Precision Medicine Approaches to Vascular Disease

Abstract

In this second of a 5-part Focus Seminar series, we focus on precision medicine in the context of vascular disease. The most common vascular disease worldwide is atherosclerosis, which is the primary cause of coronary artery disease, peripheral vascular disease, and a large proportion of strokes and other disorders. Atherosclerosis is a complex genetic disease that likely involves many hundreds to thousands of single nucleotide polymorphisms, each with a relatively modest effect for causing disease. Conversely, although less prevalent, there are many vascular disorders that typically involve only a single genetic change, but these changes can often have a profound effect that is sufficient to cause disease. These are termed “Mendelian vascular diseases,” which include Marfan and Loeys-Dietz syndromes. Given the very different genetic basis of atherosclerosis versus Mendelian vascular diseases, this article was divided into 2 parts to cover the most promising precision medicine approaches for these disease types.

Our understanding of the genetic basis of cardiovascular diseases (CVDs) has progressed rapidly over the last 2 decades, which has led to the beginnings of the precision medicine era. Precision medicine is an integrative approach that incorporates detailed information regarding an individual’s genetics, and also lifestyle, medical conditions, demographic characteristics, exposures, and other factors, which are used to devise personalized approaches to the prevention and treatment of diseases. This approach can guide patient management in many aspects, including improved diagnosis, family screening, targeted therapies guided by genotype and phenotype, risk stratification and prognosis, and assisting in reproductive decisions. These and other fundamental aspects of precision medicine are reviewed in depth in Part 1 of this Focus Seminar series.

Here in Part 2, we focus on precision medicine approaches to vascular disease. In broad terms, the most common vascular disease worldwide is atherosclerosis, which is the primary cause of coronary artery disease (CAD), peripheral vascular disease, and a large proportion of strokes and other disorders. As we explain, atherosclerosis is a highly complex disorder that involves many hundreds to thousands of single nucleotide polymorphisms (SNPs) that each impart a relatively weak effect in causing disease. On the other hand, although less prevalent, there are many vascular disorders that typically involve only a single genetic change for any given patient but with this change having a profound effect that is often sufficient to cause disease. These are referred to as “Mendelian vascular diseases.” The current article is divided into 2 parts to cover the most promising emerging precision medicine approaches for these distinct disease groups (Central Illustration).

CENTRAL ILLUSTRATION. Overview of Vascular Diseases and Respective Precision Medicine Approaches.

Given their very different genetic basis, this article is divided into sections on Mendelian versus non-Mendelian vascular diseases to cover the most promising precision medicine approaches for these disease types. GWAS = genome-wide association studies.

Miller, C.L. et al. J Am Coll Cardiol. 2021;77(20):2531–50.

MENDELIAN VASCULAR DISEASES

Nonatherosclerotic Mendelian vascular diseases include those manifesting with arterial enlargement (dilatations, ectasias, and aneurysms) and impaired tissue integrity (dis-sections and ruptures). In the absence of conventional risk factors (e.g., atherosclerosis, hypertension, chronic kidney disease, smoking, congenital lesion, drug/alcohol abuse), particularly in a younger individual, the finding of arterial dilatation should prompt evaluation for a genetic etiology.

STATE-OF-THE-ART CLINICAL CARE: HERITABLE ARTERIOPATHIES ASSOCIATED WITH ANEURYSMS AND DISSECTIONS.

Among monogenic arterial dilatation disorders, the most common are those featuring thoracic aortic aneurysm (TAA). Approximately 30% of TAA can be linked to a pathogenic variant in a set of causative genes, which primarily encode proteins related to smooth muscle cell function, extracellular matrix integrity, or the transforming growth factor-β (TGF-β) signaling pathway. Putatively damaging variants in these genes contribute to TAA through medial degeneration of the arterial wall. Inheritance is most commonly autosomal dominant. Heritable TAAs may be classified as either “syndromic” (Table 1), which are often associated with systemic and discernible traits, or “nonsyndromic” (Table 2). Classifying patients with TAA into either of these categories affects surveillance, intervention, and prognostication. Syndromes including Marfan (MFS) and Loeys-Dietz (LDS) confer a higher risk of progressive aortic dilatation and aortic emergencies than nonsyndromic causes of TAA, and thus recognition of these entities is paramount.

TABLE 1.

Summary of the Syndromic Arteriopathies

	Causative Gene(s) or Genetic Region	Primary Artery Involved	Systemic Arterial Involvement	Associated Clinical Features	Clinical Actions (Per Syndrome)	Clinical Actions (All)
Marfan syndrome	FBN1	Aortic root	Rarely AAA or craniocervical	Ectopia lentis Scoliosis Pectus deformities Pes planus Facial dysmorphisms	Consider beta-blocker or ARB Baseline and periodic surveillance tomographic imaging of entire aorta (CTA or MRA) Elective surgical repair of aortic root at 5.0 cm, or sooner if rapid growth or family history of dissection at smaller dimensions Ophthalmologic evaluations	Aortic surveillance by echocardiography (annually or more frequently if approaching surgical threshold or significant valvular insufficiency) Genetic counseling Genetic evaluation of at-risk relatives (cascade screening if causative variant is known in proband)
Loeys-Dietz syndrome	TGFBR1 (type 1) TGFBR2 (type 2) SMAD3 (type 3) TGFB2 (type 4)	Thoracic aorta	Any	Craniofacial (i.e., bifid uvula) Retinal detachment Cervical spine abnormalities Joint hypermobility	Consider beta-blocker or ARB Baseline and periodic surveillance pan-vascular imaging from head to pelvis (CTA or MRA)1 Elective surgical repair of aortic root at 4.0–4.6 cm (depending on type of LDS, rate of growth, and family history) Ophthalmologic evaluations	Pregnancy considerations for female subjects
Vascular Ehlers-Danlos syndrome	COL3A1 COLIA1 *	Any, especially mesenteric	Any	Gastrointestinal perforations Uterine rupture Early-onset varicosities Skin fragility	Baseline and periodic surveillance pan-vascular imaging Creation of "care team" and emergency management plan Counseling on lifestyle interventions to decrease risk, including avoidance of elective procedures, angiography, colonoscopy
Shprintzen-Goldberg syndrome	SKI	Aortic root†	Rare	Craniosynostosis Characteristic facies Marfanoid habitus Developmental delay Intellectual disability	Orthopedic consultation may be needed Children: consider IEP
Noonan syndrome	PTPN11 RAF7 SOS7	Aortic root	Coronary aneurysm Cerebral AVM	Short stature Pulmonic stenosis Hypertrophic cardiomyopathy Lymphedema Moyamoya Hematologic abnormalities	Cardiac imaging Hematologic evaluation Children: consider growth hormones
Turner syndrome	45,X	Thoracic aorta‡	-	Short stature Ovarian hypofunction Renal abnormalities Webbed neck Lymphedema Hypertension Metabolic abnormalities	Measure upper and lower extremity BP QT interval measurement on ECG Cardiac imaging Renal ultrasound Children: consider growth hormones Children: consider IEP
Chromosomal deletions or duplications	7q11.23 16p13.1 16p13.11 2p22 5p13	Thoracic aorta	-	Neurological disorders Complex mental disorders Characteristic facies Macrocephaly Developmental delay	Children: may need neurological or psychiatric evaluations, IEP

^*Considered a rare cause of Ehlers-Danlos syndrome due to specific alleles, but clinical validity recently called into question (21).

