点击显示 收起
1 From the Department of Social Medicine, University of Bristol, United Kingdom (NJT, DAL, IND, SE, and GDS), and the Human Genetics Division, School of Medicine, University of Southampton, United Kingdom (MC, TRG, and IND)
2 The views expressed in this publication are those of the authors and not necessarily those of any of the funding bodies. The funding bodies have had no influence over the scientific work or its publication. 3 Supported by the Department of Health and British Heart Foundation (to the British Women's Heart and Health Survery); a UK Department of Health Career Scientist Award (to DAL), by UK Medical Research Council studentships (to NJT and TRG), and by the British Heart Foundation (to MC) . 4 Reprints not available. Address correspondence to NJ Timpson, Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol, BS8 2PR UK. E-mail: n.j.timpson{at}bris.ac.uk.
ABSTRACT
Background: Variation in the perception of bitter tastes has been associated with eating behavior, body composition, and cardiovascular disease. Recent observations have implicated 2 common haplotypes of TAS2R38 in the determination of bitter compound–tasting ability.
Objective: The objectives of the study were to examine, in the British Women's Heart and Health Study cohort, any association between TAS2R38 haplotypes, coronary heart disease (CHD), CHD risk factors, and eating behavior and to examine whether the associations allow for estimation of the effects of variation in diet on the etiology of common disease.
Design: We conducted a cross-sectional study of relations between TAS2R38 haplotypes and CHD, CHD risk factors, and eating behavior in 3383 women from 23 British towns.
Results: Genotyping at P49A and V262A in TAS2R38 (rs713598 and rs1726866) allowed construction of all 4 possible haplotypes. The 2 most common haplotypes corresponded with previously identified haplotypes related to bitter compound–tasting ability. No substantial evidence of association was found between these haplotypes and CHD (odds ratio for taste-defining haplotypes: 0.97; 95% CI: 0.78, 1.2), body mass index (difference between means of taste-defining haplotypes: –0.084; 95% CI: –0.45, 0.29), or a series of physiologic and dietary characteristics. A marginally lower risk of diabetes was observed among those with the nontaster haplotype than among those with the taster haplotype (odds ratio: 0.69; 95% CI: 0.48, 1.00).
Conclusion: TAS2R38 status was not an important determinant of CHD, related risk factors, or eating behavior in the British Women's Heart and Health Study sample.
Key Words: Taste • TAS2R38 • haplotype • behavior • coronary heart disease • CHD
INTRODUCTION
Bitter-taste perception is a classically variable trait both within and between human populations (1). The prevalence of "taste blindness" (ie, a lack of sensitivity to or an inability to taste bitter chemicals) ranges from 3% in West Africa to 6–23% in China and 40% in India; 30% of the white North American populations has taste blindness (1, 2).
Investigators recently reported that haplotypic variation in the region of chromosome 7 containing the TAS2R38 taste receptor gene shows a strong association with bitter compound–tasting ability (3). As such, variation across TAS2R38 is currently recognized as highly indicative of individual bitter compound–tasting ability (3).
An indirect effect of the ability to taste bitter compounds has been observed in the consequent avoidance of a host of bitter-tasting substances (4, 5). Many of these substances have antioxidant properties, and thus the tasting ability that leads to their avoidance has been implicated in the etiology of common disorders (6, 7). Although this association provides a hypothetical explanation for the involvement of bitter compound–tasting ability in disease, there is considerable debate about the role of antioxidant vitamins in the etiology of common conditions, particularly CHD (8).
However, observations suggested that lipid pathways involved in the etiology of CHD may be affected by tasting ability (9). Although it was not consistently reported (10), a preference for sweet and high-fat food was observed to decrease with increasing perception of bitter taste (11–13), and further research highlighted relations between bitter compound–tasting ability and body mass index (BMI; in kg/m2), adiposity levels, and risk factors for cardiovascular disease (14, 15). Given such evidence, the availability of genetic markers for tasting ability may offer insight into individuals' predisposition to CHD or CHD risk traits and an opportunity to use novel approaches to the dissection of disease-environment interactions.
