Participants were recruited by the “Early Recognition and Intervention Program for Psychosis and Bipolar Disorder” project as part of the Zurich Program for Sustainable Development of Mental Health Services (ZInEP)¹ in the Canton of Zurich, Switzerland. The detailed design, inclusion, and exclusion criteria of the study were described in our previous studies (30, 31). A total of 185 individuals aged 13–35 years at HR/UHR for psychosis were assessed at baseline and were subsequently followed-up at 6, 12, 24, and 36 months for the transition to schizophrenia. At 36 months follow-up, 50% of individuals (n = 93) dropped out of the study. The dropouts were due to refusal to participate in the study in most of the cases and due to non-response after contacting in few cases. However, we are unable to give exact numbers of individuals who refused to participate or who did not respond due to missing information in several study participants. Individuals at HR for psychosis were assessed using the Schizophrenia Proneness Instrument-Child and Youth version (SPI-CY) (age < 18 years) (32, 33) or Schizophrenia Proneness Instrument-Adult version (SPI-A) (age ≥ 18 years) (34) and were included when they had one COPER basic symptom or at least two COGDIS. Individuals at UHR for psychosis were assessed using the Structured Interview for Prodromal Syndromes (SIPS) (35, 36) and were included when they met criteria for the APSS or the BLIPS or the state-trait criteria (>30% reduction in global assessment of functioning in the past year plus either schizotypal personality disorder or a first-degree relative with psychosis). The transition to schizophrenia was defined according to ICD-10 criteria (37). The severity of positive and negative symptoms was assessed using the Positive and Negative Syndrome Scale (PANSS) (38), severity of depressive symptoms with the Calgary Depression Rating scale for Schizophrenia (CDSS) (39), and anxiety symptoms with the Beck Anxiety Inventory (BAI) (40). The demographic and diagnostic characteristics of the study population are shown in Table S1 in Supplementary Material. This study was approved by the Cantonal Ethics Commission of Zurich (Ref. Nr. EK: E-63/2009) and complies with the Declaration of Helsinki. Informed written consent was obtained from adult participants and legal guardians of minors, and written assent was obtained from minors.

The participants were grouped into clinical phenotypes (cases versus controls) namely, converters to schizophrenia spectrum disorders (n = 27), i.e., schizophrenia, schizophreniform disorder, and acute psychotic disorder, versus non-converters (n = 65) at 36 months follow-up [dropouts (n = 93)], and APSS (n = 98) group versus all other help-seeking individuals (BLIPS, state-trait criteria, COGDIS, COPER; n = 87) at baseline. The transition to schizophrenia was defined according to ICD-10 criteria (37).

First, we grouped our cohort as per the factor structure of the PANSS, concentrating on negative symptoms and general psychopathology (38). As we did not find any significant differences in negative symptoms (sum score of negative (N1–N7) PANSS subscale) and general psychopathology (sum score of general psychopathology (G1–G16) PANSS subscale) across DAO, DAOA, and NRG1 SNPs, we then decided to use an exploratory approach by focusing on subgroups of psychopathology constructs defined according to the RDoC domains. The RDoC framework consists of five domains namely, negative valence systems, positive valence systems, cognitive systems, systems for social process, and arousal/regulatory systems (41). In this study, we decided to concentrate on two RDoC domains: negative valence systems and cognitive systems, due to the potential role of DAO, DAOA, and NRG1 polymorphisms in the glutamate hypothesis of schizophrenia, as it appears to explain the pathogenesis of negative symptoms and cognitive deficits better than the dopamine hypothesis (22, 23, 42) and unavailability of relevant neuropsychological scales in our study to create the positive valence domain. Negative valence systems focus on responses to aversive situations, such as fear, anxiety, and loss. Cognitive systems concentrate on various cognitive processes, such as perception, language, memory, and cognitive control. Within negative valence systems, we focused on the constructs of threat (acute and sustained) and loss. The negative valence system construct of threat was assessed by the total score of BAI with higher BAI scores pointing to increased severity of anxiety (40). The negative valence system construct loss was assessed by the response to the CDSS item 2 “hopelessness,” which can be scored as 0 (absent), 1 (mild), 2 (moderate), or 3 (severe) (39). Within cognitive systems, we chose the constructs of perception (visual and auditory) and cognitive control. The cognitive system construct visual perception was assessed by summing the following four SPI-A (and equivalent SPI-CY) items O4 (other visual perception disturbances), F1 (hypersensitivity to light), F2 (photopsia), and F3 (micropsia/macropsia), with higher scores pointing to more frequent disturbances in visual perception. The cognitive system construct of auditory perception was assessed by summing the following three SPI-A (and equivalent SPI-CY) items O5 (other acoustic perception disturbances), F4 (hypersensitivity to sounds/noise), and F5 (changed intensity/quality of acoustic stimuli), with higher scores pointing to more frequent disturbances in auditory perception. The cognitive system construct of cognitive control was obtained by summing the following four SIPS disorganization items D1, D2, D3, and D4 (35, 36), with higher scores indicating more severe disturbance in disorganized symptoms.

