Introduction
Recent years have shown significant advances through large-scale collaboration in genome-wide association studies (GWAS), which have generated replicated findings on a number of common risk alleles and copy number variants, suggesting that the risk of psychosis is polygenic (Lee et al., Reference Lee, Anttila, Won, Feng, Rosenthal, Zhu and Posthuma2019; McGrath, Mortensen, Visscher, & Wray, Reference McGrath, Mortensen, Visscher and Wray2013; Schizophrenia Working Group of the Psychiatric Genomics et al., Reference Ripke, Neale, Corvin, Walters, Farh and O'Donovan2014). Individuals that carry a higher number of risk variants have a higher risk for developing psychotic disorder. Further, there is consistent evidence from numerous twin and family studies that the risk for developing a psychotic disorder is increased in first-degree relatives of patients with the disorder (Guloksuz et al., Reference Guloksuz, Pries, Delespaul, Kenis, Luykx, Lin and van Os2019; van Os, Reininghaus, & Meyer-Lindenberg, Reference van Os, Reininghaus and Meyer-Lindenberg2017a), which suggests a familial liability to psychosis (Islam et al., Reference Islam, Khan, Quee, Snieder, van den Heuvel and Bruggeman2017). Familial liability may derive from a shared environment, i.e. the shared exposure to social adversity (e.g. bereavement). In addition, the environment of children is strongly influenced by their parents or their parents’ genes (Rutter, Moffitt, & Caspi, Reference Rutter, Moffitt and Caspi2006; van Os, Rutten, & Poulton, Reference van Os, Rutten and Poulton2008). These findings broadly support a liability-threshold model.
Evidence further suggests that socio-environmental factors play an important role in the development of psychosis (Heinz, Deserno, & Reininghaus, Reference Heinz, Deserno and Reininghaus2013; Klippel et al., Reference Klippel, Myin-Germeys, Chavez-Baldini, Preacher, Kempton, Valmaggia and Reininghaus2017; Morgan, Charalambides, Hutchinson, & Murray, Reference Morgan, Charalambides, Hutchinson and Murray2010; Rauschenberg et al., Reference Rauschenberg, van Os, Cremers, Goedhart, Schieveld and Reininghaus2017; Reininghaus et al., Reference Reininghaus, Gayer-Anderson, Valmaggia, Kempton, Calem, Onyejiaka and Morgan2016b, Reference Reininghaus, Kempton, Valmaggia, Craig, Garety, Onyejiaka and Morgan2016c; Uher, Reference Uher2014; van Os & Rutten, Reference van Os and Rutten2014; Waszczuk et al., Reference Waszczuk, Eaton, Krueger, Shackman, Waldman, Zald and Kotov2020). One targetable mechanism through which polygenic variation has been posited to impact on individuals and interact with the social environment to increase the risk for psychosis is stress sensitization (Collip, Myin-Germeys, & Van Os, Reference Collip, Myin-Germeys and Van Os2008; Rauschenberg et al., Reference Rauschenberg, van Os, Cremers, Goedhart, Schieveld and Reininghaus2017; Reininghaus et al., Reference Reininghaus, Gayer-Anderson, Valmaggia, Kempton, Calem, Onyejiaka and Morgan2016b, Reference Reininghaus, Kempton, Valmaggia, Craig, Garety, Onyejiaka and Morgan2016c). The proposition here is that the stress response is amplified in individuals with increased polygenic risk, such that they experience a greater response to even minor stressors in daily life. This process of sensitization operates at various levels of causation. At the behavioural level, the most commonly used marker of this underlying process is stress reactivity characterized by (i) stronger emotional reactions, and (ii) more intense psychotic experiences in response to minor stressors in daily life (Klippel et al., Reference Klippel, Myin-Germeys, Chavez-Baldini, Preacher, Kempton, Valmaggia and Reininghaus2017; Myin-Germeys & van Os, Reference Myin-Germeys and van Os2007; Myin-Germeys, van Os, Schwartz, Stone, & Delespaul, Reference Myin-Germeys, van Os, Schwartz, Stone and Delespaul2001; Rauschenberg et al., Reference Rauschenberg, van Os, Cremers, Goedhart, Schieveld and Reininghaus2017; Reininghaus et al., Reference Reininghaus, Kempton, Valmaggia, Craig, Garety, Onyejiaka and Morgan2016c). Minor stressors, i.e. unpleasant events, activities and social situations, as well as emotional reactions and psychotic experiences may arguably be best measured using the experience