Relations Between Plasma Lipids and Postheparin Plasma Lipases and VLDL and LDL Subfraction Patterns in Normolipemic Men and Women
Abstract VLDL1, VLDL2, IDL, and LDL and its subfractions (LDL-I, LDL-II, and LDL-III) were quantified in 304 normolipemic subjects together with postheparin plasma lipase activities, waist/hip ratio, fasting insulin, and glucose. Concentrations of VLDL1 and VLDL2 rose as plasma triglycerides (TGs) increased across the normal range, but the association of plasma TGs with VLDL1 showed a steeper slope than that of VLDL2 (P<.001). Plasma TG level was the most important determinant of LDL subfraction distribution. The least dense species, LDL-I, decreased as the level of this plasma lipid rose in the population. LDL-II in both men and women exhibited a positive association with plasma TG level in the range 0.5 to 1.3 mmol/L, increasing from about 100 to 200 mg/dL. In contrast, within this TG range the LDL-III concentration was low (≈30 mg/dL) and changed little. As plasma TGs rose from 1.3 to 3.0 mmol/L there was a significant fall in LDL-II concentration in men (r=−.45, P<.001) but not in women (r=−.1, NS). Conversely, above the TG threshold of 1.3 mmol/L there was a steeper rise in LDL-III concentrations in men than in women (P<.001); 42% of the men had an LDL-III in the range associated with high risk of heart disease (>100 mg lipoprotein/dL plasma) compared with only 17% of the women. Other influences on the LDL subfraction profile were the activities of lipases and parameters indicative of the presence of insulin resistance. Men on average had twice the hepatic lipase activity of women. This enzyme was not strongly associated with variation in the LDL subfraction profile in men, but in women it was correlated with LDL-III (r=.39, P=.001) and remained a significant predictor in multivariate analysis. Increased waist/hip ratio, fasting insulin, and glucose were correlated negatively with LDL-I and positively with LDL-III, primarily, at least in the case of LDL-III, through raising plasma TGs. On the basis of these cross-sectional observations we postulate the following model for the generation of LDL-III. Subjects develop elevated levels of large TG-rich VLDL1 for a number of reasons, including failure of insulin action. The increase in the concentration of VLDL1 expands the plasma TG pool, and this, via the action of cholesteryl ester transfer protein (which facilitates neutral lipid exchange between lipoprotein particles), promotes the net transfer of TGs into LDL-II, the major LDL species. A hepatic lipase activity in the male range (due possibly to androgen/estrogen imbalance in women) is then required to lipolyze TG-enriched LDL-II and to generate a concentration of small, dense LDL-III that exceeds the risk limit of 100 mg/dL.
- Received February 27, 1995.
- Accepted August 21, 1995.
Plasma lipoproteins are known to play a causative role in atherosclerosis and its clinical manifestation, CHD. Raised plasma cholesterol levels, particularly in the LDL fraction, correlate positively with the incidence of CHD, whereas HDL displays a negative association. The risk associated with elevations in plasma TGs, which are carried mainly in VLDL, remains controversial. However, it is increasingly clear that one way in which this plasma lipid relates to CHD is through its influence on the structure of LDL and HDL, in particular its relation to the subfraction distribution within these density classes.1 2 3 Recent advances in methods for separating and quantifying lipoprotein subclasses have improved our understanding of the role of lipoproteins in the pathophysiology of CHD. The VLDL fraction can be divided into at least two components of differing size, density, and metabolic properties.4 5 LDL is thought to consist of at least three to seven subpopulations,1 6 7 and IDL has been reported to contain larger IDL1 and smaller IDL2 particles.8 The heterogeneity observed within these density intervals may be linked since the conversion of VLDL to LDL does not occur via a single delipidation chain. Rather, parallel processing pathways operate within the delipidation cascade.9 10 For example, when delipidated, large VLDL (VLDL1, 60 to 400 Svedberg flotation units) give rise to remnants in the smaller VLDL2 (Sf 20 to 60) and IDL (Sf 12 to 20) density intervals; these are inefficiently converted to a class of LDL (Sf 0 to 12) that is cleared slowly from the plasma. Newly synthesized small VLDL (VLDL2), on the other hand, is rapidly and almost quantitatively delipidated to LDL, which is catabolized relatively rapidly.9
Metabolic studies of LDL have been complemented in recent years by examination of its subfraction distribution by the high-resolution techniques of gradient gel electrophoresis6 and density-gradient ultracentrifugation.7 Austin et al1 11 have demonstrated an association between a particular LDL phenotype and CHD risk. The atherogenic lipoprotein phenotype, defined by the predominance of small, dense LDL (“pattern B”) and moderately elevated plasma TG and low HDL-C levels, is associated with a threefold increase in CHD risk.1 11 By using the density-gradient technique to quantify individual LDL species, we have translated the atherogenic lipoprotein phenotype into a plasma concentration of LDL-III (d=1.045 to 1.065 kg/L) of greater than 100 mg lipoprotein/dL plasma12 and have shown that this, when present, gives a sevenfold increased risk in a case-control study of myocardial infarct survivors versus normal subjects. The atherogenic lipoprotein phenotype is also linked to the presence of insulin resistance as an underlying metabolic disorder.13 The insulin-resistance syndrome, which is characterized by fasting hyperinsulinemia and exaggerated insulin response to a glucose challenge,14 is considered to be an independent risk factor for CHD. Furthermore, insulin-resistant individuals, especially those with frank non–insulin-dependent diabetes mellitus, have alterations in VLDL structure and metabolism, with an abundance of the larger, TG-rich VLDL species.15 Insulin resistance is associated with an increase in central obesity as indicated by increased BMI and WHR. However, the precise effects of changes in anthropometric indices on plasma lipoprotein subfraction distributions are unknown, although they have been linked to an LDL pattern dominated by small, dense LDL.16 Epidemiological studies have further shown that atherogenic lipoprotein phenotype expression is age and sex dependent.17
In our study of a large group of subjects whose plasma lipid values spanned the normal range, we sought to understand the relations between apoB-containing lipoproteins in terms of their overall concentration, composition, and subfraction distribution. These parameters were also related to total plasma lipid levels, anthropometric indices, and gender. Our principal hypothesis was that the presence of elevated plasma TGs (possibly associated with insulin resistance) and the activity of lipolytic enzymes were together responsible for the accumulation of small, dense LDL. Further, we examined the extent to which the male-female difference in CHD risk could be explained by the variation in lipoprotein subfraction profiles.
