Selective analysis of cancer-cell intrinsic transcriptional traits defines novel clinically relevant subtypes of colorectal cancer

Claudio Isella, Francesco Brundu, Sara E. Bellomo, Francesco Galimi, Eugenia R. Zanella, Roberta Porporato, Consalvo Petti, Alessandro Fiori, Francesca Orzan, Rebecca Senetta, Carla Boccaccio, Elisa Ficarra, Luigi Marchionni, Livio Trusolino, Enzo Médico, Andrea BertottiNature Communications 2017 · DOI

TL;DR

This paper tackles the challenge of classifying colorectal cancer (CRC) by focusing on the cancer cells' intrinsic transcriptional traits, rather than being confounded by the tumor's stromal environment. They developed a novel classification system (CRIS) using patient-derived xenografts (PDXs) that identifies five distinct subtypes with unique molecular, functional, and clinical characteristics, offering improved prognostic and predictive power.

What Problem Does This Paper Address?

Traditional methods for classifying colorectal cancer (CRC) based on gene expression have been significantly influenced by the tumor's stromal microenvironment. This "stromal noise" can mask the subtle, yet crucial, transcriptional signatures inherent to the cancer cells themselves. These intrinsic cancer cell traits are thought to be more directly linked to tumor behavior, aggressiveness, and response to therapy. The authors aimed to disentangle these cancer-cell specific signals to uncover more accurate and clinically relevant CRC subtypes.

Methods

The core of their approach involved using patient-derived xenografts (PDXs). PDXs are created by implanting human tumors into immunocompromised mice. Crucially, the mouse stroma replaces the human stroma, allowing researchers to profile the human cancer cells in a relatively pure state. They used human-specific expression profiling on these PDXs to identify cancer-cell intrinsic transcriptional features.

Key techniques included:

  • Non-negative Matrix Factorization (NMF) for unsupervised clustering of PDX gene expression data to identify potential subtypes.
  • Significance Analysis of Microarrays (SAM) and Prediction Analysis for Microarrays (PAM) to identify genes that best discriminate these subtypes.
  • Nearest Template Prediction (NTP) to build a classifier based on these identified genes.
  • Gene Set Enrichment Analysis (GSEA) and Sample Set Enrichment Analysis (SSEA) to characterize the functional and phenotypic properties of the identified subtypes.
  • Development of a simplified single-sample classifier (CRIS-TSP) using the Top Scoring Pair (TSP) algorithm for potential clinical application.

Datasets

  • Training Data: A large collection of 515 patient-derived xenografts (PDXs) from 244 colorectal cancer (CRC) patients.
  • Validation Data:
    • CRC liver metastases (CRC-LM) data sets.
  • The Cancer Genome Atlas (TCGA) RNAseq data set (450 samples).
  • GSE14333 Affymetrix data set (290 samples) of primary CRCs.
  • 14 additional independent data sets, including liver metastases and CRC cell lines.
  • A data set of 241 PDXs annotated for response to cetuximab.
  • A published data set (GSE5851) of CRC liver metastases annotated for cetuximab response.
  • A publicly available and clinically annotated CRC data set (GSE14333) for prognostic analysis.
  • An independent cohort of 1,261 samples for prognostic validation.

Key Insights and Ideas

  • Stromal Confounding: The study strongly reinforces the idea that stromal components heavily influence bulk tumor gene expression, often obscuring cancer-cell intrinsic signals.
  • PDXs as a Purification Tool: PDXs are invaluable for isolating cancer-cell specific transcriptional profiles by replacing human stroma with mouse stroma.
  • Novel Subtypes (CRIS): The identification of five distinct "CRIS" (Colorectal Cancer Intrinsic Subtypes) that are not simply rearrangements of existing classifications.
  • CRIS Subtypes are Biologically Distinct: Each CRIS subtype is characterized by unique molecular alterations (mutations, copy number changes), functional pathways (e.g., WNT, EGFR, TGF-β), and phenotypic features.
  • Predictive Power: CRIS subtypes show significant potential for predicting response to therapies, particularly anti-EGFR antibodies.
  • Independent Prognostic Value: CRIS-B, in particular, emerges as a strong independent predictor of poor prognosis, even when accounting for clinical stage and stromal infiltration.