sampling method (ESM; Csikszentmihalyi & Larson, Reference Csikszentmihalyi and Larson1987; Klippel et al., Reference Klippel, Myin-Germeys, Chavez-Baldini, Preacher, Kempton, Valmaggia and Reininghaus2017; Myin-Germeys et al., Reference Myin-Germeys, Kasanova, Vaessen, Vachon, Kirtley, Viechtbauer and Reininghaus2018; Rauschenberg et al., Reference Rauschenberg, van Os, Cremers, Goedhart, Schieveld and Reininghaus2017; Reininghaus, Reference Reininghaus2018; Reininghaus, Depp, and Myin-Germeys, Reference Reininghaus, Depp and Myin-Germeys2016a; Reininghaus et al., Reference Reininghaus, Kempton, Valmaggia, Craig, Garety, Onyejiaka and Morgan2016c), a structured digital intensive longitudinal data collection technique that allows to assess moment-to-moment variation in daily life. Using this method, Myin-Germeys et al. (Reference Myin-Germeys, van Os, Schwartz, Stone and Delespaul2001) were the first to show a gradient in stress reactivity that paralleled the level of familial liability, in that stress reactivity was elevated in cases with psychotic disorder and first-degree relatives compared with controls. Recent advances in GWAS allow to investigate this further using a molecular genetic measure of polygenic risk, i.e. polygenic risk scores (PRS), that may modify individuals' response to minor stressors. Recently, a first experience sampling study did not find evidence on stress reactivity being modified by PRS in a sample of young healthy adults (Pries et al., Reference Pries, Klingenberg, Menne-Lothmann, Decoster, van Winkel, Collip and Guloksuz2020). However, previous studies have not investigated this in cases with psychotic disorder and their first-degree relatives to elucidate whether stress reactivity operates in individuals with increased familial liability, and is modified by polygenic risk, in pathways to psychosis. Furthering our understanding of this targetable mechanism is relevant as a basis for developing effective interventions.
In the current study, we aimed to investigate whether the associations of momentary stress with (i) negative affect, (ii) positive affect and (iii) psychotic experiences are modified by PRS and liability to psychosis in cases, siblings and controls. Specifically, we aimed to test the following hypotheses: (1) within each group, the magnitude of associations of momentary stress with (i) negative affect, (ii) positive affect and (iii) psychotic experiences is greater in individuals with high PRS v. individuals with low PRS (H1); and (2) the difference in magnitude of associations of momentary stress with (i) negative affect, (ii) positive affect and (iii) psychotic experiences (i.e. the difference in responses to stress) between those with high v. low levels of PRS is greater in (a) cases than in controls, (b) siblings than in controls and (c) cases than in siblings (H2).
Methods
Participants
Data were collected as part of a longitudinal study, the Genetic Risk and Outcome of Psychosis Project (GROUP) in the Netherlands and Belgium (Korver, Quee, Boos, Simons, & de Haan, Reference Korver, Quee, Boos, Simons and de Haan2012). Inclusion criteria for cases were the age between 16 and 50 years, meeting full DSM-IV criteria for a non-affective psychotic disorder and estimated level of intelligence above 70. Siblings of cases were recruited via the participating cases and included when they were aged 16–50 years. Controls were contacted through mailing lists. Inclusion criteria for controls were the age between 16 and 50 years, no lifetime psychotic disorder and no first-degree family member with a lifetime psychotic disorder. The Positive and Negative Syndrome Scale (PANSS; Kay, Fiszbein, & Opler, Reference Kay, Fiszbein and Opler1987) and a short version of the Wechsler Adult Intelligence Scale (WAIS-III; Wechsler, Reference Wechsler1977) were administered to all participants in order to assess clinical symptoms and intellectual abilities. Further, all participants completed the social functioning scale (Birchwood, Smith, Cochrane, Wetton, & Copestake, Reference Birchwood, Smith, Cochrane, Wetton and Copestake1990). Detailed information on sample characteristics and recruitment methods has been previously described (Korver et al., Reference Korver, Quee, Boos, Simons and de Haan2012).
