Internal Versus External Pressures: Effect of Housing Systems on the Zebrafish Microbiome

Abstract

Zebrafish (Danio rerio) are an attractive model organism for scientific studies, including host–microbe interactions. The organism is particularly useful for studying aquatic microbes that can colonize vertebrate hosts, including Vibrio cholerae. Previous studies have established the presence of a core zebrafish intestinal microbiome, and tracked the development of the zebrafish intestinal microbiome from the larval stage to adulthood. An unexplored matter in this host–microbe relationship is the effect of the housing system on the zebrafish intestinal and tank water microbiomes. In this study, we used 16S rRNA gene sequencing to investigate the response of zebrafish intestinal and tank water microbiomes to a change in housing conditions. Zebrafish in the separated fish tanks showed no initial differences in the structures of their intestinal microbial profiles; the same prominent bacteria were present and abundant across tanks. Immediately after the housing switch, the zebrafish intestinal microbial profiles changed in composition and structure. Within 5 days of the housing switch, the intestinal microbiome had stabilized, and changed significantly from the prehousing switch profile. This study demonstrates that although external factors can significantly perturb and alter the zebrafish intestinal microbiome, the microbiome displays a large level of selective resilience whose primary members (namely Vibrio) persist.

Introduction

The zebrafish (Danio rerio) is a tropical freshwater fish native to southern Asia. Although this cyprinid has a well-established history as a popular home aquarium fish, since the 1970s it has also been a frequently used animal model in biological research. In the past 30–40 years, the zebrafish has been used as a model to study many aspects of biology.^1–6 The first studies capitalized on the transparency of zebrafish embryos, which permitted examination of their developing anatomical structures.^3,7,8 Zebrafish transparency was further developed by researchers at the Children's Hospital Boston through the generation of the Casper fish. The Casper fish is a cross between two mutant zebrafish, nacre and roy; it lacks pigment in its skin and scales and thus is translucent even as an adult.⁹

Because of the high level of similarity between the zebrafish and human immune systems, the zebrafish has also become a popular animal model for studies of immune response and microbial pathogenesis.^10–17 The zebrafish also offers an advantage with respect to fecundity over mammalian model organisms; a single adult female zebrafish can produce several hundred eggs per week. In addition, the zebrafish matures rapidly, reaching adulthood within 3–4 months, and it can be housed economically.^9,15,18–23

The microbiome has become an increasingly important area of research in recent years.^24–26 Intestinal microbiomes critically contribute to the functioning of their host organisms. These roles include but are not limited to development and bolstering of the immune system, cell-to-cell signaling, chemical signaling used for intraspecific communication, and digestive roles where the members of the microbial community provide access to energy and nutrients that would otherwise be inaccessible to the host.^27,28 With more and more studies being conducted, the roles and significance of the microbiome are becoming increasingly clear; yet, much still remains unknown.

For zebrafish specifically, the intestinal microbiome has been shown to play major roles in fatty acid absorption into the intestinal and extraintestinal tissues, and in establishing normal homeostatic levels of neutrophils within the intestine.^29,30 Past studies have shown that the composition of the zebrafish intestinal microbiome varies across development, with zebrafish developing a more diverse intestinal microbiome as they age.³¹

The zebrafish intestinal microbiome can also be shifted by external factors such as diet and broad rearing environment. For example, Roeselers et al.³² found that the composition of the intestinal microbiome of zebrafish collected from the wild in India differed from that of zebrafish bred and raised in different laboratory environments within the United States. Nevertheless, there remained a core (i.e., consistently present) intestinal microbiome made up primarily of Aeromonas and other members of the family Aeromonadaceae and the class Gammaproteobacteria. Other groups have found that neutral processes (i.e., community assembly is a result of chance and dispersal) decrease in significance as the zebrafish develop from larvae to adults, whereas the influences of host selection and microbe–microbe interactions increase as hosts mature. Furthermore, although the zebrafish microbiome can experience variation because of external factors such as habitat and salinity, host-specific pressures within the gut are the dominant forces controlling the structure of the intestinal microbiome.^31,33,34 Collectively, these studies suggest that the structure of the zebrafish intestinal microbiome is shaped by selective pressures within the host gut, more than by the structure of microbial communities in the local and immediate environment.

In this study, we took advantage of a unique opportunity to examine if and how tank water and zebrafish intestinal microbiomes were altered after a housing switch from separated individual fish tanks to an integrated recirculating housing system that shares a central water supply. In the new system, the zebrafish were still housed in individual tanks but the water was from a shared community source that was continually monitored and adjusted to maintain the optimal aquatic environment for zebrafish. The tanks that housed the zebrafish had a constant flow of water that entered at the top of the tanks and exit out in the rear. This vacated water then passed through a four-stage filtration system and a ultraviolet sterilizer before it returned to the fish tanks.

