Modulative effects of COMT haplotype on age‐related associations with brain morphology

Abstract

Catechol‐O‐methyltransferase (COMT), located on chromosome 22q11.2, encodes an enzyme critical for dopamine flux in the prefrontal cortex. Genetic variants of COMT have been suggested to functionally manipulate prefrontal morphology and function in healthy adults. This study aims to investigate modulative roles of individuals COMT SNPs (rs737865, val158met, rs165599) and its haplotypes in age‐related brain morphology using an Asian sample with 174 adults aged from 21 to 80 years. We showed an age‐related decline in cortical thickness of the dorsal visual pathway, including the left dorsolateral prefrontal cortex, bilateral angular gyrus, right superior frontal cortex, and age‐related shape compression in the basal ganglia as a function of the genotypes of the individual COMT SNPs, especially COMT val158met. Using haplotype trend regression analysis, COMT haplotype probabilities were estimated and further revealed an age‐related decline in cortical thickness in the default mode network (DMN), including the posterior cingulate, precuneus, supramarginal and paracentral cortex, and the ventral visual system, including the occipital cortex and left inferior temporal cortex, as a function of the COMT haplotype. Our results provided new evidence on an antagonistic pleiotropic effect in COMT, suggesting that genetically programmed neural benefits in early life may have a potential bearing towards neural susceptibility in later life. Hum Brain Mapp 37:2068–2082, 2016. © 2016 Wiley Periodicals, Inc.

Abbreviations

AG

Angular gyrus

CAL

Calcarine sulcus

COMT

Catechol‐O‐methyltransferase

CSF

Cerebral spinal fluid

DMN

Default mode network

ITC

Inferior temporal cortex

LDDMM

Large deformation diffeomorphic metric mapping

MAF

Minor allele frequencies

MMSE

Mini‐mental state examination

MPRAGE

Magnetization prepared rapid gradient recalled echo

MTG

Middle temporal gyrus

OCC

Occipital

OFC

Orbitofrontal cortex

PCC

Posterior cingulate cortex

PCUN

Precuneus

SFC

Superior frontal cortex

SMG

Supramarginal gyrus

SNP

Single nucleotide polymorphism

INTRODUCTION

Converging data from structural imaging studies suggests that aberrant brain morphology plays an important role in explaining the pathophysiology of aging. For instance, a considerable age‐dependent reduction in hippocampal volumes has been identified as a hallmark of aging [Allen et al., 2005; Geerligs et al., 2014; Goodro et al., 2012; Jernigan et al., 2001; Pruessner et al., 2001; Raz et al., 2010; Scahill et al., 2003; Zhu et al., 2008]. Normal aging is also characterized by prominent cortical thinning in the prefrontal cortex and sparse cortical thinning in the temporal, parietal, and occipital cortices, suggesting an anterior‐to‐posterior gradient of decline in aging [Fjell et al., 2009; Lemaitre et al., 2012; Raz, 2000; Raz et al., 2004; Salat et al., 2004; Wang et al., 2008].

Genetic variants also contribute the variation of hippocampal and subcortical volumes, suggested by several large scale imaging genetic studies [Bis et al., 2011; Hibar et al., 2015]. We hence expect that the association of age and brain morphology might be contingent upon genetic variants [Panizzon et al., 2009; Peper et al., 2007; Thompson et al., 2001]. Catechol‐O‐methyltransferase (COMT), located on chromosome 22q11.2, is well expressed in the brain throughout the cortex, basal ganglia, amygdala etc. COMT is a major enzyme degrading catecholamine, especially dopamine [Mier et al., 2010]. Its functional single nucleotide polymorphism (SNP), COMT val158met, results in the amino acid substitution of valine (val) with methionine (met) at position 158 of COMT [Bray et al., 2003]. The val allele is suggested to determine higher COMT activity in the prefrontal cortex, which presumably leads to lower synaptic dopamine levels through enhanced degradation [Chen et al., 2004]. Thus, COMT has received great attention for understanding how its genetic variants may functionally interact with the brain morphology and function in healthy adults [Aguilera et al., 2008; Cerasa et al., 2008; Dennis et al., 2010b; Diaz‐Asper et al., 2008; Egan et al., 2001; Ehrlich et al., 2010; Honea et al., 2009a; Liao et al., 2009; Papenberg et al., 2014; Sannino et al., 2014; Tunbridge et al., 2004]. By far, studies involving young healthy adults have found that val homozygous showed greater gray matter volumes in the dorsolateral prefrontal cortex (dlPFC), cingulate, temporal and superior parietal lobes as compared to met allele carriers [Cerasa et al., 2008; Dumontheil et al., 2011; Honea et al., 2009a; Raznahan et al., 2011; Rowe et al., 2010]. Nevertheless, one would expect the smaller volumes in the aforementioned brain regions in young adults with val alleles, as val carriers have showed greater enzymatic activity and led greater dopamine degradation when compared to met homozygotes. Together, these could imply that val alleles confer neural benefits for the brain development during early adulthood. There are lack of studies that characterize the alterations in brain morphology due to the genotypes of COMT val158met across lifespan. It hence remains unclear whether val alleles would accelerate neurodegeneration in old adults in comparisons with met alleles.

