1. Diet differentially influences physiology and gut microbiome at 4 months old
To determine how common zebrafish diets differently impact fish size (length and body condition score) and the gut microbiome, we reared 176 zebrafish that were assigned one of three diets from 1- to 4-months-post fertilization (mpf) (Figure 1): Gemma, Watts and ZIRC diets. Prior to diet assignment, fish were fed a nursery diet (see methods). At 4 mpf, we selected 89 individuals across these three cohorts and collected fecal samples from each fish for microbiome profiling prior to measuring their weight and body condition score (BCS). Wilcoxon Signed-Rank Tests found that diet and sex significantly associated with weight and BCS. Female fish 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 fish had the highest mean BCS compared to fish fed Gemma- (Z = 150, P < 0.001) and Watts-diet (Z = 197, P < 0.001, Table S1.1.3). Gemma- and Watts-diet fed fish did not significantly differ from one another in terms of weight and BCS. These results indicate that ZIRC-diet contributes to heavier fish compared to Gemma- and Watts-diet fed fish.
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 clarified that ZIRC- and Watts-diet fed fish exhibited significant differences in alpha-diversity as measured by richness and Shannon Entropy (P < 0.001, Table S1.2.2). Moreover, we observed significant differences in diversity between Gemma- and Watts-diet fed fish in terms of richness (P < 0.001; Table S1.2.2), and between Gemma- and ZIRC-diet fed fish when considering the Simpson’s Index (P < 0.001; Table S1.2.2). These results indicate that diet associates with fish 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 quantified the Bray-Curtis, Canberra and Sørensen dissimilarity amongst all sample. We detected a significant 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 fish fed the same diet are more consistent in composition to one another than to fish fed other diets. Additionally, we assessed beta-dispersion, a measure of variance, in the gut microbiome community compositions for each diet group. We find the beta-dispersion levels were significantly different between the diet groups as measured by Bray-Curtis and Canberra metrics (P < 0.05; Table S1.4.1). Beta-dispersion levels were significantly reduced in Gemma-diet fed fish compared to Watts-diet fed fish when measured by Bray-Curtis metric, as well as significantly reduced compared to Watts- and ZIRC-diet fed fish when measured by Canberra metric (Table S1.4.1). These results indicate that Gemma-diet fed fish are more consistent in community composition than Watts- and ZIRC-diet fed fish at 4 mpf. Collectively, these results indicate that 4 mpf fish 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 quantified differential abundance using ANCOM-BC2. We observed 24 significantly abundant taxa at the genus level in at least one of the three diets (Table S1.5.1). Gemma-diet fed fish were enriched for Chitinibacter and were depleted of Aeromonas and Flavobacterium. Watts-diet fed fish 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 fish enriched for Cloacibacterium and Acinetobacter, and depleted of Fluviicola. Many of these taxa are identified as common members of the zebrafish gut microbiome[14, 15]. These results indicate that diet differentially supports particular members of the zebrafish microbiome community.
2. Diet impacts the successional development of the zebrafish gut microbiome
To determine how maintaining fish on different diets impacts the development of the gut microbiome, we continued to grow fish from the same diet cohorts until 7 months post fertilization (mpf; Figure 1). Microbiome samples were collected from cohort members prior to quantification of fish weight and body condition score. To determine the effect of diet on the body condition score and the gut microbiome of 7mpf fish, we conducted the same analyses as we applied to the 4 mpf fish. At 7 mpf, we find body condition score is significantly associated with diet (P < 0.05; Table S2.1.3.1). Additionally, linear regression analyses revealed statistically significant main effects of diet on gut microbiome alpha- and beta-diversity for all metrics we considered (P < 0.05; Fig 3A&B, Table S2.1.3.2-3), but an ANOVA test of beta dispersion was not significantly different between diets for any beta-diversity metric (P > 0.05; Table S2.1.3.4). These results demonstrate that diet impacts the physiology and gut microbiome of 7mpf fish.
Next, we compared our results between the 4 and 7 mpf fish to determine how diet impacts the successional development of the gut microbiome. Linear regression revealed microbial gut alpha-diversity was significantly associated with the main effect of time (P < 0.05; Table S2.2.1) for each diversity metric. However, we did not find 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 clarified that microbiome diversity was significantly different between 4 and 7 mpf Gemma- and ZIRC-diet fed fish as measured by the Shannon and Simpson’s alpha-diversity metrics (P < 0.05; Figure 3C, Table S2.2.2), but we did not find a statistically significant association between 4 and 7 mpf Watts-diet fed fish 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 fish were temporally stable, while Gemma- and ZIRC-diet fed fish diversified 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 significantly elevated between 4 and 7 mpf Gemma-diet fish when considering the Bray-Curtis and Sørensen metrics, in Watts-diet fed fish when considering the Canberra and Sørensen metrics, and in ZIRC-diet fed fish 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 zebrafish.
Differential abundance analysis revealed taxa that were significantly 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 zebrafish development. We also measured how taxon abundance changed over time within each diet (Figure S2.5.2-46.2.5). The Gemma-diet fed fish 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 zebrafish [20, 21]. The Watts-diet fed fish were uniquely depleted of Gemmobacter (Table S2.5.3). Previous work has found that Gemmobacter has a positive association with parasite exposure in infected zebrafish[22, 23]. The ZIRC-diet fed fish 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 zebrafish[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 zebrafish development.
