1. Neuroscience
Download icon

Microbiota-driven transcriptional changes in prefrontal cortex override genetic differences in social behavior

  1. Mar Gacias  Is a corresponding author
  2. Sevasti Gaspari
  3. Patricia-Mae G Santos
  4. Sabrina Tamburini
  5. Monica Andrade
  6. Fan Zhang
  7. Nan Shen
  8. Vladimir Tolstikov
  9. Michael A Kiebish
  10. Jeffrey L Dupree
  11. Venetia Zachariou
  12. Jose C Clemente
  13. Patrizia Casaccia  Is a corresponding author
  1. Icahn School of Medicine at Mount Sinai, United States
  2. BERG, United States
  3. Virginia Commonwealth University, United States
Research Article
  • Cited 45
  • Views 7,735
  • Annotations
Cite as: eLife 2016;5:e13442 doi: 10.7554/eLife.13442

Abstract

Gene-environment interactions impact the development of neuropsychiatric disorders, but the relative contributions are unclear. Here, we identify gut microbiota as sufficient to induce depressive-like behaviors in genetically distinct mouse strains. Daily gavage of vehicle (dH2O) in nonobese diabetic (NOD) mice induced a social avoidance behavior that was not observed in C57BL/6 mice. This was not observed in NOD animals with depleted microbiota via oral administration of antibiotics. Transfer of intestinal microbiota, including members of the Clostridiales, Lachnospiraceae and Ruminococcaceae, from vehicle-gavaged NOD donors to microbiota-depleted C57BL/6 recipients was sufficient to induce social avoidance and change gene expression and myelination in the prefrontal cortex. Metabolomic analysis identified increased cresol levels in these mice, and exposure of cultured oligodendrocytes to this metabolite prevented myelin gene expression and differentiation. Our results thus demonstrate that the gut microbiota modifies the synthesis of key metabolites affecting gene expression in the prefrontal cortex, thereby modulating social behavior.

https://doi.org/10.7554/eLife.13442.001

eLife digest

A combination of genes and environmental factors underlie an individual’s risk of developing a mental illness. Among the environmental factors, it is becoming clear that communication between the gut and the brain is involved, but we do not understand how these two organs communicate. Our gut contains a variety of bacteria that help us to digest food and there is some evidence that changes in these bacterial communities can influence our mental health.

Transplanting feces from one individual to the gut of another is one way to alter the communities of bacteria in the gut. Here, Gacias et al. investigated whether fecal transplants are sufficient to induce social avoidance behavior – a symptom of depression – in mice. The experiments show that introducing specific combinations of bacteria into the gut is indeed able to cause healthy adult mice to avoid social interactions. This effect was caused by changes in the “myelin” sheath that surrounds many nerve fibers and could be prevented by giving the mice antibiotics, which decreased the number of bacteria in the gut.

Further experiments revealed that the mice that became depressed after fecal transplants had higher levels of a molecule called cresol, which is produced by certain gut bacteria. Gacias et al. found that cresol is able to reduce the amount of myelin produced by brain cells. Therefore, these findings show that changing the communities of bacteria in the gut can result in the accumulation of molecules that influence social behavior. Future work will aim to identify bacteria that can reduce the amount of cresol produced in the gut, which may have the potential to treat depression.

https://doi.org/10.7554/eLife.13442.002

Introduction

Despite the diffuse prevalence of mental illness and the large efforts spent in identifying genetic elements of susceptibility, there is a need to define the role of environment—gene interactions. In addition to genetic predisposition, there is extensive epidemiologic literature emphasizing the role of environmental exposure in the development of mild to severe mood disorders. The aftermath of traumatic life events, for instance, is often characterized by the onset of severe depression or post-traumatic stress disorder (Shalev et al., 1998). The interplay between genes and environmental variables has gained recent attention, and several immunologic and lifestyle contributors have been proposed to modulate depressive symptoms. The detection of high levels of serum cytokines and the higher incidence of depression in patients with autoimmune disorders (Postal and Appenzeller, 2015; Walker et al., 2011; Moll et al., 2011; van Hees et al., 2015; Feinstein et al., 2014) has suggested a role for neuroinflammation (Godbout et al., 2008; Menard et al., 2016; Audet et al., 2014). Deficiency of specific nutrients such as omega-3 fatty acids has been reported in subsets of patients with mental illnesses (Ohara, 2005; Patrick and Ames, 2015; Poudel-Tandukar et al., 2009; Panagiotakos et al., 2010), highlighting the link between mood disorders and the bioavailability of metabolites.

There is evidence that bioactive metabolites act as mediators of gut—brain communication, as shifts in gut microbial composition impact brain neurochemistry (Cryan and Dinan, 2012; Collins et al., 2012; Desbonnet et al., 2014; Bercik et al., 2010; 2011). Indeed, psychiatric comorbidities often accompany conditions characterized by an aberrant gut microbiota composition, such as irritable bowel syndrome, functional gastrointestinal disorder, and inflammatory bowel disease (Gevers et al., 2014; Morgan et al., 2012; Haberman et al., 2014; Carroll et al., 2011; Addolorato et al., 1997). Conversely, altered gut microbiota composition and function have been reported in patients with major depressive disorders and children with autism (Jiang et al., 2015; De Angelis et al., 2015; De Angelis et al., 2013; Parracho et al., 2005). The gut microbiota is a complex microbial ecosystem that rapidly responds to environmental changes and can modulate brain development, function, and behavior (Cryan and Dinan, 2012; Collins et al., 2012; Desbonnet et al., 2014; Bercik et al., 2010; 2011; Wu et al., 2011; Daniel et al., 2014; Lax et al., 2014). These studies suggest that social behavior may be affected by abnormal interactions between gut microbiota and the brain, though the underlying mechanisms remain only partially understood.

One hypothesis for the pathogenesis of depressive-like behaviors has been suggested through studies on social isolation in mice (Liu et al., 2012; 2016; Makinodan et al., 2012), which revealed a reduction of myelinated fibers in the prefrontal cortex (PFC), associated with changes in the oligodendrocyte transcriptome (Liu et al., 2012; 2016). Myelination is a dynamic process that continues into adulthood and contributes to physiologic brain function (Liu et al., 2012; 2016; Makinodan et al., 2012; Sánchez et al., 1998; Gibson et al., 2014; McKenzie et al., 2014). Oligodendrocytes are the myelinating cells of the central nervous system (CNS), and neuropathologic and transcriptomic studies have reported downregulated oligodendroglial transcripts and reduced myelin thickness in the brains of patients with schizophrenia, major depression, and bipolar disorder (Tkachev et al., 2003; Aston et al., 2005; Katsel et al., 2005). These data highlight the role of myelin in mental illness and depressive-like behaviors, though it remains to be established whether myelination in the adult PFC and social behavior are affected by alterations in gut microbiota composition. This study characterizes the gut microbiota in mice with social avoidance behavior and demonstrates that transfer of specific bacterial taxa is sufficient to alter adult PFC myelination and results in behavioral changes consistent with a depressive-like phenotype.

Results

Non-obese diabetic (NOD) and C57BL/6 mice display differential susceptibility to develop depressive-like symptoms in response to daily gavage

Although gastric gavage and subcutaneous injections are routine, daily procedures used to administer drugs or special diet to rodents, the potential behavioral effects they may induce in mice have not been investigated. Daily gastric gavage with vehicle for two weeks (Figure 1A) was sufficient to induce social avoidance behavior in NOD mice (Figure 1B), without affecting their overall locomotor activity (Figure 1C). This depressive-like behavior induced by daily gavage was dependent on the specific mouse strain, as C57BL/6 mice were not affected (Figure 1D,E) (Moy et al., 2008). Subcutaneous injection of vehicle did not elicit any behavioral effect in either strain (Figure 1—figure supplement 1). Daily gastric gavage with an antibiotic cocktail proven to deplete the gut microbiota (Reikvam et al., 2011) failed to induce the social avoidance behavior in NOD mice (Figure 1B), and similarly had no effect on the C57BL/6 mice (Figure 1D). The antibiotic regimen was well tolerated by both NOD and C57BL/6 mice, did not impact body weight or glucose levels, and did not result in any gastric hemorrhage or visible stomach damage (Figure 1—figure supplement 2 and Figure 1—figure supplement 3). Consistent with previous reports, only chronic oral antibiotic treatment (but not subcutaneous delivery) induced enlargement of the large intestine (Figure 1—figure supplement 2), a macroscopic sign associated with microbiota depletion (Reikvam et al., 2011). Interestingly, daily gavage also induced an anxiety-like behavior in both NOD and C57BL/6 mice, as revealed by the elevated plus maze (EPM) (Figure 2B,D). However, the anxiety-like behavioral change displayed in response to daily gavage was not affected by oral antibiotic treatment (Figure 2B,D), suggesting that only the depressive-like behavior is mediated by alterations in gut microbiota. To further validate this hypothesis, we conducted the forced swim test (FST), which is considered a measure of despair-like behavior, in NOD and C57BL/6 mice after daily gavage with either vehicle or antibiotics. The despair-like behavior was induced by vehicle gavage in the NOD strain, and was prevented by oral antibiotic treatment (Figure 2C), but was not detected in the C57BL/6 mice (Figure 2E). Together, these results indicate that daily gavage of vehicle induces social avoidance and despair-like behaviors in NOD mice, but not in C57BL/6 mice, and that this effect is not observed when gavaging antibiotics orally and not subcutaneously.

Figure 1 with 3 supplements see all
The strain-specific social avoidance behavioral response to daily gavage is affected by oral antibiotic treatment.

(A) Experimental timeline: vehicle or antibiotic mix were administered daily by gastric gavage (g.g.) for 14 days. Behavioral testing was performed before (baseline) and after treatment. (B–D) Results of the Social Interaction (SI) test for NOD (B) and C57BL/6 (D) mice. Oral antibiotic treatment did not affect locomotor activity measured during the social interaction test (C,E) (3 independent experiments with 8 mice per group/experiment for a total of n=23–24 mice per condition). Data are mean ± S.E.M; *p<0.05, **p<0.01 based on one-way ANOVA with Bonferroni’s post hoc test; n.s. indicates not significant.

https://doi.org/10.7554/eLife.13442.003
Figure 2 with 1 supplement see all
The strain-specific anxiety- and despair-like behavioral responses to daily gavage are differentially affected by oral antibiotic treatment.

(A) Experimental timeline: vehicle or antibiotic mix were administered daily by gastric gavage (g.g.) for 14 days. Behavioral testing was performed before (baseline) and after treatment. Figure shows the results for the Elevated Plus Maze (EPM) and Forced Swim Test (FST) for NOD (B, C) and C57BL/6 (D, E) mice after oral treatment (g.g.). Baseline measurements for FST were performed in a separate cohort of mice (n=10) to avoid carryover effects (3 independent experiments with 8 mice per group/experiment for a total of n=24 mice per condition). Data are mean ± S.E.M; *p<0.05, **p<0.01, ***p<0.001 based on one-way ANOVA followed by Bonferroni’s post hoc test; n.s. indicates not significant.

https://doi.org/10.7554/eLife.13442.007

Behavioral differences in genetically different mice are associated with specific gut microbiota composition

To further characterize the effect of vehicle and antibiotic treatment on gut microbiota composition, we conducted 16S rRNA sequencing analysis of cecal and fecal samples collected after behavioral testing and after 14 days of treatment (Figure 3A). Unweighted UniFrac distances (Lozupone and Knight, 2005) were calculated between all pairs of fecal samples based on their microbiota composition. Based on these distances, Principal coordinate analysis (PCoA), an ordination method conceptually similar to principal component analysis, revealed a clear separation between vehicle-gavaged and baseline NOD (Figure 3) and between vehicle and antibiotic treated NOD and C57BL/6 mice (Figure 3—figure supplement 1A). PCoA analysis revealed clear differences between NOD mice before (“baseline”) and after oral treatment with antibiotics (Figure 3B), with differences also observed between samples before and after treatment with vehicle (Figure 3C). Since the depressive-like behavior was only observed in oral vehicle-treated NOD mice, we focused on identifying the specific microbiota that differ in these animals before and after treatment. Analysis of Operational Taxonomic Units (OTUs, defined as groups of 16S rRNA gene sequences with high similarity and that broadly correspond to a bacterial species) identified several taxonomic groups that were exclusively found in the vehicle-treated mice (Figure 3D and Gacias et al., 2016). These taxa represent potential candidates associated with the depressive-like phenotype observed in NOD mice. Linear discriminant analysis effect size (LEfSe) (Segata et al., 2011), a biomarker discovery method based on the Kruskal–Wallis and Wilcoxon tests, was used to identify key bacterial taxa enriched in vehicle-treated versus antibiotic-treated animals in each strain (Figure 3—figure supplement 1B–E). As expected, Proteobacteria were enriched in antibiotic-treated animals, while vehicle-treated mice had enrichment in Bacteroidetes and Firmicutes (Figure 3—figure supplement 1C,E).

Figure 3 with 1 supplement see all
Enrichment of bacterial OTUs induced by gastric gavage (g.g.) in NOD mice.

