Comparing GWAS and Brain Structure-Specific Gene Expression Profiles Identifies Psychiatric Disorder-Related Brain Structures at Different Developmental Stages

Dear Editor,

Psychiatric disorders are a group of mental disorders characterized by psychological or behavioral disabilities. Losing of the ability to work and assuming the extensive cost of long-term treatment, patients with psychiatric disorders are forced to support a heavy financial and medical burden. The growth and development of brain structures are dynamic, and different brain structures have major and specific functions to control behavior and performance. It has been demonstrated that the dysfunction of different brain structures involved in different psychiatric disorders is due to widespread alterations in the functional connectivity in the brain. Because of the dynamic alterations during brain growth and development, it is reasonable to infer that different psychiatric disorders are involved in the dysfunction of different brain structures at different life stages.

Genetic factors display a strong component in the etiology of psychiatric disorders. Gene expression is regulated by heredity, and the integration of genome-wide association studies (GWASs) and gene expression profile datasets has become a hot topic in recent studies, providing new insights into the mechanism underlying complex diseases. Based on these, it is feasible to further investigate the mechanisms of complex diseases through integrating GWASs and expression profile datasets. However, there are few studies comparing GWASs and gene expression profile datasets with the consideration of different brain structures and developmental stages for psychiatric disorders.

In this study, we explored the abnormal brain structures associated with different psychiatric disorders at different life stages, through a comparative analysis of brain structure-specific and age-specific gene expression profiles and GWAS datasets for psychiatric disorders (Supplementary materials). The mRNA expression profiles of 16 brain structures at 13 developmental stages were derived from BrainSpan (Atlas of the Developing Human Brain, http://www.brainspan.org/). To ensure the sample sizes of gene expression profile analysis, the brain transcriptomes were divided into five age groups: < 37 post-conceptional weeks (pcw), 0–3 years, 4–13 years, 14–25 years, and > 25 years. Each age group had at least three samples for different brain structures analyzed in this study. For each brain structure, the limma package was used to compare the gene expression profiles of each age group with that of all remaining samples in the same brain structure. The brain-specific differentially-expressed gene sets can be viewed as the gene expression biomarkers of the corresponding brain structure at a certain stage. Specifically in this study, the top 10% of identified differentially-expressed genes were used to generate representative gene sets specific to a certain brain structure at a certain development stage. The latest GWAS datasets of five common psychiatric disorders were downloaded from the Psychiatric Genomics Consortium website (http://www.med.unc.edu/pgc/). Briefly, these GWAS datasets included 7,387 cases and 8,567 controls for autism spectrum disorder (ASD), 20,183 cases and 35,191 controls for attention-deficit hyperactivity disorder (ADHD), 20,129 cases and 54,065 controls for bipolar disorder (BD), 33,426 cases and 54,065 controls for schizophrenia (SCZ), and 135,458 cases and 344,901 controls for major depressive disorder (MDD). The significant single-nucleotide polymorphisms (SNPs) identified by the GWASs of psychiatric disorders were mapped to genes according to their physical distances to nearby genes and methylation quantitative trait loci (meQTLs) annotation information [1, 2]. SNPs were mapped into nearby genes, which means they have an effect on the genes due to the short distance. SNPs were aligned with the SNP-target gene annotation data of meQTLs, which means that SNPs influence the genes due to the regulation of DNA methylation status at CpG sites. With the brain structure-specific gene sets at different stages (identified by gene expression profiles) and genes associated with psychiatric disorders (identified by GWASs), the gene set enrichment analysis (GSEA) approach was implemented to evaluate the functional relevance of 16 brain structures to psychiatric disorders at a certain development stage [3]. Significant enrichment was detected at a false discovery rate (FDR) < 0.05.

For annotating GWAS SNPs to nearby genes, the GSEA results of the five psychiatric disorders are summarized in Fig. 1. We identified 16, 14, 9, 15, and 7 enrichment signals related to brain structure for ASD, ADHD, BD, SCZ, and MDD, respectively, such as primary auditory cortex before 37 pcw (FDR_nearby = 6.30×10⁻³ for ASD, FDR_nearby = 5.00×10⁻⁴ for ADHD).

Fig. 1.

