Second-generation antipsychotics and metabolism alterations: a systematic review of the role of the gut microbiome

Rationale Multiple drugs are known to induce metabolic malfunctions, among them second-generation antipsychotics (SGAs). The pathogenesis of such adverse effects is of multifactorial origin. Objectives We investigated whether SGAs drive dysbiosis, assessed whether gut microbiota alterations affect body weight and metabolic outcomes, and looked for the possible mechanism of metabolic disturbances secondary to SGA treatment in animal and human studies. Methods A systematic literature search (PubMed/Medline/Embase/ClinicalTrials.gov/PsychInfo) was conducted from database inception until 03 July 2018 for studies that reported the microbiome and weight alterations in SGA-treated subjects. Results Seven articles reporting studies in mice (experiments = 8) and rats (experiments = 3) were included. Olanzapine was used in five and risperidone in six experiments. Only three articles (experiments = 4) in humans fit our criteria of using risperidone and mixed SGAs. The results confirmed microbiome alterations directly (rodent experiments = 5, human experiments = 4) or indirectly (rodent experiments = 4) with predominantly increased Firmicutes abundance relative to Bacteroidetes, as well as weight gain in rodents (experiments = 8) and humans (experiments = 4). Additionally, olanzapine administration was found to induce both metabolic alterations (adiposity, lipogenesis, plasma free fatty acid, and acetate levels increase) (experiments = 3) and inflammation (experiments = 2) in rodents, whereas risperidone suppressed the resting metabolic rate in rodents (experiments = 5) and elevated fasting blood glucose, triglycerides, LDL, hs-CRP, antioxidant superoxide dismutase, and HOMA-IR in humans (experiment = 1). One rodent study suggested a gender-dependent effect of dysbiosis on body weight. Conclusions Antipsychotic treatment-related microbiome alterations potentially result in body weight gain and metabolic disturbances. Inflammation and resting metabolic rate suppression seem to play crucial roles in the development of metabolic disorders. Electronic supplementary material The online version of this article (10.1007/s00213-018-5102-6) contains supplementary material, which is available to authorized users.


Introduction
Second-generation antipsychotics (SGAs) have been used successfully for the treatment of schizophrenia, bipolar disorders, autism spectrum disorders, major depressive disorders, tic disorder, agitation, sleeping problems, and dementia, among others (Vasan and Abdijadid 2018). The number of prescriptions for SGAs has increased worldwide for both youths and adults (Ilies et al. 2017), with the most recent cross-sectional study of 14 countries finding that quetiapine, risperidone (RIS), and olanzapine (OLZ) are t h e m o s t f r e q u e n t l y p r e s c r i b e d a t y p i c a l S G A s (Hálfdánarson et al. 2017). An alarming increase in prescriptions, particularly in youth between 15 and 19 years of age, forces us to turn our attention to the health consequences of long-term SGA treatment (Kalverdijk et al. 2017). These consequences include various cardiometabolic adverse effects, such as significant weight gain, hypertriglyceridaemia, hypercholesterolaemia, hypertension, and impaired glucose metabolism (De Hert et al. 2011;Galling and Correll 2015;Galling et al. 2016;Vancampfort et al. 2016), which are all related to metabolic syndrome and cardiovascular disease (Sjo et al. 2017). These changes emerge even after short exposure and increase with cumulative dosages, and differ between agents (Bak et al. 2014).
Overall, in people with severe mental illness, life expectancy is shortened by 10-20 years (Chang et al. 2011), predominantly due to an imbalance in the cardiometabolic system. The prevalence of metabolic syndrome was observed in approximately 30% of patients treated with SGAs (Sanchez-Martinez et al. 2017). Therefore, the American Diabetes Association and the American Psychiatric Association released consensus guidelines to monitor weight and other metabolic parameters in patients treated with SGAs. Moreover, the use of olanzapine in children is discouraged by the Food and Drug Administration because of its association with obesity (American Diabetes Association et al. 2004).
The mechanism of metabolic disruptions, including obesity, hypertension, diabetes, and atherosclerosis (Weiss and Hennet 2017), secondary to SGAs is not fully understood. However, several hypotheses have been proposed, referring to (i) illness-and lifestyle-related factors on metabolism (unhealthy diet, low physical activity, smoking) (Alvarez-Jiménez et al. 2008;Lau et al. 2016;Dayabandara et al. 2017), (ii) SGAs increasing energy intake via neurotransmitter binding in the hypothalamus (Lu et al. 2015), (iii) decreased energy expenditure due to the sedative effect of SGAs (Zimmermann et al. 2003), and (iv) genetic risk associated with the primary illness (Zhang et al. 2016). Other potentially related findings from previous research include (v) diminished insulin synthesis due to the affinity of SGAs for serotonin receptors in the pancreas, leading to diabetic-like metabolic changes (Zhang et al. 2013;Ballon et al. 2014); (vi) elevated muscle, adipose tissue, and liver insulin resistance and glucose transporter efficiency via inhibition of glucose uptake (Dwyer and Donohoe 2003;Verhaegen and Van Gaal 2017); and (vii) accelerated adipose tissue lipogenesis and elevated liver fat content in SGA-treated subjects (Chintoh et al. 2009).
A new approach discussed recently is mediation of SGAinduced adverse effects via the gut microbiota. Maier et al. (Maier et al. 2018) reported that almost one quarter of nonantibiotic drugs used in humans, predominantly antipsychotics, possess antimicrobial activity with potential to imbalance the gut ecosystem. Recently, the inhibition of Escherichia coli APC105 growth in vitro with escitalopram was shown as well as its modulatory effects toward other intestinal bacteria in animals (Cussotto et al. 2018). This might mean that the administration of psychopharmacologic drugs may mimic the effect of low-dose antibiotics and thereby be at least partly responsible for antimicrobial resistance of gut microbiota. On the other hand, Nehme et al. reported that atypical antipsychotics, including RIS, OLZ, aripiprazole, clozapine, and quetiapine, would not possess antimicrobial activity, while phenothiazines and thioxanthenes would inhibit the growth of tested bacteria at various minimum concentrations (Nehme et al. 2018).
As dysbiosis may contribute to body weight alterations and cardiometabolic outcomes (Angelakis et al. 2013;Omer and Atassi 2017;Heiss and Olofsson 2017), SGA-induced dysbiosis has been hypothesized to cause adverse metabolic effects (Kanji et al. 2018). In spite of reports linking specific changes in microbiota to weight gain and metabolic disturbances, the subject has not been comprehensively and systematically reviewed, and the mechanism underlying the potential influence of the microbiota on metabolic processes have not been discussed in detail, taking into account limitations regarding study quality. Therefore, we prepared the first systematic review (SR) investigating the following aims: (1) whether SGAs drive dysbiosis, (2) assessing whether alterations of gut microbiota composition and function affect body weight and metabolic outcome, and (3) examining the possible mechanisms of metabolic disturbances secondary to SGA treatment in rodent and human studies.

