One of the dimensions of personality with relevance to society and mental health is neuroticism (Cuijpers et al., Reference Cuijpers, Smit, Penninx, de Graaf, ten Have and Beekman2010; Lahey, Reference Lahey2009). Neuroticism can be defined as elevated stress reactivity and tendency to strong experiences of negative emotions such as sadness, anxiety, and anger, thus representing a broad negative affective dysregulation phenotype (Barlow, Ellard, Sauer-Zavala, Bullis & Carl, Reference Barlow, Ellard, Sauer-Zavala, Bullis and Carl2014). Numerous questionnaires measure this construct, and a strong association between these measures has been shown in the literature consistently (Aluja, Garcia & Garcia, Reference Aluja, Garcia and Garcia2004; De Fruyt, Van de Wiele & Van Heeringen, Reference De Fruyt, Van de Wiele and Van Heeringen2000; Draycott & Kline, Reference Draycott and Kline1995 ; Larstone, Jang, Livesley, Vernon & Wolf, Reference Larstone, Jang, Livesley, Vernon and Wolf2002; Waller, DeYoung & Bouchard, Reference Waller, DeYoung and Bouchard2016). Furthermore, the negative emotionality of the “Multidimensional Personality Questionnaire” and harm avoidance subscales of the “Temperament and Character Inventory” correlate highly with the general phenotype of neuroticism, albeit with slight differences.
Neuroticism represents a complex trait and has been linked to various genetic, neuroanatomical, and functional brain regions and is influenced further by environmental effects. Recent methods assessing the underlying genetic mechanisms of neuroticism used large genome-wide meta-analysis with populations over 100,000 (Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana and Cesarini2016; Smith et al., Reference Smith, Escott-Price, Davies, Bailey, Colodro-Conde, Ward and O’Donovan2016). Okbay et al. identified 11 variants to be associated with neuroticism, of which two tagged inversion polymorphisms in chromosomes 8 and 17 (Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana and Cesarini2016). A recent study combining epigenetic and genetic data integrated DNA methylation data (using life course consistent methylation quantitative trait loci meQTLs) and GWAS data and furthermore performed pathway analyses to identify biological pathways showing enrichment (Zhao & Liu, Reference Zhao and Liu2020). This study observed enrichment for genes with roles in 21 brain regions as well as in the reproductive and immune systems (Zhao & Liu, Reference Zhao and Liu2020). Neuroticism is a highly polygenic phenotype involving complex multifaceted neural (and likely also general physiological) circuitries and has differential utility in both population and clinical samples (Gottschalk & Domschke, Reference Gottschalk and Domschke2017; Grasby et al., Reference Grasby, Jahanshad, Painter, Colodro-Conde, Bralten, Hibar and Lind2020).
Environmental factors that seem to shape neuroticism are early developmental trauma and stressful life events (SLEs), as well as social roles and normative life transitions (Gottschalk & Domschke, Reference Gottschalk and Domschke2017; Specht et al., Reference Specht, Bleidorn, Denissen, Hennecke, Hutteman, Kandler and Zimmermann2014). Furthermore, neuroticism is influenced by an abundance of environmental factors that have implications for disease (Bucher, Suzuki & Samuel, Reference Bucher, Suzuki and Samuel2019; Lehto, Karlsson, Lundholm & Pedersen, Reference Lehto, Karlsson, Lundholm and Pedersen2019).
Although genetic and environmental effects are individually relevant, research on the interplay of genotype and environment in regard to personality formation and change is of primary interest. For instance, it has been shown that neuroticism is mediated by negative life events as well as that there is mediation by genetic effect on negative life events (Kandler, Bleidorn, Riemann, Angleitner & Spinath, Reference Kandler, Bleidorn, Riemann, Angleitner and Spinath2012). These findings underline the importance of further fine grain analysis by assessing specific environmental risk and protective factors such as childhood adversity, bullying, or parental bonding. A genome-wide interaction study detected gene–environment interactions for one SNP (rs115385310) with the broadly defined environmental risk factor of: “Felt hated by a family member as a child” (Werme, van der Sluis, Posthuma & de Leeuw, Reference Werme, van der Sluis, Posthuma and de Leeuw2021). A recent paper has posited an integrative model underlining the interplay between person-level variation and environmental variation (Wagner, Orth, Bleidorn, Hopwood & Kandler, Reference Wagner, Orth, Bleidorn, Hopwood and Kandler2020).
