Association between glutamic acid decarboxylase genes and anxiety disorders, major depression, and neuroticism

  • A Corrigendum to this article was published on 26 July 2006

Abstract

Abnormalities in the gamma-aminobutyric acid (GABA) neurotransmitter system have been noted in subjects with mood and anxiety disorders. Glutamic acid decarboxylase (GAD) enzymes synthesize GABA from glutamate, and, thus, are reasonable candidate susceptibility genes for these conditions. In this study, we examined the GAD1 and GAD2 genes for their association with genetic risk across a range of internalizing disorders. We used multivariate structural equation modeling to identify common genetic risk factors for major depression, generalized anxiety disorder, panic disorder, agoraphobia, social phobia and neuroticism (N) in a sample of 9270 adult subjects from the population-based Virginia Adult Twin Study of Psychiatric and Substance Use Disorders. One member from each twin pair for whom DNA was available was selected as a case or control based on scoring at the extremes of the genetic factor extracted from the analysis. The resulting sample of 589 cases and 539 controls was entered into a two-stage association study in which candidate loci were screened in stage 1, the positive results of which were tested for replication in stage 2. Several of the six single-nucleotide polymorphisms tested in the GAD1 region demonstrated significant association in both stages, and a combined analysis in all 1128 subjects indicated that they formed a common high-risk haplotype that was significantly over-represented in cases (P=0.003) with effect size OR=1.23. Out of 14 GAD2 markers screened in stage 1, only one met the threshold criteria for follow-up in stage 2. This marker, plus three others that formed significant haplotype combinations in stage 1, did not replicate their association with the phenotype in stage 2. Subject to confirmation in an independent sample, our study suggests that variations in the GAD1 gene may contribute to individual differences in N and impact susceptibility across a range of anxiety disorders and major depression.

Introduction

Major depression and the anxiety disorders have high lifetime prevalence and carry significant disability. Family and twin studies suggest moderate familial aggregation due primarily to genetic risk factors for these conditions.1, 2 They co-occur much more often than predicted by chance,3 and this is likely due to shared genetic risk.4 In addition, studies have consistently demonstrated associations between high levels of the personality trait neuroticism (N) in individuals and increased likelihood that they suffer from one or more of these syndromes. Twin studies suggest an overlap between the genes for N and genetic risk for depressive and anxiety symptoms and disorders.5, 6, 7

Many, but not all, linkage studies for these conditions have focused on one individual psychiatric trait or disorder at a time. However, acknowledging the difficulties inherent in identifying susceptibility genes for complex disorders like major depression and anxiety disorders and with the knowledge gleaned from advanced multivariate genetic epidemiological methods applied to large population-based twin samples, several leading research groups have begun to expand their phenotypic definitions beyond these disorder-based classifications created for clinical use.8, 9, 10 We have previously demonstrated that one can identify latent genetic risk factors that indicate shared genetic susceptibility across a range of phenotypes.11, 12, 13 Selecting subjects from the extremes of this underlying genetic risk factor should provide a powerful method for detecting genes of small effect expected to contribute to complex genetic phenotypes like major depression, anxiety disorders, and personality traits such as N.14, 15, 16

Abnormalities in the gamma-aminobutyric acid (GABA) neurotransmitter system have been noted in subjects with mood and anxiety disorders. Low plasma17 and cerebral spinal fluid18 GABA levels have been found in patients with unipolar depression when compared with controls. Magnetic resonance spectroscopy studies have shown lower occipital cortex GABA concentrations in subjects with panic disorder19 and major depression20 compared to healthy controls.

Glutamic acid decarboxylase (GAD) is the enzyme responsible for synthesizing GABA from glutamate. Its two isoforms, GAD67 and GAD65, are products of two independently regulated genes, GAD1 and GAD2, respectively, both of which are expressed in the brain. GAD1 is about 45 kB in size and located on 2q31.1, while GAD2 is about 88 kB and located on 10p12.1. They appear to differ in their intraneuronal expression, with GAD65 relatively localized to the axon terminals and GAD67 more uniformly distributed throughout the neuron.21, 22 Also, almost all of GAD67 exists in its active cofactor-bound holoenzyme form, while much of GAD65 is in the form of inactive apoenzyme that serves as a reserve pool.21, 22 One study suggests that the brain may respond to acute stress by upregulation of GAD67 synthesis, while chronic stress may stimulate enhanced availability of GAD65.23 Although many brain regions exhibit substantial expression of both genes, consistent with a dual system for controlling GABA synthesis, there does appear to be some differential distribution of these two GAD mRNAs.24 For example, in the adult rat brain, GAD65 transcripts are more abundant in hypothalamic regions and parts of the visual system, while GAD67 is more strongly expressed in neo- and cerebellar cortical regions, among others.25 Interestingly, both GAD enzymes were found to be highly expressed in the central nucleus of the macaque monkey amygdala.26

Studies of the two GAD genes in anxiety- and depression-related phenotypes are few, and the results have been mixed. Mutant GAD65 null mice demonstrate alterations in conditioned fear27 and anxiety-related behaviors.28 They did not differ in brain GABA levels from wild-type strains, but were slightly more susceptible to seizures.29 Unfortunately, it is difficult to examine the behavioral effects of GAD67 deficiency, as GAD67 null mice are born with marked deficiency of brain GABA and die of severe cleft palate soon after birth.30 In human studies, lowered GAD plasma activity has been seen in neurotic patients and those with mood disorders.31 A small family-based study in children found a modest association between GAD2 and behavioral inhibition, an anxiety-related trait.32 Among mood disorder phenotypes, while one study found no evidence of association between GAD1 and unipolar depression,33 a second study reported weak association of this gene with bipolar disorder.34

The aims of the present study are two-fold: (1) use the genetic information inherent in a large, population-based twin sample phenotyped for a broad range of related psychiatric phenotypes to select a powerful case–control sample for candidate gene association studies, and (2) apply this method to investigate possible association of the GAD1 and GAD2 genes with genetic risk shared between major depressive disorder, generalized anxiety disorder, panic disorder, agoraphobia, social phobia and N.

