HDL-subpopulation patterns in response to reductions in dietary total and saturated fat intakes in healthy subjects

ABSTRACT

Background: Little information is available about HDL subpopulations during dietary changes.

Objective: The objective was to investigate the effect of reductions in total and saturated fat intakes on HDL subpopulations.

Design: Multiracial, young and elderly men and women (n = 103) participating in the double-blind, randomized DELTA (Dietary Effects on Lipoproteins and Thrombogenic Activities) Study consumed 3 different diets, each for 8 wk: an average American diet (AAD: 34.3% total fat,15.0% saturated fat), the American Heart Association Step I diet (28.6% total fat, 9.0% saturated fat), and a diet low in saturated fat (25.3% total fat, 6.1% saturated fat).

Results: HDL₂-cholesterol concentrations, by differential precipitation, decreased (P < 0.001) in a stepwise fashion after the reduction of total and saturated fat: 0.58 ± 0.21, 0.53 ± 0.19, and 0.48 ± 0.18 mmol/L with the AAD, Step I, and low-fat diets, respectively. HDL₃ cholesterol decreased (P < 0.01) less: 0.76 ± 0.13, 0.73 ± 0.12, and 0.72 ± 0.11 mmol/L with the AAD, Step I, and low-fat diets, respectively. As measured by nondenaturing gradient gel electrophoresis, the larger-size HDL_2b subpopulation decreased with the reduction in dietary fat, and a corresponding relative increase was seen for the smaller-sized HDL_{3a, 3b}, and _3c subpopulations (P < 0.01). HDL₂-cholesterol concentrations correlated negatively with serum triacylglycerol concentrations on all 3 diets: r = −0.46, −0.37, and −0.45 with the AAD, Step I, and low-fat diets, respectively (P < 0.0001). A similar negative correlation was seen for HDL_2b, whereas HDL_{3a, 3b}, and _3c correlated positively with triacylglycerol concentrations. Diet-induced changes in serum triacylglycerol were negatively correlated with changes in HDL₂ and HDL_2b cholesterol.

Conclusions: A reduction in dietary total and saturated fat decreased both large (HDL₂ and HDL_2b) and small, dense HDL subpopulations, although decreases in HDL₂ and HDL_2b were most pronounced.

Nutrition, diet, lipoproteins, saturated fat, triacylglycerols, race, women, HDL subpopulations, cardiovascular disease

INTRODUCTION

HDL cholesterol is an important negative risk factor for coronary heart disease (1–6). It is well recognized that HDL-cholesterol concentrations are affected by dietary modifications, generally increasing with diets high in saturated fat and cholesterol and decreasing when dietary fat is replaced with carbohydrates (7–10). The HDL subpopulation is metabolically dynamic and closely related to the metabolism of triacylglycerol-rich lipoproteins (11). Within the spectrum of HDL density, there is considerable particle heterogeneity, which is linked to differences in functional properties. A common technique to classify HDL subpopulations is based on differential precipitation, HDL₂ representing a less dense and HDL₃ a more dense subpopulation (12–15). Other characterizations are based on HDL particle size analyzed by gradient gel electrophoresis, in which 5 subpopulations can be identified (HDL_2b, HDL_2a, HDL_3a, HDL_3b, and HDL_3c), and on HDL apolipoprotein (apo) specificity, with HDL particles containing only apo A-I (Lp A-I) compared with particles containing both apo A-I and apo A-II (Lp A-I/Lp A-II) isolated by immunologic methods (12, 16–19). The classifications of heterogeneity are independent, although somewhat overlapping, eg, HDL₂ contains a higher proportion of Lp A-I than Lp A-I/A-II and corresponds broadly to HDL_2b and HDL_2a. Previous studies indicated that metabolic and cardioprotective properties differ primarily between HDL subpopulations. Thus, within the HDL spectrum, the less dense HDL₂ subpopulation has been linked to the cardioprotective effect of HDL, and this subpopulation also has been found to have a negative relation to serum triacylglycerol concentrations (12, 20–25). However, reductions in HDL₃-cholesterol concentrations also have been associated with cardiovascular risk (26). It has been well established that HDL heterogeneity is affected by drastic variations in dietary fat intake (8, 12), but considerably less information is available about HDL subpopulations during smaller changes in fat intake, more typical of the average food consumption pattern.

In the Dietary Effects on Lipoproteins and Thrombogenic Activities (DELTA) Study, a carefully controlled multicenter feeding study, the effect of a stepwise reduction of total and saturated fats on plasma lipid and lipoprotein patterns in healthy subjects was investigated (27). The diets used were designed to represent variations within a range recommended to the public, and the overall variation in fat intake was, therefore, more modest than in many previous studies. As detailed elsewhere, the study was designed to simultaneously obtain results in men and women, African Americans and non-African Americans, pre- and postmenopausal women, and younger and older men. Overall results on lipids and lipoproteins were published previously (27). Notably, also in this regimen, HDL-cholesterol concentrations decreased and triacylglycerol concentrations increased after a reduction in dietary total and saturated fats. In view of the previously established association between HDL₂ and triacylglycerol (12, 20–25), we hypothesized that a modest, stepwise reduction in total and saturated fat in the diet would influence primarily the larger HDL subpopulations. In the present study, we report on diet-induced changes in HDL-subpopulation distribution and HDL particle size.

