Common laboratory diets differentially influence zebrafish gut microbiome’s successional development and sensitivity to pathogen exposure

Background: Despite the long-established importance of zebrafish (Danio rerio) as a model organism and their increasing use in microbiome-targeted studies, relatively little is known about how husbandry practices involving diet impact the zebrafish gut microbiome. Given the microbiome’s important role in mediating host physiology and the potential for diet to drive variation in microbiome composition, we sought to clarify how three different dietary formulations that are commonly used in zebrafish facilities impact the gut microbiome. We compared the composition of gut microbiomes in approximately 60 AB line adult (4- and 7-month-old) zebrafish fed each diet throughout their lifespan. Results: Our analysis finds that diet has a substantial impact on the composition of the gut microbiome in adult fish, and that diet also impacts the developmental variation in the gut microbiome. We further evaluated whether the 7-month-old fish microbiome compositions that result from dietary variation are differentially sensitive to infection by a common laboratory pathogen, Mycobacterium chelonae. Our analysis finds that the gut microbiome’s sensitivity to M. chelonae infection varies as a function of diet, especially for moderate and low abundance taxa. Conclusions: Overall, our results indicate that diet drives the successional development of the gut microbiome as well as its sensitivity to exogenous exposure. Consequently, investigators should carefully consider the role of diet in their microbiome zebrafish investigations, especially when integrating results across studies that vary by diet.


Introduction
In the effort to understand how the gut microbiome mediates vertebrate health, zebra sh (Danio rerio) have emerged as an important microbiome experimental model organism [1]. Despite the increasing use of zebra sh in microbiome research, key knowledge gaps remain about how different zebra sh husbandry practices, especially diet, in uences microbiome composition [2,3]. For example, in contrast to mice, zebra sh do not have a standard reference diet [4]. Instead, zebra sh research facilities vary by dietary husbandry practice, which can impact physiological and reproductive outcomes [5][6][7]. Given that diet plays an important role in shaping the composition of the gut microbiome in humans and across vertebrate and invertebrate animal models, such as mice and honeybees [8][9][10][11][12][13], we hypothesize that variation in dietary husbandry practice also impacts the composition of the zebra sh gut microbiome.
Quantifying this association is important because it could explain why, despite the existence of a core gut microbiome, gut microbiome composition differs across research facilities [14,15], improve efforts to integrate data across investigations, and clarify how dietary variation manifests as physiological variation.
Relatively little is known about how variation in dietary husbandry practice impacts the zebra sh gut microbiome. Prior studies that measured the impact of diet on the zebra sh gut microbiome have largely considered how substantial variation in speci c macronutrients impacts the gut microbiome (e.g., high fat versus low fat diets) [6,[16][17][18]. This variation is not typically representative of the variation in nutrient content observed across standard dietary husbandry practices [4,5]. Additionally, these studies have typically reared sh on a singular diet up to the point of experimentation, at which point sh are exposed to alternative diets. While insightful about acute effects, such experimental designs do not model the chronic dietary exposure that sh experience through husbandry. This prior work also does not typically consider how diet impacts the microbiome at different sh developmental periods, or whether dietary variation affects other characteristics of the gut microbiome, such as its sensitivity to exogenous agents (e.g., pathogens).
In this study, we sought to de ne the impact of rearing sh of different common facility diets on the gut microbiome of early adult (4-mo old) and fully mature (7-mo old) zebra sh. To do so, we reared sh throughout their lifespan on one of three different dietary husbandry practices: sh were fed either (1) the Gemma (Skretting, Fontaine les-Vervins, France) diet, which is a commercial feed widely used in zebra sh research facilities, (2) the ZIRC diet, a compound diet mixed and adopted by the Zebra sh International Research Center (ZIRC), which is one of the largest zebra sh stock centers in the world, or (3) a precisely de ned laboratory grade diet developed by Watts [5]. Overall, these diets are relatively similar from a macronutrient perspective, though they differ by formulation, ingredient sourcing, manufacturing process details, and consequently also by exact nutritional content. (Table S4.1.1). In particular, we evaluated how the microbiome differed across these groups of sh as well as over development. We also determined if these differences link to variation in sh weight and condition factor, as well as variation in how the microbiome responds to infection by one of the most common infectious agents of zebra sh research facilities, Mycobacterium chelonae [19].

