Hypertension genetics past, present and future applications

Essential hypertension is a complex trait where the underlying aetiology is not completely understood. Left untreated it increases the risk of severe health complications including cardiovascular and renal disease. It is almost 15 years since the first genome‐wide association study for hypertension, and after a slow start there are now over 1000 blood pressure (BP) loci explaining ∼6% of the single nucleotide polymorphism‐based heritability. Success in discovery of hypertension genes has provided new pathological insights and drug discovery opportunities and translated to the development of BP genetic risk scores (GRSs), facilitating population disease risk stratification. Comparing highest and lowest risk groups shows differences of 12.9 mm Hg in systolic‐BP with significant differences in risk of hypertension, stroke, cardiovascular disease and myocardial infarction. GRSs are also being trialled in antihypertensive drug responses. Drug targets identified include NPR1, for which an agonist drug is currently in clinical trials. Identification of variants at the PHACTR1 locus provided insights into regulation of EDN1 in the endothelin pathway, which is aiding the development of endothelin receptor EDNRA antagonists. Drug re‐purposing opportunities, including SLC5A1 and canagliflozin (a type‐2 diabetes drug), are also being identified. In this review, we present key studies from the past, highlight current avenues of research and look to the future focusing on gene discovery, epigenetics, gene‐environment interactions, GRSs and drug discovery. We evaluate limitations affecting BP genetics, including ancestry bias and discuss streamlining of drug target discovery and applications for treating and preventing hypertension, which will contribute to tailored precision medicine for patients.

Essential hypertension is a complex trait where the underlying aetiology is not completely understood. Left untreated it increases the risk of severe health complications including cardiovascular and renal disease. It is almost 15 years since the first genome-wide association study for hypertension, and after a slow start there are now over 1000 blood pressure (BP) loci explaining ∼6% of the single nucleotide polymorphism-based heritability. Success in discovery of hypertension genes has provided new pathological insights and drug discovery opportunities and translated to the development of BP genetic risk scores (GRSs), facilitating population disease risk stratification. Comparing highest and lowest risk groups shows differences of 12.9 mm Hg in systolic-BP with significant differences in risk of hypertension, stroke, cardiovascular disease and myocardial infarction. GRSs are also being trialled in antihypertensive drug responses. Drug targets identified include NPR1, for which an agonist drug is currently in clinical trials. Identification of variants at the PHACTR1 locus provided insights into regulation of EDN1 in the endothelin pathway, which is aiding the development of endothelin receptor EDNRA antagonists. Drug repurposing opportunities, including SLC5A1 and canagliflozin (a type-2 diabetes drug), are also being identified. In this review, we present key studies from the past, highlight current avenues of research and look to the future focusing on gene discovery, epigenetics, gene-environment interac-

Introduction
Hypertension is the leading global risk factor for morbidity and mortality, and studies have demonstrated a clear link between elevated systolic and diastolic blood pressure (BP) and cardiovascular disease (CVD) [1]. In 2015, around 7.8 million deaths were attributed to hypertension [2], and the number of individuals diagnosed with hypertension is estimated to reach 1.5 billion globally by 2025 [3]. The current guidelines proposed by the European Society of Cardiology characterize grade 1 hypertension clinically as ≥140/90 mm Hg in patients below 80 years [4], while the clini-* These authors contributed equally to this work. cal threshold defining hypertension in the United States is lower at ≥130/80 mm Hg, as stated by the American Heart Association [5]. Approximately 95% of hypertensive cases are termed as essential hypertension (EH); hypertension with an unknown cause resulting from interplay of environmental and genetic factors. The remaining 5% of cases are grouped as secondary hypertension, of which 1% are monogenic disorders [6]. Despite extensive research, BP regulation mechanisms and hypertension pathophysiology remain poorly understood, and there are issues in varied patient response and adherence to current pharmacological therapies. For decades, genomic research into hypertension and BP has provided clues into its complex genetic architecture in a bid to identify novel target mechanisms for therapeutics and personalized medicines.

