Introduction
Schizophrenia (SCZ) is a markedly heritable neurodevelopmental disorder characterized by brain functional and structural abnormalities alongside cognitive impairments (Jauhar, Johnstone, & McKenna, Reference Jauhar, Johnstone and McKenna2022). Existing heritability estimates range from 64% in familial investigations to approximately 80% in twin studies (Hilker et al., Reference Hilker, Helenius, Fagerlund, Skytthe, Christensen, Werge and Glenthøj2018; Lichtenstein et al., Reference Lichtenstein, Yip, Björk, Pawitan, Cannon, Sullivan and Hultman2009; Sullivan, Kendler, & Neale, Reference Sullivan, Kendler and Neale2003). Relatives of patients with SCZ face a significantly higher risk than individuals without a family history, possibly up to 11-fold (Le, Kaur, Meiser, & Mj, Reference Le, Kaur, Meiser and Mj2020) and manifest subtle changes in imaging traits and cognitive performance (da Motta, Pato, Barreto Carvalho, & Castilho, Reference da Motta, Pato, Barreto Carvalho and Castilho2021; Dodell-Feder, Delisi, & Hooker, Reference Dodell-Feder, Delisi and Hooker2014). The study of individuals with SCZ and those at genetic high risk (GHR) is pivotal for elucidating the genetic underpinnings of neurodevelopmental and cognitive impairment.
Since the late 19th century, pioneers have suggested that SCZ might stem from brain connectivity aberrations (Collin, Turk, & van den Heuvel, Reference Collin, Turk and van den Heuvel2016), a concept that evolved into the dysconnection hypothesis later proposed by Friston and Frith (Friston & Frith, Reference Friston and Frith1995). Advancements in neuroimaging have supported these theories, revealing anomalies in both white matter fibers and functional connectivity (FC) in patients with SCZ (Camchong, MacDonald, Bell, Mueller, & Lim, Reference Camchong, MacDonald, Bell, Mueller and Lim2011; Carreira Figueiredo, Borgan, Pasternak, Turkheimer, & Howes, Reference Carreira Figueiredo, Borgan, Pasternak, Turkheimer and Howes2022; W. Zhu, Wang, Yu, Zhang, & Zhang, Reference Zhu, Wang, Yu, Zhang and Zhang2023). The application of graph theory and network analysis to neuroimaging data further identified atypical changes in brain networks of patients with SCZ, such as altered clustering coefficients and efficiency levels (Y. Liu et al., Reference Liu, Liang, Zhou, He, Hao, Song and Jiang2008), deepening our understanding of the widespread connectivity disorder in SCZ. A crucial question for future research, highlighted by a meta-analysis, is whether disparities in functional networks represent susceptibility traits inherent in subjects genetically predisposed to SCZ (Kambeitz et al., Reference Kambeitz, Kambeitz-Ilankovic, Cabral, Dwyer, Calhoun, van den Heuvel and Malchow2016).
GHR individuals hold great value, as they can identify genetic liability across various phenotypes and reflect susceptibility. Approximately 62.5% of functional dysconnectivity is linked to genetic predisposition (Guo et al., Reference Guo, He, Liu, Linli, Tao and Palaniyappan2020; Yin et al., Reference Yin, Zhao, Li, Liu, Chen and Hong2021). Shared abnormalities were observed in both SCZ and GHR individuals, such as decreased FC, lower clustering coefficients, and higher global efficiency of networks (Lin et al., Reference Lin, Li, Dong, Wang, Sun, Shi and Lu2021; Lo et al., Reference Lo, Su, Huang, Hung, Chen, Lan and Bullmore2015). Noteworthily, despite the observed aberrant changes in brain function of GHR, they do not progress to SCZ but rather exhibit cognitive impairments (da Motta et al., Reference da Motta, Pato, Barreto Carvalho and Castilho2021).
