Elevated plasma homocysteine concentration has been suggested as a risk factor for schizophrenia, but the results of epidemiological studies have been inconsistent. The most extensively studied genetic variant in the homocysteine metabolism is the 677C>T polymorphism in the methylenetetrahydrofolate reductase (MTHFR) gene, resulting in reduced enzyme activity and, subsequently, in elevated homocysteine. A meta-analysis of eight retrospective studies (812 cases and 2113 control subjects) was carried out to examine the association between homocysteine and schizophrenia. In addition, a meta-analysis of 10 studies (2265 cases and 2721 control subjects) on the homozygous (TT) genotype of the MTHFR 677C>T polymorphism was carried out to assess if this association is causal. A 5 μmol/l higher homocysteine level was associated with a 70% (95% confidence interval, CI: 27–129) higher risk of schizophrenia. The TT genotype was associated with a 36% (95% CI: 7–72) higher risk of schizophrenia compared to the CC genotype. The performed meta-analyses showed no evidence of publication bias or excessive influence attributable to any given study. In conclusion, our study provides evidence for an association of homocysteine with schizophrenia. The elevated risk of schizophrenia associated with the homozygous genotype of the MTHFR 677C>T polymorphism provides support for causality between a disturbed homocysteine metabolism and risk of schizophrenia.
It is commonly accepted that both genetic and environmental factors contribute to the aetiology and clinical phenotype of schizophrenia. Despite numerous family, twin, and adoption studies in which evidence for hereditary and environmental contributions to schizophrenia is presented,1, 2, 3, 4 no underlying inherited mechanism has yet been identified. Although several putative susceptibility genes for schizophrenia have been identified,5 the pathogenic mechanisms remain to be resolved. Evidence of an association between schizophrenia and a specific aberrant metabolism could reveal new candidate genes and might provide more insight in the complex aetiology of schizophrenia.
Dysfunctional single-carbon transfer reactions involving the methionine–homocysteine metabolism have been proposed to be relevant in the aetiology of schizophrenia. Evolving from the original transmethylation hypothesis postulating that an abnormally (toxic) methylated metabolite might be the cause of schizophrenia,6 it was suggested subsequently that disturbances in the single-carbon metabolism itself might be causative.7 Methylation processes are vital for normal cell functioning because of their key role in protein, lipid and DNA metabolism, gene expression, synthesis of neurotransmitters and detoxification processes.8 Initial clinical evidence for an impaired methyl-carbon pathway in schizophrenic patients stems from many methionine administration studies ranging from the 1960s until the early 1990s. These studies showed that treatment with high-daily doses of L-methionine resulted in exacerbation of schizophrenic symptoms.9, 10 In addition, as compared to controls the rate of whole body methylation was found to be decreased in patients with schizophrenia.11, 12 It was suggested that these findings might reflect an enzymatic defect or a failure of a control mechanism of the single-carbon cycle in schizophrenic patients.13 The finding of a defect in methylation due to severe dysfunction of 5,10-methylenetetrahydrofolate reductase (MTHFR) resulting in hyperhomocysteinemia in a subject with schizophrenia-like symptoms14 was therefore of great interest. Homocysteine is a sulphur containing amino-acid derived from the demethylation of the essential amino-acid methionine, which is the only dietary precursor of homocysteine. The metabolism of homocysteine depends on several B-vitamins including folate, cobalamin, pyridoxine and riboflavin. MTHFR is the essential enzyme in the folate-mediated single-carbon transfer reactions. A common polymorphism exists in the catalytic domain of the MTHFR enzyme. The gene is localized on chromosome 1p36.3,15 in which a C>T substitution at position 677 results in a substitution of alanine to valine.16 The reported prevalence of the homozygous 677C>T (TT) genotype is between 0 and 32% of the population world-wide,17 ranging from 5 to 15% in Caucasian populations.18 The single amino-acid substitution results in impaired flavin adenine dinucleotide (FAD) binding, leading to loss of folate resulting, in its turn, in reduced activity of MTHFR.19 Subjects with the TT genotype have about 25% higher homocysteine levels than those with the normal homozygous (CC) genotype,18 whereby the impact of the polymorphism varies according to folate or riboflavin status.20, 21
Elucidation of an association between a genetic variant associated with elevated homocysteine levels and schizophrenia might be informative about the hypothesis that higher levels of homocysteine play a causal role in the aetiology of schizophrenia. Individual case–control studies concerning homocysteine or the 677C>T polymorphism included too few cases and controls to produce reliable evidence for an association with schizophrenia.
The aim of this study was to examine the association between homocysteine and schizophrenia by conducting a meta-analysis of case–control studies on this topic. In order to assess whether such association is causal, we have also performed a meta-analysis of all case–control observational studies with available data on the MTHFR 677C>T polymorphism and schizophrenia.
