Mitochondrial DNA copy number is associated with psychosis severity and anti-psychotic treatment

Mitochondrial pathology has been implicated in the pathogenesis of psychotic disorders. A few studies have proposed reduced leukocyte mitochondrial DNA (mtDNA) copy number in schizophrenia and bipolar disorder type I, compared to healthy controls. However, it is unknown if mtDNA copy number alteration is driven by psychosis, comorbidity or treatment. Whole blood mtDNA copy number was determined in 594 psychosis patients and corrected for platelet to leukocyte count ratio (mtDNAcnres). The dependence of mtDNAcnres on clinical profile, metabolic comorbidity and antipsychotic drug exposure was assessed. mtDNAcnres was reduced with age (β = −0.210, p < 0.001), use of clozapine (β = −0.110,p = 0.012) and risperidone (β = −0.109,p = 0.014), dependent on prescribed dosage (p = 0.006 and p = 0.026, respectively), and the proportion of life on treatment (p = 0.006). Clozapine (p = 0.0005) and risperidone (p = 0.0126) had a reducing effect on the mtDNA copy number also in stem cell-derived human neurons in vitro at therapeutic plasma levels. For patients not on these drugs, psychosis severity had an effect (β = −0.129, p = 0.017), similar to age (β = −0.159, p = 0.003) and LDL (β = −0.119, p = 0.029) on whole blood mtDNAcnres. Further research is required to determine if mtDNAcnres reflects any psychosis-intrinsic mitochondrial changes.

Psychosis is a common trait in schizophrenia (SZ), schizoaffective disorder, delusional disorder and bipolar disorder (BD) 1 , displayed as a range of symptoms reflecting dissociation from reality. Psychosis and the underlying psychiatric diagnoses have shown a robust association with somatic conditions linked to aging such as cardiovascular disease (CVD) 2 , inflammation 3 , and obesity 4 . These somatic conditions may be caused by chronic oxidative stress 5,6 or exposure to drug toxicity 7 , leading to cellular ageing 8,9 which recently has been found increased in psychotic disorders 10,11 .
The mitochondrion is the organelle for ATP production which contains multiple copies of mitochondrial DNA (mtDNA), in blood present in leukocytes and platelets. Ongoing oxidative chain reactions within its compartments make mitochondria a significant source of intracellular reactive oxygen and nitrogen species (ROS/ RNS) 12 . The molecular machinery and DNA of the mitochondria are thus at risk from the deleterious effects of oxidative stress resulting in mitochondrial dysfunction 13 .
Mitochondrial dysfunction and ROS/RNS are implicated in the induction of apoptosis through caspase activation 14,15 , inflammasome recruitment 16 and the activation of downstream cytokines and inflammatory mediators 17 . In the central nervous system, mitochondrial dysfunction could through these processes lead to neurodegeneration, the precursor to cognitive decline and dementia. Mitochondrial dysfunction has been implicated in various somatic and neuro-degenerative disorders 18,19 . Additionally, evidence of mitochondrial dysfunction in BD and SZ has been reported in magnetic resonance spectroscopy studies of small patient groups of BD and SZ 20,21 . Mitochondrial structural abnormalities have been reported in patients with BD 22 and SZ 23,24 and both diseases are associated with mtDNA mutations and polymorphisms [25][26][27][28][29][30][31][32][33][34] . Hyper-oxidative states and chronic inflammation resulting from mitochondrial dysfunction may be involved in the progression of SZ 6  Whole blood DNA extraction and mitochondrial DNA copy number (mtDNA). DNA was extracted from venous blood using a standard phenol-chloroform method (Lindblom and Holmlund) 75 followed by desalting using Illustra NAP-5 columns (GE Healthcare, Buckinghamshire, UK) and was quantified spectrophotometrically with the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). The copy number of mitochondrial DNA per nuclear genome (i.e. per cell) was determined using real-time quantitative PCR (qPCR) according to Rooney JP 76 and Venegas V 77 protocol. The relative amount of mitochondrial gene tRNA-Leu(UUR), to the nuclear single-copy gene β2-microglobulin (B2M) (mtDNA copy number) was determined using a standard curve. In brief, each DNA sample (4.0 ng) was assessed for tRNA-Leu(UUR) and B2M in triplicate within the same 384-well