Molecular Characterization of Familial Hypercholesterolemia in a North American Cohort

Abstract

Background

Familial hypercholesterolemia (FH) confers a very high risk of premature cardiovascular disease and is commonly caused by mutations in low-density lipoprotein receptor (LDLR), apolipoprotein B (APOB), or proprotein convertase subtilisin/kexin type 9 (PCSK9) and very rarely in LDLR adaptor protein 1 (LDLRAP1) genes.

Objective

To determine the prevalence of pathogenic mutations in the LDLR, APOB, and PCSK9 in a cohort of subjects who met Simon Broome criteria for FH and compare the clinical characteristics of mutation-positive and mutation-negative subjects.

Methods

Ninety-three men and 107 women aged 19 to 80 years from lipid clinics in the United States and Canada participated. Demographic and historical data were collected, physical examination performed, and serum lipids/lipoproteins analyzed. Targeted sequencing analyses of LDLR and PCSK9 coding regions and exon 26 of APOB were performed followed by detection of LDLR deletions and duplications.

Results

Disease-causing LDLR and APOB variants were identified in 114 and 6 subjects, respectively. Of the 58 LDLR variants, 8 were novel mutations. Compared with mutation-positive subjects, mutation-negative subjects were older (mean 49 years vs 57 years, respectively) and had a higher proportion of African Americans (1% vs 12.5%), higher prevalence of hypertension (21% vs 46%), and higher serum triglycerides (median 86 mg/dL vs 122 mg/dL) levels.

Conclusions

LDLR mutations were the most common cause of heterozygous FH in this North American cohort. A strikingly high proportion of FH subjects (40%) lacked mutations in known culprit genes. Identification of underlying genetic and environmental factors in mutation-negative patients is important to further our understanding of the metabolic basis of FH and other forms of severe hypercholesterolemia.

1.  Introduction

Hypercholesterolemia is one of the main causes of atherosclerosis and its clinical sequelae. Familial hypercholesterolemia (FH) is one of the most common and most severe forms of monogenic hypercholesterolemia. The disease has an autosomal codominant pattern of inheritance and is most commonly caused by mutations in the low-density lipoprotein receptor (LDLR) gene [1]. Individuals with 2 mutated LDLR alleles (FH homozygotes or compound heterozygotes) are believed to be extremely rare (~1 in 200 000 to ~1 in 1 million) [2]. However, individuals with 1 mutant allele (FH heterozygotes) are much more common, with an original estimated frequency of 1 in 500 in Western populations [2] but with recent world-wide estimates of 1 in 250 [3] and up to 1 in 70 in some populations such as French Canadians [4, 5], Afrikaners in South Africa [6], Lebanese [7], and Finns [8], where a founder effect is in place.

In addition to LDLR mutations, mutations in the apolipoprotein B (APOB) and proprotein convertase subtilisin/kexin type 9 (PCSK9) genes can cause FH phenotypes [9]. Absence or reduced functionality of the LDLR reduces clearance of plasma LDL. Similarly, gain-of-function (GOF) mutations in PCSK9 increase the ability of this protein to degrade the LDLR [10], and specific loss-of-function (LOF) mutations in ApoB-100 reduce the ability of LDL to bind to the LDLR [11]. Homozygous or compound heterozygous mutations in LDLR adaptor protein 1 are a very rare cause of FH, occurring with an estimated prevalence of <1 per million [12].

Accurate identification of FH is important because of the substantial increase in risk of coronary heart disease (CHD) in patients with this disorder [13]. A definitive, early diagnosis of patients with heterozygous FH may also offer better treatment options and improved clinical outcomes as lipid-lowering therapies are associated with significant reduction in rates of CHD [14]. The clinical diagnosis of FH is currently based on personal and family history of severe hypercholesterolemia and premature CHD and presence of tendon and cutaneous xanthomas or corneal arcus. However, other diagnostic criteria have been proposed [13, 15–18, 19]. Since genotyping provides an unequivocal diagnosis of FH, routine genotyping of patients with FH is often advocated [18, 20–22]. Investigators from different parts of the world have reported variable rates of mutation-positive FH [23–37]. Those reporting low rates of mutation-positive FH may have included possible errors in the clinical diagnosis or may have incompletely ascertained relevant disease-causing mutations. On the other hand, those reporting high rates of genotype positive FH may be from areas with a dominant founder effect.

For these reasons, we assessed the relationship between well-characterized FH and presence of causative mutations using comprehensive genotyping. Furthermore, because of a paucity of such data from North America, we determined the genetic basis of hypercholesterolemia in adults with presumptive heterozygous FH from the United States and Canada, focusing on variants in LDLR, APOB, and PCSK9. We then compared the demographics and clinical phenotypes of mutation-positive and mutation-negative subjects.

