|Home | About | Journals | Submit | Contact Us | Français|
Neurocognitive impairment was found to be heritable in individuals with schizophrenia and their relatives. However, the heritability of neurocognitive measures in families with and without schizophrenia has not been directly compared. In this study, we examined the genetic structure of clinical and neurocognitive measures in sibling pairs with and without schizophrenia to test the hypothesis that the familial aggregation of such measures may be altered by having schizophrenia.
A total of 278 subjects including patients with schizophrenia and their non-psychotic full siblings, healthy controls, and their full siblings were recruited. Heritability was estimated for working memory, episodic memory, executive function and attention, as well as clinical features, such as positive, negative and disorganization symptoms.
Many clinical and cognitive domains were impaired in subjects with schizophrenia and their non psychotic siblings. Negative symptoms, working memory, episodic memory and executive function, but not positive, disorganization symptoms and attention, were found to be significantly heritable in all sibling pairs. However, the heritability of working memory function was significantly (χ2(d.f.=6)=13.9, p=.031) decreased in proband sibling pairs (h2=.38) as compared to control sibling pairs (h2=.95). Significant genetic correlations were observed between negative symptoms and the cluster of working memory, episodic memory and executive function.
Several neurocognitive measures were heritable in sibling pairs with and without schizophrenia. However, schizophrenia reduced the heritability of working memory, perhaps due to disease-related environmental or genetic factors. Evidence for potential pleiotropy will inform future phenotypic studies.
Cognitive impairment has been suggested as a schizophrenia-related endophenotype (Mohamed et al. 1999; Gottesman and Gould 2003; Greenwood et al. 2007). Prior research suggests that neurocognitive measures may qualify as schizophrenia-related endophenotypes (Goldberg et al. 1990; Dollfus et al. 2002; Toulopoulou et al. 2007; Wedenoja et al. 2008). For example, impairments in working memory, episodic memory, and executive function are present in patients with schizophrenia across all phases of illness, and have been observed in the unaffected relatives of patients with schizophrenia (e.g., Toulopoulou et al. 2003; Snitz et al. 2006). Furthermore, variation in quantifiable traits such as working memory, have been associated with specific allelic variants (Wedenoja et al. 2008).
Heritability in the general population is an important criterion for a disease-related endophenotype, and there is much evidence to support the heritability of neurocognition. Heritability for working memory has been estimated to be 33–64% (Ando et al. 2001; Wright et al. 2001; Glahn et al. 2007; Greenwood et al. 2007; Kremen et al. 2007). Genetic studies on episodic memory and executive function have shown a wide range of heritability from near 0 to 49% (Johansson et al. 1999; Cannon et al. 2000; Polderman et al. 2006; Greenwood et al. 2007; Taylor 2007). Recent results from a multi-site study of neurocognition, the Consortium on the Genetics of Schizophrenia, have added support to the hypothesis that neurocognition is a schizophrenia-related endophenotype (Gur et al. 2007), but suggest that both genetic and environmental factors can influence neurocognitive performance in patients with schizophrenia and their relatives (Greenwood et al. 2007).
Very few studies have directly compared the heritability of neurocognitive measures in sibling pairs with and without schizophrenia. This comparison is important, as environmental factors associated with the disease process of schizophrenia could reduce the heritability of potential phenotypes, while genetic influences on the illness might actually increase heritability of such characteristics. In a recent study of the relatives of patients with schizophrenia, the heritability of neurocognitive measures was estimated in the full sample and in a subsample that excluded affected individuals with schizophrenia (Glahn et al. 2007). The results suggested that heritability estimates were increased or decreased by the inclusion of affected individuals, depending on the measure examined. However, this study did not allow the comparison of heritability in families with and without schizophrenia. Cannon and colleagues used a neuropsychological battery to examine disease versus liability predictors in twin pairs discordant for schizophrenia and control twin pairs (Cannon et al. 2000). For the disease predictors among MZ twins, they reported numerically lower familiality\Intra-Class Correlations (ICCs) in proband twin pairs than in control twin pairs, suggesting that the disease process of schizophrenia could alter the ICC of some measures such as episodic memory.
