Factor analysis of plasma lipoprotein components

Atherosclerosis, 69 (1988) 219-224 Elsevier Scientific Publishers Ireland,

219 Ltd.

ATH 04076

Factor analysis of plasma lipoprotein components I. Aursnes ‘, P. Smith 2, E.N. Christiansen ’Department

3 and K. Berg 4

ofMedicine,Ullevdl

Hospital, ’ Department ofMedicine, Red Cross Clinic, ’ Institute for Nutrition Research, and 4 Institute of Medical Genetics, University of Oslo, Oslo (Norway) (Received 9 March, 1987) (Revised, received 3 September, 1987) (Accepted 7 September, 1987)

Summary

Serum phospholipids were analyzed for their content of long-chained fatty acids together with other components of lipoproteins (total- and HDL-cholesterol, apolipoproteins A-I and B, triglycerides), in 60 coronary heart disease patients and 30 control individuals. Some of the individual variations in content of the various components showed co-variation with each other. This formed the basis for the extraction of 7 ‘factors’ by the statistical procedure ‘factor analysis’. Analysis of variance was performed with the ‘factor scores’ for subgroups of high and low age and high and low total serum cholesterol. The analysis revealed two unexpected factors which discriminated with statistical significance between young, hypercholesterolaemic patients and controls. One factor was a positive risk factor and the other a negative one. They could possibly be dependent on the existence of two at present uncharacterized subgroups of lipoproteins. These lipoproteins contained, according to the analysis, large amounts of certain fatty acids. It is suggested that fatty acid analysis might be useful in the characterization of lipoproteins that are involved in the development of atherosclerosis.

Key words: Apolipoproteins; Triglycerides

Atherosclerosis;

Cholesterol;

Introduction

Serum high density lipoproteins (HDL), quantified by their cholesterol content, are negatively associated with coronary heart disease (CHD) [1,2] and have been shown to have negative predictive The study was supported by a grant from the Ester and Aagot Johnsen Legacy, The Norwegian Council on Cardiovascular Diseases. Correspondence to: Dr. I. Aursnes, Department of Medicine, Ulleval Hospital, 0407 Oslo 4, Norway. 0021-9150/88/$03.50

0 1988 Elsevier Scientific

Publishers

Ireland,

Factor analysis; Fatty acids; Lipoproteins;

value for the occurrence of CHD (3-5). Hypothetically HDL could prevent atherosclerosis by increasing the cholesterol transport away from the arterial wall [6]. It has, however, not been shown that raising HDL-cholesterol by drug therapy reduces the risk for disease. Interest has also been focused on other components of HDL than cholesterol (i.e. apoliproteins) in the search for subfractions of the lipoprotein which might be more directly related to cholesterol deposition in the body. Ltd.

220 Some of the present authors have reported a population of myocardial infarction survivors having HDL-cholesterol levels which did not differ from control individuals of the same age and sex [7]. This prompted us to examine whether the main protein component of HDL, apoliprotein AI (apo A-I), showed any difference between the groups. We have also made a ‘factor analysis’ including all the measured lipoprotein components (fatty acids, apoliproteins, triglycerides and cholesterol). The rationale for doing this is that subfractions of lipoproteins might contain different percentages of the various fatty acids. It might therefore be possible to characterise both known and unknown lipoproteins according to their fatty acid composition. This has been done for HDL, and HDL, in that the latter was found to contain more unsaturated acids [8]. We chose, for the present purpose, to analyse the fatty acids in the plasma phospholipids only. Fatty acids on the surface of the lipoproteins may be more related to the function of lipoproteins than fatty acids in general. The factor analysis identified, as expected, the large, well known lipoprotein subfractions LDL and HDL. In addition factors emerged that were not so easily identifiable. They were not related to one of the two apoliproteins which we had measured. The factors were characterized by one or more of the fatty acids. One possible explanation for their occurrence is that they represent subfractions of lipoproteins, which, for the present purpose, we might call lipoprotein ‘factors’. It was therefore of interest to see whether these lipoprotein factors varied negatively or positively with the presence of CHD. Subjects and methods The patients were selected in order to obtain a wide span in serum cholesterol levels. Several hundred post-infarction patients were screened for coronary risk factors (hypertension, total serum cholesterol and smoking). They were given a score as described previously [7], and the two groups of patients selected for ‘study were randomly chosen from the lowest and highest quintiles of the risk factor score distribution. In order to observe the effect of age on the analysis, efforts were made to

