Association study of taste preference: Analysis in the Lithuanian population

Abstract Taste has strong evolutionary basis in the sense of survival by influencing our behavior to obtain food/medicine or avoid poisoning. It is a complex trait and varies among individuals and distinct populations. We aimed to investigate the association between known genetic factors (673 SNPs) and taste preference in the Lithuanian population, as well as to determine a reasonable method for qualitative evaluation of a specific taste phenotype for further genetic analysis. Study group included individuals representing six ethnolinguistic regions of Lithuania. Case and control groups for each taste were determined according to the answers selected to the taste‐specific and frequency of specific food consumption questions. Sample sizes (case/control) for each taste are as follows: sweetness (55/179), bitterness (82/208), sourness (32/259), saltiness (42/249), and umami (96/190). Genotypes were extracted from the Illumina HumanOmniExpress‐12v1.1 arrays’ genotyping data. Analysis was performed using PLINK v1.9. We found associations between the main known genetic factors and four taste preferences in the Lithuanian population: sweetness—genes TAS1R3, TAS1R2, and GNAT3 (three SNPs); bitterness—genes CA6 and TAS2R38 (six SNPs); sourness—genes PKD2L1, ACCN2, PKD1L3, and ACCN1 (48 SNPs); and saltiness—genes SCNN1B and TRPV1 (five SNPs). We found our questionnaire as a beneficial aid for qualitative evaluation of taste preference. This was the first initiative to analyze genetic factors related to taste preference in the Lithuanian population. Besides, this study reproduces, supports, and complements results of previous limited taste genetic studies or ones that lack comprehensive results concerning distinct (ethnic) human populations.


| INTRODUC TI ON
Taste perception is part of flavor perception, which results primarily from the combination of three discrete senses: taste, somatosensation (touch, pain, and temperature), and olfaction. (Simon et al., 2006) Taste perception is strongly evolutionary in terms of physiological behavior such as obtaining a balance of electrolytes (saltiness), acquiring energy (sweetness), synthesizing proteins (umami), and avoiding poisonous (bitterness) or rotten (sourness) substances. (Purves et al., 2001).
Taste preference has a genetic background, and evidence suggests that the perception of different tastes is a polygenic or complex trait, though some taste phenotypes (traits) were thought to be inherited as Mendelian traits. (Guo & Reed, 2001) Taste-related traits show different levels of heritability. Sweetness-related traits such as the pleasantness, frequency of consumption, and craving for sweet foods show significant heritability (40%, 50%, and 31%, respectively). (Keskitalo et al., 2007) Heritability modeling on bitter stimuli showed a common genetic factor for quinine, caffeine, and sucrose octaacetate (22%-28%) and separate and specific genetic factors for propylthiouracil (72%) and quinine (15%). (Hansen et al., 2006) Heritability may include different types of genetic variation, unidentified genetic factors, environmental factors, and interaction between those factors, which are yet to be discovered. The first demonstration of how genetic variants shape interindividual differences in human taste sense was for the bitter taste receptor TAS2R38. (Kim et al., 2003) Individuals are referred to as tasters if being PAV (Pro at 49, Ala at 262, and Val at 296) haplotype and nontaster if being AVI (Ala at position 49, Val at 262, and Ile at 296) haplotype of the receptor. (Newcomb et al., 2012) Thus, the different haplotypes within the gene contribute to the intermediate phenotypes and thereby explain the nature of the quantitative trait. (Kim et al., 2003;Mennella et al., 2011) Genetic mapping and candidate gene association studies show that taste phenotypes are influenced by allelic variation of genes involved in both peripheral and central taste processing. (Bachmanov et al., 2013) Many genome-wide association studies (GWASs) have been performed in the field of taste (Diószegi et al., 2019), but some of them lack the support of replication studies and the majority of them were performed in heterogeneous populations. This has led us to the knowledge of what is common for human populations from the genetic point of view. Nevertheless, when performing those studies, do we not miss what is specific and unique for different populations?
There are a growing number of publications emphasizing the need to analyze various ethnic populations to better understand the architecture of genetic traits.
Behaviors mediated by the taste preference, particularly food choice and intake, are among the most important ones in terms of the health problems found in developed and developing societies. Taste, chemesthetic sensation, and responsible receptors' genetics are relevant to study not only for the role in food preference, choice, and intake but also for the importance in human physiology and evolution.
Function of these genes directly contributes overall health, indirectly health policy, food industry, and even global climate inevitably linked to the food industry. (Nolden & Feeney, 2020) Moreover, taste receptors and molecules are not exclusive to the oral cavity, but are found throughout the body and have multiple functions in tissues and organs, and have a role in a wide range of diseases. (ZhuGe et al., 2020) Having in mind genetic factors that contribute to the biology of taste preference, we should not forget the environmental component, which is significant in taste preference too. (Mennella & Beauchamp, 2005; That is, a full understanding of taste is important for prevention of diseases such as hypertension, obesity, heart disease, diabetes, and some cancers. All of these diseases are strongly influenced by food choices that in turn are determined, in part, by how much we like (or dislike) the flavor of the food. (Beauchamp & Mennella, 2011) Genetic variation in taste preference influences human nutrition and health and could be used as a biomarker of predisposition to some diseases. (Bachmanov et al., 2013) As mentioned above, food choice arising from food preferences could be consciously controlled, in contrast to genetic factors. Thus, the accurate evaluation of food preference is important. It could be achieved by reasonable questionnaires. In these days of integrative and personalized medicine, it would be of great value for a clinician to have a simple tool that integrates as much as possible genotype and phenotype data.
The importance of taste research in ethnic populations and the demand to translate the results into clinical practice lead us to the aim of defining the genetic factors associated with the different taste preference in the Lithuanian population and of further evaluating the potential of using the questionnaire as a qualitative tool for food preference evaluation.

