Reduced variance in monozygous twins for multiple MR parameters: Implications for disease studies and the genetic basis of brain structure

Reduced variance in monozygous twins for multiple MR parameters: Implications for disease studies and the genetic basis of brain structure

NeuroImage 49 (2010) 1536–1544 Contents lists available at ScienceDirect NeuroImage j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l...

434KB Sizes 1 Downloads 41 Views

NeuroImage 49 (2010) 1536–1544

Contents lists available at ScienceDirect

NeuroImage j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / y n i m g

Reduced variance in monozygous twins for multiple MR parameters: Implications for disease studies and the genetic basis of brain structure Gaby S. Pell a, Regula S. Briellmann a, Kate M. Lawrence b, Deborah Glencross b, R. Mark Wellard b, Samuel F. Berkovic a,b, Graeme D. Jackson a,⁎ a b

Brain Research Institute, Neurosciences Building, Austin Health, Heidelberg West, Victoria 3081, Australia Epilepsy Research Centre, Department of Medicine, University of Melbourne, Austin Health, Heidelberg West, Victoria 3081, Australia

a r t i c l e

i n f o

Article history: Received 20 January 2009 Revised 27 August 2009 Accepted 2 September 2009 Available online 9 September 2009 Keywords: Brain structure Twins Voxel-based Genes Phenotype Genetics MRI

a b s t r a c t Twin studies offer the opportunity to determine the relative contribution of genes versus environment in traits of interest. Here, we investigate the extent to which variance in brain structure is reduced in monozygous twins with identical genetic make-up. We investigate whether using twins as compared to a control population reduces variability in a number of common magnetic resonance (MR) structural measures, and we investigate the location of areas under major genetic influences. This is fundamental to understanding the benefit of using twins in studies where structure is the phenotype of interest. Twenty-three pairs of healthy MZ twins were compared to matched control pairs. Volume, T2 and diffusion MR imaging were performed as well as spectroscopy (MRS). Images were compared using (i) global measures of standard deviation and effect size, (ii) voxel-based analysis of similarity and (iii) intra-pair correlation. Global measures indicated a consistent increase in structural similarity in twins. The voxel-based and correlation analyses indicated a widespread pattern of increased similarity in twin pairs, particularly in frontal and temporal regions. The areas of increased similarity were most widespread for the diffusion trace and least widespread for T2. MRS showed consistent reduction in metabolite variation that was significant in the temporal lobe N-acetylaspartate (NAA). This study has shown the distribution and magnitude of reduced variability in brain volume, diffusion, T2 and metabolites in twins. The data suggest that evaluation of twins discordant for disease is indeed a valid way to attribute genetic or environmental influences to observed abnormalities in patients since evidence is provided for the underlying assumption of decreased variability in twins. © 2009 Elsevier Inc. All rights reserved.

Introduction Twin studies offer the opportunity to determine the relative contribution of genes versus environment in a trait of interest. Moreover, many studies have used twins discordant for disease to distinguish between genetic and environmental influences on the morphological abnormalities associated with neurological and psychiatric disease. Conditions such as schizophrenia (Narr et al., 2002; Hulshoff Pol et al., 2002a; Styner et al., 2005), Alzheimer's disease (Luxenberg et al., 1987; Andel et al., 2005), autism (Kates et al., 2004), Tourette's syndrome (Hyde et al., 1995) and attention disorders (Reiersen et al., 2008) have been investigated in this manner. All these investigations have focused on brain volumes obtained from T1-weighted high-resolution MR images. However, there are a

⁎ Corresponding author. Fax: +613 9496 2980. E-mail address: [email protected] (G.D. Jackson). 1053-8119/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2009.09.003

host of alternative structural MR parameters describing complementary biological information. These include measurements of T2 relaxation time, water diffusion and metabolite concentration that provide unique information about structural and functional integrity. It is not yet known how these parameters vary in identical twins as compared to unrelated controls. A number of studies have investigated heritability of brain volume in healthy twins using complex computationally intensive approaches that are not readily amenable to group comparisons in disease states (for example, White et al., 2002, Thompson et al., 2001, Wright et al., 2002). A widely used alternative technique for the objective analysis of quantitative MR imaging parameters between such groups is the voxel-based (VB) approach. The technique, originally applied to volume – known as voxel-based morphometry (Ashburner and Friston, 1997) – has since been extended to the evaluation of changes in other MR parameters such as the T2 relaxation time (for example, Pell et al., 2004). The method enables the objective assessment of changes between groups of subjects across the entire imaging volume. This offers the advantage

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

in twin studies of being able to map the distribution of similarity between twin pairs when compared to healthy age-matched control pairs. The VB approach was therefore selected in this study to investigate the quantitative MRI parameters, volume, T2 relaxation time and diffusion by mapping the degree of similarity in each group. We also looked at the similarity of metabolite information obtained from magnetic resonance spectroscopy (MRS) in twins. This technique has recently been applied to a twin study of schizophrenia (Lutkenhoff et al., 2008). The aim of this study is twofold: (1) to investigate the extent to which the variance in a range of MR measures of brain structure can be reduced by controlling genetic variation; and (2) to investigate which areas of the brain are under the most genetic influence for each MR parameter. Methods Subjects Twins pairs Twenty-three pairs of healthy, monozygous twins were recruited through the Australian Twin Registry (mean age = 30 ± 5 years; 8 male twins pairs, 15 female pairs). All underwent DNA testing, and monozygosity was confirmed in all cases by a standard technique of analysis of eight highly polymorphic markers where the likelihood of dizygosity given identity of these markers is less than 0.001%. Fifteen pairs were right-handed, one pair was left-handed and the rest each consisted of a right- and left-handed twin. The twin subjects were of average height and weight and intelligence/socioeconomic status. Control pairs Twenty-three corresponding pairs of age- and gender-matched but unrelated subjects (mean age = 30 ± 5 years; 8 male pairs, 15 female pairs) were selected from our MR database comprising more than 200 subjects scanned at our institute. All subjects underwent a detailed questionnaire to exclude pre-existing neurological or psychiatric disease, with particular emphasis to exclude even mild forms of epilepsy. The structural images of all subjects (twins and unrelated control pairs) were assessed by an experienced neuroradiologist, and no abnormalities were detected in any studies. Ethics approval was obtained from the Austin Health Human Research Ethics Committee. All subjects gave written informed consent. It should be noted that the dominance of females in the twins group was matched in the controls. Imaging All MR imaging was performed on a 3-T GE LX Horizon scanner using the standard GE birdcage radiofrequency (RF) head coil. Volume The structural scan was a T1-prepared 3D high-resolution SPGR sequence (echo time, TE = 2.7 ms; repetition time, TR = 13.8 ms; flip angle = 20°; TI = 500 ms; voxel size= 0.48 × 0.48 × 2 mm). T2 relaxometry The T2 mapping sequence was a modified, optimized Carr-PurcellMeiboom-Gill (CPMG) multi-echo sequence (Pell et al., 2006a) (8 echoes, TE = 28.875–231 ms spaced at equal intervals, TR = 6.24 s, 24 slice of 5-mm thickness, no gap, in-plane voxel size = 0.9 × 1.8 mm). The slices were acquired in a plane perpendicular to the long axis of the hippocampus.

