Introduction
Although many psychological factors have been established as contributing to optimal mental health (including purpose in life, autonomy, positive affect; Keyes, Reference Keyes2007), we lack knowledge regarding the underlying neurobiological mechanisms by which such factors relate to wellbeing, which is comprised of two distinct, yet correlated, sub-constructs: subjective and psychological wellbeing. Subjective wellbeing relates to factors such as happiness, positive affect and life satisfaction, while psychological wellbeing refers to personal attributes such as autonomy, mastery, self-acceptance and finding a meaningful life purpose (Diener, Suh, Lucas, & Smith, Reference Diener, Suh, Lucas and Smith1999; Ryff & Singer, Reference Ryff and Singer2008). Although both sub-constructs have been extensively linked to positive outcomes such as decreased physical illness, greater work productivity and superior psychosocial functioning (e.g. Steptoe, Deaton, & Stone, Reference Steptoe, Deaton and Stone2015), they have often been examined independently of one another. However, it has been shown that they both contribute important and unique variance to total wellbeing (Henderson & Knight, Reference Henderson and Knight2012; Keyes & Annas, Reference Keyes and Annas2009), highlighting the importance of examining both aspects of wellbeing as one composite construct.
Reappraisal of emotional information is considered to be an important determinant of mental health outcomes including wellbeing (Davidson, Reference Davidson2004). Despite multiple behavioural studies showing an association between wellbeing and reaction time or accuracy to emotional stimuli, the corresponding evidence using neuroimaging methods is sparser and somewhat contradictory. In those neuroimaging studies available, wellbeing is usually only measured using subjective wellbeing or psychological wellbeing measures, but rarely both. Moreover, understanding whether the association between wellbeing and emotion processing is driven by genetics or environment (as can be evaluated using twin studies) would assist in understanding the underlying factors that drive this association. Our objective in this study was to address these gaps in knowledge assessing whether self-reported wellbeing (both subjective and psychological wellbeing) is associated with specific neural circuit regions probed by functional neuroimaging during the processing of salient emotional stimuli, and the extent to which such associations may be heritable and driven by shared genetics or environment.
Behavioural studies have shown a pattern of increased attentional bias to positive (i.e. happy) stimuli and reduced sensitivity to negative (i.e. fear) stimuli in individuals with higher levels of subjective wellbeing and life satisfaction (Sanchez & Vazquez, Reference Sanchez and Vazquez2014; Vittersø, Oelmann, & Wang, Reference Vittersø, Oelmann and Wang2009; Yu & Li, Reference Yu and Li2012). In our previous study with 1480 healthy twin participants from the TWIN-E cohort, we also observed an association between faster behavioural responses and the identification of happy faces in those with high wellbeing levels (Routledge et al., Reference Routledge, Williams, Harris, Schofield, Clark and Gatt2018). Prior functional magnetic resonance imaging (fMRI) studies have begun to address a complementary question, as to whether neural circuits in the human brain engaged by emotion processing are also associated with components of wellbeing – either subjective (defined as happiness and pleasure; Ryan & Deci, Reference Ryan and Deci2001) or psychological (defined as having a meaningful and purposeful life; Ryff & Singer, Reference Ryff and Singer2008). Higher levels of subjective wellbeing have been linked to greater amygdala activation in response to positive images (Cunningham & Kirkland, Reference Cunningham and Kirkland2014), and also with stronger activations in parieto-temporal regions in response to negative facial emotion stimuli (Ren, Shi, Wei, & Qiu, Reference Ren, Shi, Wei and Qiu2019). Subjective wellbeing has also been associated with resting-state fMRI metrics (e.g. regional homogeneity; Kong et al., Reference Kong, Liu, Wang, Hu, Song and Liu2015; Kong, Wang, Song, & Liu, Reference Kong, Wang, Song and Liu2016; spontaneous low frequency fluctuations; Kong, Hu, Wang, Song, & Liu, Reference Kong, Hu, Wang, Song and Liu2015) and dynamic frontal connectivity between the default mode, salience and frontal-parietal networks (Shi et al., Reference Shi, Sun, Wu, Wei, Chen, Yang and Qiu2018).
