Gut microbiome analysis of fast- and slow-growing Rainbow Trout (Oncorhynchus mykiss)

Background Diverse microbial communities colonizing the intestine of sh contribute to their growth, digestion, nutrition, and immune function. We hypothesized that the gut microbiome of rainbow trout could be associated with differential growth rates observed in sh breeding programs. If true, harnessing the functionality of this microbiome can improve protability of aquaculture. To test this hypothesis, four full-sibling families were stocked in the same tank and fed an identical diet. Two fast-growing and two slow-growing sh were selected from each family. Five different extraction methods were used to obtain DNA from feces for 16S rRNA microbiome proling. These methods were Promega-Maxwell, phenol-chloroform, MO-BIO, Qiagen-Blood, Qiagen-Stool. Methods were compared according to DNA integrity, cost, feasibility and inter-sample variation based on non-metric multidimensional scaling ordination (nMDS) clusters. Results Differences in DNA extraction methods result in signicant variation in identication of bacteria that compose the gut microbiome. Promega-Maxwell had the lowest inter-sample variation and was therefore used for the subsequent analyses. The gut microbiome was different from that of the environment (feed and water). However, feed and gut shared a large portion of their microbiome suggesting signicant contribution of the feed in shaping the gut microbiota. Beta diversity of the bacterial communities showed signicant variation between breeding families but not between the fast- and slow-growing sh. An indicator analysis determined that cellulose, amylose degrading and amino acid fermenting bacteria (Clostridium, Leptotrichia and Peptostreptococcus) as indicator taxa of the fast-growing sh. In contrary, pathogenic bacteria (Corynebacterium and Paeniclostridium) were identied as slow-growing indicator taxa. Conclusion DNA extraction methodology should be taken into account for accurate proling of the gut microbiome. Although the microbiome was not signicantly different between the fast- and slow-growing sh groups, some bacterial taxa with functional implications were indicative of sh growth rate. Further studies are warranted to explore how bacteria are transmitted and potential usage of the indicator bacteria of fast-growing sh for development of The specic objectives of our study were to determine differences in community structure of the gut microbiomes between fast- and slow-growing rainbow trout and to determine if genetics plays a role in determining the gut microbiome prole. Our results highlight differences of the gut microbiome between sh family and the bacterial taxa indicative of fast- and slow-growing rainbow trout. pooled pooled is different than that of the environment (feed and water). However, feed and gut shared a large portion of their microbiome suggesting signicant contribution of the feed in shaping the gut microbiota. Some bacterial taxa were found to be signicantly different between sh families, perhaps due to host genetics, unique early rearing environments, or vertical microbiome transmission. Although population-level microbiome differences were not found to be signicantly associated with the sh growth rate, several indicator taxa were determined in the fast- and slow-growing sh. For future studies, some of these important taxa can be investigated for potential use as probiotics to improve the gut microbiota of rainbow trout.


