Turtle growth and morphology under different habitats
Mortality was negligible in both paddy fields and ponds during the experimental period. However, only a small number of turtles were caught from the lake at 60 d, and no marked turtles were recaptured at 120 d, resulting in incomplete statistics for mortality and growth for turtles in the lake. No wild turtles were caught during sampling. A relatively small sample size (n=3) for each group was designed for turtle resource protection purposes, and it was difficult to sample from natural lakes. There were significant differences in growth among the different groups; the body weights of turtles in paddy fields and ponds were obviously higher than those in lakes (p<0.05), and divergence occurred in the early days. The growth rates of turtles were 0.76%/d, 0.68%/d and 0.40%/d for those from paddy fields, ponds and lakes, respectively, in the first 60 d. The rate was 0.72%/d and 0.62%/d for turtles from paddy fields and ponds, respectively, during the whole 120 d. The hepato-somatic index and clumpy fat index were highest in ponds, second in paddy fields and lowest in lakes (p<0.05). The gut-somatic index of weight (DSIW) for turtles from ponds was significantly higher than that for turtles from lakes and paddy fields (p<0.05). Conversely, the gut-somatic index of length (DSIL) was higher for turtles from paddy fields and lakes than those from ponds. Measured values are presented as the mean ± standard deviation, and the different superscript letters in the same row indicate significant differences (p<0.05) (Table 1).
There was no obvious trauma experienced by most turtles from the lake except occasional parasitic leeches observed on the calipash. However, more bruises or scars were observed for the turtles from ponds than those from paddy fields. The appearance, such as the color, of the carapace and plastron were different among turtles from different habitats. The carapace of turtles cultured in ponds presented a bottle green color, but the individuals from paddy fields presented a bottle green color with a slight golden yellow color, which was similar to turtles from lakes. There were significant differences in carapace width/carapace length (CW/CL) and calipash lateral width/carapace length (CLW/CL) values at 60 d and 120 d (p>0.05), but the CW/CL value was relatively higher for turtles from lakes, and CLW/CL was higher for turtles from lakes and ponds than for those from paddy fields (p<0.05) (Table 1).
Composition and diversity of turtle gut microbiota
Dominant microbesThe grouping details for samples from different habitats, culture days and intestinal segments are listed in Table 2. For gut samples, a total of 1 723 158 valid bacterial 16S rRNA gene reads were obtained, and 4 901 OTUs were identified from all samples. The observed total OTUs varied from 64~822. The total number of OTUs was significantly lower in initial groups IF and IL and higher in groups F1F and F1L from paddy fields at 60 d. The number of OTUs was 17~48, representing more than 0.01% of the total OTUs (Table S1). Significant differences were found in OTU composition among groups (Fig S1). Guts sampled at 120 d had few unique OTUs, both in the foregut and hindgut. The alpha diversity was calculated according to the composition and relative abundance of total OTUs. Generally, the alpha diversity indices of microbes in the hindgut were higher than those in the foregut. In addition, the alpha diversity was obviously lower in initial turtle guts from greenhouses (p<0.05), whereas it was obviously higher in turtle guts sampled from paddy fields than ponds and lakes (Fig. 1). The species and number of OTUs varied significantly at 60 d, different from the relatively similar results across groups obtained at 120 d. The microbial abundance was higher in samples from paddy fields than in samples from lakes and ponds during the experiment. The microbial community presented relatively high similarity in guts sampled at the same time. The PCoA (principal coordinate analysis) of the Bray-Curtis dissimilarity showed high microbial community similarity in guts from the same individual or group and significant discrepancy in samples from different habitats, sampling times and gut sections (Fig. 2). Generally, both sampling time and habitat affected the variation in the gut microbial communities.
The recognized microbes belonged to 27 phyla, 59 classes, 97 orders, 151 families, and 219 genera from all the samples based on GreenGene. The phylum and genus levels were emphasized in the analysis. Bacteroidetes, Firmicutes, Fusobacteria and Proteobacteria were the most dominant phyla, accounting for more than 95% of the total bacteria in all samples. Firmicutes was the most abundant phylum in the guts of turtles sampled from the greenhouse initially, while Proteobacteria was the most abundant phylum after cultivation in different habitats, followed by Bacteroidetes. Firmicutes and Fusobacteria commonly existed at 60 d but were rarely present at 120 d in turtles from all three habitats (Fig. 3a). Additionally, the unidentified bacteria were more abundant in turtles from lakes than those from paddy fields and ponds.
