CoChat

A tumor microenvironment-specific gene expression signature predicts chemotherapy resistance in colorectal cancer patients

Open access

The DOI matches its Crossref record; 94 of the 97 references it cites were found in a registry or an index; 17 of the 31 sources its claims cite were checked for support.

Abstract

Studies have shown that tumor microenvironment (TME) might affect drug sensitivity and the classification of colorectal cancer (CRC). Using TME-specific gene signature to identify CRC subtypes with distinctive clinical relevance has not yet been tested. A total of 18 "bulk" RNA-seq datasets (total n = 2269) and four single-cell RNA-seq datasets were included in this study. We constructed a "Signature associated with FOLFIRI resistant and Microenvironment" (SFM) that could discriminate both TME and drug sensitivity. Further, SFM subtypes were identified using K-means clustering and verified in three independent cohorts. Nearest template prediction algorithm was used to predict drug response. TME estimation was performed by CIBERSORT and microenvironment cell populations-counter (MCP-counter) methods. We identified six SFM subtypes based on SFM signature that discriminated both TME and drug sensitivity. The SFM subtypes were associated with distinct clinicopathological, molecular and phenotypic characteristics, specific enrichments of gene signatures, signaling pathways, prognosis, gut microbiome patterns, and tumor lymphocytes infiltration. Among them, SFM-C and -F were immune suppressive. SFM-F had higher stromal fraction with epithelial-to-mesenchymal transition phenotype, while SFM-C was characterized as microsatellite instability phenotype which was responsive to immunotherapy. SFM-D, -E, and -F were sensitive to FOLFIRI and FOLFOX, while SFM-A, -B, and -C were responsive to EGFR inhibitors. Finally, SFM subtypes had strong prognostic value in which SFM-E and -F had worse survival than other subtypes. SFM subtypes enable the stratification of CRC with potential chemotherapy response thereby providing more precise therapeutic options for these patients.

Overview

CoChat Summary

TL;DR

The authors built a 250-gene “Signature associated with FOLFIRI resistant and Microenvironment” (SFM) from 18 bulk RNA-seq datasets (n = 2269) and 4 scRNA-seq datasets to classify colorectal cancer (CRC) by tumor microenvironment and drug-response patterns. K-means clustering yielded six CRC subtypes (SFM-A to SFM-F) that differed in TME composition, clinicopathologic features, prognosis, microbiome patterns, and predicted sensitivity to chemotherapy and EGFR-targeted therapy.

Key findings

  • The final SFM signature contained 317 probes corresponding to 250 unique genes, selected from 2269 CRC gene expression profiles; in four scRNA-seq datasets, SFM separated malignant from nonmalignant/TME cell types.
  • K-means clustering in the discovery cohort identified 6 subtypes (SFM-A to SFM-F); subtype proportions in the main datasets were SFM-A 23%, SFM-B 23%, SFM-E 25%, SFM-C 12%, SFM-F 10%, and SFM-D 8%.
  • Chemotherapy associations were subtype-specific: SFM-E and SFM-F were enriched for FOLFIRI/FOLFOX responsiveness, while SFM-A and SFM-B were resistant; in one combined FOLFIRI dataset, 8/9 responders were classified as SFM-E or SFM-F (P = 0.0006).
  • Prognosis differed by subtype: SFM-E and SFM-F had worse survival, and in multivariate Cox analysis the strongest prognostic value was CCS3 (HR = 11.32, 95% CI 3.14–40.84), followed by SFM-F (HR = 2.48, 95% CI 1.51–3.96) and SFM-E (HR = 1.89, 95% CI 1.24–2.89).
  • The subtypes also showed distinct biology: SFM-C was enriched for MSI, BRAF mutation, hypermutation, and proximal colon phenotype, while SFM-F had stronger stroma/EMT features and higher checkpoint-marker expression.

Methods

  • Built the SFM signature by comparing sorted cell populations from GSE39395 and GSE39396 with FOLFIRI responder/nonresponder tumors from GSE62080, using differential expression with P < 0.05 and |logFC| ≥ 1; intersected results produced the final gene set.
  • Classified CRC samples into subtypes using K-means clustering on SFM expression, selecting k = 6 in the discovery dataset (GSE39582) and validating in five GEO batch, TCGA, and Renji cohort datasets.
  • Assessed subtype associations with drug response (NTP on published signatures), immune/stromal infiltration (CIBERSORT, MCP-counter, estimate), survival (Kaplan–Meier and Cox models), and microbiome patterns (PathSeq), with additional validation in four scRNA-seq datasets.