2.  Methods

A.  Study Design

This was an observational study of patients from 13 lipid clinics (9 of which were academic sites) in the United States (N = 8) and Canada (N = 5). A total of 256 individuals with documentation of fasting plasma total cholesterol ≥ 290 mg/dL (7.5 mmol/L) or LDL-cholesterol (LDL-C) ≥ 190 mg/dL (4.9 mmol/L) and triglycerides <500 mg/dL (5.6 mmol/L) were screened for eligibility to participate in the study. Documentation of lipid levels could consist of a historical untreated laboratory test result, accurate medical record, or laboratory test result obtained during screening for the study. Subjects were enrolled in the study between July 2011 and May 2012.

Eligible men and women were ≥18 years of age with a body mass index of 18.5 to 40 kg/m² and total body weight of 50 to 150 kg. Body weight was used as an exclusion criterion because of the potential effect of extreme obesity on LDL-C levels. For individuals on lipid-lowering therapy, a current screening fasting plasma LDL-C ≥ 100 mg/dL (2.6 mmol/L) was mandated for possible future eligibility for a clinical trial. Diagnosis of a definite or possible clinical diagnosis of FH phenotype was based on the Simon Broome criteria for heterozygous FH [13] and included an untreated LDL-C ≥ 190 mg/dL as well as any of the following: (a) premature CHD (males aged < 55 years and females aged < 65 years), (b) family history of premature CHD, or (c) concurrent or past history of tendon xanthomas. Subjects receiving treatment with LDL apheresis were ineligible for inclusion.

Fasting blood samples for genotyping, lipid panel, and PCSK9 protein assay were drawn at the screening visit. Genotyping was performed for all eligible subjects (n = 200). The study was conducted in accordance with the Declaration of Helsinki on Ethical Principles for Medical Research Involving Human Subjects. All subjects gave written informed consent, and approval of the protocol was obtained from local institutional review boards/independent ethics committees.

B.  Genotyping Analysis

Blood specimens for genotyping were stored at –70°C. Genomic deoxyribonucleic acid (DNA) was extracted from whole blood samples and subjected to polymerase chain reaction amplification using a Bio-Rad Tetrad 2 thermal cycler (Bio-Rad Laboratories Inc., Hercules, CA). Sequencing library pools were prepared and analyzed using Next Generation Sequencing on an Illumina HiSeq 2000 sequencing system (Illumina Inc., San Diego, CA) at Covance Genomic Laboratory (Seattle, WA). Targeted sequencing analysis of the coding regions of the LDLR (exons 1–18) and PCSK9 (exons 1–12) genes, as well as exon 26 of the APOB gene, was conducted to identify the genetic basis of the hypercholesterolemia phenotype for enrolled subjects. If no known mutations were detected in the initial sequencing analysis, additional testing for large deletions and duplications within the LDLR gene was then performed using the SALSA Multiplex Ligation-Dependent Probe Amplification P062 LDLR kit (MRC-Holland, Amsterdam, The Netherlands) as per the manufacturer’s instructions. Amplification products were separated on a 3730xl DNA Analyzer (Applied Biosystems, Carlsbad, CA), and the data were analyzed using Coffalyser software (MRC-Holland). Peak height ratios of ≤0.7 were categorized as deletions, while ratios of ≥1.3 were categorized as duplications.

All mutations in the coding regions, such as missense, frameshift, and truncation, were reported; silent, synonymous mutations were not reported. The Clinvar database and LOVD databases were queried for all the variants [38, 39], and additional primary literature was referenced to determine pathogenicity of the mutations [2, 9, 40–43]. Subjects having such mutations were defined as mutation positive, and those without were defined as mutation negative. Variants determined to be novel were absent from Clinvar, LOVD, and gnomAD databases [44].

C.  Plasma PCSK9 Assay

Plasma samples were analyzed for PCSK9 concentrations at ICON Development Solutions, LLC (Whitesboro, NY) using a validated, sensitive, and specific enzyme-linked immunosorbent assay method [45]. Plasma specimens were stored at approximately –70ºC until analysis and assayed within the 534 days of established stability data generated during validation. Calibration standard responses were linear over the range of 0.313 to 30.0 ng/mL (in buffer), using a nonweighted, 4-parameter logistic curve-fit regression. The lower limit of quantitation for PCSK9 was 0.900 ng/mL. The between-day PCSK9 assay accuracy, expressed as percent relative error, for quality control (QC) concentrations ranged from 0.858% to 16.1% for the low, medium, and high QC samples. Assay precision, expressed as the between-day percent coefficient of variation of the mean estimated concentrations of QC samples was ≤7.87% for the low (0.900 ng/mL in 5% matrix; 1.74 ng/mL after base pool correction), medium (6.00 ng/mL in 5% matrix; 6.84 ng/mL after base pool correction), and high (22.5 ng/mL in 5% matrix; 23.3 ng/mL after base pool correction) QC samples.