Search strategy and selection criteria
This study was conducted according to the requirements established in the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) protocols (Shamseer et al. 2015). Two independent authors (AM and KSZ) systematically searched PubMed/Medline/Embase/PsycInfo/ Clinicaltrials.gov from database inception until 03 July 2018. The search was conducted using the following terms identified as medical subject headings (MeSH bold font), Supplementary Concept Record terms (SCR italic font), and free text terms: (Microbiota OR Gastrointestinal Microbiome OR microbiome OR microbio*) AND (antipsych* OR neurolept* OR SGA* OR Antipsychotic agents OR Anti-Anxiety agents OR Anti-depressive Agents OR Anti-depressive Agents, Second-Generation OR Hypnotics and Sedatives OR Antimanic Agents OR Olanzapine OR Risperidone OR Atypical Antipsychotics) AND (Body Weight OR Body Weight Changes OR Body Weights and Measures OR Body Mass Index OR BMI OR Metabolism OR metabolic* OR Lipids OR tRMR OR Cholesterol OR Triglycerides OR Cholesterol, LDL OR LDL OR Fatty Acids OR Fatty Acids, Volatile OR Acetates OR Butyrates OR Butyric Acid OR Propionates OR hepatic* OR SCFA OR Toxins, Biological OR Bacterial Toxins OR Endotoxins OR Lipopolysaccharides OR Lipid A OR O Antigens OR LPS OR Glucose OR Insulin OR HOMA-IR OR Inflammation OR Cytokines OR Interleukins). Reviews, meta-analyses, and systematic reviews were omitted from the search strategy. The electronic search was supplemented by a manual review of the reference lists from eligible publications and relevant reviews.
Inclusion criteria for animal/human studies were as follows: 1. Treatment with SGAs.
2. An in vivo study.

3.
A study reporting on metabolic as well as body weight changes and alterations of microbiome composition and function (measured by direct and indirect methods).
When an animal or human study consisted of only a few experiments, only experiments fulfilling the above criteria were included and described.

Data extraction and analysis
At least two authors (AM, KSZ, IŁ) independently extracted information from each study, including details on study characteristics (e.g., study design, treatment protocol, duration, number of subjects, outcome parameters, gut microbiota analysis technique), treatment characteristics (e.g., psychopharmacological agent, dosage, duration of treatment), and subject/patient characteristics (e.g., age, sex, comorbidities, metabolic outcomes). When abstracting data from figures, WebPlot digitizer software was used (https://automeris.io/WebPlotDigitizer/). The significance of the analysed studies was arbitrarily assigned according to the following scheme: strong-germfree study, faecal transplantation, statistical significance; middle-conflicting data, lack of relevance due to small group size, data difficult to explain; weak-only co-incidence.

Risk of bias assessment
Two authors (KSŻ and IŁ) independently assessed the risk of bias using the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) Risk of Bias tool for animal studies (Hooijmans et al. 2014), except for item 9 (selective outcome reporting), as this was not assessed in any of the surveyed studies. The STROBE assessment (Vandenbroucke et al. 2014) was used for studies in humans, except for item 16 (main results: unadjusted estimates, confounder-adjusted estimates, category boundaries, translating estimates of relative risk into absolute risk for a meaningful period), as it was not applicable. Outcomes were expressed as the percentage of low-risk judgements (i.e., by dividing the low-risk score by the total number of judgements). When the number was below 16 points (50%), we arbitrarily defined the quality as low. When the results represented up to 60% of the maximum number of points, we treated the study as of moderate quality. Results up to and over 75% were considered high or very high quality, respectively. When a discrepancy occurred, a third author (WM) was involved (Supplementary Figs. S1 and S2).

Descriptive data
The initial search yielded 2340 hits; 2315 articles were excluded as duplicates or after evaluation at the title or abstract level. Out of 25 full-text articles that were reviewed, 15 were excluded for not fulfilling the inclusion criteria. Reasons for exclusion were review (n = 4), no microbiota analysis (n = 6), medications other than SGAs (n = 1), and full-text unavailability (n = 4), resulting in 10 articles that included 15 experiments in the systematic review (Fig. 1).

