Grilled and Smoked Food Carcinogen Benzo[a]pyrene Linked to Prostate Cancer Through Four Key Genes

Benzo[a]pyrene, a polycyclic aromatic hydrocarbon produced whenever organic matter burns incompletely, is one of the most ubiquitous carcinogens in the human environment. It coats charred meat, laces cigarette smoke, vehicle exhaust, and industrial soot, and the International Agency for Research on Cancer has long placed it in Group 1, its highest category of confirmed human carcinogens. What has remained far less clear is whether this molecule plays a direct role in prostate cancer, a disease that kills hundreds of thousands of men each year and whose environmental triggers continue to be debated. A new study published in Molecular Diversity by Siqi Zhu, Zhuang Li, and colleagues at Guizhou Medical University and Guizhou Provincial People’s Hospital now offers the most systematic picture yet of how BaP exposure may drive prostate tumors, combining network toxicology, machine learning, transcriptomics, single-cell sequencing, molecular docking, and laboratory validation into a single integrated pipeline.

The starting point of the investigation was a computational strategy known as network toxicology, which treats a chemical, its protein targets, and human disease genes as nodes in an interconnected graph rather than as isolated entities. The researchers first assembled a comprehensive list of proteins that benzo[a]pyrene is known or predicted to interact with, drawing on established toxicogenomic databases and target-prediction tools. They then cross-referenced this list with genes associated with prostate cancer. The overlap was striking: 975 shared targets emerged, suggesting that BaP and prostate tumors converge on an unexpectedly large common molecular territory. When these 975 genes were mapped onto known biological pathways, the enrichment analysis pointed decisively toward the PI3K-Akt signaling axis, one of the most frequently hijacked growth and survival circuits in human malignancy, along with related networks governing cell proliferation, apoptosis resistance, and immune modulation.

Identifying 975 candidate genes is a useful beginning, but cancer biology demands sharper focus. To distill the signal, the team applied differential expression analysis to prostate cancer transcriptomic data and then deployed machine learning algorithms to select which of the differentially expressed genes best separated tumor from healthy tissue. Two complementary algorithms converged on four core genes: CAV1, TWIST1, PRKCA, and GDF15. Each of these has an established biography in cancer research. CAV1 encodes caveolin-1, a scaffolding protein of membrane caveolae with documented roles in prostate cancer progression and treatment resistance. TWIST1 is a transcription factor best known for orchestrating epithelial-to-mesenchymal transition, the cellular program that allows stationary epithelial cells to become invasive and migratory. PRKCA encodes protein kinase C alpha, described in recent work as a central node of tumorigenic transcriptional networks in the prostate. GDF15, or growth differentiation factor 15, is a stress-response cytokine whose levels correlate with aggressive and castration-resistant disease.

With the four core genes in hand, the investigators turned to transcriptomic verification. The expression pattern in prostate tumor datasets was coherent and directional: TWIST1 and GDF15 were up-regulated in cancer tissue, while CAV1 and PRKCA were down-regulated. More impressive was the diagnostic performance. When the four genes were combined into a single diagnostic model, the area under the receiver operating characteristic curve reached 0.990, a value approaching the theoretical maximum of 1.0 and indicating near-perfect discrimination between cancerous and non-cancerous samples. The authors emphasize that this four-gene signature could distinguish BaP-relevant molecular states in prostate tissue with a precision that single biomarkers rarely achieve, positioning the panel as both a mechanistic fingerprint of pollutant-driven carcinogenesis and a candidate clinical diagnostic tool.

Because tumors are not merely collections of malignant cells but complex ecosystems of immune, stromal, and epithelial populations, the team next examined how the core genes behave at single-cell resolution. Single-cell RNA sequencing data revealed the specific cellular distribution of each gene within the prostate tumor microenvironment and documented immunological shifts associated with their expression. Immune infiltration analysis at the bulk-tumor level reinforced the connection: the core genes were linked to cellular growth programs and to the recruitment of immune cells into tumors. This immunological dimension matters because previous research has suggested that BaP exposure can play an immunosuppressive role during prostate cancer progression, potentially helping tumors evade surveillance. The new findings place that immune remodeling on a firmer molecular footing, tying it to genes that are themselves responsive to BaP exposure.

