Chromatin Alterations Associated With Down-regulated Metabolic Gene Expression in the Prefrontal Cortex of Subjects With Schizophrenia FREE

Dysfunction, hypoactivity, and hypometabolism of the prefrontal cortex (PFC) may contribute to the negative symptoms and cognitive deficits of schizophrenia.^1- 6 The molecular pathogenesis of prefrontal dysfunction in schizophrenia is still not clear, but the down-regulated expression of a subset of metabolic genes is thought to be involved.^7- 9 Alterations in cellular metabolism may then further compromise orderly gene expression, affecting neurotransmission,^10- 20 myelination,^10,21- 22 and other functions.^23- 24 There is evidence that in individual schizophrenic patients, PFC hypometabolism persists for decades,²⁵ suggesting that the underlying molecular mechanisms, including dysregulated metabolic gene expression, remain stable for extended periods.

In eukaryotes, the rate-limiting biochemical response that leads to the activation of gene expression involves alterations in chromatin structure.²⁶ The 4 core histones, H2A, H2B, H3, and H4, together with 147 base pairs of genomic DNA wrapped around them, compose the nucleosomes, the basic units of chromatin. Chromatin fibers are composed of arrays of nucleosomes connected by linker histones and DNA.^26- 27 Dynamic changes in chromatin conformation and accessibility of transcription factors are highly regulated by the N-terminal histone tails.^28- 29 Covalent modifications at histone N-terminal tails are differentially regulated in chromatin at sites of active gene expression compared with inactive and silenced chromatin.^28- 29 For example, the methylation of histone H3 at the arginine 17 position (H3meR17), or the phosphoacetylation of H3 at the serine 10/lysine 14 position (H3pS10-acK14), defines the open chromatin state and actual or potential transcription.^28- 34 Studies in rodents showed that treatment with (1) antipsychotic drugs that block dopamine D₂-like receptors,³⁵ (2) D₁-agonists,³⁶ or (3) the anticonvulsant and mood-stabilizing drug valproate sodium^37- 39 induces histone modification changes not only at defined genomic sequences but also on a global and genome-wide level in selected areas of the forebrain. The drug-induced, coordinated regulation of histone chemistry in whole or “bulk” chromatin is thought to have profound effects on nuclear signaling and chromatin function, including transcription and cellular differentiation.^28- 29,40 The D₂-like antagonist drugs and valproate sodium are frequently prescribed for the treatment of schizophrenia and other major psychiatric diseases,^41- 43 which raises the question of whether chromatin modifications are involved in the molecular mechanism of action of these drugs. Furthermore, in eukaryotes, cellular metabolism is tightly coupled to chromatin structure because many chromatin-remodeling complexes require adenosine triphosphatase activity.^44- 45 In addition, caloric restriction up-regulates histone deacetylase activity, which in turn leads to transcriptional silencing of chromatin.^46- 47 Given this background, we hypothesized that hypometabolism and decreased metabolic gene expression in the PFC of subjects with schizophrenia is accompanied by abnormal levels of 1 or several covalent histone modifications. These changes may reflect an adaptive response to prefrontal hypoactivity, may play a causative role in prefrontal dysfunction, or both. In any case, chromatin-related abnormalities are likely to have profound and lasting implications for cellular function and PFC circuitry. Furthermore, molecular studies on chromatin from diseased brain is pivotal to gain further insight into epigenetic factors that are thought to be involved in the etiology of the schizophrenia.^48- 52 Herein, we identify a subgroup of subjects with schizophrenia affected by down-regulated expression of 4 of 16 metabolic genes in conjunction with high levels of histone H3 methylation (H3meR17). To our knowledge, this is the first evidence that histone modifications as epigenetic regulators of gene expression may contribute to prefrontal dysfunction in schizophrenia.