Limitations

  • The paper states the study is retrospective and says the subtype classification should be tested in large prospective clinical trials.
  • The multivariate Cox models lacked some clinical covariates, because tumor grade and number of positive nodes were not publicly available; this limits certainty that SFM is an independent prognostic factor.
  • Drug response was predicted from previously published gene signatures rather than directly measured in the included datasets, so the authors say subtype-specific drug sensitivity needs future confirmation.
  • Evident from the text: some findings rely on small subgroups (for example, the Renji cohort had only four subtypes due to sample size), which may weaken stability of some subtype-specific conclusions.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41698-021-00142-x

Cites

94 of 94 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Construction of SFM signature and subtypes. a Venn diagram of differentially expressed probe sets in three datasets showing intersection of 317 probes (250 unique genes). b–e t-SNE plot of SFM gene expression profiles from four datasets (b, CRC, n = 363; c, HNSCC, n = 5502; d, melanoma, n = 4054; e, BRCA, n = 515). f–g Genes overlapped between SFM and published gene signatures. Each column indicates each gene signature and each row indicate each gene. Shared genes are indicated with black lines (f). Graph depicts the quantification of overlap across these signatures (g). h Proportion of SFM subtypes in individual and aggregated datasets. i Sankey diagram showing how each SFM subtype contributes to AJCC stage classification.
Fig. 1 Construction of SFM signature and subtypes. a Venn diagram of differentially expressed probe sets in three datasets showing intersection of 317 probes (250 unique genes). b–e t-SNE plot of SFM gene expression profiles from four datasets (b, CRC, n = 363; c, HNSCC, n = 5502; d, melanoma, n = 4054; e, BRCA, n = 515). f–g Genes overlapped between SFM and published gene signatures. Each column indicates each gene signature and each row indicate each gene. Shared genes are indicated with black lines (f). Graph depicts the quantification of overlap across these signatures (g). h Proportion of SFM subtypes in individual and aggregated datasets. i Sankey diagram showing how each SFM subtype contributes to AJCC stage classification.CC BY · PubMed Central
Fig. 2 Distinct sensitivity among SFM subtypes to FOLFIRI and FOLFOX chemotherapy regimens and RGFR inhibitors. a Heatmap showing individual response of patients to FOLFIRI and their association with subtypes in GSE62080 dataset (n = 21). P value was calculated using Chi-squared test. b Heatmap showing individual responses of patients to FOLFOX and their association with subtypes in combined three GEO datasets (GSE19860, GSE28702, and GSE69657, n = 142). P value was calculated using Chi-squared test. c Heatmap showing association of individual CRC patient’s response to FOLFIRI, FOLFOX, and EGFR inhibitors. In these analyses, samples with FDR < 0.2 were regarded as significant. d Prevalence of SFM subtypes regarded as sensitive or resistant to cetuximab in all CRC patients (n = 1752) or only in KRAS wild-type patients (n = 637). e Heatmap showing response to cetuximab, quantified expression of gene set of the EGFR pathway activity by applying GSVA and expression of individual genes of the EGFR gene set among SFM subtypes. f Box plots of the EGFR gene set and individual genes of the EGFR gene set among SFM subtypes. Box hinges represent first and third quartiles, and middle represents the median. The upper and lower whiskers extend from hinges up and down indicate the most extreme values that are within 1.5 × IQR (interquartile range) of the respective hinge.
Fig. 2 Distinct sensitivity among SFM subtypes to FOLFIRI and FOLFOX chemotherapy regimens and RGFR inhibitors. a Heatmap showing individual response of patients to FOLFIRI and their association with subtypes in GSE62080 dataset (n = 21). P value was calculated using Chi-squared test. b Heatmap showing individual responses of patients to FOLFOX and their association with subtypes in combined three GEO datasets (GSE19860, GSE28702, and GSE69657, n = 142). P value was calculated using Chi-squared test. c Heatmap showing association of individual CRC patient’s response to FOLFIRI, FOLFOX, and EGFR inhibitors. In these analyses, samples with FDR < 0.2 were regarded as significant. d Prevalence of SFM subtypes regarded as sensitive or resistant to cetuximab in all CRC patients (n = 1752) or only in KRAS wild-type patients (n = 637). e Heatmap showing response to cetuximab, quantified expression of gene set of the EGFR pathway activity by applying GSVA and expression of individual genes of the EGFR gene set among SFM subtypes. f Box plots of the EGFR gene set and individual genes of the EGFR gene set among SFM subtypes. Box hinges represent first and third quartiles, and middle represents the median. The upper and lower whiskers extend from hinges up and down indicate the most extreme values that are within 1.5 × IQR (interquartile range) of the respective hinge.CC BY · PubMed Central