History of BP genetics, early studies and results
In 1949, Page documented the multifaceted nature of EH as a result of the dysregulation of four integral systems: cardiovascular, renal, endocrine and neural [7]. The genetic contribution to hypertension was recognized 32 years later in his revision of the Mosaic Theory of Hypertension [8], following evidence from a multitude of familial studies [9][10][11] and the characterization of rare monogenic disorders of hypertension [12,13]. With a consensus on the existence of a genetic component of BP, the Platt versus Pickering debate of the 1950s considered whether hypertension was a monogenic or polygenic disorder [14]. Platt argued that rare monogenic disorders of hypertension were evidence for a monogenic nature. In contrast, Pickering postulated the Gaussian, rather than bimodal, distribution of BP throughout the population suggested BP is determined by a collection of genes and further recognized hypertension as a quantitative trait with a normal distribution, opposing the previous suggestion of hypertension as a qualitative trait distinct to normotensive BP [14]. Later studies supported the polygenic nature of hypertension and estimated the heritability of clinical systolic BP (SBP) and diastolic BP (DBP) to be 15%-40% and 15%-30% respectively [9], with rare monogenic disorders representing an extreme end of the distribution.
The Human Genome Project provided the catalyst for advances in gene mapping in the 1990s [15]. Linkage analysis was a key tool in the early years, where microsatellite genetic markers were tested for co-segregation with a trait in families, the results of these studies provided chromosomal locations of genes for traits [6]. Investigation into families with monogenic disorders of hypertension using linkage analysis facilitated the identification of the first BP-associated genes and highlighted the role of renal and adrenal pathways in BP control, as covered extensively by Raina et al. [12]. Lifton and colleagues were responsible for the majority of research into monogenic disorders of hypertension in the 1990s -of note, research on patients with Liddle syndrome identified gain-offunction mutations in SCNN1B and SCNN1G genes encoding subunits of the epithelial sodium channel (ENaC) present on collecting ducts of kidneys, establishing the role of renal sodium reabsorption in BP control [13]. Yet monogenic forms of hypertension account for a very small percentage of hypertensive cases, and in a bid to elucidate genes involved in polygenic forms of hypertension a series of candidate gene linkage studies were carried out in familial cohorts [16][17][18][19][20][21][22]. These studies yielded some promising results; however, these were often contradictory across cohorts with no single candidate gene consistently showing strong linkage with hypertension. Notably, variants of the AGT gene encoding angiotensinogen, a key player in the renin-angiotensin system of BP control, were linked to hypertension in a linkage analysis of 63 White European families [19]; this result was not replicated in a larger European study of 350 families [20]. Linkage at this locus was subsequently demonstrated across populations, in African-Caribbean [18], Mexican-American [17] and Japanese [21] cohorts, albeit a different variant, but not in others [16], highlighting there may be potential variation between ethnicities. Candidate gene studies were generally underpowered as they relied on familial cohorts for which recruitment is difficult, and there was a lack of replication data, especially from non-White European populations [23]. Furthermore, candidate genes were selected based on previously characterized BP pathways restricting the identification of novel BP genes. At the turn of the 21st century, genome-wide linkage analyses were deployed with the aim of identifying loci anywhere in the genome, a hypothesis-free approach. Various genome-wide linkage analyses were undertaken in relatively large cohorts including the Framingham Heart Study [24], the Family BP Program [25,26] and the British Genetics of Hypertension (BRIGHT) study [27]. These studies successfully identified a number of quantitative trait loci (QTLs) (regions of DNA linked to variations in the phenotype) associated with hypertension, some of which were validated in follow-up studies [28][29][30]. The identification of broad QTL regions on five chromosomes led the BRIGHT study to propose hypertension as an 'oligogenic' disorder, in which a small number of genes located in these regions provide the largest effect on the trait, with additional genes exerting smaller effects [27]. Nonetheless, interpreting linkage analysis results presented challenges and limitations; the QTLs identified spanned broad regions of DNA making the identification of the responsible gene difficult [6], and there was a lack of power to identify variants with smaller effects [31].
Alongside studies in human populations, rodent models have also been studied, and these have provided a valuable resource for understanding the polygenic nature of EH. Importantly, genetically modified mouse models have been instrumental for functional analysis and validation of BP candidate genes [32]. The inbred spontaneous hypertensive rat (SHR) strains have been effective for identifying novel genes and also providing useful physiological models of CVD, as well as a valuable tool for analysis of therapeutic efficacy and toxicity of candidate drugs [33]. The SHR family is composed of several lines of selectively inbred rats, each carrying a different genotype and expressing combinations of traits observed in human EH, as a whole mimicking the human clinical phenotype [33][34][35][36][37]. Prior to the complete sequencing of the rat genome [32], linkage analysis in these rat strains identified over 270 QTL regions associated with hypertensive traits, predominantly located on four chromosomes [34], in line with the oligogenic theory of hypertension [27]. Studies in the rat have some benefits over human studies as the genetic heterogeneity of rat strains, and their environment are easily controllable, increasing study power. Nonetheless, similar to linkage analysis in humans, these studies were limited by their sample sizes, and mapping the QTL regions to identify candidate genes remained a difficult task. Concurrent sequencing of the rat [38] and human [15] genomes provided an opportunity for comparative mapping between species, facilitating more accurate translation of QTL regions from rat studies to humans [33]. We now know that some QTL regions identified with linkage analysis in humans validate in congenic rat strains, this is a topic covered extensively by Padmanabhan and Joe in 2017 [39]. Notably, a metaanalysis of microarray data from SHR identified several genes associated with BP in the rat [40]. A number of the identified genes had previously been associated with hypertension in humans via linkage analysis or genome-wide scanning, including APOE [41], NPPA and NPPB [42], corroborating the role of these genes in BP control.

