Specific subcellular changes in oxidative stress in prefrontal cortex from patients with bipolar disorder


  • Ana C. Andreazza,

    1. Departments of Psychiatry and Pharmacology, University of Toronto, Toronto, Ontario, Canada
    2. Centre for Addiction and Mental Health, Toronto, Ontario, Canada
    Search for more papers by this author
  • Jun-Feng Wang,

    1. Department of Pharmacology and Therapeutics, University of Manitoba, Winnipeg, Manitoba, Canada
    Search for more papers by this author
  • Faraz Salmasi,

    1. Departments of Psychiatry and Pharmacology, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Li Shao,

    1. Department of Psychiatry, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Lionel T. Young

    Corresponding author
    1. Departments of Psychiatry and Pharmacology, University of Toronto, Toronto, Ontario, Canada
    2. Centre for Addiction and Mental Health, Toronto, Ontario, Canada
    • Address correspondence and reprint requests to L. Trevor Young, University of Toronto, Departments of Psychiatry and Pharmacology & Toxicology, Medical Science Building, Room 4204, 1 King's College Circle, Toronto, ON M5S 1A8, Canada. E-mail: ltrevor.young@utoronto.ca

    Search for more papers by this author


Previously, we found decreased mitochondrial complex I subunits levels and increased protein oxidation and nitration in postmortem prefrontal cortex (PFC) from patients with bipolar disorder (BD) and schizophrenia (SCZ). The objectives of this study were to replicate our findings in an independent sample of subjects with BD, and to examine more specifically oxidative and nitrosative damage to mitochondrial and synaptosomal proteins and lipid peroxidation in myelin. We isolated mitochondria, synaptosomes, and myelin using a percoll gradient from postmortem PFC from patients with BD, SCZ, and healthy controls. Levels of mitochondrial complex I and III proteins, protein oxidation (carbonylation), and nitration (3-nitrotyrosine) were assessed using immunobloting analysis. Lipid peroxidation [lipid hydroperoxides (LPH), 8-isoprostane (8-Iso), 4-hydroxy-2-nonenal (4-HNE)] were measured using colorimetric or ELISA assays. We found decreased complex I subunits levels in BD subjects compared with control (CTL), but no difference in complex III subunits. Carbonylation was increased in synaptosomes from BD group while 3-nitrotyrosine was increased in mitochondria from BD and SCZ groups. 8-Iso was found increased in the BD group while 4-HNE was increased in both SCZ and BD when compared with controls with no differences in LPH. Our results suggest that in BD mitochondrial proteins are more susceptible to potentially reversible nitrosative damage while more longstanding oxidative damage occurs to synaptic proteins.


Oxidative stress has been shown to be higher in the brain of patients with bipolar disorder (BD). Here, we demonstrated increased levels of protein oxidation in synaptosomes from postmortem prefrontal cortex from patients from BD group, while 3-nitrotyrosine was increased in mitochondria from BD and schizophrenia (SCZ) groups. Moreover, lipid peroxidation was found increased in the BD when compared with controls; suggesting that in BD mitochondrial proteins are more susceptible to potentially reversible nitrosative damage while more longstanding oxidative damage occurs to synaptic proteins.

Abbreviations used





bipolar disorder






diffusion tension imaging


electron transport chain


lipid hydroperoxides


prefrontal cortex


postmortem interval




Superoxide dismutase


voltage-dependent anion channel

The prevalence of psychiatric disorders has increased, becoming a major public health concern (Merikangas et al. 2011). NIMH statistics suggest that around 10% of adults in North-America will develop one or more mood disorders in their life-time (www.nimh.gov/statistics), whereas 2.6% of those will develop bipolar disorder (BD). Of importance, health care costs for patients with BD are higher compared to other chronic illness (i.e. asthma, coronary artery disease, and diabetes) (Williams et al. 2011), further, life years lost to disability exceed those from cancer patients (Altamura et al. 2011; Merikangas et al. 2011). Therefore, BD has become a leading health concern (Williams et al. 2010; Merikangas et al. 2011) which raises interests from a number of researchers to investigate the molecular mechanism underling this disorder with the objective of finding targets for development of biomarkers or pharmacological treatments.

