BPG is committed to discovery and dissemination of knowledge
Observational Study Open Access
Copyright: ©Author(s) 2026. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-NonCommercial (CC BY-NC 4.0) license. No commercial re-use. See permissions. Published by Baishideng Publishing Group Inc.
World J Gastrointest Oncol. Nov 15, 2026; 18(11): 122507
Published online Nov 15, 2026. doi: 10.4251/wjgo.122507
Microbial succession driven by Alistipes depletion shapes the mesenchymal and immune-excluded consensus molecular subtype 4 colorectal cancer landscape
Jian-Dong Hu, Chen-Jie Qiu, Zhen-Biao Song, Zi-Wei Xia, Hui Geng, Department of General Surgery, Changzhou Hospital of Traditional Chinese Medicine, Changzhou 213000, Jiangsu Province, China
Hai-Tao Wang, Department of Gastrointestinal Surgery, The Third Affiliated Hospital of Soochow University, Changzhou 213000, Jiangsu Province, China
ORCID number: Jian-Dong Hu (0009-0007-8728-5887); Chen-Jie Qiu (0000-0003-2632-7397); Zhen-Biao Song (0000-0002-1803-8619); Zi-Wei Xia (0009-0008-4332-5426); Hai-Tao Wang (0000-0002-8470-064X); Hui Geng (0009-0005-9647-5602).
Co-corresponding authors: Hai-Tao Wang and Hui Geng.
Author contributions: Wang HT and Geng H contributed to supervision, writing - review, and editing, and they equally to this work as corresponding authors; Hu JD contributed to the conceptualization, data curation, formal analysis, and writing - original draft preparation; Wang HT contributed to the project administration; Hu JD, Qiu CJ, and Geng H contributed to investigation; Xia ZW and Song ZB contributed to validation; Song ZB contributed to the methodology; Qiu CJ and Xia ZW contributed to the resources. All authors contributed to the article and approved the submitted version.
AI contribution statement: The author declares that no AI tools were used.
Institutional review board statement: The study was reviewed and approved by the Institutional Review Board of Changzhou Hospital of Traditional Chinese Medicine [approval No. 2023-LL-021(L)].
Informed consent statement: All study participants, or their legal guardian, provided informed written consent prior to study enrollment.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
STROBE statement: The authors have read the STROBE Statement-checklist of items, and the manuscript was prepared and revised according to the STROBE Statement-checklist of items.
Data sharing statement: Data supporting the findings of this study are available from the corresponding author upon request.
Corresponding author: Hai-Tao Wang, Professor, Department of Gastrointestinal Surgery, The Third Affiliated Hospital of Soochow University, No. 185 Jiqian Street, Changzhou 213000, Jiangsu Province, China. 15961290078@163.com
Received: April 30, 2026
Revised: June 8, 2026
Accepted: June 25, 2026
Published online: November 15, 2026
Processing time: 188 Days and 18.5 Hours

Abstract
BACKGROUND

Colorectal cancer (CRC) is characterized by prominent molecular and tumor microenvironment heterogeneity, and is classified into four subgroups according to the Consensus Molecular Subtypes (CMS) system. The CMS4 subtype presents a mesenchymal phenotype and extensive stromal infiltration, accompanied by a poor patient prognosis. Nevertheless, the microbial mechanisms driving its tumor microecology remain unclear.

AIM

To delineate the patterns of microbial succession in the tumor microecology of CMS4, and to explore its association with metabolic disturbances and tumor immune evasion.

METHODS

A total of 11 Chinese patients with CRC were enrolled, including 4 patients with CMS4, 3 with CMS3, 3 with CMS2 and 1 with CMS1. Integrated multi-omics analyses including transcriptomics, immunomics, and microbiomics were performed on tumor samples. The GhostKOALA tool was used for microbial functional annotation. Host-microbe interaction models were constructed to screen for key regulatory genes.

RESULTS

CMS4 samples displayed a sharp depletion in Alistipes indistinctus, reducing butyrate metabolism and secondary bile acid regulation. This vacancy triggered a two-step invasion in which aerotolerant Streptococcus breached the mucosal barriers and consumed local oxygen, allowing obligate anaerobic Fusobacterium to dominate the niche (fold change = 2.32). The resulting microbial community cooperated with M2 macrophages to remodel the extracellular matrix, creating dense physical barriers that excluded T cells and elevated immune exclusion scores. Host-microbe modeling further identified prostate cancer susceptibility candidate 1 as a key gene that was consistently silenced by tumor-enriched bacteria.

CONCLUSION

This study suggests that Alistipes indistinctus and Fusobacterium may be involved in mesenchymal transition and immune evasion in CMS4 CRC. The proposed microbial succession model holds potential to serve as a prognostic biomarker and therapeutic target for this high-risk subtype, and further validation is warranted.

Key Words: Colorectal cancer; Consensus molecular subtype 4; Microbial succession; Immune exclusion; Alistipes indistinctus; Fusobacterium

Core Tip: Consensus Molecular Subtype 4 colorectal cancer exhibits a unique microbial succession pattern. Depletion of Alistipes indistinctus initiates a cascade of pathogenic events, ultimately making Fusobacterium the dominant microbial taxon. This microbial shift may mediate tumor immune exclusion through extracellular matrix remodelling, indicating considerable potential for these microbial signatures to act as prognostic biomarkers and therapeutic targets for this refractory cancer subtype.



