点击显示 收起
1 From the Departments of Public Health (LFM) and Child Health (GM) and the Health Services Research Unit (AA), University of Aberdeen, Aberdeen, United Kingdom.
2 LFM was supported by a scholarship from the Institute of Applied Health Sciences, University of Aberdeen; AA was supported by the Medical Research Council and the Chief Scientist Office of the Scottish Executive Health Department. The Health Services Research Unit of the University of Aberdeen is supported by the Chief Scientist Office of the Scottish Executive Health Department; however, the views expressed herein are those of the author. 3 Reprints not available. Address correspondence to LF Masson, Department of Medicine & Therapeutics, University of Aberdeen, Polwarth Building, Foresterhill, Aberdeen AB25 2ZD, United Kingdom. E-mail: lindsey.masson{at}abdn.ac.uk.
ABSTRACT
There is wide interindividual variation in the lipid and lipoprotein responses to dietary change, and the existence of consistent hypo- and hyperresponders supports the hypothesis that responsiveness is related to genetic variation. Many studies have investigated the possibility that the heterogeneity in responsiveness to changes in dietary fat, cholesterol, and fiber intake is explained by variation in genes whose products affect lipoprotein metabolism, eg, apolipoproteins, enzymes, and receptors. A systematic review of the literature was carried out to investigate the effect of genetic variation on the lipid response to dietary intervention. A search strategy for the MEDLINE database retrieved 2540 articles from 1966 to February 2002. This strategy was adapted and performed on the EMBASE database, which retrieved 2473 articles from 1980 to week 9, 2002. Reference lists from relevant journal articles were also checked. This is the first systematic review of the literature, and it summarizes results available from 74 relevant articles. There is evidence to suggest that variation in the genes for apolipoprotein (apo) A-I, apo A-IV, apo B, and apo E contributes to the heterogeneity in the lipid response to dietary intervention. However, the effects of genetic variation are not consistently seen and are sometimes conflicting. Future studies need to have much larger sample sizes based on power calculations and carefully controlled dietary interventions and should investigate the effects of polymorphisms in multiple genes instead of the effects of polymorphisms in single genes.
Key Words: Polymorphism genotype diet lipids lipoproteins cardiovascular disease
INTRODUCTION
Cardiovascular disease has a multifactorial etiology with many established risk factors. The nonmodifiable risk factors include older age, male sex, and a family history of premature cardiovascular disease, whereas the modifiable risk factors include cigarette smoking, obesity, hypertension, physical inactivity, diabetes mellitus, elevated total cholesterol and LDL-cholesterol concentrations, and reduced HDL-cholesterol concentrations (1).
Serum cholesterol concentrations are profoundly influenced by the composition of dietary fat, with saturated fatty acids (SFAs) being the major determinant of serum cholesterol (2). Equations have been developed to predict the responses of total, LDL, and HDL cholesterol to dietary change (25), but there is wide interindividual variation in these lipid and lipoprotein responses (68). The existence of consistent hypo- and hyperresponders (9, 10) supports the hypothesis that responsiveness is related to genetic variation.
Many studies have investigated this possibility and have largely focused on genes whose products affect lipoprotein metabolism, eg, apolipoproteins, enzymes, and receptors. Although there have been several reviews of such studies (1120), none were performed in a systematic manner, which may have led to articles being omitted and introduced bias toward positive findings. For example, one review included a summary table of 28 studies that examined interactions between the apolipoprotein (apo) E genotype and diet with changes in dietary fat or cholesterol (17); however, the present systematic review identified 16 additional relevant studies that could have been included.
It is possible that conclusions from nonsystematic reviews may not be accurate because they do not take into account all of the available evidence, and there is no basis for the selection of the studies that are included in the review. This is the first systematic review of the literature. The evidence available from 74 articles describing studies that have investigated the effect of genetic variation on the lipid response to dietary interventions were examined.