Cognitive function relies on the complex neural coordination within brain networks (Shine et al., Reference Shine, Breakspear, Bell, Ehgoetz Martens, Shine, Koyejo and Poldrack2019, Reference Shine, Bissett, Bell, Koyejo, Balsters, Gorgolewski and Poldrack2016; van den Heuvel, Stam, Kahn, & Hulshoff Pol, Reference van den Heuvel, Stam, Kahn and Hulshoff Pol2009), and its impairments are associated with brain functional networks in patients with SCZ (Bassett et al., Reference Bassett, Bullmore, Meyer-Lindenberg, Apud, Weinberger and Coppola2009; He et al., Reference He, Sui, Yu, Turner, Ho, Sponheim and Calhoun2012; Menon, Palaniyappan, & Supekar, Reference Menon, Palaniyappan and Supekar2023). Executive function as a vitally cognitive domain is impaired in both SCZ and GHR individuals (da Motta et al., Reference da Motta, Pato, Barreto Carvalho and Castilho2021; Thuaire, Rondepierre, Vallet, Jalenques, & Izaute, Reference Thuaire, Rondepierre, Vallet, Jalenques and Izaute2022). Moreover, machine-learning analyses have further divulged that the more the functional brain patterns of GHR individuals approximated those of patients with SCZ, the lower their cognitive assessment scores were (Jing et al., Reference Jing, Li, Ding, Lin, Zhao, Shi and Fan2019; W. Liu et al., Reference Liu, Zhang, Qiao, Cai, Yin, Zheng and Wang2020). These studies indicate that SCZ genetic loadings may be crucial in influencing the dysconnection of SCZ brain networks and cognitive impairments.
SCZ, characterized by polygenic variations, can be initially identified via genome-wide association studies (GWASs) aimed at discerning millions of SCZ-associated single nucleotide polymorphisms (SNPs) dispersed throughout the genome (Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014). Subsequently, polygenic risk scores (PRSs) are utilized to calculate the additive genetic susceptibility of each individual, complemented by functional gene enrichment analysis (A. R. Martin, Daly, Robinson, Hyman, & Neale, Reference Martin, Daly, Robinson, Hyman and Neale2019). Our preceding study identified a correlation between SCZ-PRS and deficits in local efficiency based on structural hemispheric asymmetry in SCZ and GHR cohorts (Zhu et al., Reference Zhu, Wang, Gong, Edmiston, Zhong, Li and Tang2021). One study involving two cohorts found that SCZ-PRS is associated with a wide functional connectome in healthy controls (HCs) and a reduced connectome correlated with intelligence quotient (IQ) in SCZ (Cao, Zhou, & Cannon, Reference Cao, Zhou and Cannon2021). Additionally, the fractional amplitude of low-frequency fluctuations (fALFF) was found to mediate the association between SCZ-susceptible SNPs and working memory in a mixed SCZ and HC cohort (Luo et al., Reference Luo, Sui, Chen, Zhang, Tian, Lin and Jiang2018). However, current research into the relationship among SCZ-PRS, brain functional networks, and executive deficits is scarce. Whether and how SCZ-PRS influences cognitive impairments in SCZ and GHR individuals through the mediation of neural development in functional networks remains unclear.
To explore the relationships among SCZ-associated risk genes, function networks, and neurocognition in SCZ and GHR, this study proposes two hypotheses: first, that SCZ and GHR share altered functional networks, where these abnormal networks are associated with SCZ-related risk genes and contribute to executive deficits. Second, compared to GHR and HC, SCZ exhibits unique disease-specific alterations, characterized by broader network dysconnectivities. Based on these hypotheses, the study outlines four aims: identify common functional network alterations and executive impairments in SCZ and GHR, explore whether these shared changes are associated with risk genes, elucidate how risk genes have impacts and their biological role, and finally, identify disease-specific changes.
Methods
Participants
This study comprised a cohort of 292 participants (aged 18–55 years), comprising 100 SCZ patients, 68 GHR individuals, and 124 HCs.
The SCZ patients were recruited from two clinical centers: the First Affiliated Hospital of China Medical University and the Shenyang Mental Health Centre. The GHR participants were all first-degree relatives of patients presenting with SCZ in these two clinical centers. HCs were recruited from the local community via targeted advertisements. All participants underwent psychiatric evaluation using the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders-IV-Text Revision (DSM-IV-TR) Axis I Disorders (SCID-I). Assessments were conducted by two trained psychiatrist co-authors (S.T. and Z.Y.). SCZ participants met DSM-IV-TR diagnostic criteria. GHR individuals showed no personal history of Axis I disorders, and HCs exhibited neither personal nor familial history of Axis I disorders. Moreover, GHR underwent the Structured Interview for Prodromal Syndromes (SIPS) to confirm the absence of prodromal psychotic symptoms.
All participants were subjected to strict exclusion criteria, which encompassed the following: (1) substance abuse or dependence, including alcohol; (2) presence of any major medical condition; (3) neurological disorders; (4) history of head trauma resulting in loss of consciousness for ≥5 min; (5) contraindications for MRI; and (6) suboptimal quality of acquired MRI data.
Ethical approval for the study was obtained from the Medical Science Research Ethics Committee of the First Affiliated Hospital of China Medical University ([2012]25-1), and written informed consent was obtained from all participants.
