Bridging Health Disparities: a Genomics and Transcriptomics Analysis by Race in Prostate Cancer

Abstract

As the era of cancer genomics expands, disproportionate rates of prostate cancer incidence and mortality by race have demonstrated increasing relevance in clinical settings. While Black men are most particularly affected, as data has historically shown, the opposite is observed for Asian men, thus creating a basis for exploring genomic pathways potentially involved in mediating these opposing trends. Studies on racial differences are limited by sample size, but recent expanding collaborations between research institutions may improve these imbalances to enhance investigations on health disparities from the genomics front. In this study, we performed a race genomics analysis using GENIE v11, released in January 2022, to investigate mutation and copy number frequencies of select genes in both primary and metastatic patient tumor samples. Further, we investigate the TCGA race cohort to conduct an ancestry analysis and to identify differentially expressed genes highly upregulated in one race and subsequently downregulated in another. Our findings highlight pathway-oriented genetic mutation frequencies characterized by race, and further, we identify candidate gene transcripts that have differential expression between Black and Asian men.

Introduction

The growing number of recent investigations at the forefront of health disparities has brought to attention the added complexity of race on disease prevalence in certain cancers. Among these, prostate cancer has disparate mortalities among patients of different races, thus highlighting an area of need to further investigate these implications on treatment outcomes and drug resistance. Nearly 200 susceptible genes in prostate cancer are risk variants among patients of European ancestry. [1] While such discoveries are of great importance, incidence rates from 2014 to 2018, as presented by the American Cancer Society, stood at 172 per 100,000 cases among Black men in comparison to 99 and 55 among White and Asian men, respectively. [2] These statistics show mortality rates of 37 per 100,000 cases in Black men, 17 in White men, and 8 in Asian men. Findings are yet to be wholly defined, but the literature suggests an interplay of biological, clinical, and socioeconomic contributions.

On the contrary, the American Cancer Society’s average incidence and mortality rates from 2008 to 2012 remained 50% lower among Asian Americans and Pacific Islanders in the United States in comparison to non-Hispanic Whites. [3] Such findings speak for the lack of genomics investigations within this demographic despite having more late-stage diagnoses similar to Black prostate cancer patients. [4] Although there are few U.S. studies on the Asian race in terms of prostate cancer risk, the justifications for exploring the biology of this demographic may reveal changes that concurrently affect other higher risk race groups, particularly Black men. A recent study in China on the genomic and transcriptomic landscapes from prostate tissue samples of Chinese patients identified markers distinct from European ancestry. [5]

As the 2022 U.S. projection of cancer incidence among Black men is highest for prostate cancer, the interpretation of available genomics by race is critical for narrowing the long-existing health gap in the field. [6] Patient databases, such as The Cancer Genome Atlas (TCGA) and the Genomics Evidence Neoplasia Information Exchange (GENIE), contain thousands of tumor samples for genomic landscape discovery. These patient datasets are publicly available through cancer genomics engines such as cBioPortal. They can be filtered by age, race, and ethnicity, among many other search functions for analyzing populations of interest. [7, 8] In complement to TCGA, GENIE consists of a larger portion of race profiles as it is a collaboration between the American Association for Cancer Research (AACR) and eight different cancer institutions across the world. [9] Most recently, GENIE prostate cancer analyses have compared both actionable and high-frequency mutated genes by tumor type and race. [10, 11] However, few studies have been published on differences in copy number alterations between race groups, which may be relevant in downstream gene expression and tumor growth. [11, 12]

In the current study, using an omics approach, we draw upon the TCGA and the updated GENIE version 11 datasets, which incorporates more Black and Asian patients, to analyze mutation frequencies, copy number alterations (CNA), and gene expression by race or ancestry (Asian, Black, and White) in different tumor types (primary versus metastatic) of men with prostate cancer. We curated broad lists of pathway-specific genes with relevance to prostate cancer. Further, we present novel findings on differentially expressed genes by race from TCGA implemented by UCSC Xena genome browser. Our study provides an updated genomics and transcriptomics analysis of the GENIE and TCGA prostate cancer datasets by race.