Genotyping, imputation and polygenic risk score (PRS)
Genotype data for 2812 individuals was generated on a customized Illumina array with 570 038 SNPs. Quality control (QC) procedures were performed using PLINK v1.9 (Purcell et al., Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira, Bender and Sham2007) (see online Supplementary Material). In total, 2505 individuals passed the QC steps. Then, SNPs were imputed on the Michigan server (Das et al., Reference Das, Forer, Schönherr, Sidore, Locke, Kwong and Fuchsberger2016). After post-imputation QC, PRS were calculated for 2505 samples using schizophrenia-associated alleles and effect sizes reported in the GWAS summary statistics from the Psychiatric Genetics Consortium Schizophrenia Workgroup freeze 2 (PGC-2; Schizophrenia Working Group of the Psychiatric Genomics et al., Reference Ripke, Neale, Corvin, Walters, Farh and O'Donovan2014). To prevent potential overlap in study population to impact our results, all Dutch and Belgian individuals had been excluded from the PGC-2 GWAS to allow unbiased PRS computation. PRS were calculated using PLINK's score function for 12 GWAS p value thresholds. The PRS p t = 0.05 explained most of the variance of schizophrenia case-control status (see online Supplementary Material). Hence, it was selected to perform the following regression analyses.
ESM measures
Using ESM, momentary stress, affect and psychotic experiences were assessed over the course of six consecutive days. There is good evidence on the feasibility of ESM, especially with respect to patient samples (Myin-Germeys et al., Reference Myin-Germeys, Kasanova, Vaessen, Vachon, Kirtley, Viechtbauer and Reininghaus2018, Oorschot et al., Reference Oorschot, Kwapil, Delespaul and Myin-Germeys2009). Participants received a dedicated digital device (i.e. the PsyMate, www.psymate.eu/) and were asked to complete 10 ESM assessments per day. ESM data were only used if participants completed more than 20 assessments in total (Myin-Germeys et al., Reference Myin-Germeys, van Os, Schwartz, Stone and Delespaul2001; Reininghaus et al., Reference Reininghaus, Kempton, Valmaggia, Craig, Garety, Onyejiaka and Morgan2016c). A detailed description of the ESM measures is shown in Table 1.
Table 1. ESM measures
Note: The experience sampling methodology (ESM) was used and prompted participants 10 times a day on 6 consecutive days with a semi-random sampling scheme within a fixed, predefined time frame.
Statistical analysis
For the current analysis, the GROUP wave 3 data set with release number 7.0 and ESM data set with release number 2.0 were used. All analyses were carried out using STATA version 15.1 (StataCorp., 2017). Pairwise group comparisons were performed regarding basic group characteristics and the mean of the independent and dependent variables using one-way analysis of variance. We then fitted linear mixed models to account for the multilevel structure of ESM data using the mixed command in STATA. For each momentary stressor (composite stress, event-related, activity-related, social stress), a separate model was fitted with (i) negative affect, (ii) positive affect and (iii) psychotic experiences as outcome variables, while controlling for potential confounders (i.e. age, gender, IQ and the first three principal components to control for genetic ancestry). We added two-way (momentary stress × PRS) and three-way (momentary stress × PRS × group) interactions to test whether associations between momentary stress and (i) negative affect, (ii) positive affect and (iii) psychotic experiences were modified by PRS and group (cases, siblings, controls). In addition, multilevel mixed tobit regression models (Tobin, Reference Tobin1958) were fitted to account for skewness in data on psychotic experiences (see online Supplementary Table S3).
We used Wald tests to assess the statistical significance of each interaction term. To investigate whether associations of each momentary stressor with (i) negative affect, (ii) positive affect or (iii) psychotic experiences were greater in individuals with high v. low PRS, continuous independent variables were standardized (mean = 0, s.d. = 1) for interpreting significant interaction terms and examining the difference in associations between high (mean + 1 s.d.), and low (mean − 1 s.d.) PRS within and across groups (cases, siblings, controls) (Aiken & West, Reference Aiken and West1991; Cohen, Cohen, West, & Aiken, Reference Cohen, Cohen, West, Aiken, Cohen and Cohen2003). Specifically, we calculated linear combinations of coefficients using the lincom command in STATA testing the hypotheses that: (1) within each group, the magnitude of associations of each momentary stressor with (i) negative affect, (ii) positive affect and (iii) psychotic experiences was greater in individuals with high v. low PRS (mean ± 1 s.d. of continuous PRS) (Aiken & West, Reference Aiken and West1991; Cohen et al., Reference Cohen, Cohen, West, Aiken, Cohen and Cohen2003) (H1); and (2) the difference in magnitude of associations of each momentary stressor with (i) negative affect, (ii) positive affect and (iii) psychotic experiences in those individuals with high v. low PRS (mean ± 1 s.d. of continuous PRS) was greater in (a) cases than in controls, (b) relatives than in controls, and (c) cases than in relatives (H2). Likelihood ratio tests were used to evaluate improvement in model fit. We adjusted the significance level of likelihood ratio tests for the three-way interactions in order to correct for Type-1 error proliferation using family-wise error correction (p FWE values). The unadjusted p value was multiplied by the total number of tests, i.e. by 12 (four stress measures, three groups). Two-tailed p FWE < 0.05 was considered nominally statistically significant. We standardized continuous ESM and PRS variables (mean = 0, s.d. = 1) for interpreting significant three-way interactions.