Eligible studies were identified by searching the electronic NLM MEDLINE database for relevant reports published up to and including December 2004 using the search terms (‘homocysteine’ or ‘hyperhomocysteinemia’ or ‘MTHFR’) and (‘schizophrenia’). Additionally, the reference lists of original articles on this topic were scanned to identify articles missed by the computerized search.
Studies examining total serum or plasma homocysteine levels or the 677C>T polymorphism in the MTHFR gene were included in the meta-analysis. The following additional criteria were used to select studies for data extraction: (a) a diagnosis of schizophrenia for cases, (b) use of a case–control design and (c) publication in an English-language, peer-reviewed, indexed scientific journal. Meeting abstracts were not allowed. We also included our recent published data.22
Data extraction and synthesis
For the meta-analysis of homocysteine and schizophrenia, information was extracted from each study on: the odds ratio (OR) and 95% confidence interval (CI) as a risk estimate for the association between elevated homocysteine and schizophrenia and the number of cases and controls. If the OR and 95% CI were not provided, they were calculated from the number of cases and controls above and below a particular homocysteine concentration as a cutoff point from the control distribution in each study using Woolf's method.23 The homocysteine studies included in our meta-analysis reported different definitions of hyperhomocysteinemia by using different cutoff levels. Most studies provided the 90th percentile of homocysteine levels in the control group, while other studies reported the 95th or 75th percentile. For pooling of risk estimates, it is essential that these estimates are based on a single unit of comparison. Previous studies on homocysteine and homocysteine-related diseases reported linear dose–effects relations.24, 25, 26 In line with these findings, we calculated the OR for a 5 μmol/l increase in homocysteine level, a standard reference increment. To estimate the log OR for a 5 μmol/l increase in homocysteine, information about the normal distribution was used to calculate the expected mean level of homocysteine in the groups being compared. For example, the mean value in the top fifth of a normal distribution is 1.4 standard deviation (s.d.) above the mean, and the mean in the bottom fifth is 1.4 s.d. below the mean. Hence, the OR comparing the top to the bottom fifth is also the OR for a 2.8 s.d. difference in homocysteine levels. Wherever possible, the s.d. in controls was used to estimate the difference in μmol/l between the top and bottom fifth of a study, but if this was not reported, then a weighted average was used of the s.d. from the studies that did report it. Assuming the association was log–linear, it was then possible to calculate the OR for a unit change in homocysteine and hence for a 5 μmol/l increase in homocysteine.
For the meta-analysis of the MTHFR 677C>T polymorphism and schizophrenia, data were collected on the frequency of CC, CT and TT genotypes in cases and controls or calculated from the allele frequencies. Subjects with the TT genotype have approximately 2.7 μmol/l (25%) higher homocysteine levels compared to subjects with the CC genotype, while the mean difference of homocysteine concentration between the CT and CC genotype is only about 0.29 μmol/l.25Therefore, the TT genotype was assigned as the risk genotype, because of the relatively strong contrast in homocysteine level between the TT and CC genotype. The study-specific ORs and 95% CIs for TT compared with CC genotype were estimated using Woolf's method.23
By combining the pooled risk estimate of the association between the TT genotype and schizophrenia (OR) with the published estimate of a mean difference in plasma homocysteine between TT and CC genotype (Δhcy), we derived an equivalent risk estimate for a 5 μmol/l increase in homocysteine concentration of OR5/Δhcy (raising OR to the power of 5/Δhcy). Subsequently, we compared the derived and measured risk estimates.
Random effects model meta-analyses were performed according to the methods of DerSimonian and Laird,27 and 95% CIs were constructed by using Woolf's method.23 The significance of the pooled ORs was determined by the z-test. Heterogeneity (greater variation among study results than would be expected through change) was assessed using standard χ2 tests. The influence of individual studies on the pooled ORs was determined by sequentially removing each study and recalculating the pooled OR and 95% CI. The individual studies were ordered by the inverse of the standard error (s.e.) of the risk estimate to evaluate publication bias.28 A symmetrical inverted funnel implies absence of significant selection or publication bias. In addition, publication bias was tested by Egger's test.28 If necessary, data were completed by personal communication with the original authors. The type I error rate was set at 0.05. All statistical analyses were conducted by using Stata 8.0.
