Introduction
Anxiety disorders affect over 20% of individuals during their lifetime (Kessler et al., Reference Kessler, Avenevoli, Costello, Georgiades, Green, Gruber and Merikangas2012), have an early age of onset (Kessler et al., Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters2005) and are currently increasing sharply in prevalence (Baker, Reference Baker2020; O'Connor, Downs, Shetty, & McNicholas, Reference O'Connor, Downs, Shetty and McNicholas2020). Both genetic and environmental influences are implicated in the development of anxiety, with twin heritability estimates ranging from 20% to 60% (Ask, Torgersen, Seglem, & Waaktaar, Reference Ask, Torgersen, Seglem and Waaktaar2014; Meier et al., Reference Meier, Trontti, Purves, Als, Grove, Laine and Mors2019; Polderman et al., Reference Polderman, Benyamin, de Leeuw, Sullivan, van Bochoven, Visscher and Posthuma2015), and estimates of heritability derived from genome-wide association studies (GWASs) ranging from 1.7% in childhood and adolescence to 31% in adulthood (for a review see Ask et al., Reference Ask, Cheesman, Jami, Levey, Purves and Weber2021). Notably, the upper range of heritability estimated from common genetic variants indexed by genome-wide analyses of anxiety [single-nucleotide polymorphism (SNP) heritability] overlaps with the lower end of estimates from twin studies. This is unusual, with SNP heritability substantially less than half that of twin heritability observed for most psychiatric disorders (Yang, Zeng, Goddard, Wray, & Visscher, Reference Yang, Zeng, Goddard, Wray and Visscher2017). This might suggest a larger role for common genetic variants in anxiety in particular. Effectively, evidence-based treatments exist for anxiety disorders, notably pharmacological approaches such as antidepressants and benzodiazepines (Baldwin et al., Reference Baldwin, Anderson, Nutt, Allgulander, Bandelow, den Boer and Wittchen2014), and psychological approaches such as cognitive behavioural therapy (CBT) (Carpenter et al., Reference Carpenter, Andrews, Witcraft, Powers, Smits and Hofmann2018; Hofmann & Smits, Reference Hofmann and Smits2008). However, only ~50% of individuals respond regardless of treatment type (Clark et al., Reference Clark, Layard, Smithies, Richards, Suckling and Wright2009; Cuijpers, Cristea, Karyotaki, Reijnders, & Huibers, Reference Cuijpers, Cristea, Karyotaki, Reijnders and Huibers2016; Loerinc et al., Reference Loerinc, Meuret, Twohig, Rosenfield, Bluett and Craske2015; Rush et al., Reference Rush, Trivedi, Wisniewski, Nierenberg, Stewart, Warden and Fava2006). Individual differences in treatment response are likely to be heritable, given known genetic influences on response to many aspects of the environment, such as life events and parenting (Kendler & Baker, Reference Kendler and Baker2007). Although there is evidence for familial clustering (Franchini, Serretti, Gasperini, & Smeraldi, Reference Franchini, Serretti, Gasperini and Smeraldi1998; O'Reilly, Bogue, & Singh, Reference O'Reilly, Bogue and Singh1994) and influence of common genetic variants (Tansey et al., Reference Tansey, Guipponi, Hu, Domenici, Lewis, Malafosse and Uher2013) in response to antidepressant treatment for depression, the heritability of psychological treatment response is currently unknown. The largest studies in this field to date, both of which are genome-wide analyses, were insufficiently powered to draw strong conclusions (Coleman et al., Reference Coleman, Lester, Keers, Roberts, Curtis, Arendt and Eley2016; Rayner et al., Reference Rayner, Coleman, Purves, Hodsoll, Goldsmith, Alpers and Eley2019).