1537

Diffusion Diffusion tensor imaging (DTI) acquisitions were based on a spin echo EPI (SE-EPI) sequence. The slices were acquired axially across the whole brain volume. Since the control scans were collected as part of a different protocol at the institute, the sequence parameters differed between the control and twin pair acquisitions in the following way: Twins: Single echo SE-EPI, 52 directions + 4 b = 0 acquisitions (i.e., scans with no diffusion weighting), slice thickness = 3 mm, in-plane voxel size = 2.2 × 2.2 mm, b = 1100 s/mm2. Control pairs: Double echo SE-EPI, 28 directions + 5 b = 0 acquisitions, slice thickness = 2.5 mm, in-plane voxel size = 2.5 × 2.5 mm, b = 1200 s/mm2. The influence of the different DTI acquisitions on the comparative results was explored in this study. Image processing Images were processed using the steps of voxel-based analysis in order to warp the images into a standard space in which they could be compared. The approach can be contrasted with deformation-based morphometry which searches the deformation fields from the warping step for anatomical differences (Davatzikos et al., 1996). All images were normalized to standard space and resampled to 1 × 1 × 1 mm resolution. Smoothing was applied after pairwise subtraction (10 mm kernel). Images were checked at various stages during the processing to ensure consistent image quality. Standard SPM templates were used rather than custom templates since neuroanatomy was not expected to be differ from the default templates and no benefit has been demonstrated with the use of customized templates (Keller et al., 2004). The pre-processing steps followed the procedure commonly used for analysis of volume using voxel-based morphometry (VBM), DTI parameters (voxel-based diffusion, VBD) and T2 (voxel-based relaxometry, VBR). The specifics of the pre-processing for each modality are now described. Volume (VBM) Gray matter (GM) and white matter (WM) voxel-based analyses were performed with SPM2 (http://www.fil.ion.ucl.ac.uk/spm2), using the optimized VBM approach (Good et al., 2001). In brief, the steps included segmentation, spatial normalization of the desired segment, re-segmentation of the normalized image and modulation of the segment to compensate for warping effects. The total intracranial volume (TIV) was calculated as the sum of the three tissue segments (GM + WM + CSF) and transformed to units of liters (i.e., 106 mm3). Analysis of gray and white matter was denoted VBM (GM) and VBM (WM), respectively. To exclude inappropriate voxels from the analysis, the statistical parametric map was masked by an image generated by thresholding the appropriate SPM segment templates so that any voxel below 20% of the mean level was set to zero. Diffusion (VBD) The DTI images were processed with the FDT package in FSL (www.fmrib.ox.ac.uk/fsl/). The EPI images were corrected for eddy current distortion and the diffusion tensor model was fitted to each voxel. The fractional anisotropy (FA) and trace (mean diffusivity (MD)) were determined as invariant measures of anisotropy and mean diffusivity respectively. Warping of these images to standard space followed an optimized pathway (Pell et al., 2006b). The structural T1-weighted scan was segmented and the FA image coregistered to the gray matter segment. Coregistered images were then warped to standard space using parameters from the warp of the GM segment to SPM GM template. To exclude inappropriate voxels, the white matter mask was used for the analysis of fractional anisotropy. For the analysis of mean diffusion, a brain tissue mask was created by thresholding the union of the SPM gray and white

1538

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

matter templates at 20%. Analysis of fractional anisotropy and mean diffusion was denoted VBD (FA) and VBD (MD), respectively. T2 relaxation time (VBR) T2 maps were generated by fitting the relaxation time curves to an exponential model, i.e. S(t) = S(0).exp(− t/T2), where S(t) is the signal acquired at time, t. VBR analysis was carried out according to the procedure outlined in Pell et al. (2004). In brief, the T2 relaxation maps were spatially normalized to standard space via the T2-weighted images (3rd echo) using the SPM T2 template. The same mask as used for the DTI (MD) analysis was employed. Analysis As a measure of intra-pair similarity, pairwise subtraction of preprocessed, normalized images was performed. GM and WM volume images were divided by the TIV in order to control for global brain volume since TIV could not be included as a covariate in the analysis of paired subtractions as normally required in VBM. The similarity of twin and control pairs was then compared using the following three steps stated with their respective aims: Global analysis of standard deviation and effect size To assess whether the use of twins as compared to normal controls reduces variance, maps of the effect size and standard deviation of the similarity measure were created for each group of paired subtraction images. This was achieved by analysis of the smoothed subtraction images with a one-sample test in each group testing the null hypothesis of no intra-pair differences in image intensity. The effect size and standard deviation were calculated from the SPM output images as beta and as the calculated ratio, beta/T-statistic, respectively. As summary measures, these statistics were evaluated within global masks of the appropriate tissue segment. Voxel-based analyses of similarity Voxel-based analyses were used to assess the spatial location and distribution of areas of presumed genetic influence. This approach enables the objective assessment of changes across the whole imaging volume. In order to compare the degree of similarity between the twin and control pairs, a metric based on intra-pair signal differences was created in the following manner. The pairwise difference images were “unsigned” (the absolute operation) so that the magnitude of the differences could be used as a relative measure of similarity, and then smoothed. Similarities of control and twin pairs were assessed using a two-sample t-test with an appropriate contrast to evaluate the “difference of the differences”, i.e., the disparity between the intrapair difference images from the twin pair and its corresponding matched control pair. This can be summarized by the following expressions: Intra-pair difference: |twin1 − twin2| or |control1 − control2| within each twin or matched control pair. Inter-pair difference: twin intra-pair difference − control intra-pair difference and vice versa. The absolute subtraction used in the intra-pair difference expression gives equal weight to any disparity, whether positive or negative. A paired t-test was employed with the appropriate contrast to evaluate the inter-pair differences across the twin and matched control pairs. By using the “difference of the differences” in this way, the disparity between the intra-pair difference images from each twin pair and corresponding matched control pair was assessed using a paired t-test. The statistical threshold was p b 0.0001 (uncorrected. corresponding to T = 4.4).