plate, amplified by using Power SYBR Green in 10 µl total reaction volume. The reaction was performed on QuantStudio 7 Flex (Applied Biosystems; Life Technologies; Thermo Fisher Scientific, Waltham, MA, USA) with the following conditions: 95 °C for 15 min, followed by 40 repeats of 95 °C for 15 s and 60 °C for 1 min, followed by a dissociation stage to monitor amplification specificity. The same standard curve of control genomic human DNA (Applied Biosystems) ranging from 10 ng to 0.016 ng, was run on each plate for both genes and was used to determine the quantity of each gene for each sample. This allowed controlling for differences in the efficiencies between that of mtDNA and B2M. The gene quantities were then used to determine the M/S ratio for each sample. DNA samples with a Ct standard deviation of ≥0.3 between triplicates or a Ct value outside the standard curve were omitted from the analyses. Samples were analyzed in 12 consecutive plates. R 2 coefficients of the standard curves were above 0.99 for each primer set and 384-plate. The inter-plate coefficient of variation (CV) of M/S ratio was 8.3% calculated as a mean from three inter-plate control samples run in the 12, 384-well plates. The mean intra-plate CV of M/S ratio of the three control samples in triplicates was 4.6%. Primer binding regions were selected for low deletional <3% 78 or mutational (SNPs) exposure 2.2% 79 . No known psychiatry related SNPs are present in the primer binding mtDNA regions (Online Mendelian Inheritance in Man, OMIM ® 80 and www.mitomap.org) 79 . The probability of random somatic mutations to occur at the primer binding regions is calculated to be 0.24%. For a subset of the patients an alternative amplicon in the D-loop region of the mitochondrial genome was targeted for the quantification of mtDNA to confirm the sensitivity of the assay. This D-loop amplicon was selected based on Bai and Wong 78 and was carefully designed from mtDNA sequences that do not contain reported mutations or polymorphisms occurring in more than 1.6% of the population (http:// www.genpat.uu.se/mtDB/Polysites) 79,81 . The D-loop region has been used for mtDNA content analysis in similar investigations for mtDNA copy number quantification 11,42 . mtDNAcn measurements targeting the D-loop and tRNA-Leu(UUR) were compared and found to be highly correlated, n = 55, r = 0.996, p = 1.62E-58. The mtDNA copy number measurement success rate was 97% (594/614 samples). mtDNA copy number was corrected for platelet and leukocyte count as these variables are known to affect MS ratio. Platelets have mitochondria and accompanying mtDNA but no nuclei and hence lack a nuclear genome, which results in an overestimation of mtDNA 68 . Whole blood derived mtDNA are associated with leukocyte count as reported by several groups and thus a correction for the platelet and leukocyte have been suggested as a refinement to the mtDNA copy number measurement 68-70 . Generation of human neurons in vitro and drug treatment. The long-term neuroepithelial stem cell (NESC) line I3.2, derived from human embryonic stem cells and previously described 82 was used for the generation of human neurons in vitro. For maintenance, NESC were cultured on poly-l-Ornithine (PLO, Sigma-Aldrich, Irvine, UK) and laminin (Sigma-Aldrich, Irvine,UK) coated wells, in DMEM/F12 (Gibco) media supplemented with N2 and B27 supplements (Life Technologies; Carlsbad, CA, USA) in the presence of the growth factors EGF and FGF2-basic as previously described 82 . To induce differentiation, EGF and FGF2-basic were removed from the media (Supplementary Fig. 2A and B). Briefly, NESC were seeded on PLO/laminin at the density of around 75000-150000 cells per well, in a 48-well plate, and cultured for 7 days in DMEM/F12 media supplemented with N2 (1:100) and B27 (1:100) supplements. After one week of differentiation, media was changed to DMEM/F12 supplemented with N2 (1:100) with the addition of glial cell-derived neurotrophic factor (GDNF) 20 ng/ml, brain-derived neurotrophic factor (BDNF) 20 ng/ml, ascorbic acid 10 mM, dibutyryl adenosine 3′,5′-cyclic monophosphate sodium salt (dcAMP) 25 mM. Drug treatments started after 2 weeks of in vitro differentiation, when NESCs had a clear neuronal morphology. Differentiated neurons were characterized through immunohistochemistry stains for neuronal marker TUBB3.
