Effects of captivity and artificial breeding on microbiota in feces of the red-crowned crane (Grus japonensis)

Reintroduction of the threatened red-crowned crane has been unsuccessful. Although gut microbiota correlates with host health, there is little information on gut microbiota of cranes under different conservation strategies. The study examined effects of captivity, artificial breeding and life stage on gut microbiota of red-crown cranes. The gut microbiotas of wild, captive adolescent, captive adult, artificially bred adolescent and artificially bred adult cranes were characterized by next-generation sequencing of 16S rRNA gene amplicons. The gut microbiotas were dominated by three phyla: Firmicutes (62.9%), Proteobacteria (29.9%) and Fusobacteria (9.6%). Bacilli dominated the ‘core’ community consisting of 198 operational taxonomic units (OTUs). Both captivity and artificial breeding influenced the structures and diversities microbiota of the gut. Especially, wild cranes had distinct compositions of gut microbiota from captive and artificially bred cranes. The greatest alpha diversity was found in captive cranes, while wild cranes had the least. According to the results of ordination analysis, influences of captivity and artificial breeding were greater than that of life stage. Overall, captivity and artificial breeding influenced the gut microbiota, potentially due to changes in diet, vaccination, antibiotics and living conditions. Metagenomics can serve as a supplementary non-invasive screening tool for disease control.

Microbiota in the gut plays important roles in nutrition 17,18 , development of organs [19][20][21] and regulation of host physiology 22,23 . Both captivity and artificial breeding can influence microbial communities in guts of wildlife due to vaccination against diseases, use of antibiotics to control diseases and changes in diets and living conditions [24][25][26] . However, the effects of captivity and artificial breeding on the gut microbiota of G. japonensis are largely unknown.
The present study was conducted to examine effects of captive breeding on gut microbiota of the threatened red-crowned cranes (G. japonensis). Using next-generation sequencing of 16s rRNA gene with non-invasive collection of feces, diversities of microbes in guts of cranes from three groups, wild, captive and artificially bred, were investigated. Both captive and artificially bred groupes were further divided into two life stages: adolescent and adult. Because ages of wilds crane could not be determined in the field by use of a non-invasive approach, the wild group was excluded from the part of the study that examined effects of life stages. Microbial communities in guts were assessed in five groups of G. japonensis, including wild, captive adolescent, captive adult, artificially bred adolescent and artificially bred adult. Artificially bred cranes were the second generation of captive adults (first generation). Specific research questions included: 1) to examine the structures of microbial communities in guts of G. japonensis; 2) to evaluate the effects of captivity, artificial breeding and/or host life stage on the gut microbiota of cranes.

Results
Next generation sequencing data. After checking qualities of sequenced reads, 2,135,253 cleaned sequences with an average length of 212 ± 13 bp (Supplementary Table S1) were obtained from 108 fecal samples. The number of sequences in each sample varied from 8,030 to 108,290 (median = 13,091). A total of 1,164 OTUs representing 25 bacterial phyla were identified in feces of G. japonensis. Diversities and structures of microbial communities were compared among five groups, including artificially bred adolescents (n = 23), artificially bred adults (n = 30), captive adolescents (n = 16), captive adults (n = 17) and wild cranes (n = 22) (Supplementary Table S2). Both Chao1 and phylogenetic diversity were in accordance with the observed number of OTUs (Supplementary Figure S1 and Table S1). According to rarefaction curves (Supplementary Figure S1) and Good's coverage estimator (Supplementary Table S1), most of the abundant microbial OTUs in guts of cranes were captured by rarefaction at 8,030 reads per sample.