Introduction
The e ciency and pro tability of industrial aquaculture depends in part on the growth rate of farmed shes. Growth in farmed shes is a complex process that is directly dependent on host genetics, food quality and availability, and environmental conditions [1]. Selective breeding is one strategy that can be used to improve important phenotypic traits and help in understanding the genetic architecture and the role of molecular factors causing genetic variation among different sh [2]. Family-based selection procedures have been undertaken by the United States Department of Agriculture (USDA), National Center for Cool and Cold-Water Aquaculture (NCCCWA) to improve growth rate, llet quality and disease resistance of rainbow trout [3]. A growth-selected line was developed starting in 2002, and since then yielded a genetic gain of approximately 10% in improved growth performance per generation [4].
Microorganisms may also contribute to the productivity of farmed shes. Microorganisms making up the sh microbiome reside on the sh skin, gills, and gastrointestinal tract and likely play a crucial role in the growth rate, metabolism, and immunity of the sh host [5,6]. While host genetics has a profound role in determining the gut microbiome of humans and other mammals, it is not well studied in sh [7][8][9]. On the other hand, feed and water in which sh are reared have vital roles in shaping the gut microbiome. For example, plant and animal-based meal can widely alter the composition of the host microbiota since sh acquire their microbiome from the rst-feed they eat [10,11]. Sharp et al. reported that microbiota of the marine species can be directly inherited from ancestors and passed from generation to generation [12]. The gut in particular features a diverse microbiome contributing to the weight gain, immune development, pathogen inhibition, and various metabolic activities of the hosts [13]. Resident gut microbes are bene cial for hosts either by inhibiting pathogenic bacteria with dedicated toxins or by secreting enzymes that breakdown indigestible polysaccharides in host gut to simple monosaccharides and short chain fatty acids [14]. Gut microbes can supply compounds such as vitamin B and K to host which may improve the host energy metabolism [15].
An accurate census of bacteria from sh may allow investigation of the positive effects of the microbiota. However, pro ling of the gut microbiome is directly in uenced by many factors including the experimental design, sample collection, and processing. DNA extraction is particularly important, since microbiome analysis requires adequate quality and quantity of DNA isolated for a true representation of the host microbiome [16]. Many protocols have been commercialized for DNA extraction and previous reports demonstrate that microbiome diversity varies with different DNA extraction methods [17]. It is di cult to determine the most appropriate extraction method for the downstream microbiome analysis of a particular species. Each method has its own merits and drawbacks; for example, standardized kits are typically designed for ease of use and e ciency, but a more labor-intensive method such as phenol-chloroform extraction, despite its risk of inconsistency or contamination, can potentially produce a higher yield with better quality if performed by an experienced researcher.
In this study, we investigate how the gut microbiome of rainbow trout correlates with differential growth rates. Therefore, one objective of this research was to characterize the gut microbiome of rainbow trout using high-throughput DNA sequencing. In order to achieve this objective, we considered the effect that DNA extraction methodologies play in characterization of different microbial communities in the gut of rainbow trout.
The speci c objectives of our study were to determine differences in community structure of the gut microbiomes between fast-and slowgrowing rainbow trout and to determine if genetics plays a role in determining the gut microbiome pro le. Our results highlight differences of the gut microbiome between sh family and the bacterial taxa indicative of fast-and slow-growing rainbow trout.

Fish population
Fecal samples were collected from 15 sh representing four different genetic families. The parents of these families originated from a growthselected line at NCCCWA (year class 2014) that was previously described [4,18]. Families were produced and reared until ~18 months posthatch. Brie y, full-sibling families were produced from single-sire × single-dam mating events. All sires were siblings from a single family while dams exhibited low relatedness (coe cient of relatedness < 0.16). Eggs were reared in spring water, and water temperatures were manipulated between approximately 7-13 °C to synchronize hatch times. Each family was reared separately from hatch through approximately 20 g (7 months post hatch) when 15 sh per family were uniquely tagged by inserting a passive integrated transponder (Avid Identi cation Systems Inc., Norco, CA) into the peritoneal cavity. Tagged sh were comingled for the remainder of the grow-out period. Fish were fed a commercial shmealbased diet (42% protein, 16% fat; Ziegler Bros Inc., Gardners, PA) using automatic feeders (Arvotec, Huutokoski, Finland). Feed was provided at or just below satiation for the entire grow-out period. This study includes four families with high variance in adult body weight. From each family, four sh were selected, two that were considered fast-growing (>1952 g) and two that were slow-growing (<1572 g). Of the 16 sh selected for sampling, one slow-growing sh from family two exhibited morphological signs of disease during sample collection and was excluded from analysis, reducing the total number of samples to 15. The statistical signi cance of the rank body mass between the two groups was tested by a one-way Mann-Whitney U test with an alpha of p<0.001 (GraphPad Software, Inc., La Jolla, CA).

