点击显示 收起
1 From the Lawrence Berkeley National Laboratory, Donner Laboratory, Berkeley, CA (PTW and RMK), and the Children's Hospital Oakland Research Institute, Oakland, CA (PJB, RR, and RMK)
2 Supported in part by a grant from Dairy Management Incorporated, NIH R01 grant HL072110, and NIH Program Project grant HL-18574 from the National Heart, Lung, and Blood Institute and conducted at Lawrence Berkeley National Laboratory through the US Department of Energy under contract no. DEAC03-76SF00098. 3 Address reprint requests to PT Williams, Lawrence Berkeley National Laboratory, Donner Laboratory, 1 Cyclotron Road, Berkeley, CA 94720. E-mail: ptwilliams{at}lbl.gov.
ABSTRACT
Background: Individuals vary greatly in their lipoprotein responses to low-fat diets, with some of this variation being attributable to genes.
Objective: The purpose was to test the extent to which individual lipoprotein responses to diet can be attributed to genes in the presence of divergent exercise levels.
Design: Twenty-eight pairs of male monozygotic twins (one twin mostly sedentary, the other running an average of 50 km/wk more than the sedentary twin) went from a 6-wk 40%-fat diet to a 6-wk 20%-fat diet in a crossover design. The diets reduced fat primarily by reducing saturated and polyunsaturated fat (both from 14% to 4%) while increasing carbohydrate intake from 45% to 65%.
Results: Despite the twins' differences in physical activity, the dietary manipulation produced significantly correlated changes (P < 0.05) in the twins' total cholesterol (r = 0.56); LDL cholesterol (r = 0.70); large, buoyant LDL [Svedberg flotation rate (Sf) 7–12; r = 0.52]; apolipoprotein A-I (r = 0.49); lipoprotein(a) (r = 0.49); electrophoresis measurements of LDL-I (LDLs between 26 and 28.5 nm in diameter; r = 0.48), LDL-IIB (25.2–24.6 nm; r = 0.54), and LDL-IV (22–24.1 nm; r = 0.50); and body weight (r = 0.41). Replacing fats with carbohydrates significantly decreased the size and ultracentrifuge flotation rate of the major LDL and the LDL mass concentrations of large, buoyant LDL; LDL-I; HDL cholesterol; and apolipoprotein A-I and significantly increased concentrations of LDL-IIIA (24.7–25.5 nm) and lipoprotein(a).
Conclusions: Even in the presence of extreme differences in exercise, genes significantly affect changes in LDL, apolipoprotein A-I, lipoprotein(a), and body weight when dietary fats are replaced with carbohydrates.
Key Words: Twins • low-fat diet • high-carbohydrate diet • lipoproteins • lipoprotein(a) • physical activity • LDL subclasses • apolipoproteins • cholesterol
INTRODUCTION
The risk of coronary heart disease increases in association with higher plasma concentrations of LDL cholesterol, triacylglycerols, and lipoprotein(a) and decreases in association with higher concentrations of HDL cholesterol and apolipoprotein (apo) A-I and with the greater size and buoyancy of the LDL particles (1, 2). Low-fat, high-carbohydrate diets decrease plasma concentrations of LDL cholesterol, HDL cholesterol, and apo A-I and increase lipoprotein(a) and triacylglycerols (3). Low-fat, high-carbohydrate diets also produce a shift in the distribution of LDLs from larger, more buoyant particles to smaller, denser particles (4).
Individuals vary greatly in their lipoprotein responses to low-fat diets, with some of this variation being attributable to genes. Persons having the apo E e4 allele experience greater reductions in LDL cholesterol (5) and large, buoyant LDL [Svedberg flotation rate (Sf) 7–12] (6) with low-fat, low-cholesterol diets than do those lacking the allele. Polymorphisms in the apo B gene, the signal peptide insertion allele, the LDL receptor gene, the MN blood group, and the apo A-I promoter region are also reported to affect the LDL response to diet (5). Low-fat diets induce a greater reduction in LDL cholesterol and HDL2b (the largest HDL particles) in persons with a genetically influenced profile characterized by a predominance of small LDL particles than in those lacking this trait (7-9).
Studies of monozygotic twins provide evidence of genetic regulation in the absence of prior knowledge of the specific genes involved. Such studies provide a global test for genetic hypotheses while circumventing issues of multiple hypothesis testing that plague exploratory tests of multiple genetic loci (10). For example, overfeeding and caloric expenditure in monozygotic twins causes weight gains and losses that correlate significantly within twin pairs (11, 12). To date, however, only a small proportion of the variation in body weight has been attributed to specific genes (13).