Within the psychological wellbeing domain, individuals with greater psychological wellbeing levels have been shown to display longer (sustained) neural activity in the striatum and dorsolateral prefrontal cortex when viewing positive images (Heller et al., Reference Heller, van Reekum, Schaefer, Lapate, Radler, Ryff and Davidson2013), and increased activity in the ventral anterior cingulate cortex in response to negative information (van Reekum et al., Reference van Reekum, Urry, Johnstone, Thurow, Frye, Jackson and Davidson2007). Altogether, these findings suggest that neural circuits engaged during emotion processing are associated with wellbeing, although it is unclear to what extent heterogeneity in the results are due to differences in emotion tasks (such as the use of emotional scenes v. faces) and/or different ways of assessing subjective or psychological components of wellbeing. In this study, we used a well-standardised facial emotion task, an established measure of wellbeing that assesses both subjective and psychological components, and a well-powered sample to assess the association between wellbeing and whole-brain neural activation engaged by emotion stimuli.
By utilising a twin sample, it is also possible to quantify the amount of variance in a wellbeing–neural activation association that is due to genetic v. environmental factors. To the best of our knowledge, this has not yet been investigated despite previous studies reporting some heritability for emotion-related processes (unrelated to wellbeing). Behaviourally, the heritability of reaction times to emotional faces is small to moderate, ranging from 27% to 37% (Routledge et al., Reference Routledge, Williams, Harris, Schofield, Clark and Gatt2018). Studies utilising electroencephalography have also reported significant heritability for event-related potentials during emotional faces processing, ranging from 42% to 64%, suggesting that neural responses to such stimuli are influenced by genetic factors to some degree (Anokhin, Golosheykin, & Heath, Reference Anokhin, Golosheykin and Heath2010; Shannon, Patrick, Venables, & He, Reference Shannon, Patrick, Venables and He2013). However for fMRI, Côté et al. (Reference Côté, Beauregard, Girard, Mensour, Mancini-Marïe and Pérusse2007) reported no evidence of heritability for processing sadness (elicited via sad film excerpts) in 104 8-year-old twin pairs in two a priori defined regions of interest (medial and ventrolateral prefrontal cortex). The authors concluded that emotional experiences may instead be linked to the child's personal social relationships (i.e. unique environmental factors). Thus far, no other studies to date have examined the heritability of neural activity related to face emotion processing using fMRI. Furthermore, it is not yet known whether such regions may also be associated with wellbeing. An examination of a broader spectrum of emotions (both positive and negative) and their relationship with wellbeing levels is needed, as well as a focus on a broader age group given that emotion processing networks stabilise mostly in early adulthood (Vink, Derks, Hoogendam, Hillegers, & Kahn, Reference Vink, Derks, Hoogendam, Hillegers and Kahn2014).
The goal of the current study was to examine the association between overall wellbeing, both subjective and psychological components, and neural circuits engaged during emotion processing in a subset of healthy adult twin participants from the TWIN-E cohort (for more detail about the cohort, see Gatt et al., Reference Gatt, Korgaonkar, Schofield, Harris, Clark, Oakley and Williams2012). We used fMRI in order to examine neural activity engaged by facial expressions of emotion using a well-standardised task (Korgaonkar, Grieve, Etkin, Koslow, & Williams, Reference Korgaonkar, Grieve, Etkin, Koslow and Williams2013). Our first working hypothesis was that higher levels of wellbeing would be associated specifically with neural activity elicited by positive emotional expressions of happiness, consistent with our previous behavioural findings (Routledge et al., Reference Routledge, Williams, Harris, Schofield, Clark and Gatt2018). Our second working hypothesis was that this association would also be driven more by genetic rather than environmental factors, also as identified by this previous study.
Material and methods
Participants
A total of 263 healthy same-sex monozygotic (MZ) and dizygotic (DZ) twins from the larger TWIN-E cohort study completed the MRI component (see Gatt et al., Reference Gatt, Korgaonkar, Schofield, Harris, Clark, Oakley and Williams2012). Inclusion criteria included being a same-sex, healthy adult twin, with English as their first language, European ancestry and an absence of past/current psychiatric illness or brain injury. Thirty-three participants were dropped due to missing data or being an incomplete twin pair. The final study sample for this study consisted of 230 participants (76 MZ and 39 DZ pairs; see Table 1). The study received approval from the Human Research Ethics Committee of the University of Sydney (03–2009/11430), and all participants gave their written informed consent to participate.
