Enterocloster bolteae drives hyperglycemia-associated colorectal cancer cell proliferation via Rac/Rho-MSH6-Elongin C signaling pathway

Abstract

BACKGROUND

Diabetes mellitus (DM) is an independent risk factor for colorectal cancer (CRC). However, the microbial and molecular mechanisms linking hyperglycemia to CRC susceptibility remain incompletely understood. Gut microbiota dysbiosis is important for shaping the intestinal microenvironment under diabetic conditions.

AIM

To identify potential bacterial strains relevant to CRC risk in patients with DM.

METHODS

Fecal metagenomic sequencing was performed using samples from patients with DM and CRC (DM-CRC, n = 11), DM alone (DM, n = 8), CRC without DM (CRC, n = 10), and healthy controls (n = 11). The microbial composition and functional profiles were analyzed using comprehensive bioinformatics pipelines. Functional validation was conducted using human CRC cell lines (LoVo and RKO) cultured under high-glucose conditions and treated with Enterocloster bolteae (E. bolteae). Cell proliferation was assessed using proliferating cell nuclear antigen expression, cell counting, and Cell Counting Kit-8 assays. Key signaling pathways were investigated using quantitative polymerase chain reaction, western blotting, and siRNA-mediated gene silencing.

RESULTS

Metagenomic analysis revealed significant alterations in the gut microbial diversity and composition in the DM-CRC group compared with all other cohorts. E. bolteae was markedly enriched in DM-CRC patients and exhibited the highest relative abundance among the discriminative species. Functional enrichment analysis identified DM-CRC-specific microbial pathways associated with carcinogenesis, including RAC/RHO, MSH6, Elongin C, and TRK-related signaling. In vitro, E. bolteae supernatant significantly enhanced CRC cell proliferation under hyperglycemic conditions, accompanied by an upregulation of PCNA and increased cell viability. Notably, RAC, Rho, MSH6, and Elongin C expression levels were significantly elevated following E. bolteae exposure. Silencing RAC/RHO, MSH6, or Elongin C abolished E. bolteae-induced proliferative effects, confirming their functional involvement.

CONCLUSION

Our findings demonstrated that gut microbiota dysbiosis in DM is characterized by E. bolteae enrichment, which promotes CRC cell proliferation under hyperglycemic conditions. Mechanistically, E. bolteae appears to exert these tumor-promoting effects by activating the RAC/RHO-MSH6-Elongin C signaling pathway. Collectively, these results identify E. bolteae as a potential microbial driver and contributing factor in diabetes-associated CRC, suggesting that targeting this organism may offer novel strategies for both risk stratification and therapeutic intervention in DM.

Key Words: Enterocloster bolteae; Colorectal cancer; Diabetes; Intestinal flora imbalance; Proliferation

Core Tip: Diabetes mellitus is an established independent risk factor for colorectal cancer (CRC). We identified Enterocloster bolteae as a discriminative gut microbial signature specific to patients with diabetes mellitus and CRC. Functional pathway analysis further revealed distinctive enrichment of a carcinogenesis-related signaling pathway in this cohort, which included RAC, RHO, MSH6, and Elongin C, and this finding was subsequently validated in vitro. These findings establish a link between a diabetes-enriched gut bacterium and the activation of oncogenic pathways in CRC.

INTRODUCTION

Hyperinsulinemia and hyperglycemia promote aberrant cell proliferation[1]. Animal models of diabetes mellitus (DM) are highly susceptible to mutations in genes associated with colorectal cancer (CRC)[2]. Large-scale clinical studies have confirmed that DM is an independent risk factor for CRC and patients with DM are at a higher risk of developing CRC[3-6]. One study on the gut microbiota demonstrated that DM affects CRC susceptibility[7]. Our research group has shown that intestinal epithelial exposure to high glucose levels increases the risk for CRC[8,9].

The human gut contains approximately 2 × 10³ microbial species[10]. Intestinal microorganisms metabolize non-digestible food components, synthesize vitamins and other essential nutrients, detoxify metabolites, regulate immune responses, signal epithelial cell renewal, maintain mucosal integrity, and secrete antibacterial substances[11]. Dysbiosis may promote chronic inflammation and production of carcinogenic metabolites, leading to tumor formation[12]. Studies have shown variations in microbial abundance between intestinal tumors and adjacent non-tumor mucosa and in fecal samples between patients with CRC and controls. Further investigation of the intestinal microbiota may help elucidate its role in the etiology and diagnosis of tumors[13,14].

Our previous studies have revealed that abnormal proliferation and differentiation of intestinal epithelial cells in DM are closely associated with the intestinal gut microbial microenvironment[8,9,15,16]. To better predict CRC susceptibility in patients with DM and identify potential therapeutic strategies, we aimed to identify the most relevant bacterial strains.

MATERIALS AND METHODS

Clinical sample collection

A total of 82 fecal specimens were collected. Only 40 samples met the predefined exclusion criteria and were used for metagenomic analysis. These specimens were categorized into four groups: Patients with DM (n = 8), patients with both DM and CRC (DM-CRC, n = 11), patients with CRC without DM (n = 10), and healthy controls (n = 11). Participants ranged in age from 25 years to 65 years and met the diagnostic criteria for type 2 DM. Individuals with other organic diseases including hypertension, viral hepatitis, other infections, malignancies, or familial colon polyposis were excluded. Fecal samples from healthy individuals and patients with CRC without DM or other complications were included as controls. The collected data was used for bioinformatics analyses. The clinical data are presented in Table 1.

Table 1 Clinical data, n (%)/mean ± SD.