Clinical and cognitive data
To assess symptom severity, three scales were applied: the Brief Psychiatric Rating Scale (BPRS) (Bech, Larsen, & Andersen, Reference Bech, Larsen and Andersen1988), the 17-item version of the Hamilton Rating Scale for Depression (HAMD-17) (Hamilton, Reference Hamilton1960), and the Hamilton Rating Scale for Anxiety (HAMA) (Hamilton, Reference Hamilton1959).
To evaluate executive cognition, the Wisconsin Card Sorting Test (WCST) was completed by 56 patients with SCZ, 57 individuals at GHR, and 120 HCs, including the scores of correct responses (CR), categories completed (CC), total errors (TE), perseverative errors (PE), and nonperseverative errors (NPE).
MRI data
Image acquisition
MRI scans were conducted at the Image Institute of the First Affiliated Hospital of China Medical University utilizing a GE Signa HD 3.0-T scanner (General Electric, Milwaukee, USA). Resting-state functional MRI (rs-fMRI) data were collected with a gradient-echo planar imaging sequence (TR = 2000 ms, TE = 30 ms, flip angle = 90°, slice thickness = 3 mm, number of slices = 35, no gap, FOV = 240 × 240 mm2, matrix = 64 × 64), including 200 volumes over 400 s. To reduce noise and motion, participants used earplugs and foam pads and were instructed to keep their eyes closed without falling asleep during the scan.
Data preprocessing
Rs-fMRI data preprocessing, conducted using SPM12 (www.fil.ion.ucl.ac.uk/spm/software/spm12/) and DPABI6.1 (Advanced edition; Yan, Wang, Zuo, & Zang, Reference Yan, Wang, Zuo and Zang2016) on MATLAB 2022a, involved steps such as converting DICOM to NIFTI, removing the first 10 time points, slice timing, realignment for excessive head motion (>3 mm or 3° were excluded), spatial normalization to MNI space (3 mm voxels), Gaussian smoothing (6 mm FWHM), linear detrending, removing nuisance covariates (Friston 24, white matter, cerebrospinal fluid, global signals), and low-frequency filtering (0.01–0.08 Hz).
Construction of brain functional network
To construct a brain functional network, the Automated Anatomical Labeling (AAL) template divided the brain into 90 regions of interest (ROIs), serving as network nodes. Edges were established by FC between ROIs. Blood oxygen level-dependent signal averages from voxels in each ROI provided time series data. A 90 × 90 matrix emerged from calculating Pearson correlation coefficients among ROI pairs, later transformed into Fisher’s Z-scores. A weighted approach then assigned varying weights to edges based on FC strength (Wen et al., Reference Wen, Liu, Rekik, Wang, Chen, Zhang and He2018; Yang et al., Reference Yang, Chen, Chen, Li, Li, Castellanos and Yan2021).
Brain functional network analysis
To ensure the small-world attributes of the network and align with prior studies, a wide range of network sparsity thresholds was set: 0.09–0.30 (step size 0.01) (Su, Hsu, Lin, & Lin, Reference Su, Hsu, Lin and Lin2015; Zhang et al., Reference Zhang, Wang, Wu, Kuang, Huang, He and Gong2011). Supplementary Table 1 provides a explanation of the network metrics, as established in previous studies (Wang, Zuo, & He, Reference Wang, Zuo and He2010; Wu, Li, Zhou, Zhang, & Long, Reference Wu, Li, Zhou, Zhang and Long2020). This study assessed global efficiency (Eglob), local efficiency (Eloc), and the shortest path length (Lp), alongside nodal metrics including nodal degree centrality (Dnodal), nodal global efficiency (Enodal), and nodal local efficiency (Enodal_loc). Areas under the curve (AUCs) were calculated for these metrics across all sparsity thresholds to provide a summary scalar. The analysis was facilitated by GRETNA, a dedicated network analysis toolbox operating on the MATLAB platform (Wang et al., Reference Wang, Wang, Xia, Liao, Evans and He2015).
Genetic data
Genotyping and imputation
Genome-wide genotype data were available for 78 participants (30 SCZ and 48 GHR). Blood samples were obtained between 10:00 and 15:00, utilizing ethylenediaminetetraacetic acid anticoagulant tubes, and subsequently stored at a temperature of −80 °C until subjected to assay. Genomic DNA extraction from the whole blood samples was conducted by standard methods. The Illumina Global Screening Array-24 v1.0 BeadChip (Illumina, San Diego, CA) was employed for genome-wide variant screening, rendering data of 642,824 predetermined gene variants, alongside 53,411 custom variants. Comprehensive criteria for data exclusion and genotype imputation can be found in the Supplementary Materials (Section 1, page 1).
