Molecular Quantitative Trait Locus Mapping in Human Complex Diseases

Abstract

Mapping quantitative trait loci (QTLs) for the various molecular traits from chromatin to metabolites (i.e., xQTLs) provides insight into the locations and modes of the effects of genetic variants that influence these molecular phenotypes and the propagation of functional consequences of each variant. xQTL studies provide an indirect interrogation of the functional landscape of the molecular basis of complex diseases, including the impact of non-coding regulatory variants, the tissue specificity of regulatory elements and their contribution to disease by integrating with the genome-wide association studies (GWAS). We summarize a variety of molecular xQTL studies in human tissues and cells. In addition, using the Alzheimer’s Disease Sequencing Project (ADSP) as an example, we describe the ADSP xQTL project, a collaborative effort across the ADSP Functional Genomics Consortium (ADSP-FGC) with an ultimate goal of building a reference map of Alzheimer’s-related QTLs using existing datasets from multiple omics layers to help us study the consequences of genetic variants identified from the ADSP. xQTL studies enable the identification of the causal genes and pathways in GWAS loci, which will likely aid in the discovery of novel biomarkers and therapeutic targets for complex diseases in future development.

INTRODUCTION

Genome-wide association studies (GWAS) have discovered many genetic associations for complex human diseases and traits. However, it is still challenging to identify the causal variants and target the susceptible genes of most GWAS implicated loci since over 90% of these disease-associated variants are in non-coding regions and are not in linkage disequilibrium (LD) with any nonsynonymous coding single nucleotide polymorphisms (SNPs) (Hindorff et al., 2009; Manolio, 2010; Maurano et al., 2012, 2015). Thus, it is necessary to integrate functional genomics to help explain the effect of an individual variant on the specific gene(s). Particularly, it is important to apply functional genomics to account for the genetic influence of the variant on the organization of the human genome and epigenome, and on gene regulation, gene expression, and its protein product, i.e., the contribution of genetic effects at all major stages of gene regulation from chromatin to proteins and their metabolism process leading to specific molecular dysfunction and then human diseases.

Many GWAS genetic variants are located in gene regulatory elements (ENCODE Project Consortium, 2012; Roadmap Epigenomics Consortium et al., 2015) and are hypothesized to modulate gene regulation, through chromatin structure and modification, DNA methylation, chromatin looping, and occupancy by transcription factors and RNA-binding proteins (ENCODE Project Consortium et al., 2020). Over the last decade, technological advances have emerged for the genome-wide profiling of these molecular traits across multiple tissues and cell types. With large available datasets from multiple omics layers including bulk and single-cell transcriptomics, epigenomics, proteomics, and metabolomics, a multi-omics approach can be applied to map molecular quantitative trait loci (QTLs) across epigenome, transcriptome, proteome and metabolome, and model the propagation of functional consequences of each variant through the various levels of data, i.e., an xQTL mapping.

Recently, several studies have applied xQTL mapping to RNA sequencing, DNA methylation and histone acetylation data from the same tissue collected from subjects to generate a multi-omic xQTL resource (Chen et al., 2016; Ng et al., 2017). xQTL studies provide an indirect interrogation of the functional landscape of the molecular basis of complex diseases and provide insight into the locations and modes of the effects of functional sequences that influence molecular traits. Specifically, xQTL studies have illuminated the impact of regulatory variants, the tissue specificity of regulatory elements and the contribution of cis- versus trans-acting variation to these intermediate molecular traits.

In this article, we summarize a variety of molecular xQTL studies in human tissues and cells. In addition, using the Alzheimer’s Disease Sequencing Project (ADSP) as an example, we describe the ADSP xQTL project, a collaborative effort across the ADSP Functional Genomics Consortium (ADSP-FGC).The goal of the ADSP xQTL project is to generate a reference map of Alzheimer’s-related QTLs using existing datasets from multiple omics layers including bulk and single-cell transcriptomics, epigenomics, proteomics, lipidomics, and metabolomics generated from multiple AD relevant tissues including brain, cerebrospinal fluid (CSF), and peripheral blood/plasma (https://adsp-fgc.niagads.org/). Last, we explore how xQTLs can be used as instrumental variables to identify the causal gene(s) in each locus as well as novel biomarkers and therapeutic targets for diseases in future development.

KEY CONCEPTS

Expression quantitative trait loci (eQTLs): regions of the genome associated with the expression of a certain gene.

