The interaction of 5‐HTT variation, recent stress, and resilience on current anxiety levels in adolescents and young adults from the general population

Previous work on gene‐environment (GxE) interplay concerning anxiety has focused on the interaction of 5‐HTTLPR with childhood adversities or traumatic events whereas the impact of recent stressors is understudied, as is the integration of resilience. The current study aimed to investigate the interactive effect of 5‐HTTLPR and recent stress on anxiety in adolescents considering resilience as buffer of a GxE risk constellation.


| INTRODUCTION
Anxiety is usually adaptive in response to potentially harmful or threatening situations, but may become pathological when being out of proportion to the actual danger posed, when occurring frequently and persistently, and when leading to individual suffering or impairment (APA, 2013). Just as anxiety, stress is inevitably embedded in our life and has a pivotal impact on health. It is well-known that stress in form of major stressors, for example, childhood adversities or (non-) traumatic life events, are associated with numerous psychopathological symptoms and mental disorders including anxiety (for a review see Klauke, Deckert, Reif, Pauli, & Domschke, 2010).
Further, research indicates that also recent stress, such as ongoing daily hassles (DH) or recent chronic stress in everyday life might increase the risk for anxious mood (D'Angelo & Wierzbicki, 2003), anxiety symptoms (Barrett & Heubeck, 2000;Kanner et al., 1981) and anxiety disorders (Asselmann et al., 2017). DH comprise everyday demands and conditions perceived as irritating, frustrating or stressful (Lazarus, 1986), whereas recent chronic stress is marked by a creeping onset, prevalent long-lasting recurring stressors with uncontrollable consequences. Thus, DH or recent chronic stress occurs much more frequently than major life events, having a substantial effect regarding anxiety.
However, people differ in their susceptibility to recent stress, and susceptibility to stress may have biological roots, primarily in the serotonergic system, reflected by an interaction of stress with genetic variation on anxiety. The serotonin transporter gene linked polymorphic region (5-HTTLPR), located in the promoter region of the serotonin transporter gene (SCL6A4; chromosome 17q12) has a key role within serotonergic neurotransmission by regulating the serotonin reuptake from the synaptic cleft (Lesch, Zeng, Reif, & Gutknecht, 2003). The polymorphism comprises a short allele (S), associated with less transcription of the serotonin transporter, compared with the long allele (Lesch, Bengel et al., 1996). A singlenucleotide polymorphism rs25531 (A>G) within the long allele renders the L G allele functionally equivalent to the S allele (i.e., reduced 5-HTT availability), while the A allele results in increased 5-HTT expression (Hu et al., 2006;Wendland et al., 2006).
Whereas most of the work on gene-environment (GxE) interplay following the seminal work by Caspi et al. (2003) has focused on the interaction of 5-HTTLPR with major life events (e.g., Klauke, Deckert, Reif, Pauli, Zwanzger et al., 2011), impact of recent stressors in terms of DH or recent chronic stress are understudied, particularly with regard to anxiety. One of few studies showed that among college students S allele carriers experienced more anxious mood on days with more stress than L allele carriers (Gunthert et al., 2007). In contrary, Ming et al. (2015) found that adolescent L allele carriers exhibited more anxiety symptoms related to stressful life events, e.g.
school or friendship problems. Apart from methodological differences, the heterogeneity of the findings may to some extent be explained by positive factors counteracting a GxE risk profile.
Recently, an extended approach to the GxE interaction model to include coping (C) characteristics in GxExC model (c.f., Schiele, Herzog, Kollert, Schartner et al., 2020) has highlighted the necessity of considering both advantageous and disadvantageous influences in shaping anxiety risk. High levels of general self-efficacy were observed to buffer an otherwise increased vulnerability to anxiety as conferred by the interaction of childhood adversity and 5-HTTLPR genotype (Schiele, Ziegler et al., 2016). These results are in line with the "differential susceptibility hypothesis" (Belsky et al., 2009), which postulates that genes are neither entirely favorable nor unfavorable, but rather drive sensitivity to environmental influences as a whole.
Resilience, the process of sustaining or strengthening physiological or behavioral stability in response to stressors, may be one prominent advantageous candidate. Hjemdal, Vogel, Solem, Hagen, and Stiles (2011) showed that lower resilience was associated with higher levels of anxiety in an adolescent sample. In line, higher levels of resilience were associated with lower anxiety symptom levels in an adolescent sample (Skrove et al., 2013) and in a student sample (Haddadi & Besharat, 2010).
To date, the integration of resilience in the context of GxE models in respect to anxiety is still rare. The only study known investigating a GxE interaction while considering moderating effects of resilience-increasing factors regarding anxiety was performed in adults (Schiele, Ziegler et al., 2016), whereas to the best of our knowledge no study focused on adolescence, i.e. the developmental period when anxiety often becomes pathological (Beesdo-Baum & Knappe, 2012). As adolescence is characterized by major psychophysiological and social changes (Sawyer et al., 2012), studying the interplay of genes, stressors, and resilience factors appears crucial to improve our understanding of emerging pathological anxiety during adolescence. Therefore, the aim of the current study is to investigate the interactive effect of 5-HTTLPR and recent stress on anxiety in a general population sample of adolescents by addressing the question whether resilience might buffer a GxE risk factor constellation.