INTRODUCTION

Colorectal cancer (CRC) is one of the leading types of cancers responsible for cancer-related incidence and mortality worldwide. In 2022, over 1.9 million new cases of CRC were diagnosed (including anal cancer) and approximately 904000 deaths occurred globally, accounting for nearly 10% of the total cancer incidence and mortality worldwide[1]. Despite continuous advances in screening technologies and therapeutic modalities, marked disparities in patient prognosis remain evident. These discrepancies are closely linked to the extensive molecular and tumor microenvironment (TME) heterogeneities intrinsic to the malignancy[2,3]. The Consensus Molecular Subtype (CMS) classification system categorizes CRC into four phenotypically distinct subtypes, each with unique biological features and important clinical implications[4].

Among all CMS subgroups, the CMS4 subtype is distinguished by a predominant mesenchymal phenotype. This subtype is characterized by widespread tumor stromal infiltration, epithelial-mesenchymal transition, active angiogenesis, and the formation of a highly immunosuppressive TME[5,6]. Patients with CMS4 generally have a poor prognosis and limited benefit from conventional chemotherapy and immunotherapy, highlighting the urgent need for novel, subtype-specific therapeutic strategies[6,7].

Studies have uncovered the critical roles of gut microbiota and host-microbiota interactions in CRC development, progression, and treatment response[8,9]. However, the composition and functional characteristics of intratumoral and intestinal microbiota across different CMS classifications and their interplay with host gene regulation remain unclear. In particular, how specific microbial taxa contribute to the immune-excluded, yet immune-active nature of CMS4 warrants further investigation[10,11].

Alistipes indistinctus has previously been reported to be abundant in late-stage CRC, where it potentially promotes mast cell maturation via interleukin-6 upregulation and enhances endoplasmic reticulum protein folding by activating the protein kinase R-like endoplasmic reticulum kinase-mediated unfolded protein response, thereby supporting tumor survival[12]. Nevertheless, whether this microbial feature or other specific taxonomic successions distinctively shape the mesenchymal CMS4 microenvironment has not yet been defined.

To address these questions, we performed an integrated multi-omics profiling - incorporating RNA-sequencing, immune infiltration estimation, and microbiome characterization - of tumor, adjacent tissue, and fecal samples obtained from Chinese patients with CRC. This study systematically characterized the microbial signatures and host molecular profiles across different CMS. The primary objective was to define the unique patterns of microbial succession that shape the tumor microbial niche of CMS4, and to elucidate the underlying mechanisms connecting such microbial changes to metabolic disturbances and tumor immune evasion. By investigating subtype-specific host-microbe interactions and features of microecological dysbiosis, this study further aimed to identify novel diagnostic biomarkers and develop potential targeted therapeutic targets for the CMS4 subtype.

MATERIALS AND METHODS
Research design

Eleven patients with CRC were included in this study, and paired tumor tissues, paracancerous tissues, and fecal samples were acquired. RNA-sequencing was performed to generate host gene expression profiles, which were compared to The Cancer Genome Atlas (TCGA) data and CMS template centroids by principal component analysis (PCA) to assess consistency. Subsequent analyses included CMS subtyping, differential gene expression, immune infiltration estimation, microbiome profiling, and host-microbiota interaction modeling (Figure 1).

Figure 1
Figure 1 Workflow of the study. Tumor, adjacent tissue, and fecal samples from 11 patients with colorectal cancer were analyzed. Gene expression profiles were compared with The Cancer Genome Atlas data and Consensus Molecular Subtypes (CMS) templates by principal component analysis, and classified by CMS using CMSclassifier. Differential expression, immune infiltration, and microbiome features were examined across subtypes, with host-microbiota associations further explored to identify CMS4-specific microbial signals. CMS: Consensus Molecular Subtypes.
Patient cohort and sample collection

Clinical samples were collected from patients diagnosed with CRC at the Changzhou Hospital of Traditional Chinese Medicine. PCA[13] was performed to avoid batch effects between the 11 samples and the other samples from TCGA. The study was reviewed and approved by the Institutional Review Board of Changzhou Hospital of Traditional Chinese Medicine [approval No. 2023-LL-021(L)] and written informed consent was obtained prior to enrollment. The anatomical distributions of the tumors and sample identifiers are shown in Table 1.

Table 1 The anatomical distribution of tumors and sample identifiers.
Sex
Age
ID
Diagnosis
TNM staging
Female60ZY0437041Malignant neoplasm of the rectumT3N1M0
Female72ZY0437440Malignant neoplasm of the rectumT2N0M0
Female69ZY0438776Carcinoma at the transverse-descending colon junctionT2N1M0
Female80ZY0444496Malignant neoplasm of the transverse colonT3N0M0
Male79ZY0443954Malignant neoplasm of the sigmoid colonT3N0M0
Male85ZY0447231Malignant neoplasm of the ascending colonT3N0M0
Male83ZY0447106Malignant neoplasm of the ascending colonT3N1M0
Female68ZY0447504Malignant neoplasm of the sigmoid colonT3N0M0
Male67ZY0448159Malignant neoplasm of the ileocecal colonT4N1M0
Male48ZY0452444Malignant neoplasm of the ascending colonT3N1M0
Female65ZY0457909Carcinoma of the ileocecal regionT2N0M0
CMS subtyping and transcriptomics

RNA extraction, sequencing, and pre-processing: Total RNA was extracted from the tumor and adjacent normal tissues using the TRIzol reagent. RNA integrity was verified (RIN ≥ 7) before library construction. Qualified libraries were sequenced on the Illumina NovaSeq platform using the paired-end 150 bp (PE150) mode. Raw reads were quality-checked and aligned to the human reference genome (GRCh38) using STAR[14], and gene-level quantification was performed using featureCounts[15]. Ensembl gene IDs were mapped to Entrez IDs using the org.Hs.eg.db package. The expression matrix was log2-transformed, and the z-score was normalized across genes.