SUBJECTS AND METHODS

Subjects

The DELTA Study design and the recruitment strategy were described in detail elsewhere (27). For protocol 1, healthy, normolipidemic subjects between 22 and 67 y of age were recruited. To be eligible for participation, baseline fasting plasma lipid concentrations had to be between the 25th and 90th percentiles for total cholesterol, below the 90th percentile for plasma triacyl-glycerol, and above the 10th percentile for HDL cholesterol—all adjusted for age, race, and sex (28). Each clinical research center (Columbia University, Pennington Biomedical Research Center, Pennsylvania State University, and the University of Minnesota) recruited 25–30 subjects and a total of 118 subjects were recruited. Of these 118 subjects, 103 (57 women and 46 men) completed all 3 feeding periods. The ethnic composition of the subjects was as follows: 77 whites and 26 African Americans. Of the women, 39 were premenopausal; of the men, 30 were <40 y of age.

Study design

The study was designed as a randomized, double-blind, 3-way crossover study, in which each subject consumed 3 different diets, each for 8 wk. Each dietary period was separated by 4–6 wk. The subjects consumed 2 meals each weekday in a cafeteria; meal consumption was supervised and it was ascertained that all food provided was consumed at these meals (27). A third packaged meal and snacks were provided to all participants. Weekend meals, except for one self-selected dinner meal, were also packaged and provided to all participants.

The 3 diets were as follows: 1) an average American diet (AAD; 34.3% of energy as total fat: 15.0% saturated, 12.8% monounsaturated, and 6.5% polyunsaturated), 2) the American Heart Association (AHA) Step I diet (28.6% of energy as total fat: 9.0% saturated, 12.9% monounsaturated, and 6.7% polyunsaturated), and 3) a low-fat diet (25.3% of energy as total fat: 6.1% saturated, 12.4% monounsaturated, and 6.7% polyunsaturated fat). Details of the study design, food and menu preparation, and validation and monitoring of the diets are reported elsewhere (29). The diets were designed to provide ≈300 mg dietary cholesterol/d and the cholesterol content ranged from 267 to 285 mg/d. The diets were isoenergetic and designed to maintain body weight during the study. To ensure that the participants maintained their body weight, they were weighed twice weekly during the study and adjustments in energy intake were made if needed to maintain a stable body weight.

Analytic procedures

Blood samples were obtained weekly during the last 4 wk of each 8-wk dietary period after an overnight fast (27). Plasma and serum were isolated by centrifugation at 30000 × g for 20 min at 4°C immediately after collection. Multiple aliquots of each were stored in cryovials at −80°C until the end of the study when all samples were analyzed. Concentrations of total cholesterol, triacylglycerol, and HDL cholesterol were assayed in the weekly samples at each research center by using standard enzymatic procedures. HDL-cholesterol concentrations were determined after precipitation of apo B–containing lipoproteins (30), and LDL-cholesterol concentrations were calculated by using the Friedewald formula (31). All laboratories participated in the Centers for Disease Control and Prevention Lipid Standardization Program; additional standardization procedures were also used.

HDL-subpopulation analyses were performed at Columbia University with plasma samples obtained during weeks 7 and 8 of all 3 diets by using a modification of the procedure described by Gidez et al (32). We established previously that a steady state was obtained at 4 wk and that in each dietary period, plasma lipid concentrations remained unchanged between 4 and 8 wk (27). Briefly, heparin and magnesium chloride were added to 0.5 mL plasma at a final concentration of 1.26 g/L and 91 mmol/L, respectively. After incubation for 10 min at room temperature, samples were centrifuged at 5000 × g at 4°C in a table-top centrifuge for 1 h. The supernate was recovered and an aliquot obtained for analysis of total HDL-cholesterol concentrations. To 333 μL of the supernate, a 1/10 volume of dextran sulfate (molecular weight of 15000; Genzyme, Cambridge, MA) was added. The mixture was incubated at room temperature for 20 min and thereafter centrifuged for 30 min at 5000 × g at 4°C. The supernate, representing HDL₂, was removed and the cholesterol content was determined. HDL₃ concentrations were calculated as the difference between total HDL and HDL₂-cholesterol concentrations. For quality-control purposes, 3 individual sets of plasma samples, representing a wide variation in the HDL-subpopulation spectrum, were used. Each set of samples was divided into single-use aliquots and stored at −70°C before being used. In each assay, a set of controls was run with a 10-sample interval. The between-run CVs for HDL₂ and HDL₃ were 8% and 7%, respectively, at concentrations of 0.72 and 0.83 mmol/L, respectively.