Risk of bias
An analysis of the overall risk of bias in rodent studies was limited by restricted information being provided. Results were heterogeneous with randomization in three articles (60%) (Davey et al. 2012(Davey et al. , 2013Bahr et al. 2015b), and no information regarding potential conflicts of interest was reported in one article (20%) (Morgan et al. 2014). Other key study quality indicators were poor, and an unclear risk for most types of SYRCLE's bias was identified (Fig. S2). The reporting quality of the human studies was low (Bahr et al. 2015a) (score 13; 40.62%) and moderate (score 17; 53.12%) (Flowers et al. 2017), but a study by Yuan et al. (Yuan et al. 2018) was found to be of relatively high quality (score 20; 62.5%). For details, see Supplementary Figs. S1 and S2.   (2) Testing whether prebiotic can affect mechanism of the action of olanzapine on cortical and hippocampal NMDAR subunit proteins and transcripts.

Microbiota evaluation
Rodents Bacteria in stool were tested in five studies [four OLZ (Davey et al. 2012(Davey et al. , 2013Morgan et al. 2014;Kao et al. 2018) and one RIS study (Bahr et al. 2015b)] using widely applied 16S rRNA sequencing methods. In two OLZ studies (Davey et al. 2012;Morgan et al. 2014) information was provided regarding microbiota diversity. In one RIS study (Bahr et al. 2015b), the number of bacterial operational taxonomic units (OTUs) was reported. The abundance of bacterial phyla was analysed in two OLZ studies (Davey et al. 2012(Davey et al. , 2013 and one RIS (Bahr et al. 2015b) study, whereas bacterial classes were studied in two OLZ studies (Morgan et al. 2014;Kao et al. 2018) and bacterial genera in one RIS study (Bahr et al. 2015b). Briefly, a skewed Firmicutes/Bacteroidetes ratio was the most frequent observation of our SR, secondary to OLZ (Davey et al. 2012(Davey et al. , 2013 and RIS (Bahr et al. 2015b) treatment. Detailed data are presented in Table 1.
Reduced microbiome diversity was identified in 16 rats following OLZ treatment (both genders) (Davey et al. 2012) and 24 female mice after HFD and OLZ regimen (Morgan et al. 2014). Only one study reported data on fewer OTUs in female mice treated with RIS (Bahr et al. 2015b). OLZ treatment in rats increased the abundance of Firmicutes from 6.40 to 16.01% and decreased the abundance of Bacteroidetes from − 6.69 to − 4.30%. The effect was dose dependent and greater in females (Davey et al. 2012(Davey et al. , 2013. In addition, the abundance of Actinobacteria (females 2 mg, − 3.38%; 4 mg, − 3.57%) and Proteobacteria (females 2 mg, − 1.45%; 4 mg, − 0.83%; males 2 mg, − 2.21%) was decreased compared with vehicle-treated rodents (Davey et al. 2012). The other OLZ rodent study found an increase in the relative abundance of classes Erysipelotrichia (up to 3.40%) and Gammaproteobacteria (up to 0.45%), whereas the abundance of class Bacteroidia was reduced (− 5.30%) (Morgan et al. 2014). Only in a single study (Kao et al. 2018) OLZ administration caused no variations within microbiota composition in comparison with vehicle-treated rodents, possibly because of the short treatment duration and the dose of the administered drugs. However, OLZ was administered prior to Bgalactooligosaccharide (B-GOS) and attenuated prebiotic mode of action (↑ Bifidobacterium; ↓ Escherichia/Shigella spp., Coprococcus spp., Oscillibacter spp., C. coccoides spp., Roseuria Intestinalis cluster, and Clostridium XVIII cluster) which indirectly suggested that this SGA influenced gut microbiota. Also, acetate concentration in faeces, a by-product of gut microbiota, increased in OLZ-treated rodents, implying that microbiome structure and function could be at least partly changed by SGA (Kao et al. 2018).
Although RIS was implemented in six rodent experiments (from three articles), a microbiota analysis was performed in only one of them. An increase in the relative abundance of Firmicutes (32.6%) was found with a reciprocal decrease in the relative abundance of Bacteroidetes (− 22.40%) in drug-treated subjects (Bahr et al. 2015b). Alistipes spp. and Lactobacillus spp. were more prevalent in control-treated rodents, whereas the population of Allobaculum spp. increased (36.5%) in the RIS group (Bahr et al. 2015b). Furthermore, SGAs were shown to possess antibacterial properties in vitro; OLZ inhibited the growth of anaerobic bacteria (Bahr et al. 2015b), and diminished the growth of Escherichia coli NC101 but not Enterococcus faecalis OGIRF cultures (Morgan et al. 2014).