According to a recent study by Middleton and colleagues,⁷ a subset of 10 metabolic genes are expressed at decreased levels in the PFC of subjects with schizophrenia. This set of 10 transcripts includes malate dehydrogenase I (MDH), for which 4 studies reported either a significant decrease^7- 9 or increase¹⁰ in the PFC of subjects with schizophrenia. Therefore, we sought (1) to examine the expression of these metabolic genes in a larger cohort of schizophrenic subjects and controls and (2) to determine whether altered metabolic gene expression is linked to an abnormal histone modification profile. This study was conducted in 2 parts (Figure 1). First, 6 different histone modifications and 16 metabolic gene transcripts were profiled in 21 matched pairs of schizophrenic subjects and controls. For the overall cohort of 21 subjects with schizophrenia, we did not find significant changes in histone or transcript levels. Therefore, as the next step, we defined subgroups of subjects with schizophrenia based on histone modification levels. Then, the association between modifications and down-regulated metabolic gene expression was further tested by conducting additional studies on another 20 pairs of subjects with schizophrenia and controls collected independently of the 21 matched pairs in part 1.

The present study used the brains of 82 subjects, or 41 matched pairs of subjects with schizophrenia and controls (Table 1). All subjects were matched for age and autolysis time (±15%), and 38 of 41 pairs were also matched for sex, as described previously.⁵³ The entire matching process was finished before any experimental procedures were performed. Of the 41 pairs, 21 were obtained from a brain collection at the University of California and were used in previous studies^12,53 of GAD67 gene expression and white matter neuron distribution. Key findings reported for this postmortem collection were independently replicated.^{15- 17,54- 58} Therefore, this postmortem cohort seems to be representative of the disease. Another 20 pairs were obtained from a brain bank at the University of Maryland that includes a tissue collection from subjects diagnosed as having schizophrenia.^55,59- 60 For both brain banks, procedures for tissue collection, neuropathologic examination (to rule out degenerative and neurologic disease), diagnosing schizophrenia using DSM-IV–based diagnostic criteria, and selection of subjects with schizophrenia and controls were described in detail in previous publications.^{12,53,59- 61} Controls were defined as people without a lifetime diagnosis of serious mental illness other than alcohol abuse or substance abuse or dependence; 4 controls and 3 subjects with schizophrenia had positive toxicologic findings for alcohol, narcotics, or both. None of the cases in this study were known to be individuals with human immunodeficiency virus or hepatitis B or who were in comas or receiving artificial respiration before death. All tissue samples were fresh-frozen and kept at –75°C until further processing. All procedures were approved by the institutional review boards of the University of Massachusetts, the University of Maryland, and the University of California at Davis. From each brain, small (0.5- to 1.0-cm³) blocks of tissue were obtained from the rostral pole of the left and right frontal lobes (Brodmann area A10), which is part of the PFC and is affected by functional hypoactivity and hypometabolism.^1- 6

Tissue samples of PFC were homogenized in 0.2M sulfuric acid and incubated on wet ice in the same solution for 45 minutes at a concentration of approximately 2 mg DNA/mL. The samples were then pelleted by centrifugation, and a supernatant containing acid-soluble proteins was added to one-third volume of 100% trichloroacetic acid, mixed well to precipitate histones, pelleted, washed in 100% acetone/0.05M hydrochloride, resuspended in double-distilled H₂0, and stored at –75°C until use. Sample concentrations were determined in triplicate using a protein assay (Pierce Biotechnology Inc, Rockford, Ill) in conjunction with a microtiter plate reader (Benchmark; Bio-Rad Laboratories, Hercules, Calif), adjusted accordingly and eluted in 1× Laemmli buffer to a concentration of 10 μg/μL. In addition, equal loading was controlled by gel Coomassie blue stain. Samples were analyzed using sodium dodecyl sulfate–polyacrylamide gel electrophoresis and immunoblotting, using polyvinyl difluoride membranes (Bio-Rad Laboratories) with a 0.2-μm pore size to ensure efficient blotting of the histones, which have molecular weights in the range of 10 to 15 kDa. For immunoblotting, membranes were preblocked in Tris-buffered saline, with 0.1% Tween-20 (TBS-T) (pH, 7.5), with 5% dry milk (Bio-Rad Laboratories), incubated overnight in primary antibody in TBS-T, washed repeatedly, and incubated for 1 hour in peroxidase-conjugated secondary goat anti–rabbit antibody (1:200) (Amersham Biosciences, Piscataway, NJ), and immunoreactive bands were then visualized by chemiluminescence using a charged-coupled device camera–based system (ChemiDoc; Bio-Rad Laboratories) and film autoradiograms. All data were normalized to input (optical density per micrograms of protein). The following primary antibodies were used (working dilution): (1) anti-H3-[(phospho)Ser10,(acetyl)Lys14] = H3pS10-acK14 (1:500); (2) anti-H3-(acetyl)Lys9-Lys14 = H3acK9/14 (1:1500); (3) anti-H3-(methyl)Lys4 = H3meK4 (1:500); (4) anti-H3-(methyl)Arg17 = H3meR17 (1:150); (5) anti-H4-(acetyl)Lys8 = H4acK8 (1:6000); and (6) anti-H4-(acetyl)Lys12 = H4acK12 (1:1000). All primary antibodies were polyclonal (rabbit) and were purchased from Upstate Biotechnology Inc (Waltham, Mass). The site-specific histone modifications are shown in Figure 2.