Calculation of PRSs
PRSs were computed by multiplying the count of risk alleles by the effect size attributed to each allele, followed by the summation of the products across all SNPs for each individual (Martin et al., Reference Martin, Daly, Robinson, Hyman and Neale2019). In line with our previous study (Zhu et al., Reference Zhu, Wang, Gong, Edmiston, Zhong, Li and Tang2021), we used the 2018 GWAS results as the discovery sample (Bipolar Disorder and Schizophrenia Working Group of the Psychiatric Genomics Consortium. & Bipolar Disorder and Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2018). In this study, genetic factors associated with SCZ were analyzed in 33,426 individuals with SCZ and 32,541 controls from this dataset. Using our imputed genotyping data as the target sample, we performed p-value clumping in PRSice (www.PRSice.info) to retain strongly correlated SNPs, applying parameters of r 2 = 0.1 and a distance of 250 kb. PRSs were then calculated for each participant across eleven p-value thresholds (PTs): 0.0001, 0.001, 0.01, 0.015, 0.02, 0.025, 0.03, 0.035, 0.04, 0.045, and 0.05.
Functional enrichment analyses
Based on the dbSNP database, all SNPs within the SCZ-PRS under a certain p-value threshold most associated with network metrics were extracted and mapped to genes using position-based mapping, aligning SNPs to corresponding gene annotations by their rs names. Specifically, we utilized the latest dbSNP information available in the file All_20180423.vcf.gz, which can be downloaded from ftp://ftp.ncbi.nih.gov/snp/organisms/human_9606_b151_GRCh37p13/VCF/. This file is based on the GRCh37p13 genome build. Further details are provided on the NCBI dbVar resource page at https://www.ncbi.nlm.nih.gov/dbvar/content/org_summary/. The resultant gene lists were then uploaded into DAVID V6.8 (https://DAVID.ncifcrf.gov/) to conduct Gene Ontology analyses (Huang, Sherman, & Lempicki, Reference Huang, Sherman and Lempicki2009a; Reference Huang, Sherman and Lempicki2009b). Bonferroni correction was applied, ensuring a significance level of 0.05.
Statistical analysis
Differences in demographic, clinical, and cognitive variables
Differences across the three groups were assessed using chi-squared tests to compare sex and handedness variations and one-way analysis of variance (ANOVA) to investigate differences in age and total scores on clinical scales (BPRS, HAMD-17, and HAMA). To discern distinctions in WCST scores, a one-way analysis of covariance (ANCOVA) was used, with sex and age as covariates. In cases where post hoc comparisons were required, a pairwise analysis was performed. The least significant difference (LSD) correction was used for homogenous variance, and the Tamhane correction was used for other cases (significance at adjusted p < 0.05).
Differences in network metrics
ANCOVAs with sex and age as covariates were employed to assess differences in global network metrics (Eglob, Eloc, Lp) and the AUC for each nodal metric (Dnodal, Enodal, Enodal_loc) across the 90 nodes (significance at pFDR < 0.05).
Relationship between PRS, network metrics, and cognitive tests
In the SCZ and GHR groups, the partial correlation analysis (sex and age as covariates) was applied to investigate the relationship between SCZ-PRS, shared alterations in topological properties, and similar changes in WCST scores observed in both SCZ and GHR individuals (significance at pFDR < 0.05). Additionally, mediation analysis was executed to probe the potential impact of the common network metric alterations as mediators on the association between SCZ-PRS (causal variable) and cognitive function (outcome variable), with sex, age, and group as covariates. To conduct this analysis, the PROCESS macro within SPSS (Version 3.2, developed by Dr. Andrew F. Hayes) was utilized, and significance testing was performed using 5000 bias-corrected bootstrap samples. By summarizing the methods used by mediators, standard deviation (SD) and 95% confidence interval (CI) were utilized.
Significance was set at p < 0.05 (two-tailed) for all tests. Analyses not specifically delineated were performed using SPSS 26.0.
Results
Demographic, clinical, and cognitive characteristics
In the total sample with neuroimaging data (N = 292), no significant differences in terms of age and sex were observed among the SCZ, GHR, and HC groups. All participants were right-handed. Compared to the GHR and HC groups, the SCZ group exhibited significantly higher scores across the BPRS, HAMD-17, and HAMA total scores (see Table 1).