The GWAS era -common and rare variant discovery
The complete sequencing of the human genome in 2003 [15], paired with the development of single nucleotide polymorphism (SNP) chip arrays, enabled cost-efficient high-throughput genotyping of selected variants and powered the first genomewide association studies (GWAS) [43][44][45][46][47][48]. SNPs are single base variations in the genome which occur at different frequencies in the population. In GWAS, SNPs distributed across the genome are tested for their association with traits or diseases. GWASs have benefits over linkage analysis and candidate gene studies as they are unbiased, permit larger sample sizes and enable meta-analyses improving statistical power [49]. The first GWASs for hypertension yielded disappointing results, and no significant loci were found [43,48]. A year later in 2008, the first locus (ATP2B1) significantly associated with BP was identified in a GWAS of 1484 Japanese individuals [50]. This result was replicated in a Korean cohort of >8000 individuals [51] and in a large European ancestry cohort of nearly 30,000 individuals [45]. Over the following years, many investigators using GWAS have identified loci for the quantitative traits of systolic, diastolic and pulse pressure [52], facilitating the discovery of novel BP pathways.
SNP arrays include only a small proportion of the variants present in the genome. The development of SNP reference panels, including the 1000 Genomes Project [53] and the Haplotype Reference Consortium [54], enabled imputation and estimation of the effects of associated SNPs not featured on arrays. In tandem with SNP arrays, the establishment of accessible large-scale Biobanks (e.g., UK Biobank) containing not only genetic data but a variety of phenotypic and health-related data [55] have improved the statistical power of GWAS and enabled the detection of both common and rare BP variants [41,[56][57][58]. The first GWASs were powered for the detection of common variants; these have relatively small effect sizes on BP ranging from 0.5 to 1 mm Hg per allele. With increasing access to samples, imputation and sequencing data, investigators have also focused efforts on identifying low frequency and rare variants (minor allele frequency < 1%), which have larger effect sizes (around 1.5 mm Hg per allele). In 2011, the Exome chip was launched [59]; an array of predominantly rare and low-frequency variants mostly located in exonic (coding) regions. Four years later, a trans-ethnic meta-analysis of Exome chip data with replication in European and South Asian ancestries identified the first rare exonic variants associated with BP traits with effect sizes greater than that observed with common variants (>1.5 mm Hg per allele), mapped to four genes: RBM47, COL21A1, DBH and RRAS. In the same year, an additional trans-ethnic study reported a rare variant of the NPR1 gene, associated with a +1.1 mm Hg increase in SBP; this gene is a drug target for hypertension with a clinical trial ongoing ( Fig. 1) [57]. Nonetheless, the collective number of loci identified at this point accounted for 2.8% of the genetic heritability of BP. It was not until 2018 that the impact of large-scale Biobanks was truly demonstrated, with the identification of over 500 novel loci in a single study of one million individuals of European descent, doubling the total explained heritability from less than 3% to ∼6% [58]. More recently in 2020, a study including 1.3 million individuals using Exome chip data with replication in trans-ethnic individuals identified an additional 106 novel loci, of which 87 were rare variants [56]. To date, over 1000 loci have been significantly associated with BP, with continuous efforts being made to further unravel the genetic architecture of BP. However, elucidating the causal SNPs remains a challenge; the majority map to noncoding regions of the genome, and variants are often in linkage disequilibrium (LD) with one or more other variants, in which they are nonrandomly associated in the population.

