Familial hypercholesterolemia (FH) is a common autosomal,

Transcription

1 Clinical Expression of Familial Hypercholesterolemia in Clusters of Mutations of the LDL Receptor Gene That Cause a Receptor-Defective or Receptor-Negative Phenotype S. Bertolini, A. Cantafora, M. Averna, C. Cortese, C. Motti, S. Martini, G. Pes, A. Postiglione, C. Stefanutti, I. Blotta, L. Pisciotta, M. Rolleri, S. Langheim, M. Ghisellini, I. Rabbone, S. Calandra Abstract Seventy-one mutations of the low density lipoprotein (LDL) receptor gene were identified in 282 unrelated Italian familial hypercholesterolemia (FH) heterozygotes. By extending genotype analysis to families of the index cases, we identified 12 mutation clusters and localized them in specific areas of Italy. To evaluate the impact of these mutations on the clinical expression of FH, the clusters were separated into 2 groups: receptor-defective and receptor-negative, according to the LDL receptor defect caused by each mutation. These 2 groups were comparable in terms of the patients age, sex distribution, body mass index, arterial hypertension, and smoking status. In receptor-negative subjects, LDL cholesterol was higher ( 18%) and high density lipoprotein cholesterol lower ( 5%) than the values found in receptor-defective subjects. The prevalence of tendon xanthomas and coronary artery disease (CAD) was 2-fold higher in receptor-negative subjects. In patients 30 years of age in both groups, the presence of CAD was related to age, arterial hypertension, previous smoking, and LDL cholesterol level. Independent contributors to CAD in the receptor-defective subjects were male sex, arterial hypertension, and LDL cholesterol level; in the receptor-negative subjects, the first 2 variables were strong predictors of CAD, whereas the LDL cholesterol level had a lower impact than in receptor-defective subjects. Overall, in receptor-negative subjects, the risk of CAD was 2.6-fold that of receptordefective subjects. Wide interindividual variability in LDL cholesterol levels was found in each cluster. Apolipoprotein E genotype analysis showed a lowering effect of the 2 allele and a raising effect of the 4 allele on the LDL cholesterol level in both groups; however, the apolipoprotein E genotype accounted for only 4% of the variation in LDL cholesterol. Haplotype analysis showed that all families of the major clusters shared the same intragenic haplotype cosegregating with the mutation, thus suggesting the presence of common ancestors. (Arterioscler Thromb Vasc Biol. 2000;20:e41-e52.) Key Words: LDL receptor gene mutations familial hypercholesterolemia mutation clusters receptor phenotype and clinical expression Familial hypercholesterolemia (FH) is a common autosomal, codominant disease caused by mutations in the LDL receptor (LDL-R) gene. It is characterized by elevated plasma LDL levels, tendon xanthomas, premature coronary artery disease (CAD), and a family history of 1 or more of these. 1 In most populations, FH has proved to be extremely heterogeneous at the DNA level: 600 mutations of the LDL-R gene have been reported so far 1 3 (also see fh). Studies conducted in the rare patients with homozygous FH have shown that different types of mutations result in residual functional LDL-R activity that varies from 2% to 30%. 2,4 This mutational heterogeneity explains, to some extent, the phenotypic variation found in FH homozygotes, in whom there is a strong correlation between residual receptor activity in cultured cells and the severity of the disease. 1,2,4 A similar genotype-phenotype correlation is less clear in FH heterozygotes (with documented mutations in the LDL-R gene), because there is only a weak correlation between maximally induced LDL-R activity in cultured cells and the plasma LDL cholesterol (LDL-C) concentration. In these patients, clinical expression of FH is highly variable in terms of the severity of hypercholesterolemia and the age of onset and severity of CAD, as well as the response to diet and lipid-lowering therapy. 5 A few studies on the genotype-phenotype correlation in heterozygous FH have been carried out in isolated populations (such as the French-Canadian), Received December 17, 1999; revision accepted March 17, From the Department of Internal Medicine (S.B., L.P., M.R., S.L.), University of Genoa; the National Institute of Health (A.C., I.B.), Rome; the Institute of Internal Medicine (M.A.), University of Palermo, Palermo; the Department of Internal Medicine (C.C., C.M.), University of Rome Tor Vergata, Rome; the Department of Medical Sciences (S.M.), University of Padua, Padua; the Institute of Clinical Biochemistry (G.P.), University of Sassari, Sassari; the Department of Clinical and Experimental Medicine (A.P.), University of Naples Federico II, Naples; the Institute of Medical Therapy (C.S.), University of Rome, Rome; the Department of Biomedical Sciences (M.G., S.C.), University of Modena and Reggio Emilia, Modena; and the Department of Pediatrics (I.R.), University of Turin, Turin, Italy. Correspondence to Sebastiano Calandra, Department of Biomedical Sciences, University of Modena and Reggio Emilia, Via Campi, 287, Modena,-Italy. sebcal@unimo.it 2000 American Heart Association, Inc. Arterioscler Thromb Vasc Biol. is available at 1