(A) Experimental timeline indicating time points of fecal collection (arrows) relative to behavioral testing and treatment. (B,C) Principal coordinate analysis plots of unweighted UniFrac distances of microbiota in fecal samples at baseline and after 14 days of daily g.g. of antibiotics or vehicle in NOD mice. Each dot represents the microbiota of a sample (1 sample = feces pooled from 3–5 mice), color-coded by treatment (vehicle or antibiotic) and time-point. The percentage of variation explained by each principal coordinate (PC) is shown in parentheses. All samples were rarefied at 5000 sequences. (D) Analysis of unique Operational Taxonomic Units (OTUs) present in NOD vehicle-treated mice compared to their fecal microbiota at baseline. Figure shows representative taxa enriched in fecal samples of NOD vehicle-treated mice compared to their baseline samples. Each bar represents the microbiota of an individual sample (1 sample = 3–5 mice per cage). See Gacias et al. (2016).

https://doi.org/10.7554/eLife.13442.009

Analysis of tissue samples revealed similar differences between vehicle- and antibiotic-treated mice in both strains (p<0.01, adonis with 999 permutations). No significant changes in gut microbiota composition were detected when antibiotics were administered subcutaneously.

Modification of the gut microbiota following oral antibiotic administration induces unique changes in the medial prefrontal cortex adult myelination of NOD mice

To identify possible CNS transcriptional signatures associated with the behavioral outcomes described in the vehicle-gavaged, but not antibiotic-gavaged, NOD mice, we performed an unbiased transcriptomic analysis of the medial prefrontal cortex (mPFC) using RNA sequencing. This analysis revealed decreased expression of genes related to myelination (Figure 4 and Figure 4—figure supplements 1,2 and Gacias et al., 2016) in vehicle-gavaged NOD mice - characterized by social avoidance behavior - compared to antibiotic-treated mice, whose behavior was comparable to baseline controls (Figure 1). The differences in myelin gene transcripts in the mPFC of vehicle-gavaged NOD compared to antibiotic-treated mice were validated by quantitative real-time qPCR (Figure 4B) and immunohistochemistry (Figure 4C). These differences were detected only in NOD mice, and not in C57BL/6 mice that showed no change in social behavior with oral gavage (Figure 4D,E). The differences in myelin gene expression in the mPFC could not be attributed to a nonspecific effect of antibiotic treatment, as there were no differences observed after subcutaneous delivery (Figure 4—figure supplement 1). The regional specificity of the transcriptional changes was also assessed in NOD mice by evaluating samples from a distinct brain region, the nucleus accumbens (NAc), revealing no difference in the two treatment groups (Figure 4—figure supplement 1). These data provide further support for the relationship between defective mPFC adult myelination and depressive-like behavior, as indicated by the lower levels of myelin transcripts and reduced area of MBP immunostaining in vehicle-gavaged NOD mice exhibiting social avoidance. The results also demonstrate that the transcriptional and behavioral effects were prevented by oral antibiotic treatment.

Figure 4 with 2 supplements see all
Myelin transcripts and myelinated fibers in the medial prefrontal cortex (mPFC) of adult NOD mice with social avoidance behavior.

(A) Experimental timeline: vehicle or antibiotic mix were administered daily by gastric gavage (g.g.) for 14 days and mPFC was harvested for RNA extraction and quantitative real-time qPCR or immunohistochemsitry (B,D) qPCR of myelin transcripts after vehicle or antibiotic treatment of NOD (B) and C57BL/6 (D) mice. Values were normalized to 36b4 mRNA levels and are referred as fold change relative to vehicle-treated values (n=6 mice per group). (C,E) Representative confocal images and quantification of MBP+ fibers (red) in mPFC of NOD (C) and C57BL/6 (E) mice after vehicle or antibiotic treatment. DAPI (blue) was used as nuclear counterstain. Scale bar, 50 μm. Graph represents quantification of MBP+ fibers per surface area (n=3 for NOD; n=4 for C57BL/6). Data are mean ± S.E.M; **p<0.01, ***p<0.001 based on unpaired t test. n.s. indicates not significant.

https://doi.org/10.7554/eLife.13442.011

Transplantation of fecal microbiota from vehicle-treated NOD mice to depleted C57BL/6 recipients is sufficient to recolonize the gut and transfer transcriptional, and behavioral traits

To determine whether the social avoidance behavior and mPFC transcriptional changes induced by daily gavage of vehicle in NOD mice were caused by the enrichment of specific gut bacteria, we transferred the cecal content of vehicle-treated or antibiotic-treated NOD mice into C57BL/6 recipients, whose endogenous flora had been depleted by antibiotic treatment (Figure 5A). Social behavior in C57BL/6 depleted recipients was assessed before and after transplantation with microbiota from either vehicle-gavaged (Group I) or antibiotic-gavaged (Group II) NOD donors. The behavior of the C57BL/6 recipients resembled that of the donors: Social avoidance behavior was detected in Group I recipients, and was not observed in Group II recipients (Figure 5B,C). Intriguingly, transplantation of vehicle-gavaged NOD microbiota also transferred the transcriptional changes in the mPFC, but not in the NAc, as shown by the lower levels of myelin gene transcripts (Mag, Mog, Plp1, Mobp) in Group I mice compared to Group II recipients (Figure 5D). The functional consequences of the transcriptional changes in myelin genes were further validated by electron microscopy, and ultrastructural analysis revealed decreased myelin thickness in Group I recipients displaying a social avoidance behavior (Figure 5E). Quantification of myelin thickness relative to axonal diameter (g ratio) revealed that Group I recipients transplanted with vehicle-gavaged NOD microbiota, presented thinner myelin than Group II, recipients of antibiotic-treated NOD donors. No significant differences between the two groups were observed in the NAc (Figure 5E). The transfer of depressive-like behavior from donor to recipient was further validated by the detection of increased immobility at the FST in Group I mice compared to Group II (Figure 5—figure supplement 1).

Figure 5 with 1 supplement see all
Social avoidance behavior transfer from NOD donors to microbiota depleted C57BL/6 by fecal transplantation.

(A) Experimental timeline for donor (NOD) and transplant-recipient (C57BL/6) mice. (B,C) Results from Social Interaction (SI) tests conducted in C57BL/6 recipients before and after transplantation with either microbiota from vehicle-treated (Group I; B) or antibiotic-treated (Group II, C) NOD mice. Graphs represent the amount of time spent (seconds) in the interaction zone when a target is present. Red dashed bar represents the interaction time of the NOD donors. Data are mean ± S.E.M; *p<0.05, **p<0.01 based on a two-way ANOVA (n=12 mice per experiment, 2 replicates of 12 for a total of 24 mice per condition). (D) Graphs indicate the relative levels of myelin gene transcripts in mPFC and NAc of C57BL/6 recipients displaying (Group I) or not displaying (Group II) social avoidance behavior after transplantation with NOD microbiota (n=6–8 mice per group; *p<0.05, **p<0.01, ***p<0.001 based on unpaired t test). (E) Electron micrographs and quantified g-ratios of myelinated axons in mPFC and NAc in Group I and Group II C57BL/6 recipients after transplantation with the NOD microbiota. Scale bar, 1 μm. (n=3 per treatment and condition; statistical differences between groups were determined using two-tailed t-test; n.s. indicates not significant).

https://doi.org/10.7554/eLife.13442.014

Collectively, these findings suggest that the gut microbiota of vehicle-gavaged NOD donors was sufficient to transfer the depressive-like behavior, modulate transcript levels in the mPFC, and impact region-specific adult myelination in microbiota-depleted C57BL/6 recipients.

The genomic DNA content was measured in fecal pellets of C57BL/6 recipients to validate the depletion of the gut microbiota with 14 days of antibiotic treatment, and to evaluate the effectiveness of recolonization after transplantation (Figure 6B,D). Analysis of alpha diversity (the number of bacterial taxa present in a sample or group of samples) further confirmed the microbiota depletion (Figure 6C,E). In both groups, diversity was significantly reduced from baseline after antibiotic treatment (Figure 6C,E; p<0.01 ANOVA with Tukey’s honest significant difference (HSD) post-hoc analysis). As expected, after transplantation Group II mice still exhibited a significantly depleted diversity compared to baseline (Figure 6E; p<0.01 ANOVA with Tukey’s HSD), while bacterial diversity in Group I had recovered to levels similar to baseline and was not significantly different (Figure 6C; p=0.09, ANOVA with Tukey’s HSD). These results suggested that transfer of behavioral traits was associated with restoration of bacterial diversity to baseline levels. In order to determine the differences in microbiota compositions associated with the behavioral phenotype, we conducted PCoA analysis based on unweighted UniFrac analysis (Figure 6F). Although all pooled fecal samples from NOD donors and C57BL/6 recipients clustered together at baseline (samples on the right side of the plot), treatment with antibiotics resulted in a drastic reshaping of the bacterial communities of both NOD (middle of the plot) and C57BL/6 (bottom-left side) mice. The microbiota composition of Group II mice after transplant (which did not display social avoidance behavior) was distinct from baseline, similar to antibiotic-treated animals pre-transplant (top-left side). However, Group I recipients which displayed social avoidance behavior (#19 and #18 on the plot), had compositions that were close to those of their vehicle-treated NOD donors. In contrast, Group I recipients which did not display social avoidance behavior (#17 on the plot), clustered with Group II recipients. This result suggests that the transplant procedure was not equally effective in all Group I mice. The distance in microbiota composition between vehicle-gavaged donors and recipients was significantly correlated with the social avoidance behavior, as measured by social interaction time (Figure 6G; p=0.01). This result suggests that the ability to successfully transfer the gut microbiota from vehicle-gavaged NOD donors was significantly correlated with the transmission of the depressive-like behavior. LEfSe analysis revealed a number of taxa that were significantly different between Group I and Group II C57BL/6 recipients (Figure 6—figure supplement 1 and Gacias et al., 2016). We further refined this analysis, by identifying the specific OTUs transferred from vehicle-gavaged NOD donors to Group I recipients (Figure 6—figure supplement 2 and Gacias et al., 2016). Members of the Clostridiales order, including Lachnospiraceae and Ruminococcaceae, were among those present in equal proportions both in the donors and the recipients in Group I recipients displaying a depressive-like behavior (i.e. samples #18 and #19), while absent in Group I recipients that did not exhibit such behavior (i.e. sample#17; Figure 6F and Gacias et al., 2016). We further confirmed these taxa as potentially responsible for this phenotype by qPCR using primers specific to these bacterial groups (Figure 6—figure supplement 2C). In order to identify differences undetectable at the OTU level, we performed oligotype analysis in those OTUs established as potentially responsible for the depressive-like behavior (Segata et al., 2011; Eren et al., 2014). Oligotype analysis is an entropy-based method to identify single nucleotide differences in sequences from closely related organisms. We found that most OTUs were composed of a single high-abundance oligotype (Figure 6—figure supplement 3A–C,E,G–O) and therefore support the conclusions from the OTU-level analysis. However, we identified three OTUs that had two oligotypes with similar abundances and distribution across samples: OTU 183849 (Blautia producta, a member of the Lachnospiraceae, Figure 6—figure supplement 3D), 188840 (unidentified member within Lachnospiraceae, Figure 6—figure supplement 3F), and 4418586 (unidentified member within Clostridiales, Figure 6—figure supplement 3P). Additional inspection of these sequences revealed the oligotypes GTT and TTT from the Blautia producta OTU, as well as the TG and TT oligotypes from the Lachnospiraceae OTU, had B. producta JCM 1471 as the closest reference sequence in NCBI; the oligotypes from Clostridiales had no close reference sequence. Overall, these results show that either a single oligotype or a combination of two oligotypes with similar abundance distributions were dominant within the analyzed OTUs, which suggested they might drive the observed social phenotypes.

Figure 6 with 3 supplements see all
Effect of fecal transplantation on bacterial mass and biodiversity in microbiota depleted C57BL/6 recipients.

(A) Experimental timeline for donors (NOD) and transplanted recipients (C57BL/6). (B,C) Graphs represent fecal biomass (µg of gDNA relative to total fecal weight) of C57BL/6 recipients prior to transplantation (#1 before and #2 after 14 days of antibiotic treatment) and at end point after-transplantation (#3) with donor microbiota (n=3 pooled samples per time-point, each sample represents 1 sample = pooled feces from 3–5 mice. Data are mean ± S.E.M; *p<0.05, **p<0.01 based on one-way ANOVA with Bonferroni’s post hoc test). (C,E) Rarefaction curves comparing alpha diversity of fecal microbiota samples from C57BL/6 recipients at different experimental time-points (#1, #2, and #3). (F) Principal coordinate analysis plot of unweighted UniFrac distances of fecal samples from NOD donors and C57BL/6 mice at different time-points. (#1, #2, and #3). Each dot represents the microbiota of a sample, colored by group, treatment, and time-point (n=3 pooled samples per time-point; each sample corresponds to pooled feces from 3–5 mice). The percentage of variation explained by each principal coordinate (PC) is shown in parentheses. (E) Relationship between social interaction time and unweighted UniFrac distance to NOD donor mice (n=3) for all C57BL/6 recipients (n=10). Each point represents a single C57BL/6 animal, colored by group (light blue: Group_I, transplanted with NOD-vehicle microbiota; pink: Group_II, transplanted with NOD-antibiotic microbiota). Linear regression analysis indicates a significant correlation (p=0.0103) between the variables.

https://doi.org/10.7554/eLife.13442.016

The gut metabolome is altered in microbiota-transplanted C57BL/6 mice displaying altered social and despair-like behaviors.