In addition to COMT val158met, sequence variants lying in other loci within the COMT gene affect gene function, ultimately resulting in altered enzyme activity and adding complexity to the functional implications of COMT variations [Gothelf et al., 2014]. The 3‐SNP haplotype of COMT [upstream SNP in intro 1 (rs737865)‐val158met‐ downstream SNP in the 3’ untranslated region (rs165599)] has been found to alter the COMT expression in the human brain tissue [Bray et al., 2003], suggesting the importance of complex interactions of functional variations in COMT on the human brain. Both rs737865 and rs165599 were found to modulate cognitive functions, such as visuospatial memory, attention [Diaz‐Asper et al., 2008; Xia et al., 2012]. A landmark study has showed stronger statistical power for examining the association between prefrontal responses to working memory and the COMT haplotype than individual COMT SNPs [Meyer‐Lindenberg et al., 2006], suggesting the advantage to using haplotypes over single SNPs to capture the epistastic interactions among the COMT SNPs and hence improve statistical power over the analysis of single SNPs [Lorenz et al., 2010].

In this study, we aimed to examine whether individual COMT SNPs (rs737865, val158met, rs165599) and its haplotype module age effects on subcortical shapes and cortical thickness using a large Chinese sample comprised of adults aged from 21 years and above. We examined the subcortical shapes and group analysis of cortical thickness using an advanced brain mapping technique, large deformation diffeomorphic metric mapping (LDDMM) [Miller and Qiu, 2009]. We predicted that individual SNPs of the COMT gene (rs737865, val158met, and rs165599) would modulate the associations between age and the brain morphology, particularly in dopamine enriched regions, such as the prefrontal cortex and basal ganglia. In conjuction with the previous studies on the brain morphology in young adults [Cerasa et al., 2008; Dumontheil et al., 2011; Honea et al., 2009a; Raznahan et al., 2011; Rowe et al., 2010], our study provided new evidence of effects of COMT SNPs on brain morphology in older adults. We expected effects of COMT on the brain morphology in older adults opposite to those seen in young adults as reported in previous studies [Cerasa et al., 2008; Dumontheil et al., 2011; Honea et al., 2009a; Raznahan et al., 2011; Rowe et al., 2010]. Furthermore, we predicted that haplotype analysis should be superior in detecting such effects in comparison with individual SNPs. For this, the rs737865‐ val158met‐rs165599 haplotype was first estimated for each subject using an expectation‐maximization algorithm [Stephens and Donnelly, 2003]. Haplotype trend regression analysis was employed on cortical thickness and subcortical shapes, where estimated haplotype probabilities were used to increase statistical power and generate findings unbiased to samples with uniquely determined haplotypes. This study provides, to our knowledge, the first direct analysis of the genetic and age interactions associated with brain morphology and evidence that COMT may represent a molecular genetic marker moderating differential sensitivity to aging.

METHODS

Subjects

Two hundred and fourteen healthy Singaporean Chinese volunteers aged 21 to 80 years old were recruited (males: 93; females: 121). Volunteers with the following conditions were excluded from the recruitment: (1) major illnesses/surgery (heart, brain, kidney, lung surgery); (2) neurological or psychiatric disorders; (3) learning disability or attention deficit hyperactive disorder; (4) head injury with loss of consciousness; (5) non‐removable metal objects on/in the body such as a cardiac pacemaker; (8) diabetes or obesity; (9) a mini‐mental state examination (MMSE) score of less than 24 [Murray and Ranganath, 2007]. This study only included 174 subjects with both structural MRI scans and saliva samples. This study was approved by the National University of Singapore Institutional Review Board and all participants provided written informed consent prior to participation.

MRI Data Acquisition and Analysis

The MRI data were acquired using a 3T Siemens Magnetom Trio Tim scanner with a 32‐channel head coil at the Clinical Imaging Research Centre of the National University of Singapore. The image protocol was high‐resolution isotropic T₁‐weighted magnetization prepared rapid gradient recalled echo (MPRAGE; 192 slices, 1 mm thickness, sagittal acquisition, field of view 256 × 256 mm, matrix = 256 × 256, repetition time = 2,300 ms, echo time = 1.90 ms, inversion time = 900 ms, flip angle = 9°).

