Abstract
Prostate cancer (PCa) exhibits pronounced racial disparities in incidence, aggressiveness, and mortality, with African American (AA) men experiencing earlier onset and poorer outcomes compared to European American (EA) men. To investigate ancestry-associated molecular differences, we analyzed RNA-sequencing data (GSE104131) from high-grade (Gleason score ≥ 7 (4+3)) PCa tumors and matched adjacent normal prostate tissues of AA and EA patients using DESeq2 and Gene Set Enrichment Analysis (GSEA). Differential expression analysis revealed no individual genes significantly altered between AA and EA tumors; however, pathway-level analysis identified 33 enriched gene sets distinguishing the two groups. AA tumors displayed upregulation of keratinization and cornified envelope formation pathways, and downregulation of IgG immunoglobulin complex and endoplasmic reticulum protein-folding processes. Within-group analyses showed 574 differentially expressed genes (DEGs) in AA tumors and 330 in EA tumors (| log₂FC | > 1, FDR < 0.05). In AA tumors, notable upregulated genes included AMACR and NPY, with enrichment of xenobiotic detoxification and proliferation pathways, alongside strong downregulation of epithelial-to-mesenchymal transition (EMT) and stromal activation programs. EA tumors were characterized by upregulation of androgen-responsive epithelial programs and protein synthesis machinery, and downregulation of striated muscle development and inflammatory signaling. These findings suggest ancestry-specific differences in epithelial differentiation, immune signaling, and drug resistance mechanisms, with keratin pathway enrichment in AA tumors potentially linked to aggressive phenotypes and poorer androgen deprivation therapy response. This study underscores the utility of pathway-level approaches for detecting subtle but coordinated molecular differences and highlights the need for ancestry-informed biomarker discovery and therapeutic strategy development in PCa.
Keywords: prostate cancer, African-American men, health disparity, RNA-seq, differentially expressed genes, deregulated pathways, DESeq2, GSEA, biomarkers
Introduction
Prostate cancer (PCa) is the most common cancer in men in the United States, other than skin cancer. Mortality-wise, PCa is the second-leading cause of cancer death in American men, behind only lung cancer. About 1 in 44 men will die of prostate cancer. The American Cancer Society’s estimates for PCa in the United States for 2025 are about 313,780 new cases of PCa and about 35,770 deaths from PCa1.
Prostate cancer has the largest racial disparities of any cancer in the United States. PCa in African-American (AA) men exhibits the largest racial disparities in cancer-related mortality, with more than double the risk of death compared to other ethnic groups2. Black men have a risk of diagnosis that is 1.67 times (180 vs 108 per 100,000) and a risk of cancer death that is 2.06 times (36 vs 17 per 100,000) that of White men3.
While much of these health disparities have been ascribed to socio-economic differences and access to care4, previous studies show that AA men are at higher risk of being diagnosed with PCa, present more frequently with high-grade disease, are at increased risk of biochemical progression after radical prostatectomy (RP), and have higher expression of adverse molecular features, even in an equal access setting5. This strongly suggests that biological differences contribute to PCa racial disparities.
Recent studies further reveal that AA men with PCa exhibit unique biological profiles, including altered androgen receptor signaling, genomic instability, metabolic dysregulation, and inflammatory signaling, all of which may contribute to more aggressive disease progression6. Despite decades of research, the specific biological mechanisms underlying these disparities remain incompletely understood. There is a need for studies that integrate molecular characterization in racially diverse patient cohorts to identify biological differences that may underlie disparities in PCa7. By defining these differences at the pathway level, such research can effectively uncover biological systems perturbed in ethnic-specific tumor cells, guiding future studies toward the most promising targets for prevention, precision medicine, and therapy development8.
Building on this need, this study aims to understand the biological underpinnings of PCa racial disparities by identifying and characterizing deregulated biological pathways in African American men, who experience a higher incidence and more aggressive forms of the disease. Given prior evidence of racial disparities in prostate cancer outcomes and immune-related pathways, we hypothesize that AA prostate tumors would exhibit distinct immune signaling profiles compared to EA tumors9,10,11,12,13.
Methods
1) RNA-seq Data Acquisition and Preprocessing
RNA-seq raw count data for matched high-grade (Gleason score ≥ 7 (4+3)) prostate tumor and adjacent normal tissue samples from African-American (AA) and European-American (EA) patients were obtained from the Gene Expression Omnibus (GEO; accession GSE104131) via the National Center for Biotechnology Information (NCBI) RNA-seq counts resource14.
The original dataset comprised 32 unique biological samples: 8 tumor samples from AA patients, 8 tumor samples from EA patients, 8 adjacent normal tissue samples from AA patients, and 8 adjacent normal tissue samples from EA patients. Each sample was sequenced in two technical replicates.
The NCBI RNA-seq processing pipeline aligns sequencing runs labeled as Homo sapiens and transcriptomic to GRCh38.p13 human genome assembly (NCBI Assembly GCA_000001405.15) using HISAT2. Runs passing 50% alignment rate are then processed with Subread featureCounts to produce gene-level raw counts, with counts from multiple SRA runs for the same GEO sample summed to yield a single count value per gene per sample. Data derived from single cell samples are skipped, and values in raw count matrices are rounded for compatibility with common differential expression analysis software like DESeq2, edgeR or limma voom15.