Materials and methods

Subjects

The subjects in this study derive from the population-based Virginia Adult Twin Study of Psychiatric and Substance Use Disorders (VATSPSUD).35 The total number of complete pairs and singletons included in the genetic models were (by zygosity group): FF MZ (678 pairs, 65 singletons); FF DZ (467, 46); MM MZ (869, 230); MM DZ (653, 275); and MF DZ (1429, 462), giving a total of 9270 twin subjects. All subjects were Caucasian and born in Virginia. Approval of the local Institutional Review Board was obtained prior to the study and informed consent was obtained from all subjects prior to data collection.

Measures

We obtained lifetime psychiatric diagnoses via face-to-face or telephonic structured psychiatric interview based on the SCID.36 We used DSM-III-R37 diagnostic criteria to assess lifetime major depression, and modified DSM-III-R criteria for lifetime generalized anxiety disorder and panic disorder. Since their low prevalence had been problematic in previous analyses,38, 39 we adopted a broad diagnostic approach to these latter two disorders, respectively, reducing the minimum duration from 6 months to 1 month for generalized anxiety disorder and requiring a history of panic attacks meeting at least two criteria within 30 min. Phobia was diagnosed using an adaptation of DSM-III criteria,40 which required the presence of one or more of 22 fears which the respondent recognized as unreasonable and that, in the judgment of the interviewer, objectively interfered with the respondent's life.41 We included agoraphobia and social phobia in the model used for this study. N was assessed using the 12 items from the short form of the Eysenck Personality Questionnaire (EPQ)42 via self-report questionnaire. It was analyzed as an ordinal variable with scores from 0 to 12.

Sample selection

The LGA program43 was used to calculate the two-stage sample design that minimized the genotyping burden and had the power to detect (PTD) 80% of the markers with true effect while controlling the false discovery rate (FDR) at 0.1.44 For these calculations, we assumed a proportion of markers without effect of 0.99, selection from the 20th and 80th percentiles of an underlying liability distribution, a genetic model that included a combination of additive and dominant effects, a range of disease allele frequencies 0.1–0.9, and that the markers explained 1% of the variance of the liability distribution. LGA indicated that we needed about 350 subjects in the stage 1 and 1000 in the stage 2 sample. The estimated threshold P-value in stage 1 was about 0.1. If any of the markers genotyped in stage 1 met this critical P-value, they were then also tested in the stage 2 sample.

We used the software package Mx45 to perform a multivariate genetic analysis46, 47 in order to identify a latent phenotype reflecting genetic covariation (ie, shared genetic susceptibility) across the following six phenotypes: major depressive disorder, generalized anxiety disorder, panic disorder, agoraphobia, social phobia and N (see Hettema et al48 for details). Two common genetic factors to account for this genetic covariation were included in the model (see Figure 1). The first factor (A1), the relevant one for this study, represented the portion of genetic covariation among the psychiatric disorders due to the genes for N, while the second factor (A2) accounted for genetic influences that increase the covariation independent of N. The model-based estimates of the genetic correlations between N and the five disorders and the proportions of total and genetic variance, respectively, accounted for by this genetic factor are indicated in Table 1.

Figure 1
figure1

Latent genetic risk factors shown in circles: two additive genetic common factors (A1, A2) plus disorder-specific genetic factors (ASP) (see Hettema et al48 for the other components of the model). The first common genetic factor A1 accounts for 100% of the genetic variation of neuroticism, while the second common genetic factor A2 is independent of neuroticism. Paths that account for more than 5% of the variance in a phenotype (i.e., path loadings >0.05=0.23) are depicted with thicker lines for emphasis. We have used the genetic factor scores derived from A1 to select subjects at the extremes of the liability distribution for this study.

Table 1 Proportions of psychiatric disorder variance accounted for by genetic factor A1 estimated from twin data

To select the subjects meeting the requirements from the LGA analysis, we estimated, by maximum likelihood, genetic factor scores for each subject for this underlying genetic risk factor A1 (separately for the five gender–zygosity groups). One member from each twin pair for whom DNA was available was selected as a case or control based upon scoring above the 80th or below the 20th percentile, respectively, of the genetic factor extracted from the analysis. This produced a total sample size N=1128 consisting of 589 cases and 539 controls, of which 376 and 752 were used in stage 1 and stage 2, respectively. The cases had a mean raw N score of 6.3 (z score=1.04) and had the following frequencies of the target psychiatric illnesses: major depressive disorder (80.1%), generalized anxiety disorder (53.8%), panic disorder (20.5%), agoraphobia (14.1%) and social phobia (17.5%). The controls were free of the five disorders and had a mean raw N score of 0.55 (z score=−0.89).

