The Pennsylvania State University

The Graduate School

Department of Food Science

BITTER RECEPTOR POLYMORPHISMS INFLUENCE BITTERNESS OF NON-

NUTRITIVE SWEETENERS AND ALCOHOLIC BEVERAGE LIKING

A Thesis in

Food Science

by

Alissa Allen

© 2013 Alissa Allen

Submitted in Partial Fulfillment of the Requirements for the Degree of

Master of Science

May 2013

The thesis of Alissa Allen was reviewed and approved* by the following:

John E. Hayes Assistant Professor of Food Science Thesis Advisor

Kathleen Keller Assistant Professor of Health and Nutritional Sciences and Food Science

Joshua Lambert Assistant Professor of Food Science

Robert Roberts Professor of Food Science Head of the Department of Food Science

*Signatures are on file in the Graduate School

iii

ABSTRACT

Bitterness is largely aversive and commonly associated with lower liking and intake. However, the ability to perceive bitterness differs across individuals due to genetic variation within bitter receptor genes (TAS2Rs). The goal of the present thesis is to investigate effects of bitter receptor polymorphisms on liking and perception, of reported bitterness and sweetness of non-nutritive and the remembered liking of different alcoholic beverages. The work presented here compares genotypes for putatively functional polymorphisms in bitter receptor genes in attempt to explain individual differences. Major experimental findings include: Study 1- Polymorphisms (SNPs) in bitter receptors TAS2R9 and TAS2R31 explains 13.4% of the variation in the perceived bitterness from Acesulfame K (AceK). Study 2 - The non-nutritive sweetener rebaudioside A (RebA) elicited greater perceived bitterness than rebaudioside D (RebD), with no difference in sweetness intensity. SNPs that were previously reported in Study 1 to explain AceK bitterness were not associated with perceived bitterness of RebA or

RebD. Study 3 - Liking ratings for non-sweet alcoholic beverages were associated with

TAS2R38 diplotype. However, the diplotype failed to predict liking for sweet alcoholic beverages. This latter finding was as anticipated, as we would not expect TAS2R38 to meditate liking or intake for foods or beverages where mixture suppression removes bitterness. Together these findings expand the understanding of individual differences in taste perception with the respect to polymorphisms within taste receptor genes. With this knowledge, researchers and product developers may be better positioned to explore consequences of genetic variations, to further expand the relationship associated with intensity, liking and intake, and to develop tailored products.

iv

TABLE OF CONTENTS

LIST OF FIGURES ...... v

ACKNOWLEDGEMENTS ...... vii

Chapter 1 Literature Review ...... 1

Introduction ...... 1 Importance of tase perception ...... 4 Individual differences in perception ...... 15 Taste Receptors ...... 21 Bitterness in beverages ...... 26 Sensations from alcoholic beverages ...... 38 Conclusions ...... 41 Aims and Hypothesis ...... 45

Chapter 2 Bitterness of the non-nutritive sweetener Acesulfame Potassium vaires with polymorphisms in TAS2R9 and TAS2R3 ...... 46

Introduction ...... 47 Materials and Methods ...... 49 Results ...... 55 Discussion ...... 63

Chapter 3 Rebaudioside A and Rebaudioside D bitterness do not covary with acesulfame-K or polymorphisms in TAS2R9 and TAS2R31 ...... 70

Introduction ...... 71 Materials and Methods ...... 74 Results ...... 78 Discussion ...... 86

Chapter 4 Bitter receptor alleles differentially influence alcoholic beverage liking ...... 89

Introduction ...... 90 Materials and Methods ...... 92 Results ...... 95 Discussion ...... 100

Chapter 5 Conclusions and further steps ...... 105

References ...... 108

Appendix A: Chapter 2 Supplemental figure ...... 120

v LIST OF FIGURES

Figure 1-1: Green’s Labeled Magnitude Scale, with ‘strongest imaginable’ located at 100. For the gLMS, the top anchor is changed to ‘strongest imaginable sensation of any kind’. The gVAS only includes the ‘no sensation’ and ‘strongest imaginable sensation of any kind’. Values are not including in the participant’s scale. Figure taken from [(Green 1993)] ...... 14

Figure 1-2: Amino acid sequence and orientation of the bitter taste receptor TAS2R38. Figure is from [(Wiener 2011)] ...... 23

Figure 1-3: Sweetness and bitterness response curves for sucrose, acesulfame-K, aspartame and rebaudioside A. These figures are taken from [(DuBois, Walters et al. 1991)] ...... 32

Figure 2-1: Effect of the TAS2R9 Val187Ala polymorphism on the bitterness and sweetness of AceK and the bitterness of PROP. The bitterness of AceK was significantly different across genotype; no effect was observed for AceK bitterness or PROP bitterness (p-values provided in text). Adjectives refer to semantic labels on a general Labeled Magnitude Scale (gLMS; see text). BD refers to ‘barely detectable’...... 56

Figure 2-2: LD Plot for TAS2R SNPs on chromosomes 7 (top) and 12 (bottom). Numbers indicate rounded R-squared values and shading indicates exact R-squared values generated via Haploview ...... 57

Figure 2-3: Effect of the TAS2R31 Val240Ile allele on the bitterness and sweetness of AceK and the bitterness of PROP. The bitterness of AceK was significantly different across genotype; no effect was observed for AceK bitterness or PROP bitterness (p- values provided in text). Adjectives refer to semantic labels on a general Labeled Magnitude Scale (gLMS; see text). BD refers to ‘barely detectable’ ...... 59

Figure 2-4: Effect of AV/PA TAS2R38 diplotype. PROP bitterness differed by diplotype; differences in AceK bitterness or sweetness across diplotype were not significant (p- values provided in text)...... 61

Figure 3-1: Mean (±Std Error) gLMS ratings for bitterness and sweetness of AceK (collected previously: indicated by the grey box), RebA, RebD, aspartame, sucrose and gentiobiose are reported here. The sweetness ratings for AceK, RebA and RebD were not statistically different while bitterness was significantly different across the four non-nutritive sweeteners (see text). Adjectives refer to semantic labels on a general Labeled Magnitude Scale (gLMS; see text). BD refers to ‘barely detectable’. .. 80

vi

Figure 3-2 Scatter plots showing (top) sweetness and (bottom) bitterness for RebA and RebD, with histograms along the axes. Mean sweetness was not significantly different between RebA and RebD, although both show substantial individual differences. Mean bitterness differed, with RebD showing significantly less bitterness. Again, substantial individual differences were observed between these sweeteners...... 81

Figure 3-3: Correlations of bitterness ratings between RebA and RebD, AceK and aspartame...... 82

Figure 3-4: Effect of the TAS2R9 Val187Ala polymorphism on the bitterness of AceK, RebA and RebD. As expected, the bitterness of AceK (collected on Day 1) was significantly different across genotype for these individuals; conversely, no effect of genotype was observed for RebA or RebD (see text)...... 84

Figure 3-5: Same as Figure 2-4, except for the Val240Ile polymorphism in TAS2R31. AceK bitterness ratings were significantly different across genotype, as expected, while there was no evidence of an effect for RebA and RebD (see text)...... 85

Figure 4-1: Mean liking ratings ± standard errors for 20 alcoholic beverages ...... 96

Figure 4-2: Mean liking ratings ± SEM for AV homozygotes and PA carriers (PA/PA and PA/AV individuals) for sweet and non-sweet alcoholic beverages. Ratings were measured as part of a liking survey that also included food and non-food items to generalize the context beyond foods and beverages (see methods). The sweet group included: flavored malt beverages, sweet white wine, fruity red wine, off-dry or semi-sweet white wine, malt liquor, margaritas or daiquiris, spirits with juice or milk, spirits with energy drinks, and spirits with soda. The non-sweet group included: lager beer, pale ales, bitter beer, dry wine, scotch, vodka and martinis with gin or vodka. See text for results of repeated measures ANOVA; p-values are unadjusted t-tests...... 97

Figure 4-3: Mean liking ratings ± SEM for AV homozygotes and PA carriers for sweet beverages (n=9). See text for results of repeated measures ANOVA; p-values are unadjusted t-tests...... 99

Figure 4-3: Mean liking ratings ± SEM for AV homozygotes and PA carriers for non- sweet beverages (n=7). See text for results of repeated measures ANOVA; p-values are unadjusted t-tests...... 100

Figure A-1: Effect of the Ala227Val TAS2R31 polymorphism on the bitterness and sweetness of AceK and bitterness of PROP. The bitterness of AceK was significantly different across genotype; no effect was observed for AceK bitterness or PROP bitterness (p-values provided in text). Adjectives refer to semantic labels on the general Labeled Magnitude Scale (gLMS; see text for details). BD refers to ‘barely detectable’...... 120

vii ACKNOWLEDGEMENTS

I would like to thank John Hayes for his support and advice in the development of methods, data analysis and writing skills. Thank you to my committee for your support and contribution to my thesis.

To Samantha: I could have never gotten this far without your help. Your enthusiasm for science and work ethic was inspiring, and your friendship played a huge role in my success here at Penn

State.

To all of the Sensory Girls: I am so thankful to have all of you a part of my experience at Penn

State. I have enjoyed our wine and cheese nights, delicious food, Mondays at Mario’s, and never ending jokes. Thanks for all of your help.

To Mom, Dad and Erica: Thank you for believing in me. I know that I have missed a lot being away, but I have treasured all of the times you have visited. Football games, Mom’s flip-cup skills, making great meals, and going on walks with Mia. Thank you for your support and encouraging for me to try my hardest and to not give up on my dreams.

To Michael: Your encouragement and support has helped me to believe in myself. Thank you for talking with me all of the countless times I called you on the way too and from work. Thank you for always helping to keep my chin up and making me smile. It’s always better when we’re together.

1

Chapter 1

Literature Review

Introduction

The food industry strives to provide consumers with products that taste good. Although many factors influence consumer purchases, surveys indicate taste is the largest contributing factor, ahead of both cost and convenience (Glanz, Basil et al. 1998; IFIC 2011). Taste, in this usage, really refers to flavor, which is the integrated perception arising from taste, touch and smell input. In technical usage, taste flavor, refers to the five prototypical qualities sweet, bitter, sour, salty and umami. Taste perception may differ greatly across individuals. There are many factors that may affect taste perception, including: ethnicity, personality and age (see (Rozin &

Vollmecke 1986) for a review). This review will focus on the perception of sweet and bitter as sweetness is innately liked while bitterness often (but not always) leads to rejection of foods

(Glendinning 1994; Steiner, Glaser et al. 2001). Bitterness has been shown repeatedly to be negatively associated with liking and intake for many foods and beverages (e.g. vegetables and alcohol) (see (Drewnowski & Gomez-Carneros 2000; Duffy & Bartoshuk 2000; Tepper 1998) for reviews). This chapter will focus on how genetic differences within taste receptor genes impact taste perception. Specifically, I will be focusing on the variability in bitter taste perception and how these differences are associated with genetic variation within bitter taste receptor genes.

Taste signals are sent to the brain when compounds interact with specialized receptors expressed in taste receptor cells. These receptors can have different configurations that interact with compounds that give rise to sweet, sour, bitter, salty and umami sensations. Of all of the taste receptors, bitter taste receptors are the largest family of receptors . There are 25 bitter taste

2 receptor genes (TAS2Rs) that code the receptors (T2Rs) responsible for the ability to perceive bitterness from a wide range of compounds (Go, Satta et al. 2005; Meyerhof 2005; Meyerhof,

Batram et al. 2010). It is believed that humans have developed these diverse bitter receptors to identify a wide range of bitter toxins found in plants (Mueller, Hoon et al. 2005). The compounds that elicit bitterness are chemically diverse, differing in structure, hydrophobicity, size and charge

(see (Meyerhof 2005) ). Bitter compounds can be found in a wide range of foods, from fruits and vegetables to alcoholic beverages (Barratt-Fornell & Drewnowski 2002; Drewnowski & Gomez-

Carneros 2000), and may arise naturally during growth or can be a result of chemical reactions that take place during processing (e.g., maillard reaction, cheese production, cocoa and coffee roasting) (Meyerhof 2005; Rousseff 1990). Additionally, bitter compounds may be produced by chemical reactions that occur during storage and decay. While some bitterness is expected in some foods (e.g., scotch, cheese, cruciferous vegetables), excessive bitterness often results in rejection of the food (Rousseff 1990).

Sweet, umami, and bitter tastes are perceived by the interaction of a ligand with a G protein-coupled receptor. This causes the release of calcium from internal stores and eventually though a series of reactions a signal is sent to the brain causing the perception of bitter or sweet

(reviewed by (Margolskee 2002)).

Some bitter taste receptors are highly selective, only responding to a few specific compounds, while others are activated by a wide range of compounds (Adler, Hoon et al. 2000;

Matsunami, Montmayeur et al. 2000) (see (Meyerhof, Batram et al. 2010) for a review). There has been some success in identifying the functional group(s) that activate these receptors, although some receptors have not yet to been deorphanized (e.g. hT2R45).

Variation within bitter taste receptors genes can affect the ability to perceive bitterness from individual compounds as well as whole foods. This variation within the receptor can arise due to genetic mutations located in or near the gene (Dotson, Zhang et al. 2008; Kuhn, Bufe et al.

3

2004; Pronin, Xu et al. 2007; Roudnitzky, Bufe et al. 2011). These mutations can alter the receptor function by altering the activation site, receptor conformation or altering the sequence in the promoter region (Kim, Wooding et al. 2005). Kim and colleagues were the first to show that taste receptor function is associated with changes in DNA sequence (Kim, Jorgenson et al. 2003;

Kim, Wooding et al. 2005). Numerous other studies have reported associations between variations (polymorphisms) in bitter taste receptor genes and perception (Hayes, Wallace et al.

2011; Mennella, Pepino et al. 2005; Roudnitzky, Bufe et al. 2011; Sandell & Breslin 2006). In some cases, variations in bitter taste receptors have been used to explain difference in liking or intake of bitter foods and beverages (Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004;

Duffy, Hayes et al. 2010; Hayes, Wallace et al. 2011; Hinrichs, Wang et al. 2006; Suomela,

Sandell et al. 2012; Wang, Hinrichs et al. 2007). However, not all reports support these findings

(Gorovic, Afzal et al. 2011; Ooi, Lee et al. 2010). Nevertheless, there are additional receptors and polymorphism yet to be explored and how they impact human sensory perception and intake of bitter foods and beverages is unknown.

To determine what compounds activate individual taste receptors, in vitro methods are utilized. In vivo, advancements within genotyping have allowed for faster, and cheaper methods for genotyping individuals that can be used to associate genetic variation with human psychophysical responses. The next logical step is to explore bitter taste receptor variants and their role in liking and intake. Many studies report differences in perception, but ultimately the industry would benefit by knowing if these mutations and differences in taste perception effect health risk, consumption and presumably purchase. First, this review discusses the relevant psychophysical methods, including taste thresholds and suprathreshold scaling. Next, it will cover the perception of sweet and bitter. Then I will assess different methods used to measure sensitivity, and discuss taste receptor genetics. Subsequent thesis chapters will explore whether

4 genetic variation can explain a) bitterness perception of non-nutritive sweeteners and, b) differences in liking of different types of alcoholic beverages.

Importance of taste perception

When consumers are faced with a decision of what food products to purchase, many factors are taken into consideration, including nutrition, cost, convenience, taste and weight control. However, surveys indicate ‘taste’ (i.e. flavor) is the most important when consumers were asked to rate the importance of factor that influence purchase decisions (Glanz, Basil et al.

1998). Flavor can be defined as the collection of sensations perceived through three sensory systems, including gustation, somatosensation and olfaction. Gustation refers to sensations perceived through tastes nerves in the mouth. The prototypical tastes include: sweet, sour, salty, bitter and umami, and possibly others. Somatosensation includes the perception of texture, cooling and burn. Lastly, olfaction occurs when a volatile chemical activates receptors in the nose, regardless of whether it is perceived in the nose or mouth. The combinations of sensations from these sensory systems are integrated to form integrated flavor perceptions (Lawless 1996;

Small & Prescott 2005).

Frequently, taste profiling and sensory evaluation are executed for a particular product to inform the product developers or sensory scientist how a consumer perceives a product, and if alterations should be made to the product. Methods have been developed to answer questions about how a product or individual ingredients taste, and these methods can be used to measure taste perception or liking. However, some methods used do not account for individual differences in taste perception, as group means are often reported, obscuring individual differences.

Individuals experience and perceive different sensations when eating a food. Understanding these

5 differences when developing or reformulating a food product may be important for success in the market.

Sensory scientists often want to know how products taste, which can be done in a number of ways. Numerous sensory methods have been developed to help sensory scientists answer these questions; however, here I will be narrowing my focus to discuss the following questions: a) which product is liked or preferred over another, and b) how the products are perceived, meaning how intense the qualities are on an individual basis. Perceived intensity is often measured by providing one or more scales for the qualities of interest. These qualities can include, but not limited to, the five prototypical tastes: sweet, salty, sour, bitter and umami (see (Delwiche 1996)).

In order to measure preference and liking, other methods are used. It is important to note the difference between preference and liking (for a short review see, (Rozin & Vollmecke 1986)).

Preference is measured by having participants’ select one product out of two or more products.

Preference is often linked with liking; however, this is not always the case (Rozin & Vollmecke

1986). Liking or acceptance of a product can be determined by having individuals make ratings on a scale for one or more products.

Knowing if a product is preferred or liked over another may indicate how well a product performs relative to a competitor. However, liking or preference (i.e. hedonic) data does not inform the researcher on the reason why a product is liked over the other. More importantly, hedonic tests do not provide any insight on what the participants are experiencing when they taste and consume the product. In this respect, collecting intensity ratings, rather than liking, can provide valuable information about how the product is being perceived. It is assumed that liking is driven, at least in part, by the perceived intensity of various chemosensory qualities. However, because hedonic tests normally compare group means, it can be difficult to detect individual differences, and answer questions of why they exist, without including another variables. By measuring intensity, additional information can be collected regarding an individuals’ experience

6 from a product. This can provide additional insight for product developers to improve a product, or allow them to create multiple ideal products for different target populations.

One common way to measure the perceived intensity elicited by a product is to use a small group of trained panelists. Panelists use attributes based on pre-defined terms from the panel leader, or terms generated by the panelists themselves, to profile the product. This approach requires extensive training that is often is slow, labor intensive and costly. Ensuring the panel uses the descriptions consistently depends on constant training, similar to calibrating a machine.

However, as mentioned above, individuals differ in perception due to several different reasons.

This can complicate training, as panelists may perceive a product differently, making it difficult to come to a consensus on intensity. More importantly, taking an average of ratings of a group may conceal segmented responses. The nature of trained panels, which use standardizing ratings, reduces the ability to assess individual differences across people.

To help answer questions of individual differences, a second approach will be discussed here. This method uses untrained participants, who rate their perceived intensity for one or more qualities on a line scale without calibration. This method is often much faster than a trained panel, although some brief orientation may be required to familiarize individuals with the scale.

Collecting individuals ratings for products provide additional insight into individual perception and can be analyzed to determine if there are differences in intensity ratings. While untrained panels do not require training, a larger sample size is typically needed to make comparisons across groups (e.g. gender, age, genotype). A rating is often kept as a continuous variable (e.g. sweetness intensity) and is may be regarded as a taste phenotype for that quality (sweet phenotype).

7

Liking and intake

It is often assumed that taste perception is a driver of liking (Cardello 1995), and therefore an indirect driver of intake. More importantly, there have been many examples of foods that are perceived differently across individuals, specifically those that elicit bitterness, such as vegetables (Duffy, Hayes et al. 2010) and alcoholic beverages (Hayes, Wallace et al. 2011). (See

(Feeney 2011) for a comprehensive review.) For example, the perception of bitterness and sweetness for cruciferous vegetables (kale, brussel sprouts) is correlated with the intake of vegetables, and was mediated through reported liking of vegetables. For vegetables, individuals who experience greater perceived bitterness also report lower ratings of perceived sweetness

(presumably due to mixture suppression). Together, greater bitterness and less sweetness are associated with lower liking and intake for cruciferous vegetables, compared to individuals who report greater sweetness and less bitterness (Dinehart, Hayes et al. 2006). However, this does not extend to all products due to complex profiles of products and user expectations. For example, coffee and some alcoholic beverages elicit bitterness; however the reported bitterness does not always correlate with liking, as liking is also influenced by other non-taste effects (see discussion in (Lanier, Hayes et al. 2005). For example, factors such as dieting and perceived benefit may also influence intake, rather than strictly taste.

