Molecular Systems, Human Non-Coding Sequence Variants, and Blood Pressure

Abstract

The human genome harbors millions of non-coding sequence variants. Genome-wide association studies (GWAS) have identified thousands of robust associations linking non-coding variants to human physiological traits and complex diseases. Integrative approaches, including expression quantitative trait locus mapping, epigenomic profiling, and precise genome editing in trait-relevant cell types, enable the identification of effector genes and underlying regulatory mechanisms, such as long-range chromatin interactions, that mediate the effects of non-coding variants. Investigations of blood pressure (BP)-associated non-coding sequence variants have uncovered previously unrecognized roles of genes in BP regulation, reinforced the human genetic relevance of established BP regulatory pathways, and elucidated specific regulatory mechanisms by which non-coding variants influence gene expression and BP. Studies of orthologous non-coding genomic regions in animal models corresponding to human genomic regions harboring BP-associated variants have demonstrated substantial effects on BP, suggesting that the phenotypic impact of non-coding sequence variants may be large within human subgroups. Continued expansion of functional studies of trait-associated non-coding sequence variants, together with advances in mapping molecular quantitative trait loci and epigenomic landscapes, will provide novel insights directly relevant to human biology and disease and essential for understanding humans as molecular systems.

Graphical Abstract

1. Introduction

Complex human traits and diseases are regulated by the interplay between genetic and environmental factors. Associations of common variations in the DNA sequence, such as single-nucleotide polymorphisms (SNPs), with human traits and diseases can be identified by approaches like genome-wide association study (GWAS). Tremendous progress has been made in GWAS in the last two decades¹. The GWAS Catalog, accessed on June 29, 2025, has curated 891,200 associations for more than 15,000 traits from 7,286 publications². Genetic discoveries including GWAS findings play a prominent role in driving the development of new drugs and repurposing of existing drugs^3–5. For example, the association of variants in IL23R (Interleukin 23 Receptor) with Crohn’s disease led to the repositioning of monoclonal antibodies targeting the IL-23 receptor as Crohn’s disease interventions^6,7.

Hypertension is the No. 1 identifiable risk factor for disease burden and deaths worldwide^8,9, affecting up to 45% of the adult population in the US¹⁰. Millions who take several anti-hypertensive medications remain hypertensive¹¹. Blood pressure (BP) traits, including systolic BP (SBP), diastolic BP (DBP), and pulse pressure, are classic quantitative complex traits. GWAS have identified significant associations of several thousand independent genetic signals with BP^2,12–14. SBP differed by 16.9 mmHg, and hypertension risk differed by more than 7-fold, between individuals in the top and bottom deciles of the polygenic risk scores developed by the largest single-stage BP GWAS to date, supporting biologically significant effects of common genetic variants on BP¹⁴. It is critical to understand the mechanisms by which common genetic variants influence complex human traits such as BP.

However, one of the greatest challenges in human genetics is that more than 90% of SNPs associated with complex traits are in non-coding parts of the DNA, and many SNPs are located far from any protein-coding gene, making it difficult to pinpoint the effector gene(s) that can link a SNP to a phenotype or to investigate the functional role and mechanisms of action for trait-associated SNPs^12,15–18. Research in the last several years has led to significant progress in addressing this challenge and improving our understanding of how non-coding DNA sequence variations may influence biology and disease including BP and related traits. Such studies have enriched physiological knowledge by uncovering new genomic and epigenomic mechanisms relevant to physiology, revealing previously unknown roles of genes and pathways, and demonstrating the human genetic relevance of known pathways.

In this review, we introduce the conceptual framework of humans as molecular systems and a tree-like paradigm for understanding complex traits, describe the basic types and characteristics of non-coding DNA sequence variations and the approaches for investigating the functional role and mechanisms of action for these variations, summarize recent advances in the study of non-coding DNA sequence variations in BP regulation and the development of hypertension and related diseases, and discuss current challenges and future directions of this cutting-edge research area.

2. Humans as molecular systems

According to the current curations by recognized databases (accessed on June 29, 2025), the human genome contains 19,435 protein-coding genes, 35,901 long non-coding RNA genes, and 7,563 small non-coding RNA genes (Figure 1)¹⁹. The expression of these genes is under the control of millions of regulatory elements in the genome, including 1,063,878 curated candidate cis-regulatory elements (cCREs)²⁰. These genes encode tens of thousands of transcripts and proteins^19,21. In addition, approximately 24,000 metabolites have been detected, and more than 200,000 annotated, in humans (Figure 1)²².

Figure 1. Humans as molecular systems.

Thousands of molecules and molecular features interact with each other to give rise to emergent properties of cells, tissues, organs, and the whole human body. Figure was created in part with a licensed version of Biorender.com.

These molecules and molecular features interact with each other extensively in the context of cells, tissues, organs, and the whole human body (Figure 1). Transcripts, proteins, and metabolites interact with genomic regulatory elements to control gene expression. Transcripts and proteins interact with each other, and metabolites and ions are linked to thousands of enzymes and transporters. Genomic regulatory elements almost always function in a combinatorial manner, and transcripts, proteins, and metabolites interact among themselves. Only a subset of these molecules and molecular features is active in a specific cell state.

Such interactions result in the formation of systems of molecules and give rise to emergent properties of cells, tissues, organs, and the whole human body (Figure 1)^23–25. Cellular function and tissue, organ, and whole-body physiology are emergent properties as they are properties of the systems instead of those of each molecule or molecular feature in isolation. It is essential to understand humans as molecular systems to fully understand human physiology. Non-coding sequence variants influence the genome function and are an essential part of the human molecular systems.

3. Tree-like paradigm for understanding the regulation of complex traits

Most human traits and diseases are complex and multifactorial, involving both genetic and environmental factors. The genetic variants involved in most complex traits or diseases can number in hundreds to thousands². Yet, the number of final physiological pathways or physiological parameters that determine a trait is usually small²³. The large number of genetic variants and other molecular changes, including epigenetic changes, involved in a complex trait, can be visualized as leaves on a tree that will converge gradually into small and then large branches (Figure 2). The small branches are cellular signaling and metabolic pathways, and the large branches are physiological pathways. In the case of BP, the pathways ultimately will influence cardiac output or peripheral vascular resistance, the final determinants of BP^24,25.

Figure 2. Tree-like paradigm for understanding the regulation of complex traits.

A complex trait may involve thousands of molecular features. These molecular features converge to influence a limited number of biological pathways to regulate a few physiological determinants of the trait. Examples of genes and pathways and physiological determinants regulating blood pressure are shown on the right.

It will be clear throughout this article that the concept of convergence embodied in this tree-like paradigm is essential for placing large numbers of genetic variants and other molecular changes into a framework for understanding the molecular systems underlying complex traits. Just as important, divergence, cross-talks and feedback loops within the tree or across trees, as well as influences of environmental factors, contribute to the function of the molecular systems (Figure 2).

4. Human non-coding sequence variants

Types of human non-coding sequence variants

The most common type of human non-coding sequence variants is single nucleotide variations (SNVs), including single nucleotide polymorphisms (SNPs). These variants involve a difference in a single nucleotide base pair across individuals (Table 1). An individual may have homozygous or heterozygous allelic sequences at a SNP or SNV. Other types of non-coding sequence variants may include indels, structural variants, and tandem repeats (Table 1). In a typical individual, there are 4–5 million SNVs, most of which are SNPs, several hundred thousand indels, thousands of structural variants, and millions of polymorphic tandem repeat loci^26,27.

Table 1.

Types of human non-coding sequence variants based on DNA alteration

Type	Description
Single Nucleotide Variants (SNVs) / Single Nucleotide Polymorphisms (SNPs)	The most common type, involving a change in a single nucleotide base pair. If the variant is present in at least 1% of the population, it’s typically referred to as an SNP.
Insertions and Deletions (Indels)	Addition (insertion) or removal (deletion) of a small number of base pairs (typically less than 50 bp)
Structural Variants (SVs)	Larger-scale rearrangements of the genome (greater than 50 bp), including deletions, duplications, inversions, translocations, and copy number variations (CNVs)
Tandem Repeats (Microsatellites)	Short stretches of nucleotides repeated multiple times in a row, with varying numbers of repeats between individuals

Human non-coding sequence variants may be in a variety of functional or potentially functional genomic elements, providing indications to how the variants may influence genomic function (Table 2). Variants in gene promoters may directly influence gene expression activities. Variants in enhancers and silencers can influence gene expression through their effects on promoter activities, sometimes through long-range chromatin looping. Insulators prevent enhancers from acting on certain genes or block the spread of repressive chromatin structures. Variants in introns and 5’ or 3’ untranslated regions of mRNA may influence mRNAs through a variety of mechanisms (Table 2). Variants in or around non-coding RNA genes may affect various non-coding RNA molecules, such as microRNAs and long non-coding RNAs (lncRNAs), which themselves have regulatory roles (Table 2).

Table 2.

Types of human non-coding sequence variants based on functional elements affected

Type	Description
Regulatory Elements	Sequences that control gene expression, including promoters, enhancers, silencers, and insulators
Introns	Transcribed but later removed during RNA splicing. Variants here can affect RNA splicing. Some introns may contain regulatory elements.
Untranslated Regions (UTRs)	Regions at the 5’ and 3’ ends of messenger RNA (mRNA) that are not translated into protein but are important for mRNA stability, translation efficiency, and localization
Non-coding RNA genes	Include transfer RNAs (tRNAs), ribosomal RNAs (rRNAs), microRNAs (miRNAs), long non-coding RNAs (lncRNAs), small nuclear RNAs (snRNAs) and small nucleolar RNAs (snoRNAs), and PIWI-interacting RNAs (piRNAs). Variants can affect the transcription, processing, or function of non-coding RNA molecules.
Other	Include pseudogenes, repetitive sequences (satellite DNA and transposons), origins of DNA replication, centromeres, and telomeres.

