Department of Psychiatry and Behavioral Sciences, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan
Correspondence: Masanari Itokawa, MD, PhD, Department of Psychiatry and Behavioral Sciences, Tokyo Metropolitan Institute of Medical Science, 2-1-6 Kamikitazawa, Setagaya, Tokyo 156-8506, Japan. Email: firstname.lastname@example.org
Appropriate biological treatment and psychosocial support are essential to achieve and maintain recovery for patients with schizophrenia. Despite extensive efforts to clarify the underlying disease mechanisms, the main cause and pathophysiology of schizophrenia remain unclear. This is due in large part to disease heterogeneity, which results in biochemical differences within a single disease entity. Other factors include variability across clinical symptoms and disease course, along with varied risk factors and treatment responses. Although schizophrenia's positive symptoms are largely managed through treatment with atypical antipsychotics, new classes of drugs are needed to address the unmet medical need for improving cognitive dysfunction and promoting recovery of negative symptoms in these patients. Accumulation of toxic reactive dicarbonyls, such as methylglyoxal, are typical indicators of carbonyl stress, and result in the modification of proteins and the formation of advanced glycation end products, such as pentosidine. In June 2010, we reported on idiopathic carbonyl stress in a subpopulation of schizophrenia patients, leading to a failure of metabolic systems with plasma pentosidine accumulation and serum pyridoxal depletion. Our findings suggest two markers, pentosidine and pyridoxal, as beneficial for distinguishing a specific subgroup of schizophrenics. We believe that this information, derived from in vitro and in vivo studies, is beneficial in the search for personalized and hopefully more effective treatment regimens in schizophrenia. Here, we define a subtype of schizophrenia based on carbonyl stress and the potential for using carbonyl stress as a biomarker in the challenge of overcoming heterogeneity in schizophrenia treatment.
Schizophrenia is a chronic and disabling brain disorder, thought to be a heterogeneous syndrome. The lifetime prevalence is approximately one percent, with onset of disease frequently occurring in early adulthood.[1, 2] Numerous studies suggest oxidative stress and redox dysregulation as risk factors for the development of schizophrenia.[3-5] In addition to a genetic predisposition to disease, quantitative and qualitative resistance to stress interact with environmental elements and genetic factors, leading to a distorted metabolic balance in the body. Although several potential candidate genes and chromosomal linkage loci for schizophrenia have been identified,[7-10] little information is available on their molecular and cellular intermediates.[11-15]
Dicarbonyls, such as methylglyoxal (MG), a potent protein-glycating agent, are formed from sugars, lipids and amino acids.[16-19] This dicarbonyl accumulation, which is referred to as carbonyl stress, modifies proteins and leads to the eventual formation of advanced glycation end products (AGE). The formation of AGE is associated with three different pathways in vivo, namely the Maillard reaction, polyol pathway and lipid peroxidation.
Some reports in humans implicate increased AGE in a variety of illnesses, including diabetes,[22, 23] hemodialysis and mental illness.[25-27] Cellular removal of dicarbonyls and AGE occurs via a glutathione-dependent glyoxalase, namely, glyoxalase I (Glo1) (Fig. 1). The glyoxalase detoxification system is found in all tissues, including the brain. It is also known that this system interacts with several metabolizing cascades, some of which have been reported as possible causative factors in the etiology of mental disorders.[28, 29] Studies in humans have revealed reduced GLO1 mRNA levels in peripheral leukocytes in patients with mood disorder, genetic association of polymorphisms of GLO1 with autism and panic disorder, altered frequencies of GLO1 enzyme activity phenotypes in alcoholism, and altered GLO1 mRNA levels in post-mortem brains of patients with Alzheimer's disease.
In June 2010, we reported enhanced carbonyl stress as a disease feature in a subpopulation of schizophrenics. We found a 1.7-fold increased concentration of plasma pentosidine, a marker for AGE, and significantly decreased levels of pyridoxal, one of the three forms of vitamin B6, in schizophrenia compared with healthy control subjects. We detected heterozygous frameshift mutations with loss-of-function and moderate-effect polymorphisms in GLO1. These changes resulted in a 40–50% and 15–20% reduction in Glo1 activity in loss of function and moderate-effect mutants, respectively. Our results suggested involvement of accumulated dicarbonyls and/or pentosidine in the pathophysiology of schizophrenia patients carrying these genetic GLO1-deficits.
