Genomics of Blood Pressure and Hypertension: Extending the Mosaic Theory Toward Stratification

The genetic architecture of blood pressure (BP) now includes more than 30 genes, with rare mutations resulting in inherited forms of hypertension or hypotension, and 1477 common single-nucleotide polymorphisms (SNPs). These signify the heterogeneity of the BP phenotype and support the mosaic theory of hypertension. The majority of monogenic syndromes involve the renin-angiotensin-aldosterone system and the adrenal glucocorticoid pathway, and a smaller fraction are due to rare neuroendocrine tumours of the adrenal glands and the sympathetic and parasympathetic paraganglia. Somatic mutations in genes coding for ion channels (KCNJ5 and CACNA1D) and adenosine triphosphatases (ATP1A1 and ATP2B3) highlight the central role of calcium signalling in autonomous aldosterone production by the adrenal gland. The per-SNP BP effect is small for SNPs according to genome-wide association studies (GWAS), and all of the GWAS-identified BP SNPs explain ∼ 27% of the 30%-50% estimated heritability of BP. Uromodulin is a novel pathway identified by GWAS, and it has now progressed to a genotype-directed clinical trial. The majority of the GWAS-identified BP SNPs show pleiotropic associations, and unravelling those signals and underpinning biological pathways offers potential opportunities for drug repurposing. The GWAS signals are predominantly from Europe-centric studies with other ancestries underrepresented, however, limiting the generalisability of the findings. In this review, we leverage the burgeoning list of polygenic and monogenic variants associated with BP regulation along with phenome-wide studies in the context of the mosaic theory of hypertension, and we explore potential translational aspects that underlie different hypertension subtypes.

L'architecture g en etique de la pression art erielle (PA) comprend maintenant plus de 30 gènes, avec des mutations rares entraînant des formes h er editaires d'hypertension ou d'hypotension et 1 477 polymorphismes d'un seul nucl eotide (PSN) courants. Tout cela d emontre l'h et erog en eit e du ph enotype de la PA et soutient la th eorie « mosaïque » de l'hypertension. La majorit e des syndromes li es aux maladies monog eniques impliquent le système r enine-angiotensinealdost erone et la voie des glucocorticoïdes via les glandes surr enales, tandis qu'une plus petite fraction est due à de rares tumeurs Page's "mosaic theory" of hypertension posited that essential hypertension (HTN) is not a single disease, but several different diseases with different origins and developmentdall causing HTN and its consequences. 1 The mosaic of causes of hypertension, if it exists for essential hypertension, needs elucidation because it potentially opens new avenues for stratification, new drug development, and precision medicine. Blood pressure (BP) genomics has been one of the most challenging research areas, primarily because the inherent variability of BP and measurement errors (both human and instrument related) substantially dilute the statistical power of any discovery study. In this review, we leverage the current catalogue of polygenic variants and monogenic variants associated with BP regulation along with phenome-wide studies to determine whether there is evidence for the mosaic theory, and we explore potential translational aspects that underlie different HTN subtypes.