Functional and clinical significance of human non-coding sequence variants

The multitude of non-coding variants and the complexity of their involvement in human traits are emblematic of humans as molecular systems (Figure 1). Studies of non-coding sequence variants have opened a new dimension for understanding human biology, revealing numerous genomic and epigenomic mechanisms for the regulation of gene expression and cellular and physiological function. The 891,200 associations in the current GWAS Catalog involve tens of thousands of sequence variants, the vast majority of which are non-coding SNPs². Rarer non-coding variants including structural variants have also been associated with human traits²⁸. These non-coding variants sit at the top of the “tree” in the tree-like paradigm for understanding complex human traits (Figure 2). ENCODE (the Encyclopedia of DNA Elements project), Roadmap Epigenomics, EpiMap, GTEx (the Genotype Tissue Expression project), enhancer mapping, studies using techniques like massively parallel reporter assay (MPRA) and CRISPR screen, and other genome-scale studies have linked many non-coding variants with functional elements of the genome or with gene expression, beginning to reveal the contour of non-coding variants’ essential place within the tree of molecular networks underlying human traits^{18,20,29–33}. Several of these studies and techniques will be discussed in detail in later sections of this article.

In-depth studies of specific non-coding variants have led to numerous ground-breaking discoveries. One of the earliest examples of such studies involve the sequence variants in the promoter and introns of the HBB gene encoding the beta-globin chain of hemoglobin^34,35. These non-coding variants reduce the expression of HBB, causing beta-thalassemia. In an example of functional roles of non-coding sequence variants located far from protein-coding genes, a clustering of translocation breakpoints 1.06–1.23 Mb upstream of SOX9, microdeletions both approximately 1.5 Mb centromeric and approximately 1.5 Mb telomeric of SOX9, and a heterozygous point mutation within one of the microdeletion regions were found to underlie the development of Pierre Robin sequence, a subgroup of cleft palate³⁶. In vitro assays indicate that the mutation abrogates enhancer function and alters the binding of the transcription factor MSX1. In another example, somatic mutations in the core promoter of TERT, the gene encoding the catalytic subunit of telomerase, are found to occur in large percentages of multiple cancers. These mutations generate de novo consensus binding motifs for E-twenty-six transcription factors and increase transcriptional activity from the TERT promoter by several fold³⁷. Studies of specific non-coding variants that are relevant to the regulation of BP and related traits will be reviewed in detail in later sections of this article.

Importantly, genetic discoveries including discoveries on non-coding sequence variants can drive the development of new drugs and repurposing of existing drugs^3–5. For example, both non-coding and coding variants in IL23R were found to be associated with inflammatory bowel disease by GWAS⁶. These findings contributed to driving the repositioning of monoclonal antibodies targeting the IL-23 receptor as Crohn’s disease interventions⁷. In another example, the discovery of a synonymous sequence variant between SMN1 (encoding survival of motor neuron 2, centromeric) and its paralog SMN2 contributed to the identification of intronic splicing silencers in an intron of SMN2 mRNA as a therapeutic target for spinal muscular atrophy, leading to the development of Nusinersen, an antisense oligonucleotide as a new therapeutic for this disease^38,39. Although the variant in this case is not a non-coding variant, these studies demonstrate the potential of targeting non-coding genomic regions as a therapeutic approach. In more recent examples, several variants, including a non-coding variant upstream of ACE2 and a variant near TYK2, have been associated with the infection risk, disease severity, or critical illness of COVID-19^40,41. TYK2 encodes tyrosine kinase 2 and is a member of the Janus kinase (JAK) family. Baricitinib, an inhibitor of JAK1, JAK2, and TYK2, was tested in the RECOVERY trial and shown to reduce mortality in patients hospitalized for COVID-19⁴².

While strong examples have demonstrated the functional and clinical significance of non-coding sequence variants, the physiological role and mechanisms of action for most non-coding sequence variants remain to be investigated.

5. Approaches for investigating the role of non-coding sequence variants in physiology

Approaches for identifying and annotating trait-associated non-coding variants

The primary approach for identifying non-coding sequence variants relevant to a trait is GWAS (Table 3)^1,43. GWAS test statistical associations of common genetic variants with traits of interest in large populations. The vast majority of the genetic signals identified by GWAS are non-coding variants^12,15,18. The genetic signals are often detected by SNP arrays representing hundreds of thousands or more SNPs. Whole genome sequencing used to be too expensive for GWAS but has become more affordable and has been used in GWAS, which has the advantage of detecting more variants as well as large variants such as structural variants²⁸. Recent GWAS have been conducted in one million or more individuals, reaching saturation of common SNP-based heritability for some traits in some populations^1,44. A trait-associated genomic signal identified by GWAS may include several variants that are in high linkage disequilibrium (LD), meaning these variants are likely to be inherited together and any of them may causally contribute to the trait^15,45.

Table 3.

Approaches for identifying and annotating trait-relevant non-coding sequence variants

Technique	Description	Utility	Limitations
GWAS (Genome-Wide Association Studies)	Statistical association of common genetic variants with traits in large populations	Identifies non-coding variants linked to complex traits and diseases	Does not pinpoint causal variants or mechanisms; often identifies regions with many linked variants
Fine mapping	Statistical analysis to identify a minimum set of SNPs explaining an association signal	Narrows down the list of candidate non-coding variants	Limited by high linkage disequilibrium and small effect sizes
eQTL and other molecular QTL Mapping	Associates genetic variants with gene expression levels and other molecular features across individuals	Links non-coding variants to changes in gene expression and molecular features; identifies potential regulatory variants	Tissue- and context-dependent; effects on expression and other molecular features may be indirect
Chromatin Conformation Capture (e.g., Hi-C, Micro-C, Capture Micro-C)	Maps 3D genome architecture and physical contacts between regions	Links non-coding variants to distant gene promoters via chromatin looping	Requires high sequencing coverage; interaction may not mean regulation
Chromatin Accessibility Assays (e.g., ATAC-seq)	Profiles open chromatin regions where regulatory elements are likely to reside	Helps prioritize variants located in accessible regulatory regions	Does not directly measure function; may miss context-specific activity
Transcription factor binding and histone marks (e.g., ChIP-seq, CUT&Tag)	Profiles DNA segments bound with transcription factors or histone marks	Helps prioritize variants located in the binding regions	Does not directly measure function; may miss context-specific activity
Single-Cell and spatial Sequencing Analyses (e.g., scRNA-seq, single-cell epigenomic profiling, spatial omics)	Profiling cell type-specific gene expression and epigenomic features	Helps prioritize variants in cell type-specific context	Small sample size; challenges in cell type annotation without gene expression data
PrediXcan, S-PrediXcan, and other extensions	Computational methods for linking sequence variants to phenotypes via gene expression	Identifies genes that may mediate the effect of non-coding variants on traits	Limited by the availability of relevant gene expression data
COLOC	Bayesian statistical test of whether a GWAS signal and an eQTL share a causal variant	Generates the posterior probability of a trait and gene expression sharing a causal variant	Limited by the availability of relevant gene expression data
Deep learning models for sequence-based prediction (e.g., DeepSEA, Basset, Basenji, Enformer)	Predicts genomic regulatory function or impact of non-coding variants from DNA sequence only	In silico prioritization of functional variants based on learned sequence patterns	Model-dependent; predictions require validation

Once a non-coding genetic signal is associated with a trait, the next goals are to pinpoint the causal variants, identify the effector gene or genes (i.e., genes whose expression is directly influenced by the non-coding variants), understand the mechanisms by which the non-coding variants influence effector gene expression, and ascertain the effect of the non-coding variants and their effector genes on the trait and the physiological mechanisms involved. Genetic fine mapping can narrow down the list of potential causal variants via statistical testing to identify a minimum set of SNPs that can explain the association signal (Table 3)⁴⁶. Additional approaches are needed to achieve the above goals.

Multi-omic integration is a powerful approach for accomplishing several of these goals (Table 3). eQTL mapping, represented by the GTEx project, associates genetic variants with gene expression levels across individuals by measuring gene expression using techniques like RNA-seq and testing the associations of gene expression with genetic variants³⁰. Integration of eQTL findings with GWAS signals links trait-associated non-coding variants to changes in gene expression, potentially identifying causal variants and effector genes for these variants. Most eQTL studies have been performed in bulk tissues and tissues containing multiple cell types and may miss cell type-specific eQTLs. Cell type-specific eQTL analysis utilizing single-cell RNA-seq data is emerging^47,48. Following the same logic, QTLs can be identified for other molecular features ranging from protein and metabolite abundance to regulatory features like chromatin accessibility, DNA methylation, chromatin looping, and RNA splicing, providing further insights into the effector genes for trait-associated non-coding variants and the mechanisms by which the variants influence effector gene expression^49–56. However, genes and molecular features associated with a genetic variant may be influenced by the variant directly or indirectly.

Epigenomic mapping, even in the absence of population-based QTL analysis, can provide useful information for interpreting trait-associated non-coding variants. Chromatin conformation capture is particularly powerful for linking non-coding variants to distant genes (Table 3)^57,58. Chromatin conformation capture using techniques like Hi-C and Micro-C maps the 3D genome architecture and physical contacts between genomic regions^59–61. Many trait-associated non-coding variants are located 10s or 100s of kb from any protein-coding genes^12,15. Long-range chromatin interactions revealed by chromatin conformation capture provide plausible explanations for how non-coding variants regulate distant genes. Global chromatin conformation analysis requires high sequencing coverage, which can be mitigated by focusing the sequencing on a subset of interactions such as those involving gene promoters by incorporating a genomic region capturing step^62,63. Earlier versions of chromatin conformation capture technique allowed identification of one-to-one (3C) or one-to-many (4C) interactions⁶⁴. Techniques like ChIA-PET and HiChIP combine chromatin conformation capture with chromatin immunoprecipitation (ChIP), capturing chromatin interactions involving a specific protein such as a transcription factor or histone modification^65,66.

Other epigenomic mapping technique, such as ATAC-seq for identifying accessible chromatin regions and ChIP-seq or CUT&Tag for identifying genomic regions bound by specific transcription factors or histone marks, help to prioritize non-coding variants located in potentially regulatory regions of the genome and reveal regulatory mechanisms (e.g., specific transcription factors) that may be influenced by the variants (Table 3)^67–69. Electrophoretic mobility shift assays (EMSA) is a classic molecular biology technique for determining whether a protein (often a transcription factor), in the form of purified protein or cell nuclear extract, binds to a specific nucleic acid sequence^70,71. The functional relevance of any colocalization of non-coding variants with epigenomic features will need to be further validated.