Humans
In two human studies (Bahr et al. 2015a;Flowers et al. 2017), the bacteria in stools were tested using 16S rRNA sequencing  (Yuan et al. 2018), the copy numbers of five bacterial genera (Bifidobacterium spp., Clostridium coccoides group, Lactobacillus spp., and Bacteroides spp.) were determined by means of qPCR analysis. The difference in microbial communities between observed groups was calculated using principal coordinate analysis (PCoA). One study assessed Shannon diversity (all sample species) (Bahr et al. 2015a), and another focused on Simpson (dominant species) diversity (Flowers et al. 2017). The abundance of bacterial phyla was analysed in one study (Bahr et al. 2015a), whereas bacterial families were analysed in two human studies (Bahr et al. 2015a;Flowers et al. 2017) and bacterial genera in all human studies (Bahr et al. 2015a;Flowers et al. 2017;Yuan et al. 2018). The Firmicutes/Bacteroidetes ratio was reported in only the RIS study (Bahr et al. 2015a). Flowers et al. (2017) reported reduced Simpson diversity in females treated with SGAs which remained significant after adjusting for age, BMI, and benzodiazepine treatment (p = 0.002, β = − 4.6, R 2 = 0.12). A greater abundance of Lachnospiraceae was observed in obese patients treated with SGAs, whereas Akkermansia and Sutterella abundance was higher in controls, though only the first two differences (Lachnospiraceae and Akkermansia) remained significant (p = 0.001 and p = 0.03) after adjusting for BMI and gender. Lastly, the study found that Akkermansia were less prevalent in non-obese SGA users (p = 0.005) (Flowers et al. 2017). Bahr et al. (2015a) identified a significantly higher Shannon diversity index (0.7 points) and phylogenetic diversity in 18 male adolescents chronically (> 12 months) treated with RIS compared to psychiatric control participants. The Bacteroidetes/Firmicutes ratio was significantly lowered (0.15 vs 1.24, respectively, p < 0.05) in chronic and shortterm (1-3 months) RIS users. The tendency to decrease the Bacteroidetes/Firmicutes ratio observed in short-term RIS users seemed to correlate with the change in BMI Z-score, which is a function of both age and gender and shows the deviation from the population mean. The observed results were not significant, probably because of the small number of patients. Moreover, the authors observed that long-term treatment with RIS and significant weight gain in RIS users were associated with alterations in the gut microbiome: an increased abundance of the phylum Proteobacteria, families Erysipelotrichaceae and Ruminococcaceae, and genera Clostridium, Lactobacillus, Ralstonia, and Eubacterium and decreased abundance of the genera Prevotella and Alistipes. Interestingly, the abundance of phylum Actinobacteria and species Collinsella aerofaciensin was elevated in the RIS group without weight gain, which suggests a protective effect of these bacteria in chronic RIS users (Bahr et al. 2015a).

Rodents
The effect of OLZ administration on body weight was measured using different experimental models in four studies (experiments = 5) including C57BL/6J female mice (experiments = 2) (Morgan et al. 2014) or both genders of Sprague-Dawley rats (experiments = 3) (Davey et al. 2012(Davey et al. , 2013Kao et al. 2018). The impact of RIS on body weight was determined in three studies (experiments = 6); one study included C57BL/6J female mice (Bahr et al. 2015b), and two other studies (conference abstracts) did not report mouse gender (Grobe et al. 2015;Riedl et al. 2017). We did not conduct aggregated analysis concerning body weight, due to methodological differences and various ways of expressing measured values.
In general, administration of OLZ increased body weights in female rats (Davey et al. 2013;Kao et al. 2018) and female mice (Morgan et al. 2014). This effect was dose independent in two studies (Davey et al. 2012(Davey et al. , 2013. In one study, OLZ administration caused increased adiposity (percentage of body fat), even after correction for weight gain (Morgan et al. 2014). The body weight increases induced by OLZ was counteracted by antibiotic administration (Davey et al. 2013) and lack of bacteria (germ-free mouse model), and potentiated by an HFD (Morgan et al. 2014). The increase in body weight induced by RIS administration was not affected by antibiotics (Bahr et al. 2015b). In rodents receiving RIS, the non-aerobic resting metabolic rate (RMR) was suppressed in mice in three studies (Grobe et al. 2015;Bahr et al. 2015b;Riedl et al. 2017). One study reported no direct information regarding body weight but identified that, in RIS-treated rodents, the suppression of tRMR was not affected by cecectomy (Riedl et al. 2017). Increased fat mass and free fatty acid release and elevated expression of lipogenic enzymes were observed in 12 female rats (Davey et al. 2013). Importantly, in six female rats, administration of SGAs resulted in elevated expression of macrophage marker CD68 in adipose tissue, indicating that  body weight gains were associated with recruitment of macrophages into the fat mass (Davey et al. 2013).

Humans
In 18 male children chronically treated with RIS, the BMI Zscore increased by a mean 0.31 ± 1.11 points, and the BMI Zscore increased over 10 months of treatment (mean 0.28 ± 0.23 units) in a longitudinal study arm (Bahr et al. 2015a No other metabolic investigations were undertaken; however, Bahr et al. (2015a) performed Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) analyses and found that bacterial orthologues enriched in chronic RIS patients compared to controls were responsible for environmental information processing pathways and cellular processes, including short-chain fatty acid and tryptophan metabolism. In persons free of psychiatric treatment, there were more orthologues involved in bacterial metabolic pathways, such as vitamin metabolism. Detailed data extracted from original papers included in our systematic review are provided in Table 2.

Rodents
Most studies investigating this hypothesis have been performed in different experimental models. Morgan et al. (2014), in their germ-free experiment in gnotobiotic mice, showed that the microbiota is necessary to induce metabolic changes after OLZ treatment. One experiment a s s u m e d t h a t t h e r e l a t i v e a b u n d a n c e o f c l a s s Erysipelotrichi, which is increased by OLZ, is linked to rapid weight gain; every 1% increase in abundance resulted in a weight gain of 0.7%. The same pattern, though more pronounced, was identified relative to the class Actinobacteria (for which an OLZ effect was not observed); every 1% increase in abundance resulted in a weight gain of almost 5% (Morgan et al. 2014). In a study by Kao et al. (2018) 2-week intraperitoneal OLZ administration in female rodents elevated body weights, and prebiotic therapy attenuated this effect. The study indicated no OLZ-induced alterations of gut microbiota which means that B-GOS supplementation may prevent weight gain independently of its influence on gut microbiota. In a rodent RIS intervention (Bahr et al. 2015b), the medication caused a significant increase in weight (2.8 g) compared to control mice, and co-administration of antibiotics had no significant effect on the weight gain. When performing microbiota transfer from RIStreated mice, recipients had a 16% reduction in total resting metabolic rate (tRMR) due to a reduction in non-aerobic RMR. tRMR states for the largest portion of total energy need thus is relevant to describe metabolic outcomes (Astrup et al. 1999) Similarly, transfer of the phageome from RIS mice resulted in a significant weight gain in recipients relative to the vehicle study arm (p < 0.05) (37). Two studies obtained indirect data on the metabolic influence of microbiota changes following RIS treatment (Grobe et al. 2015;Riedl et al. 2017). In the study by Grobe et al. (2015), faecal transplants from RIS-treated rodents resulted in elevated body mass through non-aerobic RMR suppression, which was found to be unaffected by cecectomy (Riedl et al. 2017). Study conclusions are shown in Table 3.