D. Statistical Methods

The main outcome measure was prevalence of pathogenic mutations in the LDLR, APOB, and PCSK9 genes. Descriptive statistics were generated to summarize demographic characteristics including race (white, black, Asian, other), genotype characteristics of mutations, lipid profiles, medication history, and family medical history at the time of study entry. Categorical variables were reported as number and proportion of subjects in a category. Continuous variables were reported as mean (standard deviation) or median (range). Ninety-five percent confidence intervals (CIs) for the difference between the mutation-positive and mutation-negative groups were calculated using Wald’s CI procedure for categorical variables and a t-test for continuous variables.

2. Results

A. Study Populations and Demographics

The demographic characteristics of the study population at screening (ie, while subjects were receiving lipid-lowering therapy) are summarized in Table 1. We enrolled 200 unrelated subjects in this study, 93 (46.5%) were male and the mean [standard deviation] age was 52.0 [12.9] years. The majority of subjects (N = 187; 93.5%) were of white ethnic origin; 68 (34%) subjects were from Quebec, Canada.

Table 1.

Demographic Characteristics of Mutation-Positive^a (N = 120) and Mutation-Negative^b (N = 80) Subjects at the Screening Visit^c

Characteristic	Total (N = 200)	Mutation Positive (N = 120)	Mutation Negative (N = 80)	Difference (95% CI) Between Genotype Groupsd
Male gender	93 (46.5)	57 (47.5)	36 (45.0)	2.5 (–11.6, 16.6)
Age (years)	52.0 (12.9)	48.9 (13.5)	56.6 (10.4)	–7.7 (–11.2, –4.2)e
Age group
18–44 y	52 (26.0)	42 (35.0)	10 (12.5)	22.5 (11.3, 33.7)e
45–64 y	108 (54.0)	61 (50.8)	47 (58.8)	–7.9 (–21.9, 6.1)
≥65 y	40 (20.0)	17 (14.2)	23 (28.8)	–14.6 (–26.3, –2.9)e
Race
White	187 (93.5)	119 (99.2)	68 (85.0)	14.2 (6.2, 22.2)e
Black	11 (5.5)	1 (0.8)	10 (12.5)	–11.7 (–19.1, –4.2)e
Asian	1 (0.5)	0 (0.0)	1 (1.3)	–
Other	1 (0.5)	0 (0.0)	1 (1.3)	–
Fasting glucose (mg/dL)	94.8 (13.6)	94.4 (13.8)	95.5 (13.2)	–1.1 (–4.9, 2.8)
Weight (kg)	81.8 (17.1)	80.6 (17.5)	83.7 (16.4)	–3.1 (–8.0, 1.7)
Height (cm)	168.1 (10.1)	168.2 (10.5)	168.0 (9.5)	0.2 (–2.7, 3.1)
Body mass index (kg/m2)	28.8 (5.0)	28.4 (5.1)	29.6 (4.8)	–1.2 (–2.6, 0.2)
Waist circumference (cm)	95.6 (13.1)	94.6 (13.5)	97.1 (12.3)	–2.5 (–6.2, 1.3)
Diastolic BP (mm Hg)	74.4 (9.9)	72.8 (10.0)	76.7 (9.3)	–3.9 (–6.7, –1.1)e
Systolic BP (mm Hg)	122.8 (14.8)	120.6 (14.3)	126.0 (15.0)	–5.4 (–9.6, –1.2)e

Data represent number (%) or mean (SD).

–, no data; BP, blood pressure; CI, confidence interval; SD, standard deviation; y, years.

a. Excludes subjects carrying only the p.A391T or p.T726I variants of LDLR.

b. Includes subjects carrying only the p.A391T or p.T726I variants of LDLR.

c. The majority of subjects were receiving lipid-lowering therapy at screening.

d. Calculated using Wald’s confidence interval procedure for categorical variables and a t-test for continuous variables.

e. P value < 0.05.

A total of 120 subjects were classified as mutation positive; the remaining 80 subjects were classified as mutation negative. Demographic characteristics at screening were similar between the 2 groups, but age, proportion of African Americans, and diastolic and systolic blood pressures were higher in the mutation-negative subjects (Table 1).

B. Mutation Prevalence

Of the 200 study participants, 131 subjects were carriers of missense, frameshift, and/or truncation variants in LDLR. Two of the LDLR variants detected were p.A391T and p.T726I. Although the A391T variant has been detected in multiple FH cohorts,[46–49] it is a common variant (rs11669576; minor allele frequency: 4.3%–8.3%) and is not thought to be causative for FH [50, 51] Hence, the 17 subjects carrying only the LDLR p.A391T variant (heterozygotes and homozygotes) were included in the mutation-negative group. Similarly, the p.T726I variant, found in normocholesterolemic populations (rs45508991; minor allele frequency: 0.45%–0.63%), [52–54] was excluded from the analysis. Interestingly however, all 3 patients with the T726I LDLR variant also carried other disease-causing heterozygous LDLR mutations. Six subjects were carriers of a heterozygous missense mutation (p.R3527Q) in APOB associated with the FH phenotype [11]; no pathogenic variants of PCSK9 were identified in any subject. Therefore, a total of 120 (60.0%) mutation-positive subjects carried LDLR and APOB variants.