Three hundred four healthy volunteers (140 men and 164 women) aged 18 to 69 years were recruited from various sources. These included staff at the Glasgow Royal Infirmary, responders to newspaper advertisements, and healthy family members of patients who had had a coronary bypass grafting operation. All had no history or symptoms of CHD or diabetes mellitus. There was also no known history of dyslipidemia or treatment for dyslipidemia; serum urea and electrolytes were measured to exclude those with renal impairment. The study was approved by the Research Ethics Committee of the Glasgow Royal Infirmary, and each subject gave written informed consent.
Blood samples were obtained from volunteers after a 12-hour overnight fast. Blood was collected by venipuncture with potassium-EDTA (final concentration, 1 mg/mL) as anticoagulant. Plasma was harvested at 4°C by low-speed centrifugation, and aliquots for lipid and lipoprotein measurements and LDL subfractionation were used immediately. Those for insulin assay were snap-frozen at −20°C and analyzed in batches at a later time. Of the 304 volunteers, 137 underwent measurement of fasting postheparin plasma lipases. Ten minutes after the administration of 70 U heparin/kg body wt IV, 10 mL blood was collected into lithium-heparin tubes on ice. Plasma obtained from these samples was frozen immediately and stored at −70°C prior to analysis. In 191 volunteers height, weight, blood pressure, waist, and hip measurements were taken. BMI was calculated as the weight (in kilograms) divided by the height (in meters) squared, and the WHR was also determined. Waist was defined as the smallest circumference between the rib cage and iliac crest and hip as the largest circumference between waist and thigh taken in the standing position.18 Blood pressure was the average of two readings taken in the sitting position from the right arm after a 15-minute rest. Korotkov phase V was taken as the diastolic pressure.
Plasma Lipids, Lipoproteins, and Lipoprotein Fractionation
Plasma cholesterol, TG, HDL-C, VLDL-C, and LDL-C measurements were performed by a modification of the standard Lipid Research Clinics protocol.19 Total LDL was isolated by preparative ultracentrifugation at d=1.019 to 1.063 kg/L in a fixed-angle rotor (50.4 Ti; Beckman Industries Inc). VLDL1 (Sf 60 to 400), VLDL2 (Sf 20 to 60), IDL (Sf 12 to 20), and LDL (Sf 0 to 12) were prepared from plasma by a modification of the cumulative-gradient centrifugation technique described by Lindgren et al.20 The TG, free cholesterol, cholesteryl ester (calculated as the total minus free cholesterol content multiplied by 1.68 to account for the mass of the fatty acid), phospholipid, and protein contents of the lipoproteins were assayed,21 and lipoprotein concentrations were calculated as the sum of these components (expressed as milligrams per deciliter of plasma). Recoveries of the four apoB-containing lipoproteins based on the sum of their cholesterol content compared with the “non–HDL-C” measured by standard methods19 exceeded 85% in both men and women. LDL subfractions were isolated from fresh plasma by nonequilibrium density-gradient ultracentrifugation7 by using a six-step, curvilinear salt gradient. Following centrifugation for 24 hours at 40 000 rpm and 23°C in a swinging-bucket rotor (SW40; Beckman Industries Inc), the tube contents were eluted by upward displacement, and the presence of LDL fractions was detected by continuous monitoring at 280 nm. Three distinct subfractions were resolved in almost all subjects. The subfraction areas under the concentration curve were quantified (Data Graphics; Beckman Industries Inc), corrected for differences in extinction coefficient, and expressed as percentage of total LDL.7 The value for total LDL (d=1.019 to 1.063 kg/L) lipoprotein mass (free cholesterol+TG+cholesteryl ester+ phospholipid+protein), determined as above, was then used to generate individual subfraction concentrations in milligrams of lipoprotein per deciliter of plasma. The lipoprotein core-to-coat ratio was calculated as the ratio of the sum of “core” components (mass of TGs and cholesteryl esters) to the sum of “coat” constituents (mass of phospholipid, free cholesterol, and protein). This provided an index of particle size given that all lipoproteins possess a pseudomicellar spherical structure.22 Lipoprotein(a), the density interval of which overlaps that of LDL-III, was not found to contaminate significantly the LDL-III fraction. When gradient fractions were assayed by enzyme-linked immunosorbent assay (Innotest SA) in normolipidemic subjects with a wide range of lipoprotein(a) levels, ie, up to 150 mg/dL, less than 6% of the lipoprotein(a) was found in the LDL-III fraction. Lipoprotein(a) eluted from the gradient after the LDL-III peak.