Demographic, clinical, and cognitive characteristics of the SCZ, GHR, and HC
Abbreviations: BPRS, Brief Psychiatric Rating Scale; HAMD-17, the 17-item version of the Hamilton Rating Scale for Depression; HAMA, Hamilton Rating Scale for Anxiety; WCST, Wisconsin Card Sorting Test; PRS, polygenic risk scores.
*p < 0.05.
In the subsample with both neuroimaging and WCST assessments (N = 233), group differences emerged in executive function measures. The WCST scores of CR and CC were the highest in the HC group, followed by the GHR and SCZ groups. The scores of TE, PE, and NPE were the highest in the SCZ group, followed by the GHR and HC groups (see Table 1, Figure 1a).
WCST and brain functional network characteristics.
(a) was the violin diagram showing the comparison of WCST scores among groups. (b) LP was increased in the SCZ group. (c) Eglob was decreased in the SCZ group. (d) and (f) were the violin diagram showing the comparison of nodal metrics among groups. (e) and (g) were the regions showing a significant difference in nodal metrics.
Abbreviations: WCST, Wisconsin Card Sorting Test; CR, correct responses; CC, categories completed; TE, total errors; PE, perseverative errors; NPE, nonperseverative errors; Lp, the shortest length path; Eglob, global efficiency; Enodal, nodal efficiency; Dnodal, nodal degree centrality; Enodal_loc, nodal local efficiency; R, right; L, left; MCPG, median cingulate and paracingulate gyri; MTG, middle temporal gyri; CAL, calcarine fissure and surrounding cortex; LING, lingual gyri.
A significant level of pFDR < 0.05.
Within the genetic subsample with PRS data (N = 78), age remained balanced across groups, while the SCZ group had a higher female proportion than GHR. This demographic variation was accounted for in subsequent association and mediation analyses through covariance adjustment.
Brain functional network characteristics
In comparison to GHR and HC groups, the SCZ group displayed increased Lp (p = 0.006, pFDR = 0.042) and reduced Eglob (p = 0.012, pFDR = 0.042), see Figure 1b,c and Supplementary Table 2.
Regarding Enodal, significant distinctions emerged in the right median cingulate and paracingulate gyri (MCPG_R), where post hoc analysis revealed a reduction in both SCZ and GHR groups compared to the HC group (p < 0.001, pFDR < 0.001). Moreover, in MCPG_R, both SCZ and GHR groups exhibited a decreased Dnodal (p = 0.001, pFDR = 0.045). Dnodal in the right middle temporal gyrus (MTG_R) exhibited variance, with the SCZ group surpassing the GHR and HC groups (p < 0.001, pFDR < 0.001), see Figure 1d–e. Enodal_loc levels were decreased in specific brain regions within the SCZ group, including the left calcarine fissure and surrounding cortex (CAL_L), right calcarine fissure and surrounding cortex (CAL_R), left lingual gyrus (LING_L), and right lingual gyrus (LING_R) (p < 0.001, pFDR < 0.001), see Figure 1f,g and Supplementary Table 2.
Correlation between PRS and network metrics
There were 78 individuals (30 SCZ and 48 GHR) with available genetic data (demographic characteristics of participants see Supplementary Table 3). Significant correlations were identified between SCZ-PRS and two nodal metrics – Enodal and Dnodal within MCPG_R. These metrics displayed concurrent alterations in both the SCZ and GHR groups. After FDR correction, the association with Dnodal in MCPG_R remained significant across six thresholds: PT_0.015 (r = −0.274, p = 0.016, pFDR = 0.035), PT_0.025 (r = −0.267, p = 0.020, pFDR = 0.037), PT_0.035 (r = −0.298, p = 0.009, pFDR = 0.049), PT_0.040 (r = −0.288, p = 0.012, pFDR = 0.033), PT_0.045 (r = −0.293, p = 0.010, pFDR = 0.037), and PT_0.050 (r = −0.266, p = 0.020, pFDR = 0.031). However, there was no significant correlation observed in Enodal after FDR correction (Figure 2 and Supplementary Table 4). There was no association between the PRS and SCZ disease-specific alterations (Supplementary Table 5).
Association of SCZ-PRS with nodal metrics in the SCZ and GHR group.
Abbreviations: PT, p-value threshold; Enodal, nodal efficiency; Dnodal, nodal degree centrality; MCPG, median cingulate and paracingulate gyri; R, right.
*a significant level of p < 0.05, bold type was pFDR < 0.05.
