Consensus Molecular Subtypes Research Articles

Primary Colorectal Tumor Displays Differential Genomic Expression Profiles Associated with Hepatic and Peritoneal Metastases.

Despite improvements in characterization of CRC heterogeneity, appropriate risk stratification tools are still lacking in clinical practice. This study aimed to elucidate the primary tumor transcriptomic signatures associated with distinct metastatic routes. Primary tumor specimens obtained from CRC patients with either isolated LM (CRC-Liver) or PM (CRC-Peritoneum) were analyzed by transcriptomic mRNA sequencing, gene set enrichment analyses (GSEA) and immunohistochemistry. We further assessed the clinico-pathological associations and prognostic value of our signature in the COAD-TCGA independent cohort. We identified a significantly different distribution of Consensus Molecular Subtypes between CRC-Liver and CRC-peritoneum groups. A transcriptomic signature based on 61 genes discriminated between liver and peritoneal metastatic routes. GSEA showed a higher expression of immune response and epithelial invasion pathways in CRC-Peritoneum samples and activation of proliferation and metabolic pathways in CRC-Liver samples. The biological relevance of RNA-Seq results was validated by the immunohistochemical expression of three significantly differentially expressed genes (ACE2, CLDN18 and DUSP4) in our signature. In silico analysis of the COAD-TCGA showed that the CRC-Peritoneum signature was associated with negative prognostic factors and poor overall and disease-free survivals. CRC primary tumors spreading to the liver and peritoneum display significantly different transcriptomic profiles. The implementation of this signature in clinical practice could contribute to identify new therapeutic targets for stage IV CRC and to define individualized follow-up programs in stage II-III CRC.

Age-related genes affecting the immune cell infiltration in ulcerative colitis revealed by weighted correlation network analysis and machine learning.

The crosstalk between age and immunity in the context of ulcerative colitis (UC) remains incompletely understood. Our objective is to elucidate the specific age-associated genetic factors that modulate immune cell infiltration in UC, with the aim of identifying innovative therapeutic targets for the treatment of this disease. Potential batch effects between samples were removed by R package "inSilicoMerging". Unsupervised clustering analysis via the "ConsensusClusterPlus" R package was utilized to perform consensus molecular subtyping of immune subtypes in UC. The construction of a heat map was accomplished through the utilization of the R package "pheatmap", while functional enrichment analysis was executed by means of the Metascape database. The identification of the age-related gene module was achieved by performing weighted gene co-expression network analysis (WGCNA) analysis using the R package "WGCNA". The support vector machine (SVM), least absolute shrinkage and selector operation (LASSO), and random forest algorithms were performed via the "e1071", "glmnet" and "randomForest" packages in R, respectively. The diagnostic performance of the parameter was assessed using the receiver operating characteristic (ROC) curve. Correlation analysis was performed by Spearman correlation. The "XSum" package in R was employed to identify potential small-molecule drugs for UC utilizing the Connectivity Map (CMap) database. Molecular docking was performed with Autodock Vina molecular docking software. A significantly greater frequency of UC patients aged below 40 years was observed in the group with extensive disease extent as compared to those with non-extensive disease extent (70% vs. 47%; Chi-square test, p = 0.02). The application of unsupervised clustering analysis allowed for the stratification of UC patients into two distinct immune subtypes, namely cluster C1 and cluster C2. The distribution of immune subtypes was significantly different between different age categories (Chi-square test, p = 0.00219). The UC samples that were grouped under cluster C1 were distinguished by a higher abundance of macrophages and an elevated number of neutrophils relative to those in cluster C2. Based on both WGCNA and Limma analysis, 146 age-related genes were identified, which exhibited a predominant enrichment in the biological process of cellular senescence. Two age-related genes (MIDN, and PLD6) affecting the immune cell infiltration in UC were identified based on machine learning algorithms (SVM, LASSO, and random forest). The diagnostic performance of MIDN (AUC = 0.93) and PLD6 (AUC = 0.90) in discerning UC patients belonging to cluster C1 was found to be satisfactory, as demonstrated by ROC curve analysis. MIDN demonstrated a positive correlation (r = 0.50, p < 0.0001) with Neutrophil, while PLD6 exhibited a negative correlation (r = -0.52, p < 0.0001) with Neutrophil levels. The "XSum" algorithm revealed that Entinostat has therapeutic potential for UC. The docking glide score between Entinostat and MIDN, and PLD6 protein was -8.9 kcal/mol and -6.8 kcal/mol, respectively. We have identified two age-related genes, MIDN and PLD6, that are involved in immune cell infiltration in patients with ulcerative colitis. Furthermore, a small molecule drug (Entinostat) with potential therapeutic effects for UC was screened out. This study presented new perspectives on personalized clinical management and therapy research for UC.

