Introduction
Among the several explanations invoked to explain failures to find susceptibility genes, gene–environment interaction (G×E) suggests that ‘ignoring nurture may have handicapped the field's ability to understand nature’ (Moffitt et al. Reference Moffitt, Caspi and Rutter2005). G×E can be defined as the genetic control of sensitivity to environmental factors, or as the environmental control of gene expression (Kendler & Eaves, Reference Kendler and Eaves1986). Several recent studies demonstrated that the G×E approach can be of practical benefit as a tool in the search for genes connected with mental disorders (Caspi et al. Reference Caspi, Sugden, Moffitt, Taylor, Craig, Harrington, McLay, Mill, Martin, Braithwaite and Poulton2003, Reference Caspi, Moffitt, Cannon, McLay, Murray, Harrington, Taylor, Arseneault, Williams, Braithwaite, Poulton and Craig2005; Kahn et al. Reference Kahn, Khoury, Nichols and Lanphear2003; Kendler et al. Reference Kendler, Kuhn, Vittum, Prescott and Riley2005).
Childhood trauma has been repeatedly associated with psychotic features not only in schizophrenia (for review, see Read et al. Reference Read, Van Os, Morrison and Ross2005) but also in bipolar disorders (Hammersley et al. Reference Hammersley, Dias, Todd, Bowen-Jones, Reilly and Bentall2003; Neria et al. Reference Neria, Bromet, Carlson and Naz2005) and schizotypy (Yen et al. Reference Yen, Shea, Battle, Johnson, Zlotnick, Dolan-Sewell, Skodol, Grilo, Gunderson, Sanislow, Zanarini, Bender, Rettew and McGlashan2002; Berenbaum et al. Reference Berenbaum, Valera and Kerns2003). Therefore, childhood trauma may be a good candidate environmental risk factor for psychotic disorders. Schizotypy is a multidimensional construct that represents a mild imitation of the symptoms of schizophrenia (Vollema & Postma, Reference Vollema and Postma2002). Although schizotypy is the core feature of schizotypal personality disorder, it is thought to be a continuous or dimensional, rather a categorical, construct (Irwin, Reference Irwin2001). Schizotypal dimensions have been described as a potential intermediate phenotype for both schizophrenia and bipolar disorders and could be the consequence of a G×E interaction. This association is difficult to examine prior to the identification of disease genes. The study of a population at ‘high genetic risk’ that carries several unknown genes related to the disorder may improve the possibility of identifying environmental risk factors. This approach may be more appropriate than studying the interaction between an environmental factor and a specific candidate gene. Studying the impact of childhood trauma on schizotypal dimensions among unaffected relatives has several advantages. First, dimensional models of psychosis (positive, negative and disorganized symptoms) are becoming established as conceptually and clinically useful (Van Os et al. Reference Van Os, Gilvarry, Bale, Van Horn, Tattan, White and Murray2002). Second, first-degree relatives are free from the multiple artifacts that potentially confound research in schizophrenia, including the effects of long-term and usually ongoing medication treatment, multiple hospitalizations or institutionalization, and prolonged functional impairment secondary to chronic disease and social deterioration.
We investigated whether childhood trauma is correlated with the phenotypic expression of schizotypyal dimensions (positive, negative and disorganized) in two different populations of unaffected first-degree relatives. We hypothesized that the correlation between childhood trauma and schizotypal dimensions could be different among these two high-risk populations.
Method
Subjects
Consecutively admitted probands meeting DSM-IV criteria for bipolar disorder or schizophrenia were recruited from a university-affiliated hospital (Psychiatry Department, Paris XII University). Their unaffected first-degree relatives were contacted and asked to participate in the study. Relatives were interviewed with the French version of the Diagnostic Interview of Genetic Studies (DIGS; Nurnberger et al. Reference Nurnberger, Blehar, Kaufmann, York-Cooler, Simpson, Harkavy-Friedman, Severe, Malaspina and Reich1994; Preisig et al. Reference Preisig, Fenton, Matthey, Berney and Ferrero1999) to confirm the absence of diagnosis of schizophrenia or bipolar disorder. We used the validated French translation (Dumas et al. Reference Dumas, Bouafia, Gutknecht, Saoud, Dalery and d'Amato2000) of the self-rating Schizotypal Personality Questionnaire (SPQ; Raine, Reference Raine1991) to measure schizotypal dimensions in the two different groups of first-degree relatives. This questionnaire contains 74 items with yes/no answers and identifies the three classical dimensions of psychosis (positive, negative and disorganized). The SPQ can be used in the general population to identify individual differences in schizotypal dimensions.
Subjects also completed the French translation (Paquette et al. Reference Paquette, Laporte, Bigras and Zoccolillo2004) of the Childhood Trauma Questionnaire (CTQ; Bernstein & Fink, Reference Bernstein and Fink1998), a 28-item self-report questionnaire that gives scores for physical, emotional and sexual abuse, physical and emotional neglect, and a total weighted score.
The Research Ethics Board reviewed and approved the study. After providing a complete description of the study, written informed consent was obtained from all participants.