Fig. 3 Immune and TME-associated differences across SFM subtypes. a Estimate immune cells filtration using CIBERSORT in aggregated dataset. The fraction of immune cells were plotted using box plot across SFM subtypes. Box plots of fractions of endothelial cells (b) and fibroblasts (c) by MCP-counter algorithm. TGF-β response score using GSVA (d), stromal fraction using ESTIMATE (e), neoantigen production (f, g), tumor mutation burden (h, i), proportion of exhausted cells (j), and MDSC (k) using GSVA, expression differences of checkpoint biomarkers, including CD274 (l), CTLA4 (m) and PDCD1 (n), hot tumor signature score (o), IFN-γresponse score (p), and IPRES score (q) using GSVA. r Estimate IPRES signature using GSVA. Cases were annotated as “IPRES enriched” when IPRES score were >0.35 unless were annotated as “IPRES non-enriched”. The statistical difference was examined usings the Kruskal–Wallis test. (ns, P > 0.05; *P < = 0.05; **P < = 0.01; ***P < = 0.001; ****P < = 0.0001). For a–q, box hinges represent first and third quartiles, and middle represents the median. The upper and lower whiskers extend from hinges up and down indicate the most extreme values that are within 1.5 × IQR (interquartile range) of the respective hinge.
Fig. 3 Immune and TME-associated differences across SFM subtypes. a Estimate immune cells filtration using CIBERSORT in aggregated dataset. The fraction of immune cells were plotted using box plot across SFM subtypes. Box plots of fractions of endothelial cells (b) and fibroblasts (c) by MCP-counter algorithm. TGF-β response score using GSVA (d), stromal fraction using ESTIMATE (e), neoantigen production (f, g), tumor mutation burden (h, i), proportion of exhausted cells (j), and MDSC (k) using GSVA, expression differences of checkpoint biomarkers, including CD274 (l), CTLA4 (m) and PDCD1 (n), hot tumor signature score (o), IFN-γresponse score (p), and IPRES score (q) using GSVA. r Estimate IPRES signature using GSVA. Cases were annotated as “IPRES enriched” when IPRES score were >0.35 unless were annotated as “IPRES non-enriched”. The statistical difference was examined usings the Kruskal–Wallis test. (ns, P > 0.05; *P < = 0.05; **P < = 0.01; ***P < = 0.001; ****P < = 0.0001). For a–q, box hinges represent first and third quartiles, and middle represents the median. The upper and lower whiskers extend from hinges up and down indicate the most extreme values that are within 1.5 × IQR (interquartile range) of the respective hinge.CC BY · PubMed Central
Fig. 4 Survival differences across SFM subtypes. Kaplan–Meier plots of DFS and OS in training dataset (a, GSE39582) and validation dataset (b, five GEO batch; c, TCGA for DFS; and d, TCGA for OS). e Kaplan–Meier plots of Oncotype DX risk groups in stage I–IV cases. f Kaplan–Meier plots of Oncotype DX risk groups in stage II–III cases. The tick marks on the Kaplan–Meier curves indicated the censored patients. The differences between the curves were determined by the log-rank test. g Comparison of SFM subtypes and Oncotype DX classification in stage I–IV cases. h Comparison of SFM subtypes and Oncotype DX classification in stage II and III cases.
Fig. 4 Survival differences across SFM subtypes. Kaplan–Meier plots of DFS and OS in training dataset (a, GSE39582) and validation dataset (b, five GEO batch; c, TCGA for DFS; and d, TCGA for OS). e Kaplan–Meier plots of Oncotype DX risk groups in stage I–IV cases. f Kaplan–Meier plots of Oncotype DX risk groups in stage II–III cases. The tick marks on the Kaplan–Meier curves indicated the censored patients. The differences between the curves were determined by the log-rank test. g Comparison of SFM subtypes and Oncotype DX classification in stage I–IV cases. h Comparison of SFM subtypes and Oncotype DX classification in stage II and III cases.CC BY · PubMed Central
Fig. 5 Heatmap showing the relative abundance of dominant bacterial across SFM subtypes. Kruskal–Wallis test was used to test the differences across SFM subtypes.
Fig. 5 Heatmap showing the relative abundance of dominant bacterial across SFM subtypes. Kruskal–Wallis test was used to test the differences across SFM subtypes.CC BY · PubMed Central
Fig. 6 Summary of characteristics of SFM subtypes. These included pathological features, genomic markers, TME, epigenetic changes, gut microbes, prognostic value, and possible SFM subtypes-guided therapies for CRC.
Fig. 6 Summary of characteristics of SFM subtypes. These included pathological features, genomic markers, TME, epigenetic changes, gut microbes, prognostic value, and possible SFM subtypes-guided therapies for CRC.CC BY · PubMed Central