Results
Sample characteristics
The full GROUP sample consisted of 3684 participants and ESM was completed at wave 3. The analytic sample with available ESM and PRS data comprised 248 participants, i.e. 96 cases with non-affective psychosis, 79 siblings of cases and 73 controls. As the prevalence of risk alleles varies across ethnic groups, the analytic sample was selected to comprise participants of White European decent only. There were notable differences in sociodemographic and clinical characteristics of the analytic sample compared to the full GROUP sample of cases, siblings and controls (see online Supplementary Table S2).
Within the analytic sample, cases, siblings and controls differed in socio-demographic and clinical characteristics as shown in Table 2. Specifically, cases were, on average, younger (β = −1.81, 95% CI −2.243 to −1.371, p < 0.001), had lower IQ estimates (β = −10.94, 95% CI −11.74 to −10.13, p < 0.001) and comprised more men (β = −0.28, 95% CI −0.30 to −0.26, p < 0.001) than siblings. Furthermore, cases showed reduced social functioning (β = −9.32, 95% CI −11.32 to −7.31, p < 0.001) and increased positive (β = 4.65, 95% CI 3.72–5.58, p < 0.001) as well as negative symptoms (β = 3.61, 95% CI 2.78–4.45, p < 0.001) compared to siblings. The same differences were evident in cases v. controls (all p < 0.001, see Table 2). However, PRS was higher in cases than in siblings (β = 3.46, 95% CI 3.83–3.10, p < 0.001) and controls (β = 3.07, 95% CI 2.71–3.45, p < 0.001). There was some evidence that PRS in siblings was, on average, lower compared to controls (β = −0.39, 95% CI −0.77 to −0.002, p = 0.05). In addition, siblings and controls did not differ in social functioning (β = −1.18, 95% CI −3.33 to 0.98, p = 0.28) or symptoms (PANSS positive symptoms: β = −0.09, 95% CI −1.08 to 0.90, p = 0.86; PANSS negative symptoms: β = −0.16, 95% CI −0.73 to 1.05, p = 0.73).
Table 2. Sample characteristics of cases, siblings and controls
PRS, polygenic risk score; PANSS, Positive and Negative Syndrome Scale; s.d., standard deviation; v., versus; CI, confidence interval.
As shown in Table 3, cases completed fewer ESM assessments (i.e. beeps) compared to siblings (β = −0.63, 95% CI −1.07 to −0.18, p = 0.006) and controls (β = −1.50, 95% CI −2.00 to −1.01, p < 0.001), whereas siblings completed fewer assessments than controls (β = −1.78, 95% CI 2.21 to −1.34, p < 0.001). There were no differences between cases and controls (β = 0.07, 95% CI −0.06 to 0.20, p = 0.31) in interference of ESM assessments with their daily life. However, interference of ESM assessment with daily life was lower in cases than in siblings (β = −0.50, 95% CI −0.62 to −0.39, p < 0.001) and higher in siblings than in controls (β = 0.57, 95% CI 0.46–0.69, p < 0.001). Cases reported, on average, higher momentary stress (i.e. composite stress, event-related stress, activity-related stress and social stress) compared to siblings (e.g. composite stress: β = 0.30, 95% CI 0.25–0.35, p < 0.001) and controls (e.g. composite stress: β = 0.51, 95% CI 0.46–0.57, p < 0.001). Further, cases reported higher negative affect than both siblings (β = 0.68, 95% CI 0.64–0.72, p < 0.001) and controls (β = 0.71, 95% CI 0.66–0.76, p < 0.001), as well as lower positive affect than siblings (β = −0.41, 95% CI −0.47 to −0.35, p < 0.001) and controls (β = −0.52, 95% CI −0.59 to −0.45, p < 0.001). Intensity of psychotic experiences were, on average, greater in cases compared to both siblings (β = 0.39, 95% CI 0.36–0.42, p < 0.001) and controls (β = 0.39, 95% CI 0.36–0.42, p < 0.001). Although siblings and controls differed in some momentary stress measures (e.g. composite stress: β = 0.21, 95% CI 0.17–0.26, p < 0.001) and positive affect (β = −0.11, 95% CI −0.17 to −0.05, p < 0.001), this was not the case for negative affect (β = 0.03, 95% CI −0.01 to 0.08, p = 0.16) or psychotic experiences (β = −0.002, 95% CI −0.03 to 0.02, p = 0.88).