Diet differentially in uences physiology and gut microbiome at 4 months old
To determine how common zebra sh diets differently impact sh size (length and body condition score) and the gut microbiome, we reared 176 zebra sh that were assigned one of three diets from 1-to 4months-post fertilization (mpf) ( Figure 1): Gemma, Watts and ZIRC diets. Prior to diet assignment, sh were fed a nursery diet (see methods). At 4 mpf, we selected 89 individuals across these three cohorts and collected fecal samples from each sh for microbiome pro ling prior to measuring their weight and body condition score (BCS). Wilcoxon Signed-Rank Tests found that diet and sex signi cantly associated with weight and BCS. Female sh had higher weight (Z = 1,530, P < 0.001; Table S1.1.2) and BCS (Z = 1,631, P < 0.001; Table S1.1.4) compared to males. Between the three diets, ZIRC-diet fed sh had the highest mean BCS compared to sh fed Gemma-(Z = 150, P < 0.001) and Watts-diet (Z = 197, P < 0.001, Table S1.1.3). Gemma-and Watts-diet fed sh did not signi cantly differ from one another in terms of weight and BCS. These results indicate that ZIRC-diet contributes to heavier sh compared to Gemmaand Watts-diet fed sh. We next built generalized linear models (GLM) to determine if diet associated with variation in one of three measures of microbiome alpha-diversity: richness, Simpson's Index, and Shannon Entropy. An ANOVA test of these GLMs revealed that alpha-diversity varies as a function of diet for all three measures of diversity we assessed (P < 0.05; Fig 1C; Table S1.2.1). A post hoc Tukey test clari ed that ZIRC-and Watts-diet fed sh exhibited signi cant differences in alpha-diversity as measured by richness and Shannon Entropy (P < 0.001, Table S1.2.2). Moreover, we observed signi cant differences in diversity between Gemma-and Watts-diet fed sh in terms of richness (P < 0.001; Table S1.2.2), and between Gemma-and ZIRC-diet fed sh when considering the Simpson's Index (P < 0.001; Table S1.2.2). These results indicate that diet associates with sh gut microbiome diversity, and that diet may differentially impact rare and abundant microbial members of the gut.
To evaluate how diet associates with microbiome community composition, we quanti ed the Bray-Curtis, Canberra and Sørensen dissimilarity amongst all sample. We detected a signi cant clustering of microbial gut community composition based on diet as measured by all beta-diversity metrics (PERMANOVA, P < 0.05; Figure 2C, Table S1.3.1). These results indicate that microbial communities of sh fed the same diet are more consistent in composition to one another than to sh fed other diets. Additionally, we assessed beta-dispersion, a measure of variance, in the gut microbiome community compositions for each diet group. We nd the beta-dispersion levels were signi cantly different between the diet groups as measured by Bray-Curtis and Canberra metrics (P < 0.05; Table S1.4.1). Beta-dispersion levels were signi cantly reduced in Gemma-diet fed sh compared to Watts-diet fed sh when measured by Bray-Curtis metric, as well as signi cantly reduced compared to Watts-and ZIRC-diet fed sh when measured by Canberra metric (Table S1.4.1). These results indicate that Gemma-diet fed sh are more consistent in community composition than Watts-and ZIRC-diet fed sh at 4 mpf. Collectively, these results indicate that 4 mpf sh gut microbiome communities stratify by diet, but the composition of these microbial communities differ in consistency depending on diet.
Finally, to better understand the interactions between the diet and the members of the gut microbiome community, we quanti ed differential abundance using ANCOM-BC2. We observed 24 signi cantly abundant taxa at the genus level in at least one of the three diets (Table S1.5.1). Gemma-diet fed sh were enriched for Chitinibacter and were depleted of Aeromonas and Flavobacterium. Watts-diet fed sh enriched for Flavobacterium, ZOR0006, Peptostreptococcus, Cetobacterium, Tabrizicola, Cellvibrio, and unnamed genera of Microscillaceae and Chitinibacteraceae, and depleted of Crenobacter and a Sutterellaceae genus. ZIRC-diet fed sh enriched for Cloacibacterium and Acinetobacter, and depleted of Fluviicola. Many of these taxa are identi ed as common members of the zebra sh gut microbiome [14,15]. These results indicate that diet differentially supports particular members of the zebra sh microbiome community.

Diet impacts the successional development of the zebra sh gut microbiome
To determine how maintaining sh on different diets impacts the development of the gut microbiome, we continued to grow sh from the same diet cohorts until 7 months post fertilization (mpf; Figure 1).