The current study examines the effects of switching from a high-fat, low-carbohydrate diet to a low-fat, high-carbohydrate diet in monozygotic twins to assess the contribution of genes to the diet-induced changes in lipoproteins and body weight. Although it is often difficult to separate the effects of the twins' shared genotypes from the effects of their shared environment (14), the current design minimizes the effect of the shared environment by 1) deliberately choosing twins with divergent lifestyles (one physically active, one sedentary) and 2) measuring the response to an experimental manipulation of diet (as opposed to observational twin studies that may be strongly affected by the shared environment).
SUBJECTS AND METHODS
Twenty-nine pairs of male identical twins discordant for exercise participated in a crossover study of high-fat, low-carbohydrate and low-fat, high-carbohydrate diets. The twins were identified among current participants of the National Runners' Health Study (15) and from announcements distributed at footraces through the Runner's World race participation program (Rodale Press, Emmaus, PA). Criteria for eligibility were as follows: discordant for exercise (ie, either one twin was sedentary and the other was ran 32 km/wk or if both twins ran there was a 40-km/wk difference in running distance), no medication use likely to interfere with lipid metabolism, free of chronic disease, nonsmoker, and willingness to abstain from alcohol and follow the prescribed diets over the 12-wk intervention. Each twin completed a questionnaire and signed a consent form approved by the Committee for the Protection of Human Subjects at Lawrence Berkeley National Laboratory, University of California, Berkeley, CA.
The research was conducted in an outpatient setting with careful monitoring of dietary compliance. All participants were counseled by registered dietitians to follow the prescribed diets both before and during the experimental intervention. The twin pairs received, in random order, a 6-wk low-fat, solid-food diet (20% of total energy as fat, 65% as carbohydrates) and a 6-wk high-fat diet (40% fat, 45% carbohydrates) in a crossover design. The 2 experimental diets were designed to achieve a comparison of high and low fat intake by substitution of fat for carbohydrate without significant change in other major nutrients. Nutrient compositions of the diets were calculated by using the Minnesota NUTRITION DATA SYSTEM (NDS) software developed by the Nutrition Coordinating Center (version 4.01; University of Minnesota, Minneapolis, MN). Registered dietitians supplied the participants with personalized menus showing the number and size of servings for the experimental diets.
Seven-day diets were prescribed that represented 95% of the participants' total caloric intake as estimated from their baseline 4-d food records. The remaining 5% of caloric intake was provided as food combinations that matched the dietary composition of the prescribed diets and that could be consumed ad libitum in response to satiety. The prescribed diets had to be eaten in their entirety within each 7-d period. The 5% additional calories could be consumed as 120 mL (0.5 cup) of low-fat milk with 5 vanilla wafers for the low-fat diet and as 1 teaspoon (4 g) of peanut butter with 8 wheat crackers for the high-fat diet. All subjects abstained from alcohol during the study.
The staff contacted the subjects weekly during the study to verify adherence to the diet and to review the protocol. Compliance was assessed by reviewing 4-d diet records and grocery receipts. One twin pair did not complete the dietary intervention.
Twins reported to a local clinic of their choice to have their blood drawn at baseline and at the end of each 6-wk diet. All were required to have abstained for 12–14 h from all food and vigorous activity. Plasma was prepared from venous blood collected in tubes containing 1.4 mg Na2EDTA/mL. Samples were drawn only on Mondays, Tuesdays, or Wednesdays and were shipped overnight on wet ice to ensure that they were delivered to our laboratory by Thursday morning. Before starting the study, all participants received an electronic scale for measuring their own body weight. Height and weight were also measured during the clinic visits.
Lipid and lipoprotein measurements
Fasting plasma lipids were measured at baseline and after each 6-wk diet. Plasma concentrations of total cholesterol and triacylglycerols were measured by enzymatic procedures (ABA 200 instrument, Abbott Laboratories, Abbott Park, IL; 16). HDL cholesterol was measured by the dextran sulfate–magnesium precipitation of apo B–containing lipoproteins followed by enzymatic determination of cholesterol (17, 18). Plasma LDL-cholesterol concentrations were calculated from the formula of Friedewald et al (19). The laboratory remained certified by the Centers for Disease Control and Prevention lipid standardization program throughout the study. Apo A-I and B in plasma was measured by immunoturbidimetric assay (20, 21) with an ITA reagent kit (Bacton Assay Systems Inc, San Marcos, CA). Measurements are performed by using the Express 550 analyzer according to kit instructions. Calibrators and controls are assigned quantitation values based on the International Federation of Clinical Chemistry proposed Standard Reference Material SP1 and by participation in the standardization program directed by the International Federation of Clinical Chemistry and the Centers for Disease Control and Prevention. Intra- and interrun CVs were within ±5%.