Postheparin HL and LPL Assays
LPL and HL were assayed in postheparin plasma.21 Gum arabic–stabilized TG emulsion containing glycerol tri[1-14C]oleate at a specific activity of 30 μCi/mmol TG fatty acid23 was used as substrate. For the specific assay of LPL, plasma was preincubated with sodium dodecyl sulfate to inhibit HL, and pooled preheparin plasma was added to the incubation as a source of cofactor apoC-II to activate the LPL enzyme. HL was assayed in the presence of 1.0 mol/L NaCl to inactivate LPL.23 Enzyme activities are expressed in micromoles of fatty acids released per hour per milliliter of plasma.
Plasma insulin was measured by an immunoradiometric assay with two antibodies. The 125I-labeled monoclonal anti-insulin antibody was prepared from mouse hybridoma cells (Scottish Antibody Production Unit), and the solid-phase guinea pig anti-insulin antibody was coupled to Sepharose gel (Scottish Antibody Production Unit). Plasma samples were thawed immediately prior to analysis. The assay had a working range of 0.5 to 50 mU/L with an intra-assay coefficient of variation of <10% and an interassay variation of 8% to 14%. The assay showed significant cross-reaction with intact human proinsulin (60%).
Statistical analysis and manipulations were performed by using the personal computer version of minitab release 9 for Windows (Minitab Inc). All variables were assessed by drawing normality plots by using the minitab routine. The Anderson-Darling test (P<.05) for deviation from normality was employed, and in instances in which values were not normally distributed, appropriate transformations were performed to obtain a normal distribution. The following factors were subjected to loge transformation: BMI, WHR, and plasma TG, HDL-C, fasting insulin, LPL, HL, VLDL1, VLDL2, LDL-I, LDL-II, and LDL-III concentrations. Plasma VLDL-C was normalized by taking the square root, and fasting glucose was normalized by squaring it. All other variables were examined untransformed. Associations between variables were tested by calculating the Pearson correlation coefficient (r) and the coefficient of determination (r2), which was expressed as a percentage, (ie, r2 gives the percentage of variation in the dependent variable that is explained by variations in the independent variable). The significance of association between pairs of variables was determined by linear regression. Multivariate analysis was employed to determine the extent to which age, sex, general obesity (BMI), central obesity (WHR), and markers of insulin resistance (fasting glucose and insulin) explained the variability in postheparin plasma LPL and HL activity and VLDL1, VLDL2, IDL, LDL, and LDL subfraction concentrations. This was conducted by using the GLM ANOVA in minitab, which permitted the inclusion of categorical variables (ie, sex), and multiple regression, which generated an overall coefficient of determination for a given set of variables. It should be noted that the GLM gives r2 values that relate to the independent contribution of variables whereas the overall r2 determined by multiple regression is usually higher and takes into account the potential interaction between correlated variables. For consistency the same panel of variables (anthropometric indices, markers of insulin resistance, and lipases) was included as predictors in all GLM models of lipoprotein subfraction distributions. Plasma TG level was added to the models for IDL, LDL, and LDL subfractions to explore previously identified relations.1 2 12 Not all tests were performed on all subjects within the survey. For reasons of practicality, some tests were carried out on limited numbers of subjects, eg, IDL measurements, postheparin plasma lipase measurements, and fasting insulin determinations. The plasma lipid levels, age, and sex distributions of those who had IDL or lipases measured did not differ significantly from the whole group, and thus in each statistical test the maximum available information has been used.
Anthropometric indices and plasma lipid and lipoprotein levels were compared between men and women by using transformed data (where appropriate) and testing with Student’s unpaired t tests. A Bonferroni correction was employed to allow for multiple comparisons. Comparison of slopes for the regression of VLDL1, VLDL2, or LDL-III versus plasma TGs and LDL TGs versus HL in men and women was done by using the pairwise-slopes routine in minitab and tested for significance with the Mann-Whitney U test. There were 18 perimenopausal and postmenopausal women, of whom four were on hormone replacement therapy; the hormonal status of the remaining 14 was unknown. The effect that menopause and hormone replacement therapy have on lipids24 25 26 and lipases is well documented, and hence, in each instance, analyses were performed with and without these 18 subjects to ensure that the menopausal state or hormone replacement therapy did not unduly influence the results. Their inclusion did not affect the overall findings. Ninety-seven of the volunteers came from 19 different families and were related. The results were initially analyzed with all related family members excluded and then again with the 97 included. The data were not skewed by the inclusion of these family members, and hence they are present in the final analyses.
