No gene-by-environment interaction of BDNF Val66Met polymorphism and childhood maltreatment on anxiety sensitivity in a mixed race adolescent sample

ABSTRACT Background: Anxiety disorders in youth are attributable to multiple causal mechanisms, comprising biological vulnerabilities, such as genetics and temperament, and unfavourable environmental influences, such as childhood maltreatment (CM). Objective: A gene-environment (G x E) interaction study was conducted to determine the interactive effect of the BDNF Val66Met polymorphism and CM to increase susceptibility to anxiety sensitivity (AS) in a sample of mixed race adolescents. Method: Participants (n = 308, mean age = 15.8 years) who were all secondary school students and who completed measures for AS and CM were genotyped for the BDNF Val66Met polymorphism. Hierarchical multiple regression analysis was conducted to assess G x E influences on AS. Age and gender were included in the models as covariates as age was significantly associated with AS total score (p < .05), and females had significantly higher AS scores than males (p < .05). Results: A main effect of CM on AS was evident (p < .05), however, no main effect of BDNF genotype on AS was observed (p > .05). A non-significant G x E effect on AS was revealed (p < .05). Conclusions: Our results suggest that CM does not have a moderating role in the relationship between the BDNF Val66Met genotype and the increased risk of anxiety-related phenotypes, such as AS. Given the exploratory nature of this study, findings require replication in larger samples and adjustment for population stratification to further explore the role of BDNF Val66Met and CM on AS in mixed race adolescents.

Studies have demonstrated that acute and chronic stress (e.g. military training, psychological job stress, acute and repeated restraint in animals) are associated with decreased BDNF (Mitoma et al., 2008;Murakami, Imbe, Morikawa, Kubo, & Senba, 2005;Suzuki et al., 2014) and enhancement of anxiety-related behaviours (Chen et al., 2006). In adults with lifetime major depressive disorder, for example, a linear relationship between exposure to CM (i.e. emotional neglect, psychological abuse, sexual abuse) and reduced BDNF serum levels has been demonstrated in Met66 allele carriers, with the lowest BDNF levels evident in Met66 carriers reporting two or more childhood trauma types (Elzinga et al., 2011). Similarly, compared with non-trauma exposed children and adolescents, youth with CM histories (i.e. sexual abuse) have been found to exhibit significantly lower serum BDNF levels, with the lowest levels documented in those with multiple sexual assault histories (Şimşek, Yüksel, Kaplan, Uysal, & Alaca, 2015).
There is support for the role of gene-environment interaction in the aetiology of youth and adult anxiety, including panic and separation anxiety, anxiety symptoms, anxious mood and anxious temperament (Baumann et al., 2013;Chen et al., 2015;Gunthert et al., 2007;Ibarra et al., 2014;Klauke et al., 2011;Lau, Gregory, Goldwin, Pine, & Eley, 2007;Stein, Schork, & Gelernter, 2008;Vendlinski, Lemery-Chalfant, Essex, & Goldsmith, 2011). CM and AS have both consistently been implicated as risk factors for psychopathology, including anxiety disorders (Collishaw et al., 2007;Kessler, McLaughlin et al., 2010;McLaughlin et al., 2007McLaughlin et al., , 2012. We have previously determined, in a sample of secondary schoolattending adolescents, that CM is significantly and positively associated with anxiety-related temperamental traits such as AS and trait anxiety (Martin, Viljoen, Kidd, & Seedat, 2014). Further, there is evidence that interactions of life stress/early adversity (including abuse and neglect) with BDNF Val66Met gene variants predict subsequent risk for psychopathology, such as depression, in youth and adults (Carver, Johnson, Joormann, Lemoult, & Cuccaro, 2011;Chen, Li, & McGue, 2013;Gutiérrez et al., 2015;Hosang, Shiles, Tansey, McGuffin, & Uher, 2014). G x E interactions (i.e. BDNF Val66Met and early life stress exposure) have previously been found to be associated with abnormalities in brain structures, physiological indicators, deficits in cognition (i.e. poorer working memory), higher levels of depression and anxiety, and elevated temperamental traits (Gatt et al., 2009). However, data on the interactive effects of CM and genetic factors, such as the BDNF Val66Met polymorphism, on susceptibility to anxiety-related temperamental traits, such as AS, in adolescents is limited. Identifying risk markers for AS, an intermediate phenotype for anxiety disorders, in non-clinical samples is crucial to informing community-and school-level prevention strategies. Early and accurate identification is even more important in light of the wide mental health treatment gap that exists in many country settings. To our knowledge, no previous studies have investigated whether the BDNF Val66Met polymorphism interacts with CM to increase susceptibility to AS in adolescents. In addition, relative to studies that have assessed the BDNF Val66Met polymorphism in Caucasian samples, there is limited information of the allelic distribution of the BDNF Val66Met polymorphism in ethnically diverse samples. A G x E interaction study was conducted to determine the interactive effect of the BDNF Val66Met polymorphism and CM to increase susceptibility to AS in a school sample of South African mixed race adolescents.