By sampling microbial communities in the tank water and zebrafish intestines several times over 10 days before the housing switch, on or near the day of the switch, and over the 2 weeks after the switch, we observed distinct changes in both the fish tank water and zebrafish intestinal microbiomes after the change in housing systems. This study thereby contributes to the growing evidence that housing conditions can substantially impact the structure of microbiomes in animal model research, and that this variation could potentially lead to disparate results being obtained among research groups.^{31,33,35–39}

Materials and Methods

Zebrafish

The sampled fish were adult wild-type ZDR zebrafish and Casper fish. Fish from tanks 1 and 2 were zebrafish aged 10 and 24 months, respectively, at the time of collection. The Casper fish (tank 3) were aged 23 months at the time of collection. A schematic of the sample collection process is given in Figure 1. Because of the fish being placed on the new system in a staggered manner (to ensure the wellbeing of the fish in the laboratory as a whole as they were transitioned into a new and potentially inhospitable environment), a modest degree of uneven sampling resulted. The “group” sampling method (i.e., binning of contiguous sampling days into groups) was utilized to help compensate for this unevenness and to pinpoint if and when significant changes occurred to the microbiome.

FIG. 1.

Zebrafish microbiome comparisons by “Groups” and β-diversity. All Xs indicate a day that fish samples were collected (three fish were collected per indicated time point). The number of days indicates how many days the fish from the respective tank had been on the new system before being collected (i.e., an X for tank 1 on day 7 means that those three fish had been on the new system for 7 days at the time of collection). The fish in tanks 2 and 3 were moved into the new housing system on the day it was first operational; they were then sampled 2 days later. The fish in tank 1 were moved into the new housing system 2 days after it was first operational; they were sampled on that same day (4 h later), and the day after. Water samples were collected during each fish sampling event. Statistical analysis was performed by comparing the water and intestinal microbiome profiles of one “Group” with the corresponding profile of the “Group” of the closest time point (i.e., Pre vs. days 0–2, days 0–2 vs. days 3–5, and so on), controlling for tank identity although generalized linear modeling and NPMANOVA in α- and β-diversity analyses, respectively. The total number of samples were fish = 77; water = 27. The number of samples per Group were as follows: pre (−10 to −6): fish n = 26; water n = 9; days 0–2: fish = 12; water n = 4; days 3–5: fish n = 18; water n = 6; days 7–9: fish n = 14; water n = 5; day 14: fish n = 7; water n = 3. NPMANOVA or PERMANOVA, permutation multivariate analysis of variance.

Fish (three fish per tank for each individual sampling time point) were collected from the three different tanks before and after the housing switch and on or near the day of the switch (Fig. 1). For “group” 2 (days 0–2), the fish in tanks 2 and 3 had already been on the system for 2 days when the fish from tank 1 were placed on the system. The fish collected from tank 1 on day 0 were collected 4 h after being on the new system. Zebrafish from tanks 2 and 3 were collected on day 9 to balance out the total number of sampling events per tank. Before the switch, the fish were housed in separated tanks containing water filtered by reverse osmosis and maintained at pH 7.0–7.5. Tank water was conditioned with Instant Ocean Salt (Aquarium Systems, Mentor, OH) to a conductivity of 400–550 μS.

Fish were then switched over to an Aquaneering zebrafish housing system with new tanks and a recirculating community water source that automatically controls temperature, pH, and conductivity. Zebrafish were killed in 100 mL of 320 μg/mL Tricaine-S (MS-222, tricaine methanesulfonate; Western Chemical, Ferndale, WA). All animal protocols were approved by the Wayne State University Institutional Animal Care and Use Committee. The entire intestinal tract of each fish was aseptically excised. DNA was extracted from the intestinal homogenates using a DNEasy Powersoil isolation kit (Qiagen) following the manufacturer's instructions.

DNA was also extracted from 50 mL water samples from each tank (spun down at 4^°C at 10,700 g for 10 min and resuspended in 1 mL sterile phosphate-buffered saline) each day fish were collected, again using the DNEasy Powersoil isolation kit. Whereas three fish were collected per sampling event per tank to address potential interindividual variation in the gut microbiomes of fish housed together in a single tank, one 50 mL sample of water was collected to represent the microbiome in the water column of each tank per sampling event.

As the tanks had constant water circulation both before and after the housing switch, the microbial profile of the water column within tanks should be rather consistent. The cutoff for low DNA recovery yield for samples was 1 ng/μL. In contrast to the intestinal and water samples, background technical controls (i.e., blank DNA extraction kits) did not yield detectable 16S rRNA gene PCR products and were not sequenced.

DNA sequencing

DNA isolated from individual fish and water samples was used to generate 16S rRNA gene libraries. Illumina MiSeq sequencing was performed at Michigan State University using methods described previously.^40,41 The V4 region of the bacterial 16S rRNA gene was targeted for sequencing using the primers 515F (5′ GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). Raw sequence files for each sample were deposited on the NCBI Sequence Read Archive (Bioproject: ID SUB5288195).