In the processed raw count matrix downloaded from GEO, 4 samples from the original dataset were absent due to NCBI pipeline exclusions: 2 adjacent normal tissue samples from EA patients, 1 tumor sample from an EA patient, and 1 adjacent normal tissue sample from an AA patient. The final dataset analyzed, therefore, consisted of 28 biological GEO samples, all generated under uniform experimental conditions within a single study.
2) Quality Control and Normalization
All analyses were conducted in R version 4.4.3 using the DESeq2 package (version 1.46.0) for differential expression analysis16. Genes with zero total counts across all samples were removed prior to analysis. Raw counts were normalized using DESeq2’s median-of-ratios method to adjust for sequencing depth and RNA composition differences. PCA was performed on variance-stabilized count data to assess sample clustering by condition and verify the absence of batch effects. All samples originated from the same dataset (GSE104131) and were processed using a uniform protocol, minimizing the likelihood of batch effects.
This quality control ensured that observed differences in gene expression between AA and EA tumors reflected biological variation rather than technical variation, which was critical for robustly testing our hypothesis.
3) Differential Expression Analysis and GSEA Analysis
Differential expression was assessed using the Wald test, and p-values were adjusted for multiple testing with the Benjamini–Hochberg false discovery rate (FDR) method. Genes with | log2 fold change (FC) | > 1 and an adjusted p-value (FDR) < 0.05 were considered statistically significant. Gene annotations were obtained using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) Bioinformatics Resources 2021 update17.
Gene Set Enrichment Analysis (GSEA) was performed using the clusterProfiler package in R, with all genes ranked by log2 fold changes from the differential expression analysis18. Gene sets were obtained from the Human Molecular Signatures Database (MSigDB) via the msigdbr package in R for Homo sapiens19. Enrichments were considered significant at an adjusted p-value (FDR) < 0.05 and an |Normalized Enrichment Score (NES)| > 1. Representative enriched pathways were visualized using the enrichplot package in R, specifically with the gseaplot2 function20.
Results
We compared tumor and matched normal tissues within African-American (AA) and European-American (EA) groups, and between AA and EA tumors. Results are presented for each contrast below.
1) AA Tumor and EA Tumor Samples
Direct comparison of prostate cancer (PCa) samples from AA and EA patients did not yield any differentially expressed genes (DEGs) meeting the significance threshold (| log2 fold change (FC) | > 1 and adjusted p-value (FDR) < 0.05).
However, Gene Set Enrichment Analysis (GSEA) using the full ranked gene list revealed 33 significantly enriched pathways (FDR < 0.05, |Normalized Enrichment Score (NES)| > 1). Positively enriched pathways in AA tumors included GOBP keratinization (NES = 1.51, FDR = 0.0074), REACTOME formation of the cornified envelope (NES = 1.48, FDR = 0.0032), REACTOME keratinization (NES = 1.46, FDR = 0.000015), and pathways related to olfactory signaling and transduction (Table 1). Negatively enriched pathways included DURANTE adult olfactory neuroepithelium plasma cells (NES = –3.04, FDR = 0.000015), GOCC IgG immunoglobulin complex (NES = –2.73, FDR = 0.000044), and multiple ER protein-folding and transport processes such as the calnexin/calreticulin cycle (NES = –2.62, FDR = 0.0089) (Table 2).
These results suggest that while no single gene reached statistical significance, coordinated expression changes in pathways related to keratinization, immune complex formation, and ER proteostasis may underlie ancestry-associated differences in prostate cancer biology.
| Pathways | NES | FDR |
| GOBP Keratinization | 1.51 | 0.007 |
| REACTOME Formation of the Cornified Envelope | 1.48 | 0.0032 |
| REACTOME Keratinization | 1.46 | 1.48 × 10-5 |
| REACTOME Olfactory Signaling Pathway | 1.31 | 0.0021 |
| KEGG Olfactory Transduction | 1.30 | 0.0032 |
| Pathways | NES | FDR |
| DURANTE Adult Olfactory Neuroepithelium Plasma Cells | -3.04 | 1.48×10-5 |
| GOCC IgG Immunoglobulin Complex | -2.73 | 4.41×10-5 |
| GOBP Endoplasmic Reticulum to Cytosol Transport | -2.67 | 0.0042 |
| MODULE 155 | -2.65 | 0.0074 |
| REACTOME Calnexin/Calreticulin Cycle | -2.62 | 0.0089 |
2) AA Tumor and AA Adjacent Normal Samples
Differential expression analysis between AA tumor and adjacent normal samples identified 926 genes with an FDR < 0.05. Of these, 574 DEGs (330 upregulated, 244 downregulated genes) also had an | log2 FC | > 1 and were considered DEGs. The volcano plot displays all FDR-significant genes, with those exceeding both statistical and fold-change thresholds highlighted in red (Figure 2B).