Methods

TCGA and GENIE Datasets

Next-generation sequencing from tissue samples by self-reported race for TCGA Firehose Legacy and GENIE version 11 was downloaded from cBioPortal (RRID: SCR_014555). For more information on next-generation sequencing assays used, refer to the AACR GENIE data guide (https://www.aacr.org/wp-content/uploads/2022/07/GENIE_12.0-public_data_guide.pdf). No duplicates of patient data were used for calculating mutational frequencies and CNAs. Metastatic tumor subtype data was only available for GENIE cohort patient samples. More than 80% of GENIE profiles originated from the Dana Farber Cancer Institute and Memorial Sloan-Kettering Cancer Center. Institutes comprising the remaining 20% of prostate cancer genomics profiles included Institut Gustave Roussy, MD Anderson Cancer Center, Johns Hopkins Cancer Center, Netherlands Cancer Center, University Health Network, and Vanderbilt-Ingram Cancer Center. All patient profiles in this study were male. Patient age ranged from 37 to over 67 years, and the median ages were 65 and 69 years for primary and metastatic tumor types, respectively. A hybridization-based approach generated sample sequencing data. Mutation and CNA frequencies were calculated by dividing the number of patients harboring a specific gene mutation by the total number of patients sequenced for that gene. Data was separated by self-reported race, presenting the percentage of patients within each population affected by a mutation in a gene. We validated our findings using global genetic ancestry due to the limited number of self-reported African-American patients present in the TCGA cohort. Genetic ancestry is a molecular measurement of biological variation unaffected by the biases associated with socially constructed racial identities. When left to self-identify, the composition of African ancestry can vary tenfold in the African-American population. We utilized The Cancer Genetic Ancestry Atlas (RRID: SCR_003193) to categorize patients according to global genetic ancestry estimated by STRUCTURE. [13] Patients were grouped, using a threshold of 70% West African ancestry, determined by the average ancestry of the African-American population being approximately 73% West African, precisely 71% Nigerian, with varying levels of admixture. [14]

Analysis of DDR and Non-DDR Genes Relevant in Major Pathways

We selected clinically pertinent targets from the literature and those from the following commercially available screens to broaden our genomic analysis: BRCAnalysis CDX, FoundationONE CDX, ProstateNext, Invitae PCa Panel, Color Hereditary Prostate, Fulgent. These genes consisted of those in major prostate cancer pathways: androgen receptor (AR), DNA damage repair (DDR), and apoptosis/survival (TP53/MDM2, BCL2/BCL2L1/MCL1, PTEN/PI3K/AKT, MAP3K1). Much of our study was composed of the following DDR pathway genes: ATR, BARD1, CDK12, CHEK1, CHEK2, FANCA, FANCL, ATM, BRCA1, BRCA2, BRIP1, NBN, MLH1, MSH2, MSH6, PALB2, PMS2, RAD51B, RAD51C, RAD51D, and RAD54L. Clinically relevant targets from a previously conducted genomics study on GENIE version 7.0 further expanded our study: ABL1, EGFR, ERBB2, BRAF, FGFR2, FGFR3, KIT, NTRK1, NTRK2, NTRK3, PDGFRA, RET, ROS1, and ALK. [10] In addition to analyzing these actionable genes, we selected for hereditary DDR genes, including the transcription factor, HOXB13, and the cell-signaling molecule EPCAM. [15, 16] We compared mutation and CNA frequencies for genes between primary and metastatic tumor types defined by self-reported race in each figure.

Analysis of Differential Expression Profiles Generated by UCSC Xena

Differential gene expression analyses of the Genomic Data Commons (GDC) TCGA PRAD dataset were conducted through UCSC Xena (RRID: SCR_018938). [17] We first selected the prostate cancer cohort from the Genomic Data Commons Data Portal (RRID: SCR_014514) in the Xena browser (https://xenabrowser.net/heatmap/). Samples were separated by race before running the differential gene expression (DEG) pipeline. Analysis of GENIE through the DEG pipeline was unavailable due to the platform’s preuploaded datasets, including TCGA-PRAD. After selecting race subgroups, the pipeline performed a gene expression analysis of TCGA-PRAD data using a limma software analysis to compensate for smaller sample sizes. [18] Additional plots generated by this DEG pipeline included volcano plot visualization, pathway enrichment identification, principal component analyses to determine biological sample similarities, and associative rankings of small, non-coding RNA molecules, among other functions. We selected the following cohort comparisons for generating volcano plots: Black versus White, Asian versus White, Black versus Asian. Gene expression cut-off values were automatically set by the Xena DEG pipeline at + / − 1.5 for upregulated genes and downregulated genes, respectively. Gene expression of White patients functioned as a reference group for comparing non-White expression profiles. Transcriptional fold changes from the reference group were extracted from the pipeline and sorted with Venny 2.1 to isolate genes uniquely upregulated in one race and concurrently downregulated in another (i.e., genes upregulated among Black men that were simultaneously downregulated in Asians). Genes were then pooled into the following gene categories broadly based on function: non-coding regions, microRNAs, immunoglobulin coding, metabolic pathways, protein-coding regions.