Marker selection

Single-nucleotide polymorphism (SNP) markers in or near the GAD1 and GAD2 loci were initially selected using the SNP Wizard utility of the Applied Biosystems SNPbrowser Software, version 2.0, specifying the SNP Tag Selection option with haplotype R2=85%. After genotyping in a subset of our stage 1 sample, these were supplemented by additional SNPs with the aim of tagging the major haplotypes (frequency>1–2%) observed in the Caucasian panel used by the HapMap project.49 In all, we genotyped seven SNPs in stage 1 and six in stage 2 for GAD1, and 14 SNPs in stage 1 and four in stage 2 for GAD2 (see Results).

Data analysis

Pearson's χ2 test was used to test for allelic or genotypic differences between cases and controls in the two stages. Instead of analyzing the data from stage 2 separately, a more powerful test can be obtained by pooling data from both stages. A complication is that (a) only the significant markers at stage 1 are selected for the next stage and (b) the test in stage 2 is not independent from the test in stage 1 because partly the same data are used. Assuming the conventional χ2 distribution for the test statistic would result in a considerable increase in the number of false discoveries.50 To perform accurate tests on the pooled data, we therefore used a different test statistic distribution that we derived elsewhere.51, 52

It has been shown that, by focusing on (moving windows of) regions of high linkage disequilibrium (LD), haplotype-inference methods are fairly accurate even with case–control data lacking phase information.53, 54, 55 To perform haplotype analyses we used the HAPLOVIEW program56 to first find the regions of high LD (ie haplotype blocks) using the default confidence interval algorithm.57 Haplotype association analyses were then performed for markers within the same haplotype block with the Cocaphase module of the UNPHASED program,58 as well as on long-range haplotypes (see Results). UNPHASED uses the expectation–maximization algorithm59 to estimate the haplotypes and their frequencies.

The risk of false discoveries is considerable in candidate gene studies.60, 61, 62 To be better able to assess this risk, we estimated so-called q-values63 for each marker genotyped in stage 2.64, 65 In the present context, these q-values can be interpreted as the probability that a marker identified as significant is a false discovery. To estimate the q-values, we need to know the prior probability that the marker has an effect and the effect size of the marker. Because only a relatively few markers were genotyped, the prior probability cannot be estimated reliably in this study. We therefore assumed a range of possible values. Markers that, due to sampling fluctuations, have a larger effect size are more likely to be selected as significant in stage 1. In an independent sample, however, the estimate will ‘shrink’ towards the true effect size.66 To estimate the effect size, we therefore used data from stage 2 only.

Genotyping

Subjects were instructed to use standard cytology brushes to obtain buccal epithelial cells for DNA extraction. The method for DNA extraction from the cytology brushes was described previously.67 SNPs were genotyped by the 5′ nuclease cleavage assay (also called TaqMan method).68 Reactions were performed in 384-well plates with 5 μl reaction volume containing 0.25 μl of 20 × Assays-on-Demand™ SNP assay mix, 2.5 μl of TaqMan universal PCR master mix and 5 ng of genomic DNA. The conditions for PCRs were initial denaturizing at 95°C for 10 min, followed by 40 cycles of 92°C for 15 s and 55°C for 1 min. After the reaction, fluorescence intensities for reporter 1 (VIC, excitation=520± 10 nm, emission=550±10 nm) and reporter 2 (FAM, excitation=490±10 nm, emission=510±10 nm) were read by the Analyst fluorescence plate reader (LJL Biosytems, Sunnyvale, CA, USA). Genotypes were scored by a Euclidian clustering algorithm developed in our laboratory. All genotyped SNPs were checked for deviations from Hardy–Weinberg equilibrium.

Results

GAD1

Two of the four markers initially selected for the GAD1 locus met the threshold criteria in our stage 1 sample, making this gene a candidate for genotyping in stage 2. We supplemented this original set with three additional markers (markers 3, 4 and 6 in Table 2) to better characterize the haplotype structure of this gene, giving a total of seven markers genotyped in stage 1. Figure 2 depicts the relative positions of these markers with respect to the structure of the GAD1 gene.

Table 2 GAD1 individual marker association results for stage 1 (N=188 cases, 188 controls)
Figure 2
figure2

Seven SNP markers genotyped for the GAD1 locus, with exons and untranslated regions (UTRs) as indicated. Numbers in parentheses correspond to marker numbers referenced in text.

The genotype and allele frequencies and results of χ2 association tests for the seven GAD1 markers genotyped in stage 1 are listed in Table 2. All of the GAD1 markers except markers 1 and 7 met the stage 1 threshold criteria of allelic P-value <0.1 to be genotyped in stage 2. However, marker 1 significantly contributed to haplotype analyses and was therefore retained for genotyping in stage 2 together with markers 2–6. The analysis for these six markers in the stage 2 sample is shown in Table 3. The pattern of results was consistent across the two stages. None of these SNPs showed deviations from Hardy–Weinberg equilibrium.

