Individuals were selected from the Aragon Workers’ Health Study (AWHS). The AWHS is a longitudinal cohort study of cardiovascular risk factors and subclinical atherosclerosis performed among 5400 workers at the Opel Spain factory in Aragon, Spain, who have been followed up since 2009.21 Here, we studied unrelated individuals ≥ 18 years old from the AWHS, with annual or biannual determinations since 2009. The following criteria were used to select those with clinical suspicion of FH: a) participants with untreated LDL-C concentrations> 95th percentile of the Spanish population stratified by age and sex22 in at least 2 measurements; b) participants with premature CHD (< 55 years in men and <60 years in women) at baseline; c) participants under statin therapy and LDL-C> 130 mg/dL in at least 2 measurements; d) participants with untreated LDL-C concentrations> 95th percentile of the Spanish population in just 1 measurement but who also had LDL-C> 130mg/dL under statin therapy. The exclusion criteria were the presence of secondary causes of hypercholesterolemia at baseline or during follow-up, which include severe obesity (body mass index [BMI]> 35 kg/m2), poorly controlled type 2 diabetes (HbA1c> 8%), renal disease with glomerular filtration rate <30mL/min and/or macroalbuminuria, liver disease (alanine transaminase> 3 times the upper normal limit), hypothyroidism (thyroid stimulating hormone> 6 mIU/L), pregnancy, autoimmune diseases, and treatment with protease inhibitors. We defined hyperLp(a) participants as those participants with suspicion of FH who had Lp(a)> 50 mg/dL, but normal LDL-C concentrations adjusted by Lp(a) (< 95th percentile of the Spanish population)22.

All participants gave written informed consent, and the study was approved by the Aragón Ethics Committee for Clinical Research (CEICA).

Assessment of cardiovascular risk factors, personal and family history of cardiovascular disease, intake of drugs affecting lipid metabolism, and anthropometric measurements were performed in all participants at baseline and annually or biannually during the follow-up.

Lipid and lipoprotein analyses were performed on EDTA plasma samples collected after an overnight fast of at least 10hours annually or biannually since 2009. Measurements up to 2015 were included. Total cholesterol and triglyceride levels were determined by standard enzyme methods. High-density lipoprotein cholesterol (HDL-C) was measured directly by an enzyme reaction using cholesterol oxidase (UniCel DxC 800; Beckman Coulter, Inc, USA). LDL-C was calculated by Friedewald's formula. Lipid values were adjusted according to statin therapy in those participants taking lipid-lowering drugs.23 None of these participants received PCSK9 inhibitors. Lp(a) was determined by IMMAGE kinetic nephelometry (Beckman Coulter, Inc, USA). LDL-C adjusted by Lp(a) was obtained by subtracting the concentrations of Lp(a) divided by 3 to the LDL-C concentration calculated by the Friedewald formula.24,25

Whole blood genomic DNA was isolated using standard methods. LDLR (NM_000527.4), PCSK9 (NM_174936.3), APOE (NM_000041.3), STAP1 (NM_0121108.3), and LDLRAP1 (NM_015627.2) genes and exons 26 and 29 of APOB (NM_000384.2), which encodes the LDLR binding site, were sequenced by next generation sequencing with the SeqProLipo Platform (Progenika Biopharma Grifols, Spain). This platform includes point mutations, large rearrangements, and copy number variations.

To evaluate the pathogenicity of identified variants we used SIFT,26 PolyPhen-2,27 Mutation Taster,28 and PredictSNP.29 The effect of variants in potential splicing sites was predicted with FruitFly.30 To compare the frequency of identified variants in the study participants with that of the general population, we compiled the allele frequencies of the identified variants from the 1000 Genomes Project31 and ExAc Browser Data.32 Rare variants were defined as genetic variants with a frequency <1% in the general population. A rare variant was defined as a causative variant when it had been previously associated with FH or possibly pathogenic when the bioinformatic analysis prediction classified it as pathogenic.

Analyses were performed using statistical computing software R version 3.5.0.33 The distribution of the variables was analyzed with the Shapiro test. Quantitative variables with a normal distribution are expressed as mean± standard deviation and were analyzed with the ANOVA test. Variables with a skewed distribution are expressed as median and interquartile range and were analyzed with the Kruskal-Wallis test. Qualitative variables are expressed as percentages and were analyzed with the chi-square test. The level of significance was set at P <.05. Logistic binary regression was used to determine the best European Atherosclerosis Society criteria diagnosis with the presence of a mutation in candidate genes as the dependent variable.

