Abstract
Purpose: Men of African ancestry experience an excessive prostate cancer mortality that could be related to an aggressive tumor biology. We previously described an immune-inflammation signature in prostate tumors of African-American (AA) patients. Here, we further deconstructed this signature and investigated its relationships with tumor biology, survival, and a common germline variant in the IFNλ4 (IFNL4) gene.
Experimental Design: We analyzed gene expression in prostate tissue datasets and performed genotype and survival analyses. We also overexpressed IFNL4 in human prostate cancer cells.
Results: We found that a distinct interferon (IFN) signature that is analogous to the previously described “IFN-related DNA damage resistance signature” (IRDS) occurs in prostate tumors. Evaluation of two independent patient cohorts revealed that IRDS is detected about twice as often in prostate tumors of AA than European-American men. Furthermore, analysis in TCGA showed an association of increased IRDS in prostate tumors with decreased disease-free survival. To explain these observations, we assessed whether IRDS is associated with an IFNL4 germline variant (rs368234815-ΔG) that controls production of IFNλ4, a type III IFN, and is most common in individuals of African ancestry. We show that the IFNL4 rs368234815-ΔG allele was significantly associated with IRDS in prostate tumors and overall survival of AA patients. Moreover, IFNL4 overexpression induced IRDS in three human prostate cancer cell lines.
Conclusions: Our study links a germline variant that controls production of IFNλ4 to the occurrence of a clinically relevant IFN signature in prostate tumors that may predominantly affect men of African ancestry. Clin Cancer Res; 24(21); 5471–81. ©2018 AACR.
Tumor IFN signaling has recently been shown to modulate response and resistance to immune checkpoint blockade. Here, we describe a distinct and biologically relevant IFN signature in prostate tumors that has a high prevalence in African-American (AA) patients. This signature, known as “IFN-related DNA damage resistance signature” (IRDS), predicts decreased disease-free survival. Moreover, we link its occurrence to a germline variant allele, rs368234815-ΔG, within the IFNλ4 (IFNL4) gene. This IFNL4 allele controls production of a type III IFN, IFNλ4, and is a known predictor of decreased viral clearance. Together, these observations indicate that IRDS and IFNL4 rs368234815-ΔG may have a function in the tumor biology and survival of AA patients, and influence immune therapy outcomes, which should be examined in future studies.
Introduction
Prostate cancer incidence and mortality rates are highest among men of African ancestry (1–3). Environmental exposures and ancestry-specific factors may influence prostate cancer biology and cause a more aggressive disease in these men (4–11). We and others described an immune signature that is prevalent in prostate tumors of African-American (AA) patients and hypothesized that this signature affects tumor biology (5, 12–15). Here, we further investigated this immune signature and discovered an IFN signature in tumors of AA patients that is analogous to the previously described IFN-related DNA damage resistance signature, also termed IRDS (16). IRDS includes 49 IFN-stimulated genes (ISG) that are induced through activation of the JAK–STAT pathway (17). IRDS was initially identified because of its effects leading to resistance to resistance to ionizing radiation (17). It can be induced by persistent activation of the JAK–STAT pathway by various exogenous and endogenous stimuli that generate an IFN response (18, 19).
All IFNs (type I, type II, and type III) induce sets of ISGs that are overlapping but also distinct. However, among the human IFNs, the expression of the recently discovered IFNλ4, a type III IFN, is uniquely controlled by genetics. Only carriers of the ΔG allele for the germline variant rs368234815-ΔG/TT in the IFNL4 gene can produce this IFN (20). This allele is most common in individual of African ancestry (up to 80% allele frequency), although it is less common in Europeans (∼30%) and Asians (less than 10%; ref. 20). Moreover, carriers of IFNL4 rs368234815-ΔG have an impaired ability to clear certain viral infections, such as hepatitis C virus (HCV), spontaneously or after treatment (20). One of the mechanisms by which IFNλ4 renders cells refractory to an antiviral response is the induction of a persistent gene signature that resembles IRDS (20, 21). Thus, we hypothesized that because IFNλ4 is produced more commonly in men of African ancestry, it might explain the increased occurrence of IRDS in their tumors, and be of clinical importance. Accordingly, we examined whether IFNL4 rs368234815-ΔG is associated with the development of IRDS in prostate tumors and disease outcomes. Consistent with our hypothesis, we found that IFNL4 rs368234815-ΔG is significantly associated with both the presence of IRDS and increased all-cause mortality among AA prostatectomy patients.