In the current study, we characterized the heritability structure of clinical and neurocognitive measures in sibling pairs with and without schizophrenia and tested the hypothesis that the familial aggregation of these domains was influenced by the presence of the disorder. We first determined which neurocognitive deficits were present in the siblings of subjects with schizophrenia as compared to the siblings of controls (Delawalla et al. 2006). Next, we examined the familiality of these measures by estimating their heritability in all sibling pairs, control sibling pairs and proband sibling pairs with schizophrenia. The availability of sibling pairs with and without schizophrenia provided an unique opportunity to both estimate the heritability of neurocognitive measures in the non-diseased population and to determine how this heritability might be altered by schizophrenia. Such alterations might be due to the presence of shared disease vulnerability, the development of the disease in one sibling, or non-shared environmental factors associated with treatment for the disease. We also tested the hypothesis that pleiotropy (evidence of shared genes for multiple traits) could link clinical and neurocognitive features in all sibling pairs.
The subjects for this study were recruited through the Conte Center for the Neuroscience of Mental Disorders (CCNMD) at Washington University School of Medicine in St. Louis, and included: (1) probands with DSM-IV schizophrenia (PRO) (n=56); (2) their full non-psychotic siblings (PRO-SIB) (n=65); (3) healthy control participants (CON) (n=77); and (4) their siblings (CON-SIB) (n=80). All subjects gave written informed consent for participation following a complete description of the risks and benefits of participating in the study.
All subjects were diagnosed using DSM-IV criteria on the basis of a consensus between a research psychiatrist who conducted a semi-structured interview and a trained research assistant who used the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I/P) (First et al. 2001). Participants (in any group) were excluded if they: (a) met DSM-IV criteria for substance dependence or severe/moderate abuse during the 6 months preceding assessment; (b) had a clinically unstable or severe medical disorder, or a medical disorder that confounded the assessment of psychiatric diagnosis or rendered research participation dangerous; (c) had a history of head injury with documented neurological sequelae or loss of consciousness; or (d) met DSM-IV criteria for mental retardation (mild or greater in severity).
Probands were recruited from local inpatient and outpatient facilities, and were stabilized on antipsychotic medication for at least two weeks before participating in the study. Controls were recruited using local advertisements in the same community, and were required to have no lifetime history of Axis I psychotic or major mood disorders and no first-degree relatives with a psychotic disorder. CON-SIB subjects were enrolled in an identical manner to PRO-SIB subjects, and met the same general and specific inclusion and exclusion criteria. However, potential proband siblings were excluded if they had a lifetime history of any DSM-IV Axis I psychotic disorder, but not other DSM-IV Axis I disorders.
Psychopathology and cognitive function were assessed in a manner as described in Delawalla et al. (2006). Please see Table 1 (footnote) for detailed description of individual tests. Briefly, psychopathology was assessed using the Scale for the Assessment of Negative Symptoms (SANS), the Scale for the Assessment of Positive Symptoms (SAPS) (Andreasen et al., 1995), the Structured Interview for Prodromal Syndromes (SIPS) (Miller et al., 1999), and the Chapman Psychosis Proneness Scales (Chapman et al., 1995). Neurocognition was assessed using a battery of neuropsychological tests. The raw scores from the individual neuropsychological tests were first standardized by z-scores using the means and standard deviations computed across all subjects who have participated in research studies at the CCNMD at Washington University, and selected clusters of standardized z-scores were then averaged to yield four cognitive domains – working memory, episodic memory, executive function, attention (individual tests in Table 1). The reliability estimates for domain scores and individual tests are shown in Supplemental Table 1.
We compared age, gender, race, and years of education across four subject groups using a mixed-model analysis of variance (i.e., PROC MIXED in SAS 9.1; SAS Institute Inc., Cary, NC) to account for the familial correlations of siblings within the same family, since siblings members are not independent observations. F statistics or Pearson χ2 were used to examine the differences among all four subject groups. We modeled the associations of each cognitive domain with covariates (age, gender, indicator for having schizophrenia, indicator for having sibling with schizophrenia) using random effects regression models.