Design
This study was a two-tier study in a sample of secondary school students. Tier 1 employed stratified two-stage cluster sampling in which public secondary schools in Cape Town, South Africa, and students within these schools, were randomly selected. The Tier 1 sample of secondary school students was therefore representative of students attending public secondary schools in Cape Town, South Africa. Tier 1 allowed for the screening (e.g. of levels of CM and AS) and collection of salivary DNA from secondary school students from 29 public schools in Cape Town, South Africa. A description of Tier 1 methods, including clinical data pertaining to Tier 1, has previously been reported (Martin et al., 2014). Four groups of adolescents grouped according to levels of Tier 1 self-reported CM and AS, amongst others, and matched as closely as possible on age, ethnicity, gender and educational status, were included in the second tier of the study and underwent a neuropsychological and neuroimaging assessment, amongst others.

Participants
The Tier 1 sample consisted of 1149 secondary school students. DNA was extracted from 985 participants (i.e. comprising 85.7% of the total Tier 1 sample) at the same time that Tier 1 self-report measures (e.g. CM and AS) were administered. The majority of the Tier 1 sample consisted of Black (64.4%) and mixed race (32.2%) participants. As almost all Black participants (99.68%, 628/630) in the sample were Val66Val genotype carriers, the subset reported on here included only those participants that self-classified as 'mixed race' and for whom BDNF data were available (n = 308).
This study was approved by the Health Research Ethics Committee of Stellenbosch University, South Africa (ethics reference number: N10/11/370), and permission to access secondary schools and conduct this study was provided by the Western Cape Education Department. Written informed consent was obtained from parents/legal guardians and written assent was obtained from the students themselves.

Self-report measures
The following self-report questionnaires, amongst others, were administered at the secondary schools on a single occasion.
The Childhood Anxiety Sensitivity Index (CASI; Silverman, Fleisig, Rabian, & Peterson, 1991) is an 18-item instrument that measures the fear of anxiety symptoms, designed for use with school-age children and adolescents. The CASI yields a total score that can range from 18 to 54, with higher scores reflecting higher levels of AS.
The Childhood Trauma Questionnaire -Short Form (CTQ-SF; Bernstein et al., 2003) is a brief, 28item, retrospective self-report measure of the frequency of abuse (i.e. emotional, physical, sexual) and neglect (i.e. emotional, physical) experienced prior to age 18 years. The CTQ-SF yields a total score in the range 25-125, with higher scores reflecting more severe levels of abuse or neglect. Scores for each of the CTQ-SF subscales are in the range 5-25, with higher scores indicating more severe childhood trauma (Bernstein & Fink, 1998). For the purpose of the current study, we enquired about abuse and neglect experienced prior to age 12 years.

