Introduction

Numerous genome-wide association studies have reported the α-1C subunit of the L-type voltage-gated calcium channel gene (CACNA1C, 12p13.3, Chr12 : 1948979-2677376 (b36), between CACNA2D4 and ITFG2-AS1] as one of the top risk genes for schizophrenia. Many risk variants at CACNA1C have been associated with schizophrenia, some of which were replicated across multiple independent studies (O’Donovan et al., 2008; Stefansson et al., 2009; Bigos et al., 2010; Green et al., 2010; Nyegaard et al., 2010; Schizophrenia Psychiatric Genome-Wide Association Study Consortium, 2011; Ripke et al., 2013; Guan et al., 2014; He et al., 2014; Ivorra et al., 2014; Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014; Zheng et al., 2014); however, this gene remains to be comprehensively investigated; in particular, the functional roles of most risk variants have yet to be thoroughly explored.

To build on GWAS findings, it is necessary to continue focusing on this risk gene. Recently, we examined 847 single nucleotide polymorphisms (SNPs) at CACNA1C and identified a 20-SNP block at intron 3 that was consistently associated with schizophrenia across seven independent cohorts (2.5 × 10−17 ≤ P ≤ 0.049). This block was also significantly associated with CACNA1C mRNA expression in the brain across three independent European cohorts (5.1 × 10−12 ≤ P ≤ 8.3 × 10−3), and could be tagged by the most significant risk SNP, rs1006737 (meta-analysis of 17 studies: P = 1.62 × 10−42) (Wang et al., 2022). Moreover, rs1006737 was found to be significantly negatively associated with the gray matter volume (GMV) of the thalamus (P = 0.010), the surface area of isthmus cingulate cortex (P = 0.013), and the thickness of transverse temporal and superior temporal sulcus cortexes (0.005 ≤ P ≤ 0.043) (Wang et al., 2022). These pieces of evidence supported the most promising role of this variant block in schizophrenia risk.

However, CACNA1C is a large gene that is 730 kb long with 58 exons and might harbor other missing risk variant blocks for schizophrenia. Therefore, in the present study, we aimed to identify those missing risk blocks independent of the previously reported one, and explore their potential, and maybe distinct, regulatory effects on CACNA1C mRNA expression in the brain, subcortical GMVs, as well as cortical surface area and thickness, in order to complete the map of all risk variant blocks for schizophrenia across entire CACNA1C.

Usually, a GWAS examines variants across the genome unbiasedly and thus is subject to an overly conservative Bonferronic correction (α = 5 × 10−8) to control for false positives, which may lead to the loss of some important information, for example, missing critical findings with 5 × 10−8 < P < 10−5, because, for most common diseases such as schizophrenia that involves many common variants with small effect sizes, most risk effects of individual variants are difficult to reach this significance level. The present study, as a post-GWAS follow-up, aimed to preserve the information as loosely as possible (α = 0.05), but control for false positives as strictly as possible using an alternative strategy, that is, a series of replication and functional validation steps (detailed below). We examined replicable SNP-schizophrenia associations across multiple cohorts to ensure their robustness, and then demonstrated them to be biological, but not only statistical, associations by examining a series of potential regulatory effects on CACNA1C mRNA expression and brain structure.

Materials and methods Subjects

Four independent samples, comprising three European and one African-American population, were used for SNP-schizophrenia association analysis. Sample #1 was sourced from the GAIN dataset (dbGaP access number: phs000021.v3.p2), consisting of 1350 European patients with schizophrenia and 1378 healthy European controls. Sample #2 was obtained from the Ashkenazi Jewish dataset (phs000448.v1.p1), comprising 1044 Jewish patients with schizophrenia and 2052 healthy Jewish controls. Sample #3 was sourced from the Bulgarian Schizophrenia Trio Sequencing Study (phs000687.v1.p1), consisting of a total of 1826 European parent-offspring trio subjects, of which 621 were offspring with schizophrenia. Sample #4 was sourced from the GAIN dataset (phs000021.v3.p2), comprising 1195 African-American patients with schizophrenia and 954 healthy African-American controls. Samples #1 and #2 were also examined in the previous study (Wang et al., 2022).

All subjects included in the study were 18 years or older. Affected subjects met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria for schizophrenia (American Psychiatric Association, 1994). Subjects with neurological disorders, substance use disorders, or mental retardation were excluded. Controls were free of schizophrenia, schizoaffective disorder, bipolar disorder, major depressive disorder, and psychotic symptoms such as auditory hallucination and persecutory delusion. Informed written consent was obtained from all subjects before participating in the study, which was reviewed and approved by the Human Investigation Committee of Yale University. Detailed demographic data for these samples have been published in previous studies (Manolio et al., 2007; Sanders et al., 2008; Lencz et al., 2013; Fromer et al., 2014).