Genome-scale analysis of gene expression and many epigenomic features can be performed in single cells or within the spatial context of tissues^72,73. With some exceptions, single-cell and spatial omics analysis often has small sample sizes, and it is challenging to annotate cell types when co-registered gene expression data is not available, e.g., in single-cell mapping of epigenomic features. However, the resulting cell type-specific data, sometimes with a spatial context, are powerful for understanding the role of trait-associated non-coding variants because the regulation of gene expression is highly cell type specific (Table 3)^20,30,74.

The efficiency of multi-omic integration for understanding non-coding sequence variants relies on computational methods. PrediXcan is a gene-based association method that estimates the component of gene expression that is determined by an individual’s genetic profile and correlates imputed gene expression with phenotypes, revealing genes potentially linking genetic variants, especially non-coding variants, to traits⁷⁵. Several extensions have been developed⁷⁶. Notably, S-PrediXcan achieves the goals of PrediXcan by utilizing summary statistics that are more widely available from GWAS than individual genetic profiles⁷⁷. Colocalization analysis using methods like COLOC can complement techniques like S-PrediXcan by testing the probability that the trait and the gene expression share a causal variant⁷⁸. These methods often use gene expression data from studies like GTEx to train machine learning models to predict gene expression levels from genetic variants and are, therefore, limited by the availability of relevant gene expression data as discussed above for eQTL analysis.

Several deep learning methods have been developed to predict genomic regulatory function and, in some cases, effects of specific non-coding sequence variants, based on DNA sequence only. DeepSEA learns a regulatory sequence code from large-scale chromatin-profiling data, primarily transcriptional factor and histone mark binding data and DNase I hypersensitivity data from ENCODE and Roadmap Epigenomics, and uses this code to predict chromatin effects of sequence alterations at single nucleotide level⁷⁹. Basset, and its subsequent version Basenji, predicts DNA accessibility and gene expression based on DNA sequence alone^80,81. Enformer integrates information of long-range interactions into the model, resulting in more accurate predictions of variant effects on gene expression⁸².

Techniques and models for investigating the functional role of non-coding variants

Associations of non-coding variants with traits, genes, and epigenomic features provide important insights into potential functional role and mechanisms of action for non-coding sequence variants, which need to be validated by targeted experimental manipulations. Genomic editing using techniques like CRISPR-Cas are essential for ascertaining the functional role and mechanisms of action for non-coding variants (Table 4). Precise editing that creates specific allelic sequences for a variant on a native, isogenic genomic background allows the most rigorous testing of the effect of the variant. Precise editing of specific SNPs can be achieved by techniques like prime editing that mitigate some of the issues related to double-strand breaks caused by CRISPR-Cas9⁸³. For haplotypes containing many SNPs spanning a large genomic segment, a two-step technique of deletion followed by reconstitution using CRISPR-Cas9 provides an efficient method¹⁵. Prime editing-based methods are also available for deleting and inserting genomic segments of several thousand bp, although they leave behind extra sequences^84–86. When precise editing is not practical, just deleting a small non-coding genomic segment around the SNP site can still be useful as it allows the testing of the effect of the epigenetic mechanisms that the SNP may influence, e.g., long-range chromatin interaction or transcription factor binding^15,87–89. Genomic editing studies can focus on specific loci or screen many loci.

Table 4.

Techniques and models for investigating functional role of non-coding sequence variants

Technique or model	Description	Utility	Limitations
CRISPR-based Perturbation (e.g., deletion, reconstitution, prime editing)	Perturbs regulatory elements to assess their functional impact on gene expression or phenotype	Causal validation of candidate non-coding variants or regions, depending on the precision of editing	Off-target effects; limited editing efficiency; requires robust cellular or in vivo models and readouts
MPRA (Massively Parallel Reporter Assays)	Tests the regulatory activity of thousands of DNA sequences in parallel	High-throughput identification of functional regulatory variants	Artificial context; limited by reporter system sensitivity
hiPSC	Endogenous or genome-edited variants; differentiated into trait-relevant cell types or organoids	Identifies effects of variants on gene expression and cellular function in trait-relevant cellular context	Quality of differentiation; limited in vitro readouts
Primary culture cells	Endogenous or genome-edited variants	Identifies effects of variants on gene expression and cellular function in trait-relevant cellular context	Mixed genomic background; loss of cellular characteristics; limited in vitro readouts
Animal models	Genetically engineered to model human sequence variants, followed by physiological study	Validation of a variant’s effect in a living system; enables studies of physiological mechanisms	Limited sequence and physiological conservation

Genomic editing studies need to rule out off-target effects by whole genome or targeted sequencing analysis. These studies need to consider limited editing efficiency. For example, screening of dozens to hundreds of cell clones is usually required to identify and establish cell clones with a precise, desired editing outcome. Moreover, having robust, trait-relevant readouts is essential for the success of a genomic editing experiment. Techniques like CRISPR screen in culture cells are powerful only when robust and meaningful cellular readouts are available for the trait of interest⁹⁰.

Non-coding variants may influence enhancer function. The enhancer function of a genomic segment containing a specific allelic sequence of a non-coding variant can be assessed by cloning the genomic segment into a reporter system, transfecting cells with the reporter system, and analyzing the reporter expression level. Such assay can be performed for thousands of variants using a massively parallel reporter assay (MPRA), identifying regulatory variants where the genomic segment shows allele-dependent enhancer activities (Table 4)^31,91. The assay tests the function of the variants in an artificial context and is limited by the sensitivity of the reporter system and the cellular environment in which the assay is performed.

The choice of cellular models is important for functional studies of non-coding variants (Table 4). The primary consideration is whether the system would provide a physiologically relevant cellular context resembling human tissues of interest. This is critical because the effect of a SNP on gene expression can be highly cell type-specific^20,30. Other considerations include ease of transfection if genome editing is involved and how well the cells would maintain their original cellular characteristics during a long and potentially stressful culture process including clonal expansion, which is necessary for obtaining homogeneous edited cells for downstream analysis. Performing genetic engineering in hiPSCs and gene expression and functional analysis in relevant cell types or tissue organoids differentiated from edited hiPSCs represents an excellent combination of technical feasibility and biological relevance^{15,88,89,92–94}. Primary culture cells can also be useful, although they may lose original cellular characteristics during culture, especially during the stressful process of transfection and clonal expansion. Additionally, with hiPSCs, genome editing only needs to be done once to generate all differentiated cell types of interest with the desired edits on the same genomic background.

Many traits cannot be directly or meaningfully assessed in culture cells or tissue organoids. Animal models are indispensable for investigating these traits and the integrated physiological mechanisms involved (Table 4). Using animal models to target effector genes for non-coding variants can ascertain the physiological effect and potential therapeutic value of the effector genes. For direct studies of human non-coding sequence variants, a significant challenge for using animal models is the limited sequence conservation. This challenge can be mitigated with several approaches. One approach is built upon the concept of epigenetic conservation. Epigenetic conservation refers to similar epigenetic features between species at the level of genomic locus, regulatory state, or regulatory logic. While the degree of conservation varies depending on the types of genetic and epigenetic properties, substantial similarities in epigenetic features such as patterns of transcription factor or histone mark binding, DNA methylation, and chromatin accessibility and three-dimensional conformation are known to exist between humans, rodents, and other species^12,95–99. Semi-conserved, orthologous non-coding genomic regions in rodents may influence biology through molecular mechanisms analogous to those mediating the effect of the human non-coding sequence variants, especially if these orthologous regions exhibit locus-level epigenetic conservation across human and rodents. Deleting these orthologous regions in rodents is, therefore, a valuable approach for testing the physiological role of the molecular mechanisms influenced by the human non-coding variants^15,87–89.

Other potential approaches for circumventing the limited sequence conservation between humans and rodents include inserting the human variants into rodent genomes and exploring the use of species with greater sequence and physiological conservation with humans such as non-human primates.

6. Blood pressure, hypertension, and hypertensive end-organ damage

The approaches discussed in preceding sections may be applied to investigate non-coding sequence variants associated with any human trait. We will focus on the investigation of BP in the remainder of this article. In this section, we provide a brief primer of BP and hypertension intended for readers who are not familiar with these topics and to provide physiological contexts for the genetic studies reviewed in later sections. For deeper reviews of BP and hypertension, readers may refer to other review articles^100–106.

Physiological mechanisms regulating BP

BP is the product of cardiac output and total peripheral vascular resistance, and cardiac output is the product to heart rate and stroke volume. BP is tightly regulated by a complex interplay of physiological mechanisms that influence cardiac output and/or total peripheral vascular resistance to ensure adequate blood flow to tissues while preventing damage to organ systems^101,104. Figure 3 depicts representative physiological mechanisms that regulate BP.

Figure 3. Physiological mechanisms regulating blood pressure.

Blood pressure is regulated by integrated mechanisms involving the nervous system, the heart, the kidneys, blood vessels, the endocrine system, and the immune system. Mechanisms depicted in this graph are representative and are not intended to be exhaustive. HR, heart rate; SNS, sympathetic nervous system; ADH, anti-diuretic hormone; ANP, BNP, and CNP, Atrial, B-type, and C-type natriuretic peptide; RAAS, renin-angiotensin-aldosterone system; ET-1, endothelin-1; ETA and ETB, endothelin-1 receptor A and B; NO, nitric oxide; Th17, T helper 17 cells; Treg, regulatory T cells; IL-6, IL-17A, and IL-10, interleukin-6/-17A/-10. Figure was created in part with a licensed version of Biorender.com.

A primary regulator of BP is baroreflex, a rapid-acting neural mechanism¹⁰⁷. Baroreceptors in the carotid arteries and aortic arch sense changes in BP and send signals to the brainstem. In response, the autonomic nervous system adjusts heart rate, cardiac contractility, and peripheral vascular resistance. For example, a drop in BP triggers sympathetic activation, leading to increased heart rate and vasoconstriction, thereby raising BP. Similarly, epinephrine and norepinephrine produced by the adrenal gland increase heart rate and cardiac contractility, constrict blood vessels, and stimulate renal reabsorption of water and sodium, increasing BP.

