Introduction

Gastrointestinal (GI) homeostasis is central to our general health and wellbeing.1, 2, 3 However, a specific role for gut health in the regulation of mood and behaviour is now emerging,4 particularly involving bidirectional communication along the microbiome–gut-brain axis.5, 6, 7, 8, 9 Bacterial colonisation of the intestine has a major role in the post-natal development and maturation of the endocrine and immune systems,2, 7 both of which underpin central nervous system (CNS) function6 and are of particular relevance to the developing serotonergic system.10, 11

Marrying these tenets together has led to the plausible hypothesis that the microbiota might have a key neurodevelopmental role. Indeed, studies exploring this concept have emerged demonstrating that germ-free (GF) animals had an exaggerated stress response compared with their conventionally colonised (CC) counterparts and that this aberrant endocrine profile could be at least partially reversed by reconstitution of the microbiota before a critical time window later in life.12 Moreover, hippocampal protein levels of brain-derived neurotrophic factor (BDNF) were also under the influence of the resident flora.

The functional behavioural consequences of growing up germ-free have been shown to reproducibly unmask a reduced anxiety phenotype compared with colonised animals in multiple independent laboratories and using different methods of assessment.13, 14, 15 A recent study that examined the effects of microbiota transplants between strains of mice with differential exploratory behaviour profiles offered further evidence of a microbial influence on anxiety levels.15 Challenging the GI tract with pathogenic microorganisms can also elicit an increased anxiety response.16, 17 Moreover, probiotic administration studies suggest that ingestion of Bifidobacteria or Lactobacilli can beneficially alter either anxiety or depressive-like behaviours, both under pathological conditions18 and in healthy animals.19, 20 Furthermore, we have previously shown that probiotic administration influences the availability of tryptophan, the serotonin precursor.21 Collectively, these studies suggest that microbiota manipulation could be a valid therapeutic strategy to modulate CNS signalling.6

There is limited data pertaining to the control exerted by the microbiota on the hippocampal serotonergic system, which is particularly relevant to stress, anxiety and depression.11, 13 This is despite the fact that microbiota-associated hippocampal alterations across multiple other neurobiological domains have been reported.6, 12, 14 The regulation of the serotonergic system is influenced by sex,22, 23 thus it is important to investigate the influence of microbiome–gut–brain axis in mice of both sexes. Indeed, it is worth noting that many GF studies have been only carried out in female mice, which are the sex of choice in immunological, but not neurobiological, studies.24, 25

Consequently, we focused on addressing whether the microbiota effects on the CNS extended to the serotonergic system in the hippocampus and whether such effects are sex dependent. Subsequently, we determined whether the alterations that we observed could be reversed following colonisation post weaning of animals previously lacking a gut microflora.

Materials and methods

Animals

GF and CC Swiss Webster breeding pairs were supplied by Taconic (Germantown, New York, USA) and first-generation offspring were used in all experiments. GF Swiss Webster mice were housed 4–5/cage in flexible film gnotobiotic isolators under a strict 12-h light/dark cycle. CC mice were similarly housed 4–5/cage in the standard animal facility under the same controlled conditions (temperature 20–21 °C, 55–60% humidity) on the same 12 h light/dark cycle. In all cases, both GF and CC mice received the same pelleted diet after it was autoclaved (sodium dodecyl sulphate diets, product code 801010). Experiments were conducted in accordance with the European Directive 86/609/EEC and the Recommendation 2007/526/65/EC, and were approved by the Animal Experimentation Ethics Committee of University College Cork.

Experimental design

Both the GF and CC mice were euthanized in adulthood at 6–9 weeks. Cohorts of both animal types (male and female) were randomly allocated to three groups: one for high-performance liquid chromatography (HPLC) and quantitative real-time PCR (qRT-PCR) analysis (n=9–10), a second for novel-environment stress-induced corticosterone production (n=9–10) and a third for immunological assessment including splenocyte stimulation (n=5–8). In a subsequent study, a group of male GF mice were colonised (n=9) by removing them from the GF facility post weaning (3 weeks) at which time they were allowed grow to adulthood in the conventional animal facility in cages with bedding and faecal matter from CC mice, a protocol that has previously demonstrated to be effective at restoring a normal microbiota.26, 27 All the three groups (GF, CC, and colonised GF) were assessed for anxiety behaviour in the light–dark box apparatus 24 h before sacrifice for HPLC analysis.