Data analysis
All analyses were conducted using STATISTICA version 13 (StatSoftInc., 2015). Univariate normality was determined for all demographic and clinical variables. Demographic characteristics of the total sample (i.e. age, gender, current grade at school), as well as self-report variables of interest (i.e. AS, CTQ), were calculated as means (M) and standard deviations (SD) for quantitative variables and counts and associated percentages for categorical variables (e.g. gender). The severity (i.e. minimal, moderate, severe, extreme) of abuse and neglect categories of the CTQ (i.e. emotional, physical and sexual abuse, emotional and physical neglect) were computed for the total sample using recommended cut-off scores (Bernstein & Fink, 1998).
To determine if AS levels differed by gender and age, t-tests and Pearson's correlation statistic, respectively, were employed. Genotype counts (%) and the Hardy-Weinberg equilibrium (HWE) p-value was determined using the R Package SNPassoc (González et al., 2007). Demographic and clinical characteristics of the sample by genotype (i.e. Val66 homozygotes and Met66 allele carriers) were summarized as means (M) and standard deviations (SD) for quantitative variables, and counts were used for categorical variables. Independent samples t-tests and chi-square tests were used to determine any group (i.e. genotype) differences for the quantitative and categorical variables, respectively. Pearson's correlation statistic was used to assess the relationship between CTQ total score and CASI total score by genotype.
Hierarchical multiple regression analysis was used to assess the effects of BDNF genotype (coded 0 for Val66Val and 1 for Met66 allele carriers), level of CM and the interaction of BDNF genotype and CM, on AS. In the first model, gender, age, BDNF genotype and CTQ total score were included to assess the main effects of BDNF genotype and CTQ total score on the outcome, AS. In the second model, the two-way interaction term (i.e. BDNF x CTQ) was added. The F-to-remove test was used to compare the R 2 change between the first and second model to determine whether the inclusion of the interaction term resulted in a significant increase in explained variance.
Results of multiple regression analyses (i.e. model 1 and model 2) are presented in Table 3. A significant main effect of CTQ total score (p < .01) on AS was evident, however, no significant main effect of BDNF genotype on AS was observed (p > .05). In the second model, the inclusion of the interaction term did not contribute to an increase in explained variance (p > .05).

