Functional (GT)n polymorphisms in promoter region of N-methyl-d-aspartate receptor 2A subunit (GRIN2A) gene affect hippocampal and amygdala volumes

Authors


H. Inoue, Department of Neuropsychiatry, Graduate School of Medicine, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8655, Japan.
E-mail: inouehide-tky@umin.ac.jp

Abstract

The glutamate system including N-methyl-d-aspartate (NMDA) affects synaptic formation, plasticity and maintenance. Recent studies have shown a variable (GT)n polymorphism in the promoter region of the NMDA subunit gene (GRIN2A) and a length-dependent inhibition of transcriptional activity by the (GT)n repeat. In the present study, we examined whether the GRIN2A polymorphism is associated with regional brain volume especially in medial temporal lobe structures, in which the NMDA-dependent synaptic processes have been most extensively studied. Gray matter regions of interest (ROIs) for the bilateral amygdala and hippocampus were outlined manually on the magnetic resonance images of 144 healthy individuals. In addition, voxel-based morphometry (VBM) was conducted to explore the association of genotype with regional gray matter volume from everywhere in the brain in the same sample. The manually measured hippocampal and amygdala volumes were significantly larger in subjects with short allele carriers (n = 89) than in those with homozygous long alleles (n = 55) when individual differences in intracranial volume were accounted for. The VBM showed no significant association between the genotype and regional gray matter volume in any brain region. These findings suggest that the functional GRIN2A (GT)n polymorphism could weakly but significantly impact on human medial temporal lobe volume in a length-dependent manner, providing in vivo evidence of the role of the NMDA receptor in human brain development.

The N-methyl-d-aspartate (NMDA) type-glutamate receptor plays an important role in synapse formation, plasticity and maintenance (Konradi & Heckers 2003). NMDA receptors (NMDARs) are heteromers consisting of NR1 subunits and one or more of four NR2 subunits (NR2A–D). Among the NR2 subunits, NR2A and NR2B subunits predominate in the forebrain. During postnatal development, changes occur in the expression of NR2A relative to that of NR2B subunits (Loftis & Janowsky 2003; Monyer et al. 1994; Sheng et al. 1994). These changes in the NMDAR subunit composition may be associated with activity-dependent cortical development (Barth & Malenka 2001; Quinlan et al. 1999).

Recently, a variable (GT)n repeat in the 5′-regulatory region of the NMDAR NR2A subunit (GRIN2A) gene has been identified (Itokawa et al. 2003). It was shown that the repeat sequence repressed transcriptional activity in a length-dependent manner, such that the longer the repeat, the lower the promoter activity. Itokawa et al. (2003) examined the NMDAR binding sites in the parietal and temporal cortical areas of postmortem brains, using the [3H] MK801 ligand, and demonstrated that subjects with genotypes consisting of long GT repeat alleles showed few binding sites for the ligand in both these brain areas. Additionally, the (GT)n polymorphism in the promoter of GRIN2A has been reported to be associated with schizophrenia, with longer alleles more represented in schizophrenia (Iwayama-Shigeno et al. 2005; Tang et al. 2006). This association was consistent with the glutamate dysfunction hypothesis of the disorder (Harrison & West 2006; Konradi & Heckers 2003; Mohn et al. 1999), although a negative finding has also been reported (Zhao et al. 2006). Furthermore, NR2A knockout mice exhibited increased locomotor activity in a novel environment and an impairment of latent learning, which was attenuated by treatment with the antipsychotic drugs (Miyamoto et al. 2001).

Previous studies have suggested a critical role of NMDARs in hippocampal development. In addition, cellular and animal studies have suggested that NR2A subunit function and the switch from NR2B to NR2A subunits in synaptic NMDARs in the hippocampus serve as a modulator of hippocampal synaptic plasticity underlying learning and memory (Barria & Malinow 2005; Miyamoto et al. 2005; Sakimura et al. 1995).

Regional brain volume reductions have been consistently reported in patients with schizophrenia, particularly in the hippocampus as well as the dorsolateral prefrontal cortex (DLPFC) and superior temporal gyrus (STG) gray matter compared with those of normal individuals (Honea et al. 2005; Shenton et al. 2001). Another important locus for NMDA-dependent synaptic plasticity and memory function is the amygdala (Lopez de Armentia & Sah 2003; Maren & Quirk 2004; Rogan et al. 1997; Zinebi et al. 2003). Moreover, the expression of GRIN2A starts around puberty and maintains a plateau throughout adult life. This may correspond to the period of onset and subsequent chronic course of schizophrenia. Taking the above findings together, it is interesting to observe the relationship between the function of the glutamate system by means of functional polymorphisms of the NR2A subunit gene and regional brain volume.

