Revisiting Schizophrenia from an Evolutionary Perspective: An Association Study of Recent Evolutionary Markers and Schizophrenia

Abstract

The persistence of schizophrenia in human populations at a high prevalence and with a large heritability estimate despite reduced fertility and increased mortality rate is a Darwinian paradox. This may be likely if the genomic components that predispose to schizophrenia are also advantageous for the acquisition of important human traits, such as language and cognition. Accordingly, an emerging group of genomic markers of recent evolution in humans, namely human accelerated regions (HARs), since our divergence from chimpanzees, are gaining importance for neurodevelopmental disorders, such as schizophrenia. We hypothesize that variants within HARs may affect the expression of genes under their control, thus contributing to disease etiology. A total of 49 HAR single nucleotide polymorphisms (SNPs) were prioritized from the complete repertoire of HARs (n = 2737) based on their functional relevance and prevalence in the South Asian population. Test of association using 2 independent schizophrenia case-control cohorts of north Indian ethnicity (discovery: n = 930; replication: n = 1104) revealed 3 SNPs (rs3800926, rs3801844, and rs764453) from chromosome 7 and rs77047799 from chromosome 3 to be significantly associated (combined analysis: Bonferroni corrected P < .002–.000004). Of note, these SNPs were found to alter the expression of neurodevelopmental genes such as SLC25A13, MAD1L1, and ULK4; a few from the HOX gene family; and a few genes that are implicated in mitochondrial function. These SNPs may most likely alter binding sites of transcription factors, including TFCP2, MAFK, SREBF2, E2F1, and/or methylation signatures around these genes. These findings reiterate a neurodevelopmental basis of schizophrenia and also open up a promising avenue to investigate HAR-mediated mitochondrial dysfunction in schizophrenia etiology.

Introduction

Schizophrenia is a complex neuropsychiatric disorder with high prevalence (0.5%–1%) and large heritability (~80%).¹ It typically emerges in late adolescence or early adulthood (15–25 years of age in men and 20–35 years of age in women), and patients have a reduced rate of reproduction (fitness) and increased mortality rate compared to unaffected individuals.^2–5 The reduced fertility and increased mortality rate among those with schizophrenia and their progeny, coupled with high heritability should result in a decrease in prevalence according to the evolutionary concept of negative selection, but schizophrenia has a relatively stable prevalence in the population, suggesting an evolutionary paradox.⁶ One explanation for this Darwinian paradox, which has remained constant through several genetic studies, is that the genomic components that predispose to schizophrenia are also advantageous for the acquisition of important human traits, such as language, complex cognitive skills, and other favorable brain functions, including behavioral flexibility.⁷ This also derives support from the observed enrichment of recent evolutionary markers in the vicinity of schizophrenia-associated loci.^7,8 Thus, schizophrenia has been hypothesized to be a byproduct of recent evolution of the human genome from that of its closest ancestor, the chimpanzee.^3,6–11

Therefore, an emerging group of genomic markers of recent evolution in humans since our divergence from chimpanzees, such as human accelerated regions (HARs), which are highly conserved across nonhuman species but have experienced accelerated substitutions in the human genome,¹² have been gaining importance in schizophrenia. Most HARs are either intergenic (60.7%) or intronic (32.4%)¹³ but several studies suggest that HARs can function as enhancers, repressors, modifiers, and regulating the expression of genes in their vicinity during brain, heart, and limb development,^14,15 while some HARs can induce human-specific gene expression profiles.^16–18 Though protein-coding sequences have evolved across related species, large phenotypic divergence between humans and chimpanzees is believed to be driven primarily by changes in gene regulation.^19,20 Thus, despite being distributed mostly in the noncoding region, HARs could be functionally relevant for species-specific development. In a comparison made between human and chimp brains, human-specific brain connections were shown to have significantly higher expression of HAR genes in the cortical areas of human brain (P = .005, permutation testing).²¹ From these studies, it is evident that HARs capture a modest portion of the evolutionary aspects witnessed in humans.