We ranked genes by log₂ fold change among those meeting the significance threshold (FDR < 0.05) and identified 10 upregulated and 10 downregulated genes in AA tumor versus AA adjacent normal samples. Among the upregulated genes, AMACR had the highest fold change (log₂ FC = 4.96), followed by C1QTNF3-AMACR (log₂ FC = 4.60), ANKRD34B (log₂ FC = 4.42), NPY (log₂ FC = 4.03), MUC6 (log₂ FC = 3.96), CRISP3 (log₂ FC = 3.92), LOC105377527 (log₂ FC = 3.81), OR51E1 (log₂ FC = 3.75), TRC-GCA6-1 (log₂ FC = 3.61), and DLX1 (log₂ FC = 3.49) (Table 3). The top downregulated genes included IGHG1 (log₂ FC = –2.93), C1QL1 (log₂ FC = –2.82), IGHGP (log₂ FC = –2.81), WIF1 (log₂ FC = –2.81), IGKV3-20 (log₂ FC = –2.75), IGKV1-5 (log₂ FC = –2.71), CCDC54-AS1 (log₂ FC = –2.62), IGKC (log₂ FC = –2.56), OLFM4 (log₂ FC = –2.55), and NELL2 (log₂ FC = –2.50) (Table 4).
Figure 2: Plots comparing PCa and adjacent normal prostate tissue from AA patients. (A) PCA plot of AA tumor (blue) and AA adjacent normal (red) samples showing separation. The PCA plot shows the absence of batch effect due to technical issues. (B) Volcano plot shows DEGs between PCa and adjacent noncancerous prostate tissue in AA patients. The horizontal axis is the log₂ FC between PCa and adjacent noncancerous prostate tissue. The negative log10 of the adjusted p-value (FDR) obtained from DESeq2 is plotted on the vertical axis. Each gene is represented by one point on the graph. The red spots represent significant DEGs (log₂ FC > 1: upregulated; log₂ FC < -1: downregulated). The blue spots represent genes that had FDR < 0.05 but didn’t match the | log2 FC | > 1 threshold.
| Genes | Gene Annotation | log2 FC | FDR |
| AMACR | alpha-methylacyl-CoA racemase | 4.96 | 8.11×10⁻⁸ |
| C1QTNF3-AMACR | C1QTNF3-AMACR readthrough (NMD candidate) | 4.60 | 4.14×10⁻⁷ |
| ANKRD34B | ankyrin repeat domain 34B | 4.42 | 0.00293 |
| NPY | neuropeptide Y | 4.03 | 0.0207 |
| MUC6 | mucin 6, oligomeric mucus/gel-forming (gene/pseudogene) | 3.96 | 8.51×10⁻⁵ |
| CRISP3 | cysteine rich secretory protein 3 | 3.92 | 5.46×10⁻⁵ |
| LOC105377527 | uncharacterized LOC105377527 | 3.81 | 0.00320 |
| OR51E1 | olfactory receptor family 51 subfamily E member 1 | 3.75 | 0.00178 |
| TRC-GCA6-1 | tRNA-Cys (anticodon GCA) 6-1 | 3.61 | 5.84×10⁻⁶ |
| DLX1 | distal-less homeobox 1 | 3.49 | 0.00602 |
| Genes | Gene Annotation | log2 FC | FDR |
| IGHG1 | immunoglobulin heavy constant gamma 1 (G1m marker) | -2.93 | 0.00551 |
| C1QL1 | complement C1q like 1 | -2.82 | 0.0339 |
| IGHGP | immunoglobulin heavy constant gamma P (non-functional) | -2.81 | 0.00847 |
| WIF1 | WNT inhibitory factor 1 | -2.81 | 6.19×10⁻⁴ |
| IGKV3-20 | immunoglobulin kappa variable 3-20 | -2.75 | 0.00796 |
| IGKV1-5 | immunoglobulin kappa variable 1-5 | -2.71 | 0.0236 |
| CCDC54-AS1 | CCDC54 antisense RNA 1 | -2.62 | 0.0200 |
| IGKC | immunoglobulin kappa constant | -2.56 | 0.0332 |
| OLFM4 | olfactomedin 4 | -2.55 | 0.00844 |
| NELL2 | neural EGFL like 2 | -2.50 | 0.00847 |
GSEA identified a total of 920 deregulated pathways with FDR < 0.05 and | log₂ FC | > 1, including 25 upregulated and 895 downregulated pathways in PCa and adjacent normal prostate tissue from AA patients.
Among positively enriched pathways, notable examples included the LIU prostate cancer up signature (NES = 1.62, FDR = 3.72×10⁻⁴) reflects activation of established tumor-associated transcriptional programs, while GNF2 MKI67 (NES = 1.60, FDR = 0.0492) underscores enhanced proliferative activity. Additionally, upregulation of xenobiotic detoxification regulators (NES = 1.76, FDR = 0.0100) suggests potential for treatment resistance, and negative regulation of glycogen metabolic processes (NES = 1.71, FDR = 0.0315) indicates shifts in metabolic demands associated with tumorigenesis (Table 5).