Whereas conventional observational studies have suggested that dietary components are associated with CHD risk (16–24), such associations are prone to confounding factors and reverse causality (8). Mendelian randomization has been suggested as a potential method for overcoming these difficulties (25, 26). In Mendelian randomization, genetic markers predicting intermediate trait status (inherently randomized with respect to environmental factors) allow unconfounded and directed assessment of the role that lifetime exposure to environmental risk factors may play in the development of health outcomes. It was therefore hypothesized that, if genetic factors influencing dietary preference have an effect on the health status of individual persons, genetically predicted taster status could then be used to create unconfounded tests that are capable of assigning the direction of relation (ie, the ability to avoid reverse causation) of the association between eating behavior and CHD traits.
SUBJECTS AND METHODS
The British Women's Heart and Health Study
The British Women's Heart and Health Study (BWHHS) was designed to examine the causes and consequences of heart disease in women. Full details of this study were reported previously (27, 28). Between 1999 and 2001, 4286 women aged 60 to 79 y, randomly selected from 23 British towns, were interviewed and examined, and they completed medical questionnaires.
Methods used at baseline assessment were described previously (27, 28). Briefly, blood samples were taken after a minimum 6-h fast. These samples were used for assessment of insulin resistance [homeostasis model assessment (HOMA) calculated from fasting insulin and glucose concentrations] and lipids (29). Blood pressure, height and weight (used for calculating BMI), and waist and hip circumferences were measured by using standard procedures (28). Prevalent CHD was defined a medical record of myocardial infarction [based on World Health Organization diagnostic criteria (29)] or angina, a self-report of a physician's diagnosis of either of these conditions, or both (27). Prevalent diabetes was defined as a medical record of diabetes or a self-report of a physician's diagnosis. For both conditions, most cases were identified by both medical records and self-reporting (27).
In the baseline questionnaire, women were asked to indicate the frequency of consumption of a number of foods. Green vegetable consumption was divided into 6 categories: never, <1 time/wk, 1–2 d/wk, most days, 1 time/d, and >1 time/d. For the purpose of the analyses presented here, these 6 categories were collapsed into 2 categories: green vegetable avoiders (the first 3 categories of consumption frequency) and green vegetable eaters (the last 3 categories of consumption frequency). Women were also asked which type of fat they normally used for spreading. In efforts to assess preference for plain fat consumption, these groups were then also collapsed to represent those who reported the use of spreadable fat and those who reported not using spreadable fat.
Alcoholic beverage consumption also was divided into 6 categories: never, on special occasions, once or twice a month, weekends only, most days, and daily. Again, for the purpose of analyses, these 6 categories were collapsed into 2 categories: consumption of alcohol at any frequency and no consumption of alcohol.
Participants were asked for informed consent to the reviewing of their medical records and to the performance of anonymous genetic tests relating to cardiovascular disease on stored blood. Eight women declined to give consent and were not included in this study. Approval for the study protocol was obtained from local ethics committees in the 23 British towns and from the UK Multicentre Ethics Committee.
Determination of TAS2R38 genotypes
A salting-out procedure was used to extract DNA from whole blood or cell residues that had been stored in tubes with K-EDTA at –80 °C for 1–2 y (30). Quantitation was done with the use of the PicoGreen assay (Molecular Probes, Eugene, OR), and DNA concentrations were equalized by dilutions with water. Long-term stock DNA aliquots were placed in storage, and working 96-well plates of DNA dilutions to 10 ng/µL were prepared. Degenerate oligo primer (DOP) amplifications were made from dilution plates to conserve stock DNA, and 384-well polymerase chain reactions (PCRs) were performed from DOP-DNA representing 1 ng of original genomic DNA. The DOP protocol was a modified version of the method used by Cheung and Nelson (31) to minimize the loss of representation of GC-rich genomic regions.
