Technical differences between sequencing and microarray platforms impact transcriptomic subtyping of colorectal cancer.

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.

The prognostic impact of consensus molecular subtypes (CMS) and its predictive effects for bevacizumab benefit in metastatic colorectal cancer: molecular analysis of the AGITG MAX clinical trial

Methylation of Cancer-Stem-Cell-Associated Wnt Target Genes Predicts Poor Prognosis in Colorectal Cancer Patients

BackgroundBroad copy number aberrations (BCNAs) represent a common form of genome instability in colorectal cancer (CRC). CRCs show large variations in their level of aneuploidy: microsatellite-instable (MSI) tumors are known to have a near-diploid karyotype while microsatellite-stable (MSS) tumors show high level of chromosomal instability. However, MSS tumors have great heterogeneity in the number of BCNAs, with a minor percentage of samples showing an almost normal karyotype. In the present work we subdivided MSS CRCs according to a “BCNA score” and characterized their transcriptome profiles, considered as a proxy to their phenotypic features.MethodsMicrosatellite testing, genome-wide DNA copy number and whole-transcript expression analysis (HTA) were performed on 33 tumor samples and 25 normal colonic tissue samples from 32 CRC patients. 15.1% of the samples were MSI tumors (n = 5), whereas 84.9% were MSS tumors (n = 28). Gene expression data of 34 additional MSI tumors was retrieved from a public functional genomics data repository.ResultsUsing as a threshold the first quartile of the BCNA score distribution, MSS samples were classified as low-BCNA (LB, n = 7) or high-BCNA (HB, n = 21). LB tumors were enriched for mucinous CRCs and their gene-expression profile resembled that of MSI samples for what concerns a subset of genes involved in secretory processes, mucosal protection, and extracellular matrix remodeling. HB tumors were predominantly non-mucinous adenocarcinomas and showed overexpression of a subset of genes typical of surface colonocytes and EGF signaling. A large percentage of unclassified samples according to the consensus molecular subtypes (CMS) classifier was found in the LB group (43%), whereas 76% HB tumors belonged to CMS2.ConclusionsA classification of colorectal tumors based on the number of BCNAs identifies two groups of MSS tumors which differ for histopathology and gene expression profile. Such information can be exploited for its translational relevance in different aspects of CRC clinical management.

Introduction Consensus Molecular Subtypes (CMS) represent a well-established molecular stratification framework for colorectal cancer (CRC). Existing methods for CMS classification rely on a representative input cohort as a preprocessing step (CMScaller) or have difficulty generalizing to metastatic samples (CMSclassifier). We developed Tempus CMS to overcome these limitations. Our method normalizes gene expression data from input samples to a static reference cohort to enable single sample classification of both primary and metastatic tumors. We evaluated the performance of this classifier on a large, de-identified CRC cohort comprising 8,489 samples from primary and metastatic sites. Methods To normalize input data, our method shifts and scales each gene expression value based on the mean and standard deviation of the expression of that gene in the reference cohort (n=2,787 primary and metastatic CRC). CMS calls are then generated via nearest template prediction similar to CMScaller (Eide et al., 2017). The analysis cohort was selected from clinical biopsies within 30 days of primary or metastatic diagnosis excluding reference cohort samples. CMS calls were assessed by comparing subtype prevalence to reported rates and by testing for known enrichments of pathways and genomic markers. Results Subtype prevalences for lower gastrointestinal (GI) biopsies and resections (n=5,090) were comparable to reported rates for primary CRC tumors: CMS1 - 12%, CMS2 - 28%, CMS3 - 19%, and CMS4 - 33% with 7.6% no calls. Biopsies from metastatic tissue showed site-specific changes in prevalence: CMS2 increased to 43% in liver samples (n=1,715) and CMS4 was identified in 49% of samples from metastatic sites excluding the liver (n=1,684). In addition, we found that purity had an effect on prevalence: in lower GI tumors with ≥ 50% tumor purity rates of CMS2 and CMS4 were 34% and 26%, respectively, while rates in tumors with purity between 20% and 50% were 20% CMS2 and 41% CMS4. Across all tissue groups, pathway enrichment showed significant associations, including TGF-ꞵ signaling in CMS4 (adj P ≤ 4.4e-15) and cell differentiation in CMS3 (adj P ≤ 3.0e-27). Expected enrichments of CMS markers were also observed: BRAF mutations and microsatellite instability in CMS1 and KRAS mutations in CMS3. In liver tissue, higher KRAS mutation rates were noted in CMS1 (57%) and CMS3 (83%) compared to GI tissue (35% and 67%, respectively). Conclusions In this study, we developed an enhanced, single sample CMS-calling algorithm optimized for primary and metastatic sites. Application of this algorithm in a large cohort recapitulated known biology, which suggests that the tool can be used to support clinical studies requiring robust molecular stratification. Citation Format: Andrew J. Sedgewick, Tushar Chandra, Timothy Taxter, Justin Guinney. Robust single sample consensus molecular subtype classification for primary and metastatic colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5059.

