The gut microbiome is associated with brain structure and function in schizophrenia

The effect of the gut microbiome on the central nervous system and its possible role in mental disorders have received increasing attention. However, knowledge about the relationship between the gut microbiome and brain structure and function is still very limited. Here, we used 16S rRNA sequencing with structural magnetic resonance imaging (sMRI) and resting-state functional (rs-fMRI) to investigate differences in fecal microbiota between 38 patients with schizophrenia (SZ) and 38 demographically matched normal controls (NCs) and explored whether such differences were associated with brain structure and function. At the genus level, we found that the relative abundance of Ruminococcus and Roseburia was significantly lower, whereas the abundance of Veillonella was significantly higher in SZ patients than in NCs. Additionally, the analysis of MRI data revealed that several brain regions showed significantly lower gray matter volume (GMV) and regional homogeneity (ReHo) but significantly higher amplitude of low-frequency fluctuation in SZ patients than in NCs. Moreover, the alpha diversity of the gut microbiota showed a strong linear relationship with the values of both GMV and ReHo. In SZ patients, the ReHo indexes in the right STC (r = − 0.35, p = 0.031, FDR corrected p = 0.039), the left cuneus (r = − 0.33, p = 0.044, FDR corrected p = 0.053) and the right MTC (r = − 0.34, p = 0.03, FDR corrected p = 0.052) were negatively correlated with the abundance of the genus Roseburia. Our results suggest that the potential role of the gut microbiome in SZ is related to alterations in brain structure and function. This study provides insights into the underlying neuropathology of SZ.


Results
Clinical data. The cohort investigated in this study comprised 38 SZ patients and 38 NCs who did not differ in terms of age, sex, BMI or several other parameters (p > 0.05). SZ patients had higher rates of smoking (p = 0.03), longer sleep times (p = 0.007) and fewer years of education (p = 0.01). However, in terms of alcohol intake, there was a higher rate of alcohol consumption in the NC group (p < 0.001) ( Table 1).
Sequencing data and bacterial taxonomic composition. There was no significant difference between the two groups on any assessed measure of alpha diversity (Supplementary Table S1). To determine whether the www.nature.com/scientificreports/ overall gut microbiome composition differed between the two groups, we performed principal coordinate analysis (PCoA) of the Bray-Curtis distance. As shown in Fig. 1, we found a significant between-group difference in Bray-Curtis distance (pseudo-F = 1.71, p = 0.019) under 999 permutations, and the PCoA of the Bray-Curtis distance showed that the SZ and NC groups formed distinct clusters. The bacterial composition results showed that sequences from the NC group were mainly assigned to Faecalibacterium, Megamonas, Roseburia and Gemmiger at the genus level; the most abundant genus in the SZ group was also Faecalibacterium, followed by Megamonas, Ruminococcus and Akkermansia (Fig. 2a). Compared to those in the NC group, the relative abundances of Ruminococcus (p = 0.017, uncorrected) and Roseburia (p = 0.023, uncorrected) were significantly lower in the SZ group, while the relative abundances of Veillonella were significantly higher in the SZ group (Fig. 2b).

Differences in MRI indexes between the SZ patients and NCs.
We found significant differences in GMV in 16 brain regions between SZ patients and NCs. Additionally, significant differences in ReHo in 34 brain regions were found between SZ patients and NCs. Only one brain region showed significant between-group differences in ALFF. See the detailed information in Supplementary Table S2 in Figure 3 shows the significant relationship between the diversity of the microbiome and the MRI indexes in the SZ patients. Both the alpha diversity Faith_PD and the observed species were correlated with the GMV of the bilateral insula and right postcentral gyrus (p < 0.05). Meanwhile, Faith_PD showed a positive correlation with the GMV of the left inferior operculum frontal cortex (p < 0.05). Both evenness and Shannon indexes in the SZ group were positively associated with the ReHo indexes of the bilateral calcarine cortex, bilateral lingual gyrus, left superior occipital cortex and right superior parietal cortex (p < 0.05). Additionally, the evenness of alpha diversity showed positive correlations with the ReHo indexes of the right cuneus lobe, bilateral fusiform gyrus, left postcentral gyrus and left superior parietal cortex (p < 0.05). No significant correlation was detected between the microbial diversity and the ALFF index. Detailed information on the correlations between the microbial diversity and MRI indexes is listed in Supplementary Table S3. After identifying both genera and MRI indexes showing significant between-group differences, we further tested for associations between the abundance of each genus and the MRI indexes. In the SZ group, we found that the ReHo indexes in the right STC (r = − 0.35, p = 0.031, FDR = 0.039), left cuneus (r = − 0.33, p = 0.044, FDR = 0.053) and right MTC (r = − 0.33, p = 0.03, FDR = 0.052) were negatively correlated with the abundance of the genus Roseburia. The relationships between the relative abundance of Roseburia and the ReHo indexes of 3 brain regions are shown in Fig. 4.

