Author + information
- Received January 9, 2008
- Revision received May 6, 2008
- Accepted June 2, 2008
- Published online November 4, 2008.
- Fernando Civeira, MD, PhD⁎,⁎ (, )
- Estibaliz Jarauta, MD⁎,
- Ana Cenarro, PhD⁎,
- Angel L. García-Otín, PhD⁎,
- Diego Tejedor, PhD†,
- Daniel Zambón, MD, PhD‡,
- Miguel Mallen, BSc§,
- Emilio Ros, MD, PhD‡ and
- Miguel Pocoví, PhD§
- ↵⁎Reprint requests and correspondence:
Dr. Fernando Civeira, Hospital Universitario Miguel Servet, Avda. Isabel La Católica 1-3, Zaragoza 50009, Spain
Objectives The purpose of this study was to determine the frequency of mutations in the low-density lipoprotein receptor (LDLR) and apolipoprotein B (APOB) genes in consecutive patients with a clinical diagnosis of familial combined hyperlipidemia (FCH) in a nonresearch setting.
Background The lipid phenotype frequently overlaps in familial hypercholesterolemia (FH) and FCH. Detection of causative mutations in LDLR or APOB provides an unequivocal diagnosis of FH, but such genetic testing has not been systematically performed in FCH.
Methods We used Lipochip (Progenika, Derio, Spain), a microarray that includes 203 causative mutations in LDLR and 4 APOB defects, to investigate 143 unrelated FCH patients.
Results Mutations of LDLR were found in 28 patients (overall prevalence, 19.6%). No APOB defects were found. Compared with patients who had a normal LDLR gene, patients with mutations had lower waist circumference (p = 0.02); significantly (p < 0.005) higher total cholesterol, non–high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and apoB; nonsignificantly (p = 0.063) lower triglycerides; and a lower frequency of diabetes mellitus (22% vs. 0%, respectively; p = 0.002). Total cholesterol and apoB levels showed the best receiver-operator characteristics curves in the prediction of LDLR mutations, with areas under the curve (95% CI: of 0.750 (95% confidence interval [CI]: 0.647 to 0.853) and 0.744 (95% CI: 0.636 to 0.851), respectively. Total cholesterol of 335 mg/dl and apoB of 185 mg/dl were the best thresholds for diagnosis of LDLR mutations.
Conclusions Screening for LDLR defects is advisable for patients with a clinical diagnosis of FCH showing high total cholesterol or apoB levels. Diagnostic criteria for FH should not exclude patients whose personal and familial lipid values appear to fit the clinical criteria of FCH.
The low-density lipoprotein receptor (LDLR) is a cellular surface protein that recognizes and internalizes low-density lipoprotein (LDL) particles. Dysfunction of the LDLR causes familial hypercholesterolemia (FH), one of the most common inherited metabolic diseases in humans (1). Homozygous FH is a very rare condition, but the frequency of heterozygous FH is close to 1 in 500 persons in most populations. Elevated blood concentrations of LDL-cholesterol, a high risk of early-onset coronary heart disease, and extravascular lipid deposits such as arcus cornealis and tendon xanthomas characterize FH (1). Blood triglycerides in FH usually are in the normal range; thus, the typical lipid abnormality is an isolated elevation of LDL-cholesterol. However, patients with FH disclose higher triglyceride levels than do their unaffected relatives and/or age-matched subjects from the general population (2,3); and as many of 25% of heterozygous FH patients present a mixed hyperlipidemia phenotype, with elevation of both LDL-cholesterol and triglycerides (4). This subgroup of hypertriglyceridemic FH represents a diagnostic challenge, because mixed hyperlipidemia with familial presentation is characteristic of familial combined hyperlipidemia (FCH), another common inherited lipid abnormality characterized by premature coronary heart disease (5). In fact, the usual diagnostic criteria for FH contemplate that either serum triglycerides should be normal (<200 mg/dl) (6) or the diagnosis must be considered with caution in patients with hypertriglyceridemia (7). The problem is compounded by the equivocal diagnostic criteria for FCH. This disorder is associated with variable lipid phenotypes, both in affected persons and family members, with a clinical expression that is closely related to visceral fat accumulation, and the recently proposed FCH criterion of an elevated apolipoprotein (apo) B level (8) is also present in FH. Moreover, tendon xanthomas, a key diagnostic feature, are clinically detectable in only ≈40% of FH patients after the fourth decade of life (9,10).