Characteristic	DM group (n = 8)	DM-CRC group (n = 11)	DM vs DM-CRC, P value	CRC group (n = 10)	Control group (n = 11)	P value
Age (years)	55.7 ± 10.4	56.3 ± 11.5	0.7812	59.3 ± 11.8	53.5 ± 10.6	0.151
Male gender	4 (50)	5 (45.5)	0.9175	6 (60)	6 (54.5)	0.053
Smoking	2 (25)	3 (27.3)	0.8550	3 (30)	4 (36.4)	0.051
Alcohol drinking	2 (25)	3 (27.3)	0.9175	3 (30)	3 (27.3)	0.193
Family history of CRC	2 (25)	2 (18.2)	0.9175	2 (20)	2 (18.2)	0.524
Family history of colorectal polyps	3 (37.5)	2 (27.3)	0.6580	4 (40)	2 (18.2)	0.030
Regular antidiabetic drugs use	7 (87.5)	7 (63.6)	0.2681	0	0	< 0.001
BMI (kg/m2)	24.4 ± 3.9	22.2 ± 3.6	0.2498	22.5 ± 3.7	23.5 ± 3.5	0.025
FPG (mmol/L)	9.8 ± 1.2	10.1 ± 1.3	0.7211	4.1 ± 1.2	5.3 ± 1.1	< 0.001
TG (mmol/L)	3.0 ± 0.9	1.5 ± 0.9	0.0036	1.8 ± 0.9	2.1 ± 0.9	0.030
TCH (mmol/L)	5.1 ± 1.3	4.7 ± 1.3	0.4478	4.8 ± 1.3	4.9 ± 1.3	0.627
LDL (mmol/L)	3.01 ± 0.9	3.01 ± 1.0	0.6217	3.01 ± 1.0	3.01 ± 1.0	0.766
HDL (mmol/L)	0.9 ± 0.4	0.8 ± 0.3	0.6307	1.0 ± 0.3	2.0 ± 0.3	< 0.001

DM: Diabetes mellitus; CRC: Colorectal cancer; BMI: Body mass index; FPG: Fasting plasma glucose; TG: Triglyceride; TCH: Total cholesterol; LDL: Low-density lipoprotein cholesterol; HDL: High-density lipoprotein cholesterol.

The inclusion criterias: (1) Age between 18-75 years; (2) Diagnosis of DM[17] with no complications or other organic disease, and voluntary acceptance of endoscopic examination and agreement to enter the clinical trial (DM group); (3) Concurrent diagnosis of DM and CRC, with a DM diagnosis at least 5 years prior to CRC diagnosis (DM-CRC group); (4) Diagnosis of CRC according to the established criteria (CRC group); (5) No diagnosis of any other organic disease, and voluntary acceptance of endoscopic examination and agreement to enter the clinical trial (control group); and (6) None of the exclusion criteria were met, and each participant provided signed consent for both study participation and specimen retention by our research group. Participants were required to meet criteria (1) and (6), plus at least one of (2)-(5).

The exclusion criteria: Age < 18 years or > 75 years, other complications or comorbidities, hereditary diseases, severe infection, antibiotic use in the previous week before participation, interstitial tumors, neuroendocrine tumors, non-primary tumors of the colon, and lack of consent to participation.

Ethics approval and consent to participate

All participants provided written informed consent prior to the use of their clinical samples and associated research data. The overall protocol strictly adhered to the guidelines of the Institutional Review Board Committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University, approval No. 2021.174.

Bioinformatics analysis

Metagenomic analyses were performed on fecal samples from the DM (n = 8), DM-CRC (n = 11), CRC (n = 10), and control (n = 11) groups. The procedural steps were as follows: (1) Total microbial DNA was extracted from the original samples for each group; (2) DNA samples were tested for total amount and purity. Samples that met the requirements were further processed, while those that did not meet the requirements were either re-extracted or purified based on the specific deficiency; (3) Qualified DNA samples were added to a fragmentation buffer and randomly sheared using an ultrasonicator. The resulting short DNA fragments were used for library construction and quality assessment; (4) Libraries passing quality inspection were sequenced using the MGI Seq PE150 platform; (5) Raw image data files from sequencing were processed through base calling analysis to generate raw sequencing reads (raw reads); and (6) The results were stored in FASTQ file format containing sequence information (reads) and corresponding quality scores.

Bioinformatics analysis was conducted as follows: (1) Data quality control; (2) Metagnome assembly; (3) Gene prediction; (4) Species annotation; (5) Functional annotations against common databases; and (6) Dimensionality reduction analysis. For dimensionality reduction, principal component analysis and non-metric multidimensional scaling (NMDS) were performed based on species and functional abundance tables. Principal coordinate analysis was performed using Bray-Curtis distance based on functional abundance, and sample clustering analysis was also conducted. NMDS utilized Bray-Curtis dissimilarity matrices, and stress values were interpreted according to Kruskal’s classification: < 0.05 (excellent), 0.05-0.1 (good), and 0.1-0.2 (usable). When grouping information was available, linear discriminant analysis effect size multivariate statistical analysis was performed to identify differentially abundant species and functions between sample groups.

Colon cancer cell culture

The human CRC cell lines LoVo and RKO were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM, 4.5 g/L D-glucose) supplemented with 10% fetal bovine serum, 100 μg/mL streptomycin, and 100 U/mL penicillin at 37 °C in a humidified incubator until reaching a concentration of approximately 5.0 × 10⁶/high-power field. Enterocloster bolteae (E. bolteae) bacteria were cultured anaerobically at 37 °C for 24 hours. When the bacterial concentration reached 1.0 × 10⁶/high-power field, the culture was centrifuged and filtered to retain cell-free supernatant. This bacterial supernatant was used to treat the cultured cells at final dilution ratios of 1:0 (neat), 1:1, 1:10, 1:100, and 1:1000 in the culture medium. The cells were harvested at 0 hours, 24 hours, 48 hours, and 72 hours post treatment. RNA was extracted, and quantitative polymerase chain reaction (qPCR) was performed to detect the mRNA expression of proliferating cell nuclear antigen (PCNA). Cells were counted using an image-based cell counter (specific steps are described below). Cell viability was assessed using the Cell Counting Kit-8 (Bio-Rad, CA, United States) by measuring the absorbance at 450 nm.

Image-based cell counting

Cell viability was determined by mixing cell suspensions with an equal volume of trypan blue. Viable cells with intact plasma membranes excluded the dye, whereas non-viable cells with compromised membranes stained blue. Cell concentration was adjusted to a range of 5 × 10⁵ cells/mL to 1 × 10⁷ cells/mL. For image acquisition, Countstar IC1000 was used, which automatically aspirated samples into a microfluidic counting chamber and rapidly captured digital images (human, 5-180 μm) using a built-in camera under a light microscope. For image analysis, the software automatically identified particles within the image and differentiated them based on preset parameters of size and color/brightness contrast. The software analyzed each “object” within the image based on the following parameters.