Differences in taste perception are explained, to some extent, by common variations

(polymorphisms) within taste receptor genes (for examples see, (Dotson, Wallace et al. 2012;

Feeney 2011; Meyerhof, Batram et al. 2010)). Other variables have been shown to associate with perceived intensity, such as fungiform papillae (Miller & Reedy 1990) and chronic use and exposure (Prescott & Stevenson 1996). This chapter focuses on using genetic differences within bitter taste receptors, as predictors of differences in perception, which in turn may drive differences in liking. Understanding an individual’s ability to perceive particular compounds, by

8 looking at receptor genetics, or directly measuring perceived response, is an important piece in understanding the factors that drive liking and intake.

Measuring differences in taste perception

Psychophysics is the study of the quantitative relationship between physical stimuli and the perception they elicit. This review deals strictly with sensations experienced from interaction of a stimulus within the oral cavity. This includes prototypical tastes: sweet, sour, bitter, salty and umami carried by taste nerves, as well as non-taste oral sensations such as stinging, burning, cooling, etc mediated via the somatosensory system. However, this review will only focus on the prototypical tastes. Modern psychophysical methods allow the comparison of ratings across individuals, enabling us to better explore individual differences across people. Earlier methods of measuring perceived intensity did not account for individual differences, and due to the nature of the methods, ratings could not be validly compared across individuals. Contemporary methods provide greater insight into individual perceptions of a product or compound more so than liking or preference data, as it provides information about what is being perceived. For these reasons, it is important to consider methods that can accurately measure perception that can easily be compared across individuals.

Threshold vs superathreshold

There are several ways of measuring individual differences in taste perception; however,

I will discuss two distinct methods commonly used in the literature. For more extensive reviews, see (Bartoshuk 1978; Prescott & Tepper 2004). Measuring sensitivity to a particular compound is assessed by determining the concentration at which an individual can detect or recognize a taste

9 in a sample (usually a single compound in water) – these are referred to as detection or recognition thresholds, respectively. A second common way to study individual differences in perception is to measure the perceived strength of a stimulus that is at a concentration above threshold (suprathreshold intensity).

There are several ways thresholds can be measured. For both detection and recognition thresholds, samples can be presented individually or in pairs. The Harris-Kalmus threshold procedure was one of the earlier methods used to determine individual detection thresholds.

Individuals were presented with eight solutions: 4 water and 4 spiked samples with the compound of interest (often phenylthiocarbamide). If participants could correctly sort the water samples from the spiked samples, the test was repeated again with decreased concentration until the participant could no longer correctly identify the 4 spiked samples. If the participant sorted the samples incorrectly, a higher concentration was given on the subsequent trial. The threshold was identified when the individual could repeatedly sort the samples correctly (Harris & Kalmus

1949).

There are several other commonly used threshold methods, including the staircase method or up-down procedure (McBurney & Collings 1977) where individuals inform the instructor if a taste is perceived from the sample. A second method is the forced-choice method

(Wetherill & Levitt 1965), where a water control is presented in addition to a test sample, forcing the participant to choose the sample that elicits a taste. Both of these methods involve presenting test samples at different concentrations in order to determine the lowest concentration that the participant can identify a taste or correctly identify the test sample (see (Bartoshuk 1978) for a short review). Threshold procedures are extremely time consuming due to the large number of trials required to converge on a final value. However, threshold measurements are useful when determining individual sensitivity for compounds.

10

Nevertheless, individual thresholds are limited in their ability to provide information about the perceived intensity of the sensation from the stimulus (Bartoshuk 1978; Bartoshuk

2000). For instance, the bitterness intensity of a compound classically used to measure bitter sensitivity (propylthiouracil) can vary greatly, even among individuals who have similar thresholds (Bartoshuk 2000; Hayes, Bartoshuk et al. 2008). Additionally, perithreshold stimuli may not reflect doses typically found in food or beverage products. Thus, investigators interested in perception, food liking and intake began to measure perceptual differences in suprathreshold stimuli using modern psychophysical methods.

Measuring the perceived intensity at concentrations above threshold provides additional insight into differences in perception across individuals in the range relevant to foods.

Participants provide direct ratings (on line scales or category scales) based on their perceived intensity of the sample. Furthermore, unlike thresholds, multiple ratings for different qualities can be made for a single sample. This is critical for compounds that elicit more than one taste (e.g., non-nutritive sweeteners). Overall, this method is considerably faster and more efficient compared to threshold methods, as not as many stimuli are needed.

There are advantages for each of these methods. The type of experiment or hypothesis determines which type of measurement is required. The advantage of using suprathreshold method is that untrained panelists can be used for collection of individual intensity ratings. This method gives insight on the individuals’ response to a sample in a range more relevant to food.

This method led to the development of scales suitable to measure individuals perceived intensities that was able to generalize across all sensations.

11

General Labeled Magnitude Scale

Throughout the twentieth century, line and category scales were used to capture individual responses to stimuli. Ironically, the experimental psychologists were still arguing over whether people could directly rate the intensity of stimuli long after applied practitioners in the food industry were using these methods (Moskowitz & Sidel 1971; Stevens 1971; Stevens 1946).

Stevens (Stevens 1946) defined 4 levels of measurement: nominal, ordinal, interval and ratio.

Nominal measures are used for items that are classified into categories or groups – for example, coding men as 0 and women as 1. Ordinal data are rank-ordered, but the ordering does not reflect the relative size of the distance between items. An example of an ordinal scale is the grades the

USDA uses to determine the quality of meat. An interval scale has equally spaced values, but a value of zero is arbitrary – for example, each degree Fahrenheit is the same distance apart but 20 is not twice as much as 10. A ratio scale has equal distances between points and zero is meaningful. Ratio scales are commonly used in measurements, such as age or length. When using ratio measurement, one can conclude, a 50 year old is twice as old as a 25 year old (Stevens

1946).

Stevens (Stevens & Marks 1980) developed the method of magnitude estimation, which assumes participants can assign numbers with ratio properties. In practice, researchers originally assigned a number to a reference sample and the participants were asked to assign their own numbers to test samples proportionally to the control value (this was called fixed modulus magnitude estimation). This method was improved upon by allowing participants to assign their own numbers to the control samples and the test samples (Engen & Kling 1972; Stevens 1971) .

For instance, in modulus free magnitude estimation, the participant is first asked to assign a rating for a control sample and is asked to rate each subsequent sample relative to their initial value they assigned to the control sample. This method is based on the idea that if the experimental sample is

12 twice as intense as the reference sample, the participant will assign a value twice as large. Using this method, it became increasingly evident that there were differences between participants’ responses to stimuli due to findings of genetic differences (Hall, Bartoshuk et al. 1975).

Unfortunately, due the nature of magnitude estimation, allowing participants create their own numbered system, it was difficult to compare ratings between individuals (Bartoshuk 2000). For example, if a participant rated a reference sample as a 1 and an experimental sample as a 2, it is assumed the experimental sample is twice as intense for the individual. However, the same conclusion could be drawn for an individual who rated the reference a 10 and an experimental sample a 20. Thus, magnitude estimation did not give insight into the absolute differences in intensity across people, only relative differences between samples. This creates additional problems, as a wide range of numbers could be used, making magnitude estimation data difficult to analyze. Also, although numbers assigned by participants may differ, the actual intensity may be the same, as the numerical difference may merely reflect how the participants assign numbers, rather than underlying differences in perception. Due to apparent differences between ratings among participants, there was a greater need to develop a scale that generalized across individuals. Up until this point, the methods developed for measuring intensity did not allow for easy evaluation across individuals (Bartoshuk 2000). This limitation in magnitude estimation led to the creation of magnitude matching (Marks, Stevens et al. 1988; Stevens & Marks 1980).

Magnitude matching involves presenting two or more sensations (e.g. taste and sound).

By presenting the sensations in a variety of intensities, comparisons can be made across individuals (cross modality). By using a series of lights and tones, researchers were able to make comparisons across individuals, as ratings of light and sound should not differ. Rifikin and colleagues (Rifkin & Bartoshuk 1980) first applied magnitude matching to taste perception. This method improved the comparison between individuals by using sound as a non-taste standard that would be relatively consistent across participants (Bartoshuk 2000).

13

In an attempt to apply this idea of a “non-taste standard”, Green and colleagues (Green

1993; Green, Dalton et al. 1996) developed the Labeled Magnitude Scale (LMS) to generate ratio level data. The LMS also decreased the time needed to train participants and allowed for multiple qualities to be rated for a single sample. Green’s scale was based on the ‘category ratio scale’ created by Borg (Borg 1982) that was also based on ratio scales (modeling Stevens’ method of magnitude estimation). The scale originally developed by Green ranges from ‘no sensation’ (0) to

‘strongest imaginable‘ (100). Five adjectives were selected and placed along the scale at empirically defined points to help participants place their intensity ratings. These include ‘barely detectable’ (1.4), ‘weak’ (6), ‘moderate’ (17), ‘strong’ (35) and ‘very strong’ (53) (Green 1993).

Having the top of the scale (100) anchored to ‘strongest imaginable’ was the first step towards improving the comparison of data across individuals, as it was assumed that ‘strongest imaginable’ is a stable reference across people. However, Green’s original scale implicitly placed all ratings in the context of only oral sensations, meaning when rating the sweetness of a sample, the top of the scale would be the ‘strongest imaginable sweetness’. However, it was soon discovered that due to individual differences in taste perception, the anchor ‘strongest imaginable’ was not equivalent across all individuals (Bartoshuk 2000; Bartoshuk, Duffy et al.

2004). Bartoshuk, Green and others (Bartoshuk 2000) recognized this flaw and changed the top anchor to ‘strongest imaginable sensation of any kind’, resulting in a scale that is presumably independent of taste, allowing for comparison across individuals. This revised scale is referred to the general Labeled Magnitude Scale (gLMS) (Bartoshuk 2000; Bartoshuk, Duffy et al. 2004).

Since the development of the gLMS , it has become widely used across the field (e.g. (Bartoshuk,

Marino et al. 2007; Bennett, Zhou et al. 2012; Dinehart, Hayes et al. 2006; Duffy, Hayes et al.

2010; Green, Lim et al. 2010; Keast 2008; Sandell & Breslin 2006).

14

Figure 1-1: Green’s Labeled Magnitude Scale, with ‘strongest imaginable’ located at 100. For the gLMS, the top anchor is changed to ‘strongest imaginable sensation of any kind’. The gVAS only includes the ‘no sensation’ and ‘strongest imaginable sensation of any kind’. Values are not including in the participant’s scale. Figure taken from [(Green 1993)].

When using this scale, Hayes and colleagues (Hayes, Allen et al. 2013) believe it is critically important to provide participants with appropriate instructions and to have participants practice using the scale prior to the start of the test. That is, it is practice, not verbal labels that generalize the scale. This orientation (or practice session) is typically comprised of remembered oral and non-oral sensations (e.g. brightness of a well-lit room and the loudness of a whisper), encouraging participants to compare tastes and non-taste experiences on the same scale

(Bartoshuk, Duffy et al. 2004; Green & Hayes 2003). Additionally, the orientation (practice) session provides the ability to standardize of the data to a non-oral standard, such as the brightest light you have ever seen, if desired (Duffy, Peterson et al. 2004).

15

Unfortunately, participants often treat the gLMS as a category scale and frequently rate only on the adjectives, even when given specific instructions to rate anywhere along the scale

(Hayes, Allen et al. 2013). Clustered ratings have been reported for other scales (Lawless 2010).

The scale can be simplified by removing the adjectives to create a generalized Visual Analog

Scale (gVAS), as described by Dionne (Dionne, Bartoshuk et al. 2005) and Snyder (Snyder,

Prescott et al. 2006), Hayes and colleagues (Hayes, Allen et al. 2013) showed that gVAS and gLMS scales produce generally similar data, with a few key differences. Ratings were analyzed across individuals after completing two identical sessions, where gVAS and gLMS were used on alternating days. Interestingly, the gVAS produced overall higher ratings across remembered sensations and sampled stimuli. The gVAS did not exhibit categorical behavior (clustering) due to removal of adjectives, but participants reported the gLMS was easier to use than the gVAS

(Hayes, Allen et al. 2013). Additionally, the gVAS has only seen limited use in the field, compared to the widely used gLMS.

Individual differences in perception

Sweet and bitter taste

Sweet taste is innately liked (de Snoo 1937; Mennella & Beauchamp 1998). Perceived sweetness contributes to liking and intake even of foods and beverages that are not predominantly sweet (Dinehart, Hayes et al. 2006; Lanier, Hayes et al. 2005). That is, sweetness is important to consider when dealing with foods that are perceptually complex, eliciting both sweet and bitter tastes. When sweet compounds are combined with a sour or bitter compounds the perceived intensity of both taste sensations are suppressed (Bartoshuk 1975; Keast 2003; Lawless 1979).

This occurs due to mixture suppression, where the intensity for both tastes are diminished

16 compared to intensities for each separately. Measuring the perceived sweetness of a food or stimuli in addition to measuring perceived bitterness gives additional insight into liking and intake of bitter foods and beverages.

Conversely, bitterness is an innately aversive taste, as evidenced by affective facial expressions by newborns (Steiner, Glaser et al. 2001). It is believed this innate response to bitter compounds is a result of a defense mechanism used to identify potential toxic compounds

(Mueller, Hoon et al. 2005). However, Glendinning (Glendinning 1994) notes that this mechanism may have caused the rejection of nutritious and beneficial non-toxic foods that were perceived as bitter. Thus over time, thresholds for bitterness may have increased to allow for the consumption of bitter foods that may provide health benefits, even at the increased risk of ingesting a toxic food (Glendinning 1994). Bitter compounds can be found naturally in edible plants (Barratt-Fornell & Drewnowski 2002; Drewnowski & Gomez-Carneros 2000; Rozin &

Vollmecke 1986) or can also arise during storage due to aging, and also during processing (e.g. maillard browning reaction and roasting coffee) (Meyerhof 2005; Rousseff 1990).

Bitterness often associates with decreased liking (e.g. (Delwiche & Warnock 2008)), and lower intake (Zhao & Tepper 2007). For some products, some bitterness can be a desirable characteristic, such as in beer, wine and cheese (Drewnowski & Gomez-Carneros 2000; Tepper

2008). Additionally, liking for bitter foods and beverages can be learned, through repeated exposure or desirable post ingestive effects (Rozin & Vollmecke 1986). For example, coffee is a very popular beverage that is also extremely bitter to many individuals. Adding cream and sugar to coffee are common practices to mask the bitterness (Mattes 1994). Through repeated exposure and post ingestive effects, consumers may learn to like the taste over time and often reduce the amount of added cream and sugar (Cines & Rozin 1982). Foods and beverages that induce pharmacological effects after ingestion have strong potential to condition increased liking over time (e.g. caffeine in coffee and ethanol in alcoholic beverages) (Keast 2008).

17

While it is believed that bitterness helps to ward off natural plant toxins, many of these bitter compounds also have potential health benefits (Barratt-Fornell & Drewnowski 2002; Craig

1997). Consumption of certain foods (e.g. fruits and vegetables) associate with decreased risk for cardio-vascular disease (Wang, Yamamoto et al. 1995; Wong, Smith et al. 1998), diabetes (Hu,

Manson et al. 2001), and cancer (Riboli & Norat 2003; Steinmetz & Potter 1991)(see (Craig

1997; Duffy 2007) for reviews). However, many of these beneficial compounds associated with decreased health risk are found in foods that are disliked due to their bitterness (Dinehart, Hayes et al. 2006; Drewnowski & Gomez-Carneros 2000; Duffy, Hayes et al. 2010). Examples of these beneficial dietary compounds include anthocyannins, catechans, and tannins from fruits, which are commonly reported as eliciting bitterness (Drewnowski & Gomez-Carneros 2000).

Vegetable intake has been linked with decreased risk for cancer (Riboli & Norat 2003) and coronary heart disease (Joshipura, Hu et al. 2001). Some of the compounds within vegetables thought to be responsible for these health benefits belong to a group of phenolic compounds.

These compounds cause varying amounts of bitterness and astringency (Bravo 2009). For example, kale and broccoli contain quercetin, a flavonol shown to have anticarcinogenic activity and the ability to prevent the oxidation of LDL in vitro (Smith & Yang 1994). Additionally, the bitter compound goitrin is present in many cruciferous vegetables. Goitrin has a similar structure and has similar thresholds to another bitter compound (6-n-propylthiouracil) that is traditionally used to measure sensitivity to bitterness (Wooding, Gunn et al. 2010). The perceived bitterness of these two compounds is associated with greater bitterness from vegetables, decreased liking and lower intake (Bell & Tepper 2006; Dinehart, Hayes et al. 2006; Duffy, Hayes et al. 2010; Keller,

Steinmann et al. 2002; Wooding, Gunn et al. 2010). Similar relationships have been reported for other bitter foods such as alcohol (Duffy, Davidson et al. 2004; Lanier, Hayes et al. 2005) and grapefruit juice (Drewnowski, Henderson et al. 1997; Lanier, Hayes et al. 2005).

18

In products that naturally contain bitter compounds, like grapefruit juice, coffee and alcohol, some bitterness may be acceptable, while excessive bitterness is offensive (Drewnowski

& Gomez-Carneros 2000; Rousseff 1990). Bitter compounds that have been shown to be beneficial in our diet (Craig 1997; Drewnowski & Gomez-Carneros 2000) are not often incorporated into foods as many consumers find them objectionable. Also, bitter compounds naturally found in foods are often removed during processing, or by incorporating inhibitors or bitter blockers to minimize bitterness. In some cases, these bitter compounds are removed or reduced via plant breeding (Roy 1990) (see (Drewnowski & Gomez-Carneros 2000; Rousseff

1990)for examples). While removing these compounds decreases the bitterness and increases acceptability, it may potential diminish the beneficial health effects of these foods.

Measures of bitter sensitivity and acuity

In 1931, it was demonstrated that phenylthiocarbamde (PTC) perception varies across individuals, with some perceiving it as bitter, while others perceiving no taste (Fox 1932). It was hypothesized early on that this difference in ability to detect bitterness from PTC was due to a single locus mutation within a bitter taste receptor gene. Researchers found PTC useful in determining individual bitterness sensitivity, as this was the first compound discovered to vary in its perception across individuals (see (Wooding, Bufe et al. 2006). However, PTC has been largely (but not entirely) has been replaced by a synthetic drug, 6-n-propylthiouracil (PROP) to tests individuals’ ability to detect bitterness. PROP was thought to be superior to PTC, due to the sulfurous odor PTC can elicit (Bartoshuk, Duffy et al. 1994; Prescott & Tepper 2004) and because human toxicology data was readily available, due to PROP’s use as a thyroid medication.

Many studies have attempted to use PROP as a marker of chemosensory variation, to see if such variation is associated with food choice and food intake (Drewnowski, Henderson et al. 1997;

19

Duffy, Davidson et al. 2004; Fischer, Griffin et al. 1961; Hansen, Reed et al. 2006; Keast &

Roper 2007; Keller, Steinmann et al. 2002; Tepper 2008). For example, early work by Glanville and Kplan (Glanville & Kaplan 1965), reported significant correlations between PROP threshold and reported preference for five ‘strong tasting’ foods (e.g. lemon juice, horse radish and vinegar), as individuals exhibiting lower thresholds reported lower preference for these foods.

However, subsequent reports conflict, as some studies fail to find a relationship between PROP bitterness and perceived bitterness from foods, liking or intake (Drewnowski, Henderson et al.

1997; Kamerud & Delwiche 2007). These contradictory results may be due to differences the psychophysical methods used to define bitter phenotype.