^†Recently classified as "Limited" evidence for association with thoracic aortic aneurysm (21).

^‡Bicuspid aortic valve, coarctation of the aorta, thoracic aortic aneurysm, and aortic dissection.

AAA = abdominal aortic aneurysm; ARB = angiotensin receptor blocker; AVM = arteriovenous malformations; BP = blood pressure; CTA = computed tomography angiography; ECG = electrocardiogram; IEP = individualized educational plan; LDS = Loeys-Dietz syndrome; MRA = magnetic resonance angiography.

TABLE 2.

Summary of the Nonsyndromic Arteriopathies

Causative Gene(s)	Inheritance	Primary Aortopathy	Other Arteries At Risk?	Associated Clinical Features	Clinical Actions (Per Syndrome)	Clinical Actions (All)
ACTA2	Autosomal dominant	TAA and dissection*	-	Patent ductus arteriosus Bicuspid aortic valve Livedo reticularis Iris flocculi Premature CAD Moyamoya/ischemic stroke High-risk dissection in pregnancy	CAD surveillance and risk factor modification Prenatal counseling on risk of pregnancy and frequent aortic monitoring during pregnancy	Baseline tomographic imaging of entire aorta (CT imaging or MRA) Elective surgical repair of aortic root at 5.5 cm, or sooner if rapid growth, significant valvular insufficiency, or family history of dissection at smaller dimensions Aortic surveillance by echocardiography (annually or more frequently if approaching surgical threshold or significant valvular insufficiency)
MYH11	Autosomal dominant	TAA and dissection	-	Patent ductus arteriosus Focal aortic medial hyperplasia		Genetic counseling Genetic evaluation of at-risk relatives (cascade screening if causative variant is known in proband)
MYLK	Autosomal dominant	Type A and B dissection	Dissections in renal, iliac, mesenteric, subclavian, innominate	Possible absence of Preceding aneurysms
LOX	Autosomal dominant	Aortic root aneurysm	Ascending aorta	Bicuspid aortic valve
PRKG1	Autosomal dominant	Type A and B dissections	-	Possible absence of preceding aneurysms

^*Dissections at documented diameters <5.0 cm.

CAD = coronary artery disease; CT = computed tomography; MRA = magnetic resonance angiography; TAA = thoracic aortic aneurysm.

Both MFS and LDS associate with systemic, often outwardly apparent features, including scoliosis, pes planus (flat feet), and/or pectus deformities (excavatum or carinatum). Craniofacial dysmorphisms (i.e., cleft/bifid uvula) and cervical spine instability are more specific to LDS. Although often collectively referred to as “syndromic aortopathies,” these disorders are more appropriately regarded as “syndromic arteriopathies” as they (especially LDS and vascular Ehlers-Danlos syndrome [vEDS]) can cause aneurysms and dissections in virtually any arterial bed. The risk of dissection exists in vEDS even in arteries without existing aneurysmal dilatation (1), making this diagnosis challenging to achieve before an inceptive vascular catastrophe. The “nonsyndromic” heritable thoracic aortic aneurysm and dissection (hTAAD) disorders do not typically associate with noncardiovascular abnormalities and therefore often lack overt physical features. Inheritance of hTAAD disorders is autosomal dominant with variable expression and reduced penetrance.

Clinical genetic testing, typically using a targeted gene panel covering 24 to 50+ genes, can augment clinical phenotyping to potentially classify TAA as either “syndromic” or “nonsyndromic,” depending on whether a causative variant can be identified and the gene in which it resides. Furthermore, because the musculoskeletal and craniofacial dysmorphisms that aid recognition of MFS, LDS, and vEDS can overlap or be subtle/absent, genetic testing is often necessary to finalize diagnoses. Pathogenic or likely pathogenic variants in associated genes are found in almost all patients with LDS (2) and >98% of those with vEDS (3), and genetic confirmation is necessary for making these respective diagnoses. MFS is diagnosed according to criteria that incorporate clinical, family history, and genetic features (4), and ~20% of patients will not have a detectable mutation in the causally associated fibrillin 1 (FBN1) gene (5), although the medical and surgical recommendations discussed later are similar. Notably, in the absence of molecular confirmation through genetic testing, LDS may be misdiagnosed as MFS due to overlapping features (6).

Involvement of genetics specialists, including clinical geneticists and genetic counselors, can streamline the work-up of patients with TAA and typically enable collection of detailed family history data that inform risk and screening recommendations. In patients lacking overt syndromic features, genetic testing will yield an actionable result in ~30% of TAA cases. Nevertheless, a positive result may uncover a higher risk syndromic patient who otherwise would have gone undiagnosed, thus permitting tailored management. Up to 20% of nonsyndromic hTAAD cases are attributed to pathogenic variants in actin alpha 2 (ACTA2). Although a negative genetic testing result may lower the likelihood of a particular syndrome, it cannot be used to conclusively rule out any diagnosis.

Medical management of arteriopathies differs according to underlying etiology. Beta-blockers have long been advocated in MFS; however, their only randomized controlled trial in MFS was performed in 1994, which found that propranolol decreased the rate of aortic root dilatation (7). Subsequent non-randomized trials and case-control studies have been inconsistent in their findings (8,9), but a large retrospective review showed higher probability of survival in patients with MFS treated with beta-blockers (10). Based on these data, the use of beta-blockers in patients with TAA due to MFS carries a Class Ib recommendation in the American College of Cardiology/American Heart Association guidelines (11), predicated on reduction of aortic dilatation progression. These guidelines do not distinguish clinical diagnoses of MFS according to the updated Ghent criteria versus those made more specifically through identification of a pathogenic FBN1 variant. Some experts go further to advocate for initiation of beta-blockers at the time of MFS diagnosis, even in the absence of TAA (12). Both beta-blockers and angiotensin receptor blockers have been recommended for use in LDS (13), and recent data suggest this combination is also beneficial in MFS (14).

Other agents are currently under investigation in vEDS, targeting aortic biomechanical integrity (15). Although extrapolation of the aforementioned studies has led to prescription of antihypertensive agents in patients with nonsyndromic TAA, evidence supporting such therapies in normotensive TAA patients is lacking. Similarly, no evidence exists to support the prophylactic use of any medication in individuals with increased genetic risk for a nonsyndromic hTAAD who do not yet have TAA (e.g., genotype-positive relative of a patient with hTAAD due to an MYH11 variant). Thus, in normotensive patients with TAA, defining the underlying etiology, often possible only through genetic testing, determines medical management.