The present study examined the association between the GRIN2A polymorphism and regional brain volume as assessed by structural magnetic resonance imaging (MRI) in mentally healthy individuals using a combination of manual tracing and an automated method called voxel-based morphometry (VBM) (Ashburner & Friston 2000). We expected that the presence of longer (GT)n repeat would be associated with smaller volumes of such regions as the hippocampus and amygdala, which are major loci affected by NMDA-dependent synaptic processes. It was also predicted that an even smaller genotypic effect would be detected in the DLPFC and STG gray matter volumes as inferred from literature concerning patients with schizophrenia (Egan et al. 2004; Javitt et al. 1994, 1996; Marenco et al. 2006).

Materials and methods

Subjects

The subjects were 144 (96 men and 48 women) Japanese adults (age, mean ± SD: 30.8 y.o. ± 7.7), consisting of college students, hospital staff and their acquaintances. Before MRI scanning, the subjects were screened using the Structured Clinical Interview for DSM-IV Axis I Disorder, Nonpatient Edition (SCID-NP) (First et al. 1997, Japanese version; Kitamura and Okano 2003) by a trained psychiatrist (H.Y. or M.S.) to confirm that the subjects had no history of major mental illness. Other exclusion criteria were neurological illness, traumatic brain injury with any known cognitive consequences or loss of consciousness for more than 5 min, a history of substance abuse or addiction or a family history of an axis I disorder in their first-degree relatives. All were right-handed based on the Edinburgh Inventory (Oldfield 1971). The socioeconomic status (SES) and parental SES were assessed using the Hollingshead scale (Hollingshead 1957). After a thorough explanation of the study to the subjects, written informed consent was obtained. The ethical committee of the Faculty of Medicine, University of Tokyo approved this study (No. 397-1 for MRI project & No. 639–9 for genetic & imaging-genetic-association project).

MRI acquisition

The method of MRI acquisition of 1.5 mm slices was the same as that described in our previous study (Yamasue et al. 2003). The MRI data were obtained using a 1.5-Tesla scanner (General Electric Signa Horizon Lx version 8.2, GE Medical Systems, Milwaukee, WI, USA). We used three-dimensional Fourier-transform spoiled-gradient-recalled acquisition with steady state. The repetition time was 35 milliseconds, the echo time 7 milliseconds with one repetition, the nutation angle 30 degrees, the field of view 24 cm and the matrix 256 × 256 (192)× 124.

Manual tracing for hippocampus and amygdala volumetry

The amygdala and hippocampus gray matter regions of interest (ROIs) were outlined manually by one rater (H.I.) who was blind to the group status or genotype (Fig. 1). For the manual tracing, we used a software package for medical image analysis (3d slicer; software available at http://www.slicer.org), which enables a simultaneous view of orthogonal planes. The landmarks to delineate the ROIs were the same as our recent study (Yamasue et al. 2008).

Figure 1.

Two- (panels A-D) and 3- (panels E & F) dimensional images of regions of interest. The gray matter of the amygdala is labeled light green on subject's left and violet on subject's right. The gray matter of hippocampus is light blue on subject left and mulberry on subject right. A-D: Delineation of the amygdala and hippocampus in coronal slices from rostral to caudal parts of regions of interest based on MRI data of a control subject. E & F: 3-D reconstructions of the amygdala and hippocampus superimposed on the sagittal plane of the left hemisphere (panel E) and on the axial plane (panel F). The coronal lines A-D correspond to the planes of panels A-D, respectively.

For interrater reliability, two raters (H.I. and M.R.) blind to group membership, independently drew ROIs. Ten cases were selected at random, and the raters drew ROIs on every slice. The intraclass correlation coefficient was 0.87/0.85 for the left/right amygdala and 0.93/0.94 for the left/right hippocampus respectively. Intrarater reliability, computed by using all of the slices from one randomly selected brain and measured by one rater (H.I.) on two separate occasions (approximately 2 months apart), was > 0.95 for all structures.