Overall, among the pathogenic LDLR mutations detected in 114 subjects, there were a total of 58 distinct mutations. All supplementary material is located in a digital research materials repository [55]. Of these, 30 were nonsynonymous missense mutations; 18 were stop codons, frameshift, and small deletions that would result in a truncated LDLR; and 10 were large genomic rearrangements (deletion and/or duplication of 1 or more exons). Eight novel LDLR variants were identified in 8 different subjects: c.112A>T, (p.K38*); c.347G>C, (p.C116S); c.367_631del, (p.I122_I210del); c.653delGT, (p.G219Pfs*8); c.910insG, (p.D304Gfs*5); c.1927_1941del, (p.A643_L647del); Exons 4–10 dup, (p.?); and Exon 6 dup (p.?). Four other LDLR variants, c.233G>A, (p.R78H); c.1102T>C, (p.C368R); c.2167G>T, (p.E723*); and c.2230C>T, (p.R744*), which have not been published but are annotated in Clinvar or gnomAD databases were also identified. The mutations were distributed among all 18 exons and throughout the coding regions of LDLR [55]. Interestingly, one subject had a homozygous mutation (p.W87G) and another had compound heterozygous mutations (p.R410S; p.G592E) in LDLR [55].

Although missense variants in PCSK9 were detected in all 200 subjects [55], the 8 variants identified have been reported as polymorphisms with uncertain or no association with hypercholesterolemia [42, 43]. No novel or any of the previously reported GOF mutations in PCSK9 associated with the FH phenotype [9, 10, 42, 43] were detected. The LOF p.R46L variant [56] was found in 3 subjects who also had mutations in LDLR.

C. Lipid, Lipoprotein, and PCSK9 Levels

Mean serum total cholesterol, LDL-C, high-density lipoprotein–cholesterol, and apolipoprotein AI and B levels at screening (all measured on LDL-C–lowering treatment) were similar between the mutation-positive and the mutation-negative groups (Table 2). Median serum triglycerides levels were higher in the mutation-negative group (122.1 mg/dL) than in the mutation-positive group (86.1 mg/dL) (95% CI for the difference: –46.3 to –7.9 mg/dL; Table 2). In the mutation-positive group, 69% of patients had LDL-C > 130 mg/dL, 29% had LDL-C of 100 to 130 mg/dL, and only 2% had LDL-C < 100 mg/dL (all measured on LDL-C–lowering treatment) (Table 2). In the mutation-negative group, the corresponding frequencies were 64%, 35%, and 1%, respectively. Plasma PCSK9 levels were higher, but not significantly so, in the mutation-positive group compared with the mutation-negative group (median values: 78.9 ng/mL vs 43.3 ng/mL, respectively; Table 2).

Table 2.

Plasma Lipid, Lipoprotein, and PCSK9 Levels at the Screening Visit^a in Mutation-Positive^b (N = 120) and Mutation-Negative^c (N = 80) Subjects

Parameter	Total (N = 200)	Mutation Positive (N = 120)	Mutation Negative (N = 80)	Difference (95% CI) Between Genotype Groupsd
Total cholesterol (mg/dL)	227.3 (75.0)	227.3 (78.9)	227.5 (69.3)	–0.2 (–21.6, 21.2)
LDL-C (mg/dL)	167.1 (63.9)	171.3 (70.1)	160.7 (53.2)	10.6 (–7.6, 28.8)
Proportion at LDL-C:
70–<100 mg/dL	3 (1.5)	2 (1.7)	1 (1.3)	–
100–130 mg/dL	63 (31.5)	35 (29.2)	28 (35.0)	–
> 130 mg/dL	134 (67.0)	83 (69.2)	51 (63.8)	–
Triglycerides (mg/dL)
Mean (SD)	113.3 (68.6)	102.4 (63.9)	129.5 (72.5)	–27.1 (–46.3, –7.9)e
Median (range)	98.9 (8.5–494.0)	86.1 (22.2–375.0)	122.1 (8.5–494.0)
HDL-C (mg/dL)	51.5 (15.5)	52.4 (17.4)	50.1 (12.1)	2.3 (–2.1, 6.7)
ApoAI (mg/dL)	148.2 (32.6)	145.8 (35.5)	151.9 (27.3)	–6.1 (–15.7, 3.5)
ApoB (mg/dL)	133.0 (38.8)	134.3 (43.1)	131.0 (31.1)	3.3 (–8.1, 14.7)
PCSK9 (ng/mL)
Mean (SD)	125.9 (102.6)	135.6 (102.8)	111.3 (101.3)	24.4 (–4.7, 53.5)
Median (range)	67.3 (16.4–340.0)	78.9 (16.4–340.0)	43.3 (19.5–324.0)

Data represent number (%) or mean (SD); median (range) is also provided for triglycerides and PCSK9 levels.