Risk of aortic complications varies according to underlying diagnosis in TAA; at any given aortic diameter, the risk for dissection is higher in MFS (16,17) and LDS (18) than in other forms of TAA, including most nonsyndromic hTAAD (16,19). Conversely, some forms of hTAAD may be of lower risk; aortic dilatation due to pathogenic variants in fibrillin 2 (FBN2) (associated with congenital contractural arachnodactyly) is mild (20), nonprogressive, and does not seem to pose a risk for dissection (21). Therefore, the timing of prophylactic aortic surgery hinges primarily on both maximum aortic root diameter and etiologic context, with surgery advised (in the absence of other high-risk features such as accelerated growth, pregnancy, significant valvular lesions, or malignant family history) at 5.0 cm in MFS, 4.0 to 4.6 cm in LDS, and 5.5 cm for most other forms of TAA according to expert consensus guidelines (11). Contemporary management following this approach improves patient survival (10). Recommendations in vEDS are less clearly defined, as dissections often occur without preceding aortic dilation, and surgical mortality can be high.

Prognostic information can occasionally be gleaned through recognition of specific genotype-phenotype associations. For example, in MFS, patients harboring FBN1 variants affecting cysteine residues are more likely to develop aortic dilatation and ophthalmologic manifestations, whereas nonsense, frameshift, or splice variants associate more strongly with orthopedic problems (22,23). Truncating and splicing FBN1 variants are more commonly found in MFS patients with aortic dissection or in those requiring surgery compared with those without (24). Mutations in exons 24–32 of FBN1 are associated with a more severe and complete phenotype, including younger age at diagnosis and higher rate of aortic complications (23). In LDS types 1, 2, and 3, aortic dissections have occurred at dimensions of 3.9 to 4.0 cm, whereas risk in type 4 may occur at a somewhat higher dimension; additional research is needed in this area (13). Among mutations in COL3A1 responsible for vEDS, those leading to a glycine substitution, arising at a splice site or introducing an inframe insertion or deletion, are associated with a more severe phenotype, whereas the so-called null alleles leading to haploinsufficiency (responsible for <5% of pathogenic variants in this gene) are associated with a milder phenotype and complications at older ages. Gastrointestinal ruptures that can arise as one of the features of vEDS are less common in the setting of haploinsufficiency and non-glycine variants (25).

Based on these data, it is evident that defining the genetic basis for an aneurysm disorder aids clinical decision-making. Indeed, the tailoring of pharmacological and surgical management to genetic etiology improves outcomes for patients and their families.

STATE-OF-THE-ART CLINICAL CARE: NONANEURYSMAL MENDELIAN VASCULAR DISEASES.

Although rarer than those that manifest chiefly as aneurysms, a number of nonaneurysmal Mendelian vascular disorders have also been described (Table 3). These include disorders leading to vascular malformations such as hereditary hemorrhagic telangiectasia (HHT) (26) or cerebral cavernous malformations (27). Additional examples include accelerated arteriosclerosis due to excessive mineralization and fragmented elastic fibers in pseudoxanthoma elasticum (28), as well as premature ischemic stroke due to vasculopathy in cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (29).

TABLE 3.

Summary of Other Mendelian Vascular Diseases

Syndrome	Causative Gene(s)	Vascular Phenotype	Associated Clinical Features	Clinical Actions (Per Syndrome)	Clinical Actions (All)
Hereditary hemorrhagic telangiectasia	ACVRL/ALK1 ENG SMAD4 * GDF2	Arteriovenous malformation (pulmonary > hepatic > cerebral) TAA or AAA*	Mucocutaneous telangiectasias Epistaxis Gastrointestinal bleeding Pulmonary hypertension Anemia Juvenile polyposis syndrome*	Screen for orthodeoxia Echocardiography bubble study Consider imaging of brain/liver for AVMs Consider referral to hematology and/or ENT Aortic imaging and surveillance (SMAD4) Referral to gastroenterologist (SMAD4)	Genetic counseling Genetic evaluation of at-risk relatives (cascade screening if causative variant is known in proband)
Cerebral cavernous malformations	KRIT1 (CCM1) CCM2 PDCD10 (CCM3)	Multiple cavernomas	Seizures Neurological deficits Headaches Intracranial hemorrhage	Brain imaging Consider neurology referral Avoidance of antithrombotic and antiplatelet agents unless indicated for life-threatening thrombosis Monitoring during pregnancy
Pseudoxanthoma elasticum	ABCC6	Accelerated arteriosclerosis	Claudication Angina Neck and flexural area skin lesions Peau d'orange retina Angioid streaks	Baseline and surveillance with retinal specialist Avoidance of aspirin and NSAIDS Counsel against smoking (avoid vasoconstriction) Counsel on lifestyle recommendations (avoid head/eye trauma)
Autosomal dominant polycystic kidney disease	PKD1 PKD2 GANAB DNAJB11	Intracranial aneurysms TAA/dissection (rare)	Bilateral renal cysts Liver cysts Mitral valve prolapse Abdominal wall hernias	Early BP monitoring Consider aortic screening (echocardiography) Consider screening for intracranial aneurysms
CADASIL	NOTCH3	Stenotic small caliber vasculopathy of arteries and arterioles	Premature stroke Migraine with aura White matter lesions	Control of vascular risk factors Neurology referral Caution with thrombolytics and oral anticoagulants

^*Pathogenic variants in SMAD4 can cause an overlap syndrome with hereditary hemorrhagic telangiectasia, TAA, and/or juvenile polyposis syndrome and increased risk for gastrointestinal cancer (30,31).

CADASIL = cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy; ENT = ear, nose, and throat specialist; NSAIDs = nonsteroidal anti-inflammatory drugs; other abbreviations as in Tables 1 and 2

Within these disorders, important genotype-phenotype associations are worth noting. Of the 4 genes associated with HHT, only mutations in SMAD4 are additionally associated with risk for TAA (30), as well as juvenile polyposis syndrome and heightened risk of gastrointestinal malignancies (31). Therefore, recognition of this specific subtype is of critical importance, as concern and surveillance for these important risks would only be raised with molecular confirmation through genetic testing. Although screening for pulmonary arteriovenous malformations (AVMs) is recommended for those with any form of HHT regardless of symptoms, those with ENG mutations are more likely to have asymptomatic lesions detected through routine surveillance (32). Although gastrointestinal AVMs can occur in HHT, surveillance is typically dictated by symptoms; however, patients with ACVRL1 mutations are more likely to have hepatic AVMs, and screening may therefore be reasonable even in the absence of symptoms.

STATE-OF-THE-ART CLINICAL CARE: CLINICAL GENETIC TESTING FOR MONOGENIC ARTERIAL DISEASES.

Targeted gene panel sequencing is typically used for cardiovascular genetic testing, and the cost of this approach has decreased dramatically. Usually, laboratories only analyze a pre-specified panel of genes. These targeted panels can identify missense variants, splice site changes flanking the exon, and insertions/deletions (as discussed in Part 1 of this Focus Seminar series) but are not poised to capture more expansive genetic changes and may miss important variants in rare or novel genes. Whole-exome or whole-genome sequencing may occasionally be useful for further interrogation when a genetic etiology is strongly suspected (e.g., a patient with significant systemic features or notable family history) despite negative panel test results; however, it may be costly and often requires submission of samples from additional relatives to inform interpretation. Finally, foundational genetic testing, including chromosome and microarray analyses, may reveal the genetic identity of an arteriopathy as a large chromosomal duplication or deletion, which may carry additional associations warranting specialized screening and surveillance (33–35).

