Distinct gut microbiomes in Thai patients with colorectal polyps

Abstract

BACKGROUND

Colorectal polyps that develop via the conventional adenoma-carcinoma sequence [e.g., tubular adenoma (TA)] often progress to malignancy and are closely associated with changes in the composition of the gut microbiome. There is limited research concerning the microbial functions and gut microbiomes associated with colorectal polyps that arise through the serrated polyp pathway, such as hyperplastic polyps (HP). Exploration of microbiome alterations associated with HP and TA would improve the understanding of mechanisms by which specific microbes and their metabolic pathways contribute to colorectal carcinogenesis.

AIM

To investigate gut microbiome signatures, microbial associations, and microbial functions in HP and TA patients.

METHODS

Full-length 16S rRNA sequencing was used to characterize the gut microbiome in stool samples from control participants without polyps [control group (CT), n = 40], patients with HP (n = 52), and patients with TA (n = 60). Significant differences in gut microbiome composition and functional mechanisms were identified between the CT group and patients with HP or TA. Analytical techniques in this study included differential abundance analysis, co-occurrence network analysis, and differential pathway analysis.

RESULTS

Colorectal cancer (CRC)-associated bacteria, including Streptococcus gallolyticus (S. gallolyticus), Bacteroides fragilis, and Clostridium symbiosum, were identified as characteristic microbial species in TA patients. Mediterraneibacter gnavus, associated with dysbiosis and gastrointestinal diseases, was significantly differentially abundant in the HP and TA groups. Functional pathway analysis revealed that HP patients exhibited enrichment in the sulfur oxidation pathway exclusively, whereas TA patients showed dominance in pathways related to secondary metabolite biosynthesis (e.g., mevalonate); S. gallolyticus was a major contributor. Co-occurrence network and dynamic network analyses revealed co-occurrence of dysbiosis-associated bacteria in HP patients, whereas TA patients exhibited co-occurrence of CRC-associated bacteria. Furthermore, the co-occurrence of SCFA-producing bacteria was lower in TA patients than HP patients.

CONCLUSION

This study revealed distinct gut microbiome signatures associated with pathways of colorectal polyp development, providing insights concerning the roles of microbial species, functional pathways, and microbial interactions in colorectal carcinogenesis.

Key Words: Gut microbiome, Colorectal adenoma, Hyperplastic polyp, Full-length 16s rRNA, Microbial correlation networks, Predicted functional mechanisms

Core Tip: This study identified gut microbiome signatures and metabolic pathways associated with two types of colorectal polyps. It is the first report of enrichment in the sulfur oxidation pathway among patients with hyperplastic polyps (HP) and the involvement of Streptococcus gallolyticus in the secondary metabolite biosynthesis pathway among patients with tubular adenoma (TA). Additionally, analysis of microbial associations in the gut microbiomes of HP and TA patients revealed a decrease in the co-occurrence of short chain fatty acid-producing bacteria. Conversely, there was an increase in the co-occurrence of dysbiosis and colorectal cancer-associated bacteria.

INTRODUCTION

Colorectal cancer (CRC) is the third most prevalent cancer worldwide and the second leading cause of cancer-related mortality[1]. In Thailand, it is the fourth highest in terms of incidence and mortality rates; more than 20000 new cases have been reported in the 2020s[2]. The distinct molecular pathways leading to CRC are associated with various colorectal polyps. The most common pathway is the conventional adenoma-carcinoma sequence, which involves progression from tubular, tubulovillous, and villous adenomas. Another key pathway, the serrated polyp pathway, is characterized by the presence of colorectal polyps such as hyperplastic polyps (HP), sessile serrated adenoma/polyps (SSA/Ps), and traditional serrated adenoma. All the abovementioned colorectal polyps are regarded as neoplastic polyps with the potential to become malignant; they represent key phases in CRC progression[3].

The gut microbiome plays a pivotal role in connecting environmental factors to colorectal polyps because these factors are correlated with both compositional and functional changes within the collective microbial community residing in the colon. Alterations in the gut microbiome have been linked to colorectal polyps, and specific microbial species have been identified as potential drivers of oncogenesis. Microbial species commonly identified in precancerous lesions and implicated in colorectal carcinogenesis include Escherichia coli (E. coli)[4,5], Bacteroides fragilis (B. fragilis)[6], and Streptococcus gallolyticus[7]. Additionally, shifts in Fusobacterium mortiferum and pro-inflammatory bacteria (e.g., Bilophilia and Desulfovibrio), along with decreases in short chain fatty acid (SCFA)-producing bacteria such as Faecalibacterium prausnitzii and Bifidobacterium pseudocatenulatum, have been observed in adenoma patients[8,9]. Moreover, functional studies have elucidated the roles of specific microbes in colorectal polyp formation through processes such as co-metabolic dysfunction, inflammation, epigenetic alterations, and DNA damage[10,11]. Because the gut microbiome modulates the host metabolic environment, it can directly or indirectly influence mutagenesis rates, thereby influencing carcinogenesis. Importantly, gut microbiome differences have been discovered between healthy individuals and patients with serrated polyps[12,13]. It has been hypothesized that the unique microbiome alterations associated with early adenomas and premalignant colorectal polyps can serve as biomarkers for early cancer detection or the identification of individuals with a risk of colorectal polyps. Furthermore, an exploration of the microbiome alterations specific to premalignant or benign polyps would provide insights regarding the mechanisms by which specific microbes and their metabolic pathways contribute to colorectal carcinogenesis. However, there has been limited research concerning microbial alterations in colorectal polyps; substantial discrepancies in microbial markers across studies may be attributed to diverse biological factors that impact gut microbiome composition, as well as inconsistencies in microbial sequencing data processing[14,15].

