Journal of Clinical Laboratory Analysis 9:392-396 (1 995)

Further Characterization of the Plasma Lipoprotein(a) Distribution

Wendy Craig,’ Sue Poulin,’ Andrew Bostom,2 Charles Eaton,2 Joseph Laurino,* Thomas Ledue,’ and Robert Ritchie’

Foundation for Blood Research, Scarborough, Maine; ‘Department of Family Medicine, Memorial Hospital of Rhode Island, Pawtucket, Rhode Island

1

The plasma lipoprotein(a) [Lp(a)] distribu- tion in Caucasians is heavily skewed to the right, with evidence of bimodality. As there is a well-described inverse relationship be- tween apolipoprotein(a) [apo(a)] size and Lp(a) concentration, it is likely that the pres- ence of multiple apo(a) isoforms of differing frequency has a significant impact on the fi- nal distribution of Lp(a) concentrations. We have previously described an immunoblot method for examining the relationship be- tween apolipoprotein(a) [apo(a)] size and lipoprotein(a) [Lp(a)] mass among samples heterozygous for apo(a) size, thus eliminat- ing confounding by null or undetected apo(a) isoforms. In the present study, this method has been applied to examine the plasma Lp(a) distribution, independent of the effects of apo(a) isoform size and frequency. Sev- enty subjects heterozygous for apo(a) size

were studied. To take into account the in- verse relationship ( P < 0.001) between apo(a) isoform size and Lp(a) concentration, Lp(a) data associated with each apo(a) iso- form were normalized as multiples of the median Lp(a) concentration for that isoform. These apo(a) isoform-independent Lp(a) data demonstrated a strikingly multimodal distribu- tion, with five major peaks. The relative fre- quencies of Lp(a) peaks 1-5 were 17.1%, 15.0%, 35.7%, 23.6%, and 8.6%, and asso- ciated median Lp(a) concentrations were 1 .O, 6.2,15.0,21.8, and 39.6 mg/dL, respectively. Multivariate analysis demonstrated that apo(a) isoform size accounted for 23% and isoform-independent Lp(a) peaks for 59.5% of the variation in Lp(a) concentration. Fur- ther investigation of the characteristics of the apo(a) isoform-independent Lp(a) distribu- tion is warranted. o 1995 Wiley-Liss, Inc.

Key words: lipoprotein(a), phenotype, Western immunoblotting, ELISA, coronary heart disease

INTRODUCTION

Lipoprotein (a) [Lp(a)], a risk factor for coronary heart dis- ease (CHD), is similar in structure to low density lipoprotein, except that a unique protein, apolipoprotein(a) [apo(a)], is attached covalently to apo Bloo (1). Apo(a) exists as a series of > 30 genetically determined isoforms (2,3), which differ in size by increments of Kringle IV (4).

The apo(a) gene accounts for 90% of the variability in se- rum Lp(a) levels (3, and an inverse relationship between Lp(a) mass and apo(a) size is well recognized (6,7). In a recent study, Kraft et al. (8) found that intraindividual variability in apo(a) levels is almost entirely explained by variation at the apo(a) locus, but that only 46% is explained by size polymorphism, suggesting that further heterogeneity in the apo(a) gene is related to the control of serum Lp(a) levels. In support of this, Cohen et al. (9) found in a pedigree study that apo(a) levels of apparently identical size segregated with different plasma Lp(a) levels. In that family, a particular single strand DNA polymorphism of the apo(a) gene was a marker for the

different Lp(a) levels. Furthermore, Perombelon et al. (10) reported up to a 200-fold difference in the Lp(a) concentra- tion associated with the same allele.