Anthropometric Indices and Plasma Lipids and Lipases
Age, systolic blood pressure, BMI, hip circumference, plasma cholesterol, LDL-C, and VLDL-C were similar between the sexes (Table 1⇓). The men were generally heavier, but this was in keeping with their height since there was no sex difference in BMI. However, men had a significantly greater waist circumference and consequently a bigger WHR. They also had higher plasma TGs, plasma cholesterol/HDL ratio, and lower total HDL-C. The mean postheparin HL activity was twice as high in men, but there was no difference between men and women in LPL activity. Age, BMI, WHR, and plasma TGs showed significant positive associations with plasma cholesterol and LDL-C by univariate analysis of this normolipemic group of men and women (data not shown). Multivariate analysis revealed that only age and plasma TGs were significant independent predictors of LDL-C. Variation in plasma TG levels in the population correlated most strongly with BMI, WHR, fasting insulin, and LPL, and of these age, LPL and fasting insulin remained significant predictors in multivariate analysis, accounting for 20% of the variability in this lipid (data not shown). BMI, WHR, fasting insulin, LPL, HL, and plasma TGs all showed significant relationships with HDL-C in univariate analysis, but in a multivariate model containing these variables only fasting insulin, LPL, HL, and plasma TGs remained significant predictors (fasting insulin, plasma TGs, and HL were negative; LPL was positive), and together they accounted for 39% of its variability. Postheparin HL activity had significant associations with BMI, WHR, and fasting insulin in univariate analysis (Table 2⇓). By using the GLM, we ascertained that sex alone accounted for 28.5% of the variability in HL. When included in a model with age, measures of obesity, and insulin resistance, sex was still the most significant predictor (GLM [A], Table 2⇓). The multivariate model (GLM [B]), in which sex was excluded to reveal the effects of other potential factors, indicated that age and WHR were significant predictors that together accounted for 23% of HL variability. BMI and fasting insulin had significant associations with LPL activity in univariate analysis, but none were significant in the multivariate model.
Chemical Compositions of ApoB-Containing Lipoproteins
The percentage contribution of protein to lipoprotein mass (ie, grams per 100 g lipoprotein) increased steadily from 13.8±0.7% in VLDL1 to 16.6±0.5% in VLDL2, 19.8±0.3% in IDL, and 24.5±0.6% in LDL for both men and women. Relative TG content, on the other hand, decreased 10-fold, from 59.4±1.3% in VLDL1 to 38.0±1.0% in VLDL2, 12.8±0.7% in IDL, and 4.9±0.1% in LDL. The percentage of cholesteryl ester increased steadily across the lipoproteins (11.8±0.7% in VLDL1, 24.4±0.9% in VLDL2, and 32.4±0.8% in LDL), but it was maximal in IDL (43.4±0.7%); it should be noted, however, that IDL compositions were performed on only half the subjects. Free cholesterol showed a similar pattern, with a consistent increase from VLDL1 (1.3±0.2%) through VLDL2 (2.9±0.3%) and IDL (6.0±0.3%) to LDL (11.2±0.4%). Phospholipid content was relatively consistent across the lipoproteins (14.7±1.5% in VLDL1, 18.9±1.2% in VLDL2, 18.1±0.5% in IDL, and 27.2±1.3% in LDL). These data are in good agreement with previous findings.9 Significant male-female differences were found in VLDL1 phospholipid (17.7±1.9% in men versus 11.6±1.0% in women, P<.01) and VLDL2 phospholipid (21.6±1.4% in men versus 16.2±0.8% in women, P<.001) contents, but only VLDL2 phospholipid remained significantly different (P<.05) after Bonferroni correction.
Regulation of VLDL Subfraction Composition and Distribution
As plasma TG levels rose across the normal range (Fig 1⇓), both VLDL1 and VLDL2 total lipoprotein concentrations increased (r=.81, P<.001 and r=.57, P<.001 for plasma TGs versus VLDL1 and VLDL2, respectively), but the increment in VLDL1 was greater than that of VLDL2 as tested by comparison of the regression slopes (VLDL1=80.4 TG−37.9 versus VLDL2=26.3 TG+13.5, P<.001). The relationships were similar in both men and women with the ratio of VLDL1 to VLDL2 total lipoprotein rising from ≈1.0 at 0.5 mmol/L to ≈2.0 at 2.0 mmol/L plasma TGs.
In univariate analysis, measures of obesity (BMI and WHR) as well as fasting glucose and insulin levels, LPL, and HL were significantly associated with VLDL1 concentrations (Table 3⇓). Similarly, the same variables with the exception of HL also had significant relationships with VLDL2, but with the addition of age as a further factor. In multivariate analysis only HL was a significant predictor of VLDL1, while age and LPL were the significant predictors of VLDL2 concentration (Table 3⇓).
Examination of VLDL1 and VLDL2 compositions and calculation of the core-to-coat ratio in the lipoprotein particle indicated that the content and dimensions of the VLDL1 and VLDL2 particles remained relatively constant across the plasma TG range. Linear regression of core-to-coat ratios versus plasma TGs revealed the following regression equations: VLDL1 core-to-coat, 2.74−0.4 TG (r=.14, NS) and VLDL2 core-to-coat, 1.37−0.1 TG (r=.16, P<.05). This implied that the increase in VLDL TGs as well as VLDL mass was the result of increasing numbers of particles and not particle size. Likewise, the cholesterol-to-protein ratio in these VLDL subfractions was unchanged throughout the plasma TG range. Variation in the level of plasma cholesterol in this subject group was not associated with changes in the concentrations of VLDL1 and VLDL2 (data not shown).
