Molecular mosaic of colorectal cancer: Why one classification system is no longer enough?

Abstract

Colorectal cancer (CRC) is one of the most molecularly heterogeneous malignancies, with complexity that extends far beyond traditional histopathological classifications. The consensus molecular subtypes (CMS) established in 2015 brought a marked advancement in the taxonomy of CRC, consolidating six classification systems into four novel subtypes, which focus on vital gene expression patterns and clinical and prognostic outcomes. However, nearly a decade of clinical experience with CMS classification has revealed fundamental limitations that underscore the inadequacy of any single classification system for capturing the full spectrum of CRC biology. The inherent challenges of the current paradigm are multifaceted. In the CMS classification, mixed phenotypes that remain unclassifiable constitute 13% of CRC cases. This reflects the remarkable heterogeneity that CRC shows. The tumor budding regions reflect the molecular shift due to CMS 2 to CMS 4 switching, causing further heterogeneity. Moreover, the reliance on bulk RNA sequencing fails to capture the spatial organization of molecular signatures within tumors and the critical contributions of the tumor microenvironment. Recent technological advances in spatial transcriptomics, single-cell RNA sequencing, and multi-omic integration have revealed the limitations of transcriptome-only classifications. The emergence of CRC intrinsic subtypes that attempt to remove microenvironmental contributions, pathway-derived subtypes, and stem cell-based classifications demonstrates the field’s recognition that multiple complementary classification systems are necessary. These newer molecular subtypes are not discrete categories but biological continua, thus highlighting that the vast molecular landscape is a tapestry of interlinked features, not rigid subtypes. Multiple technical hurdles cause difficulty in implementing the clinical translation of these newer molecular subtypes, including gene signature complexity, platform-dependent variations, and the difficulty of getting and preserving fresh frozen tissue. CMS 4 shows a poor prognostic outcome among the CMS subtypes, while CMS 1 is associated with poor survival in metastatic cases. However, the predictive value for definitive therapy remains subdued. Looking forward, the integration of artificial intelligence, liquid biopsy approaches, and real-time molecular monitoring promises to enable dynamic, multi-dimensional tumor characterization. The temporal and spatial complexity can only be captured by complementary molecular taxonomies rather than a single, unified system of CRC classification. Such an approach recognizes that different clinical questions – prognosis, treatment selection, resistance prediction – may require different molecular lenses, each optimized for specific clinical applications. This editorial advocates for a revolutionary change from pursuing a single “best” classification system toward a diverse approach that welcomes the molecular mosaic of CRC. Only through such comprehensive molecular characterization can we hope to achieve the promise of precision oncology for the diverse spectrum of patients with CRC.

Key Words: Consensus molecular subtypes; Heterogeneity; Colorectal cancer intrinsic subtypes; Pathway-derived subtypes; Clinical translation

Core Tip: Colorectal cancer (CRC) is not a single disease but a molecular mosaic. The consensus molecular subtypes brought important progress, yet nearly a decade of experience highlights its limitations - heterogeneity, unclassified cases, and bulk transcriptomic biases. Emerging systems such as CRC intrinsic subtypes, pathway-derived subtypes, and stem cell-based frameworks reveal complementary biological perspectives. Precision oncology for CRC requires abandoning the search for one “best” classification and instead embracing a pluralistic, multi-omic, and adaptive taxonomy that aligns with distinct clinical needs.

INTRODUCTION

Kang et al’s recent study[1], is an innovative and timely step toward combining the molecular and radiological aspects of managing colorectal cancer (CRC). The authors demonstrate a paradigm change in which imaging is no longer a stand-alone diagnostic tool but rather a dynamic biomarker reflecting the tumor’s underlying molecular architecture by utilizing magnetic resonance imaging-based prediction modeling in the context of particular genetic mutations. A crucial insight is highlighted by this convergence of radiomics and genomics: CRC is a molecular mosaic characterized by various mutational, transcriptomic, and microenvironmental circumstances rather than a singular disease entity. The biological and clinical heterogeneity of CRC cannot be adequately captured by a single categorization system, whether it be histological, molecular, or radiological, as our knowledge grows.

