Genetic and Environmental Influences of White and Gray Matter Signal Contrast: A New Phenotype for Imaging Genetics?

Abstract

The estimation of cortical thickness is in part dependent on the degree of contrast in T1 signal intensity between white matter and gray matter along the cortical mantle. The ratio of white matter to gray matter signal (WM/GM contrast) has been found to vary as a function of age and Alzheimer’s disease status, suggesting a biological component to what might otherwise be labeled as a nuisance variable. The aim of the present study was to determine if measures of WM/GM contrast are genetically influenced, as well as the degree to which this phenotype may be related to the genetic and environmental determinants of cortical thickness. Participants were 514 male twins (130 monozygotic, 97 dizygotic pairs, and 60 unpaired individuals) from the Vietnam Era Twin Study of Aging. Ages ranged from 51 to 59 years. Measures of WM/GM contrast and cortical thickness were derived for 66 cortical regions of interest (ROI) using FreeSurfer-based methods. Univariate and bivariate twin analyses were used in order to estimate the heritability of WM/GM contrast, as well as the degree of shared genetic and environmental variance between WM/GM contrast and cortical thickness. WM/GM contrast was found to be significantly heritable in the majority of ROIs. The average heritability across individual ROIs was highest in the occipital lobe (.50), and lowest in the cingulate cortex (.24). Significant phenotypic correlations between WM/GM contrast and cortical thickness were observed for most of the ROIs. The majority of the phenotypic correlations were negative, ranging from −.11 to −.54. Of the 66 associations, only 17 significant genetic correlations were found, ranging from −.16 to −.34, indicating small amounts of shared genetic variance. The majority of the phenotypic correlations were accounted for by small unique environmental effects common between WM/GM contrast and cortical thickness. These findings demonstrate that like cortical thickness, WM/GM contrast is a genetically influenced brain structure phenotype. The lack of significant genetic correlations with cortical thickness suggests that this measure potentially represents a unique source of genetic variance, one that has yet to be explored by the field of imaging genetics.

The estimation of cortical thickness by manual or automated methods is dependent, to a large extent, on the degree of contrast in signal intensity between white matter and gray matter along the cortical mantle (Fischl and Dale, 2000; Narr et al., 2005). Simply put, the greater the difference in signal intensities between these tissue types, resulting in a steeper signal gradient, the more precisely the gray-white boundary can be defined. This in turn leads to a more accurate determination of cortical thickness, particularly when done in three-dimensional space. Indeed, statistical adjustment for white matter to gray matter (WM/GM) signal contrast was recently shown to increase the effect size of cortical thickness differences related to Alzheimer’s disease (AD) and improve the power to detect differences in cortical thickness due to normal aging (Westlye et al., 2009). While variability in magnetic field strength, pulse sequence, and data processing parameters can affect signal intensities in the white and gray matter, and thus influence the degree of WM/GM contrast (Han et al., 2006), there is growing evidence to suggest that this measure may also reflect intrinsic properties of the corresponding tissue.

Numerous studies have found that the ratio of white matter to gray matter signal varies as a function of increasing age, such that the degree of contrast declines later in life (Davatzikos and Resnick, 2002; Magnaldi et al., 1993; Raz et al., 1990; Salat et al., 2009; Westlye et al., 2009). Intriguingly, these contrast differences do not appear to be uniform throughout the cortex, but are predominant in the frontal, temporal, and parietal regions – areas that also demonstrate significant age-related changes in cortical thickness (Salat et al., 2009; Westlye et al., 2009). Regionally specific differences in WM/GM contrast have also been found when comparing AD to normal aging samples, suggesting that the measure may also be sensitive to AD-related neuropathology (Salat et al., 2011). The observed differences in signal contrast appear to be driven primarily by a gradual reduction in the white matter signal intensity, leading to speculation that the degree of WM/GM contrast is indicative of the integrity of the myelinated fibers that are present along the gray-white boundary (Salat et al., 2011; Salat et al., 2009; Westlye et al., 2010; Westlye et al., 2009). The presence of such age- and disease-related effects strongly suggests that there is a biological component to what might otherwise be thought of as a technical nuisance variable resulting from hardware and acquisition parameters. This conclusion is speculative, however, as it remains unclear what precise mechanism is behind the observed changes in WM/GM contrast, or whether these effects stem from the same processes responsible for age-related changes in cortical thickness.

In the present study, we examined region-of-interest (ROI) based measures of WM/GM contrast in a sample of middle-aged male twins. Utilizing the classical twin design we first determined whether WM/GM contrast was itself a heritable phenotype; that is, are individual differences in the contrast measure partially attributable to genetic factors. Reviews of early twin and family MRI studies clearly demonstrate that structural aspects of the brain (e.g., whole brain volume, gray matter volume) are under substantial genetic influence (Peper et al., 2007; Schmitt et al., 2007a). More recent studies have expanded the range of structural phenotypes to include ROI and vertex-based measures of cortical thickness (Kremen et al., 2010; Lenroot et al., 2007; Rimol et al., 2009), cortical surface area (Eyler et al., In Press; Panizzon et al., 2009; Winkler et al., 2010), as well as microstructural features of the brain’s white matter obtained through diffusion tensor imaging (DTI) (Brouwer et al., 2010; Chiang et al., 2009; Chiang et al., 2011; Kochunov et al., 2010). These studies have all found that structural brain phenotypes are under significant genetic influence. If the degree of WM/GM contrast along the cortical mantle is indicative of underlying genetic influences on tissue properties, then this measure, like the other brain phenotypes studied to date, should to some extent be heritable.

While the determination of whether or not a phenotype is heritable represents a critical step in establishing both its biological relevance and potential usefulness for future gene association studies, it is equally important to establish whether the observed genetic influences differ from those of other related phenotypes. Therefore, we also examined the degree to which measures of WM/GM contrast and cortical thickness possess common genetic and environmental influences. To date, relatively few genetically informative neuroimaging studies have examined the genetic and environmental relationships between brain phenotypes; however, those that have done so found evidence for multiple distinct sources of genetic influence (Eyler et al., 2011; Panizzon et al., 2009; Rimol et al., 2009; Schmitt et al., 2008; Schmitt et al., 2007b; Winkler et al., 2010). If WM/GM contrast and cortical thickness represent similar neuroanatomical features of the brain then a substantial degree of genetic overlap (i.e., shared genetic variance) should be present between them. Alternatively, the absence of genetic overlap would suggest that the phenotypes are biologically distinct from one another.

