Distinct Genomic Alterations in Prostate Tumors Derived from African American Men

This article is featured in Highlights of This Issue, p. 1757

Somatically acquired genetic alterations underlying higher incidence and mortality rates of prostate cancer in African American (AA) or black men compared with European American (EA) or white men, are still mostly unknown, although significant research efforts have been devoted (1–8). Most prostate cancers are considered to be indolent (nonaggressive) tumors that may not even require treatment. However, some of them are aggressive malignancies that are characterized by uncontrolled cell proliferation resulting in cancer progression, recurrence, and metastases, leading to more than 33,000 deaths, including 5,350 AA/black men in the United States annually (9, 10). Prostate cancer–specific death is primarily caused by metastasis of the cancer cells harboring driver genetic alterations that can be traced back to tumors in the prostate (11–13). Therefore, comprehensively analyzing genetic alterations in primary tumors may shed light on the mechanism responsible for cancer progression and metastasis that result in prostate cancer–specific death. Comparing genetic alterations in tumors between AA/black and EA/white men may lead to the identification of distinct aberrations that cause higher rates of mortality in AA/black men. However, extensive genome-wide analyses of somatically acquired genetic alterations in primary prostate cancer have been carried out using tumor DNA primarily from EA/white men in large cohorts (14–17), in addition to numerous pioneering studies with fewer subjects. Furthermore, deeper next-generation sequencing (NGS) of targeted variants/genes has revealed new insights on clonal expansion that may lead to cancer progression over time (13, 18, 19). In contrast, the vast majority of a limited number of genome-wide studies on somatically acquired genetic alterations in prostate cancer of AA/black men either used small cohorts and/or low-resolution approaches (1–7) or used a shallow depth or coverage (8). Together with very complex intratumor and intertumor heterogeneity in prostate cancer (12, 13, 19–21), the insufficient number of specimens has hindered the identification of somatically acquired genetic factors responsible for more aggressive prostate cancer and higher mortality rates in AA/black men.

In this work using targeted NGS with an average depth of 2,725 reads for normal and 2,522 reads for tumor DNA and the OncoScan copy-number variation (CNV) FFPE assay with more than 220,000 SNP probes across the whole genome, we first analyzed tumor and matched normal DNA from 82 AA patients. We then combine the data of somatic mutations and CNAs from additional five independent cohorts of genome-wide analysis in the public domain and analyzed somatic nucleotide mutations and CNAs in prostate cancer of a total of 171 AA/black men by comparing them with those in prostate cancer of 860 EA/white men. We highlight several genes with a higher frequency of damaging mutations or distinct patterns of CNAs that appear to contribute to aggressive features of prostate cancer in AA/black men.

All parts of this retrospective study were performed following Institution Review Board (IRB) approval. Archival formalin-fixed paraffin-embedded (FFPE) pathologic specimens of tumor and matched benign tissues from 83 AA patients with prostate cancer were obtained from the Prostate Cancer Biorepository Network (PCBN). Gleason grade (GG) and tumor–node–metastasis stages of the tumor specimens were provided by PCBN. These include one Gleason 5, 24 Gleason 6, 32 Gleason 7, five Gleason 8, and 21 Gleason 9 tumors. Detailed clinical and pathologic data for each of the cases are presented in Supplementary Table S1.

We used the GeneRead DNA FFPE Kit from Qiagen for genomic DNA isolation following manufacturer's instructions with minor modifications. Briefly, we first centrifuged the tissue specimens to the bottom of a 1.5 mL tube and added160 μL of deparaffinization solution to remove the wax. After adding 55 μL of nuclease-free water, 25 μL of buffer FTB, and 20 μL of proteinase K, the samples were incubated at 56°C for one hour or until the tissues were completely dissolved before incubating at 90°C for another hour. We then transferred the lower clear phase into a new 1.5 mL tube and added 115 μL of nuclease-free water and 35 μL of Uracil-N-Glycosilase (UNG) to remove artificially induced uracils by FFPE. After adding and mixing 2 μL of RNAse A, 250 μL of buffer AL, and 250 μL of ethanol, we transferred the lysate to the QIAamp MinElute column to bind DNA and to remove residual contaminants using AW1 and AW2 buffers and ethanol. Genomic DNA was eluted two times in ATE buffer with 20 μL each before the concentration was measured on a Qubit 3.0 Fluorometer. DNA quality was assessed using gel electrophoresis. All of the 82 pairs of samples except one that did not meet the quality control assessment were used for downstream genomic analysis.