Over the years, many different methods have been used to characterize PROP response, making it difficult to directly compare results. PROP response can be measured via threshold or suprathreshold methods. The suprathreshold method of measuring PROP phenotype is used here, as it provides additional insight in individual differences and their relationship to food intake (as mentioned above).

The suprathreshold method for measuring PROP bitterness intensity (henceforth PROP phenotype) has been developed in order to accurately measure individual differences as described by (Dinehart, Hayes et al. 2006; Duffy, Davidson et al. 2004; Hayes, Sullivan et al. 2010; Tepper,

Christensen et al. 2001). PROP bitterness ratings on a gLMS can be analyzed as a continuous variable or can be divided into three groups (see (Hayes & Duffy 2007). Ratings for the highest

PROP concentration (3.2mM) are frequently divided into three groups. The non-taster group consists of people for rate PROP below 22 on a gLMS, while the medium tasters rate the intensity between 23 and 53, with supertasters rate the intensity above a 53, although the exact cutoffs vary from study to study (Dinehart, Hayes et al. 2006; Duffy, Davidson et al. 2004; Tepper,

Christensen et al. 2001). However, this method may attenuate effect sizes due to misclassification bias. That is, two individuals who give ratings of 50 and 55 are far more similar than two

20 individuals who give ratings of 25 and 50, yet the first two would be in different taster groups, while the second two would be in the same group. Thus, using PROP phenotype as a continuous variable may help avoid this bias (Hayes & Duffy 2007). However, work in large samples indicates the classical three-group scheme may in fact reflect the natural underlying distribution

(see (Hayes & Pickering 2012)).

Notably, the bitterness perceived from PROP (PROP phenotype) correlates with bitterness perceived in vegetables (Dinehart, Hayes et al. 2006; Duffy, Hayes et al. 2010) and alcohol (Duffy, Davidson et al. 2004; Duffy, Peterson et al. 2004; Intranuovo & Powers 1998). In addition, greater PROP bitterness has also been shown to be associated with decreased liking of green tea and grapefruit juice (Drewnowski 2001; Lanier, Hayes et al. 2005), beer (Intranuovo &

Powers 1998; Lanier, Hayes et al. 2005), and cruciferous vegetables (Bell & Tepper 2006; Duffy,

Hayes et al. 2010; Keller, Steinmann et al. 2002). Conversely, the bitterness from PROP does not always correlate with the bitterness from other bitter compounds. For example, Kamerud and colleagues (Kamerud & Delwiche 2007) reported bitterness from PROP was not associated with the perceived bitterness of non-nutritive sweeteners (Kamerud & Delwiche 2007). Consistent with Kamerud’s findings, Horne et al. were unsuccessful in finding a correlation between PROP sensitivity and perceived bitterness of acesulfame K and saccharin (Horne, Lawless et al. 2002).

On the other hand, assuming that the ability to detect bitterness from PROP phenotype alone predicts the acceptability of foods, or bitterness sensitivity to all bitter compounds, is naïve, as research has shown that bitter compounds are perceived through multiple bitter taste receptors

(Meyerhof, Batram et al. 2010). Duffy and colleagues (Duffy, Hayes et al. 2009) noted that collecting multiple taste phenotypes (e.g. sucrose and quinine) gives greater insight into liking, intake and perception.

21

Taste receptors

Sweet, umami and bitter tastes are perceived when a ligand interacts with a taste receptor, which are seven-transmembered heterodimertic guanine nucleotide-binding protein (g-protein) coupled receptors (Adler, Hoon et al. 2000). When an agonist activates a receptor, it causes an intracellular unit (alpha) to release from the protein receptor, thus causing a release of calcium ions from internal stores in the cell. This calcium release triggers other ions to end the cells, resulting in depolarization, and ultimately transmission of nerve impulse to the brain causing the perception of sweet, umami or bitter (reviewed by (Margolskee 2002)). As mentioned previously, sweet and umami are perceived through heterodimers, where bitter is perceived through a single receptor, although a single ligand may activate different bitter taste receptors. .

There are 25 bitter taste receptors genes (TAS2Rs) that have been identified in humans

(Kim, Wooding et al. 2005). The TAS2Rs are located on three chromosomes, 5, 7 and 12 (5p15,

7q31 and 12p13) (Adler, Hoon et al. 2000; Bufe, Hofmann et al. 2002; Chandrashekar, Mueller et al. 2000; Matsunami, Montmayeur et al. 2000; Mueller, Hoon et al. 2005) and they encode bitter receptors (T2Rs) expressed on taste cells and elsewhere in the body. It is hypothesized that a wide variety of T2R receptors were important for our ancestors to enable them to detect a broad range of toxins found primarily in plants (Mueller, Hoon et al. 2005; Rozin & Vollmecke 1986). This thesis focuses on variations in bitterness perception, so I will describe bitter taste receptors in greater detail.

A wide variety of bitter compounds are known to activate human T2Rs in vitro (Bufe,

Breslin et al. 2005; Bufe, Hofmann et al. 2002; Chandrashekar, Mueller et al. 2000; Kim,

Jorgenson et al. 2003; Kuhn, Bufe et al. 2004; Pronin, Tang et al. 2004; Pronin, Xu et al. 2007).

For a comprehensive review refer to (Meyerhof, Batram et al. 2010). The functional structure within PTC and PROP was suggested in 1932 (Fox 1932), identified in 1951(Barnicot, Harris et

22 al. 1951) and in 2003 was confirmed (in vitro) to be a thiourea (N-C=S) moiety. This structure activates bitter taste receptors, specifically TAS2R38 (Kim, Jorgenson et al. 2003) and possibly

TAS2R4 (Chandrashekar, Mueller et al. 2000) receptors.

Taste Genetics

It is now known, for the most part, what receptors are responsible for the bitterness arising from a substantial number of bitter compounds; we can use this knowledge to begin to answer questions about variability in bitterness perception across individuals. Likewise, while it is not customary, some studies explore these relationships in real foods prior to determining the specific ligand (contrast Hayes et al 2011 with (Hofmann 2009). Some of this variation in bitterness perception is attributed to genetic mutations, or polymorphisms. Polymorphisms in

DNA can occur due to a number of reasons.

A mutation in the DNA altering the codon expression has the potential to alter the amino acid sequence. Often times, an amino acid change alters the function of the protein, depending on the location within the protein sequence. When these mutations in individual nucleotides are stable across the population (occur greater than 1%) this loci may be referred to as s single nucleotide polymorphism (SNP). However, it is important to mention that not all SNPs alter the function of the protein. SNPs that are associated with altered function of the protein are deemed functional, however they still may not be the causal SNP. Frequently, due to recombination rates,

SNPs are inherited together. Linkage disequilibrium (LD) is a term to describe SNPs that are inherited together (refer to (Weir 2008) for further review of LD). Functional SNPs are often located in the promoter region or within the coding region of a taste receptor gene. It is believed that these functional SNPs alter the orientation of the protein, making it difficult to interact with the ligand (Kim, Jorgenson et al. 2003). Tan and colleagues (Tan, Abrol et al. 2012) found

23 evidence that SNPs located in transmembrane membrane domains 3 – 6 in a GPCR have the potential to prevent hydrogen bonding with the ligand. Commonly, for a single allele, one amino acid is associated with a decrease in function, where as the other amino acid has a normal or increase response. The frequency at which functional polymorphisms occur within TAS2Rs differs across human populations (Hayes, Bartoshuk et al. 2008; Kim, Wooding et al. 2005;

Wooding, Kim et al. 2004).

Figure 1-2: Amino acid sequence depicting extracellular, transmembrane and intracellular regions of the bitter taste receptor TAS2R38. Figure is from [(Wiener 2011)].

Functional SNPs for various TAS2Rs have been discovered through in vitro and in vivo testing. Kim et al. was the first to identify the SNPs responsible for a large amount of variation in the ability to detect PTC and PROP in humans ) (Kim, Jorgenson et al. 2003; Kim, Wooding et al. 2005). Functional SNPs on TAS2R38 (Ala49Pro, Val262Ala and Ile296Val) have been consistently shown to explain the ability to detect bitterness and bitter perception from PTC and

PROP (Duffy, Davidson et al. 2004; Duffy, Hayes et al. 2010; Duffy, Peterson et al. 2004; Hayes,

24

Wallace et al. 2011; Kim, Jorgenson et al. 2003; Kim, Wooding et al. 2005; Tepper, Koelliker et al. 2008).

These three TAS2R38 SNPs form a haplotype, meaning that they are not statistically independent as they are inherited together (i.e. strong LD). The first SNP Ala49 is linked with

Val262 and Ile296, forming the AVI haplotype (Kim, Jorgenson et al. 2003). AVI homozygotes, individuals with two AVI haplotypes, have been shown to have reduced bitterness from PROP and PTC and are sometimes referred to as non-tasters, although this is a misnomer, as many AVI homozygotes will report bitterness from PROP at higher concentrations. In contrast, the taster’ haplotype PAV, Pro49, Ala262 and Val296, associates with greater bitterness from PROP (Duffy,

Davidson et al. 2004), and PTC (Kim, Jorgenson et al. 2003). This relationship is also evident with foods such as alcohol (Duffy, Davidson et al. 2004; Duffy, Peterson et al. 2004; Hayes,

Wallace et al. 2011; Intranuovo & Powers 1998; Wang, Hinrichs et al. 2007) and cruciferous vegetables (Dinehart, Hayes et al. 2006; Duffy, Hayes et al. 2010; Sandell & Breslin 2006). Other haplotypes for these SNPs (combinations of inherited amino acid sequences) are possible; however, they are rare and vary in frequency across populations (Wooding, Bufe et al. 2006).

Kim (Kim, Jorgenson et al. 2003) were the first to report frequencies for TAS2R38 haplotype across two different ethnic populations. In the European population, haplotypes frequencies were balanced with 47% of the population having the non-taster (AVI) haplotype, and 49% taster

(PAV) haplotype, with the rare AAV haplotype existing in only 3% of the population. Within the

East Asian population, the less functional haplotype (AVI) was less prevalent (30%), compared to the higher functioning (PAV) haplotype (70%). Thus, in some mixed cohorts it is important to consider ancestry when measuring haplotype frequencies and linkage disequilibrium, as allele frequencies may differ by ancestry.

It is important to recognize that other variables are associated with PROP phenotype (see

(Hayes, Bartoshuk et al. 2008), such as fungiform papillae density (Miller & Reedy 1990) and

25 body weight (Calo, Padiglia et al. 2011; Tepper 1998; Tepper & Ullrich 2002); however, these associations are beyond the scope of this review.

Polymorphisms in bitter taste receptors and their impact on liking and intake of foods and beverages

Numerous studies have suggested functional SNPs found on TAS2Rs are linked with perceived bitterness from other ingredients, foods and beverages (Bufe, Hofmann et al. 2002;

Duffy, Davidson et al. 2004; Duffy, Hayes et al. 2010; Hayes, Wallace et al. 2011; Kuhn, Bufe et al. 2004; Pronin, Xu et al. 2007; Roudnitzky, Bufe et al. 2011). In addition to the relationship between TAS2R38 polymorphisms and thioruea (PROP/PTC) bitterness, these haplotypes have been used to explain variance in bitterness perception of other bitter foods and beverages (e.g.

(Duffy, Davidson et al. 2004; Duffy, Hayes et al. 2010) (see (Feeney 2011) for review). The explanation is that individuals who experience more bitterness (PAV carriers) are less likely to consume bitter foods. Epidemiological data conflict: this effect was observed in one branch of the

European Prospective Investigation into Cancer and Nutrition (EPIC) study (Sacerdote, Guarrera et al. 2007), but no evidence of a TAS2R38 effect on vegetable intake was seen in the Diet,

Cancer and Health (DCH) study cohort (Gorovic, Afzal et al. 2011). It is important to keep in mind that this receptor is not responsible for the sensitivity to all bitter compounds, as multiple receptors respond to a wide variety of bitter ligands (Hayes & Keast 2011; Meyerhof, Batram et al. 2010).

Many bitter taste receptor genes are polymorphic, and through in vitro and in vivo studies, many of these polymorphisms have been shown to be functional (e.g. TAS2R9 (Dotson,

Zhang et al. 2008)). Psychophysical studies show SNPs in genes other than TAS2R38 are associated with the intake and liking of a variety of foods and beverages. For example, Hayes et

26 al. reported that a SNP in TAS2R19 (Arg299Cys) is significantly associated with bitterness and liking of grapefruit juice (Hayes, Wallace et al. 2011). Individuals with one or more copies of Arg at position 299, perceived less bitterness that the Cys homozygotes. Also, the Arg homozygotes perceived the greatest sweetness and liking (Hayes, Wallace et al. 2011). This same SNP was implicated in variable quinine bitterness by Reed and colleagues (Reed, Zhu et al. 2010).

Additionally, Hayes and colleagues (Hayes, Wallace et al. 2011)) reported that four SNPs within

TAS2R receptor genes (TAS2R3, TAS2R4, and TAS2R5) formed a haplotype that predicted the perceived bitterness of coffee. Although groups greatly differed in bitterness perception (~2 fold difference in bitterness), the haplotype did not predict any difference in liking ratings for coffee

Bitterness of beverages

While sweetened beverages on the market are primarily sweet, some may also elicit sensations of bitterness. For example, in diet sodas, non-nutritive sweeteners are used to replace sucrose and high fructose corn syrup to decrease the caloric content while maintaining sweetness.

Likewise, alcoholic beverages often elicit both sweet and bitter tastes. This bitterness can be attributed to ingredients such as hops in beer, phenolics in wine, and alcohol itself. These two beverage categories make up a large portion of the beverage industry, with carbonated soft drinks alone reaching $43.2 billion in sales in 2011 (Mintel 2012), and imported and domestic beer sales reaching $78 billion (Mintel 2012), and wine sales were projected to reach $40 billion in sales for

2012 (Mintel 2012). Gaining insight to the bitterness of these beverages is important to the industry (Rousseff 1990), as many reports suggest that greater perceived bitterness is associated with decreased liking (Duffy, Davidson et al. 2004; Hayes, Wallace et al. 2011; Kamerud &

Delwiche 2007) and ultimately intake (Dinehart, Hayes et al. 2006; Lanier, Hayes et al. 2005).

27

Sweeteners

Humans show a preference for sweet taste at a very young age (Mennella & Beauchamp

1998), even before birth (de Snoo 1937). It has been suggested that sweetness is an indication of calorically dense foods thus, detection and preference for sweetness was selected for evolutionarily (Chandrashekar, Hoon et al. 2006). Given the modern food supply, we no longer need to rely on this mechanism; however, we have retained our hardwired preference for sweet foods.

The sugar and sweetener market is on track to reach $5 billion in sales in 2013. Sugar is the largest segment in sweetener retail sales, followed by syrups/molasses and then sugar substitutes and honey (Mintel 2010). Sugar is a staple ingredient in homes, and is a commercially popular sweetener and bulking agent in foods and beverages. In spite of growing health concerns about added-sugars, reports show that the consumption of sweet foods and beverages has increased. Between 2005-2010 sugar sales have increased by 18% (Mintel 2010). In a national survey (1994-1995) it was reported that the average American (over the age of 2) consumes 318 kcal/day from added sugar (Nielsen, Siega-Riz et al. 2002). The World Health Organization recommends that added sugars in the diet should account for no more than 10% of dietary energy

(WHO 2003) however; it is reported that added sugar makes up 15.8% of total energy in

American diets. Of the 15.8% of added sugar, almost half (47%) is from consumption of non-diet soft drinks (Guthrie & Morton 2000).

To explore whether the consumption of sugar-sweetened beverages is associated with weight gain, Malik et al. (Malik, Schulze et al. 2006) compared 30 studies (15 cross-sectional, 10 prospective and 5 experimental), and concluded that the consumption of sugar-sweetened beverages is positively associated with weight gain, obesity and diabetes in children and adults.

However, not all reports support this finding. For example, Song and colleagues (Song, Wang et

28 al. 2012) report that total caloric intake (not total sugar consumption) is the most important dietary factor for predicting increased BMI (Song, Wang et al. 2012). However, total sugar intake makes up a large portion of total calorie intake (Berkey, Rockett et al. 2012; Bleich, Wang et al.

2009), and so it can still be considered major factor for obesity risk.

This increase in sweetened food consumption, in particular caloric soft drinks, has resulted in a nation wide health concern as obesity (Lobstein, Baur et al. 2004; Ludwig, Peterson et al. 2001) and type II diabetes (Bleich, Wang et al. 2009; Schulze, Manson et al. 2004) continue to increase (Ogden, Carroll et al. 2012). Additionally, higher rates of dental caries are also associated with consumption of sugary sodas and drinks among children (Marshall, Levy et al.

2003) and adults (Heller, Burt et al. 2001).

In non-diet sodas, bulk carbohydrate sweeteners like sucrose and high fructose corn syrup are responsible for the caloric content. Several studies report that caloric beverage intake associates with obesity (Olsen & Heitmann 2009). There are conflicting reports concerning caloric beverage intake and obesity (Pereira 2006), and it was concluded that more research is needed to confirm this relationship. Due to increased prevalence of obesity and to help manage weight, consumers are now demanding low calorie alternatives (Sandrou & Arvanitoyannis 2000;

Swann 2005). One way to achieve this is to replace some or all of the added sugar with non- nutritive sweeteners. This allows the industry a way to provide consumers with products that have reduced or zero calories, while maintaining desirable levels of sweetness. The beverage industry has had substantial success in utilizing non-nutritive sweeteners to offer diet and zero calorie beverages.

Replacing caloric beverages, such as soft drinks, with non-nutritive sweetened beverages may help to reduce weight gain (Berkey, Rockett et al. 2012; Malik, Schulze et al. 2006). For example Coca-Cola has 240 kcal per 20fl oz, compared to 0 kcal per 20fl oz for Diet Coke or

Coke Zero.

29

An early study by Tordoff and Alleva (Tordoff & Alleva 1990) suggested beverages sweetened with HFCS are associated with increased caloric intake and an increase in body weight while aspartame sweetened beverages were associated with a decrease caloric intake and a decrease in body weight of male participants. The authors concluded that non-nutritive sweetened beverages and decreased sugar intake may help to maintain body weight (Tordoff & Alleva

1990).

Raben and colleagues (Raben, Vasilaras et al. 2002) reported that in overweight individuals, non-nutritive sweetener intake was associated with a decrease in weight compared to a sucrose supplemented diet. Overweight individuals were placed into two groups, and for ten weeks consumed either sucrose sweetened foods (roughly 152g sucrose/day) or foods supplemented with non-nutritive sweeteners. Soft drink beverages provided the greatest source of sucrose or non-nutritive sweetener supplement (70 to 80%). The group that received the sucrose diet gained on average 1.6kg and the non-nutritive sweetener supplemented group decreased weight by 1.0kg over a period of 10 weeks (Raben, Vasilaras et al. 2002). Thus, they concluded individuals seeking to lose or maintain weight should are better off selecting a non-nutritive sweetened beverage over a sucrose-sweetened beverage (Raben, Vasilaras et al. 2002).

However, in epidemiological studies, diet soda intake has been positively associated with obese and overweight (Berkey, Rockett et al. 2012; Forshee & Storey 2003; Yang 2010) as shown by . In these epidemiological data, the causal direction for this association is unclear.

Conceivably, overweight and obese individuals may select diet beverages more often than lean individuals because they are attempting reduce or prevent weight gain (Berkey, Rockett et al.

2012). Alternatively however, some researchers argue that non-nutritive sweetener use may actually increase obesity risk. That is, because non-nutritive sweeteners elicit a sweetness response, our bodies expect an increase in calories. When those calories are not consumed, there is a signal to consume more calories, making it difficult to regulate calorie intake (Davidson,

30

Martin et al. 2011; Mattes & Popkin 2009). See (Mattes & Popkin 2009)) for a review of non- nutritive sweetener effects on reward systems and calorie intake. A large well-designed clinical trial is needed to resolve this issue.

Non-nutritive Sweeteners

Irrespective of conflicting epidemiological evidence, many Americans report using sugar substitutes to help lose weight or maintain weight (Mintel 2010). Between 1990 and 2000, consumption of reduced calorie beverages has increased, and up to 28% of the US population consumes foods or beverages containing non-nutritive sweeteners (for White and Hispanic populations, 22% for Black population) (Sylvetsky, Welsh et al. 2012). There are numerous non- nutritive sweeteners that are approved for use in foods and beverages to reduce energy density.