Recent targeted sequencing of HARs in consanguineous families with autism spectrum disorder (ASD) identified a significant enrichment of rare biallelic variants therein, highlighting a potential role of HAR variants in susceptibility to neurodevelopmental disorders.^12,13,22 Interestingly, these autism-associated HAR variants were shown to cause altered gene expression in mice neurospheres.¹³ Consistent with previous findings, an independent study showed that genes near HARs are mostly implicated in neurological disorder, suggesting that these regions will be valuable in the study of the pathogenesis of such disorders.²³ Among all the approved drugs, 32.42% target at least one HAR gene and, notably, a significant proportion of these HAR-gene targeting drugs are used for the treatment of neurological disorders, suggesting the implication of HAR genes in medical genetics and drug discovery.²³

A recent study using Psychiatric Genomic Consortium–genome-wide association study (PGC–GWAS) data generated from 36 989 schizophrenia cases and 113 075 controls²⁴ has indicated a probable association of HARs with schizophrenia by demonstrating that schizophrenia-related genes are in greater proximity of HARs and are also under stronger evolutionary pressure.⁸ This implies that these HARs may be important for human-specific functions and the variants in the regions may predispose to schizophrenia. We hypothesize that variants within human HARs may alter (1) the binding site of transcription factors and/or other regulatory proteins; and/or (2) epigenomic signature(s) of HARs, jeopardizing the expression of genes under their control, and, thus, can provide more insights into schizophrenia biology and may also help in identifying new drug targets.

With this background, in this study, we checked the association of 49 prioritized single nucleotide polymorphisms (SNPs) from within the complete repertoire of 2737 HARs known till date¹³ keeping in view their likely regulatory roles. Four SNPs known to alter the expression of various neurodevelopmental genes were significantly associated reiterating a neurodevelopmental etiology of schizophrenia.

Methodology

Sample Recruitment and Diagnostic Assessment

This study was approved by the institutional ethical committees of all participating institutions. The inclusion criteria for the recruitment of the participants (predominantly of north Indian origin) with schizophrenia, healthy adult and neonatal controls, have been previously described.²⁵ Briefly, patients diagnosed with schizophrenia conforming to Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria were recruited with written informed consent from the patients and accompanying relatives or caregivers as witnesses at PGIMER, Dr. RML Hospital, New Delhi.

Selection of SNPs From HARs

A total of 30 045 markers from a consolidated list of all HARs (n = 2737)¹³ were cataloged using hg19 and dbSNP146 module in UCSC table browser (https://genome.ucsc.edu/). From these, markers with minor allele frequency 0.4 < (MAF) > 0.1 in south Asian population, RegulomeDB²⁶ score <3, and/or with significant expression quantitative trait loci (eQTL) in different regions of the brain have been prioritized. A total of 49 SNPs were thus selected for genotyping (supplementary methods and supplementary table 1).

SNP Genotyping

Genotyping of prioritized SNPs was performed in a 2-stage case-control study design. For the discovery phase, nanofluidics-based Fluidigm SNPType genotyping assay was used at a commercial facility (Sandor Lifesciences Pvt. Ltd, Hyderabad, India). Significantly associated (P < .05) SNPs in the discovery cohort were then genotyped in the replication cohort using the PCR-restriction fragment length polymorphism (RFLP) method (supplementary table 2).

Genotyping QC

SNPs not in Hardy–Weinberg equilibrium (HWE; P < 10⁻³) and with >5% missing genotypes per individual and individuals with <95% SNPs successfully genotyped were removed using Plink software²⁷ (Version 1.9).

Test of Association

Pearson’s chi-square test of association was performed using —assoc option of Plink software.²⁷

Power Calculation

Power of the association study for each SNP was calculated using Quanto software^28,29 (Version 1.2.4). Parameters used for combined analysis were—study design: matched case-control; hypothesis: gene only; sample size (table 1); mode of inheritance: dominant and log additive; population risk: 0.01; OR (table 2); allele frequency of the minor allele in control group for each SNP (supplementary table 1).

Table 1.

