Precursor-dependent steroid conversion sustains androgen receptor signaling and lineage-state divergence in prostate cancer

Technical Summary: Precursor-dependent steroid conversion sustains androgen receptor signaling and lineage-state divergence in prostate cancer

Problem Statement
Castration-resistant prostate cancer (CRPC) persists despite systemic androgen deprivation, a phenomenon frequently attributed to autonomous de novo intratumoral steroidogenesis, often conceptualized as the tumor functioning as a "pseudo-gonad." While upstream inhibitors like abiraterone (targeting CYP17A1) are clinically deployed, resistance inevitably emerges. The structural and metabolic fidelity of the "tumor-as-gonad" dogma remains controversial; it is unclear whether prostate tumors possess a transcriptionally complete steroidogenic pathway capable of synthesizing androgens from cholesterol, or if they rely on converting circulating adrenal precursors via downstream enzymatic machinery. Furthermore, the relationship between this metabolic topology, lineage plasticity, and therapeutic vulnerability is poorly defined.

Methodology
The study employed an integrated multi-omics approach combining in vitro cell line characterization with large-scale clinical data analysis:

• Cell Line Models: A panel of human prostate cancer cell lines (LNCaP, VCaP, 22Rv1, DU145, PC3), non-malignant prostate epithelial cells (RWPE-1), and an adrenal steroidogenic control (NCI-H295R) were utilized.
• Functional Assays: Cell viability was assessed under serum-depleted conditions with and without exogenous dihydrotestosterone (DHT) to map androgen dependency gradients.
• Steroidomics: Targeted Liquid Chromatography-High Resolution Mass Spectrometry (LC-HRMS) was performed on conditioned media following pregnenolone supplementation to quantify absolute steroid flux and metabolic output.
• Transcriptomics: RNA sequencing (RNA-seq) was conducted across all models to analyze gene expression, pathway enrichment (GSEA, GSVA), and steroidogenic enzyme architecture.
• Clinical Cohort: Transcriptomic and clinical data from 818 patients in the TCGA-PRAD cohort were analyzed. Stratification was based on Androgen Receptor (AR) activity and lineage plasticity scores.
• Statistical Modeling: Survival analysis utilized Cox proportional hazards models and an explainable machine learning framework (SHAP) to deconstruct non-linear contributions of molecular features to patient risk, addressing limitations of standard linear models due to feature collinearity.

Key Results

1. Precursor-Dependent Topology: Contrary to the "tumor-as-gonad" hypothesis, prostate cancer models lack the transcriptional capacity for de novo steroidogenesis from cholesterol. LC-MS profiling and transcriptomics revealed a consistent, marked suppression of CYP17A1 (the rate-limiting enzyme for converting pregnenolone/progesterone to androgens) and an absence of a functional cortisol-producing branch. Instead, these cells retain downstream conversion machinery (HSD3B1, AKR1C3, SRD5A1), functioning as efficient converters of adrenal precursors (e.g., DHEA) into active androgens.
2. Lineage Divergence and Metabolic Rewiring: The study identified a continuous evolutionary trajectory between two distinct states:
   • AR-High Phenotype: Characterized by intense metabolic hyperactivity (oxidative phosphorylation, fatty acid metabolism), robust AR signaling, and dependence on downstream steroid conversion.
   • AR-Low/Plastic Phenotype: Marked by extensive lineage plasticity (EMT, mesenchymal states), loss of luminal identity, and an elevated iron-lipid redox stress profile. These cells upregulate ferroptosis defense mechanisms (GPX4, SLC7A11) to survive the loss of AR-mediated antioxidant protection.
3. Clinical Correlation: In the TCGA-PRAD cohort, AR activity was strongly correlated with downstream converting enzymes (e.g., SRD5A1, r=0.59) but showed zero correlation with CYP17A1 expression. Unsupervised clustering revealed four distinct patient archetypes ranging from AR-high/luminal to AR-low/mesenchymal.
4. Risk Prediction: Standard linear Cox models failed to isolate independent predictors due to high collinearity among metabolic and signaling features. However, a SHAP-based machine learning framework identified AR activity (24.7% relative contribution) and Plasticity score (21.3% relative contribution) as the dominant drivers of survival risk, outperforming traditional metrics like Gleason score (9.1%).

Significance and Claims
The paper claims to redefine the metabolic paradigm of CRPC by demonstrating that localized androgenic output is dictated by metabolic topology (downstream conversion efficiency) rather than pathway completeness (upstream synthesis). The authors argue that prostate tumors are not autonomous steroidogenic organs but are metabolically constrained systems that scavenge circulating adrenal precursors.

The study posits that the "cortisol void" (loss of glucocorticoid synthesis) and the preservation of downstream androgen converters are critical adaptive mechanisms. Furthermore, it establishes a mechanistic link between the loss of AR signaling and the acquisition of a specific redox vulnerability (iron-lipid stress), suggesting that lineage-plastic, AR-low tumors rely on alternative survival networks (e.g., ferroptosis resistance).

Therapeutic Implications
Based on these findings, the authors advocate for a shift in therapeutic strategy:

• Downstream Targeting: Moving beyond upstream CYP17A1 inhibition to target downstream cytosolic conversion nodes (e.g., AKR1C3, 3β-HSD) to block intracrine androgen synthesis.
• Lineage-Plasticity Vulnerabilities: Exploiting the specific redox dependencies of AR-low, plastic clones (e.g., via ferroptosis induction or targeting PI3K/AKT pathways) to eradicate treatment-resistant populations.

The authors conclude that recognizing the prostate tumor as a "broken but highly adaptive metabolic pathway" provides a framework for predicting and neutralizing the evolutionary plasticity of advanced prostate cancer.