Genetic arteriopathies typically follow autosomal dominant inheritance with variable penetrance and expressivity. Defining a genetic diagnosis in a “phenotype-positive” patient affords a straightforward method for identifying family members at increased genetic risk. Once the genetic etiology is known in the index patient (proband), this allows relatives to be genetically screened to determine whether they harbor the disease-causing variant. Baseline and surveillance vascular imaging is typically recommended for relatives identified as “genotype-positive” and may uncover critical lesions that might benefit from the disease-specific (or even genotype-specific) medical or surgical interventions and/or avoidance of high-risk activities and procedures outlined earlier. Individuals considering conception may use their genomic information for prenatal decision-making, including preimplantation genomic diagnosis with in vitro fertilization to preferentially implant only “genotypenegative” embryos.

STATE-OF-THE-ART RESEARCH: PRECISION MEDICINE FOR MENDELIAN VASCULAR DISEASES.

More widespread genetic testing for CVDs, especially arteriopathies, has led to the expansion of testing panels. Although improved awareness around genetics and increased implementation of testing add value, hTAAD panels offered through clinical laboratories increasingly include genes with limited evidence of a TAA association. This raises new challenges, as variants identified in those genes are typically reported as variants of uncertain significance (VUS). Disclosure of VUS that may not be truly causative can lead to unnecessary patient and familial anxiety, diagnostic confusion, inappropriate downstream testing, and genetic discrimination. Part 1 of this Focus Seminar series provides further discussion and explanation of VUS and the clinical challenges they pose (36).

For this reason, the Clinical Genome Resource (ClinGen) recently undertook an evaluation to determine the clinical validity of 53 “hTAAD genes.” Applying a semi-quantitative framework based on clinical, genetic, and experimental evidence, along with expert consensus, these genes were curated into evidence tiers. Nine genes (ACTA2, COL3A1, FBN1, MYH11, MYLK, SMAD3, TGF-β2, TGF-βR1, and TGF-βR2) were found to be definitively associated with hTAAD and clinically actionable. Despite common inclusion on hTAAD testing panels, 23 genes were found to have no evidence for association with TAA (21). As research on hTAAD increasingly uncovers new gene-disease links, reappraisal of the evidence for such associations will be needed to ensure that clinical genetic testing is restricted to genes with sufficient evidence to meaningfully affect patient counseling and clinical decision-making. Further research is needed to clarify the role of genetics in arteriopathies without TAA; for example, in diseases such as spontaneous coronary artery dissection, in which it is likely that a combination of Mendelian (37) and non-Mendelian (38) genetic effects contribute to disease pathogenesis.

FUTURE DIRECTIONS: INCREASING APPLICATIONS OF PRECISION MEDICINE TO MENDELIAN VASCULAR DISEASES.

Increasingly, asymptomatic genotype-positive individuals at risk for vascular disease will be ascertained through secondary findings from broader clinical testing and population genomic screening. Genomic screening is accompanying the recent proliferation of biobanks linking genomic and clinical data for precision medicine research, and their clinical utility seems promising (39). However, the undertaking of genomic screening for monogenic vascular disease will necessitate a better understanding of the prevalence and penetrance of pathogenic variants in diverse, unselected populations. Other important considerations are patient outcomes, such as the psychosocial impact of receiving a pathogenic variant result and the effect on diagnosis, management, and patient adherence to medical recommendations, particularly in the absence of symptoms or family history.

Furthermore, the impact of genomic screening goes beyond the patient to their family members and society, and must also consider financial costs and potential radiation exposure associated with lifelong surveillance imaging. Pilot biobank-based genomic screening programs will likely shed light on many of these issues. The American College of Medical Genetics and Genomics has highlighted disease-gene pairs recommended for return to patients from clinical genome or exome sequencing, regardless of the primary indication for testing (40). On this list are seven hTAAD genes, including those related to MFS, LDS, and vEDS. Therefore, sequencing performed for other indications may uncover patients at increased genetic risk for arteriopathy. As a result of these evolving recommendations, clinicians should be educated and will need to shift toward applying a genomics-first approach to patient evaluation and treatment. Finally, as these precision medicine approaches increasingly identify asymptomatic genotype-positive individuals at risk for vascular disease, this should usher in new clinical studies to investigate potential therapies that modify the natural history of these diseases, such that the need for surgery or other vascular complications can be avoided.

COMPLEX (NON-MENDELIAN) VASCULAR DISEASES

In contrast to the rarer Mendelian diseases already discussed, common complex diseases such as atherosclerosis typically involve many hundreds to thousands of genetic variants, also known as SNPs, that each impart a weak to modest effect in causing disease. Our understanding of the genetic contribution toward the risk of common complex vascular diseases is driven mostly through genome-wide association studies (GWASs). By comparing the allele frequency of naturally occurring variants (i.e., SNPs) in individuals with versus without a given disease, an increasing number of GWAS risk loci have been discovered for complex vascular diseases such CAD, stroke, and hypertension (41). These studies involve SNP genotyping of large human cohorts to discover genome-wide significant associations (p < 5 × 10⁻⁸). Reduced costs of genotyping, along with the formation of large genetic consortia, have enabled an exponential growth in vascular GWAS discoveries over the past decade (Figure 1).

FIGURE 1. Summary of GWAS to Date for Selected CVDs.

The 4 panels show summaries of representative genome-wide association studies (GWAS) for hypertension, stroke, coronary artery disease (CAD), and total cholesterol. The columns represent the study size of GWAS indicated on the x-axis, showing the generally progressive increase in size over the last decade and corresponding to the vertical axis of “Thousands” or “Millions.” The blue line represents the cumulative number of published GWAS in these fields and corresponds to the vertical axis “Count.” CVDs, cardiovascular diseases.

Because the genetic make-up of these common vascular diseases typically involves hundreds to thousands of genes, and thus the genetic disease contribution differs for essentially every individual, it is inherently more challenging to devise and implement precision medicine approaches for complex diseases than for Mendelian vascular diseases. For example, as discussed previously, for certain Mendelian arteriopathies, the knowledge of a specific genotype can already be used to screen family members, inform clinical risk, and guide decision-making. In contrast, as we discuss later in detail, research is ongoing to devise and evaluate mathematical scoring algorithms termed polygenic risk scores (PRSs) for integrating genotypes for an individual at thousands of SNPs to predict the likelihood of developing clinically significant CAD and other complex diseases. Therefore, much of this non-Mendelian section of the current article is devoted to considering ongoing research and future directions for the application of precision medicine approaches to complex vascular disorders, which, due to the inherent nature of these diseases, will take a longer time to mature and be implemented than for the Mendelian arteriopathies.