CMS classification: Subtype classification was conducted using the CMSclassifier R package, applying both random forest and single-sample predictor models[4] to ensure robust assignment. The concordance between the two methods and the Euclidean distances to the subtype centroids were calculated to confirm the reliability of the subtype calls.

Differential expression and functional enrichment analyses: To identify differentially expressed genes (DEGs) across CMS classifications, normalized count matrices were analyzed in R (version 4.2.1) using DESeq2 (version 1.36.0) and edgeR (version 3.38.2)[16,17]. A variance-stabilizing transformation was applied to normalize the count data. DEGs were defined with an adjusted P < 0.05. For each CMS classification, DEGs were visualized with volcano plots indicating log2(fold change) and -log10(padj).

Subsequent functional enrichment was performed using the clusterProfiler package (version 4.4.4) in the org.Hs.eg.db annotation database[18]. Enrichment was tested via the hypergeometric distribution of gene ontology categories (biological process, cellular component, and molecular function) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Gene lists were first mapped to Entrez IDs and then tested with enrichGO and enrichKEGG functions. For each term, ontology, pathway ID, description, gene ratio, background ratio, raw P value, adjusted P value, q value, matched gene IDs, and hit counts were recorded. Pathways with adjusted P < 0.05 were considered significant and served as references for downstream functional interpretation and host-microbiota interaction modeling.

Immune infiltration analysis

Single-sample gene set enrichment analysis was conducted via the GSVA package. A curated panel of 24 immune cell-type gene signatures was used to compute relative enrichment scores[19,20]. Additional immunotherapy indicators were evaluated, including programmed death 1 response prediction, cytolytic activity[21], immune exclusion score, and immunophenoscore[22]. Spearman correlations were calculated between gene expression and immune scores for each CMS classification.

Microbiome profiling and analysis: Total DNA was extracted from tumors, adjacent tissues, and fecal samples. Sequencing libraries were prepared and sequenced using an Illumina NovaSeq 6000 platform. Raw reads were quality-filtered, and denoising and chimera removal were performed using the DADA2 package[23] in R to generate high-quality non-chimeric amplicon sequence variants for downstream analysis.

Microbial diversity and community structure analyses: Shannon diversity indices were calculated to compare α-diversity among CMS classifications. The Mann-Whitney U test and Fisher’s exact test were used to identify differentially abundant taxa between tumor [cancer tissue (CA)] and adjacent tissue [cancer-adjacent para-carcinoma (CP)] samples[24,25]. Community structure differences were visualized using a principal coordinate analysis based on Bray-Curtis distances[26].

Host-microbiota association analysis: To assess associations between microbial abundance and host gene expression, Elastic-Net regression[27] was applied separately to CA, CP, and fecal (F) samples using the glmnet package[28] in R with α = 0.1. Normalized gene expression and relative microbial abundance matrices were used as inputs. Genes with non-zero regression coefficients were identified for overlap analysis and further visualization.

KEGG orthology annotation and Sankey visualization: To assess the functional effects of microbial dysbiosis, we performed a cross-subtype comparative analysis. While Alistipes indistinctus was prevalent in CMS1-3, its complete depletion in CMS4 (mean = 0) marked a critical functional loss. We retrieved its reference proteome from the NCBI database and assigned KEGG orthology (KO) identifiers using GhostKOALA[29]. The annotated KOs were mapped via the KEGG representational state transfer application programming interface to three core modules: Oncogenic signaling, immune regulation, and apoptosis. These KO-pathway-function relationships were visualized using a Sankey diagram to evaluate how the absence of Alistipes indistinctus contributed to the loss of homeostatic control and development of the immune-excluded CMS4 niche.

Statistical analysis

All statistical analyses were performed using R (version 4.3) and Python (version 3.9). Multiple testing corrections were applied using the Benjamini-Hochberg method[30]. Significance was defined as P < 0.05 or false discovery rate-adjusted q < 0.1. Visualization was performed using ggplot2, seaborn, and matplotlib packages. The results of the Elastic-Net regression present the most significant results as indicators of the trend genes that were regulated.

RESULTS
Sample analysis in 11 patients and CMS classification

PCA showed that the global transcriptomic architecture of the included 11 Chinese CRC samples clustered closely with the classical TCGA reference cohorts and the core template definition of the CMS framework (Figure 2A). CMS classification of all patient samples was successfully performed using the CMSclassifier R package. To test whether the Western-derived classifier translates faithfully to Chinese clinical backgrounds, we mapped the posterior probabilities of our cohort against TCGA benchmarks (Figure 2B and C). The overall predictive trajectories, particularly for the mesenchymal CMS4 subtype, showed highly consistent topological alignment between both patient groups. Additionally, cross-model comparison demonstrated that the random forest and segmented spectral projection algorithms yielded identical classification assignments for the majority of our cases (Figure 2D). Several samples (e.g., XTG, WHZ) showed inconsistent prediction results across different models, which reflected the complementarity of classification logics among these algorithms.