HDL particle size was determined in 102 subjects at Pennington Biomedical Research Center by nondenaturing gradient gel electrophoresis on samples from weeks 5, 6, 7, and 8 by using a modification of the procedure published by Blanche et al (16). Briefly, concave acrylamide gradient gels (4–30%) were cast 4 or 8 at a time by using a GSC-8 gel-casting apparatus (Pharmacia, Uppsala, Sweden). The cast gels were allowed to polymerize overnight and were thereafter used immediately or stored for ≤1 wk wrapped in moist towels in plastic bags. Plasma samples (8 μL) were electrophoresed in a Pharmacia GE-2/4 electrophoresis apparatus in 90 mmol/L tris, 80 mmol/L boric acid, 3 mmol/L EDTA buffer, pH 8.3. The gels were pre-run for 15 min at 125 V before sample application. The samples were electrophoresed at 70 V for 15 min, followed by 125 V for 24 h. After electrophoresis, the gels were stained with Oil Red O (Sigma Chemical Co, St Louis) and scanned in a model GS-670 imaging densitometer (Bio-Rad Laboratories, Hercules, CA). The gels then were counterstained with Coomassie R-250 (Sigma Chemical Co) and rescanned. The lipid distribution across 5 subpopulations was determined (16). For quality-control purposes, 2 plasma samples representing extreme HDL distributions, stored frozen in single-use aliquots in 20% sucrose, were included in each run. The CVs for HDL₂ (HDL_2b + HDL_2a) and HDL₃ (HDL_3a + HDL_3b + HDL_3c), as estimated by gradient gel electrophoresis, averaged 9.8% and 11.1%, respectively.

Statistical analysis

Linear statistical models were fitted to the lipoprotein data by using the procedure MIXED in the SAS software package (SAS Institute Inc, Cary, NC). Subjects (ID) and diets were class variables. The syntax used for describing the repeated measures and random effects was REPEATED/SUBJECT = ID and RANDOM INT DIET/SUBJECT = ID. The model for each outcome variable included the 3 diets, 4 sex-age groups (men aged <40 y, men aged ≥40 y of age, premenopausal and postmenopausal women), 2 ethnic groups (African Americans and others), 3 apo E genotype groups (E2/x, E3/3, and E4/x), 4 field centers, interactions of all the above nondietary factors with the diets, and the 3 feeding periods. The effects of diet were estimated by weighting each relevant diet term's regression coefficient by the fraction of the study population for whom that diet term applied. For example, because 25 of 103 subjects had the E4/3 or E4/4 (E4/x) apo E genotype, interactions of this factor with the diets were weighted by 0.2427 (25/103). Such weighting, based on observed frequencies, yields effects of diet for the total population that are independent of the model used.

RESULTS

As seen in Table 1, both HDL₂- and HDL₃-cholesterol concentrations decreased in a stepwise fashion as dietary saturated fat and total fat decreased. The lowest concentrations of both HDL subpopulations were observed during the low-fat diet. HDL₂-cholesterol concentrations differed significantly between all 3 diets (Table 2). Changes in HDL₃-cholesterol concentrations were smaller than those of HDL₂: concentrations with the AAD were significantly higher than those with the other 2 diets, whereas there was no significant difference in concentrations between the Step I and low-fat diets. The largest differences for both subpopulations were seen between the low-fat diet and AAD. In absolute terms, the decrease in HDL₂-cholesterol concentrations between these 2 diets was larger than that in HDL₃-cholesterol concentrations: 0.101 compared with 0.044 mmol/L, respectively. These decreases corresponded to 17.2% and 5.7%, respectively, of the HDL₂- and HDL₃-cholesterol concentrations with the AAD. Analysis of HDL size variation by gradient gel electrophoresis showed a similar pattern. The relative amounts of the largest HDL subpopulation, HDL_2b, decreased in a similar stepwise fashion with the decrease in dietary total and saturated fat (Tables 1 and 2). The changes in HDL_2a were less pronounced, but there was a small decrease between the AAD and Step I diets. Because these values are expressed in relative terms, there was a corresponding relative increase in HDL_3a, HDL_3b, and HDL_3c with a decrease in dietary fat intake. A common trend was seen for the 3 different HDL₃ particle size subpopulations, with similar relative increases in HDL_3a, HDL_3b, and HDL_3c, most pronounced for the smallest-sized subpopulation, HDL_3c.

TABLE 1

HDL-cholesterol subpopulation distribution during the dietary intervention¹

	AAD	Step I	Low fat
HDL cholesterol (mmol/L)	1.35 ± 0.29	1.25 ± 0.28	1.19 ± 0.26
HDL subpopulations (mmol/L)
HDL2	0.58 ± 0.21	0.53 ± 0.19	0.48 ± 0.18
HDL3	0.76 ± 0.13	0.73 ± 0.12	0.72 ± 0.11
HDL size (%)
HDL2b	35.0 ± 7.9	33.6 ± 7.9	33.2 ± 8.3
HDL2a	19.9 ± 2.2	19.8 ± 2.3	19.8 ± 2.3
HDL3a	19.5 ± 2.6	19.9 ± 2.7	20.1 ± 2.9
HDL3b	12.8 ± 2.9	13.3 ± 3.1	13.3 ± 3.2
HDL3c	12.8 ± 4.6	13.5 ± 4.8	13.6 ± 4.9

1

x̄ ± SD. HDL-cholesterol concentrations are from reference 27. AAD, average American diet; Step I, American Heart Association diet. To convert values to mg/dL, multiply by 38.7.