The above-mentioned scales used to tap into RDoC domains were assessed at baseline and last-available follow-up until 36 months. As there was a dropout rate of 50% at 36 months, if there were no data available at 36 months, we took the scores from the last follow-up that the individual attended (i.e., 6 or 12 or 24 months).

The study population was genotyped for DAO (rs3918347, rs4623951), DAOA (rs778293, rs3916971, rs746187), and NRG1 (rs10503929) SNPs. These SNPs were selected based on previously reported significant association with schizophrenia (10, 43). In our recent meta-analysis, we found a significant association of DAO rs4623951 [odds ratio (OR) = 0.88; minor allele: C], DAOA rs778293 (OR = 1.17; minor allele: G), DAOA rs3916971 (OR = 0.84; minor allele: T), and NRG1 rs10503929 (OR = 0.89; minor allele: C) with schizophrenia (11). In this study, the carriers of the risk allele of DAO, DAOA, and NRG1 SNPs were anticipated to have worse psychopathology scores in RDoC-negative valence and cognitive system constructs. DNA was isolated from whole blood ethylenediaminetetraacetic acid tubes collected from the study population using QIAamp DNA Blood Maxi Kit (Qiagen) as per manufacturer’s protocol. A spectrophotometer (NanoVue Plus, GE) was used to measure DNA concentrations, A260/A280, and A260/A230 ratios. The study population was genotyped for DAO (rs3918347 assay number: C_27937201_10, rs4623951 assay number: C_32177440_10, both from Applied Biosystems, USA), DAOA (rs778293 assay number: C_8704507_10, rs3916971 assay number: C_27495752_10, rs746187 assay number: C_1925241_10, all from Applied Biosystems, USA), and NRG1 (rs10503929, assay number: C_2870393_10, Applied Biosystems, USA) SNPs (44). DNA (10 ng/µl); TaqMan^® Genotyping Master Mix (Applied Biosystems, USA); and above-mentioned DAO, DAOA, and NRG1 SNP Genotyping Assays (Applied Biosystems, USA) were combined in a 384-well plate. Real-time polymerase chain reaction (PCR) was performed in a C1000™CFX384™ Thermal cycler (Bio-Rad) using TaqMan^® SNP Genotyping Assay PCR standard protocol. The allelic discrimination program of Bio-Rad CFX Manager™ Software version 2.1 was used to determine genotypes (44). Samples were run in duplicates to ensure correct results. In case of ambiguity in duplicates, genotyping was repeated in a separate run to ensure correct results. No-template controls were included in every run to exclude impurities.

RNA was isolated from whole blood collected from the study population using PAXgene Blood RNA Kit (Qiagen) according to manufacturer’s protocol. A spectrophotometer (NanoVue Plus, GE) was used to measure RNA concentrations, A260/A280, and A260/A230 ratios. RNA integrity was analyzed using Experion automated electrophoresis system (Bio-Rad) in a subset of samples to ensure RNA integrity number/RNA quality indicator >7. RNA (500 ng) was reverse transcribed using iScript™ cDNA Synthesis Kit (Bio-Rad) as per manufacturer’s protocol. In a subset of samples, negative controls were prepared with RNA using iScript™ cDNA Synthesis Kit (Bio-Rad) without reverse transcriptase enzyme as per manufacturer’s protocol. qRT-PCR was performed using cDNA, QuantiFast SYBR Green PCR kit (Qiagen), 1 µM NRG1 primer (QT00061964, Qiagen), and reference genes [β-actin (ACTB) (QT01680476), aminolevulinate synthetase (ALAS1) (QT00073122), ribosomal protein L13a (RPL13A) (QT00089915), alanyl-tRNA synthetase (AARS) (QT00054747), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (QT01192646), peptidyl prolyl isomerase A (PPIA) (QT00866137), and X-prolyl aminopeptidase1 (XPNPEP1) (QT00051471); all from Qiagen]. NRG1 mRNA levels were normalized to the references genes (44). PCR efficiencies were calculated using LinReg PCR program (45), and mean PCR efficiencies for all studied amplicons were found to be between 91 and 93%. Normalized NRG1 mRNA levels were quantified using qBASE plus software (Biogazelle), which utilizes gene-specific amplification efficiencies and allows normalization with multiple reference genes (46).

