Comparative study of gut microbiota in wild and captive Malaysian Mahseer (Tor tambroides)

Abstract Aims The aim of this study was to identify and compare the gut microbial community of wild and captive Tor tambroides through 16S rDNA metagenetic sequencing followed by functions prediction. Methods and results The library of 16S rDNA V3‐V4 hypervariable regions of gut microbiota was amplified and sequenced using Illumina MiSeq. The sequencing data were analyzed using Quantitative Insights into Microbial Ecology (QIIME) pipeline and Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt). The most abundant bacterial phyla in both wild and captive T. tambroides were Firmicutes, Proteobacteria, Fusobacteria and Bacteroidetes. Cetobacterium spp., Peptostreptococcaceae family, Bacteroides spp., Phosphate solubilizing bacteria PSB‐M‐3, and Vibrio spp. were five most abundant OTU in wild T. tambroides as compared to Cetobacterium spp., Citrobacter spp., Aeromonadaceae family, Peptostreptococcaceae family and Turicibacter spp. in captive T. tambroides. Conclusion In this study, the specimens of the wild T. tambroides contain more diverse gut microbiota than of the captive ones. The results suggested that Cetobacterium spp. is one of the core microbiota in guts of T. tambroides. Besides, high abundant Bacteroides spp., Citrobacter spp., Turicibacter spp., and Bacillus spp. may provide important functions in T. tambroides guts. Significance and impact of the study The results of this study provide significant information of T. tambroides gut microbiota for further understanding of their physiological functions including growth and disease resistance.

its artificial propagation for both conservation and aquaculture production due to its high market demand, high flesh quality and high commercial value (Ng, Abdullah, & De Silva, 2008).
Nevertheless, there is no report on the phylogenetic and functional characterization of gut microbiota of T. tambroides.
Gut microbiota can be considered as an "extra organ" due to its crucial role in intestinal development, homeostasis and immunological protection, growth and health (O'Hara & Shanahan, 2006).
The gut microbiota in vertebrate is complex and contains diverse and abundant bacteria, archaea, viruses, and fungi (Liu et al., 2015;Neuman & Koren, 2015). Gut microbiota of aquatic animals is transient and has higher fluidity than terrestrial animals; thus, changes in environmental factors such as temperature, salinity, trophic level, and host phylogeny may affect the gut microbial community (Denev, Staykov, Moutafchieva, & Beev, 2009;Guerreiro et al., 2016;Ringø et al., 2016;Sullam et al., 2012). More than 99% of environmental prokaryotes including the gut microbiota of animals are unculturable in laboratory that limits our understanding of microbial physiology, genetics, and community ecology (Schloss & Handelsman, 2005).
The development of next-generation sequencing (NGS) technology allows the recognition of discrete populations (culturable and unculturable) based on DNA sequences in the environmental samples (Konstantinidis & Rosselló-Móra, 2015;Tarnecki, Burgos, Ray, & Arias, 2017). Esposito and Kirschberg (2014) clarified that the metagenomic study means the whole genome sequencing and analysis of each member of the microbial community in an environmental sample by 16S rDNA-based sequencing should be called as metagenetic sequencing.
The objectives of this study were to identify and compare gut microbiota in wild and captive T. tambroides. Determination of core bacteria and prediction of their functions in gut microflora lead to identification of potential bacteria that could be used as probiotics to improve growth performance and disease resistance of T. tambroides in captivity.

| Fish sampling and species verification
Three captive adult T. tambroides (standard length 35.77 ± 1.39 cm, weight 960.57 ± 58.29 g) were obtained from hatchery at Agro-Biotechnology Institute (ABI) on 6 April 2015. The captive fish were obtained from the wild and reared in hatchery for 3 years.
DNA sequencing was outsourced to First BASE Laboratories Sdn.