Table 3 GAD1 individual marker association results for stage 2 (N=401 cases, 351 controls) (note: marker 7 not genotyped in stage 2 sample)

LD information for these markers is provided in Table 4. To better understand the LD structure, we constructed haplotype blocks using the default block search procedure58 in HAPLOVIEW 3.2. We established that the six GAD1 markers occur on two haplotype blocks in our sample, with markers 1 and 2 on one block and 3–6 on the second. However, as indicated in Table 4, significant LD extends across the entire GAD1 gene, suggesting the presence of longer haplotypes. We therefore analyzed association of the 2- and 4-marker haplotypes, as well as the 6-marker long-range haplotypes created by their combination.

Table 4 Pairwise linkage disequilibriuma for markers in the GAD1 gene using the entire sample (N=1128)

We present the results for the most significant haplotypes of each marker grouping by stage in Table 5, as calculated using the Cocaphase module of UNPHASED. As these haplotypes are tagged by the significant markers, they explain the single-marker results. Note the presence of both risk and protective haplotypes for each marker grouping. These results agreed with those using HAPLOVIEW to perform the analyses.

Table 5 GAD1 haplotype analysis results for stages 1 and 2 (from Cocaphase routine of unphased)

Table 6 displays the P-values pooled across both stages and the stage 2 case–control odds ratios for markers 2–6 that met stage 1 criterion, as well as for the significant haplotypes. We restricted our pooled haplotype analyses to the two long-range haplotypes from Table 4, since these represent all of the variation observed in the shorter 2- and 4-marker block-derived haplotypes. We estimate a roughly 30–40% increase in risk associated with markers 2, 3 and 6, and a 15–20% reduction in risk for markers 4 and 5. Similarly, the 1–1–1–2–2–1 haplotype increases risk by about 23%, while the 1–2–2–1–1–2 haplotype decreases risk by about 28%. Table 6 also displays the estimated false-discovery rate q, that is, the global probability that the combined results for each particular marker or haplotype occurred purely by chance, as a function of the assumed prior probability of true discovery, p0. Some of the estimated q-values were around 0.1 or less, depending on the assumed p0. This suggests that, when the corresponding P-value is used to declare significance, the expected proportion of false discoveries among the significant tests would be about 10%.

Table 6 Allele-based test statistics for GAD1 significant markers and haplotypes for entire sample pooled across both stages (N=1128)

In post hoc analyses, we explored whether the haplotypic findings for GAD1 were a result of associations with specific phenotypes within our total sample. Using Cochran–Mantel–Haenszel tests in the FREQ procedure of SAS,69 we detected significant associations (non-zero correlation) between our protective 1–2–2–1–1–2 haplotype and N (P=0.0013), major depressive disorder (P=0.011), generalized anxiety disorder (P=0.0004) and possibly panic disorder (P=0.044). We obtained similar results using the GENMOD procedure in SAS with the appropriate distributions and link functions for these phenotype variables. With similar analyses, we did not detect significant association between our risk 1–1–1–2–2–1 haplotype and any of the specific phenotypes, except for possibly agoraphobia (P=0.011 in the general association test). The relationship between mean N score for our upper and lower genetic factor score groups and copies of the 1–2–2–1–1–2 haplotype is graphed in Figure 3. We note that these analyses extend beyond our original hypotheses and do not control for factors such as multiple testing or correlated phenotypes.

Figure 3
figure3

Graph of the relationship between mean N score and number of copies of six-marker protective haplotype 1–2–2–1–1–2 identified in the GAD1 gene.

GAD2

For the GAD2 locus, we found no criterion differences between the cases and controls in stage 1 using our initial set of nine markers selected using Applied Biosystems SNPbrowser Software, version 2.0. We found, however, that these did not capture the majority of allelic variation across this locus, so we supplemented these with an additional five markers (markers 2, 4, 5, 6 and 8 in Table 7) identified using the Tagger module of HAPLOVIEW 3.2 with HapMap Phase I data. Figure 4 depicts the relative positions of these 14 markers with respect to the structure of the GAD2 gene. Table 7 lists the genotype and allele frequencies and results of χ2 association tests for the 14 GAD2 markers genotyped in stage 1. Only marker 5 met the stage 1 threshold criteria of allelic P-value <0.1, but this association did not replicate in stage 2.

Table 7 GAD2 individual marker association results for Stage 1 (N=188 cases, 188 controls)
Figure 4
figure4

Fourteen SNP markers genotyped for the GAD2 locus, with exons and untranslated regions (UTRs) as indicated. Numbers in parentheses correspond to marker numbers referenced in text.

In our stage 1 sample, GAD2 markers 4–14 formed four haplotype blocks according to the default block search procedure58 in HAPLOVIEW 3.2, while markers 1–3 did not belong to any blocks. We could identify no marker combinations within these four blocks that showed significant association in stage 1. To explore the potential effects of long-range LD, we tested combinations of markers that were not co-located within physical blocks. We found that combinations of markers 2, 6 and 8 showed significant association in stage 1, but this also did not replicate in stage 2. (See Supplementary Table 1 for the results of the four GAD2 markers genotyped in our stage 2 sample.)

