Heritability of fat intake

The considerable individual differences that undoubtedly exist in overall food intake, as well as in the intake of particular foods or specific nutrients, are partly explained by genetic variation. Relatively few studies have been undertaken to identify the genes associated with macronutrient intake, but familial aggregation of intakes has been demonstrated. The magnitude of the reported genetic effects differs from study to study (due to different populations and methods), but typically ranges from about 20 to 40 %. For fat intake measured as a percentage of energy intake, the correlation was 0·61 for monozygotic twins, 0·59 for dizygotic twins and 0·36 for siblings⁽ Reference Rankinen and Bouchard ^{9 )}. Monozygotic twins share 100 % of their genes, so the magnitude of correlations shows the degree of the influence of genes on a trait of interest. Patterns for family correlations for fat intake may depend on the method of data collection. Values from a food diary method were higher than those from FFQ, with the mean H ² values being 0·33 and 0·16, respectively. Due to methodological issues (for example, the limited number of possible responses and broad generalisation of food categories), FFQ may poorly represent actual overall intake, which can lead to both overestimation and underestimation of macronutrient intake⁽ Reference Cade, Thompson and Burley ^{10 )}, which is in this case an analysed phenotype. The results may also be affected by age-related effects, because family correlations among individuals of the same generation were higher (with a mean value of 0·40) than for individuals of different generations (with a mean value of 0·24)⁽ Reference Reed and Knaapila ^{11 )}.

The H ² values for the preference for fat or high-fat foods have been estimated in a few studies. Some results demonstrate that the preference for dietary fats is independent of genetic factors⁽ Reference Rissanen, Hakala and Lissner ^{12 )}, but high H ² (0·78) was shown for liking meat and fish⁽ Reference Breen, Plomin and Wardle ^{13 )}.

Genome-wide approach

Genome-wide linkage studies have identified several chromosome regions for macronutrient and energy intake. The Quebec Family Study has identified evidence for the presence of six quantitative trait loci (QTL) that influence total energy and macronutrient intakes. The best evidence of linkage was found at region 3q27·3, where one of the markers was linked to energy, lipid and carbohydrate intakes. A candidate gene located in this region is the adiponectin gene (ADIPOQ)⁽ Reference Choquette, Lemieux and Tremblay ^{14 )}. In the San Antonio Family Heart Study, a QTL for macronutrient consumption (total energy, total protein, total fat, and saturated and unsaturated fats) was identified on chromosome 2p22. The pro-opiomelanocortin gene (POMC) was tested as a candidate, but no association was demonstrated⁽ Reference Cai, Cole and Bastarrachea ^{15 )}. In the Health, Risk Factors, Exercise Training and Genetics (HERITAGE) Family Study, the strongest evidence of linkage for energy intake appeared on chromosomes 1p21·2 and 20q13·13, and for fat intake on chromosome 12q14·1⁽ Reference Collaku, Rankinen and Rice ^{16 )}.

More recently, a few genome-wide association studies (GWAS) have been undertaken to identify the loci associated with macronutrient intake (http://www.ebi.ac.uk/gwas/)⁽ Reference Welter, MacArthur and Morales ^{17 )}. A two-stage genome-wide association meta-analysis of macronutrient intake in populations of European descent has been carried out⁽ Reference Tanaka, Ngwa and van Rooij ^{18 )}. In this study, the H ² estimates for protein, carbohydrate and fat intakes were 17, 20 and 20–23 %, respectively. Genome-wide significant associations for fat intake were observed on 19q13·33. The minor allele of rs838145 was associated with a lower percentage of energy intake from fat. The FGF21 (fibroblast growth factor 21) gene was proposed as a candidate located in this region, and it was shown that its minor allele was associated with higher concentrations of FGF21 protein. Moreover, the BMI-increasing allele (rs1421085) of the FTO (fat mass and obesity-associated) gene was found to be associated with higher protein intake⁽ Reference Tanaka, Ngwa and van Rooij ^{18 )}. A study based on the DietGen Consortium identified rs838133 in FGF21 (19q13·33), rs197273 near TRAF family member-associated NF-κB activator (TANK) (2p24·2), and rs10163409 in FTO (16q12·2) as the top associations for the percentage of total energy intake from protein and carbohydrates⁽ Reference Chu, Workalemahu and Paynter ^{19 )}. A genome-wide significant association for the percentage of total energy intake from carbohydrates was identified in SNP in an intron of the FTO gene. The rs838133 variant at the FGF21 locus was associated with increased carbohydrate intake and decreased fat intake. A GWAS of adolescents from a French–Canadian population identified the OPRM1 (opioid receptor mu 1) gene as a fat intake gene⁽ Reference Haghighi, Melka and Bernard ^{20 )}. The minor OPRM1 allele (rs2281617) was associated with lower fat intake (by 4 %) and lower body fat mass (by about 2 kg). The rs822396 SNP in intron 1 of the adiponectin (ADIPOQ) gene, which encodes adiponectin, was identified as being associated with confectionery intake⁽ Reference Wakai, Matsuo and Matsuda ^{21 )}. This polymorphism was not directly associated with circulating adiponectin levels, but could be linked with other causative SNP. Heritability analyses have shown that the common SNP tested in this study explain a modest proportion (6–8 %) of the genetic variance in carbohydrate, protein and fat intake.