Sample collection
To characterize the gut microbiome and compare it to the surrounding water and food source, samples were collected from sh feces, water and feed. For fecal sampling, sh were anesthetized with tricane methanesulfonate (150 mg mL −1 ) (Tricaine-S, Western Chemical, Ferndale, WA) and then manually stripped to collect the fecal samples in sterile Eppendorf tubes (Eppendorf, Hauppauge, NY). Water samples of 1.5-liter volume were collected from both the inlet water source and tank (n= 4 total samples), then ltered twice through a clean sterile membrane lter with pore size 0.45μm. DNA was isolated from lters to sample bacteria present in the environment. Feed samples were also collected by taking 100 g of feed and storing it in a Ziploc bag. All samples were stored at -80 °C until DNA extraction. phenol-chloroform (Phenol: Chloroform 5:1, SIGMA) extraction method [19]. Three of the mentioned DNA extraction methods were chosen to study the gut microbiome of fast-growing versus slow-growing trout: MO BIO kit, Promega Maxwell, and phenol-chloroform extraction. More detail of the DNA extraction methods is provided in Additional le 1. Once DNA was extracted, concentration and quality were measured, and integrity of genomic DNA was checked by gel electrophoresis. All DNA extractions were stored at -80 °C until library preparation.
Before library preparation, concentrations of all DNA samples were normalized to 2 ng/μL for PCR ampli cation using a Qubit uorometer (v3.11) (Invitrogen, Carlsbad, CA). The primers 515F and 926R (Integrated DNA Technologies) (EMP; http://www.earthmicrobiome.org/empstandard-protocols/16s/), were used to target the 16S rRNA marker gene using polymerase chain reaction (PCR). The nal PCR reaction consisted of 5μL buffer, 1.5 μL 50mM MgCl 2 , 2 μL 10mM dNTP, 0.2 μL Taq polymerase, 3 μL Kb extender, 1 μL 10 μM primer, 5 μL DNA template and 7.3 μL nuclease-free water. PCR ampli cation and sample indexing was performed according to the standard earth microbiome project protocols [20]. The ampli cation conditions were 94 °C for 45 sec, 50 °C for 60 sec, 72 °C for 90 sec for 35 cycles. Ampli cation was preceded by a 10-minute preheating step at 94 °C and followed by a 10-minute elongation step at 72 °C. Ampli cation of each sample was performed in triplicate and combined to a nal volume of 75 μL. The indexed samples were then normalized (240ng/reaction) and pooled for sample puri cation purposes. The pooled amplicon was puri ed using Promega PCR puri cation kit (Promega Corporation, Madison, WI) and visualized on a 1.5 % agarose gel stained with ethidium bromide. A DNA fragment of size approximately 370 bp for each sample was excised from the DNA gel with a clean, sharp scalpel and collected in nuclease-free sterile tubes. QIAquick gel extraction kit was used to purify DNA from the resulting gel slice (Qiagen, Germantown, MD) according to the manufacturer's recommendation. The concentration of the gel-extracted library was assessed with a Qubit uorometer (Invitrogen, Carlsbard, CA) and fragment size was determined using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, California). Final qPCR-based quanti cation of the library was done using a KAPPA quanti cation kit (Roche, Pleasanton, CA).
Sequencing was done using 250bp-paired end sequencing using a 300 cycle V2 reagent cartridge on an Illumina Miseq ow cell (Illumina, Inc., San Diego, CA) according to manufacturer's instructions (Miseq System Guide) [21]. The output le was demultiplexed and converted to fastq on the Illumina MiSeq (Illumina, Inc., San Diego, CA).

Bioinformatics analyses
Sequencing data (3,972,613 raw sequences reads) were analyzed using Mothur (v.1.40.2, www.mothur.org) according to the Mothur Illumina Miseq standard operating procedure (SOP) [22] with several modi cations. After forming contigs, we determined the median length (371 bp) of the sequences. Chops.seqs was used to keep the rst 371 bp of each sequence [23]. Sequences with ambiguous base pairs were removed by using the screen. seqs command. The split.abund command was used to keep abundant sequences with greater than two reads [24]. Sequences were aligned to the SILVA v123 database and sequences that failed to align, or classi ed as Archaea, chloroplast, eukaryotic mitochondrial, or unknown sequences, were excluded from the analysis. Sequences detected by UCHIME as chimeric were removed from the analysis. The remaining sequences were clustered using VSEARCH [25] at a threshold of >97% sequence similarity. The remove. rare command was used to remove operational taxonomic units (OTUs) having less than ten reads among shared samples. Two samples (one fast-growing extracted using Promega Maxwell method and one slow-growing sh extracted using phenol chloroform method) were excluded from the analysis because sequences in these samples did not pass the quality control and ltering steps. The parameters and the command used to analyze the data are included in additional le 2.