All sequence data were processed using mothur software, v. 1.39.1^42,43 following the standard operating procedure (SOP; https://mothur.org/wiki/MiSeq_SOP) recommended by Schloss and colleagues.^42,43 In brief, sequences were removed from the data set if they (i) contained ambiguous base calls, (ii) had homopolymer runs in excess of eight bases, (iii) ended at incorrect primer positions, (iv) were designated chimeric by the uchime tool, or (v) were taxonomically classified as mitochondria, chloroplasts, archaea, eukaryotes, or were of unknown taxonomy, using the Ribosomal Database Project's trainset9_032012 reference database.^44,45

The cutoffs for sample inclusion were 3796 sequences and a Good's coverage of 98.5%. Four fish samples were thus ultimately excluded from analysis (one from tank 1 on day 10, two from tank 2 on day 14, and one from tank 3 on day 9). The 104 remaining samples (total number of samples: fish = 77; water = 27; samples by group, pre [−10 to −6]: fish n = 26; water n = 9; days 0–2: fish = 12; water n = 4; days 3–5: fish n = 18; water n = 6; days 7–9: fish n = 14; water n = 5; day 14: fish n = 7; water n = 3) were subsampled to the number of sequences in the least represented sample (3796 sequences).

Using mothur's average neighbor, split-clustering algorithm (level 3), and a 97% sequence similarity cutoff, the remaining sequences were categorized into operational taxonomic units (OTUs). A consensus taxonomy was determined for each of the OTUs (N = 1605; with 672 singletons), using a standard 80% confidence threshold.⁴⁵ Taxonomic designations, raw OTU count data, and Good's coverage values for each sample are provided as Supplementary Data. Good's coverage values averaged 99.5 ± 0.2% (standard deviation) and 99.3 ± 0.4% for the fish intestine and water samples, respectively, indicating that there was thorough sample coverage in this study.

α-Diversity analysis (i.e., diversity within an environment)

The Chao1 Index was utilized to estimate how many different bacterial OTUs were present in a given sample. The nonparametric Shannon–Wiener index and inverse Simpson index were calculated to indicate both the richness and evenness of samples (i.e., heterogeneity).^46,47 Generalized linear modeling (GLM) was performed on log-transformed data to evaluate variation in α-diversity among tanks and time points using the car, nlme, lme4, and multcomp packages in R.^48–51 The values of the indices were log-transformed before analysis with GLM to normalize the data as the residuals need to be normally distributed if tests requiring normality are to be used.⁵² Utilization of GLM enabled accounting for the three fish sampled per tank per time point. The GLM output included post hoc comparisons evaluated by multiple comparisons of mean values: Tukey contrasts.

β-Diversity analysis (i.e., diversity between two or more environments)

Microbiome composition (i.e., membership) and structure (i.e., membership plus members' relative abundances) were characterized using Jaccard and Bray–Curtis ecological similarity indices, respectively. Principal coordinates (PCoA) plots were used to visualize variation in microbiomes among samples, and the effects of tank identity and pre- or posthousing switch on microbiome composition and structure were evaluated using permutation multivariate analysis of variance (PERMANOVA or NPMANOVA).⁵³

Specifically, one-way PERMANOVAs were conducted in PAST, v.3.11 for the prehousing switch analysis,⁵⁴ and the adonis function in the Vegan package in R (http://cc.oulu.fi/~jarioksa/softhelp/vegan/html/adonis.html)⁵⁵ was used to control for tank identity (i.e., by the strata command) to evaluate the complete pre- and posthousing switch differences in the fish intestine and water microbiomes, both overall and specifically across the more defined “grouped” periods (Fig. 1).^47,53,56,57

Raw p-values are presented for all multivariate permutation tests (999 permutations). No Bonferroni corrections were applied for β-diversity analyses as they can be overly conservative when used with permutation tests.⁵⁸ SIMilarity PERcentage (SIMPER)⁵⁶ analyses were conducted to determine the contribution of specific OTUs to observed variation in microbiome structure among sample types, and Linear discriminant analysis Effect Size (LEfSe) were conducted to identify OTUs that were differentially relatively abundant among the sample types.⁵⁹

These LEfSe analyses were controlled for tank identity using the ID (i.e., tank identity) parameter. Within the LEfSe figures, all the OTUs that had an linear discriminant analysis (LDA) score >2.5 were presented (Supplementary Fig. S3B). Heat maps illustrating the microbial profiles of samples were created using Morpheus software (https://software.broadinstitute.org/morpheus/).

Results

α- and β-Diversity of the water and zebrafish intestinal microbiomes changed significantly after the housing switch

The α-diversities of the water and zebrafish intestinal microbiomes were assessed through Chao, Shannon–Weiner, and inverse Simpson indices. Before switching the water source, there was no evident variation in these metrics among the individual fish tanks, with respect to either the water or the zebrafish intestinal microbiomes (multiple comparisons of mean values: Tukey contrasts, p > 0.05).