Correlation analysis To assess the spatial location and distribution of areas of presumed genetic influence by comparison of intra-pair correlation coefficients, voxelwise correlation maps were created across the pairs of images in each group of images. This is similar to the approach used in White et al. (2002). Statistical comparison of the two groups of correlation maps was then carried out (paired t-test). The threshold of p b 0.0001 (uncorrected) was calculated using the Fisher transform (Fisher, 1921). Follow-up analyses Calculation of sample size. In order to examine the practical benefits of a difference in variance in twins, the minimum sample size necessary to detect a specified effect size with specified test characteristics, was calculated. The sample size was calculated for a fixed level of statistical power, 1 − β (0.7), significance level, α (0.05), desired level of effect size, δ (0.7) and the calculated pooled variance, s2p. The latter term was used as an estimator of the true population variance and was calculated from the SPM statistical output of separate within-group tests, either a paired t-test (twins) or a twosample t-test (matched controls), with the null hypothesis of no intrapair difference. The different tests for the twins and matched controls reflected the likely analysis approach in each case. As before, the statistical result was calculated over a mask defining the whole region of relevant signal change. The effect size was chosen to represent a magnitude of structural change that might be expected in pathology. The calculation of the sample size was then an iterative procedure since the calculation of the statistical contributions of the desired levels of Types I and II errors, α and β, are dependent on the sample size. Symmetry. The degree of cerebral asymmetry with regard to heritability was quantitatively assessed by calculation of a measure of symmetry. The thresholded statistical map from the similarity analysis was integrated over each of the two hemispheres and combined to provide a laterality index as (L − R) / (L + R), where L and R are the counts of suprathreshold voxels (i.e., voxels that show improved similarity in twins over controls) in the left and right hemispheres, respectively. Three alternative p-value thresholds were used for the counting of suprathreshold voxels (p = 10− 2, 10− 4 and 10− 5) in order to assess threshold dependence of this measure. Influence of DTI acquisition. Since different DTI acquisition parameters were used for the control and twin acquisitions, the analysis was repeated with the control pairings substituted by randomized pairings of the twins (“random pairing”) to see if this affected the outcome when compared to the twin pairs. This creates a group of equivalent “control pairs” acquired with the same imaging parameters. Magnetic resonance spectroscopy Proton MRS was carried out using a standard point-resolved spectroscopy (PRESS) sequence with two chemical shift selective pulses for water suppression (flip angle = 90°, TR/TE = 3000/30 ms, 2048 data points acquired over a spectral width of 5000 Hz). Bilateral isotropic single voxel spectra were acquired (2×2×2 cm3) in the frontal and temporal lobe. Sagittal plane, 2 cm-thick scout images (T1 spin echo) followed by 2 cm-thick coronal images, centered in the plane of the ponto-medullary junction, were acquired. Frontal lobe spectra were recorded from a region-of-interest (ROI) prescribed from a 2 cm-thick axial T1 image with the inferior plane above the ventricles and excluding scalp. An ROI in each temporal lobe was selected in the coronal plane with the lateral aspect of the hippocampus in the center of the ROI. The sagittal image was viewed to ensure that the ROI did not include petrous temporal

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

bone. After positioning of the voxel, the signal over the volume of interest was shimmed to within a line width of less than 16 Hz. Data were processed with LCModel (Provencher, 1993), using water scaling as an internal reference. The following five metabolites were quantified: N-acetylaspartate plus N-acetylaspartylglutamate (NAA), choline containing compounds (Cho), total creatine (Cr), myoinositol (mI) and the complex of the glutamine and glutamate (Glx) peaks. Concentrations were expressed as institutional units, approximating mmol/l. A basis metabolite set prepared on our scanner was used. Data from the right and left hemispheres were assessed separately. The intra-pair variability was assessed in order to test whether the twins had more similar metabolite concentrations than the unrelated controls. The measurement in twins was assessed by the relationship: (twin A − twin B) / (twin A + twin B), where “twin A” is the twin with the greater metabolite value and “twin B” is the co-twin with the smaller value. In unrelated control pairs, the calculation followed (control A − control B) / (control A + control B), where “control A” was the control of the pair with the greater value. The average intrapair variability in twins and in controls was established for all metabolites grouped together and for each metabolite separately. The differences in the intra-pair variability indices between twins and controls was assessed using paired t-tests to evaluate the intra-pair variability among the metabolites.

1539

were largely symmetrical and most widespread for the VBD (MD) analysis. A general description of the regional distribution of increased twin-pair similarity at the chosen statistical threshold for each modality now follows. VBM (gray matter) The pattern of increased twin-pair similarity in VBM (GM) analysis included largely bilateral areas in the cerebellum, temporal and frontal cortices, parahippocampal gyri and the insular cortex (Figs. 1a and b). Language and somatosensory areas were also implicated. Several areas were observed that were located more in WM than GM. This reflects likely differences in normalization and segmentation boundaries between subjects in the control pairs as a consequence of increased variability of brain structure within these pairs in comparison with twins. Cortical areas were only included to a limited extent in the thresholded maps. However, the unthresholded view (Fig. 1c) indicated that the cortex tended to increased similarity. It should be noted that the value of the statistic in areas near the edge of the brain is expected to be lowered due to the effects of normalization. VBM (white matter) Areas of increased similarity in VBM (WM) analysis included medial temporal, thalamic, inferior frontal, corpus callosum, cerebellum and anterior temporal white matter areas. The unthresholded view indicates the dominant trend across the brain of increased similarity with only small zones of edge pixels displaying the opposite trend.