BD, like its molecular basis, is a complex illness. Therefore, several hypotheses have been developed (for review Frey et al. 2013). Some of these include: alterations of neurotransmitter levels (Yatham et al. 2005; Berk et al. 2007) and calcium metabolism (Kato 2008; Schulze et al. 2012), increasing of peripheral inflammatory markers (Goldstein et al. 2009; Berk et al. 2011), and reduction of neurons and glia cells density and size (Gigante et al. 2010). Nonetheless, evidence has shown that mitochondrial dysfunction and oxidative damage may be critical to the pathophysiology of BD (Clay et al. 2010; Konradi et al. 2011, 2012; Andreazza 2012). In fact, lower levels of several mRNA encoding for mitochondrial electron transport chain (ETC.) subunits, especially those from complex I, have been reported in postmortem hippocampus (Konradi et al. 2004) and prefrontal cortex (PFC) (Iwamoto et al. 2004; Sun et al. 2006). In our previous work (Andreazza et al. 2010), we reported decreased NADH dehydrogenase (ubiquinone) iron-sulfur protein 7 levels, a mitochondrial complex I subunit, which was associated with decreased complex I activity and increased oxidative damage. A recent re-analysis of microarray studies in postmortem brain point to abnormalities in the ability of complex I subunits such as NDUFS7, NDUFS8, NDUFV1, and NDUFV2 to transfer electrons to ubiquinone in BD (Scola et al. 2013).

Brain tissue is particularly vulnerable to oxidative stress damage as the brain: (i) consumes roughly 20% of total body oxygen; (ii) has a high demand for ATP; and (iii) has a limited ability to detoxify free radicals because of low levels of catalase (Mattson and Liu 2002; Halliwell 2006). In mammalian cells, the main source of reactive oxygen species is ETC. (Halliwell and Gutteridge 2007; Adam-Vizi and Starkov 2010). Oxidative damage to biomolecules occurs when reactive oxygen species overwhelm the antioxidant capacity as described by Sies (1991). For instance, if H2O2, is not removed from the cells by antioxidant enzymes, it can react with Fe2+ and form hydroxyl radicals (OH), which react with lysine, proline, arginine, and threonine residues of proteins to form carbonyl groups (Beal 2002). Another form of oxidative damage to proteins occurs via the nitration of tyrosine induced by peroxinitrite (ONOO) forming 3-Nitrotyrosine (Beckman and Koppenol 1996; Naoi et al. 2005), with the latter being more reversible in comparison to carbonylation damage (Irie et al. 2003; Osoata et al. 2009).

Oxidative stress damage to protein, lipid, and nucleic acids has been demonstrated in BD and reviewed elsewhere (Frey et al. 2013). In a previous study (Andreazza et al. 2010), which highlights the significance of the current one, we have reported increased nitration and oxidation to proteins in mitochondria- and synaptosomes-enriched fraction from PFC of patients with BD. These results might represent a sum of the oxidative and nitrosative damages found in both mitochondria and synaptosomes. The consequences of oxidative damage to mitochondrial or synaptosomal proteins may have different impact on the cellular mechanism as 3-nitrotyrosine can be removed or repaired by denitrases (Irie et al. 2003; Osoata et al. 2009) or prevented by gluthathione antioxidant system (Kirsch et al. 2001; Andreazza et al. 2009) while formation of carbonyl groups has been described as a more lasting change in proteins associated with induction of proteasome-dependent degradation (Grune et al. 1996, 1997; Dalle-Donne et al. 2003; Wong et al. 2010). However, nitrosative modification to mitochondrial cytochrome c can switch the function of this protein from electron transfer to peroxidase activity (Nakagawa et al. 2007; García-Heredia et al. 2010; Ly et al. 2012). As nitration modification can be removed from proteins, it has been reported as post-translational modification that can control protein function and activity (Ly et al. 2012).

