Skip to main content
NCBI home page
ESMO Gastrointestinal Oncology logoLink to ESMO Gastrointestinal Oncology
. 2025 Aug 26;9:100229. doi: 10.1016/j.esmogo.2025.100229

Characterization of tumor-associated microbiome in multifocal small intestinal neuroendocrine tumors (SI-NETs)

IT Lee 1,2, A Dohlman 1,2, T Gao 1, Z Zhang 1,2, Y Kasai 3, GE Kim 4, C Thirlwell 5, E Nakakura 3, M Meyerson 1,2,6, N Mäkinen 1,2,∗
PMCID: PMC12836619  PMID: 41647997

Abstract

Background

Small intestinal neuroendocrine tumors (SI-NETs) are among the most common neoplasms of the small bowel; however, the molecular mechanisms underlying their pathogenesis are largely unknown. The multifocal nature of SI-NETs, their putative distinct genomic origins, and their enrichment in the distal ileum led us to hypothesize that environmental factors, such as pathogenic organisms, might play a role in the development of these lesions.

Materials and methods

To study the tumor-associated microbiome of multifocal SI-NETs and its potential role in pathogenesis, we used matched whole genome and transcriptome sequencing data from a cohort of 10 multifocal SI-NET patients, including 70 primary ileal NETs and their matched normal ileal mucosa and/or whole blood specimens.

Results

Microbial communities in the ileal tissue samples were primarily composed of bacteria. The most abundant genera included well-known gastrointestinal, oral, and mucosal bacteria. Ileal tissue samples from individual patients contained distinct patient-specific microbial communities. Although the microbiota composition did not show significant differences between ileal NET and normal ileal tissues, genus Propionibacterium was found to be enriched in the normal tissue specimens.

Conclusions

This study comprehensively characterizes the tissue-resident ileal microbiome of multifocal SI-NET patients. We provide clear evidence that the microbial communities in the ileum are largely patient specific, whereas our genus-level analyses suggest that SI-NET pathogenesis is unlikely driven by individual microorganisms present in the tumors at the time of surgical resection.

Key words: small intestinal neuroendocrine tumors, tumor microbiome, multifocality, ileum

Highlights

  • •

    Multifocal SI-NETs harbor tumor-associated bacteria, fungi, and viruses.

  • •

    Microbial communities in the tumors are primarily composed of bacteria.

  • •

    Microbial profiles and multivariate analyses provide strong evidence for patient-specificity of the ileal microbiome.

  • •

    No microorganisms, known or novel, are enriched in tumors compared with their matched normal tissue samples.

  • •

    Tumor-resident microorganisms are unlikely to be involved in multifocal SI-NET pathogenesis.

Introduction

Small intestinal neuroendocrine tumors (SI-NETs) are among the most common neoplasms of the small bowel, accounting for ∼40% of all small intestinal malignancies.1 SI-NETs are thought to originate from enterochromaffin cells of the digestive tract and are predominantly located in the distal ileum.2 Around one-third of the SI-NET patients are diagnosed with multiple synchronous primary tumors, but percentages as high as 54% have been reported.3,4 Currently, the only curative treatment option for SI-NETs is complete surgical resection of the tumors. Many patients, however, cannot undergo surgery as they present with extensive metastatic disease. The 5-year survival rate is <50% for patients with metastases.2

SI-NETs have a low somatic mutational burden and an unusual lack of somatic driver gene mutations.5, 6, 7 The most common genomic event identified in SI-NETs is allelic loss of chromosome (chr) 18, occurring in ∼60% of tumors,5,8,9 and the only established recurrently mutated gene is cyclin-dependent kinase inhibitor 1B (CDKN1B), affecting ∼8%-10% of tumors.6,10 Recently, we and others showed that synchronous primary tumors from the same SI-NET patient harbor distinct somatic mutational profiles, suggesting that these tumors develop independently.11, 12, 13 The low frequency of shared somatic genomic events in SI-NETs indicates that other, non-genetic mechanisms may be involved in their development.

The enrichment of SI-NETs in the distal ileum and the independent clonal origins of multifocal SI-NETs led us to hypothesize that environmental factors impact the development of these lesions. Tumors in the gastrointestinal tract are particularly likely to be influenced by their microbial environment, and consequently, the interactions between microbes and cancer cells may have an impact on cancer initiation and progression. The density of microbiota in the small bowel increases from a low density in the duodenum to a higher density in the terminal ileum.14 Additionally, the microbial populations of the small bowel have been reported to be less diverse and more dynamic than in the colon due to the short transition time, the influx of digestive enzymes and bile, and irregular food substrate delivery.15

The microbiota of the small bowel are mainly composed of bacteria belonging to the phyla Firmicutes/Bacillota, Proteobacteria, and Actinobacteria.16 Bacterial genera frequently observed in the small bowel include Lactobacillus, Clostridium, Staphylococcus, Streptococcus, and Actinomyces.14,17,18 Inconsistencies exist, however, in the reported bacterial genera and their taxonomic classification across studies because of differences in sample collection and analytical methods. It is noteworthy that only certain bacterial species, such as those from Clostridium, Lactobacillus, and Enterococcus genera, can penetrate or attach to the mucus layer of the intestinal epithelium, which may explain some of the differences seen in the bacterial communities between the small intestinal lumen and mucosa.19

Research on the microbiome of the human small bowel is challenging due to its poor accessibility, high temporal dynamics, and complexity of sample collection. Although previous studies have explored the role of the small intestinal microbiome in the development of gastrointestinal disorders, such as functional dyspepsia and inflammatory bowel disease, little is known about its role in SI-NET pathogenesis. In this study, we used whole genome and transcriptome sequencing data from a cohort of 10 multifocal SI-NET patients, including 70 tumor specimens, to study their tumor-associated microbiome and its potential role in pathogenesis.