Materials and Methods
Study design, patient data, and cell lines
To compare gene expression profiles from prostate tumors of AA and European-American (EA) patients, we analyzed gene expression data from 2 patient cohorts that were previously described in detail by us (12) and others (22). These two datasets are publicly available (GSE6956 and GSE21032) and consist of microarray data for primary prostate tumors from 33 AA and 36 EA men (12) and 24 AA (non-Hispanic) and 98 EA men (22). These men were previously untreated prostatectomy patients with exception of 16 patients in the Taylor and colleagues’ cohort who received neoadjuvant hormone therapy or chemotherapy. We examined the prevalence of two previously reported IFN-related gene signatures in the tumors, IRDS (16) and IFN -regulated genes (IRG; ref. 23). As described in the publications, IRDS and IRG include 49 and 42 ISGs, respectively. To assess the association of IRDS with disease-free survival, we evaluated the publicly available TCGA prostate cancer data accessible through the Cancer Genomics Data Server (CGDS, at http://www.cbioportal.org/public-portal) and hosted by the Computational Biology Center at Memorial-Sloan-Kettering Cancer Center through cBioPortal for Cancer Genomics. In this large dataset, recurrence-free survival was available for 491 patients who are mainly EA men (self-reported race/ethnicity: 270 white, 43 black, 6 Asian, others unknown). To examine the association between IFNL4 rs368234815-ΔG allele and IRDS in prostate tumors, we genotyped rs368234815-ΔG in available genomic DNA from 44 frozen tumors in the Wallace and colleagues’ cohort (12). We also assessed genome-wide gene expression using a linear regression model in relation to 0, 1, and 2 copies of the ΔG allele and applying an additive model. The Wallace and colleagues study has been approved by Institutional Review Boards, as described previously (12). To assess the association of IFNL4 rs368234815-ΔG with disease recurrence and overall mortality, we isolated genomic DNA from tumor-adjacent, noncancerous prostate tissues that were obtained from 197 AA patients with prostate cancer after a prostatectomy at the Cleveland Clinic (1987–2012). According to follow-up through 2014, 92 patients had a PSA-defined prostate cancer recurrence, and 29 died of various causes (Supplementary Table S1). Clinical information including age at diagnosis, tumor stage, grade (Gleason score), vital status follow-up, and cause of death was available for these subjects. Collection of tissues and patient information was reviewed and approved by the Cleveland Clinic IRB under protocol CCF IRB 313-773. Our research followed the ethical guidelines set by the Declaration of Helsinki, and informed consent was obtained from all patients in the study.
Human prostate cancer cell lines 22Rv1 and PC-3 (from EA donors) and MDA-PCa-2b (from an AA donor) were obtained from the ATCC and have been regularly authenticated using a short tandem repeat analysis with GenePrint10 and tested for Mycoplasma contamination.
IRDS in prostate tumors and associations with IFNL4-ΔG and disease recurrence
The mRNA expression data for Wallace and colleagues (12) were available in-house, although the data for Taylor and colleagues (22) were downloaded as normalized log2 data from the cBio Cancer Genomics Portal (http://cbio.mskcc.org/cancergenomics/prostate/data/). Normalized expression data were subjected to a pathway-level comparative analysis using the Sample-Level Enrichment-Based Pathway Ranking (SLEPR) method (24). SLEPR can be used to quantitatively evaluate gene set-level expression patterns at the sample level (e.g., relative expression of a gene signature in a tumor). Pathway-level enrichment assessment using SLEPR was applied to Gene Ontology annotations (GO, http://www.geneontology.org) and to the two overlapping expression signatures, IRDS and IRG. The IRDS and IRG gene lists were obtained from the two publications that described these signatures (16, 23). Supplementary Table S2 shows the Affymetrix probe set composition of IRDS for the GeneChip HG-U133A 2.0 arrays. Evaluation of sample-level upregulated genes in a pathway or predefined gene signature was performed using the one-sided MADe method (see SLEPR). Computation of pathway-level enrichment scores for each sample-level differentiated gene was performed on the basis of 2 × 2 contingency tables using the Fisher exact test. To determine the statistical significance of enrichment scores at the pathway/gene signature level for class comparison (e.g., AA vs. EA patients), a P value was calculated from 1,000 permutations. FDR Q-values were computed from permutated data. Heatmaps to visualize enrichment scores for upregulated genes in a pathway/gene signature at the sample level (e.g., in an AA tumor) were generated using gradients of red color, which indicate the enrichment level [−log (permutated P value)]. All procedures for SLEPR and visualization of findings in heatmaps are part of the WPS software (24). For the analysis of the association between tumor IRDS and IFNL4 rs368234815-ΔG, enrichment scores for IRDS were dichotomized and tumors with an enrichment score of zero (no significant enrichment above background) were defined as IRDS-negative (n = 25), whereas the other tumors were defined as IRDS-positive (n = 19).
The relationship between IRDS in prostate tumors with disease recurrence was evaluated in RNA sequencing (RNA-seq) data for the TCGA prostate cancer cohort of 491 patients. We extracted the IRDS expression profile from the TCGA RNA-seq data, leading to a 45-gene signature with all genes being measurable expressed, and applied the consensus clustering method to identify distinct tumor clusters with differential IRDS expression in the dataset. This was done using the Bioconductor ConsensusClusterPlus package (https://bioconductor.org/packages/release/bioc/html/ConsensusClusterPlus.html). This method provides a quantitative assessment for determining the number of possible clusters within the dataset. Initially, we tested 2 to 8 cluster groups. Final subgroups were defined according to their cumulative distribution functions (CDF) and the Delta area under the CDF curve, yielding three distinct clusters that represented the most robust clustering in our data. Then, we calculated the IRDS score based on IRDS expression values and assigned those to the three clusters, yielding robust differences between the clusters (low, medium, and high). We then accessed the association of these three IRDS expression groups with disease recurrence using Cox regression modeling and visualized findings with a Kaplan–Meier plot.