Crucially, we do not yet know the specific mechanisms through which anxiety develops or recovery occurs, or how genetic influences contribute to these. Furthermore, using an experimental model of specific processes underlying disordered anxiety and treatment response has the advantage of a likely reduction in heterogeneity in the dataset (Scheveneels, Boddez, Vervliet, & Hermans, Reference Scheveneels, Boddez, Vervliet and Hermans2016). One possible set of experimental mechanisms is the acquisition and subsequent extinction of fears through direct learning experiences. These processes can be measured using fear conditioning paradigms. During the fear acquisition phase, participants are exposed to two neutral stimuli, known as conditional stimuli (CS), one of which is repeatedly paired with an aversive (e.g. a loud noise) or unconditional stimulus (US). The stimulus paired with the US is referred to as the CS+. The other stimulus is never paired with anything aversive and is referred to as the CS−. Over repeated trials, participants learn that the CS+ is associated with the aversive stimulus and typically demonstrate a ‘fear’ or ‘anxiety’ response to the CS+. This initial phase models processes implicated in anxiety development. Following acquisition, participants undergo an extinction phase where both CS+ and CS− are repeatedly presented without the aversive noise. The extinction phase models exposure-based treatment of fear and anxiety disorders used in CBT (Hofmann, Reference Hofmann2008), which developed out of the fear extinction literature (Eelen & Vervliet, Reference Eelen, Vervliet, Craske, Hermans and Vansteenwegen2006).
A common way of measuring conditioning is to require participants to rate how much they expected the aversive outcome to occur with the CS+ and CS− throughout the task (Lonsdorf et al., Reference Lonsdorf, Menz, Andreatta, Fullana, Golkar, Haaker and Merz2017). These assessments can be thought of as risk-estimates. In order to measure task-related learning while controlling for general interindividual differences in responsivity, a differential metric (risk-estimates for the CS− subtracted from the risk-estimates for the CS+) is often calculated (Lonsdorf et al., Reference Lonsdorf, Menz, Andreatta, Fullana, Golkar, Haaker and Merz2017). In order to investigate the underlying contributions of genes and environment to these mechanisms, sample sizes are needed far in excess of those typically included in laboratory administered studies of fear conditioning. Notably, only one study in the largest meta-analysis of fear conditioning to date contained over 100 participants (Duits et al., Reference Duits, Cath, Lissek, Hox, Hamm, Engelhard and Baas2015). One prior twin study obtained preliminary estimates of heritability of skin conductance responses during fear acquisition and extinction (0.20–0.46), and genetic overlap between them (Hettema, Annas, Neale, Kendler, & Fredrikson, Reference Hettema, Annas, Neale, Kendler and Fredrikson2003). However, although large in terms of fear conditioning, the sample size (173 twin pairs) was small for genetic analyses, as evidenced by large confidence intervals (CIs). Furthermore, the sample was split into two groups, each of which received a different stimuli set. Thus, although this provides preliminary evidence for the heritability of fear acquisition, extinction and their overlap, a larger, more heterogeneously administered study is required to consider this question.
The current study aimed to answer two important questions: (1) to what extent are fear acquisition and extinction influenced by genetic and environmental factors; and (2) are the genetic and environmental influences on the acquisition and extinction of fear shared between all processes, or are there distinct influences on each phase? We used the recently developed Fear Learning and Anxiety Response (FLARe) smartphone app (Purves et al., Reference Purves, Constantinou, McGregor, Lester, Barry, Treanor and Eley2019) to remotely administer a differential fear conditioning and extinction paradigm to a subset of twins from the Twins Early Development Study (TEDS) (Rimfeld et al., Reference Rimfeld, Malanchini, Spargo, Spickernell, Selzam, McMillan and Plomin2019). We assessed participant discrimination between the CS+ and CS− during early and late acquisition and extinction.
