The microbiome comprises a dynamic ecological community of commensal microorganisms that inhabit our body where it interfaces with the environment. These specific microbes, which are collectively referred to as the microbiota, consist of bacteria, viruses, fungi, and protozoa; approximately equal our own cells in number; and combined pose over 200 times the number of genes as the human genome (1, reviewed in 2). Recent advances in high-throughput genetic sequencing and computational abilities reveal the richness, complexity, and essential role of the microbiome in human health. Its composition varies by anatomic region, with the gut microbiome in the distal large intestine considered the most influential for health.

After being seeded at birth by maternal vaginal bacteria in the birth canal, the neonate gut microbiota develops in a phasic manner, largely due to feeding. The gut is initially colonized by microaerophilic Proteobacteria and facultative anaerobic Actinobacteria, which consume oxygen and create a suitable niche for subsequent obligate anaerobes like Bacteroides, Clostridium, and Bifidobacterium spp. (3, 4) Breast milk stimulates the growth of bifidabacteria, but weaning results in the emergence of Firmicutes and Bacteroidetes (5). These phyla proliferate with the introduction of solid foods and eventually come to dominate the gut microbiota (5). By 2.5 to 3 years of age, the infant gut microbiota structure stabilizes and resembles the adult gut microbiota, which is also dominated by Firmicutes and Bacteroidetes (3). The developmental dynamics of the infant gut microbiota are shaped by host genes, host immunity and environmental factors, such as diet, medications, and climate (6–8).

Over the last decade, it has emerged that the human microbiome highly influences the development of the central nervous system (CNS) and the immune system. The microbiome is shaped by stress exposures from early life and, in turn, influences stress responsivity (9). Given this new information, our models of the endocrine modulation of the stress response should be updated to account for the microbiome.

The bidirectional influence of the gut microbiome and CNS occurs through the “gut-brain axis” (GBA), components of which include the vagal nerve, gut hormone signaling, immune system, tryptophan metabolism, and microbial metabolites, such as short-chain fatty acids (reviewed in 10). Activity along the GBA intersects with the HPA axis ( Figure 1 ) and may influence many psychiatric disorders, as evidenced by the association of gut dysbiosis with autism, depression, and anxiety disorders as well as functional gastrointestinal disorders (11–16). Given the purported inflammatory underpinnings for schizophrenia and its severe comorbidities with other microbiome-linked metabolic diseases, associations between schizophrenia and the microbiome are of great interest.

The neurobiology of the stress cascade and its potential for toxicity is well described. The HPA axis is the stress response system through which stress hormones and the CNS interact. Early dysregulation of the HPA axis is associated with adult stress-related disorders, including schizophrenia (18–20). Mechanistically, HPA axis dysregulation is considered to be linked to schizophrenia risk via glucocorticoid (GC) overproduction, especially during vulnerable phases of neurodevelopment. Cortisol-releasing hormone (CRH) is released from the paraventricular nucleus of the hypothalamus following physical or psychological stressors. CRH binds receptors on the anterior pituitary gland, driving release of adrenocorticotropic hormone (ACTH). This stimulates the adrenal cortex to release cortisol, the human GC hormone. Under physiological conditions increasing cortisol levels inhibit CRH release, terminating this stress cascade through a negative feedback loop. However, excess and chronic stress hyperactivate the HPA axis and cause abnormally high GC levels (21–24).

The effect of elevated GC levels on the hippocampus, the essential structure for memory and contextualizing new information, may be relevant. The hippocampus is the most commonly abnormal brain region in groups of schizophrenia cases, with progressive hippocampal volume loss a common observation (25). Increased activation, metabolism, and inflammation of the anterior hippocampus are associated with psychotic symptoms (26, 27) (reviewed in 28). The hippocampus possesses a high concentration of GC receptors that promote threat appraisal and help organize the stress response. Increased GC levels may drive overactivation and inflammation of the hippocampus and thereby promote schizophrenia (reviewed in 29–31).