For drug treatments, neurons were treated in plain DMEM/F12 media supplemented with only N2 (1:100), with the addition of either the vehicle or different concentrations of the drugs for one week before they were processed for DNA extraction. Clozapine and risperidone were dissolved in 1 M HCl before added to cell culture medium in the concentrations of 0.075 μM, 0.75 μM and 0.025 μM, 0.25 μM respectively. The vehicle used was plain DMEM/F12 media acidified by 0.0025 μM HCl, pH neutralized by CO 2 buffer. The concentrations were selected to simulate clinical target concentrations in plasma and diluted by 10-fold to simulate concentration in the cerebrospinal fluid and brain interstitial fluid [83][84][85] . At the end of treatment, cells were lysed and DNA extraction was performed using the Quick-DNA ™ Miniprep Plus Kit (Zymo Research Corp, Irvine, CA, USA).
Subsequently mtDNA copy number determination was performed as described above. Experiments were run in 3 biological replicates.
Statistical analyses. The effect of platelet and leukocyte counts on mtDNA copy number (mtDNAcn) was assessed by regressing mtDNAcn on age, platelet count and leukocyte count as follows: mtDNAcn = b 0 + b 1 (age) + b 2 (leukocyte count) + b 3 (platelet count). Thereafter, mtDNAcn was corrected for platelet and leukocyte counts by regressing mtDNAcn on the platelet to leukocyte count ratio (platelet count/leukocyte count), as SCIeNtIFIC REPORTS | (2018) 8:12743 | DOI:10.1038/s41598-018-31122-0 outlined in Hurtado-Roca, et al. 68 , generating normally distributed unstandardized residuals. The unstandardized residual variable, hereafter designated as mtDNAcn res , was treated as the dependent variable in the following regression analyses.
We evaluated the significance of psychiatric diagnosis on mtDNAcn res using an analysis of covariance (ANCOVA) adjusting for covariates suggested to influence mtDNAcn res , i.e. age, gender, smoking, alcohol intake and psychosis severity 10,[39][40][41][42]86,87 . Non-parametric Mann Whitney U test, was used to assess difference in mtD-NAcn res between patients with MetS and those without MetS.