Discussion
To the best of our knowledge, this is the first study to find a correlation between the gut microbiome and the brain structure and function in SZ. The main findings are as follows: (1) Consistent with previous studies, significant between-group differences in the gut microbiota and MRI indexes were found; (2) in SZ patients, both the alpha diversity evenness and Shannon indexes showed a positive correlation with the GMV and ReHo indexes in several brain regions; and (3) in SZ patients, the ReHo indexes in the right STC, left cuneus and right MTC were negatively correlated with the relative abundance of the genus Roseburia.
The results from 16S rRNA sequencing demonstrated significant changes in microbial composition between the two groups. The gut microbiome is a complex system that is affected by factors such as diet, age, alcohol consumption and smoking. There is evidence suggesting that exposure to alcohol results in specific modifications of the microbiome composition, bacterial overgrowth and disruption of mucosal barrier function 27 . Chronic alcohol consumption leads to an increase in Proteobacteria and a decrease in Bacteroidetes and increased intestinal permeability, resulting in translocation of LPS and endotoxins in the bloodstream, which eventually contributes to hepatic damage 28,29 .
Cigarettes produce a large number of compounds that have some kind of deleterious effects on different organs, and some of them exert an effect on intestinal health and on the microbiome. Rogers et al. reported an www.nature.com/scientificreports/ increased rate of C. difficile infection in smokers, with odds 33% greater in former smokers and 80% greater in current smokers than in never smokers 30 . Smoking withdrawal in humans results in profound changes in the microbiome, with increased microbial diversity, an increase in Firmicutes and Actinobacteria and a decrease in Bacteroidetes and Proteobacteria 31 .
In this study, shifts in taxonomic abundance in SZ were consistent with previous studies. For instance, Roseburia (order Clostridiales) and Ruminococcus (order Lactobacillales) showed depletion in SZ, which was observed in multiple previous studies on psychiatric diseases 4,32,33 . Roseburia and Ruminococcus are representative bacteria  www.nature.com/scientificreports/ that produce short-chain fatty acids (SCFAs), including butyrate and propionate, which are considered to benefit health 34 . Butyrate is produced from carbohydrates via glycolysis from the combination of two molecules of acetyl-CoA to form acetoacetyl-CoA, followed by stepwise reduction to butyryl-CoA. Roseburia can produce butyrate and propionate via the butyrate kinase route and the 1,2-propanediol pathway 35 . However, Ruminococcus lacks the ability to form butyrate from carbohydrates. The gut microbiota can produce propionate via sugar fermentation, and the main pathway is the succinate pathway. The succinate pathway is known to be present in Ruminococcus, which produces succinate rather than propionate as the end product. On the other hand, some human colonic bacteria belonging to the Negativicutes class of Firmicutes have the ability to convert succinate to propionate 36 . In addition, Veillonella can convert lactate to propionate via the succinate pathway 37 .
Microbial-derived SCFAs can cross the blood-brain barrier (BBB) and activate specific receptors in relevant brain regions pertinent to depression and anxiety-related behaviors 38 . In fact, a decrease in the relative abundance of Roseburia may be detrimental to insulin sensitivity and thus affect the concentrations of branched-chain amino acids (BCAAs) 39 . Due to the brain transporters shared between BCAAs and tryptophan, the presence of excessive BCAAs would cause decreased efficiency of the transportation of tryptophan. It has been reported that excessive consumption of BCAAs could lead to decreased cerebral concentrations of tryptophan, the precursor of serotonin, and thus result in decreased cerebral 5-hydroxytryptamine concentrations 40 . Li et al. 41 speculated that depletion of the Clostridiales taxa, which degrade BCAAs, leads to an elevated concentration of BCAAs in the circulatory system and therefore indirectly decreases cerebral serotonin concentrations, which affects mood.
In this study, we found structural abnormalities in SZ. Our results showed that the GMV of some brain regions, including the bilateral insula, frontal and temporal regions (see Supplementary Table S2 for the detailed regions) in SZ patients was decreased compared with that in NCs. These results are in line with previous studies that found that SZ showed decreased GMV in the insula 42 , superior temporal pole 43 , amygdala, anterior cingulate, and frontal cortices (superior, middle, opercular inferior, and orbital frontal gyrus) 42,43 . Van Rheenen et al. reported entire cortex volume reductions in SZ with cognitive impairments 44 . Actually, a larger cortical volume or greater gray matter density in most brain regions is often associated with better computational efficacy 45 . Thus, we inferred that cognitive impairments may be related to the reduction in GMV in SZ.