It was not possible for us leave some fish on the older system to serve as a negative control because of logistical and practical constraints. However, we observed that the α-diversity of the water and fish intestinal microbiomes within the three tanks did not change over the course of the three time points examined before the housing switch, indicating that the water and intestinal environments were originally stable over time (multiple comparisons of mean values: Tukey contrasts, p > 0.05). Although the α-diversity, specifically the heterogeneity, of water samples in the fish tanks markedly increased after the housing switch (Fig. 2), the α-diversity of fish intestinal microbiomes was not affected (Fig. 2). These data indicate that the overall richness of bacterial OTUs in the water was unchanged after the housing switch but that the dominance of certain bacterial OTUs in the tank water was reduced (particularly OTUs 1 [Vibrio] and 6 [Escherichia]) (Supplementary Fig. S1). Changes to α-diversity in the water microbiome did not seem to translate to changes within the fish intestine.

FIG. 2.

The bacterial profiles of the tank water increased in α-diversity with the housing switch but the profiles of the zebrafish intestines did not. Pre- and posthousing switch α-diversities of the zebrafish intestinal and tank water microbiomes are presented as jitter plots. Statistical tests were performed using generalized linear modeling (multiple comparisons of mean values: Tukey contrasts). Sample sizes were: fish n = 77 from three tanks; water n = 27.

With respect to β-diversity, before the housing switch, the microbiomes of the water differed among the tanks (PERMANOVA; composition: p = 0.048; structure: p = 0.026). The water microbiome in tanks 2 and 3 did not differ in composition or structure. The water microbiome in tank 1, which contained the 10-month-old zebrafish, displayed a significant difference in composition and structure from tank 3 (23-month-old Casper fish) but not tank 2 (24-month-old zebrafish). When comparing all three tanks, the fish intestinal microbiomes varied among the three tanks in composition (p = 0.023), but not structure (p = 0.212).

The zebrafish in tank 2 and the Casper fish in tank 3 had similarly composed intestinal microbiomes before the housing switch, but the intestinal microbiome composition for the zebrafish in tank 1 (the younger zebrafish) differed from that of the Casper fish in tank 3 (p = 0.025). Despite the one significant difference in the composition of the fish intestinal microbiomes at the start of the experiment, the prominent intestinal OTUs, such as OTUs 1 and 2 (both Vibrio), OTU3 (Enterobacteriaceae), and OTU4 (Cetobacterium), were consistently shared among the intestines of the fish within the three tanks (Supplementary Fig. S1).

When examining the zebrafish intestinal microbiomes in each individual tank over the three time points before the housing switch, consistent differences in composition and structure were not observed (PERMANOVA; composition: p = 0.1135 [tank 1], p = 0.3666 [tank 2], p = 0.035 [tank 3]; structure: p = 0.4554 [tank 1], p = 0.9657 [tank 2], p = 0.245 [tank 3]).

Significant changes were observed in both the composition and structure of zebrafish intestinal and tank water microbiomes after the housing switch (Fig. 3). In congruence with the α-diversity analysis, after the housing switch, the tank water microbiomes were characterized by a decrease in the relative abundance of Vibrio (OTU1; rs = −0.93834, p = 4.92E-13) and Escherichia (OTU6; rs = −0.92348, p = 6074E-12), whose abundances were negatively correlated with sample values for coordinate 1 in the PCoA plot in which tank water microbiomes before and after the housing switch clearly clustered separately (Fig. 3; Bray–Curtis).

FIG. 3.

The housing switch brought about changes in the β-diversity of water and zebrafish intestinal microbiomes. Pre- and posthousing switch β-diversities of the zebrafish intestinal and tank water microbiomes presented as PCoA plots based on Jaccard and Bray–Curtis indices. Statistical tests were one-way PERMANOVAs run using Adonis with the strata function controlling for tank identity. Sample sizes were as follows: fish n = 77 from three tanks; water n = 27. PCoA, principal coordinates.

In contrast, Enterobacteriaceae (OTU3; rs = 0.64451, p = 0.0002), Aeromonas (OTU5; rs = 0.75762, p = 4.74E-6), and Shewanella (OTU8; rs = 0.55645, p = 0.0026) were positively correlated with tank water microbiome samples for coordinate 1 (Spearman's correlation with sequential Bonferroni corrections applied), indicating that their relative abundances increased in tank water microbiomes after the housing switch. PCoA plots of all fish and water samples revealed that, with respect to microbiome composition, many of the posthousing (days 3–5 and 7–9) switch tank water microbiome samples clustered separately from the zebrafish intestinal microbiome samples (Supplementary Fig. S2).

Two-way PERMANOVAs using adonis, with the strata function controlling for tank identity, confirmed significant differences in the composition and structure of the fish intestinal and tank water microbiomes both before and after the housing switch (p = 0.001 for all tests) (Supplementary Fig. S2). When examining the β-dispersion of the fish intestine and tank water (pre vs. post) no significant difference was found in the microbial distribution of the fish intestinal structure after the housing switch (permutation test; p = 0.777); however, a significant difference was observed in the tank water (permutation test; p = 0.001).

α-Diversity of the zebrafish intestinal microbiome displayed significant changes immediately after the housing switch

Given the observation that both the fish intestinal and water microbiomes differed in composition and structure before and after the housing switch, we elucidated whether the observed differences were mercurial, that is, came about immediately after the housing switch, or were more gradual in nature, requiring the full 14 days for a complete change in diversity profiles. Utilizing the “Group” analysis approach outlined in the section “Materials and Methods” and Figure 1, the fish intestinal and water microbiomes were examined for any significant changes.

