Schizophrenia polygenic risk score in psychosis proneness

Schizophrenia (SZ) is a severe and disabling mental disorder that affects around 24 million people worldwide [1]. Family and twin studies have provided evidence of its multifactorial origin, with a strong genetic component evidenced by heritability estimates around 70–80% [2, 3]. As confirmed over the last few years by genome-wide association studies (GWAS), the genetic architecture of this disorder is highly polygenic, with the cumulative effects of a large number of genes involved. However, although numerous risk loci have already been identified, these variants still only explain a relatively small fraction of the overall heritability of SZ [4, 5].

According to the dimensional view of mental disorders, the psychosis phenotype is manifested across a dynamic continuum in which SZ represents the most extreme of a much broadly distributed clinical expression of psychosis liability expressed as schizotypy traits and psychotic-like experiences in the general population [6‐9]. SZ and its subclinical presentations are heterogeneous and this heterogeneity can be captured in a multidimensional structure, with positive, negative and disorganized symptom dimensions most commonly identified. The positive dimension involves odd beliefs ranging from trait-like features such as magical thinking to sub/clinical symptoms like delusions, unusual perceptual experiences that include illusions and hallucinations, suspiciousness and paranoia. The negative or deficit dimension comprises anhedonia, flattened affect, alogia, anergia, and disinterest in the world. And the disorganization dimension involves disruptions in the organization and expression of thought, communication, emotion, and behaviour. These dimensions range from adaptation or minimal dysfunction to overt clinical psychosis, possibly reflecting certain genetic and non-genetic etiological continuity [10‐12]. Subclinical traits are hypothesized to be genetically less complex than clinical phenotypes and more directly related to aetiological factors than categorical diagnostic groups, thus being considered interesting candidate phenotypes for the study of SZ [13‐16].

To date, little is known about the contribution of genetic risk loci for SZ to SZ-related traits in the general population [17]. In the pre-GWAS era, schizotypy and psychotic-like experiences have been considered as phenotypes in candidate gene studies attempting to identify susceptibility variants related to SZ and both have been found to be associated with previously reported genetic risk variants for this disorder [18‐23]. More recently, GWAS have laid the groundwork for the identification of the polygenicity of SZ. One of the tools that GWAS have facilitated is the calculation of Polygenic Risk Scores (PRSs), by computing the sum of an individual’s risk alleles weighted by the effect sizes of such alleles. PRSs provide an estimation of the individual genetic liability to a trait or a disorder and can be used to study the shared genetic aetiology among complex traits at the population level [24].

In the field of psychosis proneness, studies applying PRSs for SZ (SZ-PRSs) to examine the genetic overlap between SZ and its related subclinical phenotypes so far have led to controversial results. On the one hand, several previous studies failed to detect an association between SZ-PRSs and psychotic-like experiences when examining population-based samples of varying sample sizes (e.g., [25‐28]). Similarly, Nenadić and colleagues [29] tested the hypothesized association of schizotypy and SZ-PRSs in two non-clinical samples and were not able to find a significant association either. In line with these negative results, a previous study exploring the same hypothesis in a sample of male army recruits reported an inverse association with schizotypy, but follow-up analyses revealed that the association only held under stressful conditions, suggesting an environmental impact rather than a SZ-related genetic influence [30]. However, there is also a growing number of studies with significant findings that support the existence of a shared genetic background between this disorder and its related phenotypes. For instance, four studies examining relatively large general population samples found evidence of an association between SZ-PRSs and multiple measures of psychotic-like experiences [31‐34]. Likewise, Karcher and colleagues [35] recently reported an association between SZ-PRSs and distressing psychotic-like experiences in a population-based cohort of children. Regarding schizotypy, Docherty et al. [36] examined a sample of healthy individuals and found a male-specific association between SZ-PRSs and schizotypy. Additionally, van Os et al. [37] reported an association with both positive and negative schizotypy in a similar sample. It should be pointed out that the authors of this last study, as well as Zammit and colleagues [25], used interview-based measures to assess schizotypy and psychotic-like experiences instead of self-report questionnaires, which might have avoided certain phenotypic assessment biases.

Given this background, the aim of this study was to examine the contribution of SZ genetic risk variants to SZ-related traits by analyzing whether SZ-PRSs are associated with SZ-related phenotypes in a sample of non-clinical young adults. The phenotyping of participants was enriched by including the assessment of self-reported traits and psychotic-like experiences as well as face-to-face interviews of a broad range of subclinical experiences and symptoms.