Lipids are also target of oxidative stress, and their levels have consistently been shown to be higher in samples from patients with BD (blood: Andreazza et al. 2007, 2008; Versace et al. 2013 and postmortem brain: Wang et al. 2009). Recent results from our group demonstrated a relationship between peripheral levels of lipid hydroperoxide (LPH), a marker of lipid peroxidation, and decreased myelin integrity evaluated by diffusion tension imaging in patients with BD (Versace et al. 2013). Myelin is an easy target to lipid peroxidation because of the high levels of polyunsaturated acids (Versace et al. 2013). Lipid peroxidation of myelin has been implicated in neuronal damage, mitochondrial dysfunction, and decreased white matter integrity (Versace et al. 2013). As the status of lipid oxidative damage in myelin has not been previously studied in myelin from patients with BD, it would be important to analyze the lipid peroxidation in this fraction. Therefore, a more complete and nuanced understanding of the subcellular targets of oxidative stress is of particular interest to fully understand BD. Based on the evidence described above and our previous study (Andreazza et al. 2010), in this study we replicated our findings of decreased NDUFS7 complex I subunit in an independent sample of subjects and measured additional complex I and III components levels in postmortem prefrontal cortex of patients with BD or SCZ in comparison to healthy controls. Furthermore, we evaluated the levels of oxidative and nitrosative damage to proteins in both mitochondrial and synaptosomal fractions, and oxidative damage to lipids in the myelin fraction. Ultimately, the results of this study will guide the identification of oxidative stress targets and how redox balance (i.e. oxidative stress) may impact functions that are critical to the pathophysiology of BD.

Material and methods

Postmortem brain samples

The PFC (BA10, 1.0 g block) samples were generously donated by the Harvard Brain Tissue Resources (http://www.brainbank.mclean.org/). Subjects have provided consent of donation for HBTR. This study was approved by University of Toronto ethics committee (# 26048). Subjects were divided into three groups: BD (n = 16), SCZ (n = 18), and non-psychiatric controls (n = 26). This sample is described in more detail elsewhere (Banigan et al. 2013). Demographic variables, pH, postmortem interval (PMI), and medications for each group are described in the results section (Table 1). Medications prescribed at the time of death were divided into three categories: lithium, anti-depressants (including all categories), and anti-psychotics (including typical and atypical). Two senior psychiatrists using DSM-IV criteria retrospectively established diagnoses. The investigators were blind to group identity, diagnosis, and demographic variables of subjects during all experiments and measurements. Samples were coded numerically in a random manner and the code lifted only during data analyses after all experiments were completed.

Table 1. Demographic variables, pH, postmortem interval (PMI), and medications for control, bipolar disorder, and schizophrenia group
 Healthy controls (n = 26)Bipolar disorder (n = 16)Schizophrenia (= 19) 
  1. a

    One-way anova. Data are shown as mean and standard error or N and percentage.

Age60 ± 2.7958 ± 5.5159.44 ± 2.98Fs = 0.135; = 0.874a
Female11 (42.3%)13 (81.3%)6 (31.6%) 
Male15 (57.5%)3 (18.8%)13 (68.4%) 
pH6.2 ± 0.9536.31 ± 0.206.44 ± 0.12F258 = 0.832; = 0.440
PMI22.65 ± 0.7521.82 ± 1.4224.51 ± 1.64F258 = 0.384; = 0.683
Lithium 6 (37.5%)1 (5.3%) 
Anti-depressant 2 (12.5%)2 (10.5%) 
Anti-psychotics 10 (62.5%)12 (63.2%) 

Myelin, synaptosomal, and mitochondrial extraction

Cellular subfractions were extracted using the standard Percoll gradient density centrifugation method as described by Dunkley et al. (2008). Using this protocol myelin, synaptosomal, and mitochondrial fractions were separated. Postmortem brain samples were homogenized in isotonic sucrose solution (0.32 M Sucrose, 1 mM EDTA, 5 mM Tris, pH7.4) and then applied to a discontinuous Percoll gradient (3, 10, 15, 23%). After centrifugation (31 000 g, 5 min) fractions were separated according to density. The quality of extraction of specific fractions was verified in by western blotting using antibodies against mitochondrial related-protein [voltage-dependent anion channel (VDAC); Abcam, Abcam Inc, Toronto, Canada] and synaptosomal protein (synaptosome-associated 25 kDa, SNAP25; Cell Signalling Technology, Inc., Danvers, MA, USA), respectively.

