To date, multiple twin studies have been published that document the significant heritability of volume-based measures of brain structure. However, the present results clearly demonstrate that for global, lobar, and regional levels of analysis, measures of cortical volume combine at least 2 distinct sources of genetic effects, those influencing surface area and those influencing cortical thickness. We argue that the lack of a genetic correlation between these measures is consistent with the columnar organization of the cerebral cortex and the putative developmental origins of that organization as specified by the radial unit hypothesis (Rakic 1988
; Mountcastle 1997
Given these findings, it would appear that careful consideration needs to be made before utilizing any multidimensional measurement of the brain, especially in genetically informative studies. Measurements of gray matter volume conflate the contributions of thickness and surface area and therefore may not capture the basic structural elements of the cerebral cortex. The same is true for measures of gray matter density, defined as the proportion of gray matter per 3 dimensional imaging voxel (Wright et al. 1995
; Thompson et al. 2001
). Despite this, multidimensional measures have been proposed as possible endophenotypes for numerous neuropsychiatric disorders (Glahn et al. 2007
). An endophenotype is defined as a trait that is along the causal pathway from genetic predisposition to clinical outcome (Gottesman and Gould 2003
). Historically, the primary criterion for an endophenotype has been that it must be heritable, as has been repeatedly demonstrated for cortical volume (Glahn et al. 2007
; Peper et al. 2007
; Schmitt et al. 2007
). However, when a heritable trait has underlying components that are genetically distinct, it will be a less informative phenotype than the underlying components. Therefore, elucidating these distinct sets of genetic influences, when they do exist, will be a crucial step in identifying the key genetic factors that influence normal and pathological brain structure.
That surface area and thickness are genetically distinct from one another has numerous implications for continued investigations into the genetic influences of brain structure. Perhaps, the most significant of these is the need to explore the genetics of surface area to a greater degree. Like thickness, surface area is a highly heritable construct; yet, it has been largely overlooked in human imaging genetics research. Specific mutations in humans have been linked to excessive gyrification of the cortex as well as an increase in the cortical surface area (Piao et al. 2004
; Jansen and Andermann 2005
). Animal studies have also demonstrated that manipulation of specific genes can result in dramatic changes in arealization and expansion of select areas of the cerebral cortex like the primary visual area and the primary somatosensory area (Bishop et al. 2000
; Mallamaci et al. 2000
). Findings such as these would appear to suggest that the genes that influence surface area are critical to the early growth and development of the brain. The observed genetic relationship between total surface area and intracranial volume lends support to this conclusion. If this is the case, then a more focused examination of the genetics of surface area may produce new insights into disorders believed to have early developmental origins, such as schizophrenia.
In this same light, recent findings pertaining to the genetics of cortical thickness and cortical volume may have to be reconsidered. For instance, multivariate twin analyses of cortical thickness have demonstrated genetically mediated networks across the cortex in both pediatric (Schmitt et al. 2008
) and adult samples (Rimol LM, Panizzon MS, Fennema-Notestine C, Fischl B, Franz CE, Lyons MJ, Makris N, Neale MC, Pacheco J, Perry ME, Schmitt JE, Seidman L, Thermenos HW, Tsuang MT, Kremen WS, Dale AM, unpublished data). It remains to be seen if similar patterns of results will be obtained using measures of surface area. The genetic relationship between cognition and brain has been examined using measures of brain volume, cortical volume, or cortical density (Carmelli et al. 2002
; Posthuma et al. 2002
; Hulshoff Pol et al. 2006
), raising the question of which genetic influences are actually associated with cognition. Finally, the distinction between surface area and thickness will have to be addressed in genetic association studies. Defining structural phenotypes more precisely, taking into consideration the multiple latent genetic factors involved, will likely improve the ability of researchers to identify specific genes that are associated with structural differences.
Although the radial unit hypothesis suggests that the cortical surface area is influenced by the number of columns, whereas cortical thickness is influenced by the number of cells within a column (Rakic 1988
), it may be the case that the measures of surface area and thickness examined in this study reflect structural aspects other than the columnar organization of the cortex. For example, variations in cortical thickness could be due to differences in myelination of gray matter or the underlying white matter, rather than the number of cells within the column. Other explanations are also possible. Given that MRI measures do not have the resolution to examine brain structure at the cellular level, we cannot tell whether variation in thickness is due to different numbers of cells or the size of the cells.
We must acknowledge several potential limitations of this study. First, the all-male, relatively homogenous nature of this sample limits the degree to which results can be generalized to other populations. We cannot be certain that the genetic relationship between surface area and thickness would be similar in a sample of greater ethnic diversity or within a female twin cohort. Similarly, we must recognize that given the high degree of developmental change in gray and white matter during childhood and adolescence, these results may not apply to younger age cohorts (Giedd et al. 1999
; Jernigan et al. 2001
; Sowell et al. 2004
; Jernigan and Gamst 2005
). It remains an issue for future research to determine if the lack of genetic relationship observed in this middle-aged sample may be unique to this age period. The goal of the VETSA projects is to follow participants longitudinally, which may enable us to address this issue in the future.
It is also the case that while the VETSA represents a large MRI twin sample, our power to detect additive genetic effects and genetic correlations may be limited for some of the bivariate relationships. Although the power to detect genetic effects is substantially increased in multivariate twin designs, sufficient power to detect significant genetic correlations, in samples of approximately 200 twin pairs, is obtained when heritability estimates are approximately 40% (Sullivan and Eaves 2002
). Although we achieve this criterion for the majority of the bivariate analyses, there were select cases in which heritability estimates were lower than the ideal. Nevertheless, in these few cases, we were still able to demonstrate a lack of complete genetic overlap between surface area and thickness.