Single nucleotide polymorphism-disease association analysis

Genotyping, imputation [using IMPUTE2 (Howie et al., 2009)], and data cleaning have been described in detail earlier (Wang et al., 2022). The allele frequencies of SNPs were compared between individuals with schizophrenia and controls using the Fisher exact test as implemented in the program PLINK (Purcell et al., 2007), to examine the SNP-disease associations.

Although the Bonferroni correction is a commonly used method to correct for multiple comparisons, it may miss some weak but valuable SNP-disease associations. To avoid overly conservative correction and preserve the information of modest SNP-disease associations, we used a replication and functional validation design. This approach allowed us to control for false positives by requiring the associations to be replicable in the same effect direction across at least two independent cohorts and the risk SNPs to be functionally validated. Only then, we considered them as statistically significant associations if the P values for associations were close to 0.05.

cis-acting expression quantitative trait locus analysis

To investigate the potential regulatory effects of schizophrenia-risk variants on the CACNA1C mRNA expression in human postmortem brains, we conducted cis-acting expression quantitative trait locus (cis-eQTL) analysis on two independent cohorts: a UK European cohort (n = 138) (BRAINEAC dataset) (Ramasamy et al., 2014) and a European-American cohort (n = 210) (GTEx dataset) (GTEx Consortium, 2013). All subjects were free of significant neurodegenerative and neuropsychiatric disorders.

In the UK European cohort, we analyzed a total of 10 brain regions, including the cerebellar cortex, prefrontal cortex, hippocampus, medulla, occipital cortex, putamen, substantia nigra, temporal cortex, thalamus, and intralobular white matter. In the European-American cohort, we analyzed a total of 11 brain regions, including the basal ganglia (putamen, caudate, nucleus accumbens, and substantia nigra), limbic system [anterior cingulate cortex (BA24), amygdala, hippocampus, and hypothalamus], prefrontal cortex (Brodmann Area or BA9), cerebellum, and cerebellar hemisphere. To compare the normalized mRNA expression levels between different alleles of each variant, we used t-tests.

Regulatory effect of risk variants on the gray matter volumes of subcortical structures

The potential regulatory effects of schizophrenia-risk variants on the GMVs of several brain regions were analyzed in a European sample (n = 34 431; ENIGMA2 dataset) (Satizabal et al., 2019) using multiple linear regression analysis. Specifically, we focused on the nucleus accumbens, amygdala, brainstem, caudate, and thalamus. All subjects were free of neurodegenerative and neuropsychiatric disorders.

Regulatory effect of risk variants on cortical surface area and thickness

A total of 36 936 subjects, including 33 992 European (23 909 from 49 ENIGMA cohorts and 10 083 from the UK Biobank) and 2944 non-European participants (eight cohorts) (Grasby et al., 2020), were analyzed. Measures of cortical surface area and thickness were derived from in-vivo whole-brain T1-weighted MRI scans using FreeSurfer (Fischl, 2012). The potential regulatory effects of schizophrenia-risk variants on 70 traits (total surface area, average thickness, and the surface area and thickness of 34 cortical regions averaged across the right and left hemispheres) were analyzed using multiple linear regression. The analysis adjusted for the effects of sex, linear and nonlinear age effects, interactions between age and sex, ancestry [the first four Multidimensional Scaling (MDS) components], diagnostic status (when the cohort followed a case-control design), MRI acquisition orientation, scanner (when multiple scanners were used at the same site), and global measure (total surface area or average thickness).

Results A variant block across introns 36–45 of CACNA1C was significantly associated with schizophrenia across multiple samples

Independent of the previously reported risk variant block at intron 3 mentioned earlier (Wang et al., 2022), we identified a novel 17-variant block (within block: D’>0.85) (Wang et al., 2022) spanning introns 36–45 of CACNA1C (Supplementary Figure S1, Supplemental digital content 1, https://links.lww.com/PG/A305). This new variant block was significantly associated with schizophrenia in at least two of the five independent samples we studied (1.8 × 10−4 ≤ P ≤ 0.049; Table 1). The risk allele of each variant had a significantly higher frequency in the cases (or ‘transmitted’ group) than in the controls (or ‘untransmitted’ group), and this was consistent across the samples. When further defining a block with R2 > 0.85, this 17-variant block could be classified into three sub-blocks, that is, classes #1, #2, and #3 harboring one, 12, and four SNPs, respectively (Supplementary Figure S1, Supplemental digital content 1, https://links.lww.com/PG/A305 Table 1). On the basis of the three identified classes of alleles, we were able to distinguish between different functional patterns, as described below.