We also reported on a drug-naïve patient with at-risk mental state, who exhibited enhanced carbonyl stress, with high plasma pentosidine levels, suggesting the presence of stress before the onset of disease. A follow-up study showed that decreased plasma pentosidine levels were accompanied by improved psychotic symptoms on follow up, compared with initial observations. Interestingly, plasma pentosidine levels are significantly lower in outpatients than in hospitalized cases, while Glo1 activity is significantly upregulated in outpatients (Arai et al., unpublished findings). Schizophrenics with accumulated pentosidine and depleted pyridoxal showed distinct clinical features, such as a higher propensity to inpatient status, low educational status, higher frequency and longer durations of hospitalization and higher doses of anti-psychotic medication.
Carbonyl stress as identified in patients with schizophrenia is a key concept for clarifying some of the pathogenic and pathological mechanisms in schizophrenia.
A robust susceptibility gene for schizophrenia has yet to be identified, although candidate polymorphisms and chromosomal regions associated with this disease have been reported from multi-center genome-wide association studies, conducted using large-scale samples.[7, 8, 38] A serious problem in the genetic research of schizophrenia is that of heterogeneity. Disease heterogeneity complicates the search for molecular patterns and subsequently hinders the identification of disease mechanisms.
Specific biomarkers would be useful for classifying heterogeneous psychotic syndromes into homogeneous subtypes, thereby improving disease diagnosis, as well as clinical staging and prognostic information on cognitive function.[39, 40] In general, it is difficult to capture information of the brain from peripheral blood. However, multiple omics research projects using peripheral blood and urine have proven to be very effective tools for identifying new causative genes, environmental factors and disease metabolites. The classification of schizophrenia into subgroups using biomarkers is the current research goal, ultimately leading to the discovery of new classes of therapeutic agents and the advent of personalized medicine.
The condensation of carbonyl groups with amine groups leads to the formation of a reversible Schiff base. Subsequent reactions, such as oxidation and peroxidation, result in the irreversible final products known as AGE. These include Nδ-(5-hydro-5-methyl-4-imidazolon-2-yl)-ornithine (MG-H1), N-ε-(Carboxymethyl) lysine (CML), Nε-(1-carboxyethyl) lysine (CEL), 1,3-di(Nε-lysino)-4-methyl-imidazolium (MOLD) and pentosidine[42-44] (Fig. 2). Methylglyoxal (MG), glyoxal and 3-deoxyglucosone (3-DG) enhance AGE formation. Methionine sulfoxide, dityrosine and 3-nitrotyrosine are also markers for protein oxidation and nitration. The structures of AGE were determined from peripheral blood, urine and other tissue samples. Typical AGE are primarily monitored using combinations of high-performance liquid chromatography (HPLC) and mass spectrometry (LC-MS/MS).[18, 46] In our studies, pentosidine level was measured using HPLC techniques described in our previous publications.[35-37]
Discovery of Enhanced Carbonyl Stress in Schizophrenia
MG is a reactive glycating agent and a precursor of major quantitative adducts formed from proteins. It is a more reactive α-oxoaldehyde metabolite, formed after the degradation of triosephosphates glyceraldehyde 3-phosphate and dihydroxyacetone phosphate, by MG synthase in bacteria, the oxidation of acetone, ketone body metabolism, catabolism of threonine and degradation of glycated proteins. Accumulation of MG induces activation and increased degradation of proteins, DNA mutagenesis and cytotoxicity.[48, 49] Therefore, efficient metabolism and detoxification of MG is needed to prevent protein and DNA damage in cells. The glyoxalase pathway represents the major route for MG metabolism, as shown in Figure 1, and is supported by aldose reductase and aldehyde dehydrogenases.[50, 51] In the glyoxalase pathway, Glo1 catalyzes the isomerization of hemithioacetals, formed non-enzymatically from MG and reduced glutathione, to S-D-lactoylglutathione. Next, Glo 2 catalyzes the hydrolysis of S-D-lactoylglutathione, to d-lactate and glutathione (GSH).