Microbial communities in feces of G. japonensis. Predominant bacterial phyla present in feces of all G. japonensis included Firmicutes (62.9 ± 4.8%), Proteobacteria (29.9 ± 4.7%) and Fusobacteria (9.6 ± 3.0%) (Fig. 1 Association network among host-associate core microbiota in feces of G. japonensis. There were 198 OTUs shared by more than half of all samples (n = 54). These shared OTUs of fecal microbiota composed host associated 'core microbiota' , which were mainly within the Gammaproteobacteria, Bacilli and Clostridia. Core microbiota accounted for 86.1% of the microbiota in feces, and members of the class Bacilli dominated the core microbiota by accounting for 51.9% of the microbiota and 60.3% of the core microbiota. The relationship within core microbiota, co-occurrence and co-exclusion relationships among the core OTUs were inferred from the association network (Supplementary Figure S2). There were 1,135 statistically significant associations among core OTUs ('two-tailed' bootstrapped P value < 0.002). The topology of the network suggested that this network had a modular structure (modularity index, 0.48 > 0.4) 27 . The network comprised highly connected OTUs (≥5 edges per node, 62% of the total nodes) that can form a clustered topology. Core OTUs from the same order tended to co-occur more frequently (contributing 74.5% of the total associations) than those in different orders (contributing 25.5% of the total associations, Supplementary Figure S2). Association networks were collapsed at the genus level to identify key network hubs that modulated the gut microbiota. Bacillus, Brochothrix, Acinetobacter, Pseudomonas, Lactococcus, Lactobacillus and Enterococcus were defined as key hubs (highly connected genera, ≥5 genera, Fig. 2). Campylobacter, Enterococcus and Lactobacillus were mutually exclusive with other key hubs. Among the Bacilli, Lactobacillus and Enterococcus were mutually exclusive with Bacillus and Brochothrix.
Microbial communities among feces of artificial bred, captive and wild groups were distinct.
Both captivity and artificial breeding strongly influenced structures of fecal microbiota from three groups of G. japonensis (Fig. 3). Structures of fecal microbiota were also compared among artificially bred, captive and wild groups by use of the permutational multivariate analysis of variance test with unweighted UniFrac distance (PERMANOVA, Table 1). Pairwise tests revealed that compositions of microbial communities in guts of cranes from each group were each significantly different from the two other groups ( Table 2). Significant differences were also observed in both Shannon index and phylogenetic diversity among groups (Kruskal-Wallis test, P < 0.05, followed by post-hoc Mann-Whitney U tests among groups, Fig. 4). The greatest alpha diversity was found in the captive group, while the wild group had the least. Patterns of classified abundant genera (max relative abundance > 1%) varied among groups (ANOVA test, P value < 0.05, followed by the post-hoc Tukey-Kramer test among groups. P values were corrected using the Benjamin-Hochberg FDR method, Fig. 5). Twenty-five genera with significantly different relative abundances among groups were identified. Most genera belonged to one of 7 classes (Actinobacteria, Flavobacteria, Bacilli, Clostridia, Gammaproteobacteria, Fusobacteria and Epsilonproteobacteria). The relative abundances of key hubs (Bacillus, Acinetobacter, Pseudomonas, Lactococcus, Lactobacillus and Enterococcus) were significantly influenced by captivity and artificial breeding, as well as the Campylobacter.
Composition of microbial communities differed among life stages in each artificially bred or captive group. Structures of microbial communities in feces displayed significant separation between life stages in both captive and artificially bred groups (Fig. 3). PERMANOVA results (Table 1) were used as an indicator of true differences and used to classify distinct groups. Compositions of microbial communities for each sub-group were significantly different from those of the three other sub-groups ( Table 2, Artificial breeding × Life stage). Both artificially bred and captive adolescents exhibited significantly greater alpha diversities, than did adults (Fig. 4). Gut microbiota of adults contained significantly greater relative abundances of the genera Salegentibacter and Salinimicrobium. Abundances of the genera Planococcus, Turicibacter, Cellvibrio and  Pseudoalteromonas were less (Fig. 5). During each life stage, artificial breeding influenced phylogenetic diversity, relative abundance of five key hubs (Bacillus, Acinetobacter, Pseudomonas, Lactococcus and Enterococcus) and structures of the gut microbiota.
Opportunistic zoonotic bacteria in feces identified by non-invasive next-generation sequencing of 16S rRNA gene amplicon. There were 23 opportunistic enteric zoonotic pathogenic OTUs (Supplementary Table S3) representative of four classes of enteric zoonotic bacteria ( Table 3) in microbiota of feces of G. japonensis. In this study, the V4 fragment of 16s rRNA was not sufficiently long to classify Campylobacter spp. and Mycoplasma spp. to species. Not all species within Campylobacter are pathogens. Incidences and relative abundances of Campylobacter spp. and Clostridium piliforme in feces of domestic cranes were significantly greater than those of wild cranes. Captive cranes harbored the greatest relative abundance of Clostridium perfringens. The community of microbes in feces can serve as a supplementary, non-invasive tool to screen for disease control, via 16s rRNA amplicons and next-generation sequencing.

