Figure 1. Analysis of DNA copy number changes at the GAD67 -encompassing segment of chromosome 2q in sectors CA3/2 and CA1 in 15 healthy controls (Control), 15 patients with bipolar disorder (BD), and patients with schizophrenia (SZ). A, Heat map of copy number intensity (CNI) changes in the stratum oriens of sector CA3/2. Each column represents 1 sample and each row represents a specific gene, according to their physical order in the chromosome. Red indicates copy gain, and blue indicates copy loss. Copy numbers are displayed in the log2 ratio in a color code from −2 to 2. The GAD67 -encompassing segment is highlighted by the dotted line. The upper arrow points to the close-up view of CNIs at the GAD67 - encompassing segment at chromosome 2q (B). C, The CNIs of each patient are represented as squares (Control), circles (SZ), and triangles (BD), respectively. The horizontal bars indicate the mean. Mean values were compared using 1-way analysis of variance with the Tukey post hoc test. The SZ group had less copy numbers than the control group (mean [SD], 1.74 [0.27] vs 1.99 [0.05]; P < .01) but not the BD group. D, Heat map of CNIs in the stratum oriens of sector CA1. The lower arrow points to the close-up view of CNIs at the GAD67 -encompassing segment at chromosome 2q (E). F, The CNIs of each patient are represented as squares (Control), circles (SZ), and triangles (BD), respectively. The horizontal bars indicate the mean. Similar to the findings in sector CA3/2, the SZ group in sector CA1 had less copy numbers than the control group (mean [SD], 1.84 [0.15] vs 2.04 [0.09]; P < .05) but not the BD group.
Figure 2. Heat Maps of DNA copy number changes for GAD67 regulatory genes in the stratum oriens of sector CA3/2 (A) and sector CA1 (B). The copy number intensities of 6 critical GAD67 regulatory genes are compared by performing the Fisher exact test between 15 healthy controls (Controls), 15 patients with schizophrenia (SZ), and 15 patients with bipolar disorder (BD). In the heat maps, red represents copy gain, and blue represents copy loss, in units of log2. The blue curves in the middle graphs indicate the degree of gain (right) or loss (left) of each gene. The red vertical line indicates a copy number of 2. Also shown are the probabilities that changes are significant; the arrows indicate whether copy number intensities are increased (red arrows) or decreased (blue arrows).
Figure 3. Association of fold changes as a measure of gene expression and DNA copy number intensity (CNI) changes in the stratum oriens of sectors CA3/2 and CA1.The x-axis shows CNI changes, and the y-axis displays fold changes, from microarray data. Individual data points are connected with a line using linear regression analysis performed with GraphPad Prism version 4.0. In the stratum oriens of sector CA3/2, significant associations between CNI changes and fold changes are evident in patients with schizophrenia (SZ), patients with bipolar disorder (BD), and combined patients (SZ+BD), respectively. There are no significant associations between CNI changes and fold changes in any patient groups in sector CA1.
Figure 4. Association of fold changes as a measure of gene expression and DNA copy number intensity (CNI) changes for GAD67 and non- GAD67 regulatory genes in the stratum oriens of sectors CA3/2 and CA1. The x-axis shows CNI changes, and the y-axis displays fold changes, from microarray data. Individual data points are connected with a line using linear regression analysis performed with GraphPad Prism version 4.0. In sector CA3/2, both clusters showed a significant association between CNI changes and fold changes, but the GAD67 regulatory network has a higher correlation efficiency than does the non- GAD67 regulatory network. However, there is no significant association between CNI changes and fold changes in any gene cluster in sector CA1.
Sheng G, Demers M, Subburaju S, Benes FM. Differences in the circuitry-based
association of copy numbers and gene expression between the hippocampi of
patients with schizophrenia and the hippocampi of patients with bipolar disorder.Arch Gen Psychiatry. Published online February 6, 2012.
eTable 1. Demographics of subjects.
eFigure 1. Laser microdissection.
eFigure 2. Relative copy number of the GAD67 gene in genomic DNA from SZ,
BD, and healthy controls.
