Address for Correspondence Jane A. Foster, Brain-Body Institute, St. Joseph’s Healthcare, 50 Charleton Ave., Tower – T3308, Hamilton, ON L8N 4A6, Canada. Tel: 905 522 1155, x. 35993; fax: 905 540 6593; e-mail: email@example.com
Background There is increasing interest in the gut-brain axis and the role intestinal microbiota may play in communication between these two systems. Acquisition of intestinal microbiota in the immediate postnatal period has a defining impact on the development and function of the gastrointestinal, immune, neuroendocrine and metabolic systems. For example, the presence of gut microbiota regulates the set point for hypothalamic-pituitary-adrenal (HPA) axis activity.
Methods We investigated basal behavior of adult germ-free (GF), Swiss Webster female mice in the elevated plus maze (EPM) and compared this to conventionally reared specific pathogen free (SPF) mice. Additionally, we measured brain mRNA expression of genes implicated in anxiety and stress-reactivity.
Key Results Germ-free mice, compared to SPF mice, exhibited basal behavior in the EPM that can be interpreted as anxiolytic. Altered GF behavior was accompanied by a decrease in the N-methyl-D-aspartate receptor subunit NR2B mRNA expression in the central amygdala, increased brain-derived neurotrophic factor expression and decreased serotonin receptor 1A (5HT1A) expression in the dentate granule layer of the hippocampus.
Conclusions & Inferences We conclude that the presence or absence of conventional intestinal microbiota influences the development of behavior, and is accompanied by neurochemical changes in the brain.
Within the first few days of birth the infant gut is colonized by commensal intestinal microbiota, an event that heralds the beginning of a mutually beneficial lifelong relationship. In the human lower intestine, numbers of bacteria reach 1014 organisms per gram of intestinal contents with roughly 1800 different species accounted for.1 It is well-established that intestinal microbiota are essential to gastrointestinal development and function, regulate host inflammatory responses, and develop and sustain immune homeostasis.2,3 A rapidly growing body of evidence now also indicates that the microbiome acts as a metabolically active organ within the host, capable of influencing fat storage and metabolism.4–6 Gut-brain communication is important to human gastrointestinal and psychiatric illness, which is highlighted by the increased comorbidity found between anxiety disorders and both inflammatory bowel disease and the functional bowel disorders.7–10 However, while early postnatal bacterial colonization overlaps with a vulnerable time in the development of the central nervous system (CNS),11,12 to date there has been little research determining the effects of normal bacterial colonization on the development and function of the brain.
Exact mechanisms by which the brain and gut converse are not fully understood, but bi-directional communication through the autonomic nervous system, immune system and the hypothalamic-pituitary-adrenal (HPA) axis is well documented.13–15 Preclinically, intriguing findings report that subclinical doses of pathogenic bacteria alter both anxiety-like behavior and vagally mediated neural activation in the brain stems of infected mice in the absence of overt gut inflammation.16 However while these experiments are pertinent to the study of gut-brain communication, it is not clear whether non-pathogenic bacteria similarly communicate with the brain. To this end, Li et al. have more recently demonstrated that information regarding dietary induced changes to commensal microbiota may be transmitted to the brain and result in changes to learning and memory in rodents.17 These findings emphasize the importance of understanding the roles that not only intestinal microbiota may play in the communication between the gut and brain, but also diet and the nutritional status conferred by the microbiota to the host.
Based on work to date we hypothesized that commensal intestinal microbiota normally communicate with the brain and their presence influences CNS development and behavior. To test this hypothesis, we carried out experiments using germ-free (GF) mice. Previous work using these animals has demonstrated that GF mice exhibited hyperresponsive HPA axis activity following stress as compared to specific pathogen free (SPF) mice,18 however no associated behavioral changes were investigated in this study. To this end we investigated the basal behavioral phenotype in adult GF VS SPF mice. Converging lines of evidence from preclinical and clinical literature suggest that plasticity-related genes, particularly brain-derived neurotrophic factor (BDNF) in the hippocampus, influence CNS stress circuitry19–25 and stress-related behaviors26–30 and we selected to examine hippocampal BDNF mRNA expression in GF and SPF mice. Hippocampal gene expression of the serotonin receptor 5HT1A was also selected for study, as it is the most commonly linked of the serotonin receptors to emotional behavior and anxiety.31,32 Moreover, we selected the N-methyl-D-aspartate (NMDA) glutamate receptor subunits NR1, NR2A and NR2B for examination, as NMDA receptors are known to play an important role in synaptic development and plasticity, learning and memory and in the extended amygdala are thought to be involved in central expression of anxiety.33,34
Materials and Methods
Female Swiss Webster, 8-week-old GF and SPF mice were obtained from Taconic Farms Inc., Germantown, NY, USA. Germ-free mice were maintained in regular cages (16.5 × 28.1 × 12.7 cm) inside a guaranteed germ-free shipper with free access to food and water until the beginning of testing. Upon arrival, SPF mice were transferred to microisolator cages (15.2 × 26.7 × 12.7 cm) and maintained in standard housing until the beginning of testing. All animals were housed under 12 h light–12 h dark cycle, lights on at 7 am. Housing room temperature was maintained at 20 °C and humidity at 60–70%. With the exception of shipping container type, there were no differences in supplier, shipping or receiving of GF and SPF mice.