Correlation between network metrics and WCST
There were 113 individuals (56 SCZ and 57 GHR) competed WCST (demographic characteristics of participants see Supplementary Table 6). Significant correlations were established between the five subtest scores and the two nodal metrics - Enodal and Dnodal within MCPG_R (see Supplementary Table 7). After FDR correction, the Dnodal of MCPG_R was positively correlated with CR (r = 0.234, p = 0.013, pFDR = 0.033), CC (r = 0.220, p = 0.020, pFDR = 0.033), and negatively correlated with TE (r = −0.220, p = 0.020, pFDR = 0.025) (see Figure 3a–c and Supplementary Table 7). There was no association between the PRS and cognitive alterations (Supplementary Tables 8 and 9). There was no association between the WCST and SCZ disease-specific alterations (Supplementary Table 5).
Scatter plot and mediation model in SCZ and GHR groups.
(a) was scatter plots showing Dnodal in the right MCPG was positively correlated to the scores of WCST correct responses. (b) showed Dnodal was positively correlated to WCST categories completed. (d) showed Dnodal was negatively correlated WCST total errors. (d) was mediation model showing Dnodal in the right MCPG significantly mediated the association between SCZ-PRS and correct responses. (e) showed Dnodal significantly mediated the association between SCZ-PRS and categories completed. (f) showed Dnodal significantly mediated the association between SCZ-PRS and total errors. The dotted line represents a non-significant correlation.
Abbreviations: Dnodal, nodal degree centrality; MCPG, median cingulate and paracingulate gyri; R, right; WCST, Wisconsin Card Sorting Test.
A significant level of pFDR < 0.05.
Mediated moderation analysis
After mediation analysis, three mediation models were established (demographic characteristics of participants see Supplementary Table 8). Dnodal in MCPG_R fully mediated the association between the SCZ_PT_0.035 and CR (Path AB, Indirect effect = −14886.854; 95%CI: −29495.314 to −844.529, Figure 3d), between the PRS and CC (Path AB, indirect effect = −3150.880; 95%CI: −5891.580 to − 473.907, Figure 3e), and between the PRS and TE (Path AB, indirect effect = 14886.854; 95%CI: 481.107 to 29223.972, Figure 3f).
Functional enrichment analyses
To elucidate the biological underpinnings of SCZ-associated genes within SCZ-PRS, we conducted functional enrichment analyses for SCZ-PRS genes at a p-value threshold of 0.035 (PT_0.035), which has the smallest p-value and the most significant association with Dnodal in MCPG_R. A total of 30 Gene Ontology terms, mainly in biological processes, were identified for these SCZ-PRS genes. These enriched terms chiefly revolved around the regulation of signal transduction (particularly small GTPase-mediated signal transduction), Ca2+ transmembrane transport, and the modulation of synaptic transmission (Figure 4 and Supplementary Table 10).
Significant gene ontology enrichment analysis for risk genes of SCZ-PRSs.
Discussion
This study uniquely integrates genetics, brain functional networks (endophenotype), and executive function (clinical symptoms) in individuals with SCZ and those at GHR, providing valuable insights into the disease progression trajectory of polygenic hereditary disorders. The results showed how SCZ-associated risk genes influence altered functional networks and the association between executive deficits and genetically regulated alterations in functional networks. Specifically, we found that the common functional network alterations related to genetic susceptibility in both SCZ and GHR groups manifested as decreased Enodal and Dnodal in the MCPG_R. Additionally, the diminished Dnodal levels were associated with SCZ-PRS and executive deficits. Importantly, our study revealed that the effect of SCZ-PRS on the executive functions is completely mediated through altered Dnodal in MCPG_R. And the biological function of SCZ-PRS involves intracellular signal transduction, Ca2+ transmembrane transport, and modulation of synaptic transmission.
Brain network alterations in SCZ and GHR
We found common alterations in brain functional network metrics within the SCZ and GHR groups. Compared to the HC group, both SCZ and GHR groups exhibited decreased Enodal and Dnodal in MCPG_R, with no significant differences between them. The decreased Dnodal represented the reduced direct connections of the cingulate gyrus with other nodes, while the decreased Enodal indicated a declined capacity for information exchange in the cingulate gyrus. Futhermore, our findings revealed these alterations potentially represent genetic susceptibility to SCZ. The results partly align with Lo et al., who also found altered Enodal of the cingulate gyrus in both SCZ and GHR groups, specifically in the anterior and posterior cingulate cortex, and the MCPG_R (Lo et al., Reference Lo, Su, Huang, Hung, Chen, Lan and Bullmore2015).