Sheng G, Demers M, Subburaju S, Benes FM. Differences in the Circuitry-Based Association of Copy Numbers and Gene Expression Between the Hippocampi of Patients With Schizophrenia and the Hippocampi of Patients With Bipolar Disorder. Arch Gen Psychiatry. 2012;69(6):550-561. doi:10.1001/archgenpsychiatry.2011.1882
Author Affiliations: Program in Structural and Molecular Neuroscience, McLean Hospital, Belmont (Drs Sheng, Subburaju, and Benes and Mr Demers), and Program in Neuroscience (Dr Benes) and Department of Psychiatry (Drs Sheng, Subburaju, and Benes), Harvard Medical School, Boston, Massachusetts.
Context GAD67 regulation involves a network of genes implicated in schizophrenia and bipolar disorder. We have studied the copy number intensities of these genes in specific hippocampal subregions to clarify whether abnormalities of genomic integrity covary with gene expression in a circuitry-based manner.
Objective To compare the copy number intensities of genes associated with GAD67 regulation in the stratum oriens of sectors CA3/2 and CA1 in patients with schizophrenia, patients with bipolar disorder, and healthy controls.
Design Samples of sectors CA3/2 and CA1 were obtained from patients with schizophrenia, patients with bipolar disorder, and healthy controls. Genomic integrity was analyzed using microarrays, and the copy number intensities identified were correlated with the gene expression profile from a subset of these cases previously reported.
Setting Harvard Brain Tissue Resource Center at McLean Hospital, Belmont, Massachusetts.
Patients A total of 15 patients with schizophrenia, 15 patients with bipolar disorder, and 15 healthy controls.
Main Outcome Measures The copy number intensities for 28 target genes were individually examined using single-nucleotide polymorphism microarrays and correlated with homologous messenger RNA (mRNA) fold changes.
Results The copy number intensities examined using both microarrays and quantitative real-time polymerase chain reaction for the GAD67 gene were significantly decreased in sector CA3/2 of patients with schizophrenia and patients with bipolar disorder. Other genes associated with GAD67 regulation also showed changes in copy number intensities, and these changes were similar in magnitude and direction to those previously reported for mRNA fold changes in sector CA3/2 but not sector CA1. Moreover, the copy number intensities and mRNA fold changes were significantly correlated for both patients with schizophrenia (r = 0.649; P = .0003) and patients with bipolar disorder (r = 0.772; P = .0002) in sector CA3/2 but not in sector CA1.
Conclusion Insertions and deletions of genomic DNA in γ-aminobutyric acid cells at a key locus of the hippocampal circuit are reflected in transcriptional changes in GAD67 regulation that are circuitry-based and diagnosis-specific.
Copy number abnormalities have been found in genomic material from patients with many different brain disorders.1- 11 However, their significance for understanding the heritability and/or pathophysiology of such illnesses is not understood. The present study has attempted to explore this question by analyzing copy number intensities in laser-microdissected samples of the hippocampus from patients with schizophrenia and patients with bipolar disorder. In both disorders, dysfunction of γ-aminobutyric acid (GABA)–ergic cells has been inferred from the finding of decreased expression of the glutamic acid decarboxylase 67 kD isoform (GAD67), a key marker for GABAergic function.12- 15 There is a network of genes associated with the regulation of GAD67.16 This network appears to interact with other functional gene clusters, such as neurogenesis, cell cycle regulation, and the DNA damage response.17 Together, the latter 3 clusters play a critical role in maintaining genomic integrity, by detecting and repairing insertions, deletions, or perhaps even sequence rearrangements. If not repaired, such changes could theoretically result in the abnormal transcriptional regulation of genes associated with the regulation of GAD67.