The TAS2R38 (P49A and V262A) genotypes were determined by using the melting of fluorescence-labeled oligonucleotide from matched or mismatched target, which was monitored in a 384-well Odyssey post-PCR thermal ramp (Idaho Technology, Salt Lake City, UT). Detection used reduction of dabcyl quenching of fluorescence during a thermal ramping. Asymmetric PCR was performed on 2 µL of dried DOP-amplified template in 384-well white PCR plates (Abgene, Epson, United Kingdom) on a MJ Research PTC-225 DNA Engine Tetrad (Genetic Research Instrumentation Ltd, Braintree, United Kingdom). The TAS2R38 P49A variant was amplified with the use of the primers 5'-GCCAGAGGTTGGCTTGGTTTGCA-3' at 100 nmol/L and 5'-CCTGGAGTTTGCAGTGGGGTTTCTGACCA-3' at 500 nmol/L. A fluorescein isothiocyanate (FITC)-labeled probe (with 3' phosphate) of 5'-FITC-CTCAGTGCCTGCCTCT-PHOS-3' matching the wild-type sequence and an adjacent dabcyl quencher, 5'-GAGACACAGCAGCACACAATCACTGT-DABCYL-3', were included at 200 nmol/L in the PCR for the Odyssey melting assay. The TAS2R38 V262A variant was amplified with the use of the primers 5'-TGCCCAGAGGGAC-AGCTGCCATT-3' at 100 nmol/L and 5'-TGGGAAGGCACA-TGAGGACAATGAAGG-3' at 500 nmol/L. An FITC-labeled probe (with 3' phosphate) of 5'-FITC- GAAGGCAGCACAGG-ATG-PHOS-3' matching the wild-type sequence and an adjacent dabcyl quencher, 5'-GCCACAGAATCAGTAGGGGCACA-GAG-DABCYL-3', were included at 200 nmol/L in the PCR for the Odyssey melting assay.
The 5-µL portion of PCR mix also contained 1x PCR buffer, 0.2 mmol deoxynucleotide triphosphate/L, 1.5 mmol MgCl2/L, and 0.1 unit Taq DNA polymerase (all: Promega, Southampton, United Kingdom). PCR cycling conditions were 94 °C for 2 min and then 99 cycles of 94 °C for 30 s, 62 °C for 30 s, and 72 °C for 30 s, which were followed by a cycle at 72 °C for 2 min. Samples were overlaid with 5 µL Chill-Out wax (Genetic Research Instrumentation) to prevent evaporation during analysis. After PCR amplification, samples were melted from 45 °C to 75 °C in the 384-well Odyssey. LIGHTTYPER software [Gaunt TR, Hinks LJ, Christensen MB, Kiessling M, Day INM. Uses of the LightTyper in human genotype analysis: SNPs, microhaplotypes and large insertion/deletions. Biochemica (in press)] was used to analyze the change in fluorescence during melting and to group melting profiles into genotype groups. These values were then verified manually by using in-house software.
Haplotype construction
Haplotypes and predicted taster status identified in relation to previous literature (ie, a person's status as either a taster or a nontaster of bitter compounds) were summarized in Table 1. The haplotype names PAV and AVI refer to recognized taster and nontaster states, respectively (3), and are specifically derived from their protein coding sequences. In the context of this study, haplotypes AVI and PAV have been renamed AV and PA, respectively. A person was therefore designated a nontaster if he or she carried 2 copies of AV and a taster if he or she carried 1 copy of PA.
View this table:
TABLE 1. Prediction of tasting ability by TAS2R38 haplotype1
Previous results showed the existence of the variant site TAS2R38 V296I (34 bases from TAS2R38 V262A) in total linkage disequilibrium with TAS2R38 V262A in a European population (3). In light of this, it was felt appropriate that the genotyping of TAS2R38 P49A and TAS2R38 V262A alone would allow the effective tagging of common haplotypes in this region.