GCs may also influence schizophrenia through interaction with neurotrophin pathways. Neurotrophins are growth factors responsible for neuron growth, differentiation, and formation of new synapses (32). Brain-derived neurotrophic factor (BDNF), the most abundant neurotrophin, is highly active in the hippocampus, cortex, and basal forebrain, where it binds its receptor, tyrosine kinase receptor B (TrkB), to play a key role in synaptic plasticity and long-term memory formation (33). Because GC receptors and TrkB are co-expressed in the hippocampus, important crosstalk between GCs and BDNF occurs here, as threat appraisal relies on both current stress and appropriate context from memory (34). As such, GC and BDNF equilibrium remains crucial for stress response regulation throughout life. Impairment of GC receptors and TrkB in the hippocampus favors vulnerability to stress-related disorders, including schizophrenia (reviewed in 35).

These pathways are influenced by the microbiome. Gut dysbiosis can indirectly influence cortisol release and sensitivity via chronic cytokine-mediated inflammation (36–38). This proinflammatory state may be driven by microbes crossing the intestinal barrier, releasing microbial byproducts such as lipopolysaccharide (LPS), or be moderated through bacterial metabolites, such as short-chain fatty acids (39–43) (reviewed in 44). The microbiome further influences the structure and function of the amygdala, which is critical for emotion learning and social behavior, especially responses linked to anxiety and/or fear (45, 46). Studies of germ-free mice show that the absence of the microbiome during early critical developmental windows leads to chronic cortisol elevation and altered hippocampal BDNF levels (17, 47). Depleting the microbiome of previously healthy mice through antibiotics disrupts the HPA axis (36, 48, 49). Taken together, these findings suggest that a healthy microbiome is an important component of HPA axis development and that early alterations of the microbiome can affect neuroendocrine pathways throughout life.

Identifying factors to increase resilience against stress is an area of active research that may be addressed through MGBA research. Anxiety and depression-like symptoms in germ free animas as well as the transference of a depression phenotype from a human patient to a rats through fecal microbiota support the feasibility of this approach (50). Mice deficient in the CRH₁ receptor and those with increased GR activity display more resilient behaviors (51–54) and these hormones can be modulating by the gut microbiome (11, 50). Likewise, the expression of serotonergic, glutamatergic, and GABA, which are dysregulated in association with poor resilience (55), are modulated through microbiome effects in animal models (reviewed in 56). A healthy microbiome may also contribute to resilience through emotion regulation that manifests as positive emotions and optimism, cognitive flexibility, and healthy interpersonal function, attributes that are associated with active coping styles (reviewed in 57). There may be treatment role for nutritional supplementation, as stress-related behaviors and HPA dysfunction in socially isolated male mice was remedied by dietary supplementation with DHA (58) and a rat study even demonstrated that stress sensitivity from early life trauma might be remediated through long-term supplementation with an eicosapentaenoic acid (EPA)/DHA mixture (59). The overlap of findings on the M-GBA with neuroendocrine and behavioral measures with those implicated for resilience indicate opportunities to modify the impact of stress exposures and augment resilience by targeting the microbiome.

In 1934, Rosanoff and colleagues published “The Etiology of So-Called Schizophrenic Psychosis” in the American Journal of Psychology (60). This manuscript, which examined 142 pairs of twins either concordant or discordant for schizophrenia, was the first to associate birth complications with schizophrenia. In subsequent decades, schizophrenia risk during pregnancy, birth, and the neonatal period was broadly examined. Many risk factors were identified that occurred in important early developmental stages, including maternal infection, stress, and medical complications during pregnancy, fetal growth restrictions, and hypoxia during pregnancy and birth. Overall, early-life exposures have the greatest impact on the development and function of central neural circuits and the immune system (46).

Missing from this well-developed story is the impact of maternal exposures on her microbiome and the potential for vaginal dysbiosis (61, 62). The newborn’s gut microbiome is seeded by the maternal vaginal microbiome during passage through the vaginal canal (8, 63). Disruptions in maternal microbiome may cause the newborn to be seeded with a more inflammatory gut microbiome (64, 65). It is this newborn microbiome that appears to have a strong influence in driving the development of the immune system and directing neurodevelopment (17, 66–70). These important contributions to fetal development must now be included is considering the action of schizophrenia risk factors in the perinatal period.