Methods
Sample
Participants were recruited by email invitation sent to 5934 twins enrolled in the TEDS (Rimfeld et al., Reference Rimfeld, Malanchini, Spargo, Spickernell, Selzam, McMillan and Plomin2019), a longitudinal birth cohort study of twins born in England and Wales between 1994 and 1996. Twins who agreed downloaded the FLARe app (Purves et al., Reference Purves, Constantinou, McGregor, Lester, Barry, Treanor and Eley2019) via the iTunes or Google Play Stores. The experiment was completed by 2554 individuals of whom 1937 were included in the analyses after excluding those who removed their headphones, reduced phone volume <50% or exited the app during the task. See Online Supplementary Fig. S1 for a detailed illustration of participant drop-out at each stage. We explored the characteristics of twins who chose to participate in the study (n = 2554) compared to those who did not (n = 3380). Our participant groups were more likely to be older [d = 0.07; t(5909.9) = 2.78, p = 0.006], female (odds ratio 1.67; 95% CI 1.50–1.86, p < 0.001) and to have lower socioeconomic status (d = 0.08; U = 866 163, p = 0.036) than those who chose not to take part. However, there were no statistically significant differences in anxiety scores between respondents and non-respondents (d = 0.05; U = 1 065 914.5, p = 0.191).
Twin zygosity was determined by parental responses to a twin similarity questionnaire. This has 95% concordance with genotyped zygosity (Price et al., Reference Price, Freeman, Craig, Petrill, Ebersole and Plomin2000). The sample consisted of 250 complete monozygotic pairs (MZF = 180, MZM = 70), 288 dizygotic pairs (DZF = 131, DZM = 50, DZopp = 107) and 860 singletons (F = 549, M = 312). See Online Supplementary Table S1 for a description of the total sample including age, gender and self-reported clinician-provided diagnoses of any anxiety or depressive disorder.
Experimental procedure
Participants were given instructions designed to maintain optimal experimental conditions, ensure headphone usage and maximum phone volume (see online Supplementary Figs S2 and S3). They were asked to complete the task in a single, undisturbed session. The fear conditioning procedure began with a fear acquisition phase during which 12 presentations each of a large and small circle (CS) were shown in a pseudo-randomised sequence on a background image of an outdoor scene. One of these (CS+) was paired with the aversive ‘unconditional stimulus’ (US; a loud scream sound) during the final 500 ms of the trial on nine out of 12 presentations (75% reinforcement). The other was never paired with the US (CS−). The use of the large v. small circle as the CS+ was counterbalanced across participants. See online Supplementary material for trial presentation rules. Participants then had a break of at least 10-min, during which they were unable to continue with the task due to a built-in time-based restriction within the app and completed a series of online questionnaires. This was followed by a fear extinction phase. This break was enforced in order to ensure a clear temporal distinction between the acquisition and extinction phases. The same large and small circles were shown 18-times each on the background of an indoor living room scene. Neither shape was paired with the aversive stimulus. We used 18 trials in the extinction phase to ensure full extinction prior to the end of the task, given the remote delivery. After the extinction phase, participants were redirected to an external website (Qualtrics, 2019) to answer questions including whether or not they removed their headphones during the task. See Fig. 1 for a schematic diagram of the experimental procedure.
Fig. 1. Schematic diagram of fear conditioning procedures as implemented in the FLARe app. This figure shows trial and overall task structure of the fear conditioning task as implemented in the Fear Learning and Anxiety Response (FLARe) app. CS: conditional stimuli. Context: background image of an outdoor garden scene (acquisition) or indoor living room scene (extinction) displayed behind CS during each trial. US: unconditional stimulus, a loud human scream played at maximum phone volume.
Outcome measures
During each trial, participants were asked to rate how much they expected to hear the aversive stimulus on a scale ranging from 1 (completely certain NO scream will occur) to 9 (completely certain a scream WILL occur). Risk-estimates (sometimes known as expectancy ratings) for the CS− were subtracted from those of the CS+ to obtain differential scores, an index of how well participants were able to differentiate between the two experimental cues. Repetition is a core component of memory and learning (Hintzman, Reference Hintzman and Federmeier1976), and evidence from neuroimaging paradigms demonstrates that the association between fear conditioning and brain regions varies across time within phase (LaBar, Gatenby, Gore, LeDoux, & Phelps, Reference LaBar, Gatenby, Gore, LeDoux and Phelps1998). Thus, in order to assess rapid acquisition of fear learning relative to later, consolidated fear learning after several repetitions, the post hoc decision was taken to consider fear acquisition and extinction in early and late stages. Differential scores created from risk-estimates made during the first and final thirds of the acquisition phase and the first third of the extinction phase were retained to assess the initial development, consolidation and extinction of fear learning respectively (see Fig. 2). Differential scores for the last third of the extinction trials (late extinction) were not analysed in twin modelling as by this stage nearly all participants had reduced their risk-estimates for both stimuli to ~1 (see Table 1).