The present study aimed to investigate in a non-clinically ascertained sample whether SZ-PRSs were associated with SZ-related subclinical phenotypes, that is, schizotypy dimensions, psychotic-like experiences and interview-ratings of a broad range of subclinical experiences and symptoms.

When analyzing the two self-report assessments, we did not find any significant association of the polygenic burden for SZ with psychotic-like experiences or with schizotypy. In the past years, in accordance with an increasing support for the psychosis extended phenotype hypothesis (e.g., [9‐11]), many attempts have been made in order to find evidence of an overlapping genetic architecture between SZ and its related phenotypes. However, the previous literature specifically examining psychotic-like experiences and schizotypy in relation to SZ-PRSs shows inconsistent results, so the existence of a genetic overlap with these phenotypes is still unclear. Although some studies have found evidence of a shared genetic aetiology with different measures of psychotic-like experiences and schizotypy [31‐34, 36], our findings indicating a lack of an association concurs with several other studies. For instance, both Sieradzka et al. [26] and Zammit et al. [25] examined in relatively large population-based samples whether SZ-PRSs were associated with different measures of psychotic-like experiences and neither of them identified any significant association. In line with these negative results, van Os et al. [28] investigated in two independent healthy comparison samples whether SZ-PRSs were associated with psychotic-like experiences assessed with the CAPE self-report questionnaire and did not detect any significant association either. Finally, our results are consistent with those of Nenadić and colleagues [29], who were unable to identify any significant association when exploring the genetic overlap between SZ-PRSs and schizotypy in two non-clinical samples. This led them to propose that schizotypy should be regarded as a wider phenotype beyond merely harbouring risk for SZ, consistent with dimensional conceptualizations of the dual nature of schizotypy, as any other trait, as an indicator of both normal individual differences as well as behavioural risk for psychosis [57]. Therefore, specific genes giving risk for SZ may in a sense wash out in the larger pool of schizotypic individuals, which does not diminish the utility of schizotypy, but highlights that it is an interesting construct in its own right, not simply a prodromal or risk condition for SZ. Nevertheless, the comparison of previous results between these types of studies should be done with caution as the different instruments and questionnaires used for the psychometric assessment could be capturing different underlying concepts [58].

Other plausible explanations for such a lack of associations in the literature have been pointed out. Nenadić et al. [29] suggested the possibility that, in the non-clinical part of the psychosis continuum, environmental stressors may have a larger effect on the phenotypic expression of SZ-related traits than genetic predisposition, which would be in consonance with the low variance of different subclinical phenotypes that SZ-PRSs explain. Some studies have already found evidence of an environmental contribution to the expression of psychotic-like experiences and schizotypy in samples from the general population. For example, a link between psychotic-like experiences and smoking and using cannabis in general population samples has been described [59‐61]. In a similar fashion, Pries and colleagues [62] computed a score of cumulative environmental load that included childhood adversity, winter-birth, cannabis use, and hearing impairment and found that it was associated with positive, negative, and total schizotypy. Additionally, in line with these findings a previous work identified an association between SZ genetic load and positive schizotypy in a sample of male army recruits, but only at the stressed condition of military induction, which denoted an environmental influence [30]. It is also likely that in the non-clinical end of the psychosis continuum, plasticity alleles rather than risk alleles (i.e., alleles that confer sensitivity to both positive and negative environmental influences rather than alleles that only confer vulnerability in the presence of environmental adversity) play a more relevant role in the underlying pathways that lead to the expression (or not) of these subclinical manifestations. Therefore, PRSs that likely reflect susceptibility to environmental influences rather than risk to develop SZ might better capture the genetic architecture of these subclinical traits [63].

Another plausible explanation for the increasing number of negative results in previous research could be that, like SZ, its related subclinical phenotypes might also be determined by different types of genetic variants beyond common SNPs considered in PRS computation. In fact, SNP-based heritability estimates for SZ indicate that common variation only explains around 24% of the variance in SZ liability [4]. Therefore, the lack of significant associations cannot entirely rule out the possibility of a genetic overlap between these phenotypes, as it could mainly be conformed of other types of genetic variation such as copy number variants or rare variants. Finally, it could be that the PRSs constructed based on variants associated with clinically diagnosed SZ might not be able to capture the subclinical manifestations of psychosis since these typically overlap within a transdiagnostic mix of symptoms [64]. Thus, future research using PRSs built with variants associated with these subclinical symptoms in non-clinical or at-risk samples could elucidate this question.