Monogenic Syndromes
There is evidence from monogenic forms of HTN that they are caused by specific pathways perturbed by rare causal mutations in single genes resulting in an early and severe BP phenotype. 2 The monogenic forms are the most successful examples of gene mapping, with mutations in more than 31 genes now linked to perturbed gene function and BP dysregulation, enhancing our understanding of both the mechanisms and the treatment of HTN. Molecular and clinical details of monogenic syndromes have been reviewed in detail recently, and we refer to the reader to these articles. 2,3 Table 1 and Figure 1 summarise the monogenic syndromes, their causal genes, mechanisms, and management. Different monogenic syndromes all present with the same BP phenotype but are essentially separate diseases differentiated by additional clinical or laboratory characteristics and the causative genetic mutation. Although monogenic forms of HTN account for a small fraction of the public health burden of hypertension, studies of monogenic syndromes have established the genetic basis for the "known knowns" of BP regulation. These are major pathways of BP regulation, centred on sodium balance and the renin-angiotensin-aldosterone system, originally inferred from physiological studies that led to the development of almost all of the commonly used antihypertensive drugs. 4 For example, one of the most effective drugs for HTN is spironolactone, which antagonises the effect of aldosterone. It was discovered in 1957 and has been used in clinical practice since 1959. The first case of familial primary aldosteronism was described in 1966 with the discovery of the causal chimeric mutation by Lifton et al. 5 in 1992, 25 years later. Spironolactone, was recently shown to be overwhelmingly the most effective drug treatment for resistant HTN in the PATHWAY-2 randomised controlled trial. 6 Interestingly, somatic gain-of-function mutations in (KCNJ5) account for 40% of patients with an aldosteroneproducing adenoma, whereas 7% are due to mutations in the a-subunit of Na þ -K þ -adenosine triphosphatase (ATPase; ATP1A1), plasma membrane Ca 2þ -ATPase (ATP2B3), and L-type Ca 2þ channel CaV1.3 (CACNA1D). 3 Mutations in these genes are less frequent in inherited cases of primary hyperaldosteronism and raises the possibility that HTN could be due to a multiplicity of uncommon variants. 3 The majority of monogenic syndromes involve the renin-angiotensinaldosterone system and the adrenal glucocorticoid pathway with treatment directed toward ultimately reducing aldosterone and angiotensin and increasing Na þ excretion (Table 1). A smaller fraction of monogenic HTN syndromes are due to rare neuroendocrine tumours of the adrenal glands and the sympathetic and parasympathetic paraganglia: pheochromocytomas (PCCs) and paragangliomas (Table 1). Autosomal hypertension with type E brachydactyly is not related to salt reabsorption but due to mutation in the PDE3A gene resulting in enhanced activity of PDE3A leading to increased neointimal proliferation and remodelling of the arteries and neurovascular structures. 7 Although the rare monogenic syndrome may not have significant direct public health impact, the indirect global influence of drugs targeting those specific pathways among those with essential HTN is substantial. It is in this context that the potential value of the global efforts to discover the genetic basis of essential or polygenic HTN needs to be considered. Thus, monogenic syndromes fulfil Page's mosaic theory of HTN, albeit in a smaller subset of all hypertensive individuals.

Essential Hypertension
It logically follows that essential HTN may be a forme fruste of monogenic HTN, with minor variations in the monogenic genes leading to milder and later-onset HTN. Consequently, they may exhibit an underlying mosaic pattern. There are indications that this may be a possibilitydfor example, individuals of African ancestry tend to have a saltsensitive form of HTN, 8 in contrast to individuals of European ancestry; somatic mutations causing hyperaldosteronism 9 result in another subset of HTN. However, beyond these 2 examples, it remains to be established if essential HTN is a scaled-up mosaic form of monogenic HTN.
Beyond the evidence from monogenic forms of HTN, there are multiple lines of observational evidence suggesting that BP has a genetic component. 1) Family and twin studies have established that BP heritability ranges from 15% to 40% for clinic systolic blood pressure (SBP), and from 15% to 30% for clinic diastolic blood pressure (DBP); for ambulatory BP (sleep), heritability was 69% and 51% for SBP and DBP, respectively. 10, 11 2) The risk of developing HTN is significantly increased in individuals with 1 or 2 hypertensive parents, 12 and monozygotic twins show higher BP correlations than dizygotic twins. 13 And 3) in the burst of discovery in the genome-wide era, the search for common genetic variants underlying BP is based on genome-wide association studies (GWAS) which are hypothesis-generating studies that scan the entire genome to find associations between genetic variants (typically single-nucleotide polymorphisms [SNPs]) and a phenotype. The first GWAS of HTN was entirely negative, leading many to question if BP and HTN was genetically tractable or if its genetic component was trivial. 14 However, since 2007, a series of sequential GWAS of BP and HTN with sample sizes exponentially increasing to the latest study involving 1 million subjects have identified more than 1477 SNPs associated with BP traits explaining about 27% of the 30%-50% estimated heritability of BP. 2,15 All of these support a multifactorial polygenic basis for BP regulation and Somatic mutations in genes coding for ion channels (KCNJ5 and CACNA1D) and adenosine triphosphatases (ATP1A1 and ATP2B3) highlight the central role of calcium signalling in autonomous aldosterone production by the adrenal gland. The per-SNP BP effect is small for SNPs according to genome-wide association studies (GWAS), and all of the GWAS-identified BP SNPs explain w 27% of the 30%-50% estimated heritability of BP. Uromodulin is a novel pathway identified by GWAS, and it has now progressed to a genotype-directed clinical trial. The majority of the GWAS-identified BP SNPs show pleiotropic associations, and unravelling those signals and underpinning biological pathways offers potential opportunities for drug repurposing. The GWAS signals are predominantly from Europe-centric studies with other ancestries underrepresented, however, limiting the generalisability of the findings. In this review, we leverage the burgeoning list of polygenic and monogenic variants associated with BP regulation along with phenome-wide studies in the context of the mosaic theory of hypertension, and we explore potential translational aspects that underlie different hypertension subtypes. neuroendocrines des glandes surr enales et des paraganglions sympathiques et parasympathiques. Des mutations somatiques dans les gènes codant les canaux ioniques (KCNJ5 et CACNA1D) et les ad enosine triphosphatases (ATP1A1 et ATP2B3) mettent en evidence le rôle central de la signalisation du calcium dans la production autonome d'aldost erone par la glande surr enale. L'effet de PSN li es à la PA est faible pour les PSN identifi es dans les etudes d'association pang enomique (EAP), et tous les PSN de la PA identifi es par EAP expliquent w 27 % des 30 %-50 % d'h eritabilit e estim ee de la PA. L'uromoduline repr esente une nouvelle voie de signalisation identifi ee par les EAP, et elle fait d esormais l'objet d'un essai clinique bas e sur le g enotype. La majorit e des PSN li es à la PA identifi es par des EAP pr esentent des associations pl eiotropes, et l' elucidation de ces signaux et des voies biologiques sous-jacentes offre des possibilit es potentielles de r eorientation de m edicaments. Les donn ees des EAP proviennent principalement d' etudes de populations d'Europe, tandis que les autres ascendances y sont sous-repr esent ees, ce qui limite la g en eralisation des r esultats. Dans cette revue de litt erature, nous nous appuyons sur la liste emergente des variants polyg eniques et monog eniques associ ees à la r egulation de la PA ainsi que sur des etudes à l' echelle du ph enome dans le contexte de la th eorie « mosaïque » de l'hypertension, et nous explorons les aspects translationnels potentiels qui sous-tendent diff erents sous-types d'hypertension.

