Introduction

Schizophrenia is a psychiatric disorder characterized by psychotic symptoms, negative symptoms, and cognitive dysfunction1. A systematic analysis (2019) reported that schizophrenia had the highest disability weight for acute psychotic compared to other mental disorders2. It is generally accepted that antipsychotic drugs are effective in relieving psychotic symptoms by increasing dopamine activity. However, antipsychotic drug efficacy in schizophrenia seems to decline, especially in negative symptoms and depression3. Therefore, it is urgent to investigate the potential mechanism of schizophrenia and to aid an efficient treatment plan.

The gut microbial composition differs in persons with schizophrenia and controls4. Besides often discussed drug-induced microbiome changes, several studies have shown that abundances of gut microbiota species differ between untreated patients with schizophrenia and age-matched healthy volunteers5,6,7. This is probably because the function of the gut-brain axis can be regulated by the gut microbiome. The gut microbiota and brain influence each other mutually through multiple pathways, such as the immune system, tryptophan metabolism, the vagus nerve and the enteric nervous system, and microbial metabolites including short-chain fatty acids, branched-chain amino acids, and peptidoglycans8. Targeting the axis, several attempts have been made to interrupt the progression of schizophrenia, including fecal microbiota transplantation, probiotic supplements, prebiotic treatment, and antibiotics9. Furthermore, the opportunities for the microbiome to become a diagnostic tool were discussed in some diseases9,10. Nevertheless, articles have emerged that offer diverse findings recently. Zhu et al. mentioned that Streptococcus vestibularis was enriched in individuals with schizophrenia and, when transplanted to recipient mice, was found to lead to deterioration in social behaviors5. According to Nguyen et al., phylum Proteobacteria, genus Haemophilus, Sutterella, and Clostridium were found to be relatively decreased in schizophrenia subjects while the genus Anaerococcus was increased11. Unlike Nguyen, in Błażej et al.’s case-control study, generally increased Lactobacillales (order level), Bacilli (class level), and Actinobacteriota (phylum level) were found in schizophrenia patients12. Furthermore, the biological mechanisms related to gut microbiota may play a crucial role in other mental diseases, such as mild cognitive impairment, depression, and Alzheimer’s disease13,14,15,16,17. Thus, whether the causal associations between gut microbiome and schizophrenia exist remains unclear.

Mendelian randomization (MR) has an edge over other conventional observational studies aimed at assessing causal associations. MR constructs instrumental variables of exposure by harnessing genetic variants18. During gamete formation and conception, parents pass on genes to offspring randomly. Therefore, MR can investigate the relationships between exposure and outcome without disturbance of reverse causation or confounding variables19. This approach is increasingly used to infer causality for mental disorders20,21,22.

In this study, we conducted a two-sample MR analysis to identify the associations between gut microbiome and schizophrenia using statistics derived from a genome-wide association study (GWAS) from the MiBioGen and the Psychiatric Genomics Consortium (PGC).

Methods

Data sources

Summary statistics of gut microbiome were obtained from a large-scale, multi-ancestry, genome-wide meta-analysis, also known as the MiBioGen research. Including 18,340 participants from 24 cohorts over 11 countries (16 cohorts from European ancestry, n = 13,266), the study characterized microbial composition from three distinct variable regions of the 16S rRNA gene: V4, V3–V4, and V1–V2. After rarefied every sample to 10,000 reads, they took advantage of direct taxonomic binning to perform taxonomic classification. Microbiome trait Loci (mbTL) were mapped to identify microbiome quantitative trait loci (mbQTL) or microbiome binary trait loci (mbBLT)23. GWAS summary statistics of schizophrenia, obtained from PGC, included 77,285 individuals with schizophrenia and 53,386 control individuals, with most subjects being of European ancestry, containing common variant associations at 287 distinct genomic loci. It is the largest schizophrenia GWAS statistic and has an enlarged number of associated loci24.

Instrumental variable (IV)

We chose candidate instrumental variables at the genome-wide significance threshold p < 1.0 × 10−5. Independent SNPs were selected by clumping in a 10,000 Kb window and setting a linkage disequilibrium threshold of r2 < 0.001. Linkage disequilibrium was estimated by 1000 Genomes phase 3 European samples25.

Statistics analysis

The package “TwoSampleMR” was utilized for all the MR analyses. Three methods including inverse variance weighted (IVW), MR-Egger, and weighted median were applied to calculate the causal effect of gut microbiota on schizophrenia. Cochrane’s Q test in the IVW test and I2 statistics (p < 0.05 and I2 > 0.25) were carried out, aiming at assessing whether there exists potential heterogeneity. Under the premise that the association of each genetic variant with the exposure is independent of its pleiotropy, the MR-Egger provides an efficient test for meta-analysis to detect study bias from horizontal pleiotropy. The true causal effect can be estimated from the slope coefficient of Egger regression26. Then we performed MR-PRESSO analyses to detect the potential horizontal pleiotropy by removing significant outliers27. We calculated the F-statistic to assess the strength of instrumental variables. If the F-statistic of the instrumental variable was less than 10, then this instrument was considered as weak and was excluded from the MR analysis28. The weighted median can be considered a sensitivity analysis for Mendelian randomization investigations and supplementary to MR-Egger29. We prefer to use the inverse variance weighted method as a major method in the analysis because consistent and robust estimates were given under the circumstance that all of the genetic variants are valid instrumental variables.