To enable deeper NGS at a reasonable cost and still uncover the fundamental somatic changes, we identified 39 potential driver genes either with significant somatic mutations in prostatic cancer or loss-of-function leading to genomic instability (8, 13, 17–19, 22–27). Two customized panels with the probes manufactured by Agilent and Roche NimbleGen, respectively, were adopted to capture the exons of 39 selected genes for the Illumina paired-end NGS to detect somatically acquired mutations. Specifically, we first used 300 to 500 ng of genomic DNA from each of the samples and either the HaloPlex HS custom kits from Agilent or the SeqCap EZ Library SR kits from Roche for capturing targeted sequences and library preparation following manufacturer's protocols. We then performed the paired-end sequencing of 2 × 150 bp on an Illumina NextSeq500 according to manufacturer's instructions.

We employed DNA-seq analysis pipeline MuTect2 for identification of somatically acquired single-nucleotide mutations and small insertions/deletions and Genome Analysis Toolkit for germline variants with potentially damaging consequence. Briefly, paired-end reads were first aligned to the hg19 version of the human genome using Burrows-Wheeler Aligner v0.7 to generate BAM files. After sorting the BAM files using Samtools, PCR duplicates were marked using Picard, and realignment around putative gaps was performed using the Genome Analysis Toolkit (GATK) v3.2–2. Variant calling was performed by the GATK Haplotype caller. ANNOVAR (http://annovar.openbioinformatics.org/en/latest) and snpEff were used for annotating variants and for retrieving information on variants in the population-based studies such as the 1000 Genomes Project (www.1000genomes.org), NHLBI-ESP 6500 exomes, or ExAC (http://exac.broadinstitute.org/) or gnomAD (http://gnomad.broadinstitute.org/), and clinical databases such as the Human Gene Mutation Database and ClinVar. Pathogenicity of variants is defined based on American College of Medical Genetics and Genomics criteria. Specifically, pathogenic and likely pathogenic mutations are defined as (i) all protein truncating mutations unless their allele frequency is 5% or higher in any racial group in population databases or is reported as benign or likely benign in the ClinVar, and (ii) nonsynonymous changes if their allele frequency is less than 5% and reported as pathogenic and likely pathogenic mutations in the ClinVar.

We successfully sequenced a total of 81 matched tumor–normal pairs (Fig. 1). After quality control assessment, we removed four pairs with a maximum depth of less than 50 reads for normal DNA. To minimize the effect of potential false positives from deeper NGS, we next used the normal DNA in each pair as quality control and eliminated 25 mutations where the percentage of alternative alleles in normal DNA was equal or larger than the percentage of alternative alleles in matched tumor DNA. This resulted in a total of 8,800 mutations in the remaining 77 pairs of samples (Supplementary Table S2), including 22 enriched by the Agilent SureSelect with an average and a median depth of 670 and 301 reads, respectively, for tumor DNA, and 55 via the NimbleGen SeqCap with an average and a median depth 2,573 and 2,264 reads, respectively, for tumor DNA (Supplementary Table S3).

Detection of genome-wide CNAs in the tumors from 80 AA men was carried out using the Affymetrix OncoScan FFPE assay kit following manufacture's protocols provided by Affymetrix Thermo Fisher Scientific OncoScan CNV FFPE Services. We then imported the CEL files into the Affymetrix Chromosome Analysis Suite for quality control assessment and for generating OSCHP files. We next imported the OSCHP files with TuScan algorithm into the Nexus 10.0 software from BioDiscovery for CNAs and loss of heterozygosity analyses using default settings.