Long Non-Coding RNAs in Colorectal Cancer: Navigating the Intersections of Immunity, Intercellular Communication, and Therapeutic Potential.

This comprehensive review elucidates the intricate roles of long non-coding RNAs (lncRNAs) within the colorectal cancer (CRC) microenvironment, intersecting the domains of immunity, intercellular communication, and therapeutic potential. lncRNAs, which are significantly involved in the pathogenesis of CRC, immune evasion, and the treatment response to CRC, have crucial implications in inflammation and serve as promising candidates for novel therapeutic strategies and biomarkers. This review scrutinizes the interaction of lncRNAs with the Consensus Molecular Subtypes (CMSs) of CRC, their complex interplay with the tumor stroma affecting immunity and inflammation, and their conveyance via extracellular vesicles, particularly exosomes. Furthermore, we delve into the intricate relationship between lncRNAs and other non-coding RNAs, including microRNAs and circular RNAs, in mediating cell-to-cell communication within the CRC microenvironment. Lastly, we propose potential strategies to manipulate lncRNAs to enhance anti-tumor immunity, thereby underlining the significance of lncRNAs in devising innovative therapeutic interventions in CRC.

Bacterial lipopolysaccharide modulates immune response in the colorectal tumor microenvironment

Immune responses can have opposing effects in colorectal cancer (CRC), the balance of which may determine whether a cancer regresses, progresses, or potentially metastasizes. These effects are evident in CRC consensus molecular subtypes (CMS) where both CMS1 and CMS4 contain immune infiltrates yet have opposing prognoses. The microbiome has previously been associated with CRC and immune response in CRC but has largely been ignored in the CRC subtype discussion. We used CMS subtyping on surgical resections from patients and aimed to determine the contributions of the microbiome to the pleiotropic effects evident in immune-infiltrated subtypes. We integrated host gene-expression and meta-transcriptomic data to determine the link between immune characteristics and microbiome contributions in these subtypes and identified lipopolysaccharide (LPS) binding as a potential functional mechanism. We identified candidate bacteria with LPS properties that could affect immune response, and tested the effects of their LPS on cytokine production of peripheral blood mononuclear cells (PBMCs). We focused on Fusobacterium periodonticum and Bacteroides fragilis in CMS1, and Porphyromonas asaccharolytica in CMS4. Treatment of PBMCs with LPS isolated from these bacteria showed that F. periodonticum stimulates cytokine production in PBMCs while both B. fragilis and P. asaccharolytica had an inhibitory effect. Furthermore, LPS from the latter two species can inhibit the immunogenic properties of F. periodonticum LPS when co-incubated with PBMCs. We propose that different microbes in the CRC tumor microenvironment can alter the local immune activity, with important implications for prognosis and treatment response.

Unraveling dynamic interactions between tumor-associated macrophages and consensus molecular subtypes in colorectal cancer: An integrative analysis of single-cell and bulk RNA transcriptome

BackgroundAccumulating research substantiated that tumor-associated macrophages (TAMs) have a significant impact on the tumorigenesis, progression, and distant metastasis, representing a novel target for various cancers. However, the underlying dynamic changes and interactions between TAMs and tumor cells remain largely elusive in colorectal cancer (CRC). MethodsWe depicted the dynamic changes of macrophages using sing-cell RNA-seq data and extracted TAM differentiation-related genes. Next, we utilized the weighted gene co-expression network analysis (WGCNA) to acquire CMS-related modular genes using bulk RNA-seq data. Finally, we utilized univariate Cox and Lasso Cox regression analyses to identify TAM differentiation-related biomarkers and established a novel risk signature model. We employed quantitative real-time polymerase chain reaction (qRT-PCR) on CRC tissue samples and used immunohistochemistry (IHC) data frome the HPA database to validate the mRNA and protein expression of prognostic genes. The interaction of TAMs and each consensus molecular subtype (CMS) subpopulation was analyzed at the cellular level. ResultsA total of 47,285 cells from single-cell dataset and 1197 CRC patients from bulk dataset were obtained. Among those, 6400 myeloid cells were re-clustered and annotated. RNASE1, F13A1, DAPK1, CLEC10A, RPN2, REG4 and RGS19 were identified as prognostic genes and the risk signature model was established based on the above genes. The qRT-PCR analysis indicated that the expression of RNASE1 and DAPK1 were significantly up-regulated in CRC tumor tissues. The cell-cell communication analysis demonstrated complex interactions between TAMs and CMS malignant cell subpopulations. ConclusionThis study presents an in-depth dissection of the dynamic features of TAMs in the tumor microenvironment and provides promising therapeutic targets for CRC.