Using the strengths of GWAS and considering the polygenic character of neuroticism, polygenic risk score as a single genetic risk measure for neuroticism (PRS N ) has given new insights, especially in regard to GxE studies. For instance, a recent twin study used a gene–environment model to show an interaction between PRS N and rearing status (reared together versus apart) on depressive symptoms (Lehto et al., Reference Lehto, Karlsson, Lundholm and Pedersen2019). A study assessing prenatal maternal risk factors in relation to several PRS and child behavioral problems found that PRS N only predicted child internalizing behavioral problems, maternal alcohol use, and maternal anxiety during pregnancy (Ensink et al., Reference Ensink, de Moor, Zafarmand, de Laat, Uitterlinden, Vrijkotte and Middeldorp2020). An added layer of challenge is that early SLEs, for instance bullying, are partially heritable and separately associated with underlying polygenic risk scores (Schoeler et al., Reference Schoeler, Choi, Dudbridge, Baldwin, Duncan, Cecil and Pingault2019). Therefore, these gene–environment correlations also need to be taken into account.
In light of this background, we tested the contribution of gene–environment interaction to neuroticism in a unique general population twin cohort of young adults and adolescents with deep phenotyping. To elucidate the role of several key environmental factors on the development of neuroticism, we tested whether the molecular genetic risk score for neuroticism (PRS N ) interacts with environmental factors (i.e., parental bonding, bullying and childhood adversity) to influence neuroticism.
1. Methods
1.1. Sample
Data were derived from the first wave of the TwinssCan, a general population twin cohort that started including adolescent and young adult (age range = 15–35 years) twins (n = 796), their siblings (n = 43), and parents (n = 363) from April 2010 to April 2014 (Pries et al., Reference Pries, Guloksuz, Menne-Lothmann, Decoster, van Winkel, Collip and van Os2017; Pries, Klingenberg et al. Reference Pries, Klingenberg, Menne-Lothmann, Decoster, van Winkel, Collip and Guloksuz2020). The TwinssCan cohort comprises individuals fulfilling the inclusion criteria from the East Flanders Prospective Twin Survey (Derom et al., Reference Derom, Thiery, Peeters, Vlietinck, Defoort and Frijns2013; Derom et al., Reference Derom, Thiery, Rutten, Peeters, Gielen, Bijnens and Weyers2019), a prospective population-based, multi-birth registry positioned in Flanders, Belgium. Participants were excluded if they had a pervasive developmental disorder as indicated by caregivers. Sequential analysis based on sex, fetal membranes, umbilical cord blood groups, placental alkaline phosphatase, and DNA fingerprints was used to determine zygosity (Derom et al., Reference Derom, Thiery, Peeters, Vlietinck, Defoort and Frijns2013). All participants gave written informed consent, and parent(s) signed an informed consent for participants below the age of 18 years. The local ethics committee approved the study (Commissie Medische Ethiek van de Universitaire ziekenhuizen KU Leuven, Nr. B32220107766). The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
1.2. Environmental measures
1.2.1. Childhood adversity
Childhood adversity was assessed with the Dutch translation of the short version of the Childhood Trauma Questionnaire (CTQ) (Bernstein, Ahluvalia, Pogge & Handelsman, Reference Bernstein, Ahluvalia, Pogge and Handelsman1997). The CTQ comprises of five subscales: sexual abuse, emotional abuse, physical abuse, physical neglect, and emotional neglect; totaling 28 items of which 3 are Minimization/Denial validity items and 5 items per subscale. Participants were asked to rate on a scale from 1 “never” to 5 “always.” The CTQ has been validated for the Dutch population showing a Cronbach’s alpha of .91 for the physical abuse scale, .89 for emotional abuse, .95 for sexual abuse, .63 for physical neglect, and .91 for emotional neglect (Thombs, Bernstein, Lobbestael & Arntz, Reference Thombs, Bernstein, Lobbestael and Arntz2009). A continuous variable was constructed based on the total score of each participant called “childhood adversity” (CA) leaving out the validity items.