Color/brightness threshold distinguishing viable from non-viable cells: This was a critical step for differentiating viable cells from non-viable cells. Owing to the compromised membranes of non-viable cells, the trypan blue dye entered the cells, resulting in a dark, opaque appearance as solid circles in the image. The software identified these dark particles as non-viable cells by applying a predefined color intensity or brightness threshold. Due to their intact cell membranes, viable cells appeared as bright halo-like ring structures under the microscope, characterized by a bright periphery while the central nuclear area remains relatively bright.

Cell morphology parameters: As relying solely on color is insufficient, it is necessary to exclude debris that may be similar in size and shape to the cells. The software utilized predefined parameters for the cell diameter range and circularity for verification.

Size: The instrument default detection range is 5-180 μm, which can capture the majority of animal cells. Only bright particles with a diameter falling within this range were counted as viable cells.

Circularity: The software calculated circularity of each object. Only particles with a typical round or oval shape were considered. For example, an irregularly shaped but bright fragment could be excluded by this parameter.

For cell lines prone to aggregation, such as LoVo, the IC1000 offers a particularly useful feature. It intelligently identifies cell clusters and, through algorithmic analysis of the cluster area and optical density, estimates the number of individual cells contained within. This provides a correction during counting, yielding results that more closely reflect the true value.

Bolteae supernatant treatment

Bacterial strains were sequenced and identified using specific primers (Table 2). The ratio of Bolteae supernatant to culture medium (high Glu DMEM) was 0:1, 1:1, 1:10, 1:100, or 1:1000, respectively. The cells were harvested at 0 hours, 24 hours, 48 hours, and 72 hours post-treatment, followed by RNA extraction and qPCR.

Table 2 Primer sequences used in this study, sourced from Primer Bank.

Name		Sequence
Bacteria	7F1540R	5’-CAGAGTTTGATCCTGGCT-3’
	16S rDNA	5’-AGGAGGTGATCCAGCCGCA-3’
Bacteria	27F1492R	5’-AGTTTGATCMTGGCTCAG-3’
	16S rDNA	5’-GGTTACCTTGTTACGACTT-3’
Fungus	ITS1	5’-TCCGTAGGTGAACCTGCGG-3’
	ITS4	5’-TCCTCCGCTTATTGATATGC-3’
Fungus	NS1	5’-GTAGTCATATGCTTGTCTC-3’
	NS6	5’-GCATCACAGACCTGTTATTGCCTC-3’
PCNA		Forward: 5’-CCTGCTGGGATATTAGCTCCA-3’
		Reverse: 5’-CAGCGGTAGGTGTCGAAGC-3’
Rac		Forward: 5’-AAGGAGCCCCACGAGAAAAAT-3’
		Reverse: 5’-ACCGAACTTGCATTGATTCCAG-3’
Rho		Forward: 5’-CTCGCTCAGTGCGAAGACAA-3’
		Downstream: 5’-CATTCTCTGACGACATTTTCCCT-3’
Elongin C		Forward: 5’-AACCCGGACCCTAAGAAGCTA-3’
		Reverse: 5’-TCTCCGCAAGAATGTCTACTGTA-3’
MSH6		Forward: 5’-CCAAGGCGAAGAACCTCAAC-3’
		Reverse: 5’-ACCAGGGGTAACCCTCCATC-3’
TRK		Forward: 5’-ACCCCGATTCCCCTAATAGAGG-3’
		Reverse: 5’-ACCAGAACTCGGTAATCTGACT-3’
GAPDH		Forward: 5’-TCGTGGATCTGACGTGCCGCCTG-3’
		Reverse: 5’-CACCACCCTGTTGCTGTAGCCGTAT-3’

PCNA: Proliferating cell nuclear antigen; MSH6: Melanocyte-stulating hormone-6; TRK: Tropomyosin receptor kinase; GAPDH: Glyceraldehyde-3-phosphate dehydrogenase.

Total RNA isolation and qPCR detection

Total RNA was extracted using TRIzol reagent (Invitrogen, CA, United States) and treated with the DNA-free DNase treatment & removal I Kit (Thermo Fisher Scientific, Delaware, United States) to eliminate genomic DNA. For reverse transcription, 500 ng of RNA was used with a SuperScript II Reverse Transcriptase system (Invitrogen, CA, United States). cDNA (25 ng) was quantified by qPCR for each gene of interest using the SYBR green reagent (Bio-Rad, CA, United States). qPCR was performed on a Bio-Rad CFX96 platform under the following conditions: Initial denaturation at 95 °C for 30 seconds, followed by up to 40 cycles of denaturation at 95 °C for 5 seconds, annealing at 60 °C for 30 seconds, and elongation at 72 °C for 30 seconds. Fluorescence was detected after each cycle. Each sample was analyzed in triplicate. Relative mRNA expression for each gene was calculated using the 2^-ΔΔCt method (cycle threshold) and normalized to glyceraldehyde-3-phosphate dehydrogenase. The primer sequences are listed in Table 2.

Western blot analysis

For total protein extraction from cells, the lysis buffer was thawed and thoroughly mixed. Phenylmethylsulfonyl fluoride (100 mmol/L) was added to the lysis buffer at a ratio of 10 μL per 1 mL to achieve a final concentration of 1 mmol/L. Cells were washed with cold phosphate-buffered saline, lysed in the buffer on ice for 30 minutes, and centrifuged at 12000 × g for 15 minutes at 4 °C to collect the supernatant. For protein quantification, about 20 μL of 5 mg/mL bovine serum albumin protein standard solution was diluted with phosphate-buffered saline to a final volume of 100 μL, yielding a final concentration of 1.0 mg/mL. Equal amounts of protein samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred onto PVDF membranes. The membranes were blocked and then probed with a rabbit anti-mouse PCNA primary antibody (1:800, Abcam, United States) diluted in blocking buffer (5% non-fat milk powder, 0.1% Tween 20 in Tris-buffered saline) overnight at 4 °C. Proteins were detected using enhanced chemiluminescence. The integrated intensity of the protein bands was quantified by scanning densitometry and analyzed using Glyko BandScan 5.0 (Glyko, Novato, CA, United States). Data were normalized to glyceraldehyde-3-phosphate dehydrogenase as a loading control.