Humans
In humans, the data are less clear. Bahr et al. found increased BMI Z-scores in humans treated with RIS (mean 0.31 ± 1.11 points over the treatment course), whereas in controls, the scores seemed to be unchanged (Bahr et al. 2015a). Chronic treatment with RIS and a significant gain in BMI resulted in a lower Bacteroidetes/Firmicutes ratio compared with the control group. Moreover, differences in bacterial composition were observed in RIS-treated children who had BMI gains compared to those who did not. Detailed data concerning the association of differences in microbiota abundance depending on body weight gain are presented in Table 2. In the initial phase of RIS treatment lasting 10 months, BMI Z-scores increased a mean 0.28 ± 0.23 units, which seemed to correlate with a decreased Bacteroidetes/Firmicutes ratio starting 1-3 months after treatment initiation (the observed result was not significant, likely due to the small sample size).
In the second human study, SGA treatment resulted in higher BMIs, even after adjusting for patient age and gender followed by significant elevation in the abundance of Lachnospiraceae. Akkermansia counts were significantly lowered in the SGA-treated group, including non-obese patients. Surprisingly higher Lachnospiraceae and lower Akkermansia counts were observed in non-SGA-treated obese individuals (Flowers et al. 2017).
In the last human study, the authors (Yuan et al. 2018) found that at baseline Bifidobacterium spp., counts negatively correlated with serum levels of LDL and Escherichia coli count was negatively correlated with serum levels of triglycerides and hs-CRP, even after controlling for age, gender, smoking status, and disease duration. Following the treatment, a decrease in Clostridium coccoides group (since 6 weeks) and Lactobacillus spp. (since 12 weeks) and elevations in numbers of Bifidobacterium spp. (since 6 weeks) and Escherichia coli (since 6 weeks), in comparison to baseline values, were reported. When they conducted hierarchical multiple linear regression, only the differences in faecal Bifidobacterium spp. count significantly correlated with the weight changes over 24 weeks of RIS treatment (Yuan et al. 2018). A summary of the evidence from human studies is provided in Table 4.

