Association of PCSK9 Loss-of-Function Variants With Risk of Heart Failure

This case-control study investigates the association of PCSK9 variants with cardiac structure, cardiac function, and heart failure in humans.

Key Points

Question

Are PCSK9 genetic variants associated with altered cardiac structure, function, or heart failure in humans?

Findings

In this case-control study of 35 135 individuals nested within the UK Biobank, using genotyping and exome-sequencing data, common variants, the R46L variant and loss-of-function variants, serving as proxies for pharmacological PCSK9 inhibition, were identified. By combining genetic data with cardiac magnetic resonance imaging data, no significant associations were found between carrier status or genetic risk score and 11 cardiac traits.

Meaning

Results of this study suggest that there were no associations between PCSK9 genetic variants and cardiac remodeling or increased risk of heart failure in humans.

Abstract

Importance

An animal (mouse) study indicated that deficiency of proprotein convertase subtilisin/kexin type 9 (PCSK9) causes cardiac remodeling and heart failure (HF). Cardiac remodeling after PCSK9-inhibitor treatment is a concern for patients and for development of treatment directed against PCSK9.

Objective

To determine whether genetic variants in the PCSK9 gene are associated with altered cardiac structure, cardiac function, and HF in humans.

Design, Setting, Participants

This was a nested case-control study within the UK Biobank. Between March 13, 2006, and October 1, 2010, the UK Biobank enrolled 502 480 individuals aged 40 to 69 years. This study focused on a subset of those individuals, who completed cardiac magnetic resonance (CMR) imaging and had available genetic data. Analyses were conducted between November 2, 2021, and October 28, 2022.

Exposures

Carrier status of predicted loss-of-function (pLoF) PCSK9 variants, R46L missense variant, and a genetic risk score (GRS).

Main Outcomes and Measures

A total of 11 CMR imaging measurements, generated using a machine learning algorithm, and HF diagnosis.

Results

In up to 35 135 individuals with CMR images, 18 252 (52%) were female individuals, and mean (SD) age was 55.0 (7.4) years. No significant association between PCSK9 carrier status and CMR indices were found for left ventricular mass (pLoF: β = −1.01; 95% CI, −2.99 to 0.98; P = .32; R46L: β = −0.18; 95% CI, −0.55 to 0.19; P = .35; GRS: β = −0.19; 95% CI, −0.50 to 0.11; P = .22) and left ventricular ejection fraction (pLoF: β = 0.43; 95% CI, −1.32 to 2.18; P = .63; R46L: β = −0.19; 95% CI, −0.52 to 0.14; P = .26; GRS: β = −0.08; 95% CI, −0.35 to 0.20; P = .58) or HF (pLoF: odds ratio [OR], 1.14; 95% CI, 0.56-2.05; P = .69; R46L: OR, 0.99; 95% CI, 0.90-1.10; P = .91; GRS: OR, 1.04; 95% CI, 0.96-1.13; P = .32).

Conclusions and Relevance

Results of this case-control study suggest that there was no association between PCSK9 genetic variants and altered cardiac structure, cardiac function, or HF in humans.

Introduction

The efficacy of proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibition to treat hypercholesterolemia has recently attracted widespread interest.^1,2,3 Currently, monoclonal antibody and silencing RNA therapies against PCSK9 have been developed to treat hypercholesterolemia. Although the overall safety profile of PCSK9 inhibitors is favorable,^{4,5,6,7,8,9,10,11,12,13,14,15} concerns regarding potential adverse cardiac remodeling have been raised.¹⁶ A recent animal study by Da Dalt et al¹⁷ demonstrated that PCSK9 deficiency reduces running resistance and impedes cardiac lipid metabolism by introducing increased cardiomyocyte accumulation of lipid droplets and development of heart failure (HF) with preserved ejection fraction (HFpEF) in mice. To translate their findings in humans, the authors used the well-categorized R46L PCSK9 genetic variant as a proxy for PCSK9 inhibition and evaluated the association with echocardiographic-based markers of cardiac functionality. The authors reported an association between R46L carrier status and increased left ventricular mass (LVM) in 12 heterozygous carriers. Aberrant cardiac remodeling following PCSK9-inhibitor treatment is an obvious concern not only for patients treated with marketed PCSK9 inhibitors but also for ongoing development of novel treatment modalities directed against PCSK9, including oral and antisense oligonucleotide treatment. Therefore, it is important to revisit the association between PCSK9 inhibition and cardiac structure and function in a larger and more well-powered setting.