Results Global analysis of standard deviation and effect size

VBD (fractional anisotropy) Increased areas of similarity in VBD (FA) analysis were seen extensively in white matter, including temporal and frontal areas, and in the thalamic and lentiform nuclei. Several tracts were implicated including superior longitudinal and uncinate fasciculi. The unthresholded view displays the trend to increased similarity across the whole brain.

A consistent reduction in standard deviation and effect size was observed in the twin pairs in all the analyses (Table 1) with the exception of the standard deviation measure for the T2 relaxation time which was greater in the twins (see Discussion). Voxel-based analyses of similarity

VBD (mean diffusion) Areas displaying increased similarity in VBD (MD) analysis included extensive frontal areas, temporal lobe, cingulate cortex and bilateral insula. The unthresholded view displays the trend to increased similarity across the brain apart from areas close to ventricles where variance is expected to be increased.

Fig. 1 shows the results of the voxel-based analyses. The contrasts twin pair similarityb control pair similarity (twin pair similarity N control similarity pair similarity), indicate areas where the intra-pair similarity is lower (greater) in the twin pairs than in the control pairs. The signal in the absolute intra-pair difference images was consistently lower in the twin pairs as indicated by the areas displayed in the thresholded statistical parametric maps for the contrast: twin pair similarity N control pair similarity. The increased intra-twin-pair similarity had a regional distribution that varied between the different analyses. No areas of increased intra-pair difference in twins as compared to controls were observed (contrast: control pair similarity N twin pair similarity). The focal differences

VBR (T2) Areas displaying increased similarity in twin pairs were smallest in size relative to the other structural parameters investigated here, and included bilateral lentiform nuclei and insula, midbrain, temporal and parahippocampal areas. No areas were observed in the opposite

Table 1 Global measures of the effect size and standard deviation of the similarity measure (i.e., intra-pair difference between the control and twin pairs). Unit

Effect size

Standard deviation

Control VBM (GM) VBM (WM) VBD (FA) VBD (MD) VBR (T2)

a

Unitless Unitlessa Unitlessb [mm2/s]c [ms]

Twins −3

1.68 × 10 1.31 × 10− 3 3.66 × 10− 3 − 6.28 × 10− 6 1.01

% Diff −3

0.37 × 10 − 1.03 × 10− 3 1.07 × 10− 3 2.67 × 10− 6 − 0.07

− 78 − 26 − 71 − 58 −93

Control

Twins −4

1.08 × 10 1.08 × 10− 4 1.23 × 10− 4 2.76 × 10− 8 1.04

% Diff −4

0.72 × 10 0.73 × 10− 4 0.89 × 10− 4 1.99 × 10− 8 1.40

− 34 − 36 − 27 − 28 + 34

Values are calculated from the SPM output of the one-sample t-test of the paired subtraction images. Positive and negative values therefore indicate variation around the hypothesized value of zero representing no intra-pair difference. The effect size and standard deviation are calculated from the beta image and the ratio, beta/T-statistic, respectively. The percentage difference (% Diff) of the effect size is then calculated as: (|twin| − |control|) × 100 / control. VBD (FA) and VBD (MD) are the analysis of fractional anisotropy and mean diffusion trace respectively, and obtained from the DTI analysis. VBR is the analysis of T2 relaxation time data. a Volume values are equivalent to the ratio of the volume in the region (in mm3) to the TIV (in mm3). b FA values are equivalent to the typically used ratio with values between 0 and 1. c The value of diffusion trace (MD) of tissue is approximately 0.8 mm2/s.

1540

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

Fig. 1. Voxel-based analyses of the degree of similarity between twin and control pairs. The absolute intra-pair difference images are compared using paired t-test analysis (p b 0.0001 uncorrected). For example, the twin similarity N control similarity contrast in (a) highlights areas in which the twin pairs are more similar than in the control pairs. The consistent presence of signal change in this contrast indicates the benefits of using twins in such analyses. (a,b) Glass brain results are shown for the voxel-based analyses of volume, T2 and DTI parameters. (c) The unthreshold T-statistic maps showing the two contrasts (cold: controls N twins; warm: twins N controls) at the level of a representative coronal slice (y = 99 mm). Each statistical map is overlayed on the corresponding mask, for example, the gray matter mask for VBM (GM).

contrast (increased similarity in controls) but the unthresholded view indicated a trend to this direction of change in extensive areas of the cortex and ventricular areas. Correlation analysis The increased degree of correlation in the twin pairs is clearly seen by visual comparison of maps in the two groups (Figs. 2a and b). Statistical comparison of these correlation maps displayed a spatial distribution that is considerably more extensive than that exhibited by the VB analysis of similarity (Fig. 1a). For T2, areas of reduced correlation in twins were reduced in size and remained non-significant. Follow-up analyses Calculation of sample size The minimum sample size required to detect the desired effect size for twin and matched controls is shown in Table 2. The benefit of the twin design on subject numbers is clearly indicated. For a volume

study, a reduction in subject numbers of approximately 40% for a twin study is enabled. This declines to approximately 25% in the case of the DTI parameters. It should be noted that the variance values were estimated over the appropriate brain mask and will be negatively impacted by areas at, for example, the edges of the mask and around the ventricles where variance is expected to be larger. These numbers therefore reflect the lower limit of the recruitment benefit offered by a twin study. For T2, the sample size calculation was not carried out due to the unreliable measure of variance in that case (see Discussion). Symmetry The values of the laterality index (LI) parameter describing the degree of symmetry of similarity in each group are tabulated below in Table 3 for the different chosen p-value thresholds. Although the general distribution across different brain areas in all cases appeared visually to be largely symmetrical, GM and WM volumes were associated with consistent increased LI in the right and left hemispheres respectively.

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

1541

Fig. 2. Correlation maps calculated from the intra-pair difference for (a) control and (b) twin pairs at the level of coronal slice (y = 99 mm). The brighter correlation maps for the twins indicate a stronger degree of correlation in that group in those areas. (c) Statistical comparison of the correlation maps in the two groups using the Fisher transform (p b 0.0001 uncorrected). For display, correlation values below 0 were assigned the same intensity.