To study putative predictors of mtDNAcn res , regression modelling was performed as previously described 72,88 . In brief, an iterative method of regression modelling was initiated with no variables in the model, adding each variable to be tested with an entry requirement of p < 0.05 and a loss of significance at p > 0.10. Regression residuals were ensured to be normally distributed. Errors due to multiple testing were corrected for using the bonferroni method. Four multiple linear regression models were built to evaluate the effects of i) age, gender, psychosis severity (CGI-S), alcohol intake and smoking ii) drug treatment and iii) metabolic factors on mtDNAcn res . Model 1 was as follows: mtDNAcn res = b 0 + b 1 (age) + b 2 (gender) + b 3 (CGI) + b 4 (smoking) + b 5 (alcohol) + b 6 (psychiatric diagnosis). The second model looking at the effect of antipsychotic drugs and mood stabilizers on mtDNAcn res was as follows: mtDNAcn res = b 0 + b 1 (age) + b2 (gender) + b 3 (psychiatric diagnosis) + b 4 (mood stabilizer) + b n (drug), where (drug) represents the patient's treatment [yes/no] with any of n = 11 different antipsychotics. The mood stabilizer variable represents the presence of any mood stabilizer [yes/no] as more detailed information was not available. The numbers of patients on each drug and on any mood stabilizer are listed in Table 1. The mtDNAcn res of patients using a drug significant in model 2 were, for confirmation, compared to that of the all other patients using Mann-Whitney U test. The effect of clozapine and risperidone treatment on mtDNAcn res was further investigated by calculating Spearman's correlation coefficient, ρ, between the prescribed daily drug dose and the mtDNAcn res . To compare the effects of CGI-S with antipsychotic drug treatment effects we imported the significant variables from the model 1 and 2 to a third regression model. The equation of model 3 was as follows: mtDNAcn res = b 0 + b1 (proportion of life on antipsychotic treatment) + b2 (CGI) + b3 (age). This model was separately run on patients who were treated with clozapine or risperidone (model 3a, n = 139) and patients who were not on risperidone and clozapine (model 3b, n = 348). The fourth model (metabolic profile) which analyzed the effect of metabolic factors was adjusted for significant predictors of mtDNAcn res obtained from model 3. Model 4a was mtDNAcn res = b 0 + b 1 (age) + b 2 (proportion of life on antipsychotic treatment) + b 3 (gender) + b 4 (waist) + b 5 (LDL) + b 6 (HDL) + b 7 (glucose) + b 8 (Log 10 HOMA-IR) + b 9 (high BP); and model 4b was . Model 4a was run on patients who were treated with clozapine and risperidone (n = 123) and model 4b was run on patients who were not on risperidone and clozapine (n = 324). A sensitivity analysis of the Models 1, 2, 3 and 4 was performed in the patients with a diagnosis of schizophrenia. Output model indices and lack of multi-collinearity between input variables were confirmed using standard multiple linear regression (the enter method), including all input variables. Analyses were performed using IBM Statistical Package for the Social Sciences version 23, (IBM Corporation, USA). Power calculations were performed using http://biomath.info/power/ttest.htm.

Results
As previously reported we found that whole blood mtDNAcn was significantly influenced by platelet and leukocyte counts (β platelets = 0.119, p = 0.005, and β leukocytes = −0.298, p < 0.001). A regression of whole blood mtD-NAcn on the platelet to leukocyte count ratio (platelet/leukocyte) was performed to generate normally distributed unstandardized residuals (mtDNAcn res ) which were used in the following analyses as a dependent variable ( Supplementary Fig. S3).
The clinical characteristics of the psychosis patients are shown in Tables 1 and 2. The psychiatric diagnoses within the psychosis patient cohort were SZ, schizoaffective disorder, delusional disorder, psychosis unspecified, BD and other disorders with psychotic features. The effect of diagnosis on mtDNAcn res was evaluated using an analysis of covariance (ANCOVA) correcting for covariates suggested to influence mtDNAcn res , i.e. age, gender, smoking and alcohol intake. No significant difference in mtDNAcn res was observed between the six diagnosis groups (p = 0.212). Nonetheless, in subsequent regression analyses (Models 1, 2, 3 and 4) of the psychosis patients, adjustment was made for psychiatric diagnosis, in addition to age and gender, to exclude detectable confounding. Moreover, a sensitivity analysis in only those with a SZ diagnosis (n = 306) replicated the findings in Models 1-4 of the full psychosis cohort (Supplementary Table S1).

Antipsychotic drug effects on mtDNA copy number in human neurons in vitro.