TABLE 2

Differences in HDL-cholesterol subpopulation concentrations during the dietary intervention¹

	AAD − Step 1	AAD − low fat	Step I − low fat
HDL subpopulations (mmol/L)
HDL2	−0.056 ± 0.102	−0.101 ± 0.092	−0.045 ± 0.102
HDL3	−0.028 ± 0.082	−0.044 ± 0.083	−0.016 ± 0.07
HDL particle size (%)
HDL2b	−1.41 ± 2.562	−1.77 ± 2.612	−0.36 ± 2.66
HDL2a	−0.16 ± 0.82	− 0.16 ± 0.83	0.00 ± 0.84
HDL3a	0.42 ± 1.002	0.56 ± 1.092	0.14 ± 1.11
HDL3b	0.51 ± 1.032	0.57 ± 1.042	0.07 ± 0.97
HDL3c	0.65 ± 1.563	0.79 ± 1.472	0.14 ± 1.52

1

x ± SD. AAD, average American diet.

2

P < 0.001.

3

P < 0.01.

The DELTA Study recruited subjects from different age, sex, and race groups to allow estimation of possible differences in diet-induced plasma lipid concentrations. As seen in Table 3, men had lower baseline HDL₂-subpopulation concentrations (determined by precipitation) with the AAD than did women. There were no significant differences in HDL₃ between men and women; there were no other significant differences in either HDL₂ or HDL₃ between age groups. Analysis of the HDL-subpopulation size pattern showed significant differences across sex, with more HDL_2b particles and fewer HDL_3a, HDL_3b, and HDL_3c particles in women than in men (Table 4). For the different sex and age groups, HDL_2b was more numerous in premenopausal than in postmenopausal women, whereas other differences were not significant. In this study, we did not find any differences in HDL-subpopulation concentrations between African Americans and non-African Americans, by either differential precipitation or gradient gel electrophoresis. Additionally, overall HDL-cholesterol concentrations did not differ significantly between African Americans and non-African Americans (27).

TABLE 3

HDL-cholesterol subpopulation distribution in subgroups during the dietary intervention¹

	AAD		Step I		Low fat
	HDL2	HDL3	HDL2	HDL3	HDL2	HDL3
			mmol/L
Black (n = 26)	0.59 ± 0.18	0.77 ± 0.15	0.54 ± 0.20	0.75 ± 0.13	0.49 ± 0.16	0.72 ± 0.11
White (n = 77)	0.58 ± 0.22	0.76 ± 0.13	0.53 ± 0.19	0.73 ± 0.12	0.48 ± 0.19	0.71 ± 0.12
Women (n = 57)	0.68 ± 0.202	0.78 ± 0.12	0.61 ± 0.182	0.76 ± 0.13	0.58 ± 0.172	0.74 ± 0.11
Premenopausal (n = 39)	0.70 ± 0.20	0.76 ± 0.11	0.64 ± 0.19	0.75 ± 0.13	0.59 ± 0.18	0.71 ± 0.10
Postmenopausal (n = 18)	0.64 ± 0.19	0.82 ± 0.13	0.56 ± 0.13	0.79 ± 0.13	0.55 ± 0.16	0.78 ± 0.12
Men (n = 46)	0.46 ± 0.15	0.74 ± 0.14	0.43 ± 0.16	0.70 ± 0.11	0.37 ± 0.12	0.69 ± 0.12
<40 y (n = 30)	0.48 ± 0.16	0.76 ± 0.15	0.43 ± 0.14	0.71 ± 0.11	0.38 ± 0.12	0.70 ± 0.13
≥40 y (n = 16)	0.42 ± 0.15	0.70 ± 0.09	0.43 ± 0.20	0.68 ± 0.09	0.34 ± 0.11	0.67 ± 0.09

1

x ± SD. AAD, average American diet.

2

Significantly different from men, P < 0.01.

TABLE 4

HDL size subpopulation distribution in subgroups during the decrease in dietary saturated fat¹

