Neurobehavioral and neurotransmitter characterization of the control and FMT-processed rats
To test if the neurobehavioral characteristics of the recipient FH rats upon FMT were altered toward the quantified range of the donor SD rats, three anxiety-like and depression-like behavioral tests, including the forced swimming test (FST), the open-field test (OFT), and the sucrose preference test (SPT), were conducted on all four groups of rats. As shown in Fig. 1a, FST scores were significantly different between FH and SD rats (two-tailed t-test, p = 0.0183), and FMT significantly reduced the floating time of FH recipients receiving SD fecal microbiota (FH-SD rats) in comparison to FH fecal microbiota recipients (FH-FH rats; two-tailed t-test, p = 0.0468). Figure 1b showed that FH and SD rats had significantly different OFT scores (two-tailed t-test, p = 0.0178), and FMT resulted in a non-significant trend toward an increase in the central to total movement ratio for FH-SD rats compared to FH-FH rats (two-tailed t-test, p = 0.0606). For SPT, the sucrose-preference index suggested a significant trend toward a greater sucrose preference in FH-SD rats compared to FH (two-tailed t-test, p = 0.0093) and FH-FH rats (two-tailed t-test, p = 0.0122; Fig. 1c). Altogether, these results indicated that FH-SD rats presented behaviors more like SD rats than FH or FH-FH rats, which suggested that FMT of the SD gut microbiome could be responsible for the behavioral shifts in recipients.
Similarly, to determine if the neurotransmitter concentrations in recipient FH rats upon FMT had altered toward those observed for donor SD rats, we measured three major neurotransmitters (i.e., serotonin, norepinephrine, and dopamine) in all four groups of rats. Owing to the genetic dysfunctional nature of the serotoninergic system in the FH rats, the serotonin concentrations in serum was significantly lower for FH rats (two-tailed test, p = 1.9×10− 6), FH-FH rats (two-tailed test, p = 2.1×10− 6) and FH-SD rats (two-tailed test, p = 1.7×10− 6) than those for SD rats (Fig. 1d). However, the hippocampal concentrations of serotonin (Fig. 1e) were significantly higher in FH-SD rats than those in FH (two tailed t-test, p = 0.0379) and FH-FH rats (two tailed t-test, p = 0.0122), while they were similar with those in SD rats (two tailed t-test, p = 0.0715). One explanation of these results is that the transplantation of SD fecal microbiota to FH rats influenced neuromodulation through the enteric nervous system (ENS), not blood circulation. Norepinephrine (Fig. 1f) and dopamine (Fig. 1g) also had significantly higher hippocampal concentrations in FH-SD rats than FH (two tailed t-test, pnor= 0.0016 and pdop = 0.0004, respectively) and FH-FH rats (two tailed t-test, pnor=0.0081 and pdop = 0.0022, respectively). Considering that hippocampal norepinephrine concentrations in control SD rats were significantly higher than those in control FH rats (two tailed t-test, p = 0.0281), while they were similar with those in FH-SD rats (two tailed t-test, p = 0.5748), it was inferred that FMT of the SD microbiome ameliorated the situation of hippocampal norepinephrine reduction in FH rats. Interestingly, FH-SD rats showed a significant increase in hippocampal dopamine concentrations compared to SD rats (two tailed t-test, p = 0.0092; Fig. 1g), suggesting some unknown effect of FMT on this neurotransmitter.
Cytokine quantification in control and FMT-processed rats
To determine if FMT influenced immune responses in the rats, cytokines in both the serum and hippocampus of the control and FMT-processed rats were quantified. Serum cytokines IL-4 and IL-10 had significantly lower levels in FMT-processed rats compared to either FH (two tailed t-test, pIL−4 = 0.0174 and pIL−10 = 0.0079 for FH-FH rats; pIL−4 = 1.2×10− 6 and pIL−10 = 4.5×10− 7 for FH-SD rats) or SD rats (two tailed t-test, pIL−4 = 0.0083 and pIL−10 = 0.0030 for FH-FH rats; pIL−4 = 0.0022 and pIL−10 = 1.3×10− 7 for FH-SD rats). However, most serum cytokine concentrations were not significantly different following FMT or between FH-FH and FH-SD rats (Fig. S1). The levels of hippocampal cytokines IL-1b and TNF-α were significantly lower in SD rats than FH rats (two tailed t-test, pIL−1b = 0.0356 and pTNF−α = 0.0028, respectively), and they showed a non-significant trend toward lower levels in FH-SD rats compared to FH-FH rats (two tailed t-test, pIL−1b = 0.1739 and pTNF−α = 0.0598, respectively; Fig. S1). Interestingly, hippocampal IL-17A contents in FH-FH rats were significantly higher than that in FH (two tailed t-test, p = 0.0123) and FH-SD rats (two tailed t-test, p = 0.0052), suggesting that FH microbiome transplantation could induce IL-17A accumulation in FH rat hippocampus. All above results indicated that SD microbiome transplantation could induce a decrease of some hippocampal immune cytokines in FH recipients.