Total gray matter, white matter and cerebrospinal fluid volumes were calculated by the VBM procedure described below. Then, intracranial volume (ICV) was calculated by summing up the total gray matter, white matter and cerebrospinal fluid volumes. To validate this method, the ICVs of an independent sample of MRI scans for 50 adult subjects were measured by both VBM and intensity-based semiautomated segmentation procedure using analyze pc 3.0 (Yamasue et al. 2004). Then, we confirmed that the calculated intraclass correlation coefficient for the ICVs was satisfactory (0.96).

Image processing for VBM

Image processing for VBM (Ashburner & Friston 2000; Good et al. 2001), a fully automatic technique for computational analysis of differences in local tissue volume throughout the brain, was conducted using spm2 (Institute of Neurology, London, UK). The methods were the same as those of our previous study (Yamasue et al. 2008) (see Supporting Information).

Genotyping

Genomic DNA was extracted from peripheral leukocytes using a standard phenol–chloroform method. The (GT)n repeat polymorphism of the GRIN2A gene was genotyped as described by Itokawa et al. (2003), by PCR-amplifying the genomic fragments with fluorescence-labeled primers and analyzing them in an ABI 3730 sequencer equipped with genescan software (Applied Biosystems, Foster City, CA).

Statistical analyses

The effects of the GRIN2A genotype on the manually traced volumes were assessed by repeated-measures analysis of variance (ANOVA) adopting relative volumes [100 × absolute ROI volume/(ICV)] as the dependent variable, genotype as the between-subject factor and region (amygdala/hippocampus) and hemisphere (left/right) as the within-subject factors. A statistically significant level was set at P < 0.05. For testing regional specificity, the volumes of total gray matter, total white matter, CSF and ICV were also compared between genotypes using independent t-tests. Additionally, % difference was used to assess the effects of the GRIN2A genotype on the absolute and relative volumes of each region, and Pearson's r was used to calculate correlations between the genotypic indices, including the repeat number of (GT)n in longer allele and mean value of longer and shorter allele and the relative volumes of each region in the entire study sample. Effect size was the difference score for mean (s carriers) relative to mean (l/l subjects) divided by pooled SD (all subjects).

Statistical analyses of VBM were performed using an analysis of covariance model (Friston et al. 1990). To account for global anatomical variations, the ICV was treated as a confounding covariate. To detect neuroanatomical correlates of the GRIN2A genotype, the statistical analysis treated the ICV as a confounding covariate, and the GRIN2A genotype as condition. The resulting set of voxel values for each contrast constituted a statistical parametric map of the t-statistic (spm{t}). The significance of each region was corrected for multiple comparisons using the false discovery rate (FDR) (Genovese et al. 2002). The statistical significance level was set at the FDR-corrected P < 0.05. While significant effects were explored throughout the entire gray matter regions, small-volume correction was employed in predicted regions on the basis of previous studies on the locus of the NMDA-dependent plasticity and implications for its possible deficits in schizophrenia: hippocampus (Nacher & McEwen 2006), STG (Javitt et al. 1994, 1996) and DLPFC (Egan et al. 2004; Marenco et al. 2006). In contrast to the whole gray matter exploration, familywise error (FWE)-corrected P was conservatively employed to detect findings within the searched volumes (SV) that were determined according to wfu_pickatlas (a software available at http://fmri.wfubmc.edu/cms/software#PickAtlas).

Results

Figure 2 shows the allele distribution pattern of the current samples: mean [±SD] repeat number of (GT)n was 24.7 [±2.5], and the repeat number ranged from 19 to 34. The repeat number of (GT)n was divided into two classes on the basis of allele size and to ensure a statistically valid comparison; the sample size of each class was made as similar as possible [short (24 or fewer tandem repeats) and long (more than 24 repeats)]. This classification gave rise to three groups: short allele homozygotes 23 (16%), short/long heterozygotes 66 (46%), long allele homozygotes 55 (38%). The sample size of short allele homozygotes was too small to provide sufficient statistical power to draw a conclusion. Thus, the three genotypes in the GRIN2A gene were classified into two subgroups: short allele carriers (n = 89) and long allele homozygotes (n = 55). No significant difference was observed in gender, age, handedness, self SES or parental SES between the two genotype groups (Table 1). Total gray matter, total white matter, CSF, or ICV volume was also not significantly different between the two genotype groups (P > 0.51).