The most strongly downregulated pathways included the TOMLINS prostate cancer dn signature (NES = –4.16, FDR = 1.30×10⁻¹⁸), a gene set defined by reduced expression in prostate tumors relative to benign tissue (Table 6).
| Pathways | NES | FDR |
| chr15q11 | 1.78 | 3.51×10⁻⁹ |
| chr21q11 | 1.77 | 0.0149 |
| GOBP regulation of xenobiotic detoxification by transmembrane export across the plasma membrane | 1.76 | 0.0100 |
| GOBP negative regulation of glycogen metabolic process | 1.71 | 0.0315 |
| chr11q11 | 1.69 | 0.0453 |
| LIU prostate cancer up | 1.62 | 3.72×10⁻⁴ |
| GNF2 MKI67 | 1.60 | 0.0492 |
| ZHAN multiple myeloma PR up | 1.55 | 0.0330 |
| HOWARD NK cell inact monov influenza A Indonesia 05 2005 H5N1 age 18 49yo 3dy up | 1.55 | 0.0189 |
| GNF2 HMMR | 1.52 | 0.0407 |
| Pathways | NES | FDR |
| TOMLINS prostate cancer dn | -4.16 | 1.30×10⁻¹⁸ |
| GAVISH 3CA malignant metaprogram 12 EMT 1 | -3.97 | 7.53×10⁻¹⁷ |
| GAVISH 3CA metaprogram fibroblasts pericyte like | -3.82 | 1.49×10⁻¹⁴ |
| GAVISH 3CA metaprogram epithelial EMT like 1 | -3.77 | 6.52×10⁻¹⁴ |
| GAVISH 3CA metaprogram endothelial endo 6 | -3.58 | 1.37×10⁻¹¹ |
| GAVISH 3CA metaprogram epithelial interferon MHC | -3.51 | 5.77×10⁻¹¹ |
| PAPASPYRIDONOS unstable atherosclerotic plaque dn | -3.47 | 1.05×10⁻¹⁰ |
| GAVISH 3CA malignant metaprogram 18 interferon MHC II 2 | -3.45 | 1.94×10⁻¹⁰ |
| GAVISH 3CA metaprogram endothelial HEV 1 | -3.40 | 6.44×10⁻¹⁰ |
| WINTER hypoxia dn | -3.39 | 5.77×10⁻¹¹ |
3) EA Tumor and EA Adjacent Normal Samples
Differential expression analysis between the EA tumor and adjacent normal samples identified 540 genes with an FDR < 0.05. Of these, 330 DEGs (212 upregulated, 118 downregulated genes) also had an | log2 FC | > 1 and were considered DEGs. The volcano plot displays all FDR-significant genes, with those exceeding both statistical and fold-change thresholds highlighted in red (Figure 3B). Among these, we ranked genes by log₂ fold change and selected the ten most upregulated and ten most downregulated genes in EA tumor versus EA adjacent normal tissue.
The top upregulated genes included CRISP3 (log₂ FC = 5.90), a cysteine-rich secretory protein, followed by OR51E2 (log2 FC = 4.39), OR51S1 (log₂ FC = 3.63), PCAT14 (log₂ FC = 3.57), LOC105370113 (log₂ FC = 3.33), SIM2 (log₂ FC = 3.22), TDRD1 (log₂ FC = 3.17), LOC105374286 (log₂ FC = 2.95), TFF3 (log₂ FC = 2.94), and LOC107985492 (log₂ FC = 2.90) (Table 7).
The most downregulated gene was TGM4 (log₂ FC = –4.02), followed by PENK (log₂ FC = –3.18), SERPINB11 (log₂ FC = –2.84), LTF (log₂ FC = –2.61), TCAP (log₂ FC = –2.60), FAM167A (log₂ FC = –2.37), LRP2 (log₂ FC = –2.17), IQSEC3 (log₂ FC = –1.82), APOBEC3C (log₂ FC = –1.80), and PGM5-AS1 (log₂ FC = –1.79) (Table 8).
Figure 3: Plots comparing PCa and adjacent normal prostate tissue from EA patients. (A) PCA plot of EA tumor (blue) and EA adjacent normal (red) samples showing separation. It shows the absence of batch effect due to technical issues. (B) Volcano plot shows DEGs between PCa and adjacent noncancerous prostate tissue in EA patients, serving as a parallel comparison to the AA patient-specific analysis. The horizontal axis is the log₂ FC between PCa and adjacent noncancerous prostate tissue. The negative log10 of the adjusted p-value (FDR) obtained from DESeq2 is plotted on the vertical axis. Each gene is represented by one point on the graph. The red spots represent significant DEGs (log₂ FC > 1: upregulated; log₂ FC < -1: downregulated). The blue spots represent genes that had FDR < 0.05 but didn’t match the | log2 FC | > 1 threshold.
| Genes | Gene Annotation | log2 FC | FDR |
| CRISP3 | cysteine rich secretory protein 3 | 5.90 | 0.00506 |
| OR51E2 | olfactory receptor family 51 subfamily E member 2 | 4.39 | 6.15×10⁻⁷ |
| OR51S1 | olfactory receptor family 51 subfamily S member 1 | 3.63 | 0.0245 |
| PCAT14 | prostate cancer associated transcript 14 | 3.57 | 6.90×10⁻⁶ |
| LOC105370113 | uncharacterized LOC105370113 | 3.33 | 0.0420 |
| SIM2 | SIM bHLH transcription factor 2 | 3.22 | 1.25×10⁻⁵ |
| TDRD1 | tudor domain containing 1 | 3.17 | 0.00842 |
| LOC105374286 | uncharacterized LOC105374286 | 2.95 | 0.0270 |
| TFF3 | trefoil factor 3 | 2.94 | 0.0287 |
| LOC107985492 | uncharacterized LOC107985492 | 2.90 | 0.0155 |
| Genes | Gene Annotation | log2 FC | FDR |
| TGM4 | transglutaminase 4 | -4.02 | 0.0424 |
| PENK | proenkephalin | -3.18 | 0.00310 |
| SERPINB11 | serpin family B member 11 | -2.84 | 0.00698 |
| LTF | lactotransferrin | -2.61 | 0.0331 |
| TCAP | titin-cap | -2.60 | 0.0428 |
| FAM167A | family with sequence similarity 167 member A | -2.37 | 0.0165 |
| LRP2 | LDL receptor related protein 2 | -2.17 | 0.0131 |
| IQSEC3 | IQ motif and Sec7 domain ArfGEF 3 | -1.82 | 0.0156 |
| APOBEC3C | apolipoprotein B mRNA editing enzyme catalytic subunit 3C | -1.80 | 0.0218 |
| PGM5-AS1 | PGM5 antisense RNA 1 | -1.79 | 0.0493 |
GSEA identified a total of 1880 deregulated pathways with FDR < 0.05 and | log₂ FC | > 1, including 1786 upregulated and 94 downregulated pathways in PCa and adjacent normal prostate tissue from EA patients.