All experimental procedures followed the guidelines of the Canadian Council on Animal Care and were approved by the Animal Research Ethics Board, McMaster University, Hamilton, Ontario, Canada.
Behavioral testing was performed 48 h after animals arrived in the facility (n = 12 per housing condition). While a longer acclimatization period would be better for behavioral testing, this was the maximum time that was logistically possible to keep the animals in their germ-free state. Prior to behavioral testing, GF mice were transferred into microisolator cages (four per cage). Locomotor activity was tested first and 3 h later elevated plus maze (EPM) testing was completed. Mice were returned to the housing room between tests. In a separate cohort of animals (n = 12), blood and brains were collected 48 h after animals arrived in the facility for corticosterone and in situ hybridization analysis. All blood was collected between 10 am and noon.
Animals were transported to a non-colony room for behavioral testing. Plexiglas activity chambers (40 × 40 × 35 cm) were interfaced to a Digiscan 16 monitor and a computer that provided automated recording of locomotor activity using VersaMax software (AccuScan Instruments, Columbus, OH, USA). The activity chambers were covered with ventilated Plexiglas lids. Animals were left undisturbed in activity chambers for 30 min and general locomotor activity for each animal was acquired.
Elevated plus maze
Animals were transported singly to a different non-colony room. Each animal was placed in the center of the EPM and performance was videotaped throughout a 5-min session. Video recordings were analyzed. An entry was scored when all four paws were in an arm of the maze. Data were collected on the number of entries into the open and closed arms of the maze and time spent in different arms.35
In a second cohort of animals (n = 12 per housing condition), blood and brain tissue was collected 48 h after arrival in the facility, as described above. Blood was collected and processed to collect plasma, and stored at −70 °C until use. Brains were rapidly removed following decapitation, frozen in −60 °C isopentane, and stored at −70 °C until cryostat sectioning.
Corticosterone (CORT) was measured in duplicate samples using a standard radioimmunoassay kit from MP Biomedicals (Salon, CA, USA).
In situ hybridization
Standard in situ hybridization methods were used, details of which have been previously described.36,37 The BDNF riboprobe was generously provided by Drs J. Lauterborn and C. Gall, University of California Irvine, Irvine, CA. The antisense probe produced from the BDNF cDNA template is a 382 bp probe complementary to the coding region of mouse BDNF mRNA (bases 1028–1410, NM_007540). The 5HT1A receptor riboprobe was generously provided by Dr. Pat Levitt, University of Vanderbilt, Nashville, TN. NR1, NR2A and NR2B riboprobes were generated in our laboratory. NR1 primers forward 5′-GTCCTCTGCCATGTGGTTTT-3′ and reverse 5′-GGACAGGGACACATTTTGCT-3′, NR2A primers, forward 5′-CAGCTGAAGAAGATCCACTCCT-3′ and reverse 5′-GCAGTGGTTAAG ATCCCAAGAC-3′, and NR2B primers forward 5′-TAGCTATAGAGGAGCGCCAATC-3′ and reverse 5′-CTCGATTTCATCAAACTCCCTC-3′ were designed using Primer 3 online software (NR1)38 or obtained from the Allen Brain Atlas (NR2A, NR2B).39 Specificity of primers to mouse NR1, NR2A and NR2B mRNAs was confirmed using BLAST.40 PCR generated cDNA (521 bp NR1, 253 bp NR2A, 694 bp NR2B) was inserted in the pGEM T-easy expression vector (Promega, Mississauga, ON, Canada). Antisense and sense probes were transcribed from linearized plasmids with α-35S-UTP (specific activity >1000 Ci mmol−1; Perkin Elmer, Boston, MA, USA) using appropriate RNA polymerases. Hybridization of sense probes did not reveal any signal.