Regulation of IDL Composition and Concentration
The plasma concentration of IDL rose as plasma cholesterol increased across the normal range (r=.71, P<.001), and this lipoprotein accounted for about 16% of plasma cholesterol (IDL cholesterol content divided by total plasma cholesterol) (Fig 2⇓). The association between concentration of the lipoprotein and plasma TG levels was weaker (r=.41, P<.001; data not shown). In univariate analysis, age, BMI, WHR, fasting glucose, and plasma TGs were important predictors of IDL concentration (Table 4⇓). However, HL and plasma TGs were the only significant predictors in the multivariate model. The IDL core-to-coat ratio changed little across the range of plasma cholesterol levels seen in the present study (r=−.2, P=.05), but the IDL cholesterol-to-protein ratio showed a steady increase with increasing plasma cholesterol levels (r=.55, P<.001). Further analysis showed that as plasma cholesterol rose, IDL became enriched in cholesteryl esters and depleted in TGs; the cholesteryl ester–to-TG ratio exhibited a significant positive correlation with plasma cholesterol (r=.46, P<.001) (Fig 2⇓). In contrast, the cholesteryl ester–to-TG ratio in VLDL1, VLDL2, and LDL remained constant across the cholesterol range. Again, these relationships were the same in men and women.
Regulation of LDL Concentration, Composition, and Subfraction Distribution
The total LDL concentration in both men and women showed a significant positive relationship with plasma TG levels below ≈1.3 mmol/L (r=.30, P<.01) but no association above that value (r=.17, NS). (A value of 1.3 mmol/L is used throughout this article as the plasma TG threshold at which LDL-III concentrations begin to rise significantly. It is the approximate plasma TG level at which LDL-II reaches a maximum in men.) There was also, predictably, a strong positive association between plasma cholesterol and total LDL concentration (r=.73, P<.001) (data not shown). Age, BMI, WHR, and fasting insulin were significant correlates of total LDL concentration in univariate analyses (Table 4⇑), but only age and plasma TGs remained significant predictors in multivariate analysis. There was little variation in LDL composition as plasma lipid levels changed. The cholesterol-to-protein ratio (0.81±0.01) and the core-to-coat ratio (0.46±0.01) remained constant across the plasma cholesterol range in contrast to the changes seen in IDL. The percent TG content in LDL was inversely correlated with HL activity in women (r=−.48, P<.001), but this association in men was less strong (r=−.29, P<.05) (Fig 3⇓). A comparison of the regression slopes provided further support of the male-female differences (LDL TG=6.0−0.04 HL in men versus LDL TG=8.0−0.1 HL in women, P<.001).
Age, BMI, WHR, plasma TGs, fasting glucose, fasting insulin, LPL, and HL all exhibited associations with LDL-I plasma concentration in univariate analysis (Table 5⇓). In the multivariate model, age, sex, LPL, plasma TGs, and fasting glucose were significant predictors, with the latter being the most important. When all parameters in Table 5⇓ were included in a multiple regression analysis, they accounted for 26% of the variability in LDL-I. LDL-II exhibited significant correlations with age, BMI, and plasma TGs, but none of these remained significant on multivariate analysis. BMI, WHR, plasma TGs, and HL exhibited the strongest relationships with LDL-III in univariate analysis. Other factors of significance included age, fasting glucose, fasting insulin, and LPL. In the multivariate model GLM (A), sex and plasma TGs were the only significant independent predictors. When these and the other parameters in Table 5⇓ were included in multiple regression analysis, 42% of LDL-III variability was accounted for. When sex was left out of the model (GLM [B]), HL and plasma TGs were significant predictors of LDL-III; this reduced group of parameters still explained 39% of the variability in this subfraction. In an alternative approach to estimating the effect of plasma TGs and HL on LDL-III levels, the sexes were divided, and separate multiple regression analyses were performed with just these two variables. In men plasma TGs alone explained 40% of LDL-III variability (P<.0001), and HL was not a significant predictor. In contrast, in women both plasma TGs and HL were significant predictors (P=.002 and P<.0001, respectively) that together explained 37% of LDL-III variation. The sex difference was further examined by relating HL activity to LDL-III concentrations adjusted for plasma TG level; HL activity was an important correlate in women but not in men (r=.38, P<.001 and r=.05, NS, respectively).
Plasma TG level was, therefore, the most important determinant of the LDL subfraction profile in both univariate and multivariate analysis. When the changes in individual LDL subfraction distributions across the plasma TG range were examined for the whole subject group, LDL-I exhibited a decrease from mean concentration values of 100 mg/dL at a plasma TG level of 0.5 mmol/L to 40 mg/dL at a plasma TG level or 2.3 mmol/L (data not shown). LDL-II showed a positive relationship (r=.47, P<.001) at TG levels below ≈1.3 mmol/L and a negative relationship (r=−.23, P<.001) above this value.12 However, a gender difference was noted in these associations. Specifically, the LDL-II concentration in women showed no relationship (r=−.1, NS) with plasma TGs <1.3 mmol/L, but a significant negative correlation was demonstrated in men (r=−.45, P<.001) (Fig 4⇓). The concentration of LDL-III was generally low, with a mean of 30 mg/dL in the range of plasma TG levels from 0.5 to 1.3 mmol/L. Above the latter value there was a dramatic rise in LDL-III concentration in men but a smaller increase in women (Fig 4⇓). Linear regression analysis in subjects with plasma TGs >1.3 mmol/L generated equations for LDL-III concentration versus plasma TGs of LDL-III=119 TG−66.6, r=.63, in men and LDL-III=47.4 TG−11.0, r=.49, in women; the slopes of these lines differed significantly (P<.001).