Statistical analysis
To study the effect of DNA extraction methods on microbial community pro ling, Bray-Curtis distances were compared and nMDS ordination was used for visualization. To test for a signi cant effect of extraction method, we used Permutational Multivariate Analysis of Variance (PERMANOVA) on the basis of Bray-Curtis dissimilarity matrices by considering extraction technique as a xed effect and using type III sum of squares and unrestricted permutation of data with 999 permutations. SIMPROF (Similarity Pro le) was performed to test the inter-sample variation on the replicate samples with a signi cant cut off value of 0.5 (95% similarity).
Beta diversity of the gut and environmental samples were calculated using Bray-Cutis dissimilarity matrices representing pairwise (sample to sample) distances to test the variation among gut and environmental samples (feed and water). Non-metric multidimensional scaling ordination (nMDS) was used to explore the microbial communities in the fast-growing and slow-growing sh by considering the dissimilarity distance matrices among the samples. One-way PERMANOVA was used to assess the effect of sample type (feces, feed and water) as predictive of the microbiome.
To understand the effect sh growth rate on the microbiome, values from Bray-Curtis dissimilarity matrices were compared and visualized using nMDS ordination. A one-way PERMANOVA was used to determine if the growth rate or sh breeding family, both considered as xed effects, were predictive of the microbiome.
An indicator analysis was done in Mothur in order to statistically and independently select bacterial taxa that are indicative of fast-/slow-growing sh or sh breeding family. Taxa with indicator values greater than 40 and a p-value (<0.05) were considered as indicative of sh growth rate or breeding family. All data les for reproduction of the bioinformatics and statistical analyses are included in additional les 3 -8.

Results
Mean weight difference between fast and slow-growing sh The mean weight of the fast-growing sh was 2123.9 ± 105.57 g, whereas, the mean weight of the slow-growing sh was 988.6 ± 297.65 g. The mass of the fast-growing sh was signi cantly greater than that of the slow-growing sh when compared using one-way Mann-Whitney U test (p>0.05) as shown in Figure 1.
B-diversity of the fecal samples, feed and water A total 1,988 OTUs were identi ed from all gut, feed and water samples. To determine if the gut microbiome differs either from environment or feed, Bray-Curtis distance matrices were compared and the results indicated the overall gut microbiome was signi cantly different from that of the environment (water and feed) (F 3,48= 2.29 and p<0.05, R 2 =39%) Fig. 2 A).
Microbiome analysis of gut, feed and water Data analysis of the microbial communities using the Promega extraction technique revealed a total of 385 OTUs in gut, whereas feed and water samples consisted of 236 OTUs and 122 OTUs, respectively. Only 17 OTUs were shared among all samples whereas 150 OTUs were shared among feed and gut (Fig. 2 B).
The overall gut microbiota was composed of 5 phyla, 10 classes and 130 genera, while the feed consisted of 5 phyla, 11 classes and 90 genera, and the water samples consisted of 6 phyla, 36 classes and 56 genera as shown in Table 1. These numbers do not include the taxa that were considered as unclassi ed and were not assigned a speci c class or genus. Phylum Firmicutes was the most abundant phylum in the gut and was the most shared phylum among gut and feed samples. Proteobacteria was the most abundant phylum in the water samples. The most abundant phylum in the feed was Firmicutes. Detailed information about the microbiota classi cation is included in additional le 9.