Microbiome samples were collected from cohort members prior to quanti cation of sh weight and body condition score. To determine the effect of diet on the body condition score and the gut microbiome of 7mpf sh, we conducted the same analyses as we applied to the 4 mpf sh. At 7 mpf, we nd body condition score is signi cantly associated with diet (P < 0.05; . These results demonstrate that diet impacts the physiology and gut microbiome of 7mpf sh. Next, we compared our results between the 4 and 7 mpf sh to determine how diet impacts the successional development of the gut microbiome. Linear regression revealed microbial gut alphadiversity was signi cantly associated with the main effect of time (P < 0.05; Table S2.2.1) for each diversity metric. However, we did not nd a diet dependent effect on time for any alpha-diversity metric we assessed (P > 0.05; Table S2.2.1). A post hoc Tukey test clari ed that microbiome diversity was signi cantly different between 4 and 7 mpf Gemma-and ZIRC-diet fed sh as measured by the Shannon and Simpson's alpha-diversity metrics (P < 0.05; Figure 3C, Table S2.2.2), but we did not nd a statistically signi cant association between 4 and 7 mpf Watts-diet fed sh with any alpha-diversity metric (P > 0.05; Table S2.2.2). These results indicate that the alpha-diversity of the gut microbiome of Watts-diet fed sh were temporally stable, while Gemma-and ZIRC-diet fed sh diversi ed over time in diet-consistent ways.
A PERMANOVA test of the 4-and 7-mpf samples using the Bray-Curtis dissimilarity metric revealed that community composition was best explained by diet (P < 0.05; Figure 2C, Table S2.3.1), but an analysis using the Canberra measure found that variation in microbiome composition was best explained by time (P < 0.05; Fig 2D, Table S2.3.2). Given how these metrics weight the importance of abundant versus rarer taxa, respectively, these results indicate that abundant members of the microbiome community are more sensitive to the effects of diet, while rarer community members are sensitive to the effects of time. Moreover, we found beta-dispersion levels were signi cantly elevated between 4 and 7 mpf Gemma-diet sh when considering the Bray-Curtis and Sørensen metrics, in Watts-diet fed sh when considering the Canberra and Sørensen metrics, and in ZIRC-diet fed sh across all three beta-diversity metrics (P < 0.05; Table S2.4.1-3). These results indicate that abundant and rarer gut microbiome community members were differentially impacted by the effects of time depending on diet. Collectively, these results indicate that diet can have a substantial impact on how the gut microbiome successionally develops in zebra sh.
Differential abundance analysis revealed taxa that were signi cantly associated with the effects of time and diet one of the diets (Table S2.5.1). Across all three diets, the taxa that were more abundant included Fluviicola, Macellibacteroides, Bacteroides and an unnamed genus in the Barnesiellaceae family were , while taxa that were less abundant included Phreatobacter and Flavobacterium. These results indicate that irrespective of diet, the abundances of taxa change over the course of zebra sh development. We also measured how taxon abundance changed over time within each diet ( Figure  S2.5.2-46.2.5). The Gemma-diet fed sh were uniquely enriched for Exiguobacterium (Table  S2.5.2). Exiguobacterium are gram-positive facultative anaerobes in the phylum Bacillota, and are linked to fatty acid metabolism in zebra sh [20,21]. The Watts-diet fed sh were uniquely depleted of Gemmobacter (Table S2.5.3). Previous work has found that Gemmobacter has a positive association with parasite exposure in infected zebra sh [22,23]. The ZIRC-diet fed sh were uniquely enriched for Pseudomonas and Haliscomenobacter (Table S2.5.4). Pseudomonas is a common member of the gut microbiome and associated with fatty acid metabolism in zebra sh [20]. Less is known about the Haliscomenobacter genus, but an analysis of its genome revealed it is an aerobic chemoorganotroph found in aquatic systems [24]. Together, these results indicate that particular members of the gut microbiome associate with diet and zebra sh development.
To determine if sh size associated with diet across zebra sh development, we used Wilcoxon Signed-Ranks Tests to identify parameters that best explained the variation in body condition score (BCS) between 4-and 7-mpf sh. At 7mpf, the BCS signi cantly differed between sh fed different diets. However, we did not nd that BCS of sh were impacted by time (P > 0.05; Fig 2E,  Canberra and Sørensen beta-diversity metrics, there were signi cant main effects of body condition score, and signi cant interaction effects between BCS and diet (P < 0.05; Table S2.1.2.2). However, the model coe cient for the effect of body condition score and its interaction with diet is far smaller than the coe cient for the effect of diet (Table S2.1.2.2). We did not nd a signi cant association between BCS and speci c taxon abundance (Table S2.1.2.2). Collectively, these results indicate that while the gut microbiome's composition associates with BCS, the effect of diet on the gut microbiome is much stronger.
