Allelic variation in GAD1 (GAD67) is associated with ... - Nature

Molecular Psychiatry (2007) 12, 854–869 & 2007 Nature Publishing Group All rights reserved 1359-4184/07 $30.00 www.nature.com/mp

ORIGINAL ARTICLE

Allelic variation in GAD1 (GAD67) is associated with schizophrenia and influences cortical function and gene expression RE Straub, BK Lipska, MF Egan, TE Goldberg, JH Callicott, MB Mayhew, RK Vakkalanka, BS Kolachana, JE Kleinman and DR Weinberger Clinical Brain Disorders Branch, Genes, Cognition, and Psychosis Program, Intramural Research Program, National Institute of Mental Health, NIH, US Department of Health and Human Services, Bethesda, MD, USA Cortical GABAergic dysfunction has been implicated as a key component of the pathophysiology of schizophrenia and decreased expression of the gamma-aminobutyric acid (GABA) synthetic enzyme glutamic acid decarboxylase 67 (GAD67), encoded by GAD1, is found in schizophrenic post-mortem brain. We report evidence of distorted transmission of single-nucleotide polymorphism (SNP) alleles in two independent schizophrenia family-based samples. In both samples, allelic association was dependent on the gender of the affected offspring, and in the Clinical Brain Disorders Branch/National Institute of Mental Health (CBDB/NIMH) sample it was also dependent on catechol-O-methyltransferase (COMT) Val158Met genotype. Quantitative transmission disequilibrium test analyses revealed that variation in GAD1 influenced multiple domains of cognition, including declarative memory, attention and working memory. A 50 flanking SNP affecting cognition in the families was also associated in unrelated healthy individuals with inefficient BOLD functional magnetic resonance imaging activation of dorsal prefrontal cortex (PFC) during a working memory task, a physiologic phenotype associated with schizophrenia and altered cortical inhibition. In addition, a SNP in the 50 untranslated (and predicted promoter) region that also influenced cognition was associated with decreased expression of GAD1 mRNA in the PFC of schizophrenic brain. Finally, we observed evidence of statistical epistasis between two SNPs in COMT and SNPs in GAD1, suggesting a potential biological synergism leading to increased risk. These coincident results implicate GAD1 in the etiology of schizophrenia and suggest that the mechanism involves altered cortical GABA inhibitory activity, perhaps modulated by dopaminergic function. Molecular Psychiatry (2007) 12, 854–869; doi:10.1038/sj.mp.4001988; published online 1 May 2007 Keywords: GAD1; GAD67; schizophrenia; COMT; epistasis

Introduction Evidence supporting specific genetic factors in the etiology of schizophrenia has strengthened dramatically in the past few years. While archival twin and adoption studies had long established that family risk patterns were largely due to inheritance, concerns about the complexity of the risk architecture and variability in early linkage studies had in some quarters raised premature doubts that risk genes could be identified.1,2 In the last few years, however, the momentum of the positive studies has increased tremendously. Several genome-wide linkage studies have identified many regions that are now supported by data from multiple independent studies (see Correspondence: Dr RE Straub, Genes, Cognition, and Psychosis Program, National Institute of Mental Health, National Institutes of Health, 10 Center Drive, Clinical Center, Building 10, Room 4D18, MSC 1385, Bethesda, MD 20892-1379, USA. E-mail: [email protected] Received 4 May 2006; revised 7 February 2007; accepted 11 March 2007; published online 1 May 2007

Levinson3 and Owen et al.4 for review). In addition, meta-analyses have highlighted additional promising regions.5,6 More importantly, genetic associations to genes in these linkage regions have been confirmed recently in multiple samples.7–9 These genes include, among others, DISC1 on 1q,10–12 DTNBP1 (dysbindin) on 6p,13,14 NRG1 on 8p,15 catechol-O-methyltransferase (COMT) on 22q,16,17 DAOA (G72) on 13q18,19 and RGS4 on 1q.20–22 The linkage findings represented an important step, enabling targeted searches within defined regions of the genome. Each of the susceptibility genes identified thus far in these regions, when analyzed alone, accounts for only a small increase in risk, usually less than twofold. Linkage signals may be due to single susceptibility genes with relatively large effect sizes and/or operating in a relatively large percentage of families, but more likely they are due to the combined signals from multiple genes in the region, each of lesser impact and prevalence. Linkage methods alone are underpowered to reliably provide the initial detection of signals from single genes of lesser effect or genes influencing

GAD1 polymorphisms and schizophrenia RE Straub et al

liability in only a small percentage of families.1 Thus, the more powerful strategy has evolved of direct examination of allelic variation in candidate genes, those selected because of their functional characteristics or protein binding partners. This strategy has resulted in evidence of association of schizophrenia with many other genes that do not map to prominent linkage regions, for example, GRM3 on 7q,23–25 AKT1 on 14q26,27 and CHRNA7 on 15q.28 Given the low a priori odds that any gene will be a risk factor, the interpretation and validity of candidate gene results is critically dependent on existing information about the biology of the gene relative to the biology of the disease. Abnormal glutamatergic signaling has been a prominent ‘hypothesis’ about the pathophysiology of schizophrenia,29,30 and it has been suggested that a number of the genes currently implicated in schizophrenia are related to NMDA (N-methyl-D-aspartate) signaling in the brain.31–33 An equally prominent hypothesis involves gamma-aminobutyric acid (GABA) neuronal function,34–37 but this avenue has been relatively unexplored from a genetic perspective. The recent demonstration that loss of NRG1/ERBB4 signaling alters the tangential migration of cortical interneurons and reduces the number of GABAergic interneurons in post-natal cortex38 is potentially a quite important mechanistic lead, since statistical evidence suggests that these two genes may synergistically increase the risk for schizophrenia.39 The evidence of GABA dysfunction in schizophrenia is abundant. From circumstantial evidence of an association of schizophrenia with a lowered threshold for seizures,40 recent studies have found direct evidence of impaired cortical inhibition (i.e. enhanced excitability),41 and reduced transcortical inhibition of motor circuitry.42 Studies of molecular markers of GABA signaling in post-mortem brain of patients with schizophrenia have provided consistent evidence of diminished GABA activity.43,44 The findings include diminished expression of SLC6A1 (GAT-1) protein on specific axon terminals,34 increased GABA receptors containing the alpha2 subunit, which correlate with reductions in GAT-1 expression45 and reduced expression of glutamic acid decarboxylase GAD1 mRNA and of GAD67 protein,46–49 a critical enzyme in GABA biosynthesis. Reduced expression of GAD67 in the neocortex of the schizophrenic brain is one of the more consistent findings reported in post-mortem studies of schizophrenia.50 GAD67 is one of two major isoforms of GAD, an enzyme that converts glutamate to GABA.51 The other isoform, GAD65, encoded by the gene GAD2, has generally not been observed to be altered in schizophrenic brain. GAD67 is encoded by the 17 exon, 46 kb gene GAD1 located in 2q31. GAD67 exists as both homo- and heterodimers in both soluble and membrane-bound forms, primarily within the cell soma but also to a limited degree in terminals.52 In the adult, it is thought to be critical for the maintenance

of intracellular GABA reserves for metabolic activity and other functions, rather than for vesicular release. GAD67, but not GAD65, protein levels can be influenced by intraneuronal GABA levels and this occurs without a change in the level of GAD67 mRNA.53,54 There also appears to be isoform-specific physiological feedback, in that GAD67 (but not GAD65) mRNA and protein levels were increased in cortical neurons after sensory learning.55,56 GAD67 deleted homozygous mice die immediately after birth and have less than 10% of normal brain GABA levels.57 During development, it appears that GABA influences multiple processes, including proliferation, neuroblast migration and dendritic arborization, acting in a paracrine, non-synaptic mode of action.58 In fact, it now appears that dynamic remodeling of dendrites in GABAergic interneurons continues to occur in the adult visual cortex.59 Finally, there is evidence that in adult mouse hippocampal slices, GABAergic synaptic inputs depolarize proliferating progenitor cells, contributing to their later differentiation and functional incorporation into the dentate gyrus.60,61 Thus, improper regulation of GAD67, either during development or adulthood, has the potential to produce a variety of phenotypic results, some of which may increase the risk of schizophrenia. There have been three prior reports of GAD1 variation and risk for psychosis. A case–control study of association with schizophrenia using a single SNP (rs769404, marker M13 in this study) was negative.62 A second study63 reported case–control results for a Danish bipolar affective disorder (BPAD) sample and Scottish BPAD and schizophrenia samples for eight SNPs, six of which we tested here (M03, M04, M09, M13, M16 and M19 in Table 1). They found no P-values < 0.05 in the Scottish schizophrenia or BPAD samples, and in the Danish BPAD sample found genotype-wise P-values of 0.062, 0.016 and 0.096 for SNPs M03, M04 and M09, respectively. Based on the evidence of a role of GABA in schizophrenia and the consistent findings of reduced GAD67 expression in schizophrenic brain, we performed a family-based association study of variation in the GAD1 gene and risk for schizophrenia in two independent samples containing parent–child trios. Because the post-mortem data have implicated reduced expression of GAD67, we concentrated on SNPs in and around the promoter region. We report evidence that variations in the GAD1 gene are associated with schizophrenia in both of these samples. After our initial observation of association, we sought to confirm these results in another clinical data set, the National Institute of Mental Health (NIMH) childhood onset series. While we continued to pursue the series of investigations reported herein in an effort to elucidate the mechanism of the genetic association, the results in this third clinical study were reported separately.64 Thus, we now also report that variations in this gene are associated with abnormalities in cortical function that are prominent in schizophrenia and in healthy siblings at risk for

855

Molecular Psychiatry


856

Table 1

Marker and map information

SNP

NCBI dbSNP rs#

Coding strand allelesa

CBDB case MAF

CBDB control MAF

NIMH-GI case MAF

Location

M01 M02 M03 M04 M05 M06 M07 M08

rs10432420 rs7557793 rs1978340 rs872123 rs2356236 rs3762556 rs3762555 rs3749035

G/A T/C G/A T/C A/G C/G G/C A/C

0.30 0.23 0.32 0.24 0.34 0.34 0.23 0.43

0.31 0.27 0.29 0.25 0.34 0.33 0.25 0.42

0.30 0.24 0.32 0.26 0.36 0.37 0.24 0.39

50 50 50 50 50 50 50 50

Start of transcript NM_000817 M09 rs3749034 G/A 0.23 M10 rs3762553 C/G 0.30 M11 rs2270335 C/T 0.24