The replacement of caloric sweeteners by non-nutritive sweeteners is not straightforward and poses many challenges to the food and beverage industry. Sucrose does not merely add sweetness, as it also has functional properties that are not provided by non-nutritive sweeteners.

For example, sugar provides critical structure in cookies and ice cream that is not provided by non-nutritive sweeteners. However, non-nutritive sweeteners have seen substantial success in beverage applications, although even then, they do not provide the mouthfeel and viscosity of bulk carbohydrates.

The ability of non-nutritive sweeteners to provide sweetness without adding additional calories is due to their ability to elicit sweetness at very low concentrations. In other words, they are highly potent, so very little (relative to the amount of sucrose or HFCS) is needed to equal the sweetness of sucrose at concentrations used in foods in beverages. As DuBois illustrated, sucrose has a linear relationship with perceived sweetness (DuBois, Walters et al. 1991). That is, as the concentration of sucrose increases, an equal increase of sweetness is perceived. Other sugars,

31 including fructose, glucose and sugar alcohols (lactitol, maltitol and isomalt) also show this linear dose response function. However, most non-nutritive sweeteners do not follow this relationship

(DuBois, Walters et al. 1991; Hellfritsch, Brockhoff et al. 2012; Schiffman, Booth et al. 1995).

Instead, the sweetness response curves for non-nutritive sweeteners are hyperbolic, with sweetness responses increasing rapidly then asymptotically reaching a maximum sweetness as concentrations increase (DuBois, Walters et al. 1991). For this reason, referring to non-nutritive sweeteners as high intensity sweeteners is misleading, as sucrose can reach greater maximal sweetness than most non-nutritive sweeteners. Also, the dose response curves differ across non- nutritive sweeteners (DuBois, Walters et al. 1991). Due to potency differences across non- nutritive sweeteners, the amount needed to replace sucrose in a product varies; however, in general nutritionally a marginal amount is needed and therefore the caloric contribution is negligible even it the sweetener is metabolized. That is, while aspartame provides 4kcal/g (von

Rymon Lipinski 1985), the amount used is so low, these calories are not nutritionally relevant

(hence the term non-nutritive instead of non-caloric). Alternatively, some other non-nutritive sweeteners, like AceK, are not metabolized (Vetsch 1985).

32

Figure 1-3: Sweetness and bitterness response curves for sucrose, acesulfame-K, aspartame and rebaudioside A. These figures are taken from [(DuBois, Walters et al. 1991)].

However, sucrose and other caloric sweeteners have additional functional properties beyond sweetness that non-nutritive sweeteners do not provide. As summarized by Davis (Davis

1995) sucrose and other bulk sweeteners contribute to the composition of a product. For example, sucrose is essential for baking, giving bread and baked goods brownish color. Sucrose is a reducing sugar and when heated reacts with amines in the dough resulting in products that give baked goods their color, taste and aroma. Sucrose also offers antimicrobial properties, by binding water molecules, making water unavailable to microbes. Reducing sucrose in foods may alter the shelf life or safely off foods. Bulk sugars also contribute to the texture and mouthfeel of products.

By removing a bulking ingredient, the density and viscosity are reduced. These physical and chemical properties are not provided by non-nutritive sweeteners due to their relatively small size

33

(compared to sucrose) and the trivial amount of sweetener it takes to reach the desired sweetness of the product.

The beverage industry is one of the largest customers of non-nutritive sweeteners.

Compared to other food products, beverages have had relative success in replacing sucrose with non-nutritive sweeteners. That is not to say other types of products cannot use non-nutritive sweeteners successfully; however, it is more difficult to reformulate and find the best combination of ingredients to mimic the properties of sucrose, while still reducing calories and added sugar content.

There are many natural and synthetic non-nutritive sweeteners used globally. In the

United States, the FDA has approved seven sweeteners: rebaudioside A, rebaudioside and stevioside blend, acesulfame potassium, aspartame, Neotame, saccharin, and sucralose. Currently,

GRAS (generally recognized as safe) status is pending for rebaudioside D. These sweeteners are most commonly approved as food additives, with some exceptions (e.g. Stevia is labeled as a dietary supplement). To be granted approval for use in foods, ingredients must go through extensive testing to ensure that the levels used in foods and levels that would be consumed do not pose any health risks.

Saccharin was the first non-nutritive sweetener to be discovered. It was granted GRAS status in 1959 by the FDA and was approved for use as a table sweetener known at retail as Sweet

‘N Low. To give an example of the potency of saccharin, one packet (1g) of Sweet ‘N Low

(36mg saccharin) is equivalent to 2 tablespoons of sucrose (roughly 25g). Since the discovery of saccharin, many other non-nutritive sweeteners have been discovered and tested for their safety.

In 1967, acesulfame potassium (AceK) was discovered (Clauss & Jensen 1973). It provides no calories and very little is needed to replace sucrose in foods. For example, to match the sweetness of a 7.5% (w/v) sucrose sample, about 100 times less AceK is needed (0.075% w/v); however replacement is dependent on the concentration, as the relationship is not linear

34

(DuBois, Walters et al. 1991). AceK has been used across a wide range of products, with the beverage industry being the leading consumer, specifically soda (von Rymon Lipinski 1985).

Aspartame was discovered in 1965 by the chemist James Schlatter (Mazur 1976). It has been approved by over 100 countries, including the United States (US FDA 181) and has been granted GRAS status. Aspartame has been approved for use in all food products (US FDA,

1996a) including use as a table sweetener, and is available at retail under the name Equal. The sweetness of aspartame is between 160 and 220 times sweeter than sucrose on a per weight basis

(depending on the food product) (Beck 1974). In other words, to match the sweetness of a 7.5%

(w/v) sucrose solution, approximately 0.05% (w/v) of aspartame is needed.

However, many consumers distrust ‘artificial’ sweeteners, and desire non-nutritive sweeteners derived from natural sources. The Stevia plant has been used as a natural non-nutritive sweetener for centuries, but Stevia derivatives have only received regulatory approval in the US recently. Stevia is derived from the leaves of Stevia rebaudiana (Bertoni) Bertoni, which has been used as a sweetener by the Paraguayan Indians and Mestizos for centuries, and was first introduced to the Europeans in 1886 (Brandle, Starratt et al. 1998; Lewis 1992). The sweetness is derived from steviol glycosides found in the leaves of stevia. These glycosides all differ in sweetness response curves, and in the concentration found in the leaf, which has been attributed to species differences (see (Wölwer-Rieck 2012)). The most abundant glycoside in Stevia is stevioside, which is 200 to 300 times sweeter than sucrose on a weight basis (Crammer 1987), meaning 0.125% (w/v) stevioside is needed to match the sweetness of a 7.5% (w/v) sucrose solution. Interestingly, Stevia leaves are not approved for use in foods in the US. Instead, a Stevia extract containing a blend of 11 steviol glycosides is approved as a dietary supplement. The purified Stevia glycosides, stevioside (≥95% pure) and rebaudioside A (RebA) (≥97% pure), are approved by the FDA as GRAS ingredients.

35

Bitterness and off-tastes

As mentioned previously, many non-nutritive sweeteners are not only sweet, but they also elicit bitter or metallic side-tastes. Interestingly, non-nutritive sweeteners have different bitterness response curves and these functions do not parallel their respective sweetness response curves; in some cases, bitterness increases linearly with concentration (DuBois, Walters et al.

1991; Hellfritsch, Brockhoff et al. 2012; Schiffman, Booth et al. 1995). This bitterness provides industry with the difficult task of replacing sucrose with non-nutritive sweeteners, as non-sweet side-tastes may deter consumers from purchasing their product. Finding a sweetener or combination of sweeteners to minimize these effects is challenging. Blending multiple sweeteners has been shown to minimize bitterness (DuBois & Prakash 2012) (e.g. (Helgren 1957)). Other solutions to decrease bitterness is to include a low amount of sucrose, masking some of the bitterness, due to the phenomena known as mixture suppression (see Bartoshuk (1975). However, adding sucrose to minimize bitterness may add undesirable calories.

Bitterness in non-nutritive sweeteners associate with TAS2R polymorphisms

Despite the reports of bitterness from non-nutritive sweeteners (DuBois, Walters et al.

1991; Hellfritsch, Brockhoff et al. 2012; Schiffman, Booth et al. 1995), bitterness response curves averaged over panelists do not depict the wide variability in perceived bitterness, as reports show bitter taste receptor genetics mediate the ability to detect bitterness from several non-nutritive sweeteners (Horne, Lawless et al. 2002; Kuhn, Bufe et al. 2004; Roudnitzky, Bufe et al. 2011).

Early work by Bartoshuk (Bartoshuk 1979) explored individual differences in perceived bitterness from saccharin with respect to PROP thresholds. Non-tasters of PROP reported less bitterness from beverages sweetened with saccharin than PROP tasters. Interestingly, as the

36 concentration of saccharin increased, the difference in perceived bitterness attenuated between the groups. Reported sweetness did not differ between the groups. As the concentration of saccharin increased and reported sweetness reached a maximum and even decreased in sweetness at the highest concentrations (Bartoshuk 1979). However, subsequent work by Horne and colleagues

(Horne, Lawless et al. 2002) were unable to replicate Bartoshuk’s findings, and did not find an association between PROP bitterness and saccharin (Horne, Lawless et al. 2002). This discrepancy in results may be a result of the subsequent identification of a third group of PROP tasters that were previously grouped with the ‘taster’ group. This group, named ‘supertasters’, perceives greater PROP bitterness than ’medium’ tasters, (Bartoshuk, Duffy et al. 1994; Reedy,

Bartoshuk et al. 1993) and may have obfuscated the association between PROP taster group and saccharin bitterness.

Using in vitro and in vivo methods, Kuhn and colleagues (Kuhn, Bufe et al. 2004) reported that hT2R43 (previously hT2R61) and hT2R31 (previously hT2R64 and hT2R44) were responsible for the bitterness of saccharin. Interestingly, TAS2R43 is very similar in its amino acid sequence to TAS2R31, with 89% of the protein being the same base pair sequence (Pronin,

Xu et al. 2007). Functional mutations (SNPs) were identified that alter cellular response of these receptors for saccharin. Pronin and coworkers (Pronin, Xu et al. 2007) reported two SNPs in hT2R43, located at position 35 and 212, and eleven SNPs within hT2R31 (with the most variation in activation occurring at positions 35, 162, 227 and 240). There was a single SNP within hT2R31 at position 35 thought to be critical for receptor function in vitro. Pronin et al. observed in vivo that aloin and saccharin thresholds differed by the same critical polymorphism in

TAS2R31 at position 35. Individuals lacking the functional SNP Trp35 in TAS2R31 have higher thresholds for AceK bitterness than Arg35 genotypes (Pronin, Xu et al. 2007). This SNP is located in the first intracellular loop of the receptor. It was hypothesized that for TAS2R31, when

37

Tryptophan (Trp) is replaced with Arginine (Arg), this change may modify the ability of the protein to interact with the ligand by altering the structure of the receptor (Pronin, Xu et al. 2007).

Roudnitzky and colleagues (Roudnitzky, Bufe et al. 2011) also explored polymorphisms in TAS2R43 and TAS2R31 and their impact on bitterness from AceK and saccharin. 42 SNPs from

5 highly related bitter receptors on chromosome 12 (TAS2R30, TAS2R31, TAS2R43, TAS2R45 and TAS2R46) were included in the analysis. 30 SNPs were found to be significantly associated with the detection thresholds of AceK and saccharin; however only 25 were non-synonymous

(meaning the nucleotide mutation lead to a change in amino acid). Of these, two SNPs were most notable, the Ser35Trp SNP in TAS2R43 and the Arg35Trp SNP in TAS2R31, coinciding with

Pronin’s results. They were shown to be the critical SNP in bitterness detection thresholds for both AceK and saccharin in vivo. However, other SNPs within TAS2R31 and TAS2R43 were also significantly associated with bitterness thresholds. Additional analysis revealed that many SNPs were in linkage disequilibrium (inherited together) with the critical 35 SNP. Using in vitro assays

Roudnitzky reported TAS2R31 function is altered by 11 SNPs, and found that in the presence of the functional Trp35, activation can be altered depending on the amino acid located at the following loci: His45, Phe237, Arg276 and Cys281. In contrast, the four amino acids at alleles

Met162Leu, Gln217Glu, Ala227Val, and Val240Ile did not hinder response for saccharin or

AceK. In the presence of Arg at position 35 there is a reduction in activation in vitro, regardless of amino acid for the other 8 SNPs. In summary, Roudnitzky found TAS2R31 Arg35Trp is major cause of bitterness variability for saccharin and Acek (in vitro and in vivo); however, SNPs in other receptor genes may explain additional variance in AceK bitterness (Roudnitzky, Bufe et al.

2011).

Due to the genetic variations in bitter taste receptors, it is crucial for industry to consider the range of responsiveness across the target population when using a non-nutritive sweetener.

Whether SNPs within TAS2R31 might also explain differences in the perceived bitterness of other

38 non-nutritive sweeteners like Stevia extracts is unknown. Determining whether bitterness from non-nutritive sweeteners is associated across compounds may imply that bitterness is perceived through the same bitter receptor; however, comparisons of bitterness across non-nutritive sweeteners have not been examined.

Sensations from of Alcoholic beverages

Alcoholic beverages are inherently bitter, due to their ethanol content and other ingredients such as hops and anthocyanins in beer and wine respectively. It is hypothesized that the ability to perceive bitterness from alcoholic beverages may impact alcoholic beverage consumption (Dinehart, Hayes et al. 2006; Lanier, Hayes et al. 2005), and potentially alcoholism risk (Dotson, Wallace et al. 2012; Wang, Hinrichs et al. 2007). However, alcoholic beverages are not strictly bitter, but are simultaneously sweet and bitter.

Traditionally, alcoholic beverage intake is measured in standard drinks, in terms of equivalent ethanol content. One drink is considered either: 12 oz beer, 5 oz wine, or 1.5 oz of liquor (NIAAA 2005). However, the amount of bitterness and sweetness varies greatly among these beverages. For instance, a lemon drop (a type of shot) has a very different taste profile than a shot of neat whiskey. For this reason, all alcoholic beverages are not equivalent from a sensory perspective, suggesting they should be individually analyzed when considering effects of taste receptor genetics on alcohol liking and intake.

Additionally, carbonation is an important component in beer and some wines.

Carbonation causes tingling and also can be perceived as sour and has been shown to differ by concentration. Furthermore, PROP phenotype has been associated with sensitivity to carbonation levels. PROP supertasters experienced greater sourness from carbonation, followed by medium

39 tasters and non-tasters (Prescott, Soo et al. 2004). Carbonation is beyond the scope of this thesis; however, it also contributes to the complex sensations elicited by alcoholic beverages.

Astringency, a term associated with drying and puckering sensation in the mouth, is not a true taste, as it in carried to the brain via touch nerves rather than taste nerves. Nonetheless, it is also extremely important sensation to consider when discussing alcoholic beverages. Astringency is caused by polyphenols (that are naturally contained in wines) attaching to saliva proteins,

(which are responsible for coating the tongue and cheeks). When these complexes form, the salivary proteins precipitate, causing a loss of lubrication in the mouth (Bajec & Pickering 2008).

Although astringency also falls outside of the context of this review, it can often be confused with bitterness. Polyphenols, a component of wine, differs in the ratio of astringency experienced compared to bitterness (Robichaud & Noble 1990). The concentration of polyphenols in wines can be found anywhere from 1000 to 3,500 mg/L, due to differences in processing (Blanco, Auw et al. 1998).

Other ingredients in alcoholic beverages can vary by beverage type. Sucrose is added to a wide range of beverages, creating a sweeter beverage; this added sugar has the potential to suppress perceived bitterness due to mixture suppression (Lanier, Hayes et al. 2005; Martin &

Pangborn 1970).

Hops is an important ingredient added to beer both for flavor, and for functional reasons such as stability. The taste of hops can be described as bitter (Moir 2000). This bitterness is often a desired quality in some beers, making analyzing effects of taste receptor genetics with intake and reported liking less straight forward. That is, not all bitterness may be bad.

Ethanol elicits sweet and bitter taste sensations (Scinska, Koros et al. 2000). Perceived bitterness and oral irritation ratings from ethanol concentrations vary by PROP phenotype

(Bartoshuk, Conner et al. 1993; Prescott & Swain-Campbell 2000). PROP supertasters experienced greater bitterness and oral irritation than PROP non-tasters, from samples ranging

40 from 10-50% ethyl alcohol (Bartoshuk, Conner et al. 1993). This evidence suggests that the ability to perceive bitterness from ethanol may be linked with a genetic variation within

TAS2R38; however, ethanol has not been included as a screening stimulus in in vitro functionally expression experiments with various T2Rs.

Early work suggested that not all beverages are liked equally, and these differences may differ by PROP phenotype (Intranuovo & Powers 1998). Intranuovo and Powers asked participants to taste two lager beers (Budweiser and Pilsner Urquell) and rate perceived bitterness and liking. Across all PROP groups, greater bitterness was reported for Pilsner Urquell than for

Budweiser, and ratings for Budweiser were not different between PROP taste groups. For Pilsner,

PROP supertasters reported significantly higher bitterness ratings compared to medium and non- tasters. Liking ratings for the beer correlated with reported bitterness of the sampled beer

(Intranuovo & Powers 1998). Interestingly, this study has been the only study to date to compare liking ratings for sampled alcoholic beverages with different inherent levels of bitterness.

Other reports show that PROP bitterness ratings are associated with alcohol consumption

(Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004; Duffy, Peterson et al. 2004). Lanier and colleagues (Lanier, Hayes et al. 2005) reported PROP bitterness associated with increased bitterness and decreased sweetness of sampled blended whisky and beer, which in turn associated with lower alcohol intake. This clearly shows alcoholic beverages are not simply sweet or bitter but a combination of multiple taste qualities. For example, wine can simultaneously be sweet, sour, and bitter in addition to other oral somatosensations such as astringent, and irritating

(Thorngate 1997). Bitterness greatly differs across beverages, which could be attributed to a difference in alcohol content as it differs across beverage type, and also the amount of sugars added or naturally occurring in beverages. However, not all reports have been able to confirm the relationship between PROP bitterness and alcohol intake (Mattes & DiMeglio 2001). For this

41 reason, may be important to consider each beverage type individually when making comparisons between taste phenotype or genotype with consumption and liking.

Numerous studies have looked at intake frequency and total consumption of alcoholic beverages and PROP phenotype and/or polymorphisms on bitter taste receptors, including

TAS2R38 and TAS2R16 (Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004; Duffy,

Peterson et al. 2004; Hayes, Wallace et al. 2011; Wang, Hinrichs et al. 2007). Individuals reporting greater bitterness from PROP consume fewer alcoholic beverages (Duffy, Peterson et al. 2004; Intranuovo & Powers 1998). Similar conclusions were drawn for TAS2R38 genotype; several studies show AVI homozygotes consumed more alcoholic beverages, and drink alcohol more frequently than heterozygotes and PAV homozygotes (Dotson, Wallace et al. 2012; Duffy,

Davidson et al. 2004; Hayes, Wallace et al. 2011). Although looking at total consumption of alcoholic beverages is useful, especially for addiction behavior and health risk, this approach does not take into account sensory profiles of alcoholic beverages. To date, there has not been any research investigating the relationship between bitter taste receptor variants and liking ratings for different types of alcoholic beverages.

Conclusions

Psychophysical methods provide sensory scientists a means to quantify individual differences in taste perception. There have been several methods developed to measure taste perception and capture individuals’ responses to a stimulus. Perceived intensity ratings

(specifically bitterness and sweetness) have been linked with the liking or intake of foods, and even health risks (Basson, DiChello et al. 2005; Tepper, Keller et al. 2003). This information can assist the food industry in developing ideal products for their target populations. In the future, the outcome of these studies have the potential to aid in the creation and development of products

42 based on consumers taste receptor genetics, giving consumers products better tailored to their genotype.