Age was not an entry criterion for this study, but due to the nature of the recruitment process the majority of participants were young to middle-aged adults. Only a few men and postmenopausal women were over the age of 55 years. The two sexes were well matched for age and exhibited the expected differences in body habitus. It was notable that while BMI was virtually identical in both groups, WHR was higher in the men, ie, they had increased central obesity. Plasma lipids were comparable, with the lower LDL-C level in women being balanced by the higher HDL level. HL activity showed a marked sex difference, whereas LPL activity did not. A number of questions were asked of the data. First, as total plasma cholesterol and TG levels varied across the range of values seen in the subjects, how did the VLDL1, VLDL2, IDL, and LDL concentrations alter? Second, what interrelationships existed between lipases and VLDL and LDL subfractions? Third, to what extent were changes in lipoprotein levels related to insulin resistance as revealed by the presence of central obesity and/or raised fasting insulin and glucose concentrations? Fourth, did the above associations depend on gender?
Increasing plasma cholesterol from the lower to upper quintiles of the distribution (3.5 to 7.0 mmol/L) was associated with a twofold rise in plasma LDL concentration and a similar increment in IDL. No relation was seen with VLDL1 and VLDL2 concentrations. The rise in LDL from low normal to high normal was not accompanied by any perturbation in the composition of the particle, either in the core-to-coat or cholesterol-to-protein ratios. That is, the increase was attributable to greater number of particles in the circulation. In contrast, we found that IDL exhibited distinct compositional variation across the normal plasma cholesterol range. As plasma cholesterol rose, so did the cholesteryl ester–TG ratio, indicating a change in the makeup of the particle core. Either IDL particles as a whole were being increasingly modified by lipid exchange (as described for LDL and HDL2 3 27 ) as their concentration increased in the circulation (possibly because of their longer residence time), or IDL is heterogeneous and composed of two subfractions, one of which is TG rich and the other cholesterol rich.8 If the latter situation is the case, it can be postulated that at higher plasma cholesterol levels there is a selective accumulation of the cholesterol-enriched IDL particles.
As plasma TG levels rose across the range seen in this cross-sectional survey, the plasma concentration of VLDL1 increased more rapidly than that of VLDL2. Closer examination of the scattergrams suggested that while VLDL1 and VLDL2 both increased at the same rate over the plasma TG range (0.5 to 1.0 mmol/L), above this value VLDL1 continued to rise, but VLDL2 showed a smaller increment. Kinetic studies28 of VLDL TG turnover have revealed that it is the synthesis and secretion of the lipid rather than its clearance rate that controls plasma levels in normal subjects. Thus, we interpret our observations to indicate that in subjects with high-normal plasma TG levels the liver, in response to an increased abundance of intracellular TGs, releases larger VLDL1 in preference to smaller VLDL2. This allows the transport of more lipid per lipoprotein particle. The total LDL concentration in this and an earlier study12 exhibited a relationship with plasma TGs that appeared biphasic in that a positive correlation was seen at lower plasma TG levels, but above 1.3 mmol/L the association was lost, suggesting that a plateau value had been reached. These relationships between plasma TGs and VLDL1, VLDL2, and LDL concentrations were present in both men and women, and we could not detect any significant gender-based differences in the correlation coefficients. What did differ between the sexes, however, was the LDL subfraction distribution as a function of plasma TG concentration.
LDL-I concentration was influenced by a number of factors including age, features of insulin resistance (increased fasting insulin and glucose), increased obesity, plasma TGs, and HL and LPL activities. In multivariate analysis fasting glucose was the most important predictor followed by age, plasma TGs, LPL, and WHR. The metabolic mechanisms underlying these associations are not yet clear. It is possible that the negative association with plasma TGs is due to cholesteryl ester transfer protein–mediated transfer of TGs from VLDL to the larger LDL species, which would improve their potential as a substrate for HL, similar to the situation for large HDL.2 3 27 The enzyme then lipolyzes the particle to denser particles (LDL-II or LDL-III). LPL activity, on the other hand, demonstrated a positive association with LDL-I, possibly because when LPL activity is high, the circulating levels of large TG-rich lipoproteins (including chylomicrons, which were not represented in the fasting TG measurement) are kept low,2 3 27 thus limiting TG exchange into LDL-I. Alternatively, high LPL activity may favor the rapid conversion of smaller VLDL particles to LDL-I. The effect of factors associated with insulin resistance, ie, higher plasma glucose levels and WHR, which was present in both sexes is, as yet, inexplicable.