Sample collection and analysis

Unanaesthetised animals were euthanized in a random order by cervical dislocation. The brain was rapidly removed from the cranium, the hippocampus dissected out on an ice-cold plate and stored in either chilled HPLC homogenisation buffer as previously described28 or RNA later (as per the manufacturer's instructions) and stored at −80 °C for subsequent analysis. The spleen was carefully dissected out for splenocyte stimulation studies (see Supplementary information (SI) online for more details). Trunk blood was collected in tubes with ethylenediaminetetraacetic acid (EDTA) as anticoagulant, centrifuged (3500 g, 15 min, 4 °C) and the plasma stored at −80 °C for later analysis. The HPLC assay for 5-hydroxytryptamine/5-hydroxyindoleacetic acid (5-HT/5-HIAA) was based on a previously used method29 as was the HPLC assay for tryptophan and its kynurenine pathway metabolites30 (see Supplementary Information online for more details). qRT-PCR was carried out as described previously29, 31 as was the protocol for the light–dark box32 (see Supplementary Information online for more details). All samples were collected from 0900 to 1300 hours.

Novel-environment stress

CC and GF mice were individually removed from the home cage and placed in a novel cage (33 cm × 16 cm × 13 cm, L × H × W) for 30 min before sacrifice with subsequent collection of plasma for corticosterone analysis (see Supplementary Information online for more details).

Statistical analysis

The sample size was determined by a power calculation and aimed at detecting differences between groups at the 0.05 level. Data was expressed as mean±s.e.m. Statistical analysis was carried out using SPSS 18 for Windows (SPSS, Chicago, ILL, US). Statistically significant alterations were determined using a two/three-way analysis of variance followed by Bonferroni post-hoc tests or by one-way analysis of variance followed by Bonferroni post-hoc tests.

Results

Both male and female GF animals have a blunted immune response

A two-way analysis of variance (ANOVA) revealed no interaction between sex and GF status in the production of tumour necrosis factor α (TNF-α) following splenocyte stimulation with lipopolysaccharide (LPS) (F1,22=0.2, P=0.64), with a significant effect of GF status (F1,22=17.3, P<0.001). Post-hoc analysis revealed a blunted production of tumour necrosis factor α in both male (P<0.05) and female (P<0.01) GF animals compared with their CC counterparts (Figure 1a).

Figure 1
figure 1

(a) Blunted immune response in germ-free (GF) animals. Tumour necrosis factor α (TNFα) production (pg/ml) in GF and conventionally colonised (CC) animals (male and female) following stimulation of splenocytes with lipopolysaccharide (LPS; *P<0.05, male GF vs CC; $$P<0.01, female GF vs CC). (b) Exaggerated stress response in GF animals. Corticosterone release (pg/ml) following an acute stressor in male and female GF and CC animals. (**P<0.01, male GF vs CC; $$$P<0.001, female GF vs CC).

PowerPoint slide

Full size image

Both male and female GF animals have enhanced hypothalamic–pituitary–adrenal axis reactivity

There was no overall interaction between stress, sex and GF status in the release of corticosterone following a novel-environment stressor (F1,75=0.04, P=0.85). The same analysis indicated a significant interaction between stress and GF status (F1,75=21.84, P<0.001). Post-hoc analysis revealed an exaggerated production of corticosterone in both male (P<0.01) and female (P<0.001) GF animals compared with their CC counterparts (Figure 1b).

Decreased Bdnf expression in GF animals is sex specific

There was a significant GF × sex interaction for Bdnf expression levels in the hippocampus (F1,39=12.526, P=0.001). Post-hoc analysis revealed a significant decrease in the GF animals compared with their CC counterparts in male animals only (P<0.01, Figure 2a).