siRNA transfection

Cells were seeded in a 24-well plate one day prior to transfection to achieve approximately 30% confluence at the time of transfection. siRNA sequences are presented in Table 3. The Celopener reagent and RNA diluent were mixed thoroughly (using an oscillator or by pipetting at least 10 times) and allowed to stand at room temperature for 15 minutes. About 50 μL of the transfection complex was added to wells containing 450 μL of complete culture medium [high-glucose DMEM (4.5 g/L D-glucose) supplemented with 10% fetal bovine serum] or to the high-glucose culture medium supplemented with Bolteae supernatant (at a 1:10 ratio based on dose-response and time-course results of PCNA expression). The plate was gently agitated to ensure complete mixing. Cells were cultured at 37 °C in a 5% CO₂ incubator for 18-48 hours, and the culture medium was replaced 6-8 hours after transfection. siRNA silencing effects were assessed 24-72 hours post-transfection.

Table 3 siRNA primer sequences sourced from Primer Bank.

Name	Sequence
Rac-si	Forward: 5’-GCUACUUCCUCCUCAAGAAdTdT-3’
	Reverse: 5’-CAGCGGTAGGTGTCGAAGC-3’
Rho-si	Forward: 5’-CCUACAUGUUUCUGCUGAUTT-3’
	Reverse: 5’-AUCAGCAGAAACAUGUAGGTT-3’
hMSH6-si	Forward: 5’-GGCCUUGUCUGGUUUACAATT-3’
	Reverse: 5’-UUGUAAACCAGACAAGGCCTT-3’
Elongin C-si	Forward: 5’-GCACCUGAAAUUGCACUGGAAdTdT-3’
	Reverse: 5’-UUCCAGUGCAAUUUCAGGUGCdTdT-3’
GAPDH	Forward: 5’-TCGTGGATCTGACGTGCCGCCTG-3’
	Reverse: 5’-CACCACCCTGTTGCTGTAGCCGTAT-3’

hMSH6: Melanocyte-stulating hormone-6; GAPDH: Glyceraldehyde-3-phosphate dehydrogenase.

Statistical analysis

All the experiments were performed with a minimum of three independent biological replicates. Depending on the data distribution and experimental design, the data were analyzed using Student’s t-test, one-way analysis of variance, Mann-Whitney U test, Fisher’s exact test, or Spearman’s correlation test using GraphPad Prism Software (version 5). Bioinformatics evaluations (principal component analysis, NMDS, and principal coordinate analysis) were implemented on the OmicShare platform, and comprehensive data visualization was done using R software (version 4.3.3). Differences were considered statistically significant at P < 0.05.

RESULTS

Significant differences in gene distribution among different groups

Cluster analysis based on gene number and relative abundance was performed to determine the distribution of gene counts and to identify shared and unique genes. The results were visualized using a degree distribution plot (Figure 1A), unigene annotation counts (Figure 1B), a relative abundance cluster heatmap (Figure 1C), Venn diagrams (Figure 1D-F), and box plots (Figure 1G). The analysis revealed that all groups contained both shared and unique genes, and the gene distribution among different groups was significantly different.

Figure 1 Common and unique information of genes among different samples (groups). A: Degree distribution analysis of the gene catalog. The left vertical axis (frequency) represents the number of genes in the gene catalog; the right vertical axis (percentage, %) represents the percentage of genes in the gene catalog; the horizontal axis represents the degree of genes in the gene catalog; B: Heatmap of unigene annotation counts. The horizontal axis represents sample names; the vertical axis represents species information; different colors represent different numbers of unigenes; C: Horizontal relative abundance clustering heatmap. The horizontal axis represents sample information; the vertical axis represents species information. The clustering tree on the left is a species clustering tree. The values in the central heatmap correspond to Z-scores obtained by standardizing the relative abundance of each species per row. In this Z-score normalized heatmap, positive values (typically warmer colors such as red) indicate that the abundance/expression level of a given gene/species in a particular sample is higher than its mean level across all samples; higher magnitudes (e.g., +2 and +3) reflect greater deviation above the average. Negative values (typically cooler colors such as blue) indicate that abundance/expression levels below the overall mean; smaller values (e.g., -2 and -3) eflect greater deviation below the average. Values close to zero (often white) indicate that abundance/expression levels near the mean across all samples; D-F: Venn diagrams showing shared and unique genes across individual samples. Each circle represents one sample, with colors indicating group affiliation: Patients with diabetes mellitus (n = 8), patients with diabetes mellitus and colorectal cancer (n = 11), patients with colorectal cancer without diabetes (colorectal cancer, n = 10), and healthy controls (control, n = 11). Numbers in the overlapping regions denote the genes shared among samples, whereas numbers in the non-overlapping regions represent the unique genes identified in each sample; G: Box plot comparing gene counts between groups. The horizontal axis represents each group; the vertical axis represents the number of genes. Statistical differences in the number of genes among the four groups were evaluated using the Kruskal-Wallis H test. DM: Diabetes mellitus; CRC: Colorectal cancer; L.CRC: Left-sided colorectal cancer; R.CRC: Right-sided colorectal cancer.

Altered gut bacteriome biodiversity and composition in DM-CRC

Metagenomic sequencing and bioinformatic analyses revealed that the gut microbiota differed significantly among the DM-CRC, CRC, DM, and healthy control groups (P < 0.05). The difference between the DM-CRC and DM groups was also significant (P < 0.05; Figure 2). At the phylum, genus, and species levels, Eisenbergiella tayi, E. bolteae, Fusobacterium ulcerans, Faecalibacterium praussnitzii, and Akkermansia were significantly more abundant in the DM-CRC group than in the DM group. Among these, E. bolteae was the most abundant (Figure 3).