To assess the effect of clozapine and risperidone on neurons, human neurons generated in vitro from NESCs were exposed for 7 days to clozapine (0.075 μM and 0.75 μM) or risperidone (0.025 μM and 0.25 μM). For clozapine treatment, at 0.075 μM, there was a 16% reduction in mtDNA copy number compared to vehicle treated cells (p = 0.0005), and at 0.75 μM, the corresponding reduction was 25% (p = 0.0004) (Fig. 1d). For risperidone treatment, at 0.025 μM there was no change in mtDNA copy number compared to vehicle treated cells, but at 0.25 μM, there was a 14% reduction in mtDNA copy number compared to vehicle treated cells (p = 0.0126) (Fig. 1d). Thus, risperidone was found to have a reducing effect on mtDNA copy number at a concentration simulating the clinical target level in plasma but not at the concentration simulating CSF or brain interstitial target level. Clozapine was found to be associated with reduced mtDNA copy number at both doses, in a dose dependent manner. Higher doses of anti-psychotic drugs (clozapine, 75 μM and risperidone, 25 μM) were associated with massive neuronal cell death (data not shown).

Discussion
From a neuroanatomical point of view, a much reviewed discrepancy between SZ and healthy controls are aberrations in dendritic spine morphology. These have been observed in the cortical layers of SZ patients with concomitant increases in the molecular signatures of mitochondrial dysfunction 89,90 . Interestingly, compelling evidence exists to explain the link between mitochondrial dysfunction and aberrations in dendritic spine morphology as mitochondria play salient roles in dendritic spine architecture and neuronal processes which affect crucial cortical circuitry 91,92 . Mitochondrial dysfunction and oxidative stress in both brain and leukocytes are overrepresented in not only SZ but also BD type I 23,26,44,93 . There is a link between mitochondrial dysfunction and psychosis-like symptoms 53,94 , and mitochondrial dysfunction is reported to be intrinsic to the complex etiology of SZ 89 . Oxidative stress induced in dysfunctional mitochondria can cause deletions of mtDNA, and influence mitochondrial biogenesis 36 . A few studies have utilized mtDNA copy number analysis to investigate mitochondrial dysfunction and reported reduced whole blood or leukocyte mtDNA copy number in SZ and BD type I compared to healthy controls 10,43,44 . Similarly, the present study was performed to further explore whole blood mtDNA copy number, adjusted for platelet to leukocyte count ratio, in psychosis by focusing on effects of disease severity, antipsychotic drug treatment and metabolic comorbidity. The main findings of this study were that (i) mtDNA copy number was reduced with increasing psychosis severity, and (ii) the antipsychotic drugs clozapine and risperidone decreased mtDNA copy number in patient blood with similar effects on human neurons in vitro in a dose dependent manner. While this has not previously been reported, clozapine, risperidone and other antipsychotic drugs are known to be toxic to mitochondria in various transformed cell lines (neuroblastoma, adipocytes, myoblasts, hepatocytes, lymphoblasts and monocytes) in vitro 64,89 and in rodent brain 95 . Accordingly, we found reduced whole blood mtDNA copy number in psychosis patients treated with clozapine and risperidone compared to psychosis patients treated with or without other antipsychotics. We also showed an inverse correlation between prescribed oral drug dosage and mtDNA copy number. The proportion of life on antipsychotic treatment was a significant predictor of mtDNA copy number variance only in the subset of patients who were treated with clozapine and risperidone.
For patients not on clozapine or risperidone, decreasing mtDNA copy number was associated with increasing age and psychosis severity measured by the CGI-S (n = 348). Existing literature from studies of healthy individuals and non-psychotic post mortem studies support our finding that mtDNA copy number is reduced with advancing age [96][97][98] . Where psychotic features are of concern, conclusions along a similar grain were made by Li et al. 10 , who reported reduced whole blood mtDNA copy number in first-episode drug-naïve SZ patients (n = 137) compared to healthy controls and a trend for reduced mtDNA copy number to be associated with positive symptoms in SZ patients (p = 0.07). Two other studies which considered BD patients found reduced leukocyte mtDNA copy number in euthymic BD-I patients compared to BD-II and healthy controls 43,44 .
