Apo(a) Phenotypes and Lp(a) Concentrations in Offspring of Men With and Without Myocardial Infarction

Subjects

The design of EARS has been described elsewhere.²⁵ In brief, students were recruited and studied between October 1990 and February 1991. They attended universities in 14 European cities (Fig 1⇓), which for purposes of analysis were grouped a priori into five regions: Finland (Oulu and Helsinki), Great Britain (Glasgow, Scotland and Bristol, England), northern Europe (Göteborg, Sweden; Aarhus, Denmark; and Hamburg, Germany), middle Europe (Ghent, Belgium; Zurich, Switzerland; and Innsbruck, Austria), and southern Europe (Bordeaux, France; Barcelona and Reus, Spain; and Naples, Italy). Criteria for grouping were known regional differences in CHD mortality rates,²⁴ geography, and language, because interpopulation genetic differences roughly follow boundaries between major language groups. The students were between 18 and 26 years of age, and students whose fathers had a documented MI before the age of 55 were defined as cases (n=682). Each case was matched by sex and age to two students without a family history of CHD (control subjects; n=1312). The mean age of control subjects’ fathers was slightly but significantly higher than the mean age of cases’ fathers at MI (53.3±0.2 and 52.6±0.3 years, respectively; P<.01). The diagnosis of paternal nonfatal MI was based on criteria of the World Health Organization, and to ascertain fatal MI in the father we used a system involving five diagnostic subclasses.²⁵

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Map shows the locations of the 14 universities participating in the European Atherosclerosis Research Study.

Venous blood was collected after an overnight fast, and frozen serum samples were shipped to University Hospital St. Jan, Brugge, Belgium, for Lp(a) measurements. We did not have access to serum from 85 students, reducing the number of subjects for analysis to 1909 (645 cases and 1264 control subjects). Apo(a) phenotyping was performed in three laboratories (Innsbruck, Hamburg, and Aarhus), and one third of the samples from each recruitment center were shipped (frozen)²⁵ to each of the three laboratories. Apo(a) phenotypes were determined in 646 cases and 1259 control subjects (samples from 89 students were missing). Data were processed and analyzed at INSERM U.258 in Paris, France.

Lp(a) Determination

Lp(a) was quantified with a sandwich-type enzyme-linked immunosorbent assay (ELISA)²⁶ by use of a combination of a monoclonal antibody against apo(a) for coating and a stable polyclonal anti-apoB–peroxidase conjugate for detection. The conjugate was prepared from high-affinity anti-apoB immunoglobulins by use of the periodate coupling procedure. After immunization of BALB/c mice with Lp(a), we obtained 14 stable clones that were tested for specificity against LDL, plasminogen, and apo(a). We selected the 1D1 clone, an IgG1k, with high affinity and specificity for apo(a). For the ELISA, polystyrene microtiter plates were coated with the purified monoclonal antibody 1D1 at 4 μg per milliliter of coating buffer. Plates were blocked with the assay buffer, a sodium phosphate buffer (0.05 mmol/L, pH 7.4, plus 0.15 mol/L NaCl and 1 g/L casein), for 1 hour at 23°C. Between incubations, we washed the plates with assay buffer containing 0.5 mL/L added Tween-20 (Bio-Rad Labs).

For the sandwich ELISA for Lp(a), plasma samples and conjugate were diluted 2000- and 6000-fold, respectively, with the buffer used for blocking. The samples were incubated with either the immobilized antibody or the conjugate for 2 hours at 37°C. The peroxidase activity, with o-phenylenediamine as substrate, was measured by reading of the absorbance at 490 nm. Calibration curves were prepared from appropriate dilutions of purified Lp(a). Intra-assay and interassay coefficients of variation were determined at low (50 mg/L), medium (200 mg/L), or high (750 mg/L) Lp(a) concentrations. The mean coefficients of variation obtained were <3% (intra-assay) and 5% (interassay). The assay was calibrated with a highly purified Lp(a) fraction isolated from pooled human plasma. The protein content of this primary standard was determined by a Folin-Lowry assay in the presence of SDS. Lp(a) protein was converted to lipoprotein mass with the assumption that the average protein content of Lp(a) is 35%. A set of serum samples of different Lp(a) concentrations (30 to 800 mg/L) was used for internal quality control and stored in portions at −20°C for 2 years. No effect of storage was observed, provided that the samples were not thawed more than once. All Lp(a) determinations of the case-control samples were performed on frozen specimens.