The overlapping clinical presentation of FCH and FH with high triglycerides has been previously suspected in the Simon Broome Registry in the United Kingdom (11), wherein patients with a clinical diagnosis of FH but no tendon xanthomas disclosed lower blood cholesterol and higher triglycerides, blood pressure, and body mass index than did patients with tendon xanthomas. These differences suggested to the authors that the diagnostic criteria used in that cohort misclassified as FH cases some patients with FCH (11). Evidence of such misclassification was reported in a 3-generation pedigree with an apparent diagnosis of FCH in whom a defective LDLR allele was shown to cause the familial hypercholesteromic phenotype (12).
A definitive diagnosis of FH can be made by molecular testing showing a pathogenic defect in the LDLR or APOB genes. Such an approach should unequivocally clarify the diagnosis between FH and FCH in doubtful cases. To our knowledge, the LDLR and APOB genes have not been systematically screened in subjects with a clinical diagnosis of FCH. The aim of this study was to describe the frequency of LDLR and APOB mutations in a large group of unrelated subjects who had received a diagnosis of FCH, based on the presence of mixed hyperlipidemia in the index case, a variable lipid phenotype in first-degree relatives, and absence of tendon xanthomas.
From October 2005 to June 2007, consecutive patients ages ≥25 years with a clinical diagnosis of FCH attending 2 lipid clinics in Spain were recruited into a protocol approved by the local review boards and provided informed consent. The diagnosis of FCH was based on the presence of primary combined hyperlipidemia, with off-treatment serum cholesterol and triglyceride levels above the gender- and age-adjusted 90th percentile for the Spanish population (13), serum total apoB levels ≥120 mg/dl, and at least 1 first-degree relative with hyperlipidemia (total cholesterol and/or triglycerides >90th percentile). Off-treatment lipid values of first-degree family members were obtained from clinical records, and they were classified into the following categories: 1) normolipidemic (<90th percentile for total cholesterol and triglycerides); 2) type 2a (≥90th percentile total cholesterol and <90th percentile triglycerides); 3) type 2b (≥90th percentile total cholesterol and ≥90th percentile triglycerides); and 4) type 4 (<90th percentile total cholesterol and ≥90 percentile triglycerides). Gender- and age-adjusted 90th percentile cutpoints for the Spanish population were used (13). For the purpose of this study, additional inclusion criteria were serum LDL-cholesterol ≥190 mg/dl or non–high-density lipoprotein (HDL)-cholesterol ≥220 mg/dl when serum triglycerides were ≥400 mg/dl. Exclusion criteria were secondary causes of hyperlipidemia, presence of tendon xanthomas in the proband or in any family member, a previous clinical or molecular diagnosis of FH in a first-degree relative, or very high triglycerides level (>1,000 mg/dl). To exclude the presence of xanthomas uncovered by physical examination, Achilles tendons were examined by sonography with a standard protocol shared by the 2 clinics, as described (14).
Clinical and laboratory determinations
All subjects were assessed for family history of early-onset coronary heart disease, clinical history, medication use, demographic characteristics, adiposity measures, cardiovascular risk factors, and presence of tendon xanthomas. Diabetes mellitus was defined as fasting glucose level ≥126 mg/dl or treatment with antidiabetic agents.
In asymptomatic subjects, fasting blood for baseline biochemical profiles was drawn after at least 4 weeks without hypolipidemic drug treatment. In patients with prior coronary heart disease, baseline lipid values were obtained from clinical records. Cholesterol and triglycerides were determined by standard enzymatic methods. HDL-cholesterol was measured by a precipitation technique. LDL-cholesterol was estimated with the Friedewald equation when serum triglycerides were <400 mg/dl. Non–HDL-cholesterol was calculated as total cholesterol minus HDL-cholesterol. Both apoB and lipoprotein(a) were determined by using immunoturbidimetry (Unimate 3, Roche, Basel, Switzerland).
The screening of mutations in LDLR and APOB genes was performed with Lipochip, version 4 (Progenika, Derio, Spain). This platform is a microarray device that includes the diagnosis of 203 different causative mutations in the LDLR gene and 4 mutations in the binding domain of apoB. The microarray design, fabrication, and quality controls, the target deoxyribonucleic acid preparation and hybridization, the microarray scanning and quantification, and the genotyping software have been previously described in detail (15). Lipochip version 4 recognizes 88% of the 230 different LDLR mutations detected in Spain thus far. The overall specificity and sensitivity obtained for all mutations tested was 99.7% and 99.9%, respectively (16). Because the most common mutations are included in the array, after testing ≈5,000 deoxyribonucleic acid samples from patients with a clinical diagnosis of FH, we can be confident that Lipochip version 4 identifies ≈95% of genetic defects causing FH in Spain (D. Tejedor, personal communication, December 2007). Large rearrangements in the LDLR gene were analyzed using a method based on quantitative fluorescent multiplex polymerase chain reaction (17).