IFNL4 rs368234815-ΔG genotyping
Genomic DNA was isolated from frozen tissues using the DNeasy Blood and Tissue Kit, (QIAGEN). To isolate DNA from formalin-fixed, paraffin-embedded (FFPE) tissues, 5-μm sections were deparaffinized and DNA was extracted using the BiOstic FFPE Tissue DNA Isolation Kit (MO BIO Laboratories). DNA quantity and quality were determined by NanoDrop 1,000 (Thermo Fisher Scientific) and 10 ng of DNA was used for genotyping. A previously described TaqMan assay for rs368234815 (20) was purchased from Life Technologies and used on the ABI 7900 (Applied Biosystems) according to standard protocols. Genotyping success rates were 100% for the frozen tissues (44 tumors) and 98.5% for FFPE tissues (194/197), with 100% concordance among duplicates.
Expression analysis of indoleamine-2,3-dioxygenase in prostate tumors
Indoleamine-2,3-dioxygenase (IDO1) expression was measured by quantitative RT-PCR using a Life Technologies TaqMan expression assay (IDO1, Hs00984148_m1) and total RNA from prostate tumors of 21 AA and 22 EA men. Characteristics of these patients have been described previously (12). In this cohort, AA and EA patients are matched on Gleason score and pathologic stage. Relative normalized expression values were calculated as described previously (25), using the group mean Ct for the target and the endogenous control 18s RNA. Fold differences were calculated as 2−ΔΔCt. Graphs were prepared using |$\Delta {C_{\rm{t}}} = {C_{{{\rm{t}}_{{\rm{18S}}}}}} - {C_{{{\rm{t}}_{{\rm{target}}}}}}$| and plotted using GraphPad Prism 7.
Measurement of tryptophan in plasma samples
Levels of tryptophan, which is metabolized by IDO1, were measured in plasma samples from randomly selected 50 AA and 50 EA population-based controls (mean age: 63.4 and 63.7, respectively) and 50 AA and 50 EA patient with prostate cancer (mean age: 61 and 62.4, respectively). These subjects were previously recruited into the NCI-Maryland Prostate Cancer Case–Control study (26). Tryptophan metabolite concentrations were measured at SAIC/Leidos-NCI Frederick, Frederick, MD, using a described high-performance liquid chromatography method (27). Tryptophan measurements were obtained for 197 of 200 plasma samples (98.5%).
In vitro induction of an IFN signature by overexpression of IFNL4 in human prostate cancer cell lines
To examine IFNλ4-induced gene expression, the human prostate cancer cell lines, 22Rv1, PC-3, and MDA-PCa-2b, were transfected with a previously described IFNL4 expression construct that generates IFNλ4 protein with C-terminal Halo-tag (20). A GFP reporter gene was used to monitor transfection efficiency. Transfection of the IFNL4-Halo construct and empty Halo-control vector was performed in triplicates using Lipofectamine/LTX and yielded robust overexpression of IFNL4 in the IFNL4-Halo–transfected cell lines without affecting cell viability (Supplementary Fig. S1). The IFNL4 qRT-PCR was performed as described previously (20). In the experiments with 22Rv1 cells, we also added antibodies that block the activity of IFNλ4 - 20 μg/mL of goat anti–α-IL10R2 antibody (R&D Systems) that blocks one receptor of IFNλ4, or 20 μg/mL of a rabbit monoclonal anti-IFNλ4 antibody (Abcam, ab196984), that blocks IFNλ4 directly. Cells were lysed after 48 hours and total RNA was extracted with the RNeasy Kit with DNase I treatment (QIAGEN). One to 1.5 μg of total RNA was used to generate cDNA using the RT2 First Strand Kit (QIAGEN). The Human Type I IFN Response PCR array (QIAGEN) was used to evaluate expression of a panel of other relevant genes. This array includes 96 SYBR Green expression assays for type I IFNs (IFNA and IFNB) and their receptors, ISGs and molecules involved in response and resistance to signaling, as well as positive and negative controls. Differences in expression between cells transfected with the IFNL4 construct and empty vector were evaluated with a multiple comparisons-adjusted two-tailed t test. We also examined expression of two other type III IFNs to IFNL1 (Hs00601677_g1) and IFNL3 (Hs04193050), and endogenous control 18s RNA (4319413E) with Life Technologies TaqMan expression assays. For all TaqMan assays, gene expression was measured in Ct values and normalized by endogenous control 18s RNA. Analysis of the PCR Arrays was performed using the GeneGlobe Data Analysis Software, in relation to a panel of positive and negative controls included on the array.
BrdU assay
Cells were preseeded in 96-well cell culture plates (Corning Life Sciences) overnight and then cultured in either the complete culture medium or the medium without FBS, or transfected with either the IFNL4-Halo expression vector or the vector control construct. After 48-hour incubation, cells were assessed for cell proliferation using the bromodeoxyuridine (BrdU) colorimetric ELISA assay (Roche) according to the provided protocol. Eight replicates were analyzed per condition, and proliferation was calculated as relative ratio by normalization BrdU incorporation to the control.
Statistical analysis
Analyses were conducted using SAS 9.2 (SAS Institute) and R (R Foundation for Statistical Computing; http://www.r-project.org/). All statistical tests were two-sided. P < 0.05 was considered statistically significant. The nonparametric Mann–Whitney test was used for group comparisons with continuous data and the Fisher exact test for categorized data. Associations between a genotype and the occurrence of IRDS were estimated with linear regression models. Survival analysis was conducted with a multivariate Cox proportional hazards regression model under a log-additive hazards assumption. Disease-free survival was defined from the date of prostatectomy to the date of recurrence (PSA defined). Overall survival was defined from the date of the prostate cancer diagnosis to the date of death (Cleveland Clinic cohort). Survival analyses (disease-free, overall) were controlled for age (<65 or ≥65 years), stage (I, II or III, IV) and Gleason score (<7 or ≥7). Statistical significance was determined using the Wald test.