Genetic variation can serve as proxies for pharmacological inhibition and provide estimates for the long-term effects of target inhibition.^18,19 Naturally occurring genetic variations are inherited randomly at conception and remain unchanged throughout life and are therefore robust to reverse causation and residual confounding. In this study, we incorporated a large data set of both rare predicted loss-of-function (pLoF), the R46L missense variant, and common genetic variation in PCSK9, to evaluate the association of PCSK9 variants with cardiac structure and function.

Methods

Study Population

The UK Biobank (UKB) was a large population-based cohort study that enrolled 502 480 community-dwelling persons in the UK aged 40 to 69 years at recruitment between March 13, 2006, and October 1, 2010.²⁰ We conducted a nested case-control study within the UKB focusing on participants who underwent cardiac magnetic resonance imaging (CMR). The study has been conducted using the UKB resource under application number 43247. We included participants with available genetic (genotyping and/or exome-sequencing data) and CMR data. At the time of this study, approximately 500 000 participants had available genotyping data, 200 000 participants had exome-sequencing data, and 40 000 participants had CMR imaging data available.²¹ Only unrelated participants of European ancestry were included. Relatedness and ancestry definitions are provided in eMethods and eFigure 1 in the Supplement. All participants provided written informed consent, and the study complies with the Declaration of Helsinki. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Selection of Rare pLoF Variants

We used exome-sequencing data for PCSK9 from unrelated individuals of European ancestry (approximately 170 000 individuals). Information on exome-sequencing methodology, including alignment and variant calling in UKB, has been described in detail previously.²² In brief, variant filtering was applied to exclude variants with genotype quality less than 20, genotype depth less than 10, and missing genotypes greater than 0.1. Variant annotation was performed using SnpEff software (Pablo Cingolani).²³ Positional intersection with the Ensembl ENSG00000169174 transcript was used to obtain variants in PCSK9. All pLoF variants were defined as (1) variants leading to loss of a start or stop codon or to a premature stop codon; (2) open reading frameshifting indels leading to the formation of a premature stop codon; and (3) variants or indels disrupting canonical splice acceptor or donor sites. Sequence Ontology²⁴ standardized terms with putative high impact were prioritized. Furthermore, we evaluated splice site region variants with the dbscSNV database.²⁵ We restricted our analyses to pLoF with minor allele frequency less than 0.01. AdaBoost and Random Forest scores greater than 0.9 were set as splice-altering effects and classified as pLoF variants. The identified pLoF variants are provided in eTable 1 in the Supplement.

Selection of Common Genetic Variants

We used genotyping data from unrelated individuals of European ancestry to identify heterozygous and homozygous carriers of the PCSK9 R46L (rs11591147) variant. To provide complementary evidence, we used a previously published PCSK9 genetic risk score (GRS) of 7 variants that reside near or within the PCSK9 gene (eTables 2 and 3 in the Supplement).²⁶ Each of these variants are associated with lower low-density lipoprotein cholesterol (LDL-C) levels and were combined into a GRS.²⁶ The single-nucleotide variations (SNVs) were originally selected based on the following criteria: (1) a positional location of ±100 kilobase pairs of the PCSK9 gene; (2) association with LDL-C levels at genome-wide significance (P < 5 × 10⁻⁸); and (3) low linkage disequilibrium (r²<0.2) with each other. For each variant, the exposure allele was defined as the allele associated with lower LDL-C levels. The PCSK9 genetic score was weighted by the allele lowering effect on LDL-C level for each SNV, as reported in a previous LDL-C genome-wide association study,²⁷ and scaled to per 38.7 mg/dL (to convert to millimoles per liter, multiply by 0.0259) lower LDL-C levels. To allow for comparison across exposure groups (ie, binary carrier status of pLoF and R46L) and to mimic the effect of extreme inhibition, we dichotomized the GRS at the 95th percentile, resulting in 2 categories: greater than or equal to the 95th percentile vs rest.