Subgroup	HDL2b	HDL2a	HDL3a	HDL3b	HDL3c
			%
Blacks (n = 26)
AAD	36.9 ± 7.2	20.4 ± 2.1	18.7 ± 2.6	11.7 ± 2.5	12.2 ± 4.5
Step I	35.5 ± 6.4	20.3 ± 2.1	19.1 ± 2.4	12.1 ± 2.3	12.9 ± 4.4
Low fat	35.4 ± 6.7	20.4 ± 2.3	19.0 ± 2.9	12.1 ± 2.2	13.0 ± 4.3
Whites (n = 76)
AAD	34.3 ± 8.1	19.8 ± 2.2	19.8 ± 2.6	13.1 ± 3.0	13.1 ± 4.6
Step I	32.9 ± 8.2	19.6 ± 2.4	20.2 ± 2.7	13.7 ± 3.2	13.7 ± 4.9
Low fat	32.4 ± 8.7	19.5 ± 2.3	20.4 ± 2.8	13.8 ± 3.3	13.9 ± 5.0
Women (n = 57)
AAD	38.4 ± 7.72	20.3 ± 2.1	18.8 ± 2.72	11.5 ± 2.62	11.0 ± 4.02
Step I	37.1 ± 7.52	20.2 ± 2.3	19.3 ± 2.62	11.9 ± 2.72	11.5 ± 4.22
Low fat	36.7 ± 8.12	20.2 ± 2.3	19.4 ± 2.92	12.0 ± 2.82	11.7 ± 4.52
Premenopausal women (n = 39)
AAD	40.1 ± 7.53	20.3 ± 2.0	18.5 ± 2.6	11.0 ± 2.5	10.1 ± 3.4
Step I	38.8 ± 7.3	20.2 ± 2.1	18.9 ± 2.7	11.4 ± 2.5	10.7 ± 3.5
Low fat	38.5 ± 8.2	20.3 ± 2.3	19.0 ± 3.0	11.4 ± 2.7	10.7 ± 3.8
Postmenopausal women (n = 18)
AAD	34.8 ± 7.1	20.2 ± 2.5	19.6 ± 2.6	12.4 ± 2.6	12.9 ± 4.7
Step I	33.3 ± 6.7	20.2 ± 2.7	20.1 ± 2.2	13.0 ± 2.7	13.4 ± 5.1
Low fat	32.8 ± 6.6	20.0 ± 2.3	20.2 ± 2.3	13.1 ± 2.6	13.9 ± 5.1
Men (n = 45)
AAD	30.6 ± 5.9	19.5 ± 2.3	20.3 ± 2.4	14.4 ± 2.5	15.1 ± 4.2
Step I	29.1 ± 5.9	19.3 ± 2.3	20.7 ± 2.5	15.0 ± 2.7	16.0 ± 4.3
Low fat	28.7 ± 6.0	19.2 ± 2.3	20.9 ± 2.7	15.1 ± 2.8	16.0 ± 4.3
Men aged <40 y (n = 29)
AAD	31.3 ± 6.3	20.2 ± 2.1	20.4 ± 2.5	13.8 ± 2.5	14.2 ± 4.5
Step I	29.5 ± 6.3	19.9 ± 2.1	20.8 ± 2.8	14.5 ± 2.7	15.2 ± 4.6
Low fat	29.2 ± 6.6	19.8 ± 2.1	21.1 ± 3.0	14.7 ± 2.8	15.2 ± 4.3
Men aged ≥40 y (n = 16)
AAD	29.3 ± 4.9	18.3 ± 2.1	20.2 ± 2.2	15.4 ± 2.4	16.9 ± 2.8
Step I	28.4 ± 5.1	18.1 ± 2.3	20.5 ± 2.1	15.8 ± 2.5	17.2 ± 3.4
Low fat	27.9 ± 5.1	18.1 ± 2.4	20.5 ± 1.9	15.9 ± 2.5	17.6 ± 3.8

1

x̄ ± SD. AAD, average American diet.

2

Significantly different from men, P < 0.01.

3

Significantly different from postmenopausal, P < 0.01.

The overall response to the dietary changes in HDL₂- and HDL₃-cholesterol concentrations was similar among the subgroups (Table 3); in each subgroup, dietary effects were significant, as for the whole group. In all subgroups, the decrease in HDL₂-cholesterol concentrations was significantly greater than the decrease in HDL₃-cholesterol concentrations (P < 0.001). When expressed in relative terms, decreases in HDL₂-cholesterol concentrations between the low-fat diet and the AAD ranged from 14.6% to 21.3%, being slightly higher in men, independent of age. Corresponding decreases in HDL₃-cholesterol concentrations between the subgroups was small, ranging from 3.7% to 6.4%. The observed changes between diets were similar in all subgroups by gradient gel electrophoresis, with a relative decrease in HDL_2b and a relative increase in HDL_3a, HDL_3b, and HDL_3c (Table 4). Thus, the overall response of HDL subpopulations to decreases in dietary total and saturated fat, on the basis of both differential precipitation and size classification, did not differ significantly among different sex, age, or race groups.

