Gut microbial diversity of birds
To assess microbial diversity, we sequenced the V3-V4 regions of the 16S rRNA gene to identify 2,459 OTUs (operational taxonomic units) in 135 fecal samples from 43 species (37 wild bird species and 6 domestic poultry, Supplementary Fig. S1 and Supplementary Table 1). We first assessed the impact of general feeding habits on gut microbiome diversity, noting however, that the food types of omnivores vary widely (Supplementary Fig. S2 and Fig. 1a). When we grouped according to six classes of food types (fruits, corn-soy, grains, foliage, flesh, and omnivore) we found that 215 OTUs were shared by all groups. The food type with the highest number of unique OTUs was the fruit food group (271 OTUs), followed by the omnivore group (221 OTUs) (Fig. 1b). In contrast, the fewest unique OTUs were detected in the corn-soy (33 OTUs), grain (22 OTUs) and foliage (13 OTUs) food groups. Moreover, most OTUs (90%) were only detected in less than 20% of samples (Supplementary Fig. S3).
Next, we used Chao1, phylogenetic diversity (PD whole tree index) and Shannon index to illustrate bacterial richness and diversity within the communities based on the OTUs level. The alpha diversity index of the grain food group was the lowest, and significantly different from the fruit, corn-soy, flesh and omnivore food groups. For example, the grain group was significantly lower than the fruit group with the Chao1, PD and Shannon index. (FDR p < 0.05). No significant differences were observed among the other groups (Fig. 1c and Supplementary Table S2). A PCoA (principal coordinates analysis) analysis based on the Bray-Curtis distances was used to assess the differences in bacterial community structure between the samples. The results of this analysis revealed a significant clustering of gut microbiota by all diet groups, except the omnivore food group (p < 0.001), with food types separating the microbial communities along the first principal coordinate (PC1, 17.6% of variance) (Fig. 1d).
To further assess the effects of host phylogeny and diet on gut microbiome diversity, we compared the host phylogenetic tree with a UPGMA tree of the gut microbiota (Supplementary Fig. S4). Gut microbiomes of the different species mainly clustered based on food types. Furthermore, the gut microbiota of domestic poultry species fed with different food types (e.g., Gallus gallus, Meleagris gallopavo, Anas platyrhynchos and Cairina moschata) was diverse, and mainly clustered according to their food types (Supplementary Fig. S4). Based on the mantel test, both host diet (r = 0.1938, p-value = 0.0001) and phylogeny (r = 0.137, p-value = 0.0072) affect the gut microbiota of birds. MaAsLin2 analysis (adjusted p-value < 0.05) further revealed that diet plays a major role, and accounts for 92% of the microbiota features (Supplementary Fig. S5).
Predominant Bacteria Are Influenced By Dietary Differences
OTUs and genera were sparsely distributed in all samples (Supplementary Fig. S3). However, the predominant bacterial phyla present in the feces of all birds were Firmicutes (mean abundance ranged from 32.97–72.46%), Proteobacteria (12.21%~37.78%), Acitinobacteria (1.62% ~ 9.8%) and Bacteroidetes (0.07% ~ 14%) (Fig. 2a and Supplementary Table S3). At the genus level, the five most abundant genera were Lactobacillus, Clostridium, Enterococcus, Escherichia, and Turicibacter (Fig. 2b and Supplementary Table S3). Consistent with the α-diversity characteristic, the number of genera with mean relative abundance > 1% in the omnivore food group was higher than in the other groups.
Differences in abundance of the bacterial taxa was determined through a LEfSe analysis and a total of 28 taxa at different classification levels were found to have significant differnces (p < 0.05, LDA > 4) (Fig. 2c and Supplementary Fig S6). At the genus level, Lactobacillus, Leuconostoc, Clostridium and Cetobacterium were the dominant genera, and were significantly abundant in the grain, fruit, foliage and flesh food groups, respectively (p < 0.05). At the order level, a significant enrichment of Bacteroidales, was detected in the corn-soy food group. No significantly enriched taxa were observed in the omnivore food group.