To determine if fish size associated with diet across zebrafish 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 fish. At 7mpf, the BCS significantly differed between fish fed different diets. However, we did not find that BCS of fish were impacted by time (P > 0.05; Fig 2E, Table S2.1.1). These results indicate that while fish differ in BCS between diets at 7 mpf, their weight and length grow proportionally at a similar rate from 4 to 7 mpf. Interestingly, we observed a significant negative association of BCS and microbial gut diversity uniquely in fish fed the ZIRC diet as measured by Shannon Entropy and Simpson’s Index (P < 0.05; Fig 2F, Table S2.1.2.1). This result indicates that fish gut microbiomes with higher body masses are lower in diversity compared to fish with lower body mass. For Canberra and Sørensen beta-diversity metrics, there were significant main effects of body condition score, and significant interaction effects between BCS and diet (P < 0.05; Table S2.1.2.2). However, the model coefficient for the effect of body condition score and its interaction with diet is far smaller than the coefficient for the effect of diet (Table S2.1.2.2). We did not find a significant association between BCS and specific 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 influences 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 zebrafish, Mycobacterium chelonae. Mycobacteria has been reported in zebrafish 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 identifications are made using molecular methods, the identification 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 fish from each diet cohort at 4 mpf following fecal collection. These M. chelonae injected fish comprised the pathogen exposure cohort for this experiment, which we compared to the remaining, unexposed cohort of fish. At 7 mpf, we collected fecal samples from exposed and unexposed fish 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 first evaluated whether diet impacted infection outcomes, as determined by histological confirmation of infection 3.5 months following pathogen injection. We conducted a Chi-Square test to compare the infection count between fish fed the three diets. The results showed that there was a statistically significant difference in proportion of positive infection counts between the groups, X2 (2, N = 66) = 11.519, P < 0.05 (Table S3.1.1). Across all three diets, all females had infected ovaries. In contrast, we observed the following infection outcomes for male fish with extra-intestinal infections Gemma 3/12 (25%), Watts 5/24 (20.8%), and ZIRC 18/24 (70.6%) (Table 3.5.2.2.1). In male fish only, we also found a statistically significant difference in proportion of infected fish across the three diets (X2 = 11.556, df = 2, N = 53, P < 0.05; Table S3.1.2). When we conduct the same analysis with just fish sampled for microbiome analysis (Table S3.5.1.3.1), we do not observe significant effects (X2 = 4.069, df = 2, N = 44, P > 0.05; Table S3.1.3), likely due to being underpowered to detect these effects. Infections in males 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 fish 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 significant associations between infection status and body condition score based on linear regression (P > 0.05; Table S3.1.4) or any of the gut microbiome diversity and composition measures (P > 0.05; Table S3.1.5 & S3.1.6). Together, these results indicate that infection endpoints are linked to diet, but not body condition score or the gut microbiome.
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 fish found that microbial gut diversity significantly 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 significant effect (P > 0.05; Table S3.2.1). Furthermore, we used a post hoc Tukey test to clarify whether microbial gut diversity of fish differed between exposure groups by diet. Unique to ZIRC-diet fed fish, we observed microbiome diversity differed in unexposed controls compared to exposed fish as measured by all alpha-diversity metrics (P < 0.05, Table S3.2.2). Watts-diet fed fish differed in unexposed controls compared to exposed fish in terms of richness (P < 0.05, Table S3.2.2). These results suggest that the gut microbiome diversity of ZIRC-diet fed fish, and to some extent Watts-diet fed fish, are sensitive to the effects of M. chelonae exposure, but Gemma-diet fed fish are resistant to pathogen exposure. While the gut microbiomes are sensitive to the effects of pathogen exposure, we find 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 influenced microbial community composition across fish fed each diet. For each beta-diversity metric considered, PERMANOVA tests found that the main effects of diet and pathogen exposure significantly 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 coefficient effect for the interaction of diet and pathogen exposure was statistically significant 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 find significant levels of dispersion between exposed and unexposed fish 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 stratified exposed and unexposed groups of fish in both diet-robust and diet-dependent manners. Unexposed Gemma-diet fed fish were enriched for Macellibacteroides and Aurantisolimonas (Table S3.5.2), unexposed Watts-diet fed fish were enriched for an unnamed genus of Barnesiellaceae, Fluviicola, Paucibacter, and Brevibacterium (Table S3.5.3), and unexposed ZIRC-diet fed fish 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 fish included Macellibecateroides, Fluviicola, Bacteroides, Aurantisolimonas, Cerasicoccus, and three unnamed genera of Barnesiellaceae, Commonadaceae, and Sutterellaceae. Plesiomonas were more abundant in exposed fish 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 fish to that of exposed fish within each diet. Unexposed Gemma- and ZIRC-diet fed fish had significantly higher abundances of Mycobacterium to exposed (Figure 4D, Table S3.5.5). Pre-exposed Watts-diet fed fish had significantly more Mycobacterium compared to pre-exposed fish, but they did not differ significantly from unexposed control fish. 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.