In our study, samples have been obtained from the stratum oriens of sectors CA3/2 and CA1 (eFigure 1), a layer of the hippocampus in which GABA cells are the exclusive neuronal cell type. A preponderance of postmortem abnormalities in patients with schizophrenia and in patients with bipolar disorder have been found in sector CA3/2.18,19 These samples have been subjected to a microarray-based analysis of copy number intensities for 28 candidate genes involved in GAD67 regulation, neurogenesis, cell cycle regulation, and the DNA damage response. The results presented demonstrate that there are highly significant changes in the magnitude and direction of copy number intensities for specific target genes associated with GAD67 regulation in patients with schizophrenia and patients with bipolar disorder. Additionally, these copy number intensities show parallel variations in the expression of the respective messenger RNAs (mRNAs) for these target genes. To our knowledge, this is the first demonstration that disturbances in genomic integrity may contribute to GABA cell dysfunction in schizophrenia and bipolar disorder. Changes in copy number intensities vary in accordance with the disturbances in the expression of mRNAs associated with GAD67 regulation, but the patterns seen are fundamentally different for patients with schizophrenia and patients with bipolar disorder, suggesting that these changes occur in a circuitry-based and diagnosis-specific manner.
For the analyses of copy number intensities, postmortem samples of hippocampi from 15 healthy controls, 15 patients with schizophrenia, and 15 patients with bipolar disorder were obtained from the Harvard Brain Tissue Resource Center at McLean Hospital in Belmont, Massachusetts. The cases were group-matched for age, postmortem disorder, hemisphere, sex, and tissue pH (eTable). A subset of these cases (7 in each group) were described in an earlier microarray-based study,16 and the data from these cases were used to associate gene expression changes for the target genes with the respective copy number intensities obtained in the present study. Psychiatric diagnoses were established using a retrospective review of medical records and an extensive family questionnaire that includes information about medical history, psychiatric condition, and social demographics. The diagnosis of schizophrenia was made using the criteria of Feighner et al,20 whereas the diagnosis of bipolar disorder was made according to DSM-III-R criteria.
Sections (20 μm) were cut from a paraffin block using with a Jung RM2025 microtome and mounted on Leica frame slides. The sections were then deparaffinized and rehydrated in a graded series of xylene and ethanol. Nissl-stained sections were examined microscopically to ensure that each was cut in a transverse plane through the hippocampus and that all of the typical cytoarchitectonic features were present. A laser-microdissected apparatus was used to sample the stratum oriens of sectors CA2/3 and CA1. Because of the hypothesis-driven design of our study, we chose to contrast the findings in the stratum oriens of CA3/2 with those of sector CA1, which is also a layer in which GABA cells are the exclusive neuronal cell type.21 This layer in CA1 is, however, uniquely different from its homologue in CA3/2 in terms of its cytoarchitectonic detail, connectivity,22 functional integration,23 and gene expression patterns.17,21,24 For these reasons, we chose the stratum oriens of sector CA1 as a comparison site for changes in genomic integrity and mRNA expression in hippocampal GABA cells. There are approximately 3 times as many glial nuclei than neuronal nuclei in the stratum oriens of both sectors. Neurons are readily distinguishable from glia by their Nissl-positive cytoplasm and dendrites, their euchromatin content, and the size of their cell bodies, which are typically much larger. Glial nuclei show a predominance of heterochromatin, which is associated with gene silencing.25 These observations, together with our in situ hybridization studies of this layer, have suggested that very little gene expression is occurring in the glial cells at this locus.
A QIAamp DNA FFPE Tissue Kit (Qiagen) was used to extract genomic DNA from formalin-fixed, paraffin-embedded tissue. After extraction, the whole genome was amplified with a REPLI-g kit (Qiagen). DNA digestion, labeling, and hybridization were performed according to the manufacturer's instructions. In brief, genomic DNA (500 ng) is digested with Nsp I and Sty I restriction enzymes and ligated to adaptors that recognize the cohesive 4–base pair (bp) overhangs. All fragments resulting from restriction enzyme digestion were substrates for adaptor ligation. A generic primer that recognizes the adaptor sequence was used to amplify adaptor-ligated DNA fragments. Polymerase chain reaction (PCR) conditions were optimized to preferentially amplify fragments in the 200- to 1100-bp size range. The PCR amplification products for each restriction enzyme digest were combined and purified using polystyrene beads. The amplified DNA was then fragmented, labeled, and hybridized to Affymetrix Genome-Wide Human SNP Array 6.0. After the arrays were washed and stained, the copy number intensities were analyzed.