We used multiplex ligation-dependent probe amplification (MLPA) to evaluate the performance of OncoScan CNV FFPE assay by validating CNAs identified by OncoScan in nine selected genes. Briefly, five pairs of specific probes for each gene were designed according to the UCSC Genome Brower hg38 and synthesized by Integrated DNA Technologies. A unique set of all-synthetic probes was first hybridized to about 2 ng of FFPE DNA from each sample. Specific pairs of probes were then ligated to make a PCR template for high specificity. A set of fluorescent-labeled primers was next used for amplification of the templates. The PCR products were finally separated and quantified using the ABI-3500xL Genetic Analyzer. Quality control assurance was carried out by gel electrophoresis, the GeneMapper Software 5 and the Coffalyser software. CNAs were determined using a combination of the Coffalyser software and an in-house computer script. We achieved a concordance of 96% between MLPA and OncoScan for CNAs in nine genes among four different FFPE tumor–normal pairs.

It is well known that prostate tumors harbor heterogeneous intertumor and intratumor mutations in a very complex way. Although some of them occur only in a small number of cells at the beginning of tumorigenesis, the cells with driver mutations may evolve over time and eventually contribute to aggressive features of prostate cancer. To assess the spectrum of these mutations and their potential effects, we selected 39 genes of interest for prostate cancer (Fig. 2). Two customized enrichment panels with average depths of 670 and 2,573 reads for tumor DNA from Agilent and Roche, respectively, were used to successfully sequence tumor and matched normal DNA from a total of 81 AA patients, with 77 passing quality control assurances (Fig. 1; Supplementary Table S2). Among the 22 tumors sequenced using the Agilent HaloPlex HS enrichment, we identified a total of 233 mutations, with an average of 10.6 exon mutations per tumor. On the other hand, we detected a total of 8,567 exon mutations among the 55 tumors sequenced using the NimbleGen SeqCap EZ Library SR enrichment from Roche, with an average of 156 mutations per tumor in these 39 genes. Combining the results from these two platforms together, we found that the percentages of tumors harboring somatically acquired mutations were much higher than those expected in these genes, even after filtering out the ones with ≤ 1% of the mutated copies in the tumor (Fig. 2). In addition, more than 35% of AA prostate cancer harbored potentially damaging (frameshift and nonsense) mutations in their tumors for each of the following genes including APC, ATM, BRCA2, KDM6A, KMT2C, KMT2D, MED12, ZFHX3, and ZMYM3 (Fig. 2A; Supplementary Table S4). These frequencies were much higher than those reported in the Genomic Data Commons (GDC, https://portal.gdc.cancer.gov/) for prostate cancer (Fig. 2B).

While it was unknown whether any of these 39 mutated genes affected tumor phenotypes when the percentage of mutated copies in the tumor was low, an arbitrary threshold based on the depth of an alternative allele was set to filter out the tumors containing ≤ 10% mutated copies in any of the 39 genes. Using this standard, we identified the top 22 somatically mutated genes in these 77 AA men. As shown in Fig. 3A and B, our analysis revealed much higher mutation rates in the exons of ZMYM3 and FOXA1 in prostate cancer in AA men than in prostate cancer in EA/white men. Nine (11.7%) of 77 AA men harbored ZMYM3 exon mutations with most of them being either frameshift or nonsense, versus only 2.7% of EA/white men. In contrast, the FOXA 1 exon mutations identified in nine AA men (11.7% vs. 5.4% in EA/white prostate cancer) were all nonsynonymous. In addition, only 3 (3.9%) and 2 (2.6%) AA/black patients with prostate cancer in our PCBN AA cohort harbored exon mutations in SPOP and TP53, respectively, the most commonly mutated genes in EA/white patients with prostate cancer (Fig. 2B), and all were nonsynonymous (Fig. 3A).

To comprehensively search for somatically acquired genetic changes that may contribute to aggressive characteristics of prostate cancer in AA/black men, we also performed an analysis of CNAs and allelic imbalances across the whole genome using the OncoScan FFPE SNP CNV array. As shown on the CNA landscape of prostate cancer from 80 AA men in the PCBN cohort (Fig. 3C, bottom), the most common CNA was the deletion of chromosomal 6q including MAP3K7 with about 49% subjects affected; this is in contrast to the most frequent deletion of 8p including BNIP3L in EA/white men (17, 28). On the other hand, similar to what was observed in EA/white men, deletion of 13q including RB1 was among the top three most frequent genetic losses in prostate cancer of AA men with about 35% of patients affected.