3. Diet in uences gut microbiome's sensitivity to pathogen exposure Lastly, we sought to determine how diet impacts the gut microbiome's sensitivity to exogenous stressors, in particular exposure to the common pathogen of zebra sh, Mycobacterium chelonae. Mycobacteria has been reported in zebra sh from about 40% of research facilities [25].
The infection is usually only diagnosed by histology, and hence s only diagnosed to the genus level based on the presence of acid-fast bacteria. When species identi cations are made using molecular methods, the identi cation is most frequently M. chelonae [26]. It is hypothesized to be introduced through diet early in life [25,27,28]. M. chelonae forms granulomas coelomic organs, swim bladder and kidney, and in many cases it ultimately causes death. These can introduce inconsistencies in study outcomes, but the impacts on the gut microbiome are not known [25]. To clarify effect of M. chelonae infection on the gut microbiome, and whether these effects vary by diet, we injected M. chelonae into the coelomic cavities of sh from each diet cohort at 4 mpf following fecal collection. These M. chelonae injected sh comprised the pathogen exposure cohort for this experiment, which we compared to the remaining, unexposed cohort of sh. At 7 mpf, we collected fecal samples from exposed and unexposed sh to measure microbial gut diversity, composition, and taxon abundance, performed a histopathological analysis of intestinal tissue to assess infection severity, and measured body condition score.
We rst evaluated whether diet impacted infection outcomes, as determined by histological con rmation of infection 3.5 months following pathogen injection. We conducted a Chi-Square test to compare the infection count between sh fed the three diets. The results showed that there was a statistically signi cant difference in proportion of positive infection counts between the groups, X 2 (2, N = 66) = 11.519, P < 0.05 (Table S3. included the testis, coelomic cavity, swim bladder and kidney (Figure S3.1.2). With females, all showed the infections within the ovaries, with one with a coelomic infection. Colonization of the intestinal lumen by acid fast bacteria were observed in 17 exposed and 7 control sh across the diets. This result indicates that the diets considered in our study appear to dictate the progression of infection of M. chelonae, but of the samples we collected for microbiome analysis we may be underpowered to detect a difference. Next, we assessed whether infection status links to body condition score as well as measures of gut microbiome diversity and composition. We did not observe signi cant associations between infection status and body condition score based on linear regression (P > 0.05; We next considered that exposure to the pathogen could impact the gut microbiome, even though ultimate infection outcomes among exposed individuals may not. Comparing exposed to unexposed sh found that microbial gut diversity signi cantly differs between exposure groups as measured by richness and Shannon Entropy alpha-diversity metrics (P < 0.05; Figure 4A, Table S3.2.1). That said, based on linear regression, the impact of exposure on the gut microbiome alpha-diversity does not appear to differ as a function of diet, as the interaction term for these covariates did not yield a signi cant effect (P > 0.05; Table S3.2.1). Furthermore, we used a post hoc Tukey test to clarify whether microbial gut diversity of sh differed between exposure groups by diet. Unique to ZIRC-diet fed sh, we observed microbiome diversity differed in unexposed controls compared to exposed sh as measured by all alpha-diversity metrics (P < 0.05, Table S3.2.2). Watts-diet fed sh differed in unexposed controls compared to exposed sh in terms of richness (P < 0.05, Table S3.2.2). These results suggest that the gut microbiome diversity of ZIRC-diet fed sh, and to some extent Watts-diet fed sh, are sensitive to the effects of M. chelonae exposure, but Gemma-diet fed sh are resistant to pathogen exposure. While the gut microbiomes are sensitive to the effects of pathogen exposure, we nd the statistical effect of diet shaping the gut microbiome is an order of magnitude greater across all alpha-diversity metrics (P < 0.05, Table S3.2.1). Collectively, these results indicate that gut microbiome diversity is sensitive to M. chelonae exposure, but diet is the primary driver of gut microbiome diversity.