Table 1. Demographic characteristics for the final sample
Total number of the sample reflects the final sample of participants included in all analyses (fMRI, linear mixed model, univariate and multivariate twin models). COMPAS-W = total composite measure of wellbeing; DASS-42 = total measure of negative mood symptoms including depression, anxiety and stress.
Psychometric measures
Self-report scales were used to measure wellbeing (the COMPAS-W scale of wellbeing; Gatt, Burton, Schofield, Bryant, & Williams, Reference Gatt, Burton, Schofield, Bryant and Williams2014) and negative mood symptoms (the DASS-42; Lovibond & Lovibond, Reference Lovibond and Lovibond1995). More specifically, the COMPAS-W is a composite scale that measures both subjective and psychological wellbeing, while the DASS-42 measures depression, anxiety and stress symptoms (for more detail, see S1 in Supplemental Materials). Z-scores for the COMPAS-W and log-transformed scores for the DASS-42 were used in the current study to index the participant's levels of wellbeing and negative mood symptoms, respectively. These measures were part of a larger WebQ online assessment, which was a test battery of self-report questionnaires on physical and emotional health, wellbeing, and environmental factors (for a more detailed list of the measures collected, please refer to Gatt et al., Reference Gatt, Korgaonkar, Schofield, Harris, Clark, Oakley and Williams2012).
fMRI paradigm
The MRI session was part of the TWIN-E study, which consisted of five fMRI tasks that were performed within a single session in the scanner to examine cognitive and emotional functioning. For the purposes of the current study, only the parameters from the conscious viewing of facial emotion of expression task are discussed here (Korgaonkar et al., Reference Korgaonkar, Grieve, Etkin, Koslow and Williams2013; Williams et al., Reference Williams, Korgaonkar, Song, Paton, Eagles, Goldstein-Piekarski and Etkin2015).
The emotion task consisted of 240 standardised faces that depicted fear, anger, disgust, sadness, happiness or neutral expressions. Each emotion was presented in a block of eight trials of different faces but all with the same emotion, and each emotion block was repeated five times, resulting in 30 blocks in total. Each trial consisted of the presentation of a single emotional face for 500 ms, followed by an interstimulus delay of 750 ms, with a block length of 10 s. No interblock interval was included in the current task. Both the faces within block and the emotion blocks were presented in a pseudorandomised order. No responses were required from the participants; however, instructions were given at the start of the task to pay close attention to the faces and that follow-up questions may be asked. After task completion, each participant was probed on the number of different emotions they observed during the task.
Image acquisition and preprocessing
MR images were acquired using a 3 T GE signa HDx scanner (GE Healthcare, Milwaukee, WI, USA) with an eight-channel head coil at the Westmead Hospital Medical Imaging Service, Sydney. 3D T1-weighted volumes were acquired using a spoiled gradient echo (SPGR) sequence (TR = 8.3 ms; TE = 3.2 ms; flip angle = 11 degrees; inversion time = 500 ms; FOV = 256 mm; 180 sagittal slices; matrix size = 256 × 256; voxel size = 1 × 1 × 1 mm; NEX = 1; ASSET = 1.5; scanning time = 7.12 min). The functional run consisted of 120 T2*-weighted echo-planar imaging (EPI) images (TR = 2500 ms; TE = 27.5 ms; FOV = 240 mm; flip angle = 90 degrees; 40 axial slices using an interleaved sequence; matrix size = 64 × 64; voxel size = 3.75 × 3.75 × 3.5 mm; scanning time = 5.13 min). Three dummy scans were acquired at the start of the EPI acquisition.
The fMRI images were preprocessed using SPM12 software (Wellcome Trust Centre for Neuroimaging, London, UK) running in MATLAB R2018b (MathWorks, Natick, MA, USA). This included coregistration of the structural and functional EPI images to the SPM template, segmentation and spatial normalisation of the structural image to the standard Montreal Neurological Institute (MNI) space, slice-time correction of the EPI images (using middle slice as the reference; interleaved acquisition), realignment and normalisation of the images, which were resliced to a voxel size of 1.5 × 1.5 × 2 mm, before being smoothed using an isotropic Gaussian filter of 6 × 6 × 6 mm at full-width half-maximum.