186 Background: INCA33890 is an anti–PD-1/TGFβR2 bispecific antibody designed to antagonize TGFβR2/PD-1 signaling in immune cells co-expressing both targets. In the INCA33890-101 study (NCT05836324), patients (pts) with refractory non-MSI-H mCRC were treated with INCA33890 monotherapy at 3 expansion doses. Responses were observed in pts with and without active liver metastases. Here, we analyze associations between baseline biomarkers and clinical outcomes. Methods: Fresh or archival (≤3 years from study registration) baseline biopsies were required in the dose expansion part and assessed using immunohistochemistry (IHC) for CD8 and PD-L1 (SP263) and whole transcriptome and next-generation sequencing for DNA aberrations (both Tempus xT). Baseline ctDNA was analyzed using Predicine Atlas during dose escalation and expansion. Efficacy was assessed by investigators per RECIST v1.1. Results: As of July 25, 2025, efficacy data with INCA33890 monotherapy were available from 105 pts; 16 pts (15.2%) had a partial response (PR), including pts with liver and peritoneal metastases (Table). Of the 78 pts with evaluable baseline biopsies for IHC analysis, pts with PR or stable disease (SD) showed greater PD-L1 tumor area positivity scores than those with progressive disease (PD) (nominal p<0.05). mRNA expression of inflammatory genes (IFN-γ, CXCL9, CXCL10 and T-cell inflamed signature) were higher in pts with PR/SD vs PD. Pts with blood tumor mutation burden (bTMB) >20 mut/MB had a trend towards higher ORR (33.3%) vs pts with bTMB <20 mut/MB (12.5%). Consensus molecular subtypes (CMS) were assessed in 44 pts. Although CMS 4 is commonly associated with the presence of TGF-β, it was not associated with a higher ORR (7.7% [1/13]). Responses were also observed in CMS 2 (21.7% [5/23]) and CMS 1 (50% [1/2]), but not in CMS 3 (n=6). There was no relationship with tumor mRNA expression of TGFβ or TGFβ-associated signatures with clinical outcomes. In addition, SMAD4 mutations, determined by ctDNA, were present in both responders and non-responders. Mutations in KRAS (64%), TP53 (72%), and APC (47%) were the most common aberrations and had no impact on clinical outcomes. Conclusions: INCA33890 demonstrated preliminary efficacy in pts with non-MSI-H mCRC across a variety of subgroups. Higher baseline PD-L1 expression and mRNA expression of inflammatory genes enriched for clinical benefit while overexpression of TGF-β did not seem to enhance efficacy. Updated data will be presented. Clinical trial information: NCT05836324 . Efficacy across subgroups. N ORR, % DCR, % Overall 105 15.2 28.6 With/without Liver Metastases 75/30 12.0/23.3 20.0/50.0 Peritoneal Metastases 40 10.0 20.0 No Liver or Peritoneal Metastases 20 30.0 60.0 PD-L1 TAP <1% 22 9.1 22.7 PD-L1 TAP >1% 56 17.9 25.0 PD-L1 TAP >5% 13 38.5 61.5 bTMB high >20 mut/MB 15 33.3 40.0 bTMB <20 mut/MB 64 12.5 28.1