Verification

Claims the paper makes

11 Valid · 1 Not valid · 9 Not checked · 75 Own results and interpretations

52 more claims

Create a free account to see all 96 claims and to verify the 5 references still open.

See all claims — free account
  • Probably not a substantial claim. 23
  • It may state two separate claims as one. 21

Create a free account to see these claims.

Citations

Cites

Show all 94 references

Cited by

See all 41

Metadata

Authors
Zhu, X., Tian, X., Ji, L., Zhang, X., Cao, Y., Shen, C., Hu, Y., Wong, J. W. H., Fang, J.-Y., Hong, J., & Chen, H.
Affiliations
  • Shanghai Jiao Tong University (CN)
  • Cedars-Sinai Medical Center (US)
  • University of Hong Kong (HK)
Corresponding author
  • Fang, Jing-Yuan
  • Hong, Jie
  • Chen, Haoyan
Venue
npj Precision Oncology
Published
2021-02-12
Volume
5
Issue
1
Pages
7-7
ISSN
2397-768X (ISSN-L)
OpenAlex
W3128209586
Cited by
41
References
97
Citations by year
2026: 2 · 2025: 5 · 2024: 6 · 2023: 10 · 2022: 9 · 2021: 9
FWCI
3.10
Type
journal article
Language
en
Open access
Licence
Locations
  • npj Precision Oncology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Natural Science Foundation of China (201268, 81602518, 82073115, 31371273, 81874159, 81871901)
  • Shanghai Municipal Education Commission (20161309, 20152512)
  • Shanghai Jiao Tong University
Topics
  • Health Sciences › Medicine › Oncology › Colorectal Cancer Treatments and Studies
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
  • Health Sciences › Medicine › Oncology › Cancer Immunotherapy and Biomarkers
Keywords
colorectal cancer, tumor microenvironment, chemoresistance, gene expression signatures, FOLFIRI, FOLFOX, EGFR inhibitors, tumor subtypes, drug response prediction, immune cell infiltration, microsatellite instability, epithelial-mesenchymal transition, gut microbiota, prognosis, single-cell RNA sequencing, bulk RNA-seq

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: A tumor microenvironment-specific gene expression signature predicts chemotherapy resistance in colorectal cancer patients. https://app.cochat.ai/papers/10.1038/s41698-021-00142-x
The paper
Zhu, X., Tian, X., Ji, L., Zhang, X., Cao, Y., Shen, C., Hu, Y., Wong, J. W. H., Fang, J.-Y., Hong, J., & Chen, H. (2021). A tumor microenvironment-specific gene expression signature predicts chemotherapy resistance in colorectal cancer patients. npj Precision Oncology. https://doi.org/10.1038/s41698-021-00142-x

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.