fMRI analyses
At the single-subject level, blood-oxygen level-dependant responses were modelled by using a boxcar stimulus function that was convolved with a canonical haemodynamic response function. This resulted in time courses that were then applied to the general linear model, which included the six experimental regressors (Happy, Angry, Sad, Fear, Disgust, and Neutral) as well as six movement regressors (of no interest; derived from the realignment step). A high-pass filter with a frequency cut-off of 128 s was applied to remove low frequency noise, and serial autocorrelations were estimated using an AR(1) model. Statistical parametric maps for each participant were then created by calculating simple linear contrasts that were applied to the parameter estimates. These included contrasting each affective emotion (Angry, Happy, Sad, Fear, Disgust) against Neutral, in order to identify the regions specific to each emotion (while controlling for lower-level visual and face-related processes).
At the group level, the single-subject contrast images for each of the five contrasts were taken to voxel-wise random-effects analyses regardless of zygosity. Age and sex were included as covariates-of-no-interest. One sample t tests were used to examine whole brain activity for each of the emotions to determine brain regions specific for each contrast (e.g. Happy>Neutral, Angry>Neutral, Sad>Neutral, Fear>Neutral and Disgust>Neutral). Significant activation threshold was set at p < 0.05 at the cluster level using family-wise error (FWE) correction with a voxel-wise threshold of p = 0.001. Masks for each significant cluster were created for each contrast separately. This was then used to extract t-contrast values from each voxel within the mask from the single-subject contrast images via the Volumes Toolbox (https://sourceforge.net/projects/volumes.spmtools.p/), and then averaged across the significant cluster for each participant.
Statistical analyses
Linear mixed-model analysis
In order to explore the potential associations between significant functional brain activity (indexed by t-contrast values extracted from the single-subject T maps, using the β estimate differences of the conditions), wellbeing and negative mood symptoms, linear mixed models were fit using the lme4 package in R. Due to the exploratory nature of this analysis, no multiple corrections were made across the different contrasts for the number of statistical tests conducted. Age, sex, zygosity, COMPAS-W scores (measuring wellbeing) and DASS-42 scores (measuring negative mood symptoms) were included in the models as fixed-effects predictors. The DASS-42 scores were included to examine the unique associations between emotion processing and wellbeing, independent of negative mood symptoms, in light of the fact that both constructs contribute to mental health – albeit in opposing directions. In order to control for family relatedness between each twin pair, a family number was used as the random-effects predictor (e.g. the same family number was given to each twin pair, while different twin pairs all had different family numbers). Any significant associations between the brain region (indexed by the averaged t-contrast value across the cluster for each participant) and wellbeing scores from the linear mixed models were taken to univariate and multivariate twin modelling analyses. If there was no significant association between the cluster and wellbeing, no further analyses were conducted.
Univariate twin analysis
To assess the heritability of fMRI activity, wellbeing and negative mood symptoms, univariate twin analysis was conducted using the OpenMx package (version 2.13.2) in R. Using structural equation modelling, additive genetic (A), non-additive genetic (D), shared environmental (C) and non-shared environment (E) variance components for emotion-related brain activity were examined (for more detail, see S2 in Supplemental Materials). Age and sex were included as covariates. The full univariate ACE/ADE model was compared to its nested submodels (AE, CE, and E) by dropping one or two of the variance components. The significance of each component was determined by using a χ2 goodness-of-fit test, which assessed whether dropping that component resulted in a significant decrease in the model fit. To confirm the parsimony of the model, Akaike Information Criterion (AIC) values were utilised for model comparisons.
Multivariate twin analysis
Lastly, multivariate genetic modelling was used to assess the shared and non-shared genetic and environmental variance between fMRI activity, wellbeing and negative mood symptoms. More specifically, a correlated factors model was run using OpenMx in R to examine the genetic and environmental contribution to the correlations (or the covariance) between the three variables. Age and sex were included as covariates. Genetic and environmental correlations between the variables were tested by systematically constraining each correlation to zero, then comparing the new model with the previous model. Again, this was conducted by using the χ2 difference test, where a significant reduction of model fit indicated the significance of the tested correlation. This correlation was then included in the final model, which was chosen based on the AIC values.