We have previously described a method, similar to that re- ported independently by Perombelon et al. (lo), in which heterozygous samples are used to examine the Lp(a) mass/ apo(a) size relationship, thus eliminating the potential for confounding by unidentified bands or null phenotypes that may occur when single band samples are studied (1 1). Samples obviously heterozygous for apo(a) size are identified by so- dium dodecyl sulphate polyacrylamide gel electrophoresis (SDS PAGE) and immunoblotting and the relative densities of the two bands are determined by video image analysis. The isoform-associated Lp(a) levels are then calculated from relative band intensity and plasma Lp(a) level within a given

ReceivedMay 8, 1995; accepted May 11, 1995.

Address reprint requeqts to Wendy Y. Craig, Foundation for Blood Research, P.O. Box 190, Scarborough, ME 04070-0190.

0 1995 Wiley-Liss, Inc.

Distribution of Lipoprotein(a) Concentration 393

(Kirkegaard and Perry, Gaithersburg, MD). Isoforms were assigned numbers by comparison with standards. Previously, we have identified 15 isoforms using this method (1 1). In the present study, an additional six were distinguished; one larger than those seen previously and the others having mobility intermediate between previously sequential standards.

sample. In the present study we have applied this method to study the distribution of plasma Lp(a) concentration, inde- pendent of the effects of apo(a) size,

MATERIALS AND METHODS

Study Subjects

Subjects included 36 males, ages 52.2 k 6.4 years, and 34 females, ages 53. I f 6.1 years, all Caucasian and heterozy- gous for apo(a) size by SDS PAGE and immunoblotting. A further 104 subjects (72 mcn, ages 51.4 f 6.1 years, and 32 women, ages 51.2 f 9.4 years) from the same study group, having a single apo(a) isoform by SDS PAGE and immuno- blotting, were also studied. All subjects comprised a subset of patients recruited for a separate study. The apo(a) heterozy- gotes included 38 cases of premature coronary heart disease (CHD) and 32 non-CHD community-based controls and the subjects with a single apo(a) isoform included 52 cases and 52 controls.

Sample Handling and Storage

EDTA (ethylene diaminctetraacetic acid) plasma samples were obtained after a 12-hour fast. Blood was centrifuged (3,500 rpm, 5 min), separated, and plasma samples were stored in aliquots at -80°C for 3-18 months.

Assay of Lipoprotein(a) Concentration

Lp(a) was assayed in EDTA plasma by enzyme-linked immunosorbent assay (ELISA), using Apo-Tek kits from Organon-Teknika (Rockville, MD). Within-run coefficient of variation (CV) was 6.1% and total CV was 12.3%. This as- say incorporates antiapo(a) capture and antiapo B detection antibodies.

Characterization of Apo(a) lsoforms

Apo(a) isoforms were characterized in plasma by SDS PAGE and immunoblotting as described previously (1 2). With our nomenclature, low and high isoform numbers correspond to low and high apo(a) molecular weights, respectively. The method has excellent correlation with data from SDS agarose gel electrophoresis, although SDS agarose electrophoresis gives better resolution of closely migrating bands ( 1 1). Briefly, sera were separated on 3-12% gradient gels (Integrated Sepa- ration Systems. Hyde Park, MA) using an SE600 electrophore- sis unit from Hoefer (San Francisco, CA) and proteins were transferred to Immobilon-P membranes (Millipore, Bedford, MA) using a Transblot cell (BioRad, Richmond, CA). The membranes were blocked, washed, and apo(a) was detected by incubation with rabbit antihuman apo(a) (a gift from Behring, Marburg, Germany) followed by horseradish per- oxidase-conjugated goat antirabbit Ig (INCSTAR, Stillwater, MN) and 3,3’,5,5’-tetramethyl benzidine substrate

Determination of the Lp(a) Concentration Associated With Each Apo(a) Band Within Heterozygous Samples

The Lp(a) concentration associated with each isoform within a given sample was calculated from data on plasma Lp(a) concentration and relative apo(a) band intensity on immunoblots, as described previously (1 1). The calculation of isoform-associated Lp(a) concentration is not confounded by the effect of apo(a) isoform size in the Lp(a) ELISA assay, as the assay detects Lp(a) using anti-apo B. Thus adjustment of the proportions obtained by immunoblotting in order to estimate Lp(a)-apo B, as described by Perombelon et al. (1 O), was not necessary.