Apo(a) Phenotyping

Apo(a) phenotypes were determined by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) followed by immunoblotting by a method slightly modified from that described by Utermann et al.¹⁸ The samples were prepared by mixing 4 μL of serum (null phenotypes and single banded phenotypes were redetermined with 6 to 8 μL) with 50 μL of a 0.02 mol/L ethylmorpholine buffer containing 50 g/L SDS (BDH Chemicals UK). 2-Mercaptoethanol (4 μL) and a 15 g/L bromophenol blue solution in glycerol (2 μL) were added, and the samples were boiled for 10 minutes. Aliquots of 5 μL (Innsbruck) or 25 μL (Hamburg and Aarhus) were then subjected to SDS-PAGE in a discontinuous buffer and gel system²⁷ with a stacking gel containing 3.6% polyacrylamide and a resolving gel of 5.2% polyacrylamide. Acrylamide and N,N′-methylene-bis-acrylamide were purchased from LKB, and the electrophoresis equipment was either the Desaphor va. from Desaga (Hamburg and Aarhus) or the Minigel Twin from Biometra (Innsbruck). In Innsbruck, proteins were transferred to nitrocellulose filters (BA85, 0.45 μm, Schleicher and Scheull) by the method described by Towbin et al²⁸ with the Trans blot cell (Bio-Rad). Antigens were visualized on the nitrocellulose filters by a double-antibody procedure. The first antibody was a monoclonal anti-Lp(a) raised in mice,²⁹ and the second was a rabbit anti-mouse peroxidase conjugate (DAKO-PATTS).

The various apo(a) isoforms were designated according to their relative electrophoretic mobility on SDS-PAGE compared with apoB-100, with the terminology proposed by Utermann et al.⁴ F denotes a fast-migrating isoprotein smaller than apoB-100, and S1, S2, S3, and S4 denote larger isoproteins with progressively lower mobility than apoB-100. B is a band with the same mobility as apoB-100. Null type (0) designates blots showing no bands at all. As references, samples from heterozygotes with phenotypes S1S2, BS4, and FS3 were obtained from a study of healthy Caucasian Danes³⁰ and run alternately in every fourth lane. Sample isoforms were classified by comparison with the closest reference. Classification was done without knowledge of whether the sample was from a case or a control subject.

Other studies have shown that some of the 22 potential phenotypes are infrequent.³⁰ To avoid empty “cells” in the statistical comparisons of the apo(a) frequencies, we combined a priori the apo(a) phenotypes into four groups according to molecular weight (Table 1⇓). To compare our results with those of Sandholzer et al,¹⁶ we further divided the phenotypes into two classes. Class A comprised the subjects with at least one of the isoforms F, B, S1, and S2, and class B comprised the subjects with the isoforms S3, S4, and 0 (Table 1⇓).

Statistical Methods

Lp(a) concentrations in control populations and in cases and control subjects were compared by the Kolmogorov-Smirnov test and by one-way ANOVA after logarithmic transformation of data.³¹ The χ² test was used to compare distributions on apo(a) groups among cases and controls and among populations and for comparison of cases and controls with Lp(a) levels above 30 mg/dL. The Kruskal-Wallis test was used for comparison of Lp(a) concentrations between phenotype groups. As a measure of association between apo(a) class and case or control status, odds ratios with 95% confidence intervals were calculated from 2×2 tables. The Mantel-Haenszel test statistic was calculated from the stratified 2×2 tables.