METHODS
A literature search strategy was developed to identify studies that had measured the lipid and lipoprotein responses to dietary interventions in different genotype groups. The literature strategy for the MEDLINE (National Library of Medicine, Bethesda, MD) database, which was searched from 1996 to week 4 of February 2002, and the number of articles retrieved are shown in Table 1. The strategy was adapted and performed on the EMBASE database (Elsevier Science BV, Amsterdam) from 1980 to week 9 of 2002, and reference lists from relevant articles were checked. Articles were excluded if the intervention involved restriction of energy intake or overfeeding, if the intervention was combined with drug or exercise therapy, if the subjects were children or had diabetes, or if the articles did not clearly assess whether responses to diet were significantly different between genotype groups. This search generated 74 relevant articles on dietary intervention studies and 17 reviews on gene-diet interactions.
View this table:
TABLE 1 . Literature strategy for MEDLINE: from 1966 to week 4 of February 2002 and the number of articles retrieved
The original studies retrieved, according to the polymorphism studied, are summarized in Tables 27. The details given for each study include the first author, the number and sex of the subjects, the number of subjects in each genotype group, the type of dietary intervention, and the results, ie, whether there was a significantly greater response of plasma total cholesterol, LDL cholesterol, HDL cholesterol, VLDL cholesterol, or triacylglycerol in one genotype group than in another. The responses were generally in the expected direction, ie, a reduction in lipid and lipoprotein concentrations after a reduction in dietary cholesterol and SFA intakes. In most cases, the direction of the response was the same for each genotype group. In these cases, the genotype that showed the largest response to the dietary intervention is shown.
View this table:
TABLE 2 . Apolipoprotein A-I, C-III, and A-IV gene cluster1
View this table:
TABLE 3 . Apolipoprotein B gene1
View this table:
TABLE 4 . Apolipoprotein E gene1
View this table:
TABLE 5 . Genes coding for enzymes1
View this table:
TABLE 6 . LDL receptor gene1
View this table:
TABLE 7 . Other genes1
RESULTS
Apolipoprotein A-I, C-III, and A-IV gene cluster
Of 13 reports, 5 found that the presence of the apo A-I75 (G/A) A allele instead of the common G allele resulted in greater LDL-cholesterol responses to changes in dietary fat (Table 2). In addition, significant interactions between the G/A genotype and diet were found for changes in total and LDL cholesterol when subjects changed from a low-fat diet to a diet high in monounsaturated fatty acids (MUFAs) (24). When the effect of this polymorphism with the apo C-III SstI polymorphism was studied, it was found that total and LDL-cholesterol concentrations decreased most in subjects with the G/G-S1/S2 polymorphism and increased most in subjects with the G/A-S1/S1 polymorphism (24). No significant interactions between diet and other polymorphisms in the apo A-I gene were shown.
Variation at the apo C-III SstI site influenced dietary responsiveness such that after the subjects changed from a low-fat diet to a high-MUFA diet, LDL-cholesterol concentrations increased in S1/S1 subjects but decreased in S1/S2 subjects (24, 32). Variation at the apo C-III SacI or C1100T polymorphic sites has not been associated with the magnitude of the lipid response.
The evidence that exists for an interaction between diet and the apo A-IV glutamine-histidine mutation at position 360 (Gln360His) suggests that Gln/Gln subjects show significantly greater total and LDL-cholesterol responses and that Gln/His subjects show greater HDL-cholesterol responses to changes in dietary fat, cholesterol, or both. Although Wallace et al (37) found no significant differences in LDL-cholesterol responses between genotypes, dense LDL cholesterol decreased more in subjects carrying the His allele when polyunsaturated fatty acids (PUFAs) replaced SFAs in the diet. In the same study, there was a significant difference in HDL-cholesterol responses between genotype groups such that concentrations decreased in Gln/Gln subjects and increased in Gln/His subjects.