When genotype data were collected, haplotypes were constructed by using the PHASE software program (version 2.02; Internet: http://www.stat.washington.edu/stephens/phase.html; also: 32, 33). This software employs a Bayesian method for the reconstruction of chromosomal phase by using genotype data, and it generates counts and frequencies of observed haplotypes. In cases with one missing genotype, PHASE software was used to infer the haplotypic value. The underlying method in this approach is a Markov Chain-Monte Carlo procedure in which the probability of preceding observations (in this case, unambiguous phase information) allows population genetic inference about unresolved haplotypic phase.
Our prior hypothesis, which was based on the report of Kim et al (3), was that CHD trait and eating behavior phenotypes would differ according to taster status as defined by haplotypic complement (homozygous for AV = nontaster; heterozygous or homozygous for PA = taster; other = excluded). This model formed the basis of test 1, in which 15 phenotypes, some interrelated, were coded for analysis. Subsequent post hoc analysis (test 2) was a more extreme comparison between those subjects who were homozygous for AV (conventionally recognized nontasters) and those who were homozygous only for PA (a select proportion of those haplotypically defined as tasters).
Statistical analysis
On the basis of observed total allele combinations, genotype distributions for single-nucleotide polymorphisms were examined for consistency with Hardy-Weinberg equilibrium by using chi-square tests. Simple two-sample t tests with unequal variances were used to compare mean values of continuous variables by haplotype-derived category. Chi-square tests were also used to compare proportions of dichotomous variables by haplotypic category. In the case of dichotomous variables, logistic regression was used to estimate odds ratios (ORs) for outcomes in question. Given their skewed distribution, HOMA scores and triacylglycerol concentrations were log transformed to approximate a normal distribution for analyses. As such, the analyses used transformed data, but the data presented are geometric means, 95% CIs, and the proportionate difference between these haplotype-derived taster groups (ie, tasters and nontasters). All statistical analyses were performed with STATA data analysis software (version 8; Stata Corp, College Station, TX).
RESULTS
The British Women's Heart and Health Study
Of the 4286 BWHHS participants at baseline, 441 had either no blood samples taken (because venipuncture did not succeed) or had insufficient blood taken for adequate samples to be stored. An additional 37 women refused consent for the use of stored blood. Of the remaining 3808 participants, DNA from 255 was not available for assay at the time of genotyping. It is notable that these participants did not differ in any systematic way from the women for whom DNA was available. A total of 3383 samples qualified for analyses because successful genotypes had been scored for at least P49A and V262A at TAS2R3. In the BWHHS sample, the minor allele frequencies for P49A and V262A were found to be 0.4 and 0.44, respectively (Table 2). These variants were in Hardy-Weinberg equilibrium, and the frequencies observed approximately match those reported by Kim et al (3).
View this table:
TABLE 2. Allele frequencies at variant loci TAS2R38 P49A and TAS2R38 V262A observed in the British Women's Heart and Health Study sample
Haplotype construction identified 3 main haplotypic combinations corresponding with published work (3). PHASE successfully replaced missing genotypes (n = 399 over both single-nucleotide polymorphisms) with a corresponding 85% probability of accuracy in any cases. The AV haplotype [equivalent to the common AVI (nontaster) haplotype observed by Kim et al (3) and Wooding et al (34)] accounted for 56% of observed haplotypes, whereas the PA haplotype [equivalent to the common PAV (taster) haplotype also observed in the same studies] accounted for 40% of all haplotypes (Table 3).
View this table:
TABLE 3. Haplotype construction and comparative frequencies for the TAS2R38 locus in the British Women's Heart and Health Study (BWHHS) sample and as reported by Kim et al1
The rarer haplotype, AA, has been recognized previously (3) and is here termed AAV, as a novel recombinant at low frequency. In the BWHHS sample, this haplotype was observed at a frequency of 4%. Approximately 0.1% of samples were inferred by PHASE to carry the previously unrecognized haplotype PV, but these inferences and primary data have not been followed up.
