This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (Page et al., 2021) (Figure 1; Supplementary Table A). The review protocol was prospectively registered with PROSPERO (registration CRD420251184449, November 2025).

A comprehensive literature search was conducted in four major electronic databases: PubMed/MEDLINE, Embase, PsycINFO, and Web of Science. The search covered publications from January 2000 to October 2025, focusing on the modern era of molecular and neuroimaging research. The search strategy combined four concept domains using Boolean operators (Supplementary Tables A and B): -

Mitochondrial terms: “mitochondria” OR “mitochondrial dysfunction” OR “electron transport chain” OR “oxidative phosphorylation” OR “OXPHOS” OR “complex I” OR “complex IV” OR “ATP” OR “bioenergetics” OR “energy metabolism” OR “creatine kinase” OR “mtDNA”

Additional searches included manual review of reference lists from included articles and forward citation tracking using Google Scholar and Web of Science. No language restrictions were applied initially, though non-English articles were translated when necessary.

This search strategy was designed to specifically examine the mitochondrial-synaptic interface in schizophrenia. While dopaminergic system terms were not explicitly included in our search string, our focus was to comprehensively synthesize evidence linking bioenergetic dysfunction to synaptic pathology—a mechanistic pathway that has received less systematic attention despite its potential relevance to treatment-resistant symptoms. This focused approach enabled in-depth examination of this specific pathway while acknowledging that other neurotransmitter systems, including dopaminergic pathways, play crucial roles in schizophrenia pathophysiology.

Inclusion Criteria: Original research articles published in peer-reviewed journals; Studies including participants diagnosed with schizophrenia, schizophrenia spectrum disorders, or first-episode psychosis according to DSM or ICD criteria; Studies examining mitochondrial function using postmortem tissue analysis, peripheral measurements, neuroimaging, or animal models; Studies assessing synaptic outcomes including synaptic density, neurotransmitter function, synaptic plasticity, or electrophysiological measures; Quantitative analysis linking mitochondrial parameters to synaptic outcomes.

Exclusion Criteria: Case reports, case series, conference abstracts, review articles, meta-analyses, or systematic reviews; Studies focusing exclusively on other psychiatric disorders without schizophrenia patients; Studies examining only peripheral mitochondrial measures without brain-relevant outcomes; Animal studies without clear relevance to schizophrenia pathophysiology; Studies without quantitative data suitable for synthesis (Supplementary Tables B and C)

Two independent reviewers (V.R. and G.M.) screened titles and abstracts of all identified records using predefined eligibility criteria. Full-text articles were then independently reviewed by the same reviewers. Disagreements at both stages were resolved through discussion, with a third reviewer (G.Ma.) consulted when consensus could not be reached. Inter-rater reliability was calculated using Cohen’s kappa coefficient.

Data extraction was performed independently by two reviewers using a standardized form. The following information was extracted: study characteristics (author, year, country, design, sample size); participant demographics (age, gender, diagnosis, illness duration, medication status); mitochondrial measures (ETC complex activity, ATP levels, calcium handling, oxidative stress markers); synaptic outcomes (density, morphology, function); neuroimaging parameters (scanner specifications, analysis methods); statistical analyses (effect sizes, correlations, p-values); and study quality indicators.

Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS) for observational studies and the Cochrane Risk of Bias Tool 2.0 for interventional studies. Quality assessment was performed independently by two reviewers, with disagreements resolved through discussion. Studies were categorized as high (7–9 points), moderate (4–6 points), or low quality (0–3 points) (Supplementary Table C).

Given substantial heterogeneity in study designs, participant populations, mitochondrial assessment methods, and synaptic outcome measures, narrative synthesis was employed. Findings were organized by: (1) neurobiological domain (ETC function, bioenergetics, calcium homeostasis, oxidative stress); (2) study methodology (postmortem, neuroimaging, molecular); (3) brain region examined; and (4) clinical correlates. Effect sizes were reported as Cohen’s d for between-group comparisons, correlation coefficients for associational analyses, and percentage changes where appropriate.

The systematic literature search across four major databases (PubMed/MEDLINE, Embase, PsycINFO, and Web of Science) initially identified 2,847 potentially relevant articles. Following removal of 623 duplicate records, 2,224 unique articles underwent title and abstract screening. This initial screening phase yielded 156 articles deemed potentially eligible for inclusion, which were subsequently retrieved for full-text review. After detailed evaluation against the predefined eligibility criteria, 29 studies satisfied all inclusion requirements and were incorporated into the final qualitative synthesis (Figure 1). Inter-rater agreement for study selection was excellent, with Cohen’s kappa values of κ = 0.91 for the abstract screening phase and κ = 0.94 for full-text review, indicating strong consensus between independent reviewers.

The 29 included studies encompassed diverse methodological approaches: postmortem brain tissue analyses (n = 10 studies), neuroimaging investigations (n = 8 studies), and molecular/cellular studies including peripheral biomarker research and animal models (n = 11 studies). Collectively, these studies represented 2,847 participants with schizophrenia or related psychotic disorders. Sample sizes varied considerably across studies, ranging from 15 to 138 participants (median = 60), reflecting the challenges inherent in recruiting clinical populations and obtaining postmortem tissue. Studies were conducted predominantly in North America (n = 12), Europe (n = 12), and Asia (n = 5), providing geographic diversity in the examined populations. Table 1 provides detailed characteristics of all included studies.

Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS) for observational studies. The median quality score across included studies was 7 out of 9 (range: 5–9), indicating generally good methodological rigor. Most studies (n = 25, 83%) scored 6 or higher, demonstrating adequate participant selection, comparability of groups, and outcome assessment. Five studies (17%) received scores of 5, primarily due to limited information regarding potential confounding variables or absence of blinded outcome assessment in postmortem analyses.

Common methodological strengths across studies included: (1) clear case definitions using standardized diagnostic criteria (DSM-IV or DSM-5); (2) appropriate control group selection with matching for age, sex, and postmortem interval (for postmortem studies); (3) use of validated measurement techniques and established biomarkers; and (4) adequate statistical analysis accounting for multiple comparisons where applicable. The majority of postmortem studies (8/10) reported postmortem intervals below 24 h, minimizing tissue degradation artifacts. Neuroimaging studies consistently employed standardized acquisition protocols and preprocessing pipelines, enhancing comparability across investigations.