In this study, we compared gut microbiome signatures between two types of colorectal polyps: Tubular adenoma (TA, high potential for malignancy) and HP (low potential for malignancy). Our results revealed distinct gut microbiome signatures associated with each pathway of colorectal polyp development. Additionally, we identified microbes and microbial functions significantly associated with TA and HP. These findings suggest that gut microbiome signatures can serve as early biomarkers of CRC risk and help to identify potential targets for cancer prevention strategies.

MATERIALS AND METHODS

Participant recruitment and criteria

Participants in this study were volunteers who took part in the CRC Screening Development with Multiple Technologies Project 2020s, established by Chulabhorn Royal Academy in Thailand. The inclusion criteria were as follows: (1) Individuals aged 50-80 years who underwent CRC screening by colonoscopy between February 1, 2021, and June 30, 2022; (2) individuals who had not taken antibiotics within the preceding 3 mo; (3) individuals who did not use proton pump inhibitors or enemas, reported no history of constipation, or had undergone colonoscopy within the previous 1 mo; and (4) individuals who cooperated with the screening program and were willing to provide written informed consent. The exclusion criteria were as follows: (1) Incomplete clinical data; (2) loss of follow-up; (3) presence of inflammatory polyps; or (4) no stool sample collection. Each participant was recruited to take part in the study at the initial screening visit. They were interviewed by a physician, who recorded their clinical data and health history. After endoscopic examination of the large intestine, any detected polyps were removed and examined by a pathologist. In total, 152 participants were categorized into three groups based on the histopathology findings of the detected polyps: Control (CT), HP, and TA group. The CT group consisted of participants who exhibited no polyps during colonoscopy, the HP group comprised participants who had HP, and the TA group encompassed participants with at least one tubular or tubulovillous adenoma. The study protocol was approved by the Ethics Committee of Chulabhorn Research Institute and the Institutional Review Boards of Chulabhorn Royal Academy (Project Code 045/2563). Informed consent was obtained from all participants involved in the study.

Sample collection and DNA extraction

Stool samples were collected using DNA/RNA Shield Fecal Collection Tubes (Zymo Research, Irvine, CA, United States) prior to routine bowel preparation and colonoscopy. Participants were given a fecal collection kit and instructions for home stool sampling. Collected stool samples were stored at -20 °C until analysis. Microbial DNA was extracted from stool samples using the DNeasy PowerSoil Pro Kit (Qiagen, Germantown, MD, United States) in accordance with the manufacturer’s protocol. The extracted DNA was quantified using a NanoDrop and the Qubit dsDNA HS Assay Kit (both from Thermo Fisher Scientific, Waltham, MA, United States). The ZymoBIOMICS Gut Microbiome Standard (Zymo Research) was utilized as the positive control for DNA extraction, full-length 16S rRNA amplification, and sequencing.

Full-length 16S library preparation and sequencing

Full-length 16S rRNA (V1-V9) was amplified using a set of barcoded primers: (27F) 5’-GCATC/barcode/AGRGTTYGATYMTGGCTCAG-3’ and (1492R) 5’-GCATC/barcode/RGYTACCTTGTTACGACTT-3’. The barcoded primer sequences were provided by PacBio. Each polymerase chain reaction (PCR) mixture consisted of 0.5 U Q5 High-Fidelity DNA polymerase (New England Biolabs, Ipswich, MA, United States), 200 µmol/L dNTPs, 0.5 µmol/L each forward and reverse primers, and 1 ng extracted DNA. PCR was performed using a thermocycler with the following protocol: 98 °C for 30 seconds; 22 cycles of 98 °C for 10 seconds, 55 °C for 30 seconds, and 72 °C for 1 minute; and 72 °C for 2 minutes. Library constructs for full-length 16S rRNA analysis were prepared using the SMRTbell Express Template Prep Kit 2.0 (PacBio, Menlo Park, CA, United States) and Sequel Binding Kit 3.0 in accordance with the manufacturer's protocol. Sequencing was performed using the PacBio Sequel I platform.

Microbiome data analysis

Gut microbiome analysis was conducted after a series of quality control steps using the FastQC[16] and MultiQC[17] pipelines. SMRT link v11.0 (PacBio) was utilized for demultiplexing and primer removal. Next, the QIIME2 (version 2023.2) pipeline[18] was used for microbiome profiling. Specifically, the q2-dada2 plugin (version 2023.2.0)[19] within QIIME2 was implemented for read denoising with the following criteria: pooling method, pseudo; minimum and maximum sequence lengths, 1000 and 1800 bases, respectively. Amplicon sequence variants (ASVs) were subjected to filtering, chimerism screening, and base correction. Taxonomic assignment was conducted using the Greengenes2 2022.10 database[20], which was trained with a naïve Bayes classifier. Common contaminants, unclassified ASVs, spike-ins, mitochondrial sequences, and chloroplast sequences were removed. Microbiome diversity and abundance were analyzed using the PhyloSeq (version 1.42.0)[21] and microbiome (version 1.20.0)[22] packages in R software[23]. Statistical analysis was conducted by the Wilcoxon and Kruskal-Wallis rank sum tests. Microbial richness and evenness were assessed via metrics such as the Shannon index[24], Simpson index[25], phylogenetic diversity (PD)[26], and Pielou index[27]. β-diversity was determined by Bray-Curtis distance-based principal coordinate analysis (PCoA)[28,29].