Procedures for tissue fixation and immunohistochemical analysis on free-floating sections have been described previously.⁵³ The final working dilution of primary antibodies ranged from 1:250 (anti-H3pS10-acK14 and anti-H3meR17 antibodies) to 1:1000 (anti-H4acK12 antibody). We tested antibody specificity by adding synthetic peptides containing the first 21 residues of histone H3 or H4 together with a site-specific modification (Upstate Biotechnology) to the primary antibody incubation solution. Immunoreactivity was selectively abolished on immunoblots and in tissue sections by peptides (final concentration, 1.5 μg/mL) carrying the epitope recognized by the site- and modification-specific antibody.

Total RNA from approximately 100 to 200 mg of PFC tissue samples was prepared using the Trizol Reagent (Invitrogen, Carlsbad, Calif), and 10 μg was used as the template for biotin-labeled complementary DNA (cDNA) probe synthesis by annealing to GEAprimer (SuperArray Biosciences Corporation, Frederick, Md) at 70°C and then incubating for 10 minutes at 42°C with biotin-16-2′-deoxy-uridine-5′-triphosphate (GIBCO, Invitrogen Corporation, Grand Island, NY) and Moloney murine leukemia reverse transcriptase (RT) (Promega, Madison, Wis). This procedure omits polymerase chain reaction (PCR)–based amplification of the cDNA template, which could result in skewing and distortion of differences between samples. Custom-made GEArray membranes (SuperArray Biosciences Corporation) were placed in prehybridization buffer (GEAhyb) with salmon sperm DNA, 100 μg/mL, then labeled probe was added for overnight hybridization at 68°C. The membranes were washed in sodium chloride/sodium citrate (SSC)/1% sodium dodecyl sulfate with increasing stringency (2×–0.1×SSC) at 68°C and then were further processed for chemiluminescence with alkaline phosphatase–conjugated streptavidin and then incubated with chemiluminescence substrate (CDP-Star; PerkinElmer, Boston, Mass). Chemiluminescence was detected by using electrochemiluminescence autoradiography film (Amersham), and hybridization signals were quantified by densitometry using Quantity One software (Bio-Rad Laboratories).

The custom-made GEArray membranes contained 20 cDNAs spotted in duplicate and a plasmid (pUC18) as negative control (Table 2). The array included cDNAs of 17 metabolic genes, including the 10 metabolic genes that in a recent study⁷ were decreased in the PFC of patients with schizophrenia. Our array included 3 control cDNAs: (1) β-actin for internal normalization; (2) the α-subunit of calcium/calmodulin-dependent protein kinase II (CAMIIK)^12,61; and (3) the “housekeeping” ribosomal protein (RPL13A). Because the hybridization signal for 1 metabolic gene, ATP5A1, was in all cases and controls less than 2 × background, it was not considered for further analysis. Thus, the present study reports data for 16 metabolic genes.