Several studies have demonstrated that gut metabolites can impact the homeostatic host-microbiota interactions and affect behavior (Daniel et al., 2014; Hsiao et al., 2013). To determine whether altered taxa in the gut microbiota could impact the levels of metabolites, which in turn drive behavioral and transcriptional changes observed in the mPFC, we performed an unbiased metabolomic analysis of gut tissue from C57BL/6 recipients with (Group I) and without (Group II) social avoidance behavior (Figure 7). The analysis included non-targeted and targeted protocols and gas chromatography combined with time-of-flight high-resolution mass spectrometry, hydrophilic liquid chromatography coupled with high-resolution mass spectrometry and hydrophilic interaction chromatography with liquid chromatography and tandem mass-spectrometry for the study of monoamine to neurotransmitters (Tolstikov et al., 2014; Danaceau et al., 2012). After statistical corrections and normalization, we conducted Partial Least Squares-Discriminant Analysis (PLS-DA), a method that incorporates elements from principal component analysis, regression, and linear discriminant analysis, which revealed a clear separation of the overall gut metabolites between Group I and Group II (Figure 7B). A total of 382 metabolites were detected in the guts of C57BL/6 transplant recipients (Gacias et al., 2016) A first pathway impact analysis provided a visual representation of the most dramatically affected pathways between the two groups, and identified the linoleic/linolenic acid and phenylalanine/tryptophan synthetic pathways as differentially represented in the two sets of samples (Figure 7C, Table 1 and Gacias et al., 2016). Further evaluation of the metabolome using a volcano plot representing individual differences in metabolites revealed increased levels of cresol, stearamide, N-acetylasparagine, and oleamide in Group I recipients, which displayed social avoidance and despair-like behaviors (Figure 7D,E). As cresol is a highly permeable compound that was detected at high levels in the guts of Group I mice characterized by behavioral changes and impaired mPFC myelination, we treated primary cultured oligodendrocyte progenitors with increasing concentrations of cresol and tested for myelin gene expression (Figure 8). Expression of Mag, Mog, Mbp, and Cnp transcripts and the number of double-positive CNP+/OLIG2+ cells were reduced incresol-treated cultures compared to controls (Figure 8A–C). However, this effect was not due to toxicity, but rather to impaired differentiation, as indicated by the increased transcripts of immature progenitor markers (Pdgfra) and the stable OLIG2+ cell counts (Figure 8D,E).

Metabolomic analysis of gut tissue from microbiota-transplanted C57BL/6 mice.

(A) Experimental timeline. (B) 3D plot of scores between selected components generated by PLS-DA analysis comparing Group I (transplanted with microbiota from vehicle-treated NOD mice; filled circles) and Group II (transplanted with microbiota from antibiotic-treated NOD mice; open circles). (C) Metabolic pathway impact overview generated with MetaboAnalyst 3.0. Unaltered pathways have a score of 0, and the most impacted pathways have higher scores. Pathways having the least statistical significance score are uncolored, whereas pathways having a high statistical significance score are colored in red. See Gacias et al. (2016). (D) Metabolites with the greatest differential between mice with (Group I) and without (Group II) behavioral phenotype, were selected by volcano plot with a fold-change threshold of 1.5 (x axis) and t test threshold of 0.1 (y axis). Red circles represent metabolites above the threshold (Group II vs Group I); see Table 1. (E) One-way analysis of variance box and whisker plots illustrating the metabolite changes observed in Groups I and II. The y axis illustrates normalized, log transformed, and scaled peak area. Horizontal lines within the boxes represent the group means. Open circles represent excluded levels (outliers) (n=6 mice per group).

https://doi.org/10.7554/eLife.13442.020
Table 1

Summary of trends in levels of cecal metabolites in C57BL/6 transplanted mice (Group II vs Group I).

https://doi.org/10.7554/eLife.13442.021
Super PathwaySub-pathwayMetaboliteFold change (Group II vs I)p value
Amino acidPhenylalanine metabolismBenzoic Acid1.010.031786
Amino acidAlanine, aspartate and Glutamate metabolismN-acetylasparagine0.520.066953
Amino acidTryptophan metabolismXanthurenic acid7.550.072748
Amino acidUrea cycleHomocitrulline2.70.075261
Amino acidArginine and proline metabolismN-acetyl-glutamate1.680.08886
Amino acidPhenylalanine metabolismphenylpyruvate3.30.094054
CarbohydratePentose phosphate pathwaySedoheptulose-7- phosphate0.990.05242
Cofactors and vitaminsMicrobial metabolism in diverse environmentscresol0.130.019692
LipidFatty acidsHexanedioic acid42.320.0053711
LipidLong chain fatty acidLinoleic acid1.950.011101
LipidLong chain fatty amideOleamide0.460.030305
LipidLong chain fatty aciddihydroxystearic acid1.020.042845
LipidLong chain fatty amideStearamide0.090.068105
NucleotidePurine metabolismcAMP0.990.077892
  1. Data were analyzed using comprehensive global mass spectrometry-based metabolomics analysis. Additional details are provided in Experimental Procedures.

Cresol treatment decreases myelin gene expression.

(A) Transcript levels of oligodendrocyte lineage (Olig2), progenitor (Pdgfrα, Cspg4) and differentiation (Mag, Mog, Mbp, Cnp, Sox10) markers in oligodendrocyte progenitors cultured in differentiating conditions and treated with increasing concentrations of cresol (0, 10, 50 μM). DMSO was used as vehicle and negative control. Values were normalized to 36b4 mRNA levels and are referred as fold change relative to the control group (n=3 independent primary cultures). (B,C) Representative confocal images and quantification of early differentiated oligodendrocytes (CNP+/OLIG2+) after treatment with increasing concentrations of cresol (0, 10, 50 μM) for 24 hr. (D,E) Representative confocal images and quantification of oligodendrocytes (OLIG2+/DAPI+) treated with increasing concentrations of cresol (0, 10, 50 μM) for 24 hr. Scale bars, 20 μm; 10–15 fields (20×) per condition/experiment; n=2 independent primary cultures. Data are mean ± S.E.M; *p<0.05, ***p<0.001 based on one-way ANOVA with Dunnett's Multiple Comparison Test; n.s. indicates not significant

https://doi.org/10.7554/eLife.13442.022

Discussion

Our results provide strong evidence that manipulations of gut microbiota are sufficient to induce depressive-like behaviors in adult mice. The behavioral changes were detected in mice with gut microbiota enriched for the taxa Clostridiales, including the Lachnospiraceae and Ruminococcaceae families, and with increased levels of highly permeable metabolites (such as cresol) with the ability to impair oligodendrocyte differentiation and myelin gene transcription. The observation that behavioral traits were only detected in transplant recipients with effective colonization of these taxa highlights the potential molecular mechanisms by which gut microbiota impacts CNS homeostasis.

To date, several studies have focused on the relationship between microbiota composition and the development of anxiety-like behaviors. Dysbiotic microbiota induced by either pathogenic infections or antibiotic treatment has been shown to increase anxiety-like behavior in conventionally raised mice (Bercik et al., 2010; 2011; Lyte et al., 2006), while germ-free mice show reduced levels of anxiety-like behaviors compared to normal mice (Diaz Heijtz et al., 2011; Neufeld et al., 2011). In our study, anxiety-like behaviors were also shown to be affected by daily gastric manipulations affecting microbiota composition. However, social avoidance and despair-like behaviors were differentially induced by gavage in two genetically distinct strains of mice, which could be prevented by the administration of a broad-spectrum antibiotic cocktail. Daily gavage of NOD mice induced significant changes of gut bacterial communities and depressive-like behavior (social and despair-like behaviors), which was associated with enrichment of bacteria within the Clostridiales. Antibiotic treatment decreased the overall bacterial diversity and prevented the behavioral effects. Subcutaneous administration of the same antibiotic treatment failed to induce significant changes either in the microbiota composition or behavior, further highlighting the importance of a local effect of oral antibiotic treatment on these intestinal microbial communities.

To prove causality and understand whether behavioral changes observed in vehicle-gavaged NOD mice were in fact modulated by the intestinal microbiota, we transferred the cecal content of these mice (displaying a social avoidance behavior) or antibiotic-treated (with normal social behavior) NOD donors into the microbiota-depleted guts of C57BL/6 recipients. Our results demonstrate that only recipients with successful recolonization of the taxa enriched in the vehicle-treated NOD mice (e.g. Clostridiales, Lachnospiraceae and Ruminococcaceae) exhibited the social avoidance and despair-like behaviors, as well as the myelin gene expression in the mPFC of the donors. These transcriptional changes resulted in decreased adult PFC myelination in mice with transferred behavior. The microbiota of these C57BL/6 recipients showed significant differences in the abundance of several of the bacterial populations identified in the donors. Interestingly, alterations of some Lachnospiraceae and Ruminococcaceae spp. have been associated with behavioral deficits in mice (Bruce-Keller et al., 2015). Our results did not identify a single bacterium responsible for the behavioral changes induced by vehicle-gavage in the NOD mice or by transplantation in the C57Bl6 animals, suggesting that specific communities enriched in taxa from the Lachnospiraceae and Ruminococcaceae are responsible for the observed phenotype. Community-driven effects have also been reported in the induction of colonic regulatory T cells by specific mixtures of Clostridia strains in models of colitis or in cognitive and stereotypic behavioral changes induced by high-fat diet microbiota in non-obese mice (Bruce-Keller et al., 2015; Atarashi et al., 2015).

Our results also show that alterations of the microbial composition modified gut-produced metabolites and transcriptomic profiles in the mPFC, subsequently affecting behavior (Daniel et al., 2014; Hsiao et al., 2013). Microbiota composition has previously been shown to modulate anxiety-like behaviors in adult mice via changes in levels of brain-derived neurotrophic factor in the hippocampus (Bercik et al., 2011). The results of our untargeted transcriptomic analysis of the mPFC, the region responsible for the integration of external stimuli and complex behaviors (Regenold et al., 2006), identified a signature characterized by genes regulating transcription, circadian rhythm, protein phosphorylation, synapses, and myelin. Altered expression of genes related to myelin and circadian rhythm is consistent with reported white matter changes and sleep disruption in human patients with major depression (Liu et al., 2015; Landgraf et al., 2014; Lavebratt et al., 2010; Kishi et al., 2009) as well as with the reported behavioral changes detected on myelin mutant mice (Hagemeyer et al., 2012). The association of diminished myelination in mPFC with the observed social avoidance behavior is supported by recent studies describing decreased myelin gene expression and fewer myelinated fibers in the mPFC of mice after prolonged social isolation (Liu et al., 2012; 2016; Makinodan et al., 2012). Importantly, adoptive transfer of gut microbiota from NOD mice was able to recapitulate the mPFC transcriptional changes detected in recipient mice, thereby directly implicating gut microbiota as a causal factor for the induced behavioral and transcriptional changes.

One mechanism by which the gut microbiota may regulate such alterations is through the production of selective metabolites. Several recent studies have shown that a dysbiotic gut microbiota can produce neurotoxic metabolites directly impacting behavior (Hsiao et al., 2013; Persico and Napolioni, 2013; Shaw, 2010). For instance, in a mouse model for autism spectrum disorders during development, characterized by dysregulation of Lachnospiraceae, Ruminococcaceae the anxiety-like phenotype correlated with the levels of the metabolite 4-ethylphenylsulfate (4-EPS) (Hsiao et al., 2013). In our study, social avoidance behavior in adult mice was significantly associated with enrichment in Lachnospiraceae, Ruminococcaceae and Clostridiales and thedetection of high levels of cresol. This highly permeable metabolite was detected only in the gut of mice with social avoidance behavior, and was capable of preventing myelin gene expression and differentiation of oligodendrocyte progenitors into myelin-forming cells. These results suggest a potential mechanism linking CNS transcriptional changes to gut microbial homeostasis. Thereby increased intestinal production of cresol could be responsible for the behavioral changes observed in Group I transplant recipients by impacting adult myelination in the mPFC, possibly because this brain region is still capable of generating myelin after development. Several species of Clostridia have been shown to be producers of 4-EPS and cresol (Persico and Napolioni, 2013; Nicholson et al., 2012), consistent with our findings that the microbiota of transplant recipients displaying an altered behavior (social avoidance and increased despair-like behavior) was enriched with members of the Lachnospiraceae, Ruminococcaceae, and other unidentified families within the Clostridiales order. We also detected a disruption in gut biosynthesis of tryptophan, tyrosine, and phenylanine in recipient mice with behavioral changes after transplantation. This might result in changes in the systemic/CNS levels of serotonin and other neurotransmitters, as almost 90% of serotonin production occurs within the gastrointestinal tract from its precursor tryptophan (Berger et al., 2009; Yano et al., 2015). Intestinal serotonin could cross through the blood brain barrier into the brain to regulate the observed social and despair-like behaviors. Additionally, accumulating evidence suggests that alterations in the glutamatergic system impact the pathophysiology of major depressive disorders (Tokita et al., 2012). Interestingly, another metabolite that was significantly downregulated in affected transplant recipients was hexanedioic acid, also known as adipic acid, which can impact glutamate signaling by inhibiting the L-glutamate decarboxylase in the brain (Wu and Roberts, 1974). Recent work has demonstrated that epsilon toxin produced by Clostridium perfringens Type B is able to bind CNS endothelial cells and white matter tracts, inducing blood brain barrier disruption and oligodendrocyte apoptosis (Linden et al., 2015; Rumah et al., 2013; 2015). Although in our studies we could detect C. perfringens, its low abundance and the lack of demyelination at the ultrastructural level suggests that other members of the Clostridiales might be driving the behavioral outcome.