Figure 2
Figure 2 Analysis of batch effect and the result of Consensus Molecular Subtypes classification. A: Scatter plot indicating no obvious batch effect between these samples and those samples in The Cancer Genome Atlas; B and C: Heat maps comparing the Consensus Molecular Subtypes classification results; D: Diagram showing the results of Consensus Molecular Subtypes classification for the 11 samples. PC: Principal component; RF: Random forest; SSP: Segmented spectral projection; CMS: Consensus Molecular Subtypes; NA: Not available.
Transcriptomic differences and functional enrichment

Differential expression analysis revealed distinct transcriptomic profiles across various CMS (Figure 3A). In the present study, CMS1, CMS2, and CMS3 presented prominent gene expression heterogeneity. In contrast, CMS4 was predominantly characterized by global upregulation of gene expression, suggesting that this subtype possesses a unique gene regulatory pattern linked to its phenotypic features.

Figure 3
Figure 3 Differential expression and enrichment analysis. A: Volcano plots of differentially expressed genes between tumor (cancer tissue) and adjacent (cancer-adjacent para-carcinoma) tissues; B: Gene ontology and Kyoto Encyclopedia of Genes and Genomes enrichment of differentially expressed genes; C: Upset plot showing shared enriched pathways across Consensus Molecular Subtypes classifications. CMS: Consensus Molecular Subtypes.

Functional enrichment analysis further uncovered remarkable biological differences among the CMS subtypes (Figure 3B). CMS1-CMS3 were significantly enriched in pathways associated with B cell activation, complement activation, and antibody-mediated immunity, which was consistent with the humoral immune responses observed in our study. Immune-related pathways showed weak enrichment in CMS4, while pathways involved in metabolism, stress response, and membrane structure exhibited strong correlations with this subtype. The results implied that alternative mechanisms contribute to tumor progression in CMS4.

Analysis of the overlap and specificity of enriched pathways indicated that each CMS subtype harbours both shared functional characteristics and exclusive phenotypic profiles (Figure 3C). Considering the relatively small sample size of this study, the differential pathways screened for CMS1 should be interpreted with caution, since inter-individual variation may confound the results. Despite this limitation, CMS1 had the greatest number of subtype-specific enriched pathways, indicative of its unique functional signatures. The enriched pathways of the other subtypes overlapped to varying degrees, reflecting that all CMS subgroups share core biological mechanisms as well as independent subtype-specific regulatory patterns.

Immune infiltration and subtype-specific immune patterns

Single-sample gene set enrichment analysis showed divergent immune cell infiltration patterns across CMS subtypes. Based on the dataset, CMS1 exhibited high overall enrichment of adaptive and innate immune cells, manifesting the typical features of an immune-activated TME. On the contrary, CMS2 displayed consistently low infiltration levels of all types of immune cells; CMS3 showed moderate immune cell infiltration accompanied by strong intra-subtype heterogeneity. Selective expansion of innate immune cells was observed in CMS4, with macrophages and natural killer cells being the most prominent populations (Figure 4A-C).

Figure 4
Figure 4 Immune infiltration patterns and functional immunological features across Consensus Molecular Subtypes classifications. A: Z-score normalized enrichment scores of 24 immune cell types estimated by single-sample gene set enrichment analysis; B: Heatmap comparing infiltration levels of representative immune cells among Consensus Molecular Subtypes (CMS1)-CMS4; C: Boxplots showing subtype-specific differences for five significantly different immune cell types; D: Correlations between top five subtype-specific marker genes and immune infiltration within each CMS; E: Violin plots of programmed death-ligand 1 response probability, cytolytic activity score, immune exclusion score, and immunophenoscore across CMS classifications. CMS: Consensus Molecular Subtypes; NK: Natural killer; ssGSEA: Single-sample gene set enrichment analysis; PD-1: Programmed cell death protein 1; CYT: Cytolytic activity score; IPS: Immunophenoscore.

Spearman’s correlation analysis was performed between the top five non-mitochondrial genes and immune cell scores in each subtype, and distinct gene-immune cell interaction patterns were identified. Genes in CMS1 were widely positively correlated with multiple immune cell subsets. By comparison, CMS2 and CMS3 presented stronger subtype-specific characteristics, indicative of an immunosuppressive TME. Correlations between genes and immune cells in CMS4 were primarily concentrated in innate immune cell populations (Figure 4D).

Multiple immune therapy-related scores revealed that CMS1 and CMS4 had high predicted response to programmed death-ligand 1 blockade, elevated cytolytic activity, and increased immune phenotype scores. Nevertheless, CMS4 samples also had higher immune exclusion scores, suggesting that despite the activation of innate immunity in this subtype, a physical barrier hindered effective T cell infiltration (Figure 4E).

Microbiome diversity and composition

Significant differences in the taxonomic composition of microbial communities were detected among CMS subtypes and samples derived from different tissues (Figure 5A). In paracancerous tissues (CP), CMS1 and CMS3 shared similar dominant microbiota, mainly Anaerolineae and Bacilli. These microbes exert intestinal protective effects, which were consistent with the immune-activated phenotypes of the two subtypes. Microbial communities in tumor tissues (CA) showed higher overall homogenization. Notably, CMS4 was specifically enriched in Synergistia and Spirochaetia, forming a unique subtype-specific microbial structure.

Figure 5
Figure 5 Microbiome diversity, composition, and host-microbe associations across Consensus Molecular Subtypes classifications. A: Stacked bar plots showing the relative abundance of dominant microbial taxa across Consensus Molecular Subtypes (CMS) classifications and sample types; B: Heatmap illustrating correlations between selected host genes and microbial phyla in cancer tissue, cancer-adjacent para-carcinoma, and F samples; C: Mean relative abundances of the top 20 differential species across CMS1-CMS4; D: Species-level differences (cancer tissue vs cancer-adjacent para-carcinoma) showing top 20 species with the largest mean abundance shifts per CMS classification; E: Shannon diversity indices comparing α-diversity of tumor (cancer tissue), adjacent (cancer-adjacent para-carcinoma), and fecal (F) microbiomes by CMS classification. CA: Cancer tissue; CP: Cancer-adjacent para-carcinoma; CMS: Consensus Molecular Subtypes.