Methods

Participants

Data were obtained from participants in the first wave of the Vietnam Era Twin Study of Aging (VETSA), a longitudinal study of cognitive and brain aging with baseline in midlife (Kremen et al., 2006). Participants in the VETSA were drawn from the larger Vietnam Era Twin (VET) Registry, a nationally distributed sample of male-male twin pairs who served in the United States military at some point between 1965 and 1975 (Goldberg et al., 2002). Detailed descriptions of the VET Registry’s method of ascertainment and demographic characteristics have been reported on previously (Eisen et al., 1987; Henderson et al., 1990). VETSA participants are all military veterans; however, the majority did not experience combat situations during their military careers. In total, 1237 men participated in the primary VETSA project. Participants were predominantly Caucasian (89.7%), with an average age of 55.4 years (SD = 2.5), and an average education of 13.8 years (SD = 2.1). In comparison to U.S. census data, participants in the VETSA are similar in health and demographic characteristics to American men in their age range (Centers for Disease Control and Prevention, 2003). As part of the primary VETSA project, participants traveled to either the University of California San Diego (UCSD) or Boston University for a daylong evaluation consisting of physical, psychosocial, and neurocognitive assessments. Beginning in year 3 of the VETSA, MRIs were conducted on either the day before or the day following these assessments at one of two scanning sites – the UCSD Medical Center or the Massachusetts General Hospital (MGH). Informed consent was obtained from all participants prior to their participation, and scanning protocols were approved by the institutional review boards at all participating universities and hospitals.

To be eligible for the primary VETSA project both members of a twin pair had to agree to participate and be between the ages of 51 and 59 at the time of recruitment. Approximately 6% of the individuals invited to participate in the MRI component of the VETSA declined. Additional participants were excluded from the MRI study for reasons such as metal in the body (7%), claustrophobia (3%), being unable to travel to the testing site (5%), the exclusion of their co-twin (9%), and equipment problems on the scanning day (8%). In the end, approximately 59% of the invited individuals participated in the MRI study. Analyses for the present study were based on data from 514 participants: 130 monozygotic (MZ) pairs, 97 dizygotic (DZ) pairs, and 60 unpaired individuals (i.e., participants whose co-twin could either not be scanned or whose data was not usable). Zygosity for 92% of the sample was determined by analysis of 25 satellite markers that were obtained from blood samples. For the remainder of the sample zygosity was determined through a combination of questionnaire and blood group methods (Eisen et al., 1989). Within the VETSA sample, a comparison of these two approaches has demonstrated a 95% agreement rate. Participants in the VETSA MRI study are similar to the larger VETSA sample with respect to age, education, ethnicity, employment status, and self-reported health status (Panizzon et al., 2009; Kremen et al., 2010).

Image Acquisition

We have previously described in detail the acquisition parameters and post-processing methods for the VETSA MRI study (Kremen et al., 2010). Briefly, images were acquired on Siemens 1.5 Tesla scanners. Scanning sequences were specifically designed to be compatible across different scanners and vendors. Sagittal T1-weighted MPRAGE sequences were utilized with a time to inversion of 1000ms, a time echo of 3.31ms, a time repetition of 2730ms, a flip angle equal to 7 degrees, a slice thickness of 1.33mm, and a voxel size of 1.3×1.0×1.3mm. Raw DICOM MRI files from both sites were downloaded to facilities at MGH for post-processing and quality control. Of the 530 scans available, quality control procedures excluded 16 due to either scanner artifact or technical errors in image processing.

Image Processing

As detailed in our previous work (Kremen et al., 2010), cortical surface reconstruction was performed using methods based on the publicly available FreeSurfer software package (Dale et al., 1999; Fischl and Dale, 2000; Fischl et al., 2004a; Fischl et al., 1999). Briefly, the explicit reconstruction of the cortical surface involves a number of subtasks, including correction of field inhomogeneities, creation of a normalized intensity image, and removal of non-brain tissue. The resulting surface is covered with a polygonal tessellation and smoothed to reduce metric distortions. The gray/white boundary is then locally defined as the point of the steepest change along the intensity gradient, thereby enabling cortical thickness to be estimated while allowing for variability in the degree of contrast between the white matter and gray matter. This surface is subsequently deformed outwards to obtain an explicit representation of the pial surface.

The surface was then divided into distinct cortical regions of interest, and each vertex was assigned a neuroanatomical label based on 1) the probability of each label at each location in a surface-based atlas space, 2) local curvature information, and 3) other contextual information (e.g., encoding spatial neighborhood relationships between labels) (Desikan et al., 2006; Fischl et al., 2004b). Similar to the methods described by Westlye and colleagues (Westlye et al., 2009), estimates of white matter and gray matter signal intensity were obtained at a distance of 0.2 millimeters from the gray-white boundary. The cortical surface was divided into 66 ROIs (33 per hemisphere) according the Desikan et al parcellation scheme (2006), and averages of the respective vertex values were derived for each region. Measures of WM/GM contrast were calculated by dividing each ROI’s average white matter intensity value by the corresponding average gray matter intensity value.

Statistical Analysis

In order to determine the relative influence of genetic and environmental factors on WM/GM contrast, we fit univariate biometrical models (also referred to as ACE models) to the data from each of the 66 ROIs. In the classical twin design the variance of a phenotype is decomposed into the proportion attributed to additive genetic (A) influences, common or shared environmental (C) influences (i.e., environmental factors that make members of a twin pair similar to one another), and unique environmental (E) influences (i.e., environmental factors that make members of a twin pair different from one another, including measurement error) (Eaves et al., 1978; Neale and Cardon, 1992). Additive genetic influences are assumed to correlate perfectly (1.0) between monogygotic (MZ) twins because they are genetically identical. Dizygotic (DZ) twins, on the other hand, share on average 50% of their segregating DNA, and are therefore assumed to correlate .50 for additive genetic influences. The shared environment is assumed to correlate 1.0 between both members of a twin pair, regardless of their zygosity. Unique environmental influences, by definition, are uncorrelated between the members of a twin pair. The proportion of the overall variance in a phenotype that is attributable to additive genetic influences is the heritability.

In addition we fit bivariate Cholesky decomposition models for each regional WM/GM contrast estimate and its corresponding cortical thickness estimate in order to determine if the two phenotypes share common genetic and environmental determinants. The “Cholesky” model decomposes the total covariance between phenotypes into genetic and environmental components; thus, the sum of the standardized genetic and environmental covariances is equal to the phenotypic correlation. The genetic and environmental covariance estimates can also be used to calculate genetic and environment correlations. In statistical terms, the genetic correlation between two phenotypes is equal to their genetic covariance, divided by the square root of the product of their separate genetic variances (Neale and Cardon, 1992). Shared environmental and unique environmental correlations are calculated in a similar fashion using the corresponding variance and covariance estimates. Conceptually, genetic correlations represent the degree to which genetic influences of one phenotype are predictive of the genetic influences for another phenotype (Carey, 1988); the analogous definition holds for environmental correlations. Because genetic and environmental correlations can have opposing signs, resulting in phenotypic correlations that are near zero, analyses were run for all ROIs regardless of the observed phenotypic correlation between WM/GM Contrast and cortical thickness.