Materials and methods

Sample cohort

Our sample material consisted of 70 synchronous primary ileal NETs and their matched normal ileal mucosa or whole blood specimens from 10 patients with multifocal SI-NETs (Supplementary Table S1, available at https://doi.org/10.1016/j.esmogo.2025.100229). The SI-NET patients drank a clear liquid diet the day before surgery and had nothing by mouth after midnight. The patients were typically asked to also drink a laxative, such as magnesium citrate, before noon the day before surgery. The tissue samples were collected directly from the mucosal surface of a surgically removed segment of the ileum. Whole blood specimens from 3 additional patients with multifocal SI-NETs and 10 patients with a solitary SI-NET were also included in the study. This study was conducted in accordance with the Declaration of Helsinki. All patients provided informed consent in accordance with the protocols approved by the Institutional Review Boards of the Dana-Farber Cancer Institute (DFCI) and the University of California San Francisco.

Whole genome and transcriptome sequencing

Genomic DNA (gDNA) extraction, whole genome sequencing (WGS), and pre-processing of the WGS data were carried out for 70 synchronous primary ileal NETs, nine normal ileal mucosa and 16 whole blood specimens, as previously described.12 Briefly, gDNA was extracted from the whole blood samples using the QIAamp® DNA Blood Kit (Qiagen, Germantown, MD), whereas gDNA extraction of primary ileal NETs and normal ileal mucosa samples was carried out using the MagAttract® HMW DNA Kit (Qiagen, Germantown, MD). Both kits are designed to effectively remove PCR inhibitors through their purification process. The mean depth of WGS data was 81.0×  [62.5× to 171.3×] for ileal NET, 44.3×  [30.5× to 87.7×] for normal ileal mucosa, and 42.4×  [32.4× to 55.3×] for whole blood specimens (Supplementary Table S1, available at https://doi.org/10.1016/j.esmogo.2025.100229). Total RNA was extracted from 69 matched synchronous primary tumors and nine normal ileal mucosa specimens using the RNeasy kit (Qiagen, Germantown, MD). Whole transcriptome sequencing with ribosomal RNA depletion was carried out at the Center for Cancer Genomics at DFCI. Briefly, RNA libraries were sequenced on HiSeq 2500 (Illumina, San Diego, CA) to generate 100-base pair (bp) paired-end reads. Sequenced reads were aligned to the GRCh38 reference genome using STAR aligner v.2.7.5b.

Extraction of non-human sequences

PathSeq,20 from the Genome Analysis Toolkit (GATK) package v4.5.0, was used to extract non-human reads from the WGS and RNA sequencing data and align them to a microbial reference dataset. The minimum clipped read length was set to 60 to minimize the false positive rate. Paired-end reads aligning to viral decoy sequences included in the GRCh38 reference genome from the National Cancer Institute Genomic Data Commons with an alignment score >35 were appended to the PathSeq output to ensure viral reads were not excluded from downstream analysis. By default, PathSeq removes all reads aligned to the human reference genome.

The resulting output files were screened for additional human sequences by aligning the reads to a reference genome containing the updated human genome T2T-CHM13v2.0, as well as the Immuno Polymorphism database (IMGT/HLA), cloning vector sequences from the National Center for Biotechnology Information’s UniVec, a curated database of human transcripts from GENCODE human v25, and human breakpoint junction sequences, as described in Walker et al.20 WGS and RNA sequencing data were aligned to this reference with Burrows-Wheeler Aligner - Maximum Exact Matches (BWA-MEM) v0.7.17 and STAR v.2.7.5b, respectively. Paired-end reads mapping to the reference with an alignment score >35 were excluded from further analysis.

Microbial classification

After host subtraction, the remaining reads were classified with Kraken221 using the PlusPF genomic database (downloaded in June 2023), which contains complete or chromosome-level assemblies of the RefSeq archaea, bacteria, viral, plasmid, protozoa, fungi, human, and UniVec_Core sequences.22 Default parameters were used, apart from the confidence score, which was increased to 0.15. Classified read counts at each taxonomic level were normalized to the total primary read counts (in millions) in the original WGS and RNA sequencing outputs to obtain a reads per million (RPM) statistic, which was used to quantify microbial abundance in our samples. MetaPhlAn v4.0.623 was run with default parameters to identify microbes in the ileal tissue samples using marker genes, complementing and supporting microbial read classification with Kraken2. The database mpa_v_Jun23_CHOCOPhlAnSGB_202403 was used for the analysis.