Cis/trans eQTL:

the definition of a cis effect is somewhat arbitrary, but cis-acting eQTLs are typically defined to include SNPs within 1Mb upstream and downstream of the gene that is affected by that eQTL, otherwise, are trans-acting eQTLs.

xQTLs:

integrated functional genomic QTLs, i.e., regions of the genome associated with a range of molecular quantitative traits across epigenome, transcriptome, proteome and metabolome, and the propagation of functional consequences of each variant through the various levels of data.

STRATEGIC APPROACH

In this section, we present a variety of molecular traits xQTL studies (summarized in Table 1, Figure 1) by following the molecular process/architecture starting from epigenome, gene regulation, gene expression, post-transcriptional regulation, proteins and metabolites.

Table 1.

Examples of published molecular QTLs

	Name	description of the molecular trait	Example of Publication
			Journal Year	Author (PMID)
1. Gene Regulation		epignetics/epigenome
mQTLs	DNA methylation (DNAm) QTL		Cell 2016; Nat Commun. 2018	Chen (27863251); Huan (31537805);
			Nat Genet. 2021	Min (34493871)
			PLoS Genet. 2010	Gibbs (20485568)
dsQTL	DNase I sensitivity quantitative trait loci	DNase I sequencing to measure chromatin accessibility	Nature 2012	Degner (22307276)
haQTLs	histone modification (H3K9Ac)/Chromatin/		Nat Neuroscience 2017	Ng (28869584)
	eqtl, methylation and histone variation
	H3K4me1, H3K27ac		Cell 2016	Chen (27863251)
caQTLs		Chromatin accessibility QTLs	AJHG 2021	Currin (34038741)
2. Gene Expression
eQTLs	expression QTL	gene-level expression measured	Science 2007	Stranger 17289997
			Nature 2011	Kleinman (22031444)
			Cell 2013	Emilsson (23622250)
			Hum Mol Genet. 2014	Zhang (24057673)
			Nat Neurosci. 2014	Ramasamy (25174004)
			Genome Biol. 2017	Joehanes (28122634)
			Nature. 2010	Pickrell (20220758)
			Nature 2017; Science 2020	GTEx Consortium (29022597, 32913098)
eQTS	expression quantitative trait scores (between Polygenic scores and gene expression)	eQTLGen Consortium	Nat Genet. 2021	Vosa (34475573)
ct-eQTLs	cell-type-specific eQTLs		Nat Genet. 2017; Transl Psychiatry 2021	Zhernakova(27918533); Patel (33907181)
			Nat Commun. 2020	Donovan (32075962)
			Science 2020	Kim-Hellmuth (32913075)
	celltype-specific eQTLs	using single-cell RNA sequencing	Nat Genet. 2018	van der Wijst (29610479)
	single-cell eQTL	Single cell eQTL analysis	Genome Biol. 2021	Neavin (33673841)
			Elife. 2020	van der Wijst (32149610)
			Genome Biol. 2021	Cuomo (34167583)
			Cell 2021	Ota, 33930287
rare-eQTLs			Nature 2017; Science 2020; Genes 2021	Li (29022581); Ferraro (32913073); Patel (33804025)
sQTLs	splicing QTL	transcript-isoform level expression
			Nature 2010	Montgomery (20220756)
			Nature 2013	Lappalainen (24037378)
			Nat Genet. 2015	Zhang (25685889)
			Science 2016	Li (27126046)
			Nat Genet. 2018	Raj (30297968)
			Nat Commun. 2021	Garrido-Martín (33526779)
			Nat Genet 2021	Kerimov (34493866)
3. Post-transcriptional Regulation
nmdQTLs	nonsense-mediated mRNA decay (NMD)		PLoS Biol. 2008	Eberle (18447580)
	rdQTLs		PLOS Genetc. 2012	Pai 23071454
			Am J Hum Genet. 2021	Teran (34216550)
3′aQTLs	3′UTR alternative polyadenylation (APA) QTL		Nat Genet. 2021	Li (33986536)
mirQTLs	microRNA QTL		Nat Comm 2015	Huan (25791433)
			Database (Oxford) 2021	Akiyama (34730175)
m6A QTLs	N6-methyladenosine (m6A) QTL		Nat Genet. 2020	Zhang (32601472)
			Nat Genet. 2021	Xiong (34211177)
4. Proteins
pQTLs	protein QTL	protein-level expression	Nature 2018	Sun (29875488)
			Nat Commun. 2018	Yao (30111768)
			Nat Genet. 2021	Wingo (33510477)
			Nat Neurosci. 2021	Yang (34239129)
5. Metabolites
metaQTLs			Nat Genet. 2014	Shin (24816252)
			PLoS Genet. 2020	Riveros-Mckay (32150548)
6. xQTL	same variant influences multiple omic features in cis		Nat Neuroscience 2017	Ng (28869584)
	gene expression, methylation, and histone modification		Cell 2016	Chen (27863251)
			Science 2016	Li (27126046)

Figure 1.

