RE: Familial aggregation of clinical and neurocognitive features in sibling pairs with and without schizophrenia

Abstract

The recent study by Chen et al. (2009) adds to the emerging body of research providing formal estimates of the heritability of neurocognitive traits. There are now at least seven published studies in the literature, using heterogeneous samples of schizophrenia and various neurocognitive tests, covariates and methods (Glahn et al., 2007; Greenwood et al., 2007; Gur et al., 2007a,b; Husted et al., 2009; Toulopoulou et al., 2007; Tuulio-Henriksson et al., 2002). Although the findings of Chen et al. (2009), using subjects aged 9–31 years, demonstrate many similarities with the existing evidence, they also reveal some of the challenges facing this research area. Using Sequential Oligogenic Linkage Analysis Routines (SOLAR), the reported heritability estimates for four neurocognitive domains were only modest to moderate in the schizophrenia proband-unaffected sibling pairs, ranging from 0.00 to 0.43, consistent with the results of other studies. Notably, this is significantly less than comparable heritability estimates for diagnosis of schizophrenia (h2=0.81) (Toulopoulou et al., 2007). The magnitude of the Chen et al. (2009) heritability estimates for neurocognitive traits was altered by the presence of schizophrenia; an overall pattern of lower heritability estimates in the schizophrenia-sibling pairs relative to the control sibling pairs. In our recent study, using a familial form of schizophrenia associated with the NOSAP1 gene on chromosome 1 (Husted et al., 2009), we also reported modest to moderate heritability estimates (h2=0.05–0.49) for 12 neurocognitive measures, after adjusting for age and sex. However, SOLAR analyses showed that when we included diagnosis of schizophrenia as a covariate, heritability estimates either increased or decreased, depending upon the specific neurocognitive measure examined. It is clear that the expression of schizophrenia has important effects on heritability estimates and should be taken into account in these studies. Another factor that can affect these heritability estimates, not addressed by Chen et al. (2009), is general intellect. In a large sample of twin pairs, Toulopoulou et al. (2007) estimated the heritabilities of full-scale IQ and four neurocognitive domains. IQ was found to be the most highly heritable of the neurocognitive measures (h2=0.70). Lowered IQ scores in monozygotic twins discordant for schizophrenia suggested that it is the genetic vulnerability to schizophrenia, rather than the expression of schizophrenia, that reduces IQ. The domain of working memory was also highly heritable (h2=0.65). Shared genetic variance accounted for the majority of the covariance between schizophrenia and both IQ and working memory. Three other cognitive domains (processing speed, perceptual organization and verbal comprehension) were found to be moderately heritable, with a smaller portion of the observed covariance between each domain and schizophrenia genetically shared. In our recent study (Husted et al., 2009), we observed similar additive heritability for full-scale IQ, adjusting for age, sex and schizophrenia (h2=0.64). Despite the fact that IQ is a composite construct comprised of several components that overlap with a number of cognitive domains, covarying for such a measure of overall intellect when examining the heritability of a neurocognitive test can reveal interesting relationships. Our results showed that there were noteworthy differences in heritability estimates of individual neurocognitive tests in SOLAR analyses that included IQ as a covariate compared with those that did not. We reported that these differences could in part be explained by the observed bivariate genetic correlations between IQ and the individual neurocognitive tests. The bivariate correlations between IQ and our measures of working, auditory and verbal memory were very high (0.84 to 1.00), whereas the bivariate correlations between IQ and measures of processing speed were negative (−0.10 to −0.34). Indeed, we found that accounting for IQ helped sort out the bivariate genetic correlations between the neurocognitive measures and helped to identify two heritable components: memory-IQ and visuomotor-processing speed. Our results further suggest that the relationships between the heritability of cognitive domains may differ from the relationships involved in test performance itself. Like Toulopoulou et al. (2007), we believe that considering general intellect in future studies may aid in the identification of heritable neurocognitive traits related to the expression of schizophrenia. Finally, Chen et al. (2009) incorporated a correction for ascertainment bias in their SOLAR models. This ascertainment correction may be inadequate. If the shapes of the score distributions for schizophrenic probands and siblings are very different from the distributions for unaffected individuals, the heritability estimates could be affected in ways that are hard to predict. Ascertainment correction is a very difficult issue (Vieland and Hodge 1996).