Discussion
To the best of our knowledge, this is the first SR investigating the effect of SGAs on intestinal microbiota in relation to frequent metabolic adverse events associated with their use in clinical practice. This SR aimed to find answers to three questions: (1) do SGAs affect intestinal microbiota resulting in dysbiosis, (2) whether SGA-related metabolic disorders are associated with dysbiosis, and finally (3) to elucidate the mechanisms behind SGA treatment and dysbiosis leading to body weight and metabolic disturbances (Fig. 2) (Delzenne et al. 2011). Although numbers of existing rodent and human studies are limited, we found that dysbiosis secondary to SGA treatment can play a role in metabolic alterations, including weight gain.
The human gut is colonized by roughly 39 × 10 9 bacterial cells (Abbott 2016), with other species, including Archaea, eukaryotes, fungi, and viruses (Consortium et al. 2012). Dysbiosis is an imbalance in the number, composition, or function of bacteria in a given environment. Dysbiosis has been confirmed in all but one of the experimental studies in which the content of bacteria in stools was assessed. The most frequent observation was an increase in the Firmicutes/ Bacteroidetes phyla ratio. It should be emphasized that the ratio of Firmicutes/Bacteroidetes was elevated in all experimental studies investigating this parameter. These two phyla are generally dominant in the human intestinal microbiome, comprising approximately 90% of the gut microbiota (Consortium et al. 2012). The Bacteroidetes phylum has been found to synthesize acetate and propionate, while Firmicutes mainly by-product is butyrate (den Besten et al. 2013). Beneficial effects of SCFAs were shown as far as gastrointestinal functions, neuro/immune regulation, and host metabolism were concerned (Maciejewska et al. 2018;van de Wouw et al. 2018). Proper concentration of SCFAs acting via its receptors is crucial for energy homeostasis. When GPR43-deficient mice were fed with normal diet, they started to accumulate fat and SCFA-dependent activation of the receptor resulted in suppressed insulin signalling within the adipose tissue thus inhibited fat storage (Kimura et al. 2013). In 2005, Ley et al. hypothesized that differences in gut microbial ecology may be an important factor affecting energy homeostasis (Ley et al. 2005). Later, an increased Firmicutes/ Bacteroidetes ratio was observed in obese rodents and humans (Turnbaugh et al. 2009).
Elevated gut microbiota fermentative metabolism by overrepresented Firmicutes may therefore promote more intensive intestinal monosaccharide absorption, energy extraction from non-digestible food components, hepatic de novo lipogenesis, and adipocyte fatty acid storage. The analysis of the SCFA concentration represents an indirect way to analyse microbiota composition (at least a skewed Firmicutes count) and can be viewed as a marker of microbiota metabolic function.
Observations from experimental studies have only been partially confirmed in human studies. Bahr et al. (2015a) observed increased Firmicutes/Bacteroidetes ratios in both cross-sectional and longitudinal studies. In other human studies, bacterial phyla were not reported. Another result of experimental studies confirmed in the human trial was the increase in the abundance of class Erysipelotrichi, which was found to be related to the occurrence of non-alcoholic fatty liver disease (Spencer et al. 2011;Henao-Mejia et al. 2012;Raman et al. 2013). Of note, as recently demonstrated by Schwarz et al. (2018) in first-episode psychosis patients, the abundance of predominantly Lactobacillus from Firmicutes phyla was increased which correlated negatively with different clinical scores of schizophrenia. Also, after 12 months of treatment in patients with smaller alterations within gut microbiota at baseline, remission rate was more frequent. However, patients were not drug-naïve, and received antipsychotics for approximately 20 days, which at least partly confirm the higher abundance of Firmicutes following SGA treatment. We found no more data on how SGA-induced dysbiosis may affect the clinical course of the disease and consequently treatment success. To close the circle, Lactobacillus represents only a single genus within the Firmicutes phyla; thus, the effect of SGAs on Firmicutes/Bacteroidetes ratio in humans could only be speculative and requires further research.
The mechanism of dysbiosis secondary to SGAs has not been fully explained. In two studies included in the present SR, both OLZ (Morgan et al. 2014) and RIS (Bahr et al. 2015b)  (1) The gut-brain axis controls metabolism, appetite, and digestive tract functions and may become altered under dysbiosis insult.
(2) SCFAs activate G protein binding receptor (GPR) which is followed by secretion of YY peptide (PYY) resulting in lowered gut motility. Dysbiosis also induces GLP-1 resistance, followed by diminished GLP-1 receptor expression and hampered nitric oxide production. Consequently, the gut-brain peripheral axis responsible for insulin secretion and stomach emptying is inhibited.
(3) Dysbiosis is associated with loss of integrity of the gastrointestinal barrier and increased permeability of the intestinal mucosa for gut lumen antigens, including bacterial LPS.
(4) High-energy SCFAs and endotoxemia affect multiple metabolic pathways. Activation of the differentiation of peroxisomal gamma proliferator-activated receptors (PPARγ) and the pro-inflammatory endocannabinoid system takes place which regulates fatty acid synthase (FAS), enhancing hepatic de novo lipogenesis. LPS exacerbates hepatic steatosis and insulin resistance. Consequently, macrophages infiltrate adipose tissue and (5) body weight increases. The mechanisms include suppression of non-aerobic RMR (OLZ), increased free fatty acid release, and elevated hepatic expression of the lipogenic enzyme fatty acid synthase (RIS). Possible targets counteracted by prebiotics and probiotics are circled in red. FAS fatty acid synthase, GLP-1 glucagon-like peptide 1, GPR G protein binding receptor, LPS lypopolysaccharide, NO nitroic oxide, OLZ olanzapine, PPARγ peroxisomal gamma proliferatoractivated receptors, PYY YY peptide, RIS risperidone, RMR resting metabolic rate, SAA3 serum amyloid A3, SCFAs short-chain fatty acids coordinate analysis of the uniweighted UniFrac distance showed that antibiotics had a synergistic influence on gut microbiota with RIS. This effect is predominantly typical of drugs subject to enterohepatic circulation and intensely excreted in the bile (Morgan et al. 2014). Bactericidal activity causes dysbiosis via elimination of specific bacteria from the gastrointestinal tract. The observed phenomenon of a greater increase in body mass in naive SGA users (Maayan and Correll 2010) can be caused, among other reasons, by the induction of bacterial resistance in relation to the repeated use of SGAs. It should be emphasized that the observed results are not unambiguous and easy to interpret. The composition of intestinal bacteria varies among individuals and is analogous to fingerprints. Also, in individual studies (also experimental), various taxonomic groups of bacteria were analysed, and their content was analysed only in stools. The composition of bacteria in the stool is more stable and does not depend on external factors compared with the composition of bacteria in the small intestine. Changes in the microbiota of the small intestine have a much greater effect on the metabolic functions of the body. Therefore, in further experimental studies, attention should be paid to this problem.