On 2 June 2006, we identified a novel schizophrenia case with a deficiency of GLO1, and published this report in Archives of General Psychiatry (AGP) (2010). This patient was a 60-year-old man suffering from severe schizophrenia, with a brother affected by schizophrenia, a sibling who committed suicide and two maternal uncles who suffered from schizophrenia. We identified a novel frameshift mutation in this case (case #1). The c.79_80insA mutation causes a frameshift in GLO1. The adenine insertion at nucleotide 79, in exon 1 resulted in a frameshift at codon 27 and introduced a premature termination codon after aberrant translation of 15 amino acid residues (p.T27NfsX15). One year later, we found a new c.365delC mutation in another schizophrenic patient (case #2). The mutation generated a frameshift in codon 122 of exon 4 in GLO1 and a premature termination after the addition of an aberrant 27 amino acid peptide (p. P122LfsX27) (Fig. 3).
Individuals with heterozygous frameshift mutations showed an approximately 40–50% reduction of Glo1 enzymatic activity in both erythrocytes and lymphocytes, supporting the belief that GLO1 genetic mutations occur in one allele. We suspected that aberrant transcripts from the mutant allele were degraded by nonsense-mediated mRNA decay, as the predicted truncated proteins from p.T27NfsX15 and p.P122LfsX27 were not detected by immuno-blotting. Lymphoblastoid cells derived from subjects carrying heterozygous frameshift mutations showed significantly decreased levels of GLO1 mRNA, compared with those from both affected and unaffected individuals without heterozygous frameshift mutations. Glo1 enzymatic activity in heterozygous frameshift mutation carriers was lower than that of wild-type individuals (2.8 munit/106 red blood cells [RBC] in case #1, 3.0 munit/106 RBC in case #2, 6.2 munit/106 RBC in wild-type healthy control subjects).
As expected, heterozygous frameshift carriers with schizophrenia displayed increased concentrations of plasma pentosidine. Levels of plasma pentosidine detected by HPLC, a quantitatively accurate and stable technique, were 137 ng/mL in case #1 and 74.7 ng/mL in case #2. These high plasma pentosidine levels in the two patients carrying the mutation are most likely caused by reduced Glo1 function, as neither suffers from diabetes mellitus nor renal failure, two of the major causes of enhanced AGE formation.
Interestingly, these rises in pentosidine were concomitant with prominent decreases of pyridoxal, a major isoform of vitamin B6. Vitamin B6 plays an important role in various physiological functions and is converted to pyridoxine, pyridoxal and pyridoxamine in vivo. Pyridoxamine reacts with RCO by carbonyl-amine chemistry (Fig. 1). In the two GLO1 frameshift mutation carriers, pyridoxal was markedly reduced, probably due to its depletion after detoxifying RCO. We found decreased pyridoxal levels in both cases compared with wild-type individuals (<2.0 ng/mL in case #1, 2.8 ng/mL in case #2, 11.9 ng/mL in wild-type healthy controls). The depletion of pyridoxal may reflect a state of carbonyl stress in these patients.
The GLO1 gene also harbors a Glu111 (a major genotype)/Ala111 polymorphism. The frequency of the Ala111 allele exhibits high population diversity. We assayed the missense variants, Glu111/Glu111, Glu111/Ala111 and Ala111/Ala111, using 1586 schizophrenics. We identified nine homozygous Ala111 carriers, who showed approximately 20–30% lower Glo1 enzyme activity against the wild-type group. Mean values of enzymatic activity were approximately 4.7 munit/106 RBC in these patients. Patients carrying homozygous Ala111/Ala111 alleles also showed higher plasma pentosidine levels (mean value of plasma pentosidine, 98.0 ng/mL), in the absence of diabetes mellitus or renal failure.