In the structural T ₁‐weighted MRI analysis, the gray matter, white matter, cerebral spinal fluid (CSF), lateral ventricles, and subcortical structures (amygdala, hippocampus, caudate, putamen, globus pallidus, thalamus) were automatically segmented from the intensity‐inhomogeneity corrected T₁‐weighted MR images [Fischl et al., 2002]. For shape analysis, the subcortical shapes were generated using the prior shape information of an atlas that was created from 41 manually labelled individual structures via a LDDMM atlas generation procedure [Li et al., 2008]. Shape variations of individual subjects relative to the atlas were characterized by the Jacobian determinant of the deformation in the logarithmic scale, where the deformation transformed the atlas shape to be similar to subjects. This measure, termed as the “deformation map,” represents the ratio of each subject's structural volume to the atlas volume in the logarithmic scale: i.e., positive values correspond to expansion, while negative values correspond to compression of the subject's structure relative to the atlas at each anatomical location.

For thickness analysis, an inner surface was constructed at the boundary between the white and gray matter and then propagated to the outer surface at the boundary between the gray matter and CSF. Cortical thickness was measured as the distance between the corresponding points on the inner and outer surfaces [Fischl and Dale, 2000]. LDDMM cortical surface mapping algorithm was then applied to align individual cortical surfaces to an atlas cortical surface for group analysis of cortical thickness [Zhong and Qiu, 2010].

SNP Genotyping

Saliva was collected from each participant through the Oragene DNA Self‐Collection Kit (DNA Genotek Inc., Kanata, Ontario, Canada). Then, the samples were genotyped on Illumina Omni express arrays, recently shown to perform well and have better coverage than competitors in Asian populations [Jiang et al., 2013], and on Illumina Exome arrays as well as Taqman array, following the manufacturer's instructions by Expression Analysis Inc. Data were processed in GenomeStudio Genotyping Module^TM. Genotyping calls were made by the GenCall software which incorporates a clustering algorithm (GenTrain) and a calling algorithm (Bayesian model). The GenCall Score of each SNP probe and call rate of each sample were generated. The GenCall score is primarily designed as a means by which to rank and filter out failed genotypes [Oliphant et al., 2002]. Scores below 0.15 generally indicate failed genotypes and hence the genotypes with a GenCall score of less than 0.15 were not assigned genotypes [Oliphant et al., 2002]. The SNPs COMT val158met, COMT rs737865, COMT rs165599, and APOE were then extracted and used in this study. Their minor allele frequencies (MAF) were greater than 0.05. The SNPs also did not deviate from the Hardy‐Weinberg equilibrium (HWE, P < 0.001).

Haplotype Estimation

To measure linkage disequilibrium between the COMT SNPs, haplotype frequencies were estimated through use of the expectation‐maximization algorithm on individual genotypes and calculated Lewontin's D: and Pearson's correlation using Haploview software, version 4.2. We then constructed likely three SNP haplotypes in our samples on the basis of the three polymorphisms using PHASE software, version 2.1.1 [Stephens and Donnelly, 2003]. Only haplotypes with an estimated frequency >3% were used in the analysis.

Statistical Analysis

Statistical analyses for cortical thickness and subcortical shapes were performed using SurfStat [Chung et al., 2010]. To examine interaction of age with the genotype of individual SNPs on cortical thickness and subcortical shapes, a regression model included age, SNP genotypes, their interaction as main factors and gender as covariate (brain measure ∼ gender + age + SNP genotypes + age * SNP genotypes). Age was considered as a continuous variable. The genotype of SNPs was considered as a categorical variable. This model was applied to the brain measure at each vertex on the atlas surface (cortical surface for cortical thickness or subcortical surface for subcortical shape). Statistical results at all vertices on the atlas surface were corrected for multiple comparisons at a cluster level of significance, P < 0.05. This statistical analysis was run separately for individual SNPs and individual structural measures. For subcortical volumes, the aforementioned regression analysis was performed using SPSS 18 for windows.

For post hoc analysis, we first examined age effects on cortical thickness and subcortical shapes in each genotype group of individual COMT SNPs. For this, the mean values of structural measures with the significant interaction effects of age and SNP was computed for individual brain regions. Regression analysis was examined on the mean values of structural measures with age as main factor and gender as covariate in each genotype group of individual COMT SNPs. Again, age was considered as a continuous variable. Simple slope analysis was further employed to compare the slopes of age‐related decline in the structural measures among the genotype groups.

For the second post‐hoc analysis, we examined whether individual genotypes of the COMT SNPs play the same roles in young and old adults. For this, the sample was divided into three groups, Young (20–39 years old), Middle (40–54 years old) and Old (55 and above). ANCOVA analysis was performed on the structural measures with the genotypes of COMT SNPs as main factor and gender as covariate in each age group. Bonferroni's correction was applied to correct multiple comparisons in the post hoc analysis. Both analyses were run separately for individual SNPs and individual brain regions.