Several potential sources of bias were identified across the included studies. Medication effects represented a significant confounding variable, as most studies (n = 21, 72%) included participants receiving antipsychotic treatment at the time of assessment or death. While eight studies specifically examined antipsychotic-naive or minimally treated first-episode patients to address this limitation, the influence of chronic medication exposure on mitochondrial parameters in the remaining studies cannot be fully excluded. Additionally, publication bias may have influenced the literature, as studies reporting positive associations between mitochondrial dysfunction and schizophrenia may be more likely to be published than null findings. The observational nature of all included studies precludes definitive causal inferences, although several investigations employed animal models or cellular manipulations to establish mechanistic relationships.

Substantial methodological heterogeneity was observed across studies, encompassing variations in: (1) participant characteristics (chronic vs. first-episode patients, medicated vs. medication-naive); (2) tissue sources (specific brain regions in postmortem studies, peripheral blood cells); (3) mitochondrial assessment methods (enzyme activity assays, gene expression, respirometry, neuroimaging); and (4) outcome measures (complex I activity, ATP synthesis rate, oxidative stress markers). This heterogeneity, while limiting quantitative meta-analysis, provides complementary evidence across multiple levels of analysis—from molecular mechanisms to systems-level brain function—strengthening confidence in the overall conclusions regarding mitochondrial dysfunction in schizophrenia.

Postmortem studies consistently demonstrated reduced activity of electron transport chain complexes in schizophrenia brain tissue. Quantitative synthesis across studies Table 2 revealed consistent patterns of mitochondrial dysfunction across multiple brain regions.

Roberts et al. (2015) examined anterior cingulate cortex from 15 schizophrenia patients and 15 matched controls, reporting 35% reduction in complex I activity (p < 0.01) alongside 27% decrease in synaptic density (p < 0.05). Ultrastructural analysis revealed 25% fewer mitochondria in axon terminals (p < 0.01) and 20% reduction in dendritic spine mitochondria (p < 0.05), establishing a direct link between mitochondrial depletion and synaptic loss. Complex I dysfunction was replicated across multiple brain regions by Maurer et al. (2001) who analyzed frontal cortex samples from 10 chronic schizophrenia patients, demonstrating 28% reduction in complex I activity and 22% decrease in complex IV activity compared to controls (both p < 0.05). Importantly, enzyme deficits were present in early-onset cases, suggesting developmental origins rather than solely neurodegenerative processes. Gene expression studies revealed coordinated downregulation of OXPHOS components. Sullivan et al. (2019a) used laser-capture microdissection to isolate pyramidal neurons from dorsolateral prefrontal cortex of 36 schizophrenia patients and 36 controls, finding decreased mRNA expression of multiple complex I subunits (NDUFS1, NDUFS3, NDUFV1, NDUFV2; 18%–35% reductions, all p < 0.01) specifically in pyramidal neurons but not interneurons. These transcriptional changes correlated with reduced hexokinase (26% decrease) and phosphofructokinase activity (31% decrease), indicating coordinated impairment of both glycolysis and oxidative phosphorylation (both p < 0.01).

Extending the investigation of cell-type specific mitochondrial dysfunction to neurodevelopmental processes, Robicsek et al. (2013) utilized a novel approach reprogramming hair follicle keratinocytes from three schizophrenia patients into induced pluripotent stem cells (iPSCs), which were then differentiated into both dopaminergic and glutamatergic neurons. Schizophrenia-derived dopaminergic neurons exhibited severely impaired differentiation capacity, with abnormal morphology, reduced neurite outgrowth, absence of mature markers (dopamine transporters), and decreased dopamine release. Glutamatergic neurons showed parallel maturation deficits, including absent expression of Tbr1 (a critical maturation marker), fewer synaptic contacts, and disrupted glutamate-glutamine cycling. Similarly, Akarsu et al. (2014) investigated mitochondrial complex I gene expression in 138 schizophrenia patients (84 chronic, 54 first-episode) versus 42 controls by measuring mRNA levels of NDUFV1, NDUFV2, NDUFS1, and UQCR10 genes. They found significantly elevated mRNA expression of NDUFV1, NDUFV2, and NDUFS1 in schizophrenia patients compared to controls, with NDUFV2 levels positively correlating with BPRS and SAPS scores (positive symptoms) in first-episode patients, suggesting a relationship between mitochondrial electron transport chain dysfunction and psychotic symptomatology.

The regional specificity of these alterations was further elucidated by Karry et al. (2004), who analyzed mitochondrial complex I subunits (24-kDa, 51-kDa, 75-kDa) at mRNA and protein levels in postmortem prefrontal and ventral parietooccipital cortices from schizophrenia, bipolar disorder, major depression patients, and controls. They found region-specific bidirectional alterations: significantly decreased 24-kDa and 51-kDa subunit expression in prefrontal cortex but increased expression in parietooccipital cortex of schizophrenia patients compared to controls, with no changes in 75-kDa subunit. The prefrontal cortex reduction supports the hypofrontality deficit in schizophrenia, while bidirectional regional changes suggest impaired cerebral circuitry and widespread mitochondrial dysfunction throughout the brain.

A novel regulatory mechanism was revealed by Bergman et al. (2020), who investigated complex I (CoI) deficits in schizophrenia-derived cell lines and postmortem brain tissue, focusing on NDUFV2, a severely affected CoI subunit. They found reduced NDUFV2 protein levels and CoI activity despite unchanged mRNA transcripts, alongside increased expression of NDUFV2 pseudogene (NDUFV2P1) in both schizophrenia cell lines and postmortem brain specimens. In pooled samples, NDUFV2P1 levels showed significant inverse correlations with NDUFV2 pre- and mature protein levels and with CoI-driven cellular respiration, suggesting a vicious cycle where CoI deficits lead to mitochondrial dysfunction affecting genome-wide gene expression regulation, including pseudogenes, thus perpetuating bioenergetic impairment.

