Bacterial features in the different mouse GI segments
A total of 3,550,520 valid reads with a mean number of 26,496 tags for each sample were obtained from 68 samples of 12 mice (Table S1). All the sample rarefaction curves tended to approach the saturation plateau and finally, 15,180 OTUs were obtained by using 97% homology cut-off value (Fig. S1, Table S1). By investigating the microbial α diversity (observed species, Chao index, Ace index and Shannon index) of different GI segments, we found that it changed following a “U” shape (Fig. 1a, Fig. S2). Generally, the α diversity significantly lower in jejunal and ileal samples than that from the stomach, cecum, colon, and feces (Wilcoxon test, P < 0.01).
We next compared β diversity of microbes in the six parts of the GI tract according to the Bray-Curtis similarity of the genus. Bacteria displayed volatility in upper and lower GI tract (Fig. 1b). The lower GI tract including cecum, colon and feces shared similar microbial composition (Bray-Curtis similarity ≥ 0.8) and the same pattern was also observed in upper GI bacteria (Bray-Curtis similarity ≥ 0.6). However, the similarity was decreased when it was calculated between the upper and lower GI tract (Bray-Curtis similarity ≤ 0.6).
Moreover, there were 350 unique core OTUs (existing in more than 50% samples per GI segment) were observed in the whole GI tract. 276, 107, 106, 247, 274 and 251 core OTUs were found, respectively, in content from stomach, jejunum, ileum, cecum, colon and feces (Fig. 1c, red bars). About 22.9% core OTUs (n=80) were shared by the whole GI tract, and 40% (n=140) core OTUs were shared by lower GI tract and stomach, but not by jejunum and ileum (Fig. 1c, black bars).
To investigate how much GI microbiota could be represented by feces, we explored the Jaccard distances of OTUs between in fecal and the other segments (Fig. 1d). Feces shared about 82.7% OTUs with the lower GI tract, while it shared about 65.2% OTUs with the upper GI tract (Wilcoxon test, P < 0.001). The most similar GI segments with feces of microbial composition were colon who shared 88.74% OTUs, followed by cecum (76.7%), stomach (69.9%), jejunum (65.5%) and ileum (60.3%).
Finally, we explored the microbial compositional differences among GI segments. Taxonomically, 14 phyla were observed. The most abundant one was Firmicutes (51.27%) followed by Bacteroidetes (35.18%), Proteobacteria (9.17%), Actinobacteria (1.87%) and the others (Fig. S3a). Of all the phyla, only Cyanobacteria showed significant difference among the six parts of the GI tract (Kruskal-Wallis test, P < 0.001). Its abundance was lower in jejunal and ileal than the other GI segments in mice of all ages (Fig. S3b). It is worth noting that phylum variation in different GI tract was affected by aging and high-fat diet intervention. For example, we found that the older mice and the high-fat feeding ones shared the same trend in Firmicutes variation among the six GI segments: the Firmicutes abundance in jejunal and ileal was higher than the other parts, indicating that the Firmicute might be less susceptible to the jejunal environment when the mice became older or fat (Fig. 2a). However, our results showed that Bacteroidetes and Proteobacteria revealed a antagonistic trend with Firmicute among the whole GI tract in all the biological groups. It is different with Cyanobacteria whose change rule was similar in various age groups. The relative abundance of Cyanobacteria was increased from jejunal to colon (Fig. S3c). In genus level, there were 19 genera that were differently composed in among the six GI tracts according to the Kruskal-Wallis test with the cutoff P value < 0.05 (Table S2). They were clustered into two groups (Cluster1 and Cluster2) in first of which Sutterella, Aggregatibacter, Lactococcus, Lactobacillus and Streptococus were significantly enriched in the upper GI tract, while anaerobes such as Ruminococcus were significantly enriched in the lower GI tract in Cluster2 (Fig. 2c).
The bacterial community was significantly different in the elderly mice.
Four α diversity indexes including the number of observed species, Chao index, ACE index and Shannon index were compared among three age groups (Fig. 3a and Fig. S4). The observed species, Chao index and ACE index, were significantly declined in elderly mice, but this trend was only observed in the lower GI tract (Student's t-test, P < 0.05). This result might indicate that lower GI microbes were more likely to be affected by ageing, compared with the upper GI bacteria.
The whole GI microbiota composition changed significantly with ageing according to PERMANOVA analysis (permutations = 999, method=" bray", P < 0.001). When OTUs based Bray-Curtis distance was constructed to compare the GI microbial β diversity between the young-aged, middle-aged and the elderly mice, it was found that the distances between the elderly group and non-elderly group (the young-aged and middle-aged group) were larger than that between the young-aged and middle-aged group (Wilcoxon test, P < 0.001). However, microbial composition in the stomach had no significant differences among the three age groups (Fig. 3b).
To further investigate the taxonomic changes with age, we compared the microbial relative abundance between the three age groups in both phylum level and genus level. Key taxon selection was performed according to Kruskal-Wallis rank test, followed by a multiple linear regression model in which the age-group was treated as an ordinal categorical variable to enhance its robustness (Table S3, Table S4). Eventually, 5 phyla and 20 genera were observed to change marginally (Kruskal-Wallis rank test, P < 0.1) from young-aged groups to the elderly mice (Table S3, Fig. 3c). There were 3 genera, namely Allobactulum, Sutterella Lactobacillus, that clearly changed with ageing in all the GI segments. Their relative abundance was higher in upper GI tract, especially in jejunum and ileum than it in the lower GI tract (Fig. 3d). Some of ageing associated changes were also affected by GI segments. For instance, the increment of Firmicute in elderly as well as the depletion of both Proteobacteria and Bacteroidetes was more likely to be observed in ileum and jejunum than that in the other GI segments (Fig. 3c). But Coprococcus decrement and Flexispira increment in elderly mice were only observed in the cecum, colon and feces (Fig. 3d). Similarly, the abundance of Streptococcus was higher in elderly mice than that in the other age groups and it was only observed in the upper GI tract (Fig. 3d).
GI microbiota gradually changed with age when it was compared to high-fat diet treatment.
It has been reported that there is no chronological threshold or age at which the composition of the microbiota suddenly alters [16], thus, it was necessary to understand how much differences it would make separately by ageing and by other conditions such as lifestyle, disease and diet. Here, we choose high-fat diet mice as a reference group to compare the diet cased GI changes with the ones made by ageing. Cluster as well as principle coordination analysis (PCoA), based on Bray-Curtis distance of OTUs, showed that high-fat feeding mice were clustered separately from the other groups (Fig. 4ab). The intro-dissimilarity which calculated between each two of normal feeding mice was significantly lower than the inter-distance which calculated between high-fat feeding mice and normal feeding mice (Fig. 4c, Wilcoxon test, P < 0.001). In addition, we also constructed a KNN model with genus composition to classify biological groups. It was more difficult to pick one of the three age-related groups (Ycon, Mcon and Econ) than the high-fat feeding ones (Mhfd), because the accuracy to separate high-fat feeding mice from the others reached 0.94, whereas the highest accuracy to classify the three age-related groups were only 0.65 (Fig. S5). To sum up, these results suggested the healthy aging-related microbial changes were moderate, and it was not as intense as the diet/disease associated ones.
We then investigated the high-fat feeding associated genera in the GI tract (Table S4). There are 22 marginally different genera were found between the high-fat feeding and normal feeding groups (Wilcoxon test, P < 0.1). Most genera were non-overlapped with the age-related ones, and the overlapped 4 genera including AF12, Allobaculum, Bifidobacterium, Bilophila, Butyricicoccus were independently associated with different GI segments (Fig. 4d).