Haplotype trend regression analysis was used to test for interactive effects of the COMT haplotype and age on cortical thickness, subcortical volumes and shapes [Oliphant et al., 2002]. The COMT haplotype was represented using its estimated probability for individual participants using PHASE software [Stephens and Donnelly, 2003]. The haplotype trend regression model included age, probability of haplotype, and their interaction as main factors and gender as covariate (brain measure ∼ gender + age + probability of haplotype + age * probability of haplotype). Age was considered as a continuous variable. This model was applied to the brain measure at each vertex on the atlas surface (cortical surface for cortical thickness or subcortical surface for subcortical shapes). Statistical results at all vertices on the atlas surface were corrected for multiple comparisons at a cluster level of significance, P < 0.05. This statistical analysis was run separately for individual structural measures.

RESULTS

Demographics

Among 214 recruited subjects, 30 subjects with no genetic data and 32 with missing structural MRI data were excluded, resulting in a final sample size of 174 in this study. Table 1 lists the distribution of gender and MMSE scores for each age group.

Table 1.

Demographics

Age groups	Young	Middle	Older
	20–39	40–54	55 and 80
	Mean (SD)	Mean (SD)	Mean (SD)
N	54	42	78
Female, %	50.0	69.0	56.4
Age	28.7 (4.67)	47.4 (4.39)	64.0 (6.29)
MMSE scores	28.9 (1.28)	28.0 (1.48)	27.8 (1.74)

COMT Genotype Distribution and Haplotype Frequency

The genotype frequencies of individual COMT SNPs and the estimated frequencies of their haplotypes are listed in Table 2. Based on the genotype frequency of the individual SNPs, certain genotypes were grouped together for the purpose of statistical analysis with the brain data.

Table 2.

Genotype frequencies and grouping of each SNPs as well as their haplotype frequency

SNPs	Genotype frequencies (n)
	AA	AG	GG	Grouping
COMT val158met	13	66	95	‐met carriers (AA + AG)
				‐val homozygotes (GG)
COMT rs737865	80	80	14	‐G allele carriers(AG + GG)
				‐A homozygotes (AA)
COMT rs165599	49	86	39	‐AA
				‐AG
				‐GG
	ε2ε4	ε3ε2	ε3ε3	ε3ε4
APOE	5	16	125	28
Haplotype (rs737865‐ val158met‐rs165599)		Estimated Haplotype Frequency mean (SD) %
A‐met‐A		21 (0.94)
A‐val‐A		10 (0.90)
A‐val‐G		37 (1.09)
G‐met‐A		5 (0.92)
G‐val‐A		17 (0.87)
G‐val‐G		9 (1.06)

Interaction Effects of Age With Individual COMT SNPs on Cortical Thickness and Subcortical Shapes

Cortical thickness

Regression analysis revealed significant interaction effects of age with COMT val158met and COMT rs737865 but not with COMT rs165599 on cortical thickness.

Age by COMT val158met interaction effects on cortical thickness were observed predominantly in the left dorsolateral prefrontal cortex (dlPFC), left angular gyrus (AG), right superior frontal cortex (SFC) and right AG (Fig. 1A and Table 3 column 2). Further analysis revealed that age was negatively associated with cortical thickness in the left dlPFC, right SFC and bilateral AG in met carriers (Fig. 1B). In Table 3, val homozygotes also showed that age was negatively associated with cortical thickness in the right AG but not in the left dlPFC, left AG, right SFC. Additional slope analysis showed a faster decline in age‐related cortical thinning of the left dlPFC, right SFC and bilateral AG in met carriers as compared to val homozygotes (Table 3 column 4).

Figure 1.

Panels (A) and (D) show the statistical maps of interaction between age and COMT SNPs (val158met, rs737865) on cortical thickness. Scatter plots on panels (B) and (E) illustrate the relationship of age and cortical thickness within each genotype of COMT val158met and COMT rs737865, respectively. Boxplots on panels (C) and (F) illustrate the cortical thickness within genotypes of COMT val158met and rs737865 in each age groups, respectively. Abbreviation: *P < 0.05. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Table 3.