Figure 2
figure 2

(a) Brain-derived neutrotrophic factor (Bdnf) expression in the hippocampus. Decreased Bdnf expression (fold change) in the hippocampus of male germ-free (GF) mice compared with their conventionally colonised (CC) counterparts (**P<0.01, male GF vs CC). (b and c) Sex-specific alterations in serotonergic system. (b) Increased 5-hydroxytryptamine (5-HT) concentration (ng/g tissue) in the hippocampus of male GF mice compared with their CC counterparts (**P<0.01, male GF vs CC). (c) Increased 5-hydroxyindoleacetic acid (5-HIAA) concentration (ng/g tissue) in the hippocampus of male GF mice compared with their CC counterparts (***P<0.001, male GF vs CC). (d and e) Altered tryptophan availability and metabolism in GF animals. (d) Increased plasma tryptophan concentrations (ng/ml) in male GF animals compared with their CC controls (***P<0.001, male GF vs CC). (e) Decreased plasma kynurenine:tryptophan ratio in both male and female GF animals compared with their CC controls (***P<0.001, male GF vs CC; $$$P<0.001, female GF vs CC).

PowerPoint slide

Full size image

Hippocampal weight

A two-way ANOVA analysis of hippocampal weight indicated no effect of sex (F1,39=1.9, P=0.18), GF status (F1,39=0.02, P=0.9) or GF × sex interaction (F1,39=0.34, P=0.565) for this brain region.

Increased 5-HT concentrations in GF animals are sex specific

There was a trend towards a GF × sex interaction for 5-HT concentrations in the hippocampus (F1,39=3.86, P=0.057). The same analysis indicated a significant effect for GF status (F1,39=5.31, P=0.027) and sex (F1,39=4.61, P=0.038). Post-hoc analysis revealed a significant increase in 5-HT in GF compared with their CC counterparts (P<0.01) in male animals only (Figure 2b).

There was a significant interaction between GF status and sex for 5-HIAA concentrations, the main metabolite of 5-HT, in the hippocampus (F1,39=9.46, P=0.004). The same analysis revealed a significant effect for both GF status (F1,39=6.87, P=0.013) and sex (F1,39=13.43, P=0.0008). Post-hoc analysis indicated significantly increased 5-HIAA concentrations in GF compared with CC in male animals only (P<0.001, Figure 2c).

Tryptophan availability and metabolism is altered in GF animals

Although there was no significant GF × sex interaction for plasma tryptophan concentrations (F1,39=1.131, P=0.29), there was both a significant GF (F1,39=17.97, P=0.0001) and sex effect (F1,39=25.78, P<0.0001). Post-hoc analysis revealed a significant increase in plasma tryptophan availability in male (P<0.001), but not female (P=0.09), GF mice compared with their CC controls (Figure 2d).

Similarly, there was no significant GF × sex interaction (F1,39=1.3, P=0.26) in the kynurenine:tryptophan ratio, an indicator of tryptophan metabolism along the kynurenine pathway. However, there was both a significant GF (F1,39=64.78, P<0.0001) and sex effect (F1,39=14.63, P=0.0005) on this ratio. Post-hoc analysis revealed a significant decrease in this ratio in both male (P<0.001) and female (P<0.001) GF animals compared with their CC controls (Figure 2e).

GF status and serotonergic gene expression

There was an interaction between GF status and sex only for the 5-HT2C receptor gene (Table 1). Post-hoc analysis demonstrated no difference between GF animals of either sex and their respective CC controls for any of the targets assessed (Table 1).

Table 1 Gene expression data (expressed as fold change) for serotonergic targets in the hippocampus of male and female GF and CC mice
Full size table

GF status and body weight

A two-way ANOVA analysis of body weight data indicates an effect of sex (F1,39=60.269, P<0.001) but not GF status (F1,39=0.4, P=0.5). There was also a significant interaction between sex and GF status (F1,39=11.68, P=0.002). A Bonferroni post-hoc analysis revealed that there was no body weight difference between male GF animals and their CC counterparts (27.3±0.4 g vs 25.98±0.61 g, t=1.97, P>0.05) but female GF animals had a lower body weight than their CC counterparts (22.0±0.39 g vs 23.92±0.44 g, t=2.86, P<0.05).