| Sample and Procedures
The Behavior and Mind Health (BeMIND) study is a cohort study of a general population sample of adolescents and young adults from In short, an age-and sex-stratified random sample of 14-21 years old was drawn from the population registry in 2015, followed by a written invitation letter sent by the study team with a maximum of two reminder letters. All noninstitutionalized individuals aged OLLMANN ET AL. | 319 14-21 residing in a household in Dresden during the field period with sufficient German language skills were eligible to participate. Of 6321 invited subjects, 14.1% were found to be ineligible, mostly due to the fact that they were not residing under the provided address.
Of the remaining 5428 individuals, 1180 participated in the study assessments which were conducted between November 2015 and December 2016 at the study center at the Technische Universität Dresden. Participation was higher among females and among those with higher education. Total 42.8% of invited individuals did not answer the invitation letters and the nonresponder questionnaire.
Lack of time and interest were the most common given reasons for nonparticipation .
Participants underwent a standardized clinical-diagnostic assessment, an experimental assessment approximately one week later, and an ecological momentary assessment (EMA) as well as an online questionnaire assessment in between these two personal appointments. Biological or physiological data were collected during the EMA period (saliva, heart rate) and at the second personal appointment (blood/buccal, hair, anthropometric measures, and blood pressure). All participants provided written informed consent or as-  Three different measures assessed recent stress. The DH scale (Perkonigg & Wittchen, 1995) is a self-report questionnaire based on the social interview schedule (Faltermaier et al., 1985), which assesses irritating and/or frustrating demands of everyday life during the past 2 weeks using 15 four-point scaled items.

| Assessments
The perceived stress scale (PSS-4; Cohen & Williamson, 1988) is a four-item version of the PSS (Cohen, Kamarck, & Mermelstein, 1983), assessing "the degree to which situations in one's life are appraised as stressful" (Cohen et al., 1983;p.387). The items are linked to a sole latent trait measuring global stress levels, the degree to which life has been experienced as unpredictable, uncontrollable and overloaded in the past month. Each item is measured on a fivepoint Likert scale (Stächele & Volz, 2013).
The 12-item screening scale of the trier inventory for the as-

| Genotyping
Two 9 ml EDTA blood samples were collected by venipuncture using the vacuum method at baseline. Blood samples were stored immediately at −80°C. Whenever participants did not provide consent or assent to draw blood, they were asked to provide a buccal sample instead. 5-HTTLPR and the functionally related single-nucleotide polymorphism rs25531 were genotyped according to published protocols. In short, extracted DNA was amplified by polymerase chain reaction (60 s at 94°C, 60 s at 64°C, and 120 s at 72°C for 35 cycles) with the following oligonucleotide

| Statistical analysis
To improve representativeness regarding sex and age, sample weights were applied to make sure that the sample distribution of sex and age is equal to the one of the target population of the 14-21 years old people living in Dresden (for details see Beesdo-Baum, Voss et al., 2020). The sample was stratified for genotype group according to functionality (Hu et al., 2006;Wendland et al., 2006) and previous publications (Baffa et al., 2010;Baune et al., 2008;Klauke, Deckert, Reif, Pauli, Zwanzger et al., 2011;Schiele, Ziegler et al., 2016;Wang et al., 2011;Wendland et al., 2006) resulting in a high-expression group (L A L A carriers; N = 302) and a low-expression group (SS, SL G , SL A , L G L A , L G L G carriers; N = 640). A group comparison between the low-and high-expression group as well as between the excluded participants from the total sample and the analysis sample was conducted regarding sociodemographic characteristics including age (t test), sex distribution, education, social class, lifetime psychopathology (survey design-based F test; Rao & Scott, 1984)  variable was linearly rescaled to a range between zero (represents raw score of 8) and one (represents raw score of 40). Then, fractional response regression models (Papke & Wooldridge, 1996) were used, which are a viable tool if outcome data is skewed and many values occur at the lowest or highest possible outcome value. Also, the answer type of the PROMIS items suggests modeling the PROMIS score as a specific amount from a predefined maximum, that is, a fraction. Fractional logistic regression were used to determine the effects of 5-HTT genotype, stress scale (i.e., DH, STICS-SSCS, or PSS-4), and resilience (CD-RISC), as well as their interaction, on anxiety (PROMIS-ANX). The analyses were adjusted for age and sex and in a second step the presence of any lifetime mental disorder was included as covariate. Since regression coefficients of fractional response models are hard to interpret, plots of predictive margins for PROMIS-ANX from the estimated fractional response models are presented. For illustration purpose, participants above the CD-RISC sample mean were defined as "high-resilience group" and participants below the CD-RISC sample mean as "low-resilience group." Final N's in the analysis vary due to missing values in the questionnaire assessments (ranging from 18-100).