Homocysteine and schizophrenia
Data for the meta-analysis were obtained from eight published case–control studies22, 29, 30, 31, 32, 33, 34, 35 with a total number of 812 cases and 2113 control subjects. Figure 1 shows the ORs and 95% CIs of schizophrenia associated with a 5 μmol/l increase in measured plasma total homocysteine, separately for individual studies and for all studies when taken together. A 5 μmol/l increase in measured homocysteine was associated with an overall 70% (95% CI: 27–129) increased risk of schizophrenia. Sequential removing and replacement of individual studies from the calculation of the pooled OR produced risk estimates ranging from 1.58 to 1.94 with 95% CIs that always encompassed 1.0. This finding indicates that the significance of the pooled OR was not excessively influenced by any single study. There was significant heterogeneity between the results of the studies included (χ2=17.949; df=7; P=0.012), suggesting the presence of some moderating variable. Several characteristics of the included homocysteine studies are presented in Table 1, showing heterogeneity of exclusion criteria and of age and sex distributions. In Figure 1 the distribution of the ORs of the individual homocysteine studies are ordered by the inverse of the standard error of the risk estimate. It indicates that much of the evidence for an association of homocysteine and schizophrenia comes from two larger studies which deviate significantly from an OR of 1.0.32, 34 The shape of the funnel also suggests that a few smaller studies finding an inverse association may not have been published. However, the effect of these studies on the overall risk estimate is likely to be small. In addition, Egger's regression asymmetry test showed no evidence of publication bias (a=−0.9; t=−0.73; P=0.49).
MTHFR and schizophrenia
A total of nine published articles22, 31, 33, 36, 37, 38, 39, 40, 41 reported on the relationship between schizophrenia and the MTHFR 677C>T polymorphism. One article provided data on two separate studies.40 Thus, data were obtained from 10 studies involving 2265 cases and 2721 control subjects. All cases were diagnosed according to standard diagnostic systems (DSM-III, DSM-V or ICD-9). All studies used a standardized method to determine the MTHFR 677C>T genotypes. Figure 2 shows the results of individual and pooled studies for the TT genotype. Overall, the TT genotype was associated with a 36% (95% CI: 7–72) increased risk of schizophrenia compared with the CC genotype. Sequential removing and replacement of individual studies from the calculation of the pooled OR produced risk estimates ranging from 1.26 to 1.43 with 95% CIs that always encompassed 1.0. This finding indicates that the significance of the pooled OR was not excessively influenced by any single study. There was no significant heterogeneity among the results of individual studies (χ2=13.973; df=9; P=0.123). The absence of asymmetry in the distribution of the ORs of the individual studies as shown in Figure 2, and the outcome of the Egger regression asymmetry test (a=−0.06; t=−0.04; P=0.97) reduce the likelihood of publication bias.
Homocysteine and MTHFR
Based on the published estimate from a large meta-analysis on homocysteine and MTHFR studies,25 we assumed that the average homocysteine concentration is 2.7 μmol/l higher in individuals with the TT genotype than in subjects with the CC genotype. In line with this, the pooled OR of 1.36 (95% CI: 1.07–1.72) for TT versus CC genotype is equivalent to an OR of 1.77 (95% CI: 1.13–2.73) for the standard 5 μmol/l increase in homocysteine concentration (raising 1.36 to the power of 5/2.7). This risk estimate is similar to the observed OR of 1.70 in our meta-analysis.
The overall result of the present meta-analysis demonstrates that for each 5 μmol/l increase in homocysteine concentration the relative risk of schizophrenia is roughly doubled (OR=1.70; 95% CI: 1.27–2.29). This finding must be interpreted with some caution because of the observed significant heterogeneity among the homocysteine studies in this meta-analysis. However, the association of homozygosity of the 677C>T polymorphism in the MTHFR gene and schizophrenia with an OR of 1.36 (95% CI: 1.07–1.72) is suggestive for a causal relation between aberrant homocysteine metabolism and schizophrenia.
A limitation of using a meta-analytic approach for population-based observational studies is the fact that these studies may yield estimates of associations that are influenced by confounding due to age, sex or ethnic admixture (population stratification) either between studies or between cases and controls within each study.42 In this meta-analysis, only the unadjusted ORs could be calculated for each study because the adjusted ORs were not routinely provided. There might also be confounding because of the clinical heterogeneity of the patients included. Most studies were restricted to schizophrenia patients, but some studies also included cases with schizophreniform or schizoaffective disorders. It is unlikely that this has been an important moderating variable, because of the low number of included patients with these diagnoses.
Plasma homocysteine concentrations reflect genetic and environmental factors such as diet, B-vitamin intake and life-style factors like coffee consumption and smoking.43 None of the studies on homocysteine levels and risk of schizophrenia explicitly reported on all critical factors in the assessment of homocysteine. The distribution of these confounders may have differed among studies and by introducing systematic error it might explain the observed heterogeneity among studies. In order to gain further insight in the association between homocysteine and schizophrenia, it will be of utmost importance to standardize the measurement of homocysteine. The most obvious factor that contributes to heterogeneity of homocysteine levels is the folate and riboflavin status of the study population.43 In a meta-analysis on MTHFR and venous thrombosis a clear difference was seen between European studies and North-American studies,26 which could be explained by the implementation of folate enrichment of foods in North-America since 1997. In our meta-analysis, no risk estimates by continent were calculated, because of the low number of included North-American studies.