Lip and Padmanabhan
HTNdthe Pickering argument from the 1950s Platt-Pickering debate. 16 Novel pathways from GWAS Despite the plethora of common variants arising from GWAS, the biggest challenge has been linking these variants to a causal mechanism in the BP regulatory pathway. The main reason for this is because GWAS SNPs are selected for screening the genome based on linkage-disequilibrium patterns, and this results in the majority of the signals being in noncoding or intergenic regions. Nevertheless, there are examples of the value of GWAS in identification of novel pathways of BP regulation point to potential subgroups with common underlying pathways that may offer avenues for targeted screening or therapy. We start with 2 SNPs (near UMOD and PHACTR1/EDN1 genes) that have identified novel pathways with early translational potential because they involve gene products that are the targets for licensed drugs. Then we look at pharmacogenetic interactions that validate pathways targeted by current antihypertensive drugs and explore opportunities for repurposing drugs or tailoring treatment.
Uromodulin. A GWAS of BP extremes 17 identified a 5 0promoter SNP, rs13333226, near the uromodulin gene (UMOD) which is associated with BP and uromodulin excretion. UMOD is almost exclusively expressed in the thick ascending limb of the loop of Henle in the kidney, where 25% of the filtered Na þ is reabsorbed, pointing to a novel sodium-based BP pathway. A potential interaction between uromodulin and the main sodium transporter, NKCC2, in the thick ascending limb of the loop of Henle was established with transgenic mice experiments. 18,19 In Umod-knockout mice, NKCC2 shows reduced cotransporter activity with consequent greater sodium excretion and a 20 mm Hg lower BP compared with wild-type mice. 18 The role of NKCC2 was further established by Trudu et al., 19 who showed in both mice overexpressing Umod and hypertensive individuals homozygous for the UMOD increasing allele that the NKCC2 antagonist furosemide had a greater natriuretic and hypotensive effect. This result is currently being tested in a clinical trial (ClinicalTrials.gov Identifier: NCT03354897) to reposition loop diuretics in the HTN care pathway.
PHACTR1. An intronic SNP in the phosphatase and actin regulatory protein 1 (PHACTR1) gene associated with increased risk of coronary artery disease (CAD) and coronary calcification and decreased risk of migraine headache, cervical artery dissection, fibromuscular dysplasia, and HTN. 20,21 Functional analysis of this variant indicated that it is a distal regulator of endothelin (ET) 1 (EDN1), a gene located 600 kb upstream of PHACTR1. 20 Thus, this functional SNP may potentially be associated with a lifetime's exposure of at least 20% higher ET-1 precursor plasma levels, but this requires more validation. 20 Table 1.
Lip and Padmanabhan 699 Hypertension Genomics and Mosaic Theory are now licensed for the treatment of pulmonary HTN. 24 In addition, a genotype-directed use of endothelin antagonists for nonobstructive CAD may offer a solution for targeted treatment (ClinicalTrials.gov Identifier: NCT04097314).