Transcription factor expression repertoire basis for epigenetic and transcriptional subtypes of colorectal cancers

Colorectal cancers (CRCs) form a heterogenous group classified into epigenetic and transcriptional subtypes. The basis for the epigenetic subtypes, exemplified by varying degrees of promoter DNA hypermethylation, and its relation to the transcriptional subtypes is not well understood. We link cancer-specific transcription factor (TF) expression alterations to methylation alterations near TF-binding sites at promoter and enhancer regions in CRCs and their premalignant precursor lesions to provide mechanistic insights into the origins and evolution of the CRC molecular subtypes. A gradient of TF expression changes forms a basis for the subtypes of abnormal DNA methylation, termed CpG-island promoter DNA methylation phenotypes (CIMPs), in CRCs and other cancers. CIMP is tightly correlated with cancer-specific hypermethylation at enhancers, which we term CpG-enhancer methylation phenotype (CEMP). Coordinated promoter and enhancer methylation appears to be driven by downregulation of TFs with common binding sites at the hypermethylated enhancers and promoters. The altered expression of TFs related to hypermethylator subtypes occurs early during CRC development, detectable in premalignant adenomas. TF-based profiling further identifies patients with worse overall survival. Importantly, altered expression of these TFs discriminates the transcriptome-based consensus molecular subtypes (CMS), thus providing a common basis for CIMP and CMS subtypes.

Transcriptional Profiling and Consensus Molecular Subtype Assignment to Understand Response and Resistance to Anti-Epidermal Growth Factor Receptor Therapy in Colorectal Cancer.

Activating mutations in KRAS, NRAS, and BRAF are known to cause resistance to anti-epidermal growth factor receptor (EGFR) therapy; however, only approximately 40% of patients with colorectal cancer (CRC) with RASWT tumors respond to anti-EGFR treatment. We sought to discover novel biomarkers to predict response to anti-EGFR antibody treatment in CRC and to understand mechanisms of resistance to anti-EGFR therapy. Transcriptomic profiles from three clinical and two preclinical cohorts treated with cetuximab were used to assign consensus molecular subtypes (CMS) to each sample and correlated with outcomes. Restricting to RASWT patients, we observed that CMS2 tumors (canonical subtype) had significantly higher response rates relative to other CMS when treated with cetuximab combination with doublet chemotherapy (Okita et al cohort: 92% disease control rate (DCR) for CMS2, chi-square P = .04; CALGB/SWOG 80405 cohort: 90% objective response rate (ORR) for CMS2, chi-square P < .001) and with single-agent cetuximab (68%, chi-square P = .01). CMS2 tumors showed best response among right-sided (ORR = 80%) and left-sided (ORR = 92%) tumors in the CALGB/SWOG 80405 cohort. CMS2 cells lines were most likely to be sensitive to cetuximab (60%) and CMS2 patient-derived xenograft had the highest DCR (84%). We found Myc, E2F, and mammalian target of rapamycin pathways were consistently upregulated in resistant samples (enrichment score >1, false discovery rate <0.25). Inhibitors of these pathways in resistant cell lines exhibited additive effects with cetuximab. These data suggest that CRC transcriptional profiles, when used to assign CMS, provide additional ability to predict response to anti-EGFR therapy relative to using tumor sidedness alone. Notably both right-sided and left-sided CMS2 tumors had excellent response, suggesting that anti-EGFR therapy be included as a treatment option for right-sided CMS2 tumors.

Molecular characteristics of microsatellite stable early-onset colorectal cancer as predictors of prognosis and immunotherapeutic response

The incidence of early-onset colorectal cancer (EO-CRC, in patients younger than 50) is increasing worldwide. The specific gene signatures in EO-CRC patients are largely unknown. Since EO-CRC with microsatellite instability is frequently associated with Lynch syndrome, we aimed to comprehensively characterize the tumor microenvironment (TME) and gene expression profiles of EO-CRC with microsatellite stable (MSS-EO-CRC). Here, we demonstrated that MSS-EO-CRC has a similar pattern of tumor-infiltrating immune cells, immunotherapeutic responses, consensus molecular subtypes, and prognosis as late-onset CRC with MSS (MSS-LO-CRC). 133 differential expressed genes were identified as unique gene signatures of MSS-EO-CRC. Moreover, we established a risk score, which was positively associated with PD-L1 expression and could reflect both the level of tumor-infiltrating immune cells and the prognosis of MSS-EO-CRC patients. Application of this score on the anti-PD-L1 treatment cohort demonstrated that the low-risk score group has significant therapeutic advantages and clinical benefits. In addition, candidate driver genes were identified in the different-sidedness of MSS-EO-CRC patients. Altogether, MSS-EO-CRC exhibits distinct molecular profiles that differ from MSS-LO-CRC even though they have a similar TME characterization and survival pattern. Our risk score appears to be robust enough to predict prognosis and immunotherapeutic response and therefore could help to optimize the treatment of MSS-EO-CRC.