Total RNA (0.5 μg) was used for cDNA synthesis and real-time RT-PCR using the Platinum Quantitative RT-PCR ThermoScript One-Step System (Invitrogen), an Opticon 2 cycler, and Opticon software (MJ Research, Waltham), together with FAM dye-labeled TaqMan MGB probes specific for coactivator-associated arginine methyltransferase 1 (CARM1), crystallin (CRYM), cytochrome c1 (CYTOC/CYC1), malate dehydrogenase 1 (MDH), ornithine aminotransferase (OAT), and peptidyl arginine deiminase (PADI4/PAD4) and for β-actin for normalization (Taqman Assays-on-Demand Gene Expression Products; Applied Biosystems, Foster City, Calif) and according to the manufacturer’s instructions using 6mM magnesium sulfate and the following cycling protocol: 50°C for 50 minutes × 1, 95°C for 3 minutes × 1, and 95°C for 15 seconds followed by 60°C for 60 seconds × 45 cycles. Data were expressed relative to custom-made standards.

Expression of each gene was calculated to correct for potential differences in RNA input and PCR primer efficiency using the following equation^62- 63:

V = (E reference)^CT-reference / (E target)^CT-target

where V indicates the relative value of the target transcript normalized to reference transcript (β-actin); E, primer slopes; and CT, the threshold crossing cycle.

Adult 6- to 8-week-old outbred mice from a mixed genetic background, predominantly Sv129 /J and C57BL/6J, were treated for 14 days with (1) the conventional antipsychotic drug haloperidol, (2) the atypical drug risperidone, or (3) saline as a control. For each treatment, 6 animals were selected, and sex-matched littermates were matched to different treatment arms. The male-female ratio was 1:1 for each treatment group. Haloperidol and risperidone were administered intraperitoneally with saline as vehicle (injection volume, 2 mL/kg body weight) at a dose of 1 mg/kg twice daily at 8 AM and 4 PM. Two hours after the last treatment, the animals were killed and the brain was removed, and the medial rostral cortex was dissected and frozen until homogenized and further processed for histone immunoblotting as described previously herein.

Associations between histone modification or transcript levels and age, brain pH, and postmortem interval were evaluated using the Pearson correlation coefficient. All histone and gene expression data from each subject with schizophrenia are expressed relative to the matched control to test for significant changes in gene expression and histone modifications in the total cohort or in subgroups of subjects with schizophrenia. The matched-pair design was necessary because the study was conducted for 3 years in a large number of postmortem samples (N = 82), with data being collected from many different immunoblots, cDNA array membranes, and real-time RT-PCR experiments. For each experiment, the sample from a subject with schizophrenia was processed in parallel with the sample from the matched control. Subgroups of schizophrenic subjects were defined by differences in histone modification levels compared with controls. Because nothing is known about the regulation of histone modifications in human brain, our subgrouping rationale was guided by studies^64- 71 that used histone acetylation levels to subgroup certain types of carcinomas and adenomas. The significance of differences in metabolic gene transcripts in defined subgroups of subjects with schizophrenia were examined using a nonparametric Mann-Whitney test and an exact permutation distribution.⁷² The significance of case-control differences was evaluated using 1-sample t tests of the hypothesis that the mean difference was zero. Analyses were conducted using a statistical software program (StatXact version 6).^73- 74

We profiled histone modifications in the PFC using site- and modification-specific antibodies against the tail sequence of 2 core histones, H3 and H4. We focused on the 6 antibodies that fulfilled the criterion of consistent and robust immunoreactivity in our postmortem material: (1) an antibody against dimodified H3 molecules, defined by the phosphorylation of serine 10 and acetylation of lysine 14, H3pS10-acK14; (2) an antibody that binds to H3 acetylated at lysine 9 or lysine 14, H3acK9/14; (3) an antibody against H3 methylated at lysine 4, H3meK4; (4) an antibody against H3 methylated at arginine 17, H3meR17; (5) an antibody that recognizes histone H4 acetylated at lysine 8, H4acK8; and (6) an antibody against H4 acetylated at lysine 12, H4acK12. The 6 histone modifications are illustrated in Figure 2, and representative examples of immunoblots are shown in Figure 3. To confirm that these antihistone antibodies specifically label nuclei, we examined the cellular and laminar distribution pattern of immunoreactivities using histochemical analysis. Each of the 6 antihistone antibodies resulted in robust immunoreactivity of nuclei in the PFC (Figure 4). Immunoreactivity for H3meR17 was primarily confined to cortical gray matter. At higher magnification power, intense H3meR17 labeling was apparent in large, presumably neuronal nuclei (Figure 4E and F). In contrast, robust immunoreactivity for other types of histone modifications, including H4acK12, was expressed in large and small, presumably nonneuronal nuclei (Figure 4I and J).