1.2.2. Parental bonding
The Parental Bonding Instrument is a self-report questionnaire of two parenting styles, Care and Overprotection, as measured by 25 items (Gordon Parker, Tupling & Brown, Reference Parker, Tupling and Brown1979). It is designed for mother and father separately. All items are measured on a 4-point Likert scale, ranging from “very unlikely” to “very likely.” Positive items are scored as 0 = very unlikely and 3 = very likely, while negative items are scored in reverse fashion. The “Care” subscale aims to measure facets of coldness and neglect versus affection and emotional warmth, while the “Overprotection” subscale focuses on facets of independence versus control and intrusion. Optimal parenting is resembled by high care and low overprotection, while neglectful parenting is resembled by low care and low overprotection (Craissati, McClurg & Browne, Reference Craissati, McClurg and Browne2002). The instrument has demonstrated strong psychometric properties, including long-term temporal stability and high internal consistency (Cronbach’s α = .74–.95) (Parker, Reference Parker1989; Wilhelm, Niven, Parker & Hadzi-Pavlovic, Reference Wilhelm, Niven, Parker and Hadzi-Pavlovic2005). To construct a continuous measure for our analysis, we used a compound sum-score of ascending care and the inverse of ascending control, i.e., a higher sum-score means higher care and lower control, consistent with previous work (Ambruster & Witherington, Reference Ambruster and Witherington2016).
1.2.3. Bullying
The Retrospective Bullying Questionnaire (RBQ) was used to assess previous experiences of bullying (Schäfer, Reference Schäfer, Korn, Smith, Hunter, Mora-Merchán, Singer, van der Meulen and Barrett2004). The RBQ consists of 44 multiple choice items and short answer questions. Past experiences with victimization (physical, verbal, and indirect) are assessed, both in primary school as well as in secondary school. Single items consisting of five-point rating scales were used to measure frequency, intensity, and duration of each of the types of incidents. High test–retest reliability has been documented for the RBQ using Spearman correlation coefficients of primary school (r = .88) and secondary school (r = .87) (Schäfer, Reference Schäfer, Korn, Smith, Hunter, Mora-Merchán, Singer, van der Meulen and Barrett2004), and a Cronbach’s alpha of .912 (Lund & Ross, Reference Lund and Ross2021). For this analysis, in line with the previous research (Schäfer, Reference Schäfer, Korn, Smith, Hunter, Mora-Merchán, Singer, van der Meulen and Barrett2004), we used a dichotomized variable for both primary and secondary school representing whether any victimization on any subscale (physical, verbal, or indirect) occurred.
1.3. Outcome: Neuroticism
The Dutch translation of the 12-item neuroticism scale of the Eysenck Personality Questionnaire (EPQ) was used to measure neuroticism (Sanderman, Eysenck & Arrindell, Reference Sanderman, Eysenck and Arrindell1991). The EPQ comprises 12 questions representing nervousness, emotional lability, feelings of guilt, and low self-esteem, in a no (scored as 0) or yes (scored as 1) format. A sum score (range: 0–12) was constructed following a Dutch manual (Sanderman, Arrindell, Ranchor, Eysenck & Eysenck, Reference Sanderman, Arrindell, Ranchor, Eysenck and Eysenck2012).
1.4. Genotyping
As reported previously (Pries, Klingenberg et al., Reference Pries, Klingenberg, Menne-Lothmann, Decoster, van Winkel, Collip and Guloksuz2020), genotypes of the twins and their siblings were generated on two platforms: the Infinium CoreExome-24 and Infinium PsychArray-24 kits. Quality control (QC) procedures were performed using PLINK v1.9 (Purcell et al., Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira, Bender and Sham2007) in both datasets separately (see Supplementary File for details).
1.5. PRS calculation and selection
Twelve PRS N were calculated based on the GWAS meta-analysis result for the harmonized Neuroticism scores harmonized by item response theory (IRT) from the Genetics of Personality Consortium (GPC) (de Moor et al., Reference de Moor, van den Berg, Verweij, Krueger, Luciano and Arias Vasquez2015; van den Berg et al., Reference van den Berg, de Moor, McGue, Pettersson, Terracciano, Verweij and Boomsma2014), as well as UK Biobank (UKB) cohort (Fry et al., Reference Fry, Littlejohns, Sudlow, Doherty, Adamska, Sprosen and Allen2017) (see Supplementary File for details).
The PRS N threshold to be used in the analyses was selected after comparing candidate models with different PRS N thresholds based on their R 2 values. As R 2 values, we used the marginal R 2 that summarizes the goodness-of-fit of a model as the proportion of explained variance by its fixed-effects terms to the total variance in the data (Nakagawa & Schielzeth, Reference Nakagawa and Schielzeth2013; Snijders & Bosker, Reference Snijders and Bosker2012; Xu, Reference Xu2003), using the sjstats package in R (Lüdecke, Reference Lüdecke2020). For the analysis, we used PRS N with the highest R 2 value which also was significantly associated with the outcome (p < .05).