Controlled for tank identity, the OTU richness of the fish intestinal microbiomes increased immediately after the housing switch (i.e., comparing the preswitch groups with the 0–2 days postswitch groups), but then returned to preswitch levels by days 3–5 postswitch (Fig. 4). Modest reductions in fish intestinal microbiome heterogeneity were also observed between days 0–2 and 3–5 postswitch (Fig. 4). The observed changes to the α-diversities of the tank water microbiomes after the housing switch materialized more slowly (Fig. 4). Specifically, no significant changes to the tank water microbiome were observed between the preswitch “group” and the 0–2 days “group.” However, there was an increase in OTU richness of the water microbiome between the preswitch “group” and 7–9 days postswitch, and increases in water tank microbiome heterogeneity between the preswitch “group” and 3–5 days postswitch (Fig. 4).

FIG. 4.

The α-diversities of fish intestinal and water microbiomes changed soon after the housing switch. Diversities of the zebrafish intestinal and tank water microbiomes arranged by “Groups” and presented as jitter plots. Statistical tests were performed using generalized linear modeling (multiple comparisons of mean values: Tukey contrasts). *p ≤ 0.05, **p ≤ 0.005, ***p ≤ 0.0005. Sample sizes were: pre versus days 0–2: n fish = 38; water n = 13. Days 0–2 versus days 3–5: fish n = 30; water n = 10. Days 3–5 versus days 7–9: fish n = 32; water n = 11. Days 7–9 versus day 14: fish n = 21; water n = 8.

By day 14, the α-diversities of fish intestinal and water microbiome samples were not consistently different from those of samples collected before the housing switch.

Fish and water microbiomes displayed significant yet different changes in β-diversity after the housing switch

Taking the same approach outlined for the α-diversity “group” analysis, we examined the β-diversity of the different time points after the housing switch. Significant changes were observed in the composition of the fish intestinal microbiome throughout the duration of the experiment (Fig. 5). The composition of the tank water microbiome also significantly changed across the grouped time periods. However, with respect to the structure of tank water microbiomes, changes were only observed between samples from days 7 to 9 and day 14. The tank water microbiome on day 14 was significantly different from the prehousing switch tank water microbiome in both composition (p = 0.006) and structure (p = 0.008) (one-way PERMANOVA using adonis with the strata function controlling for tank identity).

FIG. 5.

Fish and water microbiomes display significant yet different changes in β-diversity after the housing switch. β-Diversities of the zebrafish intestinal and tank water microbiomes arranged by “Groups” and presented as PCoA plots based on the Jaccard and Bray–Curtis indices. Statistical tests were one-way PERMANOVAs using adonis with the strata function controlling for tank identity. Sample sizes were: pre versus days 0–2: fish n = 38; water n = 13. Days 0–2 versus days 3–5: fish n = 30; water n = 10. Days 3–5 versus days 7–9: fish n = 32; water n = 11. Days 7–9 versus day 14: fish n = 21; water n = 8.

The structure of the fish intestinal microbiome changed from before the housing switch to days 0–2 postswitch, but, after day 3 postswitch, the structure of the fish intestinal microbiome did not change (Fig. 5). When comparing the preswitch zebrafish intestinal microbiome with the days 3–5 postswitch intestinal microbiome (and to the intestinal microbiome on days 7–9 and 14), there was a significant difference in composition (p = 0.001) and structure (p = 0.045); however, the dominant bacterial taxa in the intestinal microbiome did not change (data discussed in detail hereunder).

These results for the fish intestinal microbiome are similar to what was observed in the α-diversity analyses, wherein the significant changes occurred fairly rapidly, that is, within 3–5 days of the housing switch (Fig. 4).

Fish and water microbiomes differed pre- and posthousing switch

With the specific analyses of α and β diversities complete, the final step was examining which bacterial types were driving the observed changes in both the fish intestinal and the tank water microbiomes. To visualize and elucidate these changes, heat maps, SIMPER analysis, and LEfSe were used. OTUs 1–5 accounted for 85% of the total 16S rRNA gene sequences obtained from the fish intestine, and OTUs 1–7 accounted for 90% of the total bacteria in the fish intestine (Supplementary Tables S1 and S2). Not surprisingly, one particular group of Vibrio (OTU1) was the most often detected, and the most relatively abundant, bacterium in both the fish intestines and the tank water (Fig. 6A and Supplementary Fig. S1).

FIG. 6.

Fish intestinal and water microbiomes changed after the housing switch. (A) Heat map of the OTU profiles of the fish intestinal and water microbiome samples from individual tanks. The numbers following the bacterial Genus names indicate the OTU number and the ranked average relative abundance of that OTU. All OTUs with a relative abundance >1% were included in the heat map. (B) Linear discriminant analysis Effect Size of the fish intestinal and water microbiomes indicating OTUs that increased or decreased in relative abundance after the housing change. The total number of samples were fish = 77; water = 27. OTU, operational taxonomic unit.