Several other mechanisms participate in the control of BP. A major mechanism is the renin-angiotensin-aldosterone system (RAAS)^101,108. When kidney blood flow or sodium delivery to the distal nephron in the kidneys decreases, the kidneys release renin, initiating a cascade that produces angiotensin II. Angiotensin II is a potent vasoconstrictor and stimulates aldosterone release from the adrenal glands, which promotes sodium and water reabsorption in the kidneys. Both actions of angiotensin II increase peripheral vascular resistance and blood volume and, consequently, BP. Antidiuretic hormone (ADH), synthesized in the hypothalamus and released from the posterior pituitary gland in response to low blood volume or high blood osmolarity, increases water reabsorption by the kidneys, increasing blood volume and BP. Atrial natriuretic peptide (ANP), released by the heart in response to high blood volume, promotes sodium and water excretion by the kidneys to lower BP. Nitric oxide produced by endothelial cells causes vasodilation, lowering BP. Endothelin-1 acts on endothelin receptor A in vascular smooth muscle cells to cause vasoconstriction and endothelin receptor B in the kidney to cause natriuresis. The immune system contributes to BP regulation by modulating vascular and renal functions. These are representative, but not an exhaustive list of, mechanisms that regulate BP (Figure 3).

Hypertension and hypertensive end-organ damage

Hypertension, or high BP, is a chronic medical condition where systemic BP is persistently elevated. The 2017 ACC/AHA guidelines defined hypertension as systolic BP (SBP) of ≥ 130 mmHg, diastolic BP (DBP) of ≥ 80 mmHg, or being on anti-hypertensive medications¹⁰. The pathophysiology of hypertension is complex and multifactorial (Figure 4)^{24,104,105,109}. It may involve increased sympathetic nervous system activity, leading to vasoconstriction and increased cardiac output. Overactivity of the RAAS may cause hypertension by promoting vasoconstriction (via angiotensin II) and sodium/water retention (via aldosterone), increasing blood volume and vascular resistance. Endothelial dysfunction, including impaired production of vasodilators like nitric oxide and increased release of vasoconstrictors, may lead to reduced vascular relaxation and hypertension. Vascular remodeling, i.e., structural changes in blood vessels, such as thickening of arterial walls and increased stiffness, may elevate peripheral resistance and lead to hypertension. Kidney dysfunction resulting in impaired ability of the kidneys to excrete sodium and water, may cause fluid overload and contribute to the development of hypertension. Immune mechanisms may contribute to or exacerbate the various mechanisms leading to hypertension¹¹⁰. Abnormalities in microbiota and cellular intermediary metabolism are also emerging as contributors to hypertension^109,111,112.

Figure 4. Hypertension and hypertensive end-organ damage.

Genetic and non-genetic factors contribute to the development of hypertension. These factors influence molecular mediators, leading to increased total peripheral vascular resistance or cardiac output. Hypertension is a leading risk factor for several major causes of morbidity and mortality. SNP, single-nucleotide polymorphism; TPR, total peripheral resistance. Figure was created in part with a licensed version of Biorender.com.

Affecting over one billion people globally, hypertension is a leading risk factor for cardiovascular disease and premature death^8,9,104. Sustained hypertension leads to progressive damage in various organs, known as hypertensive end-organ damage (Figure 4). In the heart, hypertension can cause left ventricular hypertrophy and contribute to heart failure (both preserved and reduced ejection fraction), coronary artery disease, and arrhythmias. In the brain, hypertension increased risk of ischemic and hemorrhagic stroke, transient ischemic attacks, and vascular dementia due to damage to cerebral blood vessels. In the kidneys, hypertension is one of the leading causes of the development and progression of chronic kidney disease (CKD), potentially leading to end-stage renal disease (ESRD). In the eyes, hypertensive retinopathy, characterized by damage to retinal blood vessels, can cause vision impairment or loss. In blood vessels, hypertension may accelerate atherosclerosis and contribute to peripheral artery disease and increased risk of aortic aneurysms and dissections.

Role of genetic factors in BP regulation and hypertension

Genetic predispositions, environmental factors like high sodium intake and physical inactivity, and comorbidities like obesity play significant roles in the development of hypertension (Figure 4)^24,106. Dozens of rare genetic mutations have been identified to cause Mendelian forms of BP abnormalities including hypertension and hypotension^113,114. Nearly all these mutations influence tubular reabsorption in the kidneys. The total pedigree-based heritability estimates for SBP, DBP and pulse pressure range from 24% to 30%¹¹⁵. The SNP-based heritability estimates for SBP, DBP and pulse pressure range from 16% to 19%, with the remaining heritability likely accounted for by rare variants^14,116,117. Common genetic variants that have been identified by GWAS as associated with BP can explain nearly 7% of SBP and DBP variation¹⁴. Polygenic risk scores constructed from all genetic variants analyzed are highly predictive of BP and hypertension risk: SBP differs by 16.9 mmHg, DBP by 10.3 mmHg, pulse pressure by 10.0 mmHg, and the sex-adjusted odds of hypertension risk by more than 7-fold between individuals in the top and bottom deciles of BP polygenic risk scores¹⁴. Common genetic variants also explain significant portions of the risk of developing diseases that hypertension may contribute to^118–120.

7. Non-coding sequence variants in blood pressure regulation and hypertension

Nearly all common genetic variants associated with BP and hypertension are non-coding variants^12,14,15. With few exceptions, the effect of these non-coding variants on BP or hypertension has not been explained by coding variants^121,122. Many of the common genetic variants associated with BP are in genomic regions 10s or 100s of kb away from any protein-coding gene^12,14,15. Integrated analysis of these common variants with other data types provides potential mechanisms for the action of these variants. The functional roles and mechanisms of action for several of these genetic signals have been experimentally validated, leading to new insights into the tree of molecular networks underlying BP regulation and the identification of potential new therapeutic targets for hypertension.

GWAS for BP

A single-stage GWAS involving 1,028,980 European individuals identified a total of 2,103 independent genetic signals associated with BP¹⁴. Polygenic risk scores (PRS) calculated by a method that integrates GWAS data with functional genomic annotations explains 11.37%, 12.12% and 7.30% of variance for SBP, DBP and pulse pressure, respectively. An earlier GWAS of over 1 million people of European ancestry identified 901 BP-associated loci in total¹²³. A trans-ethnic GWAS involving over 750,000 individuals including significant numbers of non-Hispanic blacks and Hispanics identified a total of 505 independent loci associated with one or more of the three BP traits¹³. The vast majority of BP-associated SNPs are non-coding SNPs^12,15.

Overview of functional roles and mechanisms of action for BP-associated non-coding sequence variants

Studies reviewed in the following sections, roughly in the chronological order of publications, have revealed the functional roles and mechanisms of action for several BP-associated non-coding sequence variants. An overview of these studies is provided in this section. Refer to the following sections for specific studies and citations.

A central goal of these studies is to identify the direct effector gene(s) for a BP-associated non-coding SNP or haplotype. Many studies rely on associating SNP allelic genotypes with gene expression across individuals to achieve this goal, while some studies provide definitive experimental evidence using precise genome editing. The mechanisms mediating the effect of the SNPs on their direct effector genes include allele-specific influences on genomic regulatory regions such as promoters and enhancers and long-range chromatin interactions. In studies that investigated multiple cell types, the effects of BP-associated non-coding SNPs often influence effector genes only in some cell types, highlighting the potential cell type specificity of these regulatory effects and their mechanisms.

Most studies of functional roles focus on the identified effector genes, demonstrating the influence of these genes on blood pressure in animal models. The identified effector genes often converge through intermediate phenotypes such as renal and vascular function. The physiological effect of deleting, overexpressing, or otherwise directly manipulating the genes may not represent the effect of the non-coding variants. The latter often have modest influence on the expression of their effector genes. Nevertheless, studies of effector genes can establish new components of biological pathways and identify potential new therapeutic targets.

A few studies have taken a step further to directly test and demonstrate the effect of non-coding genomic segments on phenotypes in animal models. Such studies may appear impractical given the small phenotypic effects reported by GWAS that are usually well below the detection limit of animal model studies. However, as illustrated by the studies reviewed below, the phenotypic effects in animal models can be substantial. Findings of such studies may reflect the physiological effects of the non-coding variants in subgroups of individuals as well as physiological effects of the epigenetic mechanisms impacted by the non-coding variants, providing great insights into the true molecular systems in humans.

In addition to targeted studies, broad analyses that integrate BP-associated SNPs with other types of omic data provide global views of potential mechanisms mediating the effects of the SNPs on BP and biological pathways affecting by the SNPs and a rich foundation for developing targeted functional and mechanistic studies.

UMOD and rs4293393 haplotype

rs4293393 is in LD with 28 SNPs in Europeans including rs13333226 and rs12917707, forming a haplotype spanning 14.8 kb (Figure 5)¹²⁴. The rs4293393 haplotype overlaps with the promoter region of UMOD (uromodulin), the expression of which is largely restricted to the thick ascending limb (TAL) of loop of Henle in the kidney^125,126. The TAL is responsible for reabsorbing approximately 20% of the filtered sodium primarily via NKCC2 (Na/K/2Cl cotransporter), which is the target of loop diuretics. Early GWAS using a case-control design identified significant associations between the rs13333226-G allele and a lower risk of hypertension, reduced urinary uromodulin excretion, better renal function, and reduction in risk of CVD events¹²⁷. Separately, rs4293393-T was found associated with CKD^128,129.

Figure 5. Blood pressure-associated rs4293393 risk haplotype increases UMOD expression, and greater UMOD expression causes salt-sensitive hypertension.

The rs4293393 haplotype overlaps with the UMOD promoter. Uromodulin, encoded by UMOD and expressed primarily in the thick ascending limb of Loop of Henle, increases NKCC2 activities and causes sodium retention. TAL, thick ascending limb; NKCC2, Na/K/2Cl co-transporter. Figure was created in part with a licensed version of Biorender.com.