There was a significant difference in dominant genera among initial samples and subsequent samples from different habitats. The dominant genera in the initial samples were an unclassified genus belonging to Bacteroidales, Romboutsia, Cetobacterium, Weissella, Lactococcus, Lactobacillus, Clostridium, Edwardsiella, Plesiomonas, and Sarcina. For samples from the three habitats mentioned above, the dominant genera were Cetobacterium, Chryseobacterium, Clostridium, Epulopiscium, Flavobacterium, Helicobacter, Pseudomonas, Stenotrophomonas and another unclassified genus belonging to Xanthomonadaceae. The abundance of dominant genera varied with habitat, sampling time and gut location. For turtles sampled from paddy fields, the most dominant genus in foregut samples taken at 60 d was Clostridium, and in the hindgut, it was Cetobacterium, while at 120 d, the most dominant genus was Stenotrophomonas both in the foregut and hindgut. For turtles sampled from ponds, the most dominant genera at 60 d were Flavobacterium and Cetobacterium in the foregut and hindgut, while at 120 d, the most dominant genus was also Stenotrophomonas. For turtles sampled from the lake, the most dominant genera at 60 d were Flavobacterium and Cetobacterium in the foregut and hindgut, respectively (Fig. 3b).
The dominant species in different gut locations were also distinct. In the foregut, the dominant species were Weissella cibaria, Enterococcus durans, Lactobacillus sakei, Lactococcus lactis, Lactococcus garvieae, Sarcina sp. and Pseudomonas sp., whereas in the hindgut, Clostridium sensu stricto, Romboutsia sp., Weissella cibaria, Escherichia coli, Plesiomonas shigelloides, Edwardsiella tarda, Paeniclostridium sp., Cetobacterium sp., Terrisporobacter sp. and two other unclassified species belonging to Bacteroidales were the most abundant.
Microbial communities in turtles from different habitats and at different sampling times
The microbial community was relatively complex at 60 d, especially in the foregut. At 60 d, the species of microbes were significantly more abundant in turtles from the fields, followed by those from ponds and lakes. There were 140 common species (8.2%) in the foreguts of turtles from the three different habitats (Fig. 4a); Flavobacterium sp., Pseudomonas sp., Chryseobacterium sp. and two species belonging to Xanthomonadaceae were relatively abundant. Cetobacterium somerae was more abundant in turtles from paddy fields than in those from ponds and lakes. For the hindgut, there were 205 common species (8.1%) in turtles from the three different habitats (Fig. 4b). Among these, one species belonging to Bacteroidaceae was abundant in all habitats. Cetobacterium somerae, Epulopiscium sp., Pseudomonas sp., Stenotrophomonas sp. and Flavobacterium sp. were more abundant in turtles from paddy fields and lakes than in ponds, while Clostridium sp. and Epulopiscium sp. were relatively abundant in specimens from ponds. Moreover, Chryseobacterium sp., Parabacteroides sp., Sphingobacterium faecium, Clostridium perfringens, Pseudomonas sp., Bacteroides sp. and Pseudomonas sp. commonly existed in samples from lakes and paddy fields but did not appear in pond samples. At 120 d, specific foregut microbes were more abundant in pond turtles (74%) than paddy field turtles (33.4%), and the common species accounted for 18.6%; for the hindgut, specific microbes were more abundant in paddy field turtles (44%) than pond turtles (34.4%), and the common species accounted for 26.1% (Fig. S2).
LEfSe analysis was also conducted to identify representative microbes among various groups. For the initial groups, representative genera were Weissella, Cetobacterium, Chryseobacterium, Epulopiscium, Escherichia, Flavobacterium, Lactococcus, Leuconostoc, Plesiomonas, Romboutsia, Sarcina and Stenotrophomonas. For groups cultured in different habitats, F1L contained more different species than the other groups, including members of Cetobacterium, Lactobacillaceae, Bacteroides, Parabacteroides, Plesiomonas, and several species belonging to the phylum Firmicutes presented higher LDA scores than those of the other groups. For F1F, the representative taxa were Sutterella, Bacteroides and Clostridiales. For samples from the lake, Xanthomonadaceae and Pseudomonadales were representative taxa, especially at 60 d. The representative microbes in pond turtles were numerous and belonged to various phyla, especially the phylum Proteobacteria, and some unassigned species were found turtles from this habitat (Fig. 5).
Functional predictions
The nearest sequenced taxon index (NSTI) was developed to quantify the availability of nearby genome representatives for groups (Table S2). In total, 41 predicted functional categories that represented 7 pathway maps in KEGG level 2 were indicated by PICRUSt (Fig. 6), including 330 functions on level 3 (Fig.S3). Culture period and habitats had significant effects on metabolism such as amino acid and carbohydrate metabolism, environmental and genetic information processing such as membrane transport, replication and repair. At 60 d, the functional microbiota related to amino acid and carbohydrate metabolism was distinct higher in lake samples compared to those from ponds and paddy fields(Fig. 6, Fig.S3).