Latent class analysis-derived classification improves the cancer-specific death stratification of molecular subtyping in colorectal cancer

The molecular subtypes of colorectal cancer (CRC) represent a comprehensive dissection of CRC heterogeneity. However, molecular feature-based classification systems have limitations in accurately prognosticating stratification due to the inability to distinguish cancer-specific deaths. This study aims to establish a classification system that bridges clinical characteristics, cause-specific deaths, and molecular features. We adopted latent class analysis (LCA) on 491,107 first primary CRC patients from the Surveillance, Epidemiology, and End Results (SEER) database to reveal hidden profiles of CRC. The LCA-derived classification scheme was further applied to The Cancer Genome Atlas (TCGA) to assess its effectiveness in improving the accurate stratification of molecular-based subtypes of CRC. Four classes were identified based on latent class analysis integrating demographic and clinicopathological information of CRC patients. The LCA-derived Class 1 (LCAC1) and the LCAC2 showed a high risk of dying from non-CRC, while patients in LCAC3 had a risk of dying from CRC 1.41 times that of LCAC1 (95% confidence interval [CI] = 1.39–1.43). LCAC4 had the lowest probability to die from non-CRC (hazard ratio [HR] = 0.22, 95% CI = 0.21–0.24) compared with LCAC1. Since the LCA-derived classification can identify patients susceptible to CRC-specific death, adjusting for this classification allows molecular-based subtypes to achieve more accurate survival stratification. We provided a classification system capable of distinguish CRC-specific death, which will improve the accuracy of consensus molecular subtypes for CRC patients’ survival stratification. Further studies are warranted to confirm the molecular features of LCA-derived classification to inform potential therapeutic strategies and treatment recommendations.

Genomic landscape and expression profile of consensus molecular subtype four of colorectal cancer.

Compared to other subtypes, the CMS4 subtype is associated with lacking of effective treatments and poorer survival rates. A total of 24 patients with CRC were included in this study. DNA and RNA sequencing were performed to acquire somatic mutations and gene expression, respectively. MATH was used to quantify intratumoral heterogeneity. PPI and survival analyses were performed to identify hub DEGs. Reactome and KEGG analyses were performed to analyze the pathways of mutated or DEGs. Single-sample gene set enrichment analysis and Xcell were used to categorize the infiltration of immune cells. The CMS4 patients had a poorer PFS than CMS2/3. CTNNB1 and CCNE1 were common mutated genes in the CMS4 subtype, which were enriched in Wnt and cell cycle signaling pathways, respectively. The MATH score of CMS4 subtype was lower. SLC17A6 was a hub DEG. M2 macrophages were more infiltrated in the tumor microenvironment of CMS4 subtype. The CMS4 subtype tended to have an immunosuppressive microenvironment. This study suggested new perspectives for exploring therapeutic strategies for the CMS4 subtype CRC.

Classification of colorectal cancer consensus molecular subtypes using attention-based multi-instance learning network on whole-slide images

Colorectal cancer (CRC) is the third most common and second most lethal cancer globally. It is highly heterogeneous with different clinical-pathological characteristics, prognostic status, and therapy responses. Thus, the precise diagnosis of CRC subtypes is of great significance for improving the prognosis and survival of CRC patients. Nowadays, the most commonly used molecular-level CRC classification system is the Consensus Molecular Subtypes (CMSs). In this study, we applied a weakly supervised deep learning method, named attention-based multi-instance learning (MIL), on formalin-fixed paraffin-embedded (FFPE) whole-slide images (WSIs) to distinguish CMS1 subtype from CMS2, CMS3, and CMS4 subtypes, as well as distinguish CMS4 from CMS1, CMS2, and CMS3 subtypes. The advantage of MIL is training a bag of the tiled instance with bag-level labels only. Our experiment was performed on 1218 WSIs obtained from The Cancer Genome Atlas (TCGA). We constructed three convolutional neural network-based structures for model training and evaluated the ability of the max-pooling operator and mean-pooling operator on aggregating bag-level scores. The results showed that the 3-layer model achieved the best performance in both comparison groups. When compared CMS1 with CMS234, max-pooling reached the ACC of 83.86 % and the mean-pooling operator reached the AUC of 0.731. While comparing CMS4 with CMS123, mean-pooling reached the ACC of 74.26 % and max-pooling reached the AUC of 0.609. Our results implied that WSIs could be utilized to classify CMSs, and manual pixel-level annotation is not a necessity for computational pathology imaging analysis.