The scanned images of single-nucleotide polymorphism (SNP) arrays were analyzed using the Affymetrix Genotyping Console 2.0 and the Affymetrix Genotyping Tools software. Although copy number variants are commonly used as an end point in genetic studies of large populations or in searches for risk genes, copy number intensities provide greater sensitivity and more robust findings than the rounded-off ratios that result from analyses of copy number variants. The copy number intensities were determined according to the hybridization intensity data generated from each SNP probe using dChip software.26 To calculate mean copy number values for selected regions, the output of the SNP probe intensities were exported to Excel (Microsoft). Copy number intensity changes were measured by comparing the hybridization intensities of healthy controls with those of the patients with schizophrenia or bipolar disorder.
To validate copy number intensity changes identified by SNP arrays, we performed quantitative real-time PCR for GAD67 using genomic DNA (gDNA) from 6 controls, 6 patients with schizophrenia, and 6 patients with bipolar disorder. GAD67 gene probes was purchased from Applied Biosystems (the probes' identification numbers are Hs 01242242_cn and Hs 01602806_cn). Each 20-μL assay contained 10 ng of gDNA, 900nM each of forward and reverse primers for the reference gene (RNaseP) and for the target gene, 250nM each of the VIC dye (reference)–labeled and FAM dye (target)–labeled gene-specific probe in 1× TaqMan Gene Expression Master Mix. Individual samples were run in triplicate. Real-time data were collected by the OPTICON software. Quantitative real-time PCR data were analyzed using the relative quantification or ΔΔCt method27 based on DNA copy number ratio of a target gene vs reference gene in a given patient sample relative to matched healthy control sample. Relative quantity (RQ) is 2−ΔΔCt, and the copy number is 2 × RQ.
GraphPad Prism version 4.0 software was used for statistical analysis. The Fisher exact test and the 1-way analysis of variance with the Tukey post hoc test were used to evaluate differences in copy number for each target gene in the healthy control, schizophrenia, and bipolar disorder groups. Copy number intensities were also correlated with microarray-based gene expression data obtained in an earlier study16 in which an identical subset of cases was included.
Genomic copy numbers were determined by calculating the median signal intensities of the healthy controls, the patients with schizophrenia, and the patients with bipolar disorder with respect to normal reference DNAs. As shown in Figure 1, the GAD67- containing segment at the 2q24.3-2q31.1 chromosomal locus spanned approximately 14.2 megabases. The mean copy number intensity at the GAD67 -containing locus was 28% lower in the patients with schizophrenia than in the healthy controls. Similarly, lower numbers (15%) also occurred in the CA1 sector of the patients with schizophrenia at the same locus. Patients with bipolar disorder did not show differences in copy number intensities at the 2q24.3-2q31.1 locus in either sector (Figure 1).
The analysis of copy number intensities for 28 different target genes associated with the GAD67 regulatory network (n = 14), neurogenesis (n = 4), cell cycle regulation (n = 5), and the DNA damage response (n = 5) for sectors CA3/2 and CA1 are shown in Table 1 and Table 2, respectively. The overall sequence and its specific chromosomal locus are indicated for each gene in Tables 1 and 2. The genes that were chosen as targets for our study satisfied 2 criteria: (1) they must have shown significant differences in gene expression (fold changes),17,21,24 and (2) the genes must be associated directly or indirectly with the GAD67 regulatory network that was derived from a network association analysis.21 Some of the genes that are associated with this network (such as DLX1 and DLX2)28 and that play an important role in the development and function of GABA cells were not included because they did not show expression changes, and fell below the level of detection because of their very low abundance in the stratum oriens of sectors CA3/2 and CA1 of the human hippocampus.
Significant changes in copy number intensity in GAD67 regulatory genes were observed in 9 of 14 patients with schizophrenia (64%) and in 7 of 14 patients with bipolar disorder (50%) (Tables 1 and 2). The mean CNI for this gene was also decreased by 22% and 25%, respectively, in patients with schizophrenia and patients with bipolar disorder. In sector CA1, however, 50% of patients with schizophrenia showed significant CNI changes in GAD67 regulatory genes, whereas the patients with bipolar disorder showed no changes (Tables 1 and 2).