1.6. Statistical analysis
We analyzed the association of the GPC and UKB trained PRS N , CA (Childhood Trauma Questionnaire sum score), parental bonding (separate for both parents; i.e. maternal and paternal bonding), and bullying (separately for primary and secondary school) with neuroticism (EPQ sum score) using interaction models. For the purpose of this analysis, parents and siblings were excluded. These data have a hierarchical structure due to the twin sample design, and multilevel mixed effect modeling was applied to take into account relatedness (Guo & Wang, Reference Guo and Wang2002; Hunter, Reference Hunter2021). In keeping with an earlier publication, the hierarchical structure consisted of subjects (level 1), who were part of twin pairs (level 2) (Pries, Klingenberg et al., Reference Pries, Klingenberg, Menne-Lothmann, Decoster, van Winkel, Collip and Guloksuz2020). These multilevel mixed-effects models accounted for variability associated with each level of nesting (Carlin, Gurrin, Sterne, Morley & Dwyer, Reference Carlin, Gurrin, Sterne, Morley and Dwyer2005; Simons et al., Reference Simons, Wichers, Derom, Thiery, Myin-Germeys, Krabbendam and van Os2009; Snijders & Bosker, Reference Snijders and Bosker1999). Positive skewness was observed in neuroticism. Therefore, all neuroticism scores were inverse rank normalized in accordance with previous research (Beasley, Erickson & Allison, Reference Beasley, Erickson and Allison2009). In consideration of all these factors, we used a multilevel mixed effects model with an unstructured covariance matrix using Stata (version 15.0.49). “Mixed” or “xtmelogit” commands were used, depending whether the dependent variable was continuous or dichotomous, respectively. The independent variables (PRS, CTQ, parental bonding score, and age) were standardized (mean = 0, SD = 1).
All models were controlled for known covariates (age and sex), including adjustment for ancestry using the first two genomic principal components (PCs), in keeping with previous research (Pries, van Os et al., Reference Pries, van Os, Ten Have, de Graaf, van Dorsselaer, Bak and Guloksuz2020). Interaction models included these covariates not only as additive effects but also as covariate × environment and covariate × PRS interaction terms in order to adequately control for confounding (Keller, Reference Keller2014). We report results after correction; full results are shown in tables. Additionally, we tested where there was any gene–environment correlation (rGE) present in our sample for the relevant variables.
2. Results
2.1. Sample characteristics
The total sample consisted of 778 twins of which 638 provided genetic data. This final sample consisted of 202 monozygotic twin individuals and 436 dizygotic twin individuals (319 twin pairs). Sixty percentage of the sample was female, with a mean age of 17 years (range 14–34). Table 1 reports detailed socio-demographic and sample characteristics.
2.2. Neuroticism’s variance explained by common SNPs
Twelve different PRSN thresholds (.5 – 5 × 10−8) were analyzed to determine the threshold with the highest variance explained. For the GPC PRS N , the threshold of .005 explained 4.8% of the variance in neuroticism with a p-value of .009. The UKB PRS N (threshold .2) explained 5.5% of the variance in neuroticism with a p-value of .015. There was not much heterogeneity in variance explained by the different thresholds across both sets (range 3.8% – 5.5%, as shown in Table 2). In order to explain the maximal variance, further analyses were conducted using the PRS N at p-threshold .005 (GPC) and .2 (UKB).
Note: *PRSN with highest R2 including p < 0.05 in GPC cohort. †PRSN with highest R2 including p < 0.05 in UKB cohort. Polygenic risk score for neuroticism (PRSN), genetics of personality consortium (GPC), United Kingdom Biobank (UKB).
2.3. Effects of PRSN, environmental factors, and interactions
Both PRS N and environmental factors (bullying, childhood trauma, and parental bonding) were associated with neuroticism (see Table 3). The results remained significant after correction for a priori covariates (age and sex). The directions of effect for the predictors were in the hypothesized direction. Results from the analyses using both PRS N indicated that higher neuroticism scores were associated with higher PRS N (GPC PRS N at .005: B .10, p < .01, 95% CI .03–.18; UKB PRS N at .2: B .09, p = .02, 95% CI .02–.17), higher childhood trauma (B .22, p < .01, 95% CI .15–.28), presence of bullying (primary school: B .39, p < .01, 95% CI .20–.57; Secondary school: B .43, p < .01, 95% CI .23–.63), and lower maternal and paternal bonding scores (maternal bonding: B −.21, p < .01, 95% CI −.28 – −.14; paternal bonding: B −.18, p < .01, 95% CI −.25 – −.12).
