Defining the biogeographical map and potential bacterial translocation of microbiome in human ‘surface organs’

Sample collection for microbiome profiling

We collected 1608 samples from 7 surface organs of oral cavity (6 sites), esophagus (4 sites), stomach (5 sites), small intestine (14 sites), appendix (1 site), large intestine (13 sites), and skin (10 sites), in total comprising of 53 sites in 33 subjects (Fig. 1A) who were dead due to vehicle accident, high-altitude falling, etc. (Supplementary Table 1). To minimize the post-mortem microbial changes, all samples were collected in a short duration (<1.5 h) after determination of death. Both luminal and mucosal samples were collected from the stomach, small intestine and large intestine. We parallelly introduced a set of negative controls to evaluate potential contamination (Supplementary Fig. 1A). All retrieved samples were subjected to microbial profiling by 16S v3v4 region sequencing, and samples from GI organs (n = 1030) were additionally analyzed by PacBio 16S full-length HiFi sequencing. After eliminating ASVs detected in negative controls (Supplementary Fig. 1B and Supplementary Table 2), we obtained a total of 9473 bacterial ASVs for downstream analysis (Fig. 1B). Key contaminating ASVs consist of environmental taxa (e.g., Propionibacterium (17.08%; relative abundance in negative controls), Phyllobacterium (6.12%), Deinococcus (4.87%)), and they were on average one order of magnitude higher than in mucosal samples as compared to luminal samples (Supplementary Table 2). We next applied permutational multivariate analysis of variance (PERMANOVA) to study the effect of subject’s characteristics (e.g., cause of death, length of hospitalization) on the microbiome communities. We found the length of hospitalization and antibiotic treatments had significant effects on microbiome in the oral cavity, small intestine, and large intestine, but not the cause of death (Supplementary Table 3).

A A total of 1608 samples from 53 body sites of 7 surface organs were collected from 33 subjects and were subjected to microbiome profiling. B The amount of detectable phylotypes in each organ at different taxonomic levels.

Microbial diversity varies among surface organs

We first investigated the bacterial diversity in surface organs. Significant differences in bacterial α-diversity were identified among surface organs (Fig. 2A and Supplementary Fig. 1C). The α-diversity of skin, oral cavity, and esophagus was significantly higher compared to stomach, appendix, small or large intestines, respectively (P < 0.01, Wilcoxon signed-rank test). Among seven organs, the bacteria diversity in stomach was the lowest, attributed to its low pH that limits bacterial growth. Significantly higher α-diversity in the large intestine was observed when compared to stomach or small intestine (P < 0.05).

A α-diversity of samples was grouped by organs (n = 328, 198, 110, 150, 363, 32, 427 for skin, oral cavity, esophagus, stomach, small intestine, appendix, and large intestine, respectively) and measured using the relative inverse Simpson index at the genus level. Boxplots were colored by surface organs. P values were determined using two-sided Wilcoxon signed-rank test. B α-diversity of 53 body sites in surface organs (sample size n was indicated in the button of each boxplot). Boxplots and trendlines were colored by sample types (surface or lumen). P values were determined using two-sided Wilcoxon signed-rank test. C β-diversity was measured using PCoA based on UniFrac distance. Each point (sample) was colored by its belonged organ. Community dissimilarities were tested by PERMANOVA analysis. Data are shown as Box and whisker plots (A, B) to represent the median (center line), quartiles (box), and range (whiskers) of the α-diversity for each community, excluding outliers (points outside 1.5 times the interquartile range). Source data are provided as a Source Data file.

Changes in α-diversity along the GI tract were then measured. In the upper GI tract (esophagus-stomach-duodenum), we observed that α-diversity initially falls in esophagus, reaching the bottom at the stomach, and subsequently rising at the duodenum (P < 0.05) (Fig. 2B). Meanwhile, significantly increasing trend of α-diversity was also identified in luminal samples along the lower GI tract (jejunum-Ileum-colon) (P < 0.05) (Fig. 2B), mainly attributed to the longer transit time in colon. When comparing α-diversity between mucosal and luminal samples, we observed disparities in α-diversity along the lower GI tract (Fig. 2B). Specifically, mucosal α-diversity was higher in jejunum/Ileum (P < 0.05) compared to luminal samples; while mucosal α-diversity was lower than that of luminal in the large intestine (P < 0.0001).