Heat maps of copy number intensities for specific chromosomal loci were plotted for several genes in the GAD67 regulatory network (Figure 1). As suggested by our earlier gene expression profile (GEP) study,21GAD67 expression was significantly decreased in sector CA3/2 of patients with schizophrenia and patients with bipolar disorder (Table 1). As shown in Figure 2A, the SNP arrays for GAD67, HDAC11, DAXX, PAX5, RUNX2, and CCND2 showed changes in copy number intensities in patients with schizophrenia and patients with bipolar disorder that are remarkably similar to those previously reported for a GEP study21 at this same locus of CA3/2. For example, GAD67 expression is reduced in both patients with schizophrenia and patients with bipolar disorder, HDAC11 and DAXX expression increased only in patients with schizophrenia, and both PAX5 and RUNX2 expression significantly decreased in patients with bipolar disorder. Two exceptions are PAX5, for which expression was only significantly lower in patients with bipolar disorder, and CCND2, for which transcripts were increased in patients with schizophrenia at this locus (Figure 2A). In the CA1 sector, copy number intensities for GAD67 are significantly decreased only for patients with schizophrenia, not for patients with bipolar disorder (Figure 2B), but this pattern is virtually identical to that seen for the GEP findings previously reported for the same cases.16 Copy number intensities for HDAC11 increased by 78% (Table 2), but the copy number intensities for the other genes do not show significant changes.
Copy number intensities for GAD65 (GAD2), GRIK1, RUNX2, PAX5, and SMAD4 were also decreased in the CA3/2 sector of patients with schizophrenia, whereas GRIK2, GRIK3, HDAC11, and DAXX all had higher copy number intensities. As shown in Table 3, the copy number intensity changes for these target genes in the CA3/2 sector of patients with schizophrenia were similar in direction to those reported with a GEP.16,21 In the CA3/2 sector of patients with bipolar disorder (Table 1), the copy numbers of GRIK1, GRIK2, GAD65, LEF1, DAXX, RUNX2, and PAX5 were also decreased, and most showed a direction of change consistent with the earlier GEP data for these same genes21 (Table 3).
The 4 genes in the neurogenesis cluster were chosen because they had shown significant changes in their GEPs, and our modeling suggested that they are probably important to the regulation of the cell cycle.19FGF3,29FGF9,30,31VEGF,32 and NRG133 are key players in the growth and differentiation of a wide variety of cells and tissues, including the hippocampus,34 and are believed to increase the risk for schizophrenia.35,36 In CA3/2 sector of patients with bipolar disorder, these genes showed significant decreases in copy number intensity, and these changes occurred in the same direction as those observed when a GEP was undertaken (Table 3). In the patients with schizophrenia, VEGF showed increases of copy number intensity in the CA3/2 sector, and this direction of change is similar to that seen in the GEPs for the same genes. NRG1 showed significant reductions in copy number intensity; however, the transcripts for this gene were significantly changed in the opposite direction for the GEP data (Table 3).
Copy number intensities for cell cycle regulation and the DNA damage response showed significant changes in the CA3/2 sector of patients with schizophrenia and patients with bipolar disorder (Table 1). The copy number intensity of E2F3 in patients with schizophrenia was significantly reduced, and this change was in the same direction as that observed for transcripts in the GEP study.21 The patients with bipolar disorder showed only 1 gene (MBD4) with significant changes in copy number intensity and gene expression, both of which were increased; however, this similar change was seen for this gene in sectors CA3/2 and CA1 of both groups of patients. Overall, the changes observed for copy number intensities in the cell cycle and DNA damage response categories showed much less consistency when compared with the GEP data for these same genes.
To validate the copy number intensity changes identified by SNP array, we performed quantitative PCR analysis for the candidate gene GAD67 using gDNA from microdissected tissues (obtained from 6 patients with schizophrenia, 6 patients with bipolar disorder, and 6 healthy controls). In sector CA3/2, the copy number for the GAD67 gene also decreased by 34% and 37%, respectively, in patients with schizophrenia and patients with bipolar disorder, which is in complete agreement with the SNP data. In sector CA1, the copy number for the GAD67 gene decreased by 59% in patients with schizophrenia, although there was only a 10% decrease in patients with bipolar disorder (eFigure 2). These results have demonstrated the overall validity of the copy number status determined by SNP microarrays, and they also establish that GAD67 is the gene with one of the most frequent genetic alterations in schizophrenia and bipolar disorder.