In order to investigate whether GLO1 genotypes show modified enzymatic activity, we performed functional analyses of mutants, using recombinant GFP-tagged Glo1 proteins generated by site-directed mutagenesis. Our in vitro transfection study revealed an approximately 16% reduction in enzymatic activity for Ala111 carriers over Glu111 carriers, implying an intrinsically lower enzymatic activity for the Ala111/Ala111 genotype compared with the Glu111/Glu111 genotype. Recombinants expressing either GFP-T27NfsX15 or GFP-P122LfsX27 exhibited basal protein levels, indicating that the two frameshift mutations completely abolish enzymatic activity of Glo1 protein.
Our findings suggested that markedly low Glo1 activity and a concomitant increase in carbonyl stress elicited by heterozygous frameshift mutations and the homozygous Ala111 genotype in GLO1 could be causative for schizophrenia in a small subset of patients. This provided the first evidence to support the multiple, rare variants–common disease hypothesis in schizophrenia, where the causative mechanism, in this case carbonyl stress, derived from rare large-effect mutations in a multiple affected pedigree could be replicated by a moderate-effect polymorphism in the general population.
Replication of Idiopathic Carbonyl Stress in a Subpopulation of Schizophrenia
Higher levels of plasma pentosidine were observed in 46.7% patients with schizophrenia (21 out of 45 cases), as reported in AGP (2010). On the other hand, no healthy control subjects displayed increased pentosidine accumulation. The cut-off value of pentosidine was defined as the average value of healthy controls plus 2SD, 55.2 ng/mL. The mean pentosidine level of 68.4 ng/mL was 1.7-fold higher in schizophrenic patients compared with healthy controls. In addition to pentosidine accumulation, we found that pyridoxal concentrations were also significantly reduced in patients with schizophrenia. Roughly 24% (n = 11) of 45 cases showed marked reductions in pyridoxal levels, with four cases showing less than 2 ng/mL. Mean pyridoxal values were significantly reduced in schizophrenic patients (7.5 ng/mL in schizophrenia vs 11.9 ng/mL in healthy subjects). The mean pyridoxal level in three patients carrying heterozygous frameshift mutations was approximately 3.6 ng/mL. Our results showed that pentosidine accumulation (odds ratio [OR]: 25.8, 95% confidence interval [CI]: 5.6–118.8, P < 0.0001) as well as pyridoxal reduction (OR: 10.6, 95%CI: 3.9–28.3, P < 0.0001) increased the risks for developing schizophrenia.
In a Psychiatry and Clinical Neurosciences (PCN) (2014) publication, we increased the sample size and validated our association of enhanced carbonyl stress with schizophrenia. We recruited 156 outpatients using DSM-IV criteria for schizophrenia or schizoaffective disorder, and 221 age-matched healthy control subjects. Plasma pentosidine concentrations in patients and healthy control subjects were 67.7 ng/mL and 41.9 ng/mL, respectively (Fig. 4). The mean pentosidine level from schizophrenic patients was 1.6-fold higher than that of healthy control subjects. Conversely, the mean pyridoxal value was significantly reduced in schizophrenic patients compared with controls (schizophrenia, 7.7 ng/mL; healthy control subjects, 10.2 ng/mL). We applied a cut-off point to differentiate between a carbonyl stress group and a non-stress group, set at the average value in healthy controls, plus 2SD. We found that 11.8% of patients with schizophrenia also displayed enhanced carbonyl stress. On the other hand, carbonyl stress was present in only 0.01% of healthy control subjects. These results are consistent with our previous report in AGP (2010), replicating the presence of idiopathic carbonyl stress in a subpopulation of patients with schizophrenia.
Carbonyl Stress as a Disease State Marker
We assayed alterations of pentosidine, Glo1 enzymatic activity and pyridoxal, using samples from both in- and outpatients with a follow-up period of 3–5 years. Reduced pentosidine and increased Glo1 enzymatic activity were observed in discharged cases. However, no significant changes in these parameters were observed in hospitalized patients. These levels of pentosidine and Glo1 enzymatic activity provide a suggestion that upregulation of Glo1 metabolism leads to inhibition of carbonyl stress and improves some symptoms for patients with schizophrenia. If we can find bioactives for improving dysfunctions of Glo1 metabolism, it may also prove to be of therapeutic value.