Moreover, multiple studies have consistently reported compromised cingulate gyrus function in SCZ, manifesting as reduced Enodal_loc, diminished Dnodal and clustering coefficient, and lower activation during task-based fMRI (Lynall et al., Reference Lynall, Bassett, Kerwin, McKenna, Kitzbichler, Muller and Bullmore2010; Oertel et al., Reference Oertel, Kraft, Alves, Knöchel, Ghinea, Storchak and Stäblein2019; Yan et al., Reference Yan, Tian, Wang, Zhao, Yue, Yan and Zhang2015). The collective evidence highlights a reduced involvement of the cingulate gyrus in overall brain activity in patients with SCZ and individuals at GHR, indicating its role as a genetic susceptibility marker that could contribute to preventing disease development.
Unique brain network alterations in SCZ
We also found the SCZ group exhibited unique disease-specific alterations, compared to GHR and HC. Firstly, increased Dnodal in the MTG_R of the SCZ group suggests heightened interactions between the temporal gyrus and other brain regions and potentially a compensatory response, a known phenomenon in SCZ (Lynall et al., Reference Lynall, Bassett, Kerwin, McKenna, Kitzbichler, Muller and Bullmore2010). Additionally, the SCZ group showed reduced Enodal_loc in the bilateral calcarine and lingual gyri, indicating less efficient information transfer. The role of Enodal_loc in SCZ, however, requires further exploration. Moreover, we observed lower Eglob and higher Lp in SCZ, indicating the network global information processing was diminished, aligning with previous SCZ research (Ganella et al., Reference Ganella, Bartholomeusz, Seguin, Whittle, Bousman, Phassouliotis and Zalesky2017; Ho et al., Reference Ho, Tng, Wang, Chen, Subbaraju, Shukor and Medalia2020; Zhu et al., Reference Zhu, Wang, Liu, Qin, Li and Zhuo2016). The SCZ group exhibits broader abnormalities in functional networks compared to the GHR group, manifesting as a gene-susceptibility reduction in specific brain region FC and a disease-specific decline in whole brain network efficiency.
Associations between PRS, network metrics, and executive deficits
We observed a negative correlation between decreased Dnodal in MCPG_R and SCZ-PRS scores in the SCZ and GHR groups. This implies that higher PRSs are associated with reduced connections between the cingulate gyrus and other brain regions. Concurrently, analyses utilizing UK Biobank data revealed significant associations of SCZ-PRS with fractional anisotropy, mean diffusivity, and neurite density index of cingulate gyrus, suggesting that genetic effects on multiple MRI phenotypes are located in the cingulate (Stauffer et al., Reference Stauffer, Bethlehem, Warrier, Murray, Romero-Garcia, Seidlitz and Bullmore2021). Further supporting this, recent Mendelian randomization analysis posits that genetic variations in the cingulate gyrus may be causal for SCZ (Stauffer et al., Reference Stauffer, Bethlehem, Dorfschmidt, Won, Warrier and Bullmore2023). In addition, this is complementary to our early finding in brain structural networks that PRSs are associated with Eloc deficits in SCZ and GHR populations (Zhu et al., Reference Zhu, Wang, Gong, Edmiston, Zhong, Li and Tang2021). In summary, our PRS findings suggest that aberrant brain functional networks may reflect the overall additive genetic vulnerability of SCZ.
This study demonstrated that, in both SCZ and GHR groups, Dnodal in MCPG_R was positively correlated with both CR and CC scores, while negatively correlated with TE scores. This suggests that altered brain functional networks may impact cognitive performance, aligning with previous research. Specifically, task-based fMRI investigations have revealed that prolonged task completion times in patients with SCZ are linked to reductions in the clustering coefficient and Eloc (He et al., Reference He, Sui, Yu, Turner, Ho, Sponheim and Calhoun2012). Additionally, Bassett et al. identified a correlation between impaired working memory and decreased Eglob in SCZ (Bassett et al., Reference Bassett, Bullmore, Meyer-Lindenberg, Apud, Weinberger and Coppola2009). Furthermore, machine-learning analyses revealed that the closer the functional brain patterns of GHR approximated those of SCZ, the poorer their executive function performance (Liu et al., Reference Liu, Zhang, Qiao, Cai, Yin, Zheng and Wang2020). Our findings provide additional evidence, indicating a decline in genetic susceptibility-related connectivity between the cingulate gyrus and other brain regions, correlating with reduced executive function in SCZ and GHR.