Influence of DTI acquisition The global variance of the FA intra-pair similarity in the “control” groups selected by random pairing was 0.026 ± 0.001 (cf. 0.025 in Table 1 for the true control pairs). The spatial distribution of similarity was unchanged to that shown in Fig. 1. This indicates that the different DTI acquisition schemes did not significantly influence the findings. Magnetic resonance spectroscopy Table 4 demonstrates the metabolite concentrations in twins and unrelated control pairs. The intra-pair variability measurements in metabolite concentrations were generally smaller in twin pairs than Table 2 Calculation of minimal sample to detect a specified effect size with specified test characteristics. Calculated sample size

VBM (GM) VBM (WM) VBD (FA) VBD (MD)

Control

Twins

61 61 81 63

36 37 59 49

The sample size was calculated for a fixed level of statistical power, 1 − β (0.7), significance level, α (0.05) and pooled variance, s2p (calculated from the statistical output of within-group tests, using either a paired t-test (twins) or a two-sample t-test (matched controls) and the null hypothesis of no intra-pair difference). The chosen level of effect size represented a magnitude of structural change that might be expected in pathology (0.06 for VBM and VBD (FA) and 1.2 × 10− 4 mm2/s for VBD (MD), which represented 16% of the maximum expected level in each case).

in unrelated control pairs. The only exceptions to this trend were nonsignificant increases in twins in NAA and glutamate in the frontal lobes. Across all metabolites, twins had an intra-pair variability index of average 0.06 ± 0.03, whereas controls showed an increased index of 0.08 ± 0.03 (p = 0.006). For the individual metabolites, there were significant differences (p b 0.05) in the intra-pair variability for temporal lobe NAA in both the right and left temporal lobes with less variability for that metabolite in twins compared to controls. On the right side, the twin index was 0.04 ± 0.04 whereas the control index was 0.08 ± 0.06 (p = 0.02), and on the left side, the twin index was 0.06 ± 0.04 and the control index 0.08 ± 0.06 (p = 0.05).

Table 3 Analysis of the symmetry of the similarity measure in twins.

VBM (GM) VBM (WM) DTI (FA) DTI (MD) VBR

p = 10− 2

p = 10− 4

p = 10− 5

0.01 0.08 − 0.02 − 0.06 0.06

− 0.05 0.44 − 0.1 − 0.03 − 0.18

− 0.07 0.52 − 0.07 0.07 0.0

Values of the laterality index parameter describing the degree of symmetry of similarity in each group. This measure is calculated as (L − R) / (L + R), where L and R are the counts of suprathreshold voxels (i.e., voxels that show improved similarity in twins over controls) in the left and right hemispheres, respectively. Three alternative p-value thresholds were used for the counting of the suprathreshold voxels (p = 10− 2, 10− 4 and 10− 5).

1542

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

Table 4 Analysis of MR spectroscopy (MRS) data.

LTL Cr Cho mI NAA + NAAG Glu + Gln RTL Cr Cho mI NAA + NAAG Glu + Gln LFL Cr Cho mI NAA + NAAG Glu + Gln RFL Cr Cho mI NAA + NAAG Glu + Gln

Control 1

Control 2

Intra-pair variability index

Twin 1

Twin 2

Intra-pair variability index

5.2 ± 0.6 1.7 ± 0.2 4.2 ± 1.0 7.8 ± 0.9 7.3 ± 1.6

5.5 ± 1. 1.8 ± 0. 4.5 ± 0. 7.5 ± 1. 7.9 ± 1.

0 2 9 3 6

0.09 ± 0.06 0.07 ± 0.06 0.11 ± 0.07 0.08 ± 0.06 0.10 ± 0.07

5.0 ± 0.5 1.7 ± 0.2 4.2 ± 0.8 7.3 ± 0.9 7.7 ± 0.9

5.0 ± 0.6 1.7 ± 0.2 4.1 ± 0.9 7.1 ± 0.7 7.1 ± 0.9

0.06 ± 0.04 0.06 ± 0.04 0.10 ± 0.06 0.06 ± 0.04⁎ 0.06 ± 0.06

5.2 ± 0.8 1.6 ± 0.2 4.1 ± 1.0 7.0 ± 0.9 7.8 ± 1.3

5.4 ± 1. 1.8 ± 0. 4.4 ± 1. 7.3 ± 1. 8.6 ± 1.

0 2 2 2 5

0.07 ± 0.05 0.08 ± 0.05 0.12 ± 0.06 0.08 ± 0.06 0.13 ± 0.08

4.8 ± 0.4 1.6 ± 0.2 4.1 ± 0.9 6.6 ± 0.7 7.5 ± 1.7

4.9 ± 0.6 1.6 ± 0.3 4.0 ± 1.0 6.8 ± 0.8 7.3 ± 1.2

0.06 ± 0.04 0.07 ± 0.05 0.11 ± 0.10 0.04 ± 0.04⁎ 0.11 ± 0.10

5.4 ± 0.4 1.6 ± 0.2 4.1 ± 0.8 8.1 ± 0.9 8.7 ± 1.1

5.3 ± 0. 1.7 ± 0. 3.7 ± 1. 8.1 ± 0. 8.7 ± 1.

4 2 0 8 5

0.03 ± 0.02 0.06 ± 0.05 0.09 ± 0.08 0.00 ± 0.07 0.08 ± 0.07

5.2 ± 0.5 1.6 ± 0.2 3.7 ± 0.7 8.3 ± 0.5 8.3 ± 1.6

5.4 ± 0.4 1.6 ± 0.3 3.7 ± 0.6 8.2 ± 0.7 9.4 ± 1.3

0.03 ± 0.03 0.04 ± 0.03 0.07 ± 0.06 0.03 ± 0.03 0.09 ± 0.06

5.1 ± 0.3 1.6 ± 0.2 3.8 ± 0.6 7.7 ± 0.6 8.7 ± 1.2

5.0 ± 0. 1.6 ± 0. 3.6 ± 0. 7.5 ± 0. 8.4 ± 1.