Types of genetic effects on different layers of molecular traits from chromatin to metabolites

1. Epigenetics and Gene regulation

Over the past decade, an increasing spotlight has been placed on epigenome and factors that can alter transcription/gene expression (i.e., gene regulation) such as epigenetic CpG methylation in relation to their impact on disease phenotypes (Ptak & Petronis, 2008). Consequently, studying the effects of genetic variants influencing epigenomic and gene regulation provides an opportunity to determine how these influence gene expressions, and ultimately to determine how the functional consequences of these genetic variants influence disease risk.

1.1. DNA CpG methylation QTL (mQTLs)

In 2010, Gibbs et al reported the first genome-wide QTL study of DNA methylation (DNAm) (i.e., mQTL) in human brain (Gibbs et al., 2010), which investigated the effects of common genetic variants on DNA methylation and mRNA expression in four human brain regions each from 150 individuals (600 samples total). This study showed peak enrichment for cis-eQTLs to be approximately 68,000 bp away from individual transcription start sites (TSSs); however, the peak enrichment for cis-mQTLs is located much closer, only 45 bp from the CpG site studied. Individual QTLs that appear to affect both the expression level of a gene and a physically close CpG methylation site are quite rare. Similar to this finding, Gamazon et al has explored the effects of bipolar disorder (BD) genetic variants on DNA methylation and mRNA expression in human cerebellum samples. BD variants that cis regulate both cerebellar expression and methylation of the same gene are a very small proportion of BD variants, suggesting that mQTLs and eQTLs provide orthogonal ways of functionally annotating genetic variation within the context of studies of pathophysiology in the brain (Gamazon et al., 2013). In 2018, Huan et al reported cis- and trans-mQTLs using 4,170 whole blood samples collected from a community-based cohort (Huan et al., 2019). More recently, using blood samples from 32,851 participants, Min et al. have provided a database of >270,000 independent mQTLs, of which 8.5% are trans. The authors have also shown that a substantial proportion (15–17%) of the additive genetic variance can be explained by mQTL and often influence disease phenotypes in complex ways (Min et al., 2021).

1.2. Other epigenomic QTLs

Besides large-scale mapping of mQTLs in human tissues, genome-wide search of other epigenomic QTLs has recently emerged. Using Dnase I hypersensitive sites sequencing (Dnase-seq) to measure chromatin accessibility in 70 Yoruba lymphoblastoid cell lines, Degner et al. were able to map Dnase I hypersensity QTLs (dsQTLs) and found that around 16% of the dsQTLs were eQTLs as well (Degner et al., 2012). Ng et al identified SNPs significantly associated with histone modification levels (hQTLs) by analyzing histone acetylation (H3K9Ac) data from the dorsolateral prefrontal cortex of 433 older adults (Ng et al., 2017). Furthermore, mQTLs and hQTLs may influence disease by mediating chromatin accessibility. In 2016 Chen et al. investigated mQTLs and hQTLs (H3K4me1, H3K27ac) in three major human immune cell types (CD14+ monocytes, CD16+ neutrophils, and naive CD4+ T cells) from up to 197 individuals. They found that many of these molecular-trait QTLs were colocalized with disease-associated loci (Chen et al., 2016). More recently, Currin et al mapped chromatin accessibility quantitative trait loci (caQTLs) in 20 human liver samples using ATAC-seq. The identified 3,123 caQTL variants are enriched in liver tissue promoter and enhancer states and frequently disrupt binding motifs of transcription factors expressed in liver (Currin et al., 2021).

2. Gene Expression QTLs

Gene expression molecular traits are gene-level RNA expression measurements for protein-coding genes, non-coding RNAs and their splicing transcript isoforms.

2.1. Expression quantitative trait locus (eQTLs)

Comprehensive eQTL mapping provides an important source of references for categorizing both cis and trans effects of disease-associated SNPs on gene expression, which aid in interpreting the results of GWA studies. eQTL mapping seeks to identify genetic variants that affect the gene expression and offer new functional information about the etiology of complex human diseases when combined with GWAS (Cookson et al., 2009; Montgomery & Dermitzakis, 2011). eQTL mapping was one of the earliest genome-wide molecular QTL studies. It largely began in 2007 when it became feasible to run genome-wide microarray gene expression platform for a relatively large samples and genotype data also available from the same subjects (e.g. lymphoblastoid cell lines of all 270 individuals genotyped in the HapMap) (Stranger, Forrest, et al., 2007; Stranger, Nica, et al., 2007).