FMT-mediated changes in fecal microbial taxonomy
To determine if the gut microbiota in FH-SD rats had been altered, species-level beta-diversity based on 16S rRNA amplicon sequencing data was analyzed using DEICODE. As shown in Fig. 2a, FH and FH-FH rats had significantly different beta diversity compared to FH-SD and SD rats (PERMDISP: F-statistic 0.230, p-value 0.818, n = 999 permutations; PERMANOVA: F-statistic 8.689, p-value 0.001, n = 999 permutations). The SD microbiota was significantly differentiated from the FH microbiota by the proportion of Roseburia sp. CAG 380 and Dialister sp. CAG: 357. Likewise, the FH microbiota was characterized by increased proportions of Bifidobacterium pseudolongum and Candidatus Gastranaerophilus phascolarctosicola. Log-ratio of DEICODE-feature loadings of these four species were employed further to examine the proportion of SD:FH-associated species. A significantly greater log-ratio of Roseburia sp. CAG 380 and Dialister sp. CAG: 357 (in the numerator) to Bifidobacterium pseudolongum and Candidatus Gastranaerophilus phascolarctosicola (in the denominator) between FH and SD control groups, as well as between FH-FH and FH-SD rats, were observed (Mann-Whitney Wilcox test, p < 0.05; Fig. S2). This suggested a successful transfer of the SD gut microbiota to the FH recipient rats. To further identify differentially proportional taxa and account for the compositional data, ANCOM-BC was applied. As shown in Fig. 2b, there were eight species with significantly different proportions both between the two control groups (i.e., SD vs FH) and between the two FMT-processed groups (i.e., FH-FH vs FH-SD) (effect sizes with Bonferroni-adjusted p < 0.05). These differentially proportional species were Akkermansia muciniphila, Akkermanisia muciniphila CAG:154, Bifidobacterium adolescentis, Dialister sp. CAG357, Firmicutes bacterium CAG:41, Ruminococcus sp. CAG:108, Sutterella wadsworthensis CAG:135, Veillonella sp. ACP1, and their proportions were significantly lower in SD and FH-SD rats than in FH and FH-FH rats. Interestingly, Dialister sp. CAG357 was the sole pairwise differentially proportional species that could determine the species-level beta-diversity differentiation between SD and FH-SD gut microbiota and the FH and FH-FH gut microbiota (Fig. 2a). Therefore, it was considered that the SD fecal microbiota transplantation was successful in shifting the microbiota of the recipient FH rats towards the SD-characteristic microbiota, and the significant decrease of Dialister sp. CAG357 might play a key role in the gut microbiota reassembly in FH-SD rats.
FMT-mediated changes in fecal microbial functional potential
To determine if the genetic functional potential of the recipient FH rat microbiome was altered by the donor SD rat microbiome upon FMT, MetaCyc database-mapped enzymatic reactions and pathways for the metagenomic data of the four groups were analyzed using DEICODE beta-diversity. As shown in Fig. 3a, FH and FH-FH rats had significantly different beta-diversities compared to FH-SD and SD rats in terms of enzymatic reactions (PERMDISP: F-statistic 1.420, p-value 0.179, n = 999 permutations; PERMANOVA: F-statistic 7.155, p-value 0.001, n = 999 permutations). It was found that the group of SD and FH-SD microbiomes was characterized by the gene encoding 1-deoxy-D-xylulose 5-phosphate reductoisomerase, while the clustering of FH and FH-FH microbiomes was characterized by genes encoding citrate hydro-lyase, D-threo-isocitrate hydro-lyase, sucrose phosphorylase, acetolactate synthase, butyryl-CoA dehydrogenase, isoamylase, and phosphoglucomutase. Similar to the taxonomic beta-diversity analysis, the functional beta-diversity in FH-SD rats was more similar to SD rats than to FH and FH-FH rats (Fig S3a). The metabolic pathways that defined the SD and FH-SD microbiomes were quinate degradation I and II, gallate biosynthesis, urea cycle, and carbamoyl-phosphate synthesis, while the pathways that characterized the FH and FH-FH microbiomes were L-citrulline biosynthesis, L-citrulline degradation, and L-proline biosynthesis II (from arginine). Quinate was one of several aromatic compounds that can be metabolized by microorganisms to the central intermediate protocatechuate and then be further metabolized via the β-ketoadipate pathway to acetyl-CoA and succinyl-CoA. According to a Spearman’s rank correlation analysis between the robust-Aitchison (RPCA) generated distance-ordination matrices of functional genes at the reaction level (along the X-axis) and pathway level (along the Y-axis), it was demonstrated that there was a significant association between the pathways and enzyme reactions with ρ = 0.8741 (over 999 permutations, p-value = 0.001; Fig. S3b). By applying ANCOM-BC, the pairwise differentially proportional enzyme-encoding genes and pathways in both the control groups (i.e., SD vs FH) and FMT-processed groups (i.e., FH-FH vs FH-SD) were identified. Nine metabolic pathways were pairwise differentially proportional, including glycolipid biosynthesis, chondroitin sulfate degradation, CMP-legionaminate biosynthesis, dermatan sulfate degradation, pyruvate fermentation – acetoin I, pyruvate fermentation – acetoin, starch degradation II, and zwittermicin A biosynthesis (Fig S3c). As shown in Fig. 3b, there were 29 pairwise differentially abundant enzyme genes, all of which were significantly less abundant in SD or FH-SD microbiomes than those in FH or FH-FH microbiomes, respectively (effect sizes with Bonferroni-adjusted p < 0.05). Most of the pairwise differentially abundant genes were associated with carbon metabolism. Among them, the gene encoding acetolactate synthase, which could catalyze the conversion between pyruvate and 2-acetolactate and was involved in valine and isoleucine biosynthesis and then pantothenate and CoA biosynthesis, was the only one associated with the beta-diversity differentiation between SD/FH-SD and FH/FH-FH (Fig. 3a). Hence, SD-FMT significantly changed the microbiome profile of FH rats, resulting in a significant reduction in the proportion of genes encoding acetolactate synthase.