Data analysis
Differences between groups were tested using a two-tailed t test for continuous variables and a χ2 test for discrete variables.
The relationships between schizotypal dimensions and CTQ scores were analyzed by partial correlation to control for the potential confounding influences of sex and age. Sex was used as covariable because of its influence on both CTQ and SPQ scores (Dumas et al. Reference Dumas, Bouafia, Gutknecht, Saoud, Dalery and d'Amato2000; Paquette et al. Reference Paquette, Laporte, Bigras and Zoccolillo2004). Age at interview was also used as a covariable because some studies have observed a negative correlation between age of the subjects and scores for the SPQ scale and subscales (Venables & Bailes, Reference Venables and Bailes1994; Mata et al. Reference Mata, Mataix-Cols and Peralta2005). To limit the risk of type I error due to multiple statistical tests, we examined only the total score of the CTQ, and we used the Bonferroni procedure to establish the significance threshold (p<0.05 divided by 12 statistical analyses; we retained p<0.004). We also analyzed the variables dichotomously and calculated odds ratios (ORs) and 95% confidence intervals (CIs) for schizotypy given childhood trauma in the two populations of relatives. To do that, we determined the presence or absence of childhood trauma according to the cut-off of the French validation of the CTQ. The cut-off for the schizotypal dimensions was determined as the mean score of the SPQ full-score for the whole sample.
Results
The study group was composed of 67 first-degree relatives of schizophrenic probands [31 men and 36 women, mean age (s.d.)=54.2 (15.4) years] and 71 first-degree relatives of bipolar probands [37 men and 34 women, mean age (s.d.)=53.1 (15) years]. The sex ratio (χ2=0.47, p=0.49) and the age at interview (t=0.43, p=0.67) did not differ between groups. The samples of first-degree relatives were composed of different types of relatives. In the two samples of relatives, the majority of the relatives were parents of the proband with an over-representation of parents in the sample of first-degree relatives of schizophrenic subjects (80.5% v. 59.2%). The highest completed school grade (education level) was recorded, according to the usual conventions (Pichot et al. Reference Pichot, Lebeaux, Penhouët and Simon1993) as a trichotomous variable (1=elementary school; 3=at least high school completed; 2=intermediate between 1 and 3). However, we finally decided to group the first two levels together because level 1 was infrequent in all groups. The two samples did not differ in terms of educational level (p=0.40). In the sample of schizophrenic relatives, the mean SPQ scores (full-scale and for the positive, negative and disorganized dimensions) were 10.2+7.9, 3.9+3.8, 4.7+4.4 and 1.6+1.7 respectively, and the mean CTQ score was 38.6+8.5. In the sample of bipolar relatives, the corresponding mean SPQ scores were 9.6+7.5, 3.3+3.4, 4.3+4.7 and 1.9+2 respectively, and the mean CTQ score was 39.3+9.8. The SPQ full-score and the CTQ score did not differ between the two groups (SPQ full-score: p=0.73 and CTQ full-score: p=0.95). The mean CTQ and SPQ scores were similar between males and females (p=0.58 and p=0.77 respectively). In the whole sample of relatives, we found a significant positive correlation between the CTQ score and the SPQ full-score (r=0.27, p=0.001). This correlation was fully explained by a correlation in the sample of first-degree relatives of schizophrenic subjects (r=0.43, p<0.0001), whereas no significant correlation was found in the first-degree relatives of bipolar probands (r=0.13, p=n.s.). This positive correlation was found with the positive and negative dimensions of schizotypy (positive schizotypy: r=0.41 and negative schizotypy: r=0.36) but not with the disorganized dimension (Table 1). We also investigated the partial correlation coefficients between each schizotypal dimension and childhood trauma adjusted for the two other schizotypal dimensions. Only the correlation between positive dimension and childhood trauma remains significant.
Table 1. Correlation matrix for childhood trauma and schizotypal dimensions

SPQ, Schizotypal Personality Questionnaire; CTQ, Childhood Trauma Questionnaire.
Statistical significance after Bonferroni correction:
* p<0.004.
Based on a dichotomous approach, calculation of the ORs led to similar results. The odds of having high schizotypal traits in the presence of childhood trauma were higher in the population of schizophrenic relatives (OR 3.6, 95% CI 1.09–11.8) than in the population of the bipolar relatives (OR 1.64, 95% CI 0.57–4.72), which were not significant.