Fig. 2. Average risk-estimates per trial across all participants during fear conditioning task. Average risk-estimate per stimulus, per trial, averaged across all participants. Dashed lines indicate standard error of the mean (note, these intervals are narrow). CS+, conditional stimulus paired with the aversive scream; CS−, conditional stimulus never paired with the aversive scream.
Table 1. Descriptive statistics and twin pair intraclass correlations for fear conditioning risk estimates
This table compares amean and standard deviation (s.d.) for the average risk-estimate per stimulus (CS+ or CS−) during the first third of acquisition trials (initial development), last third of acquisition trials (consolidation), first third of extinction trials (extinction) and the last third of extinction trials (late extinction), with bintraclass correlations (bootstrapped 95% CIs) between the members of twin pairs for the caverage difference in risk-estimates between the CS+ and CS− for each experimental phase (initial development, consolidation and extinction).
MZ, monozygotic twins; DZ, dizygotic twins; CS+, the conditional stimulus paired with the aversive scream; CS−, the conditional stimulus that is never paired with the aversive scream; CS differential, difference between CS+ and CS−.
Statistical analyses
Data preparation
Outcome variables (initial development, consolidation and fear extinction) were age and sex regressed to avoid artificial inflation of twin correlations (McGue & Bouchard, Reference McGue and Bouchard1984), and the residuals were normalised using square-root transformation in R.
Multivariate twin modelling
Monozygotic (MZ) twins share 100% of their segregating genes, dizygotic (DZ) twins share ~50%, while both MZ and DZ twins share their rearing environments (Rijsdijk & Sham, Reference Rijsdijk and Sham2002). Monozygotic and dizygotic twin similarity can be compared to estimate genetic and environmental influences on a trait. The degree to which MZ twins are more similar than DZ twins reflects approximately half (100–50%) of the genetic influence (labelled A; additive genetic influence). Twin resemblance not due to genetic factors is attributed to environmental factors that make twins more similar (common environmental factors, labelled C). Finally, the extent to which the MZ twin correlation differs from 1 reflects the non-shared environment and residual/error variance (labelled E).
A trivariate correlated factors solution of the Cholesky decomposition model was applied (see Fig. 3). This estimates the relative influence of latent factors A, C and E on each of the fear conditioning variables as well as the correlations between the genetic and environmental influences on each (Loehlin, Reference Loehlin1996). These parameters are then used to estimate the degree to which phenotypic correlations are attributable to genetic and environmental influences.
Fig. 3. Trivariate-correlated factors solution showing genetic and environmental influences on the initial development, consolidation and extinction of fear conditioning. This figure shows the standardised path estimates and 95% CIs for the AE trivariate-correlated factors solution of the Cholesky model. A, additive genetic effects; E, non-shared environment effects. Note that A and E present the proportion of phenotypic variance in each outcome accounted for by additive genetic and non-shared environment effects respectively. A and E for each outcome will sum to 100%. The curved paths show the correlations between the A and E factors for each outcome.
Genetic modelling was undertaken in R using the OpenMx package (Boker et al., Reference Boker, Neale, Maes, Wilde, Spiegel, Brick and Fox2011). OpenMx used full information maximum likelihood to estimate structural equation parameters, an effective method to deal with missing data (Boker et al., Reference Boker, Neale, Maes, Wilde, Spiegel, Brick and Kirkpatrick2016; Jelicić, Phelps, & Lerner, Reference Jelicić, Phelps and Lerner2009). Model fit was assessed using χ2 and Akaike's information criterion.