It is well known from many previous studies that total HDL-cholesterol concentrations correlate negatively with serum triacylglycerol concentrations (33). In the present study, there was a significant negative correlation between HDL₂-cholesterol and serum triacylglycerol concentrations on all 3 diets (Figure 1). Because serum triacylglycerol concentrations were non-normally distributed, results were log transformed before analysis. In contrast, HDL₃-cholesterol concentrations did not correlate with serum triacylglycerol concentrations on any of the diets (Figure 2). The results indicate that during each dietary regimen, higher serum triacylglycerol concentrations were associated with lower concentrations of HDL₂ but not HDL₃ cholesterol. However, after adjustment for triacylglycerol concentrations, there was still a significant difference in HDL₂-cholesterol concentrations between the 3 diets (P < 0.0001), showing a specific effect of diet on HDL₂ cholesterol. Overall, only ≈10% of the difference in HDL₂-cholesterol concentrations between the 3 diets could be explained by changes in triacylglycerol concentrations. In contrast, there was no relation of either HDL-subpopulation concentration with LDL-cholesterol or total cholesterol concentrations (data not shown). To investigate whether the diet-induced decreases in HDL-subpopulation concentrations were associated with diet-induced increases in serum triacylglycerol concentrations, individual differences in these indexes were compared. As seen in Figure 3, there was a significant relation between changes in HDL₂-cholesterol concentrations and changes in serum triacylglycerol concentrations among all 3 diets. For HDL₃-cholesterol concentrations, there was a significant correlation with changes in serum triacylglycerol only when the AAD and low-fat diet were compared, representing the overall largest differences (Figure 4). For HDL size subpopulations, we found significant correlations between triacylglycerol concentrations and HDL_2b, HDL_3a, HDL_3b, and HDL_3c with all 3 diets (Table 5). In addition, changes in HDL_2b correlated negatively with changes in triacylglycerol concentrations when the Step I or the low-fat diet was compared with the AAD, whereas there was a positive correlation between changes in HDL_3a and HDL_3b and changes in triacylglycerol. HDL_3c, the smallest HDL subpopulation, correlated weakly with triacylglycerol concentrations on all diets, and changes in HDL_3c did not relate to changes in triacylglycerol. There were virtually no changes in HDL_2a between the diets and there was no relation between HDL_2a and triacylglycerol concentrations on any of the diets. As observed for the HDL₂ and HDL₃ subpopulations, between-diet changes in HDL size subpopulations did not correlate significantly with changes in total or LDL cholesterol.

FIGURE 1.

Correlation between HDL₂-cholesterol and triacylglycerol concentrations for all subjects (n = 103) during the average American diet (AAD), the American Heart Association Step I diet, and a diet with a low total and saturated fat content (low-fat diet). To convert cholesterol and triacylglycerol to mg/dL, multiply by 38.7 and 88.5, respectively.

FIGURE 2.

Correlation between HDL₃-cholesterol and triacylglycerol concentrations for all subjects (n = 103) during the average American diet (AAD), the American Heart Association Step I diet, and a diet with a low total and saturated fat content (low-fat diet). To convert cholesterol and triacylglycerol to mg/dL, multiply by 38.7 and 88.5, respectively.

FIGURE 3.

Correlation between changes (Δ) in HDL₂-cholesterol and log triacylglycerol concentrations for all subjects (n = 103) between the American Heart Association Step I diet and the average American diet (AAD), the diet with a low total and saturated fat content (low-fat diet) and the Step I diet, and the low-fat diet and the AAD.

FIGURE 4.

Correlation between changes (Δ) in HDL₃-cholesterol and log triacylglycerol concentrations for all subjects (n = 103) between the American Heart Association Step I diet and the average American diet (AAD), the diet with a low total and saturated fat content (low-fat diet) and the Step I diet, and the low-fat diet and the AAD.

TABLE 5

Correlation between HDL size subpopulations and serum triacylglycerol concentrations¹

	HDL2b	HDL2a	HDL3a	HDL3b	HDL3c
AAD
r	−0.41	−0.05	0.43	0.43	0.21
P	< 0.001	NS	< 0.001	< 0.001	0.034
Step I
r	−0.41	−0.03	0.36	0.45	0.21
P	< 0.001	NS	< 0.001	< 0.001	0.035
Low fat
r	−0.49	−0.03	0.48	0.50	0.24
P	< 0.001	NS	< 0.001	< 0.001	0.013
Step I vs AAD
r	−0.43	−0.12	0.34	0.39	0.29
P	< 0.001	NS	< 0.001	< 0.001	0.003
Low fat vs AAD
r	−0.39	0.05	0.37	0.35	0.14
P	< 0.001	NS	< 0.001	< 0.001	NS

1

AAD, average American diet.

DISCUSSION

A major finding of the present study was that the dietary response to a reduction in total and saturated fats in the diet varied between different HDL subpopulations. We addressed these effects by measuring 2 different HDL-subpopulation properties: differential precipitation and size. Independent of the method used, the most pronounced response was seen for the less dense, larger HDL subpopulations (HDL₂ and HDL_2b), which decreased with a decrease in total and saturated fat intakes. The concentrations of this larger HDL subpopulation correlated significantly with serum triacylglycerol concentrations during all diets. In addition, the between-diet changes observed for the larger HDL subpopulations correlated significantly with between-diet changes in serum triacylglycerol concentrations. The results showed simultaneous changes in triacylglycerol, HDL_2b, and HDL₂ during dietary modification within the normal food intake range and suggest a close metabolic relation between these subpopulations.

The design of the study offered several important advantages. First, the number of subjects enrolled was larger than in most other metabolic ward studies and offered possibilities to address diversities in race and sex. In addition, the use of modest and realistic dietary changes makes the results relevant to clinical practice. Another strength was the use of 2 different and independent methods to assess HDL-subpopulation changes. Each method addresses unique properties of HDL and the results, therefore, do not necessarily overlap (12). Although concentrations of HDL subpopulations are measured in the stepwise precipitation procedure, lipid staining is used to estimate the relative concentrations of the size subpopulations. Therefore, it was reassuring that the results with stepwise precipitation were corroborated by those with lipid staining. It is important to emphasize that HDL size subpopulation data are presented in relative terms, with an increase in relative concentrations of HDL_3a, HDL_3b, and HDL_3c with the Step I and low-fat diets. When the changes in HDL subpopulations derived by precipitation were expressed in relative terms, there was an increase in the HDL₃ subpopulation from 56.6% to 59.7%. Thus, overall, both procedures gave similar results, showing reductions in the larger, less dense HDL subpopulations in response to the dietary changes.