Levels for each of the 6 histone modifications, including H3meR17, differed less than 15% between subjects with schizophrenia and controls, and these changes were not statistically significant (Figure 5A). We conclude that subjects with schizophrenia, as a group, do not show consistent alterations in global levels of methylated, acetylated, and phosphoacetylated histones in the PFC.

We measured, in the postmortem cohort, transcript levels of metabolic genes that were thought to be involved in PFC dysfunction (Table 2). Our rationale was 2-fold. First, we wanted to find out whether the reported alterations in metabolic gene expression^7- 10 are representative of a larger postmortem cohort. Second, we wanted to examine whether changes in metabolic gene expression are associated with histone modification changes. Examples of hybridization signals on the arrays are shown in Figure 6. When the total group of 21 subjects with schizophrenia was compared with the 21 controls, none of the metabolic and control genes showed a statistically significant difference between groups (Figure 5B). We conclude that in larger cohorts, the PFC of subjects with schizophrenia is not consistently affected by down-regulated metabolic gene expression.

Although little is known about the regulation of histone modifications in human brain, alterations in histone modification levels, including H3 and H4 acetylation, were used to define subgroups for certain diseases, including adenomas and carcinomas.^64- 71 In analogy to these examples from clinical pathology, we conducted subgroup analyses for each histone modification separately. We dichotomized the cohort of subjects with schizophrenia into subgroups based on modified histone levels of 1.3 or greater relative to matched controls. After the first part of the study, which was conducted on 21 matched pairs, we identified 6 subjects with schizophrenia who showed H3meR17 levels of 1.3 or greater relative to their matched controls (Schizo-H3meR17^≥1.3C). This subgroup of subjects with schizophrenia (n = 6) showed a significant decrease in the expression of 3 metabolic transcripts, CRYM, CYTOC/CYC1, and MDH (t test, P = .03), and a tendency for a decrease in OAT transcripts (P < .1) compared with the remaining 15 patients with schizophrenia. Next, we sought to confirm the association between H3 methylation and down-regulated gene expression for CRYM, CYTOC/CYC1, MDH, and OAT transcripts in the PFC of subjects with schizophrenia. To this end, we studied an additional set of 40 samples consisting of 20 matched pairs of subjects with schizophrenia and controls. Immunoreactivity for H3meR17 was again measured using immunoblotting. The messenger RNA levels for each of the 4 genes were measured using real-time RT-PCR and FAM dye-labeled TaqMan MGB probes. Of these additional 20 subjects with schizophrenia, 2 showed H3meR17 levels of 1.3 or greater relative to their matched controls. Thus, 8 (20%) of 41 subjects with schizophrenia showed levels of H3meR17 of 1.3 or greater relative to their matched controls. We first compared gene expression differences between the Schiz-H3meR17^≥1.3C subgroup (n = 8) and the remaining 33 subjects with schizophrenia (Figure 7A). For each of the 41 subjects with schizophrenia, levels of metabolic transcript were normalized to those of matched controls. The 8 subjects in the Schiz-H3meR17^≥1.3C subgroup, but not the remaining subjects with schizophrenia (the Schiz-H3meR17^<1.3C subgroup), were consistently affected by decreased metabolic gene expression (Figure 7A). Each case was scored from –4 to 0, with –4 defining cases that showed a decrease in 4 of 4 transcripts relative to matched controls. The Schiz-H3meR17^≥1.3C subgroup (n = 8) scored significantly lower compared with the Schiz-H3meR17^<1.3C subgroup (n = 33) (mean ± SEM, –3.63 ± 0.26 vs –1.7 ± 0.25; Mann-Whitney z = –3.24; P < .005, 2-tailed). We conclude that expression of a set of 4 metabolic genes (CRYM, CYTOC/CYC1, MDH, and OAT) is significantly decreased in the Schiz-H3meR17^≥1.3C subgroup. The 4 transcripts (CRYM, MDH, CYTOC/CYC1, and OAT) were strongly correlated (r = 0.55-0.83; P < .005-.0001), which indicates that they are not independently regulated. To examine whether the 4 metabolic transcripts are significantly decreased in the Schiz-H3meR17^≥1.3C subgroup, we calculated, for each subject and matched control, the mean difference in expression levels for CRYM, MDH, CYTOC/CYC1, and OAT. The Schiz-H3meR17^≥1.3C subgroup (n = 8) showed a significant deficit in expression of the 4 transcripts compared with matched controls (Δ log mean[Schiz-H3meR17^≥1.3C – Control] ± SEM, –0.25 ± 0.09; t = 2.61; P = .04, 2-tailed t test) (Figure 7B). In contrast, the remaining 33 schizophrenic subjects showed no significant deficits in expression of the 4 transcripts compared with matched controls (Δ log mean[Schiz-H3meR17^<1.3C – Control] ± SEM, 0.02 ± 0.06; t = 0.36; P = .7) (Figure 7B). We conclude that the set of 4 metabolic transcripts is significantly decreased in the Schiz-H3meR17^≥1.3C subgroup of subjects with schizophrenia compared with other subjects with schizophrenia and matched controls.