Results
Differential fear conditioning
As shown in Fig. 2, the difference between the CS+ and CS− was largest during the last four trials of acquisition, the phase we refer to as consolidation. The difference was smallest during late fear extinction when reinforcement of either stimulus had ceased. During the first four trials of acquisition (initial development), participants already showed a greater expectation of the scream occurring to the CS+ than the CS−, see first row, first section of Table 1 (meanCS+ = 6.05, s.d.CS+ = 1.63, meanCS− = 3.70, s.d.CS− = 1.56). By the consolidation phase (second row), risk-estimates had further increased for the CS+ and decreased for the CS−. Average risk-estimates reduced for both stimuli over the course of extinction. During the first third of extinction trials, ratings were similar to average risk estimates during the initial development of fear. By the end of extinction, almost all participants had ceased to report any expectation of a scream occurring for either stimulus.
Trivariate twin modelling
See Table 1 middle section for monozygotic and dizygotic twin correlations for each fear conditioning variable and Table 2 for cross-twin cross-trait genetic and phenotypic correlations. We only report results from multivariate modelling, as these are better powered compared to univariate models. The ACE model produced C estimates that were zero and non-significant (see online Supplementary Fig. S4), and we therefore tested a more parsimonious AE model. This did not result in loss of fit and was therefore selected as the final model (see online Supplementary Table S2 for model fit information).
Table 2. Cross-twin cross-trait (lower) and phenotypic (upper) correlations with proportion of variance explained by A and E
This table shows cross-twin cross-trait correlations (95% CI) for MZ (top) and DZ (bottom) twin pairs (below diagonal) and phenotypic correlations (95% CI) with the standardised proportion of phenotypic association accounted for by A/E (above diagonal). Estimate of heritability from multivariate twin analyses (95% CI) shown on the diagonal. CIs are obtained by bootstrapping over 100 iterations.
Estimates from this model are shown in Fig. 3. Initial development, consolidation and extinction of fear were all significantly influenced by additive genetic factors (A). For example, the estimate of heritability (a 2) for consolidation was 0.29 (0.18–0.39). In addition, there were significant shared genetic effects (r a) between all variables, with genetic correlations being highest between initial development and consolidation of fear [rG = 0.99 (0.97–1.00)]. In comparison, genetic correlation were significantly lower between each of the acquisition phases and extinction of fear [initial development and extinction: rG = 0.42 (0.14–0.78); consolidation and extinction: rG = 0.44 (0.05–0.75)]. The effect of non-shared environment (E) was also significant for all variables (range e 2 = 0.71–0.83), with significant non-shared environmental correlations (0.31–0.49) between all variables. See Table 2 for the standardised proportion of the phenotypic variance between variables accounted for by A and E respectively.
Discussion
We present the largest twin study of fear conditioning to date, providing evidence for genetic and non-shared environmental influences. Discriminative fear learning is significantly heritable during the initial development, consolidation and extinction of fear conditioning. We demonstrated significant contributions from non-shared environmental influences (which includes measurement error), but no evidence for the contribution of common environment. Shared genetic effects between initial development and the consolidation of fear approached unity. This indicates that the genetic variants involved in the initial development of fear learning are virtually the same as those involved in the later consolidation of fear learning. The genetic correlations between both stages of fear acquisition and extinction were moderate and significant, indicating either that some of the genetic variants implicated in fear learning also play a role in the extinction of these learned associations. However, these genetic correlations were around 0.5, indicating that roughly half the genetic factors on extinction of fear are different from those on fear learning. Although a previous study provided preliminary evidence for complete genetic overlap between acquisition and extinction of fear, these results were difficult to interpret as CIs were not included for estimates from the multivariate model (Hettema et al., Reference Hettema, Annas, Neale, Kendler and Fredrikson2003). Our study provides further support for significant shared genetic influences on acquisition and extinction. The prior study also found that both acquisition and extinction of fear were robustly heritable, with heritability estimates for fear extinction substantially greater than we find (h 2: 0.34–0.36). This may be due to differences in our paradigm (in lab v. remotely administered) and resultant measurement error, or due to differences in outcome variables (absolute v. differential expectancy ratings). However, taken together, the results from present and prior studies provide strong support for the role of genetic influences in the acquisition and extinction of fear.