Optimizing identification of consensus molecular subtypes in muscle-invasive bladder cancer: a comparison of two sequencing methods and gene sets using FFPE specimens

BackgroundMolecular subtypes predict prognosis in muscle-invasive bladder cancer (MIBC) and are explored as predictive markers. To provide a common base for molecular subtyping and facilitate clinical applications, a consensus classification has been developed. However, methods to determine consensus molecular subtypes require validation, particularly when FFPE specimens are used. Here, we aimed to evaluate two gene expression analysis methods on FFPE samples and to compare reduced gene sets to classify tumors into molecular subtypes.MethodsRNA was isolated from FFPE blocks of 15 MIBC patients. Massive analysis of 3’ cDNA ends (MACE) and the HTG transcriptome panel (HTP) were used to retrieve gene expression. We used normalized, log2-transformed data to call consensus and TCGA subtypes with the consensusMIBC package for R using all available genes, a 68-gene panel (ESSEN1), and a 48-gene panel (ESSEN2).ResultsFifteen MACE-samples and 14 HTP-samples were available for molecular subtyping. The 14 samples were classified as Ba/Sq in 7 (50%), LumP in 2 (14.3%), LumU in 1 (7.1%), LumNS in 1 (7.1%), stroma-rich in 2 (14.3%) and NE-like in 1 (7.1%) case based on MACE- or HTP-derived transcriptome data. Consensus subtypes were concordant in 71% (10/14) of cases when comparing MACE with HTP data. Four cases with aberrant subtypes had a stroma-rich molecular subtype with either method. The overlap of the molecular consensus subtypes with the reduced ESSEN1 and ESSEN2 panels were 86% and 100%, respectively, with HTP data and 86% with MACE data.ConclusionDetermination of consensus molecular subtypes of MIBC from FFPE samples is feasible using various RNA sequencing methods. Inconsistent classification mainly involves the stroma-rich molecular subtype, which may be the consequence of sample heterogeneity with (stroma)-cell sampling bias and highlights the limitations of bulk RNA-based subclassification. Classification is still reliable when analysis is reduced to selected genes.

Consensus molecular subtyping of colorectal cancer to demonstrate cetuximab benefit in right-sided CMS2 tumors, and pembrolizumab benefit in MSS CMS1 tumors.

3535 Background: Consensus Molecular Subtypes (CMS) of colorectal cancer (CRC) were first developed in 2015 using microarray-based assays but are not widely used clinically. We developed a Caris CMS classifier on whole transcriptome sequencing data (WTS) with high concordance with the previously established CMS pipeline (Guinney et al 2015), and applied to a large clinic-genomic database of CRC to investigate the utility of CMS classification in identifying patients that may respond well to therapies commonly used in CRC. Methods: Next-generation sequencing (NGS) of DNA (592-gene or whole exome) and WTS was tested on CRC patient samples (n = 12,788) at a CLIA-certified lab (Caris Life Sciences, Phoenix, AZ). Caris CMS classifier was trained against the original CMS datasets using a classic SVM model and cross-validated for optimization of the SVM parameters. Possible overtraining was evaluated by predicting CMS from an independent blinded dataset (TCGA, N = 512) with an accuracy of 88.3%. Real-world overall survival was obtained from insurance claims and calculated from tissue collection to last contact (OS); time on treatment (TOT) was from first to last of treatment time. Kaplan-Meier estimates were calculated for molecularly defined cohorts. Significance was determined as p of &lt; 0.05. Results: Among all patients, CMS1 was seen in 16%, CMS2 in 32%, CMS3 in 17% and CMS4 in 35%. MSI-H/MMRd (31%) and BRAF mut (33%) were most prevalent in CMS1 with KRAS mt the highest in CMS3 (66%). CMS2 was associated with the longest mOS (33m; 95% CI: 31m-35m), followed by CMS4 (29m; 28-31m), CMS3 (27m; 25-29m) and CMS1 (22m; 20-23m). In the microsatellite stable (MSS) tumors treated with pembrolizumab, CMS1 (N = 22, mTOT: 4.2m; 2.8-9.1m) had longer TOT than CMS2 (N = 45, mTOT: 2.1m; 1.4-3.1m), CMS3 (N = 21, 2.1m; 1.4m-3.0m) and CMS4 (N = 40; 2.1m; 1.4m-2.8m); CMS1 vs. CMS2-4 in MSS (HR: 0.58; CI: 0.34-0.97, p = 0.035). When investigating cetuximab, CMS2 had the longest mTOT (N = 189, mTOT: 6.3m; 5.4m-7.7m) among the four groups. Interestingly, although among all Ras WT tumors left-sided CRC showed longer mTOT on cetuximab (n = 457, mTOT: 5.4m; 4.7m-6.3m) than right-sided (n = 151, mTOT: 3.8m; 3.3m-4.5m), CMS2 showed similar mTOT from the left (n = 128) and the right (n = 16, mOT: 6.3m vs. 5.6m, HR = 0.896, 95% CI: 0.501-1.604, p = 0.722). Notably, CMS2 comprises 49% of left- and 15% of right-sided tumors, potentially underlying the TOT difference seen between left and right. Conclusions: A WTS based CMS classifier allows for investigation in a large real-world clinic-genomic database. We found that MSS CMS1 CRC’s may derive benefit from immunotherapy. Additionally, CMS2 subgroup of right-sided tumors may derive benefit from cetuximab. Routine CMS subgrouping of CRC provides important treatment associations that should be further investigated.