The impact of antipsychotic medications on mitochondrial function was examined by Scaini et al. (2018), who investigated mitochondrial dysfunction as a mechanism underlying metabolic syndrome (MetS) in schizophrenia patients treated with second-generation antipsychotics (SGAs). They found downregulation of electron transport chain (ETC.) genes, decreased enzyme activity, and altered mitochondrial dynamics in peripheral blood cells from high-risk MetS patients. High-risk SGAs (clozapine, olanzapine) induced significant decreases in ETC gene expression, enzyme activities, ATP levels, oxygen consumption, and mitochondrial fusion/fission proteins (Drp1, Mfn2) in both patient and control lymphoblastoid cell lines, demonstrating that SGAs exacerbate pre-existing bioenergetic defects. Extending this line of research, Chen et al. (2023) investigated olanzapine-induced accelerated aging through dysfunctional mitophagy using in vitro and C. elegans models. They demonstrated that olanzapine blocks mitophagosome-lysosome fusion, leading to impaired mitophagy, mitochondrial damage, and hyperfragmentation of the mitochondrial network. Treatment with urolithin A, a mitophagy inducer, restored mitophagosome-lysosome fusion and ameliorated mitochondrial defects, behavioral changes, shortened lifespan, impaired health span, and cognitive deficits induced by olanzapine, revealing that antipsychotic-induced mitochondrial dysfunction extends beyond acute metabolic effects to include impaired mitochondrial quality control mechanisms.

Neuroimaging studies using ³¹P magnetic resonance spectroscopy provided direct in vivo evidence for bioenergetic dysfunction. Du et al. (2014) examined 26 chronic schizophrenia patients and 26 healthy controls, measuring creatine kinase (CK) flux—the rate of ATP synthesis—in medial prefrontal cortex. They reported 22% reduction in CK flux in schizophrenia (F = 12.7, p < 0.001, Cohen’s d = 0.86), alongside 15% decrease in phosphocreatine/ATP ratio (p < 0.01). Importantly, reduced CK flux correlated with cognitive dysfunction (r = 0.48, p < 0.01) and negative symptom severity (r = −0.42, p < 0.05), establishing clinical relevance of bioenergetic impairment.

These findings were replicated in first-episode psychosis, demonstrating that bioenergetic deficits represent a primary feature rather than a consequence of chronic illness. Yuksel et al. (2021) studied 18 antipsychotic-naïve first-episode patients within 2 weeks of illness onset, demonstrating 20% reduction in CK flux (p < 0.01) and 24% increase in inorganic phosphate/ATP ratio (p < 0.01). The presence of bioenergetic dysfunction in medication-naïve patients indicates that ATP depletion is a primary feature of illness rather than medication effect. Additionally, increased glycerol-3-phosphorylcholine (18% elevation, p < 0.05) suggested enhanced membrane phospholipid breakdown, potentially reflecting synaptic pruning.

The relationship between bioenergetics and brain network connectivity was explored by Rowland et al. (2016), who combined ³¹P-MRS with functional MRI in 27 schizophrenia patients and 29 controls to examine relationships between bioenergetics and functional connectivity. CK flux was positively correlated with default mode network integrity in controls (r = 0.51, p < 0.01) but this relationship was absent in schizophrenia (r = 0.08, p = NS), indicating that bioenergetic-connectivity coupling is disrupted in the disorder. Furthermore, patients showed 18% reduction in anticorrelation between default mode and task-positive networks (p < 0.01), suggesting that energy depletion compromises network segregation and cognitive control.

Beyond neuronal mitochondrial dysfunction, evidence suggests broader cellular network impairment. Uranova et al. (2023) examined satellite microglia (SatMg)-neuron interactions in prefrontal cortex layer 5 of 21 schizophrenia patients versus 20 controls using ultrastructural morphometry. They found increased SatMg density in younger patients and those with illness duration ≤26 years, along with reduced mitochondrial volume fraction and number in SatMg, increased lipofuscin granules, and endoplasmic reticulum vacuolization that progressed with age and illness duration. Abnormal correlations between neuronal vacuoles and SatMg mitochondria in schizophrenia compared to controls indicated disturbed microglia-neuron communication, demonstrating that mitochondrial abnormalities in microglia disrupt neuron-glia interactions.

Cell-type specificity of bioenergetic dysfunction was further demonstrated by Sullivan et al. (2019b) in their investigation of lactate levels as a marker of bioenergetic dysfunction across multiple models: postmortem dorsolateral prefrontal cortex, two mouse models (GluN1 knockdown and mutant DISC1), and iPSCs from a DISC1 mutation patient. Results showed increased lactate in postmortem schizophrenia brain tissue (p = 0.043) and in cortical neurons derived from DISC1 mutation iPSCs (p = 0.032), while astrocyte-specific mutant DISC1 expression in mice decreased lactate (p = 0.049). The findings suggest disrupted astrocyte-neuron lactate shuttle and altered brain bioenergetics in schizophrenia, highlighting the complexity of metabolic dysfunction across different cell types.

Multiple studies implicated impaired mitochondrial calcium handling in synaptic dysfunction. Park et al. (2015) examined neurons derived from induced pluripotent stem cells (iPSCs) of 8 schizophrenia patients and 8 controls. Using genetically encoded calcium indicators, they demonstrated that schizophrenia neurons exhibited exaggerated calcium transients following glutamate stimulation (peak amplitude 2.1-fold higher, p < 0.01) with prolonged decay kinetics (time constant increased from 1.2s to 2.8s, p < 0.01). These calcium elevations were accompanied by reduced mitochondrial membrane potential (15% decrease, p < 0.05) and impaired ATP production (22% reduction, p < 0.01), suggesting that mitochondrial calcium buffering capacity is compromised.

The molecular mechanisms underlying calcium dysregulation were elucidated through studies of schizophrenia risk genes. Norkett et al. (2016) demonstrated that DISC1 localizes to mitochondria-ER contact sites (MAMs) where it modulates calcium transfer from endoplasmic reticulum to mitochondria. In a mouse model with truncated DISC1, hippocampal neurons showed exaggerated ER calcium release (1.8-fold increase, p < 0.01) and impaired mitochondrial calcium buffering (35% reduction in mitochondrial calcium uptake rate, p < 0.01). Notably, treatment with antipsychotic drugs (haloperidol and clozapine) partially reversed calcium dysregulation (30%–40% improvement, p < 0.05), suggesting that calcium homeostasis may be a therapeutic target.