Of the 5400 individuals from the AWHS, 4514 participants had the information required by the inclusion and exclusion criteria. A total of 886 individuals were excluded because they had incomplete recorded lipid data, less than 2 complete lipid panel and/or incomplete information about lipid-lowering drug use. Of the 4514 participants, 255 (5.65%) met the following inclusion criteria (not mutually exclusive): a) 127 participants had untreated LDL-C concentrations> 95th percentile of the Spanish population in at least 2 measurements; b) 21 participants had premature CHD at baseline (< 55 years in men and <60 years in women); c) 39 participants had LDL-C> 130mg/dL in at least 2 measurements while taking statin therapy and; d) 113 participants had untreated LDL-C concentrations> 95th percentile of the Spanish population in 1 measurement and also had LDL-C> 130mg/dL on a least another occasion with statin therapy (figure 1).

Figure 1.

Selection criteria and identified variants in candidate familial hypercholesterolemia genes. CHD, coronary heart disease; LDL, low-density lipoprotein.

(0.41MB).

Participants with high levels of LDL-C despite statin therapy and those with CHD at baseline were older and had higher BMI than the rest of the participants. Participants with statin therapy and high levels of LDL-C despite drug treatment, together with participants with high LDL-C concentrations in 1 measurement and also high levels of LDL-C with statin therapy, had higher concentrations of total cholesterol, LDL-C and triglycerides than the remaining participants. Participants with premature CHD had higher levels of Lp(a). The percentage of participants with high levels of LDL-C in at least 2 measurements who did not take any lipid-lowering therapy was 93.7% (table 1).

Sequencing analysis reported 5280 variants in candidate genes among the 255 sequenced participants (figure 1): a) 1792 variants in LDLR, 11 of which have been described as pathogenic and previously associated with FH; b) 249 variants in APOB, although none of them has been described as pathogenic or previously associated with FH; c) 1697 variants in PCSK9, only 1 of which has been described as pathogenic and previously associated with FH; d) 72 variants in APOE, 2 of which have been previously associated with type III hyperlipoproteinemia; e) 869 variants in STAP1, 3 of which have been described as possibly pathogenic by bioinformatic analysis; f) 601 variants in LDLRAP1 gene, 2 of which have been classified as possibly pathogenic by bioinformatic analysis. However, neither of these 2 possibly pathogenic variants in LDLRAP1 were present in homozygosity, a condition that has been described as necessary to cause hypercholesterolemia.

The percentage of mutations in candidate genes was similar among groups; however, participants with high levels of LDL-C in at least 2 measurements showed the highest percentage of causative mutations, exceeding 8% of cases (table 1). A total of 12 patients were carriers of 11 distinct pathogenic alleles in heterozygosity in the LDLR: 1 of them with a causative variant located in the promoter region (c.-135C>G); 7 of them had a causative variant, located in coding regions, producing an amino acid change (c.530C>T, c.826T>G, c.862G>A, c.1247G>A, c.1529C>T, c.1775G>A, and c.1816G>A); 1 of them with a causative variant, located in an intronic region, producing an alternative splicing change (c.1586+5G>A); 1 allele had 2 causative variants located in cis: (c.274C>G) and (c.313+1G>C) and 1 of them had a large rearrangement (c.941-?_1845+?del). All of them have been described as pathogenic by bioinformatic analysis and have been previously reported as a cause of FH.34–40 Only 1 patient carried a rare variant in exon 26 of the APOB gene, (c.10621A>G). However, this variant has not been previously associated as a cause of FH, has been classified as benign by bioinformatic analysis, and seems not to affect the binding region structure. A total of 5 patients carried 2 rare variants in the PSCK9 gene (c.60_65dup GCTGCT and c.743G>A), and only the first one has been previously associated with FH. Four of these patients were carriers of the c.60_65dupGCTGCT in-frame indel, which has been described as uncertain or likely pathogenic by Clin Var41 and Garcia et al.42 A total of 5 patients carried 3 rare variants in the APOE gene, none of them previously described as a cause of FH. One of them has not been previously associated with any hyperlipidemia and has been classified as unknown by bioinformatic analysis (c.335C>A). The other 2 variants, (c.460C>A, c.487C>T), were present in 3 and 1 participants, respectively. Both variants have been previously associated with type III hyperlipoproteinemia.43,44 A total of 4 patients carried 3 unknown rare variants in the STAP1 gene; 2 of them carried 1 variant located at the 5’ region (c.-60A>G) and the other 2 were carriers of 2 different variants, located in coding regions, which produce amino acid substitution (c.619G>A and c.803T>C). None of these variants have been previously associated with FH. Two patients carried 2 unknown rare variants in heterozygous state in the LDLRAP1 gene, 1 of them producing an amino acid change (c.605C>G) and the other could affect splicing (c.748-7C>G). However, mutations in LDLRAP1 only produce hypercholesterolemia when they are found in homozygosity (table 2 and table 1 of the supplementary data).