Results
IRDS is prevalent in prostate tumors of AA men
We previously described an immune-inflammation signature in tumors of AA patients that showed upregulation of genes involved in IFN signaling (12). Here, we examined whether two IFN signatures, IRDS and IRG, are detectable in prostate tumors and more prevalent in AA than EA patients. IRDS and IRG have been described independently but show considerable similarity and could be functionally equivalent, and are both associated with decreased breast cancer survival (16, 23). Our analysis found that both signatures are detected about twice as often in tumors from AA than EA men in two independent datasets, Wallace and colleagues’ and Taylor and colleagues’ (Fig. 1A and B). For example, IRDS was detected at a frequency of 67% and 42% in AA men in these two datasets, which is significantly higher than the observed frequency in EA men (33% and 18%, respectively). We also investigated IRDS and IRG in cultured prostate cancer epithelial cells from 14 AA and 13 EA men. The expression data were previously published by Timofeeva and colleagues (28). This analysis showed that the two signatures can be detected in 5 of 14 (36%) cell cultures from AA men and 2 of 13 (15%) for EA men (Supplementary Fig. S2).
Increased frequency of two IFN signatures, IRDS and IRG, in prostate tumors from AA men. A, Analysis of 33 and 36 tumors from AA and EA men, respectively, in the Wallace and colleagues’ dataset (GSE6956). Prevalence of IRG and IRDS: 55% and 67% in AA and 19% and 33% in EA. Permutated P value and FDR for difference in IRDS frequency between AA and EA tumors: P = 1.6 × 10−4; FDR = 3.7%. Red, upregulated expression. B, Analysis of 24 and 98 tumors from AA and EA men, respectively, in the Taylor and colleagues’ dataset (GSE21032). Prevalence of IRG and IRDS: 46% and 42% in AA and 17% and 18% in EA. Permutated P value and FDR for difference in IRDS frequency between AA and EA tumors: P = 0.006; FDR = 21%. Gene expression data were analyzed with the SLEPR method, as described under Materials and Methods. C, Increased mRNA expression of IDO1 in prostate tumors from AA men. IDO1 TaqMan assay was used to compare expression in AA versus EA men. Lines = mean ΔCt values. Fold expression difference is derived from 2−ΔΔCt values. D, Decreased abundance of tryptophan in plasma of AA patients with prostate cancer. P values from two-sided Mann–Whitney test. Shown are means ± SD.
Increased frequency of two IFN signatures, IRDS and IRG, in prostate tumors from AA men. A, Analysis of 33 and 36 tumors from AA and EA men, respectively, in the Wallace and colleagues’ dataset (GSE6956). Prevalence of IRG and IRDS: 55% and 67% in AA and 19% and 33% in EA. Permutated P value and FDR for difference in IRDS frequency between AA and EA tumors: P = 1.6 × 10−4; FDR = 3.7%. Red, upregulated expression. B, Analysis of 24 and 98 tumors from AA and EA men, respectively, in the Taylor and colleagues’ dataset (GSE21032). Prevalence of IRG and IRDS: 46% and 42% in AA and 17% and 18% in EA. Permutated P value and FDR for difference in IRDS frequency between AA and EA tumors: P = 0.006; FDR = 21%. Gene expression data were analyzed with the SLEPR method, as described under Materials and Methods. C, Increased mRNA expression of IDO1 in prostate tumors from AA men. IDO1 TaqMan assay was used to compare expression in AA versus EA men. Lines = mean ΔCt values. Fold expression difference is derived from 2−ΔΔCt values. D, Decreased abundance of tryptophan in plasma of AA patients with prostate cancer. P values from two-sided Mann–Whitney test. Shown are means ± SD.
Of the two signatures, IRDS and IRG, only IRDS has been characterized (16, 17, 29). We therefore focused on IRDS in our follow-up analyses. Although IRDS captures the expression of many ISGs, others are not represented by this signature. Because IDO1 is an immunosuppressive ISG with a key function in cancer biology, but not an IRDS gene (Supplementary Table S2), we examined its expression in prostate tumors from 21 AA and 22 EA patients from the Wallace and colleagues’ dataset using a qRT-PCR approach. The expression analysis showed an approximately 2-fold increased tumor expression of IDO1 in AA men when compared with EA men (Fig. 1C), consistent with the 2-fold higher prevalence of IRDS in AA men. Because IDO1 uses tryptophan as a substrate for its enzymatic activity, we examined blood tryptophan levels in AA and EA patients with prostate cancer to assess whether increased tumor IDO1 in AA patients may affect these levels (Fig. 1D). We found very similar mean blood tryptophan levels in AA men (31.9 ± 8.1 μmol/L) and EA men (32.1 ± 5.9 μmol/L) without prostate cancer, and in EA patients with prostate cancer (31.9 ± 4.5 μmol/L), but a modest decrease in blood tryptophan (by 8%–9%) among the AA cases (29.3 ± 7.6 μmol/L; P = 0.02 when compared with EA cases; P = 0.08 when compared with AA controls).