Increased 5-HT and 5-HIAA concentration in GF animals is resistant to colonisation

There was a significant overall group effect for 5-HT concentrations in the hippocampus (F2,24=4.527, P=0.0215). A post-hoc analysis demonstrated that both GF (P<0.05) and colonised GF mice (P<0.05) had elevated hippocampal 5-HT concentrations compared with their CC control animals (Figure 3a).

Figure 3
figure 3

(a and b) Hippocampal serotonergic alterations in male germ-free (GF) animals are resistant to the effects of colonization. (a) Increased hippocampal 5-hydroxytryptamine (5-HT) concentrations (ng/g tissue) are maintained in male colonised animals (*P<0.05, conventionally colonised (CC) vs GF, colonised GF). (b) Increased hippocampal 5-hydroxyindoleacetic acid (5-HIAA) concentrations (ng/g tissue) are maintained in male colonised animals (*P<0.05, CC vs GF, colonised GF). (c and d) Tryptophan availability and metabolism is restored to control levels following colonisation. (c) Colonisation of male GF mice restores plasma tryptophan concentrations (ng/ml) to control levels (*P<0.05, GF vs CC; #P<0.05, colonised GF vs GF). (d) Colonisation of male GF mice restores the plasma kynurenine:tryptophan ratio to control values (**P<0.001, GF vs CC; ###P<0.001, colonised GF vs GF). (e) Anxiety levels depend on GF status. Colonisation of male GF mice restores the number of transitions in the light–dark box to control values (**P<0.01, GF vs CC; ##P<0.01, colonised GF vs GF).

PowerPoint slide

Full size image

There was a significant overall group effect for 5-HIAA concentrations in the hippocampus (F2,22=4.904, P=0.0173). A post-hoc analysis demonstrated that both GF (P<0.05) and colonised GF mice (P<0.05) had elevated hippocampal 5-HIAA concentrations compared with their CC control animals (Figure 3b).

Colonisation restores tryptophan availability/metabolism to control levels

There was an overall group effect for plasma tryptophan concentrations (F2,25=4.293, P=0.026). A post-hoc analysis demonstrated that GF animals had elevated plasma tryptophan concentrations compared with CC control animals (P<0.05). Colonisation of the GF animals reduced plasma tryptophan concentrations compared with GF levels (P<0.05, Figure 3c).

There was an overall group effect for the kynurenine:tryptophan ratio (F2,25=18.5, P<0.0001). A post-hoc analysis demonstrated that GF animals had a reduction in this ratio compared with the CC control animals (P<0.01). Colonisation of the GF animals increased the kynurenine:tryptophan ratio compared with GF animals (P<0.001, Figure 3d).

Reduced anxiety-like behaviour in GF animals is reversed following colonisation

There was an overall group effect for the number of transitions in the light–dark box (F2,25=6.974, P=0.004). A post-hoc analysis revealed that the GF animals were less anxious than their CC controls as evidenced by an increased number of transitions (P<0.01). Colonised GF mice displayed a reduced number of transitions compared with the GF animals (P<0.01, Figure 3e).

Discussion

Bacterial colonisation of the GI tract is integral to a variety of structural, metabolic and protective functions.7 Emerging evidence now also supports a role for the microbiota in the development of the CNS.6, 12, 14, 15 In our present work, we show for, what is to our knowledge, the first time that this remit extends to a marked influence on the hippocampal serotonergic system. Our results also highlight a sex-specific effect of the microbiota on this brain region, a factor that should inform future studies in GF animals seeking to explore the CNS consequences of an absent microbiota. In addition, we have demonstrated that these neurochemical, but not behavioural, alterations are resistant to a restoration of the GI microenvironment later in life.