A drug-naïve patient with at-risk mental state (ARMS) who exhibited enhanced carbonyl stress with high plasma pentosidine levels highlighted the possibility that carbonyl stress existed before the onset of disease, as reported in PCN (2011). The patient was a 21-year-old man who first developed obsessive thoughts at age 18 and sought medical help for communication difficulties and depression. He was diagnosed with ARMS, according to Structured Interview for Prodromal Syndromes criteria, 10 months after his initial consultation. At his initial visit, his symptoms and pentosidine levels were measured, and then rechecked after 16 months of treatment. Plasma pentosidine levels showed a remarkable decrease from 113.2 ng/mL to 44.1 ng/mL. His clinical score dropped from 84 to 58 on the total Positive and Negative Syndrome Scale (PANSS), a fall promoted to a great extent by biweekly counseling and psychotherapy. Negative subscale scores and general psychopathology subscale scores also improved from 22 to 9, and from 42 to 27, respectively. Global assessment of functioning changed from 55 to 65.
Unfortunately, his latest results show increased pentosidine levels of 300 ng/mL, and he has been admitted to hospital. This finding further supports our hypothesis that carbonyl stress status can help characterize the psychosis risk state for a subset of patients with schizophrenia. Pentosidine levels may constitute a genuine biomarker for the state of disease, if carbonyl stress is confirmed as being directly linked to schizophrenic signs and symptoms.
Clinical Features of Schizophrenia Exhibiting Carbonyl Stress
Recently, we examined the clinical characteristics of patients with carbonyl stress, and psychopathological symptoms using PANSS. In this study, patients were divided into four groups according to levels of pentosidine and pyridoxal: normal pentosidine and normal pyridoxal levels (group 1), normal pentosidine and low pyridoxal (group 2), high pentosidine and normal pyridoxal (group 3) and high pentosidine and low pyridoxal (group 4). Our hope was that this classification based on biomarkers would be able to distinguish homogenous patient subsets with specific clinical characteristics from the heterogeneous disease population. As mentioned earlier, more accurate disease subdivisions should pave the way for developing personalized medicine in schizophrenia.
We evaluated the symptom severity of 49 consenting patients. Examining 30 items of PANSS, we found significant negative correlations between serum pyridoxal levels and items of hostility, emotional withdrawal, passive/apathetic social withdrawal, tension, mannerism/posturing, disorientation and active social avoidance. Furthermore, we also found significant correlations between the total general psychopathological score and serum pyridoxal levels. These correlations would support the novel therapeutic idea of pyridoxamine supplementation for carbonyl stress-related schizophrenia.
The clinical characteristics of schizophrenia with carbonyl stress suggest targeting this stress as a new therapy for psychiatric disorders. We regard pyridoxamine, a non-toxic, water-soluble vitamin B6, as a novel medicine for schizophrenia with carbonyl stress, primarily because it inhibits the formation of AGE. Recently, Katsuta et al. (2014) also investigated the relation between carbonyl stress markers and clinical characteristics of acute-stage schizophrenic patients in a cross-sectional and longitudinal study. The authors put forward vitamin B6 supplementation as a candidate for augmentation therapy in a subpopulation of schizophrenics who showed vitamin B6 depletion.
The enhanced carbonyl stress cohort, group 4, showed a more severe clinical course relative to the normal cohort, group 1 (Table 1). The proportions of subjects classified as inpatients were 23.9% in group 1 and 78.6% in group 4. Patients in group 4 showed a longer duration of hospitalization (4.2 years vs 17.4 years), higher daily doses of anti-psychotics (773.8 mg/day vs 1143.9 mg/day) and lower educational status (13 years vs 11.7 years) (Table 1). We also observed a 1.5-fold higher number of hospitalizations in group 4 compared with group 1 subjects. In this study, we could not exclude the possibility that high plasma pentosidine levels were a consequence of high doses and long exposure to anti-psychotic medication. Therefore, further studies focused on drug-naïve patients will be required to address these issues.