IRDS is associated with early disease recurrence
To examine whether IRDS is clinically relevant for patients with prostate cancer, we analyzed available gene expression and cancer recurrence data for 491 patients with prostate cancer in TCGA. We defined the occurrence of IRDS based on 45 IRDS genes with measurable expression and used the RNA-seq data to categorize patients into 3 groups with low, medium, and high IRDS expression in their tumors (Fig. 2). The Kaplan–Meier survival plot shows that an increased IRDS expression is associated with decreased disease-free survival (Fig. 2C). Using a Cox regression model, we estimate that patients with high IRDS expression in their tumors have a 2-fold increased hazard of an earlier disease recurrence [HR = 2.09; 95% confidence interval (CI), 1.07–4.10 after controlling for age, disease stage, and Gleason score], when compared with patients with low IRDS expression.
IRDS is associated with early disease recurrence in the TCGA prostate cancer cohort. A, Expression of 45 IRDS genes identifies prostate tumors with low (group 1: n = 159), medium (group 2: n = 270), and high (group 3: n = 62) expression of IRDS (sum Z-score). Hierarchical clustering based on gene expression of IRDS in 491 TCGA prostate tumors. Red, upregulated expression. B, Boxplots representing the relationship between IRDS expression and the three groups. Grouping based on IRDS sum Z score statistics. C, High IRDS expression in prostate tumors is associated with decreased disease-free survival. Kaplan–Meier survival analysis. Log-rank test (P < 0.05) and unadjusted HR from a Cox regression analysis with Ptrend.
IRDS is associated with early disease recurrence in the TCGA prostate cancer cohort. A, Expression of 45 IRDS genes identifies prostate tumors with low (group 1: n = 159), medium (group 2: n = 270), and high (group 3: n = 62) expression of IRDS (sum Z-score). Hierarchical clustering based on gene expression of IRDS in 491 TCGA prostate tumors. Red, upregulated expression. B, Boxplots representing the relationship between IRDS expression and the three groups. Grouping based on IRDS sum Z score statistics. C, High IRDS expression in prostate tumors is associated with decreased disease-free survival. Kaplan–Meier survival analysis. Log-rank test (P < 0.05) and unadjusted HR from a Cox regression analysis with Ptrend.
IFNL4 rs368234815-ΔG is associated with IRDS and a distinct expression profile in prostate tumors
IFNλ4 is a recently discovered type III IFN that induces a similar set of ISGs that constitute IRDS, and impairs an antiviral response, as shown for HCV (20). Production of IFNλ4 is genetically controlled by a genetic variant, IFNL4 rs368234815-ΔG, which is the most common among subjects of African ancestry (20). Thus, we examined whether IFNL4 rs368234815-ΔG is associated with the occurrence of IRDS in 44 prostate tumors that had available information on IRDS and the IFNL4 rs368234815-ΔG/TT genotype. As shown in Table 1, homozygote carriers of the ΔG allele (ΔG/ΔG) were significantly more likely to have IRDS-positive tumors than carriers of the combined TT/TT and TT/ΔG genotypes. When we restricted the analysis to the 23 AA patients in this cohort, we again found a significant association between the ΔG/ΔG genotype and the occurrence of IRDS in prostate tumors (OR = 8.2; 95% CI, 1.1–60.4). Of note, only AA men were carriers of the ΔG/ΔG genotype among the 44 men.
IFNL4 rs368234815-ΔG allele is associated with occurrence of IRDS in prostate tumors
. | IFNL4 genotype, N (%) . | Fisher's exact test . | OR . | |
---|---|---|---|---|
All tumors, N = 44 | TT/TT or TT/ΔG | ΔG/ΔG | P | Adjusted OR (95% CI)a |
IRDS-negative | 23 (92%) | 2 (8%) | < 0.001 | 15.7 (2.7–90.6) |
IRDS-positive | 8 (42%) | 11 (58%) | ||
Only tumors from AA men, n = 23 | ||||
IRDS-negative | 6 (75%) | 2 (25%) | 0.04 | 8.2 (1.1–60.4) |
IRDS-positive | 4 (27%) | 11 (73%) |
. | IFNL4 genotype, N (%) . | Fisher's exact test . | OR . | |
---|---|---|---|---|
All tumors, N = 44 | TT/TT or TT/ΔG | ΔG/ΔG | P | Adjusted OR (95% CI)a |
IRDS-negative | 23 (92%) | 2 (8%) | < 0.001 | 15.7 (2.7–90.6) |
IRDS-positive | 8 (42%) | 11 (58%) | ||
Only tumors from AA men, n = 23 | ||||
IRDS-negative | 6 (75%) | 2 (25%) | 0.04 | 8.2 (1.1–60.4) |
IRDS-positive | 4 (27%) | 11 (73%) |
aAdjusted for age at diagnosis and pathological stage.