As previous studies had demonstrated hippocampal alterations at both the protein and gene expression levels in GF animals,12, 14 we focused on this important brain region in our assessment of the impact of the microbiota on serotonergic function. We initially demonstrated that the consequences of growing up GF extends to a clear increase in both 5-HT and 5-HIAA, its main metabolite, over the normal levels of these neurochemicals present in the hippocampus following conventional colonisation of the GI tract. The fact that a chronic absence of the microbiota can elicit a 1.3-fold increase in 5-HT, a degree of magnitude not dissimilar to more traditional CNS-directed interventions (escitalopram and lithium),33 adds weight to recent speculation that the resident flora might represent a valid therapeutic target for CNS disorders.6 This should be tempered by our subsequent results, which demonstrated that this increase in hippocampal 5-HT concentration was maintained despite colonisation of the GF animals with a normal microbiota post weaning. In this regard, determining the impact of colonisation with specific strains of bacteria will be particularly informative.12

We then determined whether the neurochemical effects we demonstrated could also be observed at the gene expression level for key indices of serotonergic functionality. Despite the increased 5-HT concentration observed, we did not find an altered expression of the Tph2 gene, the key CNS isoform of the enzyme responsible for the synthesis of 5-HT from tryptophan,34, 35, 36 its amino-acid precursor, in either the male or female GF animals compared with their respective CC counterparts. However, this does not preclude the possibility that the intrinsic activity of this enzyme might be altered in GF animals. It is also noteworthy that the enzyme is not saturated at normal tryptophan concentrations so that enzyme-specific perturbations are not required per se to result in an increased 5-HT synthesis.37 Metabolomic studies indicate that colonisation of GF mice by the gut microbiota results in a 2.8-fold increase in plasma serotonin levels, which could theoretically also be due to a GI microbiota-mediated alteration in tryptophan metabolism.38 These findings are in agreement with the more generalised findings from metagenomic studies, which suggest the microbiome significantly enriches the metabolism of amino acids.39

Similarly, no alteration in gene expression level was found for either the SERT (serotonin transporter) gene or the range of serotonergic receptor genes evaluated (5-HT1A, 5-HT6 and 5-HT2C). Interestingly, decreased 5-HT1A receptor gene expression has been reported specifically in the dentate granule layer of the hippocampus for female GF animals14 using in-situ hybridisation. It is possible that more subtle changes in the hippocampal serotonergic system may occur at the gene expression level in male GF animals than our qRT-PCR approach can detect in whole-tissue homogenates.

It is noteworthy that many of the CNS alterations we found occurred in a sex-specific manner. Only male GF animals exhibited the serotonergic alterations described above as well as the decreased Bdnf expression. Interestingly, this sex specificity does not extend to peripheral immunological or neuroendocrine measures with both male and female GF mice having a blunted immune response to LPS stimulation and an exaggerated release of corticosterone following an acute stressor compared with their CC counterparts. The latter findings are in line with the available literature on GF status.7, 12 Future studies seeking to exploit the GF paradigm should take this into account and choose the appropriate sex of animal according to the outputs under investigation.

The mechanism surrounding these sex differences are not well understood but may relate to the well-known, but complex, influence of the oestrous cycle hormones on the CNS serotonergic system.40 For example, estrogen and its receptor (ERβ) is known to have a role in the modulation of hippocampal serotonin concentrations.41 It is possible then that this system exerts a larger influence over CNS serotonin concentrations than the resident microflora. Clearly, further studies are required to confirm this hypothesis and the use of ovarietomized animals, notwithstanding the challenges of such an approach when working with GF mice, may be necessary to address this important issue.

Only male GF animals exhibited a decrease in hippocampal Bdnf mRNA expression which is in agreement with previous data in male animals, albeit at the protein level,12 but is discordant with recent data showing increased Bdnf mRNA expression in the hippocampus of female GF mice.14 One potentially significant design difference in the latter study by Neufeld and colleagues is that their experiments were carried out on commercially sourced GF animals just 48 h following arrival at their facility.6, 14 This caveat notwithstanding, these data once again highlight the importance of sex on the modulatory CNS effects of the microbiota. Increased emphasis is being placed on the role of sex in modulating neurotransmitter systems relevant to stress, anxiety and depression.42, 43 Nonetheless, caution should be exercised in any attempt to translate sex differences in rodents to human populations.24, 25 It is also worth noting that although BNDF is an important neurotrophin for hippocampal development and repair, we did not observe any weight differences in this brain region between the GF and CC animals, regardless of sex. Nevertheless, the decreases in hippocampal BDNF in the male GF animals may be a compensatory response to the persistent elevation in 5-HT concentrations in that brain region.