Several studies have successfully used this approach to identify putative susceptibility genes at GWAS risk loci by analyzing human expression data from brain prefrontal cortex brain region (Colantuoni et al., 2011; Zhang et al., 2013), ten brain regions (Ramasamy et al., 2014), and blood (Joehanes et al., 2017; Zhang et al., 2014). A community resource of cis- and trans-acting eQTLs was created by synthesizing eQTL results from >50 datasets made available through the NHLBI GRASP database (http://apps.nhlbi.nih.gov/grasp/).This includes eQTL results from a variety of human tissues including brain, blood tissues/cells, liver, kidney, peripheral artery plaque, stomach, subcutaneous adipose, omentum and skin (Zhang et al., 2014).

It is now well established that eQTLs are abundant in a wide range of tissues, cell types and in diverse organisms, and numerous studies have implicated human regulatory eQTLs as being important contributors to phenotypic variation. Many regulatory elements are tissue specific (Heintzman et al., 2009). In order to sample the full range of biological effects of such regulatory variants, the Genotype-Tissue Expression (GTEx) project performed whole-genome sequence (WGS) and RNA-sequencing profiling for >40 tissues samples collected from the same ~800 healthy donors, providing a unique and rich resource to study tissue-specific eQTLs in reference samples (GTEx Consortium, 2020; GTEx Consortium et al., 2017).

Many eQTL studies have published their summary statistics, which can be used to gain insight into complex human traits by downstream analyses, such as fine mapping and co-localization. However, technical differences between these datasets are a barrier to their widespread use. Most recently, Kerimov et al present the eQTL catalogue (https://www.ebi.ac.uk/eqtl), a resource of quality-controlled, uniformly re-computed gene expression and splicing QTLs from 21 human studies (Kerimov et al., 2021).

2.2. Expression quantitative scores (eQTSs)

The omnigenetic model of complex disease traits proposes that disease traits are mostly controlled by a large number of small scale trans effects which influence a smaller set of core genes (Liu et al., 2019). On this basis, Võsa et al. decided to prioritize these core genes by looking for genes whose expression is associated with polygenic risk scores through meta-analysis using blood-derived expression from 31,684 individuals in the eQTLGen Consortium (Võsa et al., 2021). These expression quantitative scores (eQTSs) were presumed to be key drivers of disease phenotype due to this. The authors found that a small proportion of genes (13%) whose expression correlated with polygenic scores for 1,263 phenotypes (i.e. eQTSs) and that these genes were disproportionately trans-eQTLs, further supporting the use of the omnigenetic model and pinpointing potential drivers for those traits.

2.3. Cell type-specific eQTLs (ct-eQTLs)

Many regulatory elements are cell-type specific (Heintzman et al., 2009) and can be difficult to detect using traditional tissue specific RNA expression data alone. Zhernakova et al. were some of the first people to analyze cell type specific eQTL (ct-eQTL) by adding a cell-type interaction term to their model of gene expression (Zhernakova et al., 2017).

While bulk RNA-based methods have been successful at detecting ct-eQTLs they are fundamentally limited in their scope and accuracy. Cuomo et al. for instance found that, compared to traditional bulk-RNA based methods, they were able to find up to twice as many ct-eQTL with their optimized single-cell workflow (Cuomo et al., 2021). A genome-wide single cell eQTL analysis was first done in 2018 by van der Wijst et al. in ~25,000 peripheral blood mononuclear cells from 45 donors, identifying novel eQTL (van der Wijst et al., 2018). These methods have also been used to study cell-type specific regulation in reprogrammed fibroblasts (Neavin et al., 2021). This growth in single cell-eQTL studies has prompted researchers to create the single cell eQTLGen Consortium to further study the cellular context around disease causing genetic variants (van der Wijst et al., 2020) and to optimize study design of scRNAseq experiments for ct-eQTL analysis (Mandric et al., 2020).