3 2 7 9 2

0.04 ± 0.03 0.06 ± 0.04 0.09 ± 0.09 0.05 ± 0.04 0.08 ± 0.06

5.1 ± 0.3 1.6 ± 0.2 3.8 ± 0.6 7.6 ± 0.6 8.8 ± 1.6

4.9 ± 0.4 1.5 ± 0.2 3.8 ± 0.5 7.8 ± 0.7 8.4 ± 1.0

0.04 ± 0.03 0.05 ± 0.03 0.07 ± 0.05 0.04 ± 0.03 0.09 ± 0.05

Values of the intra-pair variability index are shown as mean ± standard deviation. Metabolites expressed as mmol/l. The following metabolites were quantified: N-acetylaspartate/ N-acetylaspartylglutamate (NAA), choline (Cho), creatine (Cr), myoinositol (mI) and glutamine/glutamate (Glx). Voxels were positioned in left/right temporal lobes (LTL/RTL) and frontal lobes (LFL/RFL). ⁎ Indicates significant difference (p b 0.05) in the intra-pair variability index between the twins and controls pairs.

Discussion In the past decade, the growth in genetics and neuroimaging has been dramatic. Structural brain imaging studies of twins lie at the interface of this effort. The benefit of twins is related to the extent to which variance in the parameters of interest is reduced by the control of genetic variation. This study indicates the degree of benefit of using twin pairs and the areas of the brain in which the magnitude of this effect is greatest for a range of MRI structural parameters. Using voxelbased analysis, we found a greater degree of structural similarity between twin pairs than between matched control pairs. The spatial extent of this increased structural similarity was largest for MR diffusion, followed by morphometry, and was least for T2 and MRS. The reduced variance in the twin pairs translates to a reduction in required subject numbers of approximately 40% for volume studies and 25% for DTI measures. Many studies have confirmed that monozygous twins display considerable structural similarities when compared to unrelated pairs (for example, Biondi et al., 1998). However, monozygous twin brains are not totally “identical” (for example, Bartley et al., 1997) and intertwin differences lie on a continuum with some features under tight genetic control and others having a greater potential for environmental influences. The use of the voxel-based approach to the study of volume changes has already been implemented in a limited number of twin studies of changes in certain disease and altered behavioral states (schizophrenia: Hulshoff Pol et al., 2006a; attention disorder: van 't Ent et al., 2007; anxiety and depression: de Geus et al., 2007; Rett syndrome: Carter et al., 2008; post-traumatic stress syndrome: Kasai et al., 2008; Alzheimer's disease: Virta et al., 2009), and it is likely that this trend will grow. A suitable design is comparison of discordant pairs using a paired t-test analysis to locate abnormal areas likely related to disease-specific effects and the comparison of discordant pairs to healthy twins or control singletons to locate areas related to genetic and environmental similarity. The voxel-based approach is a fairly coarse method for the comparison of groups of images as it relies on a fairly low resolution

spatial normalization and smoothing to prepare the images to a suitable state in which they can be compared using simple statistical techniques. It has been subject to a number of criticisms (for example, Bookstein, 2001) but if implemented and described correctly (Ridgway et al., 2008), it is a powerful technique for group comparisons of imaging data. It has the advantage of being simple to implement and allows an unbiased search of the image space for areas of abnormality. It hence offers benefits over more labor intensive approaches such as region-of-interest (ROI) drawing and sophisticated image processing methods. Analysis of the T2 relaxation time did highlight one of the limitations of the voxel-based approach. A high variability across the brain is apparent in the T2 difference maps which is largely due to large dynamic range of T2 values across tissue and CSF compartments (i.e., T2(CSF) NN T2(tissue)). This affect is exaggerated when pairing up images in the manner utilized in this study and shows up as an increase in variance especially at the boundaries between tissue types and the edge of the brain. This strongly influences the anomalous global variance measures between twins and control pairs for T2 shown in Table 1. Reports of the degree of cerebral asymmetry of heritability of brain structure are variable and appear to be sensitive to the choice of volume index with findings of no asymmetry (Wright et al., 2002) and the left hemisphere under greater genetic control (Tramo et al., 1995; Lohmann et al., 1999; specifically language areas in Thompson et al., 2001) or environmental (Geschwind et al., 2002; Carmelli et al., 2002) control. This can be contrasted with our findings of a rightward asymmetry of GM similarity and a leftward asymmetry of WM similarity in twins. This tissue-specific sensitivity requires further investigation. The extended twin model (Posthuma and Boomsma, 2000) enables the decomposition of variance of a phenotype into genetic and non-genetic environmental (common and unique) sources using analysis approaches such as the Falconer method and structural equations modeling (SEM). This approach could not be undertaken in our study since a cohort of DZ twins was not readily available. It

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

should also be noted that in a simple design comparing MZ twins to controls, genetic inferences can be made but with the understanding that shared environmental factors cannot be distinguished from genetic factors. A large body of literature has built up confirming the high heritability of overall human brain volume (for a review, see Peper et al., 2007). Most such studies have used relatively small groups of monozygous pairs although some have implemented the full-twin design (Pennington et al., 2000; Thompson et al., 2001; Hulshoff Pol et al., 2002b). Most commonly, global volume measures have been used (such as gray matter volume of whole brain or individual lobes). More recently, a number of studies have investigated focal patterns of heritability using approaches such as factor analysis (Pennington et al., 2000), path analysis (Wright et al., 2002), cortical mapping (Thompson et al., 2001), voxel-based SEM (Hulshoff Pol et al., 2006b) and automated subcortical segmentation (White et al., 2002). To summarize, a high degree of genetic involvement has been found for measures of total brain volume as well as gray matter and white matter volume. On a localized level, genetic effects have been shown to be regionally variable within the brain. High heritabilities have been found in areas including middle frontal, sensorimotor and superior and anterior temporal cortices including Wernicke's region (Thompson et al., 2001), in paralimbic structures and parietal neocortical areas (Wright et al., 2002), posterior cingulate, occipital gray matter and the connecting white matter of the superior occipitofrontal fasciculus (Hulshoff Pol et al., 2006b) and in the corpus callosum (Pfefferbaum et al., 2000). Moderate heritabilities have been reported in the hippocampus (Sullivan et al., 2001) and cerebellum (Wallace et al., 2002). General gyral patterning has been shown to have low genetic involvement. Unique environmental rather than genetic factors were shown to influence surface morphology (gyral patterns) as well as several medial brain areas such as lateral ventricles (Wright et al., 2002). Our findings of the spatial distribution of structural similarity in monozygous twins are largely in agreement with these findings but provide this information using a simple analysis method and across a range of MR parameters. In particular, our findings support the strong genetic influence in the structure of the frontal and temporal cortices (see Fig. 1). This needs to be taken into consideration in twin studies as the sensitivity to detection of pathological differences will be increased in these areas relative to others. The MRS results demonstrate a trend for reduced variability in twins in some of the measured metabolite concentrations. This effect was particularly prominent for temporal lobe NAA which is a metabolite that reflects neuronal integrity. Other temporal lobe metabolites and the frontal lobe metabolites suggest a similar but non-significant trend. While twin brains are not completely “identical”, we have demonstrated that variance in the MR parameters of diffusion, volume T2 and MRS are all under some degree of genetic control. This study has shown the distribution and magnitude of the reduced variability between identical twins. The data suggest that evaluation of twins discordant for disease will be useful for detecting subtle disease-specific effects using a range of MR parameters. Knowing the variance in the MR measure of interest and how this is genetically influenced, is fundamental to understanding the benefit of using twins in studies in both health and disease where aspects of brain structure are the phenotype of interest. Acknowledgments The authors thank the help of Professor John L. Hopper, director of the Australian Twin Registry. This work was funded by the National Health and Medical Research Council (NHMRC).