All data was analyzed using GraphPad Prism (LaJolla, CA, USA). Statistical significance was evaluated using two-way anova with Bonferroni post-tests for locomotor activity (time and housing as factors) and EPM (housing and location as factors). Two-tailed unpaired t-tests were used for comparison of BDNF and 5HT1A receptor mRNA signal in different hippocampal subregions in GF and SPF mice, and for amygdala subnuclei expression of NR1, NR2A and NR2B mRNAs. Statistical criteria for significant differences were set at P < 0.05. Data are presented as mean ± SEM.
Germ-free status was confirmed by microbiological evaluation and showed no growth for anaerobic, aerobic, and mycotic bacteria (Taconic, Germantown, NY, USA).
Total distance traveled in the activity chambers did not differ between SPF and GF mice (SPF = 9005 cm ± 106, GF = 10 654 cm ± 1455, t = 1.24, d.f. = 22, P = 0.113). Fig. 1 shows the distance traveled in the open field at 5-min intervals over the 30-m test period. No difference was observed in these values between GF and SPF mice (F1,105 = 0.86, P = 0.36).
Germ-free mice showed increased open arm exploration in the EPM compared to SPF mice (Fig. 2A; F1,44 = 22.8, P < 0.0001). Post hoc analysis showed a significant increase in time spent in the open arms by GF mice compared to SPF mice (Bonferroni, P < 0.01) and a significant decrease in time spent in the closed arm by GF mice compared to SPF mice (Bonferroni, P < 0.01). This was accompanied by an increased number of open arm entries (Fig. 2B; F1,44 = 8.23, P = 0.0063; Bonferroni P < 0.001), however, there was no significant difference between GF and SPF mice in number of entries into the closed arms (Bonferroni, P > 0.05). We conducted a minute-by-minute analysis of EPM open arm entries which revealed that GF mice continued to explore the open arms with the same frequency as the test progressed while the SPF mice showed less open arm entries per minute as the test continued (Fig. 2C). These behavioral data have been presented previously in poster form at the Neurogastroenterology & Motility meeting in 2008.41
Plasma corticosterone levels were determined in mice 48 h after arrival in the facility. Germ-free mice showed significantly higher corticosterone levels compared to SPF mice (Fig. 3; t = 2.48, d.f. = 22, P = 0.021). It is possible that elevated corticosterone in GF mice in this experiment reflects an increased stress response related to the short acclimatization period (48 h) prior to tissue collection as previous work did not detect differences in basal corticosterone in Balb/C germ-free mice.18
Altered gene expression
Housing (GF VS SPF) had a significant impact on BDNF mRNA expression in the hippocampus. Representative film images are provided for BDNF mRNA signal observed in SPF (Fig. 4A) and GF (Fig. 4B) mice. Analysis revealed a specific significant up-regulation of BDNF mRNA in the dentate gyrus of the hippocampus (Fig. 4C, t = 2.97, d.f. = 20, P = 0.0076). There were no significant differences in BDNF expression in other subregions of the hippocampus (P > 0.05). Housing (GF VS SPF) also had a significant impact on 5HT1A receptor expression in the hippocampus. Representative film images are provided for 5HT1A receptor mRNA expression observed in SPF (Fig. 5A) and GF (Fig. 5B) mice. Analysis revealed a specific significant down-regulation of 5HT1A receptor mRNA in the dentate gyrus of the hippocampus (Fig. 5C, t = 3.18, d.f. = 20, P = 0.0047). There were no significant differences in 5HT1A receptor expression in the CA1 subregion of the hippocampus (P > 0.05). Housing (GF VS SPF) had a significant impact on NR2B mRNA expression in the amygdala. Representative film images of NMDA subunit mRNA expression in SPF mice are provided (Fig. 6). As indicated by the outline in Fig. 6B, D, mRNA for the central amygdala (CeA) was measured at Bregma −1.06 mm and for the lateral amygdala (LA) and basolateral amygdala (BLA) at Bregma −1.94 mm.42 Densitometric analysis revealed a significant down-regulation of NR2B mRNA in the CeA region of the amygdala (Fig. 7C, t = 2.82, d.f. = 22, P = 0.0099). There were no significant differences in NR2B mRNA expression in other subregions of the amygdala (P ≥ 0.05). There were no significant differences between GF and SPF mice in NR1 or NR2A mRNA gene expression in any of the subnuclei of the amygdala (Fig. 7A, B; P ≥ 0.05).