Analyzing the distribution of mutations in ZMYM3 among three cohorts, including those reported by Huang and colleagues (8) and in GDC, we found that AA/black men harbored more potentially damaging mutations (77%, frameshift and nonsense) than EA/white men (58%), with most of them clustering in the 5′ portion of the gene (Fig. 4A, top). Further analyzing the data among the 160 AA/black subjects in the cohorts of GDC and Huang and colleagues (8) we found only 1.9% (three, marked by purple arrows) frameshift and nonsense mutations that were similar to 1.7% (seven) among 410 EA/white men. Therefore, the higher percentage of potentially damaging mutations of ZMYM3 was likely the result of deeper NGS used in our current analysis. In addition, prostate cancer in AA/black men apparently had much more nonsense mutations than prostate cancer in EA/white men.

It has been reported that ZMYM3 promotes BRCA1 to the location at damaged chromatin to facilitate DNA repair in U2OS cells (27). To explore a possible relationship between ZMYM3 mutation and genomic integrity in prostate cancer, we analyzed genome-wide CNAs and allelic imbalances in a tumor that harbored more than 96% frameshifted copies of ZMYM3 (Fig. 4A, marked by a pink oval circle). The results revealed 10 chromosomes with allelic imbalances (Fig. 4A, bottom right, subject 76659). These included losses of chromosomes 10, 16, 17, 18, and 21, and gains of chromosomes 8, 11, 13, and 20 (Fig. 4A, bottom left, subject 76659). One copy each of chromosomes 2 and 13 was apparently lost while the other copy was amplified resulting in allele imbalance without DNA copy-number change on chromosome 2. The only other gene with more than 96% mutated copies (pI391M) that we found was PIK3CA in two tumors (99.63% and 99.67%; Supplementary Table S2). In contrast, we found no somatically acquired CNAs and allelic imbalance in these two tumors (Supplementary Fig. S1). On the other hand, similar to ZMYM3, most of the potentially damaging mutations occurred on the 5′ portion of CDK12, although we did not find any nonsense mutations in AA/black men (Fig. 4B). One of the tumors from an AA man harbored two frameshift mutations p.607 and p.987 (marked by pink oval circles) with the mutated copies of CDK12 taking up 29% and 65%, respectively, of all the CDK12 genes (Supplementary Table S2). Further analysis of genome-wide CNAs and B-allele frequency revealed numerous DNA copy-number gains as previously reported in EA/white men and complex allelic imbalances across the whole genome in the subject 75553 (Fig. 4B, bottom panels).

To further compare the frequency of CNAs in prostate cancer of AA/black and EA/white men, we also depicted the CNA landscape of prostate cancer from 393 EA/white men in the The Cancer Genome Atlas (TCGA) cohort with a total of 498 subjects using the Nexus 10.0 software (Fig. 3C, top). We first focused on the 13 most frequently altered regions that are represented by the genes including LRP1B, RYBP, CHD1, MAP3K7, BNIP3L, MYC, CDKN2A/B, PTEN, CDKN1B, USP10, TP53, SERPINB5, and TMPRSS2-ERG (T-E), each residing in the most commonly affected region on 13 different chromosomes in prostate cancer (17, 28). It appeared that prostate cancer in AA and EA/white men share similar patterns of CNAs on these 13 chromosomes (Fig. 3C).

We then analyzed CNAs of these genes in AA/black subjects from five different cohorts in public databases (Supplementary Table S5) and added this data into our analysis, which resulted in a total of 171 EA/black and 860 EA/white men being included in current study. As shown in Supplementary Table S6, AA/black men apparently harbored a higher rate of MAP3k7 deletion at 6q15 (P = 0.0247), and lower rates of losses of PTEN at 10q23.31 (P < 0.0001), TP53 at 17p13.1 (P = 0.0002), TMPRSS2-ERG at 21q22.2–22.3 (P < 0.0001), and SERPINB5 at 18q21.33 (P = 0.009), than EA/white men. Stratifying based on Gleason sum, we found no significant difference in the frequency of CNAs at all 13 genes (regions) in Gleason ≤ 6 tumors between AA/black and EA/white men. However, AA/black men apparently harbored higher rates of MYC gain in Gleason 7 and 8 tumors (P = 0.0142, P = 0.021, respectively), LRP1B loss in Gleason 8 tumors (P = 0.0163), and deletions of MAP3K7, BNIP3L, and RB1 in Gleason ≥ 9 tumors (P = 0.0019, P = 0.0036, P = 0.0274, respectively) than EA/white men (Fig. 5A). On the other hand, the frequencies of deletions at RYBP and TMPRSS2-ERG in Gleason 7 tumors (P = 0.0314, P = 0.0035, respectively), and loss at TP53 in Gleason 7 and Gleason ≥ 9 tumors (P = 0.0207, P = 0.0006, respectively) in AA/black men was significantly lower in AA/black than in EA/white men.