SIMPER analysis indicated that OTU1 and OTU2 (both of which are Vibrio) displayed increased relative abundances in the fish intestines after the housing change (Supplementary Table S1), whereas LEfSe analysis indicated that, controlled for tank identity, OTUs 26 (Staphylococcus) and 29 (Pseudomonas) increased in relative abundance within the fish intestinal microbiomes as a consequence of the housing switch (Fig. 6B).

For the tank water microbiome, the majority of the OTUs experienced an increased relative abundance after the housing switch, with some notable exceptions being OTUs 1 (Vibrio), 6 (Escherichia), 59 and 65 (both of which are Gammaproteobactera) (Fig. 6B). Even with the decrease in OTU1, this OTU still comprised a majority of the 16S rRNA sequences obtained from the tank water samples (Fig. 6A and Supplementary Table S2). The OTUs that experienced the largest increased relative abundance after the housing switch were OTUs 5 (Aeromonas), 12 (Comamonadaceae), and 3 (Enterobacteriaceae). Of note, these OTUs did not increase in relative abundance in the fish intestine (Fig. 6B).

Examining the fish intestinal microbiome in the different “Groups,” LEfSe analysis did not detect any OTUs that experienced significant changes in relative abundance immediately after the housing switch (preswitch vs. days 0–2 postswitch). When comparing days 0–2 versus days 3–5, however, a number of OTUs experienced a change in relative abundance. These included increased abundances of OTUs 5 (Aeromonas) and 51 (Ochrobactrum), and decreased abundances of OTUs 27 (Nitratireductor), 20 (unclassified; 71% sequence similarity to Novosphingobium), and 6 (Escherichia) among others (Supplementary Fig. S3A). In the next “Group” (days 3–5 vs. days 7–9 postswitch), only OTU6 (Escherichia) displayed an increase in abundance (Supplementary Fig. S3A).

In the last “Group” (days 7–9 vs. day 14 postswitch), four bacterial groups exhibited an increase in relative abundance, whereas none exhibited a decrease. These were OTUs 2 (Vibrio), 9 (Acinetobacter), 51 (Ochrobactrum), and 29 (Pseudomonas) (Supplementary Fig. S3A).

The “Group” profiles of the tank water microbiome showed more changes in individual OTUs than those of the fish intestines. Beginning with the preswitch water versus days 0–2 postswitch, three OTUs showed a decrease in relative abundance: OTUs 45 (Blastomonas), 11 (Pelomonas), and 22 (Pseudomonas), whereas two showed an increase: OTUs 31 (Undibacterium) and 59 (Gammaproteobactera). Between days 0–2 and 3–5 postswitch, the relative abundances of 10 OTUs increased with OTUs 8 (Shewanella) and 31 (Undibacterium) exhibiting the greatest change.

Between days 3–5 and 7–9 postswitch, two OTUs experienced a decline in abundance, one of which, OTU 8 (Shewanella), displayed an increase in number in the previous “Group.” Four OTUs, including two different Acinetobacter types (OTUs 144 and 201) exhibited an increase in abundance from days 3–5 to days 7–9 postswitch. By day 14 postswitch, 16 OTUs displayed a decrease in relative abundance, whereas 4 displayed an increase. Surprisingly, the two OTUs that displayed a decrease over time in the global LEfSe analysis (OTUs 1 [Vibrio] and 6 [Escherichia]) did not show a significant decrease when examining the different “Groups” (Supplementary Fig. S3B).

Discussion

Key findings of the study

Before the housing switch, the water microbiomes differed between tanks in composition and structure. Among the fish, gut microbiomes differed in composition but not structure among tanks, suggesting that relatively abundant bacterial types in the gut are common among zebrafish regardless of the tank they reside in. Upon moving the fish into the new housing system, the composition and structure of both the fish intestinal and tank water microbiomes changed.

The “Group” analysis indicated that most of the changes to the structure of the zebrafish intestinal microbiome occurred immediately after the housing switch and that the structure of the intestinal microbiome was then relatively stable throughout the remainder of sampling; this stability in structure existed despite changes to the composition of the zebrafish intestinal microbiomes occurring throughout the duration of the experiment. This again suggests that the relatively abundant members of the zebrafish intestinal microbiome are persistent and resilient to disturbance.

For the tank water microbiome, changes in composition were also observed throughout the duration of the experiment. However, significant changes in the structure of the water microbiome occurred later, notably, after the significant structural changes in the zebrafish intestines had already occurred. Vibrio (OTU1) was the most often detected bacterium in both the fish intestines and the fish water, particularly before the housing switch. In addition, OTU1 and OTU2 (both of which are Vibrio) displayed increased abundance in the fish intestines after the housing switch. For the tank water microbiome, the vast majority of the OTUs experienced an increased relative abundance after the housing switch, with Vibrio and Escherichia becoming less relatively abundant.