We estimated heritability (h2) using a variance components analysis as implemented in the Sequential Oligogenic Linkage Analysis Routines (SOLAR) software program (version 4.0) (Almasy and Blangero 1998). The total trait variance was partitioned into two components: (a) a component due to additive polygenic effects, and (b) a component due to the effects of random environmental factors that are uncorrelated among siblings. This method applied maximum likelihood estimation to a mixed effects model that incorporated fixed effects for known covariates and variance components for genetic and random environmental effects. No shared environmental influence was assumed (cite McGue M 1983). Age and gender were included as covariates. A correction for ascertainment bias was implemented in SOLAR involving conditioning on the trait values of the probands under the assumption that the probands are randomly selected. Since the neurocognitive domains were correlated, we computed bivariate genetic and environmental correlations using SOLAR to examine possible pleiotropy where a single gene or set of genes affect more than two traits simultaneously. The bivariate genetic correlation represents the percentage of variance that is common in the two traits due to the same genes. We estimated h2 of all neurocognitive endophenotypes in all sibling pairs. To compare heritability in the two types of sibpairs, we first estimated h2 in control sibling pairs, h2 in proband sibling pairs, and h2 in all sibling pairs. A log likelihood heterogeneity test was used to examine if the derived heritability estimates differed across the two types of sibling pairs by comparing the likelihood estimates in these 3 models. The difference of −2 Log Likelihood of these models follows a χ2 distribution with degree of freedom being the difference of parameters estimated in the comparison (Rice et al. 1991).
Demographic information for the four subject groups is summarized in Table 1. The age distribution of our sample is from 9 to 31 with mean=21.1, standard error 0.26, standard deviation 4.25. The lowest to highest quartile are 19 and 24 years of age. The PRO group showed higher clinical symptoms and deficits in functioning on all neurocognitive domains when compared to the CON and CON-SIB groups (Table 2). In turn, the PRO-SIB group showed attenuated, but significant, increases in negative and disorganization symptoms but not positive symptoms, and significant deficits in working memory, episodic memory and executive function, but not attention. These data showed that having schizophrenia was associated with global neurocognitive impairment, while having the genetic vulnerability of schizophrenia was associated with more selective clinical and neurocognitive impairments. See Supplemental Table 2 for group comparisons for the individual tests within each domain.
Next, we computed the between-sibling correlations in neurocognitive domains. We found several different patterns in terms of between-sibling correlations in relationship to the group differences. Similar group differences were present for working memory, episodic memory, and executive function (see Fig 1A). For working memory and executive function (see Fig 1B/C), the moderate correlations observed in the CON/CON-SIB pairs (r=.48, p<.001; r=.38, p=.002) were decreased and not significant in the PRO/PRO-SIB pairs (r=.16, p=.25; r=.26, p=.064). However, the correlations for episodic memory (see Fig 1D) were similar in both sibpair types (r=.30, p=.017 in CON/CON-SIB pairs, r=.39, p=.004 in PRO/PRO-SIB pairs). No significant correlations were observed for attention (see Fig 1E) in either type of sibling pairs (r=.046, p=.76 in CON/CON-SIB pairs; r=.011, p=.96 in PRO/PRO-SIB pairs). This indicates that between-sibling correlations of neurocognitive measures can vary with the sibpair type (e.g., those with or without schizophrenia).
The heritability estimates for the clinical and neurocognitive domain measures are summarized in Table 3. Negative symptoms, but not positive or disorganization symptoms, were heritable. Working memory, episodic memory and executive function were all significantly heritable. Attention was found to be not heritable. Supplemental Table 3 shows the heritability estimates for individual neurocognitive tests within each domain.
Since the between-sibling correlations for some neurocognitive domains varied according to sibpair type (Figure 1), we estimated heritabilities separately for each type of sibling pairs. We observed a general pattern of higher heritability estimates in the CON/CON-SIB pairs and lower heritability estimates in the PRO/PRO-SIB pairs. The heritability estimates were numerically attenuated for negative symptoms, working memory, episodic memory, and executive function among PRO/PRO-SIB pairs as compared to CON/CON-SIB pairs. The heterogeneity log likelihood χ2 test of the difference between the two subgroups showed a statistically significant difference (χ2= 13.92 with 6 d.f., p=0.031) for working memory. The attenuation in the heritability of negative symptoms, episodic memory and executive function among proband siblings did not reach statistical significance with the heterogeneity test, although the numerical patterns were in the same directions as for working memory. Comparisons of heritabilities for individual tests within domain measures are shown in Supplemental Table 2.