We next investigated whether alterations in the peripheral availability of tryptophan might be at the root of the serotonergic alterations we observed. In support of this theory, we found that plasma tryptophan concentrations were significantly elevated in male GF animals only. Since CNS concentrations of tryptophan are predominantly determined by this circulating pool,37 our data suggests an increased supply of this amino acid precursor of serotonin to the CNS and a potential humoral route through which the microbiota can influence the CNS in addition to the recently validated neural mechanism of action.19 Previous studies have demonstrated alterations in tryptophan availability following probiotic administration.21 The fact that a similarly significant increase in plasma tryptophan concentration is not evident in the female GF animals reinforces the importance of investigating sex differences at different nodes of the microbiome–gut–brain axis and may explain why we did not see any increases in hippocampal serotonergic concentrations in our female GF animals.

Degradation of tryptophan along the kynurenine pathway actually represents the dominant metabolic fate for this amino acid37 and the ratio of kynurenine, the main metabolite, to the parent compound has been extensively used as an index of activity along this metabolic cascade.30, 44, 45, 46 In our study, this ratio indicated a reduced metabolism of tryptophan along the kynurenine pathway in male GF animals compared with their CC controls. Interestingly, this was also evident in the female GF animals where, as discussed above, it did not result in significantly altered tryptophan concentrations.

Given that the majority of central effects of GF status were unmasked in male animals, we used these for our colonisation studies where we sought to determine if restoration of a normal microbiota later in life could also reinstate normal CNS 5-HT concentrations. As indicated earlier, this proved not to be the case with the aberrant serotonergic profile maintained in the colonised GF animals. Conversely, circulating tryptophan levels were restored to control levels, accompanied by a normalisation of the kynurenine:tryptophan ratio. This reversal of plasma tryptophan availability, while reaffirming microbiota manipulation as a viable strategy for tryptophan and kynurenine pathway modulation, was apparently insufficient to impact on the hippocampal 5-HT aberrations. Thus, it appears that the neurochemical consequences of early-life intestinal microflora absence, during which time the hippocampus is particularly susceptible to stressors,47 are much more difficult to reverse with an unbiased adjustment of the GI milieu later in life. This observation also extends to the serotonergic system whose neurons are among the first to develop and may consequently be most at risk to early-life disturbances in intestinal homeostasis.10 Indeed this time period, including early adolescence, represents a critical window during which the brain is programmed for later life.48, 49, 50 However, we cannot rule out the possibility that the changes in the hippocampal serotonergic system may take a longer time to reverse than the peripheral tryptophan levels.

GF status has been most consistently associated with reductions in anxiety-like behaviours, a feature which is consistent across multiple independent laboratories using different assessment strategies.6, 13, 14, 15 Here we show once again that the male GF mice display less anxiety-like behaviours than their CC counterparts, using the light–dark box paradigm. We also report that, in contrast to the neurochemical alterations described, colonisation of the GF animals restored normal anxiety-like behaviours. Previously, this has only been demonstrated for the progeny of what have been termed ‘recolonised animals’,13 not directly in the ‘recolonised’ animals themselves. Conversely, reconstitution of the intestinal microbiota has proved insufficient to reverse the alteration in anxiety at all in female mice.51 The former study used a different mouse strain, making direct comparison with our results difficult. Irrespective of these methodological and other differences,6 the current dataset support a role for the microbiota in the regulation of anxiety-like behaviours and further implicates our resident flora as a viable therapeutic target. Further behavioural studies are urgently required to elaborate on these findings, particularly in relation to whether GF animals display altered locomotor activity or reduced behavioural inhibition but also to characterise depression-related behaviours and cognitive performance.