All analyses were performed using the raw data application of the maximum-likelihood based structural equation modeling software OpenMx (Boker et al., 2011). Measures of WM/GM contrast and cortical thickness were adjusted for the relative effects of age and scanner as part of the model fitting process. The inclusion of scanner in the model is critical to adjust for the influence of site-specific scanner hardware differences on measures of signal intensity and cortical thickness (Han et al., 2006). In the univariate analyses, the significance of the genetic and shared environment influences was tested by fixing the parameter in question to zero, and then comparing the resulting change in fit of the reduced model against that of the full model. No such test was performed for the unique environmental influences because this parameter must always be present in the model. Model comparisons were performed using the likelihood-ratio chi-square test (LRT), which was calculated as the difference in the −2 log likelihood (−2LL) of the reduced model from that of the full model. Non-significant LRT values (p > .05) indicate that a reduced model does not result in a significant change in fit relative to the comparison model, and thus provides a significance test for the parameter is question. Under certain regularity conditions, the LRT is distributed as a chi-square (χ²) with degrees of freedom (df) equal to the difference in the number of parameters between the two models (Steiger et al., 1985). However, because there is an implicit lower bound of zero for variance components, the distribution of the test statistics for the A and C parameters is distributed as a 50:50 mixture of zero and χ² with df=1 (Dominicus et al., 2006; Self and Liang, 1987). Failure to account for this mixed distribution produces p-values that are too large; however the issue is easily corrected by halving the p-values obtained from the naïve χ² with df=1 distribution. When the A and C parameters are tested simultaneously, the resulting distribution is a mixture of zero, χ² with df=1, and χ² with df=2 (Dominicus et al., 2006). In this case a more reasonable p-value can be obtained by halving the p-value generated from a χ² with df=1 distribution (Dominicus et al., 2006). Similar to the p-values, 95% confidence intervals are also affected by the boundary condition. We therefore utilized an adjustment developed by two of the co-authors (HW & MCN), which allowed the confidence intervals to be calculated in a fashion consistent with the adjusted p-values (Wu and Neale, submitted). Compared to an unadjusted CI, an adjusted CI has a higher lower limit if the point estimate is close to but not on its boundary of zero and has a higher upper limit if the point estimate of zero is obtained. The lower limit of the adjusted CI is always greater than zero when the adjusted test gives a significant result. Phenotypic, genetic and environmental correlations in the bivariate Cholesky models were also tested using the LRT method; however, due to the fact that these parameters do not involve a lower bound of zero, no adjustment of the p-values is required. Similarly, the 95% confidence intervals for these parameters required no correction.

Results

WM/GM contrast values ranged from 1.081 in the right transverse temporal cortex to 1.147 in the left caudal anterior cingulate. On average, the cingulate cortex possessed the highest contrast estimates (1.127), while the occipital lobe possessed the lowest (1.096). Age had a minimal impact on the contrast values, with only 5 of the 66 ROIs demonstrating a significant effect (p < .05). Scanner, on the other hand, was found to have a significant impact on 60 of the 66 ROIs. Effect sizes (Cohen’s d) for significant relationships ranged from .20 for the left lateral occipital cortex to .76 for the right caudal middle frontal gyrus. Average WM/GM contrast estimates, as well as the relative effects of age and scanner for all of the ROIs examined are presented in the supplemental materials (see Supplemental Table).

Heritability of White Matter/Gray Matter Signal Contrast

MZ and DZ cross-twin correlations, genetic and environmental variance components, as well as tests of significance for specific univariate model parameters are presented in Table 1. Heritability estimates (a²) for the ROI-based measures of WM/GM contrast ranged from .00 for the right posterior cingulate and the left entorhinal cortex, to .66 for the left precentral gyrus. In total, 48 ROIs were significantly heritable in the full ACE model (i.e, additive genetic influences could be fixed at zero without a significant reduction in model fit). An additional 15 ROIs had heritability estimates that were substantially larger than zero, ranging from .17 to .36. These values reached statistical significance when the shared environmental influences were constrained to be zero, making it easier to detect significant genetic influences. Given that estimates of the shared environment were zero or near zero for the majority of ROIs, this constraint could be imposed without resulting in a significant change in model fit for all but one region. On average the heritability of individual ROIs was highest in the occipital lobe (average a² = .50), followed by the parietal (average a² = .47) and frontal lobes (average a² = .46), the temporal lobes (average a² = .30), and the cingulate cortex (average a² = .24). Constraining both the genetic and shared environment parameter estimates to zero; in other words, testing a model in which only unique environmental influences accounted for the variance in WM/GM contrast, resulted in a significant change in model fit for all ROIs examined. Thus, the presence of significant familial influences on WM/GM contrast – influences attributable to either latent genetic factors or shared environmental factors – could be verified for every region.

Table 1.