LDL-II in both men and women exhibited a positive association with plasma TG level when the latter was <1.3 mmol/L12 (this value is the zenith of the quadratic regression line used to fit LDL-II to plasma TGs in men). At higher plasma TG levels (1.3 to 3.0 mmol/L) in men, there was a decrease in LDL-II12 (with a significant negative correlation), and LDL-III, which had been low when plasma TGs were <1.3 mmol/L, rose steeply. These data led us to postulate that the formation of LDL-III, possibly by remodeling of LDL-II, is increasingly favored in men as the plasma TG level rises above the 1.3 mmol/L threshold. In women the picture was different. At TG >1.3 mmol/L LDL-II did not fall, and little LDL-III was formed. Only 17% of women with plasma TGs above the threshold had an LDL-III >100 mg lipoprotein/dL plasma (the level at which significant CHD risk occurs12 ) compared with 42% in men.
The sex difference and the findings in Table 5⇑ indicate that HL may play an important role in determining the relative concentrations of LDL-II and LDL-III in plasma. HL activity is twice as high in men as women, and when LDL-III in the whole group was adjusted for variation in plasma TGs, HL activity emerged as the strongest predictor of LDL-III concentration. Generally, a combination of plasma TG >1.3 mmol/L and HL >15 U/L was required to generate LDL-III above the risk level of 100 mg lipoprotein/dL plasma, at least in the subjects we examined. Further support for the pivotal role of HL activity in influencing LDL structure can be gained from examining the composition of total LDL. In agreement with the earlier observation of Karpe et al29 on the TG content of “light LDL” (d=1.019 to 1.040 kg/L), percent LDL TG showed a significant inverse correlation with HL activity that was stronger in women than in men, indicating that in women HL activity was important in determining the TG content of LDL; women with low HL had relatively TG-enriched LDL. HL is also known to strongly influence HDL-C and HDL2 levels,27 with high HL activity associated with a low HDL2 concentration. These putative HL-driven changes in LDL and HDL subfraction distributions go a long way to explaining the differences in CHD risk between the sexes, with women having less “atherogenic” LDL-III11 12 and more “cardioprotective” HDL2.27 It is likely that the same mechanism that causes the generation of smaller, denser particles within HDL27 operates in the LDL density range. That is, the lipoprotein is first made susceptible to the action of HL by cholesteryl ester transfer protein–mediated TG transfer; we postulate that it is the VLDL1 concentration that determines the rate of TG transfer into LDL since large TG-rich VLDL is a preferred substrate for cholesteryl ester transfer protein action.30 HL then acts on LDL-II to hydrolyze the TG-enriched core and generate LDL-III. On the other hand, if the enzyme activity is low, then LDL-II remains the major species in plasma and is relatively TG enriched (as suggested by the data in Fig 3⇑). HL is regulated by sex steroid hormones, and so this final step is possibly a function of androgen/estrogen status between and within the sexes. The present findings amplify and set in context the earlier reports of Zambon et al,31 Jansen et al,32 and Watson et al21 on the role of HL in determining LDL subfraction distribution. The importance of HL becomes clear only when the subject groups exhibit a wide range for its activity, eg, when both sexes are examined.21 The previous data of Zambon et al and Jansen et al were collected in men. Our findings agree with the latter study, in which no overall independent influence of HL on LDL subfraction pattern was reported, whereas in the former report a sufficient number of subjects with a low HL were included to see the effect.
The syndrome of insulin resistance has been linked to the appearance of raised plasma TGs, lower HDL-C, and the presence of small LDL in gel electrophoresis patterns.13 In the present study we found that features of insulin resistance (increased WHR and raised fasting insulin or glucose) were correlated with LDL subfraction concentrations, particularly LDL-I (negatively) and LDL-III (positively). On the basis of the argument outlined above, it can be hypothesized that in subjects with insulin resistance the change in the LDL subfraction profile is due mainly to an elevation in plasma TG levels above the 1.3-mmol/L threshold. It is predictable that VLDL1 is specifically increased in the disorder since normally insulin acts to inhibit hepatic TG release in the form of large VLDL,15 and subjects with non–insulin-dependent diabetes mellitus have a preponderance of large, TG-rich VLDL.15 33 Furthermore, we suggest that in many subjects the generation of small, dense LDL may be the result of two hormonal effects: first, the failure of insulin action, which leads to higher VLDL1 levels, and as a consequence of this, increased neutral lipid exchange with LDL-II; second, high androgen activity, which stimulates HL activity and the conversion of TG-enriched LDL-II to LDL-III. It is noteworthy that WHR correlated with HL activity, and that insulin resistance itself and the presence of central obesity can lead to changes in androgen/estrogen balance.34
Selected Abbreviations and Acronyms
|BMI||=||body mass index|
|CHD||=||coronary heart disease|
|GLM||=||general linear model|
This work was supported in part by grant No. 190/1242 from the British Heart Foundation. Dr Tan is the recipient of a Health Manpower Development Plan scholarship from the Ministry of Health of Singapore. Dr Watson was supported by a fellowship from the Wellcome Trust. The technical assistance of Grace Lindsay, Moira Devine, Philip Stewart, and Elizabeth Murray is gratefully acknowledged. Nancy Thomson provided excellent secretarial help in the preparation of this manuscript.
Austin MA, King M-C, Vranigan KM, Krauss RM. Atherogenic lipoprotein phenotype: a proposed genetic marker for coronary heart disease risk. Circulation. 1990;82:495-506.