Table 1 - P values for single nucleotide polymorphism-schizophrenia and single nucleotide polymorphism-mRNA expression associations Class SNP Genomic position (b36) ‘CACNA1C SNP – Schizophrenia’ associations. ‘SNP - CACNA1C mRNA’ associations Effect direction of risk or effect alleles Europeans Africans Europeans Europeans Risk allele Sample #1 #2 #3 #4 BRAINEAC GTEx Case=1351 1044 621 (offspring) Risk allele 1195 Effect CRBL FCTX OCTX TCTX Effect AMYL ACC Control=1378 2052 1205 (parents) 954 allele 138 138 138 138 allele 210 210 #1 rs2238096 2620174 a – >0.05 0.021 A 0.047 a >0.05 1.6 × 10−3 0.022 0.050 – – – ↑↑– #2 rs10848680 2635439 G 0.005 – 0.020 – G 0.020 >0.05 >0.05 >0.05 A 0.050 >0.05 ↑↑↓ #2 rs216047 2636121 T 1.8 × 10−4 – 0.013 – T 0.024 >0.05 >0.05 >0.05 C 0.035 >0.05 ↑↑↓ #2 rs1894979 2637260 G 2.3 × 10−4 – 0.004 – G 0.023 >0.05 >0.05 >0.05 A 0.050 >0.05 ↑↑↓ #2 rs3794291 2639926 C 0.048 – 0.007 – C 0.026 >0.05 >0.05 >0.05 T 0.035 >0.05 ↑↑↓ #2 rs10161181 2640991 A – >0.05 0.001 a 0.011 A 0.025 >0.05 >0.05 >0.05 – – – ↑↑– #2 rs10161275 2641279 T 0.046 >0.05 0.009 – T 0.026 >0.05 >0.05 >0.05 – – – ↑↑– #2 rs11062298 2641970 A 0.039 – 0.001 a 0.015 A 0.029 >0.05 >0.05 >0.05 C 0.050 >0.05 ↑↑↓ #2 rs11062299 2641999 T >0.05 0.011 0.002 >0.05 T 0.029 >0.05 >0.05 >0.05 C 0.050 >0.05 ↑↑↓ #2 rs10848682 2642017 C >0.05 0.011 0.002 – C 0.029 >0.05 >0.05 >0.05 G 0.050 >0.05 ↑↑↓ #2 rs10774052 2642432 C 0.039 – 0.002 >0.05 C 0.036 >0.05 >0.05 >0.05 – – – ↑↑– #2 rs7315556 2644116 G >0.05 0.009 0.002 – G 0.037 >0.05 >0.05 >0.05 A 0.037 >0.05 ↑↑↓ #2 rs7310079 2645738 T >0.05 0.027 0.044 – T 0.034 >0.05 >0.05 >0.05 – – – ↑↑– #3 rs7306298 2647572 A – 0.030 0.011 – – – – – – A >0.05 0.042 ↑–↑ #3 rs10744567 2651077 G >0.05 0.034 0.015 – – – – – – G >0.05 0.020 ↑–↑ #3 rs7136355 2653737 C – 0.019 0.004 – – – – – – C >0.05 0.020 ↑–↑ #3 rs2302729 2654233 C – 0.028 0.020 >0.05 – – – – – C >0.05 0.012 ↑–↑

ACC, anterior cingulated; cerebellar cortex (CRBL), frontal cortex (FCTX), occipital cortex (OCTX), SNP, single nucleotide polymorphism; temporal cortex (TCTX), amygdale (AMYL). The italic bold lowercase risk allele is a major allele (f > 0.5). #1, #2 and #3 are classified based on R2 > 0.85 within each class. ↑ and ↓, increase or decrease risk for disease, or mRNA expression level, referencing to the disease-risk alleles.