To test the hypothesis that at least some of the neurocognitive endophenotypic and clinical domain measures are linked construct (i.e., pleiotropy), we examined the correlations between the clinical and neurocognitive domain measures in all sibling pairs. Table 4(a) shows that there were moderate-to-high correlations among all clinical and neurocognitive domains, adjusting for age, gender and sibpair types. Table 4(b) shows the strength of the genetic/familial components of the observed correlations between each pair of all seven clinical and neurocognitive domains. After adjusting for age and sex, there were significant genetic bivariate correlations between negative symptoms and three neurocognitive domains (working memory, episodic memory and executive function). In addition, there were statistically significant genetic bivariate correlations between working memory, episodic memory and executive function. For these non-converging models, we conducted alternative analyses using only gender as a covariate given that age distributions were comparable across groups (mean 20–23 years) and age was not significantly associated with the disorganization domain (p=0.80). These alternative genetic correlation estimates suggest that disorganization is also significantly associated with working memory, episodic memory and executive function, but not attention. The results suggest there might be common genetic components that influence both negative symptoms and a cluster of neurocognitive deficits (i.e. pleiotropy).
To our knowledge, this is the first study to directly compare the heritability of both clinical and neurocognitive domain measures in sibling pairs with and without schizophrenia, as well as to examine pleiotropy among these constructs. We confirmed the hypothesis that the neurocognitive domain measures were mostly heritable in healthy sibling pairs. Importantly, we observed statistically higher heritability estimates in CON/CON-SIB pairs compared to PRO/PRO-SIB pairs for working memory. Genetic pleiotropy was suggested by the strong genetic correlations between three major neurocognitive domains (working memory, episodic memory and executive function), and the strong genetic correlation between negative symptoms and the neurocognitive domain cluster consisting of working and episodic memory, and executive function.
The heritability estimates for neurocognitive measures used in the study by Greenwood and colleagues (e.g., CPT, LNS, and CVLT) were comparable to our heritability estimates in our PRO/PRO-SIB pairs, and it is notable that a number of the tests studied by COGS were included in our formulation of the working memory and episodic memory domain measures (Greenwood et al. 2007). The heritability estimate for working memory (.65) from the recent twin study in UK was also similar to our estimate using all sibpairs (Toulopoulou et al. 2007). As noted above, we found that the moderate heritability observed in the CON/CON-SIB pairs was significantly decreased in the PRO/PRO-SIB pairs for working memory, with similar patterns for executive function and episodic memory. This finding is consistent with the results of a prior twin study that found that ICC estimates for verbal episodic memory were attenuated in twin pairs discordant for schizophrenia as compared to healthy twin pairs (Cannon et al. 2000). While the results of our study provide further support for the hypothesis that domains of cognition thought to be central to the neurobiology of schizophrenia are heritable, they also suggest that this heritability is decreased in the presence of the disease and disease vulnerability.
The attenuation of the heritability of neurocognitive measures in PRO/PRO-SIB pairs could be explained in several ways. First, the neurocognitive domain measures could be subject to increased unique environmental influence in PRO/PRO-SIB pairs, such as perinatal complications or medication. Alternatively, the neurocognitive domain measures could be subject to decreased genetic/familial influence in the PRO/PRO-SIB pairs. This latter explanation could reflect the fact that genes associated with the risk of developing schizophrenia are present in the ill sibling but not the well sibling. Shared environmental effect has been reported to be small compared to heritability for working memory (Toulopoulou et al. 2007), but tracing the source of heritability estimate changes between healthy and vulnerable twin pairs remains to be studied.
The results also suggest that there is a significant genetic correlation between negative symptoms and the neurocognitive domains of working memory, episodic memory and executive function. These results need to be replicated in future genetic studies of schizophrenia. However, they are consistent with the overarching hypothesis that epigenetic factors in schizophrenia influence potential phenotypic variance (Petronis 2004).