–, no data; ApoAI, apolipoprotein AI; ApoB, apolipoprotein B; CI, confidence interval; HDL-C, high-density lipoprotein–cholesterol; LDL-C, low-density lipoprotein–cholesterol; PCSK9, proprotein convertase subtilisin/kexin type 9; SD, standard deviation.

a. The majority of subjects were receiving lipid-lowering therapy at screening.

b. Excludes subjects carrying only the p.A391T or p.T726I variants of LDLR.

c. Includes subjects carrying only the p.A391T or p.T726I variants of LDLR.

d. Calculated using a t-test.

e. P value < 0.05.

D. Personal and Family History of Cardiovascular Disease

Prevalence of CHD was similar in the mutation-positive (36.7%) versus mutation-negative (32.5%) subjects (Table 3). Compared with the mutation-positive group, the mutation-negative group had a higher prevalence of hypertension (46.3% vs 20.8%, respectively; 95% CI, –38.5% to –12.3%) and nonsignificantly higher prevalence of diabetes (12.5% vs 5.0%).

Table 3.

Personal and Family History of Cardiovascular Disease and Clinical Features Associated with Familial Hypercholesterolemia in Mutation-Positive^a (N = 120) and Mutation-Negative^b (N = 80) Subjects

Disease History/Clinical Features	Total (N = 200)	Mutation Positive (N = 120)	Mutation Negative (N = 80)	Difference (95% CI) Between Genotype Groupsc
Personal CVD history
CHD	70 (35.0)	44 (36.7)	26 (32.5)	4.2 (–9.2, 17.6)
Cerebrovascular diseased	5 (2.5)	2 (1.7)	3 (3.8)	–2.1 (–6.8, 2.7)
Hypertension	62 (31.0)	25 (20.8)	37 (46.3)	-25.4 (–38.5, –12.3)e
Diabetes mellitus	16 (8.0)	6 (5.0)	10 (12.5)	–7.5 (–15.7, 0.7)
Peripheral vascular disease	1 (0.5)	1 (0.8)	0 (0.0)	0.8 (–0.8, 2.5)
Family history
Any CVD	194 (97.0)	117 (97.5)	77 (96.3)	1.2 (–3.8, 6.3)
CHD	160 (80.0)	100 (83.3)	60 (75.0)	8.3 (–3.3, 19.9)
Cerebrovascular diseased	47 (23.5)	26 (21.7)	21 (26.3)	–4.6 (–16.7, 7.5)
Other CVDf	73 (36.5)	45 (37.5)	28 (35.0)	2.5 (–11.1, 16.1)
Other atherosclerotic-related diseaseg	5 (2.5)	3 (2.5)	2 (2.5)	0.0 (–4.4, 4.4)
Hypercholesterolemia	144 (72.0)	93 (77.5)	51 (63.8)	13.8 (0.8, 26.7)e
Severe hypercholesterolemia	14 (7.0)	9 (7.5)	5 (6.3)	1.3 (–5.8, 8.3)
Tendon xanthoma	2 (1.0)	1 (0.8)	1 (1.3)	–0.4 (–3.3, 2.5)
Clinical features of FH
Bruits	8 (4.0)	3 (2.5)	5 (6.3)	–3.8 (–9.7, 2.2)
Corneal arcus	89 (44.5)	47 (39.2)	42 (52.5)	–13.3 (–27.3, 0.7)
Pulse
Boundingh	8 (4.0)	5 (4.2)	3 (3.8)	0.4 (–5.1, 5.9)
Faint/slightly diminishedh	18 (9.0)	8 (6.7)	10 (12.5)	–5.8 (–14.3, 2.7)
Tendinous xanthomas	74 (37.0)	49 (40.8)	25 (31.3)	9.6 (–3.8, 23.0)
Tuberous xanthomas	1 (0.5)	1 (0.8)	0 (0.0)	0.8 (–0.8, 2.5)
Xanthelasmas	10 (5.0)	7 (5.8)	3 (3.8)	2.1 (–3.8, 8.0)

Data represent number (%).

CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; FH, familial hypercholesterolemia.

a. Excludes subjects carrying only the p.A391T or p.T726I variants of LDLR.

b. Includes subjects carrying only the p.A391T or p.T726I of LDLR.

c. Calculated using Wald’s confidence interval procedure.

d. Including stroke.

e. P value < 0.05.

f. “Other CVD” includes those cardiovascular diseases that cannot be classified into any of the categories.

g. “Other atherosclerotic-related disease” includes angina, cardiomyopathy, congestive heart failure, and coronary artery bypass graft.

h. The categories “Bounding” and “Faint/slightly diminished” in any site include subjects with ≥1 abnormal pulse.

Family history of hypercholesterolemia was more prevalent in the mutation-positive subjects versus mutation-negative subjects (77.5% vs 63.8%), but the family history of CHD (83.3% vs 75.0%) and cerebrovascular disease (21.7% vs 26.3%) were similar (Table 3).