Univariate ACE models for WM/GM contrast estimates and tests of nested models

Region of Interest			Standardized Variance Components						p-values
	rmz	rdz	a2	95% CI	c2	95% CI	e2	95% CI	No A	No C	No AC
Frontal Lobe
Superior Frontal Gyrus-L	.62	.31	.58	(.24; .70)	.03	(.00; .38)	.40	(.30; .52)	.0020	.4451	<.0001
Superior Frontal Gyrus-R	.61	.28	.59	(.24; .69)	.00	(.00; .36)	.41	(.31; .53)	.0021	.5000	<.0001
Middle Frontal Gyrus
Rostral Division-L	.48	.14	.45	(.19; .57)	.00	(.00; .38)	.55	(.43; .69)	.0065	.5000	<.0001
Rostral Division-R	.51	.20	.48	(.13; .59)	.00	(.00; .38)	.52	(.41; .66)	.0128	.5000	<.0001
Caudal Division-L	.59	.23	.58	(.33; .68)	.00	(.00; .36)	.42	(.32; .54)	.0006	.5000	<.0001
Caudal Division-R	.61	.18	.59	(.36; .69)	.00	(.00; .35)	.41	(.31; .53)	.0001	.5000	<.0001
Inferior Frontal Gyrus
Pars opercularis-L	.57	.09	.52	(.32; .62)	.00	(.00; .38)	.48	(.38; .62)	.0005	.5000	<.0001
Pars opercularis-R	.43	.11	.39	(.11; .52)	.00	(.00; .37)	.61	(.48; .75)	.0177	.5000	<.0001
Pars triangularis-L	.50	.02	.44	(.26; .56)	.00	(.00; .39)	.56	(.44; .72)	.0011	.5000	<.0001
Pars triangularis-R	.44	.12	.40	(.08; .53)	.00	(.00; .38)	.60	(.47; .75)	.0240	.5000	<.0001
Pars orbitalis-L	.33	.20	.18	(.00; .45)	.13	(.00; .39)	.69	(.55; .84)	.2488	.2938	<.0001
Pars orbitalis-R	.43	.00	.38	(.19; .53)	.00	(.00; .37)	.62	(.47; .78)	.0043	.5000	<.0001
Orbitofrontal Cortex
Lateral division-L	.56	.17	.49	(.15; .61)	.00	(.00; .39)	.51	(.39; .63)	.0101	.5000	<.0001
Lateral division-R	.58	.05	.52	(.34; .64)	.00	(.00; .36)	.48	(.36; .61)	.0002	.5000	<.0001
Medial division-L	.45	.08	.41	(.16; .54)	.00	(.00; .38)	.59	(.46; .74)	.0097	.5000	<.0001
Medial division-R	.55	.20	.51	(.17; .62)	.00	(.00; .40)	.49	(.38; .62)	.0077	.5000	<.0001
Frontal pole-L	.30	−.08	.24	(.01; .38)	.00	(.00; .30)	.76	(.62; .89)	.0494	.5000	.0013
Frontal pole-R	.26	.09	.24	(.00; .39)	.00	(.00; .30)	.76	(.61; .90)	.1489	.5000	.0020
Precentral gyrus-L	.70	.21	.66	(.39; .74)	.00	(.00; .34)	.34	(.26; .45)	.0000	.5000	<.0001
Precentral gyrus-R	.66	.21	.65	(.42; .73)	.00	(.00; .34)	.35	(.27; .47)	.0000	.5000	<.0001
Paracentral lobule-L	.43	.10	.41	(.18; .54)	.00	(.00; .39)	.59	(.46; .75)	.0063	.5000	<.0001
Paracentral lobule-R	.51	.24	.47	(.09; .60)	.02	(.00; .39)	.51	(.40; .66)	.0214	.4652	<.0001
Cingulate Cortex
Rostral anterior cingulate-L	.34	.07	.32	(.04; .47)	.00	(.00; .35)	.68	(.53; .83)	.0378	.5000	.0001
Rostral anterior cingulate-R	.38	.09	.33	(.00; .46)	.00	(.00; .35)	.67	(.54; .82)	.1153	.5000	<.0001
Caudal anterior cingulate-L	.28	.08	.26	(.00; .40)	.00	(.00; .31)	.74	(.60; .88)	.0929	.5000	.0011
Caudal anterior cingulate-R	.18	.02	.19	(.00; .35)	.00	(.00; .28)	.81	(.65; .97)	.1216	.5000	.0234
Rostral posterior division-L	.26	.07	.23	(.00; .37)	.00	(.00; .30)	.77	(.63; .90)	.1955	.5000	.0018
Rostral posterior division-R	.34	.34	.00	(.00; .45)	.34	(.04; .45)	.66	(.54; .79)	.5000	.0298	<.0001
Retrosplenial cortex-L	.33	.08	.29	(.00; .42)	.00	(.00; .33)	.71	(.58; .85)	.1152	.5000	.0001
Retrosplenial cortex-R	.35	.14	.28	(.00; .45)	.03	(.00; .38)	.68	(.55; .84)	.1553	.4530	<.0001
Parietal Lobe
Postcentral gyrus-L	.63	.15	.60	(.37; .69)	.00	(.00; .34)	.40	(.31; .52)	.0001	.5000	<.0001
Postcentral gyrus-R	.55	.12	.50	(.26; .59)	.00	(.00; .37)	.50	(.41; .65)	.0020	.5000	<.0001
Supramarginal gyrus-L	.53	.15	.48	(.17; .60)	.00	(.00; .38)	.52	(.40; .65)	.0080	.5000	<.0001
Supramarginal gyrus-R	.47	.05	.41	(.19; .55)	.00	(.00; .39)	.59	(.45; .73)	.0055	.5000	<.0001
Superior parietal cortex-L	.59	.16	.55	(.32; .65)	.00	(.00; .37)	.45	(.35; .58)	.0007	.5000	<.0001
Superior parietal cortex-R	.59	.05	.54	(.35; .65)	.00	(.00; .37)	.46	(.35; .60)	.0001	.5000	<.0001
Inferior parietal cortex-L	.48	.02	.42	(.23; .55)	.00	(.00; .39)	.58	(.45; .73)	.0026	.5000	<.0001
Inferior parietal cortex-R	.50	.10	.44	(.18; .57)	.00	(.00; .39)	.56	(.43; .69)	.0075	.5000	<.0001
Precuneus-L	.48	.06	.42	(.19; .55)	.00	(.00; .39)	.58	(.45; .72)	.0062	.5000	<.0001
Precuneus-R	.48	.25	.36	(.00; .56)	.09	(.00; .45)	.55	(.44; .70)	.0681	.3401	<.0001
Occipital Lobe
Lingual gyrus-L	.54	.18	.50	(.21; .62)	.00	(.00; .37)	.50	(.38; .63)	.0049	.5000	<.0001
Lingual gyrus-R	.59	.29	.46	(.10; .65)	.09	(.00; .46)	.45	(.35; .58)	.0171	.3406	<.0001
Pericalcarine cortex-L	.54	−.03	.46	(.29; .59)	.00	(.00; .36)	.54	(.41; .67)	.0008	.5000	<.0001
Pericalcarine cortex-R	.43	.07	.38	(.12; .52)	.00	(.00; .36)	.62	(.48; .77)	.0153	.5000	<.0001
Cuneus-L	.57	.16	.55	(.33; .65)	.00	(.00; .36)	.45	(.35; .59)	.0004	.5000	<.0001
Cuneus-R	.57	.15	.54	(.29; .64)	.00	(.00; .36)	.46	(.36; .60)	.0011	.5000	<.0001
Lateral occipital cortex-L	.62	.18	.58	(.34; .67)	.00	(.00; .36)	.42	(.33; .55)	.0003	.5000	<.0001
Lateral occipital cortex-R	.58	.10	.53	(.32; .64)	.00	(.00; .37)	.47	(.36; .60)	.0007	.5000	<.0001
Temporal Lobe
Lateral Aspect
Superior temporal gyrus-L	.44	.05	.39	(.17; .52)	.00	(.00; .38)	.61	(.48; .77)	.0074	.5000	<.0001
Superior temporal gyrus-R	.37	.10	.34	(.03; .47)	.00	(.00; .35)	.66	(.53; .82)	.0406	.5000	<.0001
Middle temporal gyrus-L	.48	.09	.45	(.22; .56)	.00	(.00; .37)	.55	(.44; .71)	.0033	.5000	<.0001
Middle temporal gyrus-R	.39	.12	.36	(.01; .49)	.00	(.00; .37)	.64	(.51; .79)	.0437	.5000	<.0001
Inferior temporal gyrus-L	.45	.17	.42	(.07; .55)	.00	(.00; .39)	.58	(.45; .73)	.0269	.5000	<.0001
Inferior temporal gyrus-R	.41	.18	.15	(.00; .48)	.20	(.00; .45)	.65	(.52; .79)	.2940	.2424	<.0001
Transverse temporal cortex-L	.31	.08	.29	(.00; .45)	.00	(.00; .33)	.71	(.55; .85)	.0537	.5000	.0004
Transverse temporal cortex-R	.37	.17	.36	(.00; .47)	.00	(.00; .36)	.64	(.53; .81)	.0823	.5000	<.0001
Banks sup. temporal sulcus-L	.33	.21	.24	(.00; .46)	.08	(.00; .38)	.68	(.54; .84)	.1856	.3604	<.0001
Banks sup. temporal sulcus-R	.24	−.05	.19	(.00; .34)	.00	(.00; .27)	.81	(.66; .94)	.0926	.5000	.0086
Medial Aspect
Entorhinal cortex-L	.20	.19	.00	(.00; .35)	.19	(.00; .32)	.81	(.66; .92)	.5000	.1645	.0034
Entorhinal cortex-R	.25	.02	.22	(.00; .39)	.00	(.00; .30)	.78	(.61; .91)	.0895	.5000	.0053
Parahippocampal cortex-L	.40	.22	.35	(.00; .53)	.04	(.00; .40)	.61	(.47; .77)	.0785	.4264	<.0001
Parahippocampal cortex-R	.41	.32	.05	(.00; .49)	.34	(.00; .49)	.61	(.49; .75)	.4240	.0706	<.0001
Temporal pole-L	.39	.07	.35	(.07; .48)	.00	(.00; .36)	.65	(.52; .81)	.0259	.5000	<.0001
Temporal pole-R	.33	.02	.29	(.04; .44)	.00	(.00; .34)	.71	(.56; .85)	.0350	.5000	.0003
Fusiform gyrus-L	.49	.18	.46	(.14; .57)	.00	(.00; .38)	.54	(.43; .69)	.0122	.5000	<.0001
Fusiform gyrus-R	.47	.17	.44	(.11; .55)	.00	(.00; .39)	.56	(.45; .71)	.0172	.5000	<.0001