Gut microbiome signature discovery and co-occurrence network analysis

Gut microbiome signatures were identified using the linear discriminant analysis effect size (LEfSe) discovery tool[30,31]. Before data entry into LEfSe, normalization was performed by rarefaction of 13500 reads. Class comparisons were conducted using the Kruskal-Wallis test, whereas subclass comparisons were conducted with the Wilcoxon test. Both tests used a significance threshold of P < 0.05. Differential abundance analyses based on the negative binomial distribution were performed with the DESeq2 package (version 1.38.3)[32]. The abundances were imported into the DESeq2 package, and the Wald test was used to determine statistical significance[33]. Co-occurrence correlation analyses were carried out using FastSpar (version 0.0.10) software[34]. Correlation coefficients were averaged across five inference iterations, and P values were determined by 1000 bootstrap correlations. Correlation coefficients with P values less than 0.05 and absolute correlation values greater than 0.4 were selected for visualization in Cytoscape (version 3.10.0) software[35]. Comparisons among three networks (one network per group) were performed using the DyNet application[36] in Cytoscape. Microbial species with a rewiring metric or D_n-score of ≥ 2.0, as well as an edge count of ≥ 4, were regarded as rewired nodes for each dataset in comparisons between the CT and HP groups and between the CT and TA groups.

Functional pathway and enzyme commission enrichment analyses

To explore the metabolic functions and pathways of the gut microbiome in each group, PICRUSt2 software[37] was utilized to make predictions about microbial functions within metabolic pathways. The compositions of the identified microbes were aligned with the MetaCyc database[38] to obtain estimates of their metabolic functions. Statistical Analysis for Metagenomic Profiles (STAMP; version 2.1.3) software[39] was employed to detect variations in metabolic function abundance among groups using Welch's t-test[40] with a confidence interval of 0.95 and a significance threshold of P < 0.05. Gut microbiome contributions to specific pathways were determined by collecting pathway-associated enzyme commission (EC) numbers from MetaCyc and sorting on the basis of relative abundance.

RESULTS

Participant characteristics

In total, 440 participants were invited to participate in the CRC Screening Development with Multiple Technologies Project 2020s between February 1, 2021, and June 30, 2022; of these, 152 participants were enrolled in the experiment (Figure 1). Participants were categorized into three groups: 40 in the CT group, 52 in the HP group, and 60 in the TA group. Table 1 presents the demographic and clinical characteristics of the study participants. Participants in the HP and TA groups tended to be older than participants in the CT group. The number of women did not significantly differ among the three groups, but the CT group had the lowest number of men. Most participants in the HP group had polyps in the distal colon, whereas participants in the TA group had polyps in the distal and proximal colon.

Figure 1 Flowchart of participant enrollment. CRC: Colorectal cancer; CT: Control group; HP: Hyperplastic polyps; TA: Tubular adenoma.

Table 1 Participant demographic and clinical characteristics.

Group	CT, n = 40	HP, n = 52	TA, n = 60
Age in years, median (range)	56 (52-59)	60 (51-71)	61 (52-71)
Sex
Female	31 (77.5)	32 (61.5)	31 (51.7)
Male	9 (22.5)	20 (38.5)	29 (48.3)
BMI in kg/m2, median (range)	21.7 (18-26.56)	23.4 (18.6-38.3)	24 (19-32)
Polyp location
Proximal		6 (11.5)	21 (35)
Distal		39 (75)	22 (36.7)
Both		7 (13.5)	17 (28.3)

Data are n (%). BMI: Body mass index; CT: Control; HP: Hyperplastic polyps; TA: Tubular adenoma.

Microbial community investigation

Data preprocessing with the QIIME2 pipeline revealed an average read length of 1452.95. The minimum and maximum sequence lengths were 1363 and 1788 bases, respectively. Supplementary Figure 1 shows the sequence length statistics and alpha rarefaction curve. Source Data Supplementary Table 1 lists the numbers of reads during the denoising process.

Microbiome diversities were compared among study groups using the Wilcoxon rank sum test and Kruskal-Wallis rank sum test. In terms of α-diversity, there were no significant differences among the CT, HP, and TA groups according to the Wilcoxon rank sum test. Analyses using the Shannon index (P = 0.47), PD (P = 0.85), Simpson index (P = 0.42), and Pielou index (P = 0.72) indicated that α-diversity did not significantly differ between the HP and CT groups (Figure 2A). Furthermore, α-diversity did not significantly differ between the CT and TA groups [Shannon index (P = 0.38), PD (P = 0.75), Simpson index (P = 0.29), and Pielou index (P = 0.48); Figure 2B].

Figure 2 Microbial diversity of the gut microbiome in control, hyperplastic polyps, and tubular adenoma groups. A: Box plots show α-diversity between the control (CT) and hyperplastic polyps (HP) groups; B: Box plots show α-diversity between the CT and tubular adenoma (TA) groups; C: Principal coordinate analysis (PCoA) plots show β-diversity between the CT and HP groups; D: PCoA plots show β-diversity the CT and TA groups. PD: Phylogenetic diversity.

Bray-Curtis distance-based PCoA was performed to assess β-diversity among participants in the CT, HP, and TA groups. As shown in Figure 2C and D, there were no significant differences in gut microbiome composition between the CT and HP groups or the CT and TA groups. Thus, the CT, HP, and TA groups did not demonstrate significant differences in richness and evenness at the species level.

Groupwise comparative analyses of gut microbial species: CT vs HP and CT vs TA

To identify gut microbiome signatures that could distinguish the CT group from the HP and TA groups, we compared microbial species between the CT and HP groups, and between the CT and TA groups, using the LEfSe method and DESeq2. LEfSe, a powerful tool for the discovery of high-potential signatures, combines non-parametric standard tests for statistical significance with linear discriminant analysis (LDA). The LDA model within LEfSe identifies microbial species that are differentially abundant between groups, then estimates the effect size of each significantly different microbial species[31]. In contrast, DESeq2 utilizes a negative binomial generalized linear model to estimate log fold changes between two groups, then evaluates the significance of those changes using the Wald test[33].