Consensus molecular subtypes and orphan non-coding RNAs in colorectal cancer.

e15651 Background: Colorectal cancer (CRC) is a highly heterogeneous disease with variable response to different treatment strategies. The consensus molecular subtype (CMS) classification system has been proposed as a gene expression-based framework aimed at capturing the genetic and molecular heterogeneity in CRCs and potentially stratifying patients into clinically relevant subgroups for treatment selection. We previously demonstrated that orphan non-coding RNAs (oncRNAs), a novel class of cancer enriched small non-coding RNAs (smRNA), exhibit cancer-specific expression. We also showed that oncRNAs can be used in a liquid biopsy assay for accurate early-stage detection of CRCs. We hypothesize that oncRNAs may be informative of CRC subtype-specific transcriptional signatures as defined by the CMS groups. Methods: To assess the effectiveness of oncRNAs for categorization of CMS group, we examined smRNA-seq data from The Cancer Genome Atlas (TCGA) for colon and rectal adenocarcinoma tumors ( n= 504), consisting of 77 CMS1, 216 CMS2, 71 CMS3, and 140 CMS4 tumors. CRC-specific oncRNA expression profiles were generated for all tumors in the TCGA cohort. We then trained a machine learning model on oncRNA expression profiles to predict CMS groups using a 5-fold cross validation setup. For our validation cohort, we applied our model to an independent CRC cohort ( n= 348) consisting of 56 CMS1, 101 CMS2, 59 CMS3, and 132 CMS4 CRC samples. Final CMS predictions for the validation cohort were made by averaging across the five training models. Results: Within TCGA testing folds, the model micro-averaged ROC-AUC for CMS classifications were 0.879 (95% CI: 0.86–0.90). In the validation cohort, micro-averaged AUC remained high at 0.915 (0.90–0.93). Sensitivities and AUCs for each subtype are provided in Table 1. The highest model-predicted likelihood for each CMS subtype had an accuracy of 73.6% (68.6%–78.1%), and the two most probable predictions had an accuracy of 91.1% (87.6%–93.9%). Conclusions: This study demonstrates the utility of oncRNA expression profiling to predict CMS for colorectal cancer. The generalizability of our machine learning model suggests that oncRNAs are informative of the biological differences across distinct CMS groups and may be used as a proxy for gene expression signatures to stratify colorectal cancers. Because smRNAs are more likely to be secreted into blood, we posit that an oncRNA-based liquid biopsy may open new opportunities to monitor patients and track changes in their tumor biology in real-time throughout their course of disease and treatment. [Table: see text]

Racial and genetic ancestry associations with gene expression patterns in a real-world cohort of colorectal cancer patients.