Derivation of CMR Measurements

The full CMR protocol in UKB has been described in detail previously.²¹ In brief, CMR imaging was performed on a clinical wide-bore 1.5-T scanner (MAGNETOM Aera, Syngo Platform VD13A [Siemens Healthcare]). Cardiac function was evaluated based on a combination of several sine series, including long-axis sines, and a complete short-axis stack covering the left and right ventricle was acquired at 1 slice per breath hold. For more than 43 000 UKB participants, CMR DICOM images were converted into NifTI format. CMR images were quality controlled using an automated image analysis, and annotation of cardiac imaging traits from the segmentations was subsequently performed, using a trained fully convolutional network architecture described previously.²⁸

Outcome Definitions

As primary outcomes, we used the following LV and left atrial (LA) CMR indices: LV end-diastolic volume, LV end-systolic volume, LV stroke volume, LV ejection fraction (LVEF), LV cardiac output, LVM, LA minimum volume, LA maximum volume, LA total emptying fraction, LA passive emptying fraction, and LA active emptying fraction as described in detail previously.^28,29 Volumetric measurements were indexed by body surface area calculated using the Du Bois formula.³⁰

Given the intrinsic association between heart rate and cardiac volumetric and functional indices, we also tested the 3 exposure groups for association with heart rate measured in association with CMR imaging. Because PCSK9 knockout (KO) mice displayed echocardiographic abnormalities indicative of HF in the study by Da Dalt et al,¹⁷ we explored associations with HF as a secondary outcome. Details on UKB data fields are provided in eTable 4 in the Supplement.

Statistical Analysis

Associations between carrier status (pLoF, R46L, and GRS) and LDL-C level, CMR measurements, and heart rate were evaluated using generalized linear models. Associations between carrier status and HF were tested using logistic regression. All models were adjusted for age at enrollment, sex, assessment center, and the first 5 principal components. Nonnormally distributed traits were log-transformed before association. P values were corrected for multiple comparisons using a Bonferroni P value threshold of P <.05 / 14 (P = 3.6 × 10⁻³). Our study had 80% power to detect a mean difference of 0.08 SD for R46L carriers, 0.44 SD for carriers of pLoF, and 0.09 SD for GRS greater than 95% (eMethods, eTable 5, and eFigure 2 in the Supplement).

Sensitivity Analyses

We conducted several sensitivity analyses to test the robustness of our results. We performed additional analyses including the GRS as a continuous variable, rather than dichotomizing at the 95th percentile. PCSK9 pLoFs have been associated with both increased visceral adiposity and pericardial fat mass, and both adiposity traits are known risk factors for developing HF.^31,32 To evaluate whether the association with PCSK9 variants differs across different body compositions, we conducted 3 stratified analyses. First, we categorized individuals according to body mass index ([BMI]; normal weight, ≤25; overweight, 26-30; obese, ≥30). BMI was calculated as weight in kilograms divided by height in meters squared. Second, because waist-to-hip ratio (WHR) has been shown to better reflect visceral adiposity,³³ we categorized individuals based on their WHR using sex-specific references (women: low WHR <0.80, moderate 0.80-0.86, and high >0.86; men: low <0.95, moderate 0.95-1, and high >1). Analyses with WHR were reported with and without further adjustment for BMI. Because CMR measurements were indexed for body surface area, we used unindexed measurements to avoid overcorrection. PCSK9 pLoF variants have been shown to protect against cardiovascular disease.³⁴ To test whether any lack of association with CMR measurements and HF could be confounded by the protective effects on atherosclerotic disease, we excluded individuals who had a history of coronary artery disease or stroke prior to baseline.

Exploratory Analyses

Several studies have shown that genetic variants in PCSK9 are associated with an increased risk of type 2 diabetes (T2D).^26,35,36 This risk was reported to be particularly evident in persons with impaired fasting glucose.²⁶ Because both prediabetes and T2D are important risk factors for HF,³⁷ we conducted additional exploratory analyses according to glycemic impairment. First, we tested whether the association of the 3 genetic exposure groups (pLoF, R46L and GRS) with CMR indices and HF differed in patients with and without T2D at baseline. Next, we evaluated whether the association with PCSK9 variants differed in individuals with prediabetes compared with persons with normoglycemia using data on glycated hemoglobin. Normoglycemia was defined as a hemoglobin A_1c (HbA_1c) level less than 39 mmol/mol (to convert to percentage of total hemoglobin, divide by 10.929 and add 2.15 then to convert to proportion of total hemoglobin, multiply percentage of total hemoglobin by 0.01), whereas prediabetes was defined as an HbA_1c level between 39 to 48 mmol/mol. To avoid any effects that glucose-lowering medications have on HbA_1c values, individuals with T2D prior to baseline were excluded in this analysis. In the study by Da Dalt et al,¹⁷ the KO mice developed LV hypertrophy without any accompanying change in LVEF, which was interpreted as HFpEF. To define HFpEF in the UKB, we used a combination of hospital records with CMR imaging data, as has been done previously.³⁸ Individuals were categorized as having HFpEF if they had an HF diagnosis prior to CMR imaging date and an LVEF greater than 50%.