Elastic-Net regression (α = 0.1) was applied to explore host gene-microbe correlations, and subtype- and tissue-specific association patterns were identified (Figure 5B). Genes including APOA4, MUC2, prostate, rectum and colon expressed 1 (PRAC1), REG1A, RPS4Y1 and UGT2B17 were correlated with at least one phylum-level microbial taxon in tumor tissues, paracancerous tissues, or fecal samples (F). Specifically, PRAC1 was significantly negatively correlated with various microbial taxa in tumor tissues, while such a correlation was absent in paracancerous tissues and feces. This finding confirmed the existence of tumor-specific host-microbe interactions in our cohort.

Relative abundance analysis of the top 20 differential microbial species revealed representative microbial signatures for each subtype (Figure 5C): Peptostreptococcus oralis was enriched in CMS2, Fusobacterium nucleatum dominated in CMS3, and Lachnoclostridium intestinale and Dolosigranulum pneumocystis were markedly elevated in CMS4. These microbial characteristics may be closely associated with the immunosuppressive and stroma-remodelling TME of CMS4.

Microbial composition differed evidently between tumor and paracancerous tissues (Figure 5D). Results of Mann-Whitney U test and Fisher’s exact test verified that species such as Escherichia coli and Prevotella oralis were enriched in tumor tissues, especially in CMS2 and CMS4. Collectively, these results demonstrated that each CMS subtype occupies an independent microbial niche, and gut microbiota may play a critical role in shaping the biological features of distinct CRC subtypes.

Microbial functional profiling of CMS4

Shannon diversity analysis showed that fecal samples obtained from patients with CMS4 had the lowest microbial diversity (mean Shannon index = 3.6908) among all CMS subtypes, which is a typical hallmark of gut dysbiosis (Figure 5E). Comparative analysis of microbiota across tumor tissues, paracancerous tissues, and fecal samples further clarified tissue-specific microbial distribution (Figure 5A and B). Akkermansia muciniphila was highly abundant in fecal samples, whereas Azospirillum and Parabacteroides were predominantly enriched in tumor and paracancerous tissues (Azospirillum: CA total = 23 reads, CP total = 0 reads across all three CMS4 patients) (Figure 5C and D).

To explore the biological consequences of gut dysbiosis, we conducted inter-subtype microbial differential analysis. The Alistipes genus as a whole was markedly reduced in CMS4 tumors relative to paired paracancerous tissues (CA total: 389 reads; CP total: 1613 reads; fold change = 0.24), and among which, Alistipes indistinctus was universally present and biologically active in all tissues of CMS1, CMS2, and CMS3, but completely undetectable in CMS4 sample (average relative abundance = 0), indicating that the depletion of this bacterium can be a key characteristic of CMS4. To elucidate the functional defects caused by the loss of Alistipes indistinctus, we performed proteome functional on its NCBI reference proteome annotation using the GhostKOALA tool, and 8401 KO terms were obtained. All annotated KO terms were categorized into three core functional modules via Sankey diagram visualization: Oncogenic signaling pathways, immune regulation, and apoptosis. Functional analysis suggested that the absence of Alistipes indistinctus directly disturbs pathways related to cell cycle regulation and immune evasion, which is likely a major contributor to the high aggressiveness and immune exclusion phenotype of CMS4 tumors.

Sankey diagrams were used to visualize the hierarchical relationships among KO terms, signaling pathways, and functional modules (Figure 6A). The results indicated that in a normal intestinal microenvironment and non-mesenchymal subtypes (CMS1-CMS3), Alistipes indistinctus maintains intestinal homeostasis by regulating cell cycle checkpoints and inhibiting oncogenic signaling. The complete depletion of this bacterium in CMS4 leads to the loss of its homeostatic regulatory functions and collapse of intestinal protection, thereby promoting extracellular matrix (ECM) remodelling, tumor immune evasion, and persistent activation of oncogenic pathways, ultimately forming the distinctive TME of CMS4. In summary, the depletion rather than the presence of Alistipes indistinctus acts as a potential driving factor for the formation of the CMS4 TME, featuring extensive immune cell infiltration but suppressed immune function.

Figure 6
Figure 6 Microbial features and functional pathways associated with Consensus Molecular Subtypes 4 subtype. A: Sankey diagram visualizing Kyoto Encyclopedia of Genes and Genomes orthology-pathway-function connections showing modules enriched in Alistipes indistinctus; B: Species-level differences between tumor and adjacent tissue (cancer tissue vs cancer-adjacent para-carcinoma) highlighting the top species with the largest mean abundance shifts; C: Mean relative abundances of the top 20 species identified in Consensus Molecular Subtypes 4 samples; D: Heatmap showing sample-level abundance (log10-transformed) of the top 20 species in Consensus Molecular Subtypes 4 fecal samples; E: Heatmap illustrating Alistipes indistinctus proteins mapped to colorectal cancer-related Kyoto Encyclopedia of Genes and Genomes pathways. CMS: Consensus Molecular Subtypes; CRC: Colorectal cancer; KEGG: Kyoto Encyclopedia of Genes and Genomes.