Collectively, the current report and data from other GF studies indicate that there is a discordance in the increased stress reactivity physiologically and the reduction in anxiety behaviourally, which has now been observed in three different research groups using different tasks.13, 14, 15 Interestingly, chronic treatment with a Lactobacillus strain has been shown to reduce anxiety and also reduce stress-induced corticosterone responses when given to adult CC animals.19 This suggests that the effect on anxiety is probably not due to a general adult miscommunication in the microbiome–gut–brain axis but may be developmentally primed.

Despite the fact that 5-HT has been shown to regulate the developmental ontogeny of anxiety at critical time windows during development,10, 52, 53 our current data could be interpreted as exemplifying a dissociation between the elevations in hippocampal serotonin concentrations and anxiety-like behaviours in the light–dark box. However, the relationship between serotonin and anxiety is far from being straight forward with both an anxiogenic and anxiolytic role being proposed for elevations in 5-HT, much debate still existing over the neurocircuitry involved54, 55 and yet selective serotonin reuptake inhibitors being the first choice treatment in certain anxiety disorders.56 Moreover, it is worth noting that dietary depletion of tryptophan or administration of SSRIs has been shown to have no or modest effects on light–dark box behaviour that further supports a dissociation between hippocampal 5-HT and anxiety-like behaviours in this task.57, 58, 59 Finally, as exemplified by studies with SERT knockout mice, serotonin-mediated alterations in anxiety during development can be mouse-strain dependent,60 and thus future studies, where logistically possible, should focus on whether data generated generalise to other mouse strains.

The microbiota has a broad and critical role in supporting normal digestion and host metabolism and it is interesting to consider how this function might contribute to the results presented here. A significant energy source for humans is the bacterial metabolism of dietary fiber to short-chain fatty acids, agents that are of importance for gut motility, have a trophic effect on epithelial cells, impact on immune system development and modulate enterendocrine hormone secretion.2 Moreover, systemic administration of butyrate, a histone deacetylase inhibitor, has been shown to influence BDNF expression in the hippocampus.4 However, other alternatives are also possible. For example, certain microorganisms, including Lactobacilli, are able to convert nitrate to nitric oxide, a potent regulator of both the immune and nervous systems, while others can produce neuroactive amino acids such as γ-aminobutyric acid (GABA).4 Indeed, it is likely that multiple mechanisms, including but not limited to the examples quoted above, may contribute to our results. However, it should also be noted that there was no difference in the dietary availability of proteins, fatty acids, amino acids or vitamins between the GF and CC groups.

It is also worth considering whether the differences we have reported might be related to the influence of GF status on maternal care, a factor which can have an important role in shaping behaviour in adulthood.61 However, previous reports by Sudo and colleagues,12 who assessed maternal care and did not find any effect of GF status, suggest that this does not contribute to the repertoire of perturbations we have reported. Another caveat relates to the well-known influence of cyclical female sex hormone variation on behavioural and neurochemical parameters.62 Although this can be important, the short duration of the oestrous cycle in mice coupled with the duration of the study and the fact that female mice housed together can phase into the same cycle stage63, 64 would most likely result in not having a differential predominance of animals in any one particular phase of the cycle across groups.

The current study is limited by the common difficulty all GF studies have in directly translating the results to the clinical situation where no equivalent obliteration of the microbiota can be said to exist.6 Nevertheless, it may have direct relevance to neonatal care units, where current hygienic standards in developed countries coupled to the frequent use of broad spectrum antibiotics65 may be the closest approximation to the early-life GF environment under investigation in our experiments. Indeed, the altered commensal colonisation pattern engendered by such a scenario has already been linked to the risk of developing immune-mediated diseases in western infants.65 Our studies suggest that the broader implications for the CNS should now also be considered.

In conclusion, it is emerging that a more complete understanding of the molecular, cellular and physiological basis of microbiome–gut–brain communication is required before any potential therapeutic benefits of microbiota manipulation strategies can be generated.6 We have demonstrated a sex-specific role for the microbiota in the regulation of CNS serotonergic neurotransmission profiles. These studies offer further proof of the importance of the resident flora, both as an aetiological factor and in the development of novel treatment modalities for microbiome–gut–brain axis disorders.