Note: *After adjustment for age and sex. Confidence interval (CI), polygenic risk score for neuroticism (PRSN), genetics of personality consortium (GPC), United Kingdom Biobank (UKB), childhood adversity as measured by CTQ, maternal and paternal bonding as measured by PBI, bullying in primary and secondary school as measured by RBQ.
With respect to the interaction models, we did not find statistical interactions between PRS N and environmental factors: childhood trauma, paternal and maternal bonding, and bullying (primary and secondary school), except solely for the interaction term of the GPC PRS N with maternal bonding (B − .08, p = .04, 95% CI − .15 – −.00, see Table 4). This result did not replicate using the UKB PRS N .
Note: *After adjustment for age and sex. †Significant result with p <0.05. Confidence interval (CI), polygenic risk score for neuroticism (PRSN), genetics of personality consortium (GPC), United Kingdom Biobank (UKB), childhood adversity as measured by CTQ, maternal and paternal bonding as measured by PBI, bullying in primary and secondary school as measured by RBQ.
When testing for rGE, we did not find any correlation between PRS N and environmental factors (Table 5).
Note: Confidence interval (CI), polygenic risk score for neuroticism (PRSN), genetics of personality consortium (GPC), and United Kingdom Biobank (UKB). ADDIN
3. Discussion
To the best of our knowledge, there has been sparse research investigating gene–environment interactions underlying the phenotype of neuroticism with the use of polygenic risk scores (Lehto et al., Reference Lehto, Karlsson, Lundholm and Pedersen2019; Werme et al., Reference Werme, van der Sluis, Posthuma and de Leeuw2021). In a unique adolescent and young adult twin sample, we showed that the PRS N , as well as parental bonding, exposure to bullying, and childhood trauma, were associated with phenotypic expression of neuroticism. However, there was no evidence for interaction between PRS N and any of the environmental factors.
Both PRS N were associated with neuroticism and explained 4.8% (GPC, .005 threshold) and 5.5% (UKB, 0.2 threshold) of the variance in this population, which is within the range of other publications (de Moor et al., Reference de Moor, van den Berg, Verweij, Krueger, Luciano and Arias Vasquez2015; Nagel et al., Reference Nagel, Jansen, Stringer, Watanabe, de Leeuw, Bryois and Posthuma2018; Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana and Cesarini2016; Werme et al., Reference Werme, van der Sluis, Posthuma and de Leeuw2021). The estimated variances explained in these publications ranges from .011% up to 15%. PRS N explained up to 4.2% of variance in phenotypical neuroticism in the UK Biobank cohort (UKB) (Nagel et al., Reference Nagel, Jansen, Stringer, Watanabe, de Leeuw, Bryois and Posthuma2018). In the GPC cohort, this was 15% in the two target datasets (de Moor et al., Reference de Moor, van den Berg, Verweij, Krueger, Luciano and Arias Vasquez2015). Although Okbay et al. used four cohorts (including UKB and GPC cohorts), they only found around .7% variance explained (Okbay et al., Reference Okbay, Baselmans, De Neve, Turley, Nivard, Fontana and Cesarini2016). These findings suggest that neuroticism has a highly complex genetic background including both common and rare variants.