Trudu, et al, found that the rs4293393 risk allele significantly increased UMOD transcriptional activity by about twofold in kidney cell lines in luciferase reporter assays¹³⁰. They showed that the UMOD risk variants were associated with a twofold higher UMOD transcript level in human nephrectomy samples, which was corroborated by a dose-dependent increase in urinary uromodulin levels in a large population-based cohort. Transgenic mice overexpressing uromodulin developed salt-sensitive hypertension, with blood pressure increasing with age. These mice also showed age-dependent renal lesions (tubular dilation, casts) and increased markers of kidney damage, mirroring findings in elderly human subjects homozygous for UMOD risk variants. Uromodulin overexpression in mice led to increased phosphorylation and activity of NKCC2 in the TAL. This effect was direct and dependent on membrane-anchored uromodulin and associated with increased activity of SPAK/OSR1 kinases. Hypertensive patients homozygous for the UMOD risk allele (rs4293393 TT) had higher baseline diastolic BP measured by 24-h ambulatory monitoring and showed an enhanced natriuretic and hypotensive response to furosemide, a NKCC2 inhibitor¹³⁰. Knockout of the Umod gene in mice resulted in 20 mmHg lower SBP at baseline, attenuation of salt-induced hypertension, and a leftward shift of the chronic renal function curve (pressure natriuresis curve)¹³¹.

A two-sample Mendelian randomization analysis indicates that part of the effect of urinary UMOD levels on blood pressure was mediated by estimated glomerular filtration rate (eGFR), whereas the effect on eGFR was not mediated by blood pressure¹³². A prospective clinical trial shows that, in 174 hypertensive subjects with normal baseline eGFR, treatment with the loop diuretic torsemide results in a greater and more consistent reduction of 24-hour ambulatory SBP in subjects with rs13333226-AA (−6.57 mmHg after 16 weeks of treatment) than rs13333226-AG/GG (−3.22 mmHg) genotypes¹³³.

These findings establish a likely causal link between common non-coding variants in the UMOD promoter region and uromodulin expression levels, and between uromodulin and the development of salt-sensitive hypertension and kidney damage (Figure 5). The mechanism involves uromodulin’s activation of NKCC2, which may lead to increased renal salt reabsorption and impaired pressure natriuresis (Figure 5). The non-coding variants may help to guide treatment choices for patients.

EDN1 and rs9349379

Gupta, et al, identified rs9349379, a SNP located in a non-coding region within the third intron of the PHACTR1 gene, as a likely causal variant associated with five vascular conditions including hypertension (Figure 6)⁹². The minor allele G is associated with increased risk of coronary artery disease and coronary calcification but with reduced risk of cervical dissection, migraine headache, fibromuscular dysplasia, and hypertension. This SNP resides in an enhancer marked by H3K27Ac in aortic tissues but not in non-vascular tissues analyzed by ENCODE. They found that rs9349379 regulated the expression of EDN1, a gene encoding the vasoconstrictor endothelin-1 (ET-1) and located over 600 kb away. Deleting an 88-bp region flanking rs9349379 in pluripotent stem cell-derived endothelial cells (ECs) and vascular smooth muscle cells (VSMCs) led to increased EDN1 expression, supporting its function as a repressive regulatory element. No differential expression was detected in a non-vascular cell type, neural crest progenitors. A set of isogenic human embryonic stem cell (ESC) lines were genome-edited using CRISPR-Cas9 to be either homozygous for the hypertension risk allele (A/A) or homozygous for the hypertension protective allele (G/G) at the rs9349379 locus. Differentiated vascular cells homozygous for G allele showed higher EDN1 expression and secreted more ET-1 protein. Moreover, individuals with the G allele exhibited elevated plasma levels of Big ET-1. There was no strong physical looping between rs9349379 and the EDN1 promoter based on a 4C-seq analysis, suggesting the regulation is not due to classical enhancer-promoter contact but may involve indirect long-range interactions, possibly via a super-enhancer located between the two loci⁹².

Figure 6. Hypertension-associated rs9349379 regulates EDN1 expression in vascular cells.

The hypertension-reducing allele G of rs9349379 increases EDN1 expression in vascular cells. Depending on the signaling pathways, ET-1 may cause vasoconstriction, vasodilation, or natriuresis. EDN1 and ET-1, endothelin-1; ETA and ETB, endothelin receptor A and B; NO, nitric oxide; NOS, nitric oxide synthase. Figure was created in part with a licensed version of Biorender.com.

In summary, this study shows that the non-coding SNP rs9349379 affects EDN1 expression, possibly via distal, tissue-specific enhancer activity, and that the G allele increases EDN1 expression and ET-1 protein production, which may contribute to the effect of the SNP on susceptibility to hypertension and other vascular diseases (Figure 6). The role of ET-1 in BP regulation, however, is complex (Figure 6)^134,135. In vascular smooth muscle cells, ET-1 is a potent vasoconstrictor acting on endothelin receptor A, leading to increased BP. But ET-1 can also act on endothelin receptor B in endothelial cells to cause vasodilation, lowering BP. In the kidneys, ET-1 acts on endothelin receptor B, causing natriuresis and lowering BP.

The role for the gene hosting the SNP, PHACTR1, in cardiovascular function remains unclear^136,137. rs9349379 is an eQTL for PHACTR1 in arterial tissues³⁰.

SLC4A7 and rs13082711

The non-coding SNP rs13082711 associated with BP is in LD with 125 SNPs, spanning a 134 kb region overlapping SLC4A7^124,138. SLC4A7 encodes NBCn1, an electroneutral Na⁺/HCO3⁻ co-transporter. Allelic expression imbalance analysis using rs13096477 in high LD with rs13082711, showed the risk (C) allele was associated with higher SLC4A7 expression in primary VSMCs and ECs¹³⁹. The risk allele increased the net base uptake rate and elevated steady-state intracellular pH. In the presence of Na⁺/H⁺ exchange activity, this effect persisted in VSMCs but not ECs. EMSA showed allele-dependent binding of nuclear proteins for some SNPs in the LD region, but not for the index SNP rs13082711. Bioinformatic analysis and overexpression experiments showed that the missense variant rs3755652 (Glu326Lys) in LD with rs13082711 was well-tolerated and did not affect NBCn1 activity¹³⁹.

A GWAS in Chinese identified a significant association between rs820430 and BP¹⁴⁰. rs820430 is not in LD with rs13082711 but has also been shown to affect SLC4A7 expression¹⁴¹. The rs820430-T allele increases c-Fos transcription factor binding.

PRDM6 and intronic SNPs

Several SNPs in intron 3 of PRDM6, such as rs13359291, are associated with BP, including in individuals of East Asian ancestry¹⁴². An MPRA analysis of 336 common variants in the third intron of PRDM6 in HEK293T cells identified 44 SNPs, including rs13359291, showing allele-specific regulation of PRDM6¹⁴³. Colocalization and other analyses identified several groups of SNPs in modest LD that may influence both BP and the expression of PRDM6 and suggested the presence of a super enhancer in intron 3 of PRDM6 that may bind several transcriptional factors including STAT1. The role of these non-coding variants and genomic regions in regulating PRDM6 expression was supported by a series of genomic segment deletion experiments in which deletions of a 22 kb genomic segment, the enhancer region, the STAT1 binding region, or genomic segments around specific SNPs resulted in downregulation of PRDM6 expression in HEK293T cells¹⁴³. The heterozygous disruption of Prdm6 in mice, driven by smooth muscle cell protein 22-α promoter (Prdm6^fl/+ SM22-Cre mice), did not change baseline BP but resulted in a higher number of renin-producing cells in the kidneys and the development of salt-induced hypertension that was responsive to the renin inhibitor aliskiren¹⁴³. Interestingly, PRDM6 appears to regulate a network of genes in the mouse aortas, including genes located in BP-associated loci, such as Sox6. Disruption of Sox6 reduced renin in Prdm6^fl/+ SM22-Cre mice¹⁴³.

NPR3, rs1173771 haplotype, and in vivo effects of the orthologous non-coding segment

Ren, et al, validated two independent BP-associated signals at the NPR3 (encoding natriuretic peptide receptor C, or NPR-C) locus using a GWAS involving 140,886 individuals of European ancestry from the UK Biobank cohort¹⁴⁴. They analyzed gene expression in primary arterial VSMCs from umbilical cords from 100 neonates and found that BP-elevating alleles within the rs1173771 LD block were associated with lower endogenous NPR3 mRNA and protein levels in VSMCs. The rs1173771 haplotype, spanning 17.4 kbp and containing 11 SNPs, is located approximately 23 kbp downstream of NPR3, the closest protein-coding gene, and 126 kbp from the NPR3 transcription start site (TSS) in the human genome, according to TopLD (Figure 7)^15,124. Furthermore, using formaldehyde-assisted isolation of regulatory elements (FAIRE), EMSA, ChIP, and luciferase reporter gene assays, they found that non-coding variants rs1173771 (intergenic) and rs1173747 (intronic) affected chromatin accessibility and nuclear protein binding, which may mediate the effect of the variants on NPR3 transcription¹⁴⁴. Functionally, the BP-elevating alleles were linked to increased VSMC proliferation. They also augmented angiotensin II-induced intracellular calcium flux and cell contraction in VSMCs¹⁴⁴. NPR-C has been shown to mediate the renal clearance of natriuretic peptides and the vascular effects of C-type natriuretic peptide (CNP)^145–147.

Figure 7. BP-associated rs1173771 haplotype regulates NPR3 expression, and the orthologous non-coding region influences vasoreactivity and the development of hypertension in Dahl salt-sensitive rats.

The allele-dependent effect of the haplotype on NPR3 expression is demonstrated by precise genome editing for the whole haplotype including 11 SNPs spanning 17.4 kb and may be mediated by long-range chromatin interaction. The in vivo effects were tested by deleting the orthologous non-coding region in Dahl salt-sensitive rats. NPR3 or Npr3, natriuretic peptide receptor 3 (encoding NPR-C); NPR-C, natriuretic peptide receptor C; CNP, C-type natriuretic peptide. Figure was created in part with a licensed version of Biorender.com.

Xue, et al, developed a two-step CRISPR-Cas9-mediated genome editing technique to precisely manipulate large haplotypes in hiPSCs and combined this with orthologous region deletion in animal models to study the physiological role and mechanisms of action for non-coding haplotypes (Figure 7)¹⁵. As a proof of principle, they applied this approach to the rs1173771 locus. In hiPSCs, deletion of the 17.4 kbp rs1173771 haplotype region resulted in significantly higher NPR3 mRNA expression in iPSC-derived ECs (iECs) and VSMCs (iVSMCs). Reconstituting either the BP-elevating or BP-lowering rs1173771 haplotype in hiPSCs showed that the BP-elevating haplotype led to significantly lower NPR3 expression in iECs and iVSMCs compared to the BP-lowering haplotype. Editing a single sentinel SNP, rs1173771, also demonstrated that the BP-elevating allele significantly lowered NPR3 expression in iECs and iVSMCs. Region Capture Micro-C analysis revealed that the BP-elevating rs1173771 haplotype increased chromatin contacts between the haplotype region and the NPR3 promoter in both iECs and iVSMCs, suggesting a long-range chromatin interaction mechanism for NPR3 suppression¹⁵.