Demographic details of the study cohort

Samples	Discovery		Replication
Total n = 1046 cases; 987 controls	Cases	Controls	Cases	Controls
	n = 494	n = 436	n = 552	n = 551a
Males	309	250	289	281
Females	185	186	263	270
Mean age	32.1 (±9.9)	37.43 (±13.3)	30.2 (±8.8)	28.4 (±4.6)

^a467 of these were cord blood; mean age calculated for adult controls only.

Table 2.

Test of association of human accelerated region SNPs in discovery and replication cohorts and combined analysis

rsID; hg19_chromosome: position (base change)	Discovery			Replication			Combined
	Chi sq	P value	OR (95% CI)	Chi sq	P value (Bonferroni adjusted)	OR (95% CI)	Chi sq	P value Bonferroni adjusted)	OR (95% CI)	Power (%) dominant/ log-additive
rs77047799; 3:41958895 (T>C)	5.37	.02	1.42 (1.05–1.9)	7.2	0.007 (0.03)	1.49 (1.11–1.99)	13.72	.0004 (.002)	1.45 (1.18–1.78)	92.98/96.37
rs3800926; 7:2205956 (G>A)	4.19	.04	1.25 (1.01–1.54)	6.9	0.009 (0.04)	1.32 (1.09–1.46)	9.96	.002 (.008)	1.26 (1.09–1.46)	75.11/92.30
rs3801844; 7:26829759 (T>C)	9.41	.002	1.44 (1.14–1.81)	7.73	0.005 (0.02)	1.32 (1.09–1.6)	14.75	.0001 (.0004)	1.33 (1.15–1.54)	89.04/97.12
rs764453; 7:95986043 (C>T)	14.61	.0001	1.55 (1.24–1.95)	10.18	0.001 (0.004)	1.38 (1.13–1.68)	23.59	.000001 (.000004)	1.44 (1.24–1.66)	98.32/99.87

Functional Annotation

Significantly associated SNPs were functionally annotated using FeatSNP³⁰ (http://featsnp.org; supplementary methods).

Mapping HAR SNPs to Genes

Most of the HARs are short (<1000 bp) DNA segments and evidenced to be regulatory elements that influence the transcriptional activity of the gene(s) in their vicinity.^14,31 Therefore, the mapping of SNPs to genes were done using FUMA-SNP2GENE³² module (https://fuma.ctglab.nl/snp2gene), which employs 3 approaches: (1) positional, (2) eQTL, and (3) chromatin interaction. eQTL mapping may lead to the identification of genes, which may not harbor the SNP per se but may have a functional impact. Chromatin loops on the other hand, connect regulatory elements to their target genes (supplementary methods).

Results

Study Cohort

The discovery cohort consists of 494 patients and 436 controls and the replication cohort with 552 patients and 551 controls (including 467 cord blood controls) and demographic details are presented in table 1.

Test of Association

Following quality control (QC) parameters, 54 individuals and 1 SNP were removed from the further analysis. Of the 48 SNPs tested (468 cases and 408 controls), 4 SNPs were found significantly associated (P < .04–.0001) with schizophrenia. PCR-RFLP-based genotyping of these 4 SNPs was successful in >95% of the samples in the replication cohort and all 4 SNPs were significantly associated (unadjusted P < .007–.001) and withstood Bonferroni correction (P < .01; adjusted P < .02–.004). Furthermore, association was retained in the combined analysis with ~75%–98% power in the dominant model and ~96%–99% in the log-additive model (table 2).

Functional Annotation of HAR SNPs

Functions of the 4 associated SNPs were then annotated using FeatSNP (table 3). Except rs77047799 (chromosome 3), the other 3 SNPs (chromosome 7) interfere with the binding of multiple transcription factors (TFs) and overlapped with histone ChIP-seq peaks that have methylation marks for active promoters, repressors, or enhancers, such as H3K4me1, H3K9ac, and H3K27ac in different regions of the brain (table 3; supplementary table 3).

Table 3.