Association between TAS2R38 haplotypes and CHD traits and eating behavior
There was no evidence of association between predicted taster status and CHD; BMI; waist-hip ratio; serum HDL, LDL, or triacylglycerol concentration; or HOMA score (Table 4; Table 5; Table 6; Table 7). There was a marginal tendency for the underrepresentation of diabetes among nontasters (test 1, OR: 0.69; 95% CI: 0.48, 1; P = 0.05; test 2, OR: 0.55; 95% CI: 0.35, 0.87; P = 0.01).
View this table:
TABLE 4. Relation between coronary heart disease (CHD), CHD risk variables, eating behavior, and TAS2R38 haplotypes in the British Women's Heart and Health Study1
View this table:
TABLE 5. Relation between continuous coronary heart disease (CHD) risk variables and TAS2R38 haplotypes in the British Women's Heart and Health Study1
View this table:
TABLE 6. Relation between coronary heart disease (CHD), CHD risk variables, eating behavior, and TAS2R38 haplotypes in the British Women's Heart and Health Study1
View this table:
TABLE 7. Relation between continuous coronary heart disease (CHD) risk variables and TAS2R38 haplotypes in the British Women's Heart and Health Study1
Dietary traits hypothesized to be associated with predicted bitter compound–tasting ability also failed to show significant segregation with TAS2R38 haplotypes either in test 1 or test 2. The avoidance of fats, alcohol, and green vegetables all yielded null results with respect to potential association with TAS2R38 variation. It is notable that, in the case of green vegetables, more extreme groupings of avoided vegetables were analyzed (data not shown). These groupings did not show marked differences in proportion by TAS2R38 haplotype (although these analyses were imprecise as a result of small numbers in extreme cells).
DISCUSSION
It has been hypothesized that a genetic predisposition to impaired bitter compound–tasting ability should result in a lifelong increase in exposure to a host of classically recognized CHD risk factors (13, 15), including a preference for fatty foods and an associated risk profile of body composition and lipid profile (14). However, in this study, no strong association was observed between TAS2R38 haplotypes and either CHD traits or eating behavior.
Those haplotypically predicted to be nontasters and tasters did show a difference in the prevalence of diabetes. However, the greater prevalence of diabetes in tasters was not consistent with previously reported dietary preferences of tasters and nontasters, which would predict an association in the opposite direction (13–15). Furthermore, whereas some of the 15 variables explored under a prior hypothesis (test 1) are not truly independent of each other, findings would not withstand Bonferroni correction for multiple testing.
There are 3 likely explanations for a lack of association with either CHD, CHD risk factors, or eating behavior. Whereas our study had >95% power to detect differences of 0.5 SD in continuous variables relating directly to our principal hypothesis (eg, BMI, LDL, HDL, triacylglycerols, and waist-hip ratio), there are alternative explanations for a null result. First, it is possible that errors present in the measurement of either our phenotypic or genotypic data (or both) may have resulted in misleading findings. However, the similarity of allelic and haplotypic frequencies to those already published (3) and the existence of Hardy-Weinberg equilibrium in our genetic data suggest that genotyping errors are unlikely. In phenotypic data, our observed prevalence of CHD, diabetes, and CHD risk factors are similar to those in female participants of the same age in the Health Survey for England (27). Furthermore, expected associations between established CHD risk factors and CHD are of the same magnitude and direction as those observed elsewhere (27). As such, it seems unlikely that errors in the determination of genotypes or the measurement of phenotypes are a likely explanation for our observations.
Failure to detect the influence of TAS2R38 haplotypes on eating behavior may reflect limitations in the assessment of bitter compound–tasting ability. The standard dietary questionnaire responses represent a broad analysis of dietary preferences and are far from being a fully qualitative and quantitative representation of exact dietary intake. Results from the BWHHS questionnaire relating to select dietary intake measures and vitamin C consumption showed similarities with those of the European Prospective Investigation into Cancer and Nutrition Study (35) (comparisons unpublished), but specific variants such as those of TAS2R38 may mark dietary patterns that are not evident in general questionnaires but that require more a specific approach to analysis in any future study. Further investigation would be aided by the direct testing of bitter compound–tasting ability through the use of more precise physiologic methods (36, 37).
