An Integrative Transcriptomic Analysis of a Stemness-Associated Gene Set on Ovarian Cancer Recurrence

Technical Summary: An Integrative Transcriptomic Analysis of a Stemness-Associated Gene Set on Ovarian Cancer Recurrence

Problem Statement
Ovarian cancer (OC) remains a highly lethal gynecological malignancy with a five-year survival rate below 30% in advanced stages. Despite favorable initial responses to platinum-based chemotherapy, 70–80% of patients experience disease relapse. This recurrence is frequently attributed to the persistence of chemotherapy-resistant cancer stem cells (CSCs). While specific transcription factors (e.g., NANOG, SOX2) and metabolic regulators (e.g., ALDH family) are known to maintain stemness, there is a need to identify consistent gene expression patterns that specifically characterize the transition from primary to recurrent disease. The study aims to determine whether a defined panel of six stemness-associated genes exhibits reproducible expression alterations in recurrent OC (ROC) compared to primary OC (POC), thereby elucidating the transcriptional and metabolic reprogramming underlying tumor recurrence.

Methodology
The study employed an integrative approach combining large-scale public transcriptomic data with experimental validation:

1. Data Acquisition and Cohorts:

   • TCGA Cohort: RNA-seq data (FPKM normalized) from 421 primary OC (POC) and 62 recurrent OC (ROC) samples.
   • GEO Reference: A dataset (GSM4684836) of experimentally validated ovarian cancer stem cells (OCSCs) derived from high-grade serous ovarian cancer (HGSOC) primary tumors.
   • Clinical Validation Cohort: Fresh tissue samples from 8 POC and 8 ROC patients treated at Istanbul University Faculty of Medicine, collected before first-line chemotherapy and after documented recurrence, respectively.

2. Target Gene Panel:
   The analysis focused on six genes previously implicated in stemness, self-renewal, and chemoresistance: ALDH1A1, ALDH1A2, ALDH1A3, NANOG, PROM1 (CD133), and SOX2.

3. Computational Analysis:

   • Preprocessing: FPKM values were filtered (removing genes with FPKM < 10 in >80% of samples) and log₂-transformed [log₂(FPKM+1)].
   • Differential Expression: Parametric statistical tests with False Discovery Rate (FDR) correction were used to compare POC and ROC. Significance was defined as |log₂ fold change| > 1 and adjusted p < 0.05.
   • Correlation Analysis: Pearson correlation coefficients were calculated to assess the similarity of POC and ROC transcriptomes against the OCSC reference signature.
   • Survival Analysis: A composite stemness score (average z-score of the six genes) was calculated for TCGA patients. Kaplan-Meier curves and Cox proportional hazards regression were used to evaluate Overall Survival (OS).

4. Experimental Validation:

   • Total RNA was isolated from clinical samples and converted to cDNA.
   • Gene expression was quantified using qRT-PCR with GAPDH as the reference gene.
   • Relative expression was calculated using the ΔΔCq method, and concordance between qRT-PCR and RNA-seq fold-changes was assessed via Spearman correlation.

Key Results

• Differential Expression in TCGA Data:

  • Upregulated in ROC: SOX2 and PROM1 showed significant upregulation in recurrent tumors compared to primary tumors (adjusted p < 0.01).
  • Downregulated in ROC: ALDH1A1 expression was significantly decreased in ROC compared to POC (p < 0.01).
  • No Significant Change: ALDH1A2, ALDH1A3, and NANOG did not show statistically significant differential expression in the TCGA cohort.

• Correlation with Stem Cell Signature:

  • Primary tumors exhibited a negative correlation with the OCSC signature (r = -0.59), indicating a distinct expression profile from stem cells.
  • Recurrent tumors showed a near-neutral correlation (r = -0.04), suggesting a convergence toward an OCSC-like transcriptional state during recurrence.

• Survival Analysis:

  • Patients with higher composite stemness scores showed a trend toward worse Overall Survival; however, this difference was not statistically significant (log-rank p = 0.11).

• qRT-PCR Validation:

  • Consistent Findings: SOX2 and PROM1 were significantly upregulated in the recurrent clinical cohort (p < 0.01 and p < 0.05, respectively), corroborating the TCGA findings.
  • Divergent Findings: ALDH1A1 showed a statistically significant increase in the recurrent qRT-PCR cohort (p < 0.05), contrasting with the downregulation observed in the TCGA dataset.
  • Other Genes: ALDH1A3 was significantly elevated in the qRT-PCR cohort (p < 0.05) but not in TCGA. ALDH1A2 showed a non-significant increase, and NANOG showed a moderate, non-significant increase in both datasets.
  • Concordance: A weak-to-moderate positive correlation (Spearman r = 0.42) was observed between the fold-change values of the RNA-seq and qRT-PCR platforms.

Significance and Claims
The authors claim that ovarian cancer recurrence is driven by a specific transcriptional reprogramming centered on the reactivation of a limited subset of pluripotency regulators, specifically the SOX2–PROM1 axis.

• Robust Biomarkers: The consistent upregulation of SOX2 and PROM1 across both large-scale RNA-seq and small-scale qRT-PCR datasets positions them as reliable biomarkers for recurrence-associated stem cell activation.
• Heterogeneity of Metabolic Regulators: The study highlights that ALDH isoforms exhibit heterogeneous regulation (e.g., ALDH1A1 showing opposite trends in different cohorts), suggesting their expression is highly influenced by chemotherapy exposure, tumor subtype, and microenvironmental factors, rather than serving as a uniform marker of recurrence.
• Mechanistic Insight: The shift from a negative correlation (POC) to a neutral correlation (ROC) with the OCSC signature suggests that relapse does not arise from a distinct lineage but rather from tumor cell subsets that gradually regain stem cell-like transcriptional properties.
• Therapeutic Implication: The authors propose that targeting the SOX2–PROM1 transcriptional stem cell signature could be an effective strategy to eliminate resistant cell populations and improve long-term outcomes, acknowledging that the current findings are preliminary and require further functional validation.

The paper maintains a modest tone regarding clinical applications, noting limitations such as the smaller sample size of the ROC cohort in TCGA and the limited number of validated clinical samples, which may affect the generalizability of the survival analysis and the detection of subtle transcriptional differences.