The findings linking psychosis with mtDNA copy number are bolstered by previous human post mortem studies where the number of mitochondria per tissue volume was found to be significantly decreased in oligodendrocytes of prefrontal cortex and caudate nucleus of SZ patients 23 . Mitochondrial density in the neuropils were significantly reduced in the caudate nucleus and putamen of SZ subjects compared to controls 93,99 . Moreover, psychotic positive symptoms, such as hallucinations, have been reported in subjects with mitochondrial disease 94 , and a high degree of mitochondrial myopathy encephalopathy lactic acidosis and stroke-like (MELAS) episodes have been reported in SZ and BD 31 .
To explore if the effects of clozapine and risperidone on mtDNA copy number in blood cells could also be detected in human neurons, we treated NESC-derived human neurons with clozapine and risperidone. We found that the effect was dose dependent at drug concentrations corresponding to target CSF levels during therapy 84,100 . Clozapine had a stronger effect than risperidone on mtDNA copy number at drug levels which simulate CSF or brain interstitial target levels. At clinical target plasma concentrations, there was a significant decrease (estimated at 15-25% reduction) in mtDNA copy number in the neurons exposed to any of the two drugs. These findings are supported by several studies that have shown that antipsychotics (clozapine, risperidone, haloperidol, olanzapine, quetiapine, chlorpromazine, and thiothixene) can inhibit the mitochondrial respiratory chain and cause further damage to mitochondria through oxidative stress 95,101,102 , however, the effect on mtDNA copy number was not previously studied. We detected an effect by only clozapine and risperidone but had a statistical power corresponding to 80% to detect an effect similar to that of risperidone for those drugs with a sample size of above 40 (that is haloperidol, olanzapine, zuclopenthixol, aripiprazole, perphenazine, risperidone and ziprasidone). Thus, we did not confirm a previously reported mitochondrial effect for haloperidol and olanzapine. An important limitation of this study is that while the total length of antipsychotic drug treatment was known, the specific antipsychotic drug names were available only for the antipsychotic drug treatment at sampling, and not historically. However, using prospective data starting at sampling we estimate that approximately 10% switched to another antipsychotic drug in a year. If mtDNA copy number depletion and accelerated cellular ageing occur as a result of these antipsychotic treatments, further investigation on mitochondria dysfunction as mediator between antipsychotics and associated comorbidities may be warranted. Another limitation is that the type of mood stabilizers used, in the 11% of patient cohort who were treated with them, was unknown. Amongst mood stabilizers, lithium has been shown to have beneficial effects on mitochondria through enhanced oxidative phosphorylation 103 and restoring defunct vacuolated mitochondria to their healthy baseline structure 104 . A reduction of mtDNA observed in lithium treated C. elegans with increased longevity, a finding going against the grain of understanding that a reduction of mtDNA signals cellular ageing, which was possibly explained by increasingly efficient mitochondrial bioenergetics 105 . Valproic acid, similarly has been reported to enhance mitochondrial biogenesis, accompanied with an increase in mtDNA copy number in hepatocytes 106 . Conversely lamotrigine an anti-convulsant and mood stabilizer used in more acute cases was reported to have toxic effects on mitochondria 107 .
Metabolic comorbidity and inflammation are gaining ground as bona-fide hallmarks of psychotic disorders 4 . It has been previously reported that metabolic syndrome per se is associated with a reduced mtDNA copy number 65 . We detected an association for blood mtDNA copy number to level of LDL, similarly strong as that to CGI-S in the psychosis patients not on clozapine or risperidone. The effect of metabolic syndrome on mtDNA copy number and mitochondrial functionality could be enhanced and mediated by systemic inflammation 108 . Therefore, a limitation of the present study is the lack of quantified inflammatory markers.