Genome-wide association studies on fat intake

In 2014, the results of the above-mentioned GWAS, which sought loci associated with fat intake, were published⁽ Reference Haghighi, Melka and Bernard ^{20 )}. One gene found to have a great effect on fat intake and body fat mass is the OPRM1 gene, which encodes a receptor expressed in the brain’s reward system. This opioid receptor is also the primary receptor for the endogenous opioid peptide β-endorphin, encoded by the POMC gene, which is well known for its function in appetite regulation. The rs2281617 and rs518596 polymorphisms of the OPRM1 gene have been associated with fat intake and body adiposity⁽ Reference Haghighi, Melka and Bernard ^{20 )}. Moreover, the results suggest that these polymorphisms are not functional SNP. The questions that remain are whether such associations also exist in other populations, and which of them are causative mutations. One of the SNP identified in the OPRM1 gene is the A118G functional polymorphism, which affects this gene’s mRNA and protein yield⁽ Reference Zhang, Wang and Johnson ^{22 )}. The potential role of this polymorphism in the preference for fatty foods was examined by Davis et al. ⁽ Reference Davis, Zai and Levitan ^{23 )}, who measured fat preference with the food preference questionnaire. The food preference questionnaire reflects attitudes to different food items. The researchers suggested that part of the variability in the preference for palatable foods can be explained by this polymorphism. However, this result needs to be replicated in other populations and with more ecologically valid assessments of food preferences⁽ Reference Davis, Zai and Levitan ^{23 )}. The effects of the A118G polymorphism on the frequency of high-fat food consumption, obesity and lipid metabolism have not yet been determined. Interestingly, it has been shown that prenatal exposure to cigarette smoking interacts with the OPRM1 genotype (rs2281617), with T-allele carriers showing lower fat intake in non-exposed individuals, but not in exposed individuals⁽ Reference Lee, Abrahamowicz and Leonard ^{24 )}.

Functional candidate genes

Several functional candidate gene approaches have also addressed the question of genetic variation in fat intake. Some of these have focused on genes involved in the central control of food intake, including the melanocortin 4 receptor (MC4R) gene⁽ Reference Khalilitehrani, Qorbani and Hosseini ^{25 ,} Reference Yilmaz, Davis and Loxton ^{26 )}, dopamine receptor 2 (DRD2)⁽ Reference Epstein, Temple and Neaderhiser ^{27 )}, the 5-HT_2A receptor gene⁽ Reference Aubert, Betoulle and Herbeth ^{28 )} and the FTO gene⁽ Reference Loos and Yeo ^{29 )}. The role of the FTO gene is discussed in more detail in the next section.

One gene with a great effect on body mass is the MC4R gene; this encodes a receptor for melanocortin, which binds α-MST – a product of the POMC gene. The rs17782313 polymorphism near the MC4R gene has been found to be associated with obesity among European adults⁽ Reference Xi, Takeuchi and Chandak ^{30 )}, but also with high intakes of total energy and total fat⁽ Reference Qi, Kraft and Hunter ^{31 )}. Yilmaz et al. ⁽ Reference Yilmaz, Davis and Loxton ^{26 )} showed that rs17782313 is significantly associated with depressed mood and overeating behaviours, while Khalilitehrani et al. ⁽ Reference Khalilitehrani, Qorbani and Hosseini ^{25 )} have reported an association between the same polymorphism and fat intake (Table 1).