0.24 0.29 0.25

0.18 0.32 0.26

Exon 1, 50 UTR Intron 1 Intron 1

Start codon (Met) of transcript NM_000817 M12 rs2241165 T/C 0.24 0.25 M13 rs16823181 T/C 0.41 0.41 M14 rs3828275 C/T 0.43 0.42 M15 rs10191129 C/G 0.24 0.23 M16 rs769390 A/C 0.27 0.27 M17 rs701492 C/T 0.30 0.29 M18 rs3791850 G/A 0.21 0.22 M19 rs769395 A/G 0.27 0.29

0.24 0.38 0.39 0.25 0.27 0.30 0.22 0.27

Intron 2 Exon 3, His37His Intron 3 Intron 5 Intron 6 Intron 9 Intron 12 Exon 17 (30 UTR)

Flanking Flanking Flanking Flanking Flanking Flanking Flanking Flanking

Chromosomal positionb

Intermarker distancec

Distance from M01

171485819 171490832 171495628 171496673 171497191 171497585 171497902 171498287

0 5013 4796 1045 518 394 317 385

0 5013 9809 10854 11372 11766 12083 12468

171498707 171498982 171499605 171500203

695 623 598

12888 13163 13786 14384

171500609–11 171503886 171504132 171508247 171514100 171518962 171527987 171533607 171542310

3683 246 4115 5853 4862 9025 5620 8703

14790–92 18067 18313 22428 28281 33143 42168 47788 56491

Abbreviations: CBDB, Clinical Brain Disorders Branch; MAF, minor allele frequency; NCBI, National Center for Biotechnology Information; NIMH-GI, National Institute of Mental Health Genetics Initiative; SNP, single-nucleotide polymorphism; UTR, untranslated region. a Major/minor. b From UCSC May 2004. c In base pairs.

schizophrenia, and that a 50 untranslated region (UTR) SNP affects expression in human prefrontal cortex (PFC), with a similar trend in hippocampus. Moreover, because alterations in GABA activity in brain are likely to impact on the excitability of neurons in cortical microcircuits implicated in schizophrenia,65,66 we explored the potential interaction of GAD1 risk alleles with a functional variation in COMT, another putative susceptibility gene that impacts on cortical neuronal excitability. We present preliminary evidence of an epistatic interaction of these genes. Since specific genetic variations that increase risk for complex genetic disorders such as schizophrenia often provide relatively weak statistical evidence, we have taken the approach that the interpretation of such associations can be enhanced by providing coincident biological evidence for functional effects on pathophysiological mechanisms that are implicated in the illness.67

Materials and methods Human subjects Subjects were recruited for the Clinical Brain Disorders Branch’s (CBDB) ‘sibling study’, a study of neurobiological abnormalities related to genetic risk Molecular Psychiatry

for schizophrenia.68 Subjects in the CBDB/NIMH sample included patients with schizophrenia, their available siblings (most of whom were unaffected) and controls. Participants were from 18 to 60 years of age, were above 70 in IQ and gave written informed consent of an Institutional Review Boards (IRB)approved protocol. Exclusionary criteria included significant medical problems, history of loss of consciousness for greater than 5 min, alcohol or drug abuse within the last 12 months and electroconvulsive therapy within the last 6 months. All subjects were interviewed and examined by a neurologist to rule out neurological comorbidities that might confound the study. All subjects spoke English since at least age 4, and over 98% were born and educated in the United States. Self-described ethnicity was 6.3% African American, 89.8% European American. DNA genotyping was available for 309 European American patients with schizophrenia (243 male, 66 female), 372 of their siblings (156 male, 216 female), 317 control subjects (163 male, 154 female) and 454 parents of probands. Neuropsychological testing and functional magnetic resonance imaging (fMRI) were acquired for subgroups, as described below. For all analyses shown here, only European American (labeled ‘Caucasian’ below) families were used.


An independent sample for follow-up studies to confirm family-based association with schizophrenia was obtained from the NIMH Genetics Initiative (NIMH-GI) data set consisting of only European American (‘Caucasian’) families (N = 67). This sample, which is a subset of the entire data set, consisted of DNA from selected members of nuclear families (average size = 3.04) including one or two siblings (average 1.87) with a diagnosis of schizophrenia or schizoaffective disorder (depressed subtype only), and available parents. The subset was chosen to maximize the number of potential offspring–parent trios for family-based association analyses (details available on request). Details about collection and diagnostic procedures have been described previously.69–71 Psychiatric diagnosis All subjects in the CBDB/NIMH sample were interviewed by a research psychiatrist who conducted the Structured Clinical Interview for DSM-IV (SCID-1, -II). For probands, data from psychiatric records and family members were evaluated before enrollment. All diagnoses were reviewed by 1–2 additional clinicians. Inter-rater reliability for two raters based on 20 randomly selected probands was 100% (for diagnosis of schizophrenia or schizoaffective disorder). For all analyses, subjects with the following diagnoses were included as probands: schizophrenia, schizoaffective disorder, simple schizophrenia, psychosis NOS, delusional disorder or schizoid, schizotypal or paranoid personality disorder. This is similar to a ‘broad’ diagnostic criteria used by other investigators.72 To reduce potential confounders in the cognitive analyses, we excluded a priori subjects with a history of ADD (attention deficit disorder) and dyslexia. This had no effect on the statistical significance of the genetic association findings reported below. Cognitive testing Subjects performed an extensive cognitive test battery as described previously.73,74 To reduce the number of statistical tests, effects of GAD1 genotype were examined on 15 measures, most of which we have previously reported to be related to genetic risk for schizophrenia.73,74 These included (1) four tests of episodic memory (Logical memory I and II and Verbal Paired Associates from the Wechsler Memory Scale, revised version (WMSR) and 1–5 SS scores from California Verbal Learning Test (CVLT)); (2) five measures of working memory (t-scores for categories and perseverative errors from the Wisconsin Card Sorting Task (WCST) and accuracy on the N back task (N back one, N back two, N back three); (3) one measure of attention (D0 distractible condition from the Gordon Continuous Performance Task, CPT); (4) verbal fluency for letters; (5) Trails A & B (Ocular scanning, speed, attention; Trails B also involves set switching); (6) IQ (from the Wechsler Adult Intelligence Scale, revised edition, or WAIS-R) and reading

comprehension (using the Wide Range Achievement Test, or WRAT, which may be considered a measure of premorbid IQ) were also included to assess levels of general intelligence.

857

Imaging data acquisition and analysis Whole brain BOLD fMRI data were collected on a 3 T Signa scanner using a GE-EPI RT pulse sequence (GE Medical Systems, Milwaukee, WI, USA).75–77 All fMRI data were processed and spatially normalized to a common stereotactic space (Montreal Neurologic Institute or MNI template) and analyzed using SPM99 software (Wellcome Dept. Cogn Neurol., http://www.fil.ion.ucl.ac.uk/spm/). Briefly, we generated single-subject contrast maps by contrasting the active condition against its control condition, for example, 2B (two back) versus 0B (no back) for the N back. These first level analyses for each normal control were one sample t-tests. Reported loci for genotype effects on task-related activation exceed significance levels of at P < 0.05 (corrected for multiple voxel comparisons within a region of interest encompassing the region of PFC showing a main task effect) and are reported with reference to the MNI standard space within SPM99 as in prior reports.17,78 Functional magnetic resonance imaging To test whether allelic variation in GAD1 influenced brain function independent of manifest behavior, we used fMRI to measure BOLD activation during the N back task, a measure of short-term (working) memory, in a sample of normal subjects. This task has been shown to activate the PFC in normal subjects, and patients with schizophrenia show abnormal engagement of PFC.75,76 The PFC phenotype has been imputed to altered cortical inhibition.75,79,80 We used a simple N back block design alternating between a two back (2B) working memory condition and a no back (0B) control condition. Subjects were instructed to recall the stimuli (visually presented digits 1–4) seen ‘n’ previously – two prior for 2B and the currently presented digit for 0B. Visual stimuli were presented via a back-projection screen and performance (accuracy and reaction time) recorded via the use of a button box.76 All of the normal subjects used in this analysis were matched for N back performance and reaction time, handedness (all right), age, gender and IQ. For M02, we compared 1/1s (N = 50) with 2-allele carriers (N = 52). Genotype groups sometimes differed slightly (P < 0.05) in age; therefore, an additional SPM (SPM99: www.fil.ion.ucl.ac.uk/spm/) analysis using age as a covariate was performed, and we found that it did not alter the results. Availability of the MRI scanner was the only factor determining whether fMRI was performed. Genotyping DNA was extracted from white blood cells or cerebellar brain tissue using standard methods. For clinical samples, blood was collected from all subjects as well as all available parents of patients Molecular Psychiatry


858

with schizophrenia. We selected 19 SNPs in GAD1 from NCBI databases based on position and allele frequencies (Table 1). Genotyping was performed using the Taqman 50 -exonuclease allelic discrimination assay81 (details available on request). Genotype reproducibility was routinely assessed by regenotyping all samples for selected SNPs and was generally > 99%. Genotyping completion rate was > 95% and genotyping errors were detected as Mendelization errors by the program TRANSMIT82,83 and as haplotype inconsistencies by the program MERLIN,84 which identifies improbable recombination events from dense SNP maps. Statistical analyses We measured linkage disequilibrium (LD) between markers with the D0 and r2 statistics from unrelated case and control haplotypes in parallel by use of the program LDMAX within the GOLD software package.85 For binary traits in families, single SNP association analysis was calculated using the TDTPHASE program version 2.37.86,87 We tested three-marker haplotypes for association in a sliding window across the gene using the method implemented in TDTPHASE. The E-M algorithm in TDTPHASE was used for unknown phase haplotype estimation. All P-values were computed empirically based on 100 000 permutations. The FBAT,88 family-based association test, program also was used to test for individual marker association with the clinical phenotype, and P-values are based on > 1000 permutations. The program COCAPHASE (version 2.404 from Frank Dudbridge’s web site http://www.hgmp.mrc. ac.uk/~fdudbrid/software/unphased/) was used to test for disease association with case–control data. We carried out tests of association to quantitative intermediate phenotypes using several approaches. The program SNPHAP (version 1.0, available from David Clayton’s web site http://www-gene.cimr. cam.ac.uk/clayton/software/) was also used to infer haplotype frequencies in unrelated controls and in cases from the CBDB families. Very similar frequencies were observed with the program PHASE.89 Single SNP quantitative trait analysis (i.e. with cognitive phenotypes) was performed in parent– affected child trios using the variance-components method from the version 2.4.3 of the program quantitative transmission disequilibrium test (QTDT).90 For all QTDT analyses, the orthogonal model was used. In order to protect against possible inaccuracies due to deviations from either normality or selection on the trait, empirical P-values derived from 10 000 permutations are reported. To test for epistasis, we used SPSS version 11.0 (SPSS Inc., Chicago, IL, USA) to script an unconditional logistic regression,91,92 which examined combinations of genotypes in COMT and GAD1, using affection status as the outcome variable. The ‘interaction-only’ P-value shown in Table 5 results from a test of the significance of the interaction model terms