Next, we evaluated how pathogen exposure in uenced microbial community composition across sh fed each diet. For each beta-diversity metric considered, PERMANOVA tests found that the main effects of diet and pathogen exposure signi cantly explained the variation in microbiome composition, but that the main effect of diet was consistently larger than the effect of exposure (P < 0.05; Fig 4C, Table S3.3.1). Furthermore, a PERMANOVA test found that the model coe cient effect for the interaction of diet and pathogen exposure was statistically signi cant when considering Canberra and Sørensen beta-diversity metrics, however this effect was marginal as compared to the aforementioned main effects. Moreover, a pairwise analysis of beta-dispersion did not nd signi cant levels of dispersion between exposed and unexposed sh within each diet (P > 0.05; Table S3.4.1-3). These results indicate that exposure to M. chelonae did not affect dispersion of the gut microbiome communities. Collectively, these results indicate that the gut microbiome is sensitive to pathogen exposure, but that dietary effects tend to overwhelm evidence of this sensitivity.
We also observed several microbiota that strati ed exposed and unexposed groups of sh in both dietrobust and diet-dependent manners. Unexposed Gemma-diet fed sh were enriched for Macellibacteroides and Aurantisolimonas (Table S3.5.2), unexposed Watts-diet fed sh were enriched for an unnamed genus of Barnesiellaceae, Fluviicola, Paucibacter, and Brevibacterium (Table S3.5.3), and unexposed ZIRC-diet fed sh were enriched for Macellibacteroides, Bacteroides, Mycobacterium and unnamed genera of Barnesiellaceae and Sutterelaceae (Table S3. 5.4). Across all the diets, the taxa that were more abundant in unexposed, control sh included Macellibecateroides, Fluviicola, Bacteroides, Aurantisolimonas, Cerasicoccus, and three unnamed genera of Barnesiellaceae, Commonadaceae, and Sutterellaceae. Plesiomonas were more abundant in exposed sh compared to controls (Table S3.5.1). These results indicate that pathogen exposure impacts the abundance of certain taxa within and across the diets. Next, to see if Mycobacterium species abundance differed from background, pre-exposure levels we compared Mycobacterium abundance between pre-exposure and unexposed control sh to that of exposed sh within each diet. Unexposed Gemma-and ZIRC-diet fed sh had signi cantly higher abundances of Mycobacterium to exposed ( Figure 4D, Table S3.5.5). Pre-exposed Watts-diet fed sh had signi cantly more Mycobacterium compared to pre-exposed sh, but they did not differ signi cantly from unexposed control sh. These results indicate that the abundance of taxa from the genus Mycobacterium changes in response to exposure to a pathogenic species in a diet-dependent manner.

Discussion
Zebra sh are an important emerging model organism for understanding the microbiome. Yet, there is little consistency across studies in terms of the husbandry practices used to conduct zebra sh microbiome experiments, especially in terms of diet. This lack of consistency likely stems from a dearth of knowledge about how different standard zebra sh diets impact study outcomes, both in terms of the gut microbiome's composition as well as the physiological endpoints of the host. Our study offers critical insight into how three standard zebra sh dietary formulations impacts these outcomes, nding that the zebra sh gut microbiome's development and response to pathogen exposure is sensitive to diet. These observations help clarify inconsistencies across studies, underscore the importance of considering diet when integrating data across investigations, and inform on efforts to develop standard approaches in zebra sh microbiome research.
We found that diet had a substantial impact on the structure of the gut microbiome in adult zebra sh. Previous research has found that diets with varying compositions of key macronutrients (e.g., protein, lipids and ber content) impacts zebra sh physiology and the gut microbiome [5,[16][17][18][29][30][31][32]. Moreover, diet's effect on restructuring the host's gut microbiome has been observed across an evolutionarily diverse array of vertebrate and invertebrate animal hosts[8, 9,11,12,33]. However, the nutritional compositions used in these prior studies tend to vary considerably. In particular, the feeds our study considered are far more consistent in their composition than the diets that are typically included in studies of the effect of diet on the gut microbiome (e.g., high-fast v. low-fat diets). Moreover, a unique strength of our study is that sh were fed the same diets over the vast majority of their lifespan (30 to 214 dpf), which is more consistent with a standard husbandry approach that maintains sh on a speci c diet than the relatively short-term exposures to different types of diet that are typically employed in related research. Because of these features of our experimental design, our work provides important clarity into how seemingly subtle differences in husbandry practice can result in substantial differences in the composition of the adult zebra sh gut microbiome.