Results
Demographics
Demographic characteristics for the final sample are shown in Table 1. From the initial 263 participants, 13 were dropped prior to any analyses due to missing twin data, resulting in 250 participants for fMRI analyses. Twenty participants were further dropped during statistical analyses due to missing functional activity or questionnaire data, and undetermined zygosity. In total, 152 MZ twin and 78 DZ twin participants were included in all analyses (fMRI, linear mixed modelling, univariate and multivariate twin modelling). There were no significant differences in the levels of wellbeing or negative mood symptoms between the MZ and DZ twin pairs.
Emotion-related fMRI results
The voxel-wise random-effects analyses showed significant activity for emotion in four out of the five emotion contrasts. This included inferior frontal gyri activations in Happy, Angry, Fear and Sad contrasts (all against Neutral), as well as additional cerebellar activity in the Angry condition, and insula activity in the Sad condition (see Table 2). No significant activations were observed for the Disgust>Neutral contrast.
Table 2. Significant brain regions associated with emotion-related activation
L, left; R, right; IFG, inferior frontal gyrus; k, cluster size; n.s., not significant. Regions that survive FWE correction at the cluster level (p < 0.05) are listed in order of cluster size.
Linear mixed modelling
Averaged t-contrast values for each cluster were used to examine associations between the functional activity, wellbeing (COMPAS-W) and negative mood symptoms (DASS-42). From the 250 participants that were included in the fMRI analysis, five MZ pairs and one DZ pair where one or both twins did not show any activity within this cluster, and two pairs of twins with undetermined zygosity were excluded from further analysis. Two additional MZ pairs were also dropped as one or both twin participants from each pair did not complete the COMPAS-W and/or the DASS-42 measures needed here. The results indicated a significant relationship between the right inferior frontal gyrus (IFG) cluster from the Happy>Neutral condition and wellbeing (p = 0.026; see Fig. 1 and Table 3). No other main or interaction effects were significant across the four contrasts. Therefore, only this cluster was used in the twin modelling.
Fig. 1. Voxel-wise activation of the Happy>Neutral contrast in the right inferior frontal gyrus (denoted by black circles; A), which showed a significant association with levels of wellbeing (indexed by the total COMPAS-W scores; B).
Table 3. Linear mixed-model results of the significant association between the right inferior frontal gyrus cluster (from the Happy>Neutral condition) and wellbeing (COMPAS-W)
L, left; R, right; k, cluster size; DLPFC, dorsolateral prefrontal cortex; s.e., standard error; COMPAS-W, composite measure of wellbeing; DASS-42, negative mood symptoms. p values were derived using Satterthwaite approximations for degrees of freedom, which produce acceptable Type 1 error rates (Luke, Reference Luke2017).
Univariate twin modelling
The averaged t-contrast values from the right inferior frontal cluster for the Happy>Neutral condition were entered into a univariate twin modelling analysis to assess the heritability of functional activity within this region in response to positive emotion processing. Intra-class correlations between the MZ and DZ twins indicated that an ADE model would be the most appropriate. Nested models (AE and E) were then fitted to test whether dropping parameters (for the sake of model parsimony) would still be able to explain the variance in happy emotion processing. There was no significant change of fit when an AE model was compared against the ADE model (p = 0.514) suggesting the D component could be dropped from the model. When A was dropped from the model, the fit decreased marginally (p = 0.053), suggesting that genetic factors play a role in the heritability of functional activity in the right IFG (for positive emotions). The AIC value also decreased slightly, and despite the fact that an AIC difference of less than two may be considered trivial (Millar, Reference Millar2011), the AE model was still chosen as the best fitting and parsimonious based on the marginal model fit difference. The heritability of functional activity within the right inferior cluster was found to be 20% (see Table 4).
Table 4. Univariate heritability estimates for the right inferior frontal gyrus cluster from the Happy>Neutral condition
ICC, intra-class correlation; MZ, monozygotic twins; DZ, dizygotic twins; −2LL, minus twice the log likelihood; df, degrees of freedom; AIC, Akaike's information criterion; CI, 95% confidence intervals. The variance was decomposed into additive (A), non-additive (D) and unique environmental (E) factors. The AE model (in bold) was selected as the best-fitting model based on parsimony and lowest AIC value.