STATE-OF-THE-ART CLINICAL CARE: FAMILIAL HYPERCHOLESTEROLEMIA AS A MENDELIAN CAUSE OF CAD.

Before we embark on our consideration of these many genetic variants that cause atherosclerosis and CAD, it important to first consider some rarer Mendelian causes of these diseases. In particular, although the great majority of CAD cases arise due to the effects of many SNPs and other environmental effects acting in concert, certain cases of familial CAD are secondary to rarer Mendelian disorders such as familial hypercholesterolemia (FH), which is characterized by high cholesterol levels and specifically by very high levels of low-density lipoprotein (LDL)-cholesterol. FH is inherited in an autosomal dominant pattern, and many of the lipid loci for CAD identified by GWASs contain gene targets (e.g., LDLR, APOB, PCSK9) for mutations linked to this disease. Given the need to identify FH patients early, as with the Mendelian arteriopathies, the evidence-based cascade screening of families for potential probands (e.g., through routine lipid testing and automated electronic health record searches) has been a successful approach to reduce mortality and morbidity in this high-risk group (42). The detection of FH patients by targeted genotyping or gene sequencing of causal pathogenic variants has informed tailored management of these patients with more aggressive lipid-lowering medications such as proprotein convertase subtilisin/kexin type 9 inhibitors (e.g., evolocumab, alirocumab).

STATE-OF-THE-ART RESEARCH: GWASs FOR CAD AND HYPERTENSION.

Turning to the common non-Mendelian causality of CAD, there are now at least 168 reported genome-wide significant loci (conditionally independent primary or secondary signals with p < 5 × 10⁻⁸) and >400 suggestive loci (independent loci at p < 1 × 10⁻⁴ or false discovery rate <5%) (43). As reviewed elsewhere (43,44), these loci map to genes organized in discrete biological processes and signaling pathways, such as TGF-β signaling, lipid metabolism, inflammation, cell cycle regulation, and extracellular matrix remodeling (Figure 2). As with other complex diseases, GWAS discoveries for CAD and myocardial infarction have been accelerated through meta-analyses of individual GWASs in large international consortia such as CAR-DIoGRAMplusC4D, as well as the UK Biobank. As expected, the majority of these lead associations are common (mean minor allele frequency: 0.44) with modest effect sizes (mean odds ratio: 1.07) per allele. More than one-half of the CAD loci seem to be unrelated to classical risk factors such as hypertension or hyperlipidemia, and at least one-third harbor genes with unknown functions (41,43,45). Although these discoveries have expanded our understanding of the biological mechanisms of CAD, a significant gap remains in translating this knowledge to the clinic. Extensive research will be required to now understand causal GWAS variants, genes, cell types, pathways, and phenotypes using disease-relevant preclinical model systems.

FIGURE 2. Annotated Genes and Pathways Associated With CAD and/or MI.

Genes identified by genome-wide association studies mapped to 171 coronary artery disease (CAD) and myocardial infarction (MI) risk loci are shown clustered into key pathobiological pathways. cGMP = cyclic guanosine monophosphate; ECM = extracellular matrix; NO = nitric oxide; TGF-beta = transforming growth factor-beta. The figure was produced by using BioRender. Adapted with permission from Erdmann et al. (43).

Hypertension is another common complex disease and a leading risk factor for CAD and also peripheral artery disease, stroke, heart failure, and chronic kidney disease. Although only atherosclerosis and CAD are considered in detail in the current article, much of the discussions and themes that follow are also applicable to hypertension and other complex disorders. Given that blood pressure is highly heritable, with heritable factors accounting for 30% to 50% of a given individual’s blood pressure, genetic associations can be readily measured across large populations. As for CAD, large GWAS sample sizes involving consortia such as the International Consortium for Blood Pressure (46) and the UK Biobank (47) have been necessary to make headway in identifying SNPs associated with blood pressure and hypertension. Studies conducted by these groups involved mostly European participants and detected loci for mean pulse pressure, systolic blood pressure, and diastolic blood pressure. More recent metaanalyses in the Veterans Affairs Healthcare System’s Million Veteran Program (MVP) included a trans-ethnic analysis of these blood pressure determinants in up to 776,078 individuals by combining MVP data with data from the International Consortium for Blood Pressure, UK Biobank, and Vanderbilt University’s BioVU cohort (48). With this large-scale analysis, there are now 505 independent loci associated with one or more blood pressure traits. Similar trans-ethnic case-control meta-analyses in MVP identified 19 loci for peripheral artery disease, using electronic health record-derived phenotypes, as done previously for UK Biobank (49). This work highlighted an overlap of genetic loci associated with diseases of multiple vascular beds and suggested the potential for common therapeutic interventions for differing vascular disorders (i.e., CAD and peripheral vascular disease). In summary, the identification of these loci via GWASs, and understanding their mechanisms of effect, is an important step toward precision medicine for common complex CVDs.

STATE-OF-THE-ART RESEARCH: CAD RISK LOCI HARBORING LIPID GENES AS DRUG TARGETS FOR CVD.

It is now appreciated that drugs with some form of genetic data to support their actions or effects are approximately twice as likely to obtain approval from the U.S. Food and Drug Administration (50,51). This is more obvious when the GWAS variants are missense or deleterious coding variants. A prototypical example of the GWAS approach identifying druggable CAD gene targets is the HMGCR gene, which encodes hydroxy-3-methylglutaryl-coenzyme A reductase, the target of statin therapies (44). As another example, the proprotein convertase enzyme PCSK9 harbors a missense variant (rs11591147; R46L) highly associated with CAD (52,53) and LDL-cholesterol levels (54,55). This variant is correlated with a 15% reduction in LDL-cholesterol and a 2- to 3-fold reduction in CAD risk, which mimics the potent effects of monoclonal antibody inhibition of proprotein convertase subtilisin/kexin type 9 in reducing LDL-cholesterol levels (56). Additional genetic support for non-LDL-specific CAD-associated genes such as LPA [encodes lipoprotein(a)], or triglyceride-associated genes APOC3 and ANGPTL3, have led to the rapid development of antisense oligonucleotide therapies that have achieved lipid lowering of up to 92%, 73%, and 50% through blockade of apolipoprotein (a), apolipoprotein C-III, and angiopoietin-like protein 3, respectively (57).

Importantly, additional studies that leverage natural genetic variation to understand disease causality, termed Mendelian randomization analyses, have provided causal support for lower LDL-cholesterol, triglyceride, and lipoprotein(a) levels in reducing CAD risk (58,59). This is consistent with the clinical benefits observed in large randomized control trials such as IMPROVE-IT (Improved Reduction of Outcomes: Vytorin Efficacy International Trial), while conversely, the lack of data for high-density lipoprotein cholesterol levels and cholesteryl ester transfer protein as being causal for CAD is consistent with the lack of clinical efficacy for high-density lipoprotein raising or cholesteryl ester transfer protein inhibition in CAD outcomes (58). Together, these findings highlight the value of unbiased GWAS discovery of causal pathways that could be targeted therapeutically to manage patients with hyperlipidemia and prevent CAD and related comorbidities.