Decontamination

To remove potential contaminants and false positives after classification, we implemented a two-step filtering approach. First, all microbial species that were present in >20% of the whole blood samples were excluded (Supplementary Figure S1A, available at https://doi.org/10.1016/j.esmogo.2025.100229), as implemented previously.24 Under normal physiological conditions, human blood has a low microbial biomass, therefore, we reasoned that highly prevalent microbes in these samples are likely to be technical or biological contaminants. Second, a Jaccard/Tanimoto similarity test was used to identify microbial species that co-occurred in the WGS and RNA sequencing data more often than expected by random chance (Supplementary Figure S1B, available at https://doi.org/10.1016/j.esmogo.2025.100229). We reasoned that microbes involved in SI-NET development would likely persist in the ileum and be transcriptionally active, and thus, detectable in both WGS and RNA sequencing data from matched samples. The exact Jaccard/Tanimoto significance was calculated with the ‘jaccard’ R package v0.1.0. A P value < 0.05 was considered statistically significant. Microbial species with statistically non-significant values were excluded from further analysis. Because viruses are believed to disseminate via the bloodstream, and can have either DNA or RNA genomes, this decontamination approach was only applied to bacteria and fungi. We limited our analysis to human herpesviruses, which were among the most prevalent and biologically relevant viruses present in our sample cohort.

Decontamination removed 92.3% of the unique microbial species and retained 84.5% of the unambiguous microbial reads (Supplementary Figure S1C, available at https://doi.org/10.1016/j.esmogo.2025.100229). Notably, most reads in each whole blood specimen were considered contaminants, whereas the percentage of contaminant reads varied greatly among the multifocal ileal NET and normal ileum samples (Supplementary Figure S1D, available at https://doi.org/10.1016/j.esmogo.2025.100229). This suggests that our filtering approach successfully targeted low abundance, spurious false positives.

Diversity analyses

Alpha diversity was calculated using the ‘skbio’ package v0.6.0. Beta diversity calculations were carried out with the ‘vegan’ R package v2.6-6.1. Dissimilarity statistics and principal coordinates were calculated with the vegdist function (with method ‘bray’) and cmdscale function, respectively. Permutational multivariate analysis of variance (PERMANOVA) was carried out using the adonis2 function. The dispersion of samples within groups was calculated and visualized using the betadisper function.

Differential abundance analyses

To identify microorganisms with differential abundance between ileal NET and normal ileal tissues, we carried out three complementary analyses. To quantify differential presence/absence of microbes in these tissues, we used Fisher’s exact test (scipy.stats.fisher_exact function from ‘scipy’ package v1.11.2). To adjust for covariates and account for the relative abundance of microbes, we deployed MaAsLin2 v1.18.025 and ANCOM-BC v2.6.0.26 Tissue type (‘phenotype’) and patient were set as fixed effects. False discovery rate (FDR)-adjusted q < 0.05 was considered statistically significant.

Metagenomic assembly

To further query potential microbial presence in multifocal SI-NETs, we carried out metagenomic assembly on putatively non-human reads. Reads classified as either non-human or unclassified by Kraken2 were co-assembled into contigs by SqueezeMeta v1.6.3,27 using the default settings. Kraken2 was used to classify the assembled contigs, as described earlier. The remaining unclassified contigs were analyzed with BLASTn (BLAST+ 2.15.0) with the following settings: max e-value of 10e-10, max hsps of 1, and max target seqs of 5. The results were further filtered by requiring 90% identity, and alignment length measuring at least 80% of the contig length.

Results

Multifocal SI-NETs harbor tumor-associated bacteria, fungi, and viruses

To characterize the tumor-associated microbiome of multifocal SI-NETs, we analyzed WGS data from 70 synchronous primary ileal NETs and their matched normal ileal mucosa samples from 10 multifocal SI-NET patients. After host-filtering and decontamination (see Materials and methods), tumor-resident sequencing reads were aligned to bacterial, fungal, and viral genomes across both tissue types, with bacteria being the most abundant (Figure 1A). Bacterial sequence abundances among the samples ranged broadly between 10−2 to 103 RPM, corresponding to an approximate range of 10−5 to 1 bacterial cell per human cell. Bacterial sequence abundances were significantly higher in both ileal NET and normal ileal tissues compared with whole blood samples, but did not differ between NET and normal ileal tissues. Notably, fungi were entirely absent from the whole blood samples. Presence of human Herpesviridae was consistently low across all tissue types. The alpha diversity of microorganisms varied considerably among NET and normal ileal tissues, but we did not observe any statistically significant difference in diversity between the two sample types (Figure 1B).

Figure 1.

Figure 1

Open in a new tab

Overview of the microbiome in multifocal small intestinal neuroendocrine tumor (SI-NET) patients after decontamination. (A) Kingdom-level microbial abundances in WGS data of whole blood, normal ileum, and ileal NET samples after decontamination. Each point represents the abundance of bacteria, fungi, or human Herpesviridae in a single sample. Statistical significance between sample groups is indicated by asterisks: ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001. (B) Shannon indices for the normal ileum and ileal NET samples. (C) Genus-level microbial abundances in normal ileum and ileal NET samples. Each point represents a sample with the corresponding genus on the y-axis and the abundance (in RPM) on the x-axis. Genera included in the plot are present in at least 10 samples and ordered by the median abundance (in RPM) across all samples. The size of the dot at ‘zero’ corresponds to the number of samples in which the genus was not detected. RPM, reads per million; WGS, whole genome sequencing.

To define the microbial composition of multifocal ileal NETs and normal ileum samples, we examined genera that were present in at least 10 samples and had a high median abundance (Figure 1C). Among the most abundant microbes were known common gut-colonizing bacteria, such as Parabacteroides, Clostridium, Streptococcus, and Actinomyces. The community also included oral bacteria, such as Mogibacterium and Gemella. Fungal reads corresponded to the genus Candida, and the most abundant human herpesvirus in our sample cohort was human betaherpesvirus 5.