We also found that diet impacts the developmental variation in the gut microbiome. Prior work investigating the successional development of the zebra sh gut microbiome has had inconsistent results; our efforts indicate that these inconsistencies may be attributable to the different diets utilized in these prior studies 16 [29,31,32]. While our study differed in exact length and sampling time points as compared to these prior studies, we do nd congruent trends in gut microbiome diversity to other zebra sh studies that sampled within similar developmental periods as those interrogated in our investigation. However, it is di cult to directly compare our results to these prior studies because they sampled at different time points, used a variety of diets throughout their study, used diets different from those included in our study, or did not disclose which diets were used. It is worth nothing that while our sh were fed the same diet from 30 days onward, at 114 dpf sh in our study were switched from a juvenile formulation to an adult formulation of their respective diets. These formulations differed slightly in some diets (e.g., Gemma and Watts), but in others more substantially (e.g., ZIRC), which may contribute to the variability we observed in the gut microbiome between diets across zebra sh development. Despite these limitations, we found adult zebra sh fed diets of similar nutritional composition manifest distinct gut microbiome successional patterns in community compositions across adulthood. Future work should seek consistency in diet formulations and increase sampling time points throughout zebra sh development to further clarify the successional development of zebra sh gut microbiomes.
Finally, we observed that the gut microbiome of zebra sh were sensitive to pathogen exposure, but diet was the main driver of gut microbiome structure. We ensured all sh were exposed to the pathogen by injecting Mycobacterium chelonae into the coelomic cavities of the sh at 4 mpf. We found that presence of infection was not su cient to explain associations with microbiome diversity or community composition, which is likely due to being underpowered to detect them. We found infection by diet interactions on a larger number of individuals that were assessed for histopathology, but not with the subset of sh sampled for microbiome analysis. Therefore, having a su ciently large sample size is important for observing infection effects on the gut microbiome. However, we found that gut microbiome diversi cation did not change after exposure to M. chelonae uniquely in ZIRC-diet fed sh relative to their unexposed controls. Our results contrast our prior work that found exposure to an intestinal helminth was associated with an increase in microbiome diversity [22]. One possible explanation for this discrepancy is our prior study investigated an intestinal helminth which may have different impacts on the gut microbiome associated with differences in intestinal lesion to that of a pathogenic bacterial species. For example, the nematode Pseudocapillaria tomentosa penetrates the intestinal epithelium and causes profound pathologic changes [22], whereas disease caused by Mycobacterium species in zebra sh are characterized by extra-intestinal infections [25]. Mycobacterium spp in zebra sh are hypothesized to be introduced early in life through ingestion, including diet [28,34], while sh in our study were exposed by injection into their coelomic cavities at adulthood when their gut microbiomes have been rmly established. Priority effects may have hindered the injected species of Mycobacterium from more substantially altering the gut microbiome at adulthood than if it had been introduced through a natural route during early-life microbiome assembly [35]. Future work should consider using a natural mode of infection and exposing sh to a variety of pathogens to elucidate the gut microbiome's role in mediating pathogen exposure. Furthermore, because we found that the effect of diet was far greater than pathogen exposure on shaping the gut microbiome, future studies must consider diet effects, as they may overwhelm infection effects.
In conclusion, we found diet is one of the most important factors driving variation in the zebra sh gut microbiome. Unlike prior studies, including the extensive research conducted in mammalian models, that have evaluated dietary effects on the gut microbiome using diets that fundamentally differ in macronutrient composition, our work reveals that even relatively consistent diets that are commonly selected as normal husbandry practices elicit these large impacts on microbiome composition. While the zebra sh gut microbiome differs taxonomically from other animal systems, there is a substantial amount of shared functional capacity between zebra sh and mammalian gut microbiomes [36]. Consequently, the taxa-speci c associations we found here may not directly translate to other animal systems, but the interactions between the microbiome, diet and pathogen exposure may be similar. Future work should illuminate the underlying mechanisms of the diet's in uence on zebra sh development, gut microbiome structure and the microbiome's sensitivity to pathogen exposure. Collectively, our study demonstrates that investigators should carefully consider the role of diet in their microbiome-targeted zebra sh investigations, especially when integrating results across studies that vary by diet.

Conclusions
Collectively, our study demonstrates the effect of commonly used laboratory diets on the gut microbiome of zebra sh. We reared zebra sh across their lifespan on three commonly used diets and analyzed the gut microbiome of juvenile and adult sh. Our ndings demonstrate that diet impacts the developmental trajectories of the zebra sh gut microbiome, even with similar nutritional compositions. Additionally, diets were found to sensitize the gut microbiome to pathogen exposure. These results have important implications for the practice of zebra sh husbandry and the selection of diets in microbiome studies. Our ndings will also contribute to ongoing discussions about standardizing husbandry practices, including diet, in the zebra sh research community.