2587 Background: Dysregulation of HMTs plays an important role in tumorigenesis. KMT2C and KMT2D are enzymatically active scaffold proteins that form the core of mammalian COMPASS complexes, which methylate the histone 3 lysine 4. Both KMT2C and KMT2D are involved in the regulation of gene expression. Therefore, we explored the associations of KMT2C/D loss-of-function (LOF) mutations with the expression of immune-related genes, the levels of tumor infiltrating lymphocytes (TILs), and response to immune checkpoint inhibitors (ICIs). Methods: KMT2C/D LOF mutations were defined as nonsense, frameshift, splice site variants within consensus regions, start lost, and stop lost/gained variants. An ICIs treatment cohort from the MSKCC was used for exploring the associations between KMT2C/D LOF mutations and ICIs efficacy. The RNA-Seq data obtained from the TCGA cohort was used for analysis of gene expression and the levels of TILs using CIBERSORT. Results: In MSKCC pan-cancer dataset, patients with KMT2C/D LOF mutations had a relatively longer median overall survival (OS) compared to those with non-LOF mutations, although the result did not reach statistical significance (P = 0.0832). Then we analyzed the predictive roles of KMT2C/D LOF mutations for each cancer type. The results showed that the predictive role of KMT2C/D LOF mutations for the clinical efficacy of ICIs therapy was only observed in colorectal cancer (P = 0.045). However, we did not find the associations of KMT2C/D LOF mutations with ICIs efficacy in bladder cancer, breast cancer, melanoma, glioma, head and neck cancer, renal cell carcinoma, NSCLC, and esophagogastric cancer. Consistently, analysis of TILs in colorectal cancer revealed that KMT2C/D LOF was associated with increased infiltration of several types of immune cells, such as CD8+ T cells (P = 0.0001), activated NK cells (P = 0.0001), M1 macrophage (P = 0.0005), M2 macrophage (P = 0.0115), and neutrophils (P = 0.0209). Meanwhile, regulatory T cells (Tregs) (P = 0.0048) and M0 macrophage (P = 0.0043) were dramatically decreased in KMT2C/D LOF group for colorectal cancer. Moreover, there were no significant relationships between KMT2C/D LOF and the levels of TILs in other cancer types. Our data also demonstrated that KMT2C and KMT2D could regulate the expression of more than 30 immune-related genes in colorectal cancer. Conclusions: Our data indicated that KMT2C/D LOF mutations were significantly correlated with better outcomes of ICIs therapy in colorectal cancer, suggesting it can be as a useful predictor for response to ICIs in colorectal cancer. Meanwhile, we found the associations of KMT2C/D LOF with the levels of TILs in colorectal cancer, but not in other cancer types, indicating that the efficacy of ICIs was consistent with the levels of TILs.