To address the question of whether SGA-induced dysbiosis may be responsible for metabolic malfunctions, we analysed a few experimental models. Two studies introduced antibiotic cocktails as experimental variables (Davey et al. 2013;Bahr et al. 2015b). Co-administration of antibiotics, which significantly reduced gut bacterial content, prevented dysbiosis and its metabolic consequences. On the other hand, antibiotics used only to slightly modify gut microbiota had antibacterial activity similar to SGAs but did not influence their metabolic effects. These observations confirm a potential causal relationship between dysbiosis caused by the intake of SGAs and metabolic disorders. Similar results were previously reported in mouse models (Mathur et al. 2016). As some antipsychotics have been documented to possess antimicrobial activity (Kristiansen 1979), the administration of these drugs may resemble the mode of action of low-dose antibiotic cocktails, which may be responsible for increased body mass as observed in livestock (Morgan et al. 2014). In contrast, some non-absorbable antibiotics (e.g., rifaximin) have been shown to reduce the abundance of methanogenic bacteria (Mathur et al. 2016), resulting in significant weight loss in obese individuals with diabetes (Riedl et al. 2017). The antibiotic effect on body fat composition is plausibly dose-and age-related (Cox et al. 2014). Additionally, Morgan et al. (2014) implemented HFD, which induces changes in the composition of the gut microbiota toward an obesogenic composition with subsequent metabolic consequences, including metabolic syndrome (Yang et al. 2017). However, HFD had no impact on the anthropometric indices of germ-free mice (Bäckhed et al. 2007), which may indicate that metabolic disturbances during OLZ treatment are sourced from altered gut microbiota. It was observed that both HFD and OLZ have a synergistic effect on gut microbiota composition but weight gain in mice receiving OLZ is more rapid than feeding only with HFD. This means that in the case of metabolic disorders caused by OLZ administration, in addition to the SGA-mediated obesogenic effect on gut microbiota, other factors should also be taken into consideration. Also, it is possible that such treatment is more pronounced in comparison to HFD alone. Morgan et al. (2014) confirmed the relationship between OLZ-induced dysbiosis and metabolic disorders using a germ-free model, in which they showed that the lack of bacteria in the gastrointestinal tract in mice receiving OLZ did not cause weight gain, and their conventional housing leading to the colonization of the digestive tract resulted in induction of weight gain. A similar observation was made by Bäckhed et al. (2007) who found that gnotobiotic mice had less body fat than mice housed under conventional conditions. Morgan et al. (2014) confirmed the relationship between weight gain and OLZ administration using the cross-over model, proving that the relative abundance of bacteria of the Erysipelotrichi class was related to weight gain.
Another argument for the relationship between the occurrence of dysbiosis and metabolic disorders with an increase in body weight caused by the administration of OLZ was provided by Davey et al. (2012) who observed microbiota and metabolic disorders only in female mice. In male mice, metabolic side effects and impact on bacterial abundance were minimal. In human trials included into present SR, no differences in metabolic outcome between males and females were reported (Flowers et al. 2017;Yuan et al. 2018). However, such gender-dependent discrepancies were reported earlier and may be due to drug pharmacokinetic differences (Harris et al. 1995;Beierle et al. 1999) which further support the necessity to conduct more studies. Bahr et al. and Grobe et al. provided the strongest evidence for a link between the occurrence of metabolic disorders and gut microbiota (Grobe et al. 2015;Bahr et al. 2015b). The authors observed that faecal and phage transplantation from mice treated with RIS caused weight gain and decreased rest metabolic rate. The phenotypic effect of faecal transplantation is considered as a very strong evidence of intestinal microbiota, also in terms of its effect on metabolism. For example, lean mice-derived microbiota transferred to germ-free mice resulted in lower body fat increases than the transfer of the ob/ob mice microbiome (Turnbaugh et al. 2006). Riedl et al. did not confirm this observation and found that cecectomy (associated with marked reduction of gut microbiota counts) did not influence the suppression of non-aerobic RMR caused by RIS treatment (Riedl et al. 2017). This observation may indicate that RIS is an antibacterial agent that reduces RMR, and a further reduction of bacteria count via cecectomy is not needed. However, it is also possible that the mechanism does not depend on intestinal microbiota. It should be emphasized that the results of this study come from a conference summary, and may not contain the necessary data for accurate interpretation of these observations (Riedl et al. 2017).
Human studies provide little evidence on the relationship between metabolic changes and microbiota alterations after SGA treatment. Bahr et al. (2015a) found specific microbiota alterations in weight gain of children chronically treated with RIS, although the results were not significant probably due to the small sample size. However, PcoA of unweighted UniFrac distances showed elevated phylogenetic diversity and robust differences between the overall gut microbial profiles, but no appreciable differences between significant versus nonsignificant BMI gain groups during treatment. Bioinformatic analysis (Bahr et al. 2015a) demonstrated increased butyrate and propionate metabolism in the RIS-treated group, as well as SCFA production and impairments in tryptophan metabolism. Flowers et al. (2017) found that low abundance of bacteria from the genera Akkermansia was associated with a lack of weight gain after SGA treatment. However, these results are difficult to explain. Akkermansia muciniphila may serve as a negative marker of inflammation as it was found that the abundance of this genus is reduced under the regimen of HFD and its decline correlated negatively with lipid synthesis, plasma markers of insulin resistance, cardiovascular risk, and adiposity in rodents (Schneeberger et al. 2015). In a recent study published by Yuan et al. (2018), it was concluded that weight gain in patients treated with RIS was significantly correlated with an increase of faecal Bifidobacterium spp. abundance. Bifidobacterium spp. have an antiinflammatory effect against systemic inflammation, and an increased abundance could be a compensatory reaction after weight gain and inflammation.
To answer the question whether dysbiosis caused by SGA administration is related to the occurrence of metabolic disorders, prospective clinical trials, and further experimental studies in which the same taxonomic groups of bacteria and their metabolic functions are assessed are necessary.