Multivariate twin modelling
At the multivariate level, an AE model was fitted with the right inferior frontal activity, wellbeing and negative mood scores to examine possible genetic and environmental correlations between the three variables. Here, each phenotype (or variable) is decomposed into its genetic and non-shared environmental components, and the correlations across these variables are estimated. A significant genetic correlation (r A) would indicate the extent to which the two variables share a common genetic influence, while a significant environmental correlation (r E) suggests that the relationship between the phenotypes is driven by unique environmental factors. The results showed a significant genetic correlation between wellbeing and negative mood symptoms (r A = −0.444, p = 0.037), which accounted for 41% of the phenotypic correlation between the two variables [r ph(A) = −0.405]Footnote †Footnote 1. There were no significant genetic correlations between the right inferior frontal activity and negative mood symptoms (p = 0.626) or wellbeing (p = 0.591). There was a significant environmental correlation between right inferior frontal activity and wellbeing (r E = 0.208, p = 0.047) that was almost 100% accounted for by unique environmental factors. Finally, the environmental correlation between negative mood symptoms and wellbeing was also significant (r E = −0.383, p < 0.001), which explained 59% of the phenotypic correlation (Fig. 2).
Fig. 2. AE correlated factors model for the three observed variables. Right IFG activity from the Happy>Neutral contrast was indexed by the extracted t-contrast values from the region. Negative mood symptoms and wellbeing were measured using the DASS-42 and COMPAS-W scales, respectively. The circles indicate latent factors, where A indicates the additive genetic effect, and E indicates the unique environmental effect of each component. Values on single-headed arrows indicate standardised path coefficients. Double-headed arrows indicate either genetic (r A) or environmental (r E) correlations between the latent factors with 95% confidence intervals (in square brackets). Solid and dotted lines indicate significant and non-significant correlations, respectively. R, right; IFG, inferior frontal gyrus.
Discussion
Our findings highlight that neural responses elicited during the appraisal of positive emotion stimuli are associated with greater wellbeing. Neural responsivity to facial emotion stimuli expressing happiness was centred on activation of the dorsolateral prefrontal cortex, and more specifically within the right IFG. We also observed that functional activity within this region showed 20% heritability, with the remaining variance due to unique environmental factors. Lastly, multivariate modelling revealed no common genetic variance contributing to the association between IFG activity and wellbeing, or with negative mood symptoms. In contrast, we observed that unique environment accounted for the shared variance between IFG activity and wellbeing.
These results align with previous functional neuroimaging studies that have shown a relationship between neural bases of emotion processes and specific sub-components of wellbeing (i.e. subjective wellbeing: Cunningham & Kirkland, Reference Cunningham and Kirkland2014; Ren et al., Reference Ren, Shi, Wei and Qiu2019; and psychological wellbeing: Heller et al., Reference Heller, van Reekum, Schaefer, Lapate, Radler, Ryff and Davidson2013; van Reekum et al., Reference van Reekum, Urry, Johnstone, Thurow, Frye, Jackson and Davidson2007). To this end, our finding of an association between right IFG activity and COMPAS-W adds to the evidence that there is an underlying mechanism that subserves the link between positive emotion appraisal and wellbeing. The IFG has been implicated in a range of emotion-related processes that include emotional face expressions (Jabbi & Keysers, Reference Jabbi and Keysers2008), social emotion reappraisals (Grecucci, Giorgetta, Bonini, & Sanfey, Reference Grecucci, Giorgetta, Bonini and Sanfey2013) and emotional empathy (Shamay-Tsoory, Aharon-Peretz, & Perry, Reference Shamay-Tsoory, Aharon-Peretz and Perry2009), as well as in resting-state and functional measures of wellbeing (Heller et al., Reference Heller, van Reekum, Schaefer, Lapate, Radler, Ryff and Davidson2013; Kong et al., Reference Kong, Wang, Song and Liu2016). Here, we further bridge the gap between the two fields of research and show that positive emotion-related activity within this region is associated with a measure of wellbeing that encapsulates both subjective and psychological wellbeing (but see Jo, Ou, & Kung, Reference Jo, Ou and Kung2019).