Functional annotation of associated SNPs

Sl no	rsID;hg19_chromosome:position (base change)	Transcription factor motif predicted to be interfered	Overlapped histone ChIP-seq peaks in brain tissue
1	rs77047799;3:41958895 (C>T)	No information available	No information available
2	rs3800926;7:2205956 (A>G)	TFCP2	H3K4me1, H3K27ac, H3K9ac
3	rs3801844;7:26829759 (T>C)	MAFK, SREBF2, HOXC11, HOXB13, HOXC10, HOXC13, H0XA11	H3K4me1
4	rs764453;7:95986043 (C>T)	E2F1	Nil

Mapping HAR SNPs to Genes

This was performed using FUMA, an integrative web-based platform using information from multiple biological resources to facilitate the functional annotation of GWAS results. Thirty-five genes were identified around these 4 SNPs (table 4; supplementary table 2) as briefly described below.

Table 4.

Mapping of associated SNPs to gene(s)

rsID; chromosome:position (base change)	Genes mapped to
	Positional	eQTL	Chromatin interaction
rs77047799; 3:41958895 (C>T)	ULK4	ULK4, TRAK1	TRAK1, RPSA, MOBP, EIF1B, ENTPD3, RPL14, ZNF619, CTNNB1, ZNF621, ZNF620, CCK, LYZL4, VIPR1
rs3800926; 7:2205956 (A>G)	MAD1L1	MAD1L1, FTSJ2, NUDT1, SNX8	TMEM184A, PSMG3, ELFN1, AC074389.6
rs3801844; 7:26829759 (T>C)	SKAP2	SKAP2, HOXA1, HOXA2, HOXA3, HOXA5, HOXA6, HOXA7, HOXA9	SKAP2, HOXA1, HOXA2, HOXA3, HOXA5, HOXA6, HOXA7, HOXA9, HOXA4
rs764453; 7:95986043 (C>T)	—	SLC25A13	DLX6, DLX5

• (1) rs77047799 (3:41958895:C>T): This intronic SNP of ULK4 significantly associated with that of ULK4 and TRAK1(~96 kb away). Together with TRAK1, this SNP is also mapped to CTNNB, RPSA, MOBP, EIF1B, RPL14, ZNF619, ZNF621, ZNF620, LYZL4, and VIPR (~72 kb–2.5 Mb away) by chromatin interaction approach (table 4).

• (2) rs3800926 (7:2205956:A>G): This is an intronic SNP in MAD1L1 and significantly associated with its expression together with other genes, namely FTSJ2, NUDT1, and SNX8, which are ~67–85 kb away. In chromatin interaction mapping, this SNP mapped to TMEM184A, PSMG3, ELFN1, and AC074389.6, which are ~47–62 kb away (table 4).

• (3) rs3801844 (7:26829759:T>C): This intronic SNP of SKAP2 is significantly associated with its expression along with a few HOX genes ~12–38 kb away, which are also identified by chromatin interaction mapping strategy (table 4).

• (4) rs764453 (7:95986043; C>T): Based on positional mapping, no protein-coding gene was found within the 20- kb window of this SNP. However, this SNP is significantly associated with the expression of SLC25A13, which is ~23 kb away, and was mapped via chromatin interaction approach to DLX5 and DLX6, which are ~65 kb away (table 4).

Of note, several genes among these 35 genes have been implicated in neurodevelopment based on biochemical, in vitro, animal studies etc. (see Discussion).

Discussion

Recently it has been shown that schizophrenia-related genes are in greater proximity of HARs and are also under stronger evolutionary pressure.⁸ Of the 49 prioritized HAR SNPs tested in study subjects of north Indian ancestry, using a 2-stage study design, 3 SNPs (rs3800926, rs3801844, and rs764453) from chromosome 7 and 1 SNP from chromosome 3 (rs77047799) were found to be significantly associated with schizophrenia (table 2). Functional annotation of these SNPs (table 3) revealed that all the 3 SNPs from chromosome 7 modified the binding sites for TFs, such as TFCP2, MAFK, SREBF2, HOX, and E2F1, which are known to control the expression of neurodevelopmental genes; and rs3800926 and rs3801844 were also found to overlap with histone ChIP-seq peaks that have methylation marks for active promoters, repressors, or enhancers, such as H3K4me1, H3K9ac, and H3K27ac, in different regions of the brain. These SNPs were found to alter the expression of genes, such as SLC25A13, several HOX genes, MAD1L1 and ULK4, most probably via TF binding and/or through chromatin modification of the neighboring regions (discussed later in this section).