Further studies have found associations between ct-eQTL and disease phenotypes. For instance, deconvoluted GTEx bulk tissue samples have been used to colocalize novel ct-eQTL/cell type-interaction QTL with various diseases that are masked in bulk tissue (Donovan et al., 2020; Kim-Hellmuth et al., 2020). Similarly, Patel et al. were able to find ct-eQTL which implicated novel genes in Alzheimer’s disease risk and provided further evidence for the involvement of myeloid cells in the disease (Patel et al., 2021). Most recently, Ota et al revealed a dynamic landscape of immune cell-specific gene regulation in immune-mediated diseases by performing a large-scale immune cell eQTL analysis on a dataset consisting of 28 distinct immune cell subsets from 337 patients diagnosed with 10 categories of immune-mediated diseases and 79 healthy volunteers. The authors reported dynamic variations of eQTL effects in the context of immunological conditions, as well as cell types. These cell-type-specific and context-dependent eQTLs are significantly enriched in immune disease genetic loci and implicate disease-relevant cell types, genes, and environment, which help us understand the immunogenetic functions of disease variants under in vivo disease conditions (Ota et al., 2021).

2.4. Rare-eQTLs

Most eQTL studies only look at the effect of variants that surpass a certain minor allele frequency threshold. There is usually very little statistical power to detect effects on molecular traits from rare variants due to their low allele counts in studies with small or moderate sample sizes. Despite this, Li et al. found a disproportionate quantity of rare variants near the sites of over and under-expression outlier genes (Li et al., 2017), prompting them to create a Bayesian model to predict the regulatory effect of rare variants. Ferraro et al. were able to show that this pattern of rare-variant enrichment in extreme molecular phenotypes was not only limited to RNA expression alone but also found in allelic expression and alternative splicing outliers as well (Ferraro et al., 2020). These rare-eQTLs have been implicated in disease phenotypes as well, with Patel et al. finding many rare-QTLs in genes and pathways that have been implicated in Alzheimer’s disease (Patel et al., 2021).

2.5. Alternative splicing QTLs (sQTLs)

An understanding of the genetic variation underlying transcript splicing is essential to dissect the molecular mechanisms of common disease. The effect of genetic variants on alternative RNA splicing was initially observed in 60 individuals from the HapMap project (Montgomery et al., 2010). Since then, it has been revealed that these splicing QTLs (sQTLs) are largely independent eQTLs (Lappalainen et al., 2013) and often mediate disease-relevant GWAS results in conditions such as cardiovascular diseases (Zhang et al., 2015), schizophrenia (Takata et al., 2017) and Alzheimer’s disease (Raj et al., 2018), etc. Specifically, Zhang et al performed sQTL analysis in whole blood collected from 5,257 Framingham Heart Study participants, and found many sQTL-associated genes (40%) undergo alternative splicing. In particular, the authors found 395 (4.5%) GWAS SNPs with evidence of cis-sQTLs but not gene-level cis-eQTLs. This suggests that sQTL analysis could provide additional insights into the functional mechanism underlying GWAS results. Similarly, Li et al reported that sQTLs and eQTLs tend to be independent, and eQTLs are enriched near TSSs, but the sQTLs are enriched within gene bodies and in particular within the introns they regulate (Li et al., 2016).

In 2018, Raj et al performed RNAseq to identify sources of variation in mRNA splicing in the dorsolateral prefrontal cortex (DLPFC) of 450 subjects from two aging cohorts. The authors reported that altered splicing is the mechanism for the effects of the PICALM, CLU and PTK2B susceptibility alleles, and dysregulation of mRNA splicing is a feature of Alzheimer’s disease and is, in some cases, genetically driven.

Analyzing the GTEx dataset, Garrido-Martín et al generated a comprehensive catalog of sQTLs across multiple tissues in the human genome and found that sQTLs tend to be located in post-transcriptionally spliced introns, which would function as hotspots for splicing regulation. While many variants affect splicing patterns by altering the sequence of splice sites, many more modify the binding sites of RNA-binding proteins. Genetic variants affecting splicing can have a stronger phenotypic impact than those affecting gene expression (Garrido-Martín et al., 2021). Most recently, Kerimov et al present the eQTL catalogue (https://www.ebi.ac.uk/eqtl), a resource of quality-controlled, uniformly re-computed gene expression and splicing QTLs from 21 studies on human. This resource focuses on cis-eQTLs and on cis-sQTLs where variants are associated to specific splicing events on nearby splice junctions (Kerimov et al., 2021).

3. Post-transcriptional regulation QTLs

Post-transcriptional regulation is the control of gene expression at the RNA level. It occurs between the transcription) phase and the translation) phase of gene expression. Post-transcriptional regulation is critical for the regulation of many genes across human tissues (Franks et al., 2017). It also plays a major role in cell physiology, being implicated in pathologies such as cancer and neurodegenerative diseases (Dassi, 2017).