Further examination of the gene expression data revealed that the IFNL4 rs368234815-ΔG is associated with increased expression of immune-, host defense-, and inflammation-related genes in prostate tumors (Fig. 3). Gene expression differences based on the ΔG genotype separated tumors into three distinct clusters. Cluster 1 was significantly enriched for tumors from AA men and carriers of the ΔG/ΔG genotype and showed ΔG allele dosage increase of expression of genes such as PTPRC, TNF, RSAD2, IFI44, NLRP3, CCL5, STAT1, CCL4, IFI35, IRF9, ISG15, IFNG, and MX1, among others. Many of these are IRDS genes (e.g., IFI35, IFI44, MX1, and STAT1). This observation was further explored by a pathway analysis using GO biological processes, molecular function, and cellular component annotations. The approach identified GO terms such as immune response, defense response, and response to virus, as being associated with IFNL4 rs368234815-ΔG (FDR < 5%). This observation was corroborated by transient overexpression of IFNL4 in 22Rv1, PC-3, and MDA-PCa-2b human prostate cancer cells. Expression of IFNλ4 induced an IFN signature in these cells, consistent with IRDS, under these experimental conditions (Fig. 4A–C), although not affecting cell growth (Supplementary Fig. S1). The IFNL4-transfected cells showed induction of known ISGs that are common to both IRDS and IRG, such as IFIT1-3, IFIH1, IFITM1, OAS1, OAS2, MX1, IRF7, and IRF9, when compared with vector control–transfected cells. Furthermore, when we inhibited IFNλ4 signaling with blocking antibodies targeting either the IFNλ4 receptor, IL10R2, or IFNλ4, this signature was attenuated (Fig. 4A).
IFNL4 rs368234815-ΔG is associated with a distinct expression profile in prostate tumors (n = 44). Hierarchical clustering based on expression of 1,194 transcripts yields three distinct clusters and shows a significant association between IFNL4 rs368234815-ΔG and the pattern of gene expression (red, upregulated genes). IFNL4 rs368234815 genotypes are shown above the heatmap in red for ΔG/ΔG, green for TT/ΔG, and blue for TT/TT. Cluster 1 is significantly enriched for tumors from AA men and carriers of the ΔG/ΔG genotype. Frame marks 511 probesets, representing 447 unique genes, with signal enrichment in cluster 1. GO relationships for these genes are shown to the right and include immune response, defense response and response to virus (FDR < 5% for all associations). The 1,194 transcripts were selected because their expression was associated with ΔG allele (P < 0.05) using a linear regression model where relationships with ΔG were analyzed under an additive model (coded as 0, 1, and 2 ΔG alleles).
IFNL4 rs368234815-ΔG is associated with a distinct expression profile in prostate tumors (n = 44). Hierarchical clustering based on expression of 1,194 transcripts yields three distinct clusters and shows a significant association between IFNL4 rs368234815-ΔG and the pattern of gene expression (red, upregulated genes). IFNL4 rs368234815 genotypes are shown above the heatmap in red for ΔG/ΔG, green for TT/ΔG, and blue for TT/TT. Cluster 1 is significantly enriched for tumors from AA men and carriers of the ΔG/ΔG genotype. Frame marks 511 probesets, representing 447 unique genes, with signal enrichment in cluster 1. GO relationships for these genes are shown to the right and include immune response, defense response and response to virus (FDR < 5% for all associations). The 1,194 transcripts were selected because their expression was associated with ΔG allele (P < 0.05) using a linear regression model where relationships with ΔG were analyzed under an additive model (coded as 0, 1, and 2 ΔG alleles).
IFNL4 overexpression induces an IFN signature in vitro and IFNL4 rs368234815-ΔG associates with overall survival. IFNλ4 induces an IFN signature in the 22Rv1 (A), PC-3 (B), and MDA-PCa-2b (C) human prostate cancer cells. Shown are heatmaps of genes related to the human type I IFN response after transfection of cells with an IFNL4 expression construct. Red, upregulated expression. In A, blocking antibodies were added to the culture medium targeting the IL10R2 receptor and IFNλ4. *Significantly different gene expression induced by IFNL4 overexpression versus control. D, Association of IFNL4 rs368234815-ΔG with overall survival among AA patients with prostate cancer (n = 194). Carriers of the ΔG/ΔG genotype experienced the worst survival. Kaplan–Meier survival curve by IFNL4 genotype. Log-rank test: P < 0.05.
IFNL4 overexpression induces an IFN signature in vitro and IFNL4 rs368234815-ΔG associates with overall survival. IFNλ4 induces an IFN signature in the 22Rv1 (A), PC-3 (B), and MDA-PCa-2b (C) human prostate cancer cells. Shown are heatmaps of genes related to the human type I IFN response after transfection of cells with an IFNL4 expression construct. Red, upregulated expression. In A, blocking antibodies were added to the culture medium targeting the IL10R2 receptor and IFNλ4. *Significantly different gene expression induced by IFNL4 overexpression versus control. D, Association of IFNL4 rs368234815-ΔG with overall survival among AA patients with prostate cancer (n = 194). Carriers of the ΔG/ΔG genotype experienced the worst survival. Kaplan–Meier survival curve by IFNL4 genotype. Log-rank test: P < 0.05.
IFNL4 rs368234815-ΔG is associated with overall survival of AA patients with prostate cancer
To test whether IFNL4 rs368234815-ΔG may influence either disease-free or overall survival of patients with prostate cancer, we genotyped 194 AA prostatectomy patients from the Cleveland Clinic, as described in Materials and Methods. Twenty-five (12.8%) of the patients carried the TT/TT genotype, 92 (47.4%) the TT/ΔG genotype, and 77 (39.7%) were homozygote for ΔG, which corresponds to a 63.4% ΔG allele frequency. These genotypes did not associate with the Gleason score of tumors, nor did we find an association with disease recurrence (P = 0.88) in this cohort. However, the ΔG/ΔG genotype was associated with a significantly decreased overall survival of these patients (Fig. 4D). The Cox regression model showed that each copy of the ΔG allele increased the risk of death by about 2-fold (HR = 2.07; 95% CI, 1.14–3.75 adjusted for age, disease stage, Gleason score; Ptrend = 0.01) compared with carriers of the TT/TT genotype (Table 2).