CRC is the 3^rd most common type of cancer worldwide and the 2^nd most common cause of death according to the GLOBOCAN 2022 data[2]. CRC is one of the central themes of solid cancer research, and a lot has been explored because of its high prevalence, clinical variability, and complex biology[3]. The last few years of research have shown that CRC is not just a single disease entity, but a complicated medley of molecular diversity, characterized by blending and dynamic evolution at multiple biological levels[4]. The editorial draws attention to the fundamental shift in understanding of the newer molecular classifications of CRC – the limitations of the consensus molecular subtypes (CMS), the evolution of CRC intrinsic subtypes (CRIS), pathway-derived subtypes (PDS), and numerous stem cell-based classifications. The editorial also advocates for relinquishing the quest for a single best classification system in favour of an adaptive, pluralistic molecular mosaic.

THE TRADITIONAL VIEW AND THE RISE OF CMS

The 20^th century for CRC diagnosis and management was all about histomorphology[5]. But as the molecular and genetic understanding progressed, CRC became one of the first solid tumors that was classified into novel molecular subtypes, based on characteristic biology and clinical outcome. In 2000, transcription-based systems were identified from six landmark studies[6]. These systems were recognized based on different platforms, analytic pipelines, and gene lists, highlighting little of the subtype boundaries. This is the reason CRC is known to have profound heterogeneity.

The limitations led the gastrointestinal pathologists to search for a more sophisticated classification system. So, the international CRC subtyping consortium collected data from more than 5000 tumors and, using a meta-analysis, they created the CMS in 2015[7]. This CMS classification was categorized into four subtypes[7-11].

CMS 1 (microsatellite instability-immune)

This tumor group is characterized by high microsatellite instability and hypermutation. This results from a defect in the mismatch repair of proteins. There is infiltration by immune cells, which most commonly comprises lymphocytes. This leads to an immune-related tumor microenvironment. Frequent BRAF mutations are also seen. Despite having an immune component, this group of tumors bears a poor prognosis in the metastatic setting. It shows poor response when compared to conventional chemotherapy.

CMS 2 (canonical)

It is the most common subtype. This group is characterized by chromosomal instability (CIN) with widespread DNA copy number alterations. WNT and MYC pathways are commonly activated. This group shows epithelial phenotype with intact cell adhesion. If it is localized, this group bears the best prognosis. Conventional chemotherapy usually shows excellent response, especially with fluoropyrimidine-based regimens.

CMS 3 (metabolic)

This group of tumors is characterized by prominent metabolic dysregulation, which includes metabolic enzyme expression and lipid metabolism. KRAS mutation pathway activation is common. Moderate immune infiltration is seen with less pronounced CIN compared to the CMS 2 group. Prognosis is intermediate. If metabolic pathways are targeted, therapeutic strategies may be beneficial. Sensitivity to chemotherapy is intermediate between CMS 2 and CMS 4.

CMS 4 (mesenchymal)

This group is characterized by high activation of epithelial-mesenchymal transition (EMT) genes. There is increased stromal infiltration by fibroblasts and angiogenic factors. Among all the subtypes, it has the worst prognosis. It is refractory to conventional chemotherapy and shows early metastasis and poor response to adjuvant chemotherapy treatment. Drugs that target stroma or angiogenesis may benefit this group. This classification suddenly gave hope to the clinicians, hoping for a common language, spurring trial design and retrospective translational studies to correlate molecular profiles with therapy outcomes.

THE LIMITS OF CMS: COMPLEXITY, OVERLAP, AND UNCLASSIFIED CASES

Despite the initial excitement of a novel subtyping system, the limitations of CMS have become increasingly apparent[12].

First, a group of tumors defy any particular subtype allocation and are called “mixed phenotype” – constitute about 13% of CRC[13]. This incidence becomes higher when spatial or single-cell genomics are applied. Regional CMS shifts, particularly from CMS 2 to CMS 4, are caused by zones of molecular and functional heterogeneity created by tumor budding regions and microenvironmental elements, which represent underlying cellular plasticity[13].

Second, the bulk RNA-sequencing technique used to diagnose CMS subtypes usually conflates cancer cell signals and microenvironment signals[14]. For example, CMS 1 and CMS 4 subtypes both have immune/stromal gene expression components, which may not accurately reflect their intrinsic biology and therapeutic response to treatment. This sets drawbacks for clinical translation This poses challenges for clinical translation – patients whose classification is driven by stromal signals may not benefit from medicines that target epithelial traits, and vice versa. There is a lack of practical biomarkers [like human epidermal growth factor receptor (EGFR) 2 in breast cancer] to use in daily routine practice, which prevents their adoption.