We adjusted the mtDNAcn for platelet to leukocyte ratio, but not for subpopulations of leukocytes as their counts were unavailable. While we cannot exclude that our observations might be due to changes in relative WBC subpopulation ratios, existing literature suggests that in a large patient group, total WBC count and platelet count capture the relevant information we need to correct for when making conclusions on changes in peripheral mtDNA 69,70,109 . Variation between subpopulations are rarely accounted for in mtDNAcn analyses. The mtDNA copy number variance explained by significant predictors was low, 4-6%. However, this is comparable with previous studies investigating the relationship between psychiatric disorders and cellular ageing markers (depression and telomere length) [110][111][112] . Several association tests were performed which requires correction for multiple testing. However, the regression model reported (models 1, 2 and 4 (model 3 building on 1 and 2)) were designed to test individual hypotheses. Within each model, the findings we report passed a multiple testing correction according to Bonferroni.
The focus of our study was to investigate the relationship between mtDNA copy number, psychosis severity and antipsychotic drug treatment in a psychosis outpatient clinic setting. Here the appropriate controls would have been anti-psychotic drug treatment naïve psychosis patients. However, in the Swedish healthcare setting patients who are diagnosed with the schizophrenia spectrum of disorders are always offered treatment. Therefore we were not able to obtain material from age and gender matched drug-naïve patients. A further limitation is that our study lacks healthy controls, however, previous comparisons performed have reported mtDNA copy number to be reduced in anti-psychotic treatment naïve patients of psychotic disorders (SZ and BD-I) compared to healthy controls 10,43,44 .
The patient group in our study belongs to multiple DSM-IV-based diagnostic categories, but all patients showed symptoms of severe psychosis and experienced impaired functionality. A unifying criteria of the study cohort was that, all patients were recruited from specialist psychosis outpatient clinics in Sweden, where they received anti-psychotic treatment, unless refused. Our decision to include data from all psychosis patients regardless of diagnoses is in accordance with the Research Domain Criteria (RDoC) paradigm. The RDoC is based on domains of human behavior and functionality e.g cognitive impairment in the presence of psychotic symptoms 113 . The presence of psychotic symptoms are relevant to the cognitive systems domain of the RDoC matrix, regardless of the DSM-IV criteria based diagnosis 114 .
To further address the heterogeneity of the patient cohort from the DSM-IV perspective we performed a sensitivity analysis by repeating linear modelling on the subset of the patients who had a SZ diagnosis. The main findings of the study were well replicated in the SZ cohort. Furthermore, we found no statistically significant effect of the DSM-IV-based diagnoses mtDNA copy number levels, even if this may have been due to a limited sample size within certain diagnostic groups i.e BD or delusion. In place of a symptom severity scale such as the positive and negative symptom scale (PANSS), which was unavailable in this study, we used the clinician-rated CGI-S, an indication of the patients' mental well-being, which has a significant overlap (21-60%) with PANSS 115 . The present study is limited by the unavailability of a validated symptom severity measure for psychotic cohorts, such as PANSS, in view of the heterogeneity of the patient cohort.
In conclusion, the present study describes a whole blood mtDNA copy number reduction with increasing psychosis severity, potentially driven by the use of antipsychotic drugs in those treated with clozapine and risperidone. The drug-dosage dependent reduction by clozapine and risperidone was present in both whole blood and human neurons after in vitro exposure. Our novel findings support earlier studies that reported mtDNA copy number reduction in blood from SZ and BD-I patients. However, before whole blood mtDNA copy number can be evaluated as a possible biomarker of psychosis or its progression, e.g. reflect psychosis-intrinsic mitochondrial changes, further research is required to estimate the relative contribution of other antipsychotic drugs and comorbidities to the whole blood mtDNA copy number.
Data Availability Statement. The authors report no biomedical financial interests, non-financial interests or other competing interests. The authors adhere to the data availability policy of Scientific Reports.