Gene Set Enrichment Analysis (GSEA) of Dysregulated Genes Identified by UCSC Xena Analysis Pipeline

The list of unique genes upregulated or downregulated in Black patients compared to Asian patients was used to identify whether entire pathways have higher involvement in tumorigenesis by population. The gene set comparing Black to Asian patients was utilized due to the stark contrast in incidence between the two populations. Using the Broad Institutes computational tool GSEA, we conducted a preranked analysis, comparing Xena’s dysregulated gene list to the Molecular Signatures Database (RRID: SCR_016863) hallmark gene sets annotated for use with GSEA (RRID: SCR_003199).

Statistical Analysis of Associations Between Patient Race and Gene

The associations between mutation frequencies and race in GENIE cohort were analyzed using either a chi-squared test or Mehta’s modification to Fisher’s exact test (2 × 3 tables) as appropriate. [19] P values less than 0.05 were considered to have stronger associations between that specific gene and race within the total cohort of patients. Pairwise comparisons between White and Black patients were conducted using chi-squared tests or Fisher’s exact test when the former was not valid. All p-values are presented without adjustment as the results are intended to be exploratory and hypothesis generating. Only the associations with small p-values are presented, while, in many cases, there was little association between race and mutation frequencies.

Results

Demographic Comparison of GENIE and TCGA Patient Cohorts

Among prostate cancer patients with primary tumor types, 3700 GENIE profiles (Table 1) and 500 TCGA profiles were available (Suppl Table 1). In GENIE, there were 1857 White patients (81.7%) with primary tumor-type prostate cancer, followed by 186 Black (8.2%) and 60 Asian patients (2.6%). A total of 1265 (34.2%) profiles in GENIE were of metastatic tumor type with a median age of 69 years. Among those patients with metastatic tumor type, 921 patients (72.8%) were self-reported as White, 104 were Black (8.2%), 40 were Asian (3.2%), and 180 (14.3%) did not have race demographics available. Both tumor stage and grade were available for TCGA data but not available in GENIE.

Table 1.

Demographics of GENIE cohorts with prostate adenocarcinoma based on self-reported race. Age distribution was represented by quartiles and along with median age. Patient data with unknown or unrecorded race and/or ethnicity were grouped separately for each cohort

	GENIE
Tumor sample	Primary	Metastatic
Number of patients	2272 (61.4%)	1265 (34.2%)
Age (median)
37–56	388	121
57–61	332	129
62–66	490	231
67 +	1049	776
Unknown	13	8
Median	65	69
Race
White	1857 (81.7%)	921 (72.8%)
Black	186 (8.2%)	104 (8.2%)
Asian	60 (2.6%)	40 (3.2%)
Unknown/NA	169 (7.5%)	200 (15.8%)
Ethnicity
Hispanic or Latino	92 (4.0%)	43 (3.4%)
Not Hispanic or Latino	1920 (84.5%)	977 (77.2%)
Unknown/NA	260 (11.4%)	200 (15.8%)

Comparison of Primary and Metastatic Gene Mutation Frequencies by Race Among GENIE Cohort Patients

There were 21 genes from DDR pathways and 26 genes from additional relevant pathways for comparison by race and tumor subtype identified in prostate cancer tissue. Given the large cohort in GENIE, White patients with prostate cancer functioned as an arbitrary reference point for comparative purposes. All mutation frequencies were composed of insertions, deletions, and single nucleotide variants as described through CBioPortal. Black men with primary tumors had 16/21 DDR genes with mutation frequencies and 17/21 DDR genes in metastatic disease. Among Asian men, observed mutation frequencies resulted in 14/21 DDR genes and 7/21 DDR genes in primary and metastatic tumor types, respectively. Mutation frequencies appeared in 19/26 and 14/26 non-DDR pathway genes in our panel for Black patients with primary and metastatic diseases, respectively. Asian men with primary disease displayed 13/26 non-DDR pathway genes with mutation frequencies, and 14/26 also had mutations within this cohort with metastatic tumors.