Despite these hurdles, the latest transcriptomic profiling platforms, such as NanoString, HTG EdgeSeq, and TempO-Seq, as well as targeted RNA-seq panels, can produce reliable data from formalin fixation paraffin-embedding samples[15]. These methods overcome the dependence on fresh-frozen tissue, facilitating retrospective analysis and the broader adoption of the classifiers in routine practice.

Recently, liquid biopsy has gained attention, with applications including circulating tumor DNA, exosomal RNA, and circulating tumor cells, which help monitor molecular subtype evolution and therapy response[16]. Artificial intelligence-based deconvolution algorithms are increasingly applied to bulk transcriptomic data to disentangle tumor and stromal signatures, enhancing the accuracy and reproducibility of molecular classification[17].

NEW FRONTIERS

The progressive refinement of CRC molecular taxonomy is evident from the emergence of new molecular subtypes, which include CRIS, PDS, and stem cell-based classifications[18-20]. These subtypes are designed to surmount the drawbacks of transcriptome-only bulk tumor classifications and better predict clinical behavior, therapy response, and resistance.

CRIS

CRIS exclusively focus on epithelial/tumor-intrinsic gene expression, which filters out stromal confounding factors[21,22]. By complete analysis of patient-derived xenografts, where human tumor stroma is replaced by murine components, five intrinsic subtypes have been identified.

CRIS-A (immune/hypermutated)

This subgroup has uncanny similarity with CMS 1. It has high microsatellite instability with hypermutation and POLE mutations. Because of high neoantigen load, there is strong immune signaling. This group is enriched in the BRAF V600E mutation. Interferon and immune checkpoint pathways are generally activated. Histology usually shows mucinous morphology. The tumor has intense immune infiltration by CD8+ T cells. It shows good early-stage prognosis and excellent response to immune checkpoint inhibitors (anti-programmed cell death 1, anti-cytotoxic T-lymphocyte-associated protein 4, etc.).

CRIS-B (transforming growth factor-beta/EMT/stromal-like)

Transforming growth factor-beta (TGF-β) activation signatures characterize this subgroup. There is high expression of EMT genes (ZEB1, SNAIL, TWIST) and activation of pro-angiogenic signaling. High stromal fibroblast infiltration with extracellular matrix deposition is seen. There is an overlap with CMS 4, but unlike CMS 4, CRIS-B is purely epithelial, intrinsic EMT-driven. It shows aggressive behavior with a poor prognosis and frequently presents with metastasis. Standard chemotherapy shows a limited response. This tumor group is a potential candidate for TGF-β inhibitors, stromal targeting therapies, and angiogenesis inhibitors.

CRIS-C (EGFR-activated)

This group of tumors is characterized by strong EGFR pathway activation. KRAS is wild type, and it shows MAPK/ERK pathway activation. Tumors usually have an epithelial phenotype. Anti-EGFR therapies (e.g., cetuximab, panitumumab) are generally effective in this group of tumors. EGFR inhibitors are beneficial in cases of metastatic CRC, too. If a KRAS/NRAS mutation is present, they show a poor prognosis. Targeted therapy with EGFR and MPAK blockade may be beneficial.

CRIS-D (WNT/MYC/stem-like)

This group is characterized by WNT-pathway activation (adenomatous polyposis coli mutation with β-catenin signaling). They show CIN and MYC oncogene activation. The tumor biology is of a stem-cell-like phenotype with a high proliferation index. Clinically, this group is very aggressive and shows resistance to EGFR inhibitors (despite being KRAS-wild type). WNT/MYC inhibitors could be effective in this group of tumors. Stemness-targeted strategies (Notch, Hedgehog, epigenetic therapies) are being explored.