Of note, other than these developmental genes, we also observed 5 genes, namely SLC25A13, NUDT1, FTSJ2, MOBP, and TRAK1, which are involved in mitochondrial function. Several lines of evidence have proposed that an increased metabolic rate along with changes in energy allocation was crucial in the evolution of human brain size.³³ Mitochondria is the key player in brain energy homeostasis, and, consequently, the increased energy requirement demands a coevolution of the mitochondrial system.³⁴ It is likely that HARs regulate these genes in a human-specific way to provide this high energy requirement. In a recent study, 3 pathways related to mitochondrial function—OXPHOS, mito-nuclear crosstalk, and mitochondrial dynamics have shown evidence of involvement in the pathophysiology of psychiatric diseases, particularly schizophrenia and bipolar disorder.³⁵ It is likely that HAR SNP-mediated differential regulation of the above-mentioned genes might cause mitochondrial dysfunction but its role in schizophrenia etiology currently remains unclear.

Status of the 4 Associated SNPs in Other Studies

In Schizophrenia GWASs

We checked for the association status of the 4 associated SNPs in the publicly available large-sized schizophrenia GWASs of notably different ethnicities, such as European and Chinese, namely in PGC2,²⁴ meta GWAS, including CLOZUK and PGC1 samples,³⁶ and Han Chinese³⁷ GWASs. In partial support of our findings, we observed that 1 of the 4 SNPs (rs3800926; 7:2205956:A>G) was significant in both PGC2²⁴ (P value = 8 × 10^–5) and meta GWAS, including CLOZUK and PGC1 samples³⁶ (P value = 7 × 10^–7), as well as in Han Chinese GWAS³⁷ (P value = 9 × 10^–7). SNPs in linkage disequilibrium (LD) (r² ≥ .8) with this marker in these 3 studies also showed P values ranging from 9 × 10^–4 to 7 × 10^–7 in the 3 GWASs. On the other hand, rs77047799 (3:41958895:C>T) was not found to be associated in both PGC2²⁴ (P value = .4) and meta GWAS, including CLOZUK and PGC2 samples³⁶ (P value = .1). This SNP was not covered in Han Chinese GWAS³⁷ but few SNPs in LD showed nominal association (P value <0.02). However, 2 remaining SNPs (rs3801844; 7:26829759:T>C and rs764453; 7:95986043:C>T) or their LD SNPs were not found to be significant in any of these data sets (supplementary table 4). This limited replication of our results is not unexpected considering notable differences in the linkage disequilibrium structure of rs764453, rs3801844, and rs77047799 between Caucasian, South Asian, and Chinese populations (supplementary figures 3i–iv) and consequent differences in gene–gene (SNP–SNP) interactions, which determine the course of regulation of gene expression and ultimately disease phenotype.