Finally, to evaluate if carriers of the homozygous R46L variant presented with similar results as carriers of the heterozygous variant, we ran association analyses with LA and LV traits for carriers of the homozygous and heterozygous variants separately. A list of UKB data fields that were used to define phenotypes for the sensitivity and exploratory analyses are shown in eTable 4 in the Supplement. Significance was defined as a 2-sided P value <.05. All analyses were conducted between November 2, 2021, and October 25, 2022, using R software, version 4.0.0 (R Foundation for Statistical Computing).

Results

In up to 35 135 individuals with CMR images, 16 883 (48%) were male individuals, 18 252 (52%) were female individuals, and mean (SD) age was 55.0 (7.4) years. We identified 333 individuals (0.21%) carrying a rare pLoF variant, 13 584 carriers (3.5%) of the R46L variant (13 457 heterozygous and 127 homozygous carriers), and 18 924 individuals (4.91%) with a GRS greater than or equal to the 95th percentile. Baseline characteristics of included individuals according to PCSK9 pLoF carrier status are provided in the Table.

Table. Baseline Characteristics of the UK Biobank Study Cohort According to PCSK9 Loss-of-Function Variant Carrier Status^a.

Characteristic	No. (%)
	PCSK9 loss-of-function carriers (n = 333)	Noncarriers (n = 160 747)
Age, median (IQR), y	58 (50-64)	58 (51-63)
Female sex	198 (59)	87 941 (55)
Male sex	135 (41)	72 806 (45)
Blood pressure, mean (SD), mm Hg
Systolic	140.3 (19.4)	140 (19.5)
Diastolic	82.2 (11.1)	82.2 (10.6)
BMI, mean (SD)b	27.3 (5.3)	27.4 (4.7)
Cholesterol, mean (SD), mg/dL
LDL	110.6 (29.9)	138.1 (33.4)
Total	189.9 (41.3)	221.2 (44)
Smoking	26 (7.9)	14 958 (9.3)
Medical history
Hypertension	93 (27.9)	44 167 (27.5)
Diabetes	18 (5.4)	7761 (4.8)
Coronary artery disease	16 (4.8)	6396 (4.0)
Medication
Antihypertensive medication	29 (8.7)	19 102 (11.9)
Glucose-lowering drugs	15 (4.5)	5708 (3.6)
Lipid lowering medication	20 (6)	18 825 (11.7)

Abbreviations: BMI, body-mass index; LDL, low-density lipoprotein.

SI conversion factor: To convert LDL cholesterol to millimoles per liter, multiply by 0.0259.

^a Baseline corresponds to the date of enrolment between 2006 and 2010.

^b Calculated as weight in kilograms divided by height in meters squared.

Serving as a positive control, we evaluated the association between pLoF variant carrier status, R46L variant carrier status, and the PCSK9 genetic score with LDL-C levels. As expected, the smallest effect estimate with LDL-C level was observed for the genetic score (−2.84 mg/dL; 95% CI, −3.33 to −2.35; P = 4.26 × 10⁻³⁰), and the largest effect estimate was observed for the pLoF variants (−27.81 mg/dL; 95% CI, −31.38 to −24.24; P = 1.61 × 10⁻⁵²) (eFigure 3 in the Supplement).

