Treatment effects on the host gut transcriptome.
To determine if antibiotic and probiotic-induced changes in the microbiome led changes in the host gut transcriptome, RNA-Seq was used to determine host transcript levels in the hindgut. Pairwise treatment comparisons resulted in 96 (control vs antibiotic; 35 control upregulated and 61 control downregulated), 105 (control vs probiotic; 61 control upregulated, and 44 control downregulated), 120 (antibiotic vs probiotic; 84 antibiotic upregulated, and 36 antibiotic downregulated) transcripts that were differentially expressed among treatments (Benjamini-Hochberg false-discovery rate (BH FDR) 0.1, |log2 FC| > 0.25). However, for selecting candidate genes for the OpenArray high-throughput qPCR analyses, we took a conservative approach and we only selected genes with transcripts that were significantly expressed at |log2 FC| > 1 and FDR P value < 0.05 (Supplementary Figure S3). This decreased the differentially expressed transcripts to 29 (control vs antibiotic), 29 (control vs probiotic), and 27 transcripts (antibiotic vs probiotic) (Table 2). For the control versus antibiotic group comparisons, the selected genes related to cellular process (e.g., cell activation, cell communication, cell cycle, and cell death) were upregulated and genes related to metabolism and response to stimuli and stress were downregulated in antibiotic group (Table 2). While in the control versus probiotic group comparisons, genes related to regulation of a variety of functions (regulation of meiosis, intracellular protein transport, angiogenesis, transmembrane transporter, cell adhesion, negative regulation of apoptotic process) were downregulated and genes related to post-translation modifications were over expressed in the probiotic treated fish (Table 2). Moreover, when we compared antibiotic against probiotic group transcription, genes related to cellular process (mostly apoptotic process) were over expressed in antibiotic group while genes related to cell adhesion, regulation of transcription were over expressed in probiotic group (Table 2).
OpenArray high-throughput qRT‐PCR
The LMM analysis showed PCs 4, 5, 6, 7 and 9 were significantly affected by treatment (Table 3). We identified only those genes whose contributions to the significantly affected principal component axes were important (Supplementary Figure S4) and selected them for analyses. In our analysis we also included tank, body weight, and OpenArray chip ID as random effects to correct for possible technical, environmental, and body size effects. Chip and body weight were not significant for any of the genes and were dropped from our analyses. Sire effects (nested within dam) were not significant after FDR correction. Moreover, a significant tank effect was observed for only one gene (anxa1 , p < 0.05) before FDR correction. We found no significant effects for dam‐by‐treatment or sire-by‐treatment interactions. After including FDR correction into our model, aifm3 , manf , and prmt3 still showed a significant treatment effect (Table 4).