2 2 Arterioscler Thromb Vasc Biol. September 2000 TABLE I. Selection Criteria for FH Families Selection Criteria No. of Families Collected No. of Families With Identified Mutation (% of Collected) No. of Heterozygotes With Identified Mutation No. of Heterozygotes per Family Homozygous FH phenotype (95.8%) Tendon xanthomatosis 215* 120 (55.8%) Hypercholesterolemic children 208* 58 (27.9%) Probable FH 230* 35 (15.2%) Total (39.1%) *Screening in progress. wherein a founder effect had caused the accumulation of many patients with a limited number of mutant alleles of the LDL-R gene. 6 Although these studies have provided some clue about the effect of LDL-R gene mutations on the clinical expression of FH, their results should be treated with caution, as it is likely that such patients might have other genes and environmental factors in common that might contribute to their phenotype. The marked heterogeneity of mutations of the LDL-R gene causing FH, which is present in most populations, makes it difficult to collect a sufficiently large number of patients with the same mutation and to compare the FH phenotype in groups of patients (of the same population) carrying different mutations at the LDL-R locus. Recently, Graham et al 7 have investigated genotype-phenotype correlations in heterozygous FH carrying different mutations of the LDL-R gene. They showed that frameshift (FS) mutations were associated with higher levels of LDL-C than those found in missense mutations. In an attempt to determine to what extent the type of genetic defect is an important determinant of FH phenotype, we have investigated the clinical expression of the disease in several groups (clusters) of FH heterozygotes with defined mutations of the LDL-R gene that have been identified in Italy over the last few years. Methods Patients The study included a group of 725 unrelated patients (index cases) with the clinical diagnosis of heterozygous FH. Four hundred ninety-five subjects had a diagnosis of definite FH according to the following criteria: (1) untreated LDL-C levels 95th percentile of the distribution in the Italian population, stratified for sex and age; (2) vertical transmission of hypercholesterolemia and a bimodal distribution of plasma LDL-C levels in the proband s family; and (3) the presence of subjects with the FH homozygous phenotype in the family (n 72), tendon xanthomatosis in the proband or in at least 1 first-degree relative (n 215), or severe hypercholesterolemia in some prepuberal children of the proband s family (n 208). Two hundred thirty subjects were classified as probable FH, since either no xanthomas were detected in the probands or in their family members or no sufficient family data were available (Table I). Data on smoking habits, arterial hypertension, diabetes, or other associated diseases with relevant impact on cardiovascular risk were collected. FH patients were considered to have coronary artery disease (CAD ) if they (1) had history of a documented myocardial infarction, coronary artery bypass graft, or percutaneous transluminal coronary angioplasty; (2) suffered from angina pectoris with a positive exercise ECG and thallium test or an abnormal angiogram; and (3) had silent ischemia that was detected during bicycle exercise testing and was confirmed by thallium test, stress echocardiography, or angiography. FH subjects were considered free of CAD (CAD ) if they had a negative bicycle exercise test. CAD occurring before 60 years of age was considered premature CAD. The families belonging to the FH clusters (see Results) were considered positive for the presence of tendon xanthomas and/or premature CAD if at least 1 adult member of the family was affected. Informed consent was obtained from all subjects investigated, and, in the case of children, from their parents. The study protocol was approved by the human investigation committee of each participating institution. Biochemical Analysis Fasting plasma lipid concentrations were measured before any hypolipidemic drug treatment. Total cholesterol, triglycerides, and HDL-C levels were measured enzymatically with commercial kits (Boehringer Mannheim GmbH) and an automated analyzer (Hitachi model 704); LDL-C was calculated by the Friedewald formula. Assay of LDL-R Activity in Skin Fibroblasts The assay of 125 I-labeled LDL binding, internalization, and degradation in cultured skin fibroblasts was performed as described by Goldstein et al. 8 Residual LDL-R activity (calculated as the maximum rate of saturable 125 I-LDL degradation) was expressed as a percentage of the value obtained in control fibroblasts. Analysis of the LDL-R Gene Genomic DNA was extracted from peripheral blood leukocytes by a standard procedure. 9 All DNA samples were digested by using 5 to 10 U/ g of several restriction enzymes, separated by agarose gel electrophoresis, transferred to nylon membranes, and hybridized with various LDL-R cdna probes. 10 Polymerase chain reaction (PCR) amplifications of the promoter region and exons 1 to 18 of the LDL-R gene were carried out by using the primers reported by Hobbs et al 2 in a total volume of 50 L containing 100 ng of genomic DNA, 10 mmol/l Tris-HCl, 50 mmol/l KCl, 1.5 mmol/l MgCl 2, 100 mol of each of the 4 nucleotides, 10 pmol of each primer, and 2 U of Taq DNA polymerase. Single-strand conformation polymorphism analysis was performed by using a vertical gel unit; 5 L of PCR product was mixed with an equal amount of 95% formamide, 20 mmol/l EDTA, 0.1% bromophenol blue, and 0.1% xylene cyanole; denatured at 96 C for 5 minutes; snap-cooled in 4 C ice water; and then loaded onto an 8% polyacrylamide gel containing 5% glycerol. Electrophoresis was performed at room temperature in a standard Tris-borate-EDTA buffer, ph 8.0, at 150 V for 2 hours. After electrophoresis, the gels were stained with silver. The samples showing an abnormal single-strand conformation polymorphism pattern were sequenced by using an automated fluorescence ABI Prism 310 genetic analyzer (Applied Biosystems Inc) and following the manufacturer s recommendations. Mutations identified by the automated sequencer were confirmed by manual sequencing 4 with the use of appropriate primers. Several mutations reported in Table II have been characterized previously in our laboratory. 4,10 23 The screening of some mutations in the probands family members was performed by using a variety of procedures: (1) Southern blotting (deletion of exons 2 to 12 and exons 13 to 14) 12,16 ; (2) PCR amplification of fragments of the LDL-R gene with appropriate canonic primers, followed by digestion with restriction enzymes (MspI for D200G, NlaIII for V502M, HinfI for D558Y, and PstI for