Based on the results of the analysed studies, we attempted to describe the mechanism of metabolic disorders originating from SGA treatment. Based on the current systematic review, we conclude that inflammation is critical to inducing weight gain and other metabolic alterations secondary to SGA use (Straczkowski et al. 2002;Kim et al. 2006). First, dysbiosis affects energy homeostasis of the body and lipid metabolism (Slyepchenko et al. 2016;Boulangé et al. 2016). Also, dysbiosis alters the structure and function of the intestinal barrier and may cause the translocation of bacterial antigens into the systemic circulation (Küme et al. 2017). Data on the presence of various microorganisms in extracolonic tissues and organs (Nagpal and Yadav 2017) following HFD (Wirostko et al. 1990) in obese individuals are increasing (Schwiertz et al. 2010). Bacterial lipopolysaccharide (LPS) components of gram-negative bacteria and cyanobacteria and high-energy SCFAs play major roles in energy harvesting. These molecules, among others, activate the G protein binding receptor (GPR), followed by secretion of the YY peptide (PYY), resulting in decreased intestinal motility, increased fat storage by reduced expression of the lipoprotein lipase inhibitor (fasting-induced adipose factor (FIAF)), activation of the differentiation of peroxisomal gamma proliferatoractivated receptors (PPARγ) and the pro-inflammatory endocannabinoid system, respectively, and the development of adipose mass. These components regulate fatty acid synthase (FAS), enhancing hepatic de novo lipogenesis. LPS exacerbates hepatic steatosis and insulin resistance (Marlicz et al. 2014;Jin et al. 2017;Rorato et al. 2017). Macrophages possess the ability to phagocytose LPS, migrate to peripheral tissues, and release pro-inflammatory cytokines. Consequently, adipokine synthesis is decreased and leptin and ghrelin levels increase. All of these pathways sustain systemic inflammation Kaser 2009, 2011;Park and Scherer 2011) and may contribute to metabolic alterations in patients exposed to HFD with type 2 diabetes mellitus (Burcelin 2012). Unfortunately, LPS and gut barrier function were not measured in the analysed studies.
In immune-related pathogenesis of obesity, bacterial LPS and peptidoglycans, as pro-inflammatory agents, activate pathogen recognition receptors (PRRs) on macrophages and neutrophils, and as part of the non-specific immune response (Burcelin 2012) are responsible for hyperinsulinaemia and insulin resistance (Saberi et al. 2009). Furthermore, immune-related mechanisms may contribute to gut microbial alterations. For example, mice lacking TLR5 develop dysbiosis followed by metabolic syndrome (Tilg and Kaser 2009;Tremaroli and Bäckhed 2012). Obesogenic-type dysbiosis induces inflammation within the gut and affects neurotransmitter levels and the gut-brain axis function (Collins et al. 2012). Skewed production of serotonin in the gut (Clarke et al. 2013) may be at least partly responsible for weight gain secondary to SGA treatment via microbial alterations (Collins et al. 2012). However, none of the studies included in this SR reported such an association. A study by Kao et al. (2018), however, demonstrated that the B-GOS mode of action is independent of the serotonin pathway.
Only one study in this systematic review reported elevated levels of TNF-α when co-administered with B-GOS (Kao et al. 2018). TNF-α was found to act as weight gain suppressant and influence adipocytes lipid metabolism (Langhans and Hrupka 1999;Coppack 2001). Unexpectedly, another study discovered macrophage infiltration of adipose tissue (Davey et al. 2013). However, Xu et al. (2003) discovered that macrophage infiltration of adipose tissue is associated with the development of inflammation and insulin resistance in obese individuals. In one human study (Yuan et al. 2018), elevated concentrations of hs-CRP and decreased levels of SOD in patients with SCZ were found, and these alterations were more pronounced following RIS treatment, proving that both oxidative stress and inflammation may be responsible for metabolic malfunctions. No other research evaluated the role of gut permeability in systemic inflammation. Based on data included in this systematic review, we postulate that the assessment of intestinal permeability may serve as a surrogate marker of both gut dysbiosis and metabolic alterations. This should be verified in well-controlled trials in obese individuals. However, Davey et al. (2012) have shown that OLZ administration was associated with an inflammation in female mice that can suggest that intrinsic properties of this agent may directly alter inflammatory mechanisms.
We conclude that metabolic disturbances during SGA treatment may be the consequence, at least in part, of gut dysbiosis. Numerous trials confirmed a beneficial effect of prebiotics and probiotics on gut microbiota composition, with a lower risk of metabolic and weight disturbances. We found only one study in which B-GOS administration attenuated OLZ-mediated weight gain independently of serotonin pathways and acted positively on gut microbiota composition when utilized alone. Therefore, we suggest further research, considering probiotic/ prebiotic/synbiotic therapy with concomitant SGA treatment. Such co-therapy may not only positively prevent or reduce weight gain but also modulate fasting glucose and glycated haemoglobin, dyslipidaemia, total and LDL cholesterol, and hypertension. Moreover, as discovered by Kao et al. (2018), prebiotics may elevate cortical glutamate receptor subunit mRNA expression (GluN1) in contrast to reductions of this receptor density typically seen in chronic SGAs users affecting their cognition (Krzystanek et al. 2015(Krzystanek et al. , 2016 negatively. Of particular interest is a search for target probiotic strains, such as Collinsella aerofaciens, which is increased in children treated with RIS without commensurate weight gains (Bahr et al. 2015a). Potential next-generation probiotic bacteria include Akkermansia, Bacteroides spp., and Eubacteriumhalli, as well as bacterial structural elements (cell wall proteins) and metabolites (e.g., SCFAs) (Romaní-Pérez et al. 2017;Muszyńska et al. 2018) This systematic review has at least four limitations. First, the number of studies identified and included in this review was low. Most of the included studies were conducted in rodent models with an unclear risk of bias. Thus, the results of these studies may not be fully extrapolated to humans. Second, although reliable molecular techniques were used for microbiota analyses in all of the included studies, none of the research used intestinal samples. Therefore, it is impossible to conclude the microbiota composition in various parts of the gastrointestinal tract. Third, none of the human studies were randomized and placebo controlled. Notably, SGAs have also been found to affect food intake habits (Reynolds and McGowan 2017) and, consequently, microbiota composition (Turnbaugh 2017). Although such a relationship was demonstrated in one study, and in female rodents only (Davey et al. 2012), detailed observations should be the subject of further research. Thus, our conclusions need to be cautiously considered.
In conclusion, this systematic review proves that alterations in the gut microbiota composition causing low-level inflammation and decreased energy expenditure can play a role in body weight gain during SGA treatment. Experimental research targeting the gastrointestinal microbiota to discover the exact mechanism of SGA action associated with poor metabolic outcomes and controlled prospective human studies should be initiated and followed. We truly believe that experimental and clinical studies should include an assessment of intestinal barrier integrity, markers of the generalized inflammatory process (e.g., LPS), and the effect of gut microbiota modifications (prebiotics, probiotics, antibiotics) on metabolic side effects of SGAs. Lastly, studies in the field of metabolomics should also be the next step in such experiments to. Due to individual composition and function of gut microbiota, the content of the microbiome metabolites, e.g., SCFAs or secondary bile acids, which play an important role in metabolic and cardiovascular health, would comprehensively decipher the impact of SGA-induced dysbiosis on human metabolism (Alemán et al. 2018;Chambers et al. 2018).