| Fish dissection and DNA extraction
Tor tambroides were anesthetized using 30 ppm clove oil (Neiffer & Stamper, 2009) and euthanized by pithing (Leary et al., 2013). Fish skin was disinfected with 70% ethanol prior to autopsy. The abdomen of fish was dissected using sterile instruments in laminar flow cabinet. The gut samples were removed and separated from other internal organs. The gut parts from esophagus to anus were then cut into small pieces and placed in sterile phosphate-buffered saline (PBS) (Nie, Zhou, Qiao, & Chen, 2017) followed by mixing with vortex and kept at −80°C. Gut microbiota DNA in these gut samples was extracted using PCI DNA extraction method ).

| Data analysis using Quantitative Insights into Microbial Ecology (QIIME)
The analysis of MiSeq sequencing results was done using Quantitative Insights into Microbial Ecology (QIIME ver. 1.9.0) pipeline (Caporaso, Kuczynski, et al., 2010). Adapter sequences were trimmed from the paired-end forward and reverse reads and merged. Merged reads were quality filtered at Phred Quality Score of 20 (Q20) (Cock, Fields, Goto, Heuer, & Rice, 2010). Length filter was used to remove reads shorter than 100 bp (below 20% of the library length) to avoid unspecific match that will disturb the accuracy of the calling (Edgar, 2010). Chimeric sequences were removed using RDP Gold databases as reference (Edgar, Haas, Clemente, Quince, & Knight, 2011). De novo OTU picking strategy was used as it did not cause any information lost although may be time-consuming for large datasets (Edgar, 2010;Edgar et al., 2011).
The OTUs in generated OTU BIOM file were summarized into different taxonomic levels. Taxa summary plots were plotted to show the differences in taxonomic levels of the samples. Alpha rarefactions curves were plotted to determine the adequacy of sequencing depth. Alpha diversity indexes (Chao1, Shannon and Simpson) were calculated to explain the species richness and diversity in each sample (Udayangani et al., 2017). Good's Coverage estimator was used to estimate the percentage of the total species that are represented in a sample. In beta diversity analysis, the number of sequences per sample had been rarified to equal number based on the sample which had the lowest sequences number. Principle coordinates analysis (PCoA) was used to visualize similarities or dissimilarities of data based on phylogenetic or count-based distance metrics. Weighted UniFrac was used in the PCoA analysis of this study because it accounted for differences in relative abundances of each taxon within the communities (Lozupone, Hamady, Kelley, & Knight, 2007). Mann-Whitney U test was used to determine the differences of the gut microbial communities in the wild and captive T. tambroides (Jonsson, Österlund, Nerman, & Kristiansson, 2016).

| Functions prediction using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt)
PICRUSt (ver. 1.1.0) was used to predict the metabolic functions of the microbial communities in each sample (Langille et al., 2013). Closed reference OTU picking strategy was used with the GreenGenes database (version 13.5) as reference at 97% identity threshold DeSantis et al., 2006).
The OTU table was then normalized, and the microbiota functions were predicted with referenced to Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology (KO) database (Kanehisa & Goto, 2000).

| Dissection and species verification
Autopsy of T. tambroides revealed that the gut digesta color of wild T. tambroides obtained from Kenyir Lake was green while it was brown in captive T. tambroides obtained from hatchery which was fed with commercial floating feed pellets. The cytochrome b gene of both wild and captive T. tambroides was analyzed using NCBI BLASTn and found to be 98%-99% similar to cytochrome b gene in complete mi-

| Metagenetic sequencing of wild and captive T. tambroides gut microbiota with QIIME analysis
The de novo OTU picking generated 7,749 and 9,468 OTUs for wild and captive T. tambroides gut microbiota, respectively (Table 1).
Nevertheless, the number of species found in wild T. tambroides was 501 as compared to 442 in captive ones. 304 genera were shared between wild and captive T. tambroides (Supporting Information

| Alpha diversity analysis of T. tambroides gut microbiota
The OTUs found in both wild and captive T. tambroides were reduced as the number of sequences increased at 50,000 sequences per sample (Figure 1). In Table 2, Good's Coverage confirmed that the sequencing covered up to 99% of all gut microbiota in wild and captive T. tambroides. Chao1 index showed that captive T. tambroides gut microbiota had higher species richness than wild T. tambroides.
Nevertheless, both Shannon and Simpson indexes for wild T. tambroides gut microbiota were higher than captive T. tambroides, indicated higher species diversity in fish that live in natural environment.