Table 3. Aggregate ESM scores for momentary stress, negative affect, positive affect and psychotic experiences in cases, siblings and controls
s.d., standard deviation; v., versus; CI, confidence interval.
a Adjusted for age, gender and IQ.
Stress reactivity by PRS in cases, siblings and controls
As can be seen in Table 4, we found no evidence that the association of composite momentary stress, event-related-stress, activity-related stress and social stress, on the one hand, with (i) negative affect and (ii) positive affect, on the other, was modified by PRS in cases, siblings and controls. However, there was strong evidence for three-way interaction effects of composite momentary stress × PRS × group (χ 2 = 19.70, p FWE = 0.001), activity-related stress × PRS × group (χ 2 = 22.07, p FWE < 0.001) and social stress × PRS × group (χ 2 = 17.89, p FWE = 0.001) on (iii) psychotic experiences (see Table 5). This indicated that the associations of composite momentary stress, activity-related stress, and social stress with (iii) psychotic experiences differed between individuals with high and low levels of PRS within (H1) and across groups (H2), as detailed below.
Table 4. Associations of momentary stress with negative affect and positive affect in cases, siblings and controlsa
Note: adj. β, standardized regression coefficients [continuous independent variables were standardized (mean = 0, s.d. = 1) for interpreting significant three-way interaction terms and examining the difference in associations between high (mean + 1 s.d.), and low (mean − 1 s.d.) PRS within and across groups (cases, siblings, controls)]; p FWE, family-wise error-corrected p values were computed by multiplying the unadjusted p value by the total number of tests (i.e. 4 stress measures × 3 outcomes = 12); CI, confidence interval; df, degrees of freedom; LR, likelihood ratio; s.d., standard deviation.
a Adjusted for age, gender and IQ.
b Difference in associations between those with high v. low PRS.
Table 5. Associations of momentary stress with psychotic experiences in cases, siblings and controlsa
Note: adj. β, standardized regression coefficients [continuous independent variables were standardized (mean = 0, s.d. = 1) for interpreting significant three-way interaction terms and examining the difference in associations between high (mean + 1 s.d.), and low (mean − 1 s.d.) PRS within and across groups (cases, siblings, controls)]; p FWE, family-wise error-corrected p values were computed by multiplying the unadjusted p value by the total number of tests (i.e. 4 stress measures × 3 outcomes = 12); CI, confidence interval; df, degrees of freedom; LR, likelihood ratio; s.d., standard deviation.
a Adjusted for age, gender and IQ.
b Difference in associations between those with high v. low PRS.
Within-group comparisons (H1)
While there were no differences in the magnitude of associations of momentary stress with psychotic experiences within cases with high v. low level of PRS, this was the case for siblings and controls (see Table 5). In contrast to our hypothesis, there was a weaker association of composite momentary stress (adj.β high v. low = −0.24, 95% CI −0.34 to −0.08, p = 0.003), activity-related stress (adj.β high v. low = −0.18, 95% CI −0.30 to −0.06, p = 0.003) and social stress (adj.β high v. low = −0.25, 95% CI −0.44 to −0.07, p = 0.006) with psychotic experiences in siblings with high PRS compared to siblings with low PRS. By contrast, in controls with high PRS, activity-related stress was associated with more intense psychotic experiences (adj.β high v. low = 0.10, 95% CI 0.03–0.17, p = 0.005), than in controls with low PRS. The associations of social and event-related stress with psychotic experiences did not vary by PRS in controls.
Between-group comparisons (H2)
When we examined whether PRS impacts differently on stress reactivity across groups based on differences in the magnitude of associations of momentary stress with psychotic experiences between those with high v. low PRS across groups (see Table 5), we observed consistent differences across siblings and controls as well as cases and siblings, but less consistent across cases and controls. We found evidence that the difference in associations of activity-related stress with psychotic experiences between those with high v. low PRS across groups was greatest in siblings v. controls (adj.β delta high v. low = −0.15, 95% CI −0.22 to −0.09, p < 0.001) followed by cases v. siblings (adj.β delta high v. low = 0.08, 95% CI 0.02–0.15, p = 0.01) and cases v. controls (adj.β delta high v. low = 0.07, 95% CI −0.13 to −0.01, p = 0.02). We observed the greatest differences in associations of social stress and (iii) psychotic experiences between those with high v. low PRS across groups in siblings v. controls (adj.β delta high v. low = −0.17, 95% CI −0.26 to −0.09, p < 0.001), followed by cases v. siblings (adj.β delta high v. low = 0.13, 95% CI 0.05–0.26, p = 0.001). A similar pattern of findings emerged for differences in associations of composite momentary stress and psychotic experiences between individuals with high v. low PRS across groups. There was evidence that the difference in psychotic reactivity to composite momentary stress between those with high v. low PRS varied across siblings and controls (adj.β delta high v. low = −0.16, 95% CI −0.24 to −0.09, p = 0.001) as well as cases and siblings (adj.β delta high v. low = 0.12, 95% CI 0.05–0.19, p = 0.001).