The APOE genotype, defined by the common polymorphisms p.Cys130Arg (C112R) and p.Arg176Cys (R158C), and the 2 rare variants p.Arg154Ser (R136S) and p.Lys167del (L149del), was determined by specific pyrosequencing reactions on a PSQ96MA instrument (Biotage AB, Uppsala, Sweden).
Data are presented as means (± SD) for continuous variables (medians and interquartile ranges for variables with a skewed distribution) and as frequencies or percentages for categorical variables. Differences in mean values were assessed using t tests or the Mann-Whitney U test, as appropriate. Categorical variables were compared using chi-square tests. Univariate logistic regression was used to determine the ability of clinical variables and lipid and apo concentrations to discriminate between cases with and without LDLR defects, and crude odds ratios with 95% confidence intervals (CIs) were calculated. Because of a skewed distribution, triglyceride levels were transformed to their natural logarithm for analyses. We determined the threshold values of these variables for best diagnostic performance of the presence of a LDLR defect based on analysis of receiver-operator characteristic (ROC) curves. The ROC curves were constructed by using mutation carriers as the disease group and the subjects without mutations as the nondisease group. Because all cases of diabetes mellitus occurred in subjects without genetic defects, the odds ratios calculated in the logistic regression analysis gave inconsistent results and very large CIs. For this reason, the odds ratio for diabetes was calculated by using the 2-tailed Fisher exact test.
Multivariate logistic regression analysis with the forward selection based on likelihood ratio was used to examine the independent association of different variables with the presence of FH-causing genetic defects. The independent variables used in this model were as follows: gender, age, body mass index, waist circumference, the serum levels of total cholesterol, LDL-cholesterol (or non–HDL-cholesterol), apoB, and natural logarithm triglycerides, and presence or absence of a total cholesterol level above the gender- and age-specific 95th percentile for the Spanish population (13). Diabetic subjects and subjects disclosing the APOE 2/2 genotype or rare APOE variants were excluded from the multivariate analysis. All statistical analyses were performed with SPSS software (version 13.0, SPSS, Inc., Chicago, Illinois), with significance set at p < 0.05.
A total of 143 unrelated FCH subjects, 93 men and 50 women ages 51 ± 11 years, were included in the study. A personal history of coronary heart disease and a family history of early-onset coronary heart disease were present in 30 (21%) patients and 44 (31%) patients, respectively. Mean plasma total cholesterol, triglycerides, and HDL-cholesterol levels were 326 ± 57 mg/dl, 352 ± 175 mg/dl, and 42 ± 14 mg/dl, respectively.
Frequency of LDLR mutations
A functional LDLR mutation was found in 28 (19.6%) patients. No APOB mutations were detected. The clinical and biochemical characteristics of patients with (LDLR+) and without (LDLR−) mutations are shown in Table 1. In comparison with LDLR− patients, patients shown to be LDLR+ disclosed higher serum levels of total cholesterol, LDL-cholesterol, non–HDL-cholesterol, and apoB. In contrast, triglyceride levels were higher, although not reaching statistical significance, in LDLR− patients compared with those in LDLR+ patients (p = 0.059). The frequency of mutations in women (28%) was nonsignificantly (p = 0.063) higher than in men (15%). The LDLR− patients also had a higher waist circumference, even after adjustment for gender. Diabetes was present in 26 LDLR− patients (22.6%) and was absent in LDLR+ patients.
The APOE genotype of 2 patients in the LDLR− group was E2/E2, indicating that their combined hyperlipidemia phenotype was due to dysbetalipoproteinemia. In addition, 1 E3/E3 woman and 1 E3/E3 man in the LDLR− group were heterozygous for the rare alleles L149del and S136, respectively. Both mutations have been shown to be associated with dyslipidemia (18,19). No rare APOE variants were present in the LDLR+ group.