Findings within the zebrafish intestines

With the increasing appreciation of the importance of animal gut microbiomes to overall health, any study of a pathogenic organism that colonizes the intestines of a host organism must take into account the condition of the homeostatic host microbiome. The work described here examined the changes to the zebrafish intestinal and tank water microbiomes after a housing switch from individual separate tanks to a shared recirculating system in which all tanks share a common water supply. Each of the individual tanks before the switch had its own microenvironment, whereas the recirculating system provides a very stable aquatic environment for all the tanks.

We hypothesized that this housing switch would cause changes to the fish intestinal microbiome, which is supported by the observations we described previously. We have described not only the normal resident bacteria of the zebrafish microbiome and the tank water, but also demonstrated the plasticity these microbial communities have when facing external pressures in terms of composition and structure. Although changes were observed in many different bacterial groups, the relative abundances of the primary members of the fish intestines, such as Vibrio, Enterobacteriaceae, Aeromonas, and Cetobacterium, were persistently present and relatively abundant regardless of the external pressures.

One drawback to this type of study is that the 16S rRNA gene sequencing targeting the V4 hypervariable region does not classify bacteria down to the species level. Therefore, different Vibrio species that may be present in the fish intestinal and water microbiomes in this study are unknown, and identifying them would require additional analyses. What is clear is that changing the housing system had a significant effect on the microbiomes of both the fish intestines and the tank water. Furthermore this study demonstrates, unsurprisingly, that when zebrafish or other aquatic animals are housed in separate and isolated tanks, the microbial composition of the water in those tanks varies.

For the zebrafish in separate tanks, we observed, as others have observed previously, that age and housing play significant roles in the composition of the intestinal microbiome.^31,32 This is evidenced by the fact there were no significant differences in the intestinal microbiomes between the zebrafish and Casper fish in tanks 2 and 3, both of which were notably older than the zebrafish in tank 1 whose fish had an intestinal microbiome that was significantly different in terms of composition from the Casper fish in tank 3. The lack of a significant difference between the zebrafish in tank 1 and 2 however, is likely owing to the zebrafish originating from the same source, and the fact that the older zebrafish were used for breeding the younger fish.

Furthermore, the Casper fish were initially housed in a different location than the zebrafish. Given that the intestinal microbiomes of zebrafish have been demonstrated to significantly differ based on their habitat, this could also account for the observed significance. Although we did not include a negative control for this study (i.e., leaving some of the zebrafish from each tank in the separated tanks and placing others on the new housing system), because of logistical and practical constraints, no significant changes were observed to the structures of the zebrafish intestinal microbiomes when individually examining the three time periods sampled before the housing switch, suggesting that they were generally stable.

Our data demonstrate that immediately after the housing switch various new bacteria were introduced into the fish intestines. Although the composition of the fish intestinal microbiome continued to change throughout the experiment, its structure was stable by days 3–5 postswitch. At this point, a new stable intestinal microbiome structure, that was different in structure and composition from the original pre-switch microbiome, was established. This suggests that various bacterial types were continually entering the fish intestines, and although some may have been able to persist in the gut, after days 3–5, they were not able to cause salient changes to the microbiome structure overall.

The LEfSe “Group” data analysis demonstrated large fluctuations in the bacterial profile in the first two data time points, particularly in the 0–2 versus 3–5 days range. The trend would suggest that from the moment of the switch to days 0–2 (and perhaps 3–5 days) postswitch, there were enough bacteria moving both in and out of the gut (driven either by microbial competition and/or passive efflux) for a significant change of structure to occur. Then, by days 7–9 postswitch, the fish intestinal microbiome began to stabilize.

With 85% of the zebrafish intestinal microbiome comprising five OTUs, it appears that only a few select OTUs are driving the significant changes and/or maintaining homeostasis in the zebrafish intestine. Although there were significant changes in both the composition and structure of the zebrafish intestinal microbiome after the housing switch, the majority of the intestinal microbiome still consisted of Proteobacteria. Some species of Firmicutes increased, whereas species of Actinobacteria experienced increased or decreased relative abundance. This suggests that the zebrafish microbiome allows for some flexibility and variation, but that it is ultimately constrained in its structure. This further supports the view that the structure of the microbiome in the zebrafish intestine is shaped primarily by selective pressures within the host gut.^31,33,34

It is worth noting that the bacterial OTUs comprising the majority of the microbial profile in the fish intestines and tank water differ from other studies with similar focuses. Our study is similar however, in that the majority of the bacteria we characterized in the fish intestines are Proteobacteria. Furthermore, similarity between fish intestinal and water microbiomes will unavoidably overlap because of the fish consuming water and defecating into it. We suspect that our OTU data differ from those in other studies because although the zebrafish intestines can exert internal selective pressures on the microbiome, thus ensuring the presence and survival of “essential” core bacteria, the intestinal microbiome can still vary significantly based on housing location (i.e., the Roeselers et al.³² study, which showed that zebrafish from different locations have different intestinal microbial profiles, however they still share a core intestinal microbial profile). Indeed, our data itself support this notion as we observed the intestinal microbiomes of our zebrafish to change in this experiment while the dominant members persisted and remained relatively abundant.