Altered functional networks mediating the association between SCZ-PRS and executive deficits
Our findings indicated that Dnodal in MCPG_R fully mediated the association between SCZ-PRS and executive function in both SCZ and GHR groups, suggesting that gene-regulated functional network disruption may serve as an early biomarker for executive function impairments. Our findings were similar to a study using independent component analysis, which found that fALFF mediated the relationship between SCZ-susceptible SNPs and working memory in mixed SCZ and HC cohorts, considering diagnosis as a covariate (Luo et al., Reference Luo, Sui, Chen, Zhang, Tian, Lin and Jiang2018). Another study found an association between SCZ-PRS and the functional connectome in HCs, with parallel findings of reduced connectomes and associated IQ deficits in an independent SCZ cohort (Cao et al., Reference Cao, Zhou and Cannon2021). However, the two teams failed to encompass continuous pathophysiology in one patient cohort and ignored cognitive impairment in GHR. Our study revealed that brain dysfunction mediated the association between genetic factors and cognitive deficits in both individuals with SCZ and those at GHR.
Furthermore, our functional enrichment analysis revealed that risk genes are implicated in processes like signal transduction, Ca2+ transmembrane transport, and synaptic transmission. Signal transduction, particularly small GTPase-mediated signaling, acts as a messenger for information carriage and contributes to axon guidance (Nikolic, Reference Nikolic2002). Ca2+ transmembrane transport plays a pivotal role in regulating neurotransmitter release and synaptic strength through Ca2+ levels (Neher & Sakaba, Reference Neher and Sakaba2008). Synaptic transmission regulation directly impacts information transfer between neurons (Martin, Grimwood, & Morris, Reference Martin, Grimwood and Morris2000). Overall, these risk genes significantly affect synaptic plasticity and transmission, impacting neurodevelopment and information exchange. This leads to disrupted connectivity between the cingulate gyrus and other brain regions, manifesting as noticeable declines in cognitive functions.
Limitations
This study has some limitations. First, although we observed no significant differences in cognitive performance or network metrics between medicated and unmedicated SCZ patients – likely due to the small size of the unmedicated subgroup – we cannot fully exclude antipsychotic effects on their relationships; larger, drug-naive cohorts are needed. Similarly, no differences emerged between first-episode and multiepisode – perhaps because most SCZ participants were first-episode with limited medication exposure – and no differences between in- and outpatients, suggesting these factors had minimal impact on our results. Second, its cross-sectional design limits the ability to observe longitudinal brain alterations and cognitive changes in GHR individuals, particularly whether GHR progresses to SCZ. Third, the modest SCZ sample size limits our ability to explore disease-specific brain regions fully. Similarly, WCST and PRS analyses were restricted to subgroups, although post hoc power analyses confirmed sufficient statistical power. Moreover, future studies could examine gene–environment interactions to clarify how environmental factors shape SCZ risk-gene effects on brain networks and cognition.
Conclusion
The decreased connections of the right median cingulate-paracingulate gyri with other regions were observed in both SCZ and GHR groups, potentially indicating genetic susceptibility. Additionally, these reduced connections were linked to SCZ-related risk genes and WCST scores. Crucially, the reduced involvement of the cingulate gyrus in overall brain activity mediated the effect of SCZ-related risk genes on executive deficits in SCZ and GHR groups, and these risk genes were involved in signal transduction, Ca2+ transmembrane transport, and synaptic transmission. Significantly, the SCZ group displayed broader functional network abnormalities, characterized by reduced gene susceptibility in specific regions and a disease-specific decline in whole network efficiency. Our findings provide new insights into the genetic link to neurodevelopmental mechanisms and cognitive impairment, highlight the role of the cingulate gyrus, and contribute to a deeper understanding of the genetic and neuropathological basis of SCZ.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725101177.
Data availability statement
The data supporting the study’s findings can be obtained from the corresponding author (Yanqing Tang) upon reasonable request.
Acknowledgments
We express gratitude to all participants and investigators in the study for their contributions.
Funding statement
This work was supported by the Applied Fundamental Research Program of Liaoning Province (Grant #2023JH2/101300031 to Yanqing Tang), Shenyang Science and Technology Planning Project (Grant #22–321–32-06 to Yanqing Tang), Science and Technology Innovation STI2030-Major Projects (Grant #2021ZD0200700 and #2021ZD0200600 to Yanqing Tang), Basic Scientific Research Projects of Universities of Liaoning Province (Grant #LJKMZ20221214 to Yifang Zhou), the Natural Science Foundation of Liaoning Province (Grant #2022-YGJC-40 to Lingtao Kong), and the National Natural Science Foundation of China (grant #82201689 to Xiaowei Jiang).
Competing interests
The authors declare none.
Open access