Figure 2.

Allelic frequency of (GT)n repeat in the 5’-regulatory region of GRIN2A. Since the number of repeats is presented separately for each subject, the total number of the length of the tandem repeats is shown (total allele N = 288).

Table 1.  Clinical and demographic characteristics of the study participants
 Short allele carriers (n = 89)Long allele homozygotes (n = 55)Group comparison
MeanSDMeanSD
  1. *Determined using Edinburgh Inventory (Oldfield 1971): scores greater than 0 indicate right-handedness and a score of 100 indicates strong right-handedness.

  2. The degrees of freedom varied due to unavailability of data in some subjects.

  3. Assessed using the Hollingshead scale (Hollingshead 1957). Higher scores indicate lower educational and/or occupational status.

Sex (male/female)59/3037/18chi-square [1] = 0.015, P = 0.90
Age (range)30.4 (21–59)7.431.5 (22–64)8.3t [142] = −0.86, P = 0.39
Handedness*95.812.597.55.8t [133] = −0.90, P = 0.37
Socioeconomic status1.60.641.60.65t [142] = 0.78, P = 0.44
Parental socioeconomic status2.20.712.10.80t [138] = −0.37, P = 0.72

For the genotype effects on the hippocampus and amygdala volumes, the repeated-measures ANOVA showed that there was a significant main effect of the GRIN2A genotype (F[1, 142] = 4.61, P = 0.034). There was no significant interaction between genotype and region (F[1, 142] = 2.89, P = 0.091), genotype and hemisphere (F[1, 142] = 0.23, P = 0.63), or genotype and region and hemisphere (F[1, 142] = 1.24, P = 0.27). The statistical conclusion from the main repeated-measures ANOVA was that subjects with short allele carriers had significantly larger ROI volumes than those with homozygous long alleles, without significant laterality or regional specificity among bilateral hippocampi and amygdalae. Although we cannot formally proceed to the post-hoc ANOVA separately for the hippocampus and the amygdala because the genotype-by-region interaction was not significant (P = 0.091), we tentatively conducted the analysis to elucidate the origin of this trend. There was a significant main effect of the genotype in the hippocampus (F[1, 142] = 4.65, P = 0.033) but not in the amygdala (F[1, 142] = 2.33, P = 0.13). The effect size between GRIN2A short allele carriers and long allele homozygotes on hippocampal relative volume [total: 0.37 (4.9% difference), left: 0.30 (4.2%), right: 0.39 (5.5 %)] was larger than that on the amygdala relative volume [total: 0.26 (3.1%), left: 0.28 (3.6%), right: 0.20 (2.7%)] (Table 2, Fig. 3). Finally, the repeat number of (GT)n in longer allele and the average of repeat numbers of (GT)n were correlated with the relative volume of right hippocampus in the entire study sample (longer; R = −0.167 P = 0.045, mean; R = −0.176 P = 0.035); however, there were not any other significant associations (P's > 0.30).

Table 2.  Brain volume* and GRIN2A polymorphism
inline image
Figure 3.

Manually traced hippocampal and amygdala volumes and GRIN2A genotype. Means and distributions of left and right hemispheres for relative manually traced hippocampal and amygdala volumes (absolute volume ×100/intracranial content) in which individuals with GRIN2A short allele carriers exhibited bilateral enlargement compared with individuals with GRIN2A long allele homozygotes (F [1,142] = 4.61, p = 0.034). Means are represented by solid horizontal lines drawn on each group's distribution.

Voxel-based morphometry revealed where the GRIN2A short allele carriers showed larger regional gray matter volume than those with homozygous long alleles in the predicted regions: left hippocampus ([−12, −6, −22], z = 1.74, uncorrected P = 0.041, FWE-corrected P = 0.379 with 6.2 ml SV), left DLPFC (Brodmann area 9 and 46) ([−38, 22, 40], z = 2.12, uncorrected P = 0.017, FWE-corrected P = 0.429 with 7.13 ml SV), right DLPFC ([34, 24, 42], z = 1.84, uncorrected P = 0.033, FWE-corrected P = 0.453 with 7.28 ml SV), left STG ([−66 − 36, 18], z = 2.44, uncorrected P = 0.007, FWE-corrected P = 0.190 with 17.21 ml SV) and right STG ([58, −60, 22], z = 2.48, uncorrected P = 0.007, FWE-corrected P = 0.201 with 21.01 ml SV). However, no regions reached the statistically significant level after correction for multiple comparisons.