We found that most brain regions showed lower functional brain activity in SZ patients than in NCs (Supplementary Table S2). This result is consistent with previous studies that reported altered ReHo in SZ in the bilateral STC 46 , the MTC 22,46 , the bilateral superior medial prefrontal cortex (mPFC) 20 , the right superior frontal gyrus (SFG), and the fusiform gyrus 20,22 compared with those in NCs. Most of the brain regions that showed altered ReHo, such as the STG, MTC and SFG, are related to visual and auditory perception and they are mainly located in the frontal and temporal areas (Supplementary Table S2). Several studies have reported that temporal lobe abnormalities may be related to the emergence of auditory hallucinations 47 , abnormal language processing 47 , thought disorders 48 , and other psychotic symptoms in SZ. Additionally, the frontal lobe mediates a number of important processes that may impact executive function, working memory, abstract reasoning, social behavior, empathy, self-monitoring, and impulse control in SZ 47,49 . We also found that SZ patients showed increased ALFF in the right caudate compared with NCs (Supplementary Table S2). Previous studies reported that SZ patients showed significantly increased ALFF in the right caudate nucleus 50,51 , middle temporal gyrus, inferior parietal lobule 50 , bilateral prefrontal and parietal cortex, and left superior temporal cortex 51 compared with NCs. These findings suggested that the local synchronization of spontaneous activity and the amplitude of fluctuations in SZ brains was widely disrupted, which could help explain the psychopathology of SZ.
We observed significant positive associations between the gut microbial diversity measures and the reduction in GMV in SZ (Fig. 3). Previous studies found that variations in genes were associated with GMV reduction in SZ, especially in the prefrontal cortex and anterior cingulate cortex [52][53][54] . Transcription-neuroimaging association analysis found that the expression levels of 98 genes were significantly correlated with GMV changes in SZ 55 . Although no prior study reported the relationship between GMV and the gut microbiome in SZ, decreased structural integrity of both white and gray matter regions, including the hippocampus, was observed in mice that were colonized with attention-deficit/hyperactivity disorder (ADHD) microbiota 56 . One study reported that the relative abundance of Bacteroides showed greater prominence in the cerebellum, frontal regions, and hippocampus in women, which further supports the hypothesis that microbial modulation may affect mood and behavior 57 . Interestingly, we also found several brain regions associated with PANSS scores in SZ patients (Supplementary Table S4).
Some species have been reported to be associated with brain function. Prevotella has been shown to be associated with the development of brain abscesses and other neurological syndromes via the production of IgA proteases that promote virulence and initiate an immune response 58,59 . Lin et al. reported that the genus Neisseria is negatively associated with functional network connectivity (FNC) loading, especially the FNC between the left angular gyrus and right inferior occipital gyrus, which is related to visual processing function 60 . In addition, members of the genus Neisseria, including the species Neisseria meningitidis, stimulate the immune system through a variety of mechanisms and invade the neurological nervous system during infection 61 . In this study, we found that the depletion of the genus Roseburia was significantly associated with the local synchronization of spontaneous activity of the right STC and the right MTC, which are related to auditory verbal hallucinations 62,63 and thought disturbances 64 in SZ. Dhiman et al. 65 reported that Roseburia was associated with good cognitive performance, which may further support our findings. A preliminary study on the gut microbiome and brain functional connectivity in infants revealed that alpha diversity was significantly associated with functional connectivity between the amygdala and thalamus and between the anterior cingulate cortex and anterior insula 66 , suggesting a potential pathway linking gut microbial diversity and cognitive outcomes.
We found that both evenness and Shannon of microbial diversity in SZ were positively associated with ReHo indexes of the bilateral calcarine cortex, bilateral lingual gyrus, left superior occipital cortex and right superior parietal cortex. Additionally, evenness showed positive correlations with ReHo indexes of the right cuneus lobe,  (Table S3). We inferred that the sensory and cognitive impairments might be related to the alteration of ReHo indexes and microbial alpha diversity in SZ. For example, gut microbial alpha diversity in SZ may be associated with visual hallucinations 62,67 , and this relationship could be mediated by functional brain activity in the calcarine cortex. The present study has several limitations. First, the SZ varied considerably in their medications. Specifically, the durations and types of medication for SZ were distinct. It is important to note that medication can affect microbiota composition and brain activity, which may further affect the results. Second, although we detected correlations between the MRI indexes and microbiota composition as well as diversity in the SZ group, we cannot determine the causal relationship between them. A future longitudinal study may contribute to solving this problem. Third, because the rates of smoking and alcohol consumption were low among all participants and detailed information was not collected, this study could not determine the effects of these two factors on the gut microbiome. Finally, the sample size was moderate. A larger independent sample is needed to examine the reproducibility of our findings.