secure that half of the persons in each group, and also in a control group, were below 60 years, called ‘young patients’. Three main groups were thus formed, each consisting of 30 individuals, whereof 15 were ‘young’ and 15 ‘old’. The characteristics for the ‘low risk’, ‘high risk’ and control groups were respectively (mean * SD): weight 71 + 10, 74 k 11 and 75 k 10 (kg); height 175 + 8, 179_t9 and 176 + 8 (cm); age 62+ 8, 60 * 8 and 66 k 12 (years); systolic blood pressure 125 + 16, 144 + 18 and 135 + 13 (mm Hg). Further characteristics for the 3 groups of 30 individuals each were: 6, 5 and 8 women; 8, 16 and 8 regular smokers at the time of acute infarction; 15, 13 and 0 were presently taking beta-blockers; none were taking lipid-lowering drugs; 3, 1 and 2 said they were using vegetable oils in the household, 3, 3 and 12 were taking up to 1 table spoon of cod liver oil daily and 7, 3 and 0 were reporting recent dietary changes. Seven of the ‘low risk group and 5 of the ‘high risk’ group has experienced angina before their infarction. The control group consisted of healthy present or former employees of a local bank. They reported no coronary symptoms and had a normal resting ECG. Serum triglycerides, total cholesterol and HDL cholesterol were quantified by standard enzymatic methods, the latter after precipitation with phosphotungstic acid and magnesium chloride (Boehringer Mannheim Diagnostica, F.R.G.). Apo A-I and B (apo B) were assayed by quantitative single immunodiffusion using antisera of own production [9]. The patients and the control individuals attended the clinic fasting in the morning. Ten ml of blood was drawn from an antecubital vein into a glass tube without anticoagulant. Blood was permitted to clot for 60 min at 37 o C. Serum was then harvested by centrifugation. Lipid extraction, separation of phospholipids by thin-layer chromatography and gas-chromatographic analyses of the fatty acid compositions were performed as described previously [7]. Factor analysis was performed with BMDP (Biomedical Computer Programs) from UCLA, U.S.A. The variables were 11 long-chain fatty acids, apo A-I, apo B, triglycerides and total and HDL-cholesterol. The initial factor extraction was done with the method of principal components.

221 Rotation was ‘direct oblimin’ with gamma = - 0.5. Before analysis the data were transformed by normalisation, i.e. mean = 0 and SD = 1. The data analysis was performed in two steps. At first all the measured concentrations of components were subjected to factor analysis and 7 factors were extracted by the computer. Here the individual patients and controls were supplied with a ‘factor score’ by the computer. The factor scores were given in terms of standard deviations away from the mean value for all individuals and indicates to what degree serum from one individual contains a certain factor. The factor scores for each individual factor were then divided into the 6 groups described above, according to age, risk factor score and morbidity. Student t-test was used to test all subgroups against all the others. The analysis was performed ‘a priori’, i.e. no grouping of patients were performed after the collection of data had been completed, but no measures were taken against the effect of multiple testing. The number of extracted factors was varied between 5 and 8. Variance analysis between groups were performed with the factor scores obtained with each series. It was decided to report the findings obtained with 7 factors because a particularly high degree of separation between the groups was observed when they were tested against each other with 7 operating factors.

TABLE SERUM

Results Apoliprotein A-I and cholesterol levels

Apo A-I concentration was slightly reduced in patients compared with controls (Table 1). The difference from controls reached significance only in ‘young, low risk’ patients. The level was almost identical in ‘young, high risk’ patients. This finding is in keeping with other reports [2,5]. The explanation could be that polymorphism at the apo A-I locus may be related to the serum quantity of apo A-I and to the development of atherosclerosis [lo]. There were no differences in HDL-cholesterol levels between the groups (Table 1). HDL-cholesterol and apo A-I were however correlated, r = 0.40 (patients r = 0.47, controls r = 0.38). The main results of the factor analysis are tabulated in Table 2. The factors are arranged in a sequence of decreasing ‘eigenvalue’. The latter expresses the degree of variation in the material ‘explained’ by each factor. For clarification the main components of each factor are summarized in Table 3. The results of the variance analysis are indicated by P-values and refer to differences between subgroups as regards ‘factor score’ as described under ‘Subjects and methods’. The absolute levels for factor scores for the various groups are not reported because of their limited value. Only trends in differences can form basis for discussion and future experiments.