In GWASs of Other Related and Unrelated Traits

As HARs have been reported to be associated with other neurological disorders, such as ASD,¹³ and in other human-specific developmental traits³⁸, we checked for the association of these SNPs in large-sized GWASs of related disorders, such as ASD, bipolar disorder (BPD), attention-deficit hyperactivity disorder, and brain-related traits, such as neuroticism³⁹ and educational attainment⁴⁰. Of note, the same SNP (rs3800926; 7:2205956:A>G) that was significant in both PGC2²⁴ and meta-schizophrenia GWAS³⁶ was also found to have nominal association with brain-related conditions, such as ASD, BPD, and neuroticism, but not with any unrelated traits (supplementary table 5). Importantly, P value of rs3800926 even attained genome-wide significance in education attainment GWAS⁴⁰; and rs3801844 was found to be nominally associated with depression³⁹ only. Neuroticism has been shown to increase the risk of depression and schizophrenia and found to have shared genetic risk loci.⁴¹ On the other hand, ASD, BPD, and schizophrenia show similar neuropsychiatric behavioral disturbances, with multiple overlapping genetic and environmental factors implicated in risk and course of illness.⁴² A recent study using brain transcriptome data gathered from postmortem donors affected with ASD, BPD, and schizophrenia reported shared risk genes relevant to synapse, neurotransmitter release, oxidative stress, nitric oxide synthase biosynthesis, immune response, etc.⁴³ Recent applications of whole-genome technologies have discovered both shared rare copy number variants and common SNPs that are associated with the risk of developing these disorders.⁴⁴ However, no SNP has shown association with any unrelated traits, such as height,⁴⁵ colorectal cancer,⁴⁶ eating disorder,⁴⁷ relative muscular strength,⁴⁸ 25-hydroxyvitamin D levels,⁴⁹ coronary artery disease,⁵⁰ and atrial fibrillation⁵¹ (supplementary table 5). These observations indicate that HAR SNPs, particularly rs3800926, might alter the regulation of genes that are implicated in brain-related functions warranting additional studies of HARs in human brain development.

Functional relevance of the associated SNPs:

• (1) rs3800926 (7:2205956:A>G): The risk allele “G” of this SNP reduces the binding affinity of transcription factor TFCP2. It has been implicated in disorders such as Parkinson’s related disorders, major depressive disorder, epilepsy, and Alzheimer’s.⁵² This SNP is found to be significantly associated with the differential expression of MAD1LI and FTSJ2, NUDT1, and SNX8 (https://gtexportal.org). Of these, the expression of MAD1L1, which encompasses the index SNP is highly correlated with the expression of TFCP2 in different regions of the brain (supplementary figure 1). Thus, it might be predicted that altered binding affinity of TFCP2, in turn, alters MAD1L1 expression, thus increasing the risk for schizophrenia. MAD1L1 (known as HAR3) has been suggested to play pivotal roles in human-specific or human-dominant diseases and also has been previously implicated in schizophrenia susceptibility.⁵³

• (2) rs3801844 (7:26829759:T>C): The risk allele “C” of rs3801844 was predicted to alter the binding site of TFs, namely SREBF2, MAFK, and several HOX genes (table 3). Furthermore, a significant association of this SNP with the differential expression of SKAP2 and a few HOX family genes have been observed (table 4). SKAP2 plays an essential role in the Src signaling pathway, which serves as a hub of various signaling mechanisms affecting NMDAR activation, and the suppression of such activation might increase schizophrenia susceptibility.^54–57 Though SKAP2 is a target gene of ^58–60, expression of SREBF2 is strongly correlated with SKAP2 expression in different brain tissues, such as the frontal cortex and hypothalamus (supplementary figures 2 and 4). These brain regions are found to be associated with schizophrenia.⁶¹ This indicates that the index SNP might interfere with the binding site of SREBF2 and/or MAFK and alters the expression of SKAP2, thus increasing the risk for schizophrenia. Interestingly, a frameshift variant in SKAP2 has been identified as a risk variant for schizophrenia (www.szdb.org).⁶²SREBF2 is located at 22q13.2, a region with well-established linkage to schizophrenia.⁶³ On the other hand, 2 markers located close to SREBF2 (0.8 and 1.9 Mb) were identified to be linked to schizophrenia in a linkage scan of 27 schizophrenia families,⁶⁴ reiterating its status as a strong candidate gene.^65,66 Furthermore, the association of SNPs within this gene has been reported in German population.⁶⁶HOX genes are a group of evolutionarily conserved genes that regulate early development.⁶⁷ Though conserved, HOX genes are differentially expressed in different species and probably responsible for species-specific developmental patterns and morphological diversities, probably by acquiring entirely novel regulatory elements.²⁰ Of note, these genes are found to be differentially expressed in the neural crest of chimp and humans.⁶⁸ HARs located nearby the HOX genes might be responsible for their human-specific expression and this HAR-mediated regulation can be altered by HAR SNPs, which, in turn, may increase the risk for schizophrenia. Interestingly, HOX genes are the target genes of MAFK transcription factor.^58–60 Thus, we hypothesize that the variant allele of rs3801844 alters the binding affinity of MAFK, which, in turn, modifies human-specific regulation of these HOX genes, consequently increasing the susceptibility to schizophrenia. HOX genes have been implicated in other disorders, such as mental retardation, subtypes of ASD, and epilepsy.^69,70 Additionally, de novo protein-coding mutations have been found in schizophrenia patients for HOXA1 and HOXA5 (www.szdb.org).⁷¹