Maternal infection during pregnancy is associated with the risk for schizophrenia and is a maternal stressor. A 1988 study reported an increased rate for persons who were in utero during the 1957 influenza epidemic (71). Subsequent studies replicated this finding and suggested the second trimester as the gestational risk period for schizophrenia from influenza infection, although other evidence points to the first trimester (72–74). Other maternal infections associated with the offspring’s risk for schizophrenia include rubella, varicella zoster virus, herpes simplex virus, and Toxoplasma gondii, known as TORCH agents, which can cross the placental barrier and directly infect the fetus, as can measles, polio, bacterial bronchopneumonias, and infections of the genitals and reproductive tract (75–77). Taken as a whole, infection with this group of pathogens during pregnancy is relatively common and may be an important factor for psychiatric disorder risk.

As to mechanism, there are several possibilities. One of these is direct invasion, which is consistent with the very high rate of schizophrenia following prenatal rubella, up to 20%, given rubella’s well-known propensity for neural invasion in the developing fetus (76). Supporting invasion, a mouse model of influenza infection showed persistence of influenza RNA in the brains of offspring of infected pregnant mice (78). Another possibility is indirect damage driven by maternal inflammation. During maternal infection, inflammatory cytokine levels are elevated (75) and these may disrupt fetal neurodevelopment and potentially drive schizophrenia risk. For instance, the proinflammatory cytokine IL-1β negatively regulates hippocampal neurogenesis, suggesting a possible mechanism through which chronic inflammation could affect schizophrenia susceptibility (79). Notably, maternal inflammation correlates with later childhood psychiatric symptoms (80). Other potential risk pathways include effects from maternal fever, maternal antibodies crossing the placenta and medications, such as analgesics and anti-inflammatories, taken by the mother during infection, all of which may impact fetal neurodevelopment (81–83).

However, maternal infections also alter her microbiome, potentially leading to increased production of inflammatory products released by her gut, as well as to disrupted seeding of the neonatal microbiome at birth (64, 65). Neonates born to mothers with ongoing HIV infection show decreased gut microbiome diversity including reduced levels of Prevotella, a bacterial genus linked to inflammatory regulation of stressor (84). It is possible that dysbiosis secondary to maternal infection sensitizes the neonate to further stress-related injury, including elevated schizophrenia risk. Given the data demonstrating the impact of maternal inflammation on offspring schizophrenia risk and the microbiome’s potential contributions to this inflammation, the microbiome may be a key player in schizophrenia pathogenesis.

Maternal stressors, such as depression, unwanted pregnancy, death of a partner, and exposure to war and disasters, are associated with schizophrenia in offspring (19, 85–88). For female fetuses, these external stressors are most strongly correlated with schizophrenia when they occur during the first trimester; however, male fetuses demonstrated increased schizophrenia risk through the second trimester, suggesting sex differences in critical periods (87, 88). Importantly, maternal stress during the first six months of postnatal life is associated with worse behavioral outcomes in children, suggesting that disrupted caregiving may also be a component to the schizophrenia risk posed by maternal stress (89). Additionally, prenatal nutritional deficiencies, including gross calorie deficits during famine and micronutrient deficiencies in homocysteine and vitamin D, are associated with both schizophrenia and the above-mentioned stressors (90–94), which certainly impact the microbiome composition. The short-term effects of maternal stress may act through adverse pregnancy outcomes, while the long-term effects on neurodevelopment may involve altered neonatal stress programming and gut dysbiosis (95). Maternal stress increases fetal and neonatal exposure to maternal cortisol, altering growth and behavior in humans and animal models (reviewed in 96). Stress also has well-documented effects on the microbiome, which may in turn alter inflammation and neurodevelopment in a developing neonate (62, 97–103). As an example, maternal perinatal stress increases offspring susceptibility to allergic diseases, which suggests interactivity between maternal GCs, perinatal immune development, and possible maternal dysbiosis (79). In a mouse model, prenatal maternal stress led to dysbiosis in both mother and offspring, increased IL-1β in utero, and a corresponding decrease in BDNF in offspring (104). Other experiments have shown antibiotics alter BDNF levels in dysbiotic mice, suggesting that interventions in the gut microbiome may be important in modifying risk (105).