The presence of serine instead of threonine at position 347 in the apo A-IV gene was associated with increased total and LDL-cholesterol responsiveness when subjects switched from a high-SFA diet to a National Cholesterol Education Program Step I diet (23). When the same subjects changed from the National Cholesterol Education Program Step I diet to a high-MUFA diet, subjects with the Thr/Thr genotype had a 1% decrease in total cholesterol concentrations, whereas subjects with the Ser allele had a 5% increase in total cholesterol concentrations (23). When the Thr347Ser and the apo A-I75 (G/A) genotypes were combined, carriers of the A and Ser alleles showed greater LDL-cholesterol responses to changes in dietary fat (23). However, Carmena-Ramon et al (36) investigated both the Gln360His and Thr347Ser polymorphisms and found no gene-diet or haplotype-diet interactions.
Apolipoprotein B
The evidence for an interaction between the XbaI polymorphism and diet is inconsistent (Table 3). In 2 studies, X-X- subjects showed greater LDL-cholesterol responses (42, 46), whereas Tikkanen et al (40) found that subjects carrying the X+ allele had greater total, LDL-, and HDL-cholesterol responses. However in Xu et als (92) analysis of these data, the XbaI polymorphism only explained a significant proportion of variance of the change in HDL cholesterol. Pajukanta et al (43) found no significant effect on LDL-cholesterol responsiveness, although X-X-subjects showed the greatest HDL2- and VLDL-cholesterol responses. Finally, Lopez-Miranda et al (45) studied the effect of the XbaI polymorphism in subjects with the common apo E3/3 genotype and found that X-X- subjects showed the greatest triacylglycerol response. Rantala et al (46) conducted a meta-analysis of all published dietary trials, including their own trial (46). In their analysis of 8 studies, X-X+ subjects had greater LDL responses than did X+X+ subjects and no significant differences in the responses of total or HDL cholesterol or triacylglycerol were found between genotypes.
Two of 7 interventions found that the EcoRI R- allele was associated with significantly greater total and LDL-cholesterol responses to changes in dietary fat and cholesterol (44, 46). Rantala et als (46) meta-analysis of 7 studies found that R-R- subjects had significantly greater total and LDL-cholesterol responses than did the R+R+ subjects.
Only one study found an interaction between the MspI polymorphism and response to diet. Rantala et al (46) found that M+M+ subjects had significantly greater total cholesterol responses than did M+M- subjects to a diet high in fat and cholesterol. In addition, in their meta-analysis of 5 trials, Rantala et al found that M+M+ subjects responded with significantly greater changes in plasma LDL-cholesterol and triacylglycerol concentrations than did M+M- subjects, but there were no significant effects on total or HDL-cholesterol responses.
Ten intervention studies found no significant effects of the apo B signal peptide insertion/deletion (I/D) polymorphism on dietary responsiveness; however, 2 studies reported a significantly greater responsiveness in subjects homozygous for the I allele. Lee et al (93) studied 43 men and women to compare the effects of insoluble and soluble fiber on plasma lipids. Their statistical model identified gene-diet interactions. However, they did not look specifically at differences between genotype groups; therefore, the study is not included in the tables. It was found that D/D subjects had similar decreases in HDL cholesterol after consumption of the insoluble- and soluble-fiber diets. However, I/I subjects had larger HDL-cholesterol decreases with the soluble-fiber diet, whereas I/D subjects had larger HDL-cholesterol decreases with the insoluble-fiber diet. The gene-diet interaction was significant (P = 0.021).
In response to Pajukanta et als (43) low-fat, low-cholesterol diet, I/I subjects showed the greatest decrease in HDL2. In addition, I/I and I/D subjects showed increased VLDL-cholesterol and decreased LDL-cholesterol concentrations, whereas D/D subjects showed decreased VLDL-cholesterol and increased LDL-cholesterol concentrations. The I/D polymorphism showed no significant effect on the responsiveness of total, LDL, or HDL cholesterol or triacylglycerol in a meta-analysis of 7 studies (46).
The Bsp 1261I polymorphism was not associated with the LDL response in Rantala et als study (46). However, their analysis of the data from 2 Finnish studies showed that B-B- subjects responded to a diet high in fat and cholesterol, with significantly greater increases in plasma total and LDL cholesterol and significantly smaller changes in triacylglycerol than observed in the B-B+ subjects. Finally, men homozygous for the His allele at codon 1896 may be more responsive than men with the less common Arg allele (50).