Do DNA copy number intensities reflect gene expression patterns that we have identified in patients with schizophrenia and patients with bipolar disorder within the loci of interest? In sector CA3/2 of both the schizophrenia and bipolar disorder groups, the majority of target genes showed abnormally high or low levels of mRNA expression, whereas in sector CA1, both groups showed a paucity of significant expression changes (Table 3). When decreases or increases of copy numbers are observed, the changes seen could be related to either deletions or duplications, respectively, of chromatin material. The genes showing significant correlations between copy number intensities and gene expression for the GAD67 regulatory genes (Figure 3) were generally quite different for the schizophrenia and bipolar disorder groups. As shown in Table 3, the patients with bipolar disorder showed no expression changes for GRIK3, DAXX, HDAC11, TGFB2, and TGFBR1, whereas all of the other genes showed significant expression changes. In patients with schizophrenia, RUNX2, LEF1, PAX5, and SMAD4 did not show expression changes. Overall, there was a diagnostic selectivity for the target genes associated with decreased GAD67 expression in schizophrenia vs bipolar disorder.21
When the GEP data and the copy number intensity data (Table 3) were analyzed using linear regression analyses (Figure 3), a robust correlation was observed when both groups of patients were combined (r = 0.692; P = .0001) or when the patients with schizophrenia (r = 0.649; P = .0003) and the patients with bipolar disorder (r = 0.772; P = .0002) were separately compared with the healthy controls (Figure 3). For a majority of the target genes, the changes in copy number intensities and the fold changes tended to occur in the same direction, although the magnitudes of the respective changes were, in some instances, dramatically different. For example, mRNA expression for FGF9 was decreased 11-fold in patients with bipolar disorder, whereas the loss of copy number intensity for this gene was only 20%. Disparities of this type contributed to the variance in the data and resulted in a reduction in the r correlation coefficient, even when the correlations were statistically significant.
As shown in Table 4 and Figure 4, when the target genes were separated according to those directly related to the GAD67 regulatory network and those not related to it (ie, genes associated with growth factors, cell cycle, and DNA repair), much higher correlations between copy number intensity changes and mRNA levels were observed in the former (r = 0.712; P = .0001) than in the latter (r = 0.492; P = .008) of sector CA3/2 but not sector CA1 (r = 0.067 and r = 0.225, respectively). These data suggest that the expression of genes in the GAD67 network in the CA3/2 sector is tightly linked to the status of genomic integrity at this locus.
The linear regression analyses of target genes in sector CA1 did not reveal any correlation between gene expression and copy number intensity changes. There are many genes showing notable expression changes but no copy number intensity changes, suggesting that a dissociation between copy number intensities and gene expression changes may occur in sector CA1. As with sector CA3/2, however, a subset of genes in sector CA1 not related to GAD67 regulation did not show correlations with copy number intensities (Figure 4). These results demonstrate a striking subregional difference in the association of copy number intensities with gene expression changes.
There is increasing evidence supporting the involvement of copy number variants in the etiology of schizophrenia,9,37- 39 although it is not clear what mechanisms may be related to these changes. Most of these reports have employed genome-wide screening for copy number variants, and only a few have concentrated on specific target genes.40 Although there have been preliminary attempts at linking copy number variants with expression data for specific genes in cancer research,41,42 to date, no study of the brain has attempted to associate the occurrence of copy number intensities for specific target genes with the expression of their respective mRNAs. The results described herein suggest that copy number intensity changes for specific target genes and their associated mRNAs vary not only in a diagnosis-specific way but also in a circuitry-based manner.