Discussion
A survey of bacterial communities in feces of G. japonensis provides evidence that captivity and artificial breeding significantly altered structures of the gut microbiota. Microbiota in guts of G. japonensis was dominated by Firmicutes (62.9%) which is consistent with results of previous studies of a range of avian species, such as gull 28 , Canada goose 29 , procellariform seabirds 30 and kakapo 31,32 , where members of the phylum Firmicutes dominated microbiota of guts. Firmicutes are associated with breakdown of complex carbohydrates and fatty acids, which produce energy for the host 33,34 . Relative abundances of Proteobacteria (29.9%) and Fusobacteria (9.6%) in G. japonensis were greater than those in gull, Canada goose, or kakapo, while relative abundances of Bacteroidetes (1%) and Actinobacteria (0.9%) were less 16,25,[28][29][30] . Gammaproteobacteria were present in feces of all G. japonensis as well as in feces of kakapo, turkey and the Adelie penguin (Pygoscelis adeliae) 23,32 . Similar to microbiota in guts of Pachyptila turtur and Pelecanoides urinatrix, microbiota in guts of G. japonensis contained Fusobacteria, potentially due to their diets high in fat 30,35 . In general, the top two most abundant phyla in gut microbiota of omnivorous red-crown crane were similar to that of omnivorous gull and herbivores Canada goose and kakapo. However, variations of relative abundances of the dominated phyla among avian species may be driven by the host genetics and diets 23 . The association network of host associated core microbiota suggested that succession pattern of microbial communities in G. japonensis gut were mainly driven by interactions among Lactobacillus, Enterococcus, Campylobacter, and Bacillus. Lactobacillus and Enterococcus were co-occurred, which might be due to nutritional cross-feeding and co-colonization 36,37 . Generally, most of the co-occurring species pairs or assemblages might share similar ecological characteristics 38,39 . On the other hand, Lactobacillus and Enterococcus were mutually exclusive with Bacillus, which might be due to toxic products, habitation modification, the detriment of microbial neighbors and immunomodulation 36,37 . Furthermore, Campylobacter, a potential pathogen, was mutually exclusive with Bacillus, while there were no significant associations of Campylobacter with any other taxonomic group in poultry-associated microbial network 40 .
Being held in captivity significantly influenced the assemblages of microbiota in guts of G. japonensis, which is consistent with results of previous studies on other animals, including endangered Kakapo, Antarctic seals and wild-caught rodents 31,41,42 . Specifically, domestic G. japonensis might not develop a wild-type gut microbiota during rearing in captivity. The alpha-diversity, structures of gut microbiota of captive cranes and relative abundances of key hubs were significantly different from those of wild cranes. Among the three groups, captive cranes exhibited the greatest diversity their gut microbiota, while wild cranes had the least. This might be related to transmission of bacteria, such as constant social interactions with human keepers 41 . Wild-caught animals might retain hallmarks of their previous gut microbiota for extended periods 42,43 . However, according to the shorter distance between centroids of captive and artificially bred groups than others (Fig. 3), since both captive and artificially bred cranes received a uniform diet and medical care in the reserve, captivity might have greater effects on gut microbiota than retained previous gut microbial communities of captive cranes 24,25 . Furthermore, changes in gut microbiota of domestic cranes might also be caused by  Table 2. Pair-wise, post hoc comparisons between Captivity, artificial breeding and artificial breeding × life stage using the t-statistic. t value and statistical significance are in parentheses. *P value ≤ 0.001. exposure to antibiotics. Domestic cranes are fed fish that are obtained from aquaculture facilities, which routinely use antibiotics to control disease in fish held at high population densities during aquaculture 44 . Domestic cranes are also routinely, prophylactically administered antibiotics to protect them from diseases often encountered while in captivity. Exposure to antibiotics might cause long-term changes in relative abundances of species within the microbial community of the gut. These sorts of changes have been observed in gut microbiota of humans treated with antibiotics 45 . Among the significantly different genera between captive and wild groups, Lactobacillus (class Bacilli) was the taxa with the highest fold change (461-fold greater in captive group than in wild group, Fig. 5). Lactobacillus were represented in gut microbiota community that exhibited elevated β -xylosidase and β -glucosidase levels 23 . The greater relative abundance of Lactobacillus in both captive and artificially bred groups than in the wild group might be due to the formula diet which contained relatively large amounts of sugar. Besides captivity, artificial breeding also exerted a strong influence on microbiota in guts of G. japonensis. Microbial communities in guts of artificially bred adults were distinct from those of captive adults who had hatched in the wild but were raised under the same conditions as the artificially bred adults. Artificially bred cranes lost chances of interactions with their parents, compared with other naturally hatched cranes. Artificially bred cranes were caged together since they had been hatched. Compared with wild cranes, relatively great rearing density, formula diet and frequent contact with human keepers of artificially bred cranes might increase risk of infection from the time they were neonates. The greater relative abundance of butyrate-producing Fusobacteria in artificially bred cranes might act as a feedback to defend against potentially pathogenic bacteria 30,46 , which revealed that risks of infection for cranes in the ambient environment might be greater than in the wild.