Among positively enriched pathways, the LIU prostate cancer up signature (NES = 3.74, FDR = 5.49×10⁻³⁷) reflected strong activation of previously reported tumor-associated transcriptional programs, while the GAVISH 3CA malignant metaprogram 9 unfolded protein response (NES = 3.10, FDR = 3.37×10⁻¹²) indicated heightened endoplasmic reticulum stress adaptation. Upregulation of cytoplasmic ribosomal protein pathways (NES = 2.97, FDR = 1.18×10⁻¹³) and translation initiation (NES = 2.94, FDR = 2.49×10⁻¹²) suggested an overall increase in protein synthesis demands (Table 9).
The most strongly downregulated pathways included the LIU prostate cancer dn signature (NES = –2.16, FDR = 7.33×10⁻³³), which represents transcriptional programs typically suppressed in prostate tumors, as well as multiple striated muscle contraction and muscle cell development pathways (NES ≈ –1.90), consistent with loss of muscle-related gene expression. Notably, downregulation of the HIV Tat to TLR2/4 NF-κB signaling pathway (NES = –1.88, FDR = 0.0118) points to reduced inflammatory signaling potential in EA tumors (Table 10).
| Pathways | NES | FDR |
| LIU prostate cancer up | 3.74 | 5.49×10⁻³⁷ |
| GAVISH 3CA malignant metaprogram 9 unfolded protein response | 3.10 | 3.37×10⁻¹² |
| TOMLINS prostate cancer up | 3.09 | 3.55×10⁻¹³ |
| MORF UBE2I | 3.08 | 8.38×10⁻²⁸ |
| MORF NPM1 | 2.98 | 1.64×10⁻²¹ |
| WP cytoplasmic ribosomal proteins | 2.97 | 1.18×10⁻¹³ |
| GNF2 EIF3S6 | 2.96 | 5.10×10⁻¹⁷ |
| KEGG medicus reference translation initiation | 2.94 | 2.49×10⁻¹² |
| GAVISH 3CA metaprogram epithelial androgen prostate | 2.92 | 3.24×10⁻¹⁰ |
| REACTOME SRP dependent cotranslational protein targeting to membrane | 2.87 | 2.01×10⁻¹⁴ |
| Pathways | NES | FDR |
| LIU prostate cancer dn | -2.16 | 7.33×10⁻³³ |
| WP striated muscle contraction pathway | -1.98 | 0.000365 |
| GOBP striated muscle cell development | -1.90 | 0.000123 |
| REACTOME striated muscle contraction | -1.90 | 0.00157 |
| GNF2 MYL2 | -1.88 | 0.00233 |
| KEGG medicus pathogen HIV Tat to TLR2 4 NFkB signaling pathway | -1.88 | 0.0118 |
| MODULE 387 | -1.88 | 0.00164 |
| MODULE 329 | -1.88 | 0.00146 |
| RICKMAN head and neck cancer F | -1.87 | 0.000871 |
| HP myocardial fibrosis | -1.87 | 0.0175 |
Discussion
African American (AA) men are disproportionately affected by prostate cancer, experiencing earlier onset, more aggressive disease progression, and higher mortality rates compared to European American (EA) men21,22. The racial disparities observed in prostate cancer among AA men are multifactorial, involving not only cultural and socioeconomic factors but also a significant lack of racial diversity in genomic and precision medicine research23,24.
In this study, we compared gene expression profiles between AA and EA prostate tumors, as well as between tumors and matched adjacent normal tissues within each ancestry group. The goal was to identify molecular features associated with ancestry and tumorigenesis in prostate cancer (PCa).
Direct differential expression analysis did not reveal any individual genes surpassing the defined thresholds for statistical and biological significance (| log₂ FC | > 1, FDR < 0.05). However, gene set enrichment analysis (GSEA) identified coordinated transcriptional changes across multiple biological pathways. AA tumors showed positive enrichment for keratinization25 and cornified envelope formation, which are processes related to epithelial cell differentiation, as well as for olfactory signaling pathways. Negative enrichment was observed for IgG immunoglobulin complex–related processes and ER protein-folding pathways, including the calnexin/calreticulin cycle. These findings suggest that while individual gene differences may be subtle, ancestry-associated differences in PCa can be captured at the pathway level.