Statistical results

	Age*COMT				Association of age and brain regions	Slope analysis	Post hoc among genotype groups
SNP	val158met		rs 737865
Brain regions	Cluster size	P‐valuea	Cluster size	P‐valuea	β	β	P‐value
					P‐value	P‐value
Left dlPFC	2,155	0.001	–	–	met: −0.628 (0.001)	−1.092 (0.001)	Y: met > val 0.001
					val: 0.030 (0.774)		O: val >met 0.002
Left AG	1,545	0.013	–	–	met: −0.723 (0.001)	−1.101 (0.001)	Y: met > val 0.012
					val: −0.169 (0.081)		M: met > val 0.02
							O: met < val <0.001
Right SFG	1,392	0.008	–	–	met: −0.637 (0.001)	−0.943 (0.001)	Y: met > val 0.007
					val: −0.131 (0.208)		O: met < val 0.001
Right AG	1,544	0.0126	–	–	met: −0.727 (0.001)	−0.994 (0.001)	Y: met > val 0.008
					val: −0.211 (0.031)		O: met < val <0.001
Right CAL	–	–	1,403		G: −0.635 (0.001)	−1.17 (0.001)	Y: G allele>A
				0.046	A: −0.081 (0.489)		P < 0.001
							O: G allele<A 0.025
Left anterior putamen	811	0.025	–	–	met: −0.557 (0.001)	−0.617 (0.004)	Y: met > val 0.019
					val: −0.259 (0.004)
Right anterior putamen	1,074	0.015	–	–	met: −0.559 (0.001)	−0.686 (0.001)	Y: met > val 0.010
					val: −0.227 (0.011)
Right anterior caudate nucleus	459	0.030	–	–	met: −0.520 (0.001)	−0.505 (0.031)	Y: met > val 0.032
					val: −0.257 (0.008)

The columns respectively list brain regions with significant interaction effects of age with COMT SNPs, clusters size and P‐values, β and P‐values for the associations of age and brain regions among genotype groups and the slope analysis, P‐values for post hoc analysis on genotypes in individual age groups.

Note. met = met carriers; val = val homozygotes; G = G allele carriers; A= A homozygotes. Y = Young age adults; M = Middle age adults; O = Old age adults.

Abbreviations. dlPFC—dorsolateral prefrontal cortex; AG—angular gyrus; SFC—superior frontal cortex; CAL—calcarine sulcus.

^aCorrected cluster P‐value.

A significant interaction of age by COMT rs737865 was observed in the right calcarine sulcus (CAL) (Fig. 1D and Table 3 column 2). Subsequent analysis revealed that cortical thickness in the right CAL was negatively associated with age in the G‐allele carriers of COMT rs737865 but not A‐homozygotes. Moreover, G‐allele carriers showed a faster decline in age‐related cortical thinning in the right CAL as compared to A homozygotes (Fig. 1E and Table 3 column 3 and 4).

Subcortical volumes and shapes

Interaction effects of age with COMT val158met on the subcortical volumes were observed in the same brain regions as drawn using subcortical shape analysis. We thus reported findings on the subcortical volumes in Supporting Information Table S1 of the Supplementary. The results of the shape analysis were described below as it provides detailed anatomical information.

Age by COMT val158met interaction was observed specifically in the bilateral anterior putamen and the right anterior caudate nucleus shapes (Fig. 2A and Table 3 column 2). Further analysis showed that both met carriers and val homozygotes had significant compression in the bilateral anterior putamen as well as right anterior caudate nucleus as age increases (Fig. 2B). The slope analysis revealed that met carriers showed faster compression in all the three subcortical regions as age increases when compared to val homozygotes (Table 3 column 3 and 4).

Figure 2.

Panels (A) shows the statistical maps of interaction between age and COMT SNPs (val158met, rs737865) on subcortical shapes. The scatter plots on panel (B) illustrate the relationship of age and subcortical shapes within each genotype of COMT val4680met. The boxplot on panel (C) illustrates the subcortical shapes between genotypes of COMT val4680met in each age group. Abbreviation: *P < 0.05. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

The comparisons of the genotype groups in each age group shown in Figure 2C revealed that among the young adults, met carriers had greater shapes expansion in the bilateral anterior putamen and the right caudate nucleus as compared to val homozygotes (Table 3 column 5). Among the middle and old groups, met carriers and val homozygotes had comparable bilateral anterior putamen and right anterior caudate nucleus shapes (P > 0.05).

No significant interaction of age with COMT rs165599 as well as COMT rs737865 was found on the subcortical shapes.

Interaction Effects of Age with COMT Haplotype on Cortical Thickness and Subcortical Shapes

Cortical thickness

The haplotype regression analysis adjusted for gender revealed that association between age and cortical thickness were modulated by the rs737865‐ val158met‐rs165599 haplotype. The probability of G‐met‐A haplotype modulated the negative relationship of age and left cortical thickness of the SFC, orbitofrontal cortex (OFC), dlPFC, middle temporal gyrus (MTG), posterior cingulate cortex (PCC) and lingual as well as the right cortical thickness of the SFC, dlPFC, supramarginal gyrus (SMG), AG, paracentral, precuneus (PCUN), occipital (OCC) and CAL (Fig. 3A and Table 4). Similarly, the probability of G‐val‐A haplotype modulated the negative relationship of age and the left inferior temporal cortex (ITC) (Fig. 3B).

Figure 3.