3541 Background: There is a growing incidence of colorectal cancer (CRC) among young adults and persistent disparities in outcomes by race/ethnicity across all ages. Gene expression signatures, as well as consensus molecular subtypes (CMS) derived from these, have been proposed to predict prognosis and therapy response in CRC. However, it is unclear whether gene expression or CMS are associated with racial disparities observed in CRC. We assessed whether race or genetic ancestry are associated with CMS or gene expression patterns in a deidentified cohort of 1,768 CRC patients. Methods: Patients tumors’ underwent tumor profiling with the Tempus xT NGS 648-gene assay as well as full-transcriptome RNA sequencing. We used a set of 654 ancestry-informative markers to infer genetic ancestry likelihoods for Africa (AFR), America (AMR), East Asia (EAS), Europe (EUR), and South Asia (SAS). Race/ethnicity labels, often missing in real-world data, were imputed using ancestry proportions from the literature and adjusted based on observed metadata. Gene expression data was used to assign CMS to all patients (CMS1, 2, 3, 4 and indeterminate) using CMScaller, and multinomial logistic regression was used to assess associations with race/ethnicity imputed labels and ancestry proportions. We then assessed differential expression (DE) in the MSigDB hallmark and C2 BioCarta gene sets using four separate workflows. The first two workflows used limma-voom followed by ROAST to assess DE among the imputed labels and then among the ancestry proportions (isometric log ratio transformed). The second two workflows used GSVA followed by limma. Results: Among 1,768 patients, 240 were imputed non-Hispanic (NH) Black, 94 NH Asian, 261 Hispanic/Latino/Native American (HLN), and 1,173 NH White. NH Black patients had higher odds of CMS3 vs CMS1 (OR = 2.66, p &lt; 0.001) and HLN patients had higher odds of indeterminate CMS vs CMS1 (OR = 1.90, p = 0.020), compared to NH White. AFR ancestry was significantly associated with CMS3 (OR = 1.05 per doubling in AFR proportion, p = 0.047) and indeterminate CMS (OR = 1.07, p = 0.023). In the gene set analysis, NH Black race/ethnicity was associated with over-expression of the BioCarta WNT pathway gene set. Both AFR ancestry and NH Black race/ethnicity were associated with under-expression of the MSigDB hallmark coagulation and BioCarta alternative complement gene sets and over-expression of the PITX2 pathway gene set (all significant with both ROAST and GSVA). Conclusions: We found that NH Black patients and AFR ancestry were associated with higher rates of CMS3, which is associated with KRAS mutation and was previously reported to be more common among Black patients. Indeterminate CMS associations with AFR ancestry and HLN highlights the need to use diverse patient cohorts when training unsupervised learning models to improve prognosis prediction in non-White patients.

P-109 Broaden the therapeutic opportunities for colorectal cancer: ALK inhibition benefits consensus molecular subtype 1 patients

P-109 Broaden the therapeutic opportunities for colorectal cancer: ALK inhibition benefits consensus molecular subtype 1 patients

Identifying essential genes in genome-scale metabolic models of consensus molecular subtypes of colorectal cancer.

Identifying essential targets in the genome-scale metabolic networks of cancer cells is a time-consuming process. The present study proposed a fuzzy hierarchical optimization framework for identifying essential genes, metabolites and reactions. On the basis of four objectives, the present study developed a framework for identifying essential targets that lead to cancer cell death and evaluating metabolic flux perturbations in normal cells that have been caused by cancer treatment. Through fuzzy set theory, a multiobjective optimization problem was converted into a trilevel maximizing decision-making (MDM) problem. We applied nested hybrid differential evolution to solve the trilevel MDM problem to identify essential targets in genome-scale metabolic models for five consensus molecular subtypes (CMSs) of colorectal cancer. We used various media to identify essential targets for each CMS and discovered that most targets affected all five CMSs and that some genes were CMS-specific. We obtained experimental data on the lethality of cancer cell lines from the DepMap database to validate the identified essential genes. The results reveal that most of the identified essential genes were compatible with the colorectal cancer cell lines obtained from DepMap and that these genes, with the exception of EBP, LSS, and SLC7A6, could generate a high level of cell death when knocked out. The identified essential genes were mostly involved in cholesterol biosynthesis, nucleotide metabolisms, and the glycerophospholipid biosynthetic pathway. The genes involved in the cholesterol biosynthetic pathway were also revealed to be determinable, if a cholesterol uptake reaction was not induced when the cells were in the culture medium. However, the genes involved in the cholesterol biosynthetic pathway became non-essential if such a reaction was induced. Furthermore, the essential gene CRLS1 was revealed as a medium-independent target for all CMSs.

Comprehensive transcriptomic analyses identify KDM genes-related subtypes with different TME infiltrates in gastric cancer

Histone lysine demethylases (KDMs) have been reported in various malignances, which affect transcriptional regulation of tumor suppressor or oncogenes. However, the relationship between KDMs and formation of tumor microenvironment (TME) in gastric cancer (GC) remain unclear and need to be comprehensively analyzed.In the present study, 24 KDMs were obtained and consensus molecular subtyping was performed using the "NMF" method to stratify TCGA-STAD into three clusters. The ssGSEA and CIBERSORT algorithms were employed to assess the relative infiltration levels of various cell types in the TME. The KDM_score was devised to predict patient survival outcomes and responses to both immunotherapy and chemotherapy.Three KDM genes-related molecular subtypes were Figured out in GC with distinctive clinicopathological and prognostic features. Based on the robust KDM genes-related risk_score and nomogram, established in our work, GC patients’ clinical outcome can be well predicted. Furthermore, low KDM genes-related risk_score exhibited the more effective response to immunotherapy and chemotherapy.This study characterized three KDM genes-related TME pattern with unique immune infiltration and prognosis by comprehensively analyses of transcriptomic profiling. Risk_score was also built to help clinicians decide personalized anticancer treatment for GC patients, including in prediction of immunotherapy and chemotherapy response for patients.