Differential abundance analysis further screened signature microbes of CMS4 (Figure 6B). Three microbial taxa were specifically and significantly increased solely in tumor tissues of CMS4: Parvimonas micra (JLD-CA: 169; XJF-CA: 572; XHZ-CA: 1688; total: 2429 reads), Azospirillum species 47_25 (unclassified Rhodospirillales, CA total: 23 reads; CP: Undetectable), and Massiliensis (formerly Bacteroides massiliensis). In addition, CMS4 was enriched in unclassified Fusobacterium (CA total: 15769 reads; CP total: 6800 reads; fold change = 2.32) and Bacteroides vulgatus, which were also detected in other CMS subtypes. These microbial signatures further confirmed that the CMS4 TME harbours a unique microbial ecosystem.

DISCUSSION

This study integrated CMS of CRC with multi-omics data to explore subtype-specific host-microbe interactions. Based on a Chinese patient cohort, this study combined transcriptomics, immunomics, and microbiomics analyses to illustrate the crosstalk between tumor biological characteristics and gut microecology across different CMS subtypes, with a particular focus on CMS4, a subtype with prominent clinical significance but poorly elucidated mechanisms.

Consistent with established classifications[4], CMS1 is characterized by massive infiltration of adaptive immune cells and robust humoral immune activation; CMS4 presents abundant immune cell infiltration but overall immune dysfunction. Our cohort data demonstrated that CMS4 had the lowest fecal microbial diversity. Species-level analysis revealed that the microbiota of CMS4 is dominated by mucin-degrading bacteria and opportunistic pathogens, while Azospirillum and Parabacteroides are enriched in tumor and paracancerous tissues. Such microecological imbalance is a critical driver for the formation of the immunosuppressive and stroma-rich TME in CMS4.

In this context, Alistipes indistinctus was identified as a key commensal bacterium associated with intestinal homeostasis, whose functional repertoire sustains normal intestinal physiology. Pathway analysis revealed that the proteome of Alistipes indistinctus participates in genetic information processing and carbohydrate metabolism, and modulates core biological modules including apoptosis, programmed death-ligand 1 signaling, and cell cycle (Figure 6).

Traditional views hold that microorganisms are merely passive bystanders during tumor progression. In contrast, our results support the applicability of the microbial succession-driven dysbiosis hypothesis in CMS4. We hypothesize that the complete depletion of Alistipes indistinctus in CMS4 acts as the initiating event of the pathological cascade. The resultant functional deficiency disrupts local metabolic and immune homeostasis. With the disappearance of niche competition, pathogenic microbes begin to colonize sequentially: Aerotolerant streptococci first invade the intestinal tract, consume local oxygen, and disrupt the intestinal mucosal barrier[31], eventually creating a hypoxic microenvironment that favours the massive proliferation of obligate anaerobes such as Fusobacterium.

This pattern of microbial succession was corroborated by transcriptomic results showing activated stress response and membrane remodelling pathways in CMS4. Specifically, Fusobacterium collaborates with M2-polarized macrophages to drive extensive ECM remodelling and tissue fibrosis, forming a physical barrier. This accounts for the high immune exclusion score and impaired T cell infiltration in CMS4 tumors. Therefore, the depletion of Alistipes indistinctus and the subsequent microbial succession are pivotal mechanisms underlying the formation of the “high infiltration but low immune activity” TME, as well as the poor overall prognosis associated with CMS4.

With subtype-specific microbial alterations, PRAC1 was identified as a pan-subtype conserved transcriptional regulatory hub that is uniformly modulated by tumor-enriched microbes across all CMS subgroups. PRAC1 is a tissue-specific gene predominantly expressed in the prostate, rectum, and distal colon[32], and is highly upregulated in CRC cell lines. PRAC1 is involved in region-specific transcriptional regulation in the colorectum; its expression differs between left- and right-sided CRC and is potentially correlated with transcription factors such as homeobox B13[33]. Gut microbiota can regulate host gene expression via epigenetic modification and metabolite-mediated signaling[34]. The systematic suppression of PRAC1 by pathogenic consortia suggests that it acts as a central checkpoint for microbe-mediated immune evasion. Consequently, down-regulated PRAC1 may serve as an indicator of severe mucosal dysbiosis and a putative marker for immune escape, offering a novel diagnostic or therapeutic handle that could complement current immunotherapies in high-risk CRC populations.

The relatively small sample size is the main limitation of this exploratory study, which may reduce statistical power and preclude definitive clinical conclusions. Accordingly, the model of Alistipes indistinctus depletion and the subsequent microbial succession mechanism proposed herein should be regarded as a research hypothesis rather than validated clinical diagnostic biomarkers. Nevertheless, this study adopted a strict triple-matched design (paired tumor tissues, paracancerous mucosa, and preoperative fecal samples) with patients serving as their own controls. This design largely eliminated confounders such as genetic background and living habits, and guaranteed the biological rationality of our conclusions. Further multicenter prospective validation and mechanistic experiments using germ-free animal models combined with fecal microbiota transplantation will be performed to clarify the causal relationship between Alistipes indistinctus depletion and the progression of mesenchymal-type CRC.

CONCLUSION

This integrative multi-omics study demonstrates that the aggressive mesenchymal phenotype of CRC CMS4 is linked to tumor-related microbial succession and disrupted microecological homeostasis. Depletion of the gut commensal bacterium Alistipes indistinctus leads to severe functional deficiency, which facilitates the colonization of pathogenic microbes, induces ECM remodelling, and triggers M2 macrophage-mediated tumor immune exclusion. Furthermore, pan-subtype suppression of host PRAC1 by microbiota reveals a universal regulatory pathway for microbe-driven tumor immune evasion. Restoring intestinal microecological balance and targeting this conserved host-microbe signaling axis may provide novel subtype-specific therapeutic strategies to halt tumor progression in patients with mesenchymal-type CRC.