above and beyond the main effects. Odds ratios were calculated relative to the 1/1 genotype at the GAD1 SNPs with either the 2/2 (G/G; Met/Met) genotype at rs4680 (Val158Met) or the 1/1 (A/A) genotype at rs2075507 (formerly rs2097603). Two-tailed Student’s t-tests or one-way analyses of variance (ANOVAs) were used to test if GAD1 mRNA levels differed between the genotypic groups. Spearman’s coefficients of correlation were calculated to examine if mRNA levels were correlated with age, pH or post-mortem interval (PMI). In the case of significant correlations, one-way analyses of covariance (ANCOVAs) were performed co-varying for a correlated factor. To test the effects of sex or race on dependent measures, two-way ANOVAs were used with genotype and gender as independent factors. DNA sequence analysis We identified putative promoter regions and SNP effects on transcription factor binding sites (TFBS) with the Genomatix software suite (http:// www.genomatix.de), specifically the El Dorado gene annotation database and the SNP Inspector tool of the GEMS sequence analysis package. TFBS were predetermined in the El Dorado annotation by the MatInspector program.93 Human post-mortem tissue and RNA extraction Detailed information about the CBDB/NIMH postmortem sample, including ascertainment and diagnostic procedures, neuropathology and toxicology screening, neuroleptic data, smoking history, manner of death and agonal state, as well as tissue retrieval and processing and RNA extraction and quality assessment has been published.94 Post-mortem hippocampal tissue (one hemisphere) from 76 brains of normal controls (21 women, 55 men; 24 Caucasians, 50 African Americans, two Hispanic) with mean age 41.4715.4 years, PMI 31.3714.2 h and pH 6.5970.27 and dorsolateral prefrontal cortical (DLPFC) tissue (one hemisphere) from a subset of 67 brains of the same normal controls (19 women, 48 men; 21 Caucasians, 44 African Americans, two Hispanic), with mean age 40.9715.3 years, PMI 31.1713.5 h and pH 6.5970.24 were collected at the Clinical Brain Disorders Branch, NIMH. An additional sample of hippocampal and DLPFC tissue from the brains of 32 schizophrenic cases (10 women, 22 men; 14 Caucasians, 17 African Americans, one Hispanic), with mean age 45.6715.5 years, PMI 35.9715.6 h and pH 6.4970.28 was also used in this study. Thus, the total sample comprised brain tissue from 108 subjects. The tissues were pulverized and stored at 801C. Total RNA was extracted from approximately 300 mg of tissue using the TRIZOL Reagent (Life Technologies Inc., Grand Island, NY, USA) as described previously. The yield of total RNA was determined by absorbance at 260 nm. RNA quality was assessed with a highresolution capillary electrophoresis (Agilent Technologies, Palo Alto, CA, USA). RNA integrity number (RIN) was used as an estimate of RNA quality (only


samples with RINX3.8 were included in the analysis). Total RNA (4 mg) was used in 50 ml of a reverse transcriptase reaction to synthesize cDNA, by using a SuperScript First-Strand Synthesis System for reverse transcriptase-polymerase chain reaction (RT-PCR) (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s protocol. Real-time quantitative PCR GAD67 mRNA expression levels were measured by real-time quantitative RT-PCR using an ABI Prism 7900 sequence detection system with 384-well format. Taqman probe and primers spanning exons 11–12 were purchased from Applied Biosystems (Assay-on-Demand, Cat. no. Hs00241471_m1, Applied Biosystems, Foster City, CA, USA). Each 20 ml PCR contained 6 ml of cDNA, 1 ml of 20 primer/probe mixture and 10 ml of qPCR Mastermix Plus (Eurogentec, Seraing, Belgium) containing Hot Goldstar DNA polymerase, deoxyribonucleotide triphosphates with dUTP, uracil-N-glycosylase, passive reference and optimized buffer components. PCR cycling conditions were as follows: 501C for 2 min, 951C for 10 min, 40 cycles of 951C for 15 s and 591C or 601C for 1 min. PCR data were obtained with the Sequence Detector Software (SDS version 2.0, Applied Biosystems, Foster City, CA, USA) and quantified by a standard curve method using serial dilutions of pooled cDNA derived from RNA obtained from hippocampi of 12 normal control subjects. SDS software plotted real-time fluorescence intensity. The threshold was set within the linear phase of the amplicon profiles. All measurements were performed in triplicates. Immunoblotting Human (100–130 mg) DLPFC tissue samples (1 g tissue: 10 ml buffer) were homogenized in a protease inhibitor-Tris-glycerol extraction buffer (AEBSF, 4-(2-Aminoethyl) benzenesulfonyl fluoride hydrochloride, 0.024%, aprotinin 0.005%, leupeptin 0.001%, pepstatin A 0.001%, glycerol 50%, Tris 0.6%). Protein concentration was determined using the Bradford assay. Homogenates containing 3 mg of total protein content per 10 ml were prepared for immunoblotting by diluting the homogenized samples with water and adding NuPage LDS sample buffer (Invitrogen, Carlsbad, CA, USA). Samples were then heated to 801C for 5 min to denature proteins. Ten microliters of each sample was loaded onto precast 10% Bis-Tris polyacrylamide gels (Invitrogen, Carlsbad, CA, USA), and proteins were separated by electrophoresis at 200 V for 1 h using a single power source. Each gel contained a molecular weight marker ladder SeeBlue Plus 2 (Invitrogen Carlsbad, CA, USA) and 14 samples from schizophrenics and controls. A total of four separate gels were used in one experiment (n = 53 samples). Gels were transferred onto nitrocellulose membranes at 85 V for 1 h. Membranes were blocked for an hour in 10% goat serum in Tris buffered saline (TBS) with 0.1% Tween-20 (TBS-T),

and then incubated with an anti-GAD67 (0.5 g/ml) and anti-b-actin (1:3000 dilution) antibodies. Blots were rinsed in TBS-T, incubated in a peroxidase-conjugated rabbit (GAD67) or mouse (b-actin) secondary antibody (1:10 000 dilution, Chemicon, Temecula, CA, USA) for 2 h in 5% normal goat serum in TBS-T and rinsed in TBS-T. Blots were developed in ECLplus (Amersham, Piscataway, NJ, USA) and exposed to Kodak Bio-Max film. Films were digitized using a scanner, and the resulting images were analyzed using gel plotting macros in the NIH Image software. The values for all samples were expressed as a percentage of the mean of controls on the same gel and also as ratios to b-actin.

859

Results GAD1 markers tested Marker and map information for the 19 SNPs tested in and around GAD1 is shown in Table 1. We concentrated on the 50 flanking region, under the assumption that since GAD1 expression is decreased in schizophrenic brain, at least some of the regulatory elements involved in this dysregulation would be located there. Work on the mouse homolog had shown numerous potential neuron-specific cis-regulatory elements well upstream, and that 10.2 kb of upstream sequence was sufficient to drive brain region-specific neuronal expression of a reporter gene in transgenic mice.95 Therefore, we genotyped eight SNPs in the 12 kb upstream of the transcription start site, a chromosomal region for which there are no known transcripts. None of the GAD1 SNPs or COMT SNPs (described below) deviated from Hardy–Weinberg equilibrium. Supplementary Table 1 shows the pairwise disequilibrium coefficients from the program LDMAX when applied to genotypes from 309 Caucasian cases from the CBDB/NIMH Sibling study sample. There are quite variable levels of moderate to high LD between the markers typed. Results were similar when either controls or the parents of the cases were used (data not shown). Clinical association, CBDB sample, effect of affected offspring gender The family-based TDT results from the CBDB sample using the clinical phenotype are shown in Table 2. The program TDTPHASE was used first, and to mitigate against potential errors, we followed up with FBAT and we present the results of both analyses for comparison. When transmissions to both male and female affected offspring were considered, all P-values were > 0.15. When only transmissions to affected female offspring were considered by setting male affected offspring phenotypes to unknown, three markers were positive. More strikingly, when male offspring are considered alone, there is also no evidence of association, even though they outnumber female cases nearly 4 to 1 (data not shown). Molecular Psychiatry


860

Table 2

TDT results from the CBDB family sample – effects of affected offspring gender and COMT Val158Met genotype

TDTPHASE resultsa

M#

FBAT resultsa Family subsetsb

All 269 families

Family subsetsb

Val/Val All Offspring

M01 M02 M03 M04 M05 M06 M07 M08 M09 M10 M11 M12 M13 M14 M15 M16 M17 M18 M19

P-value

T/nTc

0.31 0.76 0.74 0.95 0.24 0.45 0.98 0.40 0.88 0.97 0.99 0.97 0.30 0.67 0.88 0.81 0.56 0.16 0.90

407/424 447/451 390/396 451/452 322/341 313/325 396/395 244/230 447/444 291/290 388/387 313/312 336/316 325/317 427/424 423/419 402/411 381/360 411/409

Val/Met Met/Met All 269 families N = 76 N = 135 N = 58

Female offspring only

Val/Val

Val/Met Met/Met

N = 76 N = 135

N = 58

P-value T/nT Alleled P-value Allele P-value P-value P-value P-value Allele P-value P-value 0.03

41/23

2 0.05

2

0.01 0.01

2 2

0.05

77/59

1

0.02

1

0.02

77/56

1

0.01

1

0.03 0.006

2 2

0.02

2

0.08 0.14 0.68 0.44 0.89 0.77 0.10 0.61 0.54 0.75 0.26 0.33 0.33 0.62 0.61 0.68 0.61 0.30 0.73

0.006 0.008

2 2

0.05 0.02

0.003

1

0.002

1

0.02 0.005

2 2

0.02 0.02

0.04

2

0.05

Abbreviations: CBDB, Clinical Brain Disorders Branch; COMT, catechol-O-methyltransferase; FBAT, family-based association test; TDT, transmission disequilibrium test. a P-values not shown are P > 0.05. b Both genders included. c Transmissions and non-transmissions of the 1 allele. d Allele overtransmitted to affected individuals.