1 CONCENTRATIONS

OF CHOLESTEROL

AND

APOLIPOPROTEINS

A-I AND

B

Mean f SD. ‘Low risk’ patients

Variable

Group

Age W

1

Total cholesterol (mmol/l) Apolipoprotein B (mg/lOO ml) HDL-cholesterol (mmol/l) Apolipoprotein A-I (mg/lOO ml)

i 60 5.7* 0.8 69 k 8 1.2* 0.3 122 *19 *

* P = 0.002 for difference

groups.

between

Group

‘High risk’ patients 2

> 60 5.5* 0.7 64 k12 1.2+ 0.3 124 +20

Group

3

< 60 8.7* 1.5 76 +16 1.2+ 0.3 122 k24

Group

Control 4

> 60 8.5+ 1.3 79 f 8 1.3* 0.4 133 k18

Group

individuals 5

< 60 6.2+ 1.1 65 *lO 1.2* 0.4 142 &15 *

Group

6

> 60 6.4* 0.9 64 + 5 1.2* 0.5 132 f22

222 TABLE

2

COEFFICIENTS Variables

OF CORRELATION

AND VARIABLES

Factors

18:2(n -6) 18:3(n -3) 18:3(n -6) 20:4(n -6) 20:3(n -6) 20:5(n -3) 20:l(n -9) 20:o 22:4(n -6) 22:6(n -3) 22:5(n -3) HDL-cholesterol Total cholesterol Apo A-l Apo B Triglycerides

1

2

3

4

5

6

7

-0.11 -0.11 0.31 0.43 0.08 0.93 0.15 0.05 -0.14 0.86 0.84 - 0.01 - 0.06 - 0.04 0.00 - 0.05

- 0.02 - 0.01 -0.00 - 0.06 0.06 - 0.08 0.06 0.11 -0.19 - 0.01 -0.00 0.04 0.89 - 0.01 0.88 0.46

- 0.09 - 0.01 0.05 - 0.07 0.08 0.00 0.12 - 0.09 -0.18 - 0.09 0.03 0.80 0.12 0.82 - 0.06 -0.37

0.00 0.96 0.66 0.27 0.04 0.05 0.02 0.05 -0.10 0.01 0.03 -0.14 0.08 0.15 -0.13 0.09

0.00 - 0.05 -0.00 0.45 0.84 - 0.23 0.56 0.07 0.70 0.09 0.22 - 0.03 0.03 0.09 -0.11 0.25

0.04 0.08 - 0.08 - 0.28 -0.12 0.11 0.39 0.85 0.23 - 0.03 0.02 -0.19 0.02 0.11 0.13 -0.36

0.96 -0.05 0.23 0.20 0.08 -0.13 0.05 0.01 -0.19 0.07 0.07 0.02 - 0.01 -0.12 - 0.03 - 0.26

2.71

1.94

1.85

1.58

1.53

1.27

1.19

‘Eigenvalue’

TABLE

FOR FACTORS

3

MAIN CONTENTS OF FACTORS WITH CORRESPONDING SCORES FOR GROUPS (Same groups as in Table 1) Factor

No.

1 2 3 4 5 6 I

P-VALUES

FOR

DIFFERENCES

BETWEEN

FACTOR

Components

Statistical

Fatty acids 20 : 5( n - 3), 22 : 5( n - 3), 22 : 6( n - 3) Total cholesterol, apo B HDL-cholesterol, apo A-I Fatty acids 18 : 3( n - 6), 18 : 3( n - 3) Fatty acids 20 : 3( n - 6), 22 : 4( n - 6) Fatty acids 20 : 0, 20 : l( n - 9) negatively correlated Fatty acid 18 : 2( n - 6)

No significant differences P < OklOl, group 3 high vs group 5 P c 0.05, group 1 low vs group 5 No significant differences P i 0.05, group 3 high vs group 5 P i 0.01, group 3 low vs group 5 No significant differences

Discussion

In the present series HDL-cholesterol is not lower in patients than in controls. We have no explanation for this. Moreover, others have found total cholesterol and HDL-cholesterol levels to be inversely related [3], whereas we find similar HDL-cholesterol levels in high and low total cholesterol individuals. The observed reduction in apo A-I concentration in patients is less than what has been found

with triglycerides

significance

previously [2]. One possible explanation is the age of the patient-group under study. Since apo A-I concentration is influenced by genes [ll], it is reasonable that its expression in disease is more pronounced in relatively young patients as observed in this study. Our findings could also be biased by the fact that we observed only patients that had survived their first myocardial infarction. This, together with the relatively small number of patients under study and a somewhat higher mean age of controls