To further test the association between H3meR17 and decreased metabolic gene expression, we dichotomized the entire group of subjects with schizophrenia (N = 41) into subgroups defined by modified histone levels of 0.7 or less and greater than 0.7 compared with controls. We found no significant differences in metabolic transcripts between subgroups or relative to matched controls. Furthermore, levels of H3meR17 and metabolic gene transcripts showed no correlation (R² = 0.007-0.11) (Figure 7A). From these experiments, we conclude that significant deficits in metabolic gene transcripts in the PFC are limited to a subgroup of subjects with schizophrenia defined by high levels of H3meR17.

Regulation of H3R17 methylation includes coactivator-associated arginine methyltransferase 1 (CARM1)⁷⁵ and peptidylarginine deiminase 4 (PADI4 or PAD4), which converts methyl-Arg to citrulline, releasing methylamine.^76- 77 We measured CARM1 and PAD4 transcript levels using quantitative, real-time RT-PCR in the Schiz-H3meR17^≥1.3C subgroup and matched controls and did not observe significant differences (Δ log mean [Schiz-H3meR17^<1.3C – Control] ± SEM; CARM1: 0.03 ± 0.11; PAD4: 0.14 ± 0.12). In all case and control brains, relative levels of CARM1 transcript were much higher than those of PAD4 (CARM1/β-actin: 0.21 ± 0.06; PAD4/β-actin: 0.01 ± 0.013). We conclude that CARM1, which methylates H3 at arginine 17,⁷⁵ is expressed at robust levels in the human PFC.

After the first part of the study, we identified 6 (29%) of 21 of the subjects with schizophrenias who showed H3pS10-acK14 levels of 1.3 or greater relative to matched controls. The H3pS10-acK14 ^≥1.3C subgroup of subjects with schizophrenia (n = 6) showed a significant decrease in expression of the OAT transcript compared with the remaining 15 subjects with schizophrenia (P = .02). To confirm the association between H3 phosphoacetylation, H3pS10-acK14, and down-regulated OAT gene expression, we measured in the additional set of 20 matched pairs H3pS10-acK14 immunoreactivity and OAT cDNA levels. Of these additional 20 patients with schizophrenia, 2 showed H3pS10-acK14 levels of 1.3 or greater relative to matched controls. Thus, of 41 patients with schizophrenia, 8 (20%) showed H3pS10-acK14 levels greater than 1.3 relative to controls. Of these 8 patients in the Schiz-H3pS10-acK14^≥1.3C subgroup, 2 were also part of the Schiz-H3meR17^≥1.3C subgroup.