Association Between PCSK9 Variants and LA and LV Traits

We evaluated the association of the 3 exposure groups with 6 LV and 5 LA CMR indices. Of 385 621 genotyped individuals, up to 35 135 (9%) had available CMR measurements, and of 161 080 individuals with exome-sequencing data, up to 18 999 (12%) had available CMR measurements. In analyses including up to 41 pLoF carriers, we found no significant associations between pLoF carrier status and LV indices, eg, for LV mass (β = −1.01; 95% CI, −2.99 to 0.98; P = .32) and LV ejection fraction (pLoF: β = 0.43; 95% CI, −1.32 to 2.18; P = .63), or LA indices, beyond the threshold for multiple testing (Figure 1A and B). Of 13 584 identified as R46L carriers, at least 1100 had undergone CMR imaging. We found no significant association between R46L carrier status and CMR traits, eg, for LV mass (β = −0.18; 95% CI, −0.55 to 0.19; P = .35) and LV ejection fraction (β = −0.19; 95% CI, −0.52 to 0.14; P = .26). We also did not find any significant associations in homozygous R46L carriers, eg, for LV mass (β = −3.80; 95% CI, −8.36 to 0.75; P = .10) and LV ejection fraction (β = −0.91; 95% CI, −4.71 to 2.88; P = .64) (eTable 6 in the Supplement). Of the 18 924 individuals with a GRS greater than or equal to the 95th percentile, more than 1600 had available CMR scans. We found no significant association between the GRS and LV or LA measurements, eg, for LV mass (β = −0.19; 95% CI, −0.50 to 0.11; P = .22) and LV ejection fraction (β = −0.08; 95% CI, −0.35 to 0.20; P = .58). We also did not observe any significant associations between the GRS on a linear scale and listed outcomes in analyses including more than 32 000 individuals, eg, for LV mass (β = −0.57; 95% CI, −1.17 to 0.04; P = .07) and LV ejection fraction (β = 0.16; 95% CI, −0.37 to 0.70; P = .55) and HF (odds ratio [OR], 0.87; 95% CI, 0.74-1.03; P = .10) (eTable 7 in the Supplement).

Figure 1. Association Between Genetic Variation in PCSK9 and Cardiac Magnetic Resonance Indices.

Shown are the results from linear regression models of association between PCSK9 predicted loss-of-function (pLoF) carrier status, R46L carrier status, PCSK9 genetic risk score (GRS) at or greater than the 95th percentile, and 6 left-ventricular (A) and 5 left-atrial (B) cardiac magnetic resonance imaging indices. Point estimates are shown as dots, and lines indicate 95% CIs. LA_max indicates left atrial maximum volume; LA_min, left atrial minimum volume; LAAEF, left atrial active emptying fraction; LAPEF, left atrial passive emptying fraction; LATEF, left atrial total emptying fraction; LVCO, left ventricular cardiac output; LVEDV, left ventricular end diastolic volume; LVEF, left ventricular ejection fraction; LVESV, left ventricular end systolic volume; LVM, left ventricular mass; and LVSV, left ventricular stroke volume.

Association Between PCSK9 Variants and HF and Heart Rate

Next, given the intrinsic relationship between heart rate and cardiac volumetric and functional indices, we tested the 3 exposure groups for association with heart rate measured at the time of the CMR scan. We found no significant association between the exposure groups and heart rate (Figure 2A). Next, we evaluated whether variants in PCSK9 were associated with overt HF. We found no significant association between pLoF carrier status, R46L carrier status, or extreme genetic score and HF (pLoF: OR, 1.14; 95% CI, 0.56-2.05; P = .69; R46L: OR, 0.99; 95% CI, 0.90-1.10; P = .91; GRS: OR, 1.04; 95% CI, 0.96-1.13; P = .32) (Figure 2B).

Figure 2. Association Between Genetic Variation in PCSK9 and Heart Rate and Heart Failure.

Shown are the results from association analyses between PCSK9 predicted loss-of-function (pLoF) carrier status, R46L carrier status, PCSK9 genetic risk score (GRS) at or greater than the 95th percentile, and heart rate (A) and heart failure (B). Point estimates are shown as dots and lines indicate 95% CIs. OR indicates odds ratio.