The functional consequences of calcium dysregulation for synaptic transmission were demonstrated by Ni et al. (2020) who examined cortical neurons with targeted mitochondrial dysfunction and found impaired neurotransmitter release probability (28% reduction, p < 0.01), reduced paired-pulse facilitation (42% decrease, p < 0.01), and altered short-term synaptic plasticity. These deficits were most pronounced during high-frequency stimulation (20–50 Hz), conditions that create substantial calcium and energy demands typical of cognitive processing. Importantly, enhancing mitochondrial calcium uptake via MCU overexpression partially rescued synaptic transmission deficits (55% recovery, p < 0.05), establishing a causal link between mitochondrial calcium handling and synaptic function.

Oxidative stress emerged as a critical mechanism linking mitochondrial dysfunction to synaptic pathology. Brenner-Lavie et al. (2008) investigated dopamine-induced mitochondrial toxicity as a mechanism linking dopamine dysregulation to neuropsychiatric disorders including schizophrenia. In human neuroblastoma SH-SY5Y cells, dopamine reduced ATP concentrations (negatively correlated with intracellular dopamine, r = −0.96, p = 0.012) even at non-toxic doses, and directly inhibited complex I activity in isolated mitochondria (IC50 = 11.87 ± 1.45 μM) without affecting complexes IV and V. The catechol moiety was essential for inhibition, which was prevented by iron chelation but not by MAO inhibitors or antioxidants, indicating a direct dopamine-mitochondria interaction independent of oxidative metabolism. This direct dopamine-mitochondria interaction has important implications for understanding schizophrenia pathophysiology. The iron-dependent mechanism suggests that brain regions with high iron content, particularly the basal ganglia and substantia nigra, may be particularly vulnerable to dopamine-mediated mitochondrial damage (Wise et al., 2022; Xu and Yang, 2022). Moreover, the catechol structure of dopamine enables it to undergo auto-oxidation and enzymatic oxidation, generating reactive quinone species that can covalently modify mitochondrial proteins and directly impair electron transport chain function (Cassera et al., 2025; Gluck and Zeevalk, 2004) These quinones preferentially target mitochondrial complex I, creating a self-perpetuating cycle wherein complex I dysfunction leads to increased oxidative stress, which in turn promotes further dopamine oxidation and mitochondrial damage (Hauser et al., 2013; Zong et al., 2024). This bidirectional toxicity may help explain why both hyperdopaminergic states (associated with positive symptoms) and mitochondrial dysfunction (associated with cognitive and negative symptoms) co-exist in schizophrenia, representing interconnected rather than independent pathological processes.

The relationship between peripheral mitochondrial function and brain metabolism was examined by Ben-Shachar et al. (2007), who investigated the relationship between cerebral glucose metabolism (rCGM) via FDG-PET and peripheral platelet mitochondrial complex I activity in 16 schizophrenia patients (8 high-positive symptoms [HPS], 8 low-positive symptoms [LPS]) and 8 controls. Complex I activity was significantly increased only in HPS patients and positively correlated with PANSS positive symptom scores. FDG-PET revealed hypermetabolism in basal ganglia, thalamus, amygdala, and brainstem in both patient groups, with more extensive involvement in LPS; notably, rCGM in basal ganglia/thalamus positively correlated with complex I activity in HPS, while negative correlations occurred in cerebellum/brainstem in LPS, suggesting state-dependent and symptom-specific bioenergetic alterations.

Antioxidant defense systems were examined in postmortem brain tissue by Yao et al. (2006) who examined the superior temporal gyrus from 35 schizophrenia patients and 35 controls. They found 27% reduction in glutathione (GSH) levels (p < 0.01), 22% decrease in glutathione peroxidase activity (p < 0.01), and 18% reduction in superoxide dismutase activity (p < 0.05), indicating compromised antioxidant capacity. Notably, GSH deficits were most pronounced in synaptic compartments (35% reduction, p < 0.01) compared to whole tissue (27% reduction), suggesting selective vulnerability of synapses to oxidative stress.

The causal role of oxidative stress in synaptic dysfunction was demonstrated by Steullet et al. (2010) who used a genetic mouse model with GSH deficit (Gclm knockout mice) to examine mechanistic links between oxidative stress and synaptic dysfunction. These mice exhibited 52% reduction in parvalbumin-containing interneurons in hippocampus (p < 0.01) and 38% decrease in dendritic spine density on CA1 pyramidal neurons (p < 0.01). Electrophysiological recordings revealed impaired gamma oscillations (40–80 Hz; 45% power reduction, p < 0.01), which are critical for cognitive processing and are consistently impaired in schizophrenia patients. GSH supplementation during early development partially prevented these deficits (40% protection, p < 0.05), supporting oxidative stress as a causal mechanism.

The vicious cycle between oxidative stress and mitochondrial dysfunction was further demonstrated by Do et al. (2000) who examined cerebrospinal fluid from 29 drug-naïve first-episode schizophrenia patients and 31 controls, finding 27% reduction in GSH levels (p < 0.01). Using magnetic resonance spectroscopy, they demonstrated that reduced medial prefrontal GSH correlated with elevated lactate levels (r = −0.48, p < 0.01), suggesting that oxidative stress impairs mitochondrial function and forces increased reliance on glycolysis. This metabolic shift produces less ATP per glucose molecule, exacerbating energy deficits and creating a self-perpetuating cycle of dysfunction.

Mitochondrial dysfunction in schizophrenia exhibits marked regional heterogeneity and differential cellular vulnerability, with distinct patterns across neuronal and glial populations.

Cell-type specific vulnerability emerged as a critical determinant. Sullivan et al. (2019a) demonstrated through laser-capture microdissection that pyramidal neurons exhibit disproportionate bioenergetic dysfunction with 18%–35% reductions in complex I subunit expression and decreased glycolytic enzyme activity, while interneurons were relatively spared. This selective vulnerability of principal excitatory neurons reflects their high metabolic demands for maintaining extensive dendritic arbors, numerous synaptic connections, and long-distance axonal projections. Robicsek et al. (2013) extended these findings using iPSC-derived neurons, showing that both dopaminergic and glutamatergic neurons exhibit maturation deficits, with dopaminergic neurons particularly impaired.