Background: The recommended cut-off of < 1% positive cells to define estrogen receptor (ER) negative status in breast cancer (BC) is highly debated. Few data exist on transcriptomic features of ER-low BC, suggesting no major differences vs ER-neg. Among HER2-negative BCs, we compared gene-expression profiling (including PAM50 intrinsic subtyping) according to ER status (ER-neg: 0% vs. ER-low: 1-9%) and assessed its impact on Tumor-infiltrating lymphocytes (TILs) levels. Methods: Across two cancer centers, stage I-III HER2-negative BCs FFPE samples were reviewed for ER status. All ER-low (ER 1-9%) samples were identified and a cohort of ER-neg BC samples, matched for age and stage, was also retrieved. A small cohort of pts with intermediate ER expression (ER-int: 10-50%) was included as a control cohort. Expression of 776 BC–related genes was evaluated by nCounter® (Breast Cancer 360TM Panel) and intrinsic molecular subtyping was determined using the PAM50 subtype predictor. A False Discovery Rate (FDR) corrected unpaired two-class SAM was used to identify genes differentially expressed in different subgroups. TILs were evaluated on archival H&E slides following guidelines. Results: Of 116 stage I-III HER2-neg BC pts included, 39 had ER-neg BC, 65 had ER-low BC, and 12 had ER-int disease. PAM50 intrinsic subtype distribution was similar in ER-neg and ER-low BCs, (with an enrichment in basal-like tumors), while both subgroups differed significantly from ER-int samples (Table). As compared to ER-neg tumors, ER-low BCs showed significantly higher expression of GATA3 gene and lower expression of EDN1 and PROM1 genes by SAM analysis (FDR< 5%); however, when the analysis was limited to PAM50 basal-like tumors (N=77) we only identified two genes that were downregulated in ER-low BCs (EDN1, PROM1) as compared to ER-neg BCs. On the contrary, significant differences in transcriptomic regulation were observed between ER-low and ER-int BCs, with ER-low BCs showing significantly higher expression of 53 genes and lower expression of 398 genes, as compared to ER-int (SAM analysis FDR< 5%). We previously reported similar TIL levels in ER-low and ER-neg tumors. In both ER-low and ER-int BCs, PAM50 basal-like subtype was associated with significantly higher TILs: median TIL levels 20 (range 0-80) and 6 (range 1-40) for PAM50 Basal-like BCs versus other PAM50 subtypes respectively in ER-low samples (p< 0.001); median TIL levels 53 (range 25-80) and 5 (range 0-10) for PAM50 Basal-like BCs versus other PAM50 subtypes respectively in ER-int samples (p=0.036). No significant difference in TIL levels according to PAM50 subtyping were observed in ER-neg tumors. No significant difference in mRNA levels for PD-1 (PDCD1) and PD-L1 (CD274) genes was observed between ER-low and ER-neg tumors. Conclusions: Among patients with HER2-neg BC, gene-expression profiling is similar in ER-neg (i.e., < 1%) vs. ER-low tumors with a high prevalence of PAM50 Basal-like tumors and few differentially expressed genes. In both ER-low and ER-int HER2-neg BCs, PAM50 Basal-like subtype is also associated with higher TIL levels. Our results strongly indicate that HER2-neg/ER-low BCs and HER2-neg/ER-neg BCs are similar biological entities, thus further supporting the use of similar treatments in patients with early-stage ER-low and triple-negative BC, including immune checkpoint inhibitors. PAM50 intrinsic subtype distribution according to ER levels Citation Format: Gaia Griguolo, Claudio Vernieri, Davide Massa, Lorenzo Nicolè, Claudia Pinato, Federica Miglietta, Andrea Vingiani, Silvia Brich, Riccardo Lobefaro, Francesca Schiavi, Matteo Fassan, Giancarlo Pruneri, Valentina Guarneri, Maria Vittoria Dieci. Transcriptomic profiling of ER-negative and ER-low, HER2-negative breast cancer and its implications on immune regulation [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-03-09.

3552 Background: Prognostic biomarkers beside RAS/BRAF status are necessary to identify metastatic colorectal cancer (mCRC) patients who benefit from combined (COMB) versus sequential (SEQ) treatment with fluoropyrimidine, bevacizumab and irinotecan (randomized phase III XELAVIRI trial). Methods: mRNA was extracted from formalin-fixed paraffin embedded (FFPE) tumor tissue of 337 patients, gene expression was measured by the Nanostring PanCancer Progression Panel. Consensus molecular subtypes (CMS) classification was re-derived using a multinomial regression model. Data of Guinney et al. (Nat. Med. 2015. 21:1350-6) and FIRE-3 served as training and validation set. RAS/BRAF MUT were assessed by pyrosequencing. Median overall (OS) and progression free survival (PFS), hazard ratios (HR) and 95% confidence interval (CI) were estimated by Kaplan-Meier method and univariate Cox regression. Results: The multinomial regression model employed in the present analysis correctly predicted CMS labels in 98.3 % of the original Guinney- and 100.0 % of FIRE-3 population. In XELAVIRI, CMS subgroups were predicted as follows: CMS1: n = 62 (18.4 %); CMS2: n = 174 (51.6 %); CMS3: n = 9 (2.7 %); CMS4: n = 92 (27.3 %). A general prognostic impact of CMS was not observed when all patients were analysed. In RAS/BRAF WT mCRC patients, substantial benefit of COMB versus SEQ treatment was shown for OS and PFS in CMS2 and CMS4, but not in CMS1. Conversely, OS was significantly longer for COMB treatment in patients with RAS MUT and CMS1 mCRC, while SEQ treatment was not inferior in RAS MUT and CMS2 or CMS4 subgroups (see TABLE). Additional data for overall response rates, early tumor shrinkage and sidedness might be presented at the meeting. Conclusions: This retrospective analysis of XELAVIRI suggests that CMS may serve as biomarker that predicts response to initially combined versus less intensive sequential chemotherapy in patients with RAS/BRAF WT mCRC.[Table: see text]