Table 1. Clinical features of schizophrenics with and without carbonyl stress (modified and adapted from Miyashita et al.37)
Group 1 (n = 67)
Group 4 (n = 26)
Fold change vs group 1
Normal pentosidine and normal pyridoxal
High pentosidine and low pyridoxal
*P < 0.05 (vs group 1); **P < 0.001 (vs group 1); ***P < 0.0001 (vs group 1).
†Cut-off point for high plasma pentosidine levels, 62.9 ng/mL (the mean + 2SD of healthy controls). ‡Low pyridoxal levels, <6 ng/mL (male) and <4 ng/mL (female).
Therapeutic Potential of Pyridoxamine to Prevent Carbonyl Stress
As mentioned above, vitamin B6 consists of three components, namely, pyridoxamine, pyridoxine and pyridoxal. Of the three, only pyridoxamine is able to prevent AGE accumulation by amine-chemistry. To further validate our hypothesis, we are planning a phase II clinical trial for carbonyl stress-induced schizophrenia, using pyridoxamine. Using evidence-based medicine, we are attempting to adapt translational research into medical care for carbonyl stress-induced schizophrenia.
In addition to vitamin B6 metabolism, Glo1 metabolites are linked to other pathways, such as those involving glutathione, methionine, homocysteine, folate and amino acids, as described in AGP eFigure 1.35 Glo1 metabolism is also associated with glycolysis, the pentose phosphate pathway and lipid/protein metabolism. In our preliminary data, schizophrenics show lower folic acid concentrations and higher levels of homocysteine, relative to healthy control subjects (folic acid, χ2 = 42.21, P < 0.0001, OR = 8.84, 95%CI = 4.16–18.81; homocysteine, χ2 = 24.76, P < 0.0001, OR = 7.13, 95%CI = 3.12–16.28) (Fig. 5). Patients with higher homocysteine and lower folate levels showed high pentosidine accumulation and pyridoxal depletion. These results imply an imbalance of one-carbon metabolism in schizophrenics with altered Glo1 metabolism. These metabolic systems are functionally integrated and act homeostatically. Therefore, for us to understand the full diversity of biological systems that exists in both the diseased and healthy states, it is important to evaluate combined data from in vitro and in vivo sources.
Our study is the first investigation into GLO1 alterations and enzymatic activity in schizophrenics, and to highlight its link to idiopathic carbonyl stress, defined as pentosidine accumulation and pyridoxal depletion. Although more studies are needed to understand the molecular mechanisms precipitated by carbonyl stress in the central nervous system, carbonyl stress undoubtedly represents a new target for medication without neurotransmitter-based concepts in the treatment of schizophrenia. In particular, schizophrenic patients with low pyridoxal and high pentosidine may well benefit clinically from pyridoxamine treatment. A better understanding of the molecular mechanisms that promote the pathophysiology of carbonyl stress-related schizophrenia could drive improvements in difficult-to-treat negative symptoms and cognitive dysfunction, thereby improving the quality of life for patients.
Future omics studies combining comprehensive molecular and clinical information will lead to new discoveries, allowing for more accurate diagnostic subdivisions for schizophrenia and subsequent treatment regimens.
This study was supported by a grant from the Strategic Young Researcher Overseas Visits Program for Accelerating Brain Circulation by the Japan Society for the Promotion of Science fellowship: The collaborative study between Japan and United Kingdom on molecular mechanisms of glycation and oxidative stress in psychiatric disorders. We especially thank Professor Paul J. Thornalley and Dr Naila Rabbani. We thank members of the host research teams for technical assistance: Drs Mingzhan Xue, Jinit Masania, Fozia Shaheen and Attia Anwar in Clinical Sciences Research Laboratories, Warwick Medical School, University of Warwick, University Hospital. We also thank Hiroko Yuzawa at the Institute of Medical Sciences, Tokai University for the measurement of plasma pentosidine levels. We are grateful for the expert technical assistance of Izumi Nohara, Mayumi Arai and Nanako Obata. Additionally, we thank Ikuyo Kito and Sachie Kogoku at Tokyo Metropolitan Institute of Medical Science for assistance with the preparation of this manuscript.