Although this analysis did not take into account the sex of the examined fish, considering that there were no significant differences in the structures of the zebrafish microbiomes in the different tanks before the housing switch and that zebrafish are gonochoristic, significant structural differences in the intestinal microbiomes of male and female zebrafish are unlikely.^60,61

Findings within the tank water

Changes to the microbial profile of the water were persistent throughout the duration of the experiment (posthousing switch). Unlike the zebrafish intestine however, significant structural changes were not observed immediately after the change in housing systems. One possible reason for this is that when the new system was first brought online, roughly half the water used in the system came from existing water in the separated tanks. Second, the changes observed in the microbial profile of the zebrafish intestine differed from those of the tank water; the rate at which these changes occurred could be completely different. A third possibility is that with the lower sample number we were not able to observe changes at the same rate as the fish intestines.

The β-diversity analysis revealed significant changes occurring to the composition of the tank water microbiome at all tested intervals (as with the zebrafish intestinal microbiome). However, significant structural changes were only observed at the last time point (i.e., days 7–9 vs. day 14). In comparison with the fish intestinal microbiome, where just a few of the OTUs make up the majority of the bacterial community and drove the observed significant changes, for the water microbiome, the nine most prominent OTUs accounted for ∼66% of the total bacterial diversity of the water, and some of those OTUs were much more prominent in the water than the fish intestines.

Because a larger number of OTUs comprised small percentages of the overall bacterial profile in the water, it suggests that all these small but numerous changes to the OTU profile are what resulted in the observed significance to the tank water. It also demonstrates, unsurprisingly, that the bacterial profile of the water is much more plastic than that of the fish intestine. The global β-diversity patterns suggest that the observed changes in the water microbiome were incremental, and only when taken together did those changes become structurally significant.

Only five of the OTUs experienced a decrease in their relative abundance when examining the global profile of the tank water. Two of these OTUs were Vibrio and Escherichia, both prominent members of the water microbiome; their decreased relative abundance after the housing switch may have created opportunities for other bacterial types to grow or it may have been a reflection of competition and/or the increased influx of bacteria as a consequence of new microbiome members being introduced into the tanks. Regardless, OTU1 (Vibrio) was consistently abundant in the water throughout the experiment.

The β-diversity analyses indicated that the tank water microbiomes changed at different rates inside the three fish tanks before eventually reaching a new equilibrium by day 14 postswitch. By day 14 postswitch, all the bacteria that were increasing in number were nitrogen fixing, suggesting that it took time for these bacteria to adjust to the change in the housing system and to reestablish themselves. This may also account for the significant difference in the structure of the water microbial community day when comparing days 7–9 with day 14.

Conclusions

Although studies on other organisms that exactly mirror the experimental parameters described in this study do not seem to have been conducted, there are numerous studies that have examined the effects of the housing environment or external factors in general on other organisms' intestinal microbiomes. Overall, our data support what these other groups have observed, which is that the external environment significantly affects an organism's intestinal microbiome. For example, a study examining Zucker rats conducted by Lees et al.⁶² found that although age was the most significant source of variation in fecal microbiota composition, cage environment played a significant role in microbiota structure.

Similar trends have been observed in various species of chickens, where host-related factors (i.e., age, sex, and breed), have a large effect on intestinal microbiota.⁶³ As we report here with the zebrafish, environmental factors, such as housing, litter, feed access, and climate have an effect on the composition of the intestinal microbiome among chickens.⁶³ Another study conducted by Kubasova et al.⁶⁴ found that housing systems influence gut microbiota composition in sows but not their piglets. The study found that sows from enriched systems (i.e., containing substrates for the pigs to manipulate such as straw and extra pen space) contained a higher relative abundance of various genera of bacteria compared with sows housed in conventional (i.e., concrete floor and smaller pen space) systems.

They then concluded that the production system influenced the microbiota composition, with the ingestion of the straw being the most likely cause. Furthermore they noted that for 1- and 4-day-old piglets, the microbiota differed significantly from that of their parent sows, demonstrating that in the early days of piglets' lives, parent sows are not the most important source for intestinal microbial colonization.⁶⁴

This study is the first, to our knowledge, to examine how the zebrafish intestinal and tank water microbiomes are altered after a housing system change. The specific mechanisms behind these observed changes remain to be elucidated. A possibility however is that the tank water microbiome, being in an inchoate state in the new housing system, resulted in significant changes to the zebrafish intestinal microbiome. As time went on, certain types of bacteria, because of shedding from the fish, the increase in ammonia levels in the water, or other factors, were able to take advantage of this niche in the water, and establish a new “normal” environment. This suggests that the zebrafish intestinal microbiome is significantly influenced by the aquatic environment that houses the zebrafish and the microbiome of the zebrafish is in turn, able to influence the environment that houses them.

The information obtained from these experiments will pave the way for future studies examining the effects of aquatic pathogens on the zebrafish intestinal microbiome. In addition, this study demonstrates that the zebrafish intestinal microbiome, while robust (because of selective forces within the host gut), is susceptible to external pressures; specifically, the microbial profile of the tank water that houses the fish.

Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Data

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Table S1

Supplementary Table S2