Table 1 Candidate gene studies on fat intake and fat sensitivity

DHQ, diet history questionnaire; APOA5, apolipoprotein A-V; FTO, fat mass and obesity-associated; MC4R, melanocortin 4 receptor; OPRM1, opioid receptor mu 1; NEGR1, neuronal growth regulator 1; TMEM18, transmembrane protein 18; BDNF, brain-derived neurotrophic factor; KCTD15, K channel tetramerisation domain containing 15; ADIPOQ, adiponectin; LEP, leptin; LEPR, leptin receptor; USDA, United States Department of Agriculture; OA, oleic acid; CD36, cluster of differentiation 36.

It is worth mentioning that associations between genetic variants and traits may change over time. For example, the association between the FTO genotype and BMI strengthens during childhood and adolescence, reaching its peak at the age of 20 years. Similarly, the MC4R polymorphism shows an association with body weight that is strong during childhood and adolescence, but which weakens with increasing adult age⁽ Reference Hardy, Wills and Wong ^{32 ,} Reference Jacobsson, Almén and Benedict ^{33 )}.

FTO is the gene with the greatest effect on body mass

FTO was the first obesity-susceptibility gene to be identified in a GWAS, and is the gene with by far the largest effect and which explains the largest phenotypic variance among individuals of European ancestry⁽ Reference Frayling, Timpson and Weedon ^{34 ,} Reference Scuteri, Sanna and Chen ^{35 )}. A cluster of SNP in the first intron of FTO has been identified (linkage disequilibrium r ²>0·80); they are all significantly associated with BMI⁽ Reference Loos and Yeo ^{29 )}. Each additional minor allele is associated with a 0·39 kg/m² higher BMI and a 1·20-fold increase in the risk of obesity⁽ Reference Berndt, Gustafsson and Mägi ^{36 )}. Approximately 43 % of the population of European ancestry carry one minor allele, while 20 % carry two minor alleles⁽ Reference Loos and Yeo ^{29 )}. The FTO locus accounts for 0·34 % of interindividual variation in BMI⁽ Reference Speliotes, Willer and Berndt ^{37 )}.

The biological role of FTO has recently been described. The FTO gene encodes for the Fe(II) and 2-oxoglutarate-dependent demethylase of single-stranded DNA and RNA⁽ Reference Gerken, Girard and Tung ^{38 )}, and is capable of demethylating single-stranded DNA and RNA at m6A, m3U or 3 mT⁽ Reference Speakman ^{39 )}. It has been shown that FTO itself regulates body weight and that its overexpression leads to obesity⁽ Reference Church, Moir and McMurray ^{40 ,} Reference Merkestein, McTaggart and Lee ^{41 )}, while Fto^–/– mice show postnatal growth retardation and are resistant to high-fat diet-induced obesity⁽ Reference Fischer, Koch and Emmerling ^{42 )}. The role of FTO in the cellular sensing of amino acids has been described⁽ Reference Gulati, Cheung and Antrobus ^{43 )}. The sensing of amino acid levels in the brain regulates the orexigenic and anorexigenic pathways controlling energy balance⁽ Reference Cota, Proulx and Smith ^{44 )}. FTO is highly expressed in the hypothalamus, but data on the regulation of its expression by nutritional status are confusing⁽ Reference Olszewski, Radomska and Ghimire ^{45 –} Reference Fredriksson, Hägglund and Olszewski ^{48 )}. It has been shown in a rodent model that food deprivation up-regulates hypothalamic Fto expression, and that these changes are associated with cues related to energy intake rather than with feeding reward⁽ Reference Olszewski, Fredriksson and Olszewska ^{49 )}. It was recently described that FTO affects body weight by regulating thermogenesis⁽ Reference Claussnitzer, Dankel and Kim ^{50 ,} Reference Herman and Rosen ^{51 )}. There are direct interactions between the FTO locus and the distant IRX3 (Iroquois-related homeobox 3) and IRX5 (Iroquois-related homeobox 5) regions, which are developmental regulators affecting adipocyte differentiation. The FTO variants regulate the expression of IRX3 and IRX5, and the risk allele of the causal SNP rs1421085 disrupts a conserved motif for the ARID5B repressor, which doubles the expression of IRX3 and IRX5 during adipocyte differentiation. One result of this is a developmental shift from beige to white adipocytes and a consequent reduction in mitochondrial thermogenesis and increase in lipid storage⁽ Reference Claussnitzer, Dankel and Kim ^{50 ,} Reference Herman and Rosen ^{51 )}.