When considering early and later life stressful environmental factors, we found an association between neuroticism and parenting style. Our findings suggest that affectionate and warm parenting styles without controlling or intrusive behavior by the parents are associated with a decline in neuroticism, which can be explored further as a potential net benefit in future studies. On the other hand, having been bullied, or having experienced any form of childhood abuse or neglect was associated with an increase in neuroticism.These findings are largely consistent with previous research (Huppert, Abbott, Ploubidis, Richards & Kuh, Reference Huppert, Abbott, Ploubidis, Richards and Kuh2010; Jeronimus, Ormel, Aleman, Penninx & Riese, Reference Jeronimus, Ormel, Aleman, Penninx and Riese2013; Ono et al., Reference Ono, Takaesu, Nakai, Ichiki, Masuya, Kusumi and Inoue2017; Seki et al., Reference Seki, Shimura, Miyama, Furuichi, Ono, Masuya and Inoue2020; Takahashi et al., Reference Takahashi, Suzuki, Matsumoto, Shirata and Otani2017). By investigating positive and negative life events with the 20-item List of Threatening Experiences, a study found that distant (occurring on average a year before follow-up) negative life events were associated with increased neuroticism, while distant positive life events were associated with decreased neuroticism (Jeronimus et al., Reference Jeronimus, Ormel, Aleman, Penninx and Riese2013). The study also showed a moderating effect of childhood adversity. Specifically, those who experienced childhood adversity had higher baseline neuroticism and less increase in neuroticism after distant negative life events but more decrease in neuroticism after distant positive life events. Our results confirm that childhood adversity is associated with an increased expression of neuroticism, whereas an optimal parenting style (e.g., high care and low overprotection) is associated with decreased neuroticism. Several studies found that optimal parenting had a decreasing effect on neuroticism (Ono et al., Reference Ono, Takaesu, Nakai, Ichiki, Masuya, Kusumi and Inoue2017; Takahashi et al., Reference Takahashi, Suzuki, Matsumoto, Shirata and Otani2017). Furthermore, neuroticism can be considered a mediator of the effect of the quality of parenting on depressive symptoms (Ono et al., Reference Ono, Takaesu, Nakai, Ichiki, Masuya, Kusumi and Inoue2017). Others showed that high perceived job stress and stress response in adult employees are still indirectly influenced by both parental overprotection and care via neuroticism, underscoring the long-term effects of suboptimal parenting in development (Seki et al., Reference Seki, Shimura, Miyama, Furuichi, Ono, Masuya and Inoue2020). When assessing mental well-being in a more general sense, it was found that a high care and low overprotection parenting style were associated with mental well-being that was largely mediated by personality (Huppert et al., Reference Huppert, Abbott, Ploubidis, Richards and Kuh2010). These findings emphasize the societal impact of optimal parenting and lower neuroticism on mental well-being.
Only a few studies have thus far utilized PRS N in GxE models. In a recent study, Lehto et al. investigated the gene–environment interaction between PRS N and early-life stress using raised together or apart as a proxy for childhood adversity (Lehto et al., Reference Lehto, Karlsson, Lundholm and Pedersen2019). They found that the PRS N only had a significant association with the expression of depression in reared-together twins in their cohort of older individuals, whereas the results showed a trend toward statistical significance for neuroticism. Their G × E analyses revealed considerably stronger effect of PRS N on neuroticism in the reared-together twins, suggestive of heterogeneity in neuroticism development depending on childhood adversity. They showed an interaction between PRS N and rearing status for depressive symptoms and a similar pattern for neuroticism and loneliness. More recently, a comprehensive genome-wide gene–environment interaction study showed that although there was some benefit to this approach, interaction effects appeared to not predict much more variance in phenotypical neuroticism beyond main effects (Werme et al., Reference Werme, van der Sluis, Posthuma and de Leeuw2021). We measured childhood adversity with a retrospective questionnaire and showed a significant association with neuroticism without any interaction with the PRS N . Some of these differences may be explained due to the use of different training datasets to calculate PRS. We used the GPC as well as UKB datasets to estimate PRS N , whereas Lehto et al. used a pooled data including the UKB. It may be likely that minor variations in estimating PRS N may result in differences in the final results. Additionally, differences in the definition of childhood adversity might explain parts of the observed difference in results.
Finally, we demonstrated that the exposure to bullying during childhood was associated with higher neuroticism scores and that this exposure did not show a statistical interaction with PRS N . In this regard, findings from the ALSPAC cohort in over 5000 participants showed no association between the risk of exposure to bullying and PRS N (Schoeler et al., Reference Schoeler, Choi, Dudbridge, Baldwin, Duncan, Cecil and Pingault2019). Previous reports showed an association between bullying and neuroticism (Rosta & Aasland, Reference Rosta and Aasland2018), as well as an association between neuroticism and being an aggressor (Pabon-Carrasco et al., Reference Pabon-Carrasco, Ramirez-Baena, Jimenez-Picon, Ponce Blandon, Martinez-Montilla and Martos-Garcia2019). Furthermore, there has been plenty of research on the link between neuroticism and depression (Christensen & Kessing, Reference Christensen and Kessing2006), but research on the interaction between neuroticism and exposure to bullying yields some contradictory results. A longitudinal study showed that although neuroticism predicted both depression and social anxiety, no significant interactions were evident between neuroticism and bullying victimization (Calvete, Orue & Gamez-Guadix, Reference Calvete, Orue and Gamez-Guadix2016).