GWASs beyond European ancestry
Whilst BP genetic discovery projects continue to expand in size, they have maintained a strong bias for centring on individuals of European ancestry, and thus there is limited representation and results from other ancestral backgrounds. This is in part due to recent studies including samples from the UK Biobank which is largely European [55], with cumulative estimates of Europeans contributing 88.45% genotypes to GWASs across all traits in 2020 [60]. In order to have GWAS results that are impactful across populations, this data gap needs to be addressed. Continuing this ancestral bias in genetics could exacerbate health disparities due to ethnicity and miss the benefits of new discovery opportunities and understanding in an inclusive system. African ancestry individuals have the highest age-adjusted prevalence of hypertension [61]. Downstream issues are already occurring due to ancestral bias; using genetics for risk stratification of cardiomyopathy wrongly miscategorizes some African Americans due to their omission from control cohorts [62]. This demonstrates the importance of interrogating the population specificity of identified variants. Research across ancestries is increasing, either through sampling less studied populations or conducting trans-ethnic studies [41,63]. These studies have discovered novel and ancestry-specific loci, although they are often limited by smaller sample sizes in com-parison to European-based research [64]. An increase in diverse sampling is being addressed by ongoing establishment of national biobanks (H3Africa, BioBank Japan, the Korean Biobank, the African Genome Variation Project, Qatar Biobank and GenomeAsia 100k), and these datasets are being increasingly utilized in genomics research [65,66].