No significant associations were found between neural activity related to negative emotions and levels of wellbeing, despite previous behavioural (Sanchez & Vazquez, Reference Sanchez and Vazquez2014; Yu & Li, Reference Yu and Li2012) and neuroimaging (Chen et al., Reference Chen, Wu, Yao, Lei, Luo and Li2018; Ren et al., Reference Ren, Shi, Wei and Qiu2019) findings. This may be related to several factors, including recognition superiority and advantage conferred to happy emotion, where it has been shown that happy faces are identified more accurately and faster compared to the other basic emotions across different stimulus sets, modalities and exposure times (Calvo & Beltrán, Reference Calvo and Beltrán2013; Kirita & Endo, Reference Kirita and Endo1995). This is also in line with our previous finding of happy faces eliciting the fastest behavioural responses, which was associated with high levels of wellbeing (Routledge et al., Reference Routledge, Williams, Harris, Schofield, Clark and Gatt2018). However, the processing of negative emotions and their relationship with wellbeing may be mediated by other factors (e.g. emotion regulation and reappraisal), which are needed in order to process negative information in a rational and accurate manner (Schutte et al., Reference Schutte, Thorsteinsson, Hine, Foster, Cauchi and Binns2010). Therefore, there may have been a qualitatively different type of processing inferred for the happy condition, which was also the only positive emotion (Becker & Srinivasan, Reference Becker and Srinivasan2014). Inclusion of a positive condition may also have affected the context of the experiment, compared to previous studies that did not include any positive stimuli in their paradigms (Chen et al., Reference Chen, Wu, Yao, Lei, Luo and Li2018; Ren et al., Reference Ren, Shi, Wei and Qiu2019). Therefore, the combination of a strong representation of happy faces in the brain and the well-established link between positive emotion and high levels of wellbeing is likely to have contributed to our result of an association between happy faces and wellbeing, but not with the negative emotions. Although causal inferences cannot be made here and the results need replication, we speculate that individuals who react strongly to happy faces may have a positivity bias that leads them to be more attentive to positive stimuli, which in turn contributes to their overall wellbeing and life satisfaction, or vice versa (Bastian, Kuppens, De Roover, & Diener, Reference Bastian, Kuppens, De Roover and Diener2014; Lyubomirsky, Reference Lyubomirsky2001; Lyubomirsky, Dickerhoof, Boehm, & Sheldon, Reference Lyubomirsky, Dickerhoof, Boehm and Sheldon2011).
We also examined whether the neural correlates elicited during facial emotion processing have a genetic basis. By expanding this to an adult sample and testing both positive and negative emotions using fMRI, we found a small but marginally significant heritability estimate of 20% for the activity in the right IFG elicited for the happy faces, with the remainder being due to unique environmental influences. To our knowledge, this is the first time that the magnitude of neural activity within a specific brain region has been found to be heritable for emotion appraisal, and adds to the evidence that genetic factors have a role in influencing positive emotion processing, as observed in genetic (e.g. Wingo et al., Reference Wingo, Almli, Stevens, Jovanovic, Wingo, Tharp and Ressler2017) and behavioural studies (e.g. Routledge et al., Reference Routledge, Williams, Harris, Schofield, Clark and Gatt2018).
Finally, we conducted a multivariate twin analysis to examine whether there are any shared genetic or environmental bases between IFG activation and wellbeing (and negative mood symptoms), which contribute to the overall phenotypic expression. We did not find any evidence of a shared genetic covariance between any of the three variables. This indicates that despite the heritability of positive emotion-related neural activity (and its associations with wellbeing), there is no evidence of a shared genetic component for this phenotype and wellbeing (or negative mood) levels in our sample. This is possible even if each phenotype is highly heritable (independently of one another)Footnote 2 due to a lack of overlap of genetic effects on both, leading to a non-significant genetic correlation between the phenotypes (Røysamb & Tambs, Reference Røysamb and Tambs2016). In contrast, a novel finding we identified in the multivariate model was a significant unique environmental correlation between wellbeing and the IFG activity to happy faces (r E = 0.21). Interestingly, the environmental covariance between these two variables accounted for almost 100% of the phenotypic correlation observed, which suggests that this association mainly depends on experiences that are unique to each twin (regardless of the same family environment) and some measurement error (Plomin, Reference Plomin2011). Such unique environmental sources can include life experiences such as major illnesses or peer groups that are not shared between siblings, and tend to become much more frequent in adulthood (to the extent that most adult twins will lead separate, non-shared lives). Therefore, the presence of external events specific to each twin may have played an essential role in the expression of IFG activity and sense of wellbeing, especially considering that our participants were all adults with a mean age of 39.4 years. Again, although our study design does not allow for causal inferences, a possible explanation for this may be that individuals who are frequently exposed to positive environments (especially in adulthood, where such event occurrences may differ between twins) are biased towards the processing of positive emotions, which then has cascading effects on higher wellbeing (in line with the broaden-and-build theory of positive emotions; Fredrickson, Reference Fredrickson2004). Alternatively, it may be possible that given a particular (i.e. positive) environment, a person with higher levels of wellbeing show a disposition towards greater neural activation in response to positive emotional stimuli.