Clinical association, CBDB sample, effect of COMT Val158Met genotype Table 2 also illustrates the effect on TDT results of stratifying families into three groups based on the COMT Val158Met (SNP rs4680) genotype of the affected offspring. Transmissions to both male and female affected offspring were included in these analyses. When either the 135 families with Val/Met affected offspring or the 58 families with Met/Met offspring were analyzed, none of the GAD1 SNPs were positive. In contrast, when the 76 Val/Val offspring families were analyzed, eight of the 19 SNPs were positive. The program FBAT gave a very similar pattern of results to those from TDTPHASE, but they tended to be more significant, and, in addition, five SNPs were positive in the Val/Met subset. A similar stratification scheme based on either the functional ‘l’ (long) and ‘s’ (short) alleles at the SLC6A4 (serotonin transporter) 50 variable number of tandem repeats (VNTR)96 or on the functional BDNF Val66Met SNP rs6265,78 both of which were tested specifically afterwards as negative Molecular Psychiatry

controls, failed to yield positive results from GAD1 SNPs (data not shown). Clinical association in CBDB case–control data set Genotype-wise results for individual SNPs from the program COCAPHASE, comparing unrelated cases from the family data set to unrelated controls, are shown in Supplementary Table 2. All markers were negative, as were haplotypes in sliding windows of two, three and four markers (data not shown). We generated haplotype estimations using the program SNPHAP based on all 19 markers, run in parallel on 309 cases and 317 controls. Three common haplotypes (frequencies 0.321, 0.172 and 0.166 in cases) were found, and none of the 14 haplotypes with frequency > 0.01 in both groups differed by more than 1% between groups. There were also no case–control frequency differences in haplotypes composed of the six markers (M01, M02, M04, M07, M09 and M11) that showed the strongest LD with each other and that individually had the strongest association with the cognitive phenotypes (data not shown).


Clinical association, NIMH-GI sample Table 3 shows the FBAT results for the whole sample, and parallel analyses where only transmissions to affected female or male offspring were considered. Only SNP M03 is positive when all transmissions are considered. It is more significant in the males-only analysis and it is the same (minor) allele that is overtransmitted in the CBDB sample. In the NIMH-GI, six SNPs are positive in each of the gender-specific analyses, but none are positive in both. Most importantly, seven of the eight positive SNPs in the Val/Val family subset of the CBDB sample are positive in either the male or female analyses in the NIMH-GI sample and for all seven SNPs the overtransmitted allele is the same. We only rarely observe such a strong concordance in the pattern of positive results between these two family samples, but given that the stratification strategy was different (by Val/Met genotype versus by gender), the meaning of this is unclear. Intermediate phenotype association – cognitive phenotypes We used the program QTDT to examine the relationship between individual alleles and 15 neurocognitive phenotypes in the families in the CBDB Sibling Study and results from the 11 positive phenotypes are shown in Table 4. A number of SNPs that were in moderate to high LD with one another were associated with a broad spectrum of cognitive domains,

Table 3 FBAT results from the NIMH-GI family sample – effect of affected offspring gender M#

All offspring

Females

Males

P-value Allelea P-valueb Allele P-valueb Allele M01 M02 M03 M04 M05 M06 M07 M08 M09 M10 M11 M12 M13 M14 M15 M16 M17 M18 M19

0.85 1.00 0.03 1.00 0.51 0.51 0.68 0.19 1.00 0.41 0.85 1.00 0.27 0.24 0.85 0.82 0.64 0.83 0.49

2

0.05

0.02 0.04 0.01 0.01 0.04

0.005

2

0.02 0.02

2 2

0.03

1

0.04 0.03

1 1

1

1 2 2 1 2

Abbreviations: FBAT, family-based association test; NIMHGI, National Institute of Mental Health Genetics Initiative. a Allele overtransmitted to affected individuals. b P-values not shown are P > 0.05.

but especially so on three measures: a measure of attention (D0 distractible condition from the Gordon CPT), a measure of verbal list learning (CVLT 1–5 Std Scores) and a working memory task (N back). The clinical risk allele (allele 2 which is overtransmitted to females) at M01 is associated with poorer attention and N back performance, but better verbal learning scores. Overall, six of the eight SNPs that were associated with the clinical phenotype in the Val/Val subset produced negative results with most or all of the 15 cognitive phenotypes.

861

Intermediate phenotype association – neuroimaging We tested for effects of three SNPs (M02, M09 and M13) on the BOLD response with fMRI during the N back task in normal subjects. We tested SNP M02 because it is strongly associated with a variety of cognitive phenotypes but not with the clinical phenotype, SNP M09 because its 1 allele is associated with reduced GAD1 mRNA expression (see below) and also with the clinical phenotype in the NIMH-GI and NIMH Childhood Onset samples64 and SNP M13 because it is a coding polymorphism (His37His) associated with the clinical phenotype in both the CBDB and NIMH-GI samples. SNPs M09 and M13 were not associated with differential activation. As shown in Figure 1, for M02, 1/1 subjects (N = 50) showed greater activation in right dorsal PFC (Brodmann’s area 46/10; Talairach and Tornoux (x y z) coordinates (30 41 20), Z = 2.27, P = 0.01, k = 12) compared with 2-allele carriers (N = 52). Because the genotype groups are matched on performance and reaction time, this pattern of increased prefrontal engagement is qualitatively similar to the ‘inefficient’ prefrontal function phenotype associated with genetic risk for schizophrenia76 as well as with COMT Val alleles. It is intriguing that the 1 allele associated with greater activation also predicts poorer performance on the CVLT, WAIS (IQ) and WRAT Reading tests but better performance on D0 dist., N back, Trails and WMSR-Logical Memory 1 tests. Statistical interaction between GAD1 and COMT – case–control results To expand on the observations shown in Table 2 that in the CBDB family sample the genotype of the affected offspring at COMT Val158Met strongly affected our ability to detect association with SNPs in GAD1, we tested more directly for epistasis using the same cases and our CBDB controls. We used unconditional logistic regression91,92 to examine combinations of genotypes in COMT and GAD1, using affection status as the outcome variable. Table 5 contains only the positive (P < 0.05) results found after testing for interaction between 19 GAD1 SNPs and either COMT Val158Met or the COMT SNP rs2075507 (formerly rs2097603), which is located in a putative promoter region and has been shown to influence COMT enzyme activity.97 These two COMT SNPs exhibit only minimal LD (r2 = 0.148) with each other. Odds ratios were calculated relative to the 1/1 Molecular Psychiatry


862

Table 4

QTDT results with the cognitive phenotypes from the CBDB family sample

D0 dist.

M01 M02 M03 M04 M05 M06 M07 M08 M09 M10 M11 M12 M13 M14 M15 M16 M17 M18 M19

CVLT1–5 N back – N back – TrailsA TrailsB WAIS Est. WMSR Log. WMSR Log. WMSR Std Score two three Full Scale Mem. 1 Mem. 2 VPA1 total IQ

0.0004 0.0001

0.005 0.0004

0.003 0.001

0.009 0.003

0.022

0.0003

0.006

0.004

0.048

0.031

0.001

0.005

0.009

0.007

0.016

WRAT Reading

0.008 0.033

0.028

0.030

0.031

0.008 0.001

0.008

0.006 0.018

0.004

0.008 0.043 0.005

0.014

0.038

0.020

0.027

0.036 0.019

0.029 0.026

0.046 0.001

0.037

0.022

Abbreviations: CBDB, Clinical Brain Disorders Branch; CVLT, California Verbal Learning Test; QTDT, quantitative transmission disequilibrium test; WMSR, Wechsler Memory Scale, revised version. P-values < 0.05 are shown.

Figure 1 Effect of SNP M02 (rs7557793) genotype on the BOLD fMRI response during the N back working memory task. Results are displayed on the ‘single subject T1’ canonical image within SPM99. Healthy individuals who were genotype 1/1 (N = 50) showed an apparent abnormal increase in activation relative to 2-allele carriers (N = 52). Molecular Psychiatry


863

Table 5

Results of GAD1–COMT interaction analysis

GAD1 M#

GAD1 genotype

COMT Genotype interaction Odds ratio val158meta (rs4680) genotype P-value Type

M05 M05 M06 M10 M10

1/1 1/2 1/2 1/2 1/2

1/1 (G/G; Val/Val) 1/1 (G/G; Val/Val) 1/1 (G/G; Val/Val) 1/1 (G/G; Val/Val) 2/2 (A/A; Met/ Met)

GAD1 M#

GAD1 genotype

COMT rs2075507b Genotype interaction Odds ratio genotype P-value Type

M03 M03

2/2 (A/A) 2/2 (A/A)

1/1 (A/A) 1/2 (A/G)

0.035 0.026

Risk Protective

M06

2/2 (G/G)

1/2 (A/G)

0.040

M08 M08

1/1 (A/A) 2/2 (C/C)

1/2 (A/G) 1/2 (A/G)

M13 M13 M13

1/1 (T/T; His37His) 1/2 (A/G) 2/2 (C/C; His37His) 1/2 (A/G) 2/2 (C/C; His37His) 1/1 (A/A)

(A/A) (A/G) (C/G) (C/G) (C/G)

95% CI lower

95% CI upper

Case N

Control N

0.186 2.076 1.521 1.623 0.22

0.969 20.79 16.434 11.388 0.923

21 36 36 45 24

35 19 18 30 42

95% CI lower

95% CI upper

Case N

Control N

4.17 0.17

1.106 0.034

15.72 0.806

12 10

3 17

Protective

0.20

0.042

0.926

9

17

0.020 0.006

Protective Risk

0.45 5.53

0.227 1.65

0.883 18.56

28 28

50 24

0.002 0.002 0.027

Protective Risk Protective

0.38 6.11 0.36

0.207 1.973 0.144

0.711 18.936 0.889

32 30 9

63 27 19

0.042 0.001 0.008 0.003 0.029

Protective Risk Risk Risk Protective

0.43 6.57 5.00 4.30 0.45

Abbreviations: CI, confidence interval; COMT, catechol-O-methyltransferase; GAD, glutamic acid decarboxylase. P-values < 0.05 are shown. a The Met (2, A) allele frequency was 0.48 in cases and 0.49 in controls. b The 2 (G) allele frequency was 0.40 in cases and 0.42 in controls. This SNP is the same as rs2097603.