Risk prediction and causal mechanisms
As BP GWAS summary statistics and other datasets become publicly available to researchers (ebi.ac.uk/gwas/, genetics.opentargets.org/, phenoscanner.medschl.cam.ac.uk/), new methods for interpreting and translating these data have been developed for clinical applications and biological interpretation of findings.
Risk prediction modelling for CVD is now including genetic biomarkers. Genetic risk scores (GRSs) can be developed from combining significant risk alleles identified from GWAS. Alternatively, a more complex polygenic risk score (PRS) can be created by combining a broader range of SNPs which may not individually reach genome-wide significance but together provide an improved risk score [67]. One recent BP-GRS developed in UK Biobank (n = 392,092) combining 901 SNPs found a difference of 12.9 mm Hg (SBP) and 7.5 mm Hg (DBP) between the lowest risk decile and the highest risk decile along with a trebled risk of hypertension and an increased risk of stroke, CVD and myocardial infarction [58]. The authors of this study indicated such a risk score may have utility for early identification of individuals, at a time where lifestyle factors could be advocated to reduce BP levels. PRSs have also been used to test for responsiveness to different drug classes. One such recent study evaluated the association of a genome-wide PRS (>1 million SNPs) with antihypertensive drug responses. This was conducted across four BP drug classes based on data with a sample of n∼200 per drug using BP data before and after 4 weeks on monotherapy; however, no association was established [68]. Utilizing risk scores in this way is an area of research in its infancy, and this is reflected in the small sample sizes studied and consequently lower power which could influence the results obtained so far. Increased sample sizes would be important to account for any BP measurement variance. Furthermore, the studies in this area are predominantly conducted in individuals from European ancestries, potentially leading to gene encodes the natriuretic peptide receptor 1, atrial and brain natriuretic peptides (ANP/BNP) bind to this receptor, their binding leads to lower BP and salt extraction [105]. (a) This mechanism was first identified by injecting atrial extracts into rats resulting in natriuresis [106]. This responsible factor was soon identified, sequenced and labelled as ANP [107]. Further discoveries led to discovery of ANP binding receptors (NPR-A) which regulate cGMP with npr1 knockout mice models and those with genetargeting establishing a dose-response effect [107,108] effectiveness issues across ethnicities due to the intrinsic limitations of the PRSs [69].
Alongside the development of risk scores, Mendelian randomization (MR) is being applied widely to determine causal effects using genetic data to mimic a randomized controlled trial [70]. The MR framework uses SNPs to overcome issues with traditional observational studies such as bias due to confounders and reverse causation. As a random assortment of alleles is passed onto offspring independent of any other characteristics, SNPs in a given population should be similar in all other characteristics removing any potential confounders [70]. Since inception, MR has been regularly applied to genetic data to identify causal risk factors, drug evaluation and identifying disease mechanisms. A recent MR study used a novel approach to assess whether repurposing antihypertensive drugs would affect the risk of Alzheimer's disease [71]. The study was conducted using SNPs in genes associated with 12 antihypertensive drug classes. The selection of SNPs was based on data from www.drugbank.ca/ and www.gtexportal.org/, and these were validated in a UK Biobank SBP GWAS cohort (n = 317, 754), and the effect was estimated in an Alzheimer's disease GWAS (n = 17,008/37,154 case/control). The results indicated lowering SBP via the antihypertensive drug targets selected was unlikely to affect the risk of developing Alzheimer's disease. This work provides a blueprint for evaluating antihypertensive drug applications without conducting a full randomized controlled trial [71].
In another example, Richardson and colleagues used the principles of MR to study the association of the transcriptome across 48 tissue types with complex traits to identify candidate genes loci [72]. Their analysis applied to BP data identified possible causal associations, for example one SNP (rs1706003) which may have been overlooked by using GWAS data alone indicated the candidate gene ATP13A3 [72]. These methods serve as a reminder that a variety of approaches are available and are being constantly developed with numerous applications to utilize and interpret results from genetic studies.
Determining candidate genes and mechanisms at BP loci is key for translation to druggable targets. Before the advent of GWAS, genes and mechanisms for BP were mostly discovered using rat or mouse models. Now, GWAS is taking centre stage, and combining their results with experimental models provides additional support for drug development as illustrated by NPR1 (Fig. 1).
As mentioned previously, GWASs do not provide the causal variant or gene. Functional studies using mouse models remain a key experimental tool once a gene is identified as having strong support from bioinformatics analysis. A recent example of follow-up of GWAS loci is demonstrated with the identification of ARHGAP42 (Rho GTPase Activating Protein 42) as the result of the SNP rs633185 reported as a lead variant at this locus in a GWAS [73]. Genetically modified mice were used to establish that ARHGAP42 deficiency results in hypertension via increased response to angiotensin II and endothelin-1 [74]. These models are continuing to be applied to assess whether this candidate gene could be a valid drug target moving forward [75]. Nonetheless, functional validation represents a major challenge due to the large number of SNPs associated with BP. Candidate genes from GWAS can be evaluated using 'in vitro' systems, including techniques such as CRISPR which can be used for gene-editing of BP variants and subsequent testing in cellular models. Alternatively, mechanisms for GWAS 'candidate genes' can converge from others work studying known BP mechanisms. For example, the variant rs880315 located within an intron of CASZ1 (Castor Zinc Finger 1) is associated with hypertension in GWASs and replicates across different populations and ancestries [45,76,77]. A recent study has established that CASZ1b (short form of CASZ1) co-localizes with the mineralocorticoid receptor in the kidneys and is part of an aldosterone-dependent corepressor complex suppressing ENaCα and SGK1 which are linked to elevating BP by promoting sodium reabsorption [78].

Genetics priming drug discovery
Developing new drug treatments for EH is a key driver of BP genetics research. Increasing our armoury of therapies that effectively lower BP with minimal side effects and reduce hypertensionassociated CVD is important for personalized medicine. This is particularly necessary for hypertension as there are a large proportion of individuals that do not respond to current treatments. Evangelou et al. reported five loci (PKD2L1, SLC12A2, CACNA1C, CACNB4 and CA7) containing genes which are drug targets for several known antihypertensive drug classes [58]. These genes strongly validate the genetic approach of identifying potential drug targets. With greater than 1000 BP-associated loci now identified and genetic data being collected each year, the list of possible drug-target genes is continuously expanding. For example, in 2020 Surendran et al. reported 23 genes as potentially druggable [56]. However, only 12 of the potentially druggable genes identified by Evangelou et al. are the focus of clinical trials for BP (Table 1), including the gene EDNRA. EDNRA encodes endothelin receptor A which plays a role in the endothelin pathway, an established mechanism of BP control (Fig. 2). However, EDNRA can be considered as a drug-target, this is mentioned with caution, as currently endothelin receptor antagonists that are used clinically to treat pulmonary hypertension have had roadblocks when reaching clinical applications for EH [79]. There are also druggable genes not identified initially via GWAS but from other genetic studies which are in development, including the gene MTHFR (Fig. 3). Work is in progress to validate the mechanism by which MTHFR potentially impacts BP. It is hoped that some of these discoveries will translate to novel therapeutics.  [41,58]. One of the strongest contenders was SLC5A1, the target of canagliflozin. Canagliflozin is a SGLT2 inhibitor and originated as a therapeutic for type-2 diabetes. It has since been licensed for treatment in heart failure. Canagliflozin decreases glucose reabsorption but also reduces BP in diabetes patients, indicating potential as an anti-hypertensive therapeutic [58].
Pathway enrichment analysis is a post-GWAS test which can provide mechanistic insights and highlight organ systems and signalling pathways which could be therapeutically targeted. Notably from BP GWASs there is enrichment of genes in arteries, and TGF-ß and Notch signalling pathways are indicated [41,56]. However, a lack of enrichment in other tissues, including the kidneys which are heavily involved in BP regulation, highlights caveats in tissue enrichment analysis (noting there are limited kidney samples in public datasets) [41]. Nevertheless, pathway analyses have successfully reported genes associated with BP and other CVD pathophysiologies, including PHACTR1, informing possible interplay between disease mechanisms [41].

