Integrative analysis of TCGA transcriptomic states and DepMap dependencies prioritizes candidate vulnerabilities in immune-cold microsatellite-stable colorectal cancer

Technical Summary: Integrative Analysis of TCGA Transcriptomic States and DepMap Dependencies in MSS CRC

Problem Statement
Microsatellite-stable or microsatellite instability-low colorectal cancer (MSS/MSI-L) represents the majority of colorectal cancer (CRC) cases and is generally resistant to immune checkpoint blockade (ICB). While the "immune-cold" phenotype is a known characteristic of MSS CRC, the term often obscures substantial biological heterogeneity. Current therapeutic strategies struggle because MSS CRC resistance arises from diverse mechanisms, including true immune desert states, stromal barriers, and tumor-intrinsic evasion pathways. Furthermore, distinguishing between tumor-cell-intrinsic vulnerabilities and patient-derived microenvironmental signatures is critical for target prioritization, yet bulk transcriptomic analyses often conflate these signals. There is a need for a framework that stratifies MSS CRC into distinct transcriptomic states and integrates patient expression data with functional dependency maps to identify actionable, state-specific vulnerabilities.

Methodology
The study employed an integrative bioinformatics approach combining patient transcriptomics with functional genomics data:

1. Cohort Selection and Classification:

   • The analysis utilized RNA-seq data from 494 MSS/MSI-L tumors from the TCGA COAD/READ PanCancer Atlas cohort, classified using an MSIsensor score <3.5.
   • Tumors were stratified into four transcriptomic states based on curated gene-expression modules: MSS immune-cold (low cytotoxic/IFNγ activity), MSS hot/inflamed (high cytotoxic/IFNγ activity), MSS barrier-high (high stromal/ECM activity), and MSS intermediate.
   • Module scores were derived from z-scored expression of genes representing cytotoxic T/NK activity, IFNγ-chemokine signaling, antigen presentation, CAF/ECM barriers, and other relevant pathways.

2. Differential Analysis:

   • Pairwise comparisons were performed (MSS immune-cold vs. MSS hot/inflamed; MSS barrier-high vs. MSS immune-cold) using Mann-Whitney U tests with Benjamini-Hochberg correction to identify differentially expressed genes and pathways.

3. Dependency Integration:

   • Patient expression data were integrated with CRISPR-Cas9 gene-effect data from the Cancer Dependency Map (DepMap).
   • The analysis focused on 63 colorectal cancer models within a broader set of 1,208 cancer models.
   • Gene-effect scores (more negative values indicating stronger dependency) were calculated for candidate genes. A "common-essentiality penalty" was applied to distinguish broad cell-cycle dependencies from state-specific vulnerabilities.

4. Target Prioritization:

   • Candidates were ranked by integrating four evidence layers: (1) immune-cold association in patient data, (2) barrier-high association, (3) DepMap dependency scores in CRC models, and (4) biological/druggability annotations.
   • A key analytical distinction was made between genes representing tumor-cell-intrinsic dependencies (captured by DepMap) and those representing microenvironmental barriers (captured by patient transcriptomics but not DepMap).

Key Results

• Heterogeneity of MSS CRC: The MSS/MSI-L cohort was not uniform. Of the 494 tumors, 44.1% were classified as MSS intermediate, 20.6% as MSS immune-cold, 18.4% as MSS hot/inflamed, and 16.8% as MSS barrier-high.
• Immune-Cold Phenotype: MSS immune-cold tumors showed significant depletion of cytotoxic and inflammatory programs compared to MSS hot/inflamed tumors. Key downregulated genes included NKG7, CD8A, CXCL9, CXCL10, and LAG3, indicating reduced cytotoxic lymphocyte activity and IFNγ-mediated recruitment.
• Barrier-High Phenotype: MSS barrier-high tumors were characterized by elevated stromal and extracellular matrix (ECM) programs, specifically COL1A1, COL1A2, and COL3A1. These were interpreted as stromal-barrier markers rather than tumor-cell dependencies.
• Dependency Integration:
  • Core mitotic regulators (e.g., AURKB, AURKA) and DNA-damage response genes (e.g., ATR) showed strong dependency in CRC cell lines but were noted as potentially broad essentialities.
  • ERBB2 (HER2): Emerged as the highest-ranked clinically actionable candidate. ERBB2 expression was significantly higher in MSS immune-cold tumors compared to MSS hot/inflamed tumors and further elevated in MSS barrier-high tumors. However, DepMap data showed only modest dependency, suggesting ERBB2 marks a specific actionable subset rather than a universal dependency.
  • Other Candidates: The analysis prioritized VEGFA (angiogenic/immune-suppressive axis), PIK3CB (PI3K pathway), DNA-damage checkpoint kinases (ATR, WEE1, CHEK1), epigenetic regulators (HDAC1, HDAC3, BRD4), and apoptosis regulators (BCL2L1, MCL1).
  • Stromal Markers: Collagen genes (COL1A1/2/3) ranked highly in patient expression but were correctly identified as non-DepMap dependencies, reinforcing the necessity of separating microenvironmental barriers from tumor-cell vulnerabilities.

Significance and Claims
The paper posits that MSS CRC should not be treated as a single immunotherapy-resistant entity but rather as a collection of distinct transcriptomic states requiring different therapeutic strategies.

• Stratified Framework: The study proposes a "state-matched" therapeutic framework where MSS immune-cold tumors may require immune priming and tumor-cell sensitization, while MSS barrier-high tumors may require stromal or vascular remodeling before checkpoint blockade can be effective.
• Methodological Contribution: By integrating bulk patient transcriptomics with DepMap dependency data, the study successfully separates tumor-cell-intrinsic vulnerabilities from microenvironmental barrier signatures. This prevents the misinterpretation of stromal markers (like collagen) as druggable tumor-cell targets.
• Hypothesis Generation: The authors explicitly frame their findings as a reproducible, hypothesis-generating framework rather than a clinically validated classifier. The prioritized candidates (e.g., ERBB2, VEGFA, ATR/WEE1, HDAC/BRD4, BCL2L1/MCL1) are nominated for future experimental validation in mechanistic models (e.g., organoid-immune co-cultures) to test their ability to convert cold MSS CRC into inflamed states or sensitize tumors to therapy.
• Limitations: The authors acknowledge limitations including the use of bulk RNA-seq (lacking spatial resolution), the absence of stromal dependencies in cell-line-based DepMap data, and the lack of independent cohort validation or survival modeling. Consequently, the identified targets are presented as candidates for future validation, not confirmed therapeutic strategies.