Apolipoprotein E
Variation in the apo E gene results in the 3 common alleles (2, 3, and 4), which can produce 3 homozygous (E2/2, E3/3, and E4/4) and 3 heterozygous (E2/3, E2/4, and E3/4) genotypes. The 2, 3, and 4 alleles have frequencies of 0.08, 0.77, and 0.15, respectively, in white populations (94). The 4 studies that showed a statistically significant interaction between the variation in the apo E gene and dietary cholesterol indicate that the presence of the 4 allele results in significantly greater responses in total, LDL, or HDL cholesterol; subjects with the 2 allele showed the smallest response (Table 4).
Of the 46 interventions that involved altering the dietary fat content of the diet, significantly different total and LDL-cholesterol responses between genotype groups were reported in 8 and 11 studies, respectively, with carriers of the 4 allele tending to show the greatest responses. Although Tikkanen et al (64) found that subjects with the E4/4 phenotype showed the greatest total and LDL-cholesterol responses to dietary change, Xu et al (92) analyzed the same data and concluded that the apo E polymorphism did not explain a significant proportion of the variation of the response. Lopez-Miranda et al (71) conducted a meta-analysis of 9 studies involving 612 subjects and found that the presence of the 4 allele was associated with a significantly greater LDL response to dietary intervention.
Four studies found significantly different HDL-cholesterol responses between genotype groups: one study found that carriers of the 4 allele had the smallest HDL-cholesterol response (71), whereas the other 3 studies found that carriers of the 4 allele had the largest response (28, 54, 55, 58, 62).
Subjects with the 4 allele appear to be the most responsive to changes in dietary fat and cholesterol; however, they may not be the most responsive to changes in other aspects of the diet. For example, subjects carrying the 2 allele had the greatest total and LDL-cholesterol responses to wheat- or oat-bran supplementation (86).
Enzymes: lipoprotein lipase, hepatic lipase, and cholesterol 7-hydroxylase
Three polymorphisms in the lipoprotein lipase gene have been shown to influence the response to diet (Table 5). First, the H- allele of HindIII is associated with greater total cholesterol and triacylglycerol responsiveness (29). Second, subjects with the S447X mutation showed significantly greater decreases in LDL-cholesterol concentrations than did subjects with the common S/S genotype when PUFAs replaced SFAs in the diet. Third, subjects heterozygous for the N291S polymorphism (N/S) showed a greater mean change in triacylglycerol concentrations than did N/N subjects. The PvuII and T-93G polymorphisms have not shown a significant influence on dietary responsiveness.
Variation at the hepatic lipase NlaIII and MspI sites was not associated with the response to diet (31, 87). However, in the study by Lee et al (93) that was described above, homozygotes for valine (V/V) at position 73 had smaller decreases in LDL cholesterol with the insoluble-fiber diet than with the soluble-fiber diet, but the LDL-cholesterol responses to the 2 diets were similar for the heterozygotes for methionine. This gene-diet interaction was significant (P = 0.036). Variation in the cholesterol 7-hydroxylase gene has not been associated with dietary responsiveness (42).
LDL receptor gene
A significant interaction between diet and variation in the LDL receptor gene was shown in 2 studies (Table 6). After wheat- or oat-bran supplementation, the total and LDL-cholesterol reductions were significantly different among the HincII genotype groups, with 2/2 subjects showing the greatest reduction and 1/1 subjects showing the smallest reduction (31). Clifton et al (12) and Clifton and Abbey (14) investigated variation at a PvuII site, and subjects lacking the cutting site were found to have greater HDL responses to low-fat diets.
Other genes
Subjects homozygous for the presence of the TaqIB cutting site (B1/B1) of the cholesteryl ester transfer protein gene showed greater total and LDL-cholesterol responses to changes in the type of dietary fat than did subjects who were homozygous or heterozygous for the B2 allele in 1 of 3 studies (30) (Table 7). No evidence exists for a significant association between the I405V polymorphism and the response to changes in the amount of dietary fat and cholesterol.