Population stratification

As a potential concern for any case–control association study, we explored the possibility that these results were obtained spuriously due to population stratification. First, using self-reported ancestry data from this entirely Caucasian sample, we tested for ethnic background differences between cases and controls. While this information was available for only 592 of our 1128 subjects, we did not detect any evidence of differences in this subset of subjects. Second, we analyzed one marker in each of 33 separate genetic loci for allelic differences between cases and controls in our stage 1 sample as part of our candidate gene screening. We found differences at the P=0.05 level in only two markers, giving a rate of 6%, similar to that expected by chance. We used the software STRUCTURE70 to estimate the number of different populations represented in our stage 1 sample using a subset of 24 unlinked markers from the group of 33 mentioned above. Among the choices of population number K=1–4 tested, the most likely value was K=1, suggesting the presence of no significant subpopulations. Finally, using the method of genomic control,71 we estimated the variance inflation λ of our test statistic that may be caused by stratification. Computed on the same 24 unlinked markers, 0.98 was obtained for the estimator of λ, consistent with its expected value of 1 when no effect of stratification is present.

Discussion

In this study, we sought to test whether the GAD1 and GAD2 genes are associated with susceptibility to N, a range of anxiety disorders and major depression. This susceptibility was indexed by a latent genetic factor shared with N, the score of which we derived from multivariate twin modeling and subsequently used to select subjects at the extremes of genetic risk. We entered the resulting sample of 589 cases and 539 controls into a two-stage association study in which markers from the candidate loci were screened in stage 1, the positive results of which were tested for replication in stage 2. Significant markers or their combinations making up significantly associated haplotypes were analyzed in the combined sample, producing total P-values by estimating the overall significance level and q-values by estimating their likelihood of being a false discovery.

Out of a total of seven markers tested in the GAD1 gene, five met the threshold criterion in stage 1 of P<0.1, suggesting that this gene was a reasonable candidate for genotyping in stage 2. χ2 tests to detect allelic frequency differences between cases and controls in each stage and in the entire sample pooled across both stages produced significant P-values for several of these markers and the resulting haplotypes, suggesting that the GAD1 gene is associated with the shared genetic risk across these internalizing disorders. In particular, we identified 6-marker, long-range haplotypes that spanned the gene that were associated with our genetic factor score. The 1–2–2–1–1–2 haplotype was protective, associated with a 28% reduction in risk, while the 1–1–1–2–2–1 haplotype increased the risk by 23%. Note that these are unlikely independent findings, as these two haplotypes mirror each other about the five significant markers (2–6). After testing 14 markers in and around the GAD2 locus, we found little evidence to support an association of GAD1 gene with these phenotypes.

It is not straightforward to compare these results with other published studies that have examined the effects of the GAD genes on psychiatric illness. Only one study33 tested for association between the GAD1 gene and one of the phenotypes in the current study, major depression. The authors of this study used a smaller European-American sample of 103 cases and 125 controls and did not specify comorbidity of the cases, nor the N level. Using three SNPs different from those in the current study, they found no significant allele frequency differences between their cases and controls. A second group34 that examined GAD1 did so in a study of bipolar disorder and schizophrenia using eight SNPs (again, different from those in the current study). They reported ‘weak association’ for two promoter SNPs in their Danish sample of 82 bipolar cases and 120 controls not replicated in their Scottish sample of 200 bipolar cases and 199 controls. A family-based study in 72 children from 66 families found an association at the P=0.05 level between a dinucleotide repeat in intron 3 of the GAD2 gene and behavioral inhibition, an anxiety-related trait.32

Declaring statistical significance in a candidate gene association study is not without controversy. If one were to simply examine nominal P-values for individual tests in isolation, several markers and haplotypes achieve significance at a level P<0.05 in both stage 1 and stage 2 samples. Also, four out of six individual markers, as well as several haplotypes, are significant at P<0.01 in the pooled sample, as indicated in Table 6. In addition, markers 2, 3 and 6 would retain statistical significance at a traditional P<0.05 level in the pooled sample after Bonferroni correction for multiple testing, considering the markers alone (six markers times three tests each). There are, however, practical and theoretical limitations to the Bonferroni correction.44 For all of these reasons, we also estimated the probability that the markers genotyped at stage 2 were false discoveries (q-values). Examining Table 6, we see that, depending on the assumed value of p0, each marker's likelihood of being a false discovery may vary. Using marker 6 as an illustration, if one's success rate at identifying a priori candidates with true effects out of all available to test was 5% (p0=0.95), then only about 2% of markers declared significant at the level P=0.0006 would be predicted to be false discoveries. This increases to a 54% rate of false discoveries at this P-value if one could only pick candidates with true effect 1 out of 1000 times (p0=0.999).

We considered the possibility that these results were obtained spuriously due to population stratification by several methods. First, we tested for significant ethnic background or marker allelic differences between cases and controls, detecting none. Next, the software STRUCTURE detected no significant subpopulations in our stage 1 sample using a set of 24 unlinked markers. Finally, using this same set of markers, the genomic control methodology estimated no variance inflation of our test statistic that may be caused by stratification. We note that (1) these markers were chosen for convenience from experiments on other candidate loci, not for their power to differentiate subjects by their ancestry, and (2) this number of markers is likely insufficient to detect more modest levels of stratification.72 In addition to these investigations in the current sample, Sullivan et al73 found no evidence for stratification using 16 unlinked microsatellite markers in a case–control study of nicotine dependence in a different subset of our twin sample (n=900). Thus, at least using these methods with the stated limitations, we find no evidence that our findings derive from the effects of population stratification.

Our study, which suggests that variations in the GAD1 gene may contribute to individual differences in N and risk for several internalizing disorders, needs to be verified in an independent sample. The significance of these findings will have to await further research into the biochemical effects of GAD1 haplotypes on the expression levels of GAD67, and ultimately, GABA, either at steady state or in response to stress, and the mechanisms by which this impacts the risk for anxiety and depressive disorders.

