a smaller number of individual cells strongly express IL8, whereas many surrounding tumor cells do not. Similarly, in IL8+ prostate cancer cell lines and PDX lines, we observe less than half of all cells strongly expressing IL8, whereas surrounding cells are negative for IL8 mRNA. Because IL8 is a secreted protein, it may not be necessary for every tumor cell to produce the cytokine to induce downstream effects in neighboring cells. In addition, cell-to-cell variability including cell-cycle phase, DNA alterations, and AR expression may determine IL8 production. Corroborating our findings in whole tissue sections, both benign and tumor spots from higher-grade cases had significantly elevated IL8 expression compared with lower-grade tissues. It is of interest that this trend of increased IL8 expression in benign tissues of high-grade versus lower-grade cancer held up in the TMA analysis because those benign tissues were sampled distant to the tumor. This finding implies that there may be elevated IL8 expression across broad regions of the prostate in cases with high-grade cancer. Increased expression with grade is consistent with previous reports of IL8 serum and gene expression measurements (19, 41, 42).
Elevated serum IL8 is also associated with poorer overall survival in men with metastatic prostate cancer (18). We observed considerable IL8 mRNA expression in metastatic tumors in cells within the TME and/or in the metastatic cancer cells. Consistent with our observation in prostatectomy tissues and cell lines, not every tumor cell was positive for IL8. Most intriguing was our observation that AR− metastatic tumor cells were much more likely to express IL8 compared with AR+ metastases. Previous studies reported that androgen-responsive cells LNCaP and LAPC4 do not express IL8, whereas PC3 and DU145 do (43–45). Araki and colleagues also reported that IL8 expression is related to androgen independence in cell lines and demonstrated increased motility, invasion, matrix metalloproteinase-9, and VEGF production in IL8-transfected prostate cancer cells (45). Our study corroborates these observations, expanding the profile to include nine commonly used prostate cell lines as well as PDX lines. Enhanced IL8 mRNA expression was observed in AR-negative cell lines and PDX lines. Further, AR+ C4-2b cells did not express IL8 in spite of its androgen independence, and we observed comparable IL8 mRNA expression in LuCaP and LuCaP-CR PDX tissues, but significantly higher IL8 expression in AR− or AR-low LuCaP PDX cells. Supporting these observations, our in vitro studies demonstrated that although pharmacologic blockade of AR had no effect of IL8 mRNA expression, shRNA complete knockdown of AR resulted in increased IL8 mRNA expression. Enzalutamide activity was confirmed by PSA reduction, and we observed a modest reduction of overall nuclear AR protein in treated cells (Supplementary Fig. S7C). However, there may not have been sufficient depletion of AR in individual cells to induce IL8 expression as observed in our shRNA knockdown, where some cells appeared to completely lack AR. Altogether, these data suggest that IL8 expression correlates with the loss of functional AR and not necessarily with castration resistance per se. Our study also now demonstrates the reciprocal association between IL8 and AR in situ in patient tissues including in prostatic atrophy and metastatic prostate cancer.
In contrast to our findings of inverse expression of IL8 and AR, others have proposed that IL8 promotes prostate cancer cell proliferation, likely via induction and activation of AR (46), even in castration-resistant cells. Inhibition of IL8 signaling enhanced the ability of AR antagonist bicalutamide to reduce prostate cancer cell viability (46). AR signaling in cancer-associated fibroblast-like cells was reported to decrease IL8 expression (47). Another study reported that dihydrotestosterone reduced IL8 secretion in melanoma cells ultimately suppressing growth (17). As described earlier, IL8 potentiates proliferation and cell survival. IL8 has also been shown to be expressed in neuroendocrine cells in the prostate (48). Whereas we were unable to definitively assess an association between IL8 expression and neuroendocrine phenotype in metastatic prostate cancer due to the low number of metastatic samples in this study, the association of IL8 expression with AR loss implies this association may exist and should be the focus of future studies. Collectively, these findings suggest that IL8 may facilitate androgen-independent tumor growth due to or in concert with AR loss, thereby promoting prostate cancer cell proliferation using AR-independent mitogenic pathways.
In conclusion, our study identifies IL8 as a cytokine of interest in both primary and metastatic prostate cancer. We observed heightened IL8 expression in inflamed benign regions that might be most susceptible to tumorigenesis (e.g., inflammatory atrophy) as well as within tumor cells. Our study further corroborates a role for IL8 in prostate cancer progression, but likely via promotion of AR-independent mitogenic pathways as IL8 expression was mutually exclusive with AR expression. Future studies are needed to assess the precise relationship between IL8 and AR and whether IL8 or its receptors may be a therapeutic target for AR-negative metastatic lesions. Finally, our data highlight IL8 as a key molecule that is potentially involved in aggressive prostate cancer among both AA and EA men.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: J.P. Maynard, O. Ertunc, A.M. De Marzo, K.S. Sfanos
Development of methodology: J.P. Maynard, K.S. Sfanos
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.P. Maynard, I. Kulac, J.A. Baena-Del Valle, A.M. De Marzo
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.P. Maynard, O. Ertunc, J.A. Baena-Del Valle, K.S. Sfanos
Writing, review, and/or revision of the manuscript: J.P. Maynard, O. Ertunc, J.A. Baena-Del Valle, A.M. De Marzo, K.S. Sfanos
Study supervision: K.S. Sfanos
Acknowledgments
This work was supported by Department of Defense PCRP awards W81XWH-14-1-0364 and W81XWH-17-1-0286 (to K.S. Sfanos), Department of Defense PCRP award W81XWH-17-1-0292 (to J.P. Maynard), and Department of Defense PCRP awards W81XWH-18-2-0013, W81XWH-18-2-0015, and W81XWH-18-2-0017 PCBN (to K.S. Sfanos and A.M. De Marzo). The authors thank the patients and their families who participated in the studies at Johns Hopkins. We would like to thank and acknowledge Dr. Jody Hooper and the Legacy Gift Rapid Autopsy Program at Johns Hopkins for assistance with autopsy specimens. We thank Dr. Eva Corey and Dr. Colm Morrissey for providing the LuCaP TMAs. We also thank Dr. Paula Hurley and Dr. Michael Haffner for helpful discussions and Jessica Hicks for help with IHC.
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.