There are many taste receptors that contain SNPs that have yet to be explored in vitro and in vivo. Several groups have completed large comprehensive studies examining bitter taste receptor activation in vitro (see (Meyerhof, Batram et al. 2010) for review); however, many of these have yet to be confirmed in human behavioral studies. By examining taste receptors and their variations, we can explore their effect on taste perception from single compounds or whole food products. These results have the potential to significantly affect the way product developers design their products.

In the near future, food and beverage companies may use new insights from taste biology to formulate products that limit perceived bitterness. As this field continues to develop, knowledge of genetic polymorphisms in bitter taste receptors and the compounds that activate those receptors will be potentially important. This may provide the industry with the tools to minimize bitterness across populations. Using ingredients that show the least variation across genotypes may help to provide the most uniform product across populations. This can be achieved by selecting ingredients that do not vary greatly in bitterness perception or by further altering processing to prevent the occurrence of bitter compounds. With further testing, we will gain additional insight on bitter taste perception and how it influences liking and intake and be able to apply this knowledge to products, especially those with additional health benefits.

Another way in which the food industry may take advantage of this new knowledge is to create multiple options (formulations) of a product, tailored for different genetic groups. By offering a variety pack of products, the consumer could then find their ideal product within the set and buy that formulation in the future. An example of this sort of targeted approach can already be seen within the beverage industry today, with the existence of zero calorie, reduced calorie and regular soda. Although this may be overtly be marketing for individuals who want no and reduce

43 calorie options, it also provides consumers with beverages that have different blends of a non- nutritive sweetener offering different sensory profiles. For those who find some sweeteners to have a side-taste, the reduced calorie option may serve as an alternative to the no calorie option, as the sucrose or HFCS would mask some of the bitterness elicited by the non-nutritive sweetener. Additionally, glycosides extracted from Stevia exhibit a varied bitterness across glycosides and concentrations. Individuals who perceive bitterness from one non-nutritive sweetener may not perceive bitterness from other non-nutritive sweeteners. For instance, bitterness of AceK is not correlated with the perceived bitterness of RebA or RebD (Allen,

McGeary et al. 2013). It would be useful to know if bitterness from non-nutritive sweeteners are associated with each other and to determine which sweeteners that elicit the least amount of bitterness.

Another example of genetic market segmentation exists today for tabletop non-nutritive sweeteners. Several types of sweeteners are available at restaurants and coffee shops in different packets. Although the specific sweeteners may come from different companies, the concept is the same as above. Individuals have several options, allowing them to select the sweetener they prefer. They learn to associate different sweeteners with package color. Ultimately, off-taste and bitterness is reported to be the number one contributor in sweetener preference. However, there are a number of other factors that contribute to sweetener preference, such as: perceived safety, cost and sweetness (Mintel 2012). Allowing consumers to select their ideal product from several options may help a food company to reach a larger proportion of the total market.

Genetic testing and screening may even be utilized during product testing. Often, companies have expert sensory panels that rate the product using internal references and descriptors. However, variations in taste receptors can alter the ability to detect tastes and taste qualities of references. If a company was unaware of these inherent differences, they may use an average the group means for the ratings of a product that contains a non-nutritive sweetener,

44 resulting in an inaccurate characterization of the product. In reality there could be in fact two distinct groups of low and high responders. A clearer and more accurate picture of consumer experiences could be achieved by taking genetic differences into account when interpreting results. Using genetic screening in sensory panels would also be to their advantage to ensure their results are representative of the target population.

However, genotyping entire panels may not be financially possible for some companies.

At minimum, taste panels should be screened with ingredients that are in the test product. As the number of ligands known to activate receptors and more functional polymorphisms are discovered, it becomes increasingly evident that panelists can no longer be screened using single compounds that may or may not generalize to other bitter compounds. In order to improve the validity of panel-collected data, the specific compounds of interest must be used to ensure the panelists ability to detect the ingredient in the product. By taking into account genetic differences in taste receptors, companies will be able to improve their testing methods leading to better tasting products.

Several studies have been conducted using in vitro methods to identify ligands for all 25 bitter taste receptors (see (Meyerhof, Batram et al. 2010) for a review). Many of these findings have yet to be confirmed in vivo. Additional studies are needed to address this gap in research.

Identifying functional polymorphisms within taste receptors that mediate activation of cells or perceived intensity will expand the knowledge of factors that influence perception and aid research scientists and product developers in developing ingredients and products.

Lastly, much of the data presented in this thesis is focused on perceptual differences in model systems. Following up these psychophysical studies with consumer oriented sensory studies using products that contain the compound would add additional insight to how taste receptor genetics ultimately affects intake.

45

Aims

1. To examine the effect of bitter taste receptor SNPs within TAS2R9 and TAS2R31 on the perceived bitterness of non-nutritive sweeteners.

2. To explore the relationship between the TAS2R38 haplotype and the self-reported liking for a variety of alcoholic beverages.

Hypotheses

1. Functional SNPs within bitter taste receptors will explain the variability in perceived bitterness from some non-nutritive sweeteners, but not others.

2. Individuals with functional TAS2R38 haplotypes will associate with lower liking ratings for non-sweet alcoholic beverages, with no relationship for sweet beverages.

46

Chapter 2

Bitterness of the non-nutritive sweetener Acesulfame Potassium varies with polymorphisms in TAS2R9 and TAS2R31

Adapted from:

Allen, A. L., McGeary, J.E., Knopik, V.S., Hayes, J. E. (2013). "Bitterness of the non-nutritive

sweetener Acesulfame Potassium varies with polymorphisms in TAS2R9 and TAS2R31."

Chemical Senses. In press.

Demand for non-nutritive sweeteners continues to increase due to their ability to provide desirable sweetness with minimal calories. Acesulfame Potassium (AceK) and saccharin are well- studied non-nutritive sweeteners commonly found in food products. Some individuals report aversive sensations from these sweeteners, such as bitter and metallic side tastes. Recent advances in molecular genetics have provided insight into the cause of perceptual differences across people. For example, common alleles for the genes TAS2R9 and TAS2R38 explain variable response to the bitter drugs ofloxacin in vitro and propylthiouracil in vivo. Here, we wanted to determine whether differences in the bitterness of AceK could be predicted by common polymorphisms (genetic variants) in bitter taste receptor genes (TAS2Rs). We genotyped participants (n=108) for putatively functional single nucleotide polymorphisms (SNPs) in 5

TAS2Rs and asked them to rate the bitterness of 25mM AceK on a general Labeled Magnitude

Scale (gLMS). Consistent with prior reports, we found two SNPs in TAS2R31 were associated with AceK bitterness. However, TAS2R9 alleles also predicted additional variation in AceK bitterness. Conversely, SNPs in TAS2R4, TAS2R38 and near TAS2R16 were not significant predictors. Using one SNP each from TAS2R9 and TAS2R31, we modeled the simultaneous

47 influence of these SNPs on AceK bitterness; together, these two SNPs explained 13.4% of the variance in perceived bitterness. These data suggest multiple polymorphisms within TAS2Rs contribute to the ability to perceive the bitterness from AceK.

Introduction

Taste is the number one driver of food choices (Glanz, Basil et al. 1998; IFIC 2011).

Sweetness is innately liked by humans (reviewed by (Steiner, Glaser et al. 2001)), even prior to birth (de Snoo 1937). Accordingly, many highly liked foods contain high endogenous amounts of natural sugars, or have sugars or other sweeteners added during processing. However, while consumers continue to desire sweetened products, there is also demand for reduced added-sugar in foods, due to associated health risks such as cardiovascular disease, diabetes and obesity (Hill

& Prentice 1995; Howard & Wylie-Rosett 2002). To retain desired levels of sweetness while reducing calories, bulk carbohydrates are often replaced with non-nutritive sweeteners. Replacing sugar with non-nutritive sweeteners may help with managing energy (De la Hunty, Gibson et al.

2006; Duffy & Anderson 1998; Raben, Vasilaras et al. 2002; Tordoff & Alleva 1990), although not all evidence supports this view (Anderson, Foreyt et al. 2012; Stellman & Garfinkel 1986).

These non-nutritive sweeteners can be natural (e.g. rebaudiosideA) or synthetic (e.g. saccharin), each with varying level of potency (DuBois, Walters et al. 1991; Hayes 2008).

Acesulfame potassium (AceK) was approved by the US Food and Drug Administration for use in dry foods in 1994; approval as a general-purpose sweetener followed in 2002.

According to Mintel market research data, AceK was the most used non-nutritive sweetener in new product launches between 2004 and 2010. However, in addition to eliciting sweet sensations, many non-nutritive sweeteners also have objectionable side tastes, such as bitterness, that are experienced by some individuals but not others (Kamerud & Delwiche 2007). This bitterness is

48 concentration-dependent (Horne, Lawless et al. 2002; Schiffman, Reilly et al. 1979) so one solution to reduce aversive bitterness is to use mixtures of these high potency sweeteners, either with each other, or with bulk carbohydrate sweeteners (Hanger, Lotz et al. 1996). This method has been used with some degree of success commercially (e.g., Coke Zero), although anecdotal reports suggest some individuals still find the taste of these blends objectionable. Thus, better understanding of the mechanisms underlying this variability may facilitate improved product formulation, with the potential to substantially impact health and wellness.

There are 25 bitter taste receptor genes (TAS2Rs) in humans (Adler, Hoon et al. 2000;

Chandrashekar, Mueller et al. 2000; Meyerhof, Batram et al. 2010), and these genes are highly polymorphic compared to the rest of the genome (Kim, Wooding et al. 2005). Beginning with the deorphanization of hT2R4 and hT2R38 a decade ago (Chandrashekar, Mueller et al. 2000; Kim,

Jorgenson et al. 2003), substantial progress has been made in identifying ligands for the majority of these receptors (Meyerhof, Batram et al. 2010). With regard to non-nutritive sweeteners, Kuhn and colleagues demonstrated receptors encoded by TAS2R31 (formerly called TAS2R44) and

TAS2R43 are activated by saccharin and AceK in vitro (Kuhn, Bufe et al. 2004). Moreover, perceived bitterness of the two sulfonyl amide sweeteners is greatly reduced by cross adaptation to aristolochic acid, a purely bitter hT2R31/hT2R43 agonist (Kuhn, Bufe et al. 2004). These data are largely consistent with earlier psychophysical data showing that the bitterness of AceK and saccharin covary with each other but not with propylthiouracil (Horne, Lawless et al. 2002). Like propylthiouracil (e.g. (Hayes, Bartoshuk et al. 2008)) and grapefruit juice (Hayes, Wallace et al.

2011), the bitterness of AceK and saccharin varies across individuals, and this variation has a genetic basis (Pronin, Xu et al. 2007; Roudnitzky, Bufe et al. 2011). Roudnitzky and colleagues

(Roudnitzky, Bufe et al. 2011) generated long range haplotypes across 5 highly related TAS2Rs on chromosome 12, and identified a single common haplotype (out of seven common haplotypes) that was associated with AceK and saccharin bitterness. By mutating single residues in artificial

49 chimeric receptors, they were able to demonstrate in vitro that the Arg35Trp (R35W) polymorphism in TAS2R31 was causal. However, they also found that even in the presence of the high functioning Trp35 allele, other TAS2R31 mutations abolished the ability of hT2R31 to respond to AceK and saccharin.

As outlined above, numerous studies have demonstrated that the perception of bitter taste in humans is moderated by genetic variation in TAS2R genes (Duffy, Davidson et al. 2004;

Hayes, Wallace et al. 2011; Pronin, Xu et al. 2007; Reed, Zhu et al. 2010; Roudnitzky, Bufe et al.

2011). However, many human behavioral studies have focused on single genetic variants, which neglects the fact that both bitter tastants and bitter receptors can be highly promiscuous

(Meyerhof, Batram et al. 2010). This approach is problematic because the examination of any single variant may be obscured by noise in other variants. That is, effects of one isolated allele can be overshadowed by an aggregate effect of several other alleles that are high or low functioning (see (Lotsch, Geisslinger et al. 2009) for a nontaste example). Here, we explore the influence of putatively functional polymorphisms in multiple TAS2Rs on AceK bitterness in individual SNP analyses before considering modeling these effects simultaneously.

Materials and methods

Overview

Data presented here are part of a larger, ongoing study of the genetics of oral sensation.

This study involves four laboratory sessions across days; here, we only describe the first day of testing. On Day 1, the study was explained to participants and consent was obtained. Participants then completed a food-liking questionnaire. Next, anthropometric data and salivary DNA samples were collected, followed by digital microscopy of the anterior tongue. Participants were oriented

50 to the psychophysical scale, and sampled six perceptually complex tastants and irritants, rating them for multiple qualities. Finally, participants completed a standard propylthiouracil (PROP) phenotyping protocol with PROP, salt and tones. After leaving the laboratory, participants completed several personality questionnaires via web-form. Total time in the laboratory for session 1 was ~1 hour; all data were collected one-on-one by project staff.

Participants

Prospective participants were prescreened to ensure that they were qualified. Eligibility criteria included: between 18-45 years old, not pregnant or breastfeeding, non-smoker (had not smoked in the last 30 days), no known defects of smell or taste, no lip, cheek or tongue piercings, no history of any condition involving chronic pain, not currently taking any prescription pain medication, no reported history of choking or difficulty swallowing and no history of thyroid disease. Participants also needed to be willing to provide a DNA sample via saliva. Written informed consent was obtained from all participants. All procedures were approved by the

Pennsylvania State University Institutional Review Board (protocol number #33176).

DNA samples were available from 147 participants. Race and ethnicity was self reported using categories provided by the 1997 OMB Directive 15. To minimize potential population stratification, which can potentially cause false negatives and false positives in gene association studies (Hamer 2000), individuals with Asian (n=18), African (n=5) or unknown (n=15) ancestry were excluded from the present analyses. Thus, we report data from 108 participants (34 men) of

European ancestry, with a mean age of 27.4 (± 8.1 SD) years. Results were not substantively different in the mixed ancestry sample, but we report only the results for the European-American participants to facilitate interpretation of the LD plots.

51

Psychophysical scaling

A general Labeled Magnitude Scale (gLMS) was used to collect perceived intensity of suprathreshold stimuli (Snyder and Fast 2004). This scale ranges from 0 (‘no sensation’) to 100

(’the strongest imaginable sensation of any kind’), with intermediate descriptors at 1.4 (‘barely detectable’), 6 (‘weak’), 17 (‘moderate’), 35 (‘strong’) and 51 (‘very strong’). All participants participated in an orientation to the scale, making ratings for a list of 15 imagined or remembered sensations that included both oral and non-oral items (Hayes, Allen et al. 2013). The orientation procedure and scale instructions were intended to promote use of the scale in a generalized context not limited to oral sensations. All psychophysical data were collected using Compusense five, version 5.2 (Guelph, Ontario, Canada).

Sampling and rating of perceptually complex stimuli

Six food grade stimuli were presented in 10mL aliquots: 0.56 M potassium chloride

(Spectrum) (salty/bitter), 0.41 mM quinine HCl (Sigma-Aldrich) (bitter), 25 mM AcesulfameK

(Spectrum) (sweet/bitter), 100 mM MSG (Ajinomoto) + 50mM IMP (Ajinomoto)

(umami/savory), 0.5 M sucrose (Domino) (sweet), and 25 uM capsaicin (Sigma-Aldrich)

(burning/stinging). Only AceK data and propthiouracil (see next section) are reported here; data for the other stimuli will be reported elsewhere. A pilot study and prior experience were used to determine appropriate concentrations. The concentrations were selected to produce a sensation near ‘moderate’ on a gLMS for the main quality of the stimulus.

After being told, "You may receive stimuli causing more than one quality. Please attend to all sensations on all trials", participants swished the 10mL sample for 3s and then expectorated prior to rating. Separate ratings were obtained for sweetness, bitterness, sourness,

52 burning/stinging, savory/umami, and saltiness on the gLMS. Presentation order was a counterbalanced Williams’ design. Participants rinsed with room temperature reverse osmosis

(RO) water prior to the first sample, and between each sample. A minimum interstimulus interval

(ISI) of 30s was enforced between samples.

Measuring PROP phenotype

PROP phenotype was determined using a standard concentration series with 6-n- propthiouracil (PROP), sodium chloride (salt), and sound, as described elsewhere (e.g. (Dinehart,

Hayes et al. 2006; Duffy, Peterson et al. 2004; Hayes, Sullivan et al. 2010)). Briefly, participants rated the intensity of 10 PROP solutions, 10 salt solutions, and 25 1-kHz tones. The stimuli were blocked so that participants received 5 tones, 5 salt solutions, 5 tones, 5 salt solutions, 5 tones, 5

PROP solutions, 5 tones, 5 PROP solutions and 5 tones. Block order was fixed; stimulus order within a block was counterbalanced. Half log steps were used for PROP solutions (3.2, 1, 0.32,

0.1, 0.032 mM) and the salt solutions (0.01, 0.032, 0.1, 0.32 and 1M). The 1-kHz tones were generated with a Maico MA39 audiometer calibrated to deliver the specified SPL binaurally; stimuli ranged from 50-90 dB in 10 dB steps. The tastants were prepared with USP grade 6-n- propylthiouracil (Sigma, St Louis MO) and kosher salt in RO water. Participants rinsed with room temperature RO water between each sample, waiting a minimum of 30s before next sample.

Overall intensity ratings for each stimulus were obtained with a gLMS. Mean intensity of the top

PROP concentration (3.2 mM) was used as a continuous variable.

53

Genetic Analysis

DNA was collected from saliva, using Oragene collection kits according to manufacturer instructions (Genotek Inc, Ontario, Canada). SNPs (single nucleotide polymorphisms) in TAS2R9,

TAS2R31, and TAS2R38 were determined using Sequenom MassARRAY technology (Sequenom,

San Diego, CA). All primers were purchased from Integrated DNA Technologies (Coralville,

Iowa, USA). Genotypes were assigned automatically via MassARRAY software (Sequenom) and independently inspected by two technicians. As a standard procedure, 15% of samples are rerun to ensure reliability.

For the Arg35Trp (rs10845295) SNP in TAS2R31, attempts were made to obtain custom assays using two different technologies (Sequenom MassARRAY and custom made to order

TaqMan assays); neither approach was successful despite repeated efforts. Thus, a tag SNP approach is used here for TAS2R31, as the only published method for determining the Arg35Trp

SNP, direct sequencing, is beyond the scope of the current project.

Statistical Analysis

Data were analyzed using SAS 9.2 (Cary, NC). Prior to analyses, the smallest possible rating on the gLMS (0.5) was added to all psychophysical ratings to eliminate zeros, and data were log-transformed. For analysis of individual SNPs, analysis of variance (ANOVA) was performed via proc mixed. Posthoc comparisons were made via the Tukey-Kramer method. For the significant SNPs, we then tested for deviations from a simple additive model using the two step approach recommended by (Carey 2007). Briefly, for each SNP, two variables – alpha and delta – were created. Alpha was coded as -1 for the homozygote with the lower phenotypic mean,

0 for the heterozygote, and +1 for the homozygote with the higher mean. Delta was coded as 1 for

54 heterozygotes and 0 for homozygotes. The first regression model, which only includes alpha, tests if the SNP predicts the phenotype in a simple additive model (i.e. 1 df). The second regression model includes both the alpha and delta terms, where a significant coefficient for delta indicates deviation from additivity; that is, delta specifically tests if the heterozygote mean differs from the expected point halfway between the means of the two homozygous groups. In this way, we can test for non-additivity without assuming either a dominant or recessive model. Finally, we used multiple regression to assess the independent influence of multiple putatively functional SNPs on the phenotype simultaneously. Because we did not observe evidence of non-additivity (see results), a simple count method was used to code each SNP; a participant was given an allele score of 0, 1, or 2 corresponding to the number of putatively high function alleles – based on the results of the single SNP ANOVAs – the individual possessed (e.g. a simple additive model for each SNP). The recoded SNP variables were then used to predict the perceived bitterness of

AceK via proc reg with a separate term for each SNP.