genotype at the GAD1 SNPs with either the Met/Met genotype at Val158Met or the 1/1 genotype at rs2075507. No main effects of the individual SNPs were observed, but some combinations of genotypes greatly increased or decreased the risk of schizophrenia. For example, odds ratios of 6.57 (95% confidence interval (CI): 2.08–20.8; P = 0.001), 5.00 (95% CI: 1.5–16.4; P = 0.008) and 4.30 (95% CI: 1.62–11.4; P = 0.003) were observed for the Val/Val genotype with heterozygotes at GAD1 SNPs M05, M06 and M10, respectively. This pattern is consistent with the strong LD we observed between these three SNPs (r2 = 0.752–0.984). The putative COMT promoter SNP rs2075507 produced results of similar significance levels, but with largely different GAD1 SNPs (M03, M06, M08, M13) than those identified when Val158Met was used. Notably, these included a coding SNP (M13, His37His) with an odds ratio of 6.11 (95% Cl: 1.97–18.9; P = 0.002). SNPs M03 and M06 are in fairly strong LD (r2 = 0.662), but they are only in weak LD with the high LD (r2 = 0.874) pair of M08 and M13, an indication that rs2075507 is perhaps interacting with more than one GAD1 haplotype to elevate risk. It is worth emphasizing that five of the six GAD1 SNPs identified as interacting with COMT SNPs were clinical positives in the Val/Val subset, but all six

were strikingly negative with the cognitive phenotypes. As a negative control for interaction, we tested GAD1 SNPs with either the functional ‘l’ (long) and ‘s’ (short) alleles at the SLC6A4 (serotonin transporter) 50 VNTR96 or with the functional BDNF Val66Met SNP rs6265.78 As expected, they were both negative. GAD1 expression in post-mortem brain We used the CBDB/NIMH post-mortem brain collection, which is independent from the CBDB/NIMH Sibling Study sample, to measure GAD67 transcripts. Cases and controls did not significantly differ in age (P = 0.2), PMI (P = 0.12), pH (P = 0.8), but did trend toward significance in RIN (P = 0.08). Expression levels of GAD67 mRNA normalized by reference gene b-2-microglobulin (B2M) were inversely correlated with age (R = 0.277, P = 0.004) and positively correlated with pH (R = 0.382, P < 0.0001) and RIN (R = 0.3, P = 0.003) but not with PMI (R = 0.02, P > 0.8). There was a significant positive correlation between DLPFC GAD67 mRNA expression levels and hippocampal GAD67 mRNA expression (R = 0.373, P = 0.0002). In the hippocampus, there were no significant differences between the diagnostic groups in the Molecular Psychiatry


864

expression of B2M (F(1,107) = 0.21, P > 0.5) or GAD67 mRNA normalized by B2M (F(1,107) = 0.08, P > 0.5). This result is consistent with the report of Heckers et al.98 In the DLPFC, however, schizophrenic patients showed significantly reduced expression of GAD67 mRNA normalized by B2M (by 16%, F(1,98) = 4.5, P = 0.036), whereas B2M expression was not different between the groups (P > 0.3). In order to examine whether risk genotypes were associated with changes in GAD67 mRNA expression among controls and schizophrenics, we performed a two-way ANOVA with diagnosis and genotype as independent factors. The 50 UTR SNP M09 (rs3749034) trended toward significance in the hippocampus (F(1,104) = 3.1, P = 0.08) and reached significance in the DLPFC (F(1,98) = 5.1, P = 0.026). There were significant diagnoses by genotype interactions in both the hippocampus and DLPFC (F(1,104) = 8.6, P = 0.004 and F(1,95) = 4.7, P = 0.0006, respectively). Post hoc analysis revealed that GAD1 mRNA expression normalized by B2M was decreased in both the hippocampi and DLPFC of schizophrenics with the 1/1 genotype as compared with 2-allele carriers (both P < 0.01). We then used ANCOVA with pH as a covariate to account for differences in pH among genotypic groups, and found that the schizophrenics carrying the 1/1 genotype still showed significantly lower expression of GAD67 mRNA compared with 2-allele carriers (P < 0.01 for both hippocampus and DLPFC). Similar analysis of the expression of B2M was negative (all P > 0.6). There were no significant effects of genotype on GAD67 mRNA expression (all P > 0.05) from M03, M05, M06, M08, M13, M16, M17 or M19. We then used a subset of DLPFC samples for the assessment of GAD67 protein levels (27 controls, 26 schizophrenics) on Western blots. In this subset, controls did not differ from schizophrenics in pH or age (P > 0.5). A two-way ANOVA showed no significant effect of diagnosis or of the risk genotype at 50 UTR SNP M09, and no significant diagnosis by genotype interaction on GAD67 immunoreactivity (all F < 0.5, all P > 0.5). There was no significant correlation between protein and mRNA levels (r < 0.1, P > 0.6).

Discussion We tested 19 SNPs in the GAD1 gene in two independent samples for association with schizophrenia, and for association with biological phenotypes related to schizophrenia. Since there have been several reports of GAD67 transcript and protein downregulation in schizophrenia brain, we extended our SNP coverage well into the 50 flanking region. Moderate to high marker to marker LD was evident in a highly variable pattern, but three common 19 marker haplotypes accounted for about 66% of the total. An additional nine haplotypes accounted for a further 20%. Case–control analyses of single SNPs, as well as 19-marker and six-marker haplotypes using the clinical phenotype were uniformly negative. In Molecular Psychiatry

contrast, family-based analysis based on the same cases revealed a significant association within the CBDB sample that became evident when analysis was restricted to females receiving the susceptibility alleles. This observation may be due to the biology of females with increased risk for schizophrenia. There is ample precedence that GABAergic mechanisms are at times sexually dimorphic, both during development and adulthood. For example, it appears that GABA is responsible in part for the inhibition of luteinizing hormone-releasing hormone release before the onset of puberty in female rhesus monkeys,99 and sex differences in GABA turnover and GAD65 and GAD67 transcript levels have been observed in the rat hypothalamus.100 Finally, in the olfactory bulb of adult rats, a region of high synaptic plasticity, estrogen modulates in a cyclic manner, a truncated version of the ‘embryonic’ isoform GAD25.101 It is also likely that the effects of alterations in GABA signaling on cortical excitability differentially impact on cortical function in the males and females because of other genetic and environmental factors. The potential epistatic interaction we observed with COMT may further support this possibility, as COMT also has been reported to show greater risk effects in females than in males.102 Alternatively, since the affected males and females in the CBDB sample are from unrelated families, it is possible that the gender differences observed may actually be secondary to some combination of biases in family ascertainment or environmental factors, or to genetic heterogeneity or differential epistasis. However, we also observed that the clinical association was dependent on gender in the NIMH-GI sample, where the families often have both male and female affected offspring in the same sibship. When we first observed positive association in both the CBDB and NIMH-GI samples,103 and noted the concordance of positive SNPs and risk alleles, we genotyped samples in the NIMH Childhood Onset Schizophrenia sample to seek additional confirmation. We found further evidence for association to GAD1, but in that sample our negative markers M09, M11 and M12 were positive.64 The reasons for such allelic differences between the all ‘Caucasian’ family samples we studied here and the Childhood Onset sample are at present unclear, although this may be due in part to the latter being only about 50% Caucasian. Likewise, the post-mortem sample is only 29% Caucasian, which may explain in part the observation that the SNPs that affect expression in post-mortem brain are not the SNPs that influence the clinical and cognitive phenotypes in the CBDB sample. Of interest though, is that the 1 (G) allele at the 50 UTR SNP M09 (rs3749034), which we found associated with reduced GAD67 transcript levels in hippocampal and prefrontal regions of adult postmortem brains, is also contained in the four marker 1111 risk haplotype (our M04–M09–M11–M12) that is overtransmitted to affected individuals in the Childhood Onset sample.64


In addition to differences in ethnicity, the differing risk variants across studies are likely to reflect the use of underpowered samples combined with incomplete characterization of genetic variation, since most or of all the SNPs and haplotypes tested here are probably merely tags for a highly heterogeneous collection of risk and protective haplotypes and diplotypes of varying frequencies and penetrances. In addition, cross sample differences in associated variants may also be due to the combined effects of genetic heterogeneity, differential epistatic interactions with other susceptibility genes and variable environmental factors. For example, when we stratified the CBDB families on the COMT Val158Met genotype of the single affected offspring, the results changed dramatically. While none of the 19 GAD1 SNPs were positive in the entire sample of 269 families, eight SNPs showed association in the subset of 76 families containing Val/Val affected offspring. In addition, we found some evidence for an interaction between the Val/Val genotype and three GAD1 SNPs in a logistic regression analysis of the case–control data set. An appealing interpretation of this finding is that affected individuals with the Val/Val genotype, who are presumably relatively dopamine deficient in frontal cortex,65 are more susceptible to the biologic effects of allelic variation in GAD1. The Val allele of COMT has been consistently associated with prefrontal cognitive and physiologic deficits related to schizophrenia.17,74,104,105 The mechanism of this effect is likely related to altered dopamine signaling in PFC106 and altered excitability of glutamate and GABA neurons.66 Thus, broadly speaking, altered GABA function might be expected to exaggerate the effects of the COMT Val allele, and vice versa. Our statistical data suggesting an interaction of COMT and GAD1 are consistent with this prediction. While we have ruled out some of the more obvious artifacts (e.g. differences in sibship size, number of parents genotyped, proportion female, order of ascertainment, etc.), the mundane possibility remains that stratifying the families on COMT Val158Met genotype served also to stratify on some non-biological variable that increases the detectability of the signal from GAD1 in the Val/Val families. Such an artifact would apply to the case–control evidence for epistasis we found since the cases are probands from the families. Regardless of the underlying reasons, the data indicate that the power to detect an association between GAD1 and schizophrenia may quite strongly depend, at least in part, on other characteristics of the sample tested, such as the frequency of alleles at other genes. We note also that, consistent with a number of other susceptibility genes we are studying, extensive testing of cases versus controls yielded no evidence of association (even when stratifying this analysis on Val158Met genotype), whereas the TDT test using the very same cases was positive. We assume that these type 2 errors reflect subtle but meaningful genetic substructure107,108 in the ‘Caucasian’ case and control