References
1.  Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, Jemal A. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74:229-263.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 16785]  [Cited by in RCA: 18093]  [Article Influence: 9046.5]  [Reference Citation Analysis (55)]
2.  Meyiah A, Khan FI, Alfaki DA, Murshed K, Raza A, Elkord E. The colorectal cancer microenvironment: Preclinical progress in identifying targets for cancer therapy. Transl Oncol. 2025;53:102307.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 21]  [Reference Citation Analysis (1)]
3.  Zhai Z, Yu X, Yang B, Zhang Y, Zhang L, Li X, Sun H. Colorectal cancer heterogeneity and targeted therapy: Clinical implications, challenges and solutions for treatment resistance. Semin Cell Dev Biol. 2017;64:107-115.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 48]  [Cited by in RCA: 79]  [Article Influence: 7.9]  [Reference Citation Analysis (0)]
4.  Guinney J, Dienstmann R, Wang X, de Reyniès A, Schlicker A, Soneson C, Marisa L, Roepman P, Nyamundanda G, Angelino P, Bot BM, Morris JS, Simon IM, Gerster S, Fessler E, De Sousa E Melo F, Missiaglia E, Ramay H, Barras D, Homicsko K, Maru D, Manyam GC, Broom B, Boige V, Perez-Villamil B, Laderas T, Salazar R, Gray JW, Hanahan D, Tabernero J, Bernards R, Friend SH, Laurent-Puig P, Medema JP, Sadanandam A, Wessels L, Delorenzi M, Kopetz S, Vermeulen L, Tejpar S. The consensus molecular subtypes of colorectal cancer. Nat Med. 2015;21:1350-1356.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 4335]  [Cited by in RCA: 4086]  [Article Influence: 371.5]  [Reference Citation Analysis (28)]
5.  Peddareddigari VG, Wang D, Dubois RN. The tumor microenvironment in colorectal carcinogenesis. Cancer Microenviron. 2010;3:149-166.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 189]  [Cited by in RCA: 191]  [Article Influence: 11.9]  [Reference Citation Analysis (2)]
6.  Leonard NA, Reidy E, Thompson K, McDermott E, Peerani E, Tomas Bort E, Balkwill FR, Loessner D, Ryan AE. Stromal Cells Promote Matrix Deposition, Remodelling and an Immunosuppressive Tumour Microenvironment in a 3D Model of Colon Cancer. Cancers (Basel). 2021;13:5998.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 1]  [Cited by in RCA: 15]  [Article Influence: 3.0]  [Reference Citation Analysis (0)]
7.  Strating E, Wassenaar E, Verhagen M, Rauwerdink P, van Schelven S, de Hingh I, Rinkes IB, Boerma D, Witkamp A, Lacle M, Fodde R, Volckmann R, Koster J, Stedingk K, Giesel F, de Roos R, Poot A, Bol G, Lam M, Elias S, Kranenburg O. Fibroblast activation protein identifies Consensus Molecular Subtype 4 in colorectal cancer and allows its detection by (68)Ga-FAPI-PET imaging. Br J Cancer. 2022;127:145-155.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 3]  [Cited by in RCA: 31]  [Article Influence: 7.8]  [Reference Citation Analysis (0)]
8.  Alexander JL, Scott AJ, Pouncey AL, Marchesi J, Kinross J, Teare J. Colorectal carcinogenesis: an archetype of gut microbiota-host interaction. Ecancermedicalscience. 2018;12:865.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 19]  [Cited by in RCA: 24]  [Article Influence: 3.0]  [Reference Citation Analysis (0)]
9.  Li X, Wu D, Li Q, Gu J, Gao W, Zhu X, Yin W, Zhu R, Zhu L, Jiao N. Host-microbiota interactions contributing to the heterogeneous tumor microenvironment in colorectal cancer. Physiol Genomics. 2024;56:221-234.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 11]  [Cited by in RCA: 14]  [Article Influence: 7.0]  [Reference Citation Analysis (1)]
10.  Purcell RV, Visnovska M, Biggs PJ, Schmeier S, Frizelle FA. Distinct gut microbiome patterns associate with consensus molecular subtypes of colorectal cancer. Sci Rep. 2017;7:11590.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 151]  [Cited by in RCA: 190]  [Article Influence: 21.1]  [Reference Citation Analysis (4)]
11.  Ternes D, Karta J, Tsenkova M, Wilmes P, Haan S, Letellier E. Microbiome in Colorectal Cancer: How to Get from Meta-omics to Mechanism? Trends Microbiol. 2020;28:401-423.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 205]  [Cited by in RCA: 193]  [Article Influence: 32.2]  [Reference Citation Analysis (3)]
12.  Liu J, Huang X, Chen C, Wang Z, Huang Z, Qin M, He F, Tang B, Long C, Hu H, Pan S, Wu J, Tang W. Identification of colorectal cancer progression-associated intestinal microbiome and predictive signature construction. J Transl Med. 2023;21:373.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in RCA: 45]  [Reference Citation Analysis (0)]
13.  Gower JC. Some Distance Properties of Latent Root and Vector Methods Used in Multivariate Analysis. Biometrika. 1966;53:325-338.  [PubMed]  [DOI]
14.  Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15-21.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 48566]  [Cited by in RCA: 39975]  [Article Influence: 3075.0]  [Reference Citation Analysis (12)]
15.  Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics. 2014;30:923-930.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 25020]  [Cited by in RCA: 20404]  [Article Influence: 1700.3]  [Reference Citation Analysis (4)]
16.  Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2010;26:139-140.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 38466]  [Cited by in RCA: 32272]  [Article Influence: 2017.0]  [Reference Citation Analysis (12)]