Panels (A) and (B) illustrate statistical maps for the relationship between cortical thickness and age modulated by G‐met‐A haplotype probability and G‐val‐A haplotype probability. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Table 4.

Clusters size and P‐values for significant interaction between age and COMT haplotype

Brain regions	Cluster size (number of vertices)	Corrected cluster P‐value
G‐met‐A
Right SFC	2,409	<0.0001
Right PC	2,264	<0.0001
Right PCUN	1,630	0.00561
Right CAL	1,370	0.0223
Right OCC	1,480	0.0154
Right AG	1,905	0.00318
Right SMG	1,237	0.0352
Right dlPFC	1,232	0.0280
Left dlPFC	2,980	<0.0001
Left PCC and LING	3,585	<0.0001
Left MTC	1,795	<0.0001
Left SFC	1,387	0.00562
Left OFC	1,030	0.0371
Left amygdala	429	0.00971
A‐met‐A
Right anterior putamen	981	0.0177
Left anterior putamen	782	0.0274
G‐val‐A
Right ITC	1,929	0.00695
A‐val‐G
Right anterior putamen	909	0.0178
Right caudate nucleus	270	0.0411

Abbreviations. dlPFC—dorsolateral prefrontal cortex; AG—angular gyrus; SFC—superior frontal cortex; PC—paracentral; PCUN—precuneus; CAL—calcarine sulcus; OCC—occipital; SMG—supramarginal gyrus; PCC—posterior cingulate cortex; LING—lingual; MTC—middle temporal cortex; OFC—orbitofrontal; ITC—inferior temporal cortex.

Subcortical volumes and shapes

Interaction effects of age with A‐met‐A probabilities on the subcortical volumes were observed in the same brain regions as drawn using subcortical shape analysis. We thus reported findings on the subcortical volumes in Supporting Information Table S1.

In subcortical shapes, the probability of A‐met‐A modulated the negative relationship of age and the bilateral putamen, specifically their anterior regions (Fig. 4A and Table 4). Moreover, the probability of G‐met‐A modulated the negative relationship of age and the left amygdala shape (Fig. 4B and Table 4). Lastly, the negative association of age and right anterior caudate nucleus and right anterior putamen shapes was modulate by the probability of A‐val‐G (Fig. 4C and Table 4).

Figure 4.

Panels (A, B and C) illustrate statistical maps for the relationship between subcortical shape and age modulated by A‐met‐A haplotype probability, G‐met‐A haplotype probability and A‐val‐G haplotype probability. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

APOE Influence

Given that Apolipoprotein E4 (APOE‐ε4) allele is one of the prominent risk allele in aging [Brown et al., 2011; Honea et al., 2009b; Parasuraman et al., 2002; Plassman et al., 1997], additional analysis was carried out with APOE as covariate in the aforementioned analysis. Interaction effects of age with individual SNPs on the cortical thickness, subcortical volumes and shapes largely remained the same even after controlling for the APOE (Supporting Information Tables S2‐3 and Supporting Information Figure S1–S2). Moreover, significant interaction between age and COMT haplotype on the cortical and subcortical morphology remained the same after adjusting for the APOE (Supporting Information Tables S2‐3 and Supporting Information Figs. S3–S4).

DISCUSSION

This study examined whether individual COMT SNPs (rs737865, val158met, rs165599) and its haplotype modulate age effects on subcortical shapes and cortical thickness using a large Chinese adult sample aged 21 years and above. Our results showed an age‐related decline in cortical thickness of the dorsal pathway and age‐related shape compression in the basal ganglia as a function of the genotypes of the individual COMT SNPs. The haplotype trend regression analysis further revealed an age‐related decline in cortical thickness in the default mode network (DMN), including the PCC, PCUN, SMG and paracentral, and the ventral visual system, including the OCC and left ITC, as a function of the COMT haplotype. We further observed an antagonistic pleiotropic effect in COMT, suggesting that genetically programmed neural benefits in early life may have a potential bearing towards neural susceptibility in later life.