We performed qRT-PCR to detect DAO mRNA using DAO primers described by Verrall et al. (47) and predesigned primers [qHsaCID0011122 and qHsaCEP0058247 (Bio-Rad), Hs.PT.58.3248433 and Hs.PT.58.45768871 (IDT)]. We performed qRT-PCR to detect DAOA mRNA using primers for DAOA gene described by Benzel et al. (48), Cheng et al. (49), and pre-designed primers [QT00058863 (Qiagen), Hs.PT.58.555086 (IDT), 4331182 (ThermoFisher scientific), qHsaCEP0024792 (Bio-Rad)]. QuantiTect Whole Transcriptome Kit (207043, Qiagen) followed by qRT-PCR was also used to detect DAO and DAOA mRNA levels (44). However, we were unable to quantify DAO and DAOA mRNA with the aforementioned methods in the whole blood as either no signal was observed or genomic DNA was amplified concomitantly.

The results from SNP genotyping was analyzed using PLINK software (50). The minor allele frequency (MAF) and deviation from Hardy-Weinberg Equilibrium (HWE) was computed using PLINK software, and p < 0.05 was considered as statistically significant. The DAO (rs3918347), DAOA (rs3916971, rs778293, rs746187), and NRG1 (rs10503929) SNPs were in HWE (p > 0.05), and the MAF of DAO, DAOA, and NRG1 SNPs were similar to HapMap CEU MAF (Table S5 in Supplementary Material). The DAO rs4623951 SNP deviated from the HWE (p < 0.05; Table S4 in Supplementary Material). The differences in allele and genotype frequencies across clinical phenotypes (cases versus controls) were assessed using Chi-square test, and p < 0.05 was considered as statistically significant. The OR, 95% CI, and p value for SNP models (allelic, dominant, and recessive) across clinical phenotypes were calculated from allele/genotype frequencies using an online OR calculator,² and p value was adjusted based on the number of analyzed SNPs (Bonferroni correction, p < 0.008). The post hoc power analyses for association of DAO, DAOA, and NRG1 SNPs with converters to schizophrenia spectrum disorders versus non-converters (Table S5 in Supplementary Material) and APSS versus all other help-seeking group were conducted using Fisher’s exact test of independence in G*Power software (51), the calculated OR was used, and the alpha level was set at 0.05.

IBM^® SPSS^® Statistics (version 21) software was used for statistical analysis. Shapiro–Wilk test with Lilliefors significance correction was used to assess the normality of the distribution of NRG1 gene expression and clinical scales (BAI, CDSS, SPI-A/SPI-CY, SIPS, and PANSS). NRG1 gene expression and clinical scales showed both normal and non-normal distribution. We used non-parametric tests even for normally distributed data to maintain consistency between statistical evaluations. The differences in RDoC domains (negative valence and cognitive systems) across models (genotypic, dominant, recessive) were assessed using Mann–Whitney test (for two groups) or Kruskal–Wallis test (for >2 groups), and p values were adjusted based on the number of constructs analyzed (Bonferroni correction, p < 0.008). The differences in NRG1 mRNA levels across clinical phenotypes were assessed using Mann–Whitney test, and p < 0.05 was set as statistically significant. The post hoc power analyses for RDoC-negative valence and cognitive systems across DAO, DAOA, and NRG1 SNPs were conducted using analysis of variance test for three groups or t-test for two groups in G*Power software (51), the effect sizes were determined from means of neuropsychological scales used in RDoC-negative valence and cognitive systems, and the alpha level was set at 0.05 (Table S7 in Supplementary Material). The differences in NRG1 mRNA levels across NRG1 rs10503929 SNP genotypes was assessed using Kruskal–Wallis test, and across dominant (CC + CT, TT) and recessive models (CC, CT + TT) were assessed using Mann–Whitney test, with p < 0.05 being set as statistically significant. The correlation between NRG1 gene expression and RDoC domains (negative valence and cognitive systems) was assessed using Spearman’s rank correlation test, and p < 0.05 was considered statistically significant.

