A large body of preclinical literature highlights host-microbe interactions as a key factor in the modulation of specific brain networks related to a range of psychiatric disorders [2, 35]. Specifically, using GF animals, many studies have illustrated the critical role of a functional host microbiome in the normal expression of anxiety-, social-, fear- and depressive-like behaviours in these rodents [6, 9, 10, 36] (Additional file 8: Figure S3). Our findings indicated that two brain regions, the PFC and amygdala, known to govern the expression of fear, anxiety and social behaviours, display a large dysregulated network of miRNA expression, which may ultimately contribute to the observed behavioural changes associated with GF mice. We have also shown that exGF mice show an alteration in expression of miRNAs. Complementary approaches using antibiotics offer converging evidence for microbial regulation of miRNA expression. To our knowledge, this is the first time that the gut microbiome has been implicated in miRNA expression in the amygdala and PFC.

Many studies have shown that when GF animals are exposed to microbes early in life, between post-natal weeks 3–6, some of the observed differences can be reversed or partially reversed [6, 8, 37, 38] (Additional file 3: Figure S3). Here, when we allowed for colonization of our GF animals in a conventional facility from post-natal day 21, we found a proportion of miRNAs to have normalized expression levels. Additionally, we found that the majority of differentially regulated miRNAs between CON and GF mice showed regional specificity. These results suggest that targeting the microbiota later in life can have an effect on miRNA expression in the CNS in a brain region-specific manner. We also noted a large number of miRNAs that remained altered following post-weaning exposure to microbes. This supports the concept of critical neurodevelopmental windows during which the gut microbiota is essential in influencing brain development [39]. Previous studies have highlighted that colonization of GF mice with a conventional or reduced microbiota, or by mono-association with just one specific bacterial strain, had the capacity to reverse increased adrenocorticotropic hormone (ACTH) levels after acute stress [37], increased tryptophan levels, reduced basal anxiety [6] and altered social behaviour [8]. It is tempting to speculate that the subset of miRNAs identified that are normalized by colonization may contribute to the well-established phenotype of GF mice as they show dynamic expression, dependent on the presence or absence of a functional microbiome. However, a recent study investigating the changes in hippocampal miRNAs using a hybridized microarray approach in GF and colonized GF mice after weaning showed normalization of seven miRNAs with no restoration of behavioural changes in the open field test [40]. Similarly, we see partial reversibility in miRNAs within the amygdala and PFC but the majority remain altered in our exGF mice. A recent study has demonstrated that colonization of GF mice partially restores impaired amygdala-dependent fear memory recall in GF mice [7]. This partial recovery along with normalization of changes in miRNAs may aid in identification of microbially regulated brain-specific miRNAs. Further studies should continue to investigate more precise behavioural tests that better correlate with specific brain regions.

Within the amygdala, we found that miR-183-5p and miR-182-5p were both decreased and subsequently normalized by colonization. Both of these miRNAs have been previously linked to amygdala-dependent stress- and fear-related outputs [23, 41]. Specifically, miR-183 has been linked to regulating anxiety-related behaviours in the Indian field mouse through influencing acetylcholinesterase splicing [33]. This is also in line with the fact that GF rodents have been shown to display altered basal anxiety levels under naïve conditions and hypersecretion of corticosterone (CORT) under acute stress [37]. Clinically, miR-183 has been shown to be upregulated in whole blood samples from depressed patients on antidepressant treatment [42], highlighting that miR-183 is highly responsive to emotional stimuli. Within the lateral amygdala, miR-182 appears to be essential for long-term amygdala-dependent memory formation assessed by auditory fear conditioning [23]. Recently, it has been revealed that GF animals have impaired fear memory recall [7]. Future studies may aim to manipulate the expression of miRNAs such as miR-182 in order to normalize the amygdala-dependent memory impairments in these mice.