3 Bertolini et al Clusters of LDL-R Gene Mutations in Italy 3 TABLE II. Inventory of LDL-R Gene Mutations Causing FH in Italy Mutation Fam Ho CHe He Receptor Protein LDL-C, mmol/l (n) Allele Designation Ref. No. Exon (Ex) deletions (kb) Promoter Ex 1, 2 (20) 1 12 Not produced (8) FH Siracusa 18 Promoter Ex 1 6 ( 25) 2 4 Not produced (4) FH Bologna-1 10 Promoter Ex 1 6 (24) 1 3 Not produced (3) FH Bari-3 New Ex 2 12 (24) Not produced (43) FH Pavia 12 Ex 3 10 (12) 1 6 (1) Truncated (6) FH Massa 16 (2) Inframe deletion Ex 11, Truncated (6) FH Genoa-1 16 Ex 13, 14 (3.7) 5 21 Truncated (16) FH Chieti-1 16 Ex 13, 14 (4.7) 1 2 Truncated (2) FH Rome 16 Ex (4.5) 1 4 (1) Truncated (4) FH Potenza-1 13 (2) Inframe deletion Ex (5) Truncated (4) FH Palermo-3 4 Ex ( 25) Truncated (3) FH Catania 4 Ex 16 (2) 1 4 Inframe deletion (4) FH Padua-2 16 Exon (Ex) duplications (kb) Ex 3 6 (8) 1 9 Inframe insertion (9) FH Caltanissetta 22 Ex 9 14 (10) 2 3 Inframe insertion (3) FH Salerno 15 Ex (7) 1 3 Truncated (3) FH Bologna-2 11 Ex 16, 17 (5.5) Inframe insertion (4) FH Viterbo 10 Nonsense (cdna) 97C3T (Ex 2) Q12X (2) FH Milan C3T (Ex 2) Q14X (2) FH Milan G3T (Ex 3) 1 2 E80X (2) FH Imperia WS* 304C3T (Ex 3) 3 12 Q81X (8) FH Rapone WS* 373C3T (Ex 4) Q104X (3) FH Foggia G3T (Ex 4) 2 6 E119X (4) FH Campobasso WS* 865G3T (Ex 6) E267X (8) FH Milan C3T (Ex 15) 2 8 Q749X (8) FH Mondovì New Missense (cdna) 28T3A (Ex 1) 1 1 W-12R 7.94 FH Enna T3C (Ex 3) C68R (8) FH Catanzaro 4 346T3C (Ex 4) 1 6 C95R (6) FH Alghero G3T (Ex 4) D97Y (4) FH Naples G3T (Ex 4) 1 1 C113F 8.45 FH Alessandria New 402C3G (Ex 4) 1 5 C113W (5) FH Lucca New 415G3A (Ex 4) 2 4 D118N (4) FH Sassari A3G (Ex 4) D200G (95) FH Padua G3A (Ex 4) 4 10 C201Y (10) FH Genoa-2 New 682G3A (Ex 4) 7 16 E207K (14) FH Modena WS* 761A3C (Ex 5) 1 2 Q233P (2) FH Reggio Emilia-2 New 826T3C (Ex 6) 3 6 C255R (6) FH Ceva New 828C3G (Ex 6) 1 3 C255W (3) FH Sassari G3A (Ex 6) 1 1 E256K 6.34 FH Genoa-3 WS* 952T3C (Ex 7) 1 3 C297R (3) FH Perugia New 953G3T (Ex 7) 3 16 C297F (12) FH Trieste C3G (Ex 7) C331W (7) FH Avellino T3C (Ex 8) 2 7 C343R (7) FH Palermo-2 WS* 1109A3C (Ex 8) N349T (4) FH Reggio Calabria New 1118G3A (Ex 8) G352D (8) FH Potenza T3C (Ex 8) C358R (8) FH Naples-1 4

4 4 Arterioscler Thromb Vasc Biol. September 2000 TABLE II. Continued Mutation Fam Ho CHe He Receptor Protein LDL-C, mmol/l (n) Allele Designation Ref. No. 1195G3A (Ex 9) A378T (3) FH Nuoro A3G (Ex 9) 1 5 D394G (5) FH Padua-3 New 1301C3G (Ex 9) 1 1 T413R FH Sassari A3G (Ex 10) H464R (2) FH Milan T3A (Ex 10) 1 3 I467N (3) FH Piacenza WS* 1567G3A (Ex 10) V502M (35) FH Bari G3A (Ex 11) G528D (66) FH Palermo-1 4, G3T (Ex 12) D558Y (31) FH Casale Monferrato New 1775G3A (Ex 12) G571E (14) FH Foggia C3T (Ex 14) P664L (28) FH Frosinone-1 4 Frameshift (FS), deletion (del), insertion (ins) (cdna) del (Ex 4) 1 2 FS187Term (2) FH Asti New del, ins 6 (Ex 4) 1 8 E207del, CK ins (8) FH Chieti ins 18 (Ex 4) 1 9 CKDKSD ins (9) FH Padua-4 WS* 1067delA (Ex 8) 1 4 FS335Term (2) FH Cuneo New 1068T3A, ins 18 (Ex 8) 1 3 D335E, ECQDPD ins (3) FH Chieti insTGGC (Ex 8) 1 1 FS354Term FH Pisa delAG (Ex 10) 1 2 FS437Term (2) FH Ancona WS* insACAT (Ex 10) FS453Term (101) FH Savona WS* del (Ex 10) 1 10 FS459Term (10) FH Reggio Emilia delCT (Ex 10) FS472Term (4) FH Frosinone delG (Ex 12) 8 40 FS572Term (27) FH Sassari-1 23 Splicing IVS g3a 7 20 Inframe deletion (17) FH Olbia 23 IVS g3a (1) Inframe deletion (4) FH Agrigento 4 (2) Inframe insertion IVS g3a (1) Truncated (3) FH Avellino-2 4 (2) Inframe deletion IVS g3a (1) Truncated (3) FH Benevento 4 (2) Inframe insertion IVS g3a Inframe deletion 7.33 FH Monfalcone 4 Fam indicates number of families; Ho, homozygotes; CHe, compound heterozygotes; He, heterozygotes; Term, termination; and WS, web site LDL-C levels (adjusted for sex, age, and body mass index) are reported only for heterozygous patients whose untreated lipid values were available. *Not described before in Italy. P664L); (3) PCR amplification of fragments of the LDL-R gene with appropriate mismatched primers, followed by digestion with restriction enzymes (NdeI for g3a, StyI for E207K, RsaI for FS453, and StuI for G571E); (4) PCR amplification of a specific allele (PASA) for G528D 20 ; and (5) heteroduplex analysis for FS Details of the procedures adopted for screening these mutations will be reported elsewhere. Geographic Origin of Mutations The putative geographic origin of each LDL-R gene mutation was empirically established on the basis that individuals carrying a specific mutation were proved to have stemmed from ancestors who had been living in the same geographic area for at least 3 generations. Haplotype Analysis of the LDL-R Locus The intragenic haplotypes cosegregating with some mutant alleles (D200G, FS453, and G528D) were constructed by using diallelic markers present in the LDL-R locus. Genotyping was performed by (1) Southern blotting (BstEII, 5 ApaLI, PvuII, NcoI, and 3 ApaLI); (2) restriction enzyme digestion of PCR-amplified intron 7 (SmaI) and exons 8, 12, 13, and 15 (StuI, HincII, AvaII, and 5 MspI); and (3) single-strand conformation polymorphism analysis (1413G/A in exon 10) The polymorphic tetranucleotide microsatellite D19S394 that is 250 kb telomeric to the LDL-R gene was genotyped by using fluorescently tagged, PCR-generated fragments 29 that were electrophoresed in an Applied Biosystems automated DNA sequencer 373 A Leon and sized by Genescan 672 software (Applied Biosystems). Screening for Familial Defective Apo B-100 All probands were screened for the presence of the R3500Q mutation in the apolipoprotein (apo) B gene causing familial defective apo B Apo E Genotyping The apo E genotype was determined by PCR amplification of genomic DNA according to the procedure of Hixson and Vernier. 31