Pleiotropy
Pleiotropy occurs when a given genetic locus (SNP) influences 2 or more different phenotypes or traits. Phenomewide association studies (PheWAS) can identify statistical associations between a single variant and multiple phenotypes and thus reveal pleiotropic associations. PheWAS are usually carried out with the use of a wide range of phenotype data from electronic health records, epidemiologic studies, and clinical trials. Whereas GWAS typically investigates a single phenotype at a time, PheWAS identify all of the traits associated with a genetic variant. With near ubiquitous availability of genome-wide genotyping in most large epidemiologic cohorts and the emergence of biobanks, there is potential to identify pleiotropic SNPs to improve our understanding of the biological functions of a GWAS SNP or identify concealed pathophysiologic connections between traits previously considered as distinct. A look-up of all the 1477 BP SNPs in the GWAS catalogue and PhenoScanner 25,26 using a P value threshold of 5 Â 10 À5 showed a range of significantly associated traits. The pleiotropic traits are summarised in a word cloud (Fig. 2), where the size of the words in the word cloud indicates the weights based on the number of independent BP SNPs associated with each phenotype. There is considerable overlap of BP SNPs across traits that predispose to or correlate with type 2 diabetes mellitus (T2DM). This suggests that the co-occurrence of T2DM and HTN may manifest through shared genetic factors and shared pathways. However, detecting an independent BP effect on T2DM causation, even if it exists, may be challenging. This is highlighted by 2 recent studies that tried to establish a causal effect of BP on T2DM. One study used 28 SNPs associated with T2DM and showed that a 1 mm Hg rise in genetically determined SBP was causally associated with a 2% increased risk of hypertension. 27 However, a more recent bidirectional mendelian randomisation study of UK Biobank participants, using a larger SNP set of 134 T2DM SNPs and 233 BP SNPs, suggested that T2DM may causally affect HTN, whereas the relationship from HTN to T2DM is unlikely to be causal. 28 Nevertheless, genetics provides support for observed correlation between T2DM and HTN and the presence of shared pathways raises the possibility of combined treatment for both. The newer SGLT2 inhibitors with dual effects on both glycemic control and BP are a fortuitous example of a single drug with joint effects on BP and glycemia and support investigation of shared pathways to develop novel therapies for both conditions. 29