Resistant hypertension
Resistant hypertension, defined as high BP despite patients being on three antihypertensive drug classes [80], is not as well studied using genetic approaches. This is in part due to the fact that resistant hypertension is a complex clinical entity, in which resistant hypertension is also used as an umbrella term for any elevated BP that is nonresponsive where you can rule out nonadherence and secondary causes. This phenotype can result from a large range of diverse pathophysiologic sources and thus poses a great challenge for reliably identifying significant findings in genetic studies. Furthermore, BP GWASs have focused heavily on quantitative BP phenotypes as described therein; however, in recent years GWASs for resistant hypertension have been performed [81,82], with some new findings. For example, Rouby et al. performed, to the best of our knowledge, one of the first GWAS for resistant hypertension and used their findings to develop a GRS [81]. The GRS was based on three signals (in MSX2, IFLTD1 and PTPRD) that were found in participants taken from two randomized clinical trials (1194 White and Hispanic participants in the discovery stage from one clinical study and 585 individuals in the replication stage from an independent clinical study) [81]. The GRS is yet to be tested with researchers noting there was no cohort large enough at that time to study [81]. In contrast, Irvin et al. performed a resistant hypertension GWAS, in which they replicated results in new samples using the Million Veterans Program (n = 16,833) with replication across different ethnic groups and identified the CASZ1 locus with greatest significance [82]. Specifically, they found that rs12046278 T carriers in CASZ1, a locus also associated with quantitative BP traits, were less likely to have resistant hypertension [82]. Overall, these findings emphasize resistant hypertension GWAS is in its infancy and requires further research. New findings have the potential to improve our understanding of resistant hypertension BP biology and to optimize how BP drugs are prescribed.

BP genetics beyond GWAS -epigenetics
Whilst GWAS has enabled the discovery of a large number of loci, they only explain approximately  [110]. (a) Studies in rodent models [111,112,113] and humans [114] [41], [115], [116]). Fine-mapping of this locus confirmed rs9349379 as the causal SNP, and H3K27 acetylation analysis in aortic tissues identified this variant as a distal regulatory enhancer [117]. Deletion of 88 base pairs at the PHACTR1 [117]. These data suggested the variant acts as a distal regulator of the EDN1 6% of BP SNP heritability thus far [58]. Epigenetic changes may also have an important role in the heritability of BP and may explain some of the heritability not accounted for by SNP variation. Epigenetic changes (modifications that lead to changes in the expression of genes but do not change the DNA sequence) can be both heritable and modulated through environmental factors, for example nutrition [83]. Epigenetic mechanisms can alter the expression of specific genes through various methods including DNA methylation, which is often found at CpG dinucleotides (cytosine and guanine bases connected via a phosphodiester bond) located in promoters of genes [84]. Some known BP loci have already been shown to act through epigenetic mechanisms (see the PHACTR1 locus rs9349379 interaction with the endothelin pathway in Fig. 2). The development of arrays targeting CpG sites has allowed the investigation of epigenetic BP regulation using EWASs (epigenome-wide association studies). BP EWASs are still in the early stages of exploration compared to GWAS with only a small number of studies published each with limited sample sizes and therefore low power. One of the largest BP EWAS (17,010 individuals of European, African American, and Hispanic ancestry) reports 13 methylation loci which account for an additional 1.4%-2%