References

  1. 1

    Sullivan PF, Neale MC, Kendler KS . Genetic epidemiology of major depression: review and meta-analysis. Am J Psychiatry 2000; 157: 1552–1562.

  2. 2

    Hettema JM, Neale MC, Kendler KS . A review and meta-analysis of the genetic epidemiology of anxiety disorders. Am J Psychiatry 2001; 158: 1568–1578.

  3. 3

    Maser JD, Cloninger CR (eds) Comorbidity of Mood and Anxiety Disorders. American Psychiatric Press: Washington, DC, 1990.

  4. 4

    Middeldorp CM, Cath DC, Van Dyck R, Boomsma DI . The co-morbidity of anxiety and depression in the perspective of genetic epidemiology. A review of twin and family studies. Psychol Med 2005; 35: 611–624.

  5. 5

    Jardine R, Martin NG, Henderson AS . Genetic covariation between neuroticism and the symptoms of anxiety and depression. Genet Epidemiol 1984; 1: 89–107.

  6. 6

    Fanous A, Gardner CO, Prescott CA, Cancro R, Kendler KS . Neuroticism, major depression and gender: a population-based twin study. Psychol Med 2002; 32: 719–728.

  7. 7

    Hettema JM, Prescott CA, Kendler KS . Genetic and environmental sources of covariation between generalized anxiety disorder and neuroticism. Am J Psychiatry 2004; 161: 1581–1587.

  8. 8

    Smoller JM, Tsuang MT . Panic and phobic anxiety: defining phenotypes for genetic studies. Am J Psychiatry 1998; 155: 1152–1162.

  9. 9

    Nash MW, Huezo-Diaz P, Williamson RJ, Sterne A, Purcell S, Hoda F et al. Genome-wide linkage analysis of a composite index of neuroticism and mood-related scales in extreme selected sibships. Hum Mol Genet 2004; 13: 2173–2182.

  10. 10

    Kirk KM, Birley AJ, Statham DJ, Haddon B, Lake RI, Andrews JG et al. Anxiety and depression in twin and sib pairs extremely discordant and concordant for neuroticism: prodromus to a linkage study. Twins Res 2000; 3: 299–309.

  11. 11

    Hettema JM, Prescott CA, Myers JM, Neale MC, Kendler KS . The structure of genetic and environmental risk factors for anxiety disorders in men and women. Arch Gen Psychiatry 2005; 62: 182–189.

  12. 12

    Kendler KS, Walters EE, Neale MC, Kessler RC, Heath AC, Eaves LJ . The structure of the genetic and environmental risk factors for six major psychiatric disorders in women: phobia, generalized anxiety disorder, panic disorder, bulimia, major depression and alcoholism. Arch Gen Psychiatry 1995; 52: 374–383.

  13. 13

    Kendler KS, Prescott CA, Myers J, Neale MC . The structure of genetic and environmental risk factors for common psychiatric and substance use disorders in men and women. Arch Gen Psychiatry 2003; 60: 929–937.

  14. 14

    Van Gestel S, Houwing-Duistermaat JJ, Adolfsson R, van Duijn CM, Van Broeckhoven C . Power of selective genotyping in genetic association analyses of quantitative traits. Behav Genet 2000; 30: 141–146.

  15. 15

    van den Oord EJCG . A comparison between different designs and tests to detect QTLs in association studies. Behav Genet 1999; 29: 245–256.

  16. 16

    Schork NJ, Nath SK, Fallin D, Chakravarti A . Linkage disequilibrium analysis of biallelic DNA markers, human quantitative trait loci, and threshold-defined case and control subjects. Am J Hum Genet 2000; 67: 1208–1218.

  17. 17

    Petty F, Kramer GL, Dunnam D, Rush AJ . Plasma GABA in mood disorders. Psychopharmacol Bull 1990; 26: 157–161.

  18. 18

    Kasa K, Otsuki S, Yamamoto M, Sato M, Kuroda H, Ogawa N . Cerebrospinal fluid gamma-aminobutyric acid and homovanillic acid in depressive disorders. Biol Psychiatry 1982; 17: 877–883.

  19. 19

    Goddard AW, Mason GF, Almai A, Rothman DL, Behar KL, Petroff OA et al. Reductions in occipital cortex GABA levels in panic disorder detected with 1h-magnetic resonance spectroscopy. Arch Gen Psychiatry 2001; 58: 556–561.

  20. 20

    Sanacora G, Gueorguieva R, Epperson CN, Wu YT, Appel M, Rothman DL et al. Subtype-specific alterations of gamma-aminobutyric acid and glutamate in patients with major depression. Arch Gen Psychiatry 2004; 61: 705–713.

  21. 21

    Kaufman DL, Houser CR, Tobin AJ . Two forms of the gamma-aminobutyric acid synthetic enzyme glutamate decarboxylase have distinct intraneuronal distributions and cofactor interactions. J Neurochem 1991; 56: 720–723.

  22. 22

    Martin DL, Rimvall K . Regulation of gamma-aminobutyric acid synthesis in the brain. J Neurochem 1993; 60: 395–407.

  23. 23

    Bowers G, Cullinan WE, Herman JP . Region-specific regulation of glutamic acid decarboxylase (GAD) mRNA expression in central stress circuits. J Neurosci 1998; 18: 5938–5947.