Figure 2 Dimensionality reduction analysis based on species abundance. Each point in the figure represents a sample, and the distance between points represents the degree of difference. Samples from the same group are color-coded accordingly. The results show significant differences among the groups. A: Non-metric multidimensional scaling (NMDS) plot at the class level. A stress value less than 0.2 indicates that the NMDS analysis has reasonable reliability; B: Principal component analysis plot at the class level. The horizontal axis represents the first principal component, and the vertical axis represents the second principal component; percentages indicate the contribution of each component to sample variance; C: Principal coordinate analysis plot based on Bray-Curtis distance at the class level; D: NMDS plot at the species level. A stress value less than 0.2 indicates that the NMDS analysis has reasonable reliability; E: Principal component analysis plot at the species level. The horizontal axis represents the first principal component, and the vertical axis represents the second principal component; percentages indicate the contribution of each component to sample variance; F: Principal coordinate analysis plot based on Bray-Curtis distance at the species level; G: Bray-Curtis distance clustering tree (left) and relative abundance distribution of functions at the class level for each sample (right); H: Relative abundance distribution at the species level for each sample. DM: Diabetes mellitus; CRC: Colorectal cancer; NMDS: Non-metric multidimensional scaling; PCA: Principal component analysis; PCoA: Principal coordinate analysis.

Figure 3 Microbial community differences across groups. A: Relative abundance clustering heatmap. The clustering tree on the left represents species-level clustering. The values in the central heatmap correspond to Z-scores obtained by standardizing the relative abundance of each species. Linear discriminant analysis effect size analysis of species with intergroup differences comprises two parts; B: Linear discriminant analysis score bar chart. The horizontal axis represents group information (diabetes mellitus group, n = 8), diabetes mellitus-colorectal cancer group (n = 11), colorectal cancer group (n = 10), and control group (n = 11). The vertical axis represents species information; C: Evolutionary cladogram. Circles radiating from the inside outward represent taxonomic levels (e.g., genus or species). Each small circle at a given taxonomic level represents a specific taxon, and the diameter of the small circle is positively proportional to the relative abundance. Species with no significant differences are considered as a single group, and biomarkers of different species follow their respective group assignments. Orange nodes indicate microbial groups that play an important role in the orange module, while green nodes represent those important in the green module. The species corresponding to each letter in the cladogram are listed in the legend on the right; D: Krona chart visualizing species annotation results. Circles represent different taxonomic levels (phyla, families, genera, and species) from inside to outside. The size of each fan represents the relative proportion of the corresponding species. DM: Diabetes mellitus; CRC: Colorectal cancer; LDA: Linear discriminant analysis.

Bolteae elicites a promoting effect on colon cancer cell proliferation in DM-CRC

When the Bolteae supernatant was mixed with complete culture medium at a ratio of 1:10 and incubated for 24 hours, the expression levels of the proliferation indicator PCNA increased more significantly than those observed at other concentrations and time points (P < 0.05, Figure 4A and B). Under the same condition (1:10 dilution, 24-hour incubation), cell counts peaked, showing significant differences compared to other concentrations and time points (P < 0.05, Figure 4C-F, Supplementary Tables 1-8). Cell viability was assessed by counting live cells and performing the Cell Counting Kit-8 assay (Figure 4E-G).

Figure 4 Effects of Bolteae supernatant on proliferating cell nuclear antigen expression and proliferation of colorectal cancer cells. A and B: Quantitative polymerase chain reaction analysis of the expression level of proliferating cell nuclear antigen (PCNA) in colorectal cancer cells. There was a significant difference between the intervention group and the non-intervention group, ^aP < 0.05, n = 6. The vertical axis represents the relative abundance of PCNA in RKO (Figure 4A) and LoVo cells (Figure 4B), while the horizontal axis represents different times and intervention concentrations; C and D: Direct cell counting results. RKO colorectal cancer cells (Figure 4C) and LoVo colorectal cancer cells (Figure 4D) were counted at various time points following treatment with different concentrations of Bolteae supernatant; E and F: Cell proliferation measured using the Cell Counting Kit-8 assay. After transfection, absorbance was measured using a fluorescence spectrophotometer. Significant differences (P < 0.05) were observed compared to untreated and other intervention conditions. Spectroscopic measurements of RKO colon cancer cells (Figure 4E) and LoVo colon cancer cells (Figure 4F) at different time points after intervention with various concentrations of Bolteae supernatant are shown. Cells treated with a 1:10 ratio of the supernatant demonstrated the most significant proliferation within 24 hours; G: Representative images taken by an image-based cell counting instrument. Viable cells are distinguished by their larger and more uniform volume, while debris appears smaller and irregularly shaped. Viable cells remain unstained (bright), whereas non-viable cells are stained (dark/blue). The software quantifies the identified viable and non-viable cells and, combined with the known sample volume, automatically calculates the concentration of live cells (Figure 4C and D), accompanied by the display of the captured image. Data were analyzed using analysis of variance.

Functional analysis demonstrates differences between the groups

Kyoto Encyclopedia of Genes and Genomes (KEGG)-based functional analysis revealed functional differences between groups (Figure 5). The number of annotated unigenes varied across databases and level 1 functional categories (Figure 5A). In contrast, clustering by functional abundance separated the samples according to their functional profiles, as shown in the Bray-Curtis dendrogram and the relative abundance distribution (Figure 5B). Figure 5C and D demonstrates the associations between all experimental groups and human disease-related pathways. Figure 5E-G shows the functional divergence plots derived from the three distinct dimensionality reduction analyses at level 1. In these plots, increased interpoint distances reflect greater heterogeneity between samples, with geometric shapes representing distinct cohorts. The results consistently revealed functional disparities between the groups. The NMDS analysis yielded a stress value of 0.061, indicating a reliable ordination (according to Kruskal’s criterion, a stress value of < 0.1 suggests a good model fit).