We next compared the rest of the prostate cancer genomes between AA and EA/white men and discovered four regions of deletion (Fig. 3C, marked by green oval circles) in the PCBN AA cohort, which were not observed in prostate cancer of EA/white men in the TCGA cohort. We further analyzed the CNAs in the subjects harboring these deletions and identified a minimum overlap region (MOR) each at 2p21, 4q34.3, 7q31.31, and 15q15.1 on chromosomes 2, 4, 7, and 15, respectively. For comparison purposes, we chose a potentially relevant gene(s) in the MOR on each of these chromosomes, including THADA, NEIL3, TES, CAV1, CAV2, ING3, BMF, and BUB1B, to represent each of the four regions as shown in Fig. 5B. To investigate whether these genes were also deleted in the tumors of other prostate cancer cohorts, we finally analyzed genome-wide CNAs data published using the cohorts of Johns Hopkins Hospital (JHH; Baltimore, MD; ref. 28), Memorial Sloan Kettering Cancer Center (MSKCC, New York, NY; ref. 17), TCGA (15), Sweden (16), and dbgap (ref. 14; Supplementary Table S4). As shown in Fig. 5C, among the tumor genomes of 860 EA/white men 4.9 %, 4.7%, 2.9%, and 6.9% harbored deletions of THADA, NEIL3, ING3, and BUB1B, respectively, versus 8.8%, 8.2%, 4.1%, and 9.4% of AA/black men with deletions of these genes, respectively. Further analysis revealed that the frequencies of THADA and NEIL3 deletions in Gleason 7 tumors and BUB1B loss in Gleason 8 tumors in AA/black men were significantly higher than those in EA/white men based on Fisher exact test (Fig. 5A).

To explore the possible implications of higher frequencies of these CNAs in prostate cancer of AA/black men, we next investigated whether more frequent deletions of LRP1B, MAKP3K7, BNIP3L, RB1, THADA, BUB1B, and NEIL3, and higher rates of gain of MYC in prostate cancer of AA/black men than those in EA/white men were each correlated with the most commonly used clinical parameters, Gleason grade, and pathologic stage. As shown in Table 1, only deletions of LRP1B and BUB1B each were not associated with pathologic stage of prostate cancer at a significant level of 0.05 in AA/black men. Compared with tumors with GG ≤ 2, the rates of deletions of LRP1B at 2q22.1–2 (P = 4.37E-05), BNIP3L at 8p21.2 (P = 3.37E-04), BUB1B at 15q15.1 (P = 0.001), and NEIL3 at 4q34.3 (P = 0.0086) were each significantly higher in tumors with GG ≥ 3. More importantly, deletion of MAP3K7 or RB1 or gain of MYC was significantly associated with both higher GG and more advanced pathologic stage with OR (95% confidence interval) = 2.72 (1.45–5.11), P = 0.002; OR = 3.36 (1.75–6.46), P = 2.76E-4; OR = 2.84 (1.39–5.81), P = 0.004; respectively. These results suggest that either the pathways involved in MAP3K7, MYC, and/or RB1 or other genes in vicinity that are coaltered with these three genes play a more important role for aggressive prostate cancer in AA/black men than the genes on other chromosomes.