Levels of mitochondrial electron transport chain subunits in mitochondrial extraction

Subunit proteins (NDUFS7, NDUFS8, NDUFV1, Core2, and cytochrome c) were measured by standard immunoblotting analysis. Ten (NDUFS8, NDUFV1) or fifteen (NDUFS7, Core2 and cytochrome c) μg of proteins were loaded on to 12% acrylamide sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel and transferred to polyvinylidene difluoride membranes. Blots were incubated (2 h, 21°C) with a primary antibody, and were followed by a secondary antibody conjugated to horseradish peroxidase (1 h, 21°C). Immunoreactive bands were detected with ECL reagents and analyzed densitometrically using Versa Doc from Bio-Rad (Bio-Rad Laboratories Ltd., Mississauga, Canada). Loading control was evaluated by levels of immunoblotting the levels of VDAC (voltage-gated channel activity), an established mitochondrial loading control.

Oxidation and nitration levels in mitochondrial and synaptosomal proteins

Oxidative damage was analyzed by measuring protein carbonyl formation using the OxyBlot™ Protein Oxidation Detection kit (Millipore Co, Billerica, MA, USA, #S7151). Protein side chains are derivatized with 2,4-dinitrophenylhydrazone (DNP-hydrazone) by the reaction with 2,4-dinitrophenylhydrazine (DNPH) with carbonyl groups. The DNP-derivatized proteins were separated by polyacrylamide gel electrophoresis (12%) and western blotting, following the kit instructions. Protein bands were analyzed densitometrically and normalized against the intensity of VDAC. To evaluate nitration damage, we assessed the levels of 3-nitrotyrosine using a similar protocol. Loading controls for the 3-nitrotyrosine western blotting was βeta-actin and MemCode reversible protein stain for protein oxidation.

Measures of lipid peroxidation in myelin fraction

Three separate measures of lipid peroxidation were analyzed: LPH (Cayman; Cayman Chemical Company, Ann Arbor, MI, USA, Item-No.705003), 4-HNE (Cell Biolabs, Inc., San Diego, CA, USA; STA-338), and 8-iso (Cayman; Item No. 516351) following manufactures instructions. LPH was extracted from samples using a methanol-chloroform extraction (0°C, 1500 g, 5 min). Following incubation at 21°C with 0.1 unit of chromogen mixture per unit of chloroform extract, the samples were loaded into 96-well plates with absorbance reading at 500 nm. Absorbance was compared with a hydroperoxide standard curve to determine the amount of lipid peroxidation in samples. 4-HNE protein adducts to lysine, histidine, or cysteine were quantified by standard sandwich ELISA using an enzyme immunoassay (Kimura et al. 2005). 8-iso was quantified using ACETM Competitive EIAs with 8-iso-acetylcholinesterase conjugate as a tracer and 8-iso-specific rabbit anti-serum. As 8-iso and the tracer compete for limited anti-serum binding, the color intensity caused by tracer binding was inversely proportional to the amount of 8-iso.

Statistical methods

Statistical analyses were performed using SPSS for Windows version 19.0 software (IBM Corporation, Armonk, NY, USA). Normal distribution of data was determined using Kolmogorov-Smirnov test. The variables NDUFV1 (Z = 1.386; p = 0.043), cytochrome c (Z = 2.301; p = 0.001), NDUFS7 (Z = 1.486; p = 0.024), mitochondrial 3NT (Z = 1.373; p = 0.046), and synaptosomal 3NT (Z = 1.531; p = 0.018) did not follow Gaussian distribution. Other variables followed normal distributions including NDUFS8 (Z = 0.930; p = 0.353), Core 2 subunit (Z = 0.691; p = 0.727), mitochondrial protein carbonyl (Z = 1.114; p = 0.167), synaptosomal protein carbonyl (Z = 1.216; p = 0.104), 8-iso (Z = 0.525; p = 0.946), LPH (Z = 1.084; p = 0.192), 4-HNE (Z = 0.846; p = 0.471). To identify the differences between groups of biochemical variables we used two-way anova followed by Tukey's post-hoc test for parametric variable and Kruskal–Wallis and Mann–Whitney test for non-parametric variables. Differences were considered significant at the type I error rate of 0.05. The relation of biochemical data with gender was measured using independent t-tests, and with PMI by Pearson's correlation test. As the age, gender, pH, and PMI did not show correlation with the biochemical variables we did not use further analyses such as ancova, which would control for these effects. Patients were divided into two groups (treated or not), and compared with controls to verify the effect of medication in the biochemical parameters here assessed.