Among DDR genes in primary tumor types (Fig. 1A), a 3.33% mutation frequency of NBN was strongly observed among Asians not found in Black and White patients (P = 0.045). Among MSH2 mutations, differences in frequencies varied among all race groups (White: 1.02%; Black: 2.69%; Asian: 3.33%; P = 0.027). Several RAD genes, also, were determined to have strong statistical differences by race including RAD51B (P = 0.078), RAD51D (P = 0.029), and RAD54L (White: 0.22%; Black: 1.08%; Asian: 1.67%; P = 0.033). Mutation frequencies in patients with metastatic tumor type (Fig. 1B) were most prevalent by race in ATR (White: 1.19%; Black: 2.88%; Asian: 5%; P = 0.049) and CDK12 (White: 5.1%; Black: 14.4%; Asian: 12.5%; P = 0.0003). The mutational landscape of highly relevant non-DDR pathway genes in primary disease (Fig. 1C) was similarly compared and showed greater associations with race in BCL2L (Black: 0%; White: 0.27%; Asian: 3.33%; P = 0.019), ERBB2 (White: 0.7%; Black: 2.69%; Asian: 3.33%; P = 0.0058), and BRAF (White: 1.67%; Black: 4.3%; Asian: 1.67%; P = 0.056). Among PI3K/PTEN/AKT pathway genes, the tumor suppressor gene, PTEN, displayed a primary tumor frequency of 8.3% in Asians, 2.2% in Black men, and 6.5% in White men (P = 0.0504). Among non-DDR genes in metastatic tumor types (Fig. 1D), the gene KIT had the strongest mutation and race association (White: 0.65%; Black: 0.96%; Asian: 5%; P = 0.0400). In a paired analysis of select genes between Black and White men, TP53 had strong differences in mutation frequency among metastatic tumor type with 35% among White and 24% among Black (P = 0.0166). Metastatic patients with MYC mutations presented with 0.65% among White and 2.88% among Black (P = 0.054). The BRAF gene showed a 1.67% mutation frequency among White men and 4.3% mutation frequency among Black men with primary disease (P = 0.021).

Fig. 1.

Mutation frequencies of major pathways by self-reported race and tumor type from GENIE project v11 datasets. Data was limited to one sample per patient. Mutation frequencies of primary (A) and metastatic (B) tumor types in major DDR pathway genes were separated by race. Other major pathways as described in methods were presented by primary (C) and metastatic types as well (D). Statistical associations between race and differences in mutation frequency (P < 0.05) were determined using either chi-square test (*) or Fisher’s exact test with Mehta’s modifications (**)

Comparison of Copy Number Alteration Frequencies Within GENIE Cohort by Race and Tumor Types

Copy number alterations (CNAs) of genes investigated above were assessed for potential implications in disease susceptibility and the regulation of downstream transcriptional and translational pathways. CNAs present in primary tumor samples showed alterations in 14 genes among White, six among Black, and four among Asian patients (Fig. 2A). Alteration frequencies in metastatic tumor patients were present in 16 DDR genes for White, eight for Black, and four for Asian patients (Fig. 2B). DDR pathway gene mutations for primary tumor type (Fig. 2A) showed the MMR genes, MSH6 (P = 0.027) and MSH2 (P = 0.047), each having deletion frequencies of 3.3% in Asians. The ATM gene showed a 0.16% deletion frequency among White men with primary disease, followed by 0.54% in Black men and 1.67% in Asian men (P = 0.0536).

Fig. 2.

Copy number alteration (CNA) frequencies of major pathway genes by self-reported race in GENIE. CNA frequencies of major DDR pathway genes among patients with primary disease (A) and metastatic disease (B). CNA frequencies of non-DDR pathway genes as described in methods by primary (C) and metastatic (D) disease. Statistical associations between race and differences in mutation frequency (P < 0.05) were determined using either chi-square test (*) or Fisher’s exact test with Mehta’s modifications (**)

Among non-DDR pathways with notable mutations in metastatic disease (Fig. 2D), BRCA2 deletions presented with stronger differences by race with frequencies of 0% among Black patients, 4% among White patients, and 5% among Asians (P = 0.045). FANCA gene deletions in metastatic tumor types were also strongly associated by race (White: 2.71%; Black: 2.88%; Asian: 10%; P = 0.045). Primary tumor Asian men (Fig. 2C) had a 1.7% HOXB13 amplification frequency in comparison to 1.61% in Black men and 0.43% in White men (P = 0.044). Further, we identified strong associations with race among primary disease FGFR3 amplifications in Black men (1.08%; 0% in White and Asian men; P = 0.014) and differences in metastatic disease KIT amplifications among White (0%), Black (0.96%), and Asian (2.5%) men (P = 0.0087).