Some strengths and limitations of our study need to be mentioned. A major strength of our research was the sample that consisted of a mixed-age group ranging from 14 to 34 years of age. In personality research, it is essential to capture early arising gender typical manifestations of neuroticism, which usually manifests earlier in women than in men (as early as 12–14 years of age) (De Bolle et al., Reference De Bolle, De Fruyt, McCrae, Lockenhoff, Costa, Aguilar-Vafaie and Terracciano2015). Some differences and contrasts have even been reported earlier (Borghuis et al., Reference Borghuis, Denissen, Oberski, Sijtsma, Meeus, Branje and Bleidorn2017; McCrae, Costa & Martin, Reference McCrae, Costa and Martin2005; McCrae et al., Reference McCrae, Costa, Terracciano, Parker, Mills, De Fruyt and Mervielde2002). However, early variation is often overlooked in predominantly older cohorts (Lehto et al., Reference Lehto, Karlsson, Lundholm and Pedersen2019). Our young sample is suited to capture this early variation but may be underpowered to detect gene–environment interaction. The marginally significant finding for the interaction between PRS N (GPC) and maternal bonding, which did not replicate with PRS N (UKB), possibly indicates that some interactions might have been detected if our sample had been larger. Therefore, further research using larger cohorts is warranted. One could also note that a twin population may not be considered a complete representation of the general population, thereby making generalizations toward other populations difficult. In GWAS, the population structure has an impact on polymorphism distributions varying in regard to neuroticism as well. For instance, a study in older adults with African and White European ancestries posited that existing PRS were derived from mostly or exclusively White European samples and this limits their applicability in other ancestries (Assari et al., Reference Assari, Javanbakht, Saqib, Helmi, Bazargan and Smith2020). Other research showed that it might be possible to predict across ancestries in regards to neuroticism (Docherty et al., Reference Docherty, Moscati, Peterson, Edwards, Adkins, Bacanu and Kendler2016). Our sample was a very homogenous cohort consisting of predominantly individuals of European ancestry. Therefore, we were able to apply PRS N conveniently. However, it must be noted our findings may not be generalizable to other ethnic ancestries. Another limitation is the general caveat of self-reporting bias including social desirability, especially relevant when assessing childhood adversity, and recall bias. Given our interest in neuroticism, a phenotype measuring a propensity for negative emotion, it might be possible that people with high neuroticism might have a heightened sense of the negative in their life and tend to report subjectively intensified childhood adversities in retrospect when responding to retrospective measures. Although we cannot preclude the influence of subjective experience of adversity that might be due to neuroticism as a whole phenotype, our assessment of gene–environment correlation (rGE) shows that at least the genomic portion of neuroticism does not seem to be correlated with childhood adversity. Further research should include the use of independent raters, objective indicators, and prospective designs to generate additional robust evidence. An additional direction to take would be broadening the assessed environmental factors, ideally aiming to assess the entire exposome using an agnostic approach (Lin et al., Reference Lin, Pries, Sarac, van Os, Rutten, Luykx and Guloksuz2022).
In conclusion, we showed that PRS N , as well as parental bonding, bullying, and childhood trauma, were independently associated with the phenotypic expression of neuroticism, but there was no evidence for gene–environment interaction. Although the etiology of neuroticism including highly polygenic background and various early environmental insults is slowly disentangled, there is still no clear plausible mechanistic pathway from genetic and environmental variation via intermediate phenotypes toward clinical expressions of psychopathology.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/pen.2023.2
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request under the condition of the approval of the TwinssCan steering committee.
Funding
The authors would like to acknowledge that the East Flanders Prospective Twin Survey (EFPTS) is partly supported by the Association for Scientific Research in Multiple Births and that the TwinssCan project is part of the European Community’s Seventh Framework Program under grant agreement HEALTH-F2-2009-241909 (Project EU-GEI). B.P.F. Rutten was funded by a VIDI award (no. 91718336) from the Netherlands Scientific Organization. S. Guloksuz and B. Rutten are supported by the YOUTH-GEMs project, funded by the European Union’s Horizon Europe program under the grant agreement number: 101057182.
Competing interests
The authors have no relevant financial or nonfinancial interests to disclose.
Ethical approval
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.