Types of LDLR mutations
The clinical and biochemical characteristics, APOE genotype, and LDLR defect of the 28 LDLR+ patients is shown in Table 2. A total of 22 different mutations distributed along the whole gene were found, indicating high molecular heterogeneity. Three subjects were heterozygous for the double defect 1061-8T>C/T726I; the mutations 91G>T, 518delG, 829G>A, and 1133A>C were found in 2 subjects each. The 2 mutations (1255T>C and 1359-3C>T) have not been previously reported, although they have been found in the Spanish FH population. The proportion between negative-receptor and defective-receptor mutations was 1:3 (5 subjects had negative alleles, 16 subjects had defective alleles, and 7 patients had splicing mutations with undetermined effect on LDL receptor activity) (Online Table).
Predictors of LDLR mutations by logistic regression and ROC analyses
Crude odds ratios associated with the presence of a LDLR mutation are shown in Table 3. The variables that contributed positively were absence of diabetes, total cholesterol >95th percentile, and serum concentrations of total cholesterol, LDL-cholesterol, non–HDL-cholesterol, and apoB. Waist circumference was negatively associated with presence of a LDLR mutation. The best area under the curve in the ROC analysis was given by apoB, although without significant differences from the values obtained for total cholesterol and LDL-cholesterol.
To assess whether any combination of variables could improve the prediction of having a LDLR defect, independent associations were sought by multivariate logistic regression. The concentration of apoB was the sole independent factor associated with a LDLR mutation, with odds ratio 1.028 (95% CI: 1.013 to 1.044; p < 0.001). The area under the curve of apoB in the ROC analysis was 0.744 (95% CI: 0.636 to 0.851; p < 0.001). When apoB was excluded from the model, total cholesterol was the independent association, with odds ratio 1.019 (95% CI: 1.009 to 1.030; p < 0.001), and the area under the curve in the ROC analysis was 0.750 (95% CI: 0.647 to 0.853; p < 0.001). The ROC statistics were also used to evaluate sensitivity and specificity of thresholds for these variables. The optimal threshold was chosen so that the sum of the sensitivity and the specificity to discriminate LDLR+ from LDLR− subjects was maximal. A total cholesterol level ≥335 mg/dl and apoB ≥185 mg/dl provided the best cutoff values, with sensitivities and specificities of 61% and 72% and 61% and 74%, respectively.
Considering the association of LDLR mutations with clinical characteristics and serum cholesterol and apoB values, we calculated their frequency in various subgroups. As shown in Figure 1, the prevalence of LDLR mutations increased from 19.6% in the overall series to 41.7% among nondiabetic subjects with normal APOE genotype and a total cholesterol level ≥335 mg/dl or apoB ≥185 mg/dl.
This study shows that a substantial number of patients with a clinical diagnosis of FCH disclose a functional mutation in the LDLR gene and thus should be classified as FH bearers. These results are novel and firmly support the previously suspected notion that patients with clinical FH who present with hypertriglyceridemia and a lipid phenotype overlapping that of FCH may be misdiagnosed as having FCH (11). They also suggest that FCH is a phenotype rather than a disease, and that FH could be considered as one of its causes.
Hypertriglyceridemia is a not uncommon feature of FH that may have several causes. First, LDL particles carry small amounts of triglycerides; therefore, if there is a large number of circulating LDL particles, some degree of triglyceride elevation might be expected. Second, in addition to clearing LDL from the circulation, the LDLR removes some triglyceride-carrying very low-density and intermediate-density lipoproteins (l), and it has been clearly demonstrated that severe LDLR protein defects, especially those affecting repeat 5 of the binding domain, can interfere with intermediate-density lipoprotein uptake by liver cells (20). Third, the coexistence in the same patient of FH with other genetic lipid abnormalities, such as FCH or dysbetalipoproteinemia (21), could explain the concomitant and severe increases of cholesterol and triglycerides observed in some patients with a molecular diagnosis of FH. Lastly and probably more important, many common genetic and environmental factors contribute to triglyceride elevations in the population (22), and their influence could be magnified in the presence of LDLR defects. For example, the hypertriglyceridemic effect of a frequent functional mutation in the lipoprotein lipase gene (D9N) is higher in FH patients than in the general population (23). Our results support the existence of common polygenic and environmental triglyceride-raising factors for several reasons: the LDLR mutations found are also present in normotriglyceridemic FH; all types of mutations, not only severe ones, occurred in our FCH patients; and the observed 1 to 3 proportion of negative- to defective-receptor alleles was similar to that reported for the overall Spanish FH population (24).