Participants.
A total of 76 subjects were recruited, including 38 SZ patients recruited from the Affiliated Brain Hospital of Guangzhou Medical University and 38 age-, sex-and BMI-matched NCs recruited in Guangzhou and surrounding areas. The diagnosis of SZ was made based on the Diagnostic and Statistical Manual of Mental Disorder-IV-Text Revision (DSM-IV-TR) (SCID). Subjects with stable psychiatric symptoms for > 2 weeks and a total Positive and Negative Syndrome Scale (PANSS) score of ≥ 30 with a rate of change of ≤ 20% at 2 weeks were included in the study. Thirty-five of the SZ patients were on antipsychotic medication at the time of the study. The exclusion criteria for all participants included (1) any other current major DSM-IV-TR Axis I diagnoses; (2) any somatic diseases; (3) a history of epilepsy, except for febrile convulsions; (4) a history of having received electroconvulsive therapy in the past 6 months; (5) lactating, pregnant, or planning to become pregnant; (6) alcohol dependence or (7) noncompliance with drug administration or a lack of legal guardians.
The study protocol was approved by the ethics committees of the Affiliated Brain Hospital of Guangzhou Medical University, and written informed consent was obtained from each subject or their legal guardian prior to the study. A questionnaire was conducted among all subjects to collect general information, including age, sex, BMI, years of education, history of medication used and history of smoking and drinking.
Fecal sample collection and processing. Fresh fecal samples were collected from participants after fasting for 12 h, and all of the samples were stored at − 80 °C until DNA extraction. A total of 200 mg of each fecal sample was used for DNA extraction. The method of DNA extraction was similar to the protocol described in our previous work 4 .
MRI data collection and preprocessing. MRI data were acquired on a Philips Achieva 3 T MRI Scanner in the Affiliated Brain Hospital of Guangzhou Medical University. The functional data were obtained using an echo-planar imaging (EPI) sequence with the following parameters: repetition time (TR) = 9,000 ms, echo time (TE) = 30 ms, flip angle = 90°, field of view (FOV) = 211 mm × 211 mm, data matrix = 64 × 64, voxel size = 3.44 × 3.44 × 4.6 mm 3 . High-resolution brain structural images were obtained using a T1-weighted 3D gradient-echo sequence (TR = 8.2 ms, TE = 3.8 ms, flip angle = 7°, data matrix = 256 × 256, voxel size = 1 × 1 × 1 mm 3 ). Each study participant was instructed to keep their eyes closed, to relax but not fall asleep, and to move as little as possible.
The structural data were preprocessed using SPM 12 (https:// www. fil. ion. ucl. ac. uk/ spm/) and DPABI (version 4.3, http:// rfmri. org/ dpabi). The GMV was calculated as follows: we first segmented the original T1-weighted images into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) images. Then, the segmented GM images for all of the subjects were used to create a customized Diffeomorphic Anatomical Registration using an Exponentiated Lie algebra (DARTEL) 68 template. Afterward, the GM images were warped to the DARTEL template and spatially normalized to the Montreal Neurological Institute (MNI) space with modulation. Next, the modulated images were smoothed with a Gaussian kernel of 8 mm FWHM 69 . Finally, we extracted the GMV of 90 cerebral regions segmented by the Anatomical Automatic Labeling (AAL) atlas for each subject from the modulated and smoothed images.
The functional data were preprocessed using SPM 12 (https:// www. fil. ion. ucl. ac. uk/ spm/) and DPABI (version 4.3, http:// rfmri. org/ dpabi). For each subject, we performed the preprocessing as follows. First, we removed the first 10 time points to eliminate the nonuniform magnetic field and patient inadaptability to the environment. Then, we performed slice timing correction, and the images were realigned to the first volume for head motion correction. Subsequently, we coregistered the functional images to the individual structural images and then normalized them in Montreal Neurological Institute (MNI) standard space by using an affine transformation with the voxels being resampled to 3 × 3 × 3 mm 3 isotropic voxels. Finally, the resampled data were bandpass (0.01-0.08 Hz) filtered to reduce low-frequency drift and high-frequency physiological noise and spatially smoothed with a Gaussian kernel of 4 mm full width at half maximum (FWHM). After that, we extracted the ReHo and ALFF indexes of each subject from the preprocessed images.