GAD67 plays an important role in the activity of GABA neurons and their dysfunction in schizophrenia. Although one association study37 concluded that there was a link between the GAD67 gene and childhood-onset schizophrenia, another study,43 using a Danish cohort, provides evidence that, in this sample, there is no link to schizophrenia but, possibly, a link to bipolar disorder. Another gene, neuregulin 1 (NRG1), may be functionally involved in the regulation of GAD67 activity.44 Interestingly, NRG1 also showed significant changes in both copy number intensity and gene expression, in patients with schizophrenia and patients with bipolar disorder. This gene appears to have a weak association with schizophrenia when haplotype-based P values were included in the analyses, and there was no evidence of between-study heterogeneity.45 These studies underscore the difficulties inherently present when attempting to relate copy number intensities derived from blood samples in population studies to copy number intensities obtained from brain tissue in studies of the molecular regulation of neural circuitry. Generally speaking, the genes included in the present study cannot be thought of as genetically transmitted risk factors, even when both copy number intensities and mRNA expression show robust changes, because the study design is fundamentally different from that used in association studies. The information gathered in our study is most useful in identifying the ways in which the molecular regulation of complex circuits may be abnormal, particularly when combined with parallel rodent modeling.46
The target genes examined in our study are involved in the functional maintenance of hippocampal GABA cells that show abnormal expression of GAD67 in schizophrenia and bipolar disorder.17,21 They include networks that regulate GAD67 expression, neurogenesis, the cell cycle, and the DNA damage response. In the present study, most of these genes showed significant increases or decreases of copy number intensities at their respective chromosomal loci. The largely robust nature of these copy number intensities and their correlation with the respective expression changes suggest that they may be linked to the regulation of functional differentiation and genomic integrity in hippocampal GABA cells. Widespread deletions along the GAD67 -encompassing segment of chromosome 2q (ie, 2q24.3-2q31.1), where GAD67 is encoded,47 have been identified in patients with schizophrenia but not in patients with bipolar disorder. The fact that GAD67 expression is significantly reduced in sector CA3/2 of both disorders suggests that these broad-based deletions on the 2q chromosome may be only one of the many mechanisms involved in the decrease of GAD67 expression.
The results from quantitative PCR showed a decreased copy number intensity of GAD67 in patients with schizophrenia or bipolar disorder, which was consistent with the results from SNP arrays. These findings demonstrate that a decreased copy number intensity of GAD67 in the hippocampus could be a prevalent genetic change in the majority of patients with schizophrenia or bipolar disorder. More importantly, there was a significant correlation between DNA loss and RNA underexpression of the GAD67 gene, suggesting that the transcript level of this gene may be regulated by its DNA copy number.
In sector CA3/2, the direction of change for copy number intensities showed a high degree of correspondence with the expression changes for the respective target genes. For example, in patients with schizophrenia, HDAC11 and DAXX both showed increased copy number intensities and expression changes. In patients with bipolar disorder, however, these parameters for PAX5 and RUNX2 were both decreased. Indeed, the regression analyses provide compelling evidence to support the idea that these 2 variables are interrelated. The relationship between copy number intensities and gene expression changes was most striking for the genes directly involved in GAD67 regulation compared with the genes more indirectly involved in neurogenesis, cell cycle regulation, and the DNA damage response.
The fact that the nature of the changes in copy number intensity and gene expression for the target genes was quite different for the 2 disorders may reflect the fundamentally different nature of the molecular abnormalities found in the GABA cells of sector CA3/2 for each disorder. It seems unlikely that psychotropic medications are responsible for these differences because the 2 groups showed similar drug exposure histories (eTable). It seems more likely that the unique changes in copy number intensity and gene expression in sector CA3/2 might be related to differences in “risk genes” for the 2 respective disorders.
Unlike the copy number intensities in sector CA3/2, those in sector CA1 of patients with schizophrenia or bipolar disorder did not show any correlation with gene expression changes. Only 20% to 30% of the target genes associated with cell cycle regulation and DNA repair in sector CA1 of patients with schizophrenia or bipolar disorder showed significant changes in expression (Table 3).