Discussion
Anxiety-related temperamental traits, such as AS, trait anxiety and neuroticism, are said to be interactively impacted by genetic and environmental factors, such as CM and early life stress and trauma. This study investigated whether the BDNF Val66Met polymorphism interacted with CM to increase susceptibility to AS in a sample of mixed race adolescents. To our knowledge, this is the first study to assess the role of BDNF Val66Met in AS in a South African mixed race sample of adolescents.
Our results revealed a significant main effect of CM on AS, however, there was no significant main effect of BDNF genotype on AS. These findings suggest that the BDNF Val66Met polymorphism does not have a direct effect on AS. This finding is in line with studies that have found no significant direct association between BDNF Val66Met polymorphism and personality traits such as neuroticism or harm avoidance and anxiety disorders or mood disorders, including OCD, panic disorder, PTSD and depression (Arias et al., 2012;Chen et al., 2013;Frustaci et al., 2008;Hong et al., 2012;Minelli et al., 2011;Surtees et al., 2007;Terracciano et al., 2010), despite some studies reporting such associations (Frustaci et al., 2008;Lang et al., 2005;Min et al., 2013;Montag et al., 2010;Sen et al., 2003;Terracciano et al., 2010). Firstly, the grouping of Met66 allele carriers (i.e. Met66Met and Val66Met genotypes), as is frequently carried out in studies in which the rate of the Met66Met genotype is relatively low, such as in Caucasian samples (Gatt et al., 2009;Lehto, Maestu, Kiive, Veidebaum, & Harro, 2016;Nedic et al., 2013;Pivac et al., 2009), may introduce a bias in which a main effect of genotype is not detected due to the exclusion of   Notaras et al., 2015). Secondly, the Val66Met and Met66Met genotypes, respectively, may have dissimilar effects (Hong et al., 2012). Nevertheless, the main effect of CM on AS determined in this study provides support for the positive association between CM, including stressful life events, and AS in adolescents (McLaughlin & Hatzenbuehler, 2009;Tollenaar, Molendijk, Penninx, Milaneschi, & Antypa, 2017), a well-established cognitive risk factor for the development of anxiety disorders and associated symptoms in youth (Hishinuma et al., 2001;McLaughlin et al., 2007;Muris et al., 2001). Furthermore, our finding adds to the well-established literature demonstrating the adverse acute and longterm effects of CM or trauma on mental health and cognition in youth and adults (De Bellis, Woolley, & Hooper, 2013;Greger, Myhre, Lydersen, & Jozefiak, 2015;Irigaray et al., 2013;Taillieu, Brownridge, Sareen, & Afifi, 2016;Teicher, Ohashi, Lowen, Polcari, & Fitzmaurice, 2015).
Our results revealed a non-significant BDNF genotype x CM effect on AS. This finding is not consistent with studies that found that the low-functioning BDNF Met66 allele and CM/childhood trauma or early life adversity or stress interact to predict increased susceptibility for psychopathology, including anxiety-related temperamental traits, such as neuroticism (Gatt et al., 2009) andguilt-proneness (Szentágotai-Tətar et al., 2015), anxiety symptoms (Gatt et al., 2009) and mood disorders and associated symptoms (Aguilera et al., 2009;Carver et al., 2011;Gutiérrez et al., 2015). Our findings are also not in agreement with G x E studies that found an interactive effect of the higher functioning Val66 allele and environmental exposures (i.e. adversity, negative stressors, recent life events) on psychopathology, such as increased levels of neuroticism, harm avoidance and depression (Chen et al., 2013;Kim et al., 2009;Lehto et al., 2016). Apart from the possible confounding effects of age and gender, discrepant results across studies may in part be due to population-driven differences in BDNF Val66Met frequencies, given that the Met66 allele has consistently been found to be more common in Asian populations than in Caucasian populations (Chen et al., 2013;Petryshen et al., 2009). A further confounding factor may include phenotype heterogeneity and methodological (assessment) differences (Hong et al., 2012).
A number of study limitations should be taken into account in interpreting the current findings. First, due to the cross-sectional nature of this study, inferences about causality cannot be made. Second, our sample size of N = 308 is relatively small given the estimated sample sizes required in candidate gene studies in which functional polymorphisms are assessed and in which minor effects are expected (Duncan & Keller, 2011). Third, use of the CTQ, a retrospective self-report measure of CM, may have introduced recall bias which may have resulted in the over-or under-reporting of maltreatment frequency. Fourth, we explored one polymorphism within the BDNF gene. Finally, we did not correct for gene-environment correlation (rGEs) or population stratification. The South African mixed race population is characterized by high levels of admixture (Tishkoff et al., 2009) and ancestral diversity (i.e. Khoesan, European and Asian ethnicity) (Hemmings et al., 2016;Wright, Niehaus, Koen, Drögemöller, & Warnich, 2011), suggestive of genetic heterogeneity, which may have influenced our results. Findings of this exploratory study are therefore preliminary. Our findings add to the literature on aetiological processes that may underlie the development of anxiety-related traits, symptoms and disorder in adolescents. More specifically, our findings shed light on the role of the BDNF Val66Met polymorphism in the development of anxiety-related traits in mixed race adolescents in the context of childhood adversity. Furthermore, these results extend findings of the role of BDNF and CM on AS. Previous studies have focused on convenience samples of college students, or adults, and predominantly on the effects of the 5-HTTLPR polymorphism (Hemmings et al., 2016;Kim et al., 2009;Klauke et al., 2011;Laucht et al., 2009;Stein et al., 2008). Our results suggest that the influence of CM on adolescent AS levels is not moderated by the BDNF Val66Met polymorphism. Our findings highlight the importance of assessing gene-environment interactions in the assessment of genetic effects on anxiety-related phenotypes associated with anxiety disorders. Recommendations for future research include replication in larger samples of mixed race participants in which population stratification is corrected for and in which the Met66Met genotype is better represented (Notaras et al., 2015). Furthermore, variation across the BDNF gene would be useful to consider (Mandelman & Grigorenko, 2012). Additionally, the effect of other genetic variants, in conjunction with the BDNF Val66Met polymorphism, should be explored given findings of an epistatic effect between BDNF and the serotonin transporter genes (Martinowich & Lu, 2008;Pezawas et al., 2008). Further support for the aforementioned is reflected in findings of gene x gene and gene x gene x environment interactions which are indicative of significant interaction effects of the BDNF Val66Met polymorphism and the serotonin transporter gene (SLC6A4) on anxietyrelated traits (e.g. neuroticism and harm avoidance) (Arias et al., 2012;Terracciano et al., 2010) and depressive symptoms and disorder (Gutiérrez et al., 2015;Kaufman et al., 2006). Finally, the effects of environmental influences other than CM, such as parenting rearing practices (Ibarra et al., 2014) and general selfefficacy (Schiele et al., 2016), on AS should be explored. Investigation of the aforementioned would provide a clearer understanding of the genetic and environmental impacts on AS in mixed race adolescent samples.