It has been well established that the higher the HDL-cholesterol concentration, the lower the risk of atherogenesis, and, in epidemiologic studies, low HDL-cholesterol concentrations are a risk factor for coronary heart disease (1–6). HDL-cholesterol concentrations vary widely in the population and are affected by the action of many genetic and metabolic factors (11, 34–37). It is well established that the composition of the diet affects HDL-cholesterol concentrations, with an increase during higher intakes of saturated fat and a decrease when saturated fat is replaced with unsaturated fat or carbohydrates (7). In addition to diet, other environmental factors, such as exercise, body weight, and drug use influence HDL cholesterol (7, 11, 35). Because of this complex regulation of HDL-cholesterol concentrations, the underlying mechanisms for the cardioprotective role of HDL remains to be elucidated. Emerging data during recent years have focused attention on the role of HDL subpopulations. The HDL fraction is heterogeneous, containing particles with various contents of apolipoproteins, cholesterol, triacylglycerol, and phospholipids. Because of the complexity of HDL metabolism and composition, many subpopulation classification systems have been used. Of the HDL subpopulations measured by precipitation (HDL₂ and HDL₃), HDL₂ has been most consistently linked to the antiatherogenic effect of HDL (38–42). This subpopulation has been affected more by large changes in dietary fat intake, whereas the response in HDL₃-cholesterol concentrations has been less pronounced (43–47). However, studies in which there were more modest changes in dietary fat intake have not always shown an effect on HDL₂ (48). In some previous studies, as the replacement of dietary fat with carbohydrate increased, HDL-cholesterol concentrations decreased accordingly, particularly concentrations of the larger HDL subpopulations (7, 49, 50). The carbohydrate-induced decrease in HDL was shown to be associated with changes in apo A-I metabolism (51, 52). However, several of these studies used diets with large variations in fat and carbohydrate contents (8, 12); therefore, it was important to extend these findings to the general population by using more modest and commonly occurring dietary differences.

HDL has a complex metabolic relation to chylomicron, VLDL, and LDL metabolism (35). High triacylglycerol and low HDL-cholesterol concentrations usually occur together clinically and there is an exchange of cholesterol and triacylglycerol between HDL and triacylglycerol-rich lipoproteins in the circulation. In normolipidemic subjects, VLDL concentrations have been reported to be rate-limiting for the net transfer of cholesterol esters from HDL to VLDL, whereas cholesteryl ester transfer protein concentrations are suggested to be more important in hypertriglyceridemia (53, 54). In the present study, serum triacylglycerol concentrations increased with the stepwise replacement of fat with carbohydrate in normolipidemic subjects. This could have occurred in part because of carbohydrate-induced increases in hepatic VLDL output (55). The increased availability of VLDL-associated triacylglycerols could potentially stimulate cholesteryl ester transfer from HDL to VLDL, as suggested by the present decrease in HDL-cholesterol concentrations in parallel with the increase in triacylglycerol concentrations during a reduction in dietary fat intake. This is of particular interest because the subjects were normolipidemic, with modest changes in mean triacylglycerol concentrations between the 3 diets (AAD: 0.96 mmol/L; Step I diet: 1.04 mmol/L; low-fat diet: 1.05 mmol/L); the only significant difference was between the AAD and the other 2 diets (27). HDL₂ cholesterol has shown a stronger correlation with serum triacylglycerol than has HDL₃ cholesterol (20–25) and it is conceivable that an increased exchange between VLDL triacylglycerol and primarily HDL₂ cholesteryl ester with the Step I and low-fat diets could have contributed to the decrease in HDL₂- and HDL_2b-cholesterol concentrations. We found significant correlations not only between triacylglycerol concentrations and HDL₂- and HDL_2b-cholesterol concentrations during all 3 diets, but also between changes in triacylglycerol concentrations and in HDL₂- and HDL_2b-cholesterol concentrations, indicating a dynamic metabolic relation between triacylglycerol-rich lipoproteins and the larger, less dense HDL subpopulation in these normolipidemic subjects. However, the changes in triacylglycerol concentrations only explained a small part (10%) of the interdietary differences in HDL₂-cholesterol concentrations; after adjustment for differences in triacylglycerol concentrations, significant differences in HDL₂-cholesterol concentrations remained between the 3 diets. These results suggest that the HDL-lowering effect observed after a reduction in dietary saturated fat was due to several mechanisms.