3.1. Nonsense-mediated mRNA decay/RNA decay quantitative trait loci (nmdQTLs/rdQTLs)

Translation termination at premature termination codons (PTCs) triggers degradation of the aberrant mRNA. Normal termination codons can trigger nonsense-mediated mRNA decay (NMD) when this distance is extended; and vice versa (Eberle et al., 2008). Gene expression can be regulated post-transcriptionally by generating alternatively spliced transcripts which are degraded by NMD. In other words, NMD is a form of post-transcriptional gene regulation that can be coupled to alternative splicing. In 2012, Pai et al used a time-course study design to estimate mRNA decay rates in 70 Yoruban HapMap lymphoblastoid cell lines (LCLs), and to map genetic variation that is specifically associated with variation in mRNA decay rates across individuals. 195 such loci were identified and named RNA decay quantitative trait loci (“rdQTLs”) (Pai et al., 2012). More recently, Teran et al used RNA sequencing of the GTEx v8 cohort to compute the efficiency of NMD using allelic imbalance for 2,320 rare protein-truncating variants (PTVs) across 809 individuals in 49 tissues (Teran et al., 2021).

3.2. 3′UTR alternative polyadenylation (APA) QTL (3′aQTL/apaQTLs)

Alternative polyadenylation (APA) is an important part of post-transcriptional gene regulation as it is able to modify aspects of mRNA stability, translation, and localization (Tian & Manley, 2017). Aberrant expression of APA associated genes have been linked to both glioblastoma (Masamha et al., 2014) and idiopathic pulmonary fibrosis (Weng et al., 2019). This points towards a broad importance of APA in modulating disease status. In response to this Li et al. performed a broad analysis of APA QTL (3’aQTL) across various tissue types (Li et al., 2021). The authors found that most 3’aQTL were orthogonal to other molecular-trait QTL. The results were further validated by the authors through a CRISPR-based analysis.

3.3. microRNA quantitative trait loci (mirQTLs)

microRNAs (miRNAs) are small non-coding RNA molecules that have been found to regulate the expression and translation of mRNA (Bartel, 2009). This gene regulating ability of miRNAs comes from their complementary base pairing with their target mRNA. This property has prompted an interest in studying the effects of genetic variants in miRNA genes on disease risk, especially in the context of cancer risk (Ryan et al., 2010). This interest was initially focused on variants which modified the binding of miRNA to their mRNA targets but future studies looked at the effect of miRNA eQTL (mirQTL) on disease. Since the cis-regulation of miRNA is likely to correspond to the trans-regulation of mRNA (see Figure 2), identifying mirQTL can provide further insights into transcriptional mediators of disease. In 2015, Huan et al. identified 5,269 cis-mirQTLs from the whole blood of 5,239 individuals (Huan et al., 2015). The authors found that around half of the cis-mirQTLs were relatively distal (300–500 kb) from their respective intergenic miRNA gene, suggesting a high importance of distal regulatory elements in regulating miRNA expression. Additionally, the mirQTLs were found to be enriched for cis-mRNA-eQTLs, suggesting that many miRNAs are co-regulated with some mRNA. Most recently, a cis- and trans-mirQTL resource created from 3,448 Japanese serum samples across six dementia types was made publicly available, contributing to discovery of functional variants influencing miRNA expression, particularly in dementia studies (Akiyama et al., 2021).

Figure 2. Examples of genetic driver SNPs:

A) a SNP (black triangle) cis-regulates a miRNA gene and also trans-regulates its mRNA targets (anti-correlated); B) a SNP cis-regulates a transcription factor (TF) gene encoding a repressor/activator, and this SNP also trans-regulates its miRNA targets (anti-/positively-correlated).

3.4. N6-methyladenosine (m⁶A) RNA methylation QTLs (m⁶A QTLs)

A common post-transcriptional RNA modification is N⁶-methyladenosine (m⁶A) methylation. Variations in m⁶A methylation status have particularly been associated with various forms of cancer (Barbieri & Kouzarides, 2020). This has prompted researchers to look for genetic variants associated with m⁶A status. Zhang et al mapped the quantitative trait loci (QTLs) of m⁶A peaks in 60 Yoruba (YRI) lymphoblastoid cell lines and found that m⁶A QTLs are largely independent of expression and splicing QTLs and are enriched within the binding sites of RNA-binding proteins, RNA structure-changing variants and transcriptional features. The authors also show that m⁶A QTLs contribute to the heritability of various immune and blood-related traits at levels comparable to splicing QTLs and roughly half of expression QTLs (Zhang et al., 2020). Xiong et al. assayed 4 different tissues from 91 GTEx samples to perform a transcriptome-wide m⁶A QTL analysis (Xiong et al., 2021). The authors found that, while there was a slight enrichment for eQTL in the m⁶A QTL, most m⁶A QTL were disjoint from eQTL. Despite this, many m⁶A QTL were colocalized with GWAS hits, suggesting that m⁶A methylation may largely mediate disease status in an RNA expression-independent way.