Several studies have shown that FTO polymorphism is associated with food intake and body mass, but a comprehensive understanding of how it affects the functioning of the body still needs to be reached through investigations. Subjects homozygous for a risk A allele of FTO rs9939609 have dysregulated levels of the orexigenic hormone acylghrelin and attenuated postprandial appetite reduction⁽ Reference Karra, O’Daly and Choudhury ^{52 )}. In a 2-year trial entitled ‘Preventing Overweight Using Novel Dietary Strategies’, it was shown that dietary protein significantly modifies the genetic effects on food cravings and appetite scores; in particular, this risk A allele of FTO (rs9939609) was associated with a greater decrease in food cravings and appetite scores in participants with high-protein diet intakes⁽ Reference Huang, Qi and Li ^{53 )}. Brunkwall et al. ⁽ Reference Brunkwall, Ericson and Hellstrand ^{54 )} indicated that carriers of the risk A allele of FTO reported a higher consumption of biscuits and pastry, but a lower consumption of soft drinks, than TT genotype carriers. They thus concluded that FTO polymorphism may be associated with certain food preferences. Most studies on food intake and FTO variation were conducted in children and have shown that individuals carrying the A allele at rs9939609 consumed more fat and total energy than those not carrying the variant⁽ Reference Timpson, Emmett and Frayling ^{55 )}. Some of the studies did not find such an association. For example, Hakanen et al. ⁽ Reference Hakanen, Raitakari and Lehtimäki ^{56 )} concluded that the FTO gene is not directly associated with energy intake at age 15 years (Table 1). Similarly, there was no evidence for association between the risk A allele and dietary energy density⁽ Reference Johnson, van Jaarsveld and Emmett ^{57 )}. Hardy et al. ⁽ Reference Hardy, Racette and Hoelscher ^{58 )} also concluded that the relationship between FTO variants and BMI does not occur primarily through the mediation of food intake. On the contrary, Cecil et al. ⁽ Reference Cecil, Tavendale and Watt ^{59 )} reported an association between the risk A allele and increased energy intake. A combined analysis of over 16 000 children and adolescents has suggested that the risk A allele is associated with higher total energy intake, and that lower dietary protein intake attenuates the association between the FTO genotype and adiposity⁽ Reference Qi, Downer and Kilpeläinen ^{60 )}. Moreover, in the Multiethnic Cohort Study, percentage of energy from fat was a partial mediator of the rs8050136 effect on BMI⁽ Reference Park, Cheng and Pendergrass ^{61 )}. A randomised cross-over trial in forty overweight men showed that FTO polymorphism is related to variation in the feeling of postprandial fullness⁽ Reference Dougkas, Yaqoob and Givens ^{62 )}. Together, these may suggest that interactions between the intakes of different macronutrients and micronutrients are important for the overall results. Further studies are needed for a comprehensive understanding of FTO biology in different cell types. It could be hypothesised that, in addition to a developmental shift favouring lipid-storing adipocytes over beige adipocytes, FTO may exert its effect on body mass through the regulation of appetite and food preference⁽ Reference Loos and Yeo ^{29 )}.

Taste as a component of fat preference and fat intake

Fat consumption varies across individuals, and excess dietary fat consumption can be caused by a number of factors, including environmental triggers (for example, the broad availability of energy-dense foods) and the psychological, physiological and metabolic properties of an organism that depend on the many genes encoding hormones, enzymes and receptors involved in the regulation of food intake. However, the factors that contribute to increased fat intake are not well understood⁽ Reference Keller, Liang and Sakimura ^{91 )}.