Identification of patient-specific microbial communities in multifocal SI-NET patients

To assess whether microbial communities were similarly or differentially localized to multifocal ileal NET and normal ileal tissues, we carried out a PERMANOVA analysis to quantify the variation in microbial composition that can be attributed to tissue type. Variation in microbial abundance was not significantly dependent on the tissue type, suggesting that the composition of the ileal NET-associated microbiome does not differ significantly [Pr(>F) = 0.867, R2 = 0.005] from that of the adjacent normal tissue (Figure 2A). We also did not observe significant differences between the microbial compositions of ileal NETs with and without chr18 allelic loss [Pr(>F) = 0.212, R2 = 0.012] (Figure 2B).

Figure 2.

Figure 2

Open in a new tab

Microbial communities are patient specific in multifocal small intestinal neuroendocrine tumors (SI-NETs). (A) Principal coordinate analysis (PCoA) by tissue type. Samples are grouped by tissue type and the distributions of samples in each group are approximated by ellipses. PERMANOVA analysis P value [‘Pr(>F)’] indicates that microbial variation is only minimally dependent on tissue type, as visualized by the sizable overlap and similar shape of the NET and normal ileum ellipses. (B) PCoA by chromosome (chr) 18 allelic loss status. Ileal NETs are grouped by chr18 allelic loss status and the distributions of tumors in each group are approximated by ellipses. PERMANOVA analysis P value [‘Pr(>F)’] indicates that microbial variation is only minimally dependent on chr18 allelic loss, as visualized by the sizable overlap and similar shape of the ellipses labeled ‘WT’ and ‘Chr18 Allelic Loss’. (C) PCoA by patient. Samples are grouped by patient and the distributions of samples in each group are approximated by ellipses. PERMANOVA analysis P value [‘Pr(>F)’] indicates that microbial communities are patient specific, as visualized by the distinct patient clusters. (D) Total microbial read abundance and genus-level relative abundances. The top row displays the total microbial abundance in RPM for each sample. The bottom row displays the relative abundance of genera in each sample, colored by genus. LOH, loss of heterozygosity; N, normal ileum; P, primary tumor; RPM, reads per million; WT, wild type.

Notably, we did observe that a significant fraction (R2 = 0.335) of the variation in the tissue-associated microbial composition was explained by the patient from whom the sample was taken (Figure 2C). A graph of microbiome composition of ileal NETs from individual patients also appears consistent with the presence of patient-specific microbial populations. For example, multiple NET specimens from patient 876 were seen to harbor high relative abundances of Limosilactobacillus sequences, whereas multiple NET specimens from patient 947 harbored high relative abundances of Phocaeicola sequences (Figure 2D). Tumors from some of the patients also appeared to display gradients in microbial composition and abundance according to the order in which they were collected. For example, in patient 852, the relative abundance of Bacteroides and Phocaeicola increased gradually from tumor P11 onward, whereas in patient 848 the relative abundance of Turicibacter appeared to increase across samples as Streptococcus decreased. This pattern potentially suggests the presence of location-specific microbial communities along the length of the ileum, as tumor specimens from each patient were collected from one end of the resected ileal segment to the other. The microbial classifications by Kraken2 and the presence of patient-specific microbial populations in ileal NETs were supported by analysis with MetaPhlAn, especially in the subset of samples for which there were sufficient microbial reads for reliable detection of microbe-specific marker genes using MetaPhlAn (Supplementary Figure S2, available at https://doi.org/10.1016/j.esmogo.2025.100229).

Differential microbial abundance analysis of multifocal SI-NET and normal ileum samples

Microorganisms associated with cancer have previously been shown to differentially localize to tumors and their surrounding normal tissues. Therefore, we sought to identify microbes enriched in either ileal NET or adjacent normal tissues, suggesting a role in SI-NET tumorigenesis. We employed multiple complementary differential abundance and prevalence analyses. First, to identify presence/absence differences between tumor and normal ileum samples, we used Fisher’s exact test, which revealed that Propionibacterium was significantly enriched in the adjacent normal ileal tissues (P = 0.026, odds ratio = 24.8) (Figure 3A).

Figure 3.

Figure 3

Open in a new tab

Differential prevalence and abundance analyses of microbes in multifocal small intestinal neuroendocrine tumors (SI-NETs) and normal ileum samples. (A) Fisher’s exact test comparing microbial prevalence in ileal NET and normal ileum samples. Points represent microbial genera and their prevalence in ileal NETs (x-axis) and normal ileum samples (y-axis). Genera deviating significantly from the diagonal indicate enrichment in either ileal NET or normal ileal tissue. (B) Volcano plot of the MaAsLin2 result. The MaAsLin2 coefficient (x-axis) indicates the enrichment of genera between ileal NET (positive) and normal ileum (negative). The statistical significance threshold is represented by the horizontal line (P < 0.05). (C) Volcano plot of the ANCOM-BC result. The x-axis represents the ANCOM-BC bias-corrected log fold-change, indicating the magnitude of the differential abundance between ileal NET (positive) and normal ileum (negative). (D and E) show Propionibacterium and Staphylococcus abundances in ileal NET and normal ileum samples, respectively. While Propionibacterium and Staphylococcus were not more abundant in normal ileal tissues compared with tumor tissues, they were more prevalent in normal ileum samples.