In conclusion, our data support the concept that myelinating oligodendrocytes play a pivotal role in the pathogenic process underlying social avoidance, and define the intestinal microbiota as a potential regulator of such behavioral alterations in adult mice.

Materials and methods

Animals

Seven-week-old male C57BL/6 and NOD mice were purchased from Jackson Laboratories (Bar Harbor, ME) and housed in specific pathogen-free facilities at Mount Sinai. All procedures were performed in accordance with the Institutional Animal Care and Use Committee guidelines of the Icahn School of Medicine at Mount Sinai (#08–0676, #08–0675; LA10-00398; LA12-00193; LA12-00146).

Antibiotic treatment

A cocktail consisting of vancomycin (50 mg/kg), neomycin (100 mg/kg), metronidazole (100 mg/kg), and amphotericin B (1 mg/kg) was administered daily by gastric gavage or subcutaneous injection within a volume of 200 μL and 100 μL, respectively. Control mice received dH2O (gastric gavage) or saline (s.c.) as vehicle. Ampicillin (1 g/L) was supplemented in drinking water in the antibiotic-treated group (Reikvam et al., 2011). Antibiotics were administered daily for 14 days prior to behavioral testing. During behavioral testing, antibiotics were administered every other day and always after the behavioral tests.

Behavioral tests

All behavioral tests were recorded and tracked using Ethovision 3.0 (Noldus, Netherlands) for unbiased quantification. Overall anxiety behavior was assessed using Elevated plus maze. Social and despair-like behaviors were assessed using Social interaction and Forced swim tests. To limit carryover effects, behavioral tests were assessed in the order listed below over 14 days. Locomotor activity, Open field, Elevated plus maze, and Social interaction tests were conducted during the first week of testing with 24 hr of recovery between each task, while the Forced swim test was tested the following week.

Elevated Plus Maze

Mice were placed in the center of the maze, and behavior was recorded for 5 min. Time spent in the open and closed arenas were the dependent variables recorded by video tracking software (Ethovision 3.0, Noldus).

Social interaction test

A two-stage social interaction test was performed (Krishnan et al., 2008). In the first 2.5 min trial, each mouse was allowed to freely explore a square open-field arena (44 × 44 cm) containing a wire cage (10 × 6 cm) on one side. During the second 2.5 min trial (target present), the mouse was reintroduced into this arena now containing a social target (unfamiliar mouse) within the wire cage. Time spent interacting with target or in the corner zones was recorded by video tracking software (Ethovision 3.0, Noldus).

Forced swim test

Mice were single housed for 24 hr prior to testing and then placed in individual glass cylinders (46 cm height x 18 cm diameter) containing 15 cm of room temperature water. Sessions were videotaped for 6 min and total 'immobility' time was scored blind by a second investigator (Stratinaki et al., 2013).

DNA extraction, 16S rRNA amplification, and multiplex sequencing

All mice used for 16S rRNA sequencing were co-housed per group (3–5 mice per cage) in specific pathogen-free conditions. Fecal pellets were collected directly into sterile 1.5 mL tubes and immediately frozen and stored at -80°C. Cecal content was harvested at the end of each experiment and immediately frozen and stored at -80°C until further analysis. Fecal pellets from co-housed mice were weighted and pooled, and gDNA isolated using Powersoil DNA Isolation kit (Mo Bio Laboratories, Inc., Carlsbad, CA). DNeasy blood and tissue kit (Qiagen, Venlo, Netherlands) was used to isolate gDNA from gut tissue (cecum). For gut microbiome characterization, the V4 hypervariable region of the bacteria 16S rRNA gene was amplified using the universal primers F515 (50-CACGGTCGKCGGCGCCATT-30) and R806 (50-GGACTACHVGGGTWTCTAAT-30). A 12 bp GOLAY error-correcting barcode was added to the reverse primer to enable sample multiplexing. Reactions were performed in triplicate using the AccuPrime Taq DNA Polymerase High Fidelity system (Thermo Fisher Scientific, Waltham, MA). Unless noted, all the analyses were performed using QIIME 1.8.0 as previously described (Clemente et al., 2015). Linear discriminant analysis effect size was performed using default parameters (Segata et al., 2011). Raw data presented in Gacias et al. (2016) (doi:10.5061/dryad.31v06).

Oligotyping analysis for tissue samples

Oligotype analysis was performed on OTUs belonging to the Lachnospiraceae, Anaeroplasmataceae, and Ruminococcaceae families, and the Clostridiales order. Singletons and OTUs of low prevalence (<80% of the samples) were removed, and the sequences from the five most abundant OTUs were picked for further analysis. Entropy analysis was performed on this set of sequences to look for highly variable positions within all sequences in each OTU, and the number of oligotypes was chosen based on the entropy peaks generated (Eren et al., 2014).

Bacterial qPCR

qPCR analysis of genomic DNA extracted from tissue of C57BL/6 mice transplanted with microbiota (from vehicle- and antibiotic-treated NOD mice) was performed to quantify the total bacteria, the order of Clostridiales, and the families of Lachnospiraceae, and Ruminococcaceae in both set of animals. The primer sequences and their features are reported in Gacias et al. (2016). The reaction mixture contained 1× PerfeCTa SYBR Green FastMix, ROX (#101414–278; Quanta Biosciences, Inc., Gaithersburg, MD), 200 nM each primer, 1 μL gDNA in a total volume of 12.5 μL. Each SYBR Green PCR assay was performed in triplicate using the ABI 7900HT Real-Time PCR System (Applied Biosystems of Thermo Fisher Scientific), with the following cycling program: 5 min at 95°C, 30 s at 95°C, 45 s at 55°C/60°C, and 45 s at 72°C for 40 cycles. PCR results were analyzed using RQ Manager softwere 1.2.2 (Applied Biosystems). The annealing temperature was 55°C for all set of primers apart from Ruminococcaceae (60°C). The genome of Blautia producta ATCC 27,340 and Ruminococcus bromii ATCC 27,255 were used as reference genomes to construct the standard curves and to calculate the unknown numbers of bacterial gDNA copies in both set of animals as described previous in Tamburini et al. (Tamburini et al., 2013).

RNA isolation and qPCR

Tissue punches were taken from the mPFC or NAc and flash frozen for subsequent processing. RNA was extracted using Trizol (#15596–018; Invitrogen of Thermo Fisher Scientific) and purified with the RNeasy Micro kit (#74004; Qiagen) following the manufacturer’s protocol. RNA was reverse transcribed with qScript cDNA Supermix (#95048; Quanta Biosystems, Inc.) and qPCR was performed using Perfecta Sybr Fast Mix Rox 1250 (Quanta Biosystems, Inc.) at the Mount Sinai Shared Resource Facility (primers listed in Gacias et al., 2016). Each transcript value was calculated as the average of triplicate samples from several mice per experimental condition (typically 6–12). After normalization to 36b4, the average value for each transcript was calculated based on the values obtained in all the samples included for each experimental condition.

RNA Sequencing

RNA from the mPFC was flash frozen for subsequent processing. RNA was extracted using Trizol (Invitrogen), purified with RNeasy Micro kit (Qiagen). RNA was then used for deep sequencing analysis (RNA Seq). Samples were mapped at a rate of 79–80%. After filtering out adaptor and low-quality reads, reads were mapped using TopHat (version 2.0.8) to the mm10 mouse genome (Trapnell et al., 2009). The Cufflinks/Cuffdiff suite was used to estimate gene-level expression values as fragments per kilobase of exon model per million mapped fragments and detect differentially expressed genes at a FDR <10% and subjected to Gene Ontology enrichment.

Mouse primary oligodendrocyte cultures

Primary oligodendrocytes were prepared by sequential immunopanning and kept in undifferentiating conditions as described earlier (Watkins et al., 2008) until the onset of experiments. Briefly, oligodendrocyte progenitor cells (OPCs) were isolated from one P6 mouse pup brain using an immunopanning system enabling a purity of 95%. The dissected cortex was chopped in papain buffer, incubated for 20 min at 37°C and titrated in ovomucoid solution (CellSystems GmbH, Troisdorf, Germany). The single cell solution was centrifuged at 1000 rpm for 10 min and resuspended in panning buffer and transferred to a bacterial culture plate coated with Anti-BSL1 Griffonia simplificonia lectin (L-1100; Vector Labs, Inc., Burlingame, CA), for negative selection for 15 min, followed by a positive selection step with rat anti-mouse CD140a (10R-CD140AMS; Research Diagnostics, Inc., Flanders, NJ) as primary antibody and AffiniPure goat anti-rat IgG (H+L) (112-005-003; Dianova) as the secondary antibody for 45 min. The supernatant was aspirated, and the positive selection plate was washed with DPBS. The adherent OPCs were removed using trypsin, centrifuged for 10 min at 1000 rpm, resuspended in mouse OPC Sato medium (Watkins et al., 2008) and plated in a p100 culture plates coated with poly-d-lysine (P7886; Sigma-Aldrich, St. Louis, MO). The OPCs were cultured in a humidified incubator at 5% CO2 and 37°C with media changes every 2 d. OPCs were maintained proliferating in the presence of bFGF (20 ng/mL) and PDGF (10 ng/mL), while oligodendrocyte differentiation was induced by culturing the cells in the absence of mitogens and adding 60 nM T3 (T5516; Sigma-Aldrich) to Sato medium.

Cresol treatment

Stock solutions of Cresol (C85751; Sigma-Aldrich) were prepared in DMSO (1000-fold concentrated) and then diluted in differentiation media (SATO+T3) to give final concentrations of 10 μM and 50 μM of Cresol. Primary oligodendrocytes were plated on 0.1 mg/mL poly-d-lysine coated 6-well plates in proliferating conditions (SATO + bFGF and PDGF). Twenty-four hours after plating, cell differentiation was induced by changing the medium to SATO+T3. At this point cells were treated for 24 hr with Cresol at 10 μM or 50 μM as well as DMSO as a control. Cells were gently washed with PBS after completion of the treatment and fixed with 4% paraformaldehyde for 15 min at room temperature for immunocytochemistry experiments.

Immunohistochemistry and immunocytochemistry

Experimental animals were anesthetized and then perfused with 4% (w/v) paraformaldehyde in 0.1 M phosphate buffer. Whole brains were cryopreserved in 30% (w/v) sucrose, embedded in OCT and sectioned (14 μm). Permeabilization in blocking buffer (PGBA, 10% [v/v] normal goat serum [Vector Laboratories] and 0.5% [v/v] Triton X-100) followed by overnight incubation with primary antibody anti-MBP (clone SMI99, 1:500; BioLegend, San Diego, CA) at 4°C. After incubation with secondary fluorescent antibodies (Donkey anti-mouse Alexa Fluor 594) and nuclear counterstaining with DAPI (1:10,000; Molecular Probes of Thermo Fisher Scientific), immunoreactivity was visualized using LSM780 Meta confocal laser scanning microscope (Carl Zeiss Micro-Imaging, Jena, Germany).

Immunohistochemistry of cultured cells with CNP and OLIG2 antibodies was performed on fixed cells. Cells were grown on CC2-coated 8 well chambers (Lab-Tek, Scotts Valley, CA) for all immunocytochemistry. For staining oligodendrocyte lineage (OLIG2) and differentiation markers (CNP), cells were rinsed gently with PBS and were then fixed with 4% PFA for 15 min at room temperature. Fixed cells were first incubated with blocking/permeabilization solution (PGBA plus 10% normal goat serum, and 0.5% Triton X-100) for 1 hr at room temperature. For co-staining experiments, cells were incubated with additional primary antibodies against OLIG2 (AB9610, 1:1000; Millipore, Darmstadt, Germany) and CNP (SMI91R, 1:500; Covance, Princeton, NJ) overnight at 4°C. One-hour incubation with secondary fluorescent antibodies (Alexa Fluor 594) was performed the following day with counterstaining for DAPI (1:10000) to visualize cell nuclei.