Lp(a) Concentrations and Apo(a) Phenotypes in Control Subjects by Region

Lp(a) concentrations differed significantly between regions (P=.01). This difference was not significant in men analyzed separately (P=.23), but it was significant in women (P=.01) and in the subgroup of women not taking oral contraceptives (P=.03). In men, the median Lp(a) concentration was lowest in Finland (5.2 mg/dL; 95% CI, 3.9 to 8.5 mg/dL), highest in Great Britain (7.6 mg/dL; 95% CI, 4.6 to 15.0 mg/dL), and practically identical in the three other regions. In women, Lp(a) concentrations were also lowest in Finns (4.2 mg/dL; 95% CI, 3.5 to 5.9 mg/dL), but they were highest in southern Europeans (8.2 mg/dL; 95% CI, 6.6 to 9.9 mg/dL). The inverse relationship between apo(a) molecular weight and Lp(a) concentration was statistically significant in all five regions (P<.01) (data not shown). This relationship is also apparent after grouping of phenotypes (Table 2⇓ and Fig 2⇓). In general, the null and the large molecular weight isoforms (phenotype group IV) occurred more frequently than the small molecular weight isoforms (phenotype group I), and with one exception (northern Europeans), the rank order of phenotype group frequency was IV>III>II>I (Table 3⇓).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Graphs show comparisons of lipoprotein(a) Lp(a) concentrations in control subjects and cases within the same phenotype groups from five regions of Europe.

Because Lp(a) concentrations are largely determined by polymorphisms in the apo(a) gene locus, here expressed as apo(a) phenotypes, we investigated whether regional differences in Lp(a) concentrations in control subjects were functions of the phenotypes. Multiple pairwise comparisons of regional apo(a) phenotype frequencies indicated that there were two major geographical areas within which phenotype frequencies were similar and between which phenotype frequencies differed significantly. One of these regions was defined by the control subjects in the Finnish, middle European, and southern European universities, and the other was defined by the control subjects in the universities in Great Britain and northern Europe. After adjustment for regional differences in phenotype frequencies, the overall regional difference in Lp(a) concentration was no longer statistically significant (P=.54).

Case-Control Study

In both cases and control subjects, the distribution of Lp(a) concentration was skewed towards lower concentrations (Fig 3⇓). Although the median Lp(a) concentration was somewhat higher in cases (7.3 mg/dL; 95% CI, 6.3 to 8.1 mg/dL) than in control subjects (6.6 mg/dL; 95% CI, 6.1 to 7.2 mg/dL), this difference was not statistically significant. The similarity between the distributions of Lp(a) concentration in cases and control subjects by region and overall is apparent when they are depicted as cumulative frequency curves (Fig 3⇓).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Cumulative frequency distributions show Lp(a) concentrations in subjects with (cases) and without (control subjects) a paternal history of myocardial infarction before the age of 55 years. Data are shown from five regions of Europe and for the study population as a whole. Median concentrations of Lp(a) (in mg/dL) with 95% confidence intervals for cases and control subjects are shown on each graph. For statistical comparisons, see text.

Neither in men (P=.28) nor in women (P=.98) did Lp(a) concentrations differ significantly between cases and control subjects (data not shown). In northern Europe only, slightly higher Lp(a) concentrations in male cases than in male control subjects were borderline significant (P=.052). Moreover, significantly more northern European male cases than male control subjects had Lp(a) concentrations exceeding 30 mg/dL (P=.040). This did not pertain to northern European women (P=.29), however, and overall, Lp(a) concentrations exceeding 30 mg/dL were found in 16.5% of cases and in 15.5% of control subjects (P=.63).

In the total study population, the distribution of cases and control subjects according to the four groups of apo(a) phenotypes did not differ significantly (P=.74) (Table 3⇑). This also pertained to each region with the exception of Great Britain, where there was a significant difference (P=.035) due to a higher frequency of cases in group II and a higher frequency of control subjects in group I (Table 3⇑). When groups I and II are considered together (class A), as in the classification system proposed by Sandholzer et al,¹⁶ differences in phenotype classes between cases and control subjects were not significant (Table 3⇑). There was a tendency, however, for cases in Great Britain and northern Europe to belong to class A. Odds ratios with CIs are given in Table 3⇑. The Mantel-Haenszel test statistic was calculated from the stratified 2×2 tables (class A versus class B [group III plus group IV]) but, again, the difference was not statistically significant (Mantel-Haenszel test statistic=0.53, P=.47). Within the same phenotype groups, Lp(a) tended to be higher in cases than in control subjects in Finland and Great Britain, but these differences were not significant (Table 2⇑).