A polymorphism in the intestinal fatty acidbinding protein gene at amino acid residue 54 results in either alanine (A54) or threonine (T54). One study showed that T54/T54 subjects had greater total and LDL-cholesterol responses to a soluble-fiber diet than did the other 2 genotypes, and the heterozygotes had a significantly greater total cholesterol response to soluble fiber than did the A54/A54 homozygotes (89).
The Leu7Pro polymorphism in the neuropeptide Y gene has been investigated in one study in relation to dietary response; however, no significant effect was shown (90).
M/N blood groups have been associated with LDL-cholesterol responses to dietary change such that significantly different responses were noted between M/N subjects who responded the least and the 2 homozygote groups (blood groups M/M and N/N) who responded the most (91).
Magnitude of the response
Because of the heterogeneity in the type and duration of the interventions described, the magnitude of the lipid response to dietary interventions varied widely: in one study (46) the change in LDL cholesterol in the apo B EcoRI R-R- genotype group was as large as 59% of the baseline concentration. In the studies that showed a significant difference in response between genotype groups, the results also varied widely: in some studies, the difference in response between 2 genotype groups was 20% of the baseline lipid concentration (12, 46). However, the magnitude of these differences cannot be estimated with any accuracy, largely because most studies had only a small number of subjects in the rare genotype group.
The proportion of variance in the lipid response attributable to a single polymorphism is not likely to be > 10% (92). Therefore, individual genes contribute only a small part to the variation in the lipid response; however, when several genes are considered, the proportion of variance explained could be larger.
DISCUSSION
Evidence for a gene-diet interaction
Evidence suggests that variation in the genes for apolipoproteins A-I, A-IV, B, and E may contribute to the heterogeneity in the lipid response to dietary intervention. Specifically, carriers of the rare A allele of the apo A-I75 G/A polymorphism, the common Gln allele of the apo A-IV Gln360His polymorphism, the rare R- allele of the apo B EcoRI polymorphism, the common I allele of the apo B signal peptide I/D polymorphism, the apo B XbaI X- allele, and the apo 4 allele tend to show the greatest total and LDL-cholesterol responses to dietary change.
Many studies were unable to show significantly different responses between these genotype groups, and the genotypes showing the greatest response are not necessarily consistent between studies. For example, Tikkanen et al (40) found that subjects carrying the apo B XbaI X+ allele had greater decreases in total, LDL, and HDL cholesterol than did X-X- subjects in response to a diet low in fat and cholesterol, which contrasts with the results of other studies (42, 43, 46). Similarly, the results in Table 4 are not consistent, with E3/3 subjects showing a greater LDL response than carriers of the 4 allele after a MUFA-rich diet (73). Also, HDL-cholesterol responses have been shown to be greatest in persons carrying the 4 allele (54, 58) and also in E3/3 and E2/3 subjects (71) after changing the content of dietary fat, cholesterol, or both.
There was insufficient evidence to assess whether lipid responsiveness is affected by variation in the genes for apo C-III, lipoprotein lipase, hepatic lipase, cholesterol 7-hydroxylase, the LDL receptor, the cholesteryl ester transfer protein, or the intestinal fatty acidbinding protein. Although each of these gene products is essential in lipid metabolism, only a handful of studies have investigated variation in these genes, and most of these studies were unable to show significant gene-diet interactions.
Publication bias
Publication bias is a problem with any systematic review because "studies with results that are significant, interesting, from large well-funded studies, or of higher quality are more likely to be submitted, published, or published more rapidly than work without such characteristics" (95). Therefore, it is possible that other relevant dietary intervention studies with genotype information exist but were not included in this review because they have not been published. It is possible that the literature strategy for this systematic review missed studies because the genotype analyses were not mentioned in their title, abstract, or subject headings; a hand search of relevant journals may have identified more studies.