STATE-OF-THE-ART RESEARCH: GENETIC EVIDENCE FOR TARGETING INFLAMMATION IN CAD.

Besides the highly tractable lipid gene targets for CVD therapeutic development as discussed earlier, other genetic loci have pointed to inflammation as a disease target. For instance, the IL6R locus (encoding interleukin-6 receptor) harbors missense variants (e.g., rs2228145) associated with reduced C-reactive protein (CRP) and risk for CAD (60). Despite the lack of causal evidence for CRP on CAD risk, anti-inter-leukin-6 receptor (IL-6R) antibody (tocilizumab) treatments showed promise in Phase II trials in patients with myocardial infarction and are undergoing further evaluation in the ASSAIL-MI (Assessing the Effect of Anti-IL-6 Treatment in Myocardial Infarction) trial (NCT03004703). IL6R has been causally linked to CAD and other common vascular diseases such as abdominal aortic aneurysm and stroke, presumably by increasing the downstream effects of IL-6 to promote vascular inflammation.

The randomized controlled CANTOS (Canakinumab Anti-Inflammatory Thrombosis Outcome Study) also provided evidence that targeting interleukin-1β, via canakinumab, reduces major cardiovascular events along with IL-6 and CRP levels in patients with CAD, despite the narrow therapeutic dosing window and infection risks (61). In addition, a series of studies investigating the anti-inflammatory drug colchicine have suggested potentially relevant cardiovascular beneficial effects (62,63). Collectively, these studies are providing substantial insights into the possibility for targeted anti-inflammatory treatments for post-myocardial infarction and stable CAD patients (62–65). Although currently speculative, a potential precision medicine approach for the use of these agents, given that anti-inflammatory treatments tend to raise the risk of infection, might be to target CAD patients with particularly elevated inflammatory markers in whom the risk of infection is outweighed by the benefits of reducing inflammation.

STATE-OF-THE-ART RESEARCH: PATIENT-SPECIFIC ANTITHROMBOTIC DRUG DECISION-MAKING.

The antithrombotic P2Y12 inhibitors (e.g., clopidogrel, ticagrelor, prasugrel) are widely used to manage patients after myocardial infarction or coronary stent implantation. Given their inherent risk of bleeding complications, there has been a longstanding interest in developing personized antithrombotic regimens for patients requiring these agents. In turn, this has spurred pharmacogenomic studies of cytochrome P450 enzymes such as CYP2C19 that govern the conversion of clopidogrel to an active metabolite. Several large studies showed promise in genotype-guided therapy (66,67), consistent with the observation that 14% of patients harboring certain CYP2C19 loss-of-function variants (CYP2C19*2 and CYP2C19*3) are at increased risk for major adverse cardiovascular events such as myocardial infarction or stroke (68). A large (n = 5,302) prospective randomized trial, TAILOR-PCI (Tailored Antiplatelet Therapy Following PCI; NCT01742117), was designed to evaluate the efficacy of genotype-guided therapy versus conventional therapy in patients after percutaneous coronary intervention (67). Among all randomized patients, the primary endpoint occurred in 4.4% of the genotype-guided group and 5.3% of the conventional group (p = 0.16). This borderline result broadly typifies this area of research.

In summary of the many studies conducted, clinical trials evaluating treatment strategies guided by these stratification methods have produced mixed results, and routine genetic testing to aid in personalizing these therapies is not routinely recommended. Likewise, multiple risk scores based on demographic factors such as age, body mass index, and chronic kidney disease have been developed to guide therapeutic choices, but validation studies have yielded conflicting results, and there is limited evidence that their adoption improves clinical outcomes. Nevertheless, this remains an area of ongoing research, and future studies will need to identify how to best stratify patients to subsequently provide patient-tailored precision antithrombotic regimens (69).

STATE-OF-THE-ART RESEARCH: POLYGENIC RISK SCORES FOR CVD PREDICTION AND RISK STRATIFICATION.

Although low-frequency coding variants (0.5% to 5%) may explain more severe and early-onset CAD (70,71), as already discussed, the majority of the heritable effects on CAD at the population level are mediated through common variants (minor allele frequency >5%) (Table 4) (41). As such, a major prospect for precision approaches to CAD and other complex diseases is to develop genetic PRSs to identify high-risk patients and improve upon current prognostication and clinical management. PRSs comprise complex mathematical scoring algorithms that integrate genotypes at thousands of SNPs to predict the likelihood of developing CAD or other complex diseases (72). A weighted sum is calculated by multiplying each allele by the GWAS effect sizes, which is then subjected to p value thresholding and other statistical fine-tuning. The clinical utility of PRSs for a given trait or disease require rigorous testing for associations with clinically measured risk factors, biomarkers, and outcomes. Typically, the odds or hazard ratio for the PRS effect estimates are compared with more traditional risk metrics (e.g., Framingham Risk Score).

TABLE 4.

Estimated Heritability for Selected Cardiovascular Diseases

Trait	Twin-Based (European)	SNP-Based (European)	SNP-Based (African)	Liability Explained by Disease-Associated SNPs	Ref. #
Coronary artery disease	0.53	0.22	-	8.9%	(101–103)
Diastolic blood pressure	0.44–0.66	0.21	0.24	0.71%	(104–107)
Systolic blood pressure	0.53–0.66	0.21	0.18	1.05%	(104–107)
Cholesterol	0.51	0.15	0.53	2.84%	(105–108)
Stroke	0.0–0.32	0.34–0.42	0.35	0.01%	(109–112)

Heritability estimates were calculated by using Twin-based and single nucleotide polymorphism (SNP)-based strategies in European and African ancestry populations.

PRSs for CAD risk prediction have been evolving for several years (Supplemental Table 1 provides a summary of PRS for CVDs). Starting initially with relatively few SNPs, recent studies have shown improvement of both CAD risk stratification and prediction by incorporating a large number (~49,310) of SNPs (73). It is believed that these 49,310 variants contain more genetic information that reliably captures the lifetime of risk exposure compared with earlier PRSs. Most recently, studies using UK Biobank have tested PRSs containing millions of common variants (“genome-wide PRS”), which resulted in further improved risk stratification (74,75). Importantly, a score using 6.6 million variants captured up to a 5-fold increased risk for individuals with the top 1% of variants, comparable to the impact of rare Mendelian variants (74). In another study, Inouye et al. (75) found similar results and high discrimination accuracy from a PRS with 1.7 million variants when combined with conventional risk factors. As expected, individuals on lipid-lowering or antihypertensive medications had reduced composite risk scores, although substantial residual risk remained.

Combined monogenic (Mendelian) and polygenic risk effects on CVD risk.