Microbe-metabolite co-occurrences amongst the group-associated features
Serum and hippocampal metabolomics were performed for the four groups of rats. Comparative metabolomics analysis demonstrated that serum metabolomic profiles were relatively conserved between all four rat groups (Fig. 4a-b), but hippocampal metabolite profiles were considerably different between control rats and FMT-processed rats (Fig. 4c-d). Similarly, DEICODE-generated biplots of serum metabolomics (Fig. S4a) showed differentiation of SD and FH-SD rats to FH and FH-FH rats (PERMDISP: F-statistic 0.929, p-value 0.382, n = 999 permutations and PERMANOVA: F-statistic 4.907, p-value 0.001, n = 999 permutations), while hippocampus metabolic diversity of SD rats showed distinction with those of the FMT-processed and control FH rats (PERMDISP: F-statistic 1.113, p-value 0.276, n = 999 permutations; PERMANOVA: F-statistic 12.617, p-value 0.001, n = 999 permutations), even though the hippocampal metabolomes of FH-SD rats had the trend toward those of SD rats (Fig. S4b). These results indicated that the transplantation of SD gut microbiota to FH rats had a greater effect on the recipients’ serum metabolism than hippocampal metabolism. Among the significantly differentially abundant metabolites in control and FMT-processed rats, arachidonic acid (C20:4) in serum was the sole metabolite that was significantly pairwise differentially abundant in FH and FH-FH rats, when compared to SD and FH-SD rats. The abundance of serum C20:4 in FH-SD rats was significantly lower than that in FH-FH rats (Mann-Whitney Wilcox test, p = 0.0007), and was similar to that in the SD donor (p = 0.243). This indicated that the transplantation of SD gut microbiota reduced the serum concentration of C20:4 in FH rats.
Meanwhile, by doing the correlational analyses on the serum and hippocampal metabolomes, respectively, the associative nature of metabolites amongst themselves were identified. Interestingly, it was found that several carbon-rich fatty acids (i.e. palmitic acid (C16:0), heptadecanoic acid (C17:0), stearic acid (C18:0), and linoleic acid (C18:2)) and nitrogen-rich metabolites (e.g. 2-aminoethanol, glucopyranose, glycine, and uracil) had strongly inverse relationships in rat serum (Fig. S5a), while numerous nitrogen-rich metabolites (e.g. serine, ornithine, asparagine, and glutamine) had highly negative relationships with several brain osmolyte compounds or their precursors (i.e. N-acetylaspartic acid (NAA), inositol, and pyroglutamic acid) in rat hippocampus (Fig. S5b). Considering that the level of C20:4, a derivative of C16:0, C17:0. C18:0, and C18:2 in lipid metabolism, was pairwise significantly reduced in serum of SD and FH-SD rats as shown in Fig. 4A-B, the contents of nitrogen-rich metabolites in serum of SD and FH-SD rats were expected to be higher in comparison with those in FH and FH-FH rats, regardless of the significance, which would result in the increase of amino acids and decrease of osmolytes in the hippocampus through the blood-brain barrier (BBB). This inference was supported by the hippocampal metabolomes of the four groups of rats, where the log ratio of serine to NAA was significantly higher in FH-SD and SD rats than that in FH rats (Mann-Whitney Wilcox test, p < 0.05; Fig. S6).
To examine the co-occurrences between metabolomes in host tissues and specific bacterial species in gut microbiota, we employed MMvec, which uses neural networks to infer the nature of interactions across omics-datasets. The heatmap reflecting the conditional probabilities between the serum metabolomes and DEICODE-associated and ANCOM-BC-associated bacteria taxa along PC1 (Fig. S7a) suggested a strong likelihood of co-occurrence for all pairings with positive and higher conditional probabilities. The model showed a higher predictive accuracy of Q2 = 0.35 than the absolute null or baseline model (where no formula was used) on the cross-validation samples (Fig. S7b). Nevertheless, the hippocampal metabolomes did not show good prediction of co-occurrences with the DEICODE-associated and ANCOM-BC-associated microbes (Q2 ≈ 0), hence no visualized data were shown here. This confirmed the expectation that the intestinal microbiome more strongly influenced serum metabolites than hippocampal metabolites.