Faecal microbiome and metabolic signatures in rectal neuroendocrine tumors

Abstract

Background: The prevalence of rectal neuroendocrine tumors (RNET) has increased substantially over the past decades. Little is known on mechanistic alteration in the pathogenesis of such disease. We postulate that perturbations of human gut microbiome-metabolome interface influentially affect the development of RNET. The study aims to characterize the composition and function of faecal microbiome and metabolites in RNET individuals. Methods: We performed deep shotgun metagenomic sequencing and untargeted liquid chromatography-mass spectrometry (LC-MS) metabolomic profiling of faecal samples from the discovery cohort (18 RNET patients, 40 controls), and validated the microbiome and metabolite-based classifiers in an independent cohort (15 RNET participants, 19 controls). Results: We uncovered a dysbiotic gut ecological microenvironment in RNET patients, characterized by aberrant depletion and attenuated connection of microbial species, and abnormally aggregated lipids and lipid-like molecules. Functional characterization based on our in-house and Human Project Unified Metabolic Analysis Network 2 (HUMAnN2) pipelines further indicated a nutrient deficient gut microenvironment in RNET individuals, evidenced by diminished activities such as energy metabolism, vitamin biosynthesis and transportation. By integrating these data, we revealed 291 robust associations between representative differentially abundant taxonomic species and metabolites, indicating a tight interaction of gut microbiome with metabolites in RNET pathogenesis. Finally, we identified a cluster of gut microbiome and metabolite-based signatures, and replicated them in an independent cohort, showing accurate prediction of such neoplasm from healthy people. Conclusions: Our current study is the first to comprehensively characterize the perturbed interface of gut microbiome and metabolites in RNET patients, which may provide promising targets for microbiome-based diagnostics and therapies for this disorder.

Keywords: Gut microbiota; Metabolite; Rectal neuroendocrine tumor; Shotgun metagenomic sequencing; Untargeted liquid chromatography-mass spectrometry (LC-MS) metabolomics..

Figure 1

Faecal microbiome structure in RNET and control individuals. (A) We collected 92 stool samples from two cohort, the 58-case discovery cohort (40 healthy controls, 18 RNET patients) and the 34-case validation cohort (19 healthy controls, 15 RNET patients). Each faecal sample was subjected for deep shotgun metagenomic sequencing and LC-MS metabonomic profiling. (B-C) Microbiome community richness and Shannon index (α-diversity) at gene and species levels were measured using IGC (B) and MetaPhlAn2 (C) annotation, respectively. The dot represents one value from individual participants. Lines in the boxes indicate medians, the width of the notches is the IQR, the lowest and highest values within 1.5 times the IQR from the first and third quartiles. p values were calculated by two-sided Wilcoxon rank sum test. (D) Principal coordinates analysis (PCoA) with Bray-Curtis distance (β-diversity) of the discovery cohort based on gut metagenomic species profiles. p values were calculated by two-sided Wilcoxon rank sum test. ADONIS, R² = 0.017, p = 0.389.

Figure 2

Species-level changes in RNET microbiome community composition. (A) Relative abundance of microbial species showed significant difference between RNET patients and controls (p < 0.05, two-sided Wilcoxon rank-sum test). Boxes represent the IQRs between the first and third quartiles, and the line inside the box represents the median; whiskers represent the lowest or highest values within 1.5 times IQR from the first or third quartiles. (B) Co-occurrence (Orange) and co-excluding (Green) relationships between bacterial species in Control and RNET groups. FastSpar correlation coefficients were presented by edge width (roh < -0.2 or roh > 0.2, p < 0.05). Nodes' size (Control: blue; RNET: dark red) were scaled based on the relative abundance of each microorganism in either RNET or Control group.

Figure 3

Functional characterization of the RNET microbiome. (A) The top 15 Kyoto Encyclopedia of Genes and Genomes (KEGG) modules annotated by our in-house pipeline were differentially abundant either in control or RNET groups (LDA > 2.0, p < 0.05). Modules overlapped with those annotated by the HUMAnN2 pipeline were marked with red asterisk (*). (B) Dominant taxonomic contributors for 5 KEGG modules simultaneously enriched by our in-house and HUMAnN2 pipelines, including M00125, M00157, M00164, M00183 and M00319 were predicted using the HUMAnN2 analysis. Species are proportionally scaled within the total bar height.

Figure 4

Metabolite enrichments in patients with RNET. (A) PCoA with Bray-Curtis distance in RNET and control groups based on faecal metabolic profiles. p values were calculated by two-sided Wilcoxon rank sum test. ADONIS, R² = 0.083, p = 0.001. (B) Faecal metabolites (Log₂ FC >1 or < -1, VIP > 1.0, p < 0.05, two-sided Wilcoxon rank-sum test) differentially abundant in RNET or healthy individuals were presented as lollipop chart. The length of line indicated the relative abundance of metabolites detected by LC-MS metabolomics. VIP score was indicated by a color gradient from purple (small numerical value) to yellow (large numerical value). (C) Bar plots of the top 15 KEGG pathways based on the In (p value) that biologically enriched from RNET-specific metabolites as compared with control subjects.

Figure 5

Putative correlation of gut microbial species with metabolites in RNET patients. Heatmap of Spearman's rank correlation coefficients between differentially abundant species (LEfSe: LDA > 2.0, p < 0.05) and metabolites (Log₂ FC > 1 or < -1, VIP > 1.0, p < 0.05, q < 0.05) from RNET patients and controls. Correlation coefficients in each square represent positive (red) and negative (green) relationships. Statistically significant correlations (p < 0.05) were marked with asterisks (*). Representatively negative and positive correlations were labeled as deep blue squares.

Figure 6

Multi-omics signatures-based predication of RNET. (A) 12 faecal microbial and metabolic biomarkers from RNET and control individuals reached the lowest classifier error were obtained by the mean decrease accuracy tool from the random forests (RFs) algorithm, and ranked by their contributions to classification accuracy after permutation. The color of each biomarker indicates its enrichment in RNET (purple) or control (green) participants. (B-C) Receiver operating characteristic (ROC) curve of the RF model using discriminatory signatures (3 species, 9 metabolites) in the 58 samples of discovery cohort (B) or 34 samples of validation cohort (C). RF method was used with train function of R's caret package. For training set, five-fold cross-validation was applied with trainControl function. To compute and visualize AUC from ROC outcome, the pROC package was utilized.