Figure 5 Functional analysis based on the Kyoto Encyclopedia of Genes and Genomes database. A: Statistical chart of the number of annotated genes in each database. The numbers on the bar chart represent the number of unigenes successfully annotated. The other axis represents the level 1 functional categories within each database; B: Sample clustering analysis based on functional abundance. The left side of the figure displays the clustering tree structure based on Bray-Curtis distance. The right side shows the relative abundance distribution of functions at level 1 for each sample; C: Functional abundance clustering heatmap based on the top 5 most abundant functions per sample, using the functional annotations and abundance data from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. The horizontal axis represents sample information, and the vertical axis provides functional annotation information. The clustering tree on the left represents functional clustering. The values in the central heatmap correspond to the Z-scores obtained by normalizing the relative abundance of functions in each row; D: Top 10 KEGG level 1 functional ranked by relative abundance across samples. The Y-axis indicates the proportion of annotations within each functional category, and the X-axis shows the sample names. Relative abundance was calculated at multiple levels across databases. This figure shows the relative abundance distribution at KEGG level 1 as an example; E-G: Dimensionality reduction analysis based on functional abundance in the KEGG database. The more similar the functional composition between samples, the closer their distances in the dimensionality reduction plot. When the stress value of NMDS analysis at level 1 is less than 0.2, the non-metric multidimensional scaling analysis is considered to have a certain degree of reliability (Figure 5E). Principal component analysis at level 1 is shown, where the horizontal axis represents the first principal component, the vertical axis represents the second principal component, and the percentages represent the contribution of each component to sample differences (Figure 5F). Dimensionality reduction analysis based on Bray-Curtis distance derived from species abundance was also performed, as illustrated by principal coordinate analysis at level 1 (Figure 5G). DM: Diabetes mellitus; CRC: Colorectal cancer; PCA: Principal component analysis; NMDS: Non-metric multidimensional scaling; PCoA: Principal coordinate analysis.

Bolteae may promote tumor cell proliferative activity in CRC via targeting the RAC/RHO, MSH6, TRK, and EloginC pathway

We analyzed metabolic pathway discrepancies between the DM and DM-CRC groups using KEGG database mapping to construct a carcinogenesis-associated metabolic pathway map, in which enzymatic reactions unique to the DM-CRC group are highlighted in blue. The results identified Rac/Rho, MSH6, TRK, and Elongin C (ELOC/TCEB1) as DM-CRC-specific enzymatic markers (Figure 6).

Figure 6 Comparison of cancer-related metabolic pathways between the diabetes mellitus and diabetes mellitus-colorectal cancer groups based on Kyoto Encyclopedia of Genes and Genomes database analysis. Differences in metabolic pathway maps between the diabetes mellitus (DM) and DM-colorectal cancer groups were examined using the Kyoto Encyclopedia of Genes and Genomes database, and a metabolic pathway webpage map was generated. The figure highlights cancer-related metabolic pathways within the Kyoto Encyclopedia of Genes and Genomes pathway overview map, showing both shared and unique metabolic pathway information between the two groups. In the diagram, nodes represent chemical compounds, and edges represent enzyme-catalyzed reactions. Red indicates reactions common to both groups, while blue denotes reactions unique to the DM-colorectal cancer group. A: Pathways of colorectal cancer; B: Pathways in cancer.

Furthermore, preliminary qPCR validation demonstrated significant Rac, Rho, MSH6, and Elongin C upregulation in Bolteae-treated colon cancer cells (exposed to a 1:10 ratio of bacterial supernatant to high-glucose DMEM for 24 hours, as shown in Figure 4) compared to the controls (P < 0.05; Figure 7). No statistically significant changes in TRK expression were observed under the same conditions (P > 0.05; Figure 7).

Figure 7 Effects of Enterocloster bolteae treatment on gene expression in LoVo and RKO colorectal cancer cells. A P value < 0.05 indicates statistically significant differences. All data were normalized to the glyceraldehyde-3-phosphate dehydrogenase endogenous control. Error bars represent standard error of the mean from triplicate experiments. Statistical analysis was performed using one-way analysis of variance with Dunnett’s post-test. Colon cancer cells were treated with Enterocloster bolteae supernatant at a 1:10 ratio (bacterial supernatant to high-glucose Dulbecco’s modified Eagle’s medium) for 24 hours. Compared with controls, relative mRNA expression levels of different targets in LoVo and RKO colorectal carcinoma cell lines were assessed by quantitative real-time polymerase chain reaction. A: 16S rDNA sequence from the Ribosomal Database Project (http://rdp.cme.msu.edu/index.jsp). The comparison results indicate that the strain is Enterocloster bolteae (formerly Clostridium bolteae); B: DNA target band obtained by excising the electrophoresis band of polymerase chain reaction products; C and D: Relative mRNA expression levels of Rho, Rac, hMSH6, Elongin C (ELOC/TCEB1), and TRK in LoVo (Figure 7C) and RKO (Figure 7D) colorectal carcinoma cell lines, as assessed by quantitative real-time polymerase chain reaction.

We used siRNA to transfect two sets of colon cancer cells: One cultured in high-glucose medium and the other in high-glucose medium supplemented with E. bolteae supernatant (at a 10:1 ratio). After confirmation of successful gene silencing (Figure 8A and B, P < 0.05), we evaluated the proliferation marker PCNA in each target-silenced cell line. The results showed that in Rac/Rho-, MSH6-, and Elongin C-silenced colon cancer cells, proliferation did not further increase even with the addition of E. Boltea supernatant to the high-glucose medium (Figure 8C and D, P < 0.05).

Figure 8 Silencing of Rac, Rho, hMSH6, and Elongin C via siRNA and assessment of the silencing effect on the proliferation marker proliferating cell nuclear antigen in RKO and LoVo colorectal cancer cells. A: Quantitative real-time polymerase chain reaction quantitative analysis of Rac, Rho, hMSH6, and Elongin C expression levels in RKO cells. The vertical axis represents the relative abundance of these transcripts in colorectal cancer cells, and the horizontal axis represents different intervention conditions (no Bolteae supernatant, or a 1:10 ratio of Bolteae supernatant to culture medium); B: Quantitative real-time polymerase chain reaction quantitative analysis of Rac, Rho, hMSH6, and Elongin C expression levels in LoVo cells; C: Western blot detection of the proliferation marker proliferating cell nuclear antigen (PCNA) in RKO cells transfected with siRNA. The upper panel shows the immunoblot bands, and the lower panel presents the corresponding densitometric quantification. The vertical axis represents the integrated intensity of PCNA, while the horizontal axis indicates different intervention conditions (no Bolteae supernatant, or a 1:10 ratio of Bolteae supernatant to culture medium); D: Western blot detection of the proliferation marker PCNA in LoVo cells transfected with siRNA. PCNA: Proliferating cell nuclear antigen.