r_mz = correlation between monozygotic twins; r_dz = correlation between dizygotic twins; a² = additive genetic influences; c² = shared/common environmental influences; e² = unique environmental influences; 95% CI = corrected 95 percent confidence intervals; no A = test of the additive genetic influences can be removed from the model; no C = test of whether the shared/common environmental influences can be removed from the model; no AC = test of whether both the additive genetic and shared/common environmental influences can be removed from the model.

In order to ensure no possible bias in the WM/GM contrast heritability estimates as the result of scanner differences, additional univariate analyses were performed excluding twin pairs discordant for scanning site. For these twin pairs, the different scanners would be expected to act as unique environmental factors, making the twins appear more dissimilar from one another. Should scanner effects still be present, their removal would be anticipated to increase heritability estimates by reducing unique environmental variance. Removing these subjects from the analyses resulted in an overall small decrease in our heritability estimates. The resulting change in heritability was small; moreover, the small changes were in the opposite direction of what would be expected. We therefore concluded that the presence of scanner-discordant twins in our data did not bias our heritability estimates. Results from these analyses are available from the corresponding author upon request.

Relationship Between White Matter/Gray Matter Contrast and Cortical Thickness

Phenotypic, genetic, and environmental correlations from the bivariate analysis of WM/GM contrast and cortical thickness, as well as tests of significance for each correlation are presented in Table 2. Given the small and non-significant effects of the shared environment on nearly all WM/GM contrast measures, as well as similar effects of the shared environment on previously reported upon estimates of cortical thickness (Kremen et al., 2010), the C parameters were constrained to be zero in all models. Significant phenotypic correlations were observed for 40 of the 66 ROIs examined. The majority of these correlations were negative, ranging from −.11 to −.54; however, some positive associations were also observed, ranging from .15 to .20. Overall, the average phenotypic correlation for all ROIs was −.09. In comparison, significant genetic correlations were observed for only 17 of the 66 ROIs examined. The genetic correlations ranged in magnitude from −.89 to .25, with only values in the negative range reaching statistical significance. The strongest significant genetic correlations were observed in the left and right pericalcarine cortices (r_g = −.68 and −.89, respectively), while the remaining significant correlations were dispersed across all of the major lobes. For the remaining bivariate models, the genetic correlation could be constrained to be zero without a significant reduction in fit (p > .05), indicating the presence of minimal shared genetic variance between WM/GM contrast and cortical thickness.

Table 2.

Phenotypic, genetic, and environmental correlations between WM/GM contrast and cortical thickness