The investigators then took the study a major step forward by deleting the orthologous non-coding region in Dahl salt-sensitive (SS) rats using CRISPR-Cas9 (Figure 7)¹⁵. They identified the orthologous region using comparative genomic analysis. In the rat genome, the orthologous region is about 14.5 kbp downstream of Npr3 and approximately 76 kbp from the annotated TSS of Npr3. The deletion attenuated salt-induced hypertension, resulting in a robust SBP effect of nearly 10 mmHg. This deletion led to an upregulation of Npr3 expression and NPR-C protein levels in mesenteric arteries and improved CNP-induced vasodilation¹⁵.

A major advance made by the study by Xue, et al, is it demonstrates that non-coding genomic regions located far from protein-coding genes can have substantial physiological effects on BP¹⁵. Most BP GWAS variants had been considered impossible to study in vivo or dismissed outright because their effect sizes reported by GWAS are small, often just a fraction of 1 mmHg. The effect of rs1173771 on SBP, for example, was 0.5 mmHg according to GWAS¹⁴⁸. The study by Xue, et al, shows much larger BP effects when a non-coding genomic region orthologous to the rs1173771 haplotype region is deleted in SS rats, enabling the investigation of physiological and molecular mechanisms involved in vivo (Figure 7)¹⁵. The genomic background of SS rats, which predisposes SS rats to the development of salt-induced hypertension, and the use of a high-salt diet as a “second hit” are likely key to detecting the robust BP effects. Importantly, these findings from animal models are consistent with the long-standing notion that in individuals with specific genomic backgrounds and exposed to certain environment and lifestyle factors, individual GWAS loci may have greater effects on the phenotypes than what the GWAS suggested.

NRIP1, rs1882961, and in vivo effects of the orthologous non-coding segment

rs1882961 is associated with systolic BP¹⁴⁹. Analysis of chromatin interactions with gene promoters in endothelia-denuded human arterioles, using pan-promoter capture Micro-C, indicates that the genomic region containing rs1882961 interacts with NRIP1 promoter region 119 kb away (Figure 8)⁸⁸. NRIP1 encodes nuclear receptor interacting protein 1, also known as receptor-interacting protein 140 (RIP14), a coregulatory for most nuclear receptors including retinoic acid receptor, estrogen receptor, and thyroid hormone receptor¹⁵⁰. HiPSCs were precisely edited to contain homozygous rs1882961-C (low-BP allele) or rs1882961-T (high-BP allele). NRIP1 and SAMSN1, another local gene, were significantly upregulated in iVSMCs with the high-BP allele of rs1882961 compared with the low-BP allele. Region-capture Micro-C analysis revealed greater chromatin interactions between the rs1882961 genomic region and the promoter regions of NRIP1 and SAMSN1, but not USP25, yet another local gene that was not differentially expressed, in iVSMCs with the high-BP allele than the low-BP allele⁸⁸.

Figure 8. BP-associated rs1882961 regulates NRIP1 expression in vascular smooth muscle cells, and the orthologous non-coding region influences hypertension in Dahl salt-sensitive rats.

rs1882961, for which T is the risk or high-BP allele, regulates the expression of NRIP1 and other local genes, possibly through long-range chromatin interactions. Deletion of a 4bp orthologous non-coding genomic region reduces salt-induced hypertension by up to 15 mmHg during the dark phase of the day in Dahl salt-sensitive rats. NRIP1 and Nrip1, nuclear receptor interacting protein 1; SAMSN1, SAM Domain, SH3 Domain and Nuclear Localization Signals 1. Figure was created in part with a licensed version of Biorender.com.

Remarkably, deletion of a 4-bp segment that encompasses the rs1882961 orthologous site from the genome of the SS rat, which is 112 kbp from the TSS of Nrip1, resulted in significantly lower mean arterial pressure, SBP, DBP, and pulse pressure, but not heart rate, in female rats on a high-salt (4% NaCl) diet (Figure 8)⁸⁸. The difference of mean arterial pressure reached approximately 15 mmHg during part of the dark phase of the day (active phase for rats). Nrip1 expression in mesenteric tissue, but not the aorta, was up-regulated in response to the high-salt diet but to a lesser extent in SS rats with the 4-bp deletion. NO donor-indued vasodilation of the third-order mesenteric arteries was significantly improved in SS rats with the 4-bp deletion.

These findings demonstrate that rs1882961 regulates NRIP1 expression via long-range chromatin interaction, which may influence resistance artery functions and BP, including the development of salt-induced hypertension (Figure 8).

MPRA analysis of BP-associated SNPs in cardiovascular cells

Oliveros, et al, used MPRA to analyze 4,608 genetic variants in linkage with 135 BP loci in cardiomyocytes differentiated from an hiPSC line and in VSMCs differentiated from hTERT-immortalized adipose derived primary human mesenchymal stem cells¹⁵¹. Several hundred variants were found to be regulatory variants as they exhibited significant allelic skewing. These regulatory variants included at least one variant in 91% and 63% of the loci analyzed in cardiomyocytes and VSMCs, respectively. Some loci contained multiple regulatory variants, including loci harboring genes that had been shown, or suggested, to be involved in BP regulation such as MAP4, PDE5A, CPEB4, ULK4, CFDP1, FBN1, and ESR1. Genomic regions containing regulatory variants also significantly interacted with regions harboring BP-related genes via long-range chromatin interactions. Prime editing in aneuploid HEK293 cells was used to further study three top ranked regulatory variants. The results indicated that rs4631439 in an intron of TRAPPC9 regulated the expression of KCNK9 located 345.4 kb upstream and rs3824754 at an ultra-conserved element boundary in the promoter region of BORCS7 and the promoter of lncRNA RP11–753C18 (lnc_CYP17A1) regulated SFXN2 111.1 kb upstream and PCGF6 448.2 kb downstream, possibly via long-range chromatin interactions¹⁵¹.

Broad integration of BP GWAS SNPs with other omic data

Beginning from previously reported GWAS meta-analyses of BP traits in up to 757,601 individuals of European ancestry, van Duijvenboden, et al, conducted annotation-informed fine-mapping incorporating tissue-specific chromatin segmentation and colocalization to identify causal variants and candidate effector genes for BP traits (Figure 9)¹⁵². They identified 959 candidate genes for SBP, 904 for DBP, and 774 for pulse pressure with at least one line of evidence from complementary fine-mapping and the presence of high-confidence missense variants, colocalized eQTLs (adipose, adrenal gland, artery, kidney cortex, heart, nerve, and brain in GTEx v8³⁰), Hi-C interactions (promoter capture Hi-C data from adrenal gland, dorsolateral prefrontal cortex, hippocampus, aorta, left ventricle, right ventricle, and fat¹⁵³), and pQTLs for plasma proteins⁵⁰. By incorporating additional evidence from known phenotypic effects in mice and humans, epigenomic findings, and expression data, they further prioritized 215 SBP, 205 DBP, and 202 pulse pressure genes, which together reflect 436 unique genes¹⁵².

Figure 9. Interpreting BP-associated variants using multiomic data, and gaps in this area.

Substantial progress has been made in utilizing a variety of omic data and approaches to annotate BP-associated sequence variants and candidate genes. Such analysis improves, on a genome-scale, the understanding of how BP-associated variants influence effector genes and how the effector genes may influence BP. Several significant gaps, however, limit the power and utility of such analysis.

Keaton, et al, used S-PrediXcan to integrate BP-associated SNPs that they identified from one million individuals of European ancestry with eQTL data in five tissues (aorta, tibial artery, left ventricle, atrial appendage and whole blood) from GTEx v.7 (Figure 9)^14,30,77. They incorporated covariance matrices based on 1000 Genomes²⁶ European populations to account for LD structure. They identified 5,538 statistically significant gene-tissue combinations that are genetically predictive of BP traits (SBP, DBP, pulse pressure). These combinations correspond to 1,873 unique genes, of which 569 (30%) are nearest genes. The majority of associations were observed in arterial tissues (n = 1,503 for tibial artery; n = 1,205 for aorta). Using COLOC⁷⁸, they detected 2,793 gene-tissue pairs in which there was a statistically significant S-PrediXcan association with at least one BP trait and high posterior probability of colocalization (i.e., a single variant underlies GWAS and eQTL associations at a given locus), corresponding to a total of 1,070 distinct genes. A FUMA analysis¹⁵⁴ of the 1,070 genes revealed a total of 4,617 unique significant terms across 20 different databases of functional annotations. Some identified gene ontology annotations included endoplasmic reticulum stress and carbohydrate and/or lipid metabolism¹⁴.

An analysis of 26,585 SNPs in 1,071 LD regions identified by multi-ethnic GWAS^13,123 indicates that 17,922 SNPs are eQTLs for 3,801 genes according to HaploReg (Figure 9)¹⁵⁵, which included data from all tissues in GTEx v6 and other studies¹². BP-associated sentinel SNPs are significantly enriched for eQTLs for the 251 BP physiology genes retrieved from Gene Ontology, compared to randomly selected independent SNPs¹². AGT, ERAP1, ERAP2, LNPEP, and NEDD4L are examples of BP physiology genes that are eQTL genes for BP-associated SNPs. 787 BP-associated LD regions and 179 BP physiology genes are in the same topological associating domains (TADs), and BP-associated SNPs are more likely to be in CTCF binding regions than what would be expected from the whole genome. 63% of BP-associated SNPs are in enhancer regions defined by ENCODE²⁰, 14% of which show significantly different enhancer or promoter activities between the two alleles of each SNP according to a Survey of Regulatory Element (SuRE) reporter assay¹⁵⁶. The SuRE was performed in K562 and HepG2 cell lines, which do not have direct relevance to blood pressure regulation. Among the BP-associated SNPs, 23,091 and 1,428 are in transcriptional factor binding sites and lncRNA promoter regions, respectively, and several are predicted to alter mRNA splicing¹². rs9337951, a synonymous SNP in JCAD, was predicted to affect the secondary structure of JCAD mRNA and experimentally confirmed to influence JCAD protein abundance without changing its mRNA expression. 9,447 BP-associated SNPs have identifiable syntenic regions in the mouse genome, suggesting it is possible to investigate these SNP regions in rodent models¹².