LEfSe revealed that ten microbial species significantly differed between the CT and HP groups (Figure 3A, Source Data Supplementary Table 2). Among these, four microbial species were enriched in the HP group: Blautia A 141780 hansenii (B. hansenii), Ruminococcus C 58660 sp000433635, UBA9414 sp003458885, and Veillonella A atypica (V. atypica). Additionally, LEfSe revealed that 20 microbial species significantly differed between the CT and TA groups (Figure 3B, Source Data Supplementary Table 2). Seven microbial species were significantly enriched in the TA group, including two CRC-associated bacteria (Bacteroides H fragilis and Clostridium Q 134516 symbiosum), as well as Bacteroides nordii and Clostridium Q fessum (Figure 3B, Source Data Supplementary Table 2).

Figure 3 Linear discriminant analysis effect size and DESeq2 identified the most enriched microbial species in the control, hyperplastic polyps, and tubular adenoma groups. A: Linear discriminant analysis (LDA) effect size (LEfSe) showed significant microbial differences between the control (CT) and hyperplastic polyps (HP) groups; B: LEfSe showed significant microbial differences the CT and tubular adenoma (TA) groups (LDA score > 2); C: DESeq2 analysis graphs illustrate the log₂ fold differential abundances of microbial species between the CT and HP groups; D: DESeq2 analysis graphs illustrate the log₂ fold differential abundances of microbial species between the CT and TA groups. In the graphs, log₂ fold changes > 0 indicate an increase in the corresponding microbial species, whereas log₂ fold changes < 0 indicate a decrease. Microbial species positioned above the zero threshold demonstrated higher relative abundance in either the HP or TA group compared to the CT group.

According to DESeq2, 22 microbial species exhibited significantly different abundances in the HP group; these included dysbiosis and gastrointestinal diseases-associated bacteria, such as Mediterraneibacter gnavus (M. gnavus) and Fusobacterium varium (F. varium), as well as commensal and SCFA-producing bacteria (e.g., B. hansenii, Butyribacter sp001916135, Bifidobacterium catenulatum, and Faecalimonas umbilicata) (Figure 3C, Source Data Supplementary Table 2). Additionally, 20 microbial species showed significantly different abundances in the TA group (Figure 3D), including S. gallolyticus (a well-known CRC-associated species), M. gnavus, and F. varium. Additionally, V. atypica demonstrated significantly different abundances in the HP and TA groups. V. atypica is commonly localized in the oral cavity and has been identified in fecal samples from older patients with CRC[41,42]. Notably, the TA group had increased abundances of well-known CRC-associated bacteria. These findings suggest that increases in S. gallolyticus and B. fragilis contribute to TA development. The findings also support the notion that patients with colorectal polyps exhibit a dysregulated microbiome characterized by high abundances of potentially pathogenic bacteria. In summary, our results imply that the identified microbial species could be used as signatures for HP and TA. We also assessed the presence of pathogenic bacteria commonly associated with CRC, such as F. nucleatum and E. coli. In this study, F. nucleatum and E. coli were not detected in the taxonomic annotation (Source Data Supplementary Table 2).

Predicted functional signatures in HP and TA groups

To identify the mechanisms by which the gut microbiome influences CRC carcinogenesis and detect biologically significant differences, we examined changes in functional composition using the MetaCyc pathway database. Compared with the CT group, the HP and TA groups had 20 and eight enriched pathways, respectively (Figure 4). Enriched pathways in the HP group included cell structure biosynthesis [e.g., peptidoglycan biosynthesis I (meso-diaminopimelate containing)]; inorganic nutrient metabolism [super pathway of sulfur oxidation (Acidianus ambivalens)]; fatty acid and lipid biosynthesis (e.g., CDP-diacylglycerol biosynthesis I); cofactor, carrier, and vitamin biosynthesis; carboxylic acid biosynthesis; secondary metabolite biosynthesis; tetrapyrrole biosynthesis; and amino acid biosynthesis. Secondary metabolite biosynthesis; aromatic compound degradation; cofactor, carrier, and vitamin biosynthesis; and cell structure biosynthesis were enriched in the TA group. The main differential pathways in the TA group were related to secondary metabolite biosynthesis: Taxadiene biosynthesis (engineered), mevalonate pathway I (eukaryotes and bacteria), and the super pathway of geranylgeranyl diphosphate biosynthesis I (via mevalonate). Notably, mevalonate pathway I (eukaryotes and bacteria) overlapped between the HP and TA groups.

Figure 4 Functional differences in the gut microbiome among control, hyperplastic polyps, and tubular adenoma groups. A: PICRUSt2 demonstrated significant differences in the gut microbiome within MetaCyc pathways between the control (CT) and hyperplastic polyps (HP) groups; B: PICRUSt2 demonstrated significant differences in the gut microbiome within MetaCyc pathways between the CT and tubular adenoma (TA) groups.

Next, we investigated the contributions of microbial species to the enriched pathways in the HP and TA groups. The dominant reactions in the sulfur oxidation (Acidianus ambivalens) pathway are adenylyl-sulfate reductase (APS reductase (EC:1.8.99.2)) and the thiosulfate dehydrogenase (quinone) pathway (EC:1.8.5.2). In the HP group, 17 and 22 microbial species contributed to the APS reductase pathway (EC:1.8.99.2) and the thiosulfate dehydrogenase (quinone) pathway (EC:1.8.5.2), respectively. Parabacteroides B 862066 distasonis (P. distasonis), Bacteroides H thetaiotaomicron (B. thetaiotaomicron), and Bacteroides H cellulosilyticus were the main contributors to the thiosulfate dehydrogenase (quinone) pathway. Furthermore, the APS reductase pathway was associated with Bilophila wadsworthia (B. wadsworthia) and Desulfovibrio piger (D. piger), both sulfate-reducing bacteria (SRB) (Source Data Supplementary Table 3).