Fish Husbandry
A total of 270 30 days post fertilization (dpf) AB line zebra sh were randomly divided into eighteen 2.8 L tanks (15 sh/tank) on a single pass ow-system tanks (15 sh/ tank). During the experiment, temperature was recorded daily and ranged from 25.5-28.3°C, with the exception of two isolated overnight temperature drops below that range due to two separate power loss events that affected the source water sump heater. All other water conditions were monitored weekly, pH ranged from 7.0-7.6, total ammonia ranged from 0-0.25 ppm (measured with pH and ammonia API test kits; Mars Fishcare North America Inc. Chalfont, PA), and conductivity ranged from 109 −166 microsiemens. Light in the vivarium was provided for 14 hours/day. One plastic aquatic plant piece approximately 6 inch in length was added to each tank for enrichment when sh were 214 dpf. A stock of similarly aged Casper line sh were maintained for the duration of the experiment, with a third of the stock being maintained on each of the diet regimens matching the AB line zebra sh. These sh served as ller sh and were added to the tanks after each histological sampling time point to maintain the 15 sh/tank ratio required to maintain the prescribed diet volumes per feeding.

Diets
Fish were all fed the same nursery diet until 30 dpf, a combination of paramecia, brine shrimp, and the ZIRC Nursery Mix: Zeigler AP Larval Diet (Ziegler Bros Inc., Gardners, PA) and freeze dried rotifers. Fish were then transferred to the OSU facility and assigned randomly to one of three juvenile diets: Gemma Micro 150/300 (Skretting, Fontaine les-Vervins, France), Watts High-Fat Juvenile Mix, or ZIRC Juvenile Mix, twice daily (9 AM and 3 PM local time) until 60 dpf. From 60 dpf onward, OSU sh were not fed on weekends and 1-day holidays as per the facility institutional animal care and use protocol. The total quantity fed daily was 3% sh body weight. This continued until sh were 214 dpf and then they were transitioned to the adult version of their previously assigned juvenile diet: Gemma Micro 500 (Skretting, Fontaine les-Vervins, France), Watts Low-Fat Adult Mix, or ZIRC Adult Mix, twice daily (9 AM and 3 PM local time), except weekends and 1-day holidays. The total quantity fed daily was 3% sh body weight. The prescribed amounts of each diet regiment, for both the juvenile and adult diets were delivered by 3D printed spoons speci c to the diet and stage of life. These spoons were paired with conical tubes retro tted with leveling wires to ensure consistent feeding volumes as prescribed. All sh were only fed once, in the afternoons, on sampling days.

Diet and Pathogen Exposure
Each of the eighteen tanks was assigned one of the three diet regimens: Gemma, Watts, or ZIRC. There were three tank replicates per diet regimens for a total of nine tanks that were exposed to M. chelonae via intraperitoneal injection (3 tanks/diet with 15 sh/tank). The remaining nine tanks were similarly assigned to diet regimens and were exposed to a sterile 1X-phosphate buffered saline (PBS) solution via intraperitoneal injection. Each sh was injected with 10 uL of either the M. chelonae inoculum or saline solution. The injections were completed over the course of two days and the M. chelonae inoculum was prepared as a 0.5 McFarland each day with a target dose/ sh of 5 X 10 4 viable bacteria/ sh This target dose was chosen as we have found that it induces a higher prevalence of M. chelonae in zebra sh with minimal mortality [19,37,38].
Day 1 M. chelonae inoculum was afterwards determined by plating to be 3.1x10^3 dose per sh, while Day 2 M. chelonae inoculum was determined by plating to be 1.0x10^5 dose per sh. For ZIRC and Gemma, two tanks for ZIRC sh were injected on Day 1, and 1 tank on Day 2. For Watts, one tank was injected on Day 1 (low dose) and 2 tanks were injected on Day 2 (high dose). No signi cant difference was observed in prevalence was observed so further analyses treated the exposed sh with in each diet group together. Kinyoun's acid-fast [39]. Fish were processed into mid-sagittal sections as previously described [40]. Infection in sh were scored as positive when acid fast bacilli were observed in extra-intestinal organs [40]. A Chi-square test was used to compare positive and negative infections between sh fed each diet.