Microbial Co-occurrence Association Patterns Are Influenced By Dietary Differences
We next examined how bacterial species co-occur among the birds, which might be due to dietary differences or microbe-microbe interactions. A network contained 344 nodes and 2559 edges was constructed (Fig. 3a), with the set of topological metrics including average degree, average weighted degree, average path degree, density and average clustering coefficient listed in Supplementary Table S4. Based on the layout structure, this integrated network could be divided into 6 sub-networks, each with differing taxonomic compositions at the class level (Fig. 3a). The taxonomic information represented by each node is listed in Supplementary Table S5.
At the class level, the co-occurrence network was mainly composed by interactions of Clostridia, Bacteroidia, Gammaproteobacteria and Bacilli. Furthermore, this network consisted of six sub-networks, each with differing taxonomic compositions. Subcommunity a (SC-a), subcommunity b (SC-b) and subcommunity c (SC-c) were dominated by Clostridia and Bacteroidia; subcommunity d (SC-d) and subcommunity e (SC-e) possessed more members of Actinobacteria, Gammaproteobacteria and Alphaproteobacteria; SC-f was mostly composed of taxa from Bacilli. In addition, most of the interactions were positive, while negative interactions only appeared between Pseudomonas (classified in Gammaproteobacteria) in SC-d with Ruminococcaceae UCG-014, and between Intestinimonas and Christensenellaceae R-7 group (classified in Clostridia) in SC-a (Fig. 3a), suggesting that competitive inhibition between these pairs of communities. Interestingly, we found Parasutterella was the only gram-negative genera that co-occurred with other gram-positive bacteria in SC-a.
We then calculated the co-occurrence percentage and total abundance of each sub-microbial community to estimate the microbial coexistence in the different groups. The presence and abundance of OTUs from each sub-network differed substantially among the groups (Fig. 3b). SC-d was generally the most prevalent in all birds, while wide differences in the prevalence and abundance of SC-a and SC-f occurred among the groups, suggesting that certain host dietary specificity of these microbial consortiums. The abundance of SC-a in the corn-soy group was significantly higher than in the other groups, and the abundance of SC-d in the corn-soy group was significantly lower than in the other groups (p < 0.05) (Supplementary Table S6). This phenomenon is consistent with antagonism between SC-a and SC-b as described above.
Adaptive Evolution Of Microbial Functions To Fit Food Types
To further investigate the functional capacities of the gut microbial communities in birds, a metagenomic analysis was conducted. In total, 2,425,998 assembled genes (92.02%, 2,636,348) were identified from the prokaryotic microbes and fungi by searches against the NCBI NR database. Of this total, 1,733,474 (65.75%) and 52,110 (1.98%) were annotated in the KEGG database and CAZy database, respectively.
Detailed annotation KOs information is listed in Supplementary Figure S7. Notably, 2,508 (28.47%) of the KOs were annotated in global and overview metabolism maps and 741 KOs were annotated in carbohydrate metabolism. The number of KOs with an average relative abundance higher than 0.01% in the different groups ranged from 1,329 to 2,672, and the total abundance of those KOs in each group was higher than 85% (Supplementary Fig. S8). This indicates that the high abundance KOs cover most of the microbial functions.
To compare the microbial functions between each group, we first tested the KEGG pathway enrichment analysis based on the top abundance KOs (mean relative abundance > 0.01%) in each group. The top 20 significantly enriched KEGG metabolism pathways in each group are shown in Fig. 4a. Only 8 pathways were shared among the 6 groups. Due to differences in host diet, the metabolic pathway enrichment for each group was different. For example, lipid metabolism, including glycerolipid metabolism and fatty acid biosynthesis, was enriched, while amino acid biosynthesis functions were not, in the corn-soy group. In addition, starch and sucrose metabolism were enriched in all groups except the flesh group, a group of birds that do not intake any plant-derived polysaccharide.
Moreover, we explored the distribution of CAZymes in the different groups. The top 20 abundant CAZymes in each group are shown in a z-score normalized heatmap (Fig. 4b). This data showed that most of the high abundance CAZymes were detected in the corn-soy, flesh and grain groups. Groups that have diets containing plant-derived fiber (fruit, omni and foliage) had similar enzyme profiles and clustered together.