This obvious difference between the findings in sector CA3/2 and the findings in sector CA1 is consistent with the idea that the local tissue environment surrounding a cell may play a critical role in influencing its profile of transcriptional activity. In the hippocampus of patients with schizophrenia or bipolar disorder, the concentration of significant findings in the stratum oriens of sector CA3/2, but not sector CA1, can potentially be explained by differences in the molecular activity within neurons that comprise the microcircuitry at these 2 respective sites.24 For example, the stratum oriens of sector CA3/2, but not sector CA1, receives a rich glutamatergic projection from the basolateral amygdala, and these fibers are believed to play an important role in the integration of emotion and cognition, particularly in relation to stressful events.48 Other systems that are unique to this locus include GABAergic fibers that originate in the septal nuclei49 and contribute to the regulation of oscillatory rhythms.50
The stratum oriens is unique because the exclusive neuronal cell type in this layer is the GABAergic interneuron. These latter cells receive a robust and specific projection from the basolateral amygdala,51 which exerts an important influence on membrane properties and the action potential firing rate in fast-spiking inhibitory cells.46 Based on the present findings, basolateral amygdala projections to the stratum oriens of the CA3/2 locus may contribute to the functional and genomic integrity of GABAergic interneurons at this locus, presumably through synaptic or modulatory mechanisms. The GAD67 regulatory network includes 3 kainate-sensitive glutamate receptor (KAR) genes (GRIK1, GRIK2, and GRIK3) that encode the GluR5, 6, and 7 subunits, respectively, for this receptor. Abnormal expression and activity of KAR in the hippocampus are believed to play an important role in schizophrenia.52,53 Because the GAD67 regulatory network is linked to the canonical clusters associated with the cell cycle and the DNA damage response, the synaptically mediated influence of electrical activity generated within the trisynaptic pathway could contribute, at least theoretically, to parallel changes in copy number intensity and gene expression for the target genes associated with the regulation of GAD67 in GABA cells at this locus. These changes could include either deletions or duplications of genomic sequences in these target genes.
Most of our knowledge regarding the relationship between the sensing and repair of DNA has come from proliferating cell populations, like those active during embryogenesis.54 Little or nothing is known about the regulation of the cell cycle or the DNA damage response in postmitotic cells in the adult brain. Many believe that terminal differentiation involves an extensive reprogramming of the genome, so that genes that are relevant to interneuron function, like those associated with GAD67 regulation, are transcribed, while other genes are permanently silenced.55
The mechanisms associated with the repair of damaged DNA include pivotal target genes, such as CCND2, E2F3, MBD4, and HDAC, that may help to link the differentiation of mature GABA cells to their ability (or inability) to repair damaged DNA and preserve their genomic and functional integrity. To our knowledge, the present study provides the first evidence that deletions and/or duplications are present in terminally differentiated neurons in the adult hippocampus and that such changes may also contribute to the aberrant expression of key genes involved in the normal and abnormal functioning of hippocampal GABA cells.
In summary, the analyses reported herein demonstrate that there is considerable overlap between copy number intensities and mRNA expression for target genes associated with GAD67 regulation at a key locus within the hippocampi of patients with schizophrenia or bipolar disorder. Elucidating cell type–specific and locus-specific associations of genes comprising the GAD67 regulatory network in hippocampal GABA cells with those associated with the cell cycle and DNA repair could help to explain the presence of unique cellular endophenotypes in each of these 2 disorders.56 The genomic integrity and differentiation of tissue-specific functions in postmitotic GABA cells, and their potential relationship with age-related changes in health and disease,54 are issues that will require further study in many different forms of neuropsychiatric disease.
Correspondence: Francine M. Benes, MD, PhD, McLean Hospital, 115 Mill St, Belmont, MA 02478 (firstname.lastname@example.org).
Submitted for Publication: July 15, 2011; final revision received November 3, 2011; accepted November 15, 2011.
Published Online: February 6, 2012. doi:10.1001/archgenpsychiatry.2011.1882
Financial Disclosure: None reported.
Funding/Support: This study was supported by National Institutes of Health grants MH077175, MH42261-08, and MH/NS 077550 and the William P. and Henry B. Test Endowment (Harvard University).
Additional Contributions: We thank Cheng Li, PhD, Associate Professor of Biostatistics and Computational Biology, Harvard Medical School, for his extraordinary generosity and patience in helping us to organize our SNP database for copy number intensities and their analyses using the most appropriate statistics. We also thank Deborah L. Levy, PhD, for her expertise in genetics studies of blood samples from live subjects and her insights into how copy number variants and risk genes may interact in psychiatric disorders.