In the search for explanations for the heterogeneity in lipid responses, reviewers may tend to highlight studies showing significant effects of genetic variation while ignoring a large proportion of studies that found no such results. Studies showing nonsignificant or conflicting results cannot be ignored, especially because they outnumber the studies showing significant effects, notwithstanding the unpublished studies that could have nonsignificant and uninteresting results. Therefore, one has to ask the question "If genetic variability plays a role in the heterogeneity of lipid and lipoprotein responses to dietary change, why have so many studies been unable to demonstrate this with statistical significance?"
Possible reasons for conflicting results
There are many possible reasons why studies have been unable to show statistically significant gene-diet interactions. First, it is highly probable that lipid responses to dietary change are under polygenic control, with each gene contributing a relatively small effect. However, most studies have attempted to find only single-gene effects. For example, Gylling et al (28) found that the apo B EcoRI polymorphism had no effect on the LDL-cholesterol response to cholesterol feeding in all subjects, but in carriers of the apo 4 allele, those carrying the R+ allele showed significantly greater LDL-cholesterol responses than did the R-R- subjects. In the future, genotyping at multiple loci will be required to identify the best therapy for improving lipid and lipoprotein profiles.
Most of the studies summarized in this review lacked sufficient statistical power to detect any but a very strong effect because the sample sizes were too small, particularly for genotypes with low frequencies of the rare allele. However, many of the studies were retrospective and were not designed to examine gene-diet interactions, but data were reexamined after the availability of new information from genotype analyses. Therefore, it is perhaps not surprising that significant effects were not found in many studies because the numbers of individuals in each genotype group were so small. In many studies there were too few subjects homozygous for the rare allele to allow an analysis that would take into account differences in the response between heterozygotes and homozygotes. For apo E, where there are 6 possible genotypes, differences in the grouping of these could also lead to differences in results between studies. This illustrates that meta-analyses are important because they can detect effects with greater power and greater precision because of their inflated sample size (46, 71). In addition, in studies with small sample sizes, genotype misclassification of one individual may significantly affect the interpretation and validity of the results.
Conflicting results may also occur because of the different dietary protocols that were followed. The studies reviewed varied widely in the composition and length of the baseline and experimental diets. The dietary factors responsible for the changes seen in each genotype group are not clear because many studies modified several dietary factors, and so the dietary content in future studies should be tightly controlled and compliance must be strictly measured, ie, not only for cholesterol and the amount and type of fatty acids but also for other influential dietary components such as fiber and plant sterols. In addition, these studies investigated fasting lipid and lipoprotein concentrations; however, the effect of genetic variation may be more evident in the postprandial state than in the less-common fasting state.
Differences in the age, sex, body mass index, menopausal status, dietary backgrounds, and baseline lipid values of the participants could also have contributed to the discrepancies between the results. For instance, subjects with the 4 allele tend to have higher baseline total and LDL-cholesterol concentrations, and so greater responses in these subjects could reflect the regression to the mean phenomenon. However in Mantarri et als (65) study, baseline cholesterol concentrations were not significantly different between carriers and noncarriers of the 4 allele, showing that the greater total and LDL-cholesterol responses in the 4 group were not due to regression to the mean. It is also possible that weight change could account for differences in lipid and lipoprotein changes.
In addition, a significant effect may not reflect a causal relation but the allele may be in linkage disequilibrium with another one that does. For example, the base change that results in the XbaI site in the gene for apo B does not alter the amino acid, and so it may be in linkage disequilibrium with another functional mutation.
Conclusion
Evidence suggests that genetic variation may contribute to the heterogeneity in lipid responsiveness. At present, the evidence is limited but suggestive and justifies the need for future studies with much larger sample sizes based on power calculations, with carefully controlled dietary interventions, and that investigate the effects of polymorphisms in multiple genes rather than in single genes. Investigating gene-diet interactions will increase our knowledge of the mechanisms involved in lipid metabolism and improve our understanding of the role of diet in reducing cardiovascular disease risk.
REFERENCES