locus in CRISPR-edited stem cell-derived endothelial cells (ESC-EC) resulted in elevated expression of the endothelin-1 (EDN1) gene and ET-1 protein production compared with wild-type cell lines
of heritable BP variation. The results were validated in a cohort of 1516 individuals, along with 126 loci identified after meta-analysis [85]. Another recent EWAS has suggested potential ethnic differences in methylated sites [86]; however, this was based on a significantly smaller sample size (n = 712 comprised South Asian and European ancestries) identifying eight loci (one present in both ethnicities, seven European only). This study did not have a validation cohort; however, it included a comparison between their results and the aforementioned study, which identified some overlap (e.g., cg19693031 near the TXNIP gene) with weaker evidence of association compared to the previous study. This in part could be due to both the smaller sample size and the differing ethnic background of this study (49% South Asian descent). Understanding epigenetics and its contribution to hypertension is currently limited by technology. The current array being used in EWASs covers <2% of known CpG sites [87] as well as not detecting the effect of other common epigenetic mechanisms such as histone modification and non-coding RNA which can also be seen work in tandem with each other (e.g., lncRNA regulating DNA methylation [88]). Further research in this area will be forthcoming and will provide insights into how epigenetics may mediate the relationship between BP genetics, environmental factors and CVD.

Gene-gene and gene-environment interactions
There is a lot of ongoing work investigating genegene and gene-environment interactions, results of which may advance our understanding of why BP drugs are effective in only some patients and aid the development of targeted drugs regulating gene interactions. To the best of our knowledge, BP gene-gene interactions have only been assessed in a handful of small-scale studies [87,89,90]. For example, Meng et al. found five novel interactions (MAN1A1, LMO3, NPAP1/SNRPN, DNAL4 and RNA5SP455/KRT8P5) contributing to hypertension [91]. However, this study only included 2203 cases with a matched number of controls from a predominantly European dataset [91].
The identified genes require further research to definitively prove a link to BP modulation. There are also several other similar studies with small datasets (less than 1000 hypertensive cases per study [87,89,91]). These studies have identified novel interactions between known BP genes (e.g., between MTHFR and FGF5 [89]). This work emphasizes a need for future work scaling up gene-gene studies across populations for validation -from which gene expression may be regulated as part of more precise treatment as we gain better understanding of the interplay between genes.
There is more work on gene-environment interactions. The goal of this approach is to understand exactly how environmental factors (e.g., smoking, alcohol intake, air pollution) interact with genotypes to affect BP. These interactions may lead to preventative medicine for at-risk individuals depending on modifiable environmental factors.  [93]. Eight of these loci had significant interactions with smoking status, such as the CSMD1 locus which was also associated with BP in SHRs [93]. However, these loci were only identified in the discovery stage of the study, and, due to the small sample size of the African individuals, replication was not possible [93]. The results from this analysis and other studies [94] further highlight a need for future work to incorporate larger sample sizes of non-European populations.

Mining post-GWAS data
Post-GWAS methodology is advancing to meet growing demands for analysis of large amounts of genetic data. High-powered computational methods can account for SNP associations and their annotations in-depth, guiding hypotheses for functional follow-up, avoiding cherry-picking bias, and dissecting past LD to the causal disease genes. The methods being increasingly deployed range from machine learning, network analysis, fine-mapping, text-mining, MR and hybrid tools (Fig. 4). Machine learning is a statistical method using algorithmic rules to identify complex data patterns and make predictions. It has had investment due to its ability to identify hidden patterns within multi-omic data. Mishra [97]. As innovative methods such as these develop, their work paves the way for future studies to extract higher level information from amassing BP GWAS results and sequencing data.