Region of Interest	Phenotypic Correlations		Genetic Correlations		Unique Environment Correlation		p-values
	rp	95% CI	rg	95% CI	re	95% CI	No rp	No rg	No re
Frontal Lobe
Superior Frontal Gyrus-L	−.05	(−.16; .06)	−.16	(−.34; .01)	.18	(.00; .34)	.0682	.0668	.0463
Superior Frontal Gyrus-R	−.03	(−.14; .08)	−.17	(−.35; .01)	.22	(.05; .38)	.0271	.0689	.0113
Middle Frontal Gyrus
Rostral division-L	.04	(−.06; .15)	.10	(−.18; .37)	.00	(−.17; .17)	.7047	.4839	1.000
Rostral division-R	.02	(−.08; .13)	−.09	(−.34; .15)	.14	(−.03; .30)	.2894	.4424	.1153
Caudal division-L	.00	(−.11; .11)	−.20	(−.42; .01)	.27	(.10; .42)	.0083	.0644	.0022
Caudal division-R	−.01	(−.11; .10)	−.17	(−.38; .03)	.23	(.06; .39)	.0253	.0949	.0080
Inferior Frontal Gyrus
Pars opercularis-L	.18	(.07; .28)	.04	(−.18; .25)	.36	(.20; .50)	<.0001	.6985	<.0001
Pars opercularis-R	.18	(.08; .28)	.09	(−.23; .38)	.23	(.07; .38)	.0005	.5485	.0045
Pars triangularis-L	.07	(−.03; .18)	.13	(−.16; .40)	.03	(−.13; .20)	.3734	.3652	.6985
Pars triangularis-R	.06	(−.04; .16)	−.02	(−.35; .28)	.12	(−.05; .28)	.2645	.8875	.1573
Pars orbitalis-L	−.22	(−.31; −.12)	−.27	(−.59; .09)	−.19	(−.34; −.03)	.0001	.1269	.0221
Pars orbitalis-R	−.11	(−.21; −.01)	−.14	(−.43; .17)	−.10	(−.26; .08)	.0876	.3681	.2815
Orbitofrontal Cortex
Lateral division-L	−.15	(−.25; −.05)	−.24	(−.48; .01)	−.07	(−.24; .10)	.0177	.0599	.4274
Lateral division-R	−.09	(−.19; .02)	−.19	(−.43; .06)	.01	(−.17; .18)	.2276	.1416	.9203
Medial division-L	−.28	(−.37; −.19)	−.24	(−.50; .10)	−.31	(−.46; −.16)	<.0001	.1463	.0001
Medial division-R	−.40	(−.48; −.31)	−.64	(−.86; −.42)	−.20	(−.35; −.03)	<.0001	<.0001	.0210
Frontal pole-L	−.19	(−.28; −.09)	−.24	(−.68; .24)	−.17	(−.32; .00)	.0011	.2857	.0447
Frontal pole-R	−.25	(−.34; −.15)	.05	(−1.0; 1.0)	−.31	(−.45; −.16)	<.0001	.8875	.0001
Precentral gyrus-L	−.15	(−.25; −.04)	−.25	(−.42; −.07)	.05	(−.13; .22)	.0216	.0069	.5902
Precentral gyrus-R	−.18	(−.29; −.07)	−.33	(−.50; −.15)	.10	(−.08; .27)	.0016	.0004	.2857
Paracentral lobule-L	−.16	(−.26; −.06)	−.32	(−.55; −.08)	.00	(−.17; .17)	.0082	.0081	1.000
Paracentral lobule-R	−.23	(−.33; −.13)	−.33	(−.52; −.13)	−.11	(−.28; .06)	.0001	.0020	.2031
Cingulate Cortex
Rostral anterior cingulate-L	−.20	(−.29; −.10)	−.37	(−.81; .07)	−.13	(−.28; .04)	.0006	.0897	.1237
Rostral anterior cingulate-R	−.23	(−.32; −.13)	−.13	(−.48; .32)	−.27	(−.40; −.11)	<.0001	.5023	.0007
Caudal anterior cingulate-L	−.45	(−.53; −.37)	−.57	(−.87; −.15)	−.41	(−.53; −.27)	<.0001	.0186	<.0001
Caudal anterior cingulate-R	−.35	(−.44; −.26)	−.62	(−1.0; −.19)	−.26	(−.41; −.10)	<.0001	.0126	.0021
Rostral posterior division-L	−.01	(−.11; .09)	.09	(−.30; .52)	−.06	(−.21; .11)	.7945	.6315	.5023
Rostral posterior division-R	−.10	(−.21; .00)	−.28	(−.59; .02)	.01	(−.16; .17)	.0963	.0668	.9203
Retrosplenial cortex-L	−.05	(−.15; .06)	.05	(−.25; .37)	−.11	(−.27; .05)	.3465	.7401	.1703
Retrosplenial cortex-R	−.17	(−.26; −.07)	−.30	(−.60; .00)	−.09	(−.24; .08)	.0054	.0472	.2965
Parietal Lobe
Postcentral gyrus-L	−.06	(−.17; .05)	−.20	(−.40; .00)	.14	(−.04; .30)	.1142	.0528	.1183
Postcentral gyrus-R	−.22	(−.32; −.11)	−.34	(−.52; −.14)	−.06	(−.23; .11)	.0003	.0012	.4976
Supramarginal gyrus-L	.20	(.10; .30)	.05	(−.20; .26)	.36	(.21; .50)	<.0001	.6892	<.0001
Supramarginal gyrus-R	.15	(.05; .25)	.03	(−.25; .28)	.25	(.09; .40)	.0011	.8065	.0027
Superior parietal cortex-L	−.02	(−.13; .09)	−.24	(−.46; −.03)	.29	(.12; .44)	.0028	.0216	.0010
Superior parietal cortex-R	.06	(−.05; .17)	−.07	(−.28; .14)	.25	(.07; .41)	.0171	.5220	.0059
Inferior parietal cortex-L	.16	(.05; .26)	.10	(−.14; .33)	.23	(.05; .39)	.0017	.3897	.0104
Inferior parietal cortex-R	.20	(.10; .30)	.17	(−.09; .40)	.23	(.07; .39)	.0001	.1990	.0061
Precuneus-L	.15	(.04; .25)	−.01	(−.25; .22)	.33	(.17; .47)	.0001	1.000	.0001
Precuneus-R	.18	(.07; .28)	.09	(−.15; .31)	.27	(.11; .42)	.0002	.4348	.0014
Occipital Lobe
Lingual gyrus-L	−.21	(−.31; −.11)	−.20	(−.41; .03)	−.23	(−.39; −.06)	.0001	.0853	.0072
Lingual gyrus-R	−.18	(−.28; −.07)	−.34	(−.55; −.14)	.04	(−.13; .21)	.0027	.0011	.6171
Pericalcarine cortex-L	−.52	(−.59; −.44)	−.68	(−.86; −.48)	−.39	(−.53; −.24)	<.0001	<.0001	<.0001
Pericalcarine cortex-R	−.54	(−.61; −.47)	−.89	(−1.0; −.68)	−.34	(−.47; −.19)	<.0001	<.0001	<.0001
Cuneus-L	−.10	(−.21; .00)	−.06	(−.28; .17)	−.16	(−.32; .01)	.0550	.6242	.0706
Cuneus-R	−.25	(−.35; −.15)	−.18	(−.37; .04)	−.34	(−.48; −.18)	<.0001	.1096	.0001
Lateral occipital cortex-L	−.21	(−.31; −.11)	−.19	(−.39; .02)	−.23	(−.39; −.06)	.0001	.0680	.0071
Lateral occipital cortex-R	−.16	(−.27; −.06)	−.20	(−.41; .03)	−.13	(−.29; .04)	.0064	.0864	.1354
Temporal Lobe
Lateral Aspect
Superior temporal gyrus-L	.04	(−.07; .14)	−.07	(−.34; .20)	.13	(−.04; .30)	.2967	.6101	.1302
Superior temporal gyrus-R	.10	(−.01; .20)	.03	(−.22; .27)	.17	(.00; .33)	.0542	.8065	.0460
Middle temporal gyrus-L	.17	(.07; .27)	.25	(−.04; .53)	.11	(−.06; .27)	.0049	.0848	.1897
Middle temporal gyrus-R	.10	(−.01; .20)	−.07	(−.39; .22)	.21	(.04; .37)	.0253	.6315	.0140
Inferior temporal gyrus-L	.06	(−.04; .17)	.02	(−.29; .29)	.10	(−.07; .26)	.3499	.8875	.2636
Inferior temporal gyrus-R	−.03	(−.13; .08)	−.05	(−.31; .21)	−.01	(−.17; .15)	.8825	.7184	.9203
Transverse temporal cortex-L	−.16	(−.26; −.06)	−.22	(−.51; .09)	−.14	(−.30; .04)	.0055	.1583	.1245
Transverse temporal cortex-R	−.11	(−.22; −.01)	−.27	(−.56; .01)	.00	(−.17; .16)	.0696	.0538	1.000
Banks sup. temporal sulcus-L	−.06	(−.16; .04)	.07	(−1.0; 1.0)	−.08	(−.24; .08)	.4296	.8875	.3009
Banks sup. temporal sulcus-R	.09	(−.01; .19)	.02	(−.68; .67)	.11	(−.05; .26)	.1800	1.000	.1809
Medial Aspect
Entorhinal cortex-L	−.31	(−.40; −.22)	−.47	(−.93; −.03)	−.26	(−.40; −.10)	<.0001	.0393	.0016
Entorhinal cortex-R	−.32	(−.41; −.23)	−.58	(−1.0; −.17)	−.22	(−.37; −.06)	<.0001	.0100	.0066
Parahippocampal cortex-L	−.16	(−.26; −.05)	−.18	(−.46; .12)	−.14	(−.30; .03)	.0083	.2146	.1029
Parahippocampal cortex-R	−.17	(−.27; −.07)	−.29	(−.52; −.05)	−.07	(−.23; .10)	.0041	.0198	.4062
Temporal pole-L	−.21	(−.30; −.11)	−.33	(−.62; −.04)	−.13	(−.29; .04)	.0003	.0275	.1245
Temporal pole-R	−.04	(−.14; .06)	.20	(−.25; .80)	−.12	(−.28; .04)	.3296	.3802	.1371
Fusiform gyrus-L	.08	(−.02; .18)	.10	(−.17; .36)	.07	(−.10; .23)	.2982	.4751	.4424
Fusiform gyrus-R	.05	(−.06; .15)	−.13	(−.39; .12)	.22	(.05; .37)	.0408	.3149	.0122