Image acquisition and quantification

Images were captured with a 20× objective using an LSM 780 Metaconfocal laser scanning microscope (Carl Zeiss MicroImaging, Inc., Jena, Germany). For OLIG2 and CNP cell counts, 10–15 fields were taken per condition. For MBP area quantification, four fields were taken per mouse. Three to four mice were included per treatment condition. MBP+ area and OLIG2+/CNP+ cell counts were quantified using ImageJ (Liu et al., 2016; Rusielewicz et al., 2014). An unpaired Student’s t test or one-way ANOVA was performed to assess statistical differences between conditions as indicated in figure legends.

Electron microscopy

mPFC and NAc samples were prepared from standard electron microscopic analysis as previously described (Liu et al., 2012; 2016). Briefly, mice were transcardially perfused with 0.1 M Millonigs buffer containing 4% paraformaldehyde and 5% glutaraldehyde and post-fixed for 2 wk. Brains were harvested and the region spanning from bregma to 2.5 mm anterior to bregma was vibratome sectioned at 40 μm. Comparable sections ~1.5 mm anterior to bregma and at the level of the forceps (Liu et al., 2012) minor of the corpus callosum were selected and embedded in PolyBed resin (Polysciences), thick sectioned (1 μm) and stained with toluidine blue. Using these sections, the mPFC and the core of the NAc were identified, and both regions were thin sectioned (90 nm) and stained with uranyl acetate and lead citrate. For quantitation of myelin thickness, 10 electron micrographs were collected at 10,000× per region using a JEOL JEM 1230EX transmission electron microscope equipped with a Gatan Orius SC1000 side mount CCD camera. Using NIH Image J, the g-ratio of a minimum of 100 myelinated axons per region was calculated using the collected electron micrographs.

Blood glucose measurements

Blood samples were collected by tail snip, and blood glucose was measured using glucose strips (7080G; Bayer Contour).

Fecal transplantation protocol

At the time of transplantation, microbiota was freshly harvested from the cecum of 8–9-wk-old NOD mice treated with either vehicle or antibiotic. Cecal content was harvested, pooled, homogenized in a 1:4 in sterile solution (1x PBS: 80% glycerol, ratio 1:1), centrifuged at 800 rpm and the supernatant was collected, aliquoted, and stored at -80°C. Recipient 8-wk-old C57BL/6 mice received an oral cocktail of antibiotics (describe above in this section) once daily for 14 consecutive days prior to the transplantation. Recipients were then randomized in two groups (Group I and II), tested for social behavior, and then immediately started on the re-colonization protocol. To re-colonize the gut of C57BL/6 mice, recipient mice were orally gavaged every other day with 200 μL of cecal content isolated from the vehicle-treated or antibiotic-treated NOD mice over the subsequent 14 d (for a total of 7 times). Behavioral testing was repeated after 15 d of first transplantation, and group-blinded analysis of the results was performed. Cecal and mPFC samples were harvested at the end of the behavioral testing (14 d post-transplantation) and immediately stored at -80°C for further processing and analysis.

Tissue preparation and metabolomic analysis

Frozen tissues (30 mg) were placed in pre-chilled (-80°C) 2 mL round bottom Eppendorf tubes having a stainless steel ball in it. Next, 400 mL of a pre-chilled (-20°C) mixture of acetonitrile, isopropanol, and deionized water in proportion 3:3:2 (v/v/v) was added. Samples were homogenized using Tissue Lyser (Qiagen) at 25 Hz speed for 5 min. Samples were further centrifuged at 4°C at 12,000 rpm for 3 min. Clean supernatant was transferred into vials or 0.5 mL Eppendorf tubes (to be dried for gas chromatography combined with time-of-flight high-resolution mass spectrometry). Tissue extracts were divided in to three parts: 75 μL for gas chromatography combined with time-of-flight high-resolution mass spectrometry, 150 μL for hydrophilic liquid chromatography coupled with high-resolution mass spectrometry, and 150 μL for hydrophilic interaction chromatography with liquid chromatography and tandem mass-spectrometry. Metabolomic analyses were performed using non-targeted and targeted protocols as previously described (Tolstikov et al., 2014; Urayama et al., 2010; Zou and Tolstikov, 2008). A standard quality control sample containing a mixture of amino and organic acids was injected daily to monitor mass spectrometer response. A pooled quality control sample was obtained by taking an aliquot of the same volume of all samples from the study and injected daily with a batch of analyzed samples and to determine the optimal dilution of the batch samples and to validate metabolite identification and peak integration.

Metabolite pathway analysis

Identified metabolites were subjected to pathway analysis with MetaboAnalyst 3.0, which consists of an enrichment analysis relying on measured levels of metabolites and pathway topology, and provides visualization of the identified metabolic pathways. Accession numbers of detected metabolites (HMDB, PubChem, and KEGG Identifiers) were generated, manually inspected, and utilized to map the canonical pathways.

Data processing and statistical analysis

Behavioral and biochemical data were analyzed by unpaired, two-tailed Student’s t tests or one-way ANOVA followed by Bonferroni post hoc test, as appropriatem using Prism software (GraphPad Software, Inc., La Jolla, CA). Microbiome data were analyzed using QIIME 1.8.0 with default parameters. Statistical significance was assessed using R 3.0.2. Statistical significance for all analyses was accepted at p<0.05. Metabolomic data was analyzed as previously described in Tolstikov et al. (Tolstikov et al., 2014).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
    [Higher prevalence of depression in patients with rheumatoid arthritis--a systematic review]
    1. LT Moll
    2. L Gormsen
    3. M Pfeiffer-Jensen
    (2011)
    Ugeskrift for Laeger 173:2564–2568.
  44. 44
  45. 45
  46. 46
    Pathogenesis of depression: Insights from human and rodent studies
    1. C Ménard
    2. GE Hodes
    3. SJ Russo
    (2016)
    Neuroscience, 321, 10.1016/j.neuroscience.2015.05.053.
  47. 47
  48. 48
  49. 49
    [Omega-3 fatty acids in mood disorders]
    1. K Ohara
    (2005)
    Seishin Shinkeigaku Zasshi 107:118–126.
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78

Decision letter

  1. Peggy Mason
    Reviewing Editor; University of Chicago, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your work entitled "Microbiota-driven transcriptional changes in prefrontal cortex override genetic differences in social behavior" for consideration by eLife. Your article has been reviewed by five peer reviewers, one of whom, Peggy Mason, is a member of our Board of Reviewing Editors and a Senior Editor.

The reviewers have discussed this paper extensively with each another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The following individuals involved in review of your submission have agreed to reveal their identity; Jack Gilbert (peer reviewer).

Summary:

There was strong enthusiasm for this work which attempts to tie changes in the gut microbiota to changes in behavior and also changes in myelin-related genes, studied in the prefrontal cortex.

Essential revisions:

1) Ab-gg changes the biota of both strains but only changes NOD's behavior. The authors claim that these data support the idea that a change in gut biota alters behavior only in NODs, so in a strain specific way. An equally plausible interpretation is that a change in gut biota has nothing to do with the change in behavior. So the data are consistent with multiple interpretations.

2) "Throughout the manuscript naming the direction of changes and defining what is normal"/baseline is actively avoided and this causes major confusion. Only differences between treatment groups are reported. This leaves the data floating somewhere in space, completely un-calibrated. Control groups of unmanipulated -in any way – NODs and C57s are needed.

3) The FMT transplant to the ABx-treated C57 rats is reported in such a backhanded way as to completely confuse the reviewers who spent a great deal of time trying to decipher it. This experiment also suffers from a lack of a control – ABx-C57 untreated with gavage (the baseline for this expt).

4) Changes in myelin genes are not changes in myelin. To make the claims that the authors make, some measure of myelination is needed and some degree of specificity to the frontal cortex. And it is not clear why myelin rather than clock genes are highlighted.

Overall there is enthusiasm for the experiments and data and skepticism that the conclusions are as solid as they are presented. Acknowledgement of alternate possibilities is absolutely needed along with some softening of the conclusions.

Additional points:

Reviewer #1:

Why use NOD mice? It think you should be clearer about this when you first mention it at the beginning of the results.

The statement that "ruling out the possibility that the observed behavioralalterations in NOD mice were consequent to a systemic effect of antibiotics." Is not necessarily true right? Maybe I am wrong, but what is gavaged antibiotics interacted with the host cells in the gut, or the immune regulation – to control for this you would need to run germ free models.

It would be nice to run oligotyping on the 16S rRNA analysis of the key Lachnospiraceae, Ruminococcaceae, and Clostridiales strains anyway to explore if there were any strain specific effects that might contribute to the variance in behavioral response. Also comparison of the strain specificity between donor and recipient for the FMT, would significantly advance the case that these were transferred strains, rather than existing strains whose presence was augmented by the FMT community.

Reviewer #2:

While the NOD-derived microbiotal were associated with a number of prokaryotic genotypes and an altered metabolic profile of the gut (e.g. by mass spectroscopy and gas chromotography), no single "responsible" microbe or neurotoxic metabolite was identified. Here, the authors speculate that it may be a "community" effect, but they later discuss also specific metabolite changes (such as 4-EPS/cresol, altered tryptophan/serotonin levels, or hexanedioic acid/ altered glutamate signaling) as possible underlying causes.

While this is not the first study linking gut microbiota in mice with anxiety-like traits (the authors cite five papers and a quick search in Google Scholar adds many more papers), it is the first study to point out secondary changes of myelin gene transcripts in the CNS of antibiotica-treated mice. The same group has previously shown (Nature Neuroscience 2012) that myelination of the cortex can change as a result of social withdrawal, presumably reflecting altered neuronal activity.

I find this work very interesting and conceptually novel in the field. My only concern is that the altered expression of genes implicated in adult myelination should be more carefully discussed. We do not know whether the altered gut microbiom (or any of the metabolites derived) is "directly" causal in changing myelin gene expression of oligodendrocytes, or only "indirectly", i.e. by affecting behaviour first. Indeed, the latter had been shown by the authors' lab to be an important aspect of myelination control.

A decisive experiment would be to compare the RNA profile of oligodendrocytes in prefrontal cortex and white matter tracts, which are less likely to be affected by behavioural changes (e.g. spinal cord).

The authors should also show their immunostainings of myelinated fibers after the transplant. Does it match the strain difference observed before? A nearly 2-fold difference would be visible on Western blots.

The authors should avoid writing "it is not possible to draw a direct link between gut microbial metabolism and mPFC transcriptome". The necessary experiments have just not been carried out.

Reviewer #3:

The sc and gg vehicles produce different baseline behaviors. Is this an effect of the relative stress of the two procedures?

My one big concern is the enormous gastric damage done by the abs. From the first expt it is not clear that the causative agent is the change in microbiota or the stomach inflammation. The one expt that could be used to point to microbiota is the final one where ab-treated C57 mice get either veh or ab-treated NOD mice microbiota. But to get this we need to see the stomachs of the C57 mice in each group. While of course, fashion would have us believe that it is the microbiota, it is formally possible that the cause of the changes is metabolites from the very extreme inflammation. And that the reactions differ in NOD and C57 mice (just as the microbiota differ in these two strains).

Reviewer #4:

The premise of the paper is fascinating and the authors have done an admirable job in synthesizing a connection between gut microbiome and cortical myelin related genes. However, their hypothesis comes undone when making a direct link to behavior.

• I grappled hard and failed to follow the author's logic of the fact that antibiotic administration significantly decreased bacterial diversity of both NOD and C57BL/6 mice and yet the behavioral alterations occurred only in the NOD mice leads to the conclusion that "These findings suggest that NOD treated with vehicle might have a particular microbiota community, sensitive to antibiotic treatment, which could modulate the social and despair-like behaviors". Wouldn't the more parsimonious explanation that the behavior and microbiome be unrelated?

• Indeed the authors prove the opposite of their hypothesis is true in the next few sentences. It looks as rather that C57 mice have some strains enriched that appear to be preventive of enhanced social behavior and mood through antibiotics compared to NOD.

• The authors hypothesis also falls apart at the fecal transplantation stage – when they transplant fecal microbiota from NOD mice, which have decreased despair & increased social activity, it induces the opposite phenotype in antibiotic depleted C57 mice. This is fascinating data but difficult to explain easily. The authors somewhat try to distract the reader from this fact by now calling these groups Group 1 and Group 2.

• Overall, the authors take a rather unexpected point-of-view to interpret the data, which is highlighted by the following statement (among others throughout the manuscript): "Collectively, these findings suggest that the gut microbiota of NOD mice is sufficient to induce a reduction of the social behavior and increase the despair-like behavior independent of the genetic background and mediated through changes in the transcriptomic profile of the mPFC."

In fact, the microbiota of the NOD mice is the baseline and hence cannot be viewed as to "induce" a phenotype. The alternative interpretation, that immediately emerges from the presented data would be: “Collectively, these findings suggest that the ANTIBIOTICALLY-DEPLETED(/MODIFIED) gut microbiota of NOD mice is sufficient to induce an INCREASE of the social behavior and DECREASE in the despair-like behavior independent of the genetic background and mediated through changes in the transcriptomic profile of the mPFC."

And this finding needs to be discussed and explained by the data, i.e. the differences in microbiota and metabolome before and after antibiotic treatment and/or transplantation.