The schizophrenia-risk alleles were significantly associated with the CACNA1C mRNA expression in the brain

All 17 risk variants within the variant block showed significant associations with CACNA1C mRNA expression in different brain regions, including the cerebellar, frontal [including anterior cingulate (ACC)], occipital, temporal cortices, and amygdala (1.6 × 10−3 ≤ P ≤ 0.050). Specifically, the schizophrenia-risk allele of class #1 (including rs2238096) showed significant positive associations with CACNA1C mRNA expression in the frontal, occipital, and temporal cortices (1.6 × 10−3 ≤ P ≤ 0.050); the schizophrenia-risk alleles of class #2 showed a significant positive association with CACNA1C mRNA expression in the cerebellar cortex, but a negative association with expression in the amygdala (0.020 ≤ P ≤ 0.050); and the schizophrenia-risk alleles of class #3 showed a significant positive association with CACNA1C mRNA expression in ACC (0.012 ≤ P ≤ 0.042) (Table 1).

The schizophrenia-risk alleles were significantly associated with subcortical gray matter volumes

Seven of the 17 risk variants within the variant block (41%) showed significant association with the GMVs of subcortical structures, including the nucleus accumbens, brainstem, and caudate (0.016 ≤ P ≤ 0.048). Specifically, the schizophrenia-risk allele of class #1 was positively associated with the GMVs of the caudate and brainstem (0.017 ≤ P ≤ 0.040); the schizophrenia-risk alleles of class #2 were negatively associated with the GMVs of the nucleus accumbens (0.034 ≤ P ≤ 0.048); and the schizophrenia-risk alleles of class #3 were positively associated with the GMVs of the nucleus accumbens and brainstem (0.016 ≤ P ≤ 0.028) (Table 2).

Table 2 - P values for single nucleotide polymorphism-gray matter volumes associations in Europeans SNP Schizophrenia
-risk allele ‘SNP – GMV’ associations Effect direction risk or effect alleles Effect
allele Accumbens Brainstem Caudate 28 697 28 809 30 153 rs2238096 a a >0.05 0.017 0.040 ↑↑ rs11062298 A C 0.048 >0.05 >0.05 ↑↓ rs11062299 T C 0.045 >0.05 >0.05 ↑↓ rs10848682 C G 0.042 >0.05 >0.05 ↑↓ rs10774052 C T 0.034 >0.05 >0.05 ↑↓ rs7315556 G A 0.034 >0.05 >0.05 ↑↓ rs2302729 C C 0.028 0.018 >0.05 ↑↑

The italic lowercase risk allele is a major allele (f > 0.5). ↑ and ↓, increase or decrease risk for disease or GMV, respectively, referencing to the disease-risk alleles.

GMV, gray matter volume; SNP, single nucleotide polymorphism.

The schizophrenia-risk alleles were significantly associated with the cortical surface area

All seventeen risk variants were significantly associated with the cortical surface area of various brain regions, including frontal (pole, superior, and rostral middle), parietal (precuneus), temporal (pole), occipital (lateral, pericalcarine, lingual, and fusiform), and limbic [anterior (ACC) and posterior cingulate (PCC)] cortices (0.010 ≤ P ≤ 0.050). Specifically, the schizophrenia-risk allele of class #1 was positively (β > 0) associated with the surface area of ACC (P = 0.043), but negatively (β < 0) with the surface area of parietal (precuneus) and temporal (pole) cortices, and PCC (0.015 ≤ P ≤ 0.037); the schizophrenia-risk alleles of class #2 were positively (β > 0) associated with the surface area of parietal (precuneus) cortex and PCC (0.010 ≤ P ≤ 0.050), but negatively (β < 0) with the surface area of occipital (pericalcarine and fusiform) cortices (0.020 ≤ P ≤ 0.050); and the schizophrenia-risk alleles of class #3 were positively (β > 0) associated with the surface area of occipital (lateral, pericalcarine, and lingual) cortices (0.017 ≤ P ≤ 0.047), but negatively (β < 0) with the surface area of frontal (pole, superior, and rostral middle) cortices (0.025 ≤ P ≤ 0.048) (Table 3).

Table 3 - P values for single nucleotide polymorphism-cortical surface area associations SNP Disease-
risk allele ‘SNP-cortical surface area’ associations Effect
direction
of risk
or effect
alleles Effect
allele European Mixed Mixed European Mixed Mixed Effect
allele Mixed Mixed European Mixed Mixed Frontal Parietal Temporal Limbic Occipital Limbic Lateral Peri-calcarine Lingual Fusiform Anterior cingulate Pole Superior Rostral middle Precuneus Pole Posterior cingulate n = 32 742 35 198 35 198 32 742 35 198 35 198 35 198 35 198 32 742 35 198 35 198 rs2238096 a C >0.05 >0.05 >0.05 0.032 0.015 0.037 a >0.05 >0.05 >0.05 >0.05 0.043 ↑↓↑ rs10848680 G G >0.05 >0.05 >0.05 0.018 >0.05 0.028