| Study group
Study group included unrelated individuals representing six ethnolinguistic regions of Lithuania (West, North, and South Žemaitija and West, East, and South Aukštaitija). All self-reported healthy individuals indicated at least three generations of Lithuanian ethnicity and residency in the same ethno-linguistic region. Study participants were asked to fill in the dedicated questionnaire (see in a section "Questionnaire" below) and donate blood for DNA extraction and genotyping procedure. This is case-control genetic association study, so cases and controls were assigned as follows. Case and control groups for sweet, bitter, sour, umami, and salty food preference were determined according to the selected answers to the tastespecific questions (for more information, see the Questionnaire section below). Sample sizes (case/control) for each taste modality and preference are as follows: sweetness (55/179), bitterness (82/208), sourness (32/259), saltiness (42/249), and umami (96/190). Only few individuals overlapped between case groups; for example, the same individual assigned to the sweet taste preference case group was also assigned to the salty taste preference case group. Besides, individuals qualified as cases for particular taste preference were assigned as controls for the other; for example, the case for a sweet taste preference was also a control for a bitter, sour, umami, and/ or sweet taste preferences.

| Questionnaire
Twenty-nine questions regarding certain tastes (sweetness, bitterness, sourness, saltiness, and umami) were asked in order to evaluate the food preference of every participant in the study. Food products were assigned to a certain taste group according to the literature that was reviewed. (Feeney et al., 2011;Garcia-Bailo et al., 2009) There were six questions for sweetness, seven for bitterness, five for sourness, four for saltiness, and seven for umami (for the list of questions and possible choices, see File S1). There were three types of multiple-choice questions: Type 1: the groups of food products representing a definite taste (sweetness, bitterness, sourness, saltiness, and umami); Type 2: periodicity of consumption of a product representing one of the five tastes; and Type 3: Yes/No questions reflecting the consumption of extra products (such as sugar, salt, or pepper) to enhance a certain taste. Individuals were grouped according to their preference for sweet, sour, salty, bitter, or umami flavors.
If the sweet product group was chosen, that is, carrots, potatoes, or beetroot, while answering Type 1 questions, the individual was put into the sweet taste case group. If while answering Type 2 questions an individual admitted using certain products, that is, sweet ones, 3-5 times per week or more, that person was added to the sweet taste case group. If while answering Type 3 questions a subject indicated that he used certain taste-enhancing products, that is, sugar, that person would be placed into the sweet taste case group.
The other four taste case groups were determined in the same manner. Individuals were listed as controls if answering Type 1 questions with other food preference than the tested one, Type 2 questions with the less frequent consumption of the food tested for food preference, and (or) Type 3 questions with a contrary answer.

| Genome-wide genotyping
Genomic DNA was extracted from venous blood using either the phenol-chloroform extraction method or automated nucleic acid purification using paramagnetic particles (Freedom EVO ® Nucleic Acid Purification Workstation). The quality and quantity of purified genomic DNA were evaluated with a spectrophotometer (NanoDrop ® ND-1000 Spectrophotometer). Genome-wide genotyping following manufacturer's protocols was performed using high-density Illumina HumanOmniExpress-12v1.1 arrays (719,666 SNPs) on an Illumina HiScan ™ SQ system.
Association with five different phenotypes (taste preference for sweetness, bitterness, sourness, saltiness, and umami) was tested using the chi-square test of independence or Fisher's exact test depending on the minimal number of genotypes (the minimal number of genotypes was five for the chi-square test of independence and 0 for the Fisher's exact test). Five different genetic models (basic: genotypic, allelic and additive: Cochran-Armitage trend, dominant, and recessive) were used to evaluate association between phenotypes and known genetic factors. Genetic variants or SNPs (hereafter variants) of known candidate taste preference genes analyzed in this study are provided in Table 1. Covariates such as age, sex, body mass index (BMI), family history, or other environmental factors were not included in the analysis. A significance level (α) of 0.05 was set for this study. Permutation procedure (n = 10,000) was used to obtain empirical p-value for the chi-square test of independence or Fisher's exact test.