5 Bertolini et al Clusters of LDL-R Gene Mutations in Italy 5 Statistical Analysis Statistical analysis was performed by using the SPSS 9.0 (SPSS Inc) program. Statistically significant differences between groups or among groups for continuous variables were evaluated by Student s t test for unpaired data or ANOVA, respectively. Triglyceride values, which were not normally distributed, were logarithmically transformed before analysis. Multiple comparisons among pairs of means based on unequal sample sizes were performed by the Dunnett C method. Lipid values were adjusted for sex, age, and body mass index (BMI) by linear multiple regression analysis. When subjects were grouped by apo E genotypes, their lipid values were also adjusted separately in the receptor-defective group and the receptornegative group for the effect of each LDL-R gene mutation (ie, the lipid values for each mutation were adjusted to the grand mean of the whole group). Differences in the distributions of categorical variables were assessed by the 2 test with Yates continuity correction when necessary. Genotype distribution for the apo E polymorphism was determined by gene counting. 2 analysis was used to test for Hardy-Weinberg equilibrium and to compare the observed genotype distributions and allele numbers in CAD and CAD subjects. The proportion of the total phenotypic variability of adjusted LDL-C and HDL-C values attributable to the apo E gene polymorphism was calculated according to Boerwinkle et al. 32 The independent contribution of each variable to CAD was evaluated by logistic regression analysis, taking into account only those variables that were found to be predictive in univariate analyses. The independent categorical variables were coded as follows: 1 for male and 0 for female sex; 0 and 1 for the absence and presence, respectively, of arterial hypertension; 0, 1, 2, and 3 for nonsmoker, former smoker, current light smoker ( 10 cigarettes per day), and current heavy smoker ( 10 cigarettes per day), respectively; 0 for mutations causing a receptordefective phenotype and 1 for mutations causing a receptor-negative phenotype; 1 for genotypes 2 2 and 2 3; 2 for genotype 3 3; and 3 for genotypes 2 4, 3 4, and 4 4. Age and lipid values entered the analysis as continuous variables. Likelihood ratio statistics were used to compare models with different combinations of predictors with an inclusion significance level of 0.05 and an exclusion significance level of 0.1. Odds ratios and 95% confidence intervals are also reported. Results Mutations of the LDL-R Gene in FH Patients Table I shows the number of families (unrelated index cases) and subjects with the clinical diagnosis of definite or probable FH, in whom an LDL-R gene mutation was detected during a national survey that is still in progress. At this time, analysis of the LDL-R gene mutation has been completed in all families of patients with homozygous FH, in 42% of families with definite FH (presence of tendon xanthomas or hypercholesterolemic children), and in 15% of families with probable FH. By extending the genetic analysis to the relatives of index cases, we have been able to collect 1000 FH heterozygotes with identified mutations in the LDL-R gene. Table II shows the mutations of the LDL-R gene found in our index patients to date, the number of apparently unrelated families carrying the same mutation, and the number of subjects (heterozygotes, homozygotes, and compound heterozygotes) belonging to these families. The allele designation indicates the putative geographic origin of each mutation in our country. The table includes mutations that (1) were discovered in Italian FH patients and were reported previously by our group (see References 4 and 10 through 23 and the web site (2) were discovered in other countries and reported in the literature but also found to be present in Italy (see the web site); and (3) are reported for the first time in this article (new). During this survey, we Figure I. I Geographic location of the 3 major clusters of LDL-R gene mutations found in Italy. The number beside the G528D symbol in Sicily indicates the number of families with this mutation identified on the island. discovered that some mutations are overrepresented in specific geographic districts of the country. For example, D200G (FH Padua-1), though detected in all districts of northern Italy, is found mostly in the Veneto region and has so far not been detected in central and southern Italy. On the other hand, G528D (FH Palermo-1) is found in southern Italy and Sicily but not in northern Italy (Figure I). The identification of several apparently unrelated individuals with the same mutation allowed us to define the presence of clusters of mutations and to compare the FH phenotype among these clusters. In the present study, we arbitrarily defined a cluster as any mutation group that included 5 or more apparently unrelated families. At present, we have identified 12 such clusters (Table III). Haplotype Analysis Haplotype analysis was carried out in the largest clusters (D200G and FS453 found in northern Italy and G528D found in southern Italy and Sicily). In each cluster, the haplotype of intragenic markers cosegregating with the mutation was the same in all families, thus suggesting a common ancestor. D200G and FS453 cosegregated with the following haplotypes: SmaI ( ), StuI ( ), 1413G/A (G), HincII ( ), BstEII ( ), AvaII ( ), MspI 5 ( ), ApaLI 5 ( ), PvuII ( ), NcoI ( ), and ApaLI 3 ( ). G528D cosegregated with the following haplotype: StuI ( ), HincII ( ), BstEII ( ), AvaII ( ), ApaLI 5 ( ), PvuII ( ), NcoI ( ), and ApaLI 3 ( ).Analysis of the microsatellite D19S394, located 250 kb upstream from the LDL-R locus, showed that the FS453 mutation cosegregated in all families with a single allele (251 bp); the D200G mutation cosegregated with 3 alleles (227, 235, and 239 bp with a frequency of 0.18, 0.41, and 0.41, respectively); and the G528D mutation cosegregated with 4 alleles (243, 255, 259, and 263 bp with a frequency of 0.07, 0.41, 0.25, and 0.27, respectively). In 256 normal chromosomes, the allele frequencies were as follows: 227 (0.05), 235 (0.10), 239 (0.05), 243 (0.09), 251 (0.10), 255 (0.12), 259 (0.12), and 263 (0.10). FH Phenotype in Receptor-Defective and Receptor-Negative Clusters Table III shows the main clinical features of the clusters. Lipid values reported in the tables were adjusted for age, sex, and