There are several limitations to the current study that need to be acknowledged. Specifically, we conducted multiple linear mixed-model analyses in an exploratory manner as this was one of the first studies examining the link between fMRI brain activity to emotion and wellbeing. Therefore, our results will need replication in an independent sample. Furthermore, our heritability estimate for the right IFG activity was only marginally significant, and therefore also requires replication. This may be related to our sample size, which despite being large for MRI studies, is relatively modest for biometric twin analyses. Additionally, it has been noted that contrasting between two conditions (e.g. happy faces>neutral faces), rather than using a baseline may weaken the effects of heritability as the composition of a difference score can remove some of the genetic variance (Li et al., Reference Li, Wang, Yan, Cheung, Docherty, Sham and Chan2019). Therefore, it is unclear whether this smaller heritability estimate is due to the effect itself or simply because we examined a contrast between conditions. Lastly, our emotion task only required passive viewing of faces, rather than active responses, which may elicit stronger emotion-related neural activity. On a more general note, there are limitations regarding assumptions of the twin model that need to be mentioned, including the main assumption that MZ and DZ twins share equal environmental influences (Derks, Dolan, & Boomsma, Reference Derks, Dolan and Boomsma2006; but see Felson, Reference Felson2014). In future studies, it would be worth considering gene-by-environment interactions or gene-environment correlations that may contribute to the model effects reported here (Rathouz, Van Hulle, Lee Rodgers, Waldman, & Lahey, Reference Rathouz, Van Hulle, Lee Rodgers, Waldman and Lahey2008).
In conclusion, we sought to examine the associations between neural bases of emotion processing and wellbeing, as well as the potential heritability of the emotion-wellbeing link. We found that positive emotion was associated with wellbeing, where individuals with highest levels of wellbeing displayed the largest neural activity for happy faces in the right IFG. This activation was further found to be heritable, but there was no evidence for shared genetic correlations between the neural activity and wellbeing, or with negative mood symptoms. In contrast, a significant unique environmental correlation was observed between neural activity and wellbeing. Altogether, these results support the idea of a mutual relationship between emotion and wellbeing, and suggest that positive emotion processing may play an important role in optimal psychological functioning. Moreover, it seems that unique life events are able to influence both the neural responses to positive emotion as well as levels of wellbeing, further supporting the notion of a positive emotion-wellbeing link.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721002695.
Author contributions
HRPP conceptualised the aims of the paper, ran formal data analyses, and drafted and revised the paper. MRC aided in the statistical modelling of the twin data. JMG, LMW and PRS designed the study and oversaw the running of the larger project. JMG also conceptualised the aims of the paper, supervised analyses, and helped edit the original draft and revised versions. All authors reviewed the results, and revised and edited the final version of the paper.
Financial support
This project was supported by an Australian Research Council Linkage Grant (LP0883621). JMG, HRPP and AM were supported by a National Health and Medical Research Council (NHMRC) Project Grant (1122816). MRC was supported by the Australian Government Research Training Program Scholarship, AM was supported by the UNSW Tuition Fee Scholarship, and JJ was supported by the UNSW Scientia PhD Scholarship Scheme. MRC and AM also received the Neuroscience Research Australia top-up scholarship. PRS was supported by a NHMRC Program (1037196) and Investigator (1176716) Grants. This research was facilitated through access to Twins Research Australia, a national resource supported by NHMRC Centre of Research Excellence Grant (1079102). The authors acknowledge training provided by Twins Research Australia and QIMR Twin Modelling Workshop to aid in some of the statistical analyses used in this paper.
Conflict of interest
LMW has received advisory board fees from One Mind Psyberguide and the Laureate Institute for Brain Research unrelated to this study. JMG is a stockholder in MAP Biotech Pty Ltd. All other authors declare that they have no conflicts of interest.