4. Protein quantitative trait loci (pQTLs)

Protein QTLs (pQTLs), much like eQTLs, have been found to mediate disease risk and have a large overlap (Sun et al., 2018). For instance, in their study on cardiovascular disease, Yao et al. found that up to 190 of 372 non-redundant pQTL were colocalized with eQTL (Yao et al., 2018). pQTL variants have been associated with cardiovascular disease (Yao et al., 2018) and Alzheimer’s disease (Wingo et al., 2021). In the latter case, the effect of the disease was found to be independent of APOE4 status. Despite their frequent colocalization, many pQTL are orthogonal to other molecular trait QTL. In an analysis of cerebrospinal fluid, plasma, and brain samples, Yang et al. found that 48.0–76.6% of pQTLs did not co-localize with any eQTLs, sQTLs, mQTLs, or histone modification QTLs (Yang et al., 2021). In addition, they also found that cross tissue pQTL tended to be those which acted in cis, highlighting the importance of trans-pQTL in dissecting tissue specific pQTLs.

5. Metabolite quantitative trait loci (metaQTLs)

Concentrations of various metabolites have been linked with various disease phenotypes. Metabolite QTL (metaQTL) are therefore useful in determining what biological processes mediate disease phenotype. An early example of a metaQTL analysis was performed by Kolz et al. where several metaQTLs associated with serum uric acid levels were identified (Kolz et al., 2009). Further studies were able to find novel metaQTLs associated the concentration of various blood lipids in people of European ancestry which were further validated in non-European populations and with experiments on mouse models (Teslovich et al., 2010). Other studies, such as those done by Kettunen et al. (Kettunen et al., 2012) and Shin et al. (Shin et al., 2014), assessed metaQTL for hundreds of different metabolites and found a high degree of heritability for many metabolite concentrations. Shin et al. reported a disproportionate amount of metaQTL to be highly enriched for druggable targets, suggesting that metaQTL-linked genes may be useful in drug development. More recently, Riveros-Mckay et al. showed an effect of rare metaQTL on serum metabolite concentrations (Riveros-Mckay et al., 2020).

6. Multiple-layer Molecular trait QTLs (xQTLs)

By integrating molecular traits measured from the same tissue collected from same subjects, xQTL studies can be performed to identify variants that influence multiple omics features in cis. For example, Ng et al reported a multi-omics resource generated by applying xQTL analyses to RNA sequencing, DNA methylation and histone acetylation data from the same dorsolateral prefrontal cortex sample of 411 older adults (Ng et al., 2017). The authors found that in 9% of cases, SNP effects on RNA expression (i.e. eQTL effects) are fully mediated by epigenetic factors. By incorporating these various molecular-trait QTL to prioritize GWAS signals, they were able to identify novel loci involved in schizophrenia and bipolar disorder using an xQTL-weighted analysis approach. In addition, Chen et al carried out high-resolution genetic, epigenetic, and transcriptomic profiling in three major human immune cell types (CD14⁺ monocytes, CD16⁺ neutrophils, and naive CD4⁺ T cells) from up to 197 individuals. The authors characterized highly coordinated genetic effects on gene expression, methylation, and histone variation through QTL mapping and allele-specific analyses, and further demonstrated colocalization of xQTLs at 345 unique immune disease loci (Chen et al., 2016).

7. Application of xQTLs

Approaches that combine xQTLs and genome-wide association studies (GWAS) are offering new functional information about the etiology of complex human traits and diseases. Summary-based Mendelian Randomization (SMR) is a technique that integrates xQTL data with GWAS summary-level data of disease to determine associations between genetically determined molecular traits, such as expression and methylation, and outcomes of interest, such as diseases. This method can be used to prioritize genes/pathways underlying GWAS hits for follow-up functional studies (Zhu et al., 2016). An LD score-based approach was developed to study the cis-eQTL mediated heritability of 42 GWAS traits using GTEx data while accounting for both pleiotropy and linkage disequilibrium (Yao et al 2020). Recent Bayesian approaches have enabled simultaneous colocalization across a vast number of xQTL and GWAS traits to prioritize causal variants and identify pleiotropy (Giambartolomei et al 2018, Foley et al 2021). xQTL data can also be integrated with GWAS data at gene level to identify risk genes, gene-sets and pathways (Gerring et al, 2021, Weeks et al 2020).