6 6 Arterioscler Thromb Vasc Biol. September 2000 TABLE III. Clinical Features and Lipid Values (mmol/l) in FH Clusters Families, Families With No. of LDL-C HDL-C, TG, (% Screened) TX PCAD Subjects Mean SD Range Mean SD Mean SD Receptor-defective G571E 7 (2.9%) V502M 13 (4.8%) P664L 10 (4.4%) g3a 7 (8.7%) D200G 48 (14.9%) Total Receptor-negative Ex (3.0%) Fs (18.0%) G528D 29 (17.8%) D558Y 10 (6.7%) Fs572 8 (10.0%) E207K 7 (2.8%) Ex 13, 14 5 (1.7%) Total * TX indicates tendon xanthomas; PCAD, premature coronary artery disease; LDL-C, LDL cholesterol; HDL-C, HDL cholesterol; TG, triglycerides; Ex, exon; and FS, frameshift. The LDL receptor activity in fibroblasts of FH heterozygotes carrying Fs453, D558Y, and E207K were 44 48%, 42 46%, and 40 45%, respectively, of the values found in control fibroblasts. Lipid values were adjusted for sex, age, and body mass index. *Receptor-negative vs receptor-defective , P Receptor-negative vs receptor-defective , P BMI. We separated the clusters into 2 groups designated receptor-defective and receptor-negative, respectively, on the basis of (1) the residual LDL-R activity found by us and/or others in cultured cells (fibroblasts) of homozygous subjects ( 5% for receptor-defective, 5% for receptor-negative) 1,2,4 or heterozygous subjects ( 55% for receptor-defective, 55% for receptor-negative) carrying the same mutation and (2) the presence of a mutation leading to a frameshift and a truncated receptor (null allele). One of the striking features emerging from the analysis of these clusters is the large variability of plasma LDL-C levels observed in each cluster, regardless of the type of mutation of the LDL-R gene. The mean difference between the lowest and the highest LDL-C values was mmol/l in the receptor-defective group and mmol/l in the receptor-negative group. Table III also shows that in the receptor-negative group, the prevalence of tendon xanthomatosis in the families (defined as the number of families in which at least 1 living member had tendon xanthomatosis) and the prevalence of premature CAD (defined as the number of families in which premature CAD had been detected in at least 1 living member or had been reported as the cause of death in at least 1 deceased family member) were respectively, 2.7- and 2.1-fold those found in the receptor-defective group. Table IV shows the comparison of the clinical features and lipid values between index cases (1 index case for each unrelated family included in the major clusters shown in Table III) and between all heterozygous subjects. Index cases of the 2 groups were comparable in terms of sex distribution, mean age, BMI, smoking habits, and arterial hypertension. The unadjusted mean plasma levels of total cholesterol and LDL-C were significantly higher in the receptor-negative subjects, whereas plasma triglyceride and HDL-C levels were similar in the 2 groups. When the data were adjusted for age, sex, and BMI, plasma HDL-C levels were found to be slightly but significantly lower in the receptor-negative subjects. The percentage of index patients with tendon xanthomas and CAD was higher in the receptor-negative subjects. Considering all heterozygous subjects belonging to the clusters, the 2 groups were comparable in terms of mean age, sex distribution, arterial hypertension, and percentage of current and former smokers. BMI was found to be slightly higher in the receptor-defective group. Diabetic patients (1 patient in the receptor-defective group and 2 patients in the receptor-negative group) were not included. The differences in the unadjusted and adjusted total cholesterol, LDL-C, and HDL-C levels observed in the index patients were clearly confirmed when we compared all subjects belonging to the 2 groups (Table IV). In receptornegative subjects, the adjusted LDL-C and HDL-C levels were 18% higher and 5% lower, respectively, than those found in receptor-defective subjects. Figure II shows the age-related increase of LDL-C level in both groups; at each age interval, the mean LDL-C value was higher in the receptor-negative subjects. The differences in the prevalence of tendon xanthomas and CAD in the 2 groups were similar to those found in index cases (Table IV). In both receptor-defective and receptor-negative subjects, the genotype distributions of the apo E polymorphism were in Hardy-Weinberg equilibrium (receptor-defective: 2 3 n 18, 2 4 n 2, 3 3 n 137, 3 4 n-31, and 4 4 n 1; receptor-negative: 2 2 n 1, 2 3 n 24, 2 4 n 1, 3 3 n 230, 3 4 n 40, and 4 4 n 2), and no significant differences in allele frequencies were observed between the 2 groups (receptor-defective: , , and ; receptor-negative: , , and ). In both groups, plasma lipid values adjusted for sex, age, BMI, and

7 Bertolini et al Clusters of LDL-R Gene Mutations in Italy 7 TABLE IV. Clinical and Biochemical Features of the Index Cases and All Heterozygotes in FH Clusters the effect of each LDL-R gene mutation were found to be influenced by the apo E polymorphism. In comparison with subjects carrying the 3 3 genotype, in the receptor-defective group the 2 3 genotype was associated with lower HDL-C levels ( versus mmol/l, P 0.05) and with a tendency to higher triglyceride levels ( versus mmol/l), whereas in the receptor-negative group, 2 3 was associated with lower LDL-C levels ( versus mmol/l, P 0.05). In both groups, subjects carrying the 4 allele ( 2 4, 3 4, or 4 4 genotypes) had higher LDL-C levels than did subjects with other genotypes (receptor-defective: versus Receptor-Defective Receptor-Negative P M/F, n/n 39/46 45/55 NS 84/ /172 NS Age, y NS NS Body mass index, kg/m NS Tendon xanthomas, % 24 (28.2%) 66 (66.0%) (18.0%) 122 (40.9%) Arterial hypertension, % 13 (15.3%) 17 (17.0%) NS 20 (10.6%) 40 (13.4%) NS Current smokers, % 11 (13.0%) 15 (15.0%) NS 25 (13.2%) 29 (9.7%) NS Former smokers, % 15 (17.6%) 17 (17.0%) NS 17 (9.0%) 22 (7.4%) NS Coronary artery disease, % 17 (20.0%) 34 (34.0%) (11.1%) 59 (19.8%) 0.01 Total cholesterol, mmol/l ( ) ( ) ( 0.001) ( ) ( ) ( 0.001) LDL cholesterol, mmol/l ( ) ( ) ( 0.001) ( ) ( ) ( 0.001) HDL cholesterol, mmol/l ( ) ( ) NS ( 0.05) ( ) ( ) NS ( 0.05) Triglycerides, mmol/l ( ) ( ) NS (NS) ( ) ( ) NS (NS) Values are mean SD. Values in italics refer to all heterozygous subjects belonging to FH clusters. Lipid values shown in parentheses were adjusted for sex, age, and body mass index mmol/l in the 3 3 group and mmol/l in the 2 3 group, P 0.05; receptor-negative: versus mmol/l in the 3 3 group and mmol/l in the group, P 0.05). CAD in Receptor-Defective and Receptor-Negative Clusters Because it is well established that the prevalence of tendon xanthomas and CAD is affected by age, we looked at the distributions of these clinical features in subjects younger or older than 30 years of age. In subjects 30 years of age (71 in the receptor-defective group and 112 in the receptor- Figure II. Age-related increase of mean plasma LDL cholesterol in receptor-negative and receptordefective clusters.

8 8 Arterioscler Thromb Vasc Biol. September 2000 TABLE V. Comparison Between Subjects >30 Years of Age With ( ) or Without ( ) Coronary Artery Disease (CAD) Receptor-Defective Receptor-Negative CAD ( ) CAD ( ) P CAD ( ) CAD ( ) P M/F, n/n 42/55 12/9 NS 45/82 33/ Age, y Body mass index, kg/m NS Tendon xanthomas, % 23 (23.7%) 9 (42.9%) NS 70 (55.1%) 49 (83.1%) Arterial hypertension, % 12 (12.4%) 8 (38.1%) (13.4%) 23 (39.0%) Current smokers, % 18 (18.7%) 3 (14.3%) NS 20 (15.7%) 7 (11.9%) NS Former smokers, % 8 (8.2%) 9 (42.9%) (4.7%) 16 (27.1%) Total chol, mmol/l * LDL chol, mmol/l * HDL chol, mmol/l NS NS Triglycerides, mmol/l NS Chol indicates cholesterol. *Significantly different from receptor-defective CAD ( ) subjects. negative group), no individual was detected with CAD in either group, and only a few (2.8% in each group) were found to have tendon xanthomas. When subjects 30 years were compared (118 in the receptor-defective group and 186 in the receptor-negative group), the prevalence of tendon xanthomas and CAD in receptor-negative subjects was 2.4-fold (P 0.001) and 1.8-fold (P 0.01), respectively, greater than that observed in receptor-defective subjects. Table V compares the clinical features and lipid values in receptor-defective and receptor-negative subjects 30 years of age who were CAD( ) or CAD( ). In both groups, age, the prevalence of arterial hypertension, previous smoking habits, and total cholesterol and LDL-C levels were higher in CAD( ) patients. In the receptor-negative group, the maleto-female ratio and BMI were significantly higher in CAD( ) patients, whereas in the receptor-defective group, plasma triglycerides were higher in CAD( ) patients. The prevalence of premature CAD (occurring before 60 years of age in subjects 30) was significantly higher in receptornegative patients (23.2% versus 10.6%; , P 0.03), with a crude odds ratio (95% confidence interval) of 2.55 (1.10 to 6.41). The CAD( ) distribution according to quartiles of age was examined in subjects 30 years of age. In the first and second quartile, the prevalence of CAD was similar in the receptordefective and receptor-negative groups; in the fourth and especially the third quartile, the prevalence of CAD was much higher in the receptor-negative subjects (fourth quartile, 56.3% versus 33.3%, P 0.05; third quartile, 41.7% versus 16.1%, P 0.02). Multiple logistic regression analysis, performed separately in the 2 groups in subjects 30 years of age, showed that independent contributors to CAD were (1) male sex (P 0.02), arterial hypertension (P 0.02), and LDL-C level (P 0.007) in the receptor-defective group and (2) male sex (P 0.001), age (P ), and arterial hypertension (P 0.02) in the receptornegative group. This analysis also showed that in the receptordefective group, age had a marginal effect on CAD (P 0.06), and in the receptor-negative group, the LDL-C level did not appear to be a good predictor for the development of CAD (P 0.08). In both groups, the apo E genotype did not significantly contribute to either premature CAD or total CAD. Table VI shows the results of multiple logistic regression analysis performed in both groups taken together to compare the risk of CAD associated with receptor-negative versus receptor-defective mutations. This analysis showed that in subjects 30 years of age with receptor-negative mutations, the risk of CAD was 2.6-fold higher than that observed in subjects with receptor-defective mutations. Discussion Mutation Spectrum and Clusters In this study, we report partial characterization of the spectrum of mutations of the LDL-R gene discovered in Italy during a survey that started a few years ago and is still in progress. So far, our inventory consists of 71 mutations and includes 16 major rearrangements; 8 nonsense, 31 missense, and 11 frameshift mutations; and 5 splicing defects. During this study, we found that 4% of subjects with clinically severe FH either do not carry detectable mutations in the LDL-R gene (proximal promoter, coding sequence, and exon/intron junctions) or show a phenotype that does not TABLE VI. Parameters Associated With Coronary Artery Disease in FH Patients >30 Years of Age With Receptor-Defective and Receptor-Negative Mutations by Multiple Logistic Regression Analysis Parameter SE ( ) OR* 95% CI (OR) P Male sex Age Arterial hypertension Receptor-negative vs receptor-defective is the partial regression coefficient; SE( ), standard error of ; and CI, confidence interval. *Odds ratio (OR), defined as e, corresponds to changes in odds for males vs females, for a 1-year increment in age, for patients with arterial hypertension vs those without, and for patients with receptor-negative mutations vs patients with receptor-defective mutations, assuming that the other parameters remain constant at their mean values.