This preliminary evidence suggested that Bolteae promotes CRC cell proliferation through a functionally coordinated regulatory axis involving Rac/Rho, MSH6, and Elongin C.

DISCUSSION

This study analyzed the gene distribution of human gut microbiota across predefined groups including patients with DM-CRC, DM without CRC, CRC without DM, and healthy controls. Venn diagram analysis revealed that although the DM-CRC group shared a proportion of microbial genes with the other cohorts, it also possessed a distinct set of gut microbiota-related genes that did not overlap with any other group, indicating a unique intestinal microecological profile in patients with DM-CRC. This is consistent with the observation that species composition and dimensionality reduction analyses demonstrated clear group-specific microbial patterns. The DM-CRC group showed an increased abundance of E. bolteae, Fusobacterium ulcerans, and Faecalibacterium prausnitzii, whereas the DM group was dominated by Firmicutes, and Enterococcus avium was the predominant species in the CRC group. Notably, the dominant taxa differed markedly among groups, suggesting distinct functional microbial signatures associated with each clinical condition. Among all enriched species, E. bolteae exhibited the highest relative abundance specifically in the DM-CRC group while remaining scarce in the other cohorts, strongly implicating this bacterium as a potential contributor to tumor promotion in the diabetic intestinal microenvironment.

In this study, we found that exposure to E. bolteae supernatant significantly enhanced CRC cell proliferation under hyperglycemic conditions, highlighting the pivotal role for E. bolteae in promoting abnormal epithelial growth in the diabetic colon. Together with the metagenomic findings, these results substantiated the idea that E. bolteae facilitates the progression of diabetes-associated CRC. E. bolteae is a recognized pathogenic anaerobe that ferments carbohydrates to produce acetic and lactic acids as major metabolic end products[18]. Phylogenetically, it belongs to the phylum Clostridium and family Bolognataceae, and genomic analyses have revealed functional divergence from Clostridium difficile, particularly in the butyrate synthesis pathway and flagellar system[19]. Previous studies have reported increased abundance of E. bolteae in patients with gastrointestinal disorders, autism spectrum conditions, multiple sclerosis, neuromyelitis optica spectrum disorders, and primary Sjögren’s syndrome, where it has been associated with inflammatory gene expression, immune dysregulation, plasma cell differentiation, B-cell chemotaxis, and Th17 activation[20-23]. Despite the established involvement in immune and inflammatory diseases, the role of E. bolteae in CRC has remained largely unexplored. Our findings provided the first experimental evidence that E. bolteae acts as a protumorigenic bacterium in the diabetic intestinal microenvironment, directly promoting CRC cell proliferation and linking diabetic gut dysbiosis to colorectal carcinogenesis. Since our study provides this evidence in vitro, we acknowledge the absence of in vivo validation as a key limitation of the current work. The use of germ-free animal models or colonization studies in carcinogen-induced mouse models is essential to establish the causal role of E. bolteae in tumorigenesis. Therefore, our findings should be interpreted as hypothesis-generating, and further animal studies are warranted to confirm the protumorigenic potential of this bacterium in a complex physiological environment.

KEGG pathway analysis revealed that the gut microbiota of patients with DM-CRC exhibited a distinct functional profile compared with healthy controls, and patients with either CRC or DM alone. Notably, comparisons with the DM without CRC cohort identified several DM-CRC-specific microbial targets including RAC, Rho, Elongin C (ELOC/TCEB1), hMSH6, and TRK. RAC and Rho are well-established regulators of actin cytoskeleton dynamics, cell adhesion, and motility, playing critical roles in tumor proliferation and metastatic potential[24]. Elongin C, previously linked to von Hippel-Lindau disease through ELOC/TCEB1 mutations[25], has not yet been systematically explored in colorectal carcinogenesis, rendering its involvement a novel and unexplored avenue. hMSH6, a core DNA mismatch repair protein, is essential for genomic stability, and its dysfunction is closely associated with microsatellite instability and hereditary nonpolyposis CRC[26]. TRK receptors encoded by NTRK genes, are recognized oncogenic drivers that regulate tumor cell proliferation, differentiation, and survival through the RAS/MAPK, PI3K/AKT, and PLCγ pathway, and are established therapeutic targets in precision oncology[27]. However, in the present study, TRK expression was not significantly altered under hyperglycemic conditions, suggesting that its contribution to DM-CRC may be indirect or context-dependent. First, it is important to note that KEGG enrichment analysis is based on transcriptomic profiling and reflects the overall pathway-level activity rather than expression of a single gene. A pathway can be enriched owing to coordinated changes in multiple genes within that pathway, even if a particular gene (such as TRK) does not show significant differential expression at the mRNA level. In our dataset, the TRK pathway may have been activated by other members (e.g., neurotrophin receptors or downstream signaling components) that were not individually assessed by qPCR. Second, as hypothesized in the original discussion, the involvement of TRK signaling could be indirect or context-dependent; for instance, post-translational modifications or protein-level changes may occur without corresponding mRNA alterations. Alternatively, this discrepancy may stem from differences in sensitivity or experimental conditions between the RNA-seq and qPCR assays (e.g., different sample subsets or normalization methods). The current qPCR data do not rule out a potential role for TRK signaling, this possibility warrants further investigation.

While our work provides the first evidence linking E. bolteae to a signaling pathway, the identification of active bacterial components (e.g., through fractionation, metabolomics, or proteomics) is a critical next step. Direct bacterial contact cannot be ruled out and should be investigated using coculture systems with live bacteria. We believe that our findings provide a foundation for detailed mechanistic studies.

We acknowledge that the modest sample size is a key limitation of this study. The small cohort increases the risk of type II errors and may limit the generalizability of our findings. Furthermore, due to the strict exclusion criteria applied (e.g., exclusion of hypertension and viral hepatitis) to reduce confounding factors, our study population may not fully represent the general population of individuals with diabetes and multiple comorbidities. Therefore, the generalizability of our findings to unselected patients with diabetes requires further validation in a more diverse, large-scale, multicenter cohort.