r_p = phenotypic correlation; r_g= genetic correlation; r_e= unique environment correlation; No r_p = test of no phenotypic correlation between estimates of WM/GM contrast and cortical thickness; No r_g = test of no genetic correlation; No r_e = test of no unique environment correlation.

With respect to the unique environmental influences, 32 of the 66 ROIs were found to have significant unique environmental correlations between WM/GM contrast and cortical thickness, with roughly equal numbers of positive and negative relationships observed. The only areas that were found to demonstrate a consistent pattern of results were the left and right parietal lobes, where all but two ROIs were found to have significant unique environmental correlations. As was the case with the genetic correlations, the effects, although significant, were small and suggested only minimal shared unique environmental variance between the phenotypes.

Discussion

In the present study we examined if the degree of signal contrast between white matter and gray matter along the cortical mantle could be utilized as a genetically-informative brain imaging phenotype, a role beyond its current relegation as a technical property of MRI. Our results demonstrate that WM/GM contrast is indeed genetically influenced, with significant heritability estimates ranging from .29 to .66. In addition, the genetic and environmental determinants of WM/GM contrast were found to be largely independent of those that influence cortical thickness (i.e., there was very little shared genetic variance between WM/GM contrast and cortical thickness). Although significant phenotypic correlations were observed for 40 out of the 66 ROIs examined, only 17 genetic correlations were found to be significant. In these cases the degree of shared genetic variance tended to be small and suggested more unique genetic influences rather than common genetic influences for the phenotypes. For the vast majority of regions examined, the genetic correlation between WM/GM contrast and cortical thickness could be constrained to be zero without a significant reduction in model fit. Thus, while the degree of WM/GM contrast is certainly critical to the accurate estimation of cortical thickness, it also has minimal genetic overlap with cortical thickness, suggesting that it captures additional genetic variance that is relevant to structural properties of the cerebral cortex.

The observed heritability estimates for WM/GM contrast were similar in magnitude to those for cortical thickness previously reported from the VETSA sample (Kremen et al., 2010). These heritability estimates ranged from .00 in the right rostral and left caudal divisions of the anterior cingulate, to .75 in the left superior frontal gyrus. Average heritability estimates for the ROIs within the major lobes were .53 for the occipital lobe, .49 for the frontal lobe, .60 for the parietal lobe, .40 for the temporal lobe, and .28 for the cingulate cortex. It is important to note that the heritability estimates for cortical thickness, as well as those for WM/GM contrast, possess relatively broad 95% confidence intervals. Thus, while it may be possible to state that for a few ROIs cortical thickness is statistically more genetically influenced than WM/GM contrast, or vice versa, this statement should not be generalized to the entire brain. Additional studies of the heritability of WM/GM contrast in other twin and family MRI samples are needed in order to more precisely determine the degree to which the phenotype is genetically influenced.

Recent cross-sectional studies spanning early to older adulthood have noted that age-related differences in WM/GM contrast, or equivalent measures, appear to be driven primarily by decreasing signal intensity within the white matter (Salat et al., 2009; Westlye et al., 2009), leading to speculation that WM/GM contrast is most influenced by the degree of myelination of white matter fibers under the cortical mantle. If this is truly the case, then it will be of interest to determine the degree of phenotypic, genetic, and environmental overlap between WM/GM contrast and indicators of white matter integrity derived from additional MRI sequences such as diffusion tensor imaging (DTI). Several studies have now demonstrated that phenotypes derived from DTI are heritable (Brouwer et al., 2010; Chiang et al., 2009; Chiang et al., 2011; Kochunov et al., 2010); however, there has yet to be an examination of whether these measures are genetically or environmentally related to other brain imaging phenotypes. Should strong genetic overlap exist between WM/GM contrast measures and DTI-based measures, this could suggest that WM/GM contrast provides an indicator of white matter integrity for the fibers along the gray-white boundary. Alternatively, weak genetic overlap might suggest that WM/GM contrast captures unique structural properties that are not otherwise observable with other white matter imaging methods.