Comparison of different DNA extraction methods
To test if pro ling of the gut microbiome is directly in uenced by DNA extraction method, three replicate pools of the sh fecal samples were sequenced and analyzed using ve different extraction methods. Within non-metric dimensional scaling ordination plots, the three-replicate samples extracted with Promega clustered tightly, whereas, replicate samples of the four other extraction methods were relatively more heterogeneous (Fig. 3). PERMANOVA con rmed that DNA extraction method is predictive of the microbiome (F 4,13 = 2.4234, p<0.05, R 2 =51%).
To further investigate the effects of DNA extraction methodology on microbiome pro ling, three different methods were chosen for microbiome sequencing from individual (non-pooled) fecal samples of all available sh in the study. PERMANOVA results con rmed the signi cant effect of extraction technique on predicting microbial communities (Fig. 4 A; F 2, 42 =10.467, p<0.05, R 2 =34%). Comparative analysis of the three extraction methods revealed that phenol-chloroform had the highest OTU richness with 649 OTUs. A total of 119 OTUs overlapped between all three DNA isolation methods (Fig 4. B). Comparing the abundance of Gram-positive and Gram-negative bacteria, it was clear that the abundance of the Gram-positive is higher than that of the Gram-negative in all three DNA extraction techniques (Fig. 4 C) with the Promega kit showing the most drastic gram-positive bias. The SIMPROF test for statistically signi cant cluster and it showed that the Promega method had 95% similarity within the replicate samples forming the tightest cluster (p-value < 0.05).
Beside heterogeneity and abundance biases, other factors including yield, integrity, time durations for sample processing, amount of hazardous waste liberated were also considered during extraction comparison. Phenol-Chloroform gave the highest yield, but it is tedious, time-consuming, requires individual handling and released more hazardous waste. Whereas, Promega is semi-automated method, easy to perform in large-scale production, and shows the least inter-sample variation among the replicate samples, results release of least hazardous waste as shown in (Table  2). We decided to choose Promega for our downstream analysis of the gut microbiome.
Gut microbiome analysis of fast-and slow-growing sh Both nMDS ordination and PERMANOVA results indicated that the microbial communities did not signi cantly differ between the sh of different growth rates (p>0.05, Fig. 5 A). Both fast-and slow-growing sh possessed unique sets of OTUs and also overlapping taxa (Fig. 5 B). An indicator analysis predicted that 10 OTUs were found as indicative of growth rate (Table 3, P < 0.05). All fast-growing indicator taxa belonged to phylum Firmicutes, whereas, the slow-growing indicator taxa belonged to the Actinobacteria and Firmicutes (Table 3).
PERMANOVA results indicated that sh breeding family was predictive of the microbiome (F 3,13 =2.1673, p<0.05, R 2 =39%) (Fig 5 C). The Vennrepresentation depicted 106 OTUs shared among all the families with family 2 having the most unique OTUs (Fig. 5 D). An indicator analysis of each sh family predicted that 6 OTUs were identi ed as indicative of family 1, 3 OTUs for family 2, and 1 OTUs for family 4 ( Table 4, P<0.05).