AR amplifications were detected mainly in all races with metastatic disease; however, AR loss in primary disease type showed 1.6% among Asians with primary disease in comparison to 0% in White and Black patients (P = 0.0285).

Mutation Frequency and Copy Number Alteration Frequencies of Major Pathway Genes in TCGA Prostate Cancer Cohort

We examined the TCGA-PRAD dataset for mutation and CNA frequencies in patients with prostate cancer (Suppl Fig. 1). Due to the small number of Black (N = 7) and Asian (N = 2) patients who self-identified their race, we enriched the dataset with ancestry-ranked patients, which subsequently increased the number of Black (N = 49) and Asian (N = 11) patients, allowing for a more comprehensive comparison (Suppl Fig. 2). Compared to the GENIE dataset (Fig. 1), the ancestry-ranked TCGA-PRAD dataset displayed similar mutational and CNA frequencies (Suppl. Figure 2). In the enriched ancestry data, we observed high PTEN deletions and MYC amplifications among Asians in the GENIE cohort. On the other hand, the high frequencies of MSH and FANCA deletions among Asians were not observed in the ancestry-ranked cohort. Observable mutational and CNA frequencies among Black patients in this cohort generally followed similar trends to observations in the GENIE cohort. However, the enhanced range of genomic variation in the GENIE cohort was more apparent.

Transcriptome Breakdown of Differentially Expressed Genes Exclusively Upregulated in Black and Asian Cohorts from TCGA-PRAD

Our gene-sorting approach led us to 64 targets upregulated in the Asian cohort that were simultaneously downregulated in the Black cohort (Fig. 3A). Similarly, we identified 199 genes upregulated among Black men that were likewise downregulated in the Asian cohort. We found 174 genes upregulated in both Black and Asian men compared to White patients vs. 179 genes downregulated in both races. We further investigated the types of genes in these transcriptomes that were found to be upregulated in one race and downregulated in another race based on functionality. There were five (2.51%) proteins upregulated in Black men implicated in metabolic pathways versus one gene (1.54%) that was upregulated in Asians (Fig. 3B). Among the Asian cohort, there was one immunoglobulin-coding gene found upregulated (Fig. 3C). We then found 61 (30.7%) immunoglobulin genes upregulated in Black patients (Fig. 3D, Suppl Table 3). The largest subgroup of genes among Black and Asian samples consisted of long noncoding RNA molecules and pseudogenes with 44 genes (67.8%) in the Asian cohort and 123 genes (61.8%) in the Black cohort (Suppl Table 2, Suppl Table 4). We also noted nine (4.5%) other coding genes upregulated in Black and 16 (25%) upregulated in Asians (Suppl Table 5).

Fig. 3.

Distribution of differentially expressed genes in Black and Asian patients. A, B Gene targets specifically upregulated or downregulated among Black and Asian patients of the GDC TCGA-PRAD cohort. Distribution of transcriptome by genes uniquely upregulated in Asian (C) and Black (D) men

Pathway Analysis of Unique Genes in Isolated DEGs Extracted from UCSC Xena Analysis Pipeline

Using the MSigDB hallmark gene sets annotated in GSEA, an analysis was conducted to identify pathways unique to the upregulated or downregulated genes in Black compared to Asian patients. While the analysis identified that 30 of 50 gene sets were enriched for genes downregulated in black males, none were found to be statistically meaningful. For genes upregulated in black males, in comparison to Asian males, 20 of 50 gene sets were enriched in major pathways. Among those genes, seven had a p-value less than 0.01 and a FDR value of 25%, respectively (Table 2). Specifically, we identified E2F targets, G2M checkpoint, MYC targets versions 1 and 2, oxidative phosphorylation, mTORC1 signaling, and mitotic spindle pathways via our preranked analysis of upregulated genes in Black men with prostate cancer (Fig. 4).

Table 2.