IFNL4 rs368234815-ΔG allele is associated with decreased overall survival of AA patients with prostate cancer
Risk allele . | RAF, % . | Genotype . | N (%) . | Crude HR (95% CI) . | Adjusted HR (95% CI)a . |
---|---|---|---|---|---|
rs368234815-ΔG | 63.4 | TT/TT | 25 (12.9) | 1.00 (reference) | 1.00 (reference) |
TT/ΔG | 92 (47.4) | 1.47 (0.32–6.75) | 2.07(0.44–9.66) | ||
ΔG/ΔG | 77 (39.7) | 3.26 (0.75–14.15) | 4.28 (0.97–18.90) | ||
Total | 194 | ||||
per allele | — | 2.00 (1.09–3.67) | 2.07 (1.14–3.75) | ||
Ptrend | — | 0.02 | 0.01 | ||
ΔG/ΔG vs. TT/TT+TT/ΔG | — | 2.39 (1.14–5.01) | 2.21 (1.05–4.69) | ||
P | — | 0.02 | 0.02 |
Risk allele . | RAF, % . | Genotype . | N (%) . | Crude HR (95% CI) . | Adjusted HR (95% CI)a . |
---|---|---|---|---|---|
rs368234815-ΔG | 63.4 | TT/TT | 25 (12.9) | 1.00 (reference) | 1.00 (reference) |
TT/ΔG | 92 (47.4) | 1.47 (0.32–6.75) | 2.07(0.44–9.66) | ||
ΔG/ΔG | 77 (39.7) | 3.26 (0.75–14.15) | 4.28 (0.97–18.90) | ||
Total | 194 | ||||
per allele | — | 2.00 (1.09–3.67) | 2.07 (1.14–3.75) | ||
Ptrend | — | 0.02 | 0.01 | ||
ΔG/ΔG vs. TT/TT+TT/ΔG | — | 2.39 (1.14–5.01) | 2.21 (1.05–4.69) | ||
P | — | 0.02 | 0.02 |
NOTE: RAF = risk allele frequency in dataset (n = 194).
aMultivariable Cox regression model adjusted for age, disease stage (TNM), and Gleason score; Ptrend: ΔG/ΔG versus TT/ΔG versus TT/TT.
Discussion
In this study, we show that a subset of prostate tumors manifests a distinct IFN signature, IRDS, that has previously been linked to acquired resistance to radiation and chemotherapy, increased metastases, and poor survival of patients with breast cancer and glioblastoma (16–19, 29, 30). We show that in patients with prostate cancer, IRDS predicts decreased disease-free survival and is significantly more prevalent in tumors of AA than EA patients. Furthermore, we discovered that a germline variant in the IFNL4 gene was significantly associated with the increased prevalence of IRDS and all-cause mortality in AA prostatectomy patients, suggesting genetic predisposition to these clinical outcomes.
Upregulated IFN signaling through the JAK–STAT signaling is part of IRDS and the antiviral response (19), but is also triggered by DNA methyltransferase inhibitors (31) and occurs in various types of cancers (16, 30, 32, 33). In human triple-negative breast cancer, coinactivation of p53 and ARF coincides with an oncogenic IFNβ–STAT1–ISG15 signaling signature that is reminiscent of IRDS (33). However, in primary prostate tumors, the inactivation of p53 and ARF is rather uncommon (34, 35), indicating that the common occurrence of IRDS in primary tumors of AA patients is likely caused by another mechanism. Investigations using mouse models demonstrated that IFN signaling through STAT1 is protumorigenic in leukemia and colon cancer development through chronic inflammation–mediated carcinogenesis (36, 37). These findings are consistent with our previous reports that tumors of AA patients demonstrate a prominent immune-inflammation signature that may increase tumor aggressiveness and can be targeted with anti-inflammatory drugs (12, 26). In a study by Hardiman and colleagues (15), the effect of vitamin D supplementation on the prostate cancer transcriptome was investigated in 10 AA and 17 EA patients. In line with our findings, the authors found an immune-inflammation signature in tumors of AA patients but also showed that this signature is diminished by vitamin D treatment.
Besides increasing inflammation and disease progression, the immune-inflammation signature in AA prostate tumors may cause resistance to radiation and chemotherapy, and T-cell exhaustion. IRDS cooccurs as part of this immune-inflammation signature and has been shown to increase resistance to radiation (29), although inhibition of the JAK–STAT pathway was found to result in resensitization of docetaxel-resistant DU145 prostate cancer cells (38). Others found that the JAK–STAT signaling increases cross-resistance in myeloma cell lines (39). More recently, several investigations linked IRDS-like signatures to the clinical response to anti-CTLA4 therapy and PD-1 blockade (40, 41). Accordingly, these signatures cause immunosuppression in melanomas and resistance to anti-CTLA4 therapy but increase the response rate to PD-1 blockade. PD-L1 is an ISG and an important mediator of resistance to anti-CTLA4 therapy (40). PD-L1 expression may detrimentally affect AA patients with prostate cancer because of IRDS. Indeed, it was recently shown that expression of PD-L1 is increased in prostate tumors of AA men, when compared with EA men, using IHC (42). We did not find that PD-L1 was significantly upregulated at the transcript level in AA prostate tumors in the Wallace and colleagues and TCGA datasets, suggesting that IRDS may regulate protein stability rather than expression of PD-L1 in these tumors. However, we found that IDO1, another ISG, is upregulated on the mRNA level in AA prostate tumors, consistent with an immunosuppressive environment in these tumors. Additional measurements of plasma tryptophan levels in our study suggest that AA patients with prostate cancer may have a higher turnover of this IDO1 substrate, although the reduction of plasma tryptophan was rather modest.