To account for the relative abundance of microbes, as well as variation in patient-specific microbial communities, we next assessed the association between microbial abundance and tissue type using MaAsLin2 and ANCOM-BC, multivariable regression algorithms designed for differential abundance analysis of microbiomes. Results of MaAslin2 indicated that Propionibacterium was significantly enriched in the adjacent normal ileal tissues (P = 1.48 × 10−3), consistent with Fisher’s exact test (Figure 3B). On the other hand, ANCOM-BC indicated that Staphylococcus was significantly enriched (P = 0.044) in the adjacent normal ileal tissues (Figure 3C). However, after carrying out multiple hypothesis correction (FDR), the enrichment of Staphylococcus was no longer significant (q = 1), whereas the enrichment of Propionibacterium remained significant (q = 2.08 × 10−2). A clear majority of Kraken2-assigned Propionibacterium reads aligned to species Propionibacterium acidifaciens and Propionibacterium freudenreichii with BLASTn.

To further investigate these candidates, we examined the prevalence and abundance of Propionibacterium and Staphylococcus in our sample cohort. Propionibacterium was found in four of nine normal ileum samples (44.4%) with a mean abundance of 1.63 × 10−3 RPM, and in 8 of 70 ileal NETs (11.4%) with a mean abundance of 1.98 × 10−3 RPM. Staphylococcus was found in two of nine normal ileum samples (22.2%) with a mean abundance of 2.95 × 10−3 RPM, and in 11 of 70 ileal NETs (15.7%) with a mean abundance of 7.42 × 10−4 RPM. No statistically significant differences were observed in the abundances of Propionibacterium and Staphylococcus genera between ileal NET and normal ileum samples (P = 9.87 × 10−2 and P = 6.27 × 10−2, respectively) (Figure 3D and E).

Metagenomic assembly does not identify additional pathogens in multifocal SI-NET patients

A limitation of reference-based metagenomic analysis tools is their inability to identify unknown or uncharacterized microorganisms that do not have a complete reference genome. To detect non-reference microorganisms in multifocal SI-NETs, we carried out metagenomic and metatranscriptomic co-assembly of putative non-human sequences using SqueezeMeta. WGS and RNA sequencing data were assembled into 50 915 contigs, with lengths ranging from 200 bp up to 39 839 bp. These contigs had a mean length of 1044 bp and a median length of 527 bp. However, binning of these contigs was unsuccessful, indicating that no complete genomes could be assembled. We then classified the assembled contigs using Kraken2 and carried out a Fisher’s exact test to see if any of the assigned microbes were enriched in ileal NET or adjacent normal ileal tissues. Results indicated that Enterobacter and Firmicutes/Bacillota were significantly enriched in ileal NET samples (Figure 4A) but were no longer significant after FDR correction (q = 1 and q = 1, respectively).

Figure 4.

Figure 4

Open in a new tab

Enrichment of assembled contigs in multifocal small intestinal neuroendocrine tumors (SI-NETs) and normal ileum samples. (A) Fisher’s exact test comparing prevalence of Kraken2-classified contigs in ileal NET and normal ileum samples. Each point corresponds to a taxon, with its prevalence in ileal NETs on the x-axis and normal ileum samples on the y-axis. The classified contigs are labeled by genera, or the next lowest taxonomic rank provided by Kraken2. Taxa that deviate significantly from the diagonal indicate potential enrichment in either ileal NET or normal ileal tissue. (B) Fisher’s exact test comparing prevalence of microbes in ileal NET and normal ileum samples. Each point corresponds to an unclassified contig, with its prevalence in ileal NETs on the x-axis and normal ileum samples on the y-axis. Contigs that deviate significantly from the diagonal indicate potential enrichment in either ileal NET or normal ileal tissue.

Unclassified sequence assemblies are non-microbial

A substantial fraction of the assembled contigs remained unclassified (39.9%), potentially representing unknown microorganisms. We therefore sought to determine if any of these contigs were preferentially enriched in ileal NET or normal ileum samples. Interestingly, we found that contig ‘megahit 5332’, was significantly enriched in ileal NETs, whereas three other contigs (megahit_36853, megahit_37156, and megahit_34479) were enriched in normal ileal tissues (Figure 4B). However, after a thorough check, most unclassified contigs aligned to mammalian genomes or various grain genomes, potentially representing residual human DNA sequences or environmental contamination. For example, the four significantly enriched contigs and the top five most prevalent contigs aligned to various Homo sapiens or Mus musculus genomes with a percent identity >90% and an e-value smaller than 1e-90, indicating that these sequences are unlikely derived from undiscovered microbes.

Discussion

The absence of recurrent genomic driver alterations in multifocal SI-NETs and their enrichment in the ileum have motivated the search for non-genetic mechanisms underlying the development of these lesions, as the identification of a cause for this idiopathic cancer could greatly improve early diagnosis and prevention of SI-NETs. Infectious agents are known to play an important role in the development of various cancers in the gastrointestinal tract, such as oral, gastric, and colorectal cancers.28 Therefore, the lack of a known mechanism for SI-NET pathogenesis prompted us to search for microbial sequences in whole genome and transcriptome sequencing data from a cohort of 10 multifocal SI-NET patients, including 70 independent multifocal tumor specimens, to characterize their tumor-associated microbiomes and investigate a potential role of microorganisms in SI-NET pathogenesis.