Discussion

To the best of our knowledge, this study is the first to examine the effects of an ionotropic glutamate receptor gene polymorphism on MRI volume in humans. As postulated, the findings revealed that the individuals with long repeats in the 5′-regulatory region of GRIN2A had significantly smaller hippocampal and amygdala volumes than did short repeat carriers. Given that the (GT)n repeat suppresses promoter activity of the GRIN2A gene in vitro and protein (receptor) expression levels in vivo in a length-dependent manner (Itokawa et al. 2003), the results suggest that the attenuated NMDA neurotransmission may lead to smaller regional medial temporal structure volumes. Although speculative, these results also imply that there may be some graduation in the effect of the GRIN2A genotype on medial temporal lobe volumes, i.e. a relatively larger impact on the hippocampus than on the amygdala. It is noteworthy that long allele homozygotes did not differ significantly from short allele carriers in global brain measurements such as total gray/white matter volume or ICV, suggesting that the observed relationship may be confined to the specific brain regions in question.

On the other hand, the current study failed to reveal significant associations between genotype and regional volumes in STG and DLPFC. Because of the labor intensive nature of manual tracing, we conducted manual tracing not on STG and DLPFC but rather on the more easily defined and delimited hippocampus and amygdala. The current study showed no associations between the genotype and the volume of STG and DLPFC by using VBM. However, we cannot totally rule out the possibility that additional manual tracing analysis of these structures in future studies might detect a weak but significant effect of genotype.

The potential limitations and methodological considerations of this study are as follows. Firstly, this study includes subjects from a broad age range and of both genders. Potential confounding effects of aging and/or gender on the obtained results cannot be completely ruled out, although there were no significant age and gender ratio differences in the two genotype groups. Secondly, recent studies of ‘maging genetics', which explore the associations between genetic variations and neuroimaging measures, should be conducted and interpreted with caution. A polymorphism in a single gene relevant to neurodevelopment may only account for a limited amount of variance of brain structure and function in adults. At the point of designing the study, we should make a strong a priori hypothesis based on knowledge of the function of the selected gene and polymorphism. The sample size in the current study was relatively large. However, it is still possible that our findings have been affected by statistical fluctuations that may have stemmed from an insufficient sample size. Thus, replications in independent and larger sample sets are essential to fully confirm our findings. Third, the findings obtained by manual tracing were not confirmed by VBM. The discrepancy in results from manual tracing and VBM might derive from the difference in the characteristics of the two technologies. Voxel-based approaches like VBM have the issue of spatial normalization (Kubicki et al. 2002; White et al. 2003) and excessive correction of significance level for multiple comparisons. Therefore, the current manual tracing methodology might successfully reveal a weak effect of a single gene polymorphism on the volumes of medial temporal structures, which could not be detected by VBM. However, it cannot be totally ruled out that the manual tracing findings are a false positive correctly rejected by the VBM analysis. Forth, the present results should be interpreted cautiously. The currently observed genotype effect on regional brain volumes were weak but significant, because the effect sizes of genotype effects were smaller than those of gender effects [left hippocampus, 0.76; right hippocampus, 0.52; left amygdala, 0.36; right amygdala, 0.33 (female > male)]. Finally, the present study could not suggest any functional correlates of volumetric measures, because any behavioral or cognitive indices reflecting individual learning and memory abilities were not available in the current study population.

In conclusion, the current findings suggest that the functional (GT)n polymorphism in the promoter of GRIN2A may impact on human hippocampal and amygdala volumes and that the lower NMDAR-mediated neural transmission may lead to reduced volumes of these anatomical substrates.

Acknowledgments

This study was supported in part by grants-in-aid for scientific research (No. 18390319 & 20023009 to K.K.) from Japan Society for the Promotion of Science and the Ministry of Education, Culture, Sports, Science and Technology, Japan, and by grants-in-aid (H20-3 & Kokoro-Ippan H17-012 & H20-001 to K.K.) from the Ministry of Health, Labor and Welfare, Japan and by the National Health and Medical Research Council (Australia grant ID 237027 to M.A.R). We wish to express our gratitude to the members of the Research Resource Center at the RIKEN Brain Science Institute for the genotyping.

Ancillary