The global microbial β-diversity was also significantly different among organs (P < 0.001, PERMANOVA) (Fig. 2C). The most different microbiome was found between the large intestine and oral cavity, whilst microbiome between the stomach and esophagus was the least different (Supplementary Table 4). In the small intestine, we observed drastic intra-organ variation (Fig. 2C), and showed that intra-organ variation in the small intestine spans between stomach and large intestine clusters according to the sampling location (Supplementary Fig. 1D). Additionally, appendix microbiome was similar to the small intestine microbiome (Fig. 2C and Supplementary Table 4).

Inter-organ microbial communities are distinct

We next measured the inter-organ microbial composition. Six phyla (Proteobacteria, Firmicutes, Bacteroidetes, Actinobacteria, Fusobacteria, and Tenericutes) together occupied >98% relative abundance in each organ (Fig. 3A1 and Supplementary Fig. 2A). Abundances of these phyla were all significantly different among seven organs (Fig. 3A2). Bacteroidetes, Actinobacteria, and Fusobacteria were enriched in large intestine, skin, and oral cavity, respectively, while Proteobacteria and Firmicutes were enriched in esophagus, stomach, and small intestine. Microbial composition at genus level was then assessed (Fig. 3B). Bacteroides and Parabacteroides were predominantly enriched in small intestine, appendix, and large intestine; Porphyromonas, Prevotella, Streptococcus and Neisseria were enriched in the oral cavity; Fusobacterium was enriched in both oral cavity and appendix; and Staphylococcus, and Corynebacterium were the dominant genera in the skin. At the individual level, we observed a decreasing trend in the abundances of Staphylococcus and Corynebacterium from the skin to GI tract (Supplementary Fig. 2B). Conversely, increased abundances of Enterococcus, Ruminococcus and Bifidobacterium were observed along the GI tract. Moreover, Helicobacter was enriched in stomach and esophagus. These findings together suggested that microbial composition differs among surface organs.

A1 Abundances of six major phyla in seven organs: Proteobacteria (relative abundance: 41.31% ± 9.63%, Mean ± SD), Firmicutes (35.02% ± 8.36%), Bacteroidetes (14.10% ± 8.30%), Actinobacteria (6.21% ± 5.46%), Fusobacteria (1.65% ± 1.80%), and Tenericutes (0.37% ± 0.83%). A2 Phylum with significantly different abundance among seven organs by ANCOM-BC2 method (n = 328, 198, 110, 150, 363, 32, 427 for skin, oral cavity, esophagus, stomach, small intestine, appendix, and large intestine, respectively). B Genus with significantly different abundance among seven organs by ANCOM-BC2 method. The colormaps represents the average bacterial abundance. Data are shown as Box and whisker plots (A2) to represent the median (center line), quartiles (box), range (whiskers), and outliers (points outside 1.5 times the interquartile range). Source data are provided as a Source Data file.

Intra-organ microbial communities are heterogenous

As shown in Supplementary Fig. 2, microbes were not evenly distributed in each organ. We therefore investigated the microbiome of different intra-organ sites. β-diversity was significantly different among sites of each organ (Fig. 4). We identified the signature microbes specific to each site in an organ: Corynebacterium and Staphylococcus in the extremity cluster in skin (Fig. 4A); and Aggregatibacter in the jaws cluster of the oral cavity (Fig. 4B). In the esophagus, Helicobacter was increased from Thoracic Part (TP) to Cardiac orifice (CO), while its abundance was decreased from Fundus/Body to Pylorus (PY) in the stomach (Fig. 4C, D). In the small intestine, Prevotella were enriched in both the mucosa and lumen of duodenum, whereas Enterococcus and Bacteroides were enriched in both the mucosa and lumen of ileum (Fig. 4E, F). In the large intestine, we identified clear separation of microbiome between the right-sided and left-sided colon, attributed to the disparity in the enriched microbes (e.g., Klebsiella in the right-sided colon; Bifidobacterium and Oscillospira in the left-sided colon) (Fig. 4G, H). These data revealed the distinct microbial composition of intra-organ sites.

Intra-organ microbial communities were displayed in the left panel, measured using Constrained Correspondence Analysis (A skin; B oral cavity; C esophagus; D stomach; E mucosa of small intestine; F lumen of small intestine; G mucosa of large intestine; H lumen of large intestine and appendix). PERMANOVA analysis (adjusting for age, sex, BMI) was applied to test the significance of community dissimilarities. The arrow pointed to the direction of most rapid change towards the corresponding site. Differentially enriched microbes among intra-organ sites were displayed on the right panel. Selected microbes were colored based on their phyla. Source data are provided as a Source Data file.