Second, the association between TAS2R38 haplotypes and bitter compound–tasting ability may not be straightforward (Figure 1A). TAS2R38 is only one of the many genetic determinants likely to be involved in the pathways determining taste perception. The observation that variation at this locus correctly predicts only 80% of phenylthiocarbamide tasters (3) bears direct relation to this and is likely a testament to the existence of numerous members of the TAS2R gene family clustered on chromosomes 5, 7, and 12. Tasting ability is also likely to be influenced by many other sensory and proprioceptive pathways, and the probable result is that no single genetic marker has a great effect. In particular, these other pathways are likely to include olfactory contributions to food preference, although digestive and cognitive factors may also complicate the overall system and modify the ability to perceive bitter taste.
FIGURE 1.. The hypothesized relation between TAS2R38 haplotypes and health outcome (A) and the alternative possible explanations for the results observed in the current study (B and C). Diagram B represents a challenge to the existence of any relation between TAS2R38 genetic variation and differential ability to taste bitter compounds. Diagram C represents a challenge to the assumption that differential ability to taste bitter compounds will have any real effect on the dietary intake of an individual person.
Third, under the assumption that more accurately assessed bitter compound–tasting ability is associated with TAS2R38 haplotypes (3), the postulated relation between bitter compound–tasting ability, dietary preference, and health outcome becomes questionable (Figure 1C; 10). Given reliable phenotypic variables and accurate genetic data, our results indicate a lack of relation between tasting ability and CHD and CHD risk factors. The major implication of this (beyond the suggestion not to use genetically predicted bitter compound–tasting ability to assess the effects of dietary composition on CHD and CHD traits) is that the inherent ability to detect bitter compounds is not likely to be an important intermediate phenotype and hence a predisposing factor for CHD.
Moreover, there was a surprising lack of association between taster status and the consumption of green vegetables. Sensitivity to the bitter flavors of cruciferous vegetables is widely recognized as a common phenotype in those able to taste bitter compounds. Despite having direct questions relating to the frequency of consumption of green vegetables, we failed to observe any association with this trait. Given the lack of association with other traits, this finding may well be due to the explanations offered above, although the possibility of effect masking through sociocultural characteristics of the sample population remains. For example, older women raised in times of austerity might be expected to have less-marked food preferences. Furthermore, "debittering" through cooking or the addition of salt, sugar, or fat may effectively reduce the effect of many substances said to elicit a bitter taste response (38).
This investigation of a large, population-based sample of women has not provided support for work associating taster status with the etiology of CHD, CHD-related risk factors, or eating behavior (13–15). This lack of association between TAS2R38 gene variation and either phenotypic outcome or intermediate variable did not permit us to test potential unconfounded relations between eating behavior, CHD, and CHD risk factors by using the Mendelian randomization model (25, 26). Furthermore, larger-scale analyses in different, reasonably sized samples are required if the TAS2R38 locus is to be confirmed as having common involvement with the tasting ability and CHD risk factors. The combination of this approach with the gathering of more complete genomewide taste and olfactory receptor diversity data, and the taking of direct measures of taste perception would allow the assessment of the genetic foundations and implications of tasting ability.
ACKNOWLEDGMENTS
We thank Carol Bedford, Alison Emerton, Nicola Frecknall, Karen Jones, Rita Patel, Mark Taylor, and Katherine Wornell for collecting and entering data; all of the general practitioners and their staff members who supported the data collection; and the women who participated in the study. The British Women's Heart & Health Study is codirected by Shah Ebrahim, Peter Whincup, and Goya Wannamethee.
All authors developed the study aim and design. IND, together with SE and DAL, obtained funding for the DNA bank and genotyping. Genotyping was done by MC, and DNA bank and data management was done by TRG. NJT undertook the initial analysis and coordinated the writing of the manuscript. All authors contributed to the final version of the manuscript. NJT and IND act as guarantors. DAL is a codirector of the British Women's Heart & Health Study. None of the authors had a personal or financial conflict of interest.
REFERENCES