Exposure to antibiotics led to increased time in paradoxical anaerobism
Exposure to a cocktail of antibiotics (kanamycin, gentamicin, colistin, metronidazole, and vancomycin made fish more prone to entering paradoxical anaerobism (Fig.1). In Fig.1A, an example of paradoxical anaerobism use is provided. Immediately after introduction of ~1% ethanol (vol/vol) to the aquarium water, the fish entered a prolonged period of negligible oxygen use. Pupfish exposed to antibiotics and ethanol spent 40.5 ± 11.3% of the post-ethanol time in paradoxical anaerobism - approximately 3.6X longer than did control fish (Fig.1B). This difference was statistically significant (Welch’s t-test, p-value = 0.037). All 12 antibiotic-treated, 28 °C-acclimated pupfish used some paradoxical anaerobism when exposed to 1% ethanol (vol/vol), whereas 8 of the 12-control pupfish used paradoxical anaerobism, which is consistent with our previous results .
Antibiotic treatment was associated with the changes in community composition
Illumina sequencing yielded an average of 18,688 (range: 3,713 – 82,969) 16S rRNA gene sequences per fish gut sample (Supplementary Table 1). One of the samples, PUP9, from the antibiotic-treated group, was removed from further analysis due to low sequence count. The microbiomes were moderately diverse, with amplicon sequence variant (ASV) richness ranging from 11 – 154 and Shannon diversity values ranging from 0.67 – 6.23, and there was no significant difference in richness or diversity (Observed ASVs and Shannon index, Fig.2A; Table S1) between control and antibiotic-treated microbiomes (Welch’s t-test, observed ASVs: p=0.450 and Shannon index: p=0.897). However, beta diversity analysis based on non-metric multidimensional scaling (NMDS) using Bray-Curtis dissimilarity indicated that antibiotic treatment changed the composition of the microbiome (Fig.2B). This was further demonstrated by ANOSIM (Analysis of Similarity) analysis (R = 0.469, p = 0.001). Overall, the dominant phyla were Proteobacteria (29.9% and 46.9%), Fusobacteria (47.3% and 2.0%), Bacteroidetes (11.7% and 18.5%), Firmicutes (15% and 13.8%), Actinobacteria (4.8% and 5.3%), Patescibacteria (2.5% and 4.1%), and Dependentiae (0.0% and 2.3%); numbers in parentheses indicate mean relative abundance of each phylum in control and antibiotic-treated groups, respectively. Out of the 23 phyla detected in all samples (Fig.2C), Proteobacteria, Bacteroidetes, Actinobacteria, Patescibacteria, and Dependentiae had higher relative abundance in the antibiotic-treated group (Welch’s t-test, p < 0.05), whereas only Fusobacteria was significantly reduced due to antibiotic treatment (Welch’s t-test, p < 0.05).
SIMPER (SIMilarity of PERcentages) analysis was performed to determine the contribution made by specific ASVs to the observed dissimilarity between control and antibiotic-treated fish. SIMPER identified nine ASVs contributing to 50% of observed differences, of which Cetobacterium (Fusobacteria), Brevinema (Spirochaetes), Pseudomonas pseudomonaspeli (Proteobacteria), Ileibacterium (Firmicutes), Aeromonas (Proteobacteria), and Verrucomicrobiaceae (Verrucomicrobia) were enriched in the control group, while Escherichia-Shigella (Proteobacteria) and Flavobacterium (Bacteroidetes) ASVs were enriched in the antibiotic-treated group (Fig.3; Supplementary Table 2). ANCOM (Analysis of Composition of Microbiomes) (W=719) suggested that the loss of Cetobacterium in antibiotic-treated fish as a strong indicator of community differences. Cetobacterium was represented by two different ASVs, designated Cetobacterium and Cetobacterium 1 (Fig.3), which represent either separate species or separate populations of a single species. However, the genetic distance between these ASVs cannot be accurately assessed by small 16S rRNA gene fragments, alone. The Cetobacterium 1 ASV comprised 33% of the healthy pupfish gut microbiome, whereas the Cetobacterium ASV was present at low abundance; neither were detected in antibiotic-treated fish (Fig.3).
Cetobacterium may be involved in ethanol metabolism
To probe the possible mechanism of microbial mediation of paradoxical anaerobism, we identified microbial taxa contributing to ethanol metabolism by predicting the presence of alcohol dehydrogenase in the simulated metagenome using PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States). We identified 129 ASVs that likely have an alcohol dehydrogenase gene (adh; KO: K00001). Notably, ASVs assigned to Cetobacterium were the dominant bacteria predicted to encode Adh in control fish, indicating the potential role of Cetobacterium in clearing ethanol reaching the gut environment in these fish (Fig.4). In contrast, Flavobacterium ASVs were the dominant bacteria to predicted to encode Adh in antibiotic-treated fish. Other taxa that contributed to the adh gene pool include members of the Actinobacteria (Bifidobacterium and Mycobacterium); Proteobacteria (Acinetobacter, Delftia, Polynucleobacter, and Pseudomonas); and Firmicutes.