GWAS and Pharmacogenomics
The goal of genomics is to enable precision medicine through a greater understanding of molecular pathways that regulate BP, which can inform new drug development, personalisation of treatment, and ultimately lead to a new taxonomy of HTN. 4 But there are significant challenges in realising this goal. Current treatments of HTN have not seen any new drug approval for more than 2 decades, mainly owing to the view that current HTN management is well served by the existing set of drug classes available. Tailoring of therapy has not progressed beyond considering self-reported African ancestry and serum renin levels. 30,31 Linking all of the GWAS BP variants mapped to genes to the DrugBank 32 and Comparative Toxicogenomics Database 33 shows that all of the major antihypertensive drug classes are captured by pharmacogenetic interaction with these GWAS loci (Fig. 3). Although this may simply reflect the fact that the putative published genes mapped to GWAS SNPs were selected for plausible BP effect, this raises the possibility of exploring other genes in the region of GWAS loci for BP-lowering potential. A more attractive approach is to link pharmacogenetic interactions of GWAS signals with pleiotropic PheWAS results, because this may reveal opportunities for drug repurposing and help to define the most appropriate patient populations to benefit from a drug. Table 2 summarises a set of pharmacogenes that interact with GWAS BP genes and have BP lowering as a documented side-effect. We also show the pleiotropic associations of these loci. Some of these drugs have failed trials for other conditions and others have not been trialled for HTN. ET antagonists and riociguat currently licensed for pulmonary HTN show up as pharmacogenetic interactions with GWAS loci for BP (EDNRA and GUCY1A2, respectively) and are candidates for clinical trials for potential extension of their current indication to essential HTN. Valproic acid shows multiple interactions with a range of GWAS loci (HDAC9, SCN2A, SCN10A), indicating that it may be a candidate for repositioning. Nesiritide interacts with NPR3 and is currently not licensed because it failed in a heart failure clinical trial. PheWAS results indicate that pharmacogenes for valproic acid are also associated with adiposity traits and heart rate, pharmacogenes for ET antagonists are associated with CAD, and nesiritide pharmacogene is associated with adiposity, basal metabolic rate, height, lung function, and visceral fat traits. These raise the possibility that new drug development or drug repurposing needs to take into account multimorbid  All pleiotropic associations of GWAS BP single-nucleotide polymorphisms (SNPs) were extracted and categorised into groups of correlated traits. Some SNPs did not show any non-BP associations and were classified as nonpleiotropic. The genes linked to GWAS SNPs were determined by proximity to the SNP and cardiovascular plausibility. Only 1 gene per loci was included. Drug-gene interactions were obtained from the DrugBank and Comparative Toxicogenomics Database, and drug indications were obtained from the British National Formulary and Food and Drug Administration labelled indications.
BMI, body mass index; BMD, bone mineral density; BMR, basal metabolic rate; RBC, red blood cells; CAD, coronary artery disease; CVA, cerebrovascular accident; T2DM, type 2 diabetes mellitus; WBC, white blood cells. associations for targeting treatment to the right subset of patients. Finally, instead of repurposing drugs, pleiotropic associations open the possibility of multipurposing drugs, for example, prioritising antidepressant drugs that additionally lower BP for patients with hypertension and depression or antiepileptics that also have BP-lowering potential for patients who have both epilepsy and hypertension (Table 2).

Polygenic Risk Scores
Because the genetic make-up of an individual is largely stable from birth, genetic information has the potential to act as an early risk predictor. Essential HTN is influenced by multiple genetic variants with small individual effect sizes, so meaningful risk prediction necessitates examining the aggregated impact of these multiple variants. This is through calculation of a polygenic risk score (PRS), which is a mathematical aggregate of risk conferred by all of the SNPs significantly associated with BP. It is important to highlight that the risk information provided by the PRS is different from the risk information from genetic markers of monogenic disorders. The latter is a dichotomous result (either high or low probability of disease), whereas the former provides a wider range of probabilistic risk. In addition, the rare variant genotype points to specific biological impact of the variant, whereas the PRS is an amalgamation of numerous small-effect variants across the genome with no specific pathway implicated. A PRS constructed to use of all of the significant GWAS BP SNPs showed a significant association with stroke, CAD, heart failure, and left ventricular mass, but not for kidney function. 34 This supports the established association between HTN and cardiovascular outcomes and suggests that progression of renal damage due to HTN may continue despite control of HTN. Although there is considerable interest in the use of PRS as a biomarker for early intervention, currently there is no evidence for the clinical utility of PRS for intervention or disease prevention. It is likely that PRS currently may have limited utility because studies have been conducted in adults over the age of 40 years where disease would already have been established. However, PRS may have more value in identification of younger at-risk individuals, which merits further study. There is limited utility in personalisation of treatment or new drug discovery through PRS, primarily because it is derived from an amalgamation of all genetic variants and do not represent unique pathways.

Conclusions
Genomic studies have identified the largest set of SNPs for BP compared with other complex traits. A proportion of these may have translational potential, and the challenge is to identify them. Pleiotropic associations may point to novel underlying pathways or potentially subtypes of essential HTN. Early translational application may be through drug repositioning, followed by new drug development. PRSs look attractive, but their clinical utility needs controlled studies, and the potential ethical impacts of their widespread use exacerbating health disparities need further assessment. 35,36 Funding Sources S.P. is funded by the Medical Research Council (MR/ M016560/1; AIM-HY Study), the British Heart Foundation (BHF; PG/12/85/29925; CS/16/1/31878), and the BHF Centre of Excellence (RE/18/6/34217).

Disclosures
The authors have no conflicts of interest to disclose.