Studies of genetic variants that affect homocysteine levels would reflect long-term exposure to elevated homocysteine, and be independent of confounding and concerns about reverse causality.44 Apart from the effect modification by dietary intake of folate and riboflavin, it is unlikely that the effects of these genotypes on schizophrenia will be influenced by lifestyle or other confounders. Genetic confounding by linkage disequilibrium is possible, whereby a gene linked to the MTHFR gene controls an unknown risk factor for schizophrenia and also increases the homocysteine level. Nevertheless, no such linkage has been found thus far. The studies included in this meta-analysis involved less than a few hundred cases for each study, and so the CI around the individual ORs were wide. The present study provides more reliable evidence as to the importance of the TT genotype for the risk of schizophrenia. The concordance of the risk estimates from homocysteine studies and MTHFR 677C>T polymorphism studies concerning schizophrenia supports the hypothesis that the association of elevated homocysteine levels with this disease is in part causal.
Hyperhomocysteinemia is not specific for schizophrenia, and has been reported as a risk factor for various diseases such as cardiovascular diseases,25 neural tube defects,45 and neuropsychiatric disorders including Alzheimer's disease,46 Parkinson's disease47 and depression,48, 49, 50 stressing the pivotal role of the single-carbon mechanism. The pathogenic mechanisms underlying these moderate hyperhomocysteinemic diseases are not yet fully understood. Interestingly, accumulating evidence from in vitro and animal studies suggests that the nervous system (neuronal homeostasis) is, in particular, sensitive to folate deprivation and raised homocysteine levels.51, 52 Extra-cellular homocysteine has been found to be toxic to cultured neurons and neuron cells via stimulation of NMDA receptors, enhancing oxidative stress and inducing DNA damage eventually resulting in apoptosis and adverse effects on synaptic and glial function.51, 53, 54 Recently, experiments with mice have shown that these mechanisms are not only important in cultured embryonic premitotic neurons, but in the adult postmitotic neurons as well.55 The deleterious effect of MTHFR dysfunction on brain development was observed in individuals with a severe deficiency of this enzyme, resulting in delayed psychomotor development, mental retardation and psychiatric symptoms.56
The 70% higher risk of schizophrenia for a 5 μmol/l increase in homocysteine observed in our meta-analysis suggests a dose–response relationship. This finding provides a rational of using homocysteine levels as a biological parameter for intervention studies. A meta-analysis of short-term trials comparing the effects on homocysteine levels of different doses of B-vitamins has shown that a daily dose between 0.5 to 5 mg of folic acid was associated with a 25% lower homocysteine concentration and that vitamin B12 had an additive effect of about 7%.57 This decrease in homocysteine of 25% associated with folic acid administration is comparable to the difference in homocysteine levels between individuals with the TT and CC genotype.18, 25 Folate supplementation might therefore contribute to symptom reduction in patients with schizophrenia, as a result of lowering homocysteine levels. Thus far, only one double-blind, placebo-controlled clinical trial with folate supplements in schizophrenic patients has been published.58 Supplementation of folate improved clinical and social recovery in schizophrenia patients with low red blood cell folate at the start of the trial. Periconceptional folic acid supplementation substantially reduces the risk of neural tube defects59 possibly by protecting the developing embryo against one or more folate-related metabolic disorders. Evidence is accumulating that an aberrant neurodevelopmental process, starting in utero might also be present in the pathogenic pathway of schizophrenia.60 Our finding of the TT genotype as a risk factor for schizophrenia supports the view that homocysteine is causally related to this disorder. One can speculate if periconceptional folic acid supplementation reduces the risk of the development of schizophrenia if a disturbance in folate-dependent homocysteine metabolism is present, either in the mother or her child.
The present meta-analysis provides evidence that elevated homocysteine levels and the TT genotype of the MTHFR 677C>T polymorphism, contribute to the susceptibility of schizophrenia. Trials of folic acid and vitamin B12 supplementation for schizophrenia patients are warranted. Such clinical trials need to include a sufficiently large number of participants to have adequate power to detect improvement in schizophrenic patients. Future research should focus on genetic factors and metabolites associated with the homocysteine pathway in mother and child, because of the possible impact of this metabolism on intra-uterine brain development.
Martin den Heijer is recipient of a VENI-grant from the Netherlands Foundation of Scientific Research.