1543

References Andel, R., Crowe, M., Pedersen, N.L., Mortimer, J., Crimmins, E., Johansson, B., Gatz, M., 2005. Complexity of work and risk of Alzheimer's disease: a population-based study of Swedish twins. J. Gerontol. B Psychol. Sci. Soc. Sci. 60, 251–258. Ashburner, J., Friston, K.J., 1997. Voxel based morphometry—the methods. NeuroImage 11, 805–821. Bartley, A.J., Jones, D.W., Weinberger, D.R., 1997. Genetic variability of human brain size and cortical gyral patterns. Brain 120, 257–269. Biondi, A., Nogueira, H., Dormont, D., Duyme, M., Hasboun, D., Zouaoui, A., Chantôme, M, Marsault, C., 1998. Are the brains of monozygotic twins similar? A threedimensional MR study. Am. J. Neuroradiol. 19, 1361–1367. Bookstein, F.L., 2001. “Voxel-based morphometry” should not be used with imperfectly registered images. Neuroimage 14, 1454–1462. Carmelli, D., Swan, G.E., DeCarli, C., Reed, T., 2002. Quantitative genetic modeling of regional brain volumes and cognitive performance in older male twins. Biol. Psych. 61, 139–155. Carter, J.C., Lanham, D.C., Pham, D., Bibat, G., Naidu, S., Kaufmann, W.E., 2008. Selective cerebral volume reduction in Rett syndrome: a multiple-approach MR imaging study. Am. J. Neuroradiol. 29, 436–441. Davatzikos, C., Vaillant, M., Resnick, S.M., Prince, J.L., Letovsky, S., Bryan, R.N., 1996. A computerized approach for morphological analysis of the corpus callosum. J. Comput. Assist. Tomogr. 20, 88–97. de Geus, E.J., van't Ent, D., Wolfensberger, S.P., Heutink, P., Hoogendijk, W.J., Boomsma, D.I., Veltman, D.J., 2007. Intrapair differences in hippocampal volume in monozygotic twins discordant for the risk for anxiety and depression. Biol. Psych. 61, 1062–1071. Fisher, R.A., 1921. On the probable error of a coefficient of correlation deduced from a small sample. Metron 1, 3–32. Geschwind, D.H., Miller, B.L., DeCarli, C., Carmelli, D., 2002. Heritability of lobar brain volumes in twins supports genetic models of cerebral laterality and handedness. Proc. Natl. Acad. Sci. U.S.A. 99, 3176–3181. Hulshoff Pol, H.E., Schnack, H.G., Bertens, M.G., van Haren, N.E., van der Tweel, I., Staal, W.G., Baare, W.F., Kahn, R.S., 2002a. Volume changes in gray matter in patients with schizophrenia. Am. J. Psych. 159, 244–250. Hulshoff Pol, H.E., Posthumam, D., Baare, W.F., de Geus, E.J., Schnack, H.G., van Haren, N.E., van Oel, C.J., Kahn, R.S., Boomsma, D.I., 2002b. Twin-singleton differences in brain structure using structural equation modelling. Brain 125, 384–390. Hulshoff Pol, H.E., Schnack Mandl, R.C., Brans, R.G., van Haren, N.E., Baaré, W.F., van Oel, C.J., Collins, D.L., Evans, A.C., Kahn, R.S., 2006a. Gray and white matter density changes in monozygotic and same-sex dizygotic twins discordant for schizophrenia using voxel-based morphometry. Neuroimage 31, 482–488. Hulshoff Pol, H.E., Schnack, H.G., Posthuma, D., Mandl, R.C., Baaré, W.F., van Oel, C., van Haren, N.E., Collins, L., Evans, A.C., Amunts, K., Burgel, U, Zilles, K., de Geus, E.J, Boomsma, D.I., Kahn, R.S., 2006b. Genetic contributions to human brain morphology and intelligence. J. Neurosci. 26, 10235–10242. Hyde, T.M., Stacey, M.E., Coppola, R., Handel, S.F., Rickler, K.C., Weinberger, D.R., 1995. Cerebral morphometric abnormalities in Tourette's syndrome: a quantitative MRI study of monozygotic twins. Neurology 45, 1176–1182. Kasai, K., Yamasue, H., Gilbertson, M.W., Shenton, M.E., Rauch, S.L., Pitman, R.K., 2008. Evidence for acquired pregenual anterior cingulate gray matter loss from a twin study of combat-related posttraumatic stress disorder. Biol. Psychiatry 63, 550–556. Kates, W.R., Burnette, C.P., Eliez, S., Strunge, L.A., Kaplan, D., Landa, R., Reiss, A.L., Pearlson, G.D., 2004. Neuroanatomic variation in monozygotic twin pairs discordant for the narrow phenotype for autism. Am. J. Psych. 161, 539–546. Keller, S.S., Wilke, M., Wieshmann, U.C., Sluming, V.A., Roberts, N., 2004. Comparison of standard and optimized voxel-based morphometry for analysis of brain changes associated with temporal lobe epilepsy. NeuroImage 23, 860–868. Lohmann, G., von Cramon, D.Y., Steinmetz, H, 1999. Sulcal variability of twins. Cereb. Cortex 9, 754–763. Lutkenhoff, E.S., van Erp, T.G., Thomas, M.A., Therman, S., Manninen, M., Huttunen, M.O., Kaprio, J., Lönnqvist, J., O'Neill, J., Cannon, T.D., 2008. Proton MRS in twin pairs discordant for schizophrenia. Mol. Psych. [Electronic publication ahead of print]. Luxenberg, J.S., May, C., Haxby, J.V., Grady, C., Moore, A., Berg, G., White, B.J., Robinette, D., Rapoport, S.I., 1987. Cerebral metabolism, anatomy, and cognition in monozygotic twins discordant for dementia of the Alzheimer type. J. Neurol. Neurosurg. Psych. 50, 333–340. Narr, K.L., van Erp, T.G., Cannon, T.D., Woods, R.P., Thompson, P.M., Jang, S., Blanton, R., Poutanen, V.P., Huttunen, M., Lönnqvist, J., Standerksjöld-Nordenstam, C.G., Kaprio, J., Mazziotta, J.C, Toga, A.W., 2002. A twin study of genetic contributions to hippocampal morphology in schizophrenia. Neurobiol. Dis. 11, 83–95. Pell, G.S., Briellmann, R.S., Waites, A.B., Abbott, D.F., Jackson, G.D., 2004. Voxel-based relaxometry: a new approach for analysis of T2 relaxometry changes in epilepsy. Neuroimage 21, 707–713. Pell, G.S., Briellmann, R.S., Waites, A.B., Abbott, D.F., Lewis, D.P., Jackson, G.D., 2006a. Optimized clinical T2 relaxometry with a standard CPMG sequence. J. Magn. Reson. Imaging. 23, 248–252. Pell, G.S., Briellmann, R.S., Pardoe, H., Abbott, D.F., Jackson, G.D., 2006b. Sensitivity of the voxel-based analysis of diffusion to the warping strategy. “Proc., ISMRM 14th Annual Meeting, Seattle”, #1062. Pennington, B.F., Filipek, P.A., Lefly, D., Chabildas, N., Kennedy, D.N., Simon, J.H., Filley, C.M., Galaburda, A., DeFries, J.C., 2000. A twin MRI study of size variations in human brain. J. Cogn. Neurosci. 12, 223–232. Peper, P.S., Brouwer, R.M., Boomsma, D.I., Kahn, R.S., Hulshoff Pol, H.E., 2007. Genetic influences on human brain structure: a review of brain imaging studies in twins. Hum. Brain Mapp. 28, 464–473.