Compared with the 8 matched controls, the Schiz-H3pS10-acK14^≥1.3C subgroup (n = 8) showed a significant deficit in OAT expression (Δ log mean[Schiz-H3pS10-acK14^≥1.3C – Control] ± SEM: –0.39 ± 0.12; t = –3.21; P = .02, 2-tailed t test). However, schizophrenic subjects with H3pS10-acK14 levels of less than 1.3 (n = 33) showed no significant differences compared with matched controls (Δ log = 0.00 ± 0.05). Furthermore, we dichotomized the entire group of schizophrenic subjects (n = 41) into subgroups defined by modified histone levels of 0.7 or less and greater than 0.7 relative to controls, and no statistically significant differences were observed. Furthermore, we did not find a statistically significant association between the H3acK9/14, H3meK4, H4acK8, and H4acK12 histone modifications and metabolic gene expression. We conclude that significant deficits in multiple metabolic gene transcripts in the PFC of subjects with schizophrenia are limited to a subgroup of cases defined by high levels (≥1.3 relative to controls) of H3 methylation at arginine 17.

The alterations in 4 of 16 metabolic transcripts in the Schiz-H3meR17^≥1.3C subgroup were not accompanied by an overall change in messenger RNA levels because levels for 12 of 16 metabolic transcripts were not statistically significantly altered in these subgroups; furthermore, levels of 2 nonmetabolic transcripts, CAMIIK and RPL13A, differed less than 15% between schizophrenia subgroups or compared with matched controls.

Owing to the matching process, the differences in age, postmortem interval, and female-male ratio between schizophrenia cases and controls were less than 10%, and the differences in pH between the 2 cohorts was less than 1% (Table 1). The 8 subjects who composed the Schiz-H3meR17^≥1.3C subgroup did not show significant differences in age, sex, postmortem interval, and tissue pH compared with their matched controls and other groups (Table 3). The schizophrenia subgroups did not show significant differences in ratios of medicated and unmedicated patients, schizophrenia subtypes, or frequency of suicide. None of the 8 cases in the Schiz-H3meR17^≥1.3C subgroup and none of their matched controls were diagnosed as having alcohol or substance abuse or dependence, and none of these 16 subjects had positive results on toxicologic screening.

Furthermore, levels of H3meR17, the histone marker on which the subgrouping of the schizophrenic cohort is based, were not significantly associated with tissue pH, autolysis time, or age. Tissue pH was significantly correlated with 2 histone modifications (H3acK9/14 and H4acK12; r = 0.54 and 0.52, respectively; P < .01) but not with gene transcripts. There was a correlation between age at the time of death and glyceraldehyde-3-phosphate dehydrogenase messenger RNA levels (r = –0.39; P = .01) but no statistically significant correlations between age and other gene transcripts or histones. Autolysis time showed a weak correlation with levels of pyruvate dehydrogenase kinase, isoenzyme 2 (PDK2) transcript (r = –0.26; P < .1). Taken together, these results indicate that the down-regulation of CRYM, CYTOC/CYC1, MDH, and OAT transcripts in the Schiz-H3meR17^≥1.3C subgroup is not attributable to confounding postmortem factors, clinical variables, or medication status.

Li et al³⁵ recently showed that short-term, but not long-term, treatment with D₂-like antagonists selectively induces H3 phosphoacetylation in whole chromatin from striatal extracts. However, levels of H3 methylation, including H3meR17, remained unchanged in the striatum of drug-treated animals.³⁵ The results of the present postmortem study suggest that histone modifications in whole PFC chromatin are not affected by drug treatment. To further confirm this hypothesis, we treated adult mice for 14 days with (1) the conventional antipsychotic drug haloperidol, (2) the atypical drug risperidone, or (3) saline. In the medial rostral cortex, which in rodents is closely related to the primate PFC,^78- 79 levels of methylated H3, H3meR17, and phosphoacetylated H3, H3pS10-acK14, remained unchanged in haloperidol-and risperidone-treated animals compared with saline-treated controls (Figure 8). We conclude that antipsychotic drug treatment does not alter modified histone levels in whole chromatin from human and rodent cerebral cortices.