samples we have assembled, and that this will probably be characteristic of most studies unless it is actively controlled for. The statistical evidence for the GAD1 clinical association is not particularly strong, but perhaps this should be expected by now when analyzing genes in isolation that are neither necessary nor sufficient to generate the phenotype. The positive SNPs, and the particular risk alleles found in the CBDB and NIMH-GI samples are very similar to each other, and given that the childhood onset sample was also positive, it would appear that the signal in GAD1 from the CBDB sample is unlikely to be due to chance. There is substantial variability in the specific risk SNPs, SNP alleles, haplotypes and diplotypes found in studies of complex genetic disorders such as mental illnesses. This has lead some observers to claim that such differences should be interpreted as discrepancies. However, such a categorization would require a specific model, based on a far greater level of understanding of the variation present on risk chromosomes and their evolutionary history than is currently available. Such debates underscore the point that statistical association alone is an incomplete strategy for identifying and confirming causative genes.67 We have taken the approach that coincident data about the effect of genetic variation on biologic characteristics of the illness that are in turn plausibly related to the biology of the gene can strengthen interpretation of such statistical results.65,73,109 Thus, we have also tested the effects of variation in GAD1 on cognitive and physiologic measures related to schizophrenia and to GABA function in the cortex. We and others have consistently found that genes associated with psychiatric illnesses show greater effects at these putative intermediate phenotypes than at the clinical diagnosis.25,28,110,111 The cognitive domains of attention (CPT) and episodic and working memory (CVLT, N back) were both influenced by genetic variation in GAD1. These are well recognized as core cognitive deficits associated with schizophrenia. The WRAT Reading Score, often cited as an indicator of premorbid IQ, was also influenced. Because cognition is a behavioral manifestation of information processing at a more basic level of cortical circuitry, we have also developed assays of cortical function based on BOLD fMRI that relate to physiologic abnormalities associated with schizophrenia. With this approach, we examined the effect of genetic variation in GAD1 in normal controls matched for performance on the test administered during the fMRI experiment. Controlling for behavior allows testing for gene effects at the level of cortical information processing. We have shown that this strategy has greater power to reveal gene effects than studies of the behavior mediated by these circuits.73,77,78,111 In the BOLD fMRI assay during the prefrontally dependent working memory task, the 1 allele of SNP M02 was clearly linked to inefficient activation of right dorsal PFC in normal subjects. This is the same

865



866

allele that was significantly associated with poorer performance on the CVLT, WAIS-IQ and WRAT Reading tests, but also apparently associated with better performance on the D0 dist., N back, Trails and WMSR Logical Memory 1 tests. Surprisingly, this SNP is not associated with the clinical phenotype in any sample tested. We are uncertain of the basis for this, but suspect that it reflects differences in statistical power, distributions of cognitive measures and of other higher order sample characteristics, all in addition to what must be a variety of GAD1 haplotypes and diplotypes, each with potentially differing cognitive and clinical effects, and some in differential epistasis with other genes. We have performed many tests in this study, testing many SNPs and haplotypes on a variety of clinical and biological phenotypes, followed by a series of tests of potential genetic interactions. Virtually none of the P-values we report would survive correction for all of these tests if they were treated as independent, and without prior probability of a finding, as would be the case upon application of any biologically agnostic Bonferroni-type ‘correction’. For many reasons, we do not think that such adjustment of P-values is an appropriate or productive approach to deal with the problem of multiple testing. First, many of the SNPs are in substantial LD and thus are not independent, nor are many of the phenotypes. Second, the prior probability of SNPs in this gene impacting on expression of this gene and on intermediate phenotypes related to risk for schizophrenia is not null. Third, we have taken the approach of including supposed negative controls to test for randomness, and found no evidence of interaction between GAD1 with two other functional polymorphisms implicated in psychiatric phenotypes, that is, a VNTR in the serotonin transporter and BDNF val66met. Nevertheless, our principal findings require confirmation in other, preferably larger, schizophrenia samples, and other psychiatric phenotypes should be likewise explored. We undertook this genetic study of GAD1 because of the evidence of reduced expression of GAD67 in schizophrenic brain tissue. We predicted that the clinically positive SNPs would affect expression, and our data in post-mortem schizophrenic brain tissue lend support to this only in that the 1 allele at the 50 UTR SNP M09 is associated with decreased expression. Other SNPs showing association with the clinical phenotype did not affect expression during adulthood, at least in this post-mortem sample. By software prediction, we found that M09 was within a putative promoter region for the GAD67 transcript, and that the DNA with the 1 allele of M09 lacks the TFBS for proteins AREB6 and MYOD1 that are present on the DNA with the 2 allele. The validity of this purely in silico prediction of course remains to be determined experimentally. The 1 allele in M09 associated with reduced expression of GAD67 mRNA was also a clinical risk factor in the NIMH Childhood onset cases. However, M09 was not a risk factor in


either the CBDB or NIMH-GI samples, and yet it still strongly influenced multiple CBDB intermediate phenotypes. It is likely that the SNPs tested are monitoring several haplotypes that impact cognitive function, only some of which are relevant to the clinical phenotype and that a considerable variety of haplotypes and diplotypes exist across samples. To dissect this will require considerable additional work on much larger samples, and to facilitate higher resolution genotyping we will resequence a large number of cases. In addition, it is likely that the decreased GAD67 in schizophrenia is due also to factors both genetic and environmental that are independent of variation in GAD1. While acknowledging these various complexities and uncertainties, the fact that all three samples that we have tested to date produce positive clinical associations, and that risk SNPs and alleles are so similar in the CBDB and NIMH-GI samples and that genetic variation also impacts strongly on measures of cortical function suggests that GAD1 may be a fairly common risk factor for schizophrenia. Finally, our observation of a potential interaction with COMT adds biologic plausibility to assumptions about the mechanisms by which genes increase risk for schizophrenia7 and demonstrates the considerable importance of epistatic effects in mediating the strength of genetic association results in psychiatric disorders.

Acknowledgments We thank our patients and their families, as well as Mary Weirich (MSW) for her invaluable work in leading the recruitment team and Drs Llewellyn Bigelow and Thomas M Hyde for their tireless work in the clinical evaluation of subjects in CBDB Sibling Study and Drs Mary M Herman and Thomas M Hyde for their invaluable assistance in post-mortem tissue processing. This work was supported by the Intramural Research Program (IRP) of the NIMH, NIH.

References 1 Risch N, Merikangas K. The future of genetic studies of complex human diseases. Science 1996; 273: 1516–1517. 2 Moldin SO. The maddening hunt for madness genes. Nat Genet 1997; 17: 127–129. 3 Levinson DF. Molecular genetics of schizophrenia: a review of the recent literature. Curr Opin Psychiatry 2003; 16: 157–170. 4 Owen MJ, Williams NM, O’Donovan MC. The molecular genetics of schizophrenia: new findings promise new insights. Mol Psychiatry 2004; 9: 14–27. 5 Badner JA, Gershon ES. Meta-analysis of whole-genome linkage scans of bipolar disorder and schizophrenia. Mol Psychiatry 2002; 7: 405–411. 6 Lewis CM, Levinson DF, Wise LH, DeLisi LE, Straub RE, Hovatta I et al. Genome scan meta-analysis of schizophrenia and bipolar disorder, part II: Schizophrenia. Am J Hum Genet 2003; 73: 34–48. 7 Harrison PJ, Weinberger DR. Schizophrenia genes, gene expression, and neuropathology: on the matter of their convergence. Mol Psychiatry 2005; 10: 40–68. 8 Norton N, Williams HJ, Owen MJ. An update on the genetics of schizophrenia. Curr Opin Psychiatry 2006; 19: 158–164. 9 Straub RE, Weinberger DR. Schizophrenia genes – famine to feast. Biol Psychiatry 2006; 60: 81–83.

GAD1 polymorphisms and schizophrenia RE Straub et al 10 Fletcher JM, Evans K, Baillie D, Byrd P, Hanratty D, Leach S et al. Schizophrenia associated chromosome 11q21 translocation: identification of flanking markers and development of chromosome 11q fragment hybrids as cloning and mapping resources. Am J Hum Genet 1993; 52: 478–490. 11 Millar JK, Wilson-Annan JC, Anderson S, Christie S, Taylor MS, Semple CA et al. Disruption of two novel genes by a translocation co-segregating with schizophrenia. Hum Mol Genet 2000; 9: 1415–1423. 12 Callicott JH, Straub RE, Pezawas L, Egan MF, Mattay VS, Hariri AR et al. Variation in DISC1 affects hippocampal structure and function and increases risk for schizophrenia. Proc Natl Acad Sci USA 2005; 102: 8627–8632. 13 Straub RE, Jiang Y, MacLean CJ, Ma Y, Webb BT, Myakishev MV et al. Genetic variation in the 6p22.3 gene DTNBP1, the human ortholog of the mouse dysbindin gene, is associated with schizophrenia. Am J Hum Genet 2002; 71: 337–348. 14 Schwab SG, Knapp M, Mondabon S, Hallmayer J, BorrmannHassenbach M, Albus M et al. Support for association of schizophrenia with genetic variation in the 6p22.3 gene, dysbindin, in sib-pair families with linkage and in an additional sample of triad families. Am J Hum Genet 2003; 72: 185–190. 15 Stefansson H, Sigurdsson E, Steinthorsdottir V, Bjornsdottir S, Sigmundsson T, Ghosh S et al. Neuregulin 1 and susceptibility to schizophrenia. Am J Hum Genet 2002; 71: 877–892. 16 Li T, Ball D, Zhao J, Murray RM, Liu X, Sham PC et al. Familybased linkage disequilibrium mapping using SNP marker haplotypes: application to a potential locus for schizophrenia at chromosome 22q11. Mol Psychiatry 2000; 5: 77–84. 17 Egan MF, Goldberg TE, Kolachana BS, Callicott JH, Mazzanti CM, Straub RE et al. Effect of COMT Val108/158 Met genotype on frontal lobe function and risk for schizophrenia. Proc Natl Acad Sci USA 2001; 98: 6917–6922. 18 Chumakov I, Blumenfeld M, Guerassimenko O, Cavarec L, Palicio M, Abderrahim H et al. Genetic and physiological data implicating the new human gene G72 and the gene for D-amino acid oxidase in schizophrenia. Proc Natl Acad Sci USA 2002; 99: 13675–13680. 19 Addington AM, Gornick M, Sporn AL, Gogtay N, Greenstein D, Lenane M et al. Polymorphisms in the 13q33.2 gene G72/G30 are associated with childhood-onset schizophrenia and psychosis not otherwise specified. Biol Psychiatry 2004; 55: 976–980. 20 Chowdari KV, Mirnics K, Semwal P, Wood J, Lawrence E, Bhatia T et al. Association and linkage analyses of RGS4 polymorphisms in schizophrenia. Hum Mol Genet 2002; 11: 1373–1380. 21 Williams NM, Preece A, Spurlock G, Norton N, Williams HJ, McCreadie RG et al. Support for RGS4 as a susceptibility gene for schizophrenia. Biol Psychiatry 2004; 55: 192–195. 22 Talkowski ME, Seltman H, Bassett AS, Brzustowicz LM, Chen X, Chowdari KV et al. Evaluation of a susceptibility gene for schizophrenia: genotype based meta-analysis of RGS4 polymorphisms from thirteen independent samples. Biol Psychiatry 2006; 60: 152–162. 23 Marti SB, Cichon S, Propping P, Nothen M. Metabotropic glutamate receptor 3 (GRM3) gene variation is not associated with schizophrenia or bipolar affective disorder in the German population. Am J Med Genet 2002; 114: 46–50. 24 Fujii Y, Shibata H, Kikuta R, Makino C, Tani A, Hirata N et al. Positive associations of polymorphisms in the metabotropic glutamate receptor type 3 gene (GRM3) with schizophrenia. Psychiatr Genet 2003; 13: 71–76. 25 Egan MF, Straub RE, Goldberg TE, Yakub I, Callicott JH, Hariri AR et al. Variation in GRM3 affects cognition, prefrontal glutamate, and risk for schizophrenia. Proc Natl Acad Sci USA 2004; 101: 12604–12609. 26 Emamian ES, Hall D, Birnbaum MJ, Karayiorgou M, Gogos JA. Convergent evidence for impaired AKT1-GSK3beta signaling in schizophrenia. Nat Genet 2004; 36: 131–137. 27 Schwab SG, Hoefgen B, Hanses C, Hassenbach MB, Albus M, Lerer B et al. Further evidence for association of variants in the AKT1 gene with schizophrenia in a sample of European sib-pair families. Biol Psychiatry 2005; 58: 446–450. 28 Freedman R, Coon H, Myles-Worsley M, Orr-Urtreger A, Olincy A, Davis A et al. Linkage of a neurophysiological deficit in