Heat map of gene set enrichment analysis for the five psychiatric disorders annotating GWAS SNPs to nearby genes. ADHD attention-deficit hyperactivity disorder, ASD autism spectrum disorder, MDD major depressive disorder, BD bipolar disorder, SCZ schizophrenia, AMY amygdaloid complex, MFC anterior (rostral) cingulate (medial prefrontal) cortex, CBC cerebellar cortex, DFC dorsolateral prefrontal cortex, HIP hippocampus, ITC inferolateral temporal cortex, MD mediodorsal nucleus of thalamus, OFC orbital frontal cortex, STC posterior superior temporal cortex, IPC posteroventral parietal cortex, A1C primary auditory cortex, M1C primary motor cortex, S1C primary somatosensory cortex, V1C primary visual cortex, STR striatum, VFC ventrolateral prefrontal cortex.

For annotating GWAS SNPs to meQTL-related target genes, GSEA results of the five psychiatric disorders are summarized in Fig. 2. We found 15, 12, 22, 27, and 10 enrichment signals related to brain structure for ASD, ADHD, BD, SCZ and MDD, respectively, such as posteroventral parietal cortex before 37 pcw (FDR_meQTLs = 8.32×10⁻³ for BD, FDR_meQTLs = 5.83×10⁻⁴ for SCZ, FDR_meQTLs = 1.97×10⁻² for MDD).

Fig. 2.

Heat map of gene set enrichment analysis for the five psychiatric disorders annotating GWAS SNPs to the meQTL-related target genes (abbreviations as in Fig. 1).

Besides, several overlapped brain structure-related enrichment signals shared by annotating GWAS SNPs to nearby genes and meQTL-related target genes were found. Specific to ASD, ADHD, BD, SCZ, and MDD, 11, 10, 7, 14, and 6 overlapped brain structure-related enrichment signals were discovered, respectively, such as amygdaloid complex at 14–25 years (FDR_nearby = 1.33×10⁻³, FDR_meQTLs = 7.27×10⁻³ for ASD), cerebellar cortex before 37 pcw (FDR_nearby < 0.001, FDR_meQTLs = 2.00×10⁻³ for SCZ), and orbital frontal cortex after 25 years (FDR_nearby = 3.94×10⁻², FDR_meQTLs = 2.50×10⁻⁴ for BD) (Table 1).

Table 1.

List of the common brain structure-related enrichment signals for the five psychiatric disorders by annotating GWAS SNPs to nearby genes and meQTL-related target genes.

Brain Structures	Stages	ASD		ADHD		BD		SCZ		MDD
		FDRnearby	FDRmeQTLs	FDRnearby	FDRmeQTLs	FDRnearby	FDRmeQTLs	FDRnearby	FDRmeQTLs	FDRnearby	FDRmeQTLs
CBC	<  37 pcw	< 0.001	< 0.001	1.75×10−3	5.00×10−4	2.75×10−2	< 0.001	&< 0.001	2.00×10−3	2.25×10−3	1.25×10−3
MD	14–25 years	< 0.001	3.31×10−2			3.10×10−2	5.75×10−3	1.89×10−2	6.71×10−3
AMY	14–25 years	1.33×10−3	7.27×10−3	1.60×10−2	1.25×10−3	4.00×10−2	3.31×10−3	2.08×10−2	8.08×10−3
IPC	< 37 pcw	5.58×10−3	1.50×10−3	5.00×10−4	2.00×10−3	4.32×10−2	8.32×10−3	6.00×10−4	5.83×10−4	1.67×10−2	1.97×10−2
V1C	< 37 pcw	6.07×10−3	3.43×10−3							8.88×10−3	6.17×10−3
DFC	< 37 pcw	6.19×10−3	7.80×10−3	< 0.001	1.40×10−3			6.25×10−4	1.00×10−3	4.00×10−3	< 0.001
A1C	< 37 pcw	6.30×10−3	1.67×10−3	5.00×10−4	1.63×10−3			1.06×10−2	7.86×10−4	1.17×10−2	6.67×10−4
STC	< 37 pcw	7.25×10−3	3.00×10−3	9.00×10−4	2.69×10−3			1.38×10−3	6.67×10−4
S1C	< 37 pcw	2.48×10−2	4.00×10−3	1.67×10−3	2.99×10−2			5.00×10−4	6.25×10−4
ITC	< 37 pcw	2.82×10−2	4.80×10−3	1.50×10−3	1.42×10−3			< 0.001	7.50×10−4
MFC	< 37 pcw	1.81×10−2	3.20×10−2	9.32×10−3	3.77×10−2			< 0.001	6.11×10−4
M1C	< 37 pcw			3.17×10−3	1.53×10−2			1.5×10−3	1.09×10−2
STR	>  25 years					2.96×10−2	4.56×10−3
HIP	> 25 years					3.20×10−2	1.88×10−3
OFC	> 25 years					3.94×10−2	2.50×10−4
VFC	< 37 pcw							9.59×10−3	9.09×10−4
OFC	< 37 pcw							2.46×10−2	5.30×10−3
AMY	< 37 pcw							2.86×10−2	1.04×10−2	2.27×10−2	< 0.001