| RE SULTS
Analyzed variants were frequent in different populations and with different functions: synonymous or nonsynonymous (missense) in different genomic regions (introns or exons). Rare variants and variants from several candidate genes were not analyzed, since the genome-wide genotyping array did not include them.
HWE testing for all variants was conducted in three groups, that is, only cases, only controls, and combined case and control group. Because the genome-wide genotyping call rate was 0.97 and higher for all samples, variants with HWE p < .001 in any group were removed from further analysis. As a result, 46 variants were excluded from the analysis (sweetness-0 variants; bitterness-8 variants; sourness-30 variants; umami-4 variants; and saltiness-4 variants).
Statistically significant (p < .05) associations between SNPs and different taste preferences were shown: 3 for sweetness; 6 for bitterness; 48 for sourness; 5 for saltiness; and 1 nearly statistically significant association for umami.

| Variants associated with sweet taste preference
Three variants were significantly associated with sweet taste preference ( Table 2,

| Variants associated with bitter taste preference
Six variants were shown to be significantly associated with bitter taste preference (Table 3,  (Val262Ala)). Four more variants showed an association with bitter taste under basic allelic or genotypic models but were eliminated due to discrepancy from HWE (p < .001).

| Variants associated with sour taste preference
Analysis showed 41 variants of the ACCN1 gene statistically significantly associated with sour taste preference under basic genotypic and/or allelic models and additive recessive and/or dominant models (Table S2).
Seven significantly associated variants in other genes are provided in

| D ISCUSS I ON
Bitterness is another well-studied taste. This study confirmed the statistically significant associations of rs2274327, rs2274328, rs1832262, and rs3765964 in the CA6 gene, two coding and two intronic, respectively, and coding rs10246939 and rs1726866 in the TAS2R38 gene. The product of the CA6 gene is known as the gustin, or carbonic anhydrase 6 (CA VI), isozyme of the carbonic anhydrase secreted in saliva and milk. (Pastorekova et al., 2004) CA VI was found to be associated with bitter taste, and CA6 SNP rs2274327 has been linked with picky eating behavior in preschool-age children TA B L E 2 Statistically significant results of the analysis of the association between sweet taste preference and SNPs  (Cole et al., 2017) and implicated in taste bud function and salivary buffer capacity (Peres et al., 2010). It was postulated that CA VI may be a mechanistic link between 6-n-propylthiouracil tasting and fungiform taste papilla density and maintenance, (Melis et al.,) but a later study did not detect such an association (Feeney & Hayes, 2014 could be co-expressed with GNAT3 in some tissues, making GNAT3dependent signal transduction possible. (Imai et al., 2020) Still, this study did not find the association between GNAT3 variants (results not shown) and bitter taste preference. Variants of RGS21 (regulator of G-proteins), (Cohen et al., 2012) and TAS2R16 and TAS2R19 (TAS2 family receptors) genes were not associated with bitter taste preference either. These results imply the need for further research on the role of other G-proteins and their regulators in bitter taste pathway.
Sour taste stimuli are thought to be mainly acids (H + ), and the mechanism of signal transduction is different than it is with sweet, bitter, or umami tastes. Sour and salty taste stimuli (Na + or K + ) are transported into the taste cells through ion channels rather than receptors as sweet, bitter, and umami stimuli are. (Roper, 2007) Instead of transporting molecules, ion channels translate chemical signals into electrophysiological signals. The polycystic kidney disease 1 and polycystic kidney disease 2-like proteins PKD2L1 and PKD1L3 have been identified as sour taste-related receptors in human taste cells (Ishimaru et al., 2006), and potential ion-channel OTOP1 was present in taste cells in mouse that express Pkd2l1. (Tu et al., 2018) In our study, statistically significant association between sour taste preference and four variants (noncoding rs12360462 PKD2L1 gene; rs9925415, rs9928317, and rs4788592 PKD1L3 genes, only rs9928317 coding) was observed. This association of variants for both genes supports the involvement of these proteins in sour taste signal as heteromeric/homomeric complexes or separate parts.