These results are derived from cross-sectional data; therefore, we cannot be certain as to whether the measures of WM/GM contrast obtained in the present study are reflective of age-related changes in the phenotype, or rather are indicative of longstanding (pre-existing) tissue properties. Measures of signal contrast have been shown to be highly sensitive to the effects of age in multiple samples, with the ratios of white matter to gray matter signal intensity (or vice versa) approaching 1.0 as participants get older (Davatzikos and Resnick, 2002; Magnaldi et al., 1993; Raz et al., 1990; Salat et al., 2009; Westlye et al., 2009). Moreover, the component elements of the measure, the white matter and gray matter signal intensities, both demonstrate inverted “U” shaped patterns when examined across the lifespan, suggesting that the phenotype undergoes substantial changes both during development and then again later in life (Westlye et al., 2010). That we observed little to no effect of age on WM/GM contrast in the present study was likely a product of our rather narrow age range of 51 to 59 years. Ongoing follow-up examination of this phenotype in the second wave of the VETSA project will be able to address whether the degree of WM/GM contrast changes over the course of late middle-age, as well as the role of genetic and environmental influences on those changes.

Despite using methods designed to be compatible across different scanners and vendors, we observed significant scanner effects for nearly all measures of WM/GM contrast. While such effects certainly influence the absolute measures of WM/GM contrast, their impact on the observed heritability estimates is likely negligible. The vast majority of participants in the VETSA (roughly 95%) were scanned at the same site as their co-twin; thus, any effect of scanner on WM/GM contrast would likely be observed as a shared/common environmental influence (i.e., something that makes twins more similar to one another). By adjusting for scanner differences we eliminated any such effect and as a result made the heritability estimates (the statistic of primary interest in the present study) more precise. Statistical modeling of multi-site neuroimaging is an important area of work, and there have been a number of studies that have examined differences between scanners (including the effects of different vendors and different pulse sequences) that are important in this regard. While the more similar the initial sequence and scanner are, the more similar the resultant measures, recent work has shown that combining data from multiple sites is feasible and provides sufficient power to look at these questions, particularly when site or scanner is used as a covariate in the model (Fennema-Notestine et al., 2007; Jack et al., 2008; Kruggel et al., 2010; Stonnington et al., 2008)

There are several limitations to the present study that should be addressed. With the all-male, largely Caucasian composition of the VETSA sample, we are limited in our ability to generalize these findings to other populations. Similarly, the results observed for this sample may differ from those obtained in either younger or older cohorts where individual differences in WM/GM contrast could be markedly greater or reduced as a function of age. Although our generalizability to other groups may be limited, it should be noted that the goal of the VETSA was to characterize men in a narrow midlife age range and then follow them as they age. Thus, the ability to examine aging-related phenotypes within a specific age cohort should also be viewed as a strength of the study’s design. It is also the case that these results are not necessarily generalizable to other studies that do not use inversion recovery sequences, and the results may reflect an underestimation of age-related effects on WM/GM contrast relative to studies utilizing classical spoiled gradient echo sequences. Nevertheless, the methods used in the current study are similar, if not identical to those used by other researchers who have recently utilized contrast phenotypes to demonstrate subtle brain changes associated with normal aging and AD (Salat et al., 2011; Salat et al., 2009; Westlye et al., 2010; Westlye et al., 2009). There may also be some measurement error introduced in our boundary estimation due to our anisotropic sampling, particularly when reslicing into isotropic volumes. Previous work suggests that anisotropic acquisitions may result in a bias to underestimate cortical thickness across the surface (Wonderlick et al., 2009). Our use of sinc interpolation in resampling preserves most of the original image information without further image degradation or blurring, and therefore reduces the impact of this problem on our estimates of cortical thickness and WM/GM contrast. The methodological approach used herein is similar to many previous studies of independent samples that have demonstrated sensitivity to subtle effects, thus supporting the utility of the measures derived from images with comparable voxel size (e.g., Salat et al., 2004; Dickerson et al., 2008; Fennema-Notestine et al., 2009; Fjell et al., 2009).

Finally, it is possible that the voxel size utilized in the present study (1.3×1.0×1.3mm) may influence the heritability of the WM/GM contrast phenotype, and that a smaller voxel size could result in a more refined heritability estimate¹. Given that this is the first study to report on the heritability of WM/GM contrast, no direct comparisons are possible; however, insights may be gleaned from the comparison of the heritability estimates from the VETSA (Kremen et al., 2010) with those from a subsequent genetically-informative study. In the Genetics of Brain Structure and Function Study (GOBS), an independent sample of family-based neuroimaging data, the heritability of cortical thickness was examined utilizing the same parcellation system as the VETSA with a voxel size of 0.8 mm³ (Winkler et al., 2010). Similar to the VETSA, the ROI-based heritability estimates for cortical thickness in the GOBS demonstrated a wide range (.12 to .84; .00 to .75 in VETSA), with a roughly comparable average of the heritability estimates for all ROIs (VETSA = .46; GOBS = .41). Average heritability estimates for cortical thickness at the lobar level were also comparable across samples (Frontal: VETSA = .49, GOBS = .37; Temporal: VETSA = .40, GOBS = .41; Parietal: VETSA = .60, GOBS = .50; Occipital: VETSA = .53, GOBS = .44; Cingulate Cortex: VETSA = .28, GOBS = .37). Three of these estimates were slightly higher in the VETSA and two were slightly higher in the Winkler et al. study. These values indicate no systematic differences in the heritability estimates of the two studies, despite the differences in voxel size. Thus, the evidence argues against voxel size leading to systematically different heritability estimates for WM/GM contrast.

In addition to the above limitations, it is worth noting that conclusions drawn about the regionality of genetic and environmental determinants of WM/GM contrast may be influenced by the type of cortical parcellation system utilized. In a previous study by our group we demonstrated that heritability estimates of cortical thickness and cortical surface area derived from continuous maps of the cortical surface did not correspond to a priori ROI definitions (Rimol et al., 2009; Chen et al., 2011). This approach does not invalidate the use of cortical parcellation systems, which have established functional and anatomical significance, but rather highlights the fact that patterns of genetic and environmental influences may not adhere to predefined boundaries despite the established functional and anatomical significance of many cortical parcellation systems. Imposing these boundaries on the data may, therefore, introduce additional error into the results, subsequently reducing the heritability estimates.

In summary, the present results demonstrate that WM/GM contrast is both a heritable phenotype, and is largely genetically independent from cortical thickness. We conclude that WM/GM contrast may represent a novel phenotype with which to investigate the genetic determinants of brain structure and brain aging. Further studies are needed in order to replicate the findings in other populations and to determine how the genetic influences of WM/GM contrast may change throughout the lifespan.