As illustrated by several recent comprehensive reviews, the topic of intestinal microbiota and its impact on the nervous system43,44 has broad applications for both gastrointestinal and psychiatric illness,45,46 perhaps especially for those like the functional bowel disorders that show co-morbidity across both systems. Clearly then, the study of the impact of gut microbiota on behavior and mRNA gene expression in brain tissue is of high clinical relevance. Our data suggest that gut bacteria influence the development of behavior in mice. In this study we used the EPM, an ethologically and pharmacologically validated tool for the assessment of rodent anxiety-like behavior,47,48 to examine basal behavior of GF and SPF mice. The basal behavioral phenotype observed in EPM activity in adult GF as opposed to SPF mice is interpreted as anxiolytic. Additionally, we report central changes in gene expression in plasticity-related genes that have been established to play a role in emotional behavior in mice. Specifically we observed a significant basal decrease in NMDA receptor NR2B mRNA expression in the central amygdala, an increase in hippocampal BDNF mRNA expression and a decrease in 5HT1A receptor mRNA expression in the dentate gyrus of GF compared to SPF mice.
Our behavioral data show no differences in locomotor activity in GF and SPF mice, but increased open arm exploration in GF mice in the EPM compared to SPF mice. These data must be considered within the context of limitations to our experimental design.
1 The order and timing of behavioral testing in rodents is known to impact the outcome and in particular, testing anxiety-like behavior in the EPM is influenced by prior testing.49,50 The design of our behavioral testing was not ideal as both behavioral tests occurred on the same day, and therefore we can not be certain that the exposure to the open field did not effect the EPM behavioral data.
2 Our GF mice were exposed to SPF conditions in the 3-h period prior to EPM testing and while GF status was confirmed prior to shipping, we did not assess the microbiota status at time of behavioral testing.
3 Estrous cycle can influence stress-related behaviors51,52 and differences in estrous may have contributed variation to our data.
4 In a second group of mice, our corticosterone analysis showed higher levels in GF mice 48 h after arrival in the facility. As Sudo et al.18 have previously shown an exaggerated corticosterone response to stress, it is possible that the increased levels of plasma corticosterone reflect a stress response in GF mice to experimental conditions. Alternatively, our work was completed in female Swiss Webster mice compared to male Balb/C GF mice in previous work so both strain and sex could contribute to the differences observed.50
Considering these limitations is important, however, one would expect that these factors would increase anxiety-like behavior in mice and we did not observe this. The open arm exploration time and the open and closed arm entry data observed in our SPF are comparable to published data for these measures in Swiss Webster mice that have had longer habituation periods.53,54 In addition, based on the increased corticosterone level, one would expect to see increased anxiety-like behavior in GF mice and we saw a robust anxiolytic behavioral phenotype on several outcome measures in the EPM. These observations are in line with previous findings in male Swiss Webster mice showing that anxiety-like behaviors in the EPM are not related to corticosterone levels.55 Our molecular data provide initial insights into the neurobiological pathways underlying this behavioral phenotype.
Our finding of a downregulation in gene expression of the glutamate NMDA receptor subunit NR2B in the central amygdala of GF mice is intriguing. NMDA receptors are heteromeric complexes, and are made up of both NR1 and NR2 subunits. Previously, Sudo et al.18 used PCR and showed downregulation in NR2A mRNA in the cortex and the hippocampus of GF mice compared to SPF mice, however, we did not detect differences in subregions of the hippocampus by in situ hybridization. We extended our analysis to the NR2B subtype and to include the amygdala. The NR2B subtype is the critical receptor in amygdala synaptic plasticity and development, and in learning and memory.56 Additionally, NMDA receptor antagonists are known to block anxiety in both mice and rats.57,58 In a previous study examining ethanol-withdrawal induced anxiety in rats, administration of NMDA receptor antagonists resulted in anxiolytic-like behavior as measured in the EPM, with treated animals showing increases in both time spent in the open arm of the EPM, and open arm entries.59 This behavioral phenotype is comparable to that of our GF animals, and was accompanied by the basal decrease in NMDA receptor expression. Antagonists specific to NR2B block the acquisition of amygdala-dependent fear learning60 thereby further illustrating the role that this NMDA receptor subtype plays in the expression of anxiety, fear and CNS plasticity. It is possible that the downregulation of NR2B mRNA in our GF mice is contributing the anxiolytic-like phenotype that we observe.