Exploring how the maternal stress influences her microbiome for fetal effects relevant to schizophrenia risk may enhance our understanding of the disease and suggest new treatments or prophylactics through probiotic use (reviewed in 106). Mechanistically, the microbiome-driven effects of stress may manifest through alterations of the HPA axis during key developmental stages (107), impaired development of small intestine immune tissue and IgA production (108, 109), or alterations in gut-metabolites leading to aberrant development (110). Given that many of these downstream events are linked with schizophrenia risk, future work should aim to elicit the microbiome contributions of schizophrenia risk secondary to maternal stress.

Many obstetric complications can lead to fetal hypoxia, which carries well-known risks to medial temporal regions. With regards to schizophrenia, fetal hypoxia may be the most significant risk factor among obstetric complications, in addition to maternal infections and fetal growth restriction (111, 112). Multiple studies report increased exposure to fetal hypoxia among persons with schizophrenia (113–115). One study show fetal hypoxia predicts the risk for early onset schizophrenia even after controlling for prenatal infection and fetal growth restriction (116). Further, fetal hypoxia is associated with reduced gray matter and ventricular enlargement in cases with schizophrenia and their non-ill siblings, although not in unrelated controls (117). Mechanistically, hypoxia may have an additive effect with genetic factors hastening the onset of schizophrenia in susceptible individuals (118). Certainly hypoxia may influence the composition and function of the gut microbiome (119, 120). As described above with infection and stress, these alterations increase future susceptibility to stress by influencing systemic inflammation, stress pathways, and BDNF production. Additionally, maternal microbes may invade the fetal brain following a hypoxic episode, as has been shown in sheep (121).

In 1966, a small but significant reduction in birth weight was observed in schizophrenic patients when compared to their siblings (122), prompting consideration that fetal growth restriction was a schizophrenia risk factor. Some, but not all studies associated lower birth weight, reduced head circumference, and congenital malformations with increased schizophrenia risk (123). There are heterogeneous causes of fetal growth restriction, only some of which may be associated with the risk for schizophrenia (124).

Other perinatal obstetric complications include maternal bleeding, maternal diabetes, preeclampsia, and caesarean section birth complications (125–127) (reviewed in 128). These perinatal traumas—along with the aforementioned factors of maternal infection, maternal stress, fetal hypoxia, and fetal growth restriction—altogether present a compelling argument for a close connection between the early window of neural development and schizophrenia risk. Recent advances indicate that the vaginal microbiome suggest that it may be a key player in this relationship. After all, these traumas occur during the perinatal period, when initial microbiota seeding of the newborn’s gut by the maternal vaginal microbiome occurs during fetal passage through the birth canal.

Cortisol, the primary human “stress” hormone, is also of central relevance for the developing fetus, promoting the maturation of vital organs, including the lungs, gastrointestinal tract, liver, heart, and brain. As such, the fetal HPA axis is tightly regulated, and is responsive to minute changes in fetal plasma levels of cortisol, which easily crosses the placental barrier (reviewed in 129). Due to their high cortisol sensitivity, developing fetuses rely on the placental enzyme 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2) to inactivate maternal cortisol by converting it to less active cortisone, beginning in the second trimester (130). Thus, in early gestation, before placental 11β-HSD2 is induced, maternal hypercortisolemia has potent effects on developmental gene expression. Even after the induction of 11β-HST2, some cortisone can be reactivated through 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2), which converts cortisone back to cortisol (reviewed in 131). This effect can be heightened by factors like maternal protein malnutrition, which diminish 11β-HSD2 gene expression (132). The detrimental effects of elevated exposure to maternal cortisol go beyond fetal development to influence emotional and behavioral disturbances during infancy and childhood and in later life (85) possibly including the perinatal schizophrenia risk pathways described above (reviewed in 133). Beyond the direct association between maternal stress and schizophrenia, elevated maternal cortisol may enhance other risks. In one study, elevated maternal cortisol during the second trimester enhanced the risk for adolescent onset depression in the offspring of mothers who experienced infections during pregnancy (134).