For TAS2R38, polymorphisms at amino acid residues 49 (Ala49Pro) and 262

(Val262Ala) are known to form two common haplotypes: the Proline–Alanine (PA_) haplotype is ancestral, bestowing the ability to sense thiourea (N-C=S) compounds, while the Alanine–Valine

(AV_) variant is less functional (Wooding, Kim et al. 2004). The program Haploview (Barrett,

Fry et al. 2005) was used to examine the extent of linkage disequilibrium (LD) between each pair of markers and to determine haplotype block structure. Haplotype blocks were defined according to Solid Spine of LD criteria (Barrett, Fry et al. 2005). Haplotype pairs were assigned to each participant of European ancestry using PHASE (Stephens & Donnelly 2003; Stephens, Smith et al. 2001). PHASE estimates the probabilities of all likely pairs of haplotypes (diplotypes) assigned to each individual from genotype data. Of these, diplotypes assigned with a probability of ≥ 0.80 were selected for further analysis. LD plots (Figure 2-2) show rounded R-squared values in individual squares, and shading is used to represent the exact R-squared value, with

55 darker shades of grey indicating larger R-squared values. All genotyping, construction and assignment of haplotypes were done blind to outcome variables.

Results

TAS2R9 alleles predict AceK bitterness

Dotson and colleagues (2008) reported hT2R9 responds to the bitter drugs ofloxacin, procainamide, and pirenzepine, and this response varies with a missense polymorphism in the

TAS2R9 gene (Val187Ala; rs3741845). Here, we see evidence that this allele is also functional for

AceK bitterness, as ANOVA revealed this SNP was significantly associated with the bitterness of

AceK [F(2,102) = 4.89; p = 0.009]. Group means of logged ratings are shown in Figure 2-1; the

Ala187 homozygotes (n=37) reported less bitterness than heterozygotes (n=55) (Tukey-Kramer p

= 0.011) and the Val187 homozygotes (n=13) (Tukey-Kramer p = 0.097). The heterozygotes and

Val187 homozygotes did not differ (Tukey-Kramer p = 0.987). In the additive regression model, this SNP explained 7.0% of the variance in logged AceK bitterness (p = 0.006); there was no evidence of dominance (p = 0.16).

56

Figure 2-1: Effect of the TAS2R9 Val187Ala polymorphism on the bitterness and sweetness of AceK and the bitterness of PROP. The bitterness of AceK was significantly different across genotype; no effect was observed for AceK bitterness or PROP bitterness (p-values provided in text). Adjectives refer to semantic labels on a general Labeled Magnitude Scale (gLMS; see text). BD refers to ‘barely detectable’.

In contrast, there was no evidence that the Val187Ala allele predicted AceK sweetness

[F(2,102) = 1.73; p = 0.18]. Also, as would be expected, the Val187Ala allele did not predict the bitterness of propylthiouracil [F(2,102) = 0.42; p = 0.66]. This SNP is not in linkage disequilibrium with any other SNPs on chromosome 12 (Figure 2-2; bottom).

57

Figure 2-2: LD Plot for TAS2R SNPs on chromosome 7 (top) and 12 (bottom). Numbers indicate rounded R-squared values and shading indicates exact R-squared values generated via Haploview.

58

TAS2R31 alleles predict variation in AceK bitterness

We explored the role of three TAS2R31 SNPs previously implicated in AceK bitterness by Roudnitzky and colleagues (Roudnitzky, Bufe et al. 2011). Below, we describe individual analyses for each SNP; however, these three SNPs are in strong linkage, as shown in Figure 2-2

(bottom). Prior evidence (Roudnitzky, Bufe et al. 2011) suggests these SNPs are not causal, but are in strong disequilibrium with the causal Trp35 allele. We were unable to directly measure the

Arg35Trp polymorphism, as attempts to obtain custom primers for either Sequenom or TaqMan methods were unsuccessful, necessitating a tag SNP approach here.

In ANOVA, the Val240Ile SNP (rs10772423) was significantly associated with the bitterness of AceK [F(2,100) = 4,80; p = 0.010). As shown in Figure 2-3, the Val240 homozygotes (n = 35) reported less bitterness from AceK than the Ile240 homozygotes (n=20)

(Tukey-Kramer p = 0.010). The Val/Ile heterozygotes (n = 48) fell in the middle, as the heterozygote group mean tended to be higher than the Val240 homozygotes (Tukey-Kramer p =

0.085), but did not differ from the Ile240 homozygotes (Tukey-Kramer p = 0.38). In the additive regression model, this SNP explained 8.7% of the variance in logged AceK bitterness (p = 0.003); there was no evidence of dominance (p = 0.75). Again, we failed to find evidence this SNP predicted AceK sweetness [F(2,100) = 0.1.43; p = 0.24] or PROP bitterness [F(2,100) = 0.77; p =

0.47].

59

Figure 2-3: Same as Figure 2-1, but for the TAS2R31 Val240Ile allele. The bitterness of AceK was significantly different across genotype; no effect was observed for AceK bitterness or PROP bitterness (p-values provided in text).

We also found a similar effect pattern for the Ala227Val SNP (rs10845293) [F(2,98)=

3.55; p = 0.032]. The group means of the logged bitterness ratings for the Val227 homozygotes (n

= 33), Ala/Val heterozygotes (n=49), and Ala227 homozygotes (n= 19) are shown in

Supplemental Figure A-1. Because this SNP was in linkage with the Val240Ile SNP, we did not test for additivity or dominance. We found no evidence that the Ala227Val SNP was associated with AceK sweetness [F(2,98) = 0.15; p = 0.86] or the perceived bitterness ratings of PROP

[F(2,98) = 1.27; p = 0.29].

Finally, we also explored the influence of the Gln217Glu SNP (rs10845294) on AceK bitterness. We did not have a sufficiently large cohort to detect a significant effect [F(2,102) =

1.28; p = 0.28] given the low number of Glu217 homozygotes (n=6) compared to heterozygotes

(n=39), and Gln217 homozygotes (n=60). Previously, Roudnitzky et al (Roudnitzky, Bufe et al.

60

2011) showed that having either the Glu217 or Gln217 allele SNP did not change the activation of hT2R31 when the Trp35 allele was present. The sweetness of AceK did not significantly differ across Gln217Glu [F(2,102) = 1.41; p = 0.25], nor was this SNP associated with the reported bitterness ratings of PROP [F(2,102) = 0.08; p = 0.93].

Collectively, these results confirm that TAS2R31 contains a functional polymorphism that predicts at least some of the variation in the bitterness of AceK. Present data also suggest that hT2R31 mediates the bitterness of AceK but not propylthiouracil. Additionally, the TAS2R31 alleles do not appear to influence sweetness, at least at the concentration tested here.

TAS2R38 diplotypes predict variation in the bitterness of PROP but not AceK

Two well-characterized SNPs in TAS2R38 on chromosome 7 were measured here:

Ala49Pro (rs713598) and Val262Ala (rs1726866). These SNPs are well known to exhibit strong linkage disequilibrium and there is evidence of two common haplotypes (Wooding et al., 2004).

This was verified in our data via Haploview. Using PHASE, the following common diplotypes

(probability ≥ 0.80) were assigned in our data, AV_ homozygotes, AV_/PA_ heterozygotes, and

PA_ homozygotes, and tested whether they predicted AceK bitterness. Individuals with rare diplotypes (n=11) were excluded a priori from the analysis as these rare haplotypes are known to have intermediate phenotypes that differ both from each other and the common haplotypes (Bufe,

Breslin et al. 2005; Hayes, Bartoshuk et al. 2008; Mennella, Pepino et al. 2011), so binning them together in one group is not justified. Here, these included AA/PA (4), AA/AV (3), PV/AV (2),

PV/PA (1), and PV/PV (1) individuals. We did not observe any evidence of a relationship between common TAS2R38 diplotypes and the bitterness of AceK [F(2,94) = 1.28; p = 0.28] or sweetness of AceK [F(2,94) = 0.49; p = 0.62]. As expected, the bitterness of PROP differed with

TAS2R38 diplotype [F(2,82 )= 18.08; p <0.0001]. The AV_ homozygotes (n=19) reported

61 significantly less bitterness than the heterozygotes (n=54; Tukey-Kramer p <0.001) or PA_ homozygotes (n=12) (Tukey-Kramer p <0.001); mean bitterness for the heterozygous individuals was similar to the PA_homozygotes (Tukey-Kramer p =0.38), consistent with numerous other reports (Bufe, Breslin et al. 2005; Calo, Padiglia et al. 2011; Duffy, Davidson et al. 2004; Hayes,

Bartoshuk et al. 2008; Mennella, Pepino et al. 2010).

Figure 2-4: Same as Figure 2-1, but for the AV/PA TAS2R38 diplotype (Ala49Pro and Val262Ala). PROP bitterness differed by diplotype; differences in AceK bitterness or sweetness across diplotype were not significant (p-values provided in text).

62

Relationship between PROP bitterness and AceK bitterness and sweetness

Prior reports conflict as to whether the bitterness of sulfonyl amide sweeteners is related to the bitterness of PROP (Bartoshuk 1979; Horne, Lawless et al. 2002). Here, we found minimal evidence that PROP bitterness was predictive of AceK bitterness (R-sq = 2.9%; p =0.077).

Conversely, the bitterness of PROP was positively associated with AceK sweetness, predicting

10.2% of the variation (p < 0.001).

AceK bitterness was not predicted by putatively functional SNPs in TAS2R4 or near TAS2R16

We also tested a putatively functional SNP in TAS2R4 and a putatively functional SNP

9.4 kilobases downstream from TAS2R16 (Hayes et al. 2011). These two SNPs did not exhibit linkage disequilibrium with each other, or with the TAS2R38 haplotype (Figure 1, top). There was little to no evidence that the bitterness of AceK varied with rs2234001 in TAS2R4 [F(2,97)= 0.69; p = 0.50]. There was no evidence that the rs1308724 SNP near TAS2R16 was a significant predictor of AceK bitterness [F(2,98)= 0.16; p = 0.85].

Effect of multiple loci on AceK bitterness

While the 25 TAS2Rs are highly polymorphic, empirical evidence that these polymorphisms are functional has only been shown for five receptor genes (TAS2R4, TAS2R16,

TAS2R19, TAS2R31 and TAS2R38). Above, we tested one or more candidate SNPs in four of these genes in regard to the bitterness of AceK. Based on our findings for the individual genes above, we used a simple regression model to assess the simultaneous influence of polymorphisms in two genes – TAS2R9 (Val187Ala) and TAS2R31 (Val240Ile) – on AceK bitterness. Because the

63 effects of these two SNPs did not appear to deviate from additivity, a score of one was assigned for each putatively functional allele for each of the SNPs, resulting in two variables coded 0,1, and 2. AceK bitterness was regressed against these two recoded variables. This model explained

13.4% of the variance in bitterness (p < 0.001), and the recoded variables for Val187Ala (p

=0.021) and Val240Ile (p =0.008) were both significant.

Discussion

Current findings support the idea that not all humans perceive bitterness from AceK, which is congruent with earlier findings. Using polymorphisms found in TAS2Rs, which were previously shown to be functional for AceK and other non-nutritive sweeteners (i.e., saccharin) in vitro (Kuhn, Bufe et al. 2004; Roudnitzky, Bufe et al. 2011) and in vivo (Pronin, Xu et al. 2007), our results suggest multiple bitter taste receptors on two different chromosomes contribute to the perceived bitterness of AceK in humans. The use of multiple loci simultaneously increased our ability to explain variance in the quantitative trait measured here (AceK bitterness).

The TAS2R38 haplotype has repeatedly been shown to associate with PROP bitterness

(Bufe, Breslin et al. 2005; Duffy, Davidson et al. 2004; Duffy, Hayes et al. 2010), which we confirm here. The first SNP in this haplotype, Ala49Pro is known to be in strong linkage disequilibrium with the other two TAS2R38 SNPs, Val262Ala and Val296Ile (not measured here), such that more than 95% of individuals carry either the PAV or AVI variant (Hayes, Bartoshuk et al. 2008; Kim, Jorgenson et al. 2003). In in vitro heterologous expression systems, site directed mutation indicates that the amino acid at site 49 is the primary determinant of PROP response, with an additional influence arising from the residue at site 262; the residue at site 296 does not seem to matter, at least in cultured cells (Bufe, Breslin et al. 2005). Large-scale molecular psychophysics in vivo (Mennella, Pepino et al. 2011) partially confirm these data. Specifically, by

64 comparing humans with rare haplotypes behaviorally, Mennella and colleagues were able to tease apart the independent contributions of each site. Among those with similar diplotypes at sites 49 and 262, the 296 position also contributed to perceived bitterness, as Val296 carriers were more sensitive to PROP than Ile296. This suggests all three sites Pro49Ala, Ala262Val and Val296Ile contribute to PROP response in vivo. These data also reinforce the need to confirm data from in vitro heterologous expression systems with behavior in the whole organism (i.e. animal or human psychophysics).

Here, a two site diplotype approach (PA_ v. AV_) did not predict AceK bitterness, consistent with data showing that AceK doesn’t activate hT2R38 in heterologous expression systems (Meyerhof, Batram et al. 2010). However, psychophysical data in humans does not always agree with in vitro expression studies cells (compare (Bufe, Breslin et al. 2005) and

(Mennella, Pepino et al. 2011)). Likewise, we have reported the bitterness of grapefruit juice varies as a function of a polymorphism in TAS2R19 (Hayes, Wallace et al. 2011), while neither limonin nor naringin activate hT2R19 in functional expression systems (Meyerhof, Batram et al.

2010). Thus, it remains important to confirm negative in vitro findings from AceK psychophysically in humans.

Here, we found that PROP bitterness was positively associated with the sweetness of

AceK, but not the bitterness. PROP bitterness is a marker for overall heightened taste response across qualities (Hayes & Keast 2011). For sweetness, this is consistent with prior data showing

PROP bitterness is positively associated with the sweetness of sucrose (Hayes, Bartoshuk et al.

2008), aspartame (Dinehart, Hayes et al. 2006), AceK (Horne, Lawless et al. 2002) and sweet foods (Lanier, Hayes et al. 2005). In regard to bitterness from sulfonyl amide sweeteners, prior reports conflict. When dichotomizing individuals using threshold methods (ie tasters versus non- tasters), PROP nontasters report less suprathreshold bitterness from saccharin (Bartoshuk 1979).

Lawless and his students (Horne, Lawless et al. 2002) failed to observe a relationship between

65 saccharin and AceK bitterness and the bitterness of PROP in two experiments with relatively small sample sizes (n=30 and n=38). Here, we confirm that PROP bitterness predicts AceK sweetness but not bitterness in a large sample, suggesting the failure to find an association between the bitterness of AceK and PROP is not simply a matter of power.

The relationship between PROP bitterness and AceK sweetness, but not AceK bitterness speaks directly to the dual nature of PROP bitterness as a marker of taste function. That is, PROP bitterness confounds two separate but distinct sources of variation: TAS2R38 polymorphisms, and overall taste response (aka hypergeusia) (Hayes & Keast 2011). Previously, we showed

PROP bitterness predicts variation in the intensity of non-bitter tastants, even after controlling for

TAS2R38 genotype (Hayes, Bartoshuk et al. 2008), supporting its continued use as a marker of overall orosensory response. Moreover, molecular and behavioral data indicate TAS2R38 variation clearly predicts differential response to compounds that contain the thiourea (N-C=S) moiety (Bufe, Breslin et al. 2005; Meyerhof, Batram et al. 2010). Given this duality, it is not unreasonable that PROP bitterness should predict AceK sweetness but not bitterness in studies with small numbers of participants. That is, for AceK sweetness, PROP bitterness is capturing overall heightened taste response (Hayes & Keast 2011), as does for other sweet items (Lanier,

Hayes et al. 2005). Conversely, for AceK bitterness, the perceived intensity varies not only as a function of overall taste response, but also as a function of variation in TAS2R31 and possibly other TAS2R genes. Thus, effect size estimates for PROP should be lower or absent for bitterness than sweetness, especially if random, unmeasured, variation in TAS2R31 or TAS2R9 obfuscates weak effects in small to medium studies.

Present data support prior evidence that the Val187Ala SNP in TAS2R9 is functional

(Dotson, Zhang et al. 2008). Of 64 bitter stimuli tested in vitro by Dotson and colleagues (Dotson,

Zhang et al. 2008), only three synthetic pharmaceuticals – ofloxacin, pirenzapine, and procanimide – activated hT2R9. They reported that when Ala was replaced with Val at position

66

187, the receptor was no longer activated over a wide range of concentrations. They also note the rs3741845 SNP results in amino acid change in a region thought to alter the binding pocket of hT2R receptors (Dotson, Zhang et al. 2008). Here, we provide the first evidence that this polymorphism may contribute to the perceived bitterness of AceK, although this finding needs to be confirmed. Previously, Dotson and colleagues reported that saccharin did not activate hT2R9 in vitro; however, the top concentration used in their study was ~100 times lower than the amount required to elicit hT2R31 response in vitro (Pronin, Xu et al. 2007). This suggests our data does not directly contradict Doston et al with regard to hT2R9 and sulfonyl amide sweeteners, as their null finding may simply be a matter of dose. Notably, in their systematic screening efforts of 104 compounds in vitro, Meyerhof and colleagues were unable to identify any potential hT2R9 ligands (Meyerhof, Batram et al. 2010). Their screening battery did not include the pharmaceutical agents identified by Dotson et al., and even if they had, Meyerhof et al. used the

Val187 variant in their heterologous expression system, which would not be expected to detect the ligands tested by Dotson et al. However, we would have expected Meyerhof’s team to identify AceK as a potential hT2R9 ligand. However, their team also observed poor expression of hT2R9 receptors on the surface of their cells, and none of 104 compounds in their test battery activated hT2R9. Thus, it is currently unknown whether AceK is able to activate the Val187 hT2R9 variant in vitro at biologically relevant doses.

Paradoxically, the gain of function allele here (Val187) is the loss of function allele in the

Dotson et al report (Dotson, Zhang et al. 2008). Previously, it has been hypothesized that mutations in TAS2Rs may drive a gain of function for an alternative ligand (e.g. (Wooding, Kim et al. 2004)). Although there is no definitive example of this to date, it has long been known that

PTC tasters are nonresponsive to the bitterness of Antidesma bunius berries, while PTC nontasters report bitterness from these berries (Henkin & Gillis 1977). Recently, this finding was confirmed for the TAS2R38 genotype and Antidesma bunius berries, although the specific ligand itself was

67 not isolated (Danielle Reed; personal communication). Present results need to be confirmed in vitro, but if replicated, this would represent the first demonstrated case of a dual functional SNP that broadens the molecular receptive range of a T2R by enabling the detection an alternative ligand.

Here, we confirm that polymorphisms in the TAS2R31 gene explain variation in AceK perception in humans. Previous work indicated recognition thresholds differ across individuals

(Pronin, Xu et al. 2007; Roudnitzky, Bufe et al. 2011); here, we extend these findings to include suprathreshold intensity. The distinction between threshold and suprathreshold psychophysics is critical, as classical thresholds frequently fail to predict affective response and ingestive behavior

(Duffy, Peterson et al. 2004; Harwood, Ziegler et al. 2012; Lucas, Riddell et al. 2011). Our attempts to directly genotype individuals for the Arg35Trp (R35W) (rs10845295) were unsuccessful for technical reasons. Instead, we measured several other SNPs in TAS2R31. On the basis of single point mutation studies, Roudnitzky and colleagues found that when Ile240 was changed to Val240 in vitro, this substitution had no effect on the activation of hT2R31 by AceK.

However, in the haplotypes observed in the population, the Ile240 allele was in strong linkage disequilibrium with Arg35 (Roudnitzky, Bufe et al. 2011). Likewise, Val227 was in disequilibrium with Arg35. Thus, the SNPs measured here – Val240Ile and Ala227Val – can reasonably be used as tag SNPs for the causal SNP at residue 35.

The present study also extends prior work by looking at TAS2Rs located on separate chromosomes. This suggests that the putatively functional SNPs identified here make an independent contribution to the bitterness of AceK. That is, because they occur on distinct chromosomes, present results cannot be the result of a single haplotype across highly conserved genes. Thus, multilocus approaches like the one used here may have substantial utility when linking TAS2Rs to ingestive behavior, given the likelihood for functional recovery that may otherwise overwhelm the effects on individual SNP analyses.