Serotonin receptors are distributed throughout the CNS and receive input from neurons of the dorsal and ventral raphe. Several 5HT receptors have been implicated in anxiety-like behaviors,61–63 however, the 5HT1A receptor has received more attention in both clinical and preclinical work. 5HT1A is both a presynaptic autoreceptor and a postsynaptic receptor. Pharmacological interventions in rodents show that activation of both receptors can reduce firing of serotonergic neurons.64 Reduced firing would result in reduced 5HT synthesis, reduced 5HT release in projection areas, and reduced 5HT turnover.65,66 A role for the 5HT1A receptor in anxiety-like behavior is well supported by pharmacological studies showing an anxiogenic effect by 5HT1A agonists,67–69 and by genetic studies showing increased anxiety-like behaviors in 5HT1A-deficient mice.70–72 Hence, it is reasonable to suggest that changes in serotonergic signaling in GF mice may contribute to the altered anxiety-related phenotype observed. It is of interest to continue this work and consider how serotonergic signaling in the hippocampus and the hypothalamus might be altered in GF mice.
Up-regulation of BDNF mRNA in the dentate region of the hippocampus in the GF mice is also consistent with literature identifying a role for this molecule in anxiety-like behaviors.21,73 Recent work has demonstrated that impaired BDNF signaling in the dentate gyrus of adult mice results in a marked increase in anxiety-like behavior.26,28 Restraint stress is known to decrease expression of hippocampal BDNF,19 activate the HPA axis,74,75 and result in increased anxiety-like behavior in rodents.76 While these reports support our suggestion that higher BDNF levels are related to the observed reduction in anxiety-like behaviors in GF mice, the whole story is certainly more complex. Importantly, Sudo et al.18 observed decreased levels of BDNF protein in the hippocampus of GF mice. These differences may relate to differences at the level of mRNA and protein as a modification at the level of the transcript does not necessarily translate to a similar change at the level of the protein expression and protein levels were not examined in the current study. Alterations in housing conditions such as psychosocial enrichment can also positively impact hippocampal BDNF expression levels and suppress stress-related behaviors.77
The role of intestinal microflora in shaping a fully functional immune system is established.3,78–80 Evidence of behavioral alterations in conjunction with adaptive immune deficits have been demonstrated by Cushman et al.81 who showed that deletion of the recombinase activating gene (RAG-1) in mice which results in absent antibody synthesizing capability, caused increased exploration in the open field and decreased open arm avoidance in the EPM. However, it should be noted that RAG-1 is also expressed in the CNS82 where its function has yet to be elucidated. It is possible that the known immaturity of the adaptive immune system of GF mice contributes to the behavioral phenotype that we observed.
The sensory arm of the autonomic nervous system may be a route whereby gut microbiota affect brain function and several reports make this link. Oral ingestion of a bifidobacterium by conventional Sprague Dawley (SD) rats has been shown to result in changes to serotonin metabolism in the brain stem.83 Consumption of Lactobacillus reuteri (LR) by conventional SD rats inhibited the perception of pain consequent to visceral distension84 and administration of Lactobacillus paracasei likewise normalized antibiotic-induced visceral hypersensitivity in mice during colorectal distension.85 Rousseaux et al. have reported that ingestion of yet another commensal, Lactobacillus acidophilus, by both conventional mice and rats promoted the expression of opioid and cannabinoid receptors by gut epithelial cells. Here too, ingestion resulted in decreased sensitivity to visceral distension.86 Taken together, these reports suggest that commensal bacteria can affect nerve function and pathways even in conventional animals. Recent work in our laboratories has shown that oral treatment of SD rats with LR consistently inhibits calcium dependent potassium channels in a specific subset of enteric neurons in the myenteric plexus,87 inhibits intestinal contractility,88 and hyperexcitability in dorsal root ganglion neurons.89 These observations thus afford a direct link between luminal commensals and the enteric nervous system.
To our knowledge, ours is the first work to demonstrate an altered behavioral phenotype associated with the absence of intestinal microbiota. Further work exploring these associations may lead to novel insights into the complex roles of commensal bacteria in the development and function of the CNS. Unraveling these pathways may eventually lead to new therapeutic approaches in dealing, for example, with the known significant comorbidity between irritable bowel syndrome and psychiatric illness.9
The authors are grateful for the start-up support received from the Brain-Body Institute, St. Joseph’s Healthcare, Hamilton, operating funds from National Science and Engineering Research Council of Canada (NSERC, to JAF), and equipment funds from Canadian Foundation for Innovation (to JAF) in the conduct of this project. Graduate stipend support (to KAN) was provided by Ontario Graduate Scholarship and Ontario Graduate Scholarship in Science and Technology. J.A. Foster was the principal investigator and together with J. Bienenstock was responsible for the study design. K.M. Neufeld and J.A. Foster conducted the behavioral testing, tissue collection and processing. Data analysis was completed by K.M Neufeld and N. Kang. All investigators contributed to manuscript preparation.
The authors have no conflict of interests to disclose.