1544

G.S. Pell et al. / NeuroImage 49 (2010) 1536–1544

Pfefferbaum, A., Sullivan, E.V., Swan, G.E., Carmelli, D., 2000. Brain structure in men remains highly heritable in the seventh and eighth decades of life. Neurobiol. Aging 21, 63–74. Posthuma, D., Boomsma, D.I., 2000. A note on the statistical power in extended twin designs. Behav. Genet. 30, 147–158. Provencher, S.W., 1993. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn. Reson. Med. 30, 672. Reiersen, A.M., Constantino, J.N., Grimmer, M., Martin, N.G., Todd, R.D., 2008. Evidence for shared genetic influences on self-reported ADHD and autistic symptoms in young adult Australian twins. Twin Res. Hum. Genet. 11, 579–585. Ridgway, G.R., Henley, S.M., Rohrer, J.D., Scahill, R.I., Warren, J.D., Fox, N.C., 2008. Ten simple rules for reporting voxel-based morphometry studies. Neuroimage 40, 1429–1435. Styner, M., Lieberman, J.A., McClure, R.K., Weinberger, D.R., Jones, D.W., Gerig, G., 2005. Morphometric analysis of lateral ventricles in schizophrenia and healthy controls regarding genetic and disease-specific factors. Proc. Natl. Acad. Sci. U. S. A. 29, 4827–4872. Sullivan, E.V., Pfefferbaum, A., Swan, G.E., Carmelli, D., 2001. Heritability of hippocampal size in elderly twin men: equivalent influence from genes and environment. Hippocampus 11, 754–762. Thompson, P.M., Cannon, T.D., Narr, K.L., van Erp, T., Poutanen, V.P., Huttunen, M., Lonnqvist, J., Standertskjold-Nordenstam, C.G., Kaprio, J., Khaledy, M., Dail, R., Zoumalan, C.I., Toga, A.W., 2001. Genetic influences on brain structure. Nat. Neurosci. 4, 1253–1258.

Tramo, M.J., Loftus, W.C., Thomas, C.E., Green, R.L., Mott, L.A., Gazzaniga, M.S., 1995. Surface area of human cerebral cortex and its gross morphological subdivisions: in vivo measurements in monozygotic twins suggest differential hemisphere effects of genetic factors. J. Cogn. Neurosci. 7, 292–301. van 't Ent, D., Lehn, H., Derks, E.M., Hudziak, J.J., Van Strien, N.M., Veltman, D.J., De Geus, E.J.C., Todd, R.D., Boomsmaa, D.I., 2007. A structural MRI study in monozygotic twins concordant or discordant for attention/hyperactivity problems: evidence for genetic and environmental heterogeneity in the developing brain. NeuroImage 35, 1004–1020. Virta, J.J., Aalto, S., Järvenpää, T., Karrasch, M., Kaprio, J., Koskenvuo, M., Räihä, I., Viljanen, T., Rinne, J.O., 2009. Voxel-based analysis of cerebral glucose metabolism in mono- and dizygotic twins discordant for Alzheimer disease. Neurol. Neurosurg. Psych. 80, 259–266. Wallace, G.L., Eric, S.J., Lenroot, R., Viding, E., Ordaz, S., Rosenthal, M.A., White, T., Andreasen, N.C., Nopoulos, P., 2002. Brain volumes and surface morphology in monozygotic twins. Cereb. Cortex 12, 486–493. White, T., Andreasen, N.C., Nopoulos, P., 2002. Brain volumes and surface morphology in monozygotic twins. Cereb. Cortex 12, 486–493. Wright, I.C., Sham, P., Murray, R.M., Weinberger, D.R., Bullmore, E.T., 2002. Genetic contributions to regional variability in human brain structure: methods and preliminary results. Neuroimage 17, 256–271.