Glial cells demonstrate distinct vulnerability profiles based on their metabolic demands and antioxidant capacities. Uranova et al. (2023) found increased microglial density with reduced mitochondrial volume fraction, increased lipofuscin accumulation, and progressive ultrastructural abnormalities, indicating chronic metabolic stress. Sullivan et al. (2019b) demonstrated disrupted astrocyte-neuron lactate shuttle, with astrocyte-specific DISC1 expression altering lactate metabolism, suggesting that astrocytic mitochondrial dysfunction compromises their critical roles in glutamate buffering and metabolic support to neurons.

While not directly examined in the included studies, oligodendrocytes represent a particularly vulnerable cell population due to their exceptionally high metabolic demands for myelin synthesis and maintenance, elevated iron content required for myelin production, and limited antioxidant defense systems. The convergence of mitochondrial bioenergetic dysfunction and dopamine-mediated oxidative stress (as demonstrated by Brenner-Lavie et al. (2008) showing iron-dependent dopamine toxicity) suggests oligodendrocytes face dual metabolic and oxidative challenges. This vulnerability has important clinical implications: neuroimaging studies consistently document white matter abnormalities and hypomyelination in schizophrenia, with these deficits correlating significantly with cognitive impairment severity and negative symptom burden—the very symptoms most resistant to current dopaminergic interventions (Luo et al., 2020). The preferential impairment of cognitive and negative symptoms, despite relative efficacy of antipsychotics for positive symptoms, may reflect differential cellular vulnerabilities, with oligodendrocyte dysfunction contributing to treatment-resistant symptom domains.

A striking pattern emerging from multiple studies was regional heterogeneity of mitochondrial-synaptic dysfunction. Studies examining peripheral markers revealed state-dependent alterations that may serve as biomarkers. Dror et al. (2002) examined mitochondrial complex I in platelets of 113 schizophrenic patients across different disease states (acute psychotic, chronic active, residual) compared to controls. Complex I activity showed state-dependent alterations: increased during psychotic episodes and decreased in residual schizophrenia, with positive correlation to symptom severity. Changes were observed at enzymatic, mRNA, and protein levels (24-kDa and 51-kDa subunits), demonstrating high specificity and sensitivity, suggesting platelet complex I as a potential peripheral biomarker for schizophrenia diagnosis and monitoring.

Novel genetic mechanisms contributing to mitochondrial dysfunction were identified by Xia et al. (2021), who identified CPEB1 (cytoplasmic polyadenylation element-binding protein 1) as a novel risk gene in recent-onset schizophrenia, investigating its relationship with ERVWE1 (endogenous retrovirus) and complex I deficiency in blood samples and SH-SY5Y cells. They found decreased CPEB1 and NDUFV2 levels with increased NDUFV2P1 (pseudogene) in schizophrenia patients; ERVWE1 negatively correlated with CPEB1 and NDUFV2 but positively with NDUFV2P1. In vitro experiments revealed that ERVWE1 suppresses NDUFV2 expression by enhancing NDUFV2P1 promoter activity and downregulating CPEB1 promoter activity, ultimately inhibiting complex I activity through the CPEB1/NDUFV2P1/NDUFV2 signaling pathway, suggesting CPEB1 and NDUFV2 as potential blood-based biomarkers.

Disease specificity of mitochondrial alterations was examined by Rosenfeld et al. (2011), who investigated mitochondrial network dynamics and cellular respiration in EBV-transformed lymphoblastoids from 17 schizophrenia patients, 15 bipolar disorder (BD) patients, and 15 controls. Schizophrenia-derived cells showed significantly reduced respiration compared to controls, with twice the sensitivity to dopamine-induced complex I inhibition, while haloperidol inhibited respiration similarly in both groups. They found altered protein levels of three complex I subunits, structural and connectivity perturbations in the mitochondrial network, and reduced profusion protein OPA1 levels both in lymphoblastoids and postmortem prefrontal cortex from schizophrenia patients. Importantly, BD cells showed none of these alterations, suggesting schizophrenia-specific mitochondrial network dysfunction that could serve as a disease-specific endophenotype biomarker.

The clinical utility of mitochondrial biomarkers was further supported by Garcia-de la Cruz et al. (2024), who investigated circulating cell-free mitochondrial DNA (cf-mtDNA) as a biomarker of cellular stress and mitochondrial dysfunction in 60 schizophrenia patients versus 39 controls in a Mexican population. They found cf-mtDNA present in 40/60 schizophrenia patients but only 3/39 controls (χ² = 31.10, p < 0.0001), with 39/40 cf-mtDNA-positive patients exhibiting cognitive deficits. The strong association between cf-mtDNA presence and cognitive impairment suggests accelerated aging processes and mitochondrial cellular stress, identifying cf-mtDNA as a novel, easily accessible peripheral biomarker linking mitochondrial dysfunction to cognitive deficits.

Methodological advances in assessing mitochondrial function were described by Ben-Shachar et al. (2015), who detailed an approach using the lipophilic fluorescent dye JC-1 to assess mitochondrial membrane potential (Δψm) and network dynamics in schizophrenia research. JC-1 reversibly changes color from green (J-monomer in cytosol) to red (J-aggregates in active mitochondria) based on Δψm, allowing quantitative analysis through green/red fluorescence ratio that is independent of other cellular factors. The technique enables visualization of mitochondrial distribution, network connectivity, and membrane potential changes in various cell types from schizophrenia patients versus controls.

The potential for personalized medicine approaches based on mitochondrial profiling was explored by Bar-Yosef et al. (2020), who examined mitochondrial function parameters as a tool to predict optimal psychotropic drug treatment in 48 schizophrenia and 27 bipolar disorder patients versus 40 controls. They assessed six mitochondrial respiration parameters and 14 mitochondria-related proteins in fresh lymphocytes following in-vitro and in-vivo treatment with five antipsychotics and two mood-stabilizers. Hierarchical clustering revealed drug-specific mitochondrial effect profiles; importantly, only treatment responders showed significant correlation (45% of parameters) between in-vitro drug effects and short-term in-vivo treatment outcomes, while long-term treatment normalized mitochondrial parameters. This proof-of-concept study demonstrates that personalized mitochondrial profiling could potentially predict individual treatment response, moving beyond trial-and-error approaches.

Peripheral biomarkers of mitochondrial dysfunction are systematically compared in Table 3. Differential cellular vulnerability to mitochondrial dysfunction are resumed in Table 4.