We performed a reverse Mendelian Randomization analysis in our study to detect the potential causal effects of schizophrenia on the gut microbiome. The same methods and procedures were adopted when we conducted the reverse MR analysis.

Results

In the MR analysis of summary association statistics of gut microbiome and schizophrenia, the numbers of the instrumental variables vary from 5 to 22. The MR-PRESSO analysis identified a total of 6 SNPs as significant outliers for two of the taxa studied (class Betaproteobacteria: rs11128180, rs2321387, rs2613606; genus Eubacterium fissicatena group: rs10147907, rs151257695, rs2733072), suggesting the horizontal pleiotropy may exist in the effect of the two taxa on schizophrenia (Supplementary 1). According to the MR-PRESSO global test, no horizontal pleiotropy was found in any other MR result. After removing these outliers, the leftover SNPs were selected as instrumental variables for subsequent re-run of MR analysis.

The result of IVW suggests that nine taxa were positively correlated with schizophrenia risk, including phylum Firmicutes, class Betaproteobacteria, class Clostridia, order Clostridiales, family Prevotellaceae, genus Alloprevotella, genus FamilyXIIIUCG001, genus Hungatella, and genus Subdoligranulum (OR: 1.08–1.16, p ≤ 0.044). On the other hand, six taxa served as protective factors for schizophrenia, including family Rhodospirillaceae, family Defluviitaleaceae, family Veillonellaceae, genus Coprobacter, genus Gordonibacter, and genus Desulfovibrio (OR: 0.88–0.94, p ≤ 0.049) (Table 1 and Figs. 1A and 2A). The causal effect of class Clostridia, family Rhodospirillaceae, and genus Desulfovibrio on schizophrenia was also supported by the weighted median method. Additionally, the effect direction of the weighted median is consistent with the IVW estimate. However, none of the associations were significant after correction for FDR (p < 0.05). There is no horizontal pleiotropy since the p values of the MR-Egger regression test are all greater than 0.05. The result of Cochran’s IVW Q test revealed significant heterogeneity in the instrumental variables of class Betaproteobacteria (p = 0.010, I2 = 55.3%).

Table 1 MR results of the causal effects of the gut microbiome on schizophrenia.
Full size table
Fig. 1: Causal effects between the gut microbiome and schizophrenia (forest plot).
figure 1

A Causal effects of the gut microbiome on schizophrenia. B Causal effects of schizophrenia on the gut microbiome. CI confidence interval, OR odds ratio, P p value.

Full size image
Fig. 2: Causal effects between the gut microbiome and schizophrenia (scatter plot).
figure 2

A Causal effects of the gut microbiome on schizophrenia. B Causal effects of schizophrenia on the gut microbiome. b MR estimate, P p value.

Full size image

In the reverse MR analysis between schizophrenia and gut microbiome, we found that schizophrenia may increase the abundance of nine taxa, including class Verrucomicrobiae, family Bacteroidaceae, family Verrucomicrobiaceae, genus Akkermansia, genus Bacteroides, genus Candidatus Soleaferrea, genus Lachnospira, genus Ruminiclostridium5, and order Verrucomicrobiales (OR: 1.03–1.08, p ≤ 0.04); while schizophrenia may reduce the abundance of family Defluviitaleaceae and genus DefluviitaleaceaeUCG011 (OR: 0.94, p ≤ 7.12 × 10−3) (Table 2 and Figs. 1B and 2B). No evidence of pleiotropy and heterogeneity was found according to the MR-Egger regression intercept analysis, MR-PRESSO analysis, and Q test in IVW.

Table 2 MR results of the causal effects of schizophrenia on the gut microbiome.
Full size table

Discussion

Our study revealed causal associations between gut microbiome and schizophrenia. We identified nine bacterial features positively and six bacterial features negatively correlated with schizophrenia. In addition, there exist reverse effects from schizophrenia to 11 identified bacterial features.

We reported that the genus Desulfovibrio appeared to be associated with schizophrenia, supported by both the IVW test and the weighted median method. Previous studies highlighted the significance of the genus Desulfovibrio and its metabolism in the pathophysiology of schizophrenia30. A pilot study reported that for patients with exacerbated symptoms of schizophrenia, taking the treatment of schizophrenia with amisulpride cannot change the elevated level of Desulfovibrio in gut microbial composition31. Antipsychotic agents might result in constipation by affecting gastrointestinal motility, and the relative abundance of Desulfovibrio was higher in psychopaths with constipation32.