Microorganisms associated with cancers have been shown to differentially localize to tumor and adjacent normal tissues. For example, Fusobacterium is enriched in tumor samples of colorectal cancer patients,29 whereas Helicobacter pylori is enriched in tumor-adjacent normal tissues in gastric cancer patients.30 Regarding SI-NETs, microbes and microbial metabolites have previously been shown to influence hormonal secretion of colonic enterochromaffin cells in mice, suggesting that the microbiome could have an impact on SI-NET pathogenesis.31,32 In this study, we sought to evaluate the tumor-associated microbiome of SI-NET tissues. Although we did not identify any microorganisms enriched in SI-NETs, we found Propionibacterium to be significantly enriched in the normal ileal tissues. Most reads were derived from Propionibacterium acidifaciens and Propionibacterium freudenreichii species, one found in the human oral microbiome associated with dental caries and the other in the human gut microbiome.33,34 Propionibacterium freudenreichii is a known probiotic bacterium, which has been shown to adhere to intestinal epithelial cells and to release beneficial metabolites.34 Due to the small amount of normal ileum samples in the study, this finding should be interpreted with caution. Based on our metagenomic assembly analysis, it is unlikely that there are any unknown microbes involved in SI-NET tumorigenesis.

Our study revealed striking patient-specific microbial communities in the ileum of the multifocal SI-NET patients. Microbiome studies of fecal samples and oral swabs have long shown that the human microbiome is highly individualized.35,36 This appears to be the case for multifocal SI-NET patients as well. For example, the Turicibacter genus was highly abundant in patient 848 compared with all other patients. Bacteria from the Turicibacter genus are important members of the human gut microbiome and have been shown to modulate host bile and lipid compositions.37 The genus Limosilactobacillus, on the other hand, was highly specific for patient 876. Limosilactobacillus, which was split from the Lactobacillus genus in 2020,38 is a genomically diverse genus comprising strains that produce lactic acid and potential bacteriocins.39 Interestingly, we also observed an apparent gradient in the microbial composition of tumor samples taken along the length of the distal ileum in some of the multifocal ileal NET patients, potentially representing biogeographical variation in the small bowel. A recent study by Mulders et al. observed no differences in microbial composition or species abundance between fecal samples from patients with and without multifocal midgut NETs.40

Our study had some limitations. In addition to the small patient cohort, the depth of our WGS data limits detection to only the most abundant microbes associated with the multifocal ileal NETs, potentially overlooking less abundant microorganisms and underestimating total microbial diversity. Additionally, contamination of the microbiome data is a particular challenge with low-biomass samples, including human ileal tissue specimens.41 For example, before the removal of contaminants and false positives from our microbiome data, the raw microbial output indicated the presence of several bacterial genera, such as Pseudomonas, Corynebacterium, and Ralstonia, which have been identified as common contaminants in DNA extraction kits.42 Contaminants from sample processing and reagents can lead to subsequent erroneous attribution of microbes if not removed from the data. Additionally, microbial reference genomes can be contaminated with human sequences.43,44 Such sequence-contaminated reference genomes may align to human reads, erroneously labeling them as microbial. However, we are confident that our two-step filtering process successfully removed a large proportion of spurious low abundance microbial species from the data, resulting in credible small intestinal microbial genera.

Although others have previously studied the gut microbiome of SI-NET patients,40 to our knowledge, this is the first study profiling the tumor-associated microbiome of SI-NET patients. Our results suggest that the pathogenesis of multifocal SI-NETs is unlikely to be driven by the tumor-associated microbiome. None the less, we cannot fully eliminate the possibility of a pathogenic cause for SI-NETs, as a putatively causative pathogen could have already been cleared by the time of tissue sampling or even before disease presentation. To investigate microbes that are present at tumor initiation, a prospective microbiome study of at-risk individuals would be most appropriate. Together, our findings highlight the need for a deeper understanding of the molecular mechanisms that underlie this cancer type and encourage further research into the microbiome of the ileum.

Acknowledgments

Funding

This work was supported by the Neuroendocrine Tumor Research Foundation [grant number 628184].

Disclosure

MM declares research support from Bayer, Ono, and Janssen; patent licensing royalties from Bayer and LabCorp; and serving as a consultant for Delve Bio and Isabl; and holding equity in Bayer, Delve Bio and Isabl. All other authors have declared no conflicts of interest.

Data sharing

The WGS data analyzed in this study are accessible at the European Genome-phenome Archive website under the accession number EGAS00001006294 (https://ega-archive.org/studies/EGAS00001006294).

Supplemnetary data

Supplementary Figures
mmc1.pdf (516.7KB, pdf)
Supplementary Table 1
mmc2.xlsx (13.7KB, xlsx)