Numerous studies on somatically acquired mutations in prostate cancer using NGS have revealed a number of potential driver mutations, particularly in the patients of European ancestry (14, 18, 22, 24, 25, 29–34). It is well established that prostate tumors are very heterogeneous genetically (19–21, 35, 36) and evolve over time (11–13, 19, 37–39). Therefore, genome-wide NGS at shallow depths for primary tumors may miss some important somatically acquired mutations, especially for those occurring at later stages in a tumor cell clone within the overall tumor mass. To uncover a better picture for the most common mutations in AA prostate cancer, we used two exon capture platforms for deeper NGS with an average depth of 2,522 reads for 77 tumors and screened somatically acquired exon mutations in 39 genes. We apparently achieved a better coverage using the SeqCap EZ Library SR kits from Roche than the HaloPlex HS custom kits from Agilent (Supplementary Table S3). More importantly, deeper NGS revealed a much higher frequency of somatically acquired mutations than previously reported in prostate cancer (40), with more than 35% of AA men harboring frameshift and/or nonsense mutations in a number of cancer driver genes including APC, ATM, BRCA2, KDM6A, KMT2C, KMT2D, MED12, ZFHX3, and ZMYM3. Overall, although most of the mutated copies are present at a very low percentage in a specific tumor (Supplementary Table S2), cells harboring a particular driver mutation(s) resulting in a higher growth rate may multiply more quickly and eventually become more predominant in the tumor over time. Analyzing the phenotypical effect in tumors harboring a high percentage of mutated copies of these genes may shed light on a de novo mechanism caused by the damaging driver mutations in the progression of human prostate cancer.

Among the 77 AA tumors, only one tumor harbored more than 95% of the mutated copies with damaging consequence in any of the 39 genes that we sequenced. As shown in Supplementary Table S2, about 96% of ZMYM3 copies were frameshifted in T766659 with a Gleason sum of 8, suggesting the mutation occurred at an early stage of this tumor's development. ZMYM3 located at Xq13.1 encodes a component of histone deacetylase complexes that is involved in chromatin remodeling. It has been reported that ZMYM3 promotes BRCA1 to the location of damaged chromatin to facilitate DNA repair (27). Knockout of ZMYM3 using CRISPR–Cas9 in mice and in U2OS cells, respectively, causes defects in spindle assembly and impairs homologous recombination DNA repair leading to genomic instability (26, 27). Therefore, the imbalanced allelic frequencies on 10 different chromosomes in this AA tumor, including five deletions and four gains in net DNA copy are likely the consequence of a defective ZMYM3.

In the PCBN AA prostate cancer cohort, we found only two other tumors harboring a gene with a high percentage of mutated copies in any of the 39 sequenced genes. Both of the tumors, 44347 and 44486, harbored a p.I391M missense mutation in PIK3CA with 99.6% of copies having mutated. However, we observed no apparent allelic imbalance or DNA CNAs, suggesting no effects on genomic structure or instability from this mutation of the driver gene. On the other hand, similar to previously reported findings in prostate cancer of EA/white men mostly with castration-resistant prostate cancer (CRPC; refs. 32–34, 41), two frameshift mutations, p.607 (29%) and p.987 (65%), resulted in biallelic inactivation of CDK12 leading to genome-wide DNA copy-number gains and massive allelic imbalance in one Gleason 8 tumor of an AA man. Therefore, biallelic inactivation of CDK12, similar to that observed in CRPC, may also contribute to more aggressive prostate cancer in AA patients with primary tumors.

As a major group of somatically acquired genetic aberrations, genome-wide CNAs and their frequencies have been investigated mostly in small AA/black prostate cancer cohorts (1–7, 42) and compared with those in EA/white prostate cancer. Moreover, using exome NGS to analyze 102 AA prostate cancer, Huang and colleagues reported a more frequent gain of FASN at 17q25 along with a number of CNA loci with lower frequency in AA men than those in the TCGA cohort (8). However, we did not detect a more frequent gain of FASN in our AA cohort. A new research article reporting an association between CDNK1B deletions and increased risk of metastasis in AA men (43) was published while revising this manuscript. With a frequency of 6.3%, the deletion of CDNK1B is apparently lower than the 21.1% that we observed in our study with no difference between AA/black and EA/white men (Supplementary Table S6). In addition, SPOP was mutated at 11.2% in their cohort, in contrast to the 3.9% we observed in AA men. The discrepancy is likely derived from relatively small AA cohorts and different platforms used in these analyses.