Statistically significant association of another 44 variants in ACCN2
(rs835592, rs2272391, and rs7305558 with the highest OR = 14.5 for rs7305558) and ACCN1 (full list in Table S2) genes was detected.
This supports the assumption that acid taste pathway can start by several different channels in the taste cells. (Huque et al., 2009) It is likely that not independent variants but several haplotypes of the ACCN1 gene are responsible for variation in sour taste preference.
It is known that a heterodimer of TAS1R1 and TAS1R3 (TAS1R1/ TAS1R3) functions as an umami taste receptor in humans. (Li TA B L E 5 Statistically significant results of the analysis of the association between salty taste preference and genetic variants there was no association found with GNAT3. This could be because GNAT3 is involved in more than one different taste signal transduction pathway and is less specific. Animal models showed that the sodium-specific and amiloridesensitive epithelial sodium channel (ENaC) and the transient receptor potential cation subfamily V member 1 (TRPV1) amiloride insensitive channel are candidates for the pathway of salty taste. (Bigiani, 2020) The results of this study support this evidence as association between variants (noncoding rs12162045 and rs152733 in SCNN1B gene, OR = 0.455 and OR = 1.835, respectively; synonymous rs877610, noncoding rs8078936 and rs150908 in TRPV1 gene) and saltiness preference was observed. There is a lack of evidence for SCNN1A and SCNN1G genes being associated with saltiness and the involvement of these proteins in salty taste pathway in humans.
Our study did not identify such an association either.
The results of this study reveal only a fragment of the full spectra network elements in complex signal transduction pathways for different tastes. The food preference too is a very complex trait and depends not only on biological factors (such as age, sex, genetics), but also on culture, socio-economic status. (Davide et al., 2017;Mennella & Beauchamp, 2005).  Collaboration, 2015). Besides, the results can indicate that the questionnaire used in the study proved its value and might be a useful tool for a clinician for food preference evaluation, but further validity assessment is needed.

| CON CLUS ION
The results of the study reproduced associations of the main known genetic factors and four of the five tastes: sweetness-the genes TAS1R3, TAS1R2, and GNAT3 (three variants); bitterness-the genes CA6 and TAS2R38 (six variants); sourness-the genes PKD2L1, ACCN2, PKD1L3, and ACCN1 (48 variants); and saltiness-the genes SCNN1B and TRPV1 (five variants). Most of the associations show genetic factors that are the primary taste signal transduction pathway players in the taste bud cells (G-protein-binding receptors or ion channels), since they are very specific to particular tastes. Genetic factors encoding proteins that are further downstream of the pathway usually are not that specific and that could be one of the reasons why this particular study design failed to find the associations.
Other reasons might include the difference in genetic structure of the population, the sample size of the study, nongenetic factors that contribute to the food preference, and structure and content of the questionnaire. The lack of specific questions provides no chance to detect any significant association, as occurred in this study while analyzing the umami taste case. This study also supports some results of a few studies and complements ones that lack comprehensive results on distinct (ethnic) human populations. Finally, we found our questionnaire (based on very specific questions about nutritional habits) a beneficial aid for qualitative evaluation of taste preference.
To reliably classify individuals for food preference, there must be a sufficient number of questions including all food groups and specifying certain tastes.

TA B L E 6
Nearly statistically significant results of the analysis of the association between umami taste preference and genetic variant Note: a-chromosome; b-allele 1; c-allele 2; d-distribution of alleles or genotypes in the case group; e-distribution of alleles or genotypes in the control group; f-empirical p-value for chi-square test of independence (permutation test based on the most significant result of allelic dominant and recessive models); g-empirical p-value for Fisher's exact test (permutation test based on the most significant result of allelic dominant and recessive models); h-basic model: genotypic; i-additive model: Cochran-Armitage trend; j-basic model: allelic; k-additive model: dominant; l-additive model: recessive; m-no data available.

ACK N OWLED G M ENTS
The authors would like to thank all Lithuanian people who donated blood samples and participated in the LITGEN project. The authors are grateful to all the specialists who were involved in the recruitment of individuals, collection of blood samples, informed consent forms and questionnaires, and genomic DNA extraction. This work was in part supported by the European Social Fund under the Global Grant measure (VP1-3.1-ŠMM-07-K-01-013).

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

S TU D I E S I N VO LV I N G H U M A N S U BJ EC TS
The study conforms to the Declaration of Helsinki, US, and/or European Medicines Agency Guidelines for human subjects. Study's protocols and procedures were ethically reviewed and approved by a recognized ethical body (the Vilnius Regional Biomedical Research Ethics Committee (Permission No. 158200-05-329-79), Lithuania).

Informed written consent was obtained and documented from each
individual included in the study.

DATA AVA I L A B I L I T Y S TAT E M E N T
All necessary data is provided in the article. The raw data that support the findings of this study is available from the corresponding author upon reasonable request.