Discussion
Principal findings
The current study is the first to investigate whether momentary stress reactivity is modified by PRS in cases with non-affective psychotic disorder, siblings of cases and controls. In contrast to our hypotheses (H1, H2), we found no evidence that associations of momentary stress with (i) negative affect and (ii) positive affect were modified by PRS and group. Further, the association between momentary stress and (iii) psychotic experiences was modified by PRS within siblings and controls, but not in cases. There was strong evidence that, in contrast to our first hypothesis, siblings with high PRS reported less intense psychotic experiences in response to momentary stress compared to siblings with low PRS. However, consistent with the first hypothesis, the opposite held true within controls, as individuals in this group with high PRS showed more intense psychotic experiences in response to stress compared to those with low PRS. We further found that differences in psychotic reactivity to momentary stress between high v. low PRS varied across groups, but these differences were not consistent with those posited in H2.
Methodological considerations
The current findings should be interpreted in the light of some limitations. First, although numerous twin and family studies have suggested a high heritability for psychosis (Guloksuz et al., Reference Guloksuz, Pries, Delespaul, Kenis, Luykx, Lin and van Os2019; Islam et al., Reference Islam, Khan, Quee, Snieder, van den Heuvel and Bruggeman2017; Sullivan, Kendler, & Neale, Reference Sullivan, Kendler and Neale2003), to date, the variance explained by PRS in molecular genetic studies remains limited and it is assumed that PRS only represent a part of the genetic contributions (Wray et al., Reference Wray, Lin, Austin, McGrath, Hickie, Murray and Visscher2021). In addition, evidence for an association with psychotic symptoms in the general population remains inconsistent. While some studies have reported an association between PRS and self-reported psychotic experiences in adolescents (Pain et al., Reference Pain, Dudbridge, Cardno, Freeman, Lu, Lundstrom and Ronald2018), no evidence for such an association has been found in the adult general population (Marsman et al., Reference Marsman, Pries, ten Have, de Graaf, van Dorsselaer, Bak and van Os2020; Mistry, Harrison, Smith, Escott-Price, & Zammit, Reference Mistry, Harrison, Smith, Escott-Price and Zammit2018). Further, evidence for an association of PRS with negative symptoms remains equivocal in non-clinical populations (Mistry et al., Reference Mistry, Harrison, Smith, Escott-Price and Zammit2018). However, in clinical populations, higher negative symptoms were associated with increased PRS (Bigdeli, Peterson, Docherty, Kendler, & Fanous, Reference Bigdeli, Peterson, Docherty, Kendler and Fanous2019; Mistry et al., Reference Mistry, Harrison, Smith, Escott-Price and Zammit2018; Ruderfer & Psychiatric Genomics Consortium Bipolar/Schizophrenia Working Group, Reference Ruderfer2019). Moreover, Allardyce et al. (Reference Allardyce, Leonenko, Hamshere, Pardinas, Forty, Knott and Escott-Price2018) showed a polygenic-risk gradient across schizophrenia and bipolar disorder that increased as levels of psychotic symptoms increased. This may suggest that PRS rather represents a genetic marker for negative symptoms in schizophrenia. In the present study, the ESM psychosis measure primarily assessed positive but not negative symptoms of schizophrenia, which may explain in part the inconsistent pattern of findings observed here. In addition, internal consistency of ESM measures for negative affect and psychotic experiences were moderate, which underlines the importance of further psychometric evaluation and validation of ESM measures.
Second, the PRS as calculated in the current study assumed additive effects of individual risk alleles, which reflects a rather simple genetic model. Complex higher-order interactive associations between risk alleles were not accounted for. This may in part explain why we did not find evidence of effect modification by PRS in cases.