Co-cultures of colon cancer cells and cancer-associated fibroblasts recapitulate the aggressive features of mesenchymal-like colon cancer.

Poor prognosis in colon cancer is associated with a high content of cancer-associated fibroblasts (CAFs) and an immunosuppressive tumor microenvironment. The relationship between these two features is incompletely understood. Here, we aimed to generate a model system for studying the interaction between cancer cells and CAFs and their effect on immune-related cytokines and T cell proliferation. CAFs were isolated from colon cancer liver metastases and were immortalized to prolong lifespan and improve robustness and reproducibility. Established medium and matrix compositions that support the growth of patient-derived organoids were adapted to also support CAF growth. Changes in growth pattern and cellular re-organization were assessed by confocal microscopy, live cell imaging, and immunofluorescence. Single cell RNA sequencing was used to study CAF/organoid co-culture-induced phenotypic changes in both cell types. Conditioned media were used to quantify the production of immunosuppressive factors and to assess their effect on T cell proliferation. We developed a co-culture system in which colon cancer organoids and CAFs spontaneously organize into superstructures with a high capacity to contract and stiffen the extracellular matrix (ECM). CAF-produced collagen IV provided a basement membrane supporting cancer cell organization into glandular structures, reminiscent of human cancer histology. Single cell RNA sequencing analysis showed that CAFs induced a partial epithelial-to-mesenchymal-transition in a subpopulation of cancer cells, similar to what is observed in the mesenchymal-like consensus molecular subtype 4 (CMS4) colon cancer. CAFs in co-culture were characterized by high expression of ECM components, ECM-remodeling enzymes, glycolysis, hypoxia, and genes involved in immunosuppression. An expression signature derived from CAFs in co-culture identified a subpopulation of glycolytic myofibroblasts specifically residing in CMS1 and CMS4 colon cancer. Medium conditioned by co-cultures contained high levels of the immunosuppressive factors TGFβ1, VEGFA and lactate, and potently inhibited T cell proliferation. Co-cultures of organoids and immortalized CAFs recapitulate the histological, biophysical, and immunosuppressive features of aggressive mesenchymal-like human CRC. The model can be used to study the mechanisms of immunosuppression and to test therapeutic strategies targeting the cross-talk between CAFs and cancer cells. It can be further modified to represent distinct colon cancer subtypes and (organ-specific) microenvironments.

Integrative single-cell and bulk transcriptomes analyses identify intrinsic HNSCC subtypes with distinct prognoses and therapeutic vulnerabilities.

Tumor heterogeneity in head neck squamous cell carcinoma (HNSCC) profoundly compromises patient stratification, treatment planning and prognostic prediction, which underscores the urgent need for more effective molecular subtyping for this malignancy. Here, we sought to define the intrinsic epithelial subtypes for HNSCC by integrative analyses of single-cell and bulk RNA sequencing datasets from multiple cohorts and assess their molecular features and clinical significance. Malignant epithelial cells were identified from scRNA-seq datasets and subtyped based on differentially expressed genes. Subtype-specific genomic/epigenetic abnormalities, molecular signaling, regulatory network, immune landscape and patient survival were characterized. Therapeutic vulnerabilities were further predicted based on drug sensitivity datasets from cell lines, patient-derived xenograft models and real-world clinical outcomes. Novel signatures for prognostication and therapeutic prediction were developed by machine learning and independently validated. Three intrinsic consensus molecular subtypes (iCMS1-3) for HNSCC were proposed from scRNA-seq analyses and recapitulated in 1325 patients from independent cohorts using bulk datasets. iCMS1 was characterized by EGFR amplification/activation, stromal-enriched environment, epithelial-to-mesenchymal transition, worst survival and sensitivities to EGFR inhibitor. iCMS2 was featured by HPV+ oropharyngeal predilection, immune-hot, susceptibilities to anti-PD-1 and best prognosis. Moreover, iCMS3 displayed immune-desert and sensitivities to 5-FU and MEK, STAT3 inhibitors. Three novel, robust signatures derived from iCMS subtype-specific transcriptomics features were developed by machine learning for patient prognostication and cetuximab and anti-PD-1 response predictions. These findings reiterate molecular heterogeneity of HNSCC and advantages of scRNA-seq in pinpointing cellular diversities in complex cancer ecosystems. Our HNSCC iCMS regime might facilitate patient stratification and precision medicine.