Fasting LDL particle diameters and LDL particle subclass intervals based on particle size were calculated from calibration curves by using standards of known size (22). Analyses are based on the area within the LDL-IVB (22.0–23.2 nm), LDL-IVA (23.2–24.2 nm), LDL-IIB (24.2–24.7 nm), LDL-IIIA (24.7–25.5 nm), LDL-IIB (25.5–26.0 nm), LDL-IIA (26.0–26.5 nm), and LDL-I (26.5–28.5 nm) particle size intervals (22, 23). Analytic ultracentrifugation was used to measure concentrations of total lipoprotein mass within multiple regions for dense LDL (Sf 0–7), buoyant LDL (Sf 7–12), intermediate-density lipoproteins (IDL; Sf 12–20) and VLDL (Sf 20–400) (24).
Statistical analyses
Fifteen pairs started with the high-fat diet and 13 pairs started with the low-fat diet. Because the 2 diet sequences were not equally represented, paired t tests were not used because temporal effects would not be eliminated by the analyses. We therefore computed separately the mean lipoprotein change in switching from a high-fat to a low-fat diet and the mean lipoprotein change in switching from the low-fat to the high-fat diets and their corresponding SEs. We then calculated one-half of the differences of the mean changes and their corresponding SE (one-half of the square root of the sum of the squared SEs) to estimate separately the effect of the diet manipulation on the running twins' and the sedentary twins' lipoproteins while eliminating any temporal effects. The difference between the running and the sedentary twins' dietary responses was calculated by subtracting the lipoprotein change within each twin pair and then analyzing the calculated differences as described above. Because none of the variables responded differently in the running and sedentary twins, we also analyzed the average of the twins' responses to assess the effect of the diet on lipoproteins with greater statistical power. Twin-pair correlations of the lipoprotein responses to the diets were calculated after adjustment for the diet sequence by regression analyses. Plots of the twins' responses are presented with adjustment to represent switching from the high-fat to the low-fat diet. Statistical analyses were performed by using STATVIEW software version 5.0.1 (SAS Institute Inc, Cary, NC).
RESULTS
Baseline
The baseline characteristics of the twins are presented in Table 1. The running twins ran an average of 50 km/wk more than the sedentary twins. Correspondingly, the running twins weighed significantly less than did the sedentary twins, had significantly higher apo A-I and HDL-cholesterol concentrations in plasma, and had significantly lower apo B and triacylglycerol concentrations in plasma. LDL peak particle diameter was also significantly larger in the running twins.
View this table:
TABLE 1. Baseline screening characteristics of male monozygotic twins according to physical activity1
Consistent with their monozygosity, the twins' heights were significantly correlated (r = 0.92), as were their body masses and body mass indexes. Despite substantial differences in physical activities, the twins exhibited strong, significant correlations for apos A-I and B, triacylglycerols, total cholesterol, HDL cholesterol, LDL cholesterol, lipoprotein(a), and LDL peak particle diameter. Plasma concentrations of LDL-I, LDL-IIB, LDL-IIIA, LDL-IVA, and LDL-IVB were also significantly correlated between twins.
Switching from the high-fat to the low-fat diet
The nutrient intakes reported on the 7-d food records for the running and sedentary twins during the 2 diets are shown in Table 2. The dietary goals were achieved for both diets. The changes in mean nutrient intake when switching from the high-fat, low-carbohydrate diet to the low-fat, high-carbohydrate diet were not significantly different between the running and sedentary twins for total energy intake (mean change for the runner – the mean change for the sedentary twin ± SE: –117.69 ± 92.12 kcal/d), total fat (0.53 ± 0.82%), saturated fat (0.12 ± 0.22%), monounsaturated fat (0.19 ± 0.21%), polyunsaturated fat (0.19 ± 0.49%), carbohydrates (–1.10 ± 1.22%), protein (0.58 ± 0.51%), or dietary cholesterol (5.26 ± 15.21 mg/d).
View this table:
TABLE 2. Mean nutrient intakes with the high- and low-fat diets of male monozygotic twins according to physical activity1
As shown in Table 3, decreasing dietary fat significantly decreased HDL-cholesterol concentrations in both the running and the sedentary twins. Apo A-I also decreased significantly in the running twins and marginally in the sedentary twins. The decreases in both HDL cholesterol and apo A-I were significant when the running and sedentary twins' data were averaged, as was the increase in mean plasma lipoprotein(a) concentrations.