To minimize the effects of small cohorts with various types of heterogeneity on the frequency of CNAs, we analyzed genome-wide CNAs in five independent cohorts with a total of 171 AA/black and 860 EA/white men. In addition to the confirmation of less frequent deletion leading to TMPRSS2-ERG fusion previously reported by others, we also identified two more loci with significantly lower frequencies of CNAs and eight loci with significantly higher frequencies of CNAs in AA/black than in EA/white men. These include 12% and 11% deletions of RYBP and TP53, respectively, in Gleason 7 tumors, and 15% deletions of TP53 in Gleason ≥ 9 tumors in prostate cancer of AA/black men versus 21%, 20%, and 50%, respectively, in EA/white men. Apparently, it is hard to imagine that less frequent deletions of these genes can contribute to more aggressive features of prostate cancer in AA/black men. However, significantly more frequent deletions in the regions of 2p22.2, 4q34.3, 2q22.1–2, 15q15.1, 6q15, 8p21.2, and 13q14.2, and significantly more frequent gains in the region of 8q24.21 in AA/black than in EA/white men may lead to more aggressive prostate cancer characteristics. These more frequent deletions include the genes THADA (9.8% vs. 3.9%), and NEIL3 (9.8% vs. 3.7%), in Gleason 7 Tumors, LRP1B (44.4% vs. 10.2%) and BUB1B (33.3% vs. 4.1%) in Gleason 8 Tumors, and MAP3K7 (73.1% vs. 40.6%), BNIP3L (65.4% vs. 35.1%), and RB1 (73.1% vs. 50.9%) in Gleason 9 Tumors. The more frequent gains in AA/black versus EA/white men affect MYC with 22.8% versus 12.9% in Gleason 7 tumors, and 55.6% versus 18.4% in Gleason 8 tumors, respectively (Fig. 5A). As shown in Fig. 3C, the deletion represented by MAP3K7 affects a large genomic region on chromosomal 6q, which also includes the gene AIM1. Depletion of AIM1 has been demonstrated to promote prostate epithelial cell migration, invasion, and metastatic dissemination (44). The chromosomal translocations of the thyroid adenoma associated (THADA) gene has been reported in thyroid benign tumors and acute myeloid leukemia (45–47). The protein encoded by NEIL3 is involved in repair of damaged DNA caused by reactive oxygen species (48). More importantly, genomic deletion of MAP3K7 has been demonstrated to be associated with high-grade and early biochemical recurrence of prostate cancer (49, 50).

To further identify somatically acquired genetic changes that may contribute to more aggressive features of prostate cancer in AA/black men, we analyzed the associations between two clinical parameters (GG and pathologic stage) and CNAs with higher frequencies in AA/black than in EA/white men. The CNAs that associate with both higher GG and more advanced stage are deletions of the regions including MAP3K7 and RB1 and gain of MYC. In addition, the loss of BNIP3L, THADA, or NEIL3 is associated with more advanced pathologic stage. Therefore, from a population point of view, more frequent deletions of regions involving MAP3K7, RB1, BNIP3L, THADA, and NEIL3 and more frequent gains of the genomic region including MYC are more likely to result in more aggressive features of prostate cancer observed in AA/black than in EA/white men when compared with CNAs in other genomic regions.

In summary, the distinct CNAs described above along with more frequent damaging mutations in ZMYM3 that may lead to genomic instability all seemingly contribute to more aggressive prostate cancer and possibly higher mortality rates from the disease in AA men.

No potential conflicts of interest were disclosed.

W. Liu: Conceptualization, resources, data curation, formal analysis, supervision, methodology, writing-original draft. S.L. Zheng: Conceptualization, data curation, formal analysis, supervision, methodology, writing-original draft, writing-review and editing. R. Na: Formal analysis. L. Wei: Formal analysis, methodology. J. Sun: Formal analysis, methodology. J. Gallagher: Data curation. J. Wei: Formal analysis. W.K. Resurreccion: Writing-review and editing. S. Ernst: Data curation. K.S. Sfanos: Resources, data curation, writing-review and editing. W.B. Isaacs: Resources, writing-review and editing. J. Xu: Conceptualization, resources, supervision, writing-review and editing.

The authors thank Sameep Shah for technical support. This work is supported by Department of Defense grants W81XWH-14-1-0303 (to J. Xu.) and W81XWH-12-1-0188 (to W. Liu); by the Department of Defense Prostate Cancer Research Program, Award No. W81XWH-18-2-0013, W81XWH-18-2-0015, W81XWH-18-2-0016, W81XWH-18-2-0017, and W81XWH-18-2-0018 Prostate Cancer Biorepository Network (PCBN).

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.