68

Here, we report that our multilocus model predicts 13.4% of the variation observed in perceived bitterness of AceK. With the inclusion of two SNPs from different receptors, it would seem that other receptors, or polymorphisms not tested here may contribute to the inability to perceive bitterness from AceK in some individuals. Approximately half of our participants rated

AceK bitterness at zero. It is possible that offering participants other response options in addition to bitterness, such as “metallic” or “other”, may have captured additional off-tastes typically associated with AceK. However, we find this interpretation unsatisfying, as we would also expect untrained participants to dump any aversive, unpleasant sensations into the bitter response option

(Clark & Lawless 1994). Previous reports support that some proportion of individuals report little or no bitterness from AceK, but they do not report the exact proportion (Horne, Lawless et al.

2002).There is still a large amount of unexplained variation in our data. This could be due to other factors like fungiform papillae density (Zuniga, Davis et al. 1993), central gain (Green &

Hayes 2004) or other unmeasured polymorphisms in TAS2R genes. In particular, it seems possible that rare TAS2R31 variants may reduce response in vivo. Indeed, in vitro evidence suggests that even when the Trp35 allele is present, mutations at amino acid residues 45, 237, 276 and 281 all cause a loss of function (Roudnitzky, Bufe et al. 2011)In the future, whole gene sequencing of TAS2Rs may be required to fully explain phenotypic variation. Also, as with any candidate SNP study, we should note that unmeasured third variables can obscure SNP findings

(e.g., population stratification), and that the associations reported here may not be causal, arising instead from linkage disequilibrium with other unmeasured polymorphisms.

Conclusions

Using suprathreshold psychophysics in humans, we predicted variation in the perceived intensity of AceK bitterness using a candidate SNP approach across multiple TAS2R genes. These

69 data suggest more than one receptor is responsible for the perception of AceK bitterness. Tag

SNPs believed to be in complete linkage disequilibrium with the putatively causal SNP in

TAS2R31 predicted variation in AceK bitterness. In addition, a polymorphism Ala187Val in

TAS2R9 not been previously reported as being functional for AceK was shown to predict bitterness in vivo. Conversely, putatively functional SNPs in TAS2R4, TAS2R38 and near

TAS2R16 did not predict bitterness. Polymorphisms in two bitter receptor genes on different chromosomes both appeared to contribute to the suprathreshold bitterness of AceK, and a simple multilocus model was able to predict 13% of the variance in perceived bitterness. However, present data also suggest additional polymorphisms may contribute to the perception of AceK bitterness. More research is needed to determine if other receptors confer additional response, and whether rare polymorphisms in the genes studied here may attenuate responses to AceK.

70

Chapter 3

Rebaudioside A and Rebaudioside D bitterness do not covary with Acesulfame K or polymoprhisms in TAS2R9 and TAS2R31

Adapted from:

Allen, A. L., McGeary, J.E., Hayes, J. E. (2013). " Rebaudioside A and Rebaudioside D bitterness

do not covary with Acesulfame K or polymoprhisms in TAS2R9 and TAS2R31”. Under Review.

In order to reduce calories in foods and beverages, the food industry routinely uses non- nutritive sweeteners. Unfortunately, many are synthetically derived, and many consumers have a strong preference for natural sweeteners, irrespective of the safety data on synthetic non-nutritive sweeteners. Additionally, many non-nutritive sweeteners elicit aversive side tastes such as bitter and metallic in addition to sweetness. Bitterness thresholds of acesulfame-K (AceK) and saccharin are known to vary across bitter taste receptors polymorphisms in TAS2R31. Here we examined bitterness and sweetness perception of natural and synthetic non-nutritive sweeteners.

In a follow-up to a previous study, participants (n=122) who were previously genotyped rated sweet, bitter and metallic sensations from rebaudioside A (RebA), rebaudioside D (RebD), aspartame, sucrose and gentiobiose in duplicate in a single session. For comparison, we also present sweet and bitter ratings of AceK collected in the original experiment for the same participants. At similar sweetness levels, aspartame elicited less bitterness than RebD, which was significantly less bitter than RebA. The bitterness of RebA and RebD showed wide variability across individuals, and bitterness ratings for these compounds were correlated. However, RebA

71 and RebD bitterness did not covary with AceK bitterness. Likewise, SNPs shown previously to explain variation in the suprathreshold bitterness of AceK (rs3741845 in TAS2R9 and rs10772423 in TAS2R31) did not explain variation in RebA and RebD bitterness. Collectively, these data indicate the bitterness of RebA and RebD cannot be predicted by AceK bitterness, reinforcing our view that bitterness is not a simple monolithic trait that is high or low in an individual. This also implies consumers who reject AceK may not find RebA and RebD aversive, and vice versa.

Finally, RebD may be a superior natural non-nutritive sweetener to RebA, as it elicits significantly less bitterness at similar levels of sweetness.

Introduction

Rebaudioside A and Rebaudioside D are glycosides extracted from Stevia rebaudiana

(Bertoni) Bertoni, a perennial shrub native to Paraguay and Brazil. The Guaraní Paraguayan

Indians and Mestizos have used Stevia leaves for centuries to sweeten teas (Bertoni, 1905), beer and tobacco (Melville, 1941) (see (Lewis 1992) for a review). The sweetness is elicited by diterpenic ent-kaurene glycosides, each exhibiting a steviol aglycone (13-hydroxykaur-16-en-18- oic acid). The Joint FAO/WHO Expert Committee on Food Additives reported that 9 glycosides are found in Stevia leaves (JECFA 2010); however, since publication of this report, an additional

26 glycosides have been identified (Ohta, Sasa et al. 2010) (see (Wölwer-Rieck 2012) for a comprehensive list). These glycosides differ greatly in concentration in the leaf and are also influenced by the plants’ genotype and growing conditions. Typically, stevioside and rebaudioside A (RebA) are the two most abundant glycosides (Ohta, Sasa et al. 2010). However, whole Stevia leaves are not approved for use in foods. Currently, the only extracts from Stevia approved for use in food in the United States are purified stevioside and RebA; stevioside is approved for use as a dietary supplement while RebA has GRAS status as an approved ingredient.

72

These glycosides differ in their sensory properties, as shown by (DuBois, Walters et al.

1991; Hellfritsch, Brockhoff et al. 2012; Phillips 1989; Schiffman, Booth et al. 1995; Schiffman,

Booth et al. 1995). Although Stevia leaves and its glycoside extracts are known for their sweetness, many also elicit undesirable side-tastes, including bitterness (DuBois, Walters et al.

1991; Hellfritsch, Brockhoff et al. 2012; Schiffman, Booth et al. 1995).

The ability to detect bitterness from various bitter compounds differs greatly across individuals. Much of the variation can be explained by polymorphisms in bitter taste receptors

(Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004; Duffy, Hayes et al. 2010; Feeney

2011; Hayes, Wallace et al. 2011; Pronin, Xu et al. 2007; Roudnitzky, Bufe et al. 2011; Wang,

Hinrichs et al. 2007). Additional variance can be attributed to age (Mennella, Pepino et al. 2010), and fungiform papillae density (Bartoshuk, Duffy et al. 1994; Duffy & Bartoshuk 2000) as well as other factors (Tepper, Keller et al. 2003). Specific to non-nutritive sweeteners, variation in bitter taste receptor genes can explain differences in the bitterness arising from these sweeteners.

Kuhn and colleagues (Kuhn, Bufe et al. 2004) showed saccharin and AceK activate bitter taste receptors hT2R31 and hT2R43 (encoded by the TAS2R31 and TAS2R43 genes near chromosome 12p13.2) in vitro. Subsequently, Pronin and colleagues found polymorphisms within

TAS2R31, specifically Arg35Trp influenced bitterness perception of sampled saccharin in 55 subjects (Pronin, Xu et al. 2007). Presumably, these SNPs might also influence AceK bitterness, as earlier psychophysical evidence suggested saccharin and AceK bitterness covaries across individuals (Horne, Lawless et al. 2002). Roudnitzky (Roudnitzky, Bufe et al. 2011) confirmed this, showing that the perceived bitterness of saccharin and AceK is associated with several polymorphisms in TAS2R31, with the greatest variation explained by the Arg35Trp allele.

Individuals who expressed Trp at position 35 perceived greater bitterness than those with an Arg at this residue. This finding was confirmed in vitro, where taste receptor cells expressing Arg35 show little to no activation when exposed to saccharin or AceK (Roudnitzky, Bufe et al. 2011).

73

They also identified eight additional SNPs that further mediate the activation of TAS2R31 in varying amounts (D45H, M162L, Q127E, A226V, L237F, V240I, P276R and W281C)

(Roudnitzky, Bufe et al. 2011).

Elsewhere, Dotson et al. (Dotson, Zhang et al. 2008) reported that the Ala187Val polymorphism in TAS2R9 chromosome (12p13) influenced the activation of cells exposed to three pharmaceuticals commonly described as bitter: ofloxacin, procainamide and pirenzapine.

The Val187 allele was associated with limited to no activation, compared to the Ala187 allele.

Forthcoming data from our laboratory indicates this SNP significantly predicts the bitterness from

AceK (Allen et al. 2013). Unexpectedly, the pattern observed is the opposite of that seen in vitro by Dotson et al. for different stimui; our data suggest the Val197 allele is the high function allele, at least with respect to AceK (Allen et al. 2013).

Few studies have compared bitterness and sweetness across multiple non-nutritive sweeteners (Kamerud & Delwiche 2007). Here, we report the perceived sweetness and bitterness intensity of two natural non-nutritive sweeteners (RebA and RebD), as well as two commonly used synthetic non-nutritive sweeteners (AceK and Aspartame).

Multiple studies indicate polymorphisms in bitter taste receptor genes (TAS2Rs) influence the perception of bitterness from synthetic sulfonyl amide sweeteners like saccharin and

AceK. Likewise, anecdotal reports suggest the bitterness from natural Stevia derived sweeteners is not experienced universally. However, it remains unknown whether a) the bitterness of natural

Stevia derived sweeteners, specifically RebA and RebD, covary with the bitterness of sulfonyl amide sweeteners across individuals, and b) whether variable bitterness in RebA and RebD might be explained by genetic polymorphisms in TAS2R genes. Here, we compare the sensory profiles of Stevia derived glycosides, describe the variation in the bitterness of these stimuli across individuals, and explore whether this variability can be predicted by polymorphisms in genes

74 previously implicated in the bitterness of other non-nutritive sweeteners (i.e. TAS2R9, and

TAS2R31).

Materials and Methods

Overview

Here, we present data from a follow-up study to a larger project on the genetics of oral sensation (Project GIANT-CS; (Byrnes & Hayes 2013)). Participants who had completed the main study and had been genotyped were invited to return to our laboratory to taste a variety of tastants not originally included in the main study, including multiple non-nutritive sweeteners and disaccharides. 122 participants returned to complete the follow-up study. The study was conducted in the Sensory Evaluation Center at the Pennsylvania State University in individual testing test booths under white light. After participants were re-consented in writing for the follow-up study, they participated in a 3-minute training session in a multifunction space in our facility prior to entering the test booths. All procedures for the main project and follow-up study were approved by the Pennsylvania State University Institutional Review Board (protocol numbers #33176 and #40921). In isolated testing booths, participants rated bitter, sweet, and metallic sensations on a computerized general Labeled Magnitude Scale (gLMS).

75

Participants

Participants were screened prior to the start of the main study. Eligibility criteria included: between 18-45 years old, not pregnant or breastfeeding, non-smoker (had not smoked in the last 30 days), no known defects of smell or taste, no lip, cheek or tongue piercings, no history of any condition involving chronic pain, not currently taking any prescription pain medication, no reported history of choking or difficulty swallowing and no history of thyroid disease.

Here, we report data from one hundred and twenty-two participants (44 men), with a mean age 27.7 (±7.89) years. Self-reported race and ethnicity was collected based on criteria provided by the 1997 OMB Directive 15. This population is largely of European ancestry (n=88), with marginal representation from other ethnicities, African (n=2) and Asian (n=21), with 9 individuals selecting to not disclose their ethnicity.

Training Session

Participants were asked to rate sweetness, bitterness and metallic on the gLMS. Due to potential confusion between bitter and metallic, a brief training session was implemented to provide individuals examples of these three taste qualities. After consent was obtained, participants went through a short 2 to 3 minute training session from project staff individually or in small groups (no more than 3 to a group). Four 10 mL samples were presented labeled samples as exemplars of specific taste qualities: ‘sweet’ = 250mM sucrose (Domino), ‘bitter’= 0.05mM quinine (SAFC), ‘bitter and sweet’= 263mM sucrose and 0.5mM quinine and ‘metallic’ = 1.0mM

FeSO4 (J.T. Baker). The concentrations of the training samples were chosen to be similar to the test samples, eliciting sweet, bitter and mixture and metallic (all above threshold values). The presentation order of the training samples was fixed (sweet, bitter, sweet and bitter, and metallic).

76

At the start of the training session, participants were given a spit cup and a water cup. Participants were given the following instructions: “This training session is to help familiarize you with the taste qualities you will be asked to rate in this study. The three qualities are sweetness, bitterness and metallic”. The researcher then provided instructions on how to taste the samples: “Put the whole sample in your mouth, swish for 3 seconds and spit it out. Then rinse with water”. The samples were presented in a clear plastic medicine cup with the respective labels (sweet, bitter, sweet and bitter, and metallic). The researcher presented each sample one at a time and participants were told the taste quality(s). After tasting, participants were told, that if they experience this taste they would rate the perceived intensity on the respective scale. A mixture of sucrose and quinine was provided to give an example of a sample that may have multiple taste qualities (sweet and bitter). Prior to entering the testing booths, participants were asked to rinse with water to remove any lingering sensations.

Psychophysical Scaling of Test Stimuli

Participants reported perceived intensity of the test stimuli by rating on general Labeled

Magnitude Scale (gLMS). The gLMS anchors are 0 (‘no sensation’) to 100 (’the strongest imaginable sensation of any kind’), with descriptors at 1.4 (‘barely detectable’), 6 (‘weak’), 17

(‘moderate’), 35 (‘strong’) and 51 (‘very strong’). All participants had been trained to use the gLMS in the main study, and they were reoriented to the test before tasting test samples in the following study, this was done by asking participants to rate 6 imagined or remembered sensations on the gLMS. The orientation included both oral and non-oral items to promote ratings to be in context of all items, not just taste (Hayes, Allen et al. 2013). Data were collected using

Compusense five, version 5.2 (Guelph, Ontario, Canada).

77

Test Stimuli

Five food grade test samples were presented: 219mM sucrose (Domino), 20mM gentiobiose (Biosynth), 6.8mM aspartame (Spectrum), 1.65mM rebaudioside A (Enliten from

Corn Products International), 1mM rebaudioside D (kindly donated by J. Delwiche at PepsiCo).

Concentrations of stimuli were selected to be above bitterness threshold in pilot testing. All were prepared using reverse osmosis (RO) water. Test stimuli (10mL) were presented at room temperature under white lighting. Stimuli were presented in duplicate in clear plastic medicine cups labeled with a three-digit blinding code. Presentation order was counterbalanced in a blocked Williams Design, with each stimulus being presented once before a duplicate was presented. Participants had a 30 second break enforced between samples, and they rinsed with RO water as needed.

During the main study (Project GIANT-CS), participants sampled (in the same manner as above) a 25mM concentration of Acesulfame K (Spectrum), among other stimuli (see (Byrnes &

Hayes 2013)). Participants rated sweetness, bitterness, burning/stinging, sourness, savory/umami, and saltiness of the sample on a separate gLMS for each quality. AceK ratings from the participants (n=122) who completed the follow-up study were reanalyzed here to facilitate direct comparisons across stimuli within a single group of individuals. (Because of participant overlap, these data should not be taken as independent replication of our prior report (Allen et al. 2013).

Genetic Analysis

As reported previously (Allen et al. 2013) DNA was collected using Oragene saliva collection kits according to manufacturer instructions (Genotek Inc, Ontario, Canada). Genotypes for selected SNPs (single nucleotide polymorphisms) in TAS2R9 and TAS2R31 on chromosome 7

78 and TAS2R38 on chromosome 12 were determined using Sequenom MassARRAY technology

(Sequenom, San Diego, CA). Primers were acquired from Integrated DNA Technologies

(Coralville, Iowa, USA). Genotypes were assigned automatically via MassARRAY software

(Sequenom) and inspected by two technicians. 15% of samples are randomly selected and subjected to a secondary analysis to ensure accuracy.

A tag SNP approach was used due our inability to obtain functional primers for the

Arg35Trp (rs10845295) SNP in TAS2R31. Repeated attempts were made on multiple platforms

(Sequenom MassARRAY and made to order TaqMan assays) and these attempts all failed vendor quality control standards, requiring us to use rs10772423 as a tag SNP instead.

Statistical Analysis

Analyses were performed using SAS 9.2 (Cary, NC). SNP analyses were performed individually using analysis of variance (ANOVA) via proc mixed. Posthoc comparisons were made via the Tukey-Kramer method.

Results

Comparing sweetness and bitterness ratings of sampled non-nutritive sweeteners

Many non-nutritive sweeteners elicit both sweetness and bitterness (DuBois, Walters et al. 1991; Hellfritsch, Brockhoff et al. 2012; Kamerud & Delwiche 2007; Schiffman, Booth et al.

1995). Of these reports, many compare the sweetness and bitterness to a fixed concentration of reference stimuli. However, use of a fixed modulus (reference stimulus) may minimize real

79 individual differences across people; thus we used a generalized scale (i.e. a gLMS) here to better reflect differences across individuals, similar to work by Kamerud and Delwiche (Kamerud &

Delwiche 2007). Here, we measure the variation in perceived intensity for the sweetness and bitterness of RebA and RebD at suprathreshold concentrations, using a gLMS (without using reference samples) (Figure 3-1). By plotting sweetness of RebA and RebD simultaneously, we observed substantial variation in the perceptual sweetness intensity (Figure 3-2 (top)). Reported sweetness of RebA ranges from 0 to 41.25 and RebD sweetness ranges from 1 to 75. Figure 3-2

(bottom) shows variation bitterness ratings of RebA and RebD, with bitterness of RebA ranging from 1.5 to 77.25 while RebD ranges from 0 to 20 on a gLMS. Here, we document the wide range of individual variation in RebA and RebD sweetness and bitterness in a large genetically informed cohort.

80

Mean Sweetness and Bitterness Ratings Strong Sweetness Bitterness Moderate

10

Weak

IntensityRatings (gLMS) BD 1

AceK RebA RebD Aspertame Sucrose Gentiobiose Figure 3-1: Reported bitterness ratings (± Stder) on the gLMS for AceK, RebA, RebD, Asp, sucrose and gentiobiose. AceK ratings were reported on different days. BD, refers to barely detectable, 1.4 on the gLMS, followed by weak (6), moderate (17) and strong (35). There was no significant difference between the reported sweetness of AceK, RebA, RebD and Asp (see text for p values), with significant differences in reported bitterness for AceK, RebD and RebA (see text for p values).

To facilitate comparison across non-nutritive sweeteners, AceK ratings from the same individuals (n=122) from day 1 of the main study are also shown here (grey box in Figure 3-1). In

ANOVA comparing the non-nutritive sweeteners (i.e. excluding the disaccharides sucrose and gentiobiose), mean bitterness differed across stimuli [F (3,363) = 37.20; p <0.0001]. Perceived bitterness from RebA was greater than AceK (Tukey p <0.0001), RebD (p < 0.0001) and

Aspartame (p <0.0001). AceK was more bitter than RebD (p = 0.002) and aspartame (p <

0.0001). There was not sufficient evidence to conclude that RebD was more bitter than aspartame

(p = 0.70). Figure 2 also shows that Gentiobiose, a natural disaccharide containing aβ-(1–6)

81 linkage, was predominantly bitter with little to no sweetness, while the reverse was true for sucrose, as was expected.