In the TA group, several enzymes in enriched pathways had large contributions from S. gallolyticus; these enzymes included hydroxymethylglutaryl-CoA synthase (EC:2.3.3.10), mevalonate kinase (EC:2.7.1.36), phosphomevalonate kinase (EC:2.7.4.2), diphosphomevalonate decarboxylase (EC:4.1.1.33), and isopentenyl-diphosphate delta-isomerase (EC:5.3.3.2) (Source Data Supplementary Table 4). These findings suggested that the development of TA could be influenced by functional pathways associated with distinct gut microbiome signatures.

Analysis of microbial correlation networks in CT, HP, and TA groups

Microbial correlation networks and hub species were explored via co-occurrence network analyses that examined patterns of microbial correlation in the two colorectal polyp types. Microbial association networks were constructed for CT, HP, and TA groups using FastSpar. Table 2 presents the network properties of the CT, HP, and TA groups. The CT network exhibited the largest size in terms of the numbers of nodes and edges, as well as network diameter. These findings indicated that the CT network was more extensive and complex compared with the HP and TA networks. Moreover, the CT network contained a higher number of microbial contributors compared with the HP and TA networks, suggesting that it plays a crucial role in maintaining the overall status of the system.

Table 2 Correlation network properties of the control, hyperplastic polyp, and tubular adenoma groups.

Network properties	CT	HP	TA
Number of nodes	187	137	117
Number of edges	469	372	302
	(pos = 311, neg = 158)	(pos = 245, neg = 127)	(pos = 210, neg = 92)
Clustering coefficient	0.2	0.306	0.302
Network diameter	10	8	7
Average number of neighbors	5.613	6.698	5.650
Network density	0.035	0.064	0.055
Network centralization	0.130	0.255	0.15353

CT: Control; HP: Hyperplastic polyps; Neg: Negative; Pos: Positive; TA: Tubular adenoma.

Nodes with a high degree (> 10) were regarded as hub species in the co-occurrence networks. In the CT network (Figure 5A), 29 hub species were identified, including Coprococcus A 187866 catus (C. catus), Clostridium ramosum (C. ramosum), ER4 sp000765235, Lawsonibacter sp000177015, Sellimonas intestinalis, Blautia A 141781 caecimuris (B. caecimuris), Blautia A 141781 massiliensis (B. massiliensis), and Anaerobutyricum soehngenii (A. soehngenii) (Supplementary Figure 2A). In the HP network (Figure 5B), 27 hub species were identified, including M. gnavus, C. ramosum, Dysosmobacter sp000403435, Agathobacter rectalis (A. rectalis), B. hansenii, Clostridium Q 135853 saccharolyticum A, B. caecimuris, Enterocloster bolteae (E. bolteae), Dorea A formicigenerans (D. formicigenerans), and B. thetaiotaomicron (Supplementary Figure 2B). In the TA network (Figure 5C), 19 hub species were identified. Blautia A 141781 obeum had the highest node degree, followed by E. bolteae, Dorea A longicatena, Clostridium AQ innocuum, Dysosmobacter sp000403435, M. gnavus, D. formicigenerans, C. catus, C. ramosum, and B. caecimuris (Supplementary Figure 2C). Furthermore, the CT, HP, and TA networks contained commensal bacteria within the Blautia genus: B. massiliensis, Blautia A 141781 faecis, Blautia A 141780 argi, B. obeum, B. caecimuris, and B. wexlerae. Notably, B. hansenii was exclusively present in the HP and TA networks. Many SCFA-producing bacteria were identified as hub species, including C. catus, Faecalibacterium prausnitzii C 71358, A. rectalis, Anaerobutyricum hallii, A. soehngenii, and F. umbilicata. Some hub species were opportunistic pathogens, dysbiosis-associated bacteria, and CRC-associated bacteria, such as E. bolteae, M. gnavus, C. symbiosum, and C. ramosum.

Figure 5 Microbial interaction network of the gut microbiome. A: Co-occurrence networks were constructed at the species level using abundance data from the control (CT) group; B: Co-occurrence networks were constructed at the species level using abundance data from hyperplastic polyps (HP); C: Co-occurrence networks were constructed at the species level using abundance data from tubular adenoma (TA) group. The figure shows that all nodes had more than five connections. Each node in the network represents a single microbial species. The color of each node corresponds to its degree (i.e. number of connections with other nodes). Nodes are represented as follows, according to their degree: > 20, red; 16-19, orange; 10-15, yellow; and 6-9, pink. Edges between nodes represent correlations between those nodes; blue indicates a positive correlation, whereas red indicates a negative correlation.

The microbial correlation networks indicated that although many microbial interactions were present in all groups, some interactions were exclusive to one or two groups, suggesting that altered microbial interactions partly contribute to the distinct gut microbiome signatures of the HP and TA groups; these findings also highlight the potential role of the gut microbiome in promoting colorectal polyp development. Numerous microbial interactions present in the CT network were absent from the HP and TA networks. Specifically, the CT network showed unique microbial associations, such as negative interactions of C. catus with Lawsonibacter sp000177015 and C. ramosum. There were also positive interactions among CT-associated bacteria, such as CAG-45 sp000438375 (Lachnospiraceae) with Coprococcus A 121497 eutactus and CAG-353 sp900066885 (Ruminococcaceae). Additionally, some positive interactions between SCFA-producing bacteria such as A. hallii and A. soehngenii, and between C. eutactus and Butyribacter sp003529475, were absent from the HP and TA networks. Furthermore, the HP and TA networks exhibited specific microbial associations that were absent from the CT network, especially the interactions between commensal bacteria and SCFA-producing bacteria, as well as dysbiosis and CRC-associated bacteria. For example, positive interactions of M. gnavus with B. hansenii and F. umbilicata were observed. Positive interactions of E. bolteae with B. hansenii, Ruthenibacterium lactatiformans, and Clostridium Q 135853 saccharolyticum A, as well as positive interactions of M. gnavus with Mediterraneibacter A 155507 torques, B. thetaiotaomicron, Bacteroides caccae, Faecalimonas phoceensis, and Phocaeicola A 858004 vulgatus, were exclusively present in the HP network. In contrast, positive interactions of E. bolteae with B. fragilis and P. distasonis; B. fragilis with C. innocuum and C. ramosum; and C. ramosum with B. thetaiotaomicron were present in the TA network. These findings support the notion of dysbiosis involvement in colorectal polyp development.