The GF model has many strengths; however, it has limitations in regard to investigating the impact of altered gut microbiota later in life [4, 43–45]. Here, we further validated the lead candidate miRNAs identified in GF mice by following alternative microbiota manipulation strategy in a different species. We used brain tissue from a separate cohort of rats exposed to antibiotics during adulthood following normal development of the gut microbiota prior to weaning [30]. Behavioural assessment of these rodents post microbiota depletion showed impairments in cognition and induced depressive-like behaviours [30]. We found that alterations in miR-219a-2-3p expression in both the amygdala and PFC was a common feature of both GF status and animals rendered microbiota-deficient post-weaning by antibiotic exposure. Sequencing-based studies have found miR-219a-2-3p/miR-219-3p to be altered in the basolateral amygdala following social defeat [46]. Thus, even when the gut microbiota is present during early life, subsequent depletion via antibiotics still indicates a role of gut microbiota in miRNA regulation. Whether gut microbiota-directed interventions that produce more qualitative differences in the microbiome also differentially regulate miRNAs is an open question. Colonization of microbiota-deficient rodents with the microbiota from mice with different behavioural phenotypes, as described by Bercik et al., or incremental antibiotic doses, would also provide valuable insights [47]. Additionally, direct strain comparison between GF and antibiotic-treated mice may better reveal brain-specific miRNAs sensitive to changes in the gut microbiota.

Both strategies employed in our study to investigate the effect of the gut microbiota on miRNA expression implicated miR-206-3p as a target of the gut microbiota, which was decreased under both conditions (Figs. 2g and 4b). This particular miRNA is well validated in regard to its role in the regulation of BDNF, an essential neurotrophin, which promotes growth and development of new neurons, survival of existing ones, and has an essential role in synaptic plasticity [48]. Altered BDNF expression is a hallmark of a disturbed microbiota-gut-brain axis, and it has been repeatedly shown to be altered in GF and antibiotic-treated rodents [18, 27, 49, 50]. BDNF mRNAs are proposed to contain conserved binding sites for at least 14 different miRNA families with many predicted interactions validated in vivo [51]. Specifically, miR-206-3p has been implicated in regulating the levels of BDNF in different animal models of neurological disorders [51, 52]. We have previously shown that there is a significant increase in the mRNA levels of a specific BDNF transcript [27] in the amygdala, consistent with the finding that miR-206-3p is significantly downregulated in this region in the current study. Additionally, a decrease in miR-206-3p in our antibiotic-treated rats coincided with a decrease in BDNF mRNA [30]. Previous work has revealed that targeting BDNF signalling by miRNAs has therapeutic potential in neurodegenerative and psychiatric diseases [48]. Thus, BDNF levels may be regulated by microbiota-induced alterations in miRNA expression and thus holds potential for a new treatment strategy in many CNS-related disorders.

Our analysis of predicted mRNA targets reveals KEGG pathways enriched for neurotrophin signalling along with a prominent representation of targets enriched in GO terms for nervous system development, neurogenesis, neuronal development and differentiation. miRNA dysregulation and/or dysfunction are believed to be underlying factors contributing to neurodegenerative diseases and neurodevelopmental abnormalities [53]. In fact, GF status results in altered stress circuitry along with several well-documented behavioural alterations, cognitive impairment, impaired microglial activation and abnormally altered myelination [4]. A recent study in GF mice demonstrated that within the amygdala, there is significant volumetric expansion within the lateral, basolateral and medial nucleus when compared to CON mice [19]. Coinciding with this, GF mice display hypertrophy of aspiny interneurons and pyramidal neurons along with changes in spine density [19].

Whether this change in gross morphology of the amygdala in adulthood is related to the large change in miRNAs remains to be determined. However, linking changes in miRNAs to changes in amygdala volume has been demonstrated in a rat model of autism using valproic acid. Administration of valproic acid coincides with an enlarged amygdala and increased miR-30d and miR-181c (~ 1.2 fold increase). Functional enrichment of predicted targets of both miRNAs indicates enrichment in GO terms for tissue morphology, nervous system development and cellular development, which was confirmed by in vitro inhibition of miR-181c [54]. We also see a trend towards a significant increase in miR-181c-5p (1.4 fold increase) in our sequencing data (Additional file 1: Table S1), which may be contributing to the gross morphological changes seen in GF mice.

The exact mechanism through which the gut microbiota can influence the expression of non-coding RNAs such as miRNAs remains unclear. Currently, most evidence supports signalling through the vagus nerve [14, 55] and bacterial metabolites such as short-chain fatty acids (SCFAs), which indirectly affect the nervous system through their immunomodulatory functions [56, 57]. Recently, a study investigating the impact of the gut microbiota on myelination demonstrated that certain bacterial metabolites in vitro can impair oligodendrocyte differentiation [58].