Recent work built on published transcriptional subtypes of colorectal cancer (CRC) has shown that tumor stromal content heavily impacts CRC classification, with clinical and biological implications. Lineage-dependent transcriptional components of stromal origin could therefore dominate over more subtle, yet relevant, gene expression traits inherent to cancer cells. Based on the notion that in patient-derived xenografts (PDXs) stromal cells of the human tumor are substituted by their murine counterparts, we assessed cancer-cell intrinsic transcriptional features by human-specific expression profiling on 515 PDXs from 244 CRC patients. We identified five CRC intrinsic subtypes (CRIS) endowed with molecular, functional and phenotypic peculiarities, many of which were not previously evidenced: CRIS-A was a MIN-like, KRAS-mutated, and secretory subtype with sustained glycolytic metabolism and inflammatory traits; CRIS-B grouped a subset of poorly differentiated tumors with active TGFβ signaling and EMT features; CRIS-C clustered KRAS wild-type CIN tumors exhibiting elevated EGFR pathway activity and MYC copy number gains; CRIS-D tumors were characterized by a typical stem phenotype with high WNT signaling, associated with strong enrichment in IGF2 amplification/overexpression and FGFR autocrine stimulation; CRIS-E again displayed WNT-related features, but associated with a Paneth-like phenotype and a TP53-mutated genotype. CRIS subtypes successfully categorized independent sets of primary and metastatic CRC tumors, with limited overlap on existing transcriptional classes and unprecedented predictive and prognostic performances. Integration of stroma-centered transcriptional profiles with cancer-cell intrinsic classification by CRIS further enhanced prognostic significance. Citation Format: Andrea Bertotti, Claudio Isella, Sara E. Bellomo, Francesco Brundu, Francesco Galimi, Elisa Ficarra, Livio Trusolino, Enzo Medico. Unsupervised analysis of cancer-cell intrinsic transcriptional traits defines a new classification system for colorectal cancer with improved predictive and prognostic value. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 107.

<p>Colorectal cancer (CRC) manifests as gastrointestinal tumors with high intratumoral heterogeneity. Recent studies have demonstrated that CRC may consist of tumor cells with different consensus molecular subtypes (CMS). The advancements in single-cell RNA sequencing have facilitated the development of gene regulatory networks to decode key regulators for specific cell types. Herein, we comprehensively analyzed the CMS of CRC patients by using single-cell RNA-sequencing data. CMS for all malignant cells were assigned using CMScaller. Gene set variation analysis showed pathway activity differences consistent with those reported in previous studies. Cell–cell communication analysis confirmed that CMS1 was more closely related to immune cells, and that monocytes and macrophages play dominant roles in the CRC tumor microenvironment. On the basis of the constructed gene regulation networks (GRNs) for each subtype, we identified that the critical transcription factor ERG is universally activated and upregulated in all CMS in comparison with normal cells, and that it performed diverse roles by regulating the expression of different downstream genes. In summary, molecular subtyping of single-cell RNA-sequencing data for colorectal cancer could elucidate the heterogeneity in gene regulatory networks and identify critical regulators of CRC.</p>