View this table:
TABLE 3. Mean weight and apolipoprotein and lipoprotein concentrations of male monozygotic twins according to physical activity during the 6-wk high-fat and low-fat diets and the changes between diets1
Also shown in Table 3 are the changes in VLDL and LDL in response to decreasing fat and increasing carbohydrates. Mean LDL peak particle diameter and the LDL peak flotation rate decreased in both the sedentary and the exercising twins. Mass concentrations of large, buoyant LDL also decreased significantly in both groups. Correspondingly, changes in LDL peak particle diameter, LDL peak flotation rate, and large, buoyant LDL were strongly significant when the values for the running and sedentary twins were averaged. The additional statistical power when the values for the running and the sedentary twins were averaged resulted in significant decreases being shown in LDL-I and significant increases in LDL-IIIA. The decrease in LDL-I and increase in LDL-IIIA were significant in the sedentary twins but not the running twins (P = 0.10 for LDL-I and P = 0.07 for LDL-IIIA in the running twins). The lipoprotein responses to the diets were not significantly different between the running and sedentary twins (Table 3).
Concordance within twin pairs
Increased dietary fat did not significantly change body weight (Table 3). However, there was considerable variability in the body weight response to the diets, and the responses were significantly correlated within twin pairs (r = 0.41; Table 3). Despite the substantial differences in physical activity, changes in apo A-I were strongly correlated within twin pairs, as were changes in lipoprotein(a) (Table 3).
The strongest correlation between the running and sedentary twins' lipoproteins was the correlation in the LDL-cholesterol response when switching from a high-fat to a low-fat diet (Figure 1). The data shown in Table 3 suggest that the within-pair correlation for changes in LDL cholesterol reflects within-pair concordant changes in LDL-I and the most buoyant LDL (Sf 7–12). Changes in LDL-IIB and LDL-IV were also significantly correlated between the running twins and the sedentary twins (Table 3).
FIGURE 1.. Changes () in plasma LDL-cholesterol concentrations when switching from a 6-wk high-fat diet (40%) to a 6-wk low-fat diet (20% fat) in 28 monozygotic twin pairs discordant for physical activity. The diagonal is not a line fitted to the observations but rather a line drawn as a reference to the locus of points where the changes are identical in the twin pairs. r = 0.70, P < 0.0001.
The correlation between the twins' lipoprotein changes could not be attributed to concordance in their adherence to the dietary protocol. The correlations for changes in percentages of energy from protein and carbohydrate and dietary cholesterol were all nonsignificant (0.06 r 0.08) when switching from the high-fat, low-carbohydrate diet to the low-fat, high-carbohydrate diet. One of the twin pairs reported concordantly low changes in total and saturated fat intakes and one of the other twin pairs reported concordantly low changes in polyunsaturated fat intake. When we excluded these 2 twin pairs, the significant twin correlation was eliminated between changes in total percentage fat intake (r = 0.36 reduced to r = –0.15), percentage saturated fat intake (r = 0.58 reduced to r = 0.14), percentage monounsaturated fat intake (r = 0.36 reduced to r = 0.18), and percentage polyunsaturated fat intake (r = 0.36 reduced to r = –0.13) when switching between diets. Eliminating these 2 twin pairs had almost no detectable effect on the twin correlations for changes in apo A-I (r = 0.47), total cholesterol (r = 0.56), LDL cholesterol (r = 0.70), lipoprotein(a) (r = 0.47), LDL-I (r = 0.40), LDL-IIB (r = 0.57), LDL-IVA (r = 0.50), LDL-IVB (r = 0.49), or large, buoyant LDL mass (r = 0.58) after switching from the high-fat, low-carbohydrate diet to the low-fat, high-carbohydrate diet.
DISCUSSION
The lipoprotein changes shown in these 28 twin pairs confirm previous reports by ourselves and others that switching from a high-fat, low-carbohydrate diet to a low-fat, high-carbohydrate diet decreases HDL cholesterol and apo A-I and increases lipoprotein(a) (25-27). The diet also decreased the size and buoyancy of the LDL particle distribution as a result of reductions in LDL particles of Sf 7–12 and 26–28.5 nm diameter (LDL-I). In addition, gradient gel electrophoresis showed significant increases in LDL-IIIA. As shown in Table 3 and Figure 1, much of the LDL response in switching from a high-fat, low-carbohydrate diet may be accounted for by genes.