E. Physical Findings Related to FH

The prevalence of clinical features associated with FH was similar in the mutation-positive and the mutation-negative groups, including arterial bruits (2.5% vs 6.3%, respectively), corneal arcus (39.2% vs 52.5%), and faint or slightly diminished pulse (6.7% vs 12.5%) (Table 3). The prevalence of definite and possible FH as per the Simon Broome criteria among the mutation-positive patients was 40.8% and 59.2%, respectively, compared to 31.3% and 68.7%, respectively, in the mutation-negative group, but it was not statistically significant (95% CI of the difference: –3.8% to 23.0%) (Fig. 1). Overall, xanthomas and xanthelasmas were noted in 42.5% of subjects (47.5% of mutation-positive subjects; 35.0% of mutation-negative subjects).

Figure 1.

Prevalence of definite and possible familial hypercholesterolemia (FH) according to Simon Broome criteria among mutation-positive and mutation-negative subjects.

F. Lipid-Lowering Medications

The majority of subjects (N = 177; 88.5%) were taking lipid-lowering medication at study entry (Table 4). Statin use was documented in 80.5% of subjects (85.8% of mutation-positive subjects; 72.5% of mutation-negative subjects). The most frequently prescribed statin was rosuvastatin (N = 94; 47.0%). A higher proportion of mutation-positive subjects were receiving high-dose atorvastatin (40 or 80 mg per day) or rosuvastatin (20 or 40 mg per day) compared with mutation-negative subjects (Table 4).

Table 4.

Primary Lipid-Lowering Medication Use in Mutation-Positive^a (N = 120) and Mutation-Negative^b (N = 80) Subjects

Medicationc	Total (N = 200)	Mutation Positive (N = 120)	Mutation Negative (N = 80)
Any treatment	177 (88.5)	110 (91.7)	67 (83.8)
Ezetimibe	87 (43.5)	60 (50.0)	27 (33.8)
Fenofibrate	7 (3.5)	5 (4.2)	2 (2.5)
Nicotinic acid	23 (11.5)	17 (14.2)	6 (7.5)
Red yeast rice	3 (1.5)	2 (1.7)	1 (1.3)
Bile acid sequestrants	25 (12.5)	17 (14.2)	8 (10.0)
Colesevelam	18 (9.0)	12 (10.0)	6 (7.5)
Colestipol	5 (2.5)	3 (2.5)	2 (2.5)
Cholestyramine	2 (1.0)	2 (1.7)	0 (0.0)
Statins	161 (80.5)	103 (85.8)	58 (72.5)
Atorvastatin	40 (20.0)	31 (25.8)	9 (11.3)
40 mg/day	16 (8.0)	14 (11.7)	2 (2.5)
80 mg/day	15 (7.5)	12 (10.0)	3 (3.8)
Fluvastatin	4 (2.0)	2 (1.7)	2 (2.5)
Pitavastatin	3 (1.5)	1 (0.8)	2 (2.5)
Lovastatin	2 (1.0)	1 (0.8)	1 (1.3)
Pravastatin	8 (4.0)	3 (2.5)	5 (6.3)
Rosuvastatin	94 (47.0)	62 (51.7)	32 (40.0)
20 mg/day	16 (8.0)	13 (10.8)	3 (3.8)
40 mg/day	52 (26.0)	35 (29.2)	17 (21.3)
Simvastatin	16 (8.0)	7 (5.8)	9 (11.3)

Data represent number (%). Patients could be on more than one medication.

a. Excludes subjects carrying only the p.A391T or p.T726I variants of LDLR.

b. Includes subjects carrying only the p.A391T or p.T726I variants of LDLR.

c. World Health Organization-Drug (v02Q2) coding dictionary applied.

3. Discussion

This observational study was conducted to characterize the prevalence of disease-causing mutations in individuals with a definite or possible clinical diagnosis of FH and to investigate differences in phenotypes of mutation-positive and mutation-negative patients. As additional evidence of the veracity of the clinical diagnosis of FH in our cohort, the frequency of tendon xanthomas and on treatment LDL-C levels were much higher than the results from the FH Foundation patient registry, despite higher rates of treatment with statin and nonstatin LDL-C–lowering medications in our cohort [57, 58]. Notably, a large proportion of subjects (40.0%) diagnosed with heterozygous FH according to strict, standard clinical criteria did not have an identifiable variant explaining the hypercholesterolemic phenotype. Of the 120 subjects with genetically confirmed FH, 114 subjects were carriers of pathogenic missense, frameshift, and/or truncation mutations in LDLR, the majority of which have been previously described in patients with FH [2, 9, 40, 43]. Of the 8 novel LDLR variants, 5 were nonsense or frameshift mutations and 2 were large genomic rearrangements predicted to result in LOF of LDLR. One novel missense variant (p.C116S) affected a conserved residue mutated in known FH variants and was not represented in the single nucleotide polymorphism database. Six subjects were carriers of the most frequent APOB mutation (p.R3527Q; previously referred to as p.R3500Q) that results in familial ligand-defective apolipoprotein B-100 (FDB) [11]. No subject was a carrier of any of the GOF mutations in PCSK9 associated with autosomal dominant hypercholesterolemia [9, 42, 43].