Colorectal cancer (CRC) manifests as gastrointestinal tumors with high intratumoral heterogeneity. Recent studies have demonstrated that CRC may consist of tumor cells with different consensus molecular subtypes (CMS). The advancements in single-cell RNA sequencing have facilitated the development of gene regulatory networks to decode key regulators for specific cell types. Herein, we comprehensively analyzed the CMS of CRC patients by using single-cell RNA-sequencing data. CMS for all malignant cells were assigned using CMScaller. Gene set variation analysis showed pathway activity differences consistent with those reported in previous studies. Cell–cell communication analysis confirmed that CMS1 was more closely related to immune cells, and that monocytes and macrophages play dominant roles in the CRC tumor microenvironment. On the basis of the constructed gene regulation networks (GRNs) for each subtype, we identified that the critical transcription factor ERG is universally activated and upregulated in all CMS in comparison with normal cells, and that it performed diverse roles by regulating the expression of different downstream genes. In summary, molecular subtyping of single-cell RNA-sequencing data for colorectal cancer could elucidate the heterogeneity in gene regulatory networks and identify critical regulators of CRC.

<p>Colorectal cancer (CRC) manifests as gastrointestinal tumors with high intratumoral heterogeneity. Recent studies have demonstrated that CRC may consist of tumor cells with different consensus molecular subtypes (CMS). The advancements in single-cell RNA sequencing have facilitated the development of gene regulatory networks to decode key regulators for specific cell types. Herein, we comprehensively analyzed the CMS of CRC patients by using single-cell RNA-sequencing data. CMS for all malignant cells were assigned using CMScaller. Gene set variation analysis showed pathway activity differences consistent with those reported in previous studies. Cell–cell communication analysis confirmed that CMS1 was more closely related to immune cells, and that monocytes and macrophages play dominant roles in the CRC tumor microenvironment. On the basis of the constructed gene regulation networks (GRNs) for each subtype, we identified that the critical transcription factor ERG is universally activated and upregulated in all CMS in comparison with normal cells, and that it performed diverse roles by regulating the expression of different downstream genes. In summary, molecular subtyping of single-cell RNA-sequencing data for colorectal cancer could elucidate the heterogeneity in gene regulatory networks and identify critical regulators of CRC.</p>

Colorectal cancer (CRC) manifests as gastrointestinal tumors with high intratumoral heterogeneity. Recent studies have demonstrated that CRC may consist of tumor cells with different consensus molecular subtypes (CMS). The advancements in single-cell RNA sequencing have facilitated the development of gene regulatory networks to decode key regulators for specific cell types. Herein, we comprehensively analyzed the CMS of CRC patients by using single-cell RNA-sequencing data. CMS for all malignant cells were assigned using CMScaller. Gene set variation analysis showed pathway activity differences consistent with those reported in previous studies. Cell–cell communication analysis confirmed that CMS1 was more closely related to immune cells, and that monocytes and macrophages play dominant roles in the CRC tumor microenvironment. On the basis of the constructed gene regulation networks (GRNs) for each subtype, we identified that the critical transcription factor ERG is universally activated and upregulated in all CMS in comparison with normal cells, and that it performed diverse roles by regulating the expression of different downstream genes. In summary, molecular subtyping of single-cell RNA-sequencing data for colorectal cancer could elucidate the heterogeneity in gene regulatory networks and identify critical regulators of CRC.