Demographics and clinical variables details can be found in Table 1 and at http://www.brainbank.mclean.org/. Subjects in the study include participants with BD (n = 16), SCZ (n = 18), and non-psychiatric controls (n = 26). Age ranged from 18 to 86 years old, PMI from 13.4 to 38.3 h, and pH levels from 6 to 9. No significant differences between groups were found for these variables. Patients were prescribed psychotropic medication at the time of death including lithium (n = 7), anti-depressants (n = 4), or anti-psychotics (n = 22) (Table 1).

Decreased levels of ETC complex I subunits related proteins in mitochondrial fraction from postmortem PFC of patients with BD

Consistent with our earlier findings, decreased levels of NDUFS7 (K = 7.779; p = 0.020; Fig. 1a) were found in postmortem PFC samples from subjects with BD compared with the control (U = 94; p = 0.006) and the SCZ (U = 76; p = 0.033) groups. There were no differences comparing the levels of NDUFS7 between SCZ and controls subjects. We also measured the levels of two other complex I proteins, NDUFS8 (Fig. 1b), and NDUFV1 (Fig. 1c). There was a trend (F2,57 = 2.719; p = 0.075) towards decreased NDUFS8 levels in subjects with BD compared with healthy control (p = 0.059), but no differences in NDUFV1 levels between subject groups. We also found a positive correlation between NDUFS7 and NDUFS8 levels in all subjects (r = 0.361; p = 0.005). There were no significant differences in the levels of complex III Core2 subunit or on the electron carrier, cytochrome c, between any of the groups. There was no relationship between age, postmortem delay or pH with any of these measures except for a positive correlation between age and NDUFS8 levels (r = 0.263, p = 0.043) across subjects.

Figure 1.

Levels of Complex I subunits NDUFS7 (a), NUDFV1 (b) and NDUFS8 (c), subunit Core2 from complex III (d) and cytochrome c (e) in mitochondrial proteins from postmortem prefrontal cortex of patients with bipolar disorder (BD), schizophrenia (SCZ), and healthy controls (CTL). Panel (f) illustrated blots for each protein here evaluated. Differences between patients and controls were analyzed using Kruskal–Wallis followed by Mann–Whitney test for NDUFS7, NDUFV1 and CytC (*p < 0.05). One-way anova followed by Tukey's post-hoc test was used to analyze differences between groups for NDUFS8 and Core2 (#p = 0.075).

Oxidative damage is higher in synaptosomes while nitration damage is higher in mitochondria from postmortem PFC from patients with BD

Previous results from our laboratory demonstrated increased levels of carbonyl groups in the BD groups and elevated 3-nitrotyrosine levels in both BD and SCZ group in mitochondrial enriched fractions from postmortem PFC samples (Andreazza et al. 2010). While this tissue fraction was enriched in mitochondria, it did not separate the two main components, synaptosomes and mitochondria, representing a mix from both subcellular fractions. Thus, when we improved the separation of synaptosomes from mitochondria fraction we found increased levels of protein carbonylation in the synaptosomal fraction (F2,58 = 8.217; p = 0.001) from patients with BD (p < 0.001) (Fig. 2a) compared with both the control and SCZ group; carbonylation of proteins in the mitochondrial fraction did not differ between subject groups (F2,58 = 2.259; p = 0.114). Nitration of proteins (i.e. 3-nitrotyrosine), however, had a different profile with increased levels of 3-nitrotyrosine in mitochondria (K = 15.619; p < 0.001) from both BD (U = 74; p = 0.001) and SCZ (U = 112; p = 0.002) subject groups compared with controls (Fig. 2d), but no changes were found between any of the groups in synaptosomes (K = 1.134; p = 0.567; Fig. 2c).

Figure 2.

Protein oxidation (Carbonyl levels; a, c) and nitration (3-nitrotyrosine levels; b, d) in synaptosomal proteins (a, b) and mitochondrial proteins (c, d) from postmortem prefrontal cortex of patients with bipolar disorder (BD), schizophrenia (SCZ), and healthy controls (CTL). Differences between patients and controls were analyzed using Kruskal–Wallis and Mann–Whitney test for 3-nitrotyrosine (**p < 0.001), and one-way anova followed by Tukey's post-hoc test for protein carbonyl (*p < 0.01).