There were no significant associations between DAO, DAOA, and NRG1 SNPs with converters to schizophrenia spectrum disorders compared to non-converters at 36 months follow-up (power range: 0.05–0.31; Table S3 in Supplementary Material). However, there was a nominal association (p > 0.008) of DAO rs3918347 with APSS compared to all other help-seeking group at baseline, and the G-allele had a tendency to be a risk allele for APSS (OR = 1.84, 95% CI = 1.13–3.01, p = 0.01), with a power of 0.76 (Table S4 in Supplementary Material). There were no significant associations between the rest of the DAO, DAOA, NRG1 SNPs and APSS (power range: 0.05–0.89; Table S4 in Supplementary Material).

d-amino acid oxidase activator rs746187 recessive (CC) genotype carriers experienced nominally more hopelessness (higher item 2 score, CDSS) than CT + TT genotype carriers at the last-available follow-up time point (LA) until 36 months (p = 0.04) analyzed by Mann–Whitney test (RDoC-negative valence system: loss; Figures 1A–C; power = 0.62; effect size = 0.55; Table S8 in Supplementary Material). DAOA rs3916971 recessive (TT) genotype carriers experienced nominally more disturbances in visual perception (higher sum of O4 + F1 + F2 + F3, SPI-A) than TC + CC genotype carriers at LA until 36 months (p = 0.009) analyzed by Mann–Whitney test (RDoC cognitive system: visual perception; Figures 2A–C; power = 0.28; effect size = 0.34; Table S8 in Supplementary Material). Individuals with DAOA rs3916971 TC + CC versus TT genotype improved and had fewer visual perceptual disturbances at LA until 36 months compared to baseline, but at LA until 36 months versus baseline, they continued to have less visual perceptual disturbances than individuals with TT genotype (Figure 2C). There were no significant differences in negative valence (threat; Table S5 in Supplementary Material; power range: 0.05–0.72, Table S8 in Supplementary Material), auditory perception disturbances (Table S6 in Supplementary Material; power range: 0.05–0.56, Table S8 in Supplementary Material), and cognitive control (Table S7 in Supplementary Material; power range: 0.05–0.31, Table S8 in Supplementary Material) in individuals with DAOA SNP genotypes (rs3916971, rs778293, rs746187).

d-amino acid oxidase rs3918347 GA + AA genotype carriers experienced nominally (p = 0.04) more disturbances in auditory perception (higher sum of O5 + F4 + F5, SPI-A) than GG genotype carriers at baseline analyzed by Mann–Whitney test (RDoC cognitive system: auditory perception; Figures 3A–C; power = 0.41; effect size = 0.44; Table S8 in Supplementary Material). Individuals with DAO rs3918347 GA + AA versus GG genotype improved and had lesser auditory perceptual disturbances at LA until 36 months compared to baseline, but at LA until 36 months versus baseline, they continued to experience more auditory perceptual disturbances than individuals with CC genotype (Figure 3C). We did not find significant differences in negative valence systems (anxiousness and hopelessness; Table S5 in Supplementary Material; power range: 0.05–0.92, Table S8 in Supplementary Material), visual perception disturbances (Table S6 in Supplementary Material; power range: 0.05–0.57, Table S8 in Supplementary Material), and cognitive control (Table S7 in Supplementary Material; power range: 0.05–0.22, Table S8 in Supplementary Material) in individuals with DAO (rs3918347, rs4623951) SNP genotypes.

Neuregulin 1 rs10503929 CC + CT genotype carriers had significantly (p = 0.001) more disorganized symptoms (higher sum of the D1–D4 score, SIPS) than TT genotype carriers at baseline analyzed by Mann–Whitney test (RDoC cognitive system: cognitive control; Figures 4A–C; power = 0.99; effect size = 0.59; Table S8 in Supplementary Material). There were no significant differences in negative valence systems (anxiousness and hopelessness; Table S5 in Supplementary Material; power range: 0.07–0.51, Table S8 in Supplementary Material) and visual and auditory perception disturbances (Table S6 in Supplementary Material; power range: 0.05–0.36, Table S8 in Supplementary Material) in individuals with NRG1 (rs10503929) SNP genotypes.

To analyze the effect of age on RDoC domains across DAO, DAOA, and NRG1 SNPs, we performed an analysis of covariance with age as a covariate after transformation of not normally distributed BAI, SPI-A, and SIPS data, as described previously (52). We did not find a significant effect of age on differences in RDoC domains across DAO, DAOA, and NRG1 SNPs (data not shown).

There was a significant positive correlation between increased NRG1 mRNA levels and higher scores on the RDoC-negative valence system loss at LA until 36 months (Table S9 in Supplementary Material). Apart from this, there were no significant correlations between NRG1 mRNA levels and RDoC domains at baseline and at LA until 36 months (Table S9 in Supplementary Material). There were no statistically significant differences in NRG1 mRNA levels across clinical phenotypes (Table S10 in Supplementary Material). We did not find significant differences in NRG1 mRNA levels across NRG1 rs10503929 SNP genotypes, dominant, and recessive models analyzed by Mann–Whitney and Kruskal–Wallis tests (Table S11 in Supplementary Material). We were unable to quantify DAO and DAOA mRNA levels in the whole blood of at-risk population.