Third, in line with previous research (Pries et al., Reference Pries, Klingenberg, Menne-Lothmann, Decoster, van Winkel, Collip and Guloksuz2020; Rauschenberg, van Os, Goedhart, Schieveld, & Reininghaus, Reference Rauschenberg, van Os, Goedhart, Schieveld and Reininghaus2021c), we used a composite stress measure and adjusted for multiple testing to minimize the potential impact of type I error rate. However, sample size and number of ESM observations were fairly small and, hence, may have provided limited statistical power for detecting three-way interaction effects. Hence, careful replication of our findings is required before firm conclusions can be drawn. Another methodological limitation is that other environmental factors such as childhood trauma (Lardinois, Lataster, Mengelers, Van Os, & Myin-Germeys, Reference Lardinois, Lataster, Mengelers, Van Os and Myin-Germeys2011; Rauschenberg et al., Reference Rauschenberg, van Os, Cremers, Goedhart, Schieveld and Reininghaus2017; Reininghaus et al., Reference Reininghaus, Gayer-Anderson, Valmaggia, Kempton, Calem, Onyejiaka and Morgan2016b) or bullying experiences (Rauschenberg et al., Reference Rauschenberg, van Os, Goedhart, Schieveld and Reininghaus2021c) have been shown to impact stress reactivity, but were not included in the current analysis given the limited sample size and number of ESM observations and, thus, confounding by, or further interaction with, these factors were not taken into account.
Last, the distribution of ESM data was skewed, which violates the assumption of normally distributed residuals in linear mixed models. However, when we fitted multilevel tobit regression models in a sensitivity analysis to assess how this may have impacted our findings, these remained largely unchanged. Also, cases completed fewer ESM assessments than siblings and controls, but this did not seem to be accounted for by interference of these assessments with their daily life.
Comparison with previous research
To our knowledge, the present study is the first to investigate PRS and ESM data of a clinical sample. In many previous studies, genetic risk of psychosis has been approximated by family history, although this of course does not determine onset of the disorder (Lu et al., Reference Lu, Pouget, Andreassen, Djurovic, Esko, Hultman and Sullivan2018). Myin-Germeys et al. (Reference Myin-Germeys, van Os, Schwartz, Stone and Delespaul2001) previously observed that relatives compared to controls reported increased negative affect and psychotic experiences in response to momentary stress. Furthermore, stress sensitization has been postulated to comprise an important mechanism in pathways to psychosis (Collip et al., Reference Collip, Myin-Germeys and Van Os2008; Myin-Germeys & van Os, Reference Myin-Germeys and van Os2007). We aimed to replicate and extend these findings and underpin the proposed aetiological model.
In the current study, we found that healthy individuals with higher polygenic risk for psychosis responded with more intense psychotic experiences to daily life stress compared to those with low polygenic risk. In line with this, psychotic reactivity to momentary stress has been reported to be modified by high polygenic risk (and exposure to childhood trauma) in healthy controls (Pries et al., Reference Pries, Klingenberg, Menne-Lothmann, Decoster, van Winkel, Collip and Guloksuz2020), which, in line with the stress sensitization model suggests that polygenic risk sensitizes healthy individuals to the psychosis-inducing effects of minor stressors in daily life. However, in contrast to this hypothesis, our results suggest no evidence of stress reactivity to be modified by polygenic risk in individuals with enduring psychotic disorder. One plausible explanation for this finding may be that the psychosis-inducing effects of minor stressors in daily life may particularly operate in individuals with high PRS primarily prior to onset, in the early stages of psychosis, and attenuate over time due to the effects of illness chronicity and exposure to antipsychotic medication (van der Steen et al., Reference van der Steen, Gimpel-Drees, Lataster, Viechtbauer, Simons, Lardinois and Myin-Germeys2017).