References

  • 1.Bilimoria K.Y., Bentrem D.J., Wayne J.D., Ko C.Y., Bennett C.L., Talamonti M.S. Small bowel cancer in the United States: changes in epidemiology, treatment, and survival over the last 20 years. Ann Surg. 2009;249:63–71. doi: 10.1097/SLA.0b013e31818e4641. [DOI] [PubMed] [Google Scholar]
  • 2.Modlin I.M., Champaneria M.C., Chan A.K., Kidd M. A three-decade analysis of 3,911 small intestinal neuroendocrine tumors: the rapid pace of no progress. Am J Gastroenterol. 2007;102(7):1464–1473. doi: 10.1111/j.1572-0241.2007.01185.x. [DOI] [PubMed] [Google Scholar]
  • 3.Yamaguchi T., Manabe N., Tanaka S., et al. Multiple carcinoid tumors of the ileum preoperatively diagnosed by enteroscopy with the double-balloon technique. Gastrointest Endosc. 2005;62(2):315–318. doi: 10.1016/s0016-5107(05)01569-5. [DOI] [PubMed] [Google Scholar]
  • 4.Gangi A., Siegel E., Barmparas G., et al. Multifocality in small bowel neuroendocrine tumors. J Gastrointest Surg. 2018;22(2):303–309. doi: 10.1007/s11605-017-3586-8. [DOI] [PubMed] [Google Scholar]
  • 5.Banck M.S., Kanwar R., Kulkarni A.A., et al. The genomic landscape of small intestine neuroendocrine tumors. J Clin Invest. 2013;123(6):2502–2508. doi: 10.1172/JCI67963. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Francis J.M., Kiezun A., Ramos A.H., et al. Somatic mutation of CDKN1B in small intestine neuroendocrine tumors. Nat Genet. 2013;45(12):1483–1486. doi: 10.1038/ng.2821. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Priestley P., Baber J., Lolkema M.P., et al. Pan-cancer whole-genome analyses of metastatic solid tumours. Nature. 2019;575(7781):210–216. doi: 10.1038/s41586-019-1689-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Löllgen R.M., Hessman O., Szabo E., Westin G., Akerström G. Chromosome 18 deletions are common events in classical midgut carcinoid tumors. Int J Cancer. 2001;92(6):812–815. doi: 10.1002/ijc.1276. [DOI] [PubMed] [Google Scholar]
  • 9.Kytölä S., Höög A., Nord B., et al. Comparative genomic hybridization identifies loss of 18q22-qter as an early and specific event in tumorigenesis of midgut carcinoids. Am J Pathol. 2001;158(5):1803–1808. doi: 10.1016/S0002-9440(10)64136-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Crona J., Gustavsson T., Norlén O., et al. Somatic mutations and genetic heterogeneity at the CDKN1B locus in small intestinal neuroendocrine tumors. Ann Surg Oncol. 2015;22(suppl 3):S1428–S1435. doi: 10.1245/s10434-014-4351-9. [DOI] [PubMed] [Google Scholar]
  • 11.Zhang Z., Mäkinen N., Kasai Y., et al. Patterns of chromosome 18 loss of heterozygosity in multifocal ileal neuroendocrine tumors. Genes Chromosomes Cancer. 2020;59(9):535–539. doi: 10.1002/gcc.22850. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Elias E., Ardalan A., Lindberg M., et al. Independent somatic evolution underlies clustered neuroendocrine tumors in the human small intestine. Nat Commun. 2021;12(1):6367. doi: 10.1038/s41467-021-26581-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Mäkinen N., Zhou M., Zhang Z., et al. Whole genome sequencing reveals the independent clonal origin of multifocal ileal neuroendocrine tumors. Genome Med. 2022;14(1):82. doi: 10.1186/s13073-022-01083-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Hayashi H., Takahashi R., Nishi T., Sakamoto M., Benno Y. Molecular analysis of jejunal, ileal, caecal and recto-sigmoidal human colonic microbiota using 16S rRNA gene libraries and terminal restriction fragment length polymorphism. J Med Microbiol. 2005;54(Pt 11):1093–1101. doi: 10.1099/jmm.0.45935-0. [DOI] [PubMed] [Google Scholar]
  • 15.Booijink C.C., El-Aidy S., Rajilić-Stojanović M., et al. High temporal and inter-individual variation detected in the human ileal microbiota. Environ Microbiol. 2010;12(12):3213–3227. doi: 10.1111/j.1462-2920.2010.02294.x. [DOI] [PubMed] [Google Scholar]
  • 16.Leite G.G.S., Weitsman S., Parodi G., et al. Mapping the segmental microbiomes in the human small bowel in comparison with stool: a REIMAGINE study. Dig Dis Sci. 2020;65(9):2595–2604. doi: 10.1007/s10620-020-06173-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Wang M., Ahrné S., Jeppsson B., Molin G. Comparison of bacterial diversity along the human intestinal tract by direct cloning and sequencing of 16S rRNA genes. FEMS Microbiol Ecol. 2005;54(2):219–231. doi: 10.1016/j.femsec.2005.03.012. [DOI] [PubMed] [Google Scholar]
  • 18.Ahmed S., Macfarlane G.T., Fite A., McBain A.J., Gilbert P., Macfarlane S. Mucosa-associated bacterial diversity in relation to human terminal ileum and colonic biopsy samples. Appl Environ Microbiol. 2007;73(22):7435–7442. doi: 10.1128/AEM.01143-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Juge N. Microbial adhesins to gastrointestinal mucus. Trends Microbiol. 2012;20(1):30–39. doi: 10.1016/j.tim.2011.10.001. [DOI] [PubMed] [Google Scholar]