Statistical Methods

The relationship between isoform number (numbered se- quentially, 1-21) and Lp(a) concentration (log transformed) was examined by linear regression. Multiple regression was used to examine the relative contributions of apo(a) isoform and apo(a) isoform-independent Lp(a) distribution peaks to variation in plasma Lp(a) levels within the study group.

In order to study the apo(a) isoform-independent Lp(a) dis- tribution among subjects with different apo(a) isoforms, it was necessary first to take into account the known inverse relationship between apo(a) isoform size and Lp(a) concen- tration. This was achieved by expressing each Lp(a) level associated with a given isoform as a multiple of the median (MOM) Lp(a) value for that isoform. Where numbers were insufficient to estimate the median Lp(a) concentration for a given isoform (n < 5), data were binned with those for the next largest isoform. Lp(a) data thus normalized were then combined and their distribution studied. Due to the range of Lp(a) levels, Lp(a) MOM data were log transformed for clearer presentation of the data.

RESULTS AND DISCUSSION

Characterization of the Apo(a) Isoform- Independent Lp(a) Distribution

As illustrated in Figure 1, there was an overall inverse re- lationship between Lp(a) concentration and apo(a) size (r = 0.492, P < 0.001). Separate clusters of Lp(a) mass data were observed within each isoform class, indicating that the effect of isoform size is superimposed on a distinctly multimodal Lp(a) distribution. The data presented in Figure 1 are strik- ingly similar to those given by Perombelon et al. (10) for

Craig et at.

200

100 70 50

30 20

10 7 5

3 2

1 0.7 0.5

0.3 0.2

0.1

0

0

0

0

0

0

0 0 0 0 0

0 8 0 0 1 0

I J

8 - " t i e Q

Q

0 0

0

0 8

0 0 0 0 0

0

O O 0 0 0

0 0

0

I I I I I I I I I I I 1 I l l I I I I I I

-1 1 2 3 3A 4 4A 5 5A 6 6A 7 7A 8 8A 9 10 11 12 13 14

Apo(a) lsoforrn Number

Fig. 1. Relationship between Lp(a) concentration and apo(a) size among heterozygous samples.Apo(a) isoform number and band-specific Lp(a) levels were determined in 70 sera that were demonstrably heterozygnus for apo(a) by SDS PAGE and imrnunoblotting, as described in Methods. There was a significant inverse relationship (r = 0.476, P < 0.001) between Lp(a) concentration and apo(a) size.

isofom-associated Lp(a) concentrations. However, we have interpreted these findings somewhat differently. Perombelon et al. (10) focused on relatively large areas of clustering within the graph, which they interpreted as evidence for the pres- ence of at least two subpopulations of apo(a) alleles with dif- ferent sizes and expression. In contrast, we focus on the presence of periodicity parallel with the regression line de- scribing the apo(a) size/Lp(a) mass relationship.

To examine the L,p(a) concentration distribution without confounding by the observed between-isoform differences in Lp(a) level (Table l), isoform-associated Lp(a) data were normalized between isoforms as described in Materials and Methods. Thus data from Figure 1 could be condensed into a single distribution; this apo(a) isoform-independent Lp(a) concentration distribution is illustrated in Figure 2.As the sepa- rate distributions of cases and controls were superimposable (not shown), data from these subgroups were combined. The

TABLE 1. Lipoprotein(a) Levels Associated With Different Apolipoprotein(a) Isoforms

Apolipoprotein(a) Frequency Median isofom-associated isoform group" (n) lipoprotein(a) (mg/dL)

-1, I, 2 5 70.6 3 12 34.8 J,3a 8 40.9 5 ,4a 13 11.2 6, Sa 20 17.4 7,6a 22 12.6 8, l a 21 13.4 9,8a 14 9.0 10 10 12.3 11 8 3.2 12. 14 1 2.9

n 4 in a given group, data were binned with those for the next larg- est isoform.