The neurotrophin BDNF is also critical for neurodevelopment. Elevated levels of BDNF are reported in fetuses with severe growth restriction as well as those with macrosomia in the context of maternal diabetes (135). Mechanistically, it is proposed that BDNF is neuroprotective in the developing fetus through anti-inflammatory mechanisms (136). In vivo animal models demonstrate that BDNF can reduce hypoxic brain injury through modulation of inflammatory cytokines and promotion of microglial activation (137). Given BDNF’s protective role in the developing brain, it is possible that downregulation of BDNF could exacerbate schizophrenia risk in the perinatal window.

Colonization of a newborn neonate gut is normally seeded by the vaginal microbiome during birth, as described, along with maternal vaginal, skin, and oral and fecal bacterial strains (8, 63, 138). These vaginal contributions are transient and by four months post-birth, the infant’s gut microbiome is more similar to the maternal gut microbiome ( Figure 2 ) (139, reviewed in 140). Neonates born via caesarean section lack exposure to the maternal vaginal microbiome and demonstrate a higher prevalence of maternal oral and skin microbes. They are also more likely to develop immune-related disorders (8, 64, 65, 141, 142). Disruptions of the maternal vaginal microbiome via infection, stress, or other pathways may lead to neonatal dysbiosis (65, reviewed in 143). Pre-term birth, caesarean sections, steroid use, and antibiotic use are also associated with dysbiosis in the newborn infant (144, 145).

The initial development of the microbiome, including its seeding at birth and development through very early childhood, is important for the development of a healthy core microbiome that is resistant to later perturbation. Given that dysregulation of the microbiome can cause pathogenic inflammation, dysbiosis in the perinatal window may lead to long-term inflammatory dysregulation (146–148). Further studies are needed to determine how maternal flora may influence immune development and schizophrenia risk in their offspring.

Childhood onset of schizophrenia is rare, but a number of neurologic and psychiatric features are already present in childhood. Likewise, a number of traumatic exposures in childhood are associated with schizophrenia risk. The possibility that these presentations reflect the interactions of stress signaling and neurotrophic effects that may be influenced or modulate by the microbiome effects, which is currently being explored.

As with early childhood, broad impairments in neuromotor development, cognitive function, and behavior often mark individuals at risk for schizophrenia (reviewed in 154). As the adolescent matures, poor coordination, balance, and perceptual-motor and visual-motor functioning may become more apparent in a subgroup of cases (152, 168, 173). Cognitively, lower intelligence and especially a decrease in intellectual function mark schizophrenia risk (151, 153, 169, 244). There is impairment of individual domains including arithmetic and spelling, formal thought disorders, attention difficulties, increased distractibility, poor executive functioning, and general learning and memory difficulties (152, 153, 169, 173, 245). Behaviorally, aggression, withdrawal, and generally poor social competence and peer relations are also concerning, with psychiatric symptoms including affective flattening and anxiety often present (149, 151, 174, 175, 246, 247).

As discussed earlier, studies of exposures do not usually distinguish between pre-pubescent children and post-pubescent adolescents. The aforementioned exposures of sexual, physical, and emotional abuse as well as infection similarly convey schizophrenia risk among adolescents. However, trauma may have different long-term outcomes post-puberty, and its potential effect on schizophrenia risk merits further study. Additionally, the increased risk-taking behavior exhibited at this stage may be influenced by early trauma and influence further trauma exposures. New exposures, such as recreational drug use, may also contribute to schizophrenia risk.