We show that down-regulated metabolic gene expression in the PFC is not a consistent feature of subjects diagnosed as having schizophrenia. A subgroup of subjects with schizophrenia shows decreased expression of multiple metabolic transcripts in conjunction with high levels of open chromatin-associated histone H3 methylation, H3meR17 (Figure 9). These alterations were not explained by clinical characteristics, medication, or postmortem factors. In schizophrenic subjects with high levels of H3meR17 in prefrontal chromatin, multiple metabolic pathways, including ornithine-polyamine metabolism (CRYM and OAT), mitochondrial electron transport (CYTOC/CYC1), and the tricarboxylic acid cycle (MDH) were affected. Molecular alterations in these metabolic pathways were recently reported for other cohorts with schizophrenia^7- 9 and subjects with bipolar disorder.⁸¹

The methylation of histone H3 at arginine 17 is associated with the open chromatin state and transcriptional activation.^28- 29,31 Therefore, dysregulation of H3meR17 in prefrontal chromatin, as observed in a subgroup of subjects with schizophrenia, could affect widespread chromosomal regions. Presently, it remains unclear whether this increase in H3meR17 reflects a compensatory mechanism to overcome the deficit in expression of metabolic and other gene transcripts.

Because decreased metabolism leads to a deficit in high-energy phosphate molecules in the PFC of subjects with schizophrenia,⁸⁰ a breakdown in orderly metabolic activity may further affect the chromatin-remodeling machinery of the nucleus that depends on adenosine triphosphate.^{44- 45,82- 83} Thus, in a vicious cycle, decreased chromatin-remodeling could render genomic DNA less accessible for the large multienzyme transcription complexes and could alter the pattern of acetylation, methylation, and phosphorylation at the NH₂-terminal tails of the core histones that define the functional state of chromatin.^26- 35 These maladaptive changes may result in additional chromatin defects,⁸⁴ including abnormal DNA methylation⁸⁵ and repair.⁸⁶ Therefore, hypoactivity and hypometabolism^{1- 6,25,80,87} and dysregulated metabolic gene expression in the PFC of subjects with schizophrenia may eventually lead to long-lasting molecular imprints in prefrontal chromatin, resulting in transcriptional dysregulation,^7- 24,52 alterations in cytoarchitecture^88- 91 and neuronal morphology,^92- 95 and oligodendrocyte loss.⁹⁶ It remains the subject of future studies to elucidate the molecular mechanisms that link histone modifications and other epigenetic determinants of gene expression to cortical dysfunction in schizophrenia.

Correspondence: Schahram Akbarian, MD, PhD, Brudnick Neuropsychiatric Research Institute, Department of Psychiatry, University of Massachusetts Medical School, Worcester, MA 01604 (Schahram.akbarian@umassmed.edu).

Submitted for Publication: December 5, 2003; final revision received November 10, 2004; accepted January 14, 2005.

Funding/Support: This work was supported by the National Alliance for Research on Schizophrenia and Depression, Great Neck, NY; the National Alliance for Autism Research, Princeton, NJ; and the Worcester Foundation for Biomedical Research (Dr Akbarian). The brain collection at Maryland Psychiatric Research Center was supported by grant MH60744 from the National Institutes of Health, Bethesda, Md (Dr Roberts).

Previous Presentation: This study was presented in part at the 2003 International Congress on Schizophrenia Research; March 30, 2003; Colorado Springs, Colo.

Acknowledgment: We thank the members of the Maryland Brain Collection: Terri U’Prichard, MA, for permissions and interviews, Sami Daoud, MD, for dissections, and David Fowler, MD, and Jack Titus, MD (Office of the Chief Medical Examiner), for their support and cooperation.