To further characterize the microbial community structure and key microbial associations among patients with different types of polyps, dynamic changes in interactions between the CT and HP networks, and between the CT and TA networks, were explored via DyNet analyses that identified synchronized and rewired nodes across the two datasets. The rewired node score was determined by the D_n-score, which reflects the altered interactions of microbial species across the synchronized networks in the two datasets. DyNet visualization of the synchronized CT and HP networks revealed 92 rewired nodes; 70 rewired nodes were present in both datasets (Figure 6A). Similarly, the synchronized CT and TA networks exhibited 108 rewired nodes; 83 rewired nodes were present in both datasets (Figure 6B). Additionally, the synchronized networks showed seven and two rewired nodes that were exclusive to the HP and TA datasets, respectively. DyNet visualization revealed that unique rewired nodes in the HP group consisted of SCFA-producing bacteria, commensal bacteria, and CRC-associated bacteria. For example, A. rectalis, Ruminococcus D bicirculans, Blautia A 141780 stercoris, Eubacterium I ramulus, and S. gallolyticus were identified as unique rewired nodes in the HP group. In contrast, A. rectalis and Megamonas funiformis were identified as unique rewired nodes in the TA group. Many interactions involving rewired nodes exclusive to the HP group were between SCFA-producing bacteria and commensal bacteria. Examples include interactions of A. rectalis with F. prausnitzii, Lachnospira eligens, and A. soehngenii; E. ramulus with C. catus; R. bicirculans with A. hallii; and A. rectalis with D. formicigenerans. However, rewired nodes exclusive to the TA group displayed fewer interactions between SCFA-producing bacteria and commensal bacteria compared with nodes in the HP group. Examples include interactions of A. rectalis with C. catus. Notably, S. gallolyticus was identified as a unique rewired node in the HP dataset, demonstrating interactions with SCFA-producing bacteria and commensal bacteria such as F. prausnitzii, B. faecis, B. caecimuris, Dysosmobacter sp000403435, Anaerotignum lactatifermentans, and Amedibacillus dolichus. These findings suggest that co-occurrence patterns and microbial interactions differ between the HP and TA groups, which could also describe the development of the two colorectal polyp types and their distinct malignancy landscapes.

Figure 6 Co-occurrence networks and DyNet visualization of synchronized co-occurrence networks. A: The control (CT) and hyperplastic polyps (HP) networks; B: The CT and tubular adenoma (TA) networks. Nodes represent microbial species, whereas edges represent correlation coefficients between microbial species. Green edges are only present in the CT network; red nodes and edges are exclusive to the HP and TA groups; and white nodes and gray edges are shared between the CT and HP groups, as well as the CT and TA groups.

DISCUSSION

The development of colorectal polyps is significantly influenced by alterations in gut microbiome community composition and ecology. It typically arises from an imbalance in the gut microbiome community and the colonization of microbial species that trigger chronic inflammation, eventually leading to the multistep process of polyp formation[43]. Here, we investigated differences in gut microbiome communities between individuals with and without colorectal polyps, which provided insights concerning the microbial species, functions, and mechanisms that impact colorectal polyp development through the conventional adenoma-carcinoma sequence and the serrated polyp pathway. Furthermore, we identified multiple differentially abundant species in HP patients and TA patients; many of these species have been associated with CRC, dysbiosis, and colorectal adenoma[44-50].

In the present study, the overall gut microbiome compositions did not significantly differ among the CT, HP, and TA groups. However, there were variations in the microbial species associated with each group. These findings are consistent with the results of previous cohort studies that did not demonstrate differences in overall gut microbiome composition between normal samples and samples from patients with adenoma[13,15]. However, previous studies regarding colorectal polyps have yielded inconsistent results regarding community diversity[51,52]. Some studies showed no differences in diversity, whereas others revealed increased diversity in patients with polyps. These discrepancies may be influenced by factors such as sample size, statistical power, or the presence of population-specific microbial drivers or pathogens. Another possible explanation is that the gut microbiome associated with colorectal polyps is similar to the gut microbiome of healthy individuals[52,53]. The discrepancies also may have arisen from the limited taxonomic coverage and reliance on reference genomes during whole metagenomic sequencing taxonomic profiling[54,55].

Multiple studies have shown that some pathogenic bacteria, such as Fusobacterium, B. fragilis, and E. coli, are highly abundant in patients with colorectal polyps[4,5,7-9,56]. For example, biofilm formation and virulence factors from E. coli, responding to environmental changes in the mucosa, can induce genotoxic effects as well as inflammatory and neoplastic processes. This occurs through the activation of DNA damage, oxidative stress, and the NF-κB and STAT3 signaling pathways[57]. We did not identify E. coli and Fusobacterium in the HP and TA groups. These results are consistent with previous findings concerning colorectal adenomas[44,47]. Furthermore, they underscore the importance of developing polyp-specific biomarkers that are specifically associated with colorectal adenomas.