Success of reintroduction of captive birds into the wild may depend on development of microbial assemblies in the gut 26 . Separation of structures of microbial communities in feces between the adolescent and adult cranes suggests G. japonensis gut microbiota transition following the change of life stage. Adolescent G. japonensis exhibited significantly greater alpha diversities than did adults. Changes of structures and alpha-diversities of gut microbiota during host development have been reported in fishes 19 , mammals 47,48 and kakapos 31 . Adolescent animals were mainly influenced by exposure to bacteria in their environment during early development 19 , especially for development of the immature immune system 49 . One possible way to manage this development would be to treat individuals with probiotics and/or "microbe" transplants to maintain relatively stable gut microbiota 50,51 . Figure 5. Patterns of abundance and mean relative abundances of the most abundant, assignable genera varied among groups. Criteria for abundant assignable genera are mean relative abundance > 1% and with genus annotation (annotation confidence score > 60%). Key hubs (determined by network analysis) were in blue, and Campylobacter (mutually exclusive with key hubs) were in red. Mean relative abundance was tested with ANOVA, P values < 0.05, followed by post hoc Tukey-Kramer tests among groups. P values were corrected using the Benjamin-Hochberg FDR method.
Scientific RepoRts | 6:33350 | DOI: 10.1038/srep33350 In the present study, we delineate the diversities and structures of fecal microbiota of wild, captive and artificially bred red-crown cranes. The gut microbiota of G. japonensis in captivity and artificial breeding groups were significantly altered from the wild group, which might increase the risk of infection and further cause the failure of reintroduction. Building upon the current observations, further studies may combine multiple "omics" technologies to identify the effects of diet change and use of antibiotics on the health of cranes. Fecal microbiota analysis could be a useful tool to assess the health status of hosts and to improve conservation and management strategies for this threatened species, if the references of gut microbiota of healthy cranes were characterized.

Methods
Ethics statement. All methods, procedures and, experiments were conducted in accordance with the guidelines issued by the Ethics Committee of Nanjing University. All experimental designs and methods for sampling have been approved by the Administration of the Yancheng National Natural Reserve in China.
Sample description. Samples of feces were collected from both domestic and wild G. japonensis at the Yancheng National Nature Reserve (32.614~34.476N, 119.857~121.096E) on January 10, 2013. Cranes in captivity received a uniform diet and medical care in the reserve. Also, based on regulations for breeding programs for birds, individual cranes had the same amount of interaction with the same keepers. The captive cranes (17 individuals) and the artificially bred cranes (28 individuals) were kept separated in different enclosures according to age. Artificially bred cranes were the second generation of captive adult cranes (first generation). Each bonded, pair of adults used for artificial breeding (>3 years old), was housed in a single enclosure before and during breeding. Artificially bred adolescents (<3 years) were reared together in two separate enclosures (one for individuals of about 1-year or age and the other for individuals of approximately 2-years of age. Cranes were classified as adolescents (<3 years, sexually immature) or adults (>3 years, sexually mature) based on morphology of both feathers and topknot of adults 52 . Feces of wild G. japonensis were collected from 11 individuals at an area in which they foraged. Samples of feces were placed into sterile 5-mL storage tubes and transferred on ice to the laboratory. A total of 108 samples of feces were collected (Supplementary Table S2). Samples were vacuum freeze-dried, homogenized and stored at − 80 °C prior to extraction of DNA.