• In their analysis of microbiota in NOD vs. C57Bl6J (under control conditions) it should be made clear, which bacteria are more abundant in NOD mice because it would be important to know which bacteria are depleted in NOD mice that lead to the behavioral performance increase. And this finding needs to be discussed and explained by the data, i.e. the differences in microbiota and metabolome before and after antibiotic treatment and/or transplantation.

• It may be helpful to state what would be expected and then look for deviations from the expectation: Microbe-depleted C57 mice are expected to show a similar microbial profile to the respective donor, i.e. NOD vehicle-treated or NOD antibiotic treated. The authors should provide and discuss evidence that is or isn't the case. Unexpected interpretation happens from this point onwards. Interpretation of this data very much depends on what the baseline is, what is viewed as "normal" and what as "altered".

• It would benefit readability of the manuscript a lot if the author would describe the direction of the change they mean rather than say that behavior was "altered" in many sentences. The same is true for "differences" in the microbiota composition. The use of such terms without mentioning the direction of the change make the findings sound very vague and make it more difficult to interpret the data

• How are the mouse strains different genetically? How could this affect the microbiota present in each strain and subsequently the reaction to the antibiotic? This study needs to take gene-environment interactions into account.

• Why use NOD mice at all, it is a strange choice for behavioral genetic studies; does their diabetes interfere with the interpretation of phenotype?

• The authors’ assertion that social & despair-related behavior is "mediated through changes in the transcriptomic profile of the mPFC" is not proven at all. Correlation dos not mean causation.

• Regarding the myelination data itself the authors show an increase in mRNA and increased intensity of myelin staining. Is intensity staining the best method, does this mean there is more myelination, as in more axons, or are the fibers that are myelinated more heavily myelinated. Looking at structure using electron microscopy would provide an answer. What is the alteration at the level of oligodendrocytes? Have any of the myelin related changes any functional relevance to demyelination-related disorders or to functional myelin responses.

• The Introduction is lacking a coherent flow and would benefit from significant editing. I found the first paragraph of Introduction, whilst interesting, overlong and not really pertinent to study at hand.

• The relevance of current study to schizophrenia & bipolar is not clear and the authors are perhaps conflating this relevance in Introduction. This is further expanded in the Discussion that compares the data with that from an autism model. If this is focused on social behavior, as title but not introduction indicate, then it’s easy to bring all of this together, but currently it is a number of 'characters' in search of a coherent story.

• An example of the lack of flow is going from the statement "Our laboratory has recently demonstrated that mouse models of depression were characterized by down-regulation of oligodendroglial transcripts and myelin thickness in the medial prefrontal cortex (mPFC), thereby supporting a role for myelin in the pathogenesis of depressive disorders” to "Indeed, manipulations of intestinal microbiota have recently emerged as a novel approach to influence brain function and behavioral outcomes" – this is a strange transition.

• The authors state that "Antibiotic regimen was well tolerated" other than bodyweight, which is a relatively crude measure, was there any other change in secretomotor function, defecation rates etc.

• Motor activity needs to be shown at individual time-bins as the overall effect seems to be different in NOD mice if not reaching statistical significant when collapsed overall.

• Is the myelination effects specific to the cortex; Other brain areas such as the amygdala and hippocampus would be equally worthy of investigation and at least addressed.

• The myelination data is intriguing but it’s somewhat strange that the authors chose to follow up on the least-significantly enriched functional category (myelination) in their transcriptome analysis; further scientific justification is needed.

• Were antibiotic treatments continued during the 1.5 weeks behavioral testing?

• y-axis in Figure 4g unclear – what is MBP+ fibers mean pixel intensity? How is this measured?

Figure 4: it would be worthwile to graphically show the NOD vehicle treated mice for gene expression.

• Results section needs to be cleared up to enhance readability.

• When describing Figure 5d, the authors state "characterized by lower levels of myelin gene transcripts […] in Group I and higher in Group II", which is confusing as this was not compared to baseline in C57Bl6. They should state that transcription of these genes was increased in Group II compared to Group I as shown in the figure.

• The color legend in Figure 6c needs to show the Group names (Group I or Group II), to avoid confusion about the interpretation of the data. Also the description of this figure panel states, "significantly distinct clustering of C57BL/6 transplanted mice displaying social avoidance and an increase in despair-like behaviors compared to normal behaving controls" although, according to the data presented in figure 1, the "normal" behaving animals show more social behavior and less despair-like behavior.

• A list of the differentially expressed genes should be added to supplemental material, together with a list of terms from functional enrichment analysis.

• The Discussion focuses mainly on the metabolomics data, but does not help explaining how the observed phenotypes are connected to a depleted microbiota. Here the authors show that it is the bacteria that remain after antibiotics (and are therefore overtaking/-growing) are necessary for the myelin increase – that's what the fecal transfer shows. if it was the bacteria that are gone who are important for the myelin changes than you would also see the effects in antibiotic C57Bl6 mice. This however, very much depends on the genetic background of the two strains and it's interaction with the microbiome – which is not discussed in this manuscript at all.

Figure 7d in the last sentence of the Results section should be 7e and this sentence is again misleading, or at least again a rather surprising conclusion similar to the above.

• Not clear why the outcome of the increased myelin gene expression in transplantation of NOD microbiota to C57BL/6 is not shown at the protein level.

• Why are they using two different reference genes in the qPCR data?

• Images don't elude to which part of the medial PFC they imaged. It should be easy enough with the staining to highlight that. What layer of the PFC has more myelin?

Reviewer #5:

1) The notion that the antibiotic treatments alter myelinating cells in the PFC is based on gene expression analysis RT-PCR and histological data showing changes on MBP+ fibers. While RNA levels provides strong evidence for changes in transcription or RNA stability in oligodendrocytes, to conclude that the antibiotic treatment affects myelin, immunostaining for MBP+ fibers is not sufficient. Morphological analysis like the ones the Casaccia group has done in prior studies (e.g. electron microscopy of myelin thickness, quantification of oligodendrocyte and oligodendrocyte progenitor cell density) will be necessary.

2) It would be very important to determine if the changes in myelin gene expression are unique to the PFC or this phenomenon is widespread throughout the brain. The authors should evaluate the expression of the myelin genes affected by antibiotic treatment in a brain region not linked to social/emotional regulation. The interpretation of the findings would be very different if this is a specific or a general phenomenon.

3) There are some confusing aspects regarding the effects of treatments on behavior. For example, in the immobility tests in Figure 1B, it seems that the only group with altered behavior is the oral-vehicle group. The behavior of mice treated with oral antibiotics is the same as the one with SI or SI-vehicle (and similar to all C57BL/6). Furthermore, in the measurements of social interactions, it seems that the oral vehicle reduces and the oral-antibiotic increases the time in the interaction zone compared to both groups treated by SI. Difference in behaviors between mice treated by SI vs oral, such as those in the elevated plus maze (Figure 2B and C) leaves me wondering if in some cases lack or presence of effects of antibiotics could be due to a particular susceptibility of a mouse strain to a manipulation such as gavage or injections, and not really to a difference in the changes in microbiota. This issue would be clarified by presenting data on control mice (not treated at all).

4) Did the authors monitor for onset of diabetes in NOD mice (i.e by detecting evidence of glycosuria)? This strain spontaneously develops diabetes in approximately 20-30% of males and the incidence of disease has been linked to changes in the microbiome (See Wen et al. Nature 2008; PMID 18806780). It would be important to know if the behavioral changes with antibiotic treatment in the NOD mice coincide with metabolic disease to determine if there is an interaction among these variables.

5) The authors speculate that neurotoxic metabolites produced by a combination of gut microbia may affect PFC function. This interpretation would be strengthened if these metabolites were measured and detected in the CNS in vehicle and antibiotic treatment mice.

https://doi.org/10.7554/eLife.13442.025

Author response

Essential revisions:

1) Ab-gg changes the biota of both strains but only changes NOD's behavior. The authors claim that these data support the idea that a change in gut biota alters behavior only in NODs, so in a strain specific way. An equally plausible interpretation is that a change in gut biota has nothing to do with the change in behavior. So the data are consistent with multiple interpretations.

2) "Throughout the manuscript naming the direction of changes and defining what is normal"/baseline is actively avoided and this causes major confusion. Only differences between treatment groups are reported. This leaves the data floating somewhere in space, completely un-calibrated. Control groups of unmanipulated -in any way – NODs and C57s are needed.

3) The FMT transplant to the ABx-treated C57 rats is reported in such a backhanded way as to completely confuse the reviewers who spent a great deal of time trying to decipher it. This experiment also suffers from a lack of a control – ABx-C57 untreated with gavage (the baseline for this expt).

We believe that these comments are legitimate, given the fact that we had not clearly explained that we had referred all our analysis to baseline (without explicitly showing the data). We have now modified all the figures to include those data and revised the text accordingly. In order to summarize the overall experimental results, we would like to summarize the results.

We present evidence that untreated NOD (NOD baseline = no phenotype) had no depressive-like phenotype, while NOD mice gavaged with vehicle vehicle showed the behavioral phenotype (NOD vehicle = phenotype), which was not observed when the animals were gavaged with antibiotics (NOD antibiotic gavage = no phenotype). This suggested that the behavioral effect of gavage requires the microbiota of the NOD mice, since it was not observed when it was depleted by the chronic antibiotic treatment. The behavioural effect of daily gavaging observed in NOD was not detected in C57Bl mice, which showed no depressive-like phenotype either untreated or gavaged with vehicle or antibiotic. This indicated that the behavioral effects induced by daily gavaging were dependent both on the specific mouse strain and its microbiota.

We then asked the question of whether transplanting the microbiota from one strain responsive to gavaging (i.e. NOD) to a non-responsive one (i.e. C57) would be sufficient to transfer the behavioural response to daily gavaging. In order to do so, we depleted (using an antibiotic regimen) the microbiota of unmanipulated baseline C57Bl mice and then transplanted them with the microbiota from NOD mice with a depressive-like phenotype (which had been previously induced by vehicle-gavaging). This transplantation was sufficient to induce in the C57Bl recipients the same social avoidance behavior detected in the vehiclegavaged NOD donors. This phenomenon was not observed in C57Bl recipients that received the microbiota of antibiotic-gavaged NOD donors. Overall, these results suggest that the microbiota of vehicle-gavaged NOD mice was sufficient to elicit a depressive-like phenotype when transferred to a different strain of mice with a normal baseline behavior and antibiotically depleted microbiota.

Finally, we sought to identify the bacterial taxa that were associated with this behavioral phenotype. We therefore performed two types of analyses: the first one (presented in the new Figure 3 and associated table of OTUs) focused on differences between vehicle-gavaged NOD (with behavioral phenotype) and untreated NOD baseline mice. This analysis demonstrated the presence of specific taxa in vehicle-gavaged NOD (displaying a phenotype), that were not found at baseline (= no phenotype), and we identify as potential candidates to explain the depressive-like phenotype in vehicle-gavaged NOD mice. In addition, we reasoned that if this behavior was transferable through the microbiota, then the same relevant bacteria present in the NOD donors with the behavioral phenotype should be detected in C57Bl recipients that showed the social avoidance behavior after transplantation. For this reason, we focused on the similarities between the microbiota of vehicle-gavaged NOD donors (which exhibited a depressive phenotype) and C57Bl recipients transplanted with the microbiota of such NOD donors (which showed the depressive-like symptoms after transplantation, but not before). This analysis is presented in the new Figure 6, which identifies members of the Clostridiales, Ruminococcacea and Lachnospiraceae as the main bacterial taxa related to the depressive-like behavior transferred from vehicle-gavaged NOD donors to C57bl recipients. We further validated the enrichment of these taxa in the samples by RT-PCR, confirming the relation of these bacteria with the phenotype.

4) Changes in myelin genes are not changes in myelin. To make the claims that the authors make, some measure of myelination is needed and some degree of specificity to the frontal cortex. And it is not clear why myelin rather than clock genes are highlighted.

The reviewer is correct in his/her determination. In the original submission we had included not only transcript levels but also immunohistochemical data. In response to the comment related to regional specificity, we now show that the transcriptional and ultrastructural changes occur in the PFC and not in other related brain regions. The reason we focused on myelin is because our group (Liu et al., Nature Neuroscience 2012; Liu et al., J, Neurosci., 2016) and others (Matinaken, Science, 2012) had previously reported the importance of PFC myelination in mice with depressive-like symptoms.

Reviewer #1:

Why use NOD mice? It think you should be clearer about this when you first mention it at the beginning of the results.

Gene-environment interactions are known to affect the development of neuropsychiatric disorders, however the relative impact of these two variables and the potential underlying mechanisms of pathogenesis remain elusive. We simply wanted to use two genetically distinct strains with clear behavioral differences to evaluate the gut microbiota as environmental variable modulating behavior.

The statement that "ruling out the possibility that the observed behavioralalterations in NOD mice were consequent to a systemic effect of antibiotics." Is not necessarily true right? Maybe I am wrong, but what is gavaged antibiotics interacted with the host cells in the gut, or the immune regulation – to control for this you would need to run germ free models.

We have rewritten the Discussion to better reflect our results and we have moderated our conclusions.