A preference for a certain food is defined as the selection of one food item over others when liking is the basis for the selection, though it may be only one of the motives⁽ Reference Birch ^{92 )}. A greater preference for fatty foods, as well as increases in the consumption of such food, have been documented in obese subjects⁽ Reference Drewnowski and Holden-Wiltse ^{93 )}. It should be borne in mind that food preferences and food intake may not be correlated with each other⁽ Reference Drewnowski and Hann ^{94 )}. Since food preferences are just one component of the food decision-making process, they usually provide only an approximation of actual food consumption⁽ Reference Drewnowski and Hann ^{94 ,} Reference van Meer, Charbonnier and Smeets ^{95 )}. Food choices depend on genetic and environmental determinants, with the latter including food availability and accessibility, as well as the social and cultural environment, but also several economic factors. Individual biological predispositions depend on several mechanisms – partly dependent on genotype – which determine the regulation of appetite as well as taste and sensory sensitivity⁽ Reference Contento ^{96 ,} Reference Wansink ^{97 )}. The sensory qualities of food are critical to dietary preferences, and taste may be one of the most important determinants of food choice⁽ Reference Garcia-Bailo, Toguri and Eny ^{98 )}. The fat content of food contributes to sensory properties that can guide food choice and energy intake⁽ Reference Running, Mattes and Tucker ^{99 )}. However, sensory responses alone do not predict food consumption⁽ Reference Drewnowski ^{100 )}. As mentioned by Mattes, several recent observations have drawn attention to the links between oral fat detection, fat intake, lipid metabolism and chronic disease risk⁽ Reference Mattes ^{101 )}.

Another important question is whether body mass affects fat sensitivity. However, experimental support for a hypothesis relating fat taste to fat intake and BMI remains equivocal⁽ Reference Running, Mattes and Tucker ^{99 )}. The most commonly tested hypothesis states that decreased fatty acid sensitivity leads to increased fat intake and obesity, and the results of several studies have supported this hypothesis⁽ Reference Stewart, Feinle-Bisset and Golding ^{102 –} Reference Stewart, Newman and Keast ^{105 )}. However, factors other than adiposity status – including genotype, salivary composition, and habitual or acute dietary fat intake – may influence fat taste intensity ratings⁽ Reference Tucker, Nuessle and Garneau ^{106 )}. In obese individuals, portion control or a low-fat diet may increase fat sensitivity, but the low-fat diet had the greatest effect⁽ Reference Newman, Bolhuis and Torres ^{107 )}. It has been reported that obesity may shift the preference for oily solutions and orosensory detection of NEFA in diet-induced obesity (DIO) rats and mice⁽ Reference Chevrot, Bernard and Ancel ^{108 )}. The results suggest that, during behavioural tests, obese animals have a lesser ability to detect fatty acids through a cluster of differentiation 36 (CD36)-mediated mechanism than do lean animals.

Fat taste

Taste is a chemical sense whose mechanism involves chemical stimulation of sensory cells contained in taste buds⁽ Reference Reed and Xia ^{109 )}. The primary and commonly accepted tastes are sweet, bitter, salty, sour and umami⁽ Reference Ikeda ^{110 )}. In recent years, researchers have proposed another taste corresponding to fat, called pinguis⁽ Reference Mattes ^{101 )} or oleogustus⁽ Reference Running, Craig and Mattes ^{111 )}. Although, some questions still remain (as pointed out by Besnard et al. ⁽ Reference Besnard, Passilly-Degrace and Khan ^{112 )}), accumulating evidence suggests that humans can taste fatty acids and that dietary fat consumption may be partially regulated by an oral detection mechanism⁽ Reference Newman, Haryono and Keast ^{113 )}. The main compounds in dietary lipids are TAG⁽ Reference Lawson, Williamson and Champagne ^{114 )}, but there are reports that the primary stimuli for orosensory fat perception are fatty acids. The first evidence that taste receptors are activated by NEFA was provided in 1997 by Gilbertson et al. ⁽ Reference Gilbertson, Fontenot and Liu ^{115 )}. Moreover, Kawai & Fushiki⁽ Reference Kawai and Fushiki ^{116 )} have demonstrated that NEFA bind directly to receptor CD36, at the same time disproving the idea that TAG could be recognised by CD36. The key protein involved in the conversion of TAG to NEFA in the oral cavity is lingual lipase. In rodents, this enzyme has strong lipolytic activity and plays a primary role in fat detection⁽ Reference Garcia-Bailo, Toguri and Eny ^{98 )}. In vivo assays suggest that in humans the functional activity of LP is very weak (2 μmol/min·per l)⁽ Reference Roberts, Montgomery and Carey ^{117 )}, or even absent⁽ Reference Kulkarni and Mattes ^{118 ,} Reference Voigt, Stein and Galindo ^{119 )}. The secretion of lingual lipase is stimulated by chewing, which may suggest that lingual lipase contributes to oral fat detection in the case of fatty foods, which require a greater oral processing effort⁽ Reference Roberts, Montgomery and Carey ^{117 )}.