Furthermore, the clinical data shown in Table 1 revealed no significant differences among the four groups in terms of high-density lipoprotein cholesterol, fasting plasma glucose, antidiabetic drug use, and family history of polyps. Despite these intergroup differences existed, their potential impact on the present study was considered negligible. It is well established that patients with DM exhibit heightened susceptibility to CRC, and substantial disparities in gut microbiota composition between DM and DM-CRC populations have been extensively documented[8,9,15,16]. The present study was specifically designed to characterize the differences in intestinal flora between the DM-CRC and DM groups. Reexamination of the baseline data presented in Table 1 showed no statistically significant intergroup differences in high-density lipoprotein cholesterol, fasting plasma glucose, antidiabetic medication history, or family history of polyps (all P > 0.05). These findings effectively excluded these variables as potential confounders in the analysis of CRC-associated dominant microbiota in the diabetic population. Although previous studies have suggested associations between factors such as body mass index, hypertension, smoking, and alcohol consumption, and CRC risk, these are currently regarded as correlative rather than definitive etiological factors[28-30]. In our cohort, none of these factors were significantly enriched between the two groups. Furthermore, both the univariate and Spearman’s correlation analyses confirmed the absence of potential confounding effects of these variables. The results of our additional correlation analyses, as presented in Table 4, indicated that there were no significant differences between the DM and DM-CRC groups in terms of fasting plasma glucose, antidiabetic drug use, or family history of polyps (P > 0.05). However, triglyceride levels emerged as a notable confounding factor, showing a significant difference between the two groups (P < 0.05, Table 4).

Table 4 Spearman correlation analysis between the diabetes mellitus-colorectal cancer group and the diabetes mellitus group.

Variable	r	95%CI	P value
Age	0.06827	-0.4104 to 0.5175	0.7812
Male	0.02548	-0.4454 to 0.4854	0.9175
Smoking	-0.04495	-0.5001 to 0.4297	0.8550
Drinking	0.02548	-0.4454 to 0.4854	0.9175
CRC history	0.02548	-0.4454 to 0.4854	0.9175
Hypertension	-0.2955	-0.6690 to 0.1973	0.2194
C polyps history	-0.1086	-0.5466 to 0.3760	0.6580
Antidiabetic drugs	-0.2676	-0.6520 to 0.2263	0.2681
BMI	-0.2776	-0.6582 to 0.2159	0.2498
FPG	0.08770	-0.3940 to 0.5316	0.7211
TG	-0.6334	-0.8487 to -0.2379	0.0036
TCH	-0.1852	-0.5992 to 0.3069	0.4478
HDL	-0.1179	-0.5532 to 0.3679	0.6307
LDL	-0.1951	-0.6057 to 0.2976	0.4236
Abundance of E. bolteae	0.8563	0.6498-0.9451	< 0.0001

CI: Confidence interval; CRC: Colorectal cancer; BMI: Body mass index; FPG: Fasting plasma glucose; TG: Triglyceride; TCH: Total cholesterol; HDL: High-density lipoprotein cholesterol; LDL: Low-density lipoprotein cholesterol.

Current evidence does not provide information on the direct correlation between triglycerides and the gut microbiota in CRC. Evidence suggests that the “triglyceride-glucose index” is a reproducible routine laboratory test that may serve as a potential biomarker for CRC[31]. However, these findings do not elaborate on the specific relationship between triglycerides or the triglyceride-glucose index and the gut microbiota. Therefore, based on the abundance profiles of the dominant bacteria observed in this study, we propose that although triglycerides levels demonstrate statistical significance, they are unlikely to be the primary drivers of the distinct microbial signature characteristics of the DM-CRC group. Future studies should conduct multivariate adjusted analyses to rule out interference from these factors. To control for this potential confounder, we performed a median stratification analysis based on triglyceride levels (median = 2.6 mmol/L). This analysis revealed that after stratification, there was no significant difference in triglyceride levels between the DM and DM-CRC groups in either the subgroup with triglycerides ≥ 2.6 mmol/L or the subgroup with triglycerides < 2.6 mmol/L (P > 0.05). Importantly, the abundance of E. bolteae remained significantly associated with the DM-CRC group (P < 0.05) in both strata (Table 5). Nevertheless, the generalizability of our conclusions is limited, highlighting the need for validation in large-scale multicenter cohorts along with multivariate adjusted analyses to control for potential confounders.

Table 5 Spearman correlation analysis between the diabetes mellitus group and the diabetes mellitus-colorectal cancer group after stratification based on median triglyceride level.

Characteristic		r	95%CI	P value
TG ≥ 2.6 mmol/L	TG	-0.3242	-0.8132 to 0.4332	0.3946
	Abundance of Bolteae	0.9376	0.7248-0.9871	0.0002
TG < 2.6 mmol/L	TG	-0.2871	-0.7764 to 0.4181	0.4213
	Abundance of Bolteae	0.9133	0.6676-0.9796	0.0002

TG: Triglyceride; CI: Confidence interval.

Collectively, the integration of KEGG metabolic mapping with functional validation indicated that E. bolteae is functionally involved in hyperglycemia-associated CRC remodeling through the RAC/Rho-MSH6-Elongin C signaling pathway, with TRK acting as a potential auxiliary component. Although these findings establish a functional link between diabetic gut dysbiosis and tumor proliferative signaling, they are preliminary and underscore the need for further studies to identify specific effector molecules and validate the direct interactions.

CONCLUSION

The gut microbiota of patients with DM-CRC demonstrated altered biodiversity and compositional shifts. E. bolteae might be a contributing factor highlighting the presence of diabetic colonic dysbiosis, as it promotes the aberrant proliferation of intestinal cancer cells under high-glucose conditions. Following interventions with E. bolteae, the accumulation of Rac/Rho, hMSH6, and Elongin C was observed in cancer cells, supporting this association. Moreover, siRNA-mediated knockdown of these factors markedly inhibited the proliferation of intestinal cancer cells, demonstrating that they are critically involved in the proliferative effects induced by E. bolteae.

This study identified E. bolteae as a potential driver of diabetes-associated CRC. Under hyperglycemic conditions, E. bolteae promoted CRC cell proliferation through the RAC/Rho-MSH6-Elongin C signaling pathway, providing a mechanistic link between diabetic gut dysbiosis and colorectal carcinogenesis and highlighting a potential target for microbiome-based risk stratification and therapeutic intervention. These preliminary findings require further in vivo studies to identify specific effector molecules and validate direct interactions.