Furthermore, in contrast to our hypotheses, siblings with high PRS appeared to be resilient to the exposure of stress as they reported less intense psychotic experiences in response to momentary stress. In the present study, siblings largely grew up in a shared environment with cases, which may have substantially contributed to familial liability. Having a close relative coping with psychotic experiences may affect individuals’ interpretation of their own experiences. As families experiencing mental health-related problems have more exposure to mental health-related information and mental health services, this may increase health literacy (Hurley, Swann, Allen, Ferguson, & Vella, Reference Hurley, Swann, Allen, Ferguson and Vella2020). Thus, siblings may better recognize early warning signs or show health-promoting behaviour when distressed. A high PRS for schizophrenia in relatives indicated an increased polygenic risk in individuals with increased familial liability to psychosis. There is evidence from previous research that PRS is higher in relatives than controls (van Os et al., Reference van Os, Pries, Delespaul, Kenis, Luykx, Lin and Guloksuz2020; van Os et al., Reference van Os, van der Steen, Islam, Guloksuz, Rutten, Simons and Investigators2017b), though, notably, in the present study a high PRS in relatives reflected a low PRS in cases of our sample. The marginal difference in PRS between siblings and controls in the present study may result from selecting the specific analytic sample used for the current analysis. Another explanation for the finding that stress reactivity is reduced in siblings may point to a genetic resilience factor that may have mitigated polygenic and environmental vulnerability. In fact, Hess, Tylee, Mattheisen, Borglum, and Glatt (Reference Hess, Tylee, Mattheisen, Borglum and Glatt2019) proposed a polygenic resilience score for schizophrenia, i.e. a heritable gene variation that reduces the penetrance of risk loci. To this end, they identified healthy relatives with high PRS for psychosis and compared this sample to risk-matched cases showing that the resilience score increases in unaffected individuals as their PRS increases. The specific role of buffering or protective factors in pathways to psychosis such as a polygenic resilience score for schizophrenia or a resilience-enhancing social environment (Gayer-Anderson & Morgan, Reference Gayer-Anderson and Morgan2013), and their impact on targetable mechanisms needs to be elucidated further as a basis for improving prevention (Rauschenberg et al., Reference Rauschenberg, Schick, Goetzl, Roehr, Riedel-Heller, Koppe and Reininghaus2021a, Reference Rauschenberg, Schick, Hirjak, Seidler, Paetzold, Apfelbacher and Reininghaus2021b; Reininghaus et al., Reference Reininghaus, Depp and Myin-Germeys2016a), treatment (Schick et al., Reference Schick, Paetzold, Rauschenberg, Hirjak, Banaschewski, Meyer-Lindenberg and Reininghaus2021; van Aubel et al., Reference van Aubel, Bakker, Batink, Michielse, Goossens, Lange and Myin-Germeys2020) and, ultimately, outcomes of psychosis.
Conclusion
In contrast to previous propositions, the current work provided no evidence that polygenic risk impacts reactivity to minor stressors in daily life in individuals with enduring non-affective psychotic disorder and prolonged exposure to antipsychotic medication. Our findings tentatively suggest that polygenic risk may operate in different ways than previously assumed and amplify stress reactivity in unaffected individuals but take on the role of a resilience factor in relatives by attenuating their stress reactivity. Stress reactivity may reflect a putative mechanism underlying polygenic risk and resilience to psychosis. Targeting this putative mechanism in individuals' daily life through novel ecological momentary interventions as experimental manipulation method is an important next step (Reininghaus et al., Reference Reininghaus, Depp and Myin-Germeys2016a). This not only promises to further our understanding of how this mechanism impacts individuals with varying levels of risk but will also pave new ways to the prevention of, and treatment for, psychosis.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721004761.
Acknowledgements
The authors would like to thank all participants of the GROUP study. Furthermore, we would like to thank all research personnel involved in the project, in particular Joyce van Baaren, Erwin Veermans, Ger Driessen, Truda Driesen, Erna van ’t Hag.
Financial support
Ulrich Reininghaus was supported by a Heisenberg professorship (no. 389624707) funded by the German Research Foundation (DFG). Inez Myin-Germeys was supported by an FWO Odysseus grant (G0F8416N). Ruud van Winkel is supported by Research Foundation Flanders (FWO) (grant number G063518N, Senior Clinical Fellowship 1803616N) and the King Baudoin Foundation (Chair of Transition Psychiatry). The infrastructure for the GROUP study was funded by the Geestkracht programme of the Dutch Health Research Council (ZON-MW, 10-000-1002) and funds from participating universities and mental health care organizations (Amsterdam: Academic Psychiatric Centre of the Academic Medical Center and the mental health institutions: GGZ Ingeest; Arkin, Dijk en Duin; GGZ Rivierduinen; Erasmus Medical Centre and GGZ Noord Holland Noord. Groningen: University Medical Center Groningen and the mental health institutions: Lentis, GGZ Friesland; GGZ Drenthe; Dimence; Mediant; GGNet Warnsveld; Yulius Dordrecht and Parnassia psycho-medical center The Hague. Maastricht: Maastricht University Medical Centre and the mental health institutions: GGZ Eindhoven en De Kempen; GGZ Breburg; GGZ Oost-Brabant; Vincent van Gogh voor Geestelijke Gezondheid; Mondriaan; Virenze riagg; Zuyderland GGZ; MET ggz; Universitair Centrum Sint-Jozef Kortenberg; CAPRI University of Antwerp; PC Ziekeren Sint-Truiden; PZ Sancta Maria Sint-Truiden; GGZ Overpelt and OPZ Rekem. Utrecht: University Medical Center Utrecht and the mental health institutions: Altrecht; GGZ Centraal and Delta).
Conflict of interest
None.
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