Findings from ENCODE and GTEx are useful for linking BP GWAS SNPs to genomic regulatory regions and eQTLs (Figure 9). Genomic regulatory features and eQTLs show significant tissue specificity^20,30, making findings from BP-relevant tissues especially valuable for understanding potential mechanisms of action for BP-associated SNPs. Utilizing eQTL data obtained from 356 tubular and 303 glomerular samples, taking into account predicted cell type fractions in each sample, Sheng, et al, prioritized 88 genes for the 340 GWAS loci associated with SBP¹³, where causal genetic variants associated with blood pressure and gene expression were shared¹⁵⁷. Cell type-interacting eQTLs obtained from analyzing this eQTL dataset highlighted effects of SBP-associated SNPs rs4292 and rs6687360 on ACE and AGT expression, respectively, in the proximal tubule, possibly via changes in transcription factor binding¹⁵⁷.

Eales, et al, identified kidney eQTLs from an analysis of kidney samples from 430 white European individuals collected after elective nephrectomies or before renal transplantation¹⁵⁸. Approximately 31% of independent BP GWAS loci that they analyzed (252 loci) contained kidney eQTL SNPs that are associated with 418 genes. SBP and DBP ranked as the top traits with the most significant enrichment of kidney eQTL SNPs among 24 GWAS phenotypes examined including cardiometabolic traits but not kidney function traits. Mendelian randomization analysis indicated that approximately 15% of BP GWAS signals may be driven by expression, alternative splicing or DNA methylation of a kidney gene¹⁵⁸. Most of these genes are not known to be involved in the physiological regulation of BP, human hypertension or the kidney.

Arterioles account for 60%−70% of total peripheral vascular resistance. SNPs associated with SBP, DBP and pulse pressure are consistently enriched in chromatin contact regions in human intact or endothelia-denuded arterioles identified by Micro-C or pan-promoter capture Micro-C analyses, compared to SNPs linked to non-arteriole-related traits⁸⁸. BP-associated SNPs form several thousand chromatin interactions with genes in arterioles, most of which have not been previously reported as locus genes for the SNPs or revealed through eQTL analysis by GTEx in large arteries⁸⁸. Moreover, BP-associated SNPs are more likely to be near expressed genes and in chromatin loops and accessible chromatin regions in iPSC-derived ECs and VSMCs⁹³.

Analysis of single-nucleus RNA-seq data from multiple organs from three major models of hypertension (mice treated with Ang II, SS rats, and SHR) has linked BP-associated SNPs to specific cell types and transcriptional programs broadly⁸⁹. By calculating the enrichment of SNP-related genes within the differential expression patterns of each cell type in the animal models, this analysis revealed expected associations, such as VSMCs from mesenteric artery and endothelial cells from the hypothalamus and kidney linked to multiple BP traits, as well as novel associations, including myelinating oligodendrocytes, astrocytes, and microglia from the hypothalamus linked to multiple BP traits, and left ventricle cardiomyocytes and fibroblasts linked to BP⁸⁹.

In summary, integration of GWAS loci with various omics data and findings from studies using techniques like MPRA suggest that BP GWAS SNPs may regulate hundreds of genes that do not yet have an established role in BP regulation (Figure 9). The regulatory effects are often cell type-specific and may sometimes be mediated by long-range chromatin interactions.

8. Non-coding sequence variants in hypertension-related end-organ damage

GWAS have identified many non-coding sequence variants associated with diseases that hypertension may contribute to¹⁵. Hypertension is not the only risk factor for these diseases, and with few exceptions, GWAS have not specifically investigated aspects of these diseases attributable to hypertension, although the connection with BP is sometimes examined. Here, we provide a brief overview of GWAS of stroke, heart failure, and CKD, highlighting non-coding variants and the connections to BP. This brief overview is intended to highlight potential roles of non-coding variants in the development of hypertension-related end-organ damage. The overview is not intended to be comprehensive for these diseases or imply that these diseases result specifically from hypertension.

A cross-ancestry GWAS meta-analyses of 110,182 patients who have had a stroke and 1,503,898 control individuals identified association signals for stroke and its subtypes at 89 independent loci¹¹⁸. Fine mapping identified 110 and 16 credible set-trait pairs in European and East Asian participants, each of which having a 95% posterior probability of containing a causal variant. Only 1.2% of the variants within credible sets were exonic variants for coding RNA. In silico mutagenesis analysis identified 78 predicted variant-transcript-model sets comprising 13 causal regulatory variants and 19 transcripts. An example is the G allele of rs12476527 in the 5′ UTR of KCNK3, which was predicted to increase KCNK3 expression in kidney cortex tubule cells and has been associated with higher SBP in addition to increased stroke risk. Among all vascular traits, SBP in Europeans and SBP and DBP in East Asians showed some of the strongest genetic correlations with stroke¹¹⁸.

A GWAS meta-analysis of all-cause heart failure in 207,346 individuals with and 2,151,210 individuals without heart failure identified 9,990 variants at 176 loci where genetic associations reached genome-wide significance¹¹⁹. Analysis of pleiotropic genetic associations identified several heart failure loci colocalized with BP loci. Approximately two-thirds of lead loci (115/176) did not contain a fine-mapped variant affecting a protein-coding sequence. GWAS-prioritized genes overlapped most substantially with SBP in addition to coronary artery disease, BMI and atrial fibrillation. Among carriers of truncating variants in TTN (TTNtv), which accounts for the majority of the rare-variant burden heritability of heart failure, the prevalence of heart failure ranged from 8.3% in the lowest PRS decile to 28.1% among the highest decile, suggesting that common-variant background modifies heart failure risk among carriers of rare pathogenic variants¹¹⁹.

Another GWAS meta-analysis of heart failure analyzed 1,946,349 individuals in the Heart Failure Molecular Epidemiology for Therapeutic Targets (HERMES) consortium¹⁵⁹. These individuals included 153,174 diagnosed cases of HF, of which 44,012 had a nonischemic etiology. Among these, 5,406 and 3,841 had nonischemic heart failure with reduced and preserved ejection fraction, respectively¹⁵⁹. The analysis identified 66 loci associated with heart failure, 46 (70%) of which were associated with at least one nonischemic subtype. Fine mapping identified 70 credible sets containing 547 putative causal variants at 47/66 HF loci. In Mendelian randomization analysis, genetically predicted high SBP and higher body mass index (BMI) were associated with risk of all HF phenotypes, with SBP having the largest effect on ni-HFrEF, while BMI having the greatest magnitude of effect on ni-HFpEF¹⁵⁹.

Large GWAS are limited for CKD but available for indices of kidney function such as GFR estimated based on serum creatinine (eGFRcrea). A multiancestry GWAS involving 2.2 million individuals identified 1,026 independent loci associated with eGFRcrea¹⁶⁰. Loci such as FGF5 are colocalized with BP traits. Integration of eGFRcrea GWAS signals with 32 types of data prioritized 24,437 regulatory variants targeting 1,060 genes. A convergence of coding and regulatory variations in specific genes was observed, with 601 of the 782 genes disrupted by 1363 coding variants also targeted by regulatory variants¹⁶⁰. An earlier GWAS involving 625,219 individuals including 64,164 CKD cases identified 23 genome-wide-significant loci associated with CKD¹²⁰. Lower genetic risk scores based on the combined effect of the 147 eGFRcrea index SNPs likely relevant for kidney function (as supported by association with BUN) were associated with higher odds ratios (ORs) of chronic renal failure, glomerular diseases, acute renal failure, as well as hypertensive diseases¹²⁰.

Some non-coding variants associated with these end-organ damage traits have been subjected to targeted functional and mechanistic studies. A chromosome 1 locus tagged by the SNP rs580698 is associated with heart failure¹⁶¹. The locus is close to ACTN2, encoding a structural cardiac protein inside the sarcolemma. Fine mapping identified more than 100 SNPs in LD at this locus, which overlaps with an enhancer region suggested by Roadmap Epigenomics and ENCODE and is in contact with ACTN2 promoter. This locus has not been identified as an eQTL for any nearby gene, and its association with heart failure cannot be explained by any coding variants in ACTN2 even though rare mutations in ACTN2 have been associated with cardiomyopathy and heart failure. Genomic deletion of this enhancer region resulted in approximately 50% reduction of ACTN2 during the differentiation of human embryonic stem cells to cardiomyocytes, supporting a role of the non-coding variants in regulating ACTN2 expression and, thereby, heart failure¹⁶¹.

9. Sex differences

Sex differences are well-recognized for BP regulation and the development, progression, and severity of hypertension and response to treatment^162–165. These clinical differences have been attributed to various factors, including sex chromosomes, hormones, sex differences in behavior and environmental exposures^164–167. Given the strong influence of sex on BP, it is likely that sex will influence the effect of SNPs on BP. Indeed, recent sex-stratified GWAS of BP using the UK Biobank resource have identified 412 and 142 loci that reach genome-wide significance in females and males only, respectively¹⁶⁸. In addition, several loci showed sexually dimorphic effects defined as differences in effect size and directions of the association with BP according to sex. Estrogen receptor 1 (ESR1) exhibited the highest level of differential TF binding site enrichment when comparing all sex-specific loci to non-sex-specific loci¹⁶⁸. In the All of Us cohort, females with low SBP polygenic risk scores are less likely, and females with high SBP polygenic risk scores are more likely, to develop hypertension, compared with their respective male counterparts¹⁶⁹. The interplay between sex and BP GWAS loci is an important topic for further investigation (Figure 9).

10. Gaps in knowledge and future directions

Non-coding sequence variants clearly play an important role in the regulation of human traits and diseases including BP, hypertension, and hypertension-related end-organ damage. However, critical knowledge gaps remain (Figure 9).