Microbial community differences between lumen and mucosa

The availability of paired lumen-mucosa samples allowed us to investigate the microbial difference between the two sample types. Using 16S v3v4 dataset, mucosal microbial communities were all significantly different from luminal microbial communities in stomach, small intestine and large intestine (P < 0.0001 for all, PERMANOVA) (Figs. 2B and 5A). To decipher the microbial relationships between lumen and mucosa, we used logistic regression and identified 33, 52, and 47 mucosal/luminal-associated microbes in stomach, small intestine and large intestine, respectively. In the stomach, 60% (9/15) of mucosal-enriched genera were members of Firmicutes; whilst major gastric juice-enriched microbes belonged to Firmicutes (47%, 7/15) and Proteobacteria (47%, 7/15; e.g., Helicobacter) (Fig. 5B). In the small intestine, Firmicutes occupied 50% (19/38) of mucosal-enriched microbes (e.g., Coprococcus and Clostridium), whereas 43% (6/14) of luminal-enriched microbes belonged to Proteobacteria (Fig. 5C). Akkermansia and Bifidobacterium, two beneficial microbes in humans, were also enriched in the intestinal mucosa. In the large intestine, 81% (13/16) of mucosal-enriched microbes belonged to Firmicutes (Fig. 5D). Among luminal-enriched microbes, 42% (13/31) were members of Firmicutes, followed by Bacteroidetes (29%, 9/31). We then conducted similar analysis using 16S full-length dataset (Supplementary Fig. 3) in order to validate the above observations. We found that consistent lumen/mucosa-enriched bacteria were identified along the GI tract, including the stomach, small intestine, and large intestine (Supplementary Fig. 4). Moreover, in both small intestine and large intestine, we observed nine mucosal-enriched genera and seven luminal-enriched genera that are mucosal- and luminal-associated microbes (Fig. 5E), respectively.

A Microbial dissimilarities between mucosal and luminal samples of the stomach, small or large intestines, measured using Constrained Correspondence Analysis. PERMANOVA analysis was applied to test the significance of mucosal samples compared to luminal samples. Each point represented an individual sample and was colored by sample types (mucosal or luminal sample) and shaped according to its originated site of the surface organ. Significant mucosa-enriched and lumen/gastric juice-enriched microbes in different sites of (B) stomach, (C) small or (D) large intestines, measured using logistic regression model. Beta values represented the magnitude of difference in relative abundance between paired luminal and mucosal samples, and the degree of consistency among subjects. Points were colored by sites. Selected microbes (FDR < 0.05) were colored based on their belonged phyla. E Shared mucosa-enriched (left) or lumen-enriched (right) microbes between the small intestine and large intestine. Source data are provided as a Source Data file.

Functional capacities of microbiome differ among organs

Microbial functional attributes in surface organs were analyzed. Different pathways with significant enrichment were identified in each organ (Supplementary Fig. 5A and Supplementary Data 1): aerobic respiration in the skin; nucleoside and nucleotide biosynthesis/degradation (e.g., adenosine and guanosine) in the oral cavity; fatty acid metabolism (e.g., gondoate biosynthesis) in the esophagus, stomach, and small intestine; and pentose phosphate pathway including glucose/sugars catabolism in the appendix and large intestine. Comparative analysis of metabolic pathways revealed several carbohydrates degradation pathways that are significantly enriched in the small intestine (e.g., sucrose degradation) and large intestine (e.g., glycogen degradation of bacteria) (P < 0.05), respectively (Supplementary Fig. 5B). Amino acid synthesis (e.g., L-isoleucine, L-aspartate, L-histidine, and L-arginine) were significantly enriched in both the lower GI tract (appendix, small intestine and large intestine) and skin (P < 0.05) compared to other organs. Collectively, we revealed the differential microbial functional traits among surface organs.

Intra-organ microbial interaction network reflects organ specificity

To uncover microbial interplay in each organ, we calculated pairwise microbial interactions in each organ using SECOM method (Supplementary Fig. 6). We observed that each organ has its own patterns of microbial interactions (Supplementary Fig. 7 and Supplementary Data 2). Significantly different microbial interactions were observed among organs, with more co-exclusive relationships in oral cavity and large intestine, and more co-occurrent relationships in other organs (Supplementary Fig. 8A). We also used SPARCC method, which showed consistent findings (data not shown). Twenty-eight organ-specific microbial interactions was identified by both SECOM and SparCC (Supplementary Table 5), showing that microbial correlations were different among GI organs, for example, Bacteroides showed strong co-exclusive relationship with other microbes in the large intestine but strong co-occurrence with the same microbes in the upper GI organs (Supplementary Fig. 8B). These results implied that the microbiome in each organ habitat exhibits distinct microbe-environment relationships, suggestive of impact from host factors such as pH level and nutrient availability.