It would be nice to run oligotyping on the 16S rRNA analysis of the key Lachnospiraceae, Ruminococcaceae, and Clostridiales strains anyway to explore if there were any strain specific effects that might contribute to the variance in behavioral response. Also comparison of the strain specificity between donor and recipient for the FMT, would significantly advance the case that these were transferred strains, rather than existing strains whose presence was augmented by the FMT community.

We thank the reviewer for this valuable suggestion. It is indeed possible that some of the differences were attributable to effects at the strain level rather than at the OTU level, and we performed this analysis on all OTUs found to be transferred from the vehicle-gavaged NOD mice to the C57Bl that were not found in the recipients before transplantation (see Figure 6A for a description of this analysis). Results of the oligotyping analysis are presented in Figure 6—figure supplement 3. Most of the OTUs associated with the behavioural phenotype were composed of a single oligotype at high abundance (Figure 6—figure supplement 3, panels A to C and G to O). Three OTUs had two oligotypes with similar distribution of abundance across samples: Blautia producta, an unidentified member of the Lachnospiraceae, and an unidentified member of the Clostridiales. Further inspection of the sequences associated with these oligotypes revealed Blautia producta JCM 1471 as the closest reference sequence in NCBI, while the Clostridiales oligotypes had no close reference sequence. Overall, these results suggest that either a single oligotype or a combination of two oligotypes are dominant within the analyzed OTUs and might drive the observed depressive-like phenotypes.

Reviewer #2

While the NOD-derived microbiotal were associated with a number of prokaryotic genotypes and an altered metabolic profile of the gut (e.g. by mass spectroscopy and gas chromotography), no single "responsible" microbe or neurotoxic metabolite was identified. Here, the authors speculate that it may be a "community" effect, but they later discuss also specific metabolite changes (such as 4-EPS/cresol, altered tryptophan/serotonin levels, or hexanedioic acid/ altered glutamate signaling) as possible underlying causes.

While this is not the first study linking gut microbiota in mice with anxiety-like traits (the authors cite five papers and a quick search in Google Scholar adds many more papers), it is the first study to point out secondary changes of myelin gene transcripts in the CNS of antibiotica-treated mice. The same group has previously shown (Nature Neuroscience 2012) that myelination of the cortex can change as a result of social withdrawal, presumably reflecting altered neuronal activity.

I find this work very interesting and conceptually novel in the field. My only concern is that the altered expression of genes implicated in adult myelination should be more carefully discussed. We do not know whether the altered gut microbiom (or any of the metabolites derived) is "directly" causal in changing myelin gene expression of oligodendrocytes, or only "indirectly", i.e. by affecting behaviour first. Indeed, the latter had been shown by the authors' lab to be an important aspect of myelination control.

This is a legitimate comment and for this reason, in this resubmission, we have focused on a careful and detailed analysis of one of the metabolites which was found to be increased in the gut of mice with the behavioral phenotype. In Figure 7, we identify cresol as an important metabolite which was detected at higher levels in the gut of C57Bl recipients transplanted with the microbiota of vehicle-gavaged NOD than in C57Bl transplanted with the microbiota of antibiotic-gavaged NOD and therefore asked whether this metabolite could directly modulate myelin gene expression in cultured oligodendrocytes. In Figure 8 we show that cresol treatment of cultured oligodendrocyte progenitors reduced myelin gene transcripts as it impaired their ability to differentiate. These results are of high relevance and provide a molecular explanation also for the effect of the microbiota on the ultrastructural data in the PFC. Our group and others have previously shown that social avoidance behavior is associated with defective adult myelination in the PFC and the data in this manuscript identify a metabolite, cresol, as a metabolic intermediate with the ability to pass the blood brain barrier and modulate myelin gene expression in oligodendrocyte lineage cells.

A decisive experiment would be to compare the RNA profile of oligodendrocytes in prefrontal cortex and white matter tracts, which are less likely to be affected by behavioural changes (e.g. spinal cord).

We did perform this analysis and show that the changes in myelin genes are detected in the prefrontal cortex, but not in other brain regions such as the nucleus accumbens (Figure 5D and supplemental). This suggests that the effect of the microbiota on behavior is consequent to molecular changes occurring in brain regions affecting social behavior, rather than to non-specific changes throughout the brain.

The authors should also show their immunostainings of myelinated fibers after the transplant. Does it match the strain difference observed before? A nearly 2-fold difference would be visible on Western blots.

In order to provide a clear evaluation of the effect of transplantation of NOD microbiota to C57Bl recipient mice, we conducted a blinded ultrastructural assessment of the PFC myelination. The new data are now presented in Figure 5E which reveal changes in myelination in the PFC and not in the NAC (Figure 5E). These ultrastrictural data are of high relevance, as we and others have previously shown that.

The authors should avoid to write "it is not possible to draw a direct link between gut microbial metabolism and mPFC transcriptome". The necessary experiments have just not been carried out.

We have modified the text accordingly.

Reviewer #3:

The sc and gg vehicles produce different baseline behaviors. Is this an effect of the relative stress of the two procedures?

As previously mentioned, we believe that this question was consequent to the omission of the untreated baseline group in our original submission. The sc group behaved like the untreated group, only the NOD mice gavaged with vehicle showed a depressive-like behavior, which was not observed when they were gavaged with antibiotics, suggesting that the phenotype was related to a specific microbiota.

My one big concern is the enormous gastric damage done by the abs. From the first expt it is not clear that the causative agent is the change in microbiota or the stomach inflammation. The one expt that could be used to point to microbiota is the final one where ab-treated C57 mice get either veh or ab-treated NOD mice microbiota. But to get this we need to see the stomachs of the c57 mice in each group.

We have now included the gross anatomy of the stomachs of these mice to address the reviewer’s concern on gastric damage. While the size of the large intestine was affected by the antibiotic treatment, we did not detect changes in weight, size or presence of any substantial macroscopic damage in the gavaged mice. We also would like to mention that neither the weight nor the blood glucose levels were altered by the procedure.

Reviewer #4:

The premise of the paper is fascinating and the authors have done an admirable job in synthesizing a connection between gut microbiome and cortical myelin related genes. However, their hypothesis comes undone when making a direct link to behavior.

[…]

• It may be helpful to state what would be expected and then look for deviations from the expectation: Microbe-depleted C57 mice are expected to show a similar microbial profile to the respective donor, i.e. NOD vehicle-treated or NOD antibiotic treated. The authors should provide and discuss evidence that is or isn't the case. Unexpected interpretation happens from this point onwards. Interpretation of this data very much depends on what the baseline is, what is viewed as "normal" and what as "altered".

As discussed above, we believe that the reviewer’s concerns are legitimate since we had not previously explicitly shown the data at baseline. We have now entirely modified the figures, reorganized the text and we hope that the new data and presentation of the manuscript will allow the message to come across clear, solid and consistent. To start with, we would like to clarify the general conclusions drawn from of our results:

(1) NOD baseline = no phenotype

(2) NOD vehicle gavage = behavioral phenotype

(3) NOD antibiotic gavage = no phenotype

(4) C57 baseline = no phenotype

(5) C57 vehicle gavage = no phenotype

(6) C57 antibiotic gavage = no phenotype

At baseline, neither NOD nor C57 animals exhibited a depressive-like behavior (1,4). Gavaging with vehicle induced a phenotype in NOD (2) but not in C57 (5). Gavaging with antibiotics did not result in a phenotype in NOD nor in C57 (3,6). From these findings, we reason that the behavioral phenotype must be related to differences in the microbiota of vehicle-gavaged NOD compared to untreated baseline NOD. Treatment with antibiotics was used as a control to demonstrate that the depressive-like behavior observed in NOD mice requires not only gavaging but also a specific microbiota, as its depletion through antibiotics does not lead to a phenotype. By analyzing bacteria found in vehicle-gavaged NOD mice not present at baseline we identified a first set of taxa potentially responsible for inducing the behavioral changes observed (Figure 3).

We then ask the question of whether this microbiota associated with a social avoidance behavior in NOD mice could be transferred to C57 mice. Our results in this second experiment are as follows:

(7) C57 transplanted with the NOD vehicle gavage = behavioral phenotype

(8) C57 transplanted with the NOD antibiotic gavage = no phenotype

From these results, we examined the OTUs found in vehicle-gavaged animals (phenotype) that were not found at baseline (no phenotype) and that were also found in transplanted C57 animals (7, which also exhibit the behavioral phenotype) but not before transplantation (no phenotype), as shown in Figure 6A. We argue that these OTUs are responsible for the depressive-like symptoms observed in donor NOD and recipient C57 mice. Again, we use antibiotics as a control (8) to demonstrate that transplantation from antibiotic-gavaged NOD into C57 does not result in a phenotype. Analysis of the specific OTUs inducing a phenotype identified members of the Clostridiales, Lachnospiraceae, and Ruminococcaceae as candidates for the depressive-like behavior (Figures 6B,C, Figure 6—figure supplement 2C,D and Figure 6—figure supplement 3).

• It would benefit readability of the manuscript a lot if the author would describe the direction of the change they mean rather than say that behavior was "altered" in many sentences. The same is true for "differences" in the microbiota composition. The use of such terms without mentioning the direction of the change make the findings sound very vague and make it more difficult to interpret the data.

We agree with the Reviewer that the initial omission of the baseline data might have resulted in a description of the data that appeared quite convoluted and perhaps not easily accessible. In this revised version, however, the description of the data is accurately depicted in terms of enrichment or depletion relative to baseline levels or by comparing levels before and after transplantation. We sincerely hope that the extensive text and figure revisions have adequately addressed the Reviewer’s concerns.

• How are the mouse strains different genetically? How could this affect the microbiota present in each strain and subsequently the reaction to the antibiotic? This study needs to take gene-environment interactions into account.

As the reviewer highlights, the main point of this study is to address gene-environment interactions, not to define the genetic differences between strains, which might entail an entirely different approach and the use of congenic lines. While we cannot exclude the possibility that the genetic information, which is clearly distinct in the two strains, could affect the differences in the microbiota composition, the main message of our study is that microbiota can transfer behavioral and transcriptional traits, regardless of the genotype of the recipient mice.

• Why use NOD mice at all, it is a strange choice for behavioral genetic studies; does their diabetes interfere with the interpretation of phenotype?

We would respectfully note that the emphasis of the manuscript was not to study the genetic effect on behavior. As such the selection of two mouse strains that had previously been reported to exhibit behavioural differences (Moy et al., 2008) does not seem “strange” to us, but rather adequate to address the effect of environment/gene interaction. Diabetes develops very late in NOD and to exclude any potential confounder related to metabolic alterations we have included monitoring of blood glucose levels throughout the experimental paradigm. Since no changes in blood glucose levels were detected, we conclude that the NOD were not diabetic and as such is not likely that diabetes could interfere with interpretation of our results.

• The author's assertion that social & despair-related behavior is "mediated through changes in the transcriptomic profile of the mPFC" is not proven at all. Correlation dos not mean causation.

We agree with the reviewer and have modified this statement in the text.

• Regarding the myelination data itself the authors show an increase in mRNA and increased intensity of myelin staining. Is intensity staining the best method, does this mean there is more myelination, as in more axons, or are the fibers that are myelinated more heavily myelinated. Looking at structure using electron microscopy would provide an answer. What is the alteration at the level of oligodendrocytes? Have any of the myelin related changes any functional relevance to demyelination-related disorders or to functional myelin responses.

We now include not only transcript levels (measured by RT-PCR), and protein (measured by immunohistochemistry), but also a quantitative ultrastructural analysis and evidence of regional differences in myelination. It is worth mentioning that the selective effect of microbiota in the PFC is not surprising if we consider that this is one region characterized by adult ongoing myelination, and thereby the more likely region to respond to changes in active metabolites like cresol, which have the ability to decrease myelin gene expression in oligodendrocyte progenitors.

• The Introduction is lacking a coherent flow and would benefit from significant editing. I found the first paragraph of Introduction, whilst interesting, overlong and not really pertinent to study at hand.

[…]

• An example of the lack of flow is going from the statement "Our laboratory has recently demonstrated that mouse models of depression were characterized by down-regulation of oligodendroglial transcripts and myelin thickness in the medial prefrontal cortex (mPFC), thereby supporting a role for myelin in the pathogenesis of depressive disorders” to "Indeed, manipulations of intestinal microbiota have recently emerged as a novel approach to influence brain function and behavioral outcomes" – this is a strange transition.

The three comments above have been addressed by entirely re-writing the Introduction. We sincerely hope that the revised text will address the reviewer’s comments. In addition, we believe the overall readability of the manuscript has dramatically improved.

• The authors state that "Antibiotic regimen was well tolerated" other than bodyweight, which is a relatively crude measure, was there any other change in secretomotor function, defecation rates etc.

We disagree with the reviewer regarding this comment as, even though we have not measured the secremotor function or defection rate, the careful assessment of body weight, body condition (including dehydration), overall stomach appearance and weight, collectively indicate an overall healthy state which would not have been detected in case of severe diarrhea or strong intolerance to the antibiotic regimen.

• Motor activity needs to be shown at individual time-bins as the overall effect seems to be different in NOD mice if not reaching statistical significant when collapsed overall.