MSigDB hallmark gene sets enriched in genes differentially upregulated in GDC TCGA Black patients in comparison to Asian patients. Size: the number of genes represented in the gene set. Enrichment score (ES): a measure of how overrepresented the gene in a pathway is on both extremes of the ranked gene list

Gene set	Size	Enrichment score (ES)	Normalized ES	NOM p-value	FDR q-value
HALLMARK_E2F_TARGETS	195	− 0.56	− 2.25	0	0
HALLMARK_G2M_CHECKPOINT	190	− 0.5	− 2.03	0	0
HALLMARK_MYC_TARGETS_V1	194	− 0.47	− 1.85	0	0.003
HALLMARK_MYC_TARGETS_V2	58	− 0.53	− 1.8	0	0.002
HALLMARK_MTORC1_SIGNALING	194	− 0.36	− 1.48	0	0.037
HALLMARK_OXIDATIVE_PHOSPHORYLATION	184	− 0.37	− 1.48	0	0.031
HALLMARK_UNFOLDED_PROTEIN_RESPONSE	106	− 0.36	− 1.35	0.016	0.085
HALLMARK_MITOTIC_SPINDLE	198	− 0.31	− 1.22	0	0.209

Fig. 4.

Enriched pathways for genes upregulated in TCGA Black patients. Pathways were strongly associative with race if P < 0.01 with an FDR < 25%

Discussion

The implication of race continuously remains a challenge in treatment development as specific populations are more heavily impacted by mortality in prostate cancer. Recent race additions to genomic profiles have expanded opportunities to dissect uncharted biological differences, partially contributing to this disparity. Our study compares mutation and CNA profiles that vary by tumor type and race to continue investigations on the genomics front as patient populations are expanded. Even with the recent increase in patients selected for sequencing in project GENIE, it should be noted that clinical sequencing may not be entirely representative of larger populations, especially among Black and Asians. However, preliminary analyses may guide more strongly associated interpretations as enriched sequencing becomes available. We found overlaps between genes investigated in our study and previous genomic and CNA alteration publications. [10, 11, 20, 21] These include genes in DDR pathways such as MSH2 and MSH6 and canonical prostate cancer genes like AR, TP53, and PTEN.

Among DDR pathway genes, our data demonstrated strong associations of homozygous deletions frequencies of MSH6 and MSH2 varying by race. Although the frequencies of these mutations were all less than 5%, they remained highest among Asian patients in primary disease while Black patients displayed no frequencies of MSH6 and MSH2. We also observed a 3.6% mutation frequency of MSH2 in Asians with primary disease, which had no mutation frequency in metastatic disease of this cohort. Despite preliminary observations, high MSH2 deletions and mutations, particularly among Asian patients with primary disease, may demonstrate potential pathway and disease markers of interest within this cohort. Similarly, the high deletion frequency of the homologous recombination repair gene, FANCA, as observed in Asian patients with metastatic disease, supported previous findings on the genetic risk of prostate cancer associated with this gene. [22] The higher prevalence of FANCA deletions may serve as an additional biomarker for predicting the genetic risk of Asian patients. Further, this loss of function mutation has shown efficacy in reducing specific drug resistance phenotypes, such as in the case of olaparib resistance and cisplatin hypersensitivity. [23, 24] Strong associations between race and the RAD family genes were also implicated in our analysis similar to previously described mutation frequencies in Kohaar et al. 2022 that compared African American and European American patients. [25]

Our findings on AR displayed very low mutation frequencies among men of all races with primary prostate cancer tumor types and similar frequencies in metastatic tumors as observed in a previous study on earlier GENIE datasets. [11] Amplification frequencies of AR among Asians with metastatic tumors were greatest (33.3%), followed by Black (26%) and White patients (25.4%). These trends support a previous study conducted by Stopsack et al. where the greater alterations were observed among Asian patients over Black and White patients. [20] However, we observed a small percentage of AR deletions among Asians with primary disease previously not found. It should be noted that variations in AR mutations, particularly among metastatic tumors, may vary depending on treatment exposure and should be further investigated for the likelihood of genomic variation. We acknowledge the limitations of project GENIE on the potential effects of these unknown factors in interpreting genomics findings. Further, it has been determined that patients of African descent are more likely to contain short CAG repeat sequences, which should be considered for future studies on AR activity by race. [26]

Our findings demonstrated high rates of mutations among all races in the tumor suppressor gene, TP53, previously found to have a higher mutation frequency in metastatic disease. [27] A paired race analysis using a chi-squared test between Black and White men of metastatic TP53 mutations demonstrated a 36% frequency among White men and a 24% frequency among Black men (P = 0.0166). We did not determine strong associations of varying mutation frequencies of TP53 in primary disease by race like Stopsack et al. between Black men and White men. We further discovered strong variations across all race groups in this study among metastatic tumor mutations of TP53 with the greatest frequencies in Asians (43%), followed by White (36%), and Black patients (24%). Given these unexpected mutational frequencies in metastatic tumor types, future studies similarly identifying these discrepancies may contribute toward narrowing health disparities on the genomic front.