| Beta diversity analysis of wild and captive T. tambroides gut microbiota
The PCoA plots in Figure 3

| Predicted metabolic functions using PICRUSt
PICRUSt analyses revealed a total of 293 predicted functions where 277 functions existed in both samples (Supporting Information   Figure 5).

| D ISCUSS I ON
Although previous studies (Esa et al., 2008;Sati et al., 2013) mostly used cytochrome c oxidase subunit I (COI) for mahseer species identification, all six T. tambroides used in this study for metagenetic analysis were identified using mitochondrial Cytochrome b (CytB) gene. Comparison of CytB gene and COI gene showed that CytB gene is more accurate to construct phylogeny trees and reveal evolutionary relationships, and it gave better resolution during separating species based on sequence data (Tobe, Kitchener, & Linacre, 2011). Although Hampala showed considerable geographical variation in coloration and morphological characteristics, the mitochondrial cytochrome b gene sequencing was able to resolve phylogenetic relationship of Hampala fishes (Ryan & Esa, 2006).
It was necessary to accurately verify the species of the mahseer fish used in this study prior to MiSeq sequencing since other fishes such as Tor spp. and Neolissochilus spp. are morphological similar to T. tambroides (Laskar et al., 2013). Identification of fish species based on morphological appearances is also subjective and can lead to misidentification.
Due to the different types of the feeds, eating habits and habitats of the wild and captive T. tambroides, it is anticipated that their gut microbiota community will be different (Li et al., 2014;Ringø et al., 2016). The wild T. tambroides in Kenyir Lake lives in natural environment that contains various types of algae belonging to cyanophytes, bacillariophytes and chlorophytes (Rouf, Phang, & Ambak, 2010).
Digesta of all three wild T. tambroides were green in color indicating these fish may be fed in various kinds of algae or plant as food in Kenyir Lake. In contrast, the gut digesta of captive T. tambroides was brown due to the formulated pellet diet which consists of complete nutrients from animal and plant sources. Therefore, T. tambroides in captivity would grow faster than wild T. tambroides at the same age.  (Dehler et al., 2017).
Higher species richness of gut microbiota was found in Atlantic salmon parr exposed to open water natural environment than in captive reared ones (Dehler et al., 2017). Although the weight and sizes of captive T. tambroides were higher compared to wild T. tambroides used in this study, it was anticipated that their gut microbiota will be highly influenced by the living environment, feed, and feeding habits.
Firmicutes and Bacteroidetes are able to degrade wide range of polysaccharides (Cockburn & Koropatkin, 2016). This may explain the higher percentages of Firmicutes and Bacteroidetes in wild T. tambroides gut microbiota as their habitat in Kenyir Lake contains huge amount of periphyton algae (Rouf et al., 2010). Cell wall of green algae contains various polysaccharides such as cellulose, pectins, hemicelluloses, lignin, and others (Domozych et al., 2012).
Digesta in the wild T. tambroides guts used for this study appeared to be green in color indicated these fish consumed a lot of algae or plant as food which had cell walls made of polysaccharides. In PCoA plots, sample data points cluster together indicated high similarity of gut microbial population among the samples. One of the data of captive T. tambroides sample was not cluster close to the rest as captive T. tambroides samples that were obtained from separated tanks in hatchery where the different microbial community in the tank water could contribute to these dissimilarities. Bacterial communities in water affected Nile tilapia larvae gut microbial communities (Giatsis et al., 2015).  Roeselers et al., 2011;Tsuchiya, Sakata, & Sugita, 2007;Van Kessel et al., 2011). However, the effects of this species have never been tested in fish. This may be due to the fact that Cetobacterium spp. is obligate anaerobe that will die under normal atmospheric condition thus hinder the possibility of using this species as probiotics in aquaculture production. There was a report stated that bacteria-mediated cobalamin biosynthesis was supported by the presence of cobalamin synthesizers such as Bacteroides, Lactobacillus, and Cetobacterium (Koo et al., 2017). Besides, C. somerae was reported to produce vitamin B 12 which also known as cobalamin (Tsuchiya et al., 2007).
The number of Bacillus spp. was higher in captive T. tambroides gut samples. The Bacillus spp. may originate from probiotics capsules that were added into the tanks few years ago. Bacillus species have been widely used as probiotics in aquaculture industry. Bacillus spp.