| Descriptive statistics
Descriptive characteristics of the analysis sample are given in Table 1. Genotype groups did not differ significantly regarding age

| Effect of 5-HTT genotype, recent stress, and resilience on anxiety
To present the results of the analyses precisely, beta coefficients of fractional logit models are reported. These beta coefficients are not to be confused with the usual beta coefficients from linear regression models. For interpretation, please see the figures depicting predictive margins calculated from the fractional logit models.    | 323

| DISCUSSION
The aim of the current study was to examine the interactive effect of SCL6A4 genotype and recent stress on anxiety, by addressing the question whether resilience is able to buffer a GxE risk constellation in adolescents and young adults. Resilience was found to interact with recent stress, such as DH and recent chronic stress, and 5-HTTLPR regarding anxiety. Specifically, adolescents carrying the more active L A L A genotype reported consistently higher levels of anxiety when they experienced more DH or recent chronic stress and had low levels of resilience. When the resilience scores were high, L A L A carriers reported the lowest anxiety scores in spite of DH or recent chronic stress.
These findings are in line with a study reporting that adolescent carriers of the L allele exhibited more anxiety symptoms related to stressful life events, for example, school or friendship problems (Ming et al., 2015). Contrary, Gunthert et al. (2007) found that the S allele rather than the L allele modified the effect of current daily stress on anxious mood in college students. These contradictory findings regarding the allelic direction of association may, however, be partly due to moderating positive influence as highlighted by the present study (c.f., Klauke, Deckert, Reif, Pauli, Zwanzger et al., 2011;Schiele, Ziegler et al., 2016). Consequently, a previous study has focused on the positive end that may counteract the deleterious impact of life stressors and serve to compensate for a biological risk profile. The interplay between anxiety traits and childhood trauma was found to be moderated by self-efficacy in 5-HTTLPR/rs25531 L A L A genotype carriers. Carriers of the L A L A genotype scored highest on different measures of anxiety if they had experienced childhood maltreatment, but only when general self-efficacy was low. When general self-efficacy was high, L A L A carriers had the lowest anxiety levels despite childhood adversities (Schiele, Ziegler et al., 2016). The present study showed for the first time, that beside major stressors minor stressors like DH and recent chronic stress interact with 5-HTTLPR in a GxE risk factor constellation and can be buffered by resilience.
The findings of the present study support the "differential susceptibility hypothesis" (Belsky et al., 2009) stating that a given genotype (in this case L A L A ) does not transfer vulnerability for anxiety per se, but is rather subject to positive as well as negative environmental influences. In relation to this study, L A L A carriers reported higher levels of anxiety in the presence of more DH or recent chronic stress when resilience was low but at the same time, L A L A carriers were able to counterbalance the negative effects of DH and recent chronic stress if their resilience was high. This illustrates that the L A L A genotype might rather constitute a "plasticity" factor than a "risk" genotype. Some limitations need to be considered regarding the present findings. Only cross-sectional data was used for the current analysis.
Although anxiety was assessed temporally after the stressexposures, given the close proximity (few days) between the assessment of anxiety and stress or hassles, overlap in time scales cannot be excluded in most cases. Thus, future studies should focus on clearly prospective-longitudinal research. In addition, genetic data or information on the Caucasian descent for some participants with genetic data was not available, reducing the overall available sample size for the current analysis. Finally, the total BeMIND sample is regional and characterized by a relatively high educational level restricting the generalization of the results to adolescents and young adults with other backgrounds.
Despite the limitations, our findings from a general population sample of adolescents and young adults emphasize the focus on DH and recent chronic stress in a GxE framework, which is in line with other studies also pointing towards the influence of DH on anxiety (Barrett & Heubeck, 2000;D'Angelo & Wierzbicki, 2003). In further studies, additional factors like epigenetic processes modulating gene function and temporally dynamic processes susceptible to environmental influences, such as daily life stress respect to anxiety (for a review see Gottschalk et al., 2020;Schiele & Domschke, 2018) in particular SLC6A4 DNA methylation (Domschke et al., 2014;Duman & Canli, 2015;Kang et al., 2013;Roberts et al., 2014;Schiele, Kollert et al., 2019) have to be considered.
In conclusion, the present study suggests that minor stressors, such as DH or recent chronic stress, in adolescent and young adults with low levels of resilience are in concert with higher levels of anxiety in carriers of the more active L A L A genotype. However higher levels of resilience buffer against the background of a genetic risk constellation resulting in the lowest level of anxiety in L A L A carriers with high levels of resilience despite the existence of DH or recent chronic stress. The integration of resilience in the GxE model regarding anxiety is broadening the scope by the incorporation of a "differential susceptibility" and "plasticity" framework. The consideration of positive and negative environmental influences along with genetic make-up carries potential for further research and clinical practice. Testing the effects of early interventions to build resilience targeted at individuals with higher DH or chronic stress and genetic susceptibility may be useful to prevent an escalation of distress and associated unfavorable health outcomes.

ACKNOWLEDGMENTS
The Behavior and Mind Health (BeMIND) study is part of the research program "The epidemiology of functional and dysfunctional behavioral and psychological factors in health and disease (EBP)" funded by the