Sensitivity Analyses

We conducted several sensitivity analyses to test the robustness of our associations. First, we found no evidence to support that the PCSK9 variants had any differential association with LV and LA indices or HF across different body compositions, including BMI, WHR, or WHR adjusted for BMI categories (eg, for LV mass, pLoF: P for interaction = .86; R46L: P for interaction = .57; GRS: P for interaction = .84; and for LV ejection fraction, pLoF: P for interaction = .72; R46L: P for interaction = .76; GRS: P for interaction = .73; and for HF, pLoF: P for interaction = .41; R46L: P for interaction = .48; GRS: P for interaction = .77) (eTable 8, 9, and 10 in the Supplement). The results did not materially change when excluding individuals with cardiovascular disease or T2D at baseline, nor did we find any differential associations across individuals with normoglycemia or prediabetes at baseline, eg, for LV mass (pLoF: P for interaction = .09; R46L: P for interaction = 0.65; GRS: P for interaction = 0.62) and LV ejection fraction (pLoF: P for interaction = .92; R46L: P for interaction = 0.99; GRS: P for interaction = 0.62) and HF (R46L: P for interaction=0.93; GRS: P for interaction = 0.69) (eTable 11, 12, and 13 in the Supplement). To specifically test the association with HFpEF, we created a proxy phenotype by combining data on a subset with available hospital records for HF and LVEF measurements. In total, we identified 132 individuals with HFpEF and 34 628 controls. We found no statistically significant associations between the genetic exposure groups (R46L and GRS) and HFpEF (R46L: OR, 0.20; 95% CI, 0.01-0.90; P = .11; GRS: OR, 0.72; 95% CI, 0.26-1.60; P = .48) (eTable 14 in the Supplement). Finally, we found no significant associations between R46L homozygous carrier status and CMR traits or HF, eg, for LV mass (β = −3.80; 95% CI, −8.36 to 0.75; P = .10) and LV ejection fraction (β = −0.91; 95% CI, −4.71 to 2.88; P = .64) and HF (OR, 0.80; 95% CI, 0.20-2.14; P = .71) (eTable 6 in the Supplement).

Discussion

In this case-control study nested within the UK Biobank, using genetic instruments as proxies for therapeutic inhibition, we investigated the association of genetic variants in PCSK9 with the human heart. We evaluated the association of both common and rare genetic variants with 11 CMR indices and risk of HF. We found no evidence to support that genetic variants in PCSK9 are associated with aberrant cardiac structure and function.

In the randomized, double-blinded, placebo-controlled Further Cardiovascular Outcomes Research With PCSK9 Inhibition in Subjects With Elevated Risk (FOURIER) study, with a median follow-up of 2.2 years, evolocumab was shown to lower LDL-C levels by 59% and reduce the risk of cardiovascular death, myocardial infarction (MI), and stroke by 20% when compared with placebo.^4,11 Moreover, no significant between-group differences in adverse events were found in the study. Electrocardiograms (ECGs) were also collected and evaluated continuously for approximately 5000 participants, and no trends indicative of clinically important effects of evolocumab on ECGs were found.⁴ Furthermore, no difference in cardiovascular death or hospitalization for worsening HF was found between the evolocumab and placebo group. In the Evaluation of Cardiovascular Outcomes After an Acute Coronary Syndrome During Treatment With Alirocumab (ODYSSEY OUTCOMES) trial, with a median follow-up of 2.8 years, alirocumab reduced nonfatal MI by 14%, stroke by 27%, and unstable angina by 39% compared with placebo, and mortality rate was found to be significantly lower with alirocumab than with placebo treatment (3.5% vs 4.1%).¹³ However, no significant difference in hospitalization for congestive HF was found (1.9% vs 1.9%). In a post hoc analysis of the ODYSSEY OUTCOMES trial, White et al³⁹ investigated the effect of alirocumab on CV outcomes and HF in patients with and without a history of HF. The study showed that alirocumab reduced the risk of major adverse cardiovascular events in patients without a history of HF but not in those with a history of HF. Alirocumab was not shown to decrease or increase hospitalizations for HF; however, rates of hospitalization for HF were numerically lower with alirocumab in patients without HF and numerically higher with alirocumab in patients with a history of HF.