The phosphatase and tensin homologue (PTEN) is a well-published tumor suppressor protein whose loss of expression has clinical relevance in the progression of various cancer types, including prostate cancer. In particular, PTEN mutations are present in more than 50% of metastatic tumors, and screenings for this gene have application in treatment prognosis. [28] We showed that Black men displayed higher PTEN deletion frequencies than Asian men; however, White men displayed the greatest percentage of alterations similar to findings in the recent report by Stopsack et al. 2021. [20] Higher mutation frequencies in PTEN and AKT1 were observed among Asian men with primary disease than Black men, suggesting more prevalent associations of the PTEN/PI3K/AKT pathway with Asians. Prostate cancer genomics among Asian patients is the least investigated, likely due to this demographic’s historically low disease prevalence. Further, the limited studies conducted represent mainly Chinese and Japanese men, and even fewer studies go beyond patients of East Asian descent. [29] In contrast to Asians with metastatic disease, Black patients with metastatic prostate cancer had higher percentages of PTEN mutations in this study. However, Black patients demonstrated a much lower mutation rate when compared to the whole GENIE patient pool, with 20% and 50% frequencies for primary and castration-resistant disease, respectively. Race-dependent genetic factors in driving tumor progression and affecting drug response are of interest for investigation. For example, two recent studies demonstrated that Black men who received abiraterone had improved overall survival compared to non-Hispanic White men with metastatic CRPC (mCRPC), and ipatasertib plus abiraterone greatly improved progression-free survival in mCRPC patients with PTEN loss. [28, 30, 31] Future prospective studies should determine genetic drivers of differential abiraterone outcomes by race.

We also present novel findings on a subset of differentially expressed genes by race based on RNA sequencing data of patients in the TCGA-PRAD cohort. UCSC Xena’s genomics visualization engine supports a wide range of these datasets to accommodate sequencing assays such as whole genome, single cell, DNA methylation, and transposase-accessible chromatin (ATAC). [17] Its web-browser based user interface allows for the centralized analysis of genomics and transcriptomics of both public and private datasets through various visualization modes. This has guided in the discovery of several molecular drivers of the disease. [32–34] However, previous studies have primarily focused their attention on DEGs among tumor samples from White patients. [35] In contrast, we selected RNA sequencing data from White patients as a reference group. We identified candidate genes specifically upregulated in one race group and downregulated in another. We found five coding genes (UGT2B17, GSTM1, AMPD1, CTRB2, HAO1) with metabolic functions downregulated among Asians and upregulated in Black men. Previous studies have suggested that the disproportionately worse survival seen among Black patients is related to biological differences in drug metabolism. [36–38] In particular, uridine diphosphoglucoronosyl-transferase 2B17 (UGT2B17) was upregulated in Black men in our study and in a recent publication conducted on predicting metabolic differences by race. [36] The study found UGT2B17 as a potential prognostic biomarker for docetaxel and cisplatin resistance. UGT2B17 also has functionality in the catabolism of androgens, and its gene expression correlates with disease progression in prostate cancer. [39] To date, its overexpression is not characterized by race. Our study alongside Liu et al. demonstrates strong associations of UGT2B17 gene expression among Black men that is also downregulated among Asians. [36]

We also found that the metabolic gene, GSTM1, was particularly upregulated among Black patients in our DEG analysis and simultaneously downregulated among Asians. This common subtype of glutathione-S-transferase (GST) enzymes is involved with the metabolism of carcinogens, electrophilic structures, therapy agents, and toxins, in addition to implications in the progression of prostate cancer. [40] As such, lower expression of the gene would be detrimental, resulting in improper detoxification of reactive molecules and leading to cancer progression; indeed, reduced GSTM1 gene expression is found in prostate cancer patients. [41] Genetic polymorphisms in the gene also can result in nonfunctional enzymes and may contribute to prostate cancer development. GSTM1 null genotypes may explain the variation in prostate cancer incidence among different racial groups, placing Asians at increased risk for prostate cancer. [42]