Interestingly, we identified 2 subjects with homozygous or compound heterozygous FH (1.0% of the total cohort) [55], whose LDL-C levels remained high after treatment. These data indicate that genotypic homozygous and compound heterozygous FH may be far more common than previously thought, as also suggested by other investigators [17, 18]. Furthermore, treated LDL-C levels in these patients may be indistinguishable from those seen in patients with heterozygous FH.

The increased use of genetic testing has revealed that a significant proportion (up to 70% in some studies) of patients with a clinical diagnosis of FH do not have mutations in any of the loci known to cause FH [23–37]. The mutation detection rate in FH is largely dependent on ascertainment strategy, diagnostic criteria, and genotypic approaches as well as the ethnicity of the population being tested [59]. In the study by Ahmad et al. [60], using techniques similar to those used in this study to identify pathogenic gene variants in FH, 66% of the patients were mutation negative. This cohort included a large number of African Americans, 77% of whom were mutation negative. Our study included only 11 African American subjects. Interestingly, 90% of them were mutation negative, whereas among the vast majority of our subjects who were non-Hispanic white (93.5%), the mutation-negative frequency was 36%. There is a paucity of data on FH prevalence in African Americans, who are more often unaware of their hypercholesterolemic status than the white population [61]. These data add to the notion that the lack of genotypic confirmation of FH status may be more common among understudied populations such as African Americans.

Our study included 68 subjects from Quebec where the prevalence of FH is higher due to founder effects. As expected, 49 of the 68 subjects (72.1%) were mutation positive, including 20 carriers of a large deletion encompassing the promoter and exon 1 (LDLR-1P, exon 1 del), and 8 carriers of the p.W87G mutation (including 1 homozygote), both common mutations in the French-Canadian population [62]. But even in this population, 27.9% of subjects with a clinical diagnosis of FH did not have genotypic confirmation of FH. After excluding 68 subjects from Quebec, of the remaining 132 patients, 54% were mutation positive and 46% were mutation negative. These data are not significantly different compared to the overall results of 60% mutation positive and 40% mutation negative among 200 patients (P = 0.3; 95% CI of difference; –17% to 5%). Thus, the overall conclusions of our study remain the same even after exclusion of the subjects from Quebec.

Another limitation of our study is related to mandating a current screening fasting plasma LDL-C ≥ 100 mg/dL (2.6 mmol/L) for individuals on lipid-lowering therapy for possible future eligibility for a clinical trial. This criterion appears to have contributed to the exclusion of up to 56 out of 256, that is, 22% of the study participants in whom the prevalence of disease-causing mutations remains unknown. This could have introduced some bias in our results, but this exclusion may have actually enriched the study cohort in patients with genetically confirmed FH, since patients with LDL-C < 100 mg/dl on lipid-lowering therapy may be less likely to have FH. Limitations of our genotyping methodology might have contributed to the underdetection of molecular defects causing hypercholesterolemia in this cohort. We did not sequence other genes such as APOE, ABCG5, ABCG8, and LDLRAP, which may be rarely responsible for hypercholesterolemia. Synonymous variants in LDLR, APOB, and PCSK9 coding regions were not considered as causative of hypercholesterolemia, as it is extremely rare to see an effect of such variants on protein levels or function [63]. Since we only captured mutations in the coding regions of LDLR, PCSK9, and APOB, the splice-site variants were missed. We limited our genotyping to exon 26 of APOB. Although mutations outside exon 26 of APOB have been identified in rare cases of FH, subsequent segregation analyses have not identified a causal role for these variants [64, 65]. This was an exploratory study, and we consider the results as hypothesis generating. No power or sample size calculations were made.

Familial combined hyperlipidemia (FCH), an inherited hyperlipidemia with an estimated prevalence of 1.0% to 2.0% in the general population and characterized by increased total cholesterol and/or triglycerides levels, has a complex, polygenic mode of inheritance, making the definitive diagnosis of this disorder and its differentiation from FH difficult [66, 67]. Xanthomas and xanthelasmas were slightly, but not significantly, more prevalent in mutation-positive subjects (47.5%) than mutation-negative subjects (35.0%), although total cholesterol, LDL-C, and high-density lipoprotein–C levels were comparable. Definite FH according to Simon Broome criteria was present in 40.8% of mutation-positive and in 31.3% of mutation-negative subjects. This difference was not statistically significant. However, historical untreated lipid values were not collected as part of the study protocol although this information was used by individual investigators to help establish the diagnosis of FH.