Within AA patients, differential expression analysis identified 574 DEGs (330 upregulated, 244 downregulated) meeting both statistical and fold-change thresholds. The most upregulated genes included AMACR26 and C1QTNF3-AMACR, both previously implicated in PCa, as well as genes involved in metabolic processes (e.g., GSTA1) and neural/olfactory signaling (e.g., NPY, OR51E1)27,28. The most downregulated genes included immunoglobulin components (IGHG1, IGHGP, IGKV3-20, IGKV1-5, IGKC) and WIF1, a known Wnt pathway inhibitor29.
Pathway analysis revealed enrichment of prostate cancer–associated transcriptional signatures (LIU prostate cancer up) and proliferation markers (MKI67)30, along with pathways regulating xenobiotic detoxification and glycogen metabolism. These findings align with established roles of metabolic reprogramming and treatment resistance in tumor biology. Downregulated pathways included the TOMLINS prostate cancer dn signature and multiple epithelial-to-mesenchymal transition (EMT)–associated gene programs (GAVISH 3CA malignant metaprogram 12 EMT 1, GAVISH 3CA epithelial EMT like 1) as well as fibroblast/pericyte-like stromal programs and hypoxia-related gene sets.
Within EA patients, 330 DEGs (212 upregulated, 118 downregulated) met both statistical and fold-change thresholds. Upregulated genes included CRISP3, OR51E2, and PCAT14, all previously associated with PCa biology31,32,33. Downregulated genes featured TGM4, PENK, and SERPINB11, among others. Pathway enrichment indicated strong upregulation of multiple prostate cancer–specific signatures (LIU prostate cancer up, TOMLINS prostate cancer up), unfolded protein response, androgen–prostate epithelial programs, and protein synthesis machinery. Downregulated pathways were predominantly related to striated muscle development and contraction, myocardial fibrosis, and inflammatory signaling.
Implications
This study identifies distinct transcriptomic differences between African American and European American prostate tumors, highlighting ancestry-associated variation in pathways linked to keratinization, xenobiotic detoxification, immune complex formation, and stromal–epithelial interactions. The enrichment of keratinization-related pathways in African American tumors aligns with prior reports linking keratin family gene expression to aggressive disease subtypes and poorer androgen deprivation therapy response. Upregulation of xenobiotic detoxification pathways suggests an enhanced drug efflux capacity, potentially contributing to treatment resistance. Conversely, the downregulation of immunoglobulin complex pathways and TLR–NFκB signaling in European American tumors indicates suppression of specific inflammatory and immune processes, which may influence tumor–immune interactions. Collectively, these findings underscore the potential biological underpinnings of racial disparities in prostate cancer outcomes and point toward ancestry-informed therapeutic strategies and biomarker development.
Limitations
This study has several limitations that should be considered when interpreting the results. First, the RNA-seq data were obtained from a publicly available dataset (GSE104131) with a relatively small sample size (n = 28), which may limit the statistical power to detect subtle differences and increase the possibility of false negatives in differential expression analysis. Second, the analysis was restricted to a single cohort and lacked validation in an independent patient population, which is necessary to confirm the reproducibility and generalizability of the findings. Third, although all samples were processed under uniform experimental conditions, unmeasured biological variables such as treatment history, comorbidities, or tumor microenvironment heterogeneity could have influenced gene expression patterns. Fourth, the study design was observational and cross-sectional, precluding causal inferences regarding the functional role of the identified pathways in disease progression or racial disparities. Finally, while pathway-level analyses provided insights into coordinated biological changes, these approaches rely on existing gene set annotations, which may not capture novel or context-specific biology, particularly in underrepresented populations. Future work incorporating larger, racially diverse cohorts, multi-omics integration, and functional validation will be essential to further elucidate the biological mechanisms underlying prostate cancer disparities.
Conclusion
These findings reinforce the value of pathway-level analysis in detecting ancestry-associated molecular differences in PCa. They indicate both shared and distinct biological processes underlying tumorigenesis in AA and EA patients, which have the potential to inform future investigations into ancestry-specific disease mechanisms and the development of tailored therapeutic strategies and biomarkers.