Transcriptome-wide association study (TWAS), a method that systematically investigate the relationship between genetically predicted gene expression and disease risk, provides a powerful approach to identify disease risk genes and can uncover possible causal genes at loci identified by GWAS (Gamazon et al., 2015; Gusev et al., 2016). Using this approach, Li et al identified 66 genes whose predicted expression or splicing levels in dorsolateral prefrontal cortex (DLFPC) and peripheral monocytes are significantly associated with Parkinson’s disease (PD) risk (Li et al., 2019). Additionally, considering the essential role of epigenetic features in predicting gene expression, Yao et al developed an epigenetic element-based TWAS considering both genetic and epigenetic effects on gene expression as a powerful method to identify specific genes and mechanisms that underlie PD genetic risk factors (Yao et al., 2021). Applying the same idea to integrate pQTL results, Wingo et al performed a brain proteome-wide association study and implicated novel proteins in depression pathogenesis (Wingo et al., 2021). Various statistical methods have also been developed for multi-tissue extensions to TWAS (Barbeira et al 2019, Hu et al 2019, Zhou et al 2020), with applications to neurodevelopmental disorders such as Autism (Rodiguez-Fontenla et al 2021) and schizophrenia (Wu et al 2021).

8. Future Development

The multi-omic xQTL approach has enabled researchers to map the propagation of functional consequences of each variant and identify the causal gene(s) and pathways in GWAS loci. To establish links between tissue-specific cellular effects and diseases of interest, we need improved study designs which make use of technological advances in resolving the omics, cell history and state, population of origin and diverse endophenotypes.

With more single-cell level omics data available, cell-type-specific xQTLs will be studied extensively to provide insights into the architecture of disease and the landscape of gene regulation in the cell-specific level. In addition, deconvolution methods are very useful for estimating cell fraction/composition for cell-type-specific QTL study, especially for less available tissues, e.g., brain. However, optimal new methods are still needed for handling sequencing differences in snRNA-seq/sn-ATAC-seq data. Additional reference single-cell/single-nuclei level omics data across tissue types /brain regions, disease stages and diverse ancestry/ethnic groups are needed.

Studying common eQTL variant regulating gene expression in healthy reference samples allows us to distinguish the causal effects leading to the phenotype from the reactive effects that emerge to the individuals after developing the phenotype (Schadt et al., 2005). xQTL studies directly performed on the subjects of the complex trait study will be informative. An increasing number of studies are interrogating gene expression in well-phenotyped samples (Montgomery & Dermitzakis, 2011). Recently, sex-biased eQTLs in cis (sb-eQTLs) and high-confidence population biased eQTLs (pb-eQTLs) separately for individuals of European or African ancestry are made available from GTEX v8 (https://www.gtexportal.org/home/datasets). In addition, the ability to cluster the genetic effects on pathways (i.e. their molecular phenotypic attributes) may reveal a level of complexity in the manifestation of causality that was previously unknown.

Finally, the development of reference xQTL datasets for diseases of interest will strengthen our ability to interpret personalized genomes and will provide a valuable framework for the biological understanding of phenotypic variability and disease risk. For example, the ADSP xQTL project, a new collaborative effort across the ADSP Functional Genomics Consortium, is designed to generate a reference map of Alzheimer’s-related xQTLs to determine the effect of genetic variation on a variety of molecular traits. This approach will help us study the consequences of genetic variants identified from the ADSP, clarify the roles of different genes in Alzheimer’s, and find potential new biomarkers and therapeutic targets. The ADSP xQTL project leverages existing datasets from multiple omics layers including bulk and single-cell transcriptomics for eQTLs/sQTLs/ct-eQTLs, epigenomics for mQTLs, proteomics for pQTLs, and metabolomics for metaQTLs from multiple tissues including brain, CSF, and plasma. This xQTL reference map will be made available to the general scientific community in 2022.

CONCLUDING REMARKS

The growth of molecular-trait QTLs in recent years has provided researchers with a useful tool in unraveling the biology of complex disease. The increasingly broad coverage of these studies in terms of molecular-traits and tissue types studied is enabling the discovery of more disease-relevant genes and pathways through their integration with disease GWAS results. This process will likely aid in the discovery of novel treatment targets and biomarkers for complex diseases.

DATA AVAILABILITY

Data derived from public domain sources.