29

30

31 32

33 34

35

36

37

38

39

40 41

42

43

44 45

46

47

48

49

50

schizophrenia to a chromosome 15 locus. Proc Natl Acad Sci USA 1997; 94: 587–592. Goff DC, Coyle JT. The emerging role of glutamate in the pathophysiology and treatment of schizophrenia. Am J Psychiatry 2001; 158: 1367–1377. Konradi C, Heckers S. Molecular aspects of glutamate dysregulation: implications for schizophrenia and its treatment. Pharmacol Ther 2003; 97: 153–179. Cloninger CR. The discovery of susceptibility genes for mental disorders. Proc Natl Acad Sci USA 2002; 99: 13365–13367. Harrison PJ, Owen MJ. Genes for schizophrenia? Recent findings and their pathophysiological implications. Lancet 2003; 361: 417–419. Moghaddam B. Bringing order to the glutamate chaos in schizophrenia. Neuron 2003; 40: 881–884. Lewis DA. GABAergic local circuit neurons and prefrontal cortical dysfunction in schizophrenia. Brain Res Brain Res Rev 2000; 31: 270–276. Benes FM, Berretta S. Gabaergic interneurons. Implications for understanding schizophrenia and bipolar disorder. Neuropsychopharmacology 2001; 25: 1–27. Kalkman HO, Loetscher E. GAD(67): the link between the GABAdeficit hypothesis and the dopaminergic- and glutamatergic theories of psychosis. J Neural Transm 2003; 110: 803–812. Wassef A, Baker J, Kochan LD. GABA and schizophrenia: a review of basic science and clinical studies. J Clin Psychopharmacol 2003; 23: 601–640. Flames N, Long JE, Garratt AN, Fischer TM, Gassmann M, Birchmeier C et al. Short- and long-range attraction of cortical GABAergic interneurons by neuregulin-1. Neuron 2004; 44: 251–261. Norton N, Moskvina V, Morris DW, Bray NJ, Zammit S, Williams NM et al. Evidence that interaction between neuregulin 1 and its receptor erbB4 increases susceptibility to schizophrenia. Am J Med Genet B Neuropsychiatr Genet 2006; 141: 96–101. Hyde TM, Weinberger DR. Seizures and schizophrenia. Schizophr Bull 1997; 23: 611–622. Daskalakis ZJ, Christensen BK, Chen R, Fitzgerald PB, Zipursky RB, Kapur S. Evidence for impaired cortical inhibition in schizophrenia using transcranial magnetic stimulation. Arch Gen Psychiatry 2002; 59: 347–354. Mattay VS, Callicott JH, Bertolino A, Santha AK, Tallent KA, Goldberg TE et al. Abnormal functional lateralization of the sensorimotor cortex in patients with schizophrenia. Neuroreport 1997; 8: 2977–2984. Costa E, Davis JM, Dong E, Grayson DR, Guidotti A, Tremolizzo L et al. A GABAergic cortical deficit dominates schizophrenia pathophysiology. Crit Rev Neurobiol 2004; 16: 1–23. Lewis DA, Hashimoto T, Volk DW. Cortical inhibitory neurons and schizophrenia. Nat Rev Neurosci 2005; 6: 312–324. Benes FM, Vincent SL, Marie A, Khan Y. Up-regulation of GABAA receptor binding on neurons of the prefrontal cortex in schizophrenic subjects. Neuroscience 1996; 75: 1021–1031. Akbarian S, Kim JJ, Potkin SG, Hagman JO, Tafazzoli A, Bunney Jr WE et al. Gene expression for glutamic acid decarboxylase is reduced without loss of neurons in prefrontal cortex of schizophrenics. Arch Gen Psychiatry 1995; 52: 258–266. Hashimoto T, Volk DW, Eggan SM, Mirnics K, Pierri JN, Sun Z et al. Gene expression deficits in a subclass of GABA neurons in the prefrontal cortex of subjects with schizophrenia. J Neurosci 2003; 23: 6315–6326. Volk DW, Austin MC, Pierri JN, Sampson AR, Lewis DA. Decreased glutamic acid decarboxylase67 messenger RNA expression in a subset of prefrontal cortical gamma-aminobutyric acid neurons in subjects with schizophrenia. Arch Gen Psychiatry 2000; 57: 237–245. Woo TU, Walsh JP, Benes FM. Density of glutamic acid decarboxylase 67 messenger RNA-containing neurons that express the N-methyl-D-aspartate receptor subunit NR2A in the anterior cingulate cortex in schizophrenia and bipolar disorder. Arch Gen Psychiatry 2004; 61: 649–657. Torrey EF, Barci BM, Webster MJ, Bartko JJ, Meador-Woodruff JH, Knable MB. Neurochemical markers for schizophrenia, bipolar

867



868 51 52

53

54

55

56

57

58 59

60

61

62

63

64

65

66

67 68

69

70

71

disorder, and major depression in postmortem brains. Biol Psychiatry 2005; 57: 252–260. Pinal CS, Tobin AJ. Uniqueness and redundancy in GABA production. Perspect Dev Neurobiol 1998; 5: 109–118. Kanaani J, Lissin D, Kash SF, Baekkeskov S. The hydrophilic isoform of glutamate decarboxylase, GAD67, is targeted to membranes and nerve terminals independent of dimerization with the hydrophobic membrane-anchored isoform, GAD65. J Biol Chem 1999; 274: 37200–37209. Rimvall K, Martin DL. The level of GAD67 protein is highly sensitive to small increases in intraneuronal gamma-aminobutyric acid levels. J Neurochem 1994; 62: 1375–1381. Rimvall K, Sheikh SN, Martin DL. Effects of increased gammaaminobutyric acid levels on GAD67 protein and mRNA levels in rat cerebral cortex. J Neurochem 1993; 60: 714–720. Gierdalski M, Jablonska B, Siucinska E, Lech M, Skibinska A, Kossut M. Rapid regulation of GAD67 mRNA and protein level in cortical neurons after sensory learning. Cereb Cortex 2001; 11: 806–815. Lech M, Skibinska A, Siucinska E, Kossut M. Learning-induced plasticity of cortical representations does not affect GAD65 mRNA expression and immunolabeling of cortical neuropil. Brain Res 2005; 1044: 266–271. Asada H, Kawamura Y, Maruyama K, Kume H, Ding RG, Kanbara N et al. Cleft palate and decreased brain gamma-aminobutyric acid in mice lacking the 67-kDa isoform of glutamic acid decarboxylase. Proc Natl Acad Sci USA 1997; 94: 6496–6499. Represa A, Ben Ari Y. Trophic actions of GABA on neuronal development. Trends Neurosci 2005; 28: 278–283. Lee WC, Huang H, Feng G, Sanes JR, Brown EN, So PT et al. Dynamic remodeling of dendritic arbors in GABAergic interneurons of adult visual cortex. PLoS Biol 2006; 4: e29. Tozuka Y, Fukuda S, Namba T, Seki T, Hisatsune T. GABAergic excitation promotes neuronal differentiation in adult hippocampal progenitor cells. Neuron 2005; 47: 803–815. Ge S, Goh EL, Sailor KA, Kitabatake Y, Ming GL, Song H. GABA regulates synaptic integration of newly generated neurons in the adult brain. Nature 2006; 439: 589–593. De LV, Muglia P, Masellis M, Jane DE, Wong GW, Kennedy JL. Polymorphisms in glutamate decarboxylase genes: analysis in schizophrenia. Psychiatr Genet 2004; 14: 39–42. Lundorf MD, Buttenschon HN, Foldager L, Blackwood DH, Muir WJ, Murray V et al. Mutational screening and association study of glutamate decarboxylase 1 as a candidate susceptibility gene for bipolar affective disorder and schizophrenia. Am J Med Genet B Neuropsychiatr Genet 2005; 135: 94–101. Addington AM, Gornick M, Duckworth J, Sporn A, Gogtay N, Bobb A et al. GAD1 (2q31.1), which encodes glutamic acid decarboxylase (GAD67), is associated with childhood-onset schizophrenia and cortical gray matter volume loss. Mol Psychiatry 2005; 10: 581–588. Weinberger DR, Egan MF, Bertolino A, Callicott JH, Mattay VS, Lipska BK et al. Prefrontal neurons and the genetics of schizophrenia. Biol Psychiatry 2001; 50: 825–844. Winterer G, Weinberger DR. Genes, dopamine and cortical signal-to-noise ratio in schizophrenia. Trends Neurosci 2004; 27: 683–690. Weiss KM, Terwilliger JD. How many diseases does it take to map a gene with SNPs? Nat Genet 2000; 26: 151–157. Egan MF, Goldberg TE, Gscheidle T, Weirich M, Bigelow LB, Weinberger DR. Relative risk of attention deficits in siblings of patients with schizophrenia. Am J Psychiatry 2000; 157: 1309–1316. Cloninger CR, Kaufmann CA, Faraone SV, Malaspina D, Svrakic DM, Harkavy-Friedman J et al. Genome-wide search for schizophrenia susceptibility loci: The NIMH Genetics Initiative and Millennium Consortium. Am J Med Genet 1998; 81: 275–281. Faraone SV, Matise T, Svrakic D, Pepple J, Malaspina D, Suarez B et al. Genome scan of European-American schizophrenia pedigrees: results of the NIMH genetics initiative and millennium consortium. Am J Med Genet 1998; 81: 290–295. Kaufmann CA, Suarez B, Malaspina D, Pepple J, Svrakic D, Markel PD et al. NIMH Genetics Initiative Millennium Schizo-