In 24 (9.41%) participants with criteria for suspicion of FH due to LDL-C> 95th percentile, when the cholesterol transported in Lp(a) was analyzed, their LDL-C were no longer> 95th percentile. None of these hyperLp(a) participants was a carrier of a pathogenic mutation in FH candidate genes.

Table 3 shows the clinical characteristics and lipid profile according to the genetic classification including hyperLp(a) participants. Participants carrying LDLR mutations had the highest levels of total and LDL-C. In addition, participants carrying PCSK9 mutations had significantly higher levels of total and LDL-C than those carrying STAP1 mutations and participants classified with hyperLp(a) (P <.001 and P <.001, respectively). Moreover, participants carrying STAP1 mutations had the lowest levels of total and LDL-C, and had lower levels of total and LDL-C than participants without mutation in candidate genes.

Mutations in the LDLR gene were responsible for 75% of some mutations in candidate genes, vs 25% of those in participants who carried some mutation in the PSCK9 gene. Figure 2 shows the percentage of mutations according to groups by baseline LDL-C levels. The higher the percentage, the greater the increase in LDL-C: 0.40% of participants with baseline LDL-C <190mg/dL, 1.20% of participants with baseline LDL-C from 190-220mg/dL, and 4.80% of participants with baseline LDL-C> 220mg/dL had a mutation in candidate genes. Considering those patients with a mutation in candidate FH genes plus those with hyperLp(a), almost 16% of the group with criteria for suspicion had monogenic disease.

Figure 2.

Percentage of participants with causative mutations according to baseline levels of low-density lipoprotein cholesterol. LDL-C, low-density lipoprotein cholesterol.

(0.07MB).

Logistic binary regression was used to study the association of clinical data with 2 response variables: mutation in any gene and mutation in the LDLR gene. The analysis showed that baseline LDL-C, LDL-C adjusted by Lp(a) and total cholesterol levels were significantly associated with the presence (yes/no) of any mutation in candidate genes (OR, 1.023; P <.001; OR, 1.024; P <.001 and OR, 1.0.17; P <.001 respectively, table 4A-C). Nevertheless, no criterion showed a significant association with the presence of any mutation in candidate genes. The presence of mutation in the LDLR (yes/no) was significantly associated with baseline total and LDL-C (OR, 1.009; P=.010 and OR, 1.0.16; P=.0488, respectively) and the criteria for suspicion of FH: untreated LDL-C concentrations above the 95th percentile in 1 measurement and also LDL-C> 130mg/dL while taking statins (P=.011, table 4D).

Table 5 shows the number of participants with a mutation based on the criteria they met, positive and negative predictive values and sensitivity and specificity regarding the sample selected due to the presence of at least 1 criterion. The first criterion had the highest positive predictive value (10.2%) to detect the presence of functional mutations in candidate genes (LDLR, PCSK9 and STAP1 genes), especially to detect mutations in the LDLR and PCSK9 genes (6.3% and 2.4%). Nevertheless, the fourth criterion showed the highest positive predictive value to detect mutations in the LDLR gene (8.0%), showing in this selected sample higher specificity than the first criterion (0.510 and 0.572, respectively).

Our study has some limitations. First, AWHS is a cohort with a high prevalence of men, obesity, hypertension, and hypercholesterolemia, which may not be representative of the general population. However, because it is a relatively young population there, will be no bias in the prevalence of FH mutations. Second, we report some mutations, c.743G>A in the PCSK9 gene and c.10621A>G in APOB, with uncertain significance. However, both are described as neutral or benign by bioinformatic analysis and the ClinVar database, so there are not enough data to classify them as pathogenic variants. Sequencing analysis of the APOB gene included only exon 26 and exon 29, while sequencing the entire APOB gene could report other variants as a cause of FH phenotype. Nevertheless, hypercholesterolemia due to APOB defects would require that the protein would have defective binding to the LDL receptor, since mutations causing only lower expression or improper folding of ApoB would produce hypocholesterolemia. Therefore, the variants causing FH are expected to be located in the coding region of the ApoB binding domain (in exon 26 and 29). In fact, most APOB mutations described as a cause of FH have been identified in exon 26.49 Finally, possibly pathogenic mutations in the PCSK9 gene (c.60_65dupGCTGCT) and in the STAP1 gene (c.-60A>G, c.619G>A and c.803T>C) need further studies to gain greater insight into their role in the FH phenotype.