References
- American Cancer Society. Key statistics for prostate cancer. https://www.cancer.org/cancer/types/prostate-cancer/about/key-statistics.html (2024). [↩]
- A. Elbialy, A. Sood, S.-J. Wang, P. Wang, A. Fadiel, A. V. Parwani, S. Huang, G. Shvets, N. Putluri, J. Li, X. Liu. Unveiling racial disparities in prostate cancer using an integrative genomic and transcriptomic analysis. Cell Insight. 4, 100238 (2025). https://doi.org/10.1016/j.cellin.2025.100238 [↩]
- R. Gulati, Y. A. Nyame, J. M. Lange, J. E. Shoag, A. Tsodikov, R. Etzioni. Racial disparities in prostate cancer mortality: a model-based decomposition of contributing factors. J. Natl. Cancer Inst. Monogr. 2023(62), 212–218 (2023). https://doi.org/10.1093/jncimonographs/lgad018 [↩]
- A. J. Vickers, B. Mahal, O. O. Ogunwobi. Racism does not cause prostate cancer, it causes prostate cancer death. J. Clin. Oncol. 41(12), 2151–2154 (2023). https://doi.org/10.1200/JCO.22.02203 [↩]
- Farrell, G. Petrovics, D. G. McLeod, S. Srivastava. Genetic and molecular differences in prostate carcinogenesis between African American and Caucasian American men. Int. J. Mol. Sci. 14(8), 15510–15531 (2013). https://doi.org/10.3390/ijms140815510 [↩]
- M. Kim, P. Tamukong, G. C. Galvan, Q. Yang, A. De Hoedt, M. R. Freeman, S. You, S. Freedland. Prostate cancers with distinct transcriptional programs in Black and White men. Genome Med. 16, 92 (2024). https://doi.org/10.1186/s13073-024-01361-0 [↩]
- B. A. Mahal, T. Gerke, S. Awasthi, H. R. Soule, J. W. Simons, A. Miyahira, S. Halabi, D. George, E. A. Platz, L. Mucci, K. Yamoah. Prostate cancer racial disparities: A systematic review by the Prostate Cancer Foundation Panel. Eur Urol Oncol 5(1), 18-29 (2022). https://doi.org/10.1016/j.euo.2021.07.006 [↩]
- P. Creixell, J. Reimand, S. Haider, G. Wu, T. Shibata, M. Vazquez, V. Mustonen, A. Gonzalez-Perez, J. Pearson, C. Sander, B. J. Raphael, D. S. Marks, B. F. F. Ouellette, A. Valencia, G. D. Bader, P. C. Boutros, J. M. Stuart, R. Linding, N. Lopez-Bigas, L. D. Stein. Mutation Consequences and Pathway Analysis Working Group of the International Cancer Genome Consortium. Pathway and network analysis of cancer genomes. Nat Methods 12(7), 615–621 (2015). https://doi.org/10.1038/nmeth.3440 [↩]
- J. S. Myers, A. K. von Lersner, C. J. Robbins, Q.-X. A. Sang. Differentially expressed genes and signature pathways of human prostate cancer. PLoS One 10(12), e0145322 (2015). https://doi.org/10.1371/journal.pone.0145322 [↩]
- E. A. Pudova, G. S. Krasnov, A. A. Kobelyatskaya, M. V. Savvateeva, M. S. Fedorova, V. S. Pavlov, K. M. Nyushko, A. D. Kaprin, B. Y. Alekseev, D. Y. Trofimov, G. T. Sukhikh, A. V. Snezhkina, A. V. Kudryavtseva. Gene expression changes and associated pathways involved in the progression of prostate cancer advanced stages. Frontiers in Genetics 11, 613162 (2021). https://doi.org/10.3389/fgene.2020.613162 [↩]
- L. Albarracín-Navas, M. Almonte-Becerril, E. Guerrero, J. Rivadeneira, M. Telechea-Fernández, E. Guzmán, F. Calderón, M. J. Hernández-Leal, T. Otzen, C. Manterola, G. Duque, Á. L. Riffo-Campos. Differential protein-coding gene expression profile in patients with prostate cancer. Biomedicines 12, 2509 (2024). https://doi.org/10.3390/biomedicines12112509 [↩]
- W. Rayford, A. T. Beksac, J. Alger, M. Alshalalfa, M. Ahmed, I. Khan, U. G. Falagario, Y. Liu, E. Davicioni, D. E. Spratt, E. M. Schaeffer, F. Y. Feng, B. Mahal, P. L. Nguyen, R. B. Den, M. D. Greenberger, R. Bradley, J. M. Watson, M. Beamer, L. Stamatakis, D. J. Carmen, S. Awasthi, J. Hwang, R. Weil, H. Merisaari, N. Mohamed, L. A. Deane, D. Chakravarty, K. K. Yadav, K. Yamoah, S. S. Nair, A. K. Tewari. Comparative analysis of 1152 African-American and European-American men with prostate cancer identifies distinct genomic and immunological differences. Communications Biology 4, 670 (2021). https://doi.org/10.1038/s42003-021-02140-y [↩]
- S. Ramakrishnan, E. Cortes-Gomez, S. R. Athans, K. M. Attwood, S. R. Rosario, S. J. Kim, D. E. Mager, E. G. Isenhart, Q. Hu, J. Wang, A. Woloszynska. Race-specific coregulatory and transcriptomic profiles associated with DNA methylation and androgen receptor in prostate cancer. Genome Medicine 16, 52 (2024). https://doi.org/10.1186/s13073-024-01323-6 [↩]
- E. L. Teslow, B. Bao, G. Dyson, C. Legendre, C. Mitrea, W. Sakr, J. D. Carpten, I. Powell, A. Bollig-Fischer. Exogenous IL-6 induces mRNA splice variant MBD2_v2 to promote stemness in TP53 wild-type, African American PCa cells. Molecular Oncology 12(7), 1138–1152 (2018). https://doi.org/10.1002/1878-0261.12316 [↩]