Knowledge on the transduction mechanism of fat taste is limited and most information has come from animal studies. Fatty acid perception is mediated by the proposed CD36 receptor, the G-protein-coupled receptors GPR120 and GPR40⁽ Reference DiPatrizio ^{120 )}, and transient receptor potential channel type M5 (TRPM5)⁽ Reference Liu, Shah and Croasdell ^{121 )}. CD36, also known as FAT (fatty acid translocase), is an integral membrane protein with high affinity for long-chain fatty acids⁽ Reference Baillie, Coburn and Abumrad ^{122 )}. It is found on the apical side of the lingual taste-bud cells⁽ Reference Laugerette, Passilly-Degrace and Patris ^{123 )} and plays an important role in dietary lipids perception. In a mouse model it has been proved that in gustatory cells linoleic acid, by binding to CD36, induces the Src-PTK phosphorylation and initiates Ca signalling⁽ Reference El-Yassimi, Hichami and Besnard ^{124 )}. GPR40 and GPR120, members of the GPCR family, are specifically expressed in the mouse gustatory epithelium of the tongue⁽ Reference Cartoni, Yasumatsu and Ohkuri ^{125 )}. Both CD36 and GPR120 exhibit similar binding specificities for long-chain fatty acids; however, GPR120 can only bind to specific types of fatty acid, while CD36 has a higher affinity for ligands and can respond to many types of fatty acid⁽ Reference Reed and Xia ^{109 )}. Mice lacking CD36, GPR120 or GPR40 have diminished preference for fatty acids⁽ Reference Cartoni, Yasumatsu and Ohkuri ^{125 ,} Reference Martin, Passilly-Degrace and Gaillard ^{126 )}. However, GPR40 is not abundant in human gustatory tissues, while GPR120 is present in gustatory and non-gustatory epithelia⁽ Reference Galindo, Voigt and Stein ^{127 )}. For that reason a rodent model may not be appropriate to explore human fat taste transduction. Stimulation of the above-mentioned receptors leads to the generation of a specific signal and initiation of a second-messenger cascade, which in turn results in activation of the afferent nerve fibre, transferring the signal to the brain⁽ Reference El-Yassimi, Hichami and Besnard ^{124 ,} Reference Khan and Besnard ^{128 )}.

Orosensory sensitivity may be modulated by endogenous factors (endocannabinoids) and by hormones, as well as by fat consumption or appetite. Recent studies have shown that paracrine signalling within the taste buds may be regulated by fat sensitivity⁽ Reference Besnard, Passilly-Degrace and Khan ^{112 )}. It has been demonstrated that glucagon-like-peptide-1 (GLP-1) has a significant impact on taste sensitivity in mice⁽ Reference Shin, Martin and Golden ^{129 )}. GLP-1 and its receptor (GLP-1R) are expressed in two populations of mammalian taste cells: a subset of type II cells and a subset of type III cells⁽ Reference Shin, Martin and Golden ^{129 )}. Disruption of the Glp-1r gene may lead to a significant deterioration in fatty acid detection, as has been confirmed in two bottle preference tests in mice and through the licking test⁽ Reference Martin, Passilly-Degrace and Chevrot ^{130 )}. The molecular mechanism behind GLP-1’s modulation of sensitivity to fatty acids is not yet fully understood, although it has been speculated that the intact Glp-1r gene may increase sensitivity to fatty acids by regulating lingual CD36 during eating⁽ Reference Martin, Passilly-Degrace and Gaillard ^{126 )}.