CRIS-E (metabolic/KRAS-driven)

This group is characterized by enrichment of KRAS-driven mutation. There is a strong association with high metabolic reprogramming, such as glycolysis and oxidative phosphorylation. PI3K-AKT-mTOR pathway activation is frequent. This group of tumors shows metabolic plasticity, which can adapt under therapeutic stress (e.g., chemotherapy, targeted therapy, hypoxia, or nutrient deprivation). Because of the KRAS-driven mechanism, they do not respond to EGFR inhibitors. Metabolic inhibitors that target glycolysis, glutaminolysis, oxidative phosphorylation, etc., can be effective. Trials are being explored using KRAS G12C inhibitors and combination metabolic therapies.

PDS

PDS focus on pathway-level biological activity, which provides a direct link to the molecular mechanisms and clinical phenotypes[22,23].

PDS 1: Canonical/LGR5+ stem-rich tumors

It shows elevated cell-cycle-related pathways and canonical stem cell enrichment. The WNT pathway is upregulated. This group shows a favorable prognosis. It correlates with the CMS 2 subtype. Standard chemotherapy is effective.

PDS 2: Regenerative/ANXA1+ stem-rich tumors

A regenerative stem cell population with ANXA1+ expression characterizes it. It is strongly associated with TGF-β pathway activation, interferon responses, EMT, etc. CMS 1 and CMS 4 categories fall under this category. The BRAF V600E mutation is a common event.

PDS 3: Slow-cycling, differentiated tumor

These are previously overlooked slow-cycling subsets with CMS 2 subtypes. It shows reduced stem cell populations and differentiated lineages. In locally advanced disease, it carries the worst prognosis.

STEM CELL-BASED CLASSIFICATIONS

It is a sophisticated approach to understanding tumor heterogeneity. Cancer stem cells are a distinct population of tumor cells characterized by self-renewal capacity, multipotency, and tumor-initiating ability. Stem cell-based methods identify important participants in tumor growth, recurrence, and inter-patient variability, which enhances genomic and transcriptome subtyping[24].

When combined, CRIS, PDS, and stem cell-based classifications enhance CRC molecular taxonomy by catching subtleties beyond bulk gene expression and opening up new possibilities for precision oncology, such as resistance mitigation, robust prognostication, and informed therapy selection.

By allowing for more accurate patient stratification and treatment selection, complementary molecular classifications of CRC present a promising paradigm for incorporation into clinical trials and standard practice. These classifications, such as immune checkpoint inhibitors for immune-rich CMS 1 or anti-EGFR agents for CRIS-C subtypes, can be used to define molecularly homogeneous subgroups in clinical trials, enhancing the capacity to detect treatment effects and identify responders based on subtype-specific vulnerabilities. In order to categorize cancers into these complementary subtypes and make individualized treatment decisions based on the molecular and microenvironmental context of the tumor, routine clinical practice could use multiplexed molecular testing platforms, such as formalin fixation paraffin-embedding-compatible assays or liquid biopsies. Furthermore, incorporating liquid biopsies with real-time molecular monitoring enables dynamic surveillance of therapeutic resistance and tumor progression, enabling adaptive therapy adjustments. In order to improve prognosis and patient outcomes through customized therapies, our pluralistic molecular taxonomy approach promotes the creation of biomarker-driven clinical trials and precision oncology workflows that represent the intricate, heterogeneous biology of CRC.

TOWARD A MOLECULAR MOSAIC – ABANDONING THE “SINGLE SYSTEM” DOGMA

The molecular intricacy and dynamics of CRC thwart any attempt at single-system classification. Different clinical, prognostic, and research concerns are optimized for by each biological lens, including stem cell-based, CMS, CRIS, and PDS. It is now evident that the only practical route to precision oncology in CRC is to embrace complementary taxonomies, each tailored for specific settings and uses, rather than aiming for a universal taxonomy.

Such diversity acknowledges the demands of modern care: Distinct genetic predictors are needed for adjuvant therapy selection, resistance prediction, immunotherapy candidate identification, and relapse monitoring. An adaptive, evolving approach that integrates multi-omics, geographical, and longitudinal data shows the most promise because no single categorization can fulfill all functions.

CONCLUSION

An adaptable, pluralistic molecular framework incorporating real-time molecular dynamics, evaluating the illness from different perspectives, and adjusting classification to clinical requirements is currently necessary for CRC research and clinical care. Precision oncology can only effectively address the problem of CRC heterogeneity and improve patient outcomes by acknowledging the shortcomings of single-system categorization and embracing a heterogeneous biological mosaic.