223 compared with patients, might explain the lack of differences in HDL-cholesterol between the groups. The factor analysis performed in this study lends the opportunity to study the distribution of both known and possibly unknown lipoproteins in the material. One recognizes factor 2 as LDL and factor 3 as HDL. Factor 2 separates groups 3 and 4 extremely well from groups 5 and 6 due to the selection of the patients for the study. Factor 3 did not separate the groups significantly better than apo A-I alone. Dietary effects may underlie some of the factors constructed by the computer program. Particularly factor 1 which contains n-3 fatty acids might reflect more or less fish eating or fish oil consumption in the individuals. The factor does not discriminate well between the various groups, although somewhat higher scores for ‘low risk patients are observed. The differences in factor 4 between the groups did not reach statistical significance. Factor 5 is a positive risk factor which discriminates (P < 0.05) between young, hypercholesterolaemic patients and controls. Factor 5 is positively correlated with triglycerides (Table 2) and is therefore likely to reside in the very low density lipoprotein fraction. The triglyceride level has been reported by some workers to be a risk factor for coronary disease [12]. Factor 5 is indicative of a lipoprotein containing large amounts of fatty acids 20 : 3 and 22 : 4 in its phospholipids. Factor 6 is a negative risk factor, with discriminating power between groups 3 and 5. Absolute difference in mean scores between groups 3 and 5 for factor 6 was almost 1 standard deviation. Factor 6 is negatively correlated with triglycerides (Table 2). The analysis also shows that its phospholipids are dominated by the fatty acids 20 : 0 and 20 : 1. These two acids constitute a small fraction, each less than l%, of the fatty acids in plasma phospholipids. The fact that they express themselves in the analysis would mean that they make up a large fraction of the fatty acids in the hypothetical lipoprotein that may be hidden behind factor 6. Lastly factor 7 is extracted. It is dominated by the fatty acid 18 : 2 and does not discriminate between the groups.

We conclude that fatty acid determinations in lipoprotein fractions might be useful tools in the analysis of genetic and dietary variation in lipid levels in man and thereby in identifying people of various risk categories. Sofar, however, only minimal contribution has been made as to the question of identifying unknown lipoproteins which can distinguish CHD patients from others. Acknowledgements The present report is based on data from the ‘WARE’ study, Red Cross Clinic, Oslo. Statistical analysis was performed by Rolf Volden at The Norwegian Computing Center, Oslo, Norway. References Caste&, W.P., Doyle, J.T., Gordon, T., Hames, C.G., Hjortland, MC., Hulley, S.B., Kagan, A. and Zukel, W.J., HDL cholesterol and other lipids in coronary heart disease. The Cooperative Lipoprotein Phenotyping Study, Circulation, 55 (1977) 767. Berg, K. and Barresen, A.L., Serum-high-density-lipoprotein and atherosclerotic heart-disease, Lancet, i (1976) 499. Miller, N.E., Forde, O.H., Thelle, D.S., Mjos, O.D., The Tromso Heart Study. High-density lipoprotein and coronary heart-disease: A prospective case-control study, Lancet, ii (1977) 965. Kannel, W.B., Castelli, W.B. and Gordon, T., Cholesterol in the prediction of atherosclerotic disease. New perspective based on the Framingham study, Ann. Int. Med., 90 (1979) 85. Ishicowa, T., Fidge, N., Thelle, D.S., Forde, O.H. and Miller, N.E., The Tromse Heart Study: Serum apolipoprotein Al concentrations in relation to future coronary disease, Eur. J. Clin. Invest., 8 (1978) 179. Miller, G.J. and Miller, N.E., Plasma-high-density-lipoprotein concentration and development of ischaemic heart-disease, Lancet, i (1975) 16. Aursnes, LA., Darum, H.P., Smith, P., Amesen, H., Christiansen, E.N., Norum, K.R., Fischer, S. and Weber, P.C., Low and high risk coronary patients discriminated by blood platelet fatty acid composition, Scand. J. Clin. Lab. Invest., 46 (1986) 115. Devitt, M.M. and Headon, D.R., Fatty acyl composition of the major lipid classes of HDL, and HDL, of human serum, Clin. Chim. Acta, 155 (1986) 309. 9 Borresen, A.L. and Berg, K. Presence of ‘free’ apo A-I in serum: Implications for immunological quantification of HDL and its apoproteins, Artery, 7 (1980) 139. 10 Karathanasis, SK., Norum, R.A., Zannis, V.I. and Breslow, J.L., An inherited polymorphism in the human apoliprotein

224 A-l gene locus related to the development of atherosclerosis, Nature, 301 (1983) 718. 11 Berg, K., Genetics of coronary heart disease. In: A.G. Steinberg, A.G. Beam, A.G. Motulsky and B. Childs (Eds.), Progress in Medical Genetics, Vol. 5, W.B. Saunders Company, Philadelphia, PA, 1983.

12 Carlson, L.A. and B&tiger, L.E., Risk factors for ischaemic heart disease in men and women. Results of the 19-year follow-up of the Stockholm Prospective Study, Acta Med. Stand., 218 (1985) 207.