• (3) rs764453 (7:95986043; C>T): Binding site for E2F1 predicted to be disrupted by the risk allele “T” of this SNP (table 3). E2F1 is a member of E2F family of transcription factors that are expressed during the embryonic development of the central nervous system and plays a pivotal role in cell proliferation, differentiation, and apoptosis.⁷² Differential requirements for E2F1 in regulating the proliferation of embryonic versus neonatal neural progenitors suggested an indispensable role for E2F1 in cortical and adult neurogenesis,⁷³ and other E2Fs cannot compensate for the loss of E2F1.⁷⁴ Interestingly, this SNP was found to be associated significantly with the differential expression of SLC25A13, which is a target gene of E2F1.^58,59,75SLC25A13 is a member of the mitochondrial carrier family and associated with altered neurogenesis and neuropsychiatric disorders.⁷⁶ Alterations in mitochondrial morphology, brain energy metabolism, and mitochondrial enzyme activity are involved in the pathophysiology of different neuropsychiatric disorders, such as schizophrenia, depression, and bipolar disorder.^77–79SLC25A13 found to be expressed differentially (false discovery rate-corrected P value = 2.51 × 10^–6) in schizophrenia cases with respect to controls in a study that includes RNA sequencing data from brain samples from ~500 individuals with schizophrenia and ~2000 controls.⁸⁰ Thus, it might be predicted that the risk allele of rs764453 destroys the TF-binding site of E2F1, which, in turn, decreases the expression of SLC25A13 interfering with neurogenesis and increasing the susceptibility to schizophrenia. Chromatin interaction mapping identified 2 genes namely DLX5 and DLX6 (table 3). These genes encode 2 homeobox TFs and, during development, Dlx5/6 involved in the differentiation of certain GABAergic neurons.⁸¹ Adolescent mice, heterozygous for a generalized deletion of Dlx5 and Dlx6 (Dlx5/6^+/−), present traits reminiscent of human schizophrenia⁸².

• (4) rs77047799 (3:41958895:C>T): This SNP is significantly associated with the expression of ULK4 and TRAK1 in different brain regions that are associated with schizophrenia (table 4; supplementary table 3). ULK4 plays a major role in the regulation of GABAergic signaling and corticogenesis and has been implicated in schizophrenia and related disorders.⁸³TRAK1 is involved in the regulation of endosome-to-lysosome trafficking, including endocytic trafficking of EGF-EGFR complexes and GABA-A receptors,⁸⁴ and these pathways that have been implicated in schizophrenia etiology.⁸⁵ Furthermore, when TRAK1 is O-glycosylated, it abolishes mitochondrial motility and is crucial for recruiting O-GlcNAc transferase to the mitochondrial surface of neuronal processes.⁸⁶ Together with DISC1, it has been implicated in schizophrenia in various studies.⁹¹ Disruption of the DISC1-Miro-TRAK complex has been shown to inhibit mitochondrial transport in neurons and impaired development of neuronal dendrites.⁹¹ Furthermore, several de novo mutations have been reported for TRAK1 in schizophrenia cases (www.szdb.org).^71,88TRAK1 is also associated with GSK3β, another gene implicated in schizophrenia.⁸⁹

In summary, based on the direct and indirect effects of the 4 significantly associated HAR SNPs from among the prioritized subset in this study, it seems likely that HARs may indeed be associated with schizophrenia. Furthermore, the overlap of HAR-associated genes and their targets with diverse neurodevelopmental phenotypes¹³ suggest that genome-wide mutational and functional characterization of HARs may substantiate their role in the regulation of brain size, structure, and/or other cognitive or social traits in humans.