Dominance of
Clostridium spp. was higher in wild T. tambroides gut. Clostridium butyricum was reported as a potential probiotic that has strong adhesion and antagonistic activity against A. hydrophila and Vibrio anguillarum (Pan et al., 2008).
Bile is essential for digestion and absorption of fats and removal of excess cholesterol, bilirubin, drugs, and toxic compounds (Kanehisa, Tanabe, Sato, & Morishima, 2017 reported to possess the capability to produce iso-bile acids Hirano, Masuda, Oda, & Mukai, 1981). Enzymes from gut microbiota may contribute significantly to bile acid metabolism and essential for bile acid homeostasis in the host and contributed to host health (Long, Gahan, & Joyce, 2017). Higher bile secretion functions of gut microbiota in wild T. tambroides may offer protection to the fishes that exposed to natural environment. In this study, the Clostridium spp. was 2.09% in wild samples as compared to 0.42% in captive ones.
Carbohydrate needed to be digested to monosaccharides prior to absorption in the small intestine (Kanehisa et al., 2017). Some fishes may able to digest mono-, di-, and oligosaccharides but not for indigestible complex carbohydrates such as hemicellulose and cellulose which usually plenty in plants (Krogdahl, Hemre, & Mommsen, 2005).
High carbohydrate and high lipid diets have been widely used in aquaculture to reduce cost, but they also caused excessive lipid accumulation in the fish liver (Xie et al., 2017). In contrast, wild T. tambroides may consume algae, fruits, small fishes, and crustaceans. Thus, the gut microbiota in captive T. tambroides showed higher carbohydrate metabolism function. Many Bacteroides spp. such as Bacteroides thetaiotaomicron are capable of metabolize polysaccharides in gut (Ravcheev, Godzik, Osterman, & Rodionov, 2013). Besides, these bacteria also can provide energy from indigestible polysaccharides comprising part of the host diet (Schwalm & Groisman, 2017).
Lysine biosynthesis evolved separately into two pathways which are diaminopimelic acid (DAP) and aminoadipic acid (AAA) pathways (Liu, White, & Whitman, 2010). Lysine is an essential amino acid for living organism especially those consume vegetarian or low animal protein diet. Lysine biosynthesis functions were higher in wild T. tambroides gut microbiota, and this could be due to consumption of microalgae that was abundant in Kenyir Lake. PICRUSt predictions were made based on available genome sequences of bacteria thus some OTUs that lack of closely related genomes may be underpredicted (Salinas & Magadán, 2017). Lysine produced by gut microbiota was reported to be absorbed at the host's small intestines (Metges, 2000). bacteria that may be use as potential probiotics. This requires the development of specific media and analysis of growing conditions.

ACK N OWLED G M ENTS
The author truthfully appreciates the funding for this study from Ministry of Science, Technology and Innovation (MOSTI) Malaysia.
The author gratefully thanking Malaysia Genome Institute (MGI) for the laboratory facility for MiSeq sequencing and computer facility for bioinformatics data analysis.
F I G U R E 5 Box-plots showed three significant predictive functions (bile secretion, carbohydrate metabolism and lysine biosynthesis) of gut microbial communities in wild and captive Tor tambroides

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interests.

AUTH O R S CO NTR I B UTI O N
Anjas and Tan conceived of the study conception and experimental design. Tan, Marilyn, and Nazrien were responsible for the acquisition of data. Anjas, Tan, Natrah, and Iswan analyzed and interpreted the data. Anjas and Tan drafted the manuscript. All authors discussed the results and contributed to the final manuscript.

DATA ACCE SS I B I LIT Y
All sequences were also submitted to NCBI Sequence Read Archive