Refining Colorectal Cancer Classification and Clinical Stratification Through a Single-Cell Atlas

Genome Biology, Tempus-authored Manuscript
Authors Ateeq M. Khaliq, Cihat Erdogan, Zeyneb Kurt, Sultan Sevgi Turgut, Miles W. Grunvald, Tim Rand, Sonal Khare, Jeffrey A. Borgia, Dana M. Hayden, Sam G. Pappas, Henry R. Govekar, Audrey E. Kam, Jochen Reiser, Kiran Turaga, Milan Radovich, Yong Zang, Yingjie Qiu, Yunlong Liu, Melissa L. Fishel, Anita Turk, Vineet Gupta, Ram Al-Sabti, Janakiraman Subramanian, Timothy M. Kuzel, Anguraj Sadanandam, Levi Waldron, Arif Hussain, Mohammad Saleem, Bassel El-Rayes, Ameen A. Salahudeen, & Ashiq Masood


Colorectal cancer (CRC) consensus molecular subtypes (CMS) have different immunological, stromal cell, and clinicopathological characteristics. Single-cell characterization of CMS subtype tumor microenvironments is required to elucidate mechanisms of tumor and stroma cell contributions to pathogenesis which may advance subtype-specific therapeutic development. We interrogate racially diverse human CRC samples and analyze multiple independent external cohorts for a total of 487,829 single cells enabling high-resolution depiction of the cellular diversity and heterogeneity within the tumor and microenvironmental cells.


Tumor cells recapitulate individual CMS subgroups yet exhibit significant intratumoral CMS heterogeneity. Both CMS1 microsatellite instability (MSI-H) CRCs and microsatellite stable (MSS) CRC demonstrate similar pathway activations at the tumor epithelial level. However, CD8+ cytotoxic T cell phenotype infiltration in MSI-H CRCs may explain why these tumors respond to immune checkpoint inhibitors. Cellular transcriptomic profiles in CRC exist in a tumor immune stromal continuum in contrast to discrete subtypes proposed by studies utilizing bulk transcriptomics. We note a dichotomy in tumor microenvironments across CMS subgroups exists by which patients with high cancer-associated fibroblasts (CAFs) and C1Q+TAM content exhibit poor outcomes, providing a higher level of personalization and precision than would distinct subtypes. Additionally, we discover CAF subtypes known to be associated with immunotherapy resistance.


Distinct CAFs and C1Q+ TAMs are sufficient to explain CMS predictive ability and a simpler signature based on these cellular phenotypes could stratify CRC patient prognosis with greater precision. Therapeutically targeting specific CAF subtypes and C1Q + TAMs may promote immunotherapy responses in CRC patients.