ADHD attention-deficit hyperactivity disorder, ASD autism spectrum disorder, MDD major depressive disorder, BD bipolar disorder, SCZ schizophrenia, AMY amygdaloid complex, MFC anterior (rostral) cingulate (medial prefrontal) cortex, CBC cerebellar cortex, DFC dorsolateral prefrontal cortex, ITC inferolateral temporal cortex, MD mediodorsal nucleus of thalamus, OFC orbital frontal cortex, STC posterior superior temporal cortex, IPC posteroventral parietal cortex, A1C primary auditory cortex, M1C primary motor cortex, STR striatum, S1C primary somatosensory cortex, VFC ventrolateral prefrontal cortex, V1C primary visual cortex, HIP hippocampus.

It has been demonstrated that different psychiatric disorders involve the dysfunction of different brain structures, due to widespread alterations in functional connectivity. For instance, Moberget et al. reported that cerebellar cortical volume is significantly reduced in patients with SCZ compared with healthy controls [4]. Ishida et al. detected abnormalities in the cerebellum of BD patients [5]. ADHD has been found to be associated with volumetric reductions of the frontal regions [6]. Besides, there is growing evidence supporting the hypothesis that the age of onset of brain dysfunction ranges from early childhood for ASD to late adolescence and early adulthood for SCZ [7]. Psychiatric disorders usually begin at certain stages of life with specific gene expression patterns that contribute to brain dysfunction during the development of psychiatric disorders [8]. Therefore, we combined the brain structures and developmental stages to explore the genetic mechanisms of psychiatric diseases.

For SCZ, after comparing the GSEA results of nearby and meQTL-related genes, the significant developmental stages were < 37 pcw and 14–25 years, consistent with previous studies showing that the age of onset for SCZ typically begins at late adolescence or early adulthood [9]. Our results also suggested that dysfunction of the cerebellar cortex and dorsolateral prefrontal cortex contributed to the development of SCZ. One study that found a significant difference in cerebellar volume between SCZ patients and healthy controls implied that this abnormality is specific to SCZ [10]. Moreover, a previous study found that long-interval cortical inhibition of dorsolateral prefrontal cortex in SCZ patients is significantly lower than in healthy cases (P = 0.004) [11].

For BD, two stages (14–25 and > 25 years) were both significant in the GSEA results of nearby and meQTL-related genes. A previous study found that the typical age of onset for BD is young adulthood, consistent with our results [12]. Moreover, we discovered that the abnormal development of orbital frontal cortex and the striatum greatly contributed to the etiology of BD. Of note, there is supporting evidence that dysfunctional orbital prefrontal cortex is associated with mania, which is included in BD, using high-sensitivity positron emission tomography [13]. By comparing the transcriptomic sequencing between 18 BD patients and 17 controls and co-expression network analysis, the module with the highest genetic association signal for BD was also enriched in the gene expression of dorsal striatum medium spiny neurons, implying that the etiology of BD is associated with striatal function at the gene level [14]. In addition, novel isoforms of the PDE10A gene, usually expressed in the striatum, have been associated with BD [15].

Psychiatric disorders usually begin at certain stages of life with specific gene expression patterns, contributing to brain dysfunction during the development of psychiatric disorders. A comparative analysis of the genetic loci associated with psychiatric disorders and gene expression patterns specific to brain structures can be used to evaluate the functional relevance of different brain structures to psychiatric disorders. Our research supplies a method that can realize the localization of brain structures related to the different developmental stages of psychiatric disorders. This study also provides a new approach to studying the pathogenic mechanisms underlying complex mental diseases.

Nevertheless, there were three limitations in our study. First, we analyzed the differently-expressed genes in the differential developmental stages of 16 brain areas. However, due to the limitation of cases, other brain structures were not analyzed. Second, the data used in this study were obtained from Europeans. There are differences among ethnic groups in the mechanisms of psychiatric disorders. For this reason, careful interpretation is needed while applying our results to other groups. Third, although in the GSEA results we used FDR < 0.05 to control the false-positive rate, we did not control the false-positive rate when calculating the differently expressed genes specific to a brain structure at a certain development stage. In further research, this issue requires attention.

In this study, we aimed to evaluate the functional relevance of different brain structures in five common psychiatric disorders at different developmental stages. Our results provide novel clues for understanding the biological mechanisms of these disorders considering different brain structures at different ages.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Conflict of interest

The authors claim that there are no conflicts of interest.