9 Bertolini et al Clusters of LDL-R Gene Mutations in Italy 9 segregate with the LDL-R gene haplotype within the family. This discovery is not surprising, as other studies have reported that a certain proportion of putative FH subjects do not have detectable mutations at the LDL-R locus From the mutation list shown in Table II, the conclusion that emerges is that Italy is like the United Kingdom, 7,33 Germany, 36 and Japan, 37 where many mutations have produced a highly heterogeneous picture, in contrast to the situation found in other countries (like Norway, Finland, and, to some extent, Denmark), where few mutations account for a large proportion of FH patients For each index case, we tried to retrieve as much family data as possible to define the geographic origin of the ancestor likely to carry the mutation identified in the index case. This family history allowed us to locate clusters of mutations in specific areas of the country and to identify several clusters. To have a sufficient number of subjects for the comparison of lipid values and clinical features in patients carrying different mutations, we decided to pool the clusters according to the predicted effect of each mutation on LDL-R function (receptor-defective versus receptor-negative clusters; Table III). This analysis showed that even in the setting of our large allelic heterogeneity, heterozygous FH patients who carry receptor-negative mutations have a more severe atherogenic lipid profile (higher LDL-C and lower HDL-C) than do patients with receptordefective mutations. These results agree with previous observations made in population samples living in other geographic settings. In French-Canadian FH heterozygotes carrying a large deletion ( 15 kb) in the promoter and the first exon of the LDL-R gene (receptor-negative phenotype), the mean serum cholesterol level was much higher than in heterozygous patients with the W66G mutation (receptordefective phenotype). 6 Afrikaner FH patients with a common receptor-defective mutation (D206E) have significantly lower cholesterol levels and milder disease than do Afrikaners carrying a receptor-negative mutation (V408M). 41 Vohl et al 42 compared FH heterozygotes carrying null alleles, FH heterozygotes carrying defective alleles, and non-fh subjects. They found that carriers of null alleles had the highest plasma total cholesterol and LDL-C levels and the highest total cholesterol to HDL-C ratio, whereas the carriers of defective alleles had intermediate levels between null-allele carriers and non-fh patients. Similar results, reported in other studies, 7,43,44 suggest that in FH heterozygotes (as previously documented in FH homozygotes), 1,4,6 the difference in the magnitude of increase in plasma cholesterol and LDL-C levels largely derives from differences in the nature of the mutation in the LDL-R gene. Prevalence of CAD In index cases as well as in all heterozygous FH patients, the prevalence of CAD in receptor-negative subjects (of both sexes) was 1.7-fold higher than that found in subjects carrying receptor-defective mutations. Furthermore, when subjects 30 years old were further stratified by age, we found that in receptor-negative patients, CAD occurred at a younger age than in receptor-defective persons. This observation is in keeping with that of Vohl et al, 42 who showed that in French-Canadian FH heterozygotes, the development of CAD as estimated from the age at first coronary angiography or at first revascularization occurred at an earlier age in FH patients carrying a null mutation compared with carriers of defective alleles. Recently, Gaudet et al 44 compared the expression of CAD assessed by coronary angiography in young (25 to 49 years) versus middle-aged (50 to 64 years) male FH heterozygotes with receptor-negative mutations ( 15-kb deletion, Y468X or R329X) and with missense mutations (W66G, E207K, or C646Y); in this latter group, 92% of individuals carried the W66G mutation that produces a receptor-defective phenotype. In both age groups, carriers of receptor-negative mutations had higher plasma cholesterol and LDL-C levels than did carriers of missense mutations, and their relative odds of being affected by CAD before the age of 50 was 3-fold greater than that observed in patients with missense mutations (in whom the receptor-defective phenotype was prevalent). Multiple logistic regression analysis performed in our subjects who were 30 years of age showed that in both receptor-defective and receptor-negative groups, independent contributors to CAD were male sex, age, and arterial hypertension (Table VI). The plasma level of LDL-C was a strong independent contributor to CAD in the receptor-defective group but was a weaker contributor in the receptor-negative group. This latter observation suggests that in our receptornegative subjects, any increase in LDL-C above certain threshold level (eg, 8.0 mmol/l LDL-C) contributes less to the development of CAD than do age, male sex, and arterial hypertension. Taking into consideration all of these factors, the multiple logistic regression analysis showed that in both sexes 30 years of age, the risk of CAD is 2.6-fold higher in patients carrying receptor-negative mutations. In conclusion, compared with receptor-defective mutations, receptor-negative mutations are more strongly associated with premature CAD, this relationship being independent of common risk factors such as sex, age, and hypertension, as well as the other risk factors such as smoking habits, diabetes, and obesity, 45 which were not contributors to CAD in our study. Interindividual Variability of LDL-C Level As previously reported, 6 we found wide variations in plasma lipid and lipoproteins, even in patients carrying the same mutation. Table III shows that the wide interindividual variation in LDL-C level persisted even after adjustment for sex, age, and BMI and was present in all clusters regardless of the type of mutation and its biological effect (receptornegative and receptor-defective). These variations may be due to genetic and/or environmental factors that affect expression of the wild-type allele (in the case of receptornegative mutations) or both alleles (in the case of receptordefective mutations) or influence other pathways of lipoprotein metabolism. Several groups have investigated the role of common polymorphisms of several candidate genes (such as apo E, apo B, and the LDL-R) that are known to be associated with variations in plasma LDL, HDL, and triglycerides in the general population. In light of the point discussed above, the most reliable studies appear to be those conducted on FH patients with identified mutations of the LDL-R gene. Dallongeville et al 46 were the first to show that LDL-C levels were lower in the E3/2 subset than in the E3/3 or E4/3 subset of heterozygous FH subjects with the 15-kb French-Canadian mutation. These observations were ex-