The implications of reducing the HDL cholesterol and specific HDL-subpopulations with low-fat diets are controversial (33). Previous studies suggest that a decrease in overall HDL- or HDL₂- and HDL_2b-cholesterol concentrations in conjunction with a diet-induced increase in serum triacylglycerol concentrations did not indicate an increased risk of coronary heart disease (7, 34, 49, 50, 56, 57). In addition, several intervention studies showed that reductions in saturated fat and cholesterol intakes are associated with a decreased incidence in atherosclerotic cardiovascular disease and improved mortality (58–60). Thus, reductions in HDL cholesterol, primarily in HDL₂ and HDL_2b cholesterol, may not be harmful if coupled with reductions in LDL cholesterol. In parallel with the changes in HDL subpopulations, there was a reduction in LDL cholesterol and apo B concentrations in the present study, clearly compatible with a decreased risk of coronary heart disease (27). The decrease in LDL cholesterol and apo B concentrations, suggesting a decrease in the number of circulating LDL particles, may offset any potential negative effects of a decrease in overall HDL, HDL₂, or HDL_2b concentrations.

In conclusion, a gradual decrease in total and saturated fat intakes resulted in a significant decrease in HDL₂- and HDL_2b-cholesterol concentrations in all age, sex, and race groups, whereas changes in HDL₃ cholesterol and HDL_3a-, HDL_3b-, and HDL_3c-cholesterol concentrations were less prominent. Although the quantitative effect of changes in serum triacylglycerol concentrations on differences in HDL subpopulations was limited, there was a close association between changes in concentrations of HDL₂ and HDL_2b and changes in serum triacylglycerol concentrations. The results indicate that dietary changes suggested to be prudent for a large segment of the population will primarily affect the concentrations of the most prominent antiatherogenic HDL subpopulation. However, the simultaneous decrease in the atherogenic LDL subpopulation will most likely offset any potential negative effect on cardiovascular risk.

The assistance of the Nutrition Unit and the nursing staff of the Irving Center for Clinical Research at Columbia University is gratefully acknowledged. We thank the DELTA investigators, listed by affiliation, for making this study possible: Columbia University—Henry N Ginsberg, Rajasekhar Ramakrishnan, Wahida Karmally, Lars Berglund, Maliha Siddiqui, Niem-Tzu Chen, Steven Holleran, Colleen Johnson, Roberta Holeman, Karen Chirgwin, Kellye Stennett, Lencey Ganga, Tajudeen T Towolawi, Minnie Myers, Colleen Ngai, Nelson Fontanez, Jeffrey Jones, Carmen Rodriguez, and Norma Useche; Pennington Biomedical Research Center—Michael Lefevre, Paul S Roheim, Donna Ryan, Marlene M Windhauser, Catherine M Champagne, Donald Williamson, Richard Tulley, Ricky Brock, Deonne Bodin, Betty Kennedy, Michelle Barkate, Elizabeth Foust, and Deshoin York; Pennsylvania State University—Penny Kris-Etherton, Satya S Jonnalagadda, Janice Derr, Abir Farhat-Wood, Vikkie A Mustad, Kate Meaker, Edward Mills, Mary-Ann Tilley, Helen Smiciklas-Wright, Madeline Sigman-Grant, Jean-Xavier Guinard, Pamela Sechevich, C Channa Reddy, Andrea M Mastro, and Allen Cooper; University of Minnesota—Patricia Elmer, Aaron Folsom, Nancy Van Heel, Christine Wold, Kay Fritz, Joanne Slavin, and David Jacobs; University of North Carolina at Chapel Hill—Barbara Dennis, Paul Stewart, CE Davis, James Hosking, Nancy Anderson, Susan Blackwell, Lynn Martin, Hope Bryan, W Brian Stewart, Jeffrey Abolafia, Malachy Foley, Conroy Zien, Szu-Yun Leu, Marston Youngblood, Thomas Goodwin, Monica Miles, and Jennifer Wehbie; Mary Imogene Bassett Hospital—Thomas A Pearson and Roberta Reed; University of Vermont—Russell Tracy and Elaine Cornell; Virginia Polytechnic and State University—Kent K Stewart and Katherine M Phillips; Southern University—Bernestine B McGee and Brenda Williams; Beltsville Agricultural Research Center—Gary R Beecher, Joanne M Holden, and Carol S Davis; National Heart, Lung, and Blood Institute—Abby G Ershow, David J Gordon, Michael Proschan, and Basil M Rifkind.

FOOTNOTES

*

See corresponding editorial on page 949.

2

Supported by grants 5-U01-HL-49644, 5-U01-HL-49648, 5-U01-HL-49649, 5-U01-HL-49651, and 5-U01-HL-49659 from the National Heart, Lung, and Blood Institute and grants M-01-RR-00645 and M01-RR-00400 from the National Center for Research Resources, National Institutes of Health. The following companies donated products used in this study: Bertoli USA, Best Foods, Campbell Soup Co, Del Monte Foods, General Mills, Hershey Foods Corp, Institute of Edible Oils and Shortenings, Kraft General Foods, Land O'Lakes, McCormick Inc, Nabisco Foods Group, Neomonde Baking Co, Palm Oil Research Institute, Park Corp, Procter and Gamble, Quaker Oats, Ross Laboratories, Swift-Armour and Eckrich, Van Den Bergh Foods, Cholestech, and Lifelines Technology, Inc.