A 20-μ L reaction including Platinum ® Taq polymerase (Life Technologies, CA, USA) was used for each PCR amplification. Amplification was conducted in a SureCycler 8800 Thermal Cycler (Agilent Technologies, CA, USA) under the following conditions: initial denaturation at 94 °C for 2 min followed by 25 cycles at 94 °C for 15 s, 50 °C for 30 s and 68 °C for 30 s, with a final extension at 68 °C for 7 min. To minimize potential bias during PCR triplicate reactions were performed for each sample. Products of PCR were checked to assess their size and specificity by use of electrophoresis on 1.2% w/v agarose gels, and purification was performed using the MinElute Gel Extraction Kit (Qiagen, CA, USA). Purified products were quantified using Qubit ™ dsDNA HS Assay Kits (Invitrogen, CA, USA) and then adjusted to 10 ng/μ L with molecular-grade water. All purified products were Pathogen Prevalence, mean and range of relative abundance (%)  Table 3. Incidences, means and ranges of relative abundance of opportunistic enteric zoonotic pathogens in feces of cranes. Mean relative abundance was tested with ANOVA, followed by post hoc, Tukey-Kramer tests among groups. P values were corrected using the Benjamin-Hochberg FDR method. Mean relative abundance that significantly differed from wild group is in bold (*P FDR < 0.05, **P FDR < 0.01, ***P FDR < 0.001).
Scientific RepoRts | 6:33350 | DOI: 10.1038/srep33350 pooled at an equal ratio for subsequent sequencing using an Ion Torrent Personal Genome Machine (PGM, Life Technologies, CA, USA) according to the manufacturer's instructions.
Bioinformatics and statistical analysis. Sequenced amplicons were analyzed following the UPARSE pipeline 53 . Quality control: PGM adaptor sequences were automatically removed on the ION server (version 3.6.2). Primer sequences and low-quality reads were discarded from the raw output by use of the QIIME toolkit 54 (split_libraries.py script: average quality cut-off value: 20; sliding window size: 50 bp; maximum length of homopolymer: 8; minimum sequence length: 200 bp; maximum sequence length: 260 bp). Denoising, chimera removal and OTU-picking: Cleaned reads were denoised and clustered into OTUs with a sequence similarity cut-off of 97%. Chimeric and singleton OTUs (sizes less than 2) were removed. After evaluation of the quality of the sequencing and application of a chimeric reads filter, a total of 2,135,253 clean sequences with an average length of 212 ± 13 bp were obtained (Supplementary Tables S1). Taxonomic assignment and cleaning: Taxonomy was assigned to representative sequences using the RDP classifier 55 against the Greengenes database (bacterial OTUs) 56 . A small number of unexpected archaeal and chloroplast sequences were removed (QIIME filter_taxa_ from_otu_table.py script: -p k__Bacteria -n c__Chloroplast). Alpha diversity: Alpha rarefaction was performed using the Shannon, Phylogenetic Diversity (PD), Chao1 and observed species metrics (Supplementary Table S1). Beta diversity and clustering: All samples were rarefied at an equal sequencing depth (8,030 reads/sample) to reduce biases resulting from differences in sequencing depth. Principal coordinate analysis (PCoA) was performed to cluster the microbial communities based on the unweighted UniFrac distance.
Comparisons of phylogenetic diversity and evenness were performed, using the Kruskal-Wallis test, followed by post hoc Mann-Whitney U tests in the R environment. Pairwise comparisons of abundant OTUs (relative abundance > 1%) were performed using ANOVA followed by post-hoc Tukey-Kramer tests. P values were corrected using Benjamin-Hochberg FDR methods. PCoA and PERMANOVA were used to analyze the bacterial community data with PRIMER 7 software 57,58 . The statistical significance was set as P value < 0.05, and the number of permutation test replicates was set at 9,999. An association network of core OTUs was generated using SparCC 59 with 500 bootstraps to assign P values. Core OTUs were OTUs that were found in more than 50% of all samples. The microbial "core" association network was filtered to include only correlations with a correlation ρ > 0.7 and a 'two-tailed' P value < 0.002. The network was displayed and analyzed with Cytoscape 60 . To identify key genera that modulated the gut microbiota, the network was collapsed at the level of genus with ThematicMap (http://apps.cytoscape.org/apps/thematicmap#cy-app-releases-tab). Some topological parameters were calculated using the Network Analyzer 61 to describe the complex pattern of interrelationships among nodes.