There were several limitations to our study. First, we focused on neurocognitive domain measures instead of individual tests (see supplemental data for individual test results) because the domain scores were more reliable (see Supplemental Table 1). One could debate which tests should be included in which domain, and further research is needed to understand the way in which results may differ across specific tests thought to assess the same domain (e.g., differential reliability, construct validity, etc.) Furthermore, the variation in assessment tools of clinical symptoms between this study and the others may explain the difference between our findings and others. For example, disorganization symptoms were reported to be heritable by Cardno et al (Cardno et al. 2001), but not in our study which used different assessment instruments. Based on Table 3, disorganization symptom domain was marginally heritable with p=0.082, a trend level significance. In addition to the variances due to instruments and study populations, another source of variation may be the sibpair types. The studies suggesting disorganization to be most heritable had concordant schizophrenia sibling pairs, while we used a discordant sibling pair design consisted of the proband with schizophrenia and a healthy young sibling (mean age 22 years of age) who may or may develop schizophrenia later on. As such, the variance in positive and disorganization symptoms in healthy siblings may be limited and could have reduced the heritability estimates. Second, heritability was calculated as proportion of the total variance shared between sibling pairs. Unlike a twin study design where variances due to genes and unique environmental factors can be statistically separated, our estimation of heritability must be interpreted as familiality due to both shared genes and shared environments of the sibling pairs. Third, other factors which could influence the assessments of clinical and neurocognitive functioning, such as antipsychotic and other psychotropic medications, were not studied in the current analyses. Further understanding of the familiality of clinical and cognitive function requires studies of non-medicated sibling pairs and concordant unaffected sibling pairs. In addition, our current sample is susceptible to ascertainment bias because these are voluntary patients with siblings who would both consent to volunteer. Another limitation is the power. The statistical power in heritability estimation is a function of number of families, sibship size, type I error, and hypothesized heritability. In the current study with approximately 130 sibpairs, the power of detecting h2 of 0.4 or higher is 0.7–0.8 (Klein 1974; Shaw 1987). Therefore, the size of our sample may not possess the sufficient power to detect a smaller heritability than 0.40. The issue of power limits one’s interpretation of negative findings. However, we were able to find significant heritability for most neurocognitive measures in our sample. When a higher statistical threshold is applied to adjust for multiple comparisons, most results remain significant, especially for the heritability estimates. When only a p-value less than .01 is considered significant, the bivariate genetic correlations between executive function and working memory, episodic memory and working memory, and executive function and negative symptoms remain significant. Because education may be associated with neurocognitive functioning, we conducted other analyses with age, gender and years of education as covariates and found similar results. The heritability estimates our results are similar and the differences are described in Table 3 and Table 4(b). Finally, multiple comparisons is always a concern, so the magnitude of actual p values were provided when possible in each table for more conservative significance standards.
Despite these limitations, our results suggest that familial aggregation of clinical and neurocognitive domains in healthy sibling pairs is stronger than that in sibling pairs affected by schizophrenia, especially for working memory, and that there was a significant genetic correlation between negative symptoms and three of our four neurocognitive domains. In future studies, this study model can be applied to test for the presence of specific genetic or environmental factors, while observing changes due to specific genetic and unique environmental influences. For example, the effects of candidate genes for specific or general cognitive features on the observed between-sibling correlations could be tested. Also, known environmental risk factors, such as a history of obstetrical complications, could be included in such models. In addition, one could include markers of brain structure or function as covariates in the estimation of heritability for specific cognitive domains.
We thank the staff of the Administration/Assessment and Biostatistical Cores of the CCNMD at Washington University School of Medicine for collection of the clinical and imaging data and data management. Dr. Li-Shiun Chen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Role of the Funding Source: This research was supported by NIH grants P50 MH071616 and R01 MH56584. The NIH had no further role in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.
Conflict of Interest
Dr. Csernansky has received research grants from the NIMH and NIA, and royalties from Medtronic for a patent held jointly with Washington University School of Medicine, has been a paid consultant for Eli Lilly and Sanofi-Aventis, and has received speakers’ honoraria from Janssen Pharmaceutica, Eli Lilly and Bristol-Myers Squibb. Dr. Deanna Barch has received grants from the NIMH, NIA, NARSAD and the McDonnel Center for Systems Neuroscience. All other authors declare that they have no conflict of interest.
Contributors: Drs. Csernansky, Barch, Thompson designed the study and wrote the protocol. Drs. Chen and Rice undertook the statistical analyses. Dr. Chen wrote the first draft of the manuscript. All authors contributed to and have approved the final manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.