Colorectal cancer (CRC) is an aggressive disease and the leading cause of cancer death, characterised by high heterogeneity and various risk factors related to its etiology (1-3). During carcinogenesis, adenomatous polyps, which represent most premalignant lesions (85-90% of sporadic CRC), can develop into CRC (4-6). About 20-25% of CRC patients are diagnosed with metastatic disease, which is associated with poorer survival rates. CRC can spread to various tissues, including lymph nodes, liver, lungs, peritoneum, bones, and the central nervous system (7). The liver is the most common site for detecting metastatic CRC and is involved in 25% to 50% of cases. However, a rare form of CRC with bone metastasis occurs in 3% to 7% of patients; these individuals often have worse survival outcomes, with shorter survival times and limited treatment options (8-10). TGF-β and BMP signalling pathways are crucial mechanisms in tissue homeostasis, promoting cell proliferation and differentiation during crypt formation in the intestine (11-15). Dysregulation of these pathways in intestinal cells can impair their tumour-suppressing functions and facilitate tumour development (16-18). In particular, mutations in TGFBR2, BMPR1A, and SMAD4 have been identified as contributing to CRC carcinogenesis (2, 3, 19-23). Patients with consensus molecular subtype (CMS) 4, categorised based on gene expression signature, are associated with prominent activation of TGF-β, along with stromal infiltration, epithelial to mesenchymal transition (EMT), and angiogenesis, leading to poorer survival outcomes (19). In silico analysis of gene expression levels of TGF-β and BMP signalling components across CRC cell lines using the DepMap database, along with examination of key proteins and their phosphorylation statuses through Western blots, revealed that ligand activation occurs at various levels, with notable activation of upstream receptors such as TGFBR2, ACVR1B, and BMPR1A. Increased expression of signal transduction genes, including SMAD1, SMAD2, SMAD3, and SMAD5, was also observed. Interestingly, although phosphorylation of SMAD1/5/8 was detected in SW620 cell lines (a metastatic CRC cell line), SMAD4, a central element in signal transduction, was found to be downregulated. Apart from the differential expression of TGF-β and BMP signalling in CRC, the investigation of gene expression in these pathways within disease-free bone marrow (BM) cells using the Stemformatics database was also investigated to provide some insight into whether this morphogenic pathway is involved in metastatic spread to the bone. Here we demonstrated high levels of TGF-β and BMP signalling pathways in the bone marrow associated cells, indicating maintenance of homeostasis and serving as a baseline reference for further research. Assessment of epithelial SMAD4 expression in colorectal polyps using immunohistochemistry (IHC) and digital weighted histoscoring with QuPath revealed that low SMAD4 levels in adenomatous polyps correlated with higher grades of dysplasia, different histological subtypes, the presence of metachronous polyps, and served as a prognostic marker. This marker indicated an increased risk of developing metachronous polyps, particularly in the tubulovillous polyp subtype. Additionally, transcriptomic analysis of tubulovillous polyps showed that upregulation of genes involved in protein deubiquitination occurs in polyps with low SMAD4 levels, along with a likely enrichment in tyrosine metabolism, PPAR signalling, arginine and proline metabolism, leukocyte transendothelial migration, and basal cell carcinoma. In CRC, lower epithelial SMAD4 expression was strongly associated with higher tumour stages and increased tumour stroma. Moreover, lower SMAD4 expression in CRC tumours had prognostic significance, predicting decreased cancer-specific survival in CRC patients, especially in right-sided tumours. A combination of tumour SMAD4 levels and stroma percentages suggested that the worst survival outcome was in patients with low SMAD4 expression in the tumour and high stroma content. Transcriptomic analysis also identified downregulation of SOD3 and enrichment of aminopeptidase activity in this group. Differential expression patterns at gene and protein levels, along with phosphorylation activity, were observed across the complex crosstalk of MAPK/ERK, WNT/β-catenin, TGF-β/BMP, and PI3K signalling pathways in CRC cell lines. However, the mechanisms regulating SMAD4 activity in CRC remain unclear.Testing the TGF-β and BMP signalling inhibitor (LDN-212854) in combination with the standard chemotherapy (Fluoropyrimidine; 5-FU) showed synergistic effects on CRC cell viability, cell cycle arrest, cell proliferation, and cell recovery in an in vitro 2D study. Conditioned media from BM and hepatic cells influence changes in CRC behaviour. The 3D bioprinted SW620 spheroids in 2% alginate and 8% gelatin hydrogel supported physiological interactions, spheroid survival and growth, and were used for drug screening, demonstrating efficacy of the LDN-212854/5-FU combination in a more mechanophysical 3D system. Further development of this in vitro 3D model to become multicell by incorporating metastatic CRC culture with BM and hepatic niches would enhancefuture CRC research and drug discovery.