Although the great majority of CAD cases arise due to polygenic influences, as discussed, FH and other Mendelian disorders can also lead to CAD. Given the combined risk of both rare Mendelian and common GWAS variants, it is plausible that individuals harboring both types of variants are at greatest risk for developing CVD events. Indeed, among 725 individuals with a causal FH mutation, those in the highest CAD PRS tertile had a greater prevalence of events compared with those in the lowest tertile (40.9% vs. 24.7%) (76). This finding suggests that in the setting of a severe monogenic disease such as FH, the polygenic background can modify CVD risk. Similar observations were made in a whole-genome sequencing analysis of early-onset myocardial infarction, in which those with FH mutations and high CAD PRS had higher LDL-cholesterol levels (77). Finally, a study combining data from the UK Biobank and Color Genomics further evaluated the relationship of monogenic and polygenic risk for CAD and identified a large gradient of risk (5% for FH variant noncarriers to 78% for FH variant carriers) (78). These results support the notion that incorporation of PRSs for CAD and other CVDs may increase the precision of risk estimation even in those harboring monogenic variants and could potentially inform more individualized treatment decisions and/or clinical trial designs.

Clinical utility of PRSs for CAD and current challenges.

A number of studies have evaluated the performance of “genome-wide” variant CAD PRSs in various populations, with conflicting results. Some of the differences may be attributed to differences in study composition, including ancestry, age of participants, and/or focus on prevalent CAD versus incident CAD. For instance, in the MESA (Multi-Ethnic Study of Atherosclerosis) cohort, a genome-wide CAD PRS only had a modest association with incident CAD and no added value above conventional risk factors (e.g., American College of Cardiology/American Heart Association pooled cohort equations) (79). Similarly, a genome-wide CAD PRS study in the UK Biobank only had incremental increases in predictive ability over the pooled cohort equations after adjusting for age and sex (80). Therefore, at the current time, PRSs based on GWAS risk variants have yet to establish clinical utility and thus remain as research tools. It is possible that the addition of other -omics-based molecular profiles (e.g., plasma protein levels [81]) and personalized whole-genome sequencing-based discovery of rare and somatic mutations related to CVD risk will enhance PRS predictions. However, we believe that significant improvements are needed before the widespread adoption of PRSs in clinical decision-making and precision medicine approaches is justified.

FUTURE DIRECTIONS IN GWAS AND PRS RESEARCH.

Strikingly, nearly 80% of all GWAS participants are of European ancestry, although this group only constitutes 16% of the global population (82). This lack of diversity in available GWAS datasets results in poor generalizability of European-derived PRSs across racial and ethnic groups, with ~1.6-, 1.7-, 2.5-, and 4.9-fold lower prediction accuracy in Hispanic/Latino American, South Asian, East Asian, and African subjects, respectively. A study from the large multi-ethnic eMERGE (Electronic Medical Records and Genomics) cohort compared both a “restricted” and “genome-wide” ancestry-specific CAD PRS to predict incident CAD in individuals of European ancestry, African ancestry, and Hispanic ethnicity (83). Although PRSs derived from European ancestry adults performed similarly in adults of Hispanic ethnicity, they had poor associations with incident CAD in individuals of African ancestry. A combination of differences likely explains these discrepancies, including patient selection bias, allele frequency, effect size, associations between different SNPs (known as “linkage disequilibrium”), and gene-environment interactions. Although improved statistical methods accounting for admixture and cryptic relatedness can reduce population stratification effects, a more straightforward solution is to increase diversity in GWASs for CVD and other traits.

Early CAD GWAS meta-analyses combined European and South Asian ancestry individuals, which identified mostly shared loci between these populations, supporting combined trans-ethnic analyses to improve discovery (84). This was largely consistent with meta-analyses of multi-ancestry cohorts including European, East Asian, South Asian, and African-American individuals (85). It is worth noting that some loci (e.g., PHACTR1, CXCL12, COL4A1/COL4A2) had stronger evidence of association despite smaller samples in South Asian subjects compared with other ancestry groups. However, until recently, compared with the large European cohorts, there have been limited studies of similar sizes to elucidate true ancestry-specific signals. A recent meta-analysis of 168,228 Japanese individuals (25,892 case subjects and 142,336 control subjects) identified 8 novel CAD risk loci (86). In this same study, a trans-ethnic meta-analysis revealed an additional 37 novel loci, which improved CAD PRS prediction compared with loci identified from Japanese or European GWASs alone.

Despite the higher prevalence of CAD, CAD-related mortality, and hypertension risk factors in African-American subjects, large-scale GWAS data in this population are still lacking. Future meta-analyses of the MVP datasets along with cohorts from other studies with African-American and Hispanic participants should increase the discovery of ancestry-specific loci in ancestrally diverse and underrepresented populations (87). With increased diversity of CAD GWASs and their findings, we can better use PRS combined with clinical predictors to target high-risk individuals in non-European populations while also mitigating ongoing health care disparities (88). This is particularly important given that pathogenic GWAS variants identified in individuals of European ancestry may actually be protective in non-European ancestry individuals, and vice versa.

STATE-OF-THE-ART RESEARCH: MOVING BEYOND GWASs TO A GENE REGULATORY NETWORK FOCUS FOR CVD.

Despite the increasing number of risk alleles identified by GWASs at genome-wide significance (p < 5 × 10⁻⁸), there are many sub-genomewide significant associations (i.e., p < 1 × 10⁻⁴ but not p < 5 × 10⁻⁸) that may or may not be important for CAD and cardiometabolic diseases. Many of these suggestive loci harbor genes that could play fundamental roles in the initiation or progression of CVD (41). In fact, by considering the total variation in GWAS datasets, at least 50% of the genetic variability of CAD and cardiometabolic diseases can be captured (89). This is consistent with the fact that, as discussed, genome-wide PRSs with millions of variants seem to offer improved prediction for CVDs. Collectively, this finding argues that there is likely to be more disease-relevant information to extract from GWAS datasets outside the genome-wide significant risk loci. The key question is, how should this information be leveraged?

CVDs are systemic disorders that are driven by both inherited genetic and environmental risk factors. Thus, research strategies are now being devised and implemented that leverage multi-dimensional, multiorgan -omics data collected from diverse populations of healthy and diseased subjects to unravel these complex causal disease mechanisms. These studies can pinpoint potentially actionable changes in metabolic organs that collectively drive disease-causing processes. Using CAD as an example, by considering the full spectrum of molecular interactions across primarily metabolic tissues (e.g., intestine [lipid absorption, microbiome], liver [lipid and glucose processing], adipose [inflammation and lipid storage], skeletal muscle [glucose uptake and diabetes], and the arterial wall [atherosclerosis]), we can track the complex etiology of CAD across different patients. With these novel multi-omics studies, we can begin to identify more potent system-wide therapeutic targets and biomarkers that more accurately diagnose, treat, and prevent cardiometabolic disease and CAD (41). This work will depend on the maturation of bulk and single-cell ribonucleic acid and protein assay technologies toward the clinic, to accurately measure patient transcriptomes and proteomes across accessible human tissues (e.g., biopsy samples) and whole blood.