Currently, there is no firm evidence that a viral infection is a cause of prostate cancer (43), although a meta-analysis of 26 tissue-based case–control studies reported an association between human papillomavirus positivity and the disease (44). In addition, RNASEL, a gene that protects against viral pathogens, may have an important tumor suppressor function in prostate cancer (45, 46). Besides viral infections, spontaneous reactivation of endogenous retroviruses may also lead to occurrence of IRDS. We reported the upregulation of the human endogenous retrovirus type K envelope protein in prostate tumors of AA patients (47). Moreover, it was recently shown that treatment with DNA methyltransferase inhibitors can lead to reactivation of endogenous retroviruses and an IRDS-like signature in patients with melanoma that increases the sensitivity to immune checkpoint therapy (31). Western diet may also upregulate IFN signaling (48).
With this study, we provide the first evidence that IFNL4 rs368234815-ΔG could be a predisposition factor for IRDS in patients with prostate cancer. Although this effect is not exclusive to AA men, but based on a much higher ΔG allele frequency in AA compared with EA patients (e.g., 62% in AA vs. 34% in EA in the NCI-Maryland Prostate Cancer Case–Control Study), this genetic predisposition would be most important for men of African ancestry. This stark difference in the ΔG allele frequency between AA and EA patients could account for some of the known health disparity in outcomes of prostate cancer. IFNλ4 protein is only produced in individuals with IFNL4 rs368234815-ΔG and attenuates antiviral responses through negative regulation of IFN signaling (21). We showed that IFNλ4 induces an IFN signature in three human prostate cancer cell lines, but could not detect IFNL4 expression in human prostate tumors based on TCGA RNA-seq data. IFNL4 expression is known to be low and transient, but IFNλ4 is potent even at very low levels (21, 49). Finally, IFNL4 rs368234815-ΔG was associated with overall survival of AA prostatectomy patients, indicating a broader biological effect of this variant. We also found similar association with overall survival of AA prostatectomy patients for an intronic IFNL4 variant, rs12979860-T, which is in high linkage disequilibrium with rs368234815-ΔG (Supplementary Fig. S3). However, in contrast to IRDS, we did not find an association of IFNL4 rs368234815-ΔG with decreased disease-free survival. Perhaps, our analysis of 194 genotyped AA prostatectomy patients was underpowered to detect this association, or the impact of IFNL4 rs368234815-ΔG on disease recurrence is weaker than the effect of IRDS. Alternatively, IFNL4 rs368234815-ΔG may mainly affect outcomes of late-stage therapies such as chemotherapy and immunotherapy, or may have a detrimental impact on morbidities arising from long-term androgen ablation therapy, explaining the observed association of this genotype with overall survival.
Thus, future research is needed to define the role of IFNλ4 in IRDS among patients with prostate cancer. Yet, based on our findings, IFNL4 rs368234815-ΔG may have potential clinical use for identifying patients with prostate cancer with increased sensitivity to immune checkpoint blockade therapy and thereby predicting disease outcomes.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: W. Tang, C.A. Loffredo, R.H. Silverman, G.R. Stark, S. Ambs
Development of methodology: W. Tang, T.A. Wallace, C. Magi-Galluzzi, S.V. Jordan, L. Prokunina-Olsson
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Tang, C. Magi-Galluzzi, T.H. Dorsey, O.O. Onabajo, A. Obajemu, C.A. Loffredo, E.A. Klein
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Tang, M. Yi, C.A. Loffredo, R.M. Stephens, L. Prokunina-Olsson, S. Ambs
Writing, review, and/or revision of the manuscript: W. Tang, T.A. Wallace, M. Yi, C. Magi-Galluzzi, S.V. Jordan, C.A. Loffredo, R.H. Silverman, G.R. Stark, E.A. Klein, L. Prokunina-Olsson, S. Ambs
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Tang, T.H. Dorsey
Study supervision: S. Ambs
Acknowledgments
This research was supported by the Intramural Research Program of the Center for Cancer Research (ZIA BC010499 and ZIA BC010624, to S. Ambs) and Division of Cancer Epidemiology and Genetics (ZIA CP010201, to L. Prokunina-Olsson), NCI, NIH, the Prostate Cancer Foundation 2013 Ben Franklin-PCF Special Challenge Award (G.R. Stark, S. Ambs, and E.A. Klein), the Cleveland Clinic Prostate Cancer Center of Excellence Award (to R.H. Silverman, G.R. Stark, C. Magi-Galluzzi, and E.A. Klein), and the National Institute of Allergy and Infectious Diseases (NIAID), NIH grant R01AI135922 (to R.H. Silverman). We thank the Cooperative Prostate Cancer Tissue Resource (CPCTR) for providing tissue specimens and supporting data. We would also like to thank personnel at the University of Maryland and the Baltimore Veterans Administration Hospital for their contributions with the recruitment of subjects.
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.