To test whether the carcinogen could physically engage its putative targets, the researchers performed molecular docking, a computational technique that predicts how a small molecule fits into the binding pocket of a protein. The docking simulations showed that benzo[a]pyrene binds favorably to the core targets, providing a structural rationale for the associations uncovered by the network analysis. Docking results are inherently approximate and do not prove physiological binding in living cells, but within the study’s multi-layered design they serve as an important plausibility check, bridging the gap between statistical gene associations and the physical chemistry of pollutant-protein interactions.

Perhaps the most consequential portion of the work is its external validation. Computational pipelines in toxicology are sometimes criticized for generating elegant networks that evaporate under experimental scrutiny, so the authors tested their predictions with laboratory and dataset evidence outside the discovery pipeline. The validation confirmed that BaP may promote the progression of prostate cancer and reproduced the expected expression shifts: GDF15 up-regulated and PRKCA down-regulated in prostate cancer contexts linked to BaP exposure. The convergence of the computational prediction and the external evidence strengthens the causal narrative considerably, indicating that the four-gene axis is not an artifact of a single database or analytical choice but a reproducible molecular pattern.

The study arrives amid a growing wave of network toxicology applied to environmental carcinogens in urological cancers. Recent publications have used similar frameworks to implicate polycyclic aromatic hydrocarbons in reproductive health outcomes, to trace the oncogenic pathways of aristolochic acids across prostate, kidney, and bladder cancers, and to dissect the contributions of phthalates such as diethyl phthalate and DEHP to prostate carcinogenesis. Epidemiological work has also lent real-world weight to the hypothesis: occupational exposure to polycyclic aromatic hydrocarbons has been associated with elevated prostate cancer risk in case-control studies. What distinguishes the new paper is the breadth of its validation stack, extending from machine-learning biomarker selection through single-cell immunology to molecular docking and external experimental confirmation, all focused on a single Group 1 carcinogen that nearly every person encounters daily through diet, air, and tobacco smoke.

The implications cut in two directions. Clinically, a four-gene diagnostic panel with an AUC of 0.990 suggests that pollutant-driven molecular signatures could eventually complement prostate-specific antigen testing, which suffers from well-documented problems of overdiagnosis and poor specificity. If GDF15, for example, is already being explored as a circulating marker of response to docetaxel chemotherapy in metastatic castration-resistant prostate cancer, integrating exposure-linked gene panels into diagnostic algorithms could help identify which tumors are environmentally fueled and potentially which patients might benefit from interventions targeting the PI3K-Akt pathway, where several inhibitors are already under investigation in hormone-related cancers. Public health officials, meanwhile, may find in these results additional mechanistic justification for limiting BaP exposure through food preparation guidance, tobacco control, and air quality regulation, since the molecular evidence now traces a plausible route from charred protein and diesel soot to the transcriptional circuits of prostate tumors.

The authors are careful about scope. Their conclusions are framed as BaP potentially promoting prostate cancer development through regulation of CAV1, TWIST1, GDF15, and PRKCA, with causal certainty limited by the correlative nature of much of the transcriptomic evidence and by the approximate nature of docking predictions. The DU145 prostate cancer cell line used in their experiments is commercially available, and the study received support from the Guizhou Provincial Health Commission and related provincial research programs. Still, the work exemplifies a methodological shift that is transforming environmental oncology: instead of asking whether a chemical damages DNA, researchers now map the full network of its molecular conversations with human tissue, then test the strongest nodes experimentally. For a carcinogen as widespread as benzo[a]pyrene, and a cancer as prevalent as prostate cancer, that map may prove to be one of the most valuable public health documents of the decade.

Subject of Research: Molecular mechanisms linking benzo[a]pyrene exposure to prostate cancer

Article Title: Exploring the molecular mechanism of benzo[a]pyrene affecting prostate cancer based on network toxicology and external validation

Article References: Zhu, S., Li, Z., Jiang, K., Sun, F., & Zhu, J. (2026). Exploring the molecular mechanism of benzo[a]pyrene affecting prostate cancer based on network toxicology and external validation. Molecular Diversity. https://doi.org/10.1007/s11030-026-11729-6

Image Credits: AI Generated

DOI: 10.1007/s11030-026-11729-6

Keywords: benzo[a]pyrene, prostate cancer, network toxicology, machine learning, CAV1, TWIST1, PRKCA, GDF15, PI3K-Akt pathway, single-cell analysis, molecular docking, environmental carcinogens