As we move toward leveraging multi-dimensional, multiorgan -omics data, a concurrent evolution is poised to occur in how we conceptualize complex diseases. Rather than considering single genes and pathways, as has been the typical approach until now, cutting edge research is increasingly considering entire networks of hundreds or thousands of genes that act in concert to cause disease pathogenesis. Referred to as “systems biology,” this approach has great potential to leverage GWAS data beyond candidate genes, by inferring genetically and environmentally regulated disease-driving gene networks (“gene-regulatory networks” [GRNs]). GRNs are empirically derived from transcriptome and proteome data by constructing weighted co-expression networks and subnetworks, which have been found to be active both within and between tissues (41,90–92). Unlike individual genes or pathways, GRNs explain a large portion of clinical variation inherent in complex disorders such as CAD. In fact, we recently showed that the genetic regulation of GRNs is responsible for a large fraction of the heritability of CAD, which is not accounted for by lead SNPs identified by GWASs (91).

Using advanced computational methods, it has been possible to understand that certain genes within a GRN exert profound effects on most of the other genes in the network. As a simplistic analogy, this is somewhat like an “apex” gene at the top of a pyramid of other genes: when the top apex gene is affected, most of the other downstream genes are also affected. The apex genes that exert the most powerful effects on an entire GRN are more correctly termed “key driver” genes (93,94). Perturbation experiments of key driver genes, by increasing or inhibiting their expression in vitro or in vivo, have shown their efficacy in modulating the activity of entire GRNs as well as downstream phenotypes associated with these networks, including CAD (94,95). This latter characteristic has prompted the term “key disease driver” (41,94). As an exciting recent research development, several studies have shown that key disease drivers are likely to be particularly promising therapeutic targets. For example, in a recent study, the transcription factor MAFF was identified as a novel central regulator of an atherosclerosis/CAD-relevant network in the liver (96). MAFF triggered context-specific expression of the receptor for LDL and other genes known to affect CAD risk, thus identifying it as a possible treatment target.

FUTURE DIRECTIONS: PATIENT-SPECIFIC NETWORK APPROACHES TO GUIDE THERAPIES AND PREDICT CLINICAL OUTCOMES.

The causal nature of GRNs and their key drivers provides an exciting new mechanistic framework to decipher complex disease biology not accounted for by individual GWAS risk loci. This framework holds great promise for understanding many differing aspects of disease pathobiology, including heritability (91), drug toxicity (97), disease variability from one individual to another, gene-environment interactions, and organ-to-organ communication across organ borders (98). Furthermore, this can be extended to other -omics features such as plasma proteomic, metabolomic, and lipidomic datasets, to better understand endocrine and other signaling in CVDs (81). In particular, plasma proteomics also seems to hold great promise for predicting risk of disease, and may complement the PRS as a personalized approach to disease detection and risk stratification (81,99).

The long-term goal in the evolution of these emerging pipelines will be to integrate the framework of multi-organ GRNs with detailed clinical datasets and outcomes to infer patient-specific trajectories and responses to medication and lifestyle changes. In the near-term, we can already integrate multi-omic regulatory networks with GWASs and Mendelian variation to improve the interpretability and clinical utility of PRS for CVDs (Figure 3). Recently, the integration of genetic loss-of-function variants, ribonucleic acid-sequencing data from the STARNET dataset, and electronic health record data for CVD patients in the BioMe Biobank, identified known (e.g., APOC3) and novel (e.g., DGAT2) targets amenable to therapeutic intervention (39). In another study, it was recently shown that GRNs in the atherosclerotic arterial wall exhibit clear differences between male and female subjects, raising the possibility of sex-specific GRN-guided therapies to manage and treat CVD (100). Finally, yet another study has indicated the potential of these multi-omic network approaches to predict toxicity or, specifically in this case, to explain the observation that individuals treated with cholesterol-lowering statins have an increased risk of developing diabetes. By integrating gene expression, genotype, metabolomic, and clinical data from the STARNET study, the authors identified a glucose-and lipid-determining GRN showing inverse relationships with lipid and glucose traits (97). Key drivers of this network influenced lipid and glucose levels in inverse directions, thus potentially explaining the aforementioned statin side effect. Along these lines, as we fully develop this systems-based understanding of the genetic architecture of complex CVDs, this is expected to open the door to multiple future precision medicine approaches that improve differing aspects of individual patient risk stratification and management.

FIGURE 3. Integration of Mendelian Genetics, GWAS, and GRNs to Improve PRS.

This schematic shows a proposed pathway to integrate Mendelian genetics, genome-wide association studies (GWAS), and multi-omic regulatory networks to improve interpretability, accuracy, and clinical utility of polygenic risk scores (PRS). EHR = electronic health record; GRN = gene regulatory network. The figure was produced by using BioRender.

CONCLUSIONS

Clinical vascular diseases are diverse in their genetic and pathological basis, phenotypes, and clinical management, as well as their current and emerging precision medicine approaches. At least for the Mendelian vascular disorders, precision medicine approaches are already a part of contemporary clinical management, with genetic diagnostic testing, cascade screening of family members, and tailored therapeutic approaches such as the timing of prophylactic aortic surgery based on aortic size and genetic etiology. For atherosclerosis and CAD, the profound complexity of the thousands of genes and loci that are involved has meant that very different approaches will be required to achieve individualized precision medicine. Nevertheless, PRS and a network- and systems-based holistic understanding of disease are some of the many exciting new prospects for precision care in patients with common complex disorders. These orthogonal approaches to individualized precision medicine for vascular disease are collectively driving major advances in patient care. As perhaps the most exciting aspect for both clinicians and patients, as we integrate high-throughput genetic and multi-omics analyses in the coming years, the scope and utility of these vascular precision medicine approaches will only continue to increase.

HIGHLIGHTS.

• Inherited vascular diseases include those associated with arterial enlargement (dilatation, ectasia, or aneurysm) or impaired arterial wall integrity (dissection or rupture).

• Arterial dilatation, particularly in young patients, should prompt evaluation for a genetic etiology.

• Atherosclerosis generally arises due to the interplay of environmental and lifestyle factors, and hundreds to thousands of single nucleotide polymorphisms, each with a relatively modest disease-causing effect.

• Although the genetics of atherosclerosis and Mendelian arteriopathies may differ, integrated methods are emerging to apply precision medicine to these diseases.

ABBREVIATIONS AND ACRONYMS

AVM

arteriovenous malformation

CAD

coronary artery disease

CRP

C-reactive protein

CVD

cardiovascular disease

FBN1

fibrillin 1

FH

familial hypercholesterolemia

GRN

gene regulatory network

GWAS

genome-wide association study

HHT

hereditary hemorrhagic telangiectasia

hTAAD

heritable thoracic aortic aneurysm and dissection

IL-6R

interleukin-6 receptor

LDL

low-density lipoprotein

LDS

Loeys-Dietz syndrome

MFS

Marfan syndrome

MVP

Million Veteran Program

MYH11

smooth muscle myosin heavy chain 11

PRS

polygenic risk score

SNP

single nucleotide polymorphism

TAA

thoracic aortic aneurysm

TGF-β

transforming growth factor-β

vEDS

vascular Ehlers-Danlos syndrome

VUS

variant of uncertain significance