Microbial inter-organ relations exist in GI organs

By sampling a large set of intra-individual sites, we attempt to characterize the microbial inter-organ relations (i.e., bacterial translocation) along the GI tract. We re-sequenced the samples using 16S full-length sequencing, which provides higher taxon resolution than 16S v3v4 region. We then measured the presence of bacterial ASVs (the exact sequence variants; relative abundance >0.1%) in the intra-individual organs using 16S v3v4 and full-length data, respectively. The ASVs were collapsed to species level and the species prevalence among individuals were calculated. Consistent results between the full-length sequencing and v3v4 region sequencing were found (Fig. 6A and Supplementary Fig. 9A). We discovered that oral pathogens (prevalence >50%; e.g., Neisseria spp. and F. nucleatum) were less prevalent in the GI tract (<50%), especially the lower GI tract (Fig. 6A). We then applied correlation analysis to indicate the co-enrichment or co-depletion of bacteria in multiple organs. We observed fewer bacteria with positive correlations (P < 0.05) between oral cavity and lower GI organs than that between oral cavity and upper GI organs (esophagus and stomach) (Fig. 6B and Supplementary Fig. 9B). These suggest that the oral-to-lower GI contribution is limited (5.5% ± 3.95% of oral bacteria) in healthy individuals. Moreover, Bacteria with positive correlations were more distinguishable within upper or lower GI organs (e.g., esophagus and stomach: ratio = 0.53; SI and LI: ratio = 0.51) than between upper and lower GI organs (e.g., esophagus and LI: ratio = 0.13) (Fig. 6B), supporting the evidence for the restricted bacteria translocation from the upper GI to lower GI organs in healthy individuals.

A Bacterial prevalence in each organ by 16 S full-length sequencing. ASVs with relative abundance >0.1% (~10 sequencing reads) were considered as present on the organ. Red dot represents >50% prevalence. B The ratio of bacteria with positive correlations between each pair of organs among the prevalent bacteria (n = 76). The abundance of N. mucosa and F. nucleatum were plot in the right-side, with dash line links the same individual (P < 0.05, correlation analysis). Data are shown as Box and whisker plots to represent the median (center line), quartiles (box), range (whiskers), and outliers (points outside 1.5 times the interquartile range). Two-tailed Spearman correlation, Partial Spearman correlation, and two-tailed Pearson correlation were used simultaneously. C ASVs simultaneously present in the intra-individual upper GI (left) or intra-individual lower GI tract (right). Areas labeled in red represent the presence of ASV on all the organs from the same individual (relative abundance >0.1% for all). Light red color: one type of ASV of a particular species shared among organs from the same individuals. Dark red color: >1 ASVs of a particular species shared among organs from the same individuals. Bacterial prevalence in the oral cavity was displayed on left side of the plot. Source data are provided as a Source Data file.

Some high prevalent bacteria in an organ were also prevalent (>50%) in other organs as shown in Fig. 6A. We therefore asked if these bacteria were simultaneously present (relative abundance > 0.1%) in the upper GI or lower GI organs from the same individual (core microbial species, defined as species that coexisted in different organs of the same individual). Indeed, there were ASVs co-existed in all upper GI or lower GI organs intra-individually (Fig. 6C), which was independently verified by 16S v3v4 data (Supplementary Fig. 10A). On the other hand, unique bacterial signatures were found in the upper GI or lower GI tract. For example, S. salivarius and H. pylori in the upper GI, and Bacteroides spp. (e.g., B. vulgatus and B. caccae) and R. gnavus in the lower GI. Shared signatures including E. faecium, K. pneumoniae, and Enterobacteriaceae spp. (E. coli, E. flexneri, and E. sonnei) were also found between the upper GI and lower GI tract (Fig. 6C). Moreover, correlation analysis confirmed their inter-organ relations in the lower GI and upper GI organs, respectively (Supplementary Fig. 10B). Our result thus suggests a microbiome core with significant inter-organ relations co-existed in different organs of the intra-individual GI tract.