Key Results

  • CRIS Classification: Five distinct CRC intrinsic subtypes (CRIS-A through CRIS-E) were identified.
  • CRIS-A: Mucinous, glycolytic, associated with MSI or KRAS mutations.
  • CRIS-B: TGF-β pathway activity, epithelial-mesenchymal transition (EMT), linked to poor prognosis.
  • CRIS-C: Elevated EGFR signaling, sensitive to EGFR inhibitors.
  • CRIS-D: WNT activation, IGF2 overexpression/amplification.
  • CRIS-E: Paneth cell-like phenotype, TP53 mutations.
  • High Classification Accuracy: The CRIS classifier achieved high confidence assignment rates (e.g., 94% in PDXs, >90% in TCGA and GSE14333).
  • Therapeutic Prediction: CRIS-C was significantly enriched in cetuximab-sensitive tumors and depleted in resistant ones, showing predictive power independent of known genetic resistance markers.
  • Prognostic Prediction: CRIS-B was identified as a negative prognostic predictor, with a significant association with worse disease-free survival (DFS). This was independent of tumor stage and chemotherapy.
  • Integration with Stromal Data: Combining CRIS-B status with CAF (cancer-associated fibroblast) score provided even greater prognostic power, allowing for refined risk stratification.

Limitations and Future Work

  • PDX Limitations: The authors acknowledge that PDXs might introduce some biases due to xenotransplantation. They suggest future work using alternative methods like cell sorting or microdissection to validate findings on pure cancer cells.
  • Clinical Translation: While a simplified classifier (CRIS-TSP) was developed, its performance outside the training set was less robust. Further validation and standardization are needed for clinical application.
  • Stromal Influence: Despite focusing on intrinsic traits, the authors emphasize that stromal components remain important and suggest integrating both intrinsic and stromal signatures for optimal patient management.
  • Prospective Validation: The prognostic implications of CRIS subtypes, particularly the combined CRIS-B/CAF score, require prospective validation in larger cohorts.
  • Therapeutic Targeting: The identified subtypes and their associated pathways (e.g., TGF-β, WNT, IGF2) represent potential targets for novel therapeutic strategies, which warrant further investigation.

Summary updated

Abstract

Stromal content heavily impacts the transcriptional classification of colorectal cancer (CRC), with clinical and biological implications. Lineage-dependent stromal transcriptional components could therefore dominate over more subtle expression traits inherent to cancer cells. Since in patient-derived xenografts (PDXs) stromal cells of the human tumour are substituted by murine counterparts, here we deploy human-specific expression profiling of CRC PDXs to assess cancer-cell intrinsic transcriptional features. Through this approach, we identify five CRC intrinsic subtypes (CRIS) endowed with distinctive molecular, functional and phenotypic peculiarities: (i) CRIS-A: mucinous, glycolytic, enriched for microsatellite instability or KRAS mutations; (ii) CRIS-B: TGF-β pathway activity, epithelial-mesenchymal transition, poor prognosis; (iii) CRIS-C: elevated EGFR signalling, sensitivity to EGFR inhibitors; (iv) CRIS-D: WNT activation, IGF2 gene overexpression and amplification; and (v) CRIS-E: Paneth cell-like phenotype, TP53 mutations. CRIS subtypes successfully categorize independent sets of primary and metastatic CRCs, with limited overlap on existing transcriptional classes and unprecedented predictive and prognostic performances.
BibTeX
@inproceedings{isella2017selective,
  author = {Claudio Isella and Francesco Brundu and Sara E. Bellomo and Francesco Galimi and Eugenia R. Zanella and Roberta Porporato and Consalvo Petti and Alessandro Fiori and Francesca Orzan and Rebecca Senetta and Carla Boccaccio and Elisa Ficarra and Luigi Marchionni and Livio Trusolino and Enzo Médico and Andrea Bertotti},
  title = {{Selective analysis of cancer-cell intrinsic transcriptional traits defines novel clinically relevant subtypes of colorectal cancer}},
  booktitle = {Nature Communications},
  year = {2017},
  doi = {10.1038/ncomms15107},
}
Figure