Our study is the first to identify the overexpression of AMPD1, CTRB2, and HAO1 in prostate cancer. While the functional significance of the previously mentioned genes in prostate cancer remains to be determined, AMPD1 was identified as a novel biomarker for predicting disease outcome and immune response in HER2-positive breast cancer; whereas CTRB2 confers risk of pancreatic cancer and inactivation of HAO1 results in progression of hyperoxaluria, respectively. [43–45] We found that in the case of DEGs upregulated among Asians and simultaneously downregulated among Black patients, no metabolic genes were overexpressed in Asians. Together, our findings align with previous analyses of DEGs by race comparisons on survival within metabolic and inflammatory pathways. [36, 46–48] For instance, docetaxel treatment among Black men improved survival rates of this group to be comparable to that of White men despite the survival typically being lower for the former. [49]

Our GSEA analysis provided further evidence of the correlation between drug resistance and metabolism, as the genes identified as upregulated in Black men displayed strongly associated enrichment in MYC targets, versions 1 and 2, and oxidative phosphorylation. Prior work by Kumar et al. suggests that the higher protein levels of Myc have downstream effects, lowering cytochrome c expression. [50] The authors noted that cytochrome c deficiency increased glycolysis rates and mitochondrial dysfunction, an occurrence supported by our pathway analysis. Further, prostate cancer risk variants in the 8q24 region, which houses both MYC and long noncoding RNAs, were identified as unique to those with African ancestry. [32] A recent study examined on the relationship between genealogy and self-reported race on alterations, promoting the importance of genetic ancestry in health disparities. [20, 51] Collectively, studies on the association between African ancestry, Myc variants, and mitochondrial function may address drug resistance and aggressive disease in Black patients.

Because prostate cancer has the most characterized heritability of the common cancers, [52] this discovery suggests that risk variants could be passed through generations, linked with local ancestry. Black men are often diagnosed at younger ages, indicating that heredity may play a more important role in the population. [53] In addition, self-identified racial categories do not define an individual’s ancestral background. Black patients have high rates of ancestral admixture, largely due to the transatlantic slave trade, and range from having 20 to 100% West African ancestry (with an average of about 80%) [14]. Admixture seems to be one of several determinants in prostate cancer incidence. In Nigeria, 23.3 per 100,000 males are affected by prostate cancer, whereas 247.3 per 100,000 Black males are diagnosed in the USA. [54] Future work could build upon our initial findings, identifying whether specific percentages of West African ancestry correlate to disease incidence or progression. Further, ancestry-based genomics analyses have determined significance between ancestry and self-reported race with tumor mutations. [55] However, we also acknowledge studies reporting that small sample size may limit determining racial significance depending on the analysis. [21] The TCGA prostate cancer population is of considerable size; however, even the most updated cohort is mainly comprised of White patients with only seven and two self-reported Black and Asian patients, respectively. Previous studies utilizing TCGA-PRAD have identified race-specific changes in miRNAs, protein expression, and epigenetic variations as well as clinical outcomes. [56–64]

Health disparities in the treatment of aggressive diseases such as prostate cancer continue to affect patients of different races as several variables spanning from clinical to biological factors are involved. For instance, a study showed that Black men with prostate cancer are less likely to receive a direct treatment plan than White patients even as disease aggressiveness increased. [65] This is one of several examples of higher mortality among Black over White or Asian patients, thus emphasizing the prevalent health disparity. In addition to treatment disparities, ongoing genomics studies have implied biological and ancestral differences that may drive differences in mortality and survival rates. As updated datasets become publicly available that incorporate race demographics, it will become increasingly more urgent to continue observing trends in genomic alterations that have notable biological implications, given the present limitations in racial diversity among prostate cancer patients.

Data availability

Data analyzed in this study were obtained from CBio-Portal for GENIE v11 (https://www.cbioportal.org/genie/). [9] TCGA Firehose Legacy data was also obtained from CBioPortal (https://www.cbioportal.org/study/summary?id=prad_tcga). Data obtained for gene expression analysis of GDC TCGA PRAD were obtained from the UCSC Xena genome browser (https://xenabrowser.net).