Discussion
The salmonid aquaculture industry can bene t from development of fast-growing sh. Selective breeding is one strategy that can be used to improve sh growth and help in understanding its genetic architecture [26]. On the other hand, the environmental factors, particularly diet, have immense role in improving this phenotypic trait. Microbiota of the sh gut also have a signi cant role in determining host health and, in some instances, growth rate [27]. The microbiota and its host have an intimate and sometimes mutually bene cial relationship. The environmental conditions and host diet have profound role in shaping the gut microbiota [28]. In addition, recent studies have illustrated that host genetics have direct impact on gut microbial composition in many species including human and rodents [9] and to some extent chicken [29]. In uence of the host genetics in determining the microbiota of a sh has yet to be characterized [9,30].
In this study, the DNA extraction methodology comparison was performed to optimize the extraction methodology and apply this to the comparison of fast-and slow-growing sh gut microbiomes. Five different extraction techniques including bead beating and semi-automated methods were included. The effects of the DNA extraction methods were assessed on the basis of DNA quantity, quality and the inter-sample variation in microbial communities between replicates. The concentration and the quality of the DNA varied signi cantly between the DNA extraction techniques. The MOBIO, Qiagen blood and Qiagen stool gave relatively low yield, whereas Promega Maxwell kit that uses automated method resulted in a higher yield in comparison to the other three kits which is consistent with previous reports [31]. In comparison, phenolchloroform, being a robust method, uses a stringent lysis step and produced the highest DNA yield and highest microbial diversity. This is likely due to this method being able to effectively lyse the cell walls of both the Gram-positive and Gram-negative bacteria. However, the phenolchloroform method resulted in higher inter-sample variation, was the most labor-intensive, and produces more hazardous waste when compared to the Promega method. It has been proven that the bead-beating methods result in identi cation of greater microbial diversity than non-beating methods [32]. MOBIO method involves bead beating to physically lyse cell wall of bacteria and increase the number of the microbial species identi ed but showed relatively high inter-sample variation among replicates. Promega Maxwell, semi-automated method seems biased toward the Gram-positive bacteria, perhaps, due to addition of lysozyme enzymes which induces lysis of the Gram-positive bacterial cell wall or decreased lysis of Gram-negative bacteria. The Promega method showed the least inter-sample variation among replicates. Similar is the case with Qiagen-stool, Qiagen-Blood and Tissue kit, since both kits gave su cient yield and integrity but resulted higher inter-sample variation among replicates.
This study showed that the sh gut microbiome differs from that of the feed and environment (tank water and inlet water). Regardless of the DNA extraction method, most of sh microbiomes clustered together separately from that of the water samples. Conversely, microbiomes of the feed samples clustered closer to those of the gut samples and shared higher number of OTUs compared to that of the inlet and tank water samples. Various studies have been done in rainbow trout to see the in uence of feed on shaping the gut microbiome, and it has been shown that the diversity and abundance of microbiomes in rainbow trout increases gradually after the rst feed they eat. For example, plant-based feed may increase the abundance of phylum Firmicutes whereas marine based feed increases the abundance of phylum Proteobacteria [33]. These ndings suggest that sh may acquire at least some of their microbiome from the feed they eat [10,33,34]. Regarding the water samples, the gut samples shared relatively higher number of OTUs with the tank water compared to that shared with the inlet water. The tank water likely contains bacteria that has leached from feces, or fecal particulates, which may explain the higher OTUs shared among gut and tank water in comparison to inlet water source. Fish were also reared in partial reuse water (60% recycled) which may contribute to the sharing of OTUs between the water and gut samples.
We found that certain taxa were indicators of the sh growth rate and sh breeding family. The indicator taxa associated with slow growth rate seem to be harmful/pathogenic bacteria whereas the indicator taxa of fast-growing sh seem to have a mutual bene cial relationship with the host. Corynebacterium and Paeniclostridium which are known pathogens [35] were more prevalent in slow-growing sh. The toxins produced by these bacteria cause swelling and abdominal discomfort due to uid accumulation and sometimes also lead to development of circumscribed lesions and lethargic behavior [36]. Conversely, bacteria belonging to phylum Firmicutes; Lactobacillus, Lactococcus, family Propionibacteriaceae and phylum Bacteriodetes were signi cantly more abundant in the fast-growing sh. These bacteria produce microbial metabolites such as short-chain fatty acids during glucose fermentation [37]. Such fatty acids can improve growth rate of rainbow trout [3]. These bacteria, perhaps, can be used as probiotics since they produce enzymes for fatty acid degradation, help in breakdown of food and produce valuable nutrients and energy [38][39][40][41]. These microbiomes also induce mucus production which acts as a barrier for pathogenic bacteria and sometimes also leads to production of antimicrobial peptides. In addition, the Bacteriodetes produce inhibitory substances like bacteriocin which initiates pathogenic bacterial cell lysis or growth inhibition [38]. Lactobacillus has been proven to inhibit the pathogens and, therefore, used as preservatives for food storage since they can induce the barrier function in the host epithelium against pathogens [42].
Moreover, families Lachnospiraceae, Planococcaceae, Leptotrichiaceae, and Peptostreptococcaceae belonging to the phylum Firmicutes were indicator taxa for the fast-growing sh in this study. The high abundance of phylum Firmicutes in host is associated to obesity or increase body weight in host as the bacteria belonging to this phylum can take part in lipoprotein degradation, fatty acid accumulation in adipose tissue, in addition, these bacteria consists of nutrient transporter gene [43][44][45]. Bacteria belonging to class Lachnospiraceae reside in digestive tract, produce butyric acid, aid in amino acid fermentation, protein digestion, absorption of fatty acids, associated with weight gain and prevention of different diseases due to microbial and host epithelial cell growth [46,47]. On the other hand, bacteria like Sellimonas, Clostridium, Peptostreptococcus in fast-growing sh can take part in fermentation of different amino acids, lactates and sugars. Clostridium are more likely to produce cellulase enzyme and result in degradation of the cellulolytic bers, whereas other bacteria like Selenomonas produce propionate which initiates gluconeogenesis and also produces xylosidase enzymes that aid in digestion of the animal feeds. The most widely prevalent and statistically signi cant indicator taxa of the fast-growing sh, Peptostreptococcus and Clostridium, are more likely be involved in amino acid fermentation that ultimately leading to amino acid absorption in host gut. Leptorichia, the most abundant taxa in gut of all the fast-growing sh are cellulose-degrading bacteria, therefore, amylase and cellulase activity are expected to be more prominent in the host inhabiting this bacteria [48]. Similarly, the class Enterobacteriaceae was found to be signi cantly abundant taxonomical class in most of the fast-growing sh. E. coli belonging to class Enterobacteriaceae group has proven to be associated with weight gain in human infants [49].
Strategies such as manipulation of plant and animal sources for nutrition have implications for the gut microbiome of sh [50,51]. For example, plant-based feed may constitute different microbiota in comparison to animal-based feed. Many enzymes producing bacteria residing inside gastrointestinal tract produce enzymes like amylase, lipase, cellulase aid in the digestion of the different metabolites [52]. The most abundant phylum Firmicutes found in gut are more importantly associated to digest polysaccharide such as cellulose, hemicelluloses and xylan. On the other hand, Bacteroidetes and other Lactic Acid Bacteria induce production of antimicrobial agents to inhibit the growth of other pathogenic bacteria in gut [53] improving the host immune system and metabolism.
Finally, although most of microbiota were shared among the sh families, some unique taxa were characteristic for each family which suggests that genetics is a contributing factor that affecting the gut microbiome. Unique taxa included Trueperiolla, Kocuria, Lactobacillus, Propionibacteriaceae, and Lactococcus. Kocuria has been reported to induce the protective immune system in rainbow trout by inhibiting pathogenic bacteria like Vibrio [54]. Differences in microbiota among the families suggest that genetics affects the susceptibility of the sh to colonize speci c members of the microbiota from the environment. However, it should also be acknowledged that early periods of development occurred in different tanks speci c to each family. Although all four tanks were positioned sequentially, utilized the same water source (inlets came originated from the same pipe), and consumed identical feed, it is unknown if the microbial communities within each tank differed and, if so, how they could have persisted through the subsequent 12-month grow-out period. It is also unknown if there is vertical microbiome transmission from the parents to progeny or if maternal fecal contamination of eggs during manual egg stripping contributes to the offspring microbiome. Further research is needed to validate familial differences and determine the contribution of genetic and environmental factors to development of the gut microbiome.