  24. 24

    Esclapez M, Tillakaratne NJ, Kaufman DL, Tobin AJ, Houser CR . Comparative localization of two forms of glutamic acid decarboxylase and their mRNAs in rat brain supports the concept of functional differences between the forms. J Neurosci 1994; 14(Part 2): 1834–1855.

  25. 25

    Feldblum S, Erlander MG, Tobin AJ . Different distributions of GAD65 and GAD67 mRNAs suggest that the two glutamate decarboxylases play distinctive functional roles. J Neurosci Res 1993; 34: 689–706.

  26. 26

    Pitkanen A, Amaral DG . The distribution of GABAergic cells, fibers, and terminals in the monkey amygdaloid complex: an immunohistochemical and in situ hybridization study. J Neurosci 1994; 14: 2200–2224.

  27. 27

    Stork O, Yamanaka H, Stork S, Kume N, Obata K . Altered conditioned fear behavior in glutamate decarboxylase 65 null mutant mice. Genes Brain Behav 2003; 2: 65–70.

  28. 28

    Kash SF, Tecott LH, Hodge C, Baekkeskov S . Increased anxiety and altered responses to anxiolytics in mice deficient in the 65-kDa isoform of glutamic acid decarboxylase. Proc Natl Acad Sci USA 1999; 96: 1698–1703.

  29. 29

    Asada H, Kawamura Y, Maruyama K, Kume H, Ding R, Ji FY et al. Mice lacking the 65 kDa isoform of glutamic acid decarboxylase (GAD65) maintain normal levels of GAD67 and GABA in their brains but are susceptible to seizures. Biochem Biophys Res Commun 1996; 229: 891–895.

  30. 30

    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.

  31. 31

    Kaiya H, Namba M, Yoshida H, Nakamura S . Plasma glutamate decarboxylase activity in neuropsychiatry. Psychiatry Res 1982; 6: 335–343.

  32. 32

    Smoller JW, Rosenbaum JF, Biederman J, Susswein LS, Kennedy J, Kagan J et al. Genetic association analysis of behavioral inhibition using candidate loci from mouse models. Am J Med Genet 2001; 105: 226–235.

  33. 33

    Lappalainen J, Sanacora G, Kranzler HR, Malison R, Hibbard ES, Price LH et al. Mutation screen of the glutamate decarboxylase-67 gene and haplotype association to unipolar depression. Am J Med Genet B Neuropsychiatr Genet 2004; 124: 81–86.

  34. 34

    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.

  35. 35

    Kendler KS, Prescott CA . A population-based twin study of lifetime major depression in men and women. Arch Gen Psychiatry 1999; 56: 39–44.

  36. 36

    Spitzer RL, Williams JBW . Structured Clinical Interview for DSM-III-R (SCID). Biometrics Research Department, New York State Psychiatric Institute: New York, 1985.

  37. 37

    American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, revised 3rd edn. American Psychiatric Association: Washington, DC, 1987.

  38. 38

    Hettema JM, Prescott CA, Kendler KS . A population-based twin study of generalized anxiety disorder in men and women. J Nerv Ment Dis 2001; 189: 413–420.

  39. 39

    Kendler KS, Gardner CO, Prescott CA . Panic syndromes in a population-based sample of male and female twins. Psychol Med 2001; 31: 989–1000.

  40. 40

    American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 3rd edn. American Psychiatric Association: Washington, DC, 1980.

  41. 41

    Kendler KS, Myers J, Prescott CA, Neale MC . The genetic epidemiology of irrational fears and phobias in men. Arch Gen Psychiatry 2001; 58: 257–265.

  42. 42

    Eysenck HJ&ESBG. Manual of the Eysenck Personality Questionnaire. Hodder and Stoughton: London, 1975.

  43. 43

    Robles JR, van den Oord EJ . lga972: a cross-platform application for optimizing LD studies using a genetic algorithm. Bioinformatics 2004; 20: 3244–3245.

  44. 44

    van den Oord EJ, Sullivan PF . A framework for controlling false discovery rates and minimizing the amount of genotyping in the search for disease mutations. Hum Hered 2003; 56: 188–199.

  45. 45

    Neale MC, Boker SM, Xie G, Maes HH . Mx: Statistical Modeling, 5th edn. Department of Psychiatry, Medical College of VA of VA Commonwealth University: Richmond, VA, 1999.

  46. 46

    Kendler KS, Neale MC, Kessler RC, Heath AC, Eaves LJ . The genetic epidemiology of phobias in women. The interrelationship of agoraphobia, social phobia, situational phobia, and simple phobia. Arch Gen Psychiatry 1992; 49: 273–281.

  47. 47

    Neale MC, Cardon LR . Methodology for Genetic Studies of Twins and Families. Kluwer Academic Publishers BV: Dordrecht, The Netherlands, 1992.

  48. 48

    Hettema JM, Neale MC, Myers JM, Prescott CA, Kendler KS . A population-based twin study of the relationship between neuroticism and internalizing disorders. Am J Psychiatry 2006; 163: 857–864.

  49. 49

    Barrett JC, Fry B, Maller J, Daly MJ . The International HapMap Project. Nature 2003; 426: 789–796.