In line with the role of COMT in regulating synaptic dopamine levels [Mier et al., 2010], our findings revealed that individual COMT SNPs modulates the association of age and the morphology of the dopamine‐enriched brain regions, including those in the dorsal pathway and the basal ganglia. We showed that met carriers had greater age‐related neurodegeneration in these brain regions compared to val homozygotes. This is inconsistent with previous findings based on Caucasians [Meyer‐Lindenberg et al., 2006], that is, met allele of COMT val158met was suggested as a “smarter” allele principally due to its positive effects on both intermediate and cognitive endophenotypes when compared to val allele. Several studies based on Chinese samples have found better performances over a range of cognitive processes such as attention, working memory, mental state of intention in val allele carriers [Qian et al., 2009; Wang et al., 2013; Xia et al., 2012; Zhang et al., 2007]. Moreover, healthy adults with val alleles showed a greater hippocampal volume as compared to val homozygotes [Wang et al., 2013]. However, the examination of COMT influence on the brain morphology has thur far been limited to hippocampus volume or prefrontal function in healthy young adults [Meyer‐Lindenberg et al., 2006; Wang et al., 2013]. Our study provided new evidences of the COMT effects on age‐related associations with both the cortical and subcortical brain morphology in a Chinese sample. Interestingly, our findings showed an antagonistic pleiotropy phenomenon, the concept originally from evolutionary biology. It suggests that certain genes may impact fitness differently during different stages of life. It has been substantially supported by the findings of APOE on behavioural and neural endophenotypes across lifespan [Borghesani et al., 2008; Dennis et al., 2010a; Filbey et al., 2010; Hayden et al., 2009; Jochemsen et al., 2012; Lind et al., 2006; Mondadori et al., 2007; Xu et al., 2009]. Young adults with APOE ε4 showed better cognitive performances than their non‐carriers counterparts. Moreover, young adults with APOE ε4 also showed more bilaterally activity in the medial temporal lobe, medial frontal cortex and hippocampus [Dennis et al., 2010a; Filbey et al., 2010; Filbey et al., 2006; Jochemsen et al., 2012; Mondadori et al., 2007]. On the other hand, old adults with APOE ε4 showed poorer cognitive performance and less activity in the hippocampus and frontal cortex than non‐carriers of APOE ε4 [Borghesani et al., 2008; Hayden et al., 2009; Lind et al., 2006; Xu et al., 2009]. We provided new evidence of the COMT antagonistic pleiotropy phenomenon on the brain morphology across lifespan. Our results suggested a genetically programmed COMT‐related neural benefits, such as thicker cortex in young adults and a potential bearing towards neural susceptibility in older adults. Moreover, the observed antagonistic pleiotropy phenomenon appears to be more apparent in the cortical regions than the basal ganglia regions.

Most notably, haplotypes analysis, compared to analysis on individual SNPs, showed improved statistical power in detecting effects of COMT genetic variants on the association between age and brain morphology. The haplotypes analysis provided the findings consistent with those derived from the individual SNPs but considerably stronger in terms of the cluster size (see Tables 3 and 4). Moreover, in addition to the dorsal visual system detected from COMT val158met, the haplotype analysis also uncovered the age X COMT haplotype effects on cortical thickness in the brain regions part of the DMN and the ventral visual system.

The dorsal and ventral systems are involved in working memory [Ungerleider et al., 1998]. Previous studies suggested that both systems are vulnerable to age [Hof and Mobbs, 2001]. The ventral visual system comprises of the visual cortex that projects in the ITC, which eventually terminates in the ventral prefrontal cortex [Murray and Wallace, 2011]. It is critical for providing information on object's representation. Older adults, as compared to young adults, showed decreased object recognition accuracy [Nobre and Kastner, 2014]. Additionally, during route encoding task, older subjects also showed reduced activity in the fusiform as compared to young adults. The dorsal pathway, on the other hand, comprises of the visual cortex that projects in the posterior parietal cortex, which eventually terminates in the dlPFC. The pathway has been found to be critical for processing spatial information on objects [Murray and Wallace, 2011]. Deficit in encoding spatial information was shown in older adults [Meulenbroek et al., 2004; Schiavetto et al., 2002]. Decreased involvement of the inferior parietal cortex was in relation to deficit in object location processing in older adults when compared to younger adults. Beyond age effects on both systems, young adults showed that dlPFC‐related working memory performance [Honea et al., 2009a; Mier et al., 2010; Nyberg et al., 2014; Roffman et al., 2008] and functional activation to working memory in the inferior parietal cortex [Gibbs and D'Esposito, 2005] are as a function of COMT val158met. While dopamine is well expressed in both brain regions, one expects that the COMT manipulation of the functions of these brain regions may be via its regulation of dopamine availability. Our study employed haplotype analysis and further uncovered that genetic variants of COMT modulate the association of age and brain morphology in both dorsal and visual systems, which may suggest a crucial molecular mechanism via dopamine system in aging.