Distribution of Lipoprotein(a) Concentration 395

26 - 24 - 22 -

20 - 18 - 16 - 14 - 12 - 10 -

8 -

6 -

4 -

2 -

-

- - - - - - -

- - - - -

_.

- - - - - -

- - 0 . 7 I d - . , . . . . . . . . . I . . .

0.0 0.2 0.4 0.6 0.8

log Lp(a) MOM

Fig. 2. Isofom-independent distribution of Lp(a) concentration. Apo(a) isofom-specific Lp(a) concentrations were normalized hetween isofoms by expressing data as multiples of the median (MOM) value for the relevant isofom, as described in Materials and Methods.

apo(a) isoform-independent Lp(a) distribution is markedly multimodal, with five peaks (four major and one comprising several minor peaks combined). When this same analysis was applied to single band samples from the same study popula- tion (n = 104), a similar pattern was obtained (data not shown), although individual peaks were broader, most likely due to the probable inclusion of unidentified heterozygous samples with homozygous samples in this group.

Characteristics of Apo(a) Isoform-Independent Lp(a) Peaks

Based on Figure 2, the data were divided into five discrete peaks, with log Lp(a) MOM ranges of -2.1--0.55 (peak 1; minor peaks combined), -0,549--0.2, -0.149-0.15, 0.1514.45, and 0.451-0.7 (peaks 2-5, respectively). The characteristics of these five peaks are given in Table 2. There was a 39.6-fold difference in median apo( a) isoform-associ- ated Lp(a) concentration between peaks 1 and 5. Lp(a) peak 3 was the most frequent (35.7%) and peak 5 was the least frequent (8.6%).

The data indicate that there are a number of discrete deter- minants of Lp(a) concentration, in addition to the effect of apo(a) isoform size and thus are consistent with previous re- ports suggesting that genetic regulation of Lp(aj levels oc- curs in addition to that determined by apo(a) size (5,8-11,13). The mechanism for this regulation cannot be deduced from the present data, but might be related to the findings of Rader

TABLE 2. Characteristics of Peaks in Apo(a) Isoform- IndeDendent Lda) Distribution

Apo(a) isoforn-independent Lp(a) distribution peaks

1 2 3 4 5

Lipoprotein(a) (m.ddL) Median 1 .o 6.2 15.0 21.8 39.6 Range 0.2-6.2 1.4-15.5 2.3-83.3 4.3-174.8 9.2-58.4

Frequency n 24 21 so 33 12 w 17.1 15.0 35.7 23.6 8.6

396 Craig et al.

et al. (1 3), who reported that variation in Lp(a) levels among individuals with the same isoform is caused primarily by dif- ferences in Lp(a) production rate.

Contributors to Variability in Plasma Lp(a) Concentration

Multiple variable linear regression analysis demonstrated that apo(a) isoform number and Lp(a) peak number together accounted for 82.5% of the variation in band-specific Lp(a) levels in the present study group. Apo(a) isoform differences explained 23% of the variation, and Lp(a) peak number ex- plained 59.5%.

The results are consistent with those of others (5 ,8 ) . who found that the apo(a) gene accounts for 90% or more of the variation in Lp(a) levels. Estimates of the contribution of apo(a) isoform size to the variation in Lp(a) concentrations have varied. Kraft et al. (8) report that DNA size polymor- phism accounts for 46% of variation in Lp(a) concentrations, whereas Boerwinkle et al. ( 5 ) report a value of 69%. Our present estimate (23% of the variation in Lp(a) concentra- tion explained by apo(a) isoform size) is the lowest to date. This discrepancy may be related to the nature of the Lp(a) assay used or to the fact that confounding by null or unde- tected isoforms is eliminated in the present analysis.