Furthermore, we observed substantial variation in differential microbial species among the HP, TA, and CT groups. The present study revealed a significant increase in M. gnavus abundance among HP and TA patients, whereas the abundances of B. fragilis and S. gallolyticus were only significantly increased in TA patients. M. gnavus abundance is elevated in various gastrointestinal diseases, including inflammatory bowel disease, irritable bowel syndrome, CRC, Crohn's disease, and ulcerative colitis[50,58]. This elevated abundance may be associated with inflammation and bowel neoplasia[58]. M. gnavus is capable of metabolizing primary bile acids, which are not absorbed by the small intestine, into secondary bile acids[59]. Elevated levels of secondary bile acids can induce oxidative and nitrosative stresses, DNA damage, apoptosis, and mutations in host cells. Furthermore, secondary bile acids interact with the farnesoid X receptor in an antagonist manner, leading to enhanced Wnt signaling in the conventional adenoma-carcinoma sequence[60]. A meta-analysis of case reports and case series from 1970 to 2010 indicated that approximately 60% of patients with S. gallolyticus infections also had concurrent colon adenomas or carcinomas, a rate significantly higher than the percentage observed in the general population[61]. Additional studies have shown associations of colorectal adenoma or carcinoma with S. gallolyticus infection[62].

The current study offers additional evidence to support the strong association between S. gallolyticus and colorectal adenomas. It has been observed that S. gallolyticus is closely linked to the transformation of colorectal mucosa into adenoma, potentially through mechanisms such as epithelial barrier invasion or virulence factor release. S. gallolyticus enhances inflammation and tumorigenesis by targeting NF-κB and Wnt/β-catenin signaling, upregulating β-catenin levels, and inducing inflammation via cytokines (e.g., interleukin-1, interleukin-8, and cycolooxygenase-2)[63-65]. Therefore, S. gallolyticus is presumed to participate in neoplastic transformation. Our findings indicate that S. gallolyticus is a key contributor to the mevalonate pathway, which is substantially enriched in patients with TA. This pathway plays a crucial role in the biosynthesis of compounds such as isopentenyl pyrophosphate, farnesyl pyrophosphate, and geranylgeranyl pyrophosphate, which serve as building blocks for various essential biomolecules, including lipoproteins, dolichol, ubiquinone, and cholesterol-derived products (e.g., steroid hormones, oxysterols, vitamin D, and bile acids). These metabolites play important roles in the regulation of cellular metabolism[66]. However, isopentenyl pyrophosphate, farnesyl pyrophosphate, and geranylgeranyl pyrophosphate can also contribute to inflammation-mediated tumor growth through oncogenic activation of Ras[66]. Previous studies have revealed that mevalonate pathway inhibition can impede the growth and proliferation of colon cancer cell lines[67]. The present study revealed a novel link between S. gallolyticus and the mevalonate pathway, which involved cell signaling in carcinogenesis. Notably, mevalonate pathway activity was predicted to substantially increase in the HP and TA groups compared with the CT group. This finding strongly implies that the mevalonate pathway plays a key role in colorectal polyp formation. Furthermore, the present study revealed significant increases in the abundances of B. caecimuris, C. symbiosum, and B. fragilis among patients with TA. B. caecimuris is a commensal bacterium in the human gut, and there is no evidence linking it to colorectal polyps[68]. However, B. caecimuris has been detected in fecal samples from CRC patients[69]. The abundance of C. symbiosum is increased in colorectal adenoma, making it a promising biomarker for the noninvasive detection of colorectal adenoma[70,71]. B. fragilis, a mucin-degrading bacterium[72], can adhere to intestinal mucus and utilize it as a nutrient source for growth[73]. B. fragilis produces a metalloprotease that alters signaling pathways and induces the production of reactive oxygen species, resulting in DNA damage and E-cadherin cleavage[74,75]. These results indicated that TA patients predominantly had microbial species associated with inflammation and the adenoma-carcinoma sequence. Our findings highlight the distinct contributions of the gut microbiome to the conventional adenoma-carcinoma sequence and the serrated polyp pathway. In the conventional adenoma-carcinoma sequence, the presence of pathogens and inflammation-enhancing microbial species in an inflammatory environment can promote the development of colorectal carcinogenesis. Our findings also suggest that the microbial species detected in this study may be useful for identifying patients with a high risk of colorectal adenoma.

Functional analysis provides valuable insights into the complex mechanisms underlying the development of serrated polyps. In the present study, we predicted enrichment of the APS reductase pathway in HP patients; this pathway contains a microbial enzyme that metabolizes sulfate to sulfite. Our findings suggest that the APS reductase pathway is associated with SRB, including D. piger and B. wadsworthia, which can increase H₂S levels. Previous studies concerning gut microbiome alterations among individuals with colorectal adenomas have revealed increased levels of SRB such as Bilophila, Desulfovibrio, and B. wadsworthia in patients with adenomatous polyps[76]. Endogenous production of H₂S primarily occurs through the metabolic activities of SRB and other bacteria, which metabolize inorganic sulfur compounds such as sulfate and sulfite (commonly found in processed food preservatives), as well as organic sulfur compounds (e.g., cysteine or taurine, present in red meat)[77]. There is emerging evidence that the metabolism of organic sulfur by SRB in the human gut may link diets high in red and processed meat to increased risks of early-onset adenomas[78,79]. H₂S can damage the mucosal layer by disrupting disulfide bonds, which causes the mucus layer to become less viscous and more permeable. These changes allow toxic compounds and microbial species from the gut lumen to directly interact with the epithelial cell surface, leading to cellular damage, triggering immune responses, and promoting inflammation[79,80]. Chronic inflammation is frequently associated with gastrointestinal cancers, and individuals with colitis have an increased risk of cancer[81]. Therefore, mucin restoration and mucosal barrier strengthening are therapeutic objectives during chronic inflammation, particularly in patients with extensive colitis. This approach is likely to reduce neoplastic processes in the intestinal epithelium and improve health outcomes.