Increased levels of 4-hydroxynonenal and 8-iso in myelin fraction from postmortem PFC from patients with BD

Myelin is largely composed of lipids (70–85%, dry mass) and is an excellent target for lipid peroxidation (O'Brien and Sampson, 1965). Significant differences were found in 8-iso levels between groups (F2,52 = 4.394; p = 0.017) with a significant increase in BD (p = 0.016) compared with controls. 4-HNE was also significantly different between groups (F2,53 = 7.360; p = 0.002) with significant increases in both BD (p = 0.009) and SCZ (p = 0.004) (Fig. 3a–c) compared with controls. In contrast, LPH was not significantly different between subject groups (F2,53 = 0.283; p = 0.755), which suggests that later rather than earlier markers of lipid peroxidation occurs in postmortem PFC from patients with BD. After controlling for covariates (age, gender, PMI, and PH), the results for lipid peroxidation markers remained the same. In addition, 8-iso showed a positive correlation with 4-HNE levels and a negative correlation with complex III core 2 subunit and cytochrome c (Fig. 3d–f).

Figure 3.

Levels of lipid peroxidation in myelin subcellular fraction from postmortem prefrontal cortex of patients with bipolar disorder (BD), schizophrenia (SCZ), and healthy controls (CTL). (a) Lipid hydroperoxyde (LPH); (b) 8-isoprostane (8-iso); (c) 4-hydronxynonenal (4-HNE). (d) Correlation between 8-Isoprostane and 4-HNE levels, (e) Correlation between 8-Isoprostane and complex I subunit NDUFV1 and (f) Correlation between 8-iso and core 2 subunit levels. Differences between patients and controls were analyzed using one-way anova followed by Tukey's post-hoc test; *p < 0.05, **p < 0.01.

SOD2 and GPx1 were not different between subject groups

Superoxide dismutase (SOD) reacts with O2 to form H2O2, which is then converted into H2O by GPx. If H2O2 cannot be converted into H2O, it is able to react with Fe++ and form OH, inducing protein oxidation. Similarly, O2 can react with NO and form ONOO, leading to nitrosative damage. Therefore, we evaluated the levels of SOD and GPx in both fractions. SOD is a family with t major isoforms. SOD1 is located in the cytoplasm, SOD2 is in both the cytoplasm and the mitochondria, while SOD3 is an extracellular enzyme. Accordingly, we analyzed only SOD1 and SOD2, but not SOD3 (McCord and Fridovich 1988). For GPx enzymes, we decide to measure the most abundant GPx in the brain which is the GPx1 (Power and Blumbergs, 2009). There were no differences in SOD2 or GPx1 levels between groups in either mitochondria (or synaptosomes Figure S1). In addition, we measured the levels of SOD1 in synaptosomes and found no differences between groups (Figure S2).

Effect of medication at the time of death on oxidative stress markers

There was no effect of treatment with anti-psychotic, lithium or anti-depressants at the time of death on levels of complex I or III subunits and protein nitration or carbonylation levels. There was no effect of treatment with any of these psychotropic drugs on nitration of synaptosomal proteins or on any of the mitochondrial measures or on lipid peroxidation in the myelin fraction.


Mitochondrial dysfunction, which leads to oxidative damage to cells, in PFC and hippocampus has been identified as an important contributing factor to the pathophysiology of BD (Wang 2007; Clay et al. 2010; Gigante et al. 2010; Andreazza 2012). In this study, we replicated the finding of decreased NDUFS7 in PFC from patients with BD in an independent sample of postmortem brain tissue. We demonstrated that mitochondrial proteins are more vulnerable to nitration damage in both BD and SCZ, while synaptosomal proteins are more prone to oxidation particularly in BD. It was revealed that myelin fraction is a target of lasting lipid peroxidation in these illnesses. Together, these results support the hypothesis that mitochondrial dysfunction occurs in prefrontal cortex in patients with BD and that it leads to nitrosative damage in mitochondria and lasting oxidative damage in synapses. Our findings also suggest that mitochondrial dysfunction may subsequently induce oxidative damage to white matter, and that this damage can be targeted by the specific treatments for this disorder. All of these findings will facilitate our understanding of the specific molecular pathophysiology of BD.