In summary, elimination of the effects of apo(a) isoform size and frequency reveals a distinctly multimodal Lp(a) dis- tribution. The peaks of this distribution may represent dis- crete determinants of plasma Lp(a) concentration that act independently of the effect of apo(a) isoform size.

ACKNOWLEDGMENTS

The authors thank Louis M. Neveux for performing the sta- tistical analyses, and Nancy Healy for technical assistance. This work was supported in part by grant 9307843s from theAmeri- can Heart Association, Rhode Island Affiliate (CBE).

REFERENCES 1. 2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

Utemann G: The mysteries of Lp(a). Scienre 246:904-910. 1989. Kamboh M1, Svitko CM, Williams ER, Ferrell RE, Pollitzer WS: Hypervariable polymorphism of apo(a) in blacks and whites as reflected by phenotyping. Chem Phys Lipids 671683283-292, 1994. Marcovina SM, Zhang ZH, Gaur VP, Albers JJ: Identification of 34 apolipoprotein(a) alleles between American blacks and whites.Biochem Biophys Res Comrnun 191:1192-1196, 1993. Gavish D, Azrolan N, Breslow JL: Plasma Lp(aj concentration is in- versely correlated with the ratio of kringle IV/kringle V encoding do- mains in the apo(a) gene, J Clin Invest 84:2021-2027, 1989. Boerwinkle E, Leffert CC, Lin J, Lacher C, Chiesa G, Hobbs HH: Apolipoprotein(a) gene accounts for greater than 90% of the variation in plasma lipoprotein(a) concentrations. J Clin Invest 9052-60, 1992. Utermann G, Menzel HJ, Kraft HG, Duba HC, Kemmler H, Seitz C: Lp(a) glycoprotein phenotypes. Inheritance and relation to Lp(a)-lipo- protein concentrations in plasma. J Clin Invest 80:458-465, 1987. Gaubatz JW, Ghanem KI, Guevara J, Jr, Nava ML, Patsch W, Morrisett JD: Polymorphic forms of human apolipoprotein[a]: Inheritance and relationship of their molecular weights to plasma levels of lipoprotein[a]. JLipidRes 31:603413, 1990. Kraft HG, Kochl S, Menzel HJ, Sandholzer C, Utermann G: The apolipoprotein(a) gene: A transcribed hypervariable locus controlling plasma lipoprotein(a) concentration. Hunt Genet 90:220-230, 1992. Cohen JC, Chiesa G, Hobbs HH: Sequence polymorphisms in the apolipoprotein(a) gene: Evidence for dissociation between apolipoprotein(aj size and plasma lipoprotein(a) levels. J Clin In- vest 91:1630-1636, 1993. Perombelon YFN, Soutar AK, Knight B L Variation in lipoprotein(a) concentration associated with different apolipoprotein(a) alleles. J Clin Invest 93:1481-1492, 1994. Craig WY, Poulin SE, Ledue TB, Kamboh MI: Apolipoprotein(a): A comparison of isoforms identified by sodium dodecyl sulfate-polyacryl- amide gel electrophoresis or by sodium dodecyl sulfate-agarose gel elec- trophoresis. Electrophoresis 14: 1038-1041, 1993. Craig WY, Poulin SE, DiGeorge AM, Forster NR, Neveux LM, Ledue TB, Ritchie RF: The molecular weights of twelve apolipoprotein(a) variants, determined using haptoglobin 2-2 polymer molecular weight standards. Appl Theor Electrophoresis 3:241-246, 1993. Rader DJ, Cain W, Zech LA, Usher D. Brewer HB Jr: Variation in lipoprotein(a) concentrations among individuals with the same apolipoprotein(a) isoform is determined by the rate of lipoprotein(a) production. J Clin Invest 9 1:443+!47, 1993.