Recreational drugs exploration is frequent in adolescence and many carry a significant risk for psychosis, particularly cannabis. By their first psychotic episode, approximately half of patients will have a history of cannabis use and one-third meet criteria for cannabis use disorder (248). Alcohol use is similarly high among individuals who have experienced their first psychotic episode, and there is elevated use of cocaine, amphetamine, barbiturate, and other drugs. Cause and effect associations of cannabis and psychosis are well described, although some schizophrenia-susceptible individuals may self-medicate to reduce the anxiety surrounding the presentation of schizophrenia symptoms, with this drug-seeking behavior may further exacerbate their risk for the disorder (249). Chronic exposure to tetrahydrocannabinol (THC), an active ingredient in cannabis, can disrupt neurodevelopmental maturation dependent on endocannabinoid pathways and may lead to overactivation of a pro-hallucinogenic pathway of 5-HT2A receptors, which may promote schizophrenia onset in susceptible individuals (250).

Substance abuse can dysregulate the HPA axis. Alcohol and nicotine use induce cortisol production, and long-term use can cause chronic cortisol elevation and dysregulation similarly to previously described trauma (251–254). Additionally, the gut microbiome is dysregulated by psychostimulants, alcohol, and opioids (255–259) (reviewed in 260). Microbiome influences on addiction are an active area of research. Microglial function is shaped by the microbiome and altered by drugs of abuse (233, 261). Likewise, BDNF dysregulation by dysbiosis is associated with altered behavioral response to cocaine and alcohol (256, 262, 263). While more work is needed to establish causal relationships, these findings suggest multiple ways in which the microbiome may influence addiction behaviors.

While a number of risk factors for schizophrenia are identified, recent evidence points to protective factors. Specifically, aerobic exercise (AE) is hypothesized to play an important protective role against stress induced effects. AE induces a cascade of molecular and cellular processes that support brain plasticity and growth of new vasculature and trigger the processes through which neurotrophins mediate neural plasticity (reviewed in 275–278). Among neurotrophins, BDNF is the most susceptible to regulation by physical exercise (279–281), with synthesis and release into the blood circulation increasing in a dose-response manner (282, 283). Consistent with these findings, Voss et al. (284, 285) found increased connectivity between the bilateral parahippocampus and the bilateral middle temporal gyrus was linked to BDNF increase in AE subjects. A recent meta-analysis (286) of 29 studies (N = 1111 healthy subjects) examined the effect of exercise on BDNF in three exercise paradigms: 1) a single session of exercise; 2) a session of exercise following a program of regular exercise; and 3) resting BDNF levels following a program of regular exercise. Results demonstrated a moderate effect size for increases in BDNF following a single session of exercise (Hedges’ g = .46, p < .001). Further, regular exercise intensified the effect of a session of exercise on BDNF levels (Hedges’ g = .59, p = .02). Finally, results indicated a small effect of regular exercise on resting BDNF levels (Hedges’ g = .27, p = .005). Examination of moderator effects across paradigms found that subjects’ age was not significantly related to changes in BDNF following exercise, but sex significantly moderated the effect of exercise on BDNF levels, such that studies with more women showed less BDNF change resulting from exercise.

Consistent with these reports, findings indicate individuals with schizophrenia tend to have highly sedentary lifestyle characterized by low aerobic fitness which was highly correlated with poor cognitive functioning and symptoms (287). These findings parallel reports among individuals at clinical high risk for psychosis indicating lower levels of fitness, less physical activity, as well as more barriers to exercise (288–292). Yet, a pilot AE RCT indicated engagement in AE led to 11.0% increase aerobic fitness (293) as well as BDNF vs. a 1.9% in the TAU subjects (294) (reviewed in 295). A hierarchical multiple regression analysis indicated that, after controlling for age, sex, changes in anti-psychotic and SSRIs, and changes in menstrual cycle phase, BDNF changes independently predicted changes in cognitive function (b = .38, t = 2.06, p = .05) (296). Notably, improvements in cognitive functioning were associated with intensity of AE activity (294).

Exercise alters the composition and functional capacity of the gut microbiome independent of diet (reviewed by 28). As the effects of AE on BDNF production are further studied in schizophrenia, examination of how the microbiome influences this pathway may be illuminating.