LDLR mutations are rare in subjects with mixed hyperlipidemia and diabetes
One important observation of our study is the striking difference in the frequency of type 2 diabetes mellitus between the groups with and without LDLR mutations. Clearly, FCH is closely linked with obesity, insulin resistance, and diabetes (25); and the presence of these metabolic abnormalities coexisting with elevated blood cholesterol and triglycerides levels is highly predictive of the absence of LDLR defects in patients with combined hyperlipidemia.
The results of APOE genotyping are interesting, inasmuch as 4 of 115 patients (3.5%) with a clinical diagnosis of FCH and no LDLR mutations had molecular defects in APOE that are characteristic of type III hyperlipidemia or dysbetalipoproteinemia, which also presents with combined hyperlipidemia overlapping that of FCH (21). These findings indicate that APOE genotyping is mandatory before a diagnosis of FCH is considered.
Strengths and limitations
This study has the strengths of a cross-sectional design in a sizeable cohort of FCH patients defined by strict clinical criteria and assessed for LDLR defects with up-to-date techniques of molecular testing. There are also limitations to our study. One limitation is that we did not perform family studies to examine segregation and genotype–phenotype interactions of the detected LDLR mutations. Another limitation is that the observed frequency of LDLR defects may not be extrapolated to other FCH cohorts showing less severe lipid phenotypes. The present series has an inherent selection bias toward more severe lipid abnormalities and may not be representative of FCH patients within the general population because patients are referred to our lipid clinics mainly by general practitioners because of severe dyslipidemia or refractoriness to hypolipidemic treatment. The frequency of LDLR mutations in FCH patients with milder lipid phenotypes is probably lower than that reported here. An alternative explanation for the relatively high frequency of LDLR mutations in our FCH patients could be that the genetic defects were not pathogenic. However, this is highly improbable for several reasons: most of the detected LDLR defects have been previously reported as causative mutations in FH; some of them generate truncated proteins and/or their deleterious effects have been confirmed “in vitro”; the loss of receptor function has been demonstrated in family studies for many of them; and none of the mutations found have been observed in normolipidemic populations.
Two important conclusions can be derived from this work. First, pending the availability of a genetic or biochemical marker for a reliable diagnosis of FCH, the exclusion of LDLR gene defects is highly recommended for patients with this clinical diagnosis who present with severe elevations of total cholesterol or apoB levels. This exclusion could be particularly important for research purposes and to limit the heterogeneity observed in many biochemical and genetic studies of FCH cohorts. Second, the diagnostic criteria for FH should not exclude patients with combined hyperlipidemia. In this situation, the total cholesterol or apoB level might be a better diagnostic tool than LDL-cholesterol, and values above 335 or 185 mg/dl, respectively, should raise the suspicion of FH and prompt a search for LDLR mutations.
For a supplemental table on the description of the LDLR mutations found in subjects with mixed hyperlipidemia, please see the online version of this article.
Supported by grants from the Spanish Ministry of Health FISS PI05/0247, PI06/0365, PI06/1402, and RTIC C06/01 (RECAVA), and the Spanish Ministry of Education and Science (SAF2005-07042). Dr. Tejedor is employed by Progenika Biopharma, the company that commercializes the microarray (Lipochip) for genetic diagnosis of familial hypercholesterolemia in Spain.
- Abbreviations and Acronyms
- confidence interval
- familial combined hyperlipidemia
- familial hypercholesterolemia
- low-density lipoprotein
- low-density lipoprotein receptor gene
- receiver-operator characteristic
- Received January 9, 2008.
- Revision received May 6, 2008.
- Accepted June 2, 2008.
- American College of Cardiology Foundation
- Goldstein J.L.,
- Hobbs H.H.,
- Brown M.S.
- Defesche J.
- Veerkamp M.J.,
- de Graaf J.,
- Hendriks J.C.,
- Demacker P.N.,
- Stalenhoef A.F.
- Civeira F.,
- Castillo S.,
- Alonso R.,
- et al.
- Junyent M.,
- Gilabert R.,
- Zambón D.,
- et al.
- Tejedor D.,
- Castillo S.,
- Mozas P.,
- et al.
- Tejedor D.,
- Castillo S.,
- Mozas P.,
- et al.
- Heath K.,
- Day I.N.,
- Humpries S.E.
- Russell D.W.,
- Brown M.S.,
- Goldstein J.L.
- Wittekoek M.E.,
- Moll E.,
- Pimstone S.N.,
- et al.