Fecal Collection
Five sh from each tank at 4-and 7-months post fertilization sampling time points were randomly selected for fecal sampling. Fecal material was collected from individual sh at the same sample intervals as outlined for the growth parameters. Fecal collection was set up the day before growth parameter sampling. Fish were transferred to 1.4 L tanks (1 sh/tank) containing ~0.4 L of sh water at least 30 minutes after the last feeding of the day. Fish were left to defecate overnight and all fecal material was collected from each tank the following morning in a 1.5ml microcentrifuge tube. Fecal samples were immediately spun at 10k rpm for 2 minutes, excess tank water was removed, and samples were snap frozen on dry ice and stored at -80 ˚C until processing.

16S Sequencing
Microbial DNA was extracted from zebra sh fecal samples and 16S rRNA gene sequence libraries were produced and analyzed following established approaches [41].

Statistical Analysis
All microbiome DNA sequence analyses and visualizations were conducted in R (v 4.2.1) [42]. Fastq les were processed in using the DADA2 R package (v 1.18.0) [43]. Brie y, forward and reverse reads were trimmed at 250 and 225 bp, respectively, subsequently merged into contigs, and subject to amplicon sequence variant (ASV) identi cation. ASVs unannotated at the Phylum level were removed to result in 2029 remaining detected ASVs. We used Wilcoxon Signed-Ranks Tests to identify parameters that best explained the variation in weight and body condition scores. Alpha-diversity was calculated using the estimate_richness function (Phyloseq v 1.38.0) and transformed using Tukey's Ladder of Powers using methods described previously [41]. After transformation, scores were normalized from 0 to 1 by dividing each score by the maximum value, which allowed us to compare results across alpha-diversity metrics using general linear models (GLMs). Post-hoc Tukey Tests evaluated pairwise comparisons of models using multcomp (v1.4-2) glht function [44]. We corrected for multiple tests using Benjamini-Hochberg correction [45]. Two-way ANOVA assess these GLMs. Beta-diversity models were generated using methods described previously [41]. Brie y, we evaluated three beta-diversity metrics-Bray-Curtis, Canberra, and Sørensen and resolved the relationship between experimental parameters and beta-  Afterwards, a cohort of sh from each diet were exposed to Mycobacterium chelonae. 5) Three months later when sh were 214 dpf, body size measurements were conducted on all sh and fecal samples were collected from a random selection of ve sh per tank (n = 89). Histopathology check was conducted to assess infection burden on all sh. ordination based on the Bray-Curtis dissimilarity of gut microbiome composition. The analysis shows that physiology and gut microbiome composition signi cantly differs between the diets. "ns" indicates not signi cantly different, *, **, *** indicates signi cant differences below the 0.05, 0.01, and 0.001 levels, respectively.

Figure 3
Development is associated with altered microbiome composition. (A) Shannon Entropy of diversity shows that gut microbiome diversity signi cantly differs between Watts-diet fed sh to sh fed the Gemma-and ZIRC-diets in 7 months post fertilization (mpf) zebra sh. (B) Capscale ordination based on the Bray-Curtis dissimilarity of gut microbiome composition in 7 mpf zebra sh. (C) Shannon Entropy for diversity shows microbial gut diversity increases with development in 4 to 7 mpf zebra sh fed the Gemma-and ZIRC-diets, but not Watts-diet fed sh. Capscale ordination of gut microbiome composition based on the (D) Bray-Curtis dissimilarity by diet and (E) Canberra measure by time. (F) Body condition score negatively associates with gut microbiome diversity as measured by Simpson's Index across 4 and 7 mpf zebra sh fed. the ZIRC diet. The analysis shows that sh size and gut microbiome composition signi cantly differs between the diets across development, and there may be diet-dependent link with physiology. A "ns" indicates not signi cantly different, "*" indicates signi cant differences below the 0.05 level.

Figure 4
Exposure to Mycobacterium chelonae inhibits diversi cation of gut microbiome. (A) Shannon Index for diversity of pre-exposed 4 month post fertilization (mpf), 7 mpf exposed and unexposed sh, and (B) for exposure groups within each diet. Capscale ordination based on the Bray-Curtis dissimilarity of gut microbiome composition of sh by (C) diet. (D) Log fold change of Mycobacteriumof pre-exposed, exposed and unexposed sh within each diet as calculated by ANCOM-BC. Values are in reference to exposed sh within each diet. The analysis shows gut microbiome's sensitivity to pathogen exposure is linked to diet, but Mycobacterium's abundance is diet-dependent. A "ns" indicates not signi cantly different, and * indicates signi cant differences below the 0.05. An "X" indicates a group is signi cantly differentially abundant compared to the exposed treatment reference group.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.