Discussion
We found a significant positive correlation between history of childhood trauma and schizotypal dimensions in unaffected first-degree relatives of schizophrenic subjects but not in relatives of bipolar probands. This correlation was mainly due to a strong association with the positive dimension of psychosis. To our knowledge, this is the first study to explore the influence of an environmental risk factor in two populations with different genetic backgrounds. Even though schizotypy was shown to be elevated in relatives of bipolar subjects (Kety et al. Reference Kety, Wender, Jacobsen, Ingraham, Jansson, Faber and Kinney1994; Schürhoff et al. Reference Schürhoff, Laguerre, Szöke, Méary and Leboyer2005), suggesting some shared genetic vulnerability factors, the only difference between our two populations was that one is at genetic risk for schizophrenia and the other for bipolar disorders. We observed that the impact of childhood trauma on the expression of schizotypal dimensions was strongest among subjects who were at high genetic risk for schizophrenia, suggesting that susceptibility genes specific to schizophrenia may interact with childhood trauma to induce the emergence of schizotypal dimensions, mainly positive psychotic features. This is in accordance with several previous reports (for review, see Read et al. Reference Read, Van Os, Morrison and Ross2005) showing a high influence of childhood trauma on positive dimension in schizophrenia and schizotypy. In line with our results, Spauwen et al. (Reference Spauwen, Krabbendam, Lieb, Wittchen, Jansen and Van Os2006), using a follow-up design, found an association between self-reported trauma and psychosis proneness. The strength of the association increased with a narrower psychosis definition. These authors speculated that extended exposure to trauma may increase the risk for positive symptoms through direct effects on dopamine function. Along similar lines, Read et al. (Reference Read, Van Os, Morrison and Ross2005) suggested that early, prolonged and severe traumas increase the risk for later positive symptoms through lasting effects on the hypothalamic–pituitary–adrenal axis. Even if associations between negative dimension and childhood trauma have been reported previously (Lysaker et al. Reference Lysaker, Meyer, Evans, Clementz and Marks2001; Resnick et al. Reference Resnick, Bond and Mueser2003), this association is more difficult to interpret. A plausible explanation is that traumatic avoidance, emotional numbing and reduced responsiveness may resemble negative dimension and may be interpreted as a reaction to the trauma. Conversely, in rodents and non-human primates, trauma (for example, the forced swim test) has been used as a model of the negative symptoms of schizophrenia such as flattening of affect and avolition (Corbett et al. Reference Corbett, Zhou, Sorensen and Mondadori1999). Concerning the disorganized dimension, as in previous studies, we did not show any correlation with childhood trauma (Read & Argyle, Reference Read and Argyle1999; Hammersley et al. Reference Hammersley, Dias, Todd, Bowen-Jones, Reilly and Bentall2003). However, this could also be explained by the small variation of this dimension. Our study could be considered as an example of how an environmental factor interacting with specific susceptibility genes can influence susceptibility to a disorder or to a particular dimension of a disorder. Our results can suggest that the same dimensions have different genetic backgrounds in bipolar disorders and schizophrenia. However, there is also a more plausible explanation. The same genes cause susceptibility for a common phenotype in bipolar disorders and schizophrenia, in particular the positive dimension (Berrettini, Reference Berrettini2003; Schürhoff et al. Reference Schürhoff, Szöke, Méary, Bellivier, Rouillon, Pauls and Leboyer2003). To become manifest, this susceptibility needs to interact with some environmental factors. However, other factors, disorder-specific, can exert a protective or permissive role on the action of these environmental factors. For example, traits that have been associated with genetic susceptibility to bipolar disorders (i.e. cyclothymic, hyperthymic or depressive temperaments) might have a protective effect on the emergence of positive schizotypal traits in the presence of childhood trauma.
The possibility that susceptibility genes for schizophrenia may increase exposure to childhood trauma can be excluded because rates of childhood trauma were similar among the two groups of relatives. However, because of the design of our study, two alternative hypotheses could be proposed. Indeed, our results could also be explained by shared family environmental factors among the schizophrenia relatives rather than genetic factors; this could only be ruled out by using an adoption study design. In other words, the childhood trauma could be a proxy for other environmental risks such as social disadvantage or family discord/chaos specific to schizophrenic families. Moreover, because of the retrospective assessment of the childhood trauma, we cannot exclude the possibility that subjects reported a high level of childhood trauma because they have a high level of schizotypal features. Nevertheless, the fact that no association is found in the relatives of bipolar probands suggests that a bias cannot entirely explain the association found in relatives of schizophrenic subjects between trauma and schizotypal dimensions. It is also interesting to note that even though the relatives tend to be defensive in their response, the use of the SPQ discriminates the two samples of relatives. Overall, our results support the assumption that the clinical expression of the vulnerability induced by environmental risk factors depends on G×E interactions. Speculation about the implications of this finding should be guarded until these results are replicated as we cannot formally exclude the possibility that childhood trauma interacted with other specific environmental risk factors.
In conclusion, our findings suggest that susceptibility genes specific to schizophrenia may influence response to pathogenic environments. It is clear that if a gene's connection to the disorder is conditional to the environment, this will have the natural consequence of diminishing the researchers' capacity to detect the association between the gene and the disorder. Research to uncover specific environmental risk factors for diseases may be more effective if we recruit samples on the basis of their high genetic loading and if we investigate which genes discriminate between exposed individuals who did, versus did not, become ill.
Acknowledgments
This work was supported by grants from the Agence Nationale de la Recherche (Manage-BP), the Fondation pour la Recherche sur le Cerveau, the RTRS (Réseau Thématique Translationel de Recherche et de Soins) and the Institut de la Santé et de la Recherche Médicale (INSERM). We thank Marie-José De Sousa and Emmanuelle Abadie for their technical assistance, and an anonymous reviewer for helpful comments.
Declaration of Interest
None.