Likewise, the DMN was identified as major brain regions influenced by the interaction of age and COMT haptotype. The DMN is primarily associated with monitoring of self‐related mental processes such as recollection of past experiences, imagining future‐related events known as episodic memory as well as working memory [Addis et al., 2007; Buckner and Carroll, 2007]. It is also highly implicated in aging [Bluhm et al., 2008; Damoiseaux et al., 2008; Esposito et al., 2008]. Previous studies have found that decreased connectivity strength between the PCC and mPFC confers a detrimental impact on episodic memory, working memory and executive functions in the elderly [Damoiseaux et al., 2008; Mevel et al., 2013; Sambataro et al., 2010; Spreng and Grady, 2010]. Moreover, the pivotal structures in the DMN, such as the precuneus and PCC, were highly susceptible to the effect of dopamine during working memory task in young healthy adults [Braskie et al., 2008; Stokes et al., 2011]. Although there is evidence for a role of COMT in regulating age‐related episodic memory, working memory [Lisman and Grace, 2005; Nagel et al., 2008], putative impacts of the COMT genetic variants on intermediate endophenotype have thus far been limited to the examination using resting state fMRI. The examination of COMT influence on the functional connectivity in healthy young adults, however, is relatively unclear. While there was a study which observed a stronger functional connectivity between PCC and prefrontal regions in the val homozygous as compared to their met allele carriers counterparts [Jang et al., 2012], several others found no or decreased connectivity between prefrontal regions and the PCC in the val homozygous [Liu et al., 2010; Tian et al., 2013; Tunbridge et al., 2013]. The inconsistency might be due to the use of single SNP for analysis and samples of different age groups. This warrants more investigation of COMT effect on the DMN. To our best of knowledge, none has looked at the influence of the COMT haplotype on the structural basis of aging brain. Our study provided new evidence of a modulative role of the COMT gene on age‐related associations with structural brain regions involved in the DMN. Overall, the study of COMT haplotype, which characterize the linkage disequilibrium structure of several SNP markers, showed stronger association with phenotypic regions that are highly involved in aging and under the influence of dopamine availability than individual markers. Furthermore, it is also superior in identifying potential complex interactions of variants in COMT on cortical and subcortical regions that may not be easily inferred from single SNP analysis alone.

Our study employed the advanced brain mapping technique, LDDMM, and examined the subcortical shapes as a function of interaction between COMT and age. Using shape analysis, we were able to provide additional information with respect to the detailed anatomical information, which is beyond volumetric analysis. Our study highlighted that the shape variations of the anterior segment of the caudate nucleus and putamen can be explained by the interaction of age with COMT val158met and its haplotype. Previous study has shown that the anterior caudate nucleus played a role in goal‐directed decision‐making by integrating performance information and demand of cognitive control during instrumental learning. The anterior putamen plays a role in tracking conditions that would lead to correct responses [Brovelli et al., 2011]. Interestingly, we also observed a role of COMT haplotype on the shapes of amygdala in our sample. The amygdala was also highly susceptible to age and an age–related decrease of amygdala volume was associated with poor socioemotional memory score in the old adults [Cassidy and Gutchess, 2012]. Moreover, the dopamine D1 and D2 are expressed in amygdala in addition to numerous innervation of dopaminergic fibers [Asan, 1998; Ito et al., 2008]. A previous study had further showed a positive correlation between dopamine release and amygdala signal change in response to affective stimuli [Takahashi et al., 2010], suggesting an influential effect of dopamine on amygdala function. Thus by examining subcortical shapes, specific regions and their function contributing to the molecular basis of aging can be better defined and understood.

The strength of this study includes a unique opportunity to examine individual SNPs as well as its haplotype of COMT on brain morphology, including cortical thickness, subcortical volumes and shapes, in adults aged across the lifespan. Using estimated haplotype probability in statistical analysis on imaging data increases statistical power and generates results unbiased to samples with uniquely determined haplotypes. Perhaps one major limitation of our study, and one common to most studies of imaging genetics, is that our current lack of knowledge of the relation between the genetic variants and region‐specific variation in COMT gene expression and dopamine signaling and also their temporal patterns. Nevertheless, our findings are consistent with the apparent increased level of dopamine signaling in the prefrontal cortex and basal ganglia [Braskie et al., 2008]. In addition, our study did not examine gene and gene interaction on the brain morphology in aging, which is most likely present. Furthermore, due to the limited sample in this study, certain genotypes were grouped based on their frequency for the purpose of statistical analysis with the brain data. This may reduce the opportunity to explore effects of individual genotypes on aging brain. Finally, our statistical models were limited to linear, which may not be true in reality. Hence, we also presented the post‐hoc analysis where age was considered as a discrete variable. This partially solved the problem of the linearity assumption but the cutoff of the age intervals may drive the results.

In conclusion our results showed that genetic variants of COMT modulate the association of age and brain morphology of the dorsal visual pathway and basal ganglia that are dopaminergically rich. Additionally, COMT shows an antagonistic pleiotropy phenomenon that confers a differential impact of COMT on an aging brain. We further highlighted the advantages of relating probabilistic haplotype to brain morphology where results further revealed that both ventral visual pathway and DMN were also modulated by COMT. Studying the COMT genetic variations with aging alterations of brain anatomy provides a better understanding of the molecular mechanism in the aging brain morphology.

Supporting information

Supporting Information