Findings should be considered in light of some limitations. First, the remote delivery of the experiment necessitates lower experimental control compared to laboratory delivery. This is particularly reflected in the number of individuals who removed their headphones during the task, and thus excluded from analyses. Second, the sample was drawn from a population-based twin cohort. In future studies, it will be important to test whether there are differences in the genetic and environmental contributors to fear conditioning between individuals with anxiety disorders compared to healthy individuals.
Genetic influences on fear acquisition
We explored genetic influences on two stages of fear acquisition. Initial fear development was calculated using risk-estimate differences for the CS+ and CS− during the first third of acquisition trials. Fear consolidation was indexed by the risk-estimates differences during the final third of trials. Although initial development and consolidation shared all of the same genetic influences, genetic factors accounted for more variance in fear consolidation. A persistent reduced capacity to distinguish between cues paired and never paired with threat is similar to the tendency to interpret neutral stimuli as threatening (negative interpretation bias). This cognitive bias is a core feature of anxiety in adults and children (Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg, & van IJzendoorn, Reference Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg and van IJzendoorn2007; Dudeney, Sharpe, & Hunt, Reference Dudeney, Sharpe and Hunt2015; Lau & Waters, Reference Lau and Waters2017; Lester, Lisk, Carr, Patrick, & Eley, Reference Lester, Lisk, Carr, Patrick and Eley2019). Generalised anxiety disorder (GAD), in particular, is associated with interpretation biases (Eysenck, MacLeod, & Mathews, Reference Eysenck, MacLeod and Mathews1987; Eysenck, Mogg, May, Richards, & Mathews, Reference Eysenck, Mogg, May, Richards and Mathews1991; Hayes, Hirsch, Krebs, & Mathews, Reference Hayes, Hirsch, Krebs and Mathews2010; Mathews, Richards, & Eysenck, Reference Mathews, Richards and Eysenck1989; Mogg et al., Reference Mogg, Bradley, Miller, Potts, Glenwright and Kentish1994), which we have previously shown to be heritable (Brown et al., Reference Brown, Waszczuk, Zavos, Trzaskowski, Gregory and Eley2014; Eley et al., Reference Eley, Gregory, Lau, McGuffin, Napolitano, Rijsdijk and Clark2008; Lau, Belli, Gregory, & Eley, Reference Lau, Belli, Gregory and Eley2014). In sum, the consolidation of fear learning is probably more relevant than initial fear learning to understanding anxiety disorders. Notably, the heritability of fear consolidation (29%) is similar to estimates of GAD from previous twin studies (Hettema, Neale, & Kendler, Reference Hettema, Neale and Kendler2001). This supports the possibility that the consolidation of fear conditioning is a relevant mechanism underlying disordered anxiety.
Genetic and environmental influences on fear extinction
The modest but significant heritability of fear extinction (15%) suggests that a relatively small degree of variation in fear extinction in healthy individuals is influenced by genetic factors. The only prior twin study of fear conditioning to date (Hettema et al., Reference Hettema, Annas, Neale, Kendler and Fredrikson2003) estimated the heritability of fear extinction of the CS+ specifically to be 36% (95% CI 18–52), and the CS− 34% (95% CI 16–50). The CIs for the current study include estimates up to 25% (95% CI 4–25). It is possible that the range of overlap between CIs from the prior and current study (16–25) may represent the ‘true’ range of heritability. There have been no twin studies of treatment response for anxiety, and two genome-wide meta analyses of CBT treatment response have failed to derive significant heritability estimates despite having statistical power to detect heritability of 30% or greater (Coleman et al., Reference Coleman, Lester, Keers, Roberts, Curtis, Arendt and Eley2016; Rayner et al., Reference Rayner, Coleman, Purves, Hodsoll, Goldsmith, Alpers and Eley2019). This study thus adds to our limited knowledge about the relative role for genes and the environment in fear extinction, a mechanism underpinning exposure-based treatment, a core component of CBT for anxiety disorders.