Previous work from our laboratory and others has shown decreased mRNA expression of several mitochondrial complex I subunits, including NDUFS7 (Iwamoto et al. 2005; Sun et al. 2006), NDUFS1 (Konradi et al. 2004), NDUFV1 (Konradi et al. 2004), NDUFS8 (Sun et al. 2006), and NDUFS3 (Konradi et al. 2004; Iwamoto et al. 2005). Complex I is formed by 44–46 subunits with four subcomplexes (Lenaz 2001; Green and Kroemer 2004; Reeve et al. 2008), including λ subcomplex located in the hydrophilic arm through iron-sulphur clusters, which transfers electrons from NADH to ubiquinone (Janssen et al. 2006; Scola et al. 2013). The transfer starts by NDUFV1 subunit, followed by NDUFS1, NDUFS8, and NDUFS7 (Janssen et al. 2006; Scola et al. 2013). To understand if impaired λ subcomplex is associated with BD we previously re-analyzed data from microarray studies from patients with BD or SCZ and found abnormal expression of λ subcomplex subunits including NDUFS1, NDUFS7, NDUFV1, and NDUFS8 in BD but not SCZ. This suggests that impairment of the electron transfer process in complex I is specific to BD (Scola et al. 2013). In this study, we found decreased NDUFS7 levels and a trend toward decreased levels of NDUFS8 in BD; a positive correlation between NDUFS7 and NDUFS8, but no significant differences for NDUFV1; and complex III core II subunit and cytochrome c. NDUFS7 and NDUFS8 are important in electron transfer and further exploration of the function of these subunits may be relevant for understanding BD.

In an earlier study, we reported increased levels of protein oxidation and nitration in protein extracts enriched with mitochondria and synaptosomes from postmortem PFC from patients with BD and SCZ (Andreazza et al. 2010). Here, we separated synaptosomal and mitochondrial proteins to understand the role of oxidative damage to proteins in subcellular fractions. We successfully extracted both fractions, in addition to myelin, and found increased levels of protein oxidation in synaptosomal proteins from BD with a trend for increased levels in SCZ group. Meanwhile, nitration levels were increased in mitochondrial proteins in both BD and SCZ group. These results suggest that mitochondrial and synaptosomal related proteins are affected differently by oxidative stress. One explanation for these findings may be related to different levels of activity of antioxidant enzymes. However, in this study we found no difference in the expression of SOD and GPx in either mitochondrial or synaptosomal fraction, which suggests that other processes are involved in the regulation of nitration or oxidation to protein in these samples.

Myelin is composed of 70–80% lipids and is rich in polyunsaturated fatty acids, which are preferential targets for lipid peroxidation because of the easy access to carbon-carbon double bonds (Rydén et al. 2009). Lipid peroxidation occurs as a chain reaction, starting with the formation of LPH and progressing until two hydroperoxide radicals react and form a non-radical species, such as 4-HNE and 8-iso (Halliwell and Gutteridge 1984, 2007; Halliwell 1992, 2006; Bongarzone et al. 1995). LPH can be detoxified from the cells by antioxidant enzymes, such as glutathione peroxidase and glutathione-S-transferase, which helps prevent advanced damage to lipids (Halliwell and Gutteridge 1984, 2007; Halliwell 1992, 2006; Bongarzone et al. 1995). Recently, our group reported that increased serum levels of LPH present a positive relationship with impairment of myelin fraction detected by diffusion tension imaging (Versace et al. 2013). Here, we demonstrated increased levels of end-products (8-iso and 4-HNE) in myelin fraction from PFC of patients with BD, which is consistent with this imaging data. The present data provide direct evidence of lipid peroxidation in the myelin fraction from subjects with BD.