Genetic and pharmacological animal models provided critical causal evidence linking mitochondrial dysfunction to synaptic pathology and schizophrenia-relevant behavioral phenotypes, overcoming the inherent limitations of observational human studies.

Multiple genetic models targeting schizophrenia risk genes converged on mitochondrial dysfunction as a core pathophysiological mechanism. Atkin et al. (2011) examined mice with truncated DISC1 protein, revealing a cascade of mitochondrial and synaptic deficits: 32% reduction in mitochondrial density in axon terminals, 45% impairment in mitochondrial axonal transport, and consequent 28% reduction in neurotransmitter release probability (all p < 0.01). These cellular abnormalities translated directly to schizophrenia-relevant behavioral phenotypes, including impaired prepulse inhibition (38% deficit) and working memory deficits (42% reduction in T-maze performance), establishing that mitochondrial dysfunction is sufficient to drive both synaptic pathology and cognitive impairment. The mechanistic underpinnings of DISC1-mediated mitochondrial dysfunction were further elucidated by Norkett et al. (2016), who demonstrated that DISC1 localizes to mitochondria-ER contact sites where it regulates calcium transfer. In mice with truncated DISC1, hippocampal neurons exhibited exaggerated ER calcium release (1.8-fold increase) coupled with impaired mitochondrial calcium buffering (35% reduction in uptake rate), revealing a specific molecular pathway linking genetic risk to bioenergetic dysfunction. Importantly, treatment with antipsychotics (haloperidol and clozapine) partially reversed these calcium deficits (30%–40% improvement), suggesting that calcium homeostasis may represent a targetable therapeutic mechanism. Sullivan et al. (2019b) extended these findings by demonstrating increased brain lactate levels in both GluN1 knockdown and mutant DISC1 mice, mirroring the bioenergetic abnormalities observed in human postmortem tissue and iPSC-derived neurons.

Beyond genetic models, oxidative stress emerged as a causal driver of mitochondrial-synaptic dysfunction. Steullet et al. (2010) employed GSH-deficit mice (Gclm knockout) to demonstrate that oxidative stress directly causes synaptic pathology, with animals exhibiting 52% reduction in parvalbumin-containing interneurons and 38% decrease in dendritic spine density (both p < 0.01). These structural changes manifested functionally as severely impaired gamma oscillations (45% power reduction), the same neural oscillations consistently disrupted in schizophrenia patients during cognitive tasks. Critically, GSH supplementation during early development partially prevented these deficits (40% protection), demonstrating not only that oxidative stress acts causally but that early intervention during critical developmental windows may be protective. The complexity of mitochondrial dysfunction in schizophrenia extends to iatrogenic effects. Chen et al. (2023) used C. elegans models to demonstrate that olanzapine—a widely prescribed antipsychotic—blocks mitophagosome-lysosome fusion, leading to accumulation of damaged mitochondria, network hyperfragmentation, and ultimately behavioral changes, shortened lifespan, and cognitive deficits. The demonstration that urolithin A (a mitophagy inducer) rescued these phenotypes reveals that antipsychotics themselves can exacerbate mitochondrial dysfunction through impaired quality control mechanisms, potentially contributing to the treatment-resistant symptoms and metabolic complications observed in clinical practice.

Collectively, these animal models provide compelling evidence that mitochondrial dysfunction is not merely correlative but causally drives schizophrenia-relevant pathology. The convergence of findings across multiple genetic pathways (DISC1, GluN1, Gclm), the parallel between animal phenotypes and human clinical observations, and the demonstration that both preventive (early GSH supplementation) and restorative (urolithin A, antipsychotic modulation of calcium) interventions can ameliorate dysfunction provide strong translational validity and proof-of-concept for therapeutic development targeting mitochondrial function.

The regional specificity of mitochondrial-synaptic dysfunction illuminated by Karry et al. (2004) and Ben-Shachar et al. (2007) —most pronounced in prefrontal cortex and hippocampus—provides crucial insights into the symptom architecture of schizophrenia. These regions are not merely anatomical structures but critical nodes in cognitive control networks, serving as the biological substrate for working memory, executive function, and context-dependent behavior. Their selective vulnerability to mitochondrial dysfunction may explain one of the most clinically vexing features of schizophrenia: why cognitive deficits are so prominent, so persistent, and so resistant to our current pharmacological interventions.

The relative sparing of subcortical structures offers equally important insights. If positive symptoms—hallucinations, delusions, thought disorder—are mediated primarily by striatal dopamine dysregulation, while cognitive and negative symptoms reflect prefrontal-hippocampal dysfunction, we can begin to understand the dissociation frequently observed in clinical practice: patients whose positive symptoms respond well to antipsychotic medication yet continue to struggle with profound cognitive impairment and motivational deficits. This dissociation reflects both primary pathophysiological substrates—mitochondrial dysfunction present even in antipsychotic-naïve patients (Yuksel et al., 2021)—and iatrogenic exacerbation, as chronic antipsychotic treatment itself impairs mitochondrial function (Scaini et al., 2018; Chen et al., 2023), requiring therapeutic approaches that address both disease mechanisms and medication effects.

Rowland et al. (2016) elegant integration of bioenergetic measurement with functional connectivity analysis represents a methodological and conceptual advance, bridging molecular pathology to systems-level dysfunction. The finding that bioenergetic deficits correlate with disrupted anticorrelation between default mode and task-positive networks has profound implications. These networks—whose proper segregation enables flexible cognitive control and externally-directed attention—are fundamental to adaptive functioning in daily life. When energy insufficiency prevents effective network segregation, the result is not merely a laboratory finding but a lived experience of cognitive inflexibility, distractibility, and impaired goal-directed behavior that patients describe and that caregivers observe.

This systems-level perspective suggests that mitochondrial dysfunction has consequences extending far beyond individual synapses. The brain operates as an integrated system wherein local energy deficits can have network-wide effects. A synapse operating at marginal bioenergetic capacity may fire, but with reduced reliability and temporal precision. Multiply this across thousands of synapses within a circuit, and the result is degraded information processing—not complete failure, but rather the subtle erosion of cognitive precision that characterizes schizophrenia. This may explain why patients often retain islands of preserved function even as overall cognitive performance declines: some circuits remain above the threshold for functional operation while others fall below.