Regarding the CAARMS interview, our findings seem to identify an association between the polygenic risk for SZ and the presence of motor abnormalities. We found an association with the Motor/physical Change subscale, which evaluates subjectively experienced difficulties with movement and objective signs of catatonia, including: subjective complaints of impaired motor functioning; informant reported or observed changes in motor functioning; subjective complaints of impaired bodily sensation; and subjective complaints of impaired autonomic functioning [51]. Individuals reporting some degree of motor abnormalities on the Motor Change subscale presented higher SZ-PRSs, which suggests that non-clinical individuals with a higher end of polygenic burden for SZ already present some degree of motor dysfunction despite being functional young adults, in comparison with those subjects with a low SZ polygenic load. In fact, when we divided our sample into quartiles of increasing PRS, a trend towards an increase in motor changes with increasing SZ polygenic load could be observed, where the risk of presenting some motor abnormalities was three times higher for individuals in the highest quartile than those in the lowest quartile. Given that SZ is known to affect men and women differently [65], the association analyses were also conducted based on gender. The association between the SZ polygenic burden and the presence of motor abnormalities found in the whole sample was also detected in the female subsample, although in this case all significance was lost after FDR correction. Regarding men, no significant association was found. However, given that the size of the male subsample was considerably small (N = 64), we cannot rule out the possibility that this association also exists in men. Moreover, it could be observed that the association was stronger for the whole sample than for the female subset only, which suggests that men were actually contributing to the significance of the association, rather than diminishing it. Nevertheless, this should be considered cautiously, given that we did not have specific hypotheses about this domain relative to the other domains tapped by the CAARMS.

Some studies have estimated that up to 80% of patients with SZ present some motor anomalies, already observed very early during premorbid development in most of these patients [66]. These abnormalities have been associated with poorer psychopathological, cognitive, and social outcomes [67‐69]. In fact, motor impairment constitutes a key transdiagnostic feature indexing disease severity [70, 71] and a risk factor for conversion to psychosis [72, 73]. Consistent with recent claims of a severity continuum of psychopathology, in which established psychosis (e.g., SZ) might index the extreme end of this continuum [74], SZ-PRSs might also be reflective of a severity score and thus, be more likely to detect most severe manifestations in a non-clinical sample of young adults. The association between SZ-PRSs and the presence of motor abnormalities might be detecting those individuals with poorer functioning and with increased liability for transitioning to clinical at-risk states. It remains to be established how these results connect with neurodevelopmental processes and if PRSs are able to finally shed a light in the interplay between genetic risk, neurodevelopmental processes, and subclinical traits in the general population.

The findings of the present study should be interpreted with caution bearing in mind some limitations. On the one hand, given the face-to-face nature of part of the psychometric assessment, our sample was relatively small for the type of investigation carried out and a slight lack of statistical power has to be acknowledged. On the other hand, the sex-stratified analyses performed may be biased since more than 70% of the participants were women and the size of the male subsample (N = 64) will likely have affected our statistical power. Additionally, we had no knowledge of the developmental and drug use history of the participants, which could have added an interesting insight to the analyses carried out. Finally, due to the non-clinical nature of the sample analyzed in this study, most participants reported relatively low scores on the psychometric scales. On the CAARMS interview, especially, a clear floor effect could be observed as this interview was initially designed to evaluate help-seeking individuals rather than non-clinical subjects—although we highlight that the present sample was oversampled for both positive and negative schizotypy and psychotic-like experiences from a larger unselected sample. Thus, the CAARMS variables were dichotomized, which might have led to a loss of statistical power [75]. Even so, it is worth noting that we were still able to detect a robust association with the Motor Change subscale of this interview.

In conclusion, our results seem to suggest that schizotypy and psychotic-like experiences share less genetic variability with SZ than initially hypothesized and that SZ-PRSs are able to capture the subjective motor changes measured with the CAARMS interview presented in some healthy subjects from the general population. Further studies in larger non-clinical samples are required in order to finally unravel the shared genetic background between this disorder and its subclinical phenotypes and to ensure the replicability of the association with motor abnormalities found in the present study. Moreover, the replication of this finding across levels of severity expression (e.g., at risk mental states and first episode psychosis) would allow us to both understand more in depth the nature of this association and test whether, as it would be expected, this association becomes more evident at greater severity expressions of the extended psychosis phenotype that presumably index increasing genetic load for SZ.