Our results did not demonstrate an association of treatment with lithium or anti-psychotics at the time of death on the oxidative stress damage markers evaluated here. Both lithium and anti-psychotics have been reported to modulate oxidative stress as reviewed by Andreazza and Young (2013). Indeed, lithium has been reported to have antioxidant properties as demonstrated by its ability to attenuate cell death induced by rotenone, a complex I inhibitor, or H2O2 (King and Jope 2005; Lai et al. 2006). Tan and Wang (2011) showed that lithium inhibits the ability of amphetamine to both decrease complex I activity and to form 4-HNE adducts to proteins in rat frontal cortex. Both typical and atypical anti-psychotics were shown to inhibit complex I activity in SH-SY5Y neuroblastoma (Brenner-Lavie et al. 2008), lymphoblastoid cell lines from patients with SCZ or healthy controls (Rosenfeld et al. 2011), Additionally, chronic treatment with haloperidol or clozapine increases the levels of protein carbonylation in rat brain (Agostinho et al. 2007, 2011). How effects in animal studies translate to findings in patients remains an unanswered question. There are also many questions about how much can be interpreted from the knowledge that patients were prescribed anti-psychotics at the time of death. Future studies are necessary to investigate further the effect of lithium and anti-psychotics in oxidative stress parameters.

Bipolar disorder and schizophrenia share clinical features and genetic vulnerability (Scola et al. 2013). Therefore, it is expected that there will be shared or common biological findings underlying their pathophysiology. Indeed, our lab previously reported increased 3-nitrotyrosine levels in both BD and SCZ subjects as well as elevated levels of carbonylation in patients with BD in mitochondrial enriched fractions from postmortem PFC samples (Andreazza et al. 2010). While this tissue fraction was enriched in mitochondria, the two main components, synaptosomes and mitochondria, were not separated. Even when using a more specific method of extraction, we were able to confirm the specificity of protein oxidation for the BD subjects while changes to nitration were found in both BD and SCZ. In support of these results, increased levels of 3-nitrotyrosine have also been reported in blood samples from patients with either BD (Andreazza et al. 2009) or SCZ (Dietrich-Muszalska et al. 2012). The complexity of this scenario is highlighted by Scola et al. (2013) where they suggested that both BD and SCZ may share downregulation of complex I subunits; however, when subunits were analyzed separately or grouped by function, the findings for BD appear to be particularly marked on acceptance of electrons, while the findings in SCZ are less specific. Future studies might focus on identifying specific protein targets for oxidative or nitrosative damage in each illness to further clarify the shared and unique pathophysiology of each disorder.

While interpreting our results, it is important to reflect on the limitations of our sample and methods. Postmortem brain samples are valuable tools to study psychiatric illness as they allow direct measurement of affected tissue. This approach, however, is accompanied by a number of potential issues such as the effect of PMI, pH, and storage conditions of the tissue, which could affect the results of this and others studies. To control for these factors, we have correlated our results with PMI and pH levels and did not find significant relationships with these variables and the main outcomes of this study. Previous studies using animals demonstrated that oxidative damage to proteins do not suffer alterations until 40 h of PMI (Crecelius et al. 2008). In addition, the results reported here are only from the analysis of a single region; the PFC and other brain regions maybe altered differently. Nonetheless, the main findings from the literature are been from the PFC. This study was able to replicate, in a new set of postmortem brain samples, our previous findings of increased levels of oxidative damage to proteins from PFC. It will be necessary to evaluate oxidative damage to protein in different brain areas to further understand the region-specificity of oxidative stress in psychiatry disorders in future studies.

In conclusion, our results provide compelling evidence that oxidative damage might be involved in synaptic modifications in BD and nitration of tyrosine in mitochondrial regulation. As well as, lasting modification lipid peroxidation are present in myelin fraction and correlate with redox modification of complex I observed in mitochondrial proteins. Future studies are necessary to investigate which protein are targets of oxidative or nitrosative modifications and how lipid peroxidation might contribute to this process. Oxidative damage to protein induces alterations in protein conformation, which can change the physiological protein interaction. Studies investigating the impact of oxidative stress to protein networks are crucial to further understand the involvement of oxidative stress in psychiatry. Ultimately, deepening our knowledge on oxidative stress mechanism in the brain will guide us towards rational design of new drugs and/or biomarkers for psychiatric illness.


This study is supported by grants from the Canadian Institutes of Health Research (A.C.A, L.T.Y. and J.F.W.), and NARSAD Young Investigator awards (A.C.A.). Specimens were donated by Harvard Brain Tissue Resource Center courtesy of Drs. Francine M. Benes, John C. Hedreen and Wilson Woo, which is supported in part by PHS grant number R24 MH068855.

Disclosure of potential conflicts of interest

There are no potential conflicts of interest.