We have now included in the revised Figure 1 the measurement of the locomotor activity in NOD and C57BL/6 during the SI test.

• Is the myelination effects specific to the cortex; Other brain areas such as the amygdala and hippocampus would be equally worthy of investigation and at least addressed.

We provide evidence for regional specificity in the PFC. Social behavior is known to be affected by changes in myelination in this area. We have now also included also a detailed comparative transcriptional and ultrastructural analysis at the level of the nucleus accumbens, which was not affected by the manipulation of the microbiota. The issue of regional specificity has been linked to the fact that the PFC is one of the very few regions that displays ongoing active myelination in the adult brain and, therefore, we believe more deeply affected by increased levels of microbiota-produced metabolites, like cresol.

• The myelination data is intriguing but its somewhat strange that the authors chose to follow up on the least-significantly enriched functional category (myelination) in their transcriptome analysis; further scientific justification is needed.

We believe that in this revised version we have provided a very convincing argument for the rational behind our interest in the characterization of myelination. We and others had previously described behavioural changes associated with myelination deficits in the PFC. We have also recently shown that enhancement of adult myelination counteracts the effect of social isolation in the induction of social avoidance behavior. All these data and additional rationale is now provided in the Introduction and Discussion of the revised text.

• Were antibiotic treatments continued during the 1.5 weeks behavioral testing?

Yes, the antibiotic regimen was continued throughout the behavioral testing period. To clarify this point we have updated the Materials and methods for this section.

• y-axis in Figure 4g unclear – what is MBP+ fibers mean pixel intensity? How is this measured?

Using image J we have calculated the amount of NF+ axons that were also immunoreactive for MBP+ staining. This method has been used in other published papers such as Liu J et al. (J Neuroscience 2016) and Rusielewicz T et al. (Glia 2014).

Figure 4- it would be worthwile to graphically show the NOD vehicle treated mice for gene expression.

We have generated new graphs in order to show the vehicle treated group bar instead of a dashed line.

• Results section needs to be cleared up to enhance readability.

The Result section has been re-written to render the data more accessible and to improve flow and Readability.

• When describing Figure 5d, the authors state "characterized by lower levels of myelin gene transcripts […] in Group I and higher in Group II", which is confusing as this was not compared to baseline in C57Bl6. They should state that transcription of these genes was increased in Group II compared to Group I as shown in the figure.

We believe there might be some confusion in the data interpretation. Group I and Group II refer to microbiota depleted C57BL/6 recipients that received either the microbiota of the vehicle-gavaged NOD (which displayed the social avoidance phenotype, compared to baseline, as shown in the new Figure 1) or that of the antibioticgavaged NOD (which did not display behavioral changes compared to baseline, as shown in Figure 1). The results are now clearly explained in the Result section of the text.

• The color legend in Figure 6c needs to show the Group names (Group I or Group II), to avoid confusion about the interpretation of the data. Also the description of this figure panel states, "significantly distinct clustering of C57BL/6 transplanted mice displaying social avoidance and an increase in despair-like behaviors compared to normal behaving controls" although, according to the data presented in figure 1, the "normal" behaving animals show more social behavior and less despair-like behavior.

We sincerely hope that the extensive revisions have eliminated these points of contention.

• A list of the differentially expressed genes should be added to supplemental material, together with a list of terms from functional enrichment analysis

The above mentioned data have been uploaded in Dryad as supplemental table.

• The Discussion focuses mainly on the metabolomics data, but does not help explaining how the observed phenotypes are connected to a depleted microbiota. Here the authors show that it is the bacteria that remain after antibiotics (and are therefore overtaking/-growing) are necessary for the myelin increase – that's what the fecal transfer shows. if it was the bacteria that are gone who are important for the myelin changes than you would also see the effects in antibiotic C57Bl6 mice. This however, very much depends on the genetic background of the two strains and it's interaction with the microbiome – which is not discussed in this manuscript at all.

There are two points that are made by the Reviewer. The first one relates to the potential mechanisms linking metabolomics to microbiota and observed phenotype and the second one refers to the possibility that the behavioral changes might be caused by “depletion “rather than “enrichment” for specific taxa. In response to this comment, we now show that cresol, one of the metabolite that we find enriched in the gut of mice with a behavioral phenotype, is capable of impairing myelin gene expression in cultured oligodendrocytes. The transcriptional changes in myelin genes and ultrastructural differences in PFC myelination were only detected in mice with high cresol, which were also characterized by social avoidance behavior and the presence of high levels of Lachnospiraceae, Clostridiales and Ruminococcaceae. The second concept which relates to depletion versus enrichment of specific bacterial taxa has been addressed experimentally in Figure 6 and associated supplement, which clearly shows that the behavioral changes induced by fecal transplantation were detected only in recipients with a high level of bacterial diversity and enrichment in Lachnospiraceae, Clostridiales and Ruminococcaceae.

Figure 7d in the last sentence of the Results sections should be 7e and this sentence is again misleading, or at least again a rather surprising conclusion similar to the above.

The text has been entirely reworded to describe the new changes

• Not clear why the outcome of the increased myelin gene expression in transplantation of NOD microbiota to C57BL/6 is not shown at the protein level.

We provide ultrastructural data that allow a much more detailed level of resolution. Since the changes are regional, we do not believe that myelin preparations obtained from the whole brain would have allowed us to detect these regional specific differences.

• Why are they using two different reference genes in the qPCR data?

This was a typographical error. We only used the 36b4 as reference gene throughout the analysis, as now clearly indicated in the text.

• Images don't elude to which part of the medial PFC they imaged. It should be easy enough with the staining to highlight that. What layer of the PFC has more myelin?

We have now provided a schematic of the region that we have analyzed.

Reviewer #5:

1) The notion that the antibiotic treatments alter myelinating cells in the PFC is based on gene expression analysis RT-PCR and histological data showing changes on MBP+ fibers. While RNA levels provides strong evidence for changes in transcription or RNA stability in oligodendrocytes, to conclude that the antibiotic treatment affects myelin, immunostaining for MBP+ fibers is not sufficient. Morphological analysis like the ones the Casaccia group has done in prior studies (e.g. electron microscopy of myelin thickness, quantification of oligodendrocyte and oligodendrocyte progenitor cell density) will be necessary.

In agreement with the Reviewer’s request we have now included electron microscopy of the prefronatal cortex of mice after transplantation.

2) It would be very important to determine if the changes in myelin gene expression are unique to the PFC or this phenomenon is widespread throughout the brain. The authors should evaluate the expression of the myelin genes affected by antibiotic treatment in a brain region not linked to social/emotional regulation. The interpretation of the findings would be very different if this is a specific or a general phenomenon.

In agreement with the Reviewer’s request we have now included images and transcriptional analysis of the Nucleus Accumbens and we show that no changes can be detected in this brain region.

3) There are some confusing aspects regarding the effects of treatments on behavior. For example, in the immobility tests in Figure 1B, it seems that the only group with altered behavior is the oral-vehicle group. The behavior of mice treated with oral antibiotics is the same as the one with SI or SI-vehicle (and similar to all C57BL/6). Furthermore, in the measurements of social interactions, it seems that the oral vehicle reduces and the oral-antibiotic increases the time in the interaction zone compared to both groups treated by SI. Difference in behaviors between mice treated by SI vs oral, such as those in the elevated plus maze (Figure 2B and C) leaves me wondering if in some cases lack or presence of effects of antibiotics could be due to a particular susceptibility of a mouse strain to a manipulation such as gavage or injections, and not really to a difference in the changes in microbiota. This issue would be clarified by presenting data on control mice (not treated at all).

As described before, we agree entirely with the Reviewer. There was confusion that originated from the omission of the control mice, and hope the reviewed manuscript clearly presents our results and conclusions.

4) Did the authors monitor for onset of diabetes in NOD mice (i.e by detecting evidence of glycosuria)? This strain spontaneously develops diabetes in approximately 20-30% of males and the incidence of disease has been linked to changes in the microbiome (See Wen et al. Nature 2008; PMID 18806780). It would be important to know if the behavioral changes with antibiotic treatment in the NOD mice coincide with metabolic disease to determine if there is an interaction among these variables.

We have monitored glucose levels and body weight in NOD mice and did not detect any changes during the time of our experimentation. The presence of normoglycemia (as shown in Figure 1—figure supplement 3) indicates the absence of diabetes.

5) The authors speculate that neurotoxic metabolites produced by a combination of gut microbia may affect PFC function. This interpretation would be strengthened if these metabolites were measured and detected in the CNS in vehicle and antibiotic treatment mice.

This is a good point. However, to perform this experiment would require a substantial financial and time investment that the lab cannot afford at this moment. In order to address the potential direct effect of the highly cell permeable metabolite cresol on the cell of interest (i.e. myelinating cells), we have directly treated cultured oligodendrocytes with cresol and showed that this metabolite has the ability to impair oligodendrocyte differentiation and decrease myelin gene expression.

https://doi.org/10.7554/eLife.13442.026

Article and author information

Author details

  1. Mar Gacias

    Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    MG, Approved the final version submitted, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    mar.gacias-monserrat@mssm.edu
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0003-1583-1643
  2. Sevasti Gaspari

    Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    SG, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  3. Patricia-Mae G Santos

    Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    P-MGS, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  4. Sabrina Tamburini

    1. Icahn Institute for Genomics and Multiscale Biology, Icahn School of Medicine at Mount Sinai, New York, United States
    2. Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    ST, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  5. Monica Andrade

    1. Icahn Institute for Genomics and Multiscale Biology, Icahn School of Medicine at Mount Sinai, New York, United States
    2. Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    MA, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  6. Fan Zhang

    Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    FZ, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  7. Nan Shen

    1. Icahn Institute for Genomics and Multiscale Biology, Icahn School of Medicine at Mount Sinai, New York, United States
    2. Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    NS, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  8. Vladimir Tolstikov

    BERG, Framingham, United States
    Contribution
    VT, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  9. Michael A Kiebish

    BERG, Framingham, United States
    Contribution
    MAK, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  10. Jeffrey L Dupree

    Department of Anatomy and Neurobiology, Virginia Commonwealth University, Richmond, United States
    Contribution
    JLD, Approved the final version submitted, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article, Contributed unpublished essential data or reagents
    Competing interests
    The authors declare that no competing interests exist.
  11. Venetia Zachariou

    1. Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, United States
    2. Department of Pharmacology and Systems Therapeutics, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    VZ, Approved the final version submitted, Conception and design, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
  12. Jose C Clemente

    1. Icahn Institute for Genomics and Multiscale Biology, Icahn School of Medicine at Mount Sinai, New York, United States
    2. Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, United States
    3. Immunology Institute, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    JCC, Approved the final version submitted, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-3970-9856
  13. Patrizia Casaccia

    1. Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, United States
    2. Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, United States
    Contribution
    PC, Approved the final version submitted, Conception and design, Analysis and interpretation of data, Drafting or revising the article, Contributed unpublished essential data or reagents
    For correspondence
    patrizia.casaccia@mssm.edu
    Competing interests
    The authors declare that no competing interests exist.

Funding

National Multiple Sclerosis Society (RG 4890 A10/2)

  • Patrizia Casaccia

National Institutes of Health (R37NS42925)

  • Patrizia Casaccia

National Institutes of Health (P30 NS047463)

  • Jeffrey L Dupree

National Institutes of Health (P30 CA016059)

  • Jeffrey L Dupree

National Institutes of Health (R01NS52738)

  • Patrizia Casaccia

National Institutes of Health (R01 NS086444-01)

  • Venetia Zachariou

other (GCO14-0560)

  • Jose C Clemente

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Jia Liu (Icahn School of Medicine at Mount Sinai), Dr. Sarah Moyon (Icahn School of Medicine at Mount Sinai), and Dr. Julia Patzig (Icahn School of Medicine at Mount Sinai) for critically reading the manuscript. This work was supported by grant R37NS42925, R01NS52738 and NMSS RG to PC, by the SUCCESS philanthropic grant GCO14-0560 to JCC and by R01NS086444-01 to VZ. Electron microscopy was performed at the VCU - Dept. of Anatomy & Neurobiology Microscopy Facility, supported, in part, by funding from NIH-NINDS Center Core Grant P30 NS047463 and, in part, by funding from NIH-NCI Cancer Center Support Grant P30 CA016059. Computing was partially supported by the Department of Scientific Computing at the Icahn School of Medicine at Mount Sinai.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols of the Icahn School of Medicine at Mount Sinai (#08-0676, #08-0675; LA10-00398; LA12-00193; LA12-00146).

Reviewing Editor

  1. Peggy Mason, Reviewing Editor, University of Chicago, United States

Publication history

  1. Received: December 3, 2015
  2. Accepted: April 7, 2016
  3. Accepted Manuscript published: April 20, 2016 (version 1)
  4. Accepted Manuscript updated: May 3, 2016 (version 2)
  5. Version of Record published: May 16, 2016 (version 3)

Copyright

© 2016, Gacias et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 7,735
    Page views
  • 2,350
    Downloads
  • 45
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

  1. Microbes in the gut influence the social behaviour of mice.

    1. Neuroscience
    Jad Barchini et al.
    Research Article Updated