The ReHo index was calculated as follows: first, the ReHo index of each voxel was denoted by Kendall's coefficient of concordance (KCC) of the time series of this voxel with its 26 nearest neighbors 53 . Then, the raw ReHo index of each voxel was divided by the global mean ReHo index for each subject to reduce the global effects of variability across the participants 70 . Subsequently, the individual ReHo maps were partitioned into 90 cerebral regions by the AAL atlas, and the mean ReHo index of each region was acquired by averaging the ReHo indexes www.nature.com/scientificreports/ within that region. Finally, we obtained the ReHo index of each brain region segmented by the AAL atlas for each subject.
The ALFF index was calculated as follows: first, each voxel of the time series was converted to the frequency domain by using fast Fourier transformation 33 . Then, the square root of the power spectrum was calculated and averaged across a predefined frequency range. ALFF is the averaged square root, which reflects the absolute intensity of spontaneous brain activity. Finally, the whole brain voxel average ALFF was divided to reduce the global effects of variability across the subjects to achieve standardization 71 . Finally, we obtained the ALFF index of 90 cerebral regions by the AAL atlas for each subject.
Statistical analyses of bioinformatics and brain function. Sequencing of the V4 region of the 16S rRNA gene was performed on the Illumina MiSeq platform. The sequence data were processed to concatenate reads into tags according to the overlapping relationship by using QIIME2 72 . The raw sequencing results were demultiplexed and quality controlled by applying the DADA2 73 algorithm to generate feature sequences. The output features were rarefied to 13,500 sequences per sample, which was the lowest value in the dataset. Features containing fewer than 2 sequences or those present in less than 20% of the subjects were filtered out. The microbial community structure was characterized using measures of alpha diversity and beta diversity. The alpha diversity indexes we selected were evenness, Faith's phylogenetic diversity (Faith_PD), observed species and the Shannon index. Since the sequence number of fecal samples from one SZ subject was lower than the set sampling depth (13,500), this sample was dropped from the alpha diversity analysis. The differences in diversity between groups were calculated using the nonparametric Kruskal-Wallis H test in QIIME2. The Bray-Curtis dissimilarity of beta diversity indicates differences in taxa composition between samples based on quantitative species abundance data, which may be presented in a distance matrix. Output matrices were ordinated and visualized using the vegan package from R 74 . A classifier for taxonomy analysis was trained based on sequences and taxonomic results from the Greengenes database (http:// green genes. lbl. gov).
The taxonomic table was normalized to the relative abundances at different taxa levels, and 1321 features and 153 genera were obtained. All differential abundances at the genus level were tested using the Mann-Whitney U test. Two-sample t-tests were performed on the MRI indexes to compare the differences in functional brain activity between the SZ patients and the NCs. To determine the association between the differential abundance at the genus level and the MRI indexes, we further calculated the residuals of the relative abundances of those taxa and MRI indexes with significant group differences, controlling for age, sex and years of education by the 'vglm' function in the VGAM package 75 . Pearson's correlations were then calculated between the residuals of the relative abundances of the altered genera and the different MRI indexes. The significance of all tests was set at p < 0.05 or FDR corrected p < 0.05 (two-sided).
Ethical approval. The study protocol was approved by the ethics committees of the Affiliated Brain Hospital of Guangzhou Medical University. Written informed consent was obtained from each subject before the study.
Informed consent. Informed consent was obtained from all individual participants included in the study.

Conclusion
In summary, the current study indicated significant correlations between the gut microbiome and brain structure and function in SZ patients. We found that both Faith_PD and observed species of alpha diversity were significantly correlated with GMV of the bilateral insula and right postcentral gyrus. Additionally, Faith_PD showed a positive correlation with GMV of the left inferior operculum frontal cortex. Both evenness and Shannon indexes of alpha diversity in SZ were positively associated with ReHo indexes of the bilateral calcarine cortex, bilateral lingual gyrus, left superior occipital cortex and right superior parietal cortex. Furthermore, evenness showed positive correlations with ReHo indexes of the right cuneus lobe, bilateral fusiform gyrus, left postcentral gyrus and left superior parietal cortex. Additionally, we found that the ReHo indexes in the right STC, left cuneus and right MTC were negatively correlated with the abundance of the genus Roseburia. These findings demonstrated that the properties of the gut microbiome might be associated with alterations in brain structure and function in SZ patients.

Data availability
The datasets generated and analyzed in the current study are available from the corresponding author upon reasonable request.