The functional role and mechanisms of action remain unknow for most trait-associated non-coding sequence variants other than the associations and correlations based on GWAS and other omic data. The small population-wide effect sizes reported by GWAS have discouraged investigators from studying individual variants. The largest allelic effect size for a GWAS signal for SBP is 1.23 mmHg, and the median allelic effect size for the top 100 GWAS signals for SBP is 0.46 mmHg (Figure 10)¹⁴. However, combined effects of genetic variants, as reflected by polygenetic risk scores, are large (Figure 10)¹⁴. Importantly, recent studies have demonstrated that non-coding genomic segments in animal models orthologous to human genomic segments harboring trait-associated sequence variants can have much larger phenotypic effects than GWAS suggest ^15,88,89. The BP effects of these individual non-coding genomic segments, which are observed in animal models with permissive genomic backgrounds and exposed to informative stimuli, are comparable with or exceed the effects observed in clinical trials of anti-hypertensive medications and reducing daily sodium intake from 150 to 50 mmol (Figure 10)^170,171. Targeted functional and mechanistic studies of individual trait-associated genetic signals, as well as their effector genes including their therapeutic potential, is likely a fruitful research direction as GWAS have already provided evidence supporting the relevance of these genetic signals and genes to human biology or disease. Such studies can be carried out using robust interventional techniques and versatile systems like precise genome editing in hiPSCs, which can be complemented and enhanced by studies in model organisms in vivo when possible.

Figure 10. Non-coding genomic segments can have substantial effects on BP.

While population-wide effects of individual GWAS signals on BP are small, combined effects of common variants, as illustrated by polygenic risk scores, are large. Moreover, BP effects of non-coding genomic regions orthologous to the human genomic regions harboring single GWAS signals can be large in animal models with permissive genome backgrounds and exposed to informative stimuli such as high-salt diets. The sizes of these effects are comparable with or exceed the effects of anti-hypertensive medications and reducing daily sodium intake from 150 to 50 mmol in clinical trials. See text for references.

While targeted experimental studies of individual variants remain essential for advancing our understanding of how non-coding variants influence biology, one of the next frontiers is to investigate combinatorial effects of variants. Given the high frequency of the risk allele for many disease-associated SNPs, individuals likely harbor the risk alleles of multiple SNPs. As such, combinatorial effects of variants may better represent real-life effects of variants than effects of individual variants. Although the concept of epistasis was proposed more than 100 years ago and extensive evidence for epistasis has been found in yeast and other model organisms, the magnitude of the role of epistasis in human genetics is less certain^172–175. Nonetheless, multiple SNPs have been reported to show combinatorial effects on enhancers that influence a common target gene¹⁷⁶. In addition, combinatorial effects of multiple GWAS signals that transcend small effects of individual GWAS signals have been reported, for example, for psoriasis¹⁷⁷. New approaches for prioritizing specific SNP combinations from a practically infinite number of possible combinations^178,179 and for experimentally investigating such SNP combinations will likely revolutionize the understanding of the effects of common genetic variants and their utility for precision medicine.

Just as SNP x SNP and gene x gene interactions may alter the effects of individual SNPs or genes on a trait, SNP (gene) x environment interactions have a well-established role in determining and regulating complex traits¹⁸⁰. The importance of SNP (gene) x environment interactions is highlighted by the studies discussed earlier in this article where deletion of a non-coding segment of a rat genome orthologous to a human genomic segment harboring a BP-associated non-coding SNP substantially changes BP in the rat model. In both the study of the rs1173771 haplotype region and the study of rs1882961, the BP effects in the rat model become prominent only when the rats are exposed to a high-salt diet^15,88. It will be challenging to identify the environmental setting that amplifies the effect of a specific SNP or combination of SNPs. However, gene–environment-wide association studies may detect the role of some SNP (gene) x environment interactions in regulating human traits¹⁸¹. Targeted studies of non-coding variants in cell or animal models should continue to consider and incorporate informative environmental stimuli.

The value of eQTL data for identifying effector genes of non-coding sequence variants is abundantly clear from the studies reviewed throughout this article. Yet, the eQTL data currently available is highly limited. While eQTL data is available for dozens of tissues, the data is lacking for some tissues that are highly relevant to BP and hypertension, such as resistance arteries including arterioles. Functional genomic insights obtained from the analysis of large arteries may not be applicable to arterioles, leaving a critical knowledge gap⁸⁸. Even when eQTL data is available, data replication has been absent or limited. In addition, most of the available eQTL data was obtained from the analysis of bulk tissues and tissues containing multiple cell types^30,157. The expression of many genes and the regulation of gene expression are highly cell type specific, suggesting that bulk tissue-based studies likely miss many meaningful eQTLs^20,29. Cell type-specific eQTL data is emerging but remains scarce⁴⁷. Furthermore, pQTLs allow the linking of non-coding sequence variants to protein abundance that is more physiologically relevant than RNA abundance, but nearly all currently available pQTLs are based on protein abundance in plasma or serum^{49,50,182–184}. Very few studies have examined pQTLs in tissues. Identification of eQTLs in trait-relevant tissues, cell type-specific eQTLs, and tissue pQTLs is urgently needed and will transform the understanding of how trait-associated non-coding sequence variants influence the trait.

Similarly, epigenomic data has been highly valuable for understanding the mechanisms of action for non-coding sequence variants but will benefit from improved data quality, tissue relevance to traits of interest, and cell type specificity^20,29. Single-cell techniques for analyzing chromatin conformation, DNA accessibility, DNA methylation, and transcription factor and histone mark binding profiles, sometimes jointly, are evolving^69,185–189. Improvement of these techniques and their application to analyze trait-relevant tissues will significantly advance the understanding of non-coding sequence variants associated with BP and other traits. Moreover, data analysis methods are critical for gaining meaningful insights from epigenomic data in conjunction with GWAS and other data types. New methods for integrating and parsing epigenomic and other data types, including methods that incorporate deep learning as well as dynamic modeling, will be essential for driving the study of non-coding sequence variants forward^25,190–193.

A major promise of studying the genetics of complex diseases is it may provide a novel basis for tailoring treatments to patients’ genetic characteristics, which may improve treatment efficacy and reduce adverse effects¹⁹⁴. Sequence variants, including non-coding variants, could alter drug metabolism, transport, or drug targets, leading to clinically meaningful differences in treatment outcomes^195,196. The finding discussed earlier in this article that patients with the rs13333226-AA variant, which elevates UMOD expression and NKCC2 activity, are more sensitive to the BP-lowering effect of a loop diuretic is an example of that¹³³. In addition to pharmacogenetic and pharmacogenomic applications, the genetics of complex diseases, including polygenic risk scores, may provide new opportunities to recognize disease subtypes and elucidate the mechanisms underlying the subtypes, leading to improved, more precise disease management beyond drug treatment. Success in achieving these translational goals will require progress in several areas including accurate and comprehensive assessment of non-genetic factors influencing individual health and integration of genetic and non-genetic data across large populations. A better understanding of the functional roles and mechanisms of action for disease-associated non-coding variants will contribute to achieving these goals.

11. Summary and conclusions

In summary, the human genome harbors millions of non-coding sequence variants. Thousands of these variants are associated with human physiological traits and diseases. These variants collectively explain substantial fractions of trait variability. Population-wide phenotypic effects of individual variants are often small but combined effects of sequence variants are large. Moreover, studies of orthologous non-coding genomic regions in animal models suggest phenotypic effects of individual non-coding sequence variants could be large in subgroups of humans.

Several approaches, including eQTL mapping, can link non-coding sequence variants to genes that the variants may regulate. Precise editing, followed by gene expression and functional analysis in trait-relevant cellular and organismal contexts, enables definitive testing of the functional role and mechanisms of action for non-coding sequence variants. Non-coding sequence variants can influence gene expression by altering the activities of gene promoters, enhancers, RNA splicing, and other mechanisms. For variants located far from the genes that they regulate, long-range chromatin interactions between the variant sites and gene promoters provide a plausible mechanistic basis.

Several non-coding sequence variants associated with BP have been investigated with targeted experiments. These investigations uncover previously unrecognized roles of genes in BP regulation or the human genetic implications of known BP regulatory pathways and reveal mechanisms by which non-coding variants regulate genes and BP. Prominent examples include the findings that BP-associated variants in the promoter influence UMOD expression, and UMOD influences BP and the development of salt-induced hypertension, probably by regulating NKCC2. Distant variants associated with BP regulate the expression of NPR3 in vascular cells, possibly via long-range chromatin interactions. An orthologous non-coding genomic region regulates Npr3 expression, alter vasoreactivity, and influences SBP by nearly 10 mmHg in SS rats fed a high-salt diet. Genome-scale analysis and data integration points to hundreds of genes and dozens of signaling and metabolic pathways that may mediate the effect of non-coding variants on BP, contributing to the tree-like regulatory network that underlies BP regulation.

Future studies of functional roles and mechanisms of action for non-coding sequence variants, including those associated with BP, hypertension, and hypertensive end-organ damage, will continue to elucidate the human molecular systems. Precise investigations of individual variants, as well as investigations at the new frontier of combinatorial effect of variants, are poised to provide novel insights directly relevant to human biology and disease. Advances in mapping molecular QTLs and epigenomic landscapes, especially at the cell type-specific level, and cutting-edge computational methods, will be critical for driving forward this exciting field of research.

Clinical highlights.

• The GWAS Catalog, accessed on June 29, 2025, has curated 891,200 associations for more than 15,000 traits from 7,286 publications. The vast majority of these associations involve non-coding sequence variants.

• Genetic discoveries including GWAS findings play a prominent role in driving the development of new drugs and repurposing of existing drugs.

• Population-wide phenotypic effects of individual variants are often small, but combined effects of sequence variants are large. Moreover, studies of orthologous non-coding genomic regions in animal models suggest phenotypic effects of individual non-coding sequence variants may be large in subgroups of humans.

• Several non-coding sequence variants associated with BP have been investigated with targeted experiments. These investigations uncover previously unrecognized roles of genes in BP regulation, highlight the human genetic relevance of established BP regulatory pathways, and reveal mechanisms by which non-coding variants regulate genes and BP.

• BP-associated variants in the promoter influence UMOD (Uromodulin) expression, and UMOD influences BP and the development of salt-induced hypertension, probably by regulating NKCC2. These variants influence treatment response to loop diuretics.

• A non-coding genomic region orthologous to the human genomic region harboring BP-associated variants that regulate NPR3 (Natriuretic Peptide Receptor 3) influences systolic BP by nearly 10 mmHg in Dahl salt-sensitive rats fed a high-salt diet.