72

73

74

75

76

77

78

79

80 81

82 83

84

85 86

87

88

89

90

91

92

93

phrenia Consortium: linkage analysis of African-American pedigrees. Am J Med Genet 1998; 81: 282–289. Straub RE, MacLean CJ, O’Neill FA, Burke J, Murphy B, Duke F et al. A potential vulnerability locus for schizophrenia on chromosome 6p24–22: evidence for genetic heterogeneity. Nat Genet 1995; 11: 287–293. Egan MF, Goldberg TE, Gscheidle T, Weirich M, Rawlings R, Hyde TM et al. Relative risk for cognitive impairments in siblings of patients with schizophrenia. Biol Psychiatry 2001; 50: 98–107. Goldberg TE, Egan MF, Gscheidle T, Coppola R, Weickert T, Kolachana BS et al. Executive subprocesses in working memory: relationship to catechol-O-methyltransferase Val158Met genotype and schizophrenia. Arch Gen Psychiatry 2003; 60: 889–896. Callicott JH, Mattay VS, Verchinski BA, Marenco S, Egan MF, Weinberger DR. Complexity of prefrontal cortical dysfunction in schizophrenia: more than up or down. Am J Psychiatry 2003; 160: 2209–2215. Callicott JH, Egan MF, Mattay VS, Bertolino A, Bone AD, Verchinksi B et al. Abnormal FMRI response of the dorsolateral prefrontal cortex in cognitively intact siblings of patients with schizophrenia. Am J Psychiatry 2003; 160: 709–719. Hariri AR, Mattay VS, Tessitore A, Kolachana B, Fera F, Goldman D et al. Serotonin transporter genetic variation and the response of the human amygdala. Science 2002; 297: 400–403. Egan MF, Kojima M, Callicott JH, Goldberg TE, Kolachana BS, Bertolino A et al. The BDNF val66met polymorphism affects activity-dependent secretion of BDNF and human memory and hippocampal function. Cell 2003; 112: 257–269. Callicott JH, Ramsey NF, Tallent K, Bertolino A, Knable MB, Coppola R et al. Functional magnetic resonance imaging brain mapping in psychiatry: methodological issues illustrated in a study of working memory in schizophrenia. Neuropsychopharmacology 1998; 18: 186–196. Callicott JH. An expanded role for functional neuroimaging in schizophrenia. Curr Opin Neurobiol 2003; 13: 256–260. McGuigan FE, Ralston SH. Single nucleotide polymorphism detection: allelic discrimination using TaqMan. Psychiatr Genet 2002; 12: 133–136. Clayton D, Jones H. Transmission/disequilibrium tests for extended marker haplotypes. Am J Hum Genet 1999; 65: 1161–1169. Clayton D. A generalization of the transmission/disequilibrium test for uncertain-haplotype transmission. Am J Hum Genet 1999; 65: 1170–1177. Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin – rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet 2002; 30: 97–101. Abecasis GR, Cookson WO. GOLD – graphical overview of linkage disequilibrium. Bioinformatics 2000; 16: 182–183. Dudbridge F, Koeleman BP, Todd JA, Clayton DG. Unbiased application of the transmission/disequilibrium test to multilocus haplotypes. Am J Hum Genet 2000; 66: 2009–2012. Spielman RS, Ewens WJ. A sibship test for linkage in the presence of association: the sib transmission/disequilibrium test. Am J Hum Genet 1998; 62: 450–458. Horvath S, Xu X, Lake SL, Silverman EK, Weiss ST, Laird NM. Family-based tests for associating haplotypes with general phenotype data: application to asthma genetics. Genet Epidemiol 2004; 26: 61–69. Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet 2001; 68: 978–989. Abecasis GR, Cardon LR, Cookson WO. A general test of association for quantitative traits in nuclear families. Am J Hum Genet 2000; 66: 279–292. Cordell HJ, Todd JA, Hill NJ, Lord CJ, Lyons PA, Peterson LB et al. Statistical modeling of interlocus interactions in a complex disease: rejection of the multiplicative model of epistasis in type 1 diabetes. Genetics 2001; 158: 357–367. Cordell HJ. Epistasis: what it means, what it doesn’t mean, and statistical methods to detect it in humans. Hum Mol Genet 2002; 11: 2463–2468. Quandt K, Frech K, Karas H, Wingender E, Werner T. MatInd and MatInspector: new fast and versatile tools for detection of


94

95

96

97

98

99 100

101

102

consensus matches in nucleotide sequence data. Nucleic Acids Res 1995; 23: 4878–4884. Lipska BK, Deep-Soboslay A, Shannon-Weickert C, Hyde TM, Martin CE, Herman MM et al. Critical factors in gene expression in postmortem human brain: focus on studies in schizophrenia. Biol Psychiatry 2006; 60: 650–658. Kobayashi T, Ebihara S, Ishii K, Kobayashi T, Nishijima M, Endo S et al. Structural and functional characterization of mouse glutamate decarboxylase 67 gene promoter. Biochim Biophys Acta 2003; 1628: 156–168. Collier DA, Stober G, Li T, Heils A, Catalano M, Di Bella D et al. A novel functional polymorphism within the promoter of the serotonin transporter gene: possible role in susceptibility to affective disorders. Mol Psychiatry 1996; 1: 453–460. Chen J, Lipska BK, Halim N, Ma QD, Matsumoto M, Melhem S et al. Functional analysis of genetic variation in catechol-Omethyltransferase (COMT): effects on mRNA, protein, and enzyme activity in postmortem human brain. Am J Hum Genet 2004; 75: 807–821. Heckers S, Stone D, Walsh J, Shick J, Koul P, Benes FM. Differential hippocampal expression of glutamic acid decarboxylase 65 and 67 messenger RNA in bipolar disorder and schizophrenia. Arch Gen Psychiatry 2002; 59: 521–529. Terasawa E, Fernandez DL. Neurobiological mechanisms of the onset of puberty in primates. Endocr Rev 2001; 22: 111–151. Searles RV, Yoo MJ, He JR, Shen WB, Selmanoff M. Sex differences in GABA turnover and glutamic acid decarboxylase (GAD(65) and GAD(67)) mRNA in the rat hypothalamus. Brain Res 2000; 878: 11–19. Krizbai IA, Katarova Z, Szabo G, Parducz A, Wolff JR. Modulation of the truncated GAD25 by estrogen in the olfactory bulb of adult rats. Neuroreport 2000; 11: 791–794. Shifman S, Bronstein M, Sternfeld M, Pisante-Shalom A, Lev-Lehman E, Weizman A et al. A highly significant association between a COMT haplotype and schizophrenia. Am J Hum Genet 2002; 71: 1296–1302.

103 Straub RE, Egan MF, Goldberg TE, Callicott JH, Hariri AR, Balkissoon R et al. GAD1, which encodes glutamate decarboxylase 1 (GAD 67), is associated with adult onset schizophrenia in two independent samples. Am J Med Genet (Neuropsychiatr Genet) 2003; 122B: 177. 104 Mattay VS, Goldberg TE, Fera F, Hariri AR, Tessitore A, Egan MF et al. Catechol O-methyltransferase val158-met genotype and individual variation in the brain response to amphetamine. Proc Natl Acad Sci USA 2003; 100: 6186–6191. 105 Rosa A, Peralta V, Cuesta MJ, Zarzuela A, Serrano F, MartinezLarrea A et al. New evidence of association between COMT gene and prefrontal neurocognitive function in healthy individuals from sibling pairs discordant for psychosis. Am J Psychiatry 2004; 161: 1110–1112. 106 Tunbridge EM, Bannerman DM, Sharp T, Harrison PJ. Catechol-omethyltransferase inhibition improves set-shifting performance and elevates stimulated dopamine release in the rat prefrontal cortex. J Neurosci 2004; 24: 5331–5335. 107 Helgason A, Yngvadottir B, Hrafnkelsson B, Gulcher J, Stefansson K. An Icelandic example of the impact of population structure on association studies. Nat Genet 2005; 37: 90–95. 108 Gardner M, Gonzalez-Neira A, Lao O, Calafell F, Bertranpetit J, Comas D. Extreme population differences across Neuregulin 1 gene, with implications for association studies. Mol Psychiatry 2006; 11: 66–75. 109 Egan MF, Goldberg TE. Intermediate cognitive phenotypes associated with schizophrenia. Methods Mol Med 2003; 77: 163–197. 110 Freedman R, Olincy A, Ross RG, Waldo MC, Stevens KE, Adler LE et al. The genetics of sensory gating deficits in schizophrenia. Curr Psychiatry Rep 2003; 5: 155–161. 111 Hariri AR, Goldberg TE, Mattay VS, Kolachana BS, Callicott JH, Egan MF et al. Brain-derived neurotrophic factor val66met polymorphism affects human memory-related hippocampal activity and predicts memory performance. J Neurosci 2003; 23: 6690–6694.

869

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http:// www.nature.com/mp)