Determination of fat sensitivity and its relation to body mass

There is interindividual variability in fat taste sensitivity, but there are methodological challenges involved in testing fat sensitivity, such as the lack of a commonly accepted test method. The specific testing methods and stimulus vehicle used vary across research groups, which makes comparison very difficult. In this way, several aspects of testing procedures may affect the results and contribute to the variability seen in them⁽ Reference Running, Mattes and Tucker ^{99 )}. Tucker et al. ⁽ Reference Tucker and Mattes ^{131 ,} Reference Tucker, Edlinger and Craig ^{132 )} claim that repeated testing is required to properly assess associations between fat taste and outcomes such as BMI or food intake. Recently, a reliable and reproducible method of assessing oral chemoreception using an emulsion of milk and 18 : 1 has been proposed⁽ Reference Haryono, Sprajcer and Keast ^{133 )}. However, this method requires several unstable milk emulsions and may be too complex for large studies⁽ Reference Running, Mattes and Tucker ^{99 )}.

There is also a biological component to the individual ability to detect fatty acids, but the H ² of fatty acid perception is unclear⁽ Reference Winkler, Justice and Graff ^{7 )}. Pepino et al. ⁽ Reference Pepino, Love-Gregory and Klein ^{134 )} (Table 1) analysed twenty-two obese subjects with different CD36 genotypes (rs1761667) and showed that GG homozygotes had an eight-fold lower oral detection threshold for oleic acid than did AA homozygotes; this was associated with lower gene expression. Although fat intakes and fat preferences were similar among subjects with different genotypes, the small number of individuals involved may explain this result. It could be assumed that the effect size of this single genotype on fat intake is too small to detect in such a group. Moreover, potential differences between lean and obese subjects were not considered in this study. Obese women with the CD36 GG genotype (rs1761667) exhibited an oral detection threshold for oleic acid that was over three times lower than that of individuals with the AA genotype⁽ Reference Mrizak, Šerý and Plesnik ^{135 )}. In one study alone, associations between gene polymorphism, fat sensitivity, fat preference (though not fat intake) and body weight were examined at the same time. Three polymorphisms of the CD36 gene (rs1761667, rs3840546 and rs1527483) were reported to be associated with the outcomes in the study of Keller et al. ⁽ Reference Keller, Liang and Sakimura ^{91 )} (Table 1). Participants were presented with salad dressings with three different fat concentrations and asked to rate perceived oiliness, fat content and creaminess on a visual analogue scale. As nose clips were not used in this experiment (to imitate a real eating experience), the ratings were based on both taste and smell. The test used in this study was thus not a discrimination test. Moreover, salad dressings are not pure stimuli and may not accurately reflect true fat sensitivity. On the other hand, a discrimination test with the use of salad dressing mimics food choices made during natural eating occasions. It was found that alleles of rs1761667, rs1527483 and rs3840546 were associated with perceived creaminess, fat content ratings and body weight, respectively For a few polymorphisms of CD36, no associations were detected with the examined traits, so the results were inconsistent, as might be anticipated for polymorphisms separated by about 2 kb⁽ Reference Keller, Liang and Sakimura ^{91 )}. Interestingly, the Cd36 gene is expressed in the olfactory epithelium of mice, and Cd36-deficient animals display impaired preference for a lipid mixture odour⁽ Reference Xavier, Ludwig and Nagai ^{136 )}. Humans are able to detect slight differences between milk samples with varying grades of fat, and this ability is not affected by BMI or dairy intake⁽ Reference Boesveldt and Lundström ^{137 )}. It could thus be hypothesised that the effect of the CD36 polymorphism on food choices involves sensing fats in the oral cavity and through olfactory perception, which might have played a role in the study of Keller et al. ⁽ Reference Keller, Liang and Sakimura ^{91 )}. In summary, little is known so far about the genetic component of fat sensing. Methodological issues, including the lack of a rapid and valid test method for fatty acid sensitivity, may be the reason why no GWAS on this trait has been performed.