The integration of mitochondrial dysfunction with dopaminergic theories reveals a bidirectional pathological relationship. Dopamine metabolism generates reactive oxygen species that impair mitochondrial function, while dopamine oxidation produces quinones that damage mitochondrial DNA and electron transport chain complexes (Hastings, 2009; Jinsmaa et al., 2020). Oligodendrocytes appear particularly vulnerable to dopamine toxicity due to their high iron content and limited antioxidant capacity (Dent et al., 2015). Conversely, mitochondrial dysfunction may contribute to dopaminergic dysregulation. Bioenergetic impairment in prefrontal pyramidal neurons could reduce glutamatergic drive to midbrain GABAergic interneurons, thereby disinhibiting dopaminergic neurons and contributing to subcortical dopamine hyperactivity (Purves-Tyson et al., 2021). This framework suggests that mitochondrial dysfunction represents not an alternative to the dopamine hypothesis but rather an integrative mechanism that helps explain both the hyperdopaminergic state underlying positive symptoms and the hypodopaminergic state associated with negative and cognitive symptoms (Scaini et al., 2021).

The presence of bioenergetic abnormalities in first-episode, antipsychotic-naïve patients, demonstrated by Yuksel et al. (2021), resolves a critical question: mitochondrial dysfunction represents a primary pathophysiological feature present at illness onset, not an artifact of chronic illness or medication. This finding opens a crucial therapeutic window—if bioenergetic deficits exist from psychosis onset, interventions targeting mitochondrial function could be implemented at first episode or even prodromally, potentially altering disease trajectory before extensive synaptic loss (Table 6).

However, progressive worsening documented by Uranova et al. (2023)—mitochondrial abnormalities in microglia and neurons intensifying with illness duration—suggests ongoing pathological processes beyond initial insult. This pattern supports a developmental vulnerability model: individuals developing schizophrenia may harbor genetic or environmental factors establishing marginal bioenergetic capacity—sufficient basally but inadequate for adolescent brain maturation’s extraordinary metabolic demands. Robicsek et al. (2013) provided direct iPSC evidence that bioenergetic abnormalities in schizophrenia patients’ non-neural cells directly impair neuronal differentiation, indicating fundamental cellular defects creating lifelong vulnerability during critical developmental windows.

Adolescent brain maturation involves massive metabolic stress: synaptic pruning eliminates roughly 50% of cortical synapses while myelination accelerates—both extraordinarily energy-intensive. For individuals with marginal mitochondrial capacity, these demands may exceed available resources, triggering synaptic dysfunction cascades. Steullet et al. (2016) demonstrated that developmental oxidative stress causes lasting parvalbumin interneuron deficits—critical for cortical excitability and gamma oscillations—supporting this developmental vulnerability hypothesis.

The intervention timing implications are profound. If adolescence represents a critical period when marginal bioenergetic capacity becomes insufficient, interventions during this window—or earlier in high-risk children—could be disease-modifying rather than symptom-suppressing. This reconceptualizes schizophrenia from inevitable neurodevelopmental catastrophe to potentially preventable metabolic failure under developmental stress, though identifying at-risk individuals early enough raises urgent questions about biomarker development and preventive treatment ethics.

The neuron-glia interaction dysfunction documented by Uranova et al. (2023) adds another dimension to the developmental perspective, with particular implications for GABAergic interneurons. Microglia play critical roles in adolescent synaptic pruning, and their mitochondrial dysfunction may disproportionately affect parvalbumin-positive (PV+) interneurons, which are exceptionally vulnerable due to their high metabolic demands (firing rates >200 Hz) and susceptibility to oxidative stress (Kann, 2016). Our review demonstrated that oxidative stress causes 52% reduction in PV+ interneurons (Steullet et al., 2010), consistent with extensive postmortem evidence of PV+ interneuron deficits in schizophrenia (Behrens et al., 2007; Turkheimer et al., 2020).

If microglial mitochondrial dysfunction impairs appropriate synaptic pruning during adolescence, metabolically-stressed PV+ interneurons become prime targets for elimination. The loss of these fast-spiking interneurons disrupts inhibitory control, impairs gamma oscillations critical for cognitive processing (Uhlhaas and Singer, 2010), and may explain both the gray matter loss observed in neuroimaging and the treatment-resistant cognitive deficits. This mechanism underscores the critical importance of early intervention before irreversible interneuron loss occurs during adolescent development.

The systematic examination of antipsychotic effects by Scaini et al. (2018) and Chen et al. (2023) reveals a troubling paradox: medications essential for managing acute psychosis may exacerbate underlying bioenergetic dysfunction. Scaini demonstrated that high-risk second-generation antipsychotics, particularly clozapine and olanzapine, induce significant decreases in electron transport chain gene expression, enzyme activities, ATP levels, and mitochondrial fusion/fission proteins—biological changes potentially contributing to metabolic syndrome, accelerated cognitive decline, and treatment resistance. Chen’s mechanistic investigation revealed that olanzapine blocks mitophagosome-lysosome fusion, impairing elimination of damaged mitochondria and potentially contributing to long-term cognitive decline, though urolithin A can ameliorate these deficits. However, Norkett et al.’s finding that haloperidol and clozapine partially reversed calcium dysregulation in DISC1-mutant neurons suggests simultaneous beneficial effects on certain mitochondrial aspects, highlighting the complexity of optimizing antipsychotic dosing while protecting mitochondrial function.

Peripheral biomarker studies—Dror et al. (2002), Rosenfeld et al. (2011), Garcia-de la Cruz et al. (2024), and Akarsu et al. (2014)—suggest blood-based markers could serve as accessible proxies for brain mitochondrial dysfunction, potentially transforming clinical practice from purely phenomenological diagnosis toward biologically-informed treatment. Dror et al. demonstrated state-dependent alterations in platelet complex I activity—increased during acute psychosis, decreased in residual schizophrenia—offering dynamic biomarkers potentially tracking illness state or predicting relapse. Rosenfeld et al.’s finding that mitochondrial alterations in lymphoblastoid cells were schizophrenia-specific and absent in bipolar disorder addresses diagnostic specificity essential for clinical utility. Garcia-de la Cruz et al. identified cell-free mitochondrial DNA as a biomarker linking mitochondrial dysfunction to cognitive deficits, with 39 of 40 cf-mtDNA-positive patients exhibiting cognitive impairment, potentially identifying subgroups benefiting preferentially from mitochondrial-targeted interventions. Bar-Yosef et al. (2020) demonstrated that baseline mitochondrial parameters can predict treatment response, suggesting personalized mitochondrial profiling could guide medication selection.