  50. 50

    Lowe CE, Cooper JD, Chapman JM, Barratt BJ, Twells RC, Green EA et al. Cost-effective analysis of candidate genes using htSNPs: a staged approach. Genes Immun 2004; 5: 301–305.

  51. 51

    Bukszár J, Van den Oord EJCG . An asymptotic approximation for Pearson's statistic in two-stage genetic designs where data are pooled. Biometrics 2005 (in press).

  52. 52

    Bukszár J, Van den Oord EJCG . Accurate and efficient power calculations for 2 × m tables in unmatched case–control designs. Statist Med 2005 (e-published 18 July 2005).

  53. 53

    Niu T, Qin ZS, Xu X, Liu JS . Bayesian haplotype inference for multiple linked single-nucleotide polymorphisms. Am J Hum Genet 2002; 70: 157–169.

  54. 54

    Qin ZS, Niu T, Liu JS . Partition-ligation-expectation-maximization algorithm for haplotype inference with single-nucleotide polymorphisms. Am J Hum Genet 2002; 71: 1242–1247.

  55. 55

    Stram DO . Tag SNP selection for association studies. Genet Epidemiol 2004; 27: 365–374.

  56. 56

    Barrett JC, Fry B, Maller J, Daly MJ . Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 263–265.

  57. 57

    Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, Blumenstiel B et al. The structure of haplotype blocks in the human genome. Science 2002; 296: 2225–2229.

  58. 58

    Dudbridge F . Pedigree disequilibrium tests for multilocus haplotypes. Genet Epidemiol 2003; 25: 115–121.

  59. 59

    Excoffier L, Slatkin M . Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol 1995; 12: 921–927.

  60. 60

    Colhoun HM, McKeigue PM, Davey SG . Problems of reporting genetic associations with complex outcomes. Lancet 2003; 361: 865–872.

  61. 61

    Freimer N, Sabatti C . The use of pedigree, sib-pair and association studies of common diseases for genetic mapping and epidemiology. Nat Genet 2004; 36: 1045–1051.

  62. 62

    Van den Oord EJCG . Controlling false discoveries in candidate gene studies. Mol Psychiatry 2005; 10: 230–231.

  63. 63

    Storey J, Tibshirani R . Statistical significance for genome-wide studies. Proc Natl Acad Sci USA 2003; 100: 9440–9445.

  64. 64

    Wacholder S, Chanock S, Garcia-Closas M, El Ghormli L, Rothman N . Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J Natl Cancer Inst 2004; 96: 434–442.

  65. 65

    Van den Oord EJCG, Sullivan PF . False discoveries and models for gene discovery. Trends Genet 2003; 19: 537–542.

  66. 66

    Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG . Replication validity of genetic association studies. Nat Genet 2001; 29: 306–309.

  67. 67

    Straub RE, Sullivan PF, Ma Y, Myakishev MV, Harris-Kerr C, Wormley B et al. Susceptibility genes for nicotine dependence: a genome scan and followup in an independent sample suggest that regions on chromosomes 2, 4, 10, 16, 17 and 18 merit further study. Mol Psychiatry 1999; 4: 129–144.

  68. 68

    Livak KJ . Allelic discrimination using fluorogenic probes and the 5′ nuclease assay. Genet Anal 1999; 14: 143–149.

  69. 69

    SAS Institute. SAS/STAT Software: Version 8. SAS Institute Inc.: Cary, NC, 1999.

  70. 70

    Pritchard JK, Stephens M, Donnelly P . Inference of population structure using multilocus genotype data. Genetics 2000; 155: 945–959.

  71. 71

    Devlin B, Roeder K . Genomic control for association studies. Biometrics 1999; 55: 997–1004.

  72. 72

    Freedman ML, Reich D, Penney KL, McDonald GJ, Mignault AA, Patterson N et al. Assessing the impact of population stratification on genetic association studies. Nat Genet 2004; 36: 388–393.

  73. 73

    Sullivan PF, Neale MC, Silverman MA, Harris-Kerr C, Myakishev MV, Wormley B et al. An association study of DRD5 with smoking initiation and progression to nicotine dependence. Am J Med Genet 2001; 105: 259–265.

Download references

Acknowledgements

This work was supported by NIH grants MH-40828, MH-65322, MH-20030, DA-11287, MH/AA/DA-49492 to KSK and K08 MH-66277-1, as well as a Junior Faculty Research Award from the Anxiety Disorders Association of America, an NARSAD Young Investigator Award and a Pfizer/SWHR Scholars Award to JMH. We acknowledge the contribution of the Virginia Twin Registry, now part of the Mid-Atlantic Twin Registry (MATR), to ascertainment of subjects for this study. The MATR, directed by Drs J Silberg and L Eaves, has received support from the National Institutes of Health, the Carman Trust and the WM Keck, John Templeton and Robert Wood Johnson Foundations.

Author information

Correspondence to J M Hettema.

Additional information

Preliminary results from this study were presented at the XIIIth World Congress on Psychiatric Genetics, October 14–18, 2005 in Boston, MA, USA.

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

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Hettema, J., An, S., Neale, M. et al. Association between glutamic acid decarboxylase genes and anxiety disorders, major depression, and neuroticism. Mol Psychiatry 11, 752–762 (2006). https://doi.org/10.1038/sj.mp.4001845

Download citation

Keywords

  • glutamic acid decarboxylase
  • depression
  • anxiety
  • neuroticism
  • association study
  • genetics

Further reading