The high estimate of the non-shared environment (e 2 = 85%) influencing fear extinction in this study partially reflects high measurement error given our task (Purves et al., Reference Purves, Constantinou, McGregor, Lester, Barry, Treanor and Eley2019). Furthermore, the use of difference scores as an outcome, which, while reducing the total number of variables examined, aggregates the error variance of each individual stimuli, and thus inflates the overall non-shared environmental estimate.
There are several plausible contributors to variable treatment response over and above direct genetic influences. These include specific life experiences, and individual characteristics such as unemployment, low educational attainment and poor interpersonal relationships that are also known to reduce treatment efficacy (DeRubeis et al., Reference DeRubeis, Cohen, Forand, Fournier, Gelfand and Lorenzo-Luaces2014; Mojtabai, Reference Mojtabai2017; Newman, Llera, Erickson, Przeworski, & Castonguay, Reference Newman, Llera, Erickson, Przeworski and Castonguay2013; Renaud, Russell, & Myhr, Reference Renaud, Russell and Myhr2014). Specific disorder profiles, including greater symptom severity, comorbidity with other mental health disorders and poor treatment adherence are associated with poor treatment response in anxious adults and children (Hudson et al., Reference Hudson, Lester, Lewis, Tropeano, Creswell, Collier and Eley2013, Reference Hudson, Rapee, Lyneham, McLellan, Wuthrich and Schniering2015; Rayner et al., Reference Rayner, Coleman, Purves, Hodsoll, Goldsmith, Alpers and Eley2019; Wergeland et al., Reference Wergeland, Fjermestad, Marin, Bjelland, Haugland, Silverman and Heiervang2016). Understanding modifiable personal risk factors may help identify areas for intervention that could serve as a precursor or adjunct to therapy to further enhance the effect of exposure within a personalised medicine framework.
Genetic and environmental overlap between fear acquisition and extinction
There were moderate genetic (r a = 0.41–0.44) and non-shared environment (r e = 0.31–0.41) correlations between fear acquisition and extinction. If fear extinction is considered a credible marker underlying treatment response (Craske, Hermans, & Vervliet, Reference Craske, Hermans and Vervliet2018), then at least some of the genetic markers underlying treatment response are the same as those underlying anxiety development, and approximately half the genes relevant to extinction should be identified in GWAS of anxiety. This ‘cause informs cure’ (Uher, Reference Uher2008) perspective would mean that findings from the rapidly growing anxiety genetics field would be relevant to understanding genetic influences on psychological treatment response.
We note genetic correlations between fear acquisition and extinction were below 0.5, indicating that distinct influences exist. Crucially, this indicates understanding the genetics of anxiety disorder is not sufficient to fully understand the genetics of psychological treatment response. It continues to be important to find ways of investigating treatment response as a distinct phenotype in large samples of individuals with sufficient depth of phenotyping to explore the role of variance in a range of environmental and individual factors within genetically informed designs.
Conclusions
We presented robust evidence for genetic influences on fear acquisition and extinction, shedding light on possible mechanisms underlying known genetic influences on anxiety. We have also shown moderate genetic overlap between fear acquisition and extinction. This indicates that genetic susceptibility to developing fears is likely to have a moderate influence on reducing learned fears, for example through exposure-based treatment. Future studies using genotyped samples with fear conditioning or treatment outcome data will enable more detailed investigation of the genetics of the development and treatment of anxiety.
In conclusion, these findings represent a significant advance in the genetics of fear acquisition and extinction, two key processes underlying the development and exposure-based treatment of anxiety disorders.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721002580.
Financial support
K.L. Purves was funded during this work by the Alexander von Humboldt Foundation and the UK Medical Research Council. T. McGregor is supported by the UK Medical Research Council (MR/N013700/1). T.C. Eley and G. Breen are part-funded by a programme grant from the UK Medical Research Council (MR/V012878/1). The TEDS study has received repeated programme grant funding from the UK Medical Research Council (programme grant at time of data collection MR/M021475/1). This study presents independent research part-funded by the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.
Conflict of interest
The author(s) declare no conflicts of interest with respect to the authorship or the publication of this article.
Ethical standards
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.
Open access