- National Center for Biotechnology Information. RNA-seq counts. https://www.ncbi.nlm.nih.gov/geo/info/rnaseqcounts.html (2024). [↩]
- M. I. Love, W. Huber, S. Anders. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology 15, 550 (2014). https://doi.org/10.1186/s13059-014-0550-8. [↩]
- B. T. Sherman, M. Hao, J. Qiu, X. Jiao, M. W. Baseler, H. C. Lane, T. Imamichi, W. Chang. DAVID: a web server for functional enrichment analysis and functional annotation of gene lists (2021 update). Nucleic Acids Res 50(W1):W216-W221 (2022). https://doi.org/10.1093/nar/gkac194. [↩]
- S. Xu, E. Hu, Y. Cai, Z. Xie, X. Luo, L. Zhan, W. Tang, Q. Wang, B. Liu, R. Wang, W. Xie, T. Wu, L. Xie, G. Yu. Using clusterProfiler to characterize multiomics data. Nature Protocols 19, 3292–3320 (2024). https://doi.org/10.1038/s41596-024-01020-z. [↩]
- I. Dolgalev. msigdbr: MSigDB gene sets for multiple organisms in a tidy data format. R package version 24.1.0 (2025). Available at: https://CRAN.R-project.org/package=msigdbr. [↩]
- G. Yu. enrichplot: Visualization of functional enrichment result. R package version 1.28.4 (2025). doi:10.18129/B9.bioc.enrichplot. Available at: https://bioconductor.org/packages/enrichplot. [↩]
- J. Farrell, D. Young, Y. Chen, J. Cullen, I. L. Rosner, J. Kagan, S. Srivastava, D. G. McLeod, I. A. Sesterhenn, S. Srivastava, G. Petrovics. Predominance of ERG-negative high-grade prostate cancers in African American men. Mol Clin Oncol 2, 982–986 (2014). doi:10.3892/mco.2014.378 [↩]
- C. M. Zeigler-Johnson, E. Spangler, M. Jalloh, S. M. Gueye, H. Rennert, T. R. Rebbeck. Genetic susceptibility to prostate cancer in men of African descent: implications for global disparities in incidence and outcomes. Can J Urol 15, 3872–3882 (2008). PMID: 18304397 [↩]
- J. W. Lillard Jr., K. A. Moses, B. A. Mahal, D. J. George. Racial disparities in Black men with prostate cancer: a literature review. Cancer 128, 3787–3795 (2022). doi:10.1002/cncr.34433 [↩]
- W. Liu, L. Zheng, R. Na, L. Wei, J. Sun, J. Gallagher, J. Wei, W. K. Resurreccion, S. Ernst, K. S. Sfanos, W. B. Isaacs, J. Xu. Distinct genomic alterations in prostate tumors derived from African American men. Molecular Cancer Research 18, 1815–1824 (2020). https://doi.org/10.1158/1541-7786.MCR-20-0648 [↩]
- J. O. Mori, J. White, I. Elhussin, B. M. Duduyemi, B. Karanam, C. Yates, H. Wang. Molecular and pathological subtypes related to prostate cancer disparities and disease outcomes in African American and European American patients. Frontiers in Oncology 12, 928357 (2022). https://doi.org/10.3389/fonc.2022.928357 [↩]
- N. Jiang, S. Zhu, J. Chen, Y. Niu, L. Zhou. A-Methylacyl-CoA racemase (AMACR) and prostate-cancer risk: A meta-analysis of 4,385 participants. PLOS ONE 8(10), e74386 (2013). https://doi.org/10.1371/journal.pone.0074386 [↩]
- A. Pronin, V. Slepak. Ectopically expressed olfactory receptors OR51E1 and OR51E2 suppress proliferation and promote cell death in a prostate cancer cell line. J Biol Chem 296, 100475 (2021). https://doi.org/10.1016/j.jbc.2021.100475. [↩]
- T. Henry, K. Wang, J. Pluznick, O. Zhuravel, S. Zozulya, V. Slepak. Characterization of novel agonists and antagonists for ectopically expressed olfactory receptor OR51E1. J Pharmacol Exp Ther 389(Suppl 3), 283 (2024). https://doi.org/10.1124/jpet.283.946950. [↩]
- Z. Xia, D. Du, Z. Zhang, Z. Liu, Z. Hu, X. Li, X. Guo, Z. He. WIF1 and DKK3 in prostate cancer: from molecular pathways to therapeutic targets: a narrative review. Transl Androl Urol 13(11):2601-2616 (2024). https://doi.org/10.21037/tau-24-304. [↩]
- Z. Song, Q. Zhou, J.-L. Zhang, J. Ouyang. Marker Ki-67 is a potential biomarker for the diagnosis and prognosis of prostate cancer based on two cohorts. World J Clin Cases 12(1):32-41 (2024). https://doi.org/10.12998/wjcc.v12.i1.32. [↩]
- M. Volpert, L. Furic, J. Hu, A. E. O’Connor, R. J. Rebello, S. Keerthikumar, J. Evans, D. J. Merriner, J. Pedersen, G. P. Risbridger, P. McIntyre, M. K. O’Bryan. CRISP3 expression drives prostate cancer invasion and progression. Endocr Relat Cancer 27(7):415-430 (2020). https://doi.org/10.1530/ERC-20-0092. [↩]
- M. T. Thomsen, M. Busk, D. Zhang, C. L. Chiu, H. Zhao, F. J. Garcia-Marques, A. Bermudez, S. Pitteri, M. Borre, J. D. Brooks, J. R. Nyengaard. The olfactory receptor OR51E2 regulates prostate cancer aggressiveness and modulates STAT3 in prostate cancer cells and in xenograft tumors. BMC Cancer 25, 535 (2025). https://doi.org/10.1186/s12885-025-13928-0. [↩]
- Y. Yan, J. Liu, Z. Xu, M. Ye, J. Li. lncRNA PCAT14 is a diagnostic marker for prostate cancer and is associated with immune cell infiltration. Dis Markers. 2021 Dec 31;2021:9494619. https://doi.org/10.1155/2021/9494619. [↩]