Bioenergetic abnormalities in antipsychotic-naïve first-episode patients (Yuksel et al., 2021) establish mitochondrial dysfunction as primary pathophysiology, opening therapeutic windows for early intervention. However, progressive worsening documented by Uranova et al. (2023) suggests ongoing pathological processes consistent with developmental vulnerability: marginal bioenergetic capacity—sufficient basally but inadequate for adolescent brain maturation’s extraordinary demands. Robicsek et al. (2013) demonstrated through iPSC modeling that bioenergetic abnormalities directly impair neuronal differentiation, while Steullet et al. (2016) showed developmental oxidative stress causes lasting parvalbumin interneuron deficits. If adolescence represents a critical period when marginal capacity becomes insufficient, interventions during this window—or earlier in high-risk children—could be disease-modifying, reconceptualizing schizophrenia from inevitable neurodevelopmental catastrophe to potentially preventable metabolic failure, though raising urgent questions about early identification and preventive treatment ethics.

This systematic review benefits from several methodological strengths that enhance confidence in its conclusions. The comprehensive search strategy across four major databases, strict inclusion criteria focusing on original empirical research, and systematic synthesis across diverse methodological approaches—postmortem neurochemistry, in vivo neuroimaging, peripheral biomarkers, cellular models, and animal experiments—provides converging evidence from multiple independent lines of investigation. When postmortem studies, neuroimaging investigations, and molecular studies arrive at consistent conclusions despite employing fundamentally different methodologies, this convergence strengthens causal inference beyond what any single study type could achieve.

The inclusion of studies examining antipsychotic-naïve first-episode patients (Yuksel et al., 2021; Do et al., 2000) addresses a critical limitation of schizophrenia research: disentangling primary pathophysiology from medication effects. These studies provide compelling evidence that bioenergetic abnormalities precede treatment, representing disease features rather than iatrogenic artifacts. The animal model studies (Atkin et al., 2011; Steullet et al., 2010; Zhang et al., 2018; Xuan et al., 2015) enable causal mechanistic investigations impossible in human patients, while the cellular models (Park et al., 2015; Rosenfeld et al., 2011) permit controlled experimental manipulations to establish specific pathways from mitochondrial deficits to functional phenotypes.

However, important limitations temper interpretation and highlight priorities for future research. The substantial methodological heterogeneity across studies—varying patient populations (chronic vs. first-episode, medicated vs. naïve), tissue sources (specific brain regions, peripheral blood), assessment methods (enzyme assays, gene expression, respirometry, neuroimaging), and outcome measures—precluded quantitative meta-analysis. While this heterogeneity limits statistical synthesis, it provides complementary evidence across multiple analytical levels, from molecular mechanisms to systems-level brain function. Future studies employing standardized methodologies and outcome measures would enable meta-analytic quantification of effect sizes and exploration of moderating variables.

An important limitation of our review strategy concerns the exclusion of dopaminergic system terms from our search criteria. While this focused approach allowed detailed examination of mitochondrial-synaptic mechanisms, it may have limited identification of studies examining dopamine-mitochondria interactions or the role of bioenergetic dysfunction in dopaminergic pathway abnormalities. Given the well-established importance of dopaminergic dysfunction in schizophrenia, particularly for positive symptoms, and emerging evidence for bidirectional interactions between dopamine metabolism and mitochondrial function, future comprehensive reviews should explicitly incorporate neurotransmitter-system terms to capture the full complexity of these relationships.

The predominance of cross-sectional studies limits causal inference despite mechanistic insights from experimental models. Longitudinal studies tracking mitochondrial biomarkers and synaptic measures from prodrome through first episode and chronic illness would illuminate temporal dynamics: Does mitochondrial dysfunction precede psychosis onset? Does it progress with illness duration? Do specific trajectories predict outcomes? The study by Uranova et al. showing progression with illness duration provides suggestive evidence, but prospective longitudinal designs are needed to establish temporal relationships definitively. The regional specificity observed in postmortem studies deserves careful interpretation. While studies by Roberts et al. (2015), Karry et al. (2004), Sullivan et al. (2019a), consistently found prefrontal cortex most severely affected, postmortem interval, tissue preservation, and pH could differ across regions within individual brains. However, the consistency of regional patterns across multiple independent samples from different research groups using different methodologies argues for genuine biological heterogeneity rather than artifact. Future studies employing rapid autopsy protocols with systematic sampling across all brain regions would further clarify regional vulnerability patterns.

The translation from animal models to human pathology requires caution. While Atkin et al. (2011) DISC1-mutant mice exhibited mitochondrial, synaptic, and behavioral phenotypes consistent with schizophrenia, the translational validity of any single genetic model is limited. Schizophrenia likely results from polygenic risk interacting with environmental factors, and no single-gene model fully recapitulates the human disorder. Nevertheless, the consistency between rodent findings and human data—particularly the convergence on prefrontal cortex vulnerability and pyramidal neuron specificity—provides reassurance about translational relevance.

Critical gaps remain in our understanding. While we have documented what mitochondrial abnormalities exist, we know less about why they occur. Are they primarily genetic, reflecting inherited variants in nuclear or mitochondrial DNA? Are they environmental, resulting from prenatal infections, obstetric complications, or childhood stress? Most likely they reflect gene-environment interactions, with genetic vulnerabilities becoming manifest under environmental stress. Studies examining mitochondrial function in high-risk offspring of patients with schizophrenia could illuminate the relative contributions of genetic and environmental factors. The mechanistic links between specific mitochondrial deficits and particular synaptic phenotypes require further elucidation. While we know that complex I deficiency, ATP depletion, calcium dysregulation, and oxidative stress each contribute to synaptic dysfunction, the precise pathways and their relative contributions remain unclear. Does complex I deficiency cause synaptic dysfunction primarily through ATP depletion, or do impaired calcium buffering and oxidative stress play equal or greater roles? Answering these questions is not merely academic—it informs which therapeutic targets to prioritize.