10 10 Arterioscler Thromb Vasc Biol. September 2000 panded by Ferrieres et al, 47 who found that the contribution of the apo E polymorphism was different between the sexes. Women with the 2 3 genotype had lower total cholesterol, LDL-C, and apo B levels than did women with the 3 3 genotype, whereas in men, the 2 2 genotype was associated with higher VLDL-C than was the 3 3 genotype. Overall, 20% of the variability for the various lipid traits among these FH patients was explained by apo E genotype. Surprisingly, these effects of apo E polymorphism were not observed in a large group of FH heterozygotes carrying the North Karelia mutation, 48 although serum LDL-C levels were found to be lower in subjects with the apo 4 2 genotype in comparison with subjects with other genotypes. Betard et al 49 examined whether the phenotypic variation in lipoprotein and apo B levels observed in a sample of French-Canadians carrying the 15-kb deletion was associated with genetic variability of the wild allele (ie, nondeletion LDL-R gene). They found that some haplotypes, defined by using pairs of restriction fragment length polymorphisms of the LDL-R gene, contributed to quantitative variations in HDL-C and LDL-C levels in women but not in men. However, in FH subjects carrying the North Karelia mutation, Vuorio et al 48 failed to find a significant association between PvuII and AvaII polymorphisms of the LDL-R gene or XbaI polymorphism of the apo B gene and plasma lipids and the prevalence of CAD. In the present study, we found that in both receptornegative and receptor-defective groups, apo E genotype influences plasma LDL-C levels. Compared with the 3 3 genotype, 2 2 and 2 3 genotypes have lower LDL-C levels, whereas the opposite was true for genotypes containing the 4 allele. Overall, apo E genotype accounted for 4.4% and 4.2% of the variation in plasma LDL-C level (adjusted for sex, age, BMI, and the effect of each LDL-R gene mutation) in receptor-defective and receptor-negative groups, respectively. Therefore, apo E polymorphism explains only part of the large interindividual variability of plasma lipids and lipoproteins observed in each cluster of our series, thus suggesting the presence of other genetic and environmental factors. The important role played by environmental factors has emerged from a recent study by Pimstone et al, 50 who compared lipid values and the prevalence of CAD in FH Chinese subjects, with defined LDL-R gene mutations, living in China and Canada. Higher LDL-C levels (and a higher prevalence of tendon xanthomas and CAD), not accounted for by an effect of the LDL-R gene mutation, were observed in FH heterozygotes living in Canada than in those living in China. These differences could be ascribed, at least in part, to the striking contrast in dietary fat consumption and level of physical exercise between subjects living in Canada and those living in China. It is reasonable to assume that these factors have a role to play in our series, too. A systematic study of these factors in the main clusters of our series is now in progress. Our preliminary dietary survey, however, has failed to show striking differences in dietary habits (in terms of total and saturated fat consumption) similar in magnitude to that reported in the study of Pimstone et al. 50 A recent comparison (1994 to 1996) of dietary habits between northern Italy (where, according to tradition, the diet is presumed to contain more meat and dairy products) and southern Italy (where a Mediterranean-type diet is presumed to be consumed) has revealed that the differences in total and saturated fat intake are fairly small (35% versus 33% and 13% versus 12%, respectively; A. D Amicis and A. Turrini, National Institute for Food and Nutrition, personal communication, 2000). In this context, we need a very large number of genotyped FH subjects to be able to ascertain the contribution of diet to the interindividual variability of LDL-C level. Common Ancestor The identification of clusters of families carrying the same mutation in specific areas of the country raises the question of the origin of these mutations (common ancestor versus recurrent mutations). We examined the haplotype cosegregating with each mutation in the largest clusters (D200G, FS453, and G528D). D200G has been reported in other European countries in patients with different ethnic backgrounds, suggesting the hypothesis that D200G is a recurrent mutation. However, the observation that in our cluster, all patients share the same intragenic haplotype and the mutation is mostly restricted to the northeastern areas of Italy strongly suggests that D200G is identical by descent in all patients of the cluster (which also includes 4 identified homozygotes). The G528D is a common mutation in southern Italy and Sicily (the cluster includes 3 apparently unrelated homozygotes and 2 related compound heterozygotes) but has been reported in other European countries, too. Interestingly, G528D is fairly common in Greece, where it accounts for 23% of unrelated FH heterozygotes and shows 91% linkage disequilibrium with a single D19S394 microsatellite allele (257 bp). 51 In our series, all patients carrying the G528D mutation share the same intragenic haplotype, but the mutation was found to be linked with 4 different alleles of the D19S394 microsatellite. The finding that in our patients both D200G and G528D are in linkage disequilibrium with several alleles of the polymorphic microsatellite D19S394 telomeric to the LDL-R gene may be explained in several ways: (1) a de novo occurrence of these mutations in some probands; (2) a series of recombination events between the mutations in the LDL-R locus and the D19S394 locus; or (3) replication slippage of the D19S394 variable number of tandem repeats during meiosis. If we assume that recombinations between the 2 loci have occurred over time, we may estimate that numerous generations have elapsed since the common ancestor, given that there is an expected 1/400 chance of recombination between the D19S394 locus and LDL-R locus in a single generation. 29 Apparently, the D200G and G528D mutations might date back to thousands of years ago. The FS453 mutation cosegregates in all 31 families not only with the same intragenic haplotype but also with a single allele of the D19S394 microsatellite (251 bp); in this case, we can assume that this mutation is considerably more recent, as it is limited to a restricted geographic area of the country. Acknowledgments This work was supported by Telethon (project No. E. 0947), by a grant from the University of Genoa, and by a grant from the Italian National Institute of Health. References 1. Goldstein JL, Hobbs HH, Brown MS. Familial hypercholesterolemia. In: Scriver CR, Beaudet AL, Sly WS, Valle I, eds. The Metabolic and