17.  Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:550.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 82919]  [Cited by in RCA: 70368]  [Article Influence: 5864.0]  [Reference Citation Analysis (54)]
18.  Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16:284-287.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 30559]  [Cited by in RCA: 27638]  [Article Influence: 1974.1]  [Reference Citation Analysis (38)]
19.  Yoshihara K, Shahmoradgoli M, Martínez E, Vegesna R, Kim H, Torres-Garcia W, Treviño V, Shen H, Laird PW, Levine DA, Carter SL, Getz G, Stemke-Hale K, Mills GB, Verhaak RG. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 7649]  [Cited by in RCA: 7370]  [Article Influence: 566.9]  [Reference Citation Analysis (12)]
20.  Bindea G, Mlecnik B, Tosolini M, Kirilovsky A, Waldner M, Obenauf AC, Angell H, Fredriksen T, Lafontaine L, Berger A, Bruneval P, Fridman WH, Becker C, Pagès F, Speicher MR, Trajanoski Z, Galon J. Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human cancer. Immunity. 2013;39:782-795.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 3609]  [Cited by in RCA: 3325]  [Article Influence: 255.8]  [Reference Citation Analysis (14)]
21.  Rooney MS, Shukla SA, Wu CJ, Getz G, Hacohen N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell. 2015;160:48-61.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 3179]  [Cited by in RCA: 2969]  [Article Influence: 269.9]  [Reference Citation Analysis (12)]
22.  Charoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder D, Hackl H, Trajanoski Z. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. Cell Rep. 2017;18:248-262.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 3970]  [Cited by in RCA: 3702]  [Article Influence: 411.3]  [Reference Citation Analysis (10)]
23.  Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJ, Holmes SP. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods. 2016;13:581-583.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 30465]  [Cited by in RCA: 22037]  [Article Influence: 2203.7]  [Reference Citation Analysis (27)]
24.  Fisher RA. On the Interpretation of χ 2 from Contingency Tables, and the Calculation of P. J R Stat Soc. 1922;85:87-94.  [PubMed]  [DOI]
25.  Mann HB, Whitney DR. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. Ann Math Statist. 1947;18:50-60.  [PubMed]  [DOI]
26.  Bray JR, Curtis JT. An Ordination of the Upland Forest Communities of Southern Wisconsin. Ecol Monogr. 1957;27:325-349.  [PubMed]  [DOI]  [Full Text]
27.  Zou H, Hastie T. Regularization and Variable Selection Via the Elastic Net. J R Stat Soc Ser B Stat Methodol. 2005;67:301-320.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 9695]  [Cited by in RCA: 7666]  [Article Influence: 365.0]  [Reference Citation Analysis (2)]
28.  Engebretsen S, Bohlin J. Statistical predictions with glmnet. Clin Epigenetics. 2019;11:123.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 950]  [Cited by in RCA: 894]  [Article Influence: 127.7]  [Reference Citation Analysis (7)]
29.  Kanehisa M, Sato Y, Morishima K. BlastKOALA and GhostKOALA: KEGG Tools for Functional Characterization of Genome and Metagenome Sequences. J Mol Biol. 2016;428:726-731.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 1737]  [Cited by in RCA: 2964]  [Article Influence: 269.5]  [Reference Citation Analysis (3)]
30.  Haynes W.   Benjamini-Hochberg Method. In: Dubitzky W, Wolkenhauer O, Cho KH, Yokota H. Encyclopedia of Systems Biology. New York: Springer, 2013.  [PubMed]  [DOI]  [Full Text]
31.  Parker BJ, Wearsch PA, Veloo ACM, Rodriguez-Palacios A. The Genus Alistipes: Gut Bacteria With Emerging Implications to Inflammation, Cancer, and Mental Health. Front Immunol. 2020;11:906.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 281]  [Cited by in RCA: 1396]  [Article Influence: 232.7]  [Reference Citation Analysis (5)]
32.  Liu XF, Olsson P, Wolfgang CD, Bera TK, Duray P, Lee B, Pastan I. PRAC: A novel small nuclear protein that is specifically expressed in human prostate and colon. Prostate. 2001;47:125-131.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 30]  [Cited by in RCA: 34]  [Article Influence: 1.4]  [Reference Citation Analysis (0)]
33.  Xie B, Bai B, Xu Y, Liu Y, Lv Y, Gao X, Wu F, Fang Z, Lou Y, Pan H, Han W. Tumor-suppressive function and mechanism of HOXB13 in right-sided colon cancer. Signal Transduct Target Ther. 2019;4:51.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 18]  [Cited by in RCA: 30]  [Article Influence: 4.3]  [Reference Citation Analysis (0)]
34.  Garrett WS. Cancer and the microbiota. Science. 2015;348:80-86.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 1216]  [Cited by in RCA: 1131]  [Article Influence: 102.8]  [Reference Citation Analysis (9)]
Footnotes

Peer review: Externally peer reviewed.

Peer-review model: Single blind

Specialty type: Oncology

Country of origin: China

Peer-review report’s classification

Scientific quality: Grade B, Grade C

Novelty: Grade B, Grade B

Creativity or innovation: Grade B, Grade C

Scientific significance: Grade C, Grade C

P-Reviewer: Kim KG, PhD, South Korea; Schirosi L, PhD, Japan S-Editor: Hu XY L-Editor: A P-Editor: Yang YQ

Write to the Help Desk