Whereas our previous studies held total caloric intake constant or manipulated calorie intake to hold body weight constant (4, 6, 7, 8), here we prescribed 95% of caloric intake and allowed each subject to supplement his diet with food combinations in accordance with individual preferences to achieve satiety while maintaining the nutrient composition of the diets. This method more realistically reflects the implementation of these diets in free-living unsupervised populations. This approach was taken because weight and lipoprotein changes that occur with real-life exposure to these diets may differ from those observed when caloric intake or body weights are forced to remain constant. For example, reductions in dietary fat have been reported by others to increase triacylglycerol and the ratio of total cholesterol to HDL cholesterol under weight-maintenance conditions but not under ad libitum conditions leading to weight loss (28).
The unique study design showed significant within-pair correlations in the twins' lipoprotein responses to the dietary manipulations despite their divergent lifestyles. The strongest correlation was for changes in LDL cholesterol (Figure 1). Although several genes have been linked to LDL-cholesterol change during dietary manipulation (5), these are unlikely to account for the 49% of the variance in LDL-cholesterol change that our study attributed to the twins' genes or shared environment. Analytic ultracentrifugation and gradient gel electrophoresis suggest that the concordance in the twins LDL-cholesterol response involves large, buoyant LDL particles of Sf 7–12 and large LDL particles of the LDL-I subclass. The agreement among 3 independent LDL measurements involving 3 separate methods confirms the concordant LDL-cholesterol response to the diet.
Diet-induced changes in the LDL-IIB subclass were also significantly correlated within twin pairs, as were changes in LDL-IV. The LBL-IVB subclass is a relatively minor portion of the LDL distribution that was recently shown to have an independent association with coronary disease progression (29). The discontinuity in the concordance of the monozygotic-twin diet response between LDL-IIB and LDL-IV shown in Table 3 is similar to the discontinuities we reported when LDL subclasses are correlated with atherosclerosis (29) and other lipoproteins (30).
The high monozygotic correlation for lipoprotein(a) measured cross-sectionally is consistent with the finding that >90% of the variation in lipoprotein(a) concentrations is accounted for by the apolipoprotein(a) gene (31). Our data (Table 3) also suggest a strong genetic influence on the lipoprotein(a) response to diet.
We recognize that free-living populations could be less likely to follow controlled diets than are subjects for whom food is supplied. However, we have now completed several studies of men and women with similar dietary protocols (4, 6, 7, 9). Our success in implementing these studies is reflected both in diet records and by the finding that mean lipid responses conform with those predicted from previous controlled feeding studies (32).
We defined divergent lifestyles with respect to different levels of physical activity. As shown in Table 1, runners weighed significantly less than did their sedentary twins, had lower plasma concentrations of apo B and triacylglycerols, higher plasma concentrations of apo A-I and HDL cholesterol, and larger LDL peak particle diameter. Although these lipoprotein and weight differences are well documented between vigorously active and inactive men (33-35), the data in Table 1 show that these differences persist when genetic effects are controlled for, which is an important consideration because the lipoprotein response to exercise is affected by genes (36). Genes presumably also partially explain why sedentary men with high HDL-cholesterol concentrations run longer weekly distances when enrolled in a training program than do those with low HDL-cholesterol concentrations (37, 38). The running twins also had higher concentrations of lipoprotein(a) than did their sedentary brothers, which has not been consistently observed by others (39-41) but may have been discernible in our study design because we matched for genotype [ie, Table 1 shows a strong genetic concordance for lipoprotein(a) values].
Our results suggest there are genes that strongly influence the LDL-cholesterol response to diet, even in the presence of large differences in physical activity. These genes appear to primarily affect the dietary response of the larger, more buoyant LDL particles. Previous studies have indicated that these particles are more strongly associated with changes in saturated fat intake than are other LDL species (42). Even the most physically active men are susceptible to the effects of diet on HDL cholesterol, apo A-I, and large, buoyant LDL concentrations and the size and buoyancy of the predominant LDL particles. The prominent role genes play in regulating lipoprotein response to diet is evident whether following ab libitum dietary choices (Table 1) or large dietary perturbations in carbohydrate and fat consumption, regardless of the level of physical activity (Table 3). Moreover, our analyses support earlier observations indicative of the genetic regulation of weight change after environmental perturbation (11, 12). On the basis of these results, we believe that detailed analyses using genetic association or linkage studies are warranted to identify the causes of the associations of diet with lipoprotein and weight.
ACKNOWLEDGMENTS
We acknowledge the valuable contribution of Susan Fernstrom for planning and implementing the dietary intervention and for dietary assessment.
PTW designed the study, directed the recruitment, analyzed the data, and wrote the manuscript. PJB directed the lipoprotein analyses, RR directed the intervention and coordinated the collection of laboratory samples, and RMK secured funding, was the principal investigator of the study, helped design the study, and contributed critically to the scientific content of the manuscript.
REFERENCES