Despite that multiple clinical trials of PCSK9 inhibitors have consistently displayed an overall beneficial cardiovascular safety profile,^2,4,13,14,40 a recent animal study showed that PCSK9-deficient mice displayed abnormal cardiac lipid metabolism, increased thickness of the LV posterior wall, and signs of HFpEF.¹⁷ To investigate whether these observations could be translated into humans, we evaluated the association of PCSK9 genetic variants with cardiac remodeling and overt HF using state-of-the-art CMR imaging data. LoF variants, predicted to largely or entirely abolish the function of the gene product, are an informative class of genetic variation. These variants are natural models for lifelong reduction of the target gene product and can provide information about the potential toxicity of therapeutic inhibitors targeting the product of these genes. As such, we studied pLoF variant carriers and homozygous R46L variant carriers as counterparts to KO mice. We found no significant differences in LVEF or LVM in carriers compared with noncarriers, nor did we find evidence of altered volumes or function in pLoF carriers or R46L homozygotes. In addition to the animal data, Da Dalt et al¹⁷ also investigated the association between R46L carrier status and LVM measured using echocardiography. The authors reported a small, albeit statistically significant, increase in LVM in a small cohort of 12 human carriers. We reevaluated this association in a more well-powered setting, with more than 100-fold more R46L carriers (n = 1237). Again, no significant association with LV CMR imaging indices was found. Last, because atrial volumetric indices are key parameters in diagnosing HFpEF, we evaluated the association between genetic proxies and 5 atrial volumetric and functional indices. We found no significant association between PCSK9 variants and LA CMR indices. Complementary to the CMR-derived measurements, we also investigated whether PCSK9 variants were associated with the risk of overt HF. We observed no significant association between PCSK9 variants and HF risk.

Limitations

The results of our study should be interpreted within the context of its limitations. First, by investigating the association of LoF variants, we not only evaluated the condition of reduced circulating PCSK9 produced in the liver but also the PCSK9 expressed in all organs. Second, we were not able to evaluate the association of PCSK9 variants with B-type natriuretic peptide levels, as B-type natriuretic peptide level measurements are not available in the UKB. Third, we acknowledge that the sensitivity analysis of HFpEF is likely underpowered and that some degree of misclassification cannot be excluded, ie, individuals with HF and recovered LVEF may be misclassified as HFpEF. Fourth, our study included individuals of European descent only, which limits the generalizability to non-European ethnicities.

Conclusions

In conclusion, in this case-control study nested within the UKB, results suggest that PCSK9 variants, proxying lifelong PCSK9 reduction, were not associated with any detrimental outcomes of cardiac structure and function. Our long-term results are in line with safety data from the shorter-term trials that indicate that a partial to complete reduction of the PCSK9 protein in humans is unlikely to result in the severe phenotypes observed in model organisms.

Supplement.

eMethods

eTable 1. Loss-of-Function Variants in PCSK9

eTable 2. Variants in PCSK9 Included in the Genetic Score and Their Association With LDL-C in the Global Lipids Genetics Consortium

eTable 3. Linkage Disequilibrium Matrix for Variants Included in the PCSK9 Genetic Score

eTable 4. UK Biobank Data Fields Used in Regression Analyses

eTable 5. Power Calculation Mean Effect Size at 80% Power and Alpha .05

eTable 6. Association Between R46L Variant Carrier Status and Cardiac Indices

eTable 7. Association Between Genetic Risk Score on a Linear Scale, LDL-C Levels and Cardiac Indices

eTable 8. Association Between Genetic Variation in PCSK9 and Cardiac Indices Across BMI Categories

eTable 9. Association Between Genetic Variation in PCSK9 and Cardiac Indices Across Waist-to-Hip Categories

eTable 10. Association Between Genetic Variation in PCSK9 and Cardiac Indices Across Waist-to-Hip Categories Adjusted for BMI

eTable 11. Association Between Genetic Variation in PCSK9 and Cardiac Indices in Persons Without Cardiovascular Disease

eTable 12. Association Between Genetic Variation in PCSK9 and Cardiac Indices in Persons Without Type 2 Diabetes

eTable 13. Association Between Genetic Variation in PCSK9 and Cardiac Indices Across HbA_1c Categories (Normal and Prediabetes)

eTable 14. Association Between Genetic Variation in PCSK9 and Heart Failure With Preserved Ejection Fraction

eFigure 1. Principal Component Plot

eFigure 2. Power Calculation Plot

eFigure 3. Association Between PCSK9 Predicted Loss-of-Function (pLoF) Carrier Status, R46L Carrier Status, ≥95th Percentile PCSK9 Genetic Risk Score (GRS) and LDL-C Levels

eReferences