Conclusions
This study showed that DNA extraction methodology should be taken into account in accurate pro ling of the gut microbiome. The sh gut microbiome is different than that of the environment (feed and water). However, feed and gut shared a large portion of their microbiome suggesting signi cant contribution of the feed in shaping the gut microbiota. Some bacterial taxa were found to be signi cantly different between sh families, perhaps due to host genetics, unique early rearing environments, or vertical microbiome transmission. Although population-level microbiome differences were not found to be signi cantly associated with the sh growth rate, several indicator taxa were determined in the fast-and slow-growing sh. For future studies, some of these important taxa can be investigated for potential use as probiotics to improve the gut microbiota of rainbow trout.
Declarations Tables   Table 1. Number of phyla, classes and genera in gut, feed and water samples using the Promega DNA extraction method. Due to technical limitations, Table 4 has been placed in the Supplementary Files section. Figure 1 Mean weight of fast-growing and slow-growing sh. The error bars indicate standard deviation.

Figure 2
A) nMDS representation of bacterial community of sh gut, feed and water samples using all three extraction methods (stress value=0.14). Feed sample clustered together with fecal samples, whereas most of the water samples clustered apart from fecal samples. B) Venn-diagram of common and unique OTUs in gut, feed and water samples. nMDS representation of three replicate pooled samples using 5 different extraction methods (stress value=0.12). Each extraction method is signi cantly different (p<0.05). SIMPROF analysis tested for signi cant distinct cluster. One of the phenol-chloroform samples did not pass the QC and was excluded from the analysis.

Figure 4
A) nMDS representation of the fecal samples using three different extraction methods. Samples were clustered on the basis of Bray-Curtis distance matrices (stress value=0.13). B) Venn Diagram depicting the common and unique OTUs in three different extraction methods, P:C indicates phenol chloroform C) Abundance of Gram-positive and Gram-negative bacteria on rainbow trout gut using three different extraction methods.

Figure 5
A) nMDS representation of the fast-and slow-growing sh using Promega extraction method (stress value=0.07). B) Venn-diagram depicting the common and unique OTUs in fast-growing and slow-growing rainbow trout C) nMDS representation of the sh family on the basis of dissimilarity matrices (stress value=0.07). Most of the samples from family 1 were clustered apart from families 2, 3 and 4. D) Venn representation of the common and unique OTUs among four different families.