In addition to higher blood pressure and the prevalence of hypertension in the mutation-negative group compared to mutation-positive group; the mutation-negative subjects were older, had higher serum triglycerides, a trend toward greater prevalence of diabetes, and reduced prevalence of family history of hypercholesterolemia. All these differences in the phenotypes point to possible differences in the pathogenesis of hypercholesterolemia in the 2 groups, with the possibility of a polygenic cause of hypercholesterolemia in some mutation-negative patients compatible with FCH. Unfortunately, we did not analyze polygenic risk scores as our genotyping was limited to the 3 FH-causing genes. The clinical relevance of higher plasma PCSK9 levels in the mutation-positive group compared with the mutation-negative group is unclear, but it is likely attributable to the reciprocal inverse relationship between LDLR and PCSK9 [68, 69].

In our cohort, 91.7% of mutation-positive and 83.8% of mutation-negative subjects were taking 1 or more lipid-lowering medications. Despite this, total cholesterol and LDL-C levels remained high, with two-thirds of subjects (69% of mutation-positive and 64% of mutation-negative) having an LDL-C > 130 mg/dL. This may reflect the lack of consensus on which therapeutic lipid targets are appropriate in patients with heterozygous FH or difficulty in reducing LDL-C levels despite use of high statin doses in patients with FH who have severe untreated hypercholesterolemia. Specific and uniform guidelines for the management of LDL-C in patients with FH—irrespective of whether they have a molecular diagnosis—will facilitate improved management of these individuals at high risk of CHD and other atherosclerotic vascular complications [14, 17, 18, 70].

Early diagnosis and effective treatment of FH is especially important owing to the increased risk of premature CHD and atherosclerotic vascular disease. Genetic confirmation of the diagnosis suggests lifelong exposure to hypercholesterolemia,[71] thus justifying aggressive lipid lowering for the probands and a search for affected family members (cascade screening). Selective genotyping may also be useful in genetic counseling of individuals with FH [72]. However, plasma LDL-C levels show large variability in heterozygous FH, even among subjects with identical mutations, presumably due to the effects of other lipid-modifying genetic and nongenetic factors. Some investigators have reported normal levels of LDL-C among individuals with genotypic FH—particularly among those identified through cascade genetic screening, however, some low frequency LDLR variants may not have been excluded [73]. On the other hand, genotyping of suspected patients with FH in the clinical setting is complex, costly, and may prove to be counterproductive since a lack of genetic confirmation may lead to unwarranted loosening of intervention strategies.

The identification of underlying factors, including genetic and environmental, in mutation-negative patients is important to further our understanding of the genetic and metabolic basis of hypercholesterolemia. While some mutation-negative patients may harbor novel rare variants in as yet unknown FH genes, others may accumulate multiple common variants in lipid-modulating loci as suggested by some investigators [74]. It is also possible that epigenetic or nongenetic factors are at play in the development of a hypercholesterolemia whose severity overlaps with FH. Future studies are needed to identify the causes of severe hypercholesterolemia in mutation-negative FH subjects.

4.  Conclusions

In this North American cohort, 60% of clinically diagnosed patients with FH were confirmed to have disease-causing mutations in the LDLR and APOB genes, and none had PCSK9 mutations. There were some differences among the phenotypes of mutation-positive and mutation-negative subjects, but overall it was not possible to predict the presence of mutations based on phenotype, including the presence or absence of tendon xanthomas that are considered diagnostic for FH in the absence of sitosterolemia and cerebrotendinous xanthomatosis. Genotyping may help in the confirmation of the clinical diagnosis and in the early initiation of aggressive medical management, but it may not be desirable if mutation-negative subjects are left untreated or denied novel therapies. Future studies are needed to elucidate the risk of atherosclerotic vascular disease among the mutation-positive and mutation-negative subjects with a clinical diagnosis of FH and to understand the pathogenesis of hypercholesterolemia in the mutation-negative group.

Glossary

Abbreviations

APOB

apolipoprotein B

CHD

coronary heart disease

FH

familial hypercholesterolemia

GOF

gain of function

LDL-C

low-density lipoprotein–cholesterol

LDLR

LDL receptor

LOF

loss of function

PCSK9

proprotein convertase subtilisin/kexin type 9

QC

quality control.

Additional Information

Disclosure Summary: Dr. Garg has served on advisory boards, and has received research grants from Pfizer Regeneron, Akcea and Aegerion. Dr. Fazio has served as a consultant to Amgen, Amarin, Aegerion, Akcea, Esperion, and Novartis. Dr. Duell has served as a consultant to Akcea, Astra-Zeneca, Esperion, RegenxBio, Retrophin, Regeneron/Sanofi and has received institutional support for research from Esperion, Regeneron, RegenxBio, and Retrophin. Drs. Udata, Joh, Sirota, Dettling, Liang, Garzone, Gumbiner, and Wan are or were employees of Pfizer with ownership of stock in Pfizer; Mr. Riel was an employee of Pfizer when this study was conducted. Dr. Baass serves on advisory boards, is a consultant for, and has received research grants from Sanofi/Regeneron and Amgen; has received research grants from Astra-Zeneca and Merck; and is a speaker for Akcea.