A previous study involving a Thai population revealed decreased B. thetaiotaomicron abundance among individuals with adenoma, whereas the abundance was increased among individuals with CRC. Conversely, P. distasonis abundance was increased among patients with adenoma and patients with CRC[82]. The present study showed no significant difference in the relative abundances of B. thetaiotaomicron and P. distasonis between the polyp groups (HP and TA) and the CT group. Both microbial species contribute to thiosulfate oxidation within the sulfur oxidation pathways. Specifically, they are involved in the enzymatic process known as thiosulfate:quinone oxidoreductase, which facilitates thiosulfate oxidation and subsequent tetrathionate production. These intermediates in the sulfur cycle may serve as key sites for electron transfer and energy generation[83]. Nevertheless, it is important to note that the present study specifically focused on the contributions of B. thetaiotaomicron and P. distasonis to metabolic pathways that exhibited significant differences compared with the CT group, rather than conducting a comprehensive evaluation of all pathways.

Co-occurrence network analysis revealed fewer interactions between beneficial microbial species in the HP and TA groups, which may be associated with the occurrence of colorectal polyps. Notably, positive interactions among beneficial microbial species such as C. catus, Dysosmobacter sp000403435, and the butyrate-producing genus Eubacterium were diminished. Additionally, the HP and TA groups showed fewer negative associations between C. catus and opportunistic pathogens such as E. bolteae. These results suggest that decreased interactions among beneficial microbial species contribute to colorectal polyp formation. Furthermore, synchronized network analysis demonstrated differences in co-occurrence patterns between the HP and TA groups. For example, the TA group exhibited fewer occurrences of beneficial microbial species compared with the HP group; they also displayed higher levels of co-occurrence involving CRC-associated bacteria. These microbial species are associated with the inflammation that leads to the progression of HP and TA. SCFAs can reduce gut inflammation by promoting gut barrier integrity and permeability through various mechanisms that also help to maintain homeostasis[84-86]. Therefore, the co-occurrence of SCFA-producing bacteria in patients with HP may enable the maintenance of a good health status[56]. In contrast, a decrease in the co-occurrence of SCFA-producing bacteria, accompanied by an increase in the co-occurrence of CRC-associated bacteria, may be involved in colorectal adenoma formation and colorectal carcinogenesis in TA patients. This hypothesis is supported by a previous report in which patients with cystic fibrosis, who carry adenomas and have a high risk of CRC, exhibited reductions in SCFA-producing bacteria and an increased relative abundance of B. fragilis[87].

Colorectal polyps can develop through two main genetic pathways. The conventional adenoma-carcinoma sequence is characterized by mutations in the adenomatous polyposis coli gene, chromosomal instability, or microsatellite instability, as well as the absence of CpG island methylator phenotype (CIMP) alterations; this pathway leads to TA. The alternative pathway, the serrated polyp pathway, is mainly characterized by BRAF mutations and high numbers of CIMP alterations; this pathway leads to HP, SSA/Ps, and traditional serrated adenoma. There is substantial histological overlap between SSA/Ps and HP; SSA/Ps constitute up to 30% of all colon cancers. SSA/Ps can develop either as primary tumors or evolve from HP[3], suggesting CRC susceptibility in HP patients. Our investigation of gut microbiome characteristics in different types of colorectal polyps revealed that HP and TA had distinct gut microbiome signatures. This comprehensive analysis of gut microbiome signatures has provided valuable insights concerning gut microbiome contributions to colorectal polyp development, particularly with respect to gut microbiome effects on carcinogenesis. In the serrated polyp pathway, dysbiosis and gastrointestinal disease-associated bacteria, along with the inflammation-inducing sulfur oxidation pathway, contribute to the establishment of a tumor microenvironment. However, this phenomenon is counterbalanced by increased abundances and co-occurrence of SCFA-producing bacteria in HP patients. Furthermore, we speculate that the increased abundance of CRC-associated bacteria mediating the mevalonate pathway involves inflammation and cell proliferation in TA patients, suggesting that such bacteria contribute to the conventional adenoma-carcinoma sequence.

The present study utilized long-read 16S rRNA sequencing, which offers higher resolution for assignments of microbial identity at the species or strain levels. However, it is important to acknowledge the limitations of this study. In particular, it had a small sample size, the results should be confirmed in another cohort, and experimental validation is needed. Furthermore, during bioinformatics analysis, ASVs were aggregated to the species level and then used as input for alpha and beta diversity analyses, as well as differential abundance and co-occurrence analyses.

CONCLUSION

This study utilized differential abundance, co-occurrence, and differential pathway analyses to characterize the gut microbiome signatures in colorectal polyps. The differential abundance analysis identified candidate microbial species that could serve as biomarkers for colorectal polyps. The co-occurrence analysis provided insights concerning the dynamic changes in microbial correlation networks among the CT, HP, and TA groups. The differential pathway analysis predicted functional pathways and determined the roles of microbial species in metabolic function during colorectal polyp development. The results highlight the importance of numerous pathways in colorectal polyp development, offering evidence to support interventions and treatment in the context of CRC carcinogenesis. Furthermore, analyses of the dynamic changes between the CT group and colorectal polyp groups (HP and TA) enhanced the understanding of gut microbiome interactions within the community. Specifically, our findings suggest that HP patients have an increased risk of CRC; more effective strategies are needed to identify and manage such patients.

ACKNOWLEDGEMENTS

In memory of Assistant Professor Dr. Bunchorn Siripongpreeda — his robust leadership and project management made this work possible. We sincerely thank Watsawan Sangkaew, Watcharapol Khoonin, Piangpit Kamolphan, Teeramade Thidseesang, Yanee Poomjaroen, Supanan Makkram, Suyada Sukho, Phaktheema Phoothong, Suchada Panaiem, Chanisara Suebwongdit, Napatsawan Opad, and Tulyawat Prasongmaneerut for assistance with participant recruitment. We especially thank Thanrada Popraros, Napong Leelasithorn, Kawin Chinpong, Tanabhorn Kookongkritsadakorn, Kamonwan Soonklang, Penpisut Homsuwan, and Kerati Jirawattanapalin for assistance with collecting patient information from questionnaires. We also thank Chirayu Auewarakul, Thaniya Sricharunrat, and Montri Gururatsakul for providing medical advice.