  • 20.Walker M.A., Pedamallu C.S., Ojesina A.I., et al. GATK PathSeq: a customizable computational tool for the discovery and identification of microbial sequences in libraries from eukaryotic hosts. Bioinformatics. 2018;34(24):4287–4289. doi: 10.1093/bioinformatics/bty501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Wood D.E., Lu J., Langmead B. Improved metagenomic analysis with Kraken 2. Genome Biol. 2019;20(1):257. doi: 10.1186/s13059-019-1891-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Goldfarb T., Kodali V.K., Pujar S., et al. NCBI RefSeq: reference sequence standards through 25 years of curation and annotation. Nucleic Acids Res. 2025;53(D1):D243–D257. doi: 10.1093/nar/gkae1038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Blanco-Míguez A., Beghini F., Cumbo F., et al. Extending and improving metagenomic taxonomic profiling with uncharacterized species using MetaPhlAn 4. Nat Biotechnol. 2023;41(11):1633–1644. doi: 10.1038/s41587-023-01688-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Dohlman A.B., Arguijo Mendoza D., Ding S., et al. The cancer microbiome atlas: a pan-cancer comparative analysis to distinguish tissue-resident microbiota from contaminants. Cell Host Microbe. 2021;29(2):281–298.e5. doi: 10.1016/j.chom.2020.12.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Mallick H., Rahnavard A., McIver L.J., et al. Multivariable association discovery in population-scale meta-omics studies. PLoS Comput Biol. 2021;17(11) doi: 10.1371/journal.pcbi.1009442. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Lin H., Peddada S.D. Analysis of compositions of microbiomes with bias correction. Nat Commun. 2020;11(1):3514. doi: 10.1038/s41467-020-17041-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Tamames J., Puente-Sánchez F. SqueezeMeta, a highly portable, fully automatic metagenomic analysis pipeline. Front Microbiol. 2019;9:3349. doi: 10.3389/fmicb.2018.03349. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Cao Y., Xia H., Tan X., et al. Intratumoural microbiota: a new frontier in cancer development and therapy. Signal Transduct Target Ther. 2024;9(1):15. doi: 10.1038/s41392-023-01693-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Kostic A.D., Gevers D., Pedamallu C.S., et al. Genomic analysis identifies association of Fusobacterium with colorectal carcinoma. Genome Res. 2012;22(2):292–298. doi: 10.1101/gr.126573.111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Lehr K., Nikitina D., Vilchez-Vargas R., et al. Microbial composition of tumorous and adjacent gastric tissue is associated with prognosis of gastric cancer. Sci Rep. 2023;13(1):4640. doi: 10.1038/s41598-023-31740-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Yano J.M., Yu K., Donaldson G.P., et al. Indigenous bacteria from the gut microbiota regulate host serotonin biosynthesis. Cell. 2015;161(2):264–276. doi: 10.1016/j.cell.2015.02.047. [Erratum in: Cell. 2015;163:258] [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Bellono N.W., Bayrer J.R., Leitch D.B., et al. Enterochromaffin cells are gut chemosensors that couple to sensory neural pathways. Cell. 2017;170(1):185–198.e16. doi: 10.1016/j.cell.2017.05.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Obata J., Fujishima K., Nagata E., Oho T. Pathogenic mechanisms of cariogenic Propionibacterium acidifaciens. Arch Oral Biol. 2019;105:46–51. doi: 10.1016/j.archoralbio.2019.06.005. [DOI] [PubMed] [Google Scholar]
  • 34.Rabah H., Rosa do Carmo F.L., Jan G. Dairy propionibacteria: versatile probiotics. Microorganisms. 2017;5(2):24. doi: 10.3390/microorganisms5020024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Human Microbiome Project Consortium Structure, function and diversity of the healthy human microbiome. Nature. 2012;486(7402):207–214. doi: 10.1038/nature11234. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Zhou X., Shen X., Johnson J.S., et al. Longitudinal profiling of the microbiome at four body sites reveals core stability and individualized dynamics during health and disease. Cell Host Microbe. 2024;32(4):506–526.e9. doi: 10.1016/j.chom.2024.02.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Lynch J.B., Gonzalez E.L., Choy K., et al. Gut microbiota Turicibacter strains differentially modify bile acids and host lipids. Nat Commun. 2023;14(1):3669. doi: 10.1038/s41467-023-39403-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Zheng J., Wittouck S., Salvetti E., et al. A taxonomic note on the genus Lactobacillus: description of 23 novel genera, emended description of the genus Lactobacillus Beijerinck 1901, and union of Lactobacillaceae and Leuconostocaceae. Int J Syst Evol Microbiol. 2020;70(4):2782–2858. doi: 10.1099/ijsem.0.004107. [DOI] [PubMed] [Google Scholar]
  • 39.Ksiezarek M., Grosso F., Ribeiro T.G., Peixe L. Genomic diversity of genus Limosilactobacillus. Microb Genom. 2022;8(7) doi: 10.1099/mgen.0.000847. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Mulders M.C.F., Audhoe A.S., Van Koetsveld P.M., et al. Midgut neuroendocrine tumor patients have a depleted gut microbiome with a discriminative signature. Eur J Cancer. 2024;197 doi: 10.1016/j.ejca.2023.113472. [DOI] [PubMed] [Google Scholar]
  • 41.Karstens L., Asquith M., Davin S., et al. Controlling for contaminants in low-biomass 16S rRNA gene sequencing experiments. mSystems. 2019;4(4) doi: 10.1128/mSystems.00290-19. e00290-19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Salter S.J., Cox M.J., Turek E.M., et al. Reagent and laboratory contamination can critically impact sequence-based microbiome analyses. BMC Biol. 2014;12:87. doi: 10.1186/s12915-014-0087-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Breitwieser F.P., Pertea M., Zimin A.V., Salzberg S.L. Human contamination in bacterial genomes has created thousands of spurious proteins. Genome Res. 2019;29(6):954–960. doi: 10.1101/gr.245373.118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Steinegger M., Salzberg S.L. Terminating contamination: large-scale search identifies more than 2,000,000 contaminated entries in GenBank. Genome Biol. 2020;21(1):115. doi: 10.1186/s13059-020-02023-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figures
mmc1.pdf (516.7KB, pdf)
Supplementary Table 1
mmc2.xlsx (13.7KB, xlsx)

Articles from ESMO Gastrointestinal Oncology are provided here courtesy of Elsevier

RESOURCES