Packard CJ, Shepherd J. Metabolic basis of the link between plasma triglycerides and coronary heart disease. In: Born GVR, Schwartz CJ, eds. New Horizons in Coronary Heart Disease. London, England: Current Science; 1993:5.1-5.11.
Deckelbaum RJ, Granot E, Oschry Y, Rose L, Eisenberg S. Plasma triglyceride determines structure-composition in low- and high-density lipoproteins. Arteriosclerosis.. 1984;4:225-231.
Patsch W, Patsch JR, Kostner GM, Saller S, Braunsteiner H. Isolation of subfractions of human very low density lipoproteins by zonal ultracentrifugation. J Biol Chem.. 1978;253:4911-4915.
Kuchinskiene Z, Carlson LA. Composition, concentration and size of low density lipoproteins and of subfractions of very low density lipoproteins from serum of normal men and women. J Lipid Res.. 1982;23:762-769.
Krauss RM, Burke DJ. Identification of multiple subclasses of plasma low density lipoproteins in normal humans. J Lipid Res.. 1982;23:97-104.
Griffin BA, Caslake MJ, Yip B, Tait GW, Packard CJ, Shepherd J. Rapid isolation of low density lipoprotein subfractions from plasma by density gradient ultracentrifugation. Arteriosclerosis.. 1990;83:59-67.
Musliner TA, Giotas C, Krauss RM. Presence of multiple subpopulations of lipoproteins of intermediate density in normal subjects. Arteriosclerosis.. 1986;6:79-87.
Packard CJ, Munro A, Lorimer AR, Gotto AM, Shepherd J. Metabolism of apolipoprotein B in large triglyceride-rich very low density lipoproteins of normal and hypertriglyceridemic subjects. J Clin Invest.. 1984;74:2178-2192.
Demant T, Shepherd J, Packard CJ. Very low density lipoprotein apolipoprotein B metabolism in humans. Wien Klin Wochenschr.. 1988;66:703-712.
Griffin BA, Freeman DJ, Tait GW, Thomson J, Caslake MJ, Packard CJ, Shepherd J. Role of plasma triglyceride in the regulation of plasma low density lipoprotein (LDL) subfractions: relative contribution of small, dense LDL to coronary heart disease. Atherosclerosis.. 1994;106:241-253.
Reaven GM, Chen YDI, Jeppesen J, Maheux P, Krauss RM. Insulin resistance and hyperinsulinaemia in individuals with small dense low density lipoprotein particles. J Clin Invest.. 1993;92:141-146.
Reaven GM. Banting lecture 1988: role of insulin resistance in human disease. Diabetes.. 1988;37:1595-1607.
Taskinen M-R, Packard CJ, Shepherd J. Effect of insulin therapy on metabolic fate of apolipoprotein B-containing lipoproteins in NIDDM. Diabetes.. 1990;39:1017-1027.
Selby JV, Austin MA, Newman B, Zhang D, Quesenberry CP, Mayer EJ, Krauss RM. LDL subclass phenotypes and the insulin-resistance syndrome in women. Circulation. 1993;88:381-387.
Austin MA, King M-C, Vranizan KM, Krauss RM. Atherogenic lipoprotein phenotype: a proposed genetic marker for coronary heart disease risk. Circulation. 1990;82:495-505.
Ashwell M, Chinn S, Stalley S, Garrow JS. Female fat distribution: a simple classification based on two circumference measurements. Int J Obes.. 1982;6:289-302.
Lipid Research Clinics Program. Manual of Laboratory Operations, I: Lipid and Lipoprotein Analysis. Bethesda, Md: National Institutes of Health; 1975:75. US Dept of Health, Education, and Welfare publication NIH 75-628.
Lindgren FT, Jensen LC, Hatch FT. The isolation and quantitation analysis of serum lipoproteins. In: Nelson GJ, ed. Blood Lipid and Lipoproteins: Quantitation, Composition and Metabolism. New York, NY: Wiley-Interscience; 1972:181-274.
Watson TDG, Caslake MJ, Freeman DJ, Griffin BA, Hinnie J, Packard CJ, Shepherd J. Determinants of low-density lipoprotein subfraction distribution and concentrations in young normolipidemic subjects. Arteriosclerosis.. 1994;14:902-910.
Watson TDG, Burns L, Packard CJ, Shepherd J. Selective determination of lipoprotein lipase and hepatic triglyceride lipase in heparinised plasma from horses. Am J Vet Res.. 1992;54:771-775.
Patsch JR, Prasad S, Gotto AM, Patsch W. High density lipoprotein2: relationship of the plasma levels of this lipoprotein species to its composition to the magnitude of postprandial lipemia and to the activities of lipoprotein lipase and hepatic lipase. J Clin Invest.. 1987;80:341-347.
Stalenhoef AFH, Demacker PNM, Lutterman JA, Laar AV. Plasma lipoproteins, apolipoproteins and triglyceride metabolism in familial hypertriglyceridemia. Arteriosclerosis.. 1986;6:387-394.
Eisenberg S. Preferential enrichment of large-sized very low density lipoprotein populations with transferred cholesteryl esters. J Lipid Res.. 1985;26:487-493.
Zambon A, Austin MA, Brown BG, Hokanson JE, Brunzell JD. Effect of hepatic lipase on LDL in normal men and those with coronary artery disease. Arterioscler Thromb.. 1993;13:147-153.