Challenges and outlook
Genetics research is starting to provide a more comprehensive understanding of how individual pathways and systems contribute to hypertension. However, the specifics of how BP regulatory mechanisms interact with each other (e.g., epigenetic factors acting in known BP pathways, gene-gene, gene-environment, and gene-microbiome interactions) are largely unexplored. A recent review by Weber et al. describes the regulation of hydrogen sulfide metabolism and interactions with epigenetic factors and gut microbiota and their contribution to hypertension [98]. From their research they put forward the theory that hydrogen sulfide metabolism can be dysregulated and affect renal function and BP, by gut microbiota (specifically bacterial activity that increases urea blood concentration) and microRNA modulation of angiotensin II [98]. This study integrated results from multiple modalities -epigenetics, metabolomics and the microbiome -becoming one of the few studies placing pathway research into a broader picture of BP biology. As research advances, intersectional stud-ies such as this will need to become commonplace and fit into the overall paradigm of how genetic studies are validated by research interlaying findings from different studies, and extending their insights into translational research for hypertension (Fig. 5).
From a clinical perspective, increased access to genetic data is providing many applications with clinical utility. One example is the use of GRSs for coronary heart disease (CHD) for prediction and prevention of disease [99]. Kullo et al. conducted a clinical trial (n = 203) in participants with no CHD. Participants were randomized to receive a 10-year risk prediction using a clinical risk score (CRS) alone or a CRS and CHD GRS and results showed the latter group had lower cholesterol levels and were more likely to receive statins. This study had used shared decision making based on the patient's overall risk [99]. Recently, a study by Weale et al. developed a PRS for atherosclerotic CVD including CHD [100]. They presented an integrated risk tool that performs 10-year risk prediction across diverse ethnic and ancestry groups. They found improved prediction of CHD across individuals of not only self-reported White ethnicities but Black/African, American/Black, Caribbean/Black, African and South Asian (Indian, Bangladeshi or Pakistani) ethnicities -with PRS effect sizes in these ethnicities being significant and of comparable size to those seen in individuals of White ethnicities [100]. These positive results provide for the first time a validated PRS tool that can generalize across ancestries and presents a roadmap example for similar studies that may be appropriate for informing hypertension treatment using BP GRSs.
The pharmacogenetics of BP drugs, understanding if a person's drug response depends on genetic influences, has also gained attention -but this is an area which has room for improvement before offering clinical potential. For example, Arnett et al. developed the GenHAT (Genetics of Hypertension Associated Treatments) study, an investigation of hypertension genetic variants and their interactions with antihypertensive treatments in relation to CHD [101]. This study was conducted over 4.9 years in 39,114 hypertensive cases across ethnicities -with individuals grouped depending on their antihypertensive medication class -providing one of the largest and most diverse pharmacogenetic BP studies. However, the results were not as promising as expected, with, for example, results exploring AGT gene interactions across hypertension drug classes finding 'no gene by treatment interactions to be statistically significant' [102]. This study and others suggest that the pharmacogenetics of BP requires further work, which aligns with the overarching need for increased translational research before the genetics of BP can provide clinical benefits and advance hypertension treatments.
In conclusion, early genetic investigations of hypertension successfully established rodent models and advances in the human genome project have permitted discoveries for genes causing both monogenic forms of hypertension and EH. In recent years new lines of genetic analyses have taken shape, with epigenetics now being a hot topic. However, whilst the genetic insights and progress made so far have been vast, it is far from being comprehensive. There are now large whole genome datasets coming online and being collated into databases such as in gnomAD [103], the UK Biobank [55] and Genomics England's 100,000 genomes project [104], which provide full coverage of the genome at scale for the first time. Research in hypertension is beginning to focus on non-European data, as translational findings need to benefit all populations equally and improve rather than worsen existing health disparities. Research also needs to expand in broader directions, spanning the metabolome, microbiome and incorporation of environmental factors. Such work will provide a more complete view of complex BP biology.
As intersectional experiments develop, they may present an opportunity to connect findings with growing gene-gene and gene-environment research that may be the key to unlocking BP insights awaiting discovery. Overall, the genetics of hypertension is providing new information on underlying physiology, it is building towards an endgame for precision medicine, that may ultimately lead to lower BP as an important cofactor for decreasing CVD and potentially offer large global impact.

Institute of Health Research Barts Biomedical
Research Centre for PhD funding. Claudia C. Cabrera and Patricia B. Munroe also wish to acknowledge the National Institute of Health Research Barts Biomedical Research Centre for their support. All the authors thank Prof. Andrew Tinker, Dr. Ajay K Gupta and Dr. Emma Forton Magavern (Clinical Pharmacology, Barts and the London School of Medicine and Dentistry) for their critical review of the draft manuscript. Figures 1-4 were created with Biorender.com.