Our result unveiled that phylum Firmicutes was positively correlated with schizophrenia. A study utilizing the polyriboinosinic-polyribocytidylic (Poly I:C) mouse model reported that the increased level of phylum Firmicutes which activate the immune system might account for neuroplasticity reduction in cortical areas in patients with schizophrenia33. However, recent studies yielded conflicting results. Numerous studies have compared the composition of the intestinal microflora in schizophrenia patients and healthy volunteers, showing that schizophrenia patients have a lower relative abundance of Firmicutes at the phylum level34,35. Xiang et al. observed an increased abundance of Firmicutes in the fecal microbiota of patients with schizophrenia after 14–19 days of antipsychotic treatment36.

As shown in MR results, class Betaproteobacteria is also a risk factor for schizophrenia. In a case-control study, the cognitive function was positively correlated with class Betaproteobacteria in patients with schizophrenia, suggesting that the alternations of the gut microbiome may partly account for the cognitive impairment in schizophrenia6.

We also demonstrated that the class Clostridia and order Clostridiales were causally associated with a higher risk of schizophrenia, which was in line with previous studies37,38, He et al. carried out a series of experiments and observed that the abundance of the order Clostridiales was elevated in subjects at ultra-high risk for psychosis, which produced more Short Chain Fatty Acids (SCFAs). These SCFAs contribute to the activation of microglial cells and the membrane dysfunction confirmed by an increase in choline levels39. Conversely, Li et al. pointed out that the depletion of Clostridiales might mediate schizophrenia through an imbalance in amino acid metabolism and carbohydrate metabolism40. In social isolation rats (an animal model for schizophrenia), decreases in the class Clostridia are probably associated with the increase in several indices of “anxiety-like” behavior and the dysfunction in learning and memory41.

Along with previous studies, the family Prevotellaceae was found to be positively correlated with the risk of schizophrenia in our study37,42. A recent comparative study by Chen et al. involved the gut microbiota structure in schizophrenia patients with violence and without violence. They found more enriched Family Prevotellaceae in schizophrenia patients with violence than in those without violence, which would probably be instrumental to a deeper understanding of the etiology of violence43. Lee et al.’s research reported a differential abundance of Prevotella and Prevotellaceae NK3B31 group in patients with the deficit subtype of schizophrenia compared with those with non-deficit schizophrenia and healthy controls. Correlations between these alternations of gut microbiome and cognitive performance were found in patients who participated in the study44.

Family Veillonellaceae was reported to exert protective effects in schizophrenia in our findings. Schizophrenia patients with violence had impoverished Veillonellaceae compared to those without violence, implying its involvement in violent behaviors43. By contrast, another study observed that the family Veillonellaceae was increased in patients with schizophrenia. In addition, lower glutamate and higher glutamine and GABA in the hippocampus were observed in the mice receiving schizophrenia microbiome fecal transplants, which indicates that the pathology of schizophrenia may be relevant to the alternation of neurochemistry caused by the change of intestinal microbiome42.

In our MR analysis, we identified the genus Family XIII UCG00 and the genus Hungatella that were positively associated with schizophrenia, while the genus Coprobacter was negatively associated with schizophrenia. A broadly similar result of a recent comparative study showed that higher abundances of the Family XIII AD3011 group, Hungatella, and Coprobacillus were found in patients with schizophrenia. The interactions between inflammatory cytokines, gut permeability, lipid metabolism, impaired glucose homeostasis, and gut microbiota alterations probably mediate cognitive impairment, severe psychotic symptoms, and worse social functioning. For example, the level of zonulin appeared to regulate the performance of attention45. Borkent et al. also argued that gastrointestinal barrier dysfunction which increases intestinal permeability serves as a potential cause for schizophrenia46. Several attempts have been made to improve the efficacy of treatments in schizophrenia, such as fecal microbiota transplantation47, probiotic supplement48, and prebiotic treatment49, further supporting the role of the gut microbiome and gut-brain axis in schizophrenia. Zheng et al. reported that antipsychotic medication like amisulpride may exert its therapeutic action by influencing the gut microbiome31. Similarly, it was hypothesized that both probiotics and repetitive transcranial magnetic stimulation (rTMS) alleviate schizophrenia symptoms by influencing the gut microbiome50,51,52. These previous studies show us a promising avenue for the therapy of psychiatric conditions by manipulating the gut-brain axis.

Notably, Mendelian Randomization uses inherited genetic variation underlining each trait as an instrumental variable to detect the association between gut microbiome and schizophrenia. These genetic variants are inherited from parents and will not change after birth, which makes confounding factors less likely to affect these variables16,17.

Nevertheless, the study comes with several limitations. First, the majority of participants in the GWAS data had European ancestry, thus our study may not be suitable for the non-European population. Second, none of our MR results remained significant after the multiple correction. Third, in our study, we did not take some vital environmental factors into consideration, including the effects of diet and medication on the disease progression.

Conclusion

In summary, we comprehensively unveiled the bidirectional causal relationship between gut microbiome and schizophrenia. Our findings probably provide novel insights into the pathology and treatment of schizophrenia.