Figure 3-2: Scatter plots showing (top) sweetness and (bottom) bitterness for RebA and RebD, with histograms along the axes. Mean sweetness was not significantly different between RebA and RebD, although both show substantial individual differences. Mean bitterness differed, with RebD showing significantly less bitterness. Again, substantial individual differences were observed between these sweeteners.

82

The mean sweetness for the four non-nutritive sweeteners fell between moderate and strong on a gLMS. In ANOVA, the sweetness ratings differed across stimuli [F(3,363) = 4.53; p

= 0.004], and this effect was primarily driven by the ratings for aspartame, which was sweeter than RebA, RebD and AceK (all p’s <0.05). Notably, the sweetness of AceK, RebA and RebD were not statistically different from one another (all p’s > 0.8).

Figure 3-3 shows correlations for the perceived bitterness of RebA with the bitterness of

RebD, AceK and Aspartame. The bitterness of RebA and RebD were significantly correlated with each other (R-sq=0.32; p<0.0001). Notably, RebA bitterness is not significantly correlated with perceived bitterness of AceK (R-sq=0.03; p=0.055). Although the association between RebA and

Asp was significant, the amount of variance explained was low (R-sq=0.08; p<0.0013).

Figure 3-3: Correlations of bitterness ratings between RebA and RebD, AceK and aspartame.

83

TAS2R9 SNP explains AceK bitterness but not bitterness of RebA and RebD

Previously, bitterness ratings of AceK collected from day 1 (Allen et al. 2013) were associated with a SNP in TAS2R9, Val187Ala (rs3741845: chr. 7). Here, to allow direct comparisons for these same polymorphisms across different non-nutritive sweeteners, we have reanalyzed the AceK ratings taken from Day 1 for the 122 individuals reported here. (Because of participant overlap, these data should not be taken as independent replication of our prior report

(Allen et al. 2013); they are included here to facilitate comparisons with RebA and RebD). Figure

3-4 shows sweet and bitter ratings for AceK, RebA and RebD as a function of the Val187Ala

SNP in TAS2R9. Here, a primarily European mixed ancestry cohort contained 44 Ala homozygotes, 59 heterozygtoes and 13 Val homozygtoes (genotype frequencies did not vary from

Hardy Weinberg equilibrium X2 = 1.06 p = 0.30). The perceived bitterness from AceK was predicted by the amino acid at position 187, with Val homozygotes and heterozygotes perceiving significantly more bitterness than Ala homozygotes (7.88±2.6, 7.38±1.2 and 1.70±1.4 respectively) [F(2,113) = 5.10; p = 0.0076]. There was no significant difference between Val homozygotes and heterozygotes (Tukey p = 0.98). Ala homozygotes had lower mean bitterness compared to heterozygotes (Tukey p = 0.009) and a similar trend was observed between Ala homozygotes and Val homozygotes (p = 0.10). AceK sweetness did not differ by Val187Ala genotype [F(2,113) = 0.91; p=0.40].

In this cohort, the bitterness of RebA and RebD were not explained by Val187Ala allele.

RebA bitterness did not differ across genotypes with Val homozygotes rating 13.9±2.5, heterozygotes 9.6±1.2, and Ala homozygotes 8.8±1.4 [F(2,113) = 1.68; p = 0.19]. Similarly,

RebD bitterness did not differ [F(2,113) = 0.41; p = 0.66] across Val homozygotes, heterozygotes and Ala homozygotes (2.7±1.1, 1.7±0.5 and 1.2±0.6 respectively).

84

Figure 3-4: Effect of the TAS2R9 Val187Ala polymorphism on the bitterness of AceK, RebA and RebD. As expected, the bitterness of AceK (collected on Day 1) was significantly different across genotype for these individuals; conversely, no effect of genotype was observed for RebA or RebD (see text).

TAS2R31 SNPs explain AceK bitterness but not bitterness of RebA and RebD

Polymorphisms in TAS2R31 (chr. 7) have shown to mediate the bitterness response to sulfonyl amide sweeteners (Allen, McGeary et al. ; Pronin, Xu et al. 2007; Roudnitzky, Bufe et al. 2011). Figure 3-5 shows the bitterness ratings of RebA, RebD and AceK across Val240Ile genotypes. Here, Val240Ile (rs10772423) failed to explain variation in the bitterness of RebA

85

[F(2,113) = 1.49; p = 0.23]. Similarly, Val240Ile did not explain variation in RebD bitterness

[F(2,113) = 0.12; p = 0.89], which was quite low (1.9±0.9, 1.8±0.5 and 2.2±0.6, for Ile homozygotes, heterozygotes and Val homozygotes, respectively). AceK ratings from day 1 were included to make the point that Val240Ile genotypes explained the bitterness of AceK in these same individuals [F(2,113) = 5.33; p = 0.006]. Ile homozygotes (n=19) and homozygotes (n=52) perceived greater bitterness compared to the non-functioning, Val homozygotes (n=45) (genotype frequencies did not vary from Hardy Weinberg equilibrium X2 = 0.141 p = 0.71). Conversely,

AceK sweetness ratings did not differ across genotype [F(2,113) = 0.52; p = 0.60].

Figure 3-5: Same as Figure 4, except for the Val240Ile polymorphism in TAS2R31. AceK bitterness ratings were significantly different across genotype, as expected, while there was no evidence of an effect for RebA and RebD (see text).

86

Previously, we reported a second SNP on TAS2R31, Ala227Val (rs10845293) associated with the bitterness of AceK in a European-American cohort (Allen et al. 2013).

The Ala227Val SNP was not associated with the bitterness ratings of RebA [F(1,108) = 1.25; p = 0.29] or RebD [F(2,108) = 0.18; p = 0.84] here.

Discussion

Although the main function of non-nutritive sweeteners is to provide sweetness with negligible calories, they often elicit undesirable bitter and metallic sensations (DuBois, Walters et al. 1991; Ellis 1995; Helgren, Lynch et al. 1955), limiting their utility. Moreover, many non- nutritive sweeteners are synthetic which are less desirable to many consumers, resulting in greater interest in natural non-nutritive sweeteners like Stevia and Monkfruit extracts. Currently, purified

RebA extract (from the leaves of Stevia rebaudiana) is approved for use in foods in the United

States as a GRAS ingredient. However, other glycosides within Stevia also elicit a wide range of sweet and bitter sensations (Hellfritsch, Brockhoff et al. 2012). Hellfritsch and colleagues recently described sweetness and bitterness response functions for 9 different Stevia glycosides.

To generate these response curves, individuals were trained using sweet and bitter references

(sucralose and rubusoside) at fixed concentrations. Their data suggested RebD might be a superior sweetener to RebA, as it has a steeper sweetness response curve with lower bitterness compared to RebA. Here, we confirm these data using a different psychophysical approach (i.e. measuring individual perceived intensities without providing a standard reference). We did not generate a dose response function; rather, a concentration was selected to provide at least moderate sweetness on a gLMS. At isosweet concentrations, RebD elicited significantly lower perceived bitterness than RebA. Our data, coupled with that of Hellfritsch et al. suggest it may be

87 desirable to commercialize RebD as a GRAS ingredient, or to selectively breed for Stevia plants with higher RebD content.

Many non-nutritive sweeteners exhibit a hyperbolic dose response function, which limits their maximal sweetness. This is why we use the phrase non-nutritive sweeteners rather than

‘high-intensity sweeteners’; the later is misleading given the linear dose-response function for bulk mono- and disaccharides (DuBois, Walters et al. 1991; Hayes 2008). At the concentration used here for RebA (1.65mM), we are in the flat region of the sweetness response curve (DuBois,

Walters et al. 1991; Hellfritsch, Brockhoff et al. 2012). Conversely, the bitterness response curve for RebA continues to increase linearly. That is, it may be possible to use a lower dose and still achieve the desired level of sweetness with lower levels of bitterness. However, because we were specifically interested in the covariation of bitterness across sweeteners, we intentionally selected concentrations that would ensure participants perceived some bitterness. Overall, aspartame bitterness was very low, consistent with previous reports (DuBois, Walters et al. 1991; Kamerud

& Delwiche 2007).

Roudnitzky and colleagues (Roudnitzky, Bufe et al. 2011) showed that AceK bitterness recognition thresholds are associated with polymorphisms in TAS2R31. Subsequently, we extended this work using a tag SNP approach, showing that TAS2R31 polymorphisms could also explain variation in the suprathreshold bitterness of AceK (Allen et al. 2013). Here, we show that a tag SNP thought to be in linkage disequilibrium with the functional Arg35Trp SNP in TAS2R31 does not explain variability in the bitterness perception of RebA or RebD. While activation of

TAS2R31 with AceK has been well documented (Kuhn, Bufe et al. 2004; Meyerhof, Batram et al.

2010; Roudnitzky, Bufe et al. 2011), we are unaware of any reports showing RebA and RebD activate TAS2R31 in vitro or in vivo. Likewise, putatively functional polymorphisms in TAS2R9 did not predict the bitterness of RebA or RebD. The inability of TAS2R31 and TAS2R9 SNPs to predict RebA and RebD bitterness would be expected given the absence of a relationship between

88

AceK bitterness and RebA/RebD bitterness. It is likely that the bitterness from RebA and RebD occur via other bitter receptors, possibly hT2R4 and hT2R14 (Hellfritsch, Brockhoff et al. 2012).

Additional work is needed to determine if SNPs in these genes might explain variable bitterness from RebA and RebD.

Conclusions

Here we show that stevioside extracts from Stevia differ in their perceived bitterness at suprathreshold concentrations that are isosweet. Purified RebD may be a superior natural non- nutritive sweetener compared to RebA, as it elicits less bitterness than RebA. Functional SNPs in

TAS2R31 known to explain variation in AceK and saccharin bitterness were unable to explain variability perceived bitterness of RebA or RebD. This is consistent with psychophysical data showing that AceK bitterness was not correlated with the perceived bitterness of RebA or RebD.

These data are consistent with our position that bitterness is not a simple monolithic trait that is high or low in an individual (Hayes, Bartoshuk et al. 2008; Hayes, Wallace et al. 2011). They also suggest that aversive sensations from RebA and RebD may not generalize to AceK and vice versa. Whether variable bitterness in Stevia extracts can be explained by polymorphisms in other bitter taste genes (TAS2Rs) remains to be determined.

89

Chapter 4

Bitter receptor alleles differentially influence alcoholic beverage liking

Adapted from:

Allen, A. L., McGeary, J.E., Hayes, J. E. (2013). " Bitter receptor alleles differentially influence

alcoholic beverage liking." Under review.

The taste of alcohol may be especially important during initial exposure, both in animal models (e.g. sucrose fading) and humans (e.g. alcopops). Prior data indicate bitter alleles associate with lower alcohol intake; this relationship presumably occurs via liking. However, alcoholic beverages are bitter and sweet. Bitterness is not always aversive, and perceptual masking can occur with the addition of other ingredients or mixers. Here, relationships between bitter receptor gene polymorphisms and self-reported liking of alcoholic beverages with different sensory profiles were tested. Participants (n=133) reported liking for 20 beverages based on prior experience. They were genotyped for two TAS2R38 polymorphisms, Ala49Pro (rs713598) and

Val262Ala (rs1726866), resulting in three common haplotypes: AV/AV, AV/PA, or PA/PA. We compared AV homozygotes to PA carriers as prior work suggests one copy of the PA haplotype is sufficient to depress intake. TAS2R38 genotypes associated with liking, with PA carriers reporting lower liking than AV homozygotes. We observed an interaction with beverage type.

TAS2R38 variation predicted liking in non-sweet alcoholic beverages, but no effect was observed in sweet alcoholic beverages. Protective effects of TAS2R38 genotype on intake may depend on

90 beverage type. Alcoholic beverages are not merely ethanol delivery systems; rather they are a wide range of stimuli with strikingly different sensory properties. Sweetened alcoholic beverages may potentially mitigate effects of sensory genetics on intake. Present data informs public health efforts regarding alcopops: protective effects of sensory genetics on liking (and thus intake) may be reduced during key developmental windows when sweetened beverages are readily available.

Introduction

Bitterness is innately aversive, causing rejection in neonatal nonhuman primates and human infants (Steiner, Glaser et al. 2001). However, adult humans often learn to like and ingest bitter beverages (e.g. beer and coffee), presumably due to learning resulting from postingestive pharmacological consequences. Moreover, humans have 25 different bitter receptor genes

(TAS2Rs) that are broadly and narrowly tuned to detect a diverse range of potentially toxic plant compounds (Meyerhof, Batram et al. 2010; Wooding, Gunn et al. 2010). Of these, a handful of

TAS2Rs are believed to contain functional polymorphisms (Hayes, Wallace et al. 2011; Hinrichs,

Wang et al. 2006; Reed, Zhu et al. 2010; Roudnitzky, Bufe et al. 2011; Soranzo, Bufe et al.

2005). Presumably, the evolutionary pressure for these polymorphisms arose from diet and the local plant environment that necessitated greater or lesser responsiveness to bitter stimuli for survival. The most widely studied bitter receptor gene, TAS2R38, has been shown to be protective against alcohol intake, at least in some populations (Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004; Hayes, Wallace et al. 2011; Wang, Hinrichs et al. 2007). Moreover, when gene-gene interactions between multiple alcohol related genotypes (e.g. ADH1B, ADH1C, ADH7, ALDH2, and TAS2R38) are considered (Mustavich, Miller et al. 2010), it appears taste may be especially protective during early exposure, as other metabolic genetic risk factors may never come into play if alcohol is not actually ingested due to aversive taste properties.

91

The TAS2R38 receptor gene contains three SNPs (Single Nucleotide Polymorphisms) that are inherited together in distinct haplotypes (Kim, Jorgenson et al. 2003). The three SNPs in this genotype are Pro49Ala, Ala262Val and Val296Ile. The two most common haplotypes are PAV and AVI (49% and 47%, respectively) in a population of European ancestry (Kim, Jorgenson et al. 2003). Some other combinations are observed infrequently (AAV, AAI and PVI) and they typically result in intermediate phenotypic sensitivity (Bufe, Breslin et al. 2005; Cannon, Baker et al. 2005; Hayes, Bartoshuk et al. 2008; Kim, Jorgenson et al. 2003; Kim, Wooding et al. 2005).

Data in vitro and in vivo suggest the first SNP (Pro49Ala) is the primary but not sole determinant of bitter taste phenotype (Bufe, Breslin et al. 2005; Mennella, Pepino et al. 2011). The ancestral

PAV haplotype associates with increased bitter response (‘tasting’) while the AVI haplotype associates with reduced response (‘non-tasting’); thus, heterozygotes show an intermediate response to certain bitter ligands like 6-n-propylthiouracil (PROP) (e.g. (Hayes, Bartoshuk et al.

2008).

Multiple alcohol related phenotypes (drinking frequency, summed quantity-frequency, and high risk drinking) have been previously associated with variation in TAS2R38 genotype

(Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004; Hayes, Wallace et al. 2011; Wang,

Hinrichs et al. 2007). However, ethanol and alcoholic beverages are not merely bitter; rather, they are simultaneously sweet and bitter (Berg, Filipello et al. 1955; Di Lorenzo, Kiefer et al. 1986;

Lemon, Brasser et al. 2004). Polyphenols in wine, hops in beer, as well as ethanol itself, give rise to this bitterness. The classical ‘protection hypothesis’ suggests that greater bitterness drives lower liking and intake of alcoholic beverages (Duffy, Davidson et al. 2004; Hayes, Wallace et al.

2011). Indeed, Lanier and colleagues found that greater endogenous bitterness and lower sweetness from sampled alcoholic beverages associated with reduced intake (Lanier, Hayes et al.). It is assumed that the link between TAS2R38 and intake is mediated via liking, but there is a gap in the literature, as this idea has never been formally tested. A standard drink has consistent

92 amount of ethanol: a 5oz glass of white wine, a 12 oz lager, and a 1.5 oz shot of tequila all deliver the same amount of active ingredient (NIAAA 2005). Thus, most prior work appropriately focuses on the amount of ethanol consumed without regard to the specific type of beverage.

However, in our everyday lives, it is readily apparent that beer and daiquiris are very different sensory (and culinary) experiences. The goal of the present study was not to characterize abuse, or measure potential relationships between TAS2R38 and alcohol dependence. Rather, we were testing the specific hypothesis that the influence of TAS2R38 genotype on liking of alcoholic beverages varies across different beverages in a reportedly healthy population.

Materials and Methods

Participants

Ninety women and 43 men with a mean age of 26.6±7.6(SD) were used in the final analysis (total n =133); data from 13 additional participants were not included due to rare haplotypes. Participants provided self-report of their ethnicity using the categories provided in

OMB Directive 15. Of participants who chose to disclose their ethnicity, 90 were Caucasian, 16 were Asian, and 4 were African or African-American; 23 participants chose not to report any ethnicity.

These data were collected as part of a larger study on the genetics of oral sensation. Each participant completed a screening questionnaire before the first session to ensure that they meet the study inclusion criteria. Eligibility criteria included: age between 18-45 years, not pregnant or breastfeeding, non-smoker (had not smoked in the last 30 days), no known defects of smell or taste, no lip, cheek or tongue piercings, no history of any condition involving chronic pain, not currently taking any prescription pain medication, no reported history of choking or difficulty

93 swallowing and no history of thyroid disease. Participants also needed to be willing to provide a saliva sample to obtain DNA. Written informed consent was obtained from all participants. All procedures were approved by the Pennsylvania State University Institutional Review Board

(protocol number #33176). Alcohol misuse, abuse or dependence were not assessed, as the goal of the larger study was focused broadly on oral sensation, and food and beverage liking, not alcohol related phenotypes.

Liking survey

Participants completed a 63 item hedonic survey with 27 foods and 20 alcoholic beverages (Byrnes & Hayes 2013). The survey included 16 non-food items to encourage participants to make ratings within the context of all experiences, not just foods and beverages

(Byrnes & Hayes 2013; Duffy, Hayes et al. 2009; Harrington, Kennedy et al. 2012). The scale used was an unstructured hedonic line scale with labels at each anchor. The right end of the scale was labeled “strongest liking of any kind” (100), and the left end was “strongest disliking of any kind” (-100) with ‘neutral” in the middle (0). Participants were asked to rate each food or non- food item on the scale. If they had not experienced a specific item previously, they were instructed to not make a rating and skip ahead to the next item. Alcoholic beverages on the survey

(with their exact wording) included: ‘pale ales (Bass, Summer Sam Adams)’, ‘scotch or whiskey

(straight or with ice)’, ‘vodka or gin martini’, ‘malt liquor (Colt 45, Old English)’, ‘lagers

(Budweiser, Stella, Heineken, Corona)’, ‘dry cider (Magners, Strongbow)’, ‘sweet white wine’,

‘blended whiskey (straight or with ice)’, ‘bitter beer (IPAs, stouts)’, ‘off-dry or semi-sweet white wine’, ‘vodka (straight or with ice)’, ‘spirits with soda (rum n’ coke, 7&7)’, ‘flavored malt beverages (Skyy Blue, Smirnoff Ice, Mike’s)’, ‘fruity red wine’, ‘dry wine (red or white)’,

‘margaritas or daiquiris’, ‘shooters (lemon drop, kamikaze)’, ‘spirits with energy drinks (redbull

94 and vodka)’, ‘spirits with juice or milk (White Russian)’, ‘fortified wine (port)’ and ‘brandy or cognac’. All hedonic data were collected using Compusense five, version 5.2 (Guelph, Ontario,

Canada).

Genetic Analysis

DNA was collected using Oragene saliva collection kits following manufacturer instructions (Genotek Inc, Ontario, Canada). SNPs (single nucleotide polymorphisms) within

TAS2R38 were determined using Sequenom MassARRAY technology (Sequenom, San Diego,

CA). Primers were purchased from Integrated DNA Technologies (Coralville, Iowa, USA).

Genotypes were assigned via MassARRAY software (Sequenom) and independently inspected by two technicians. As a standard procedure, 15% of samples are rerun to ensure reliability.

Statistical Analysis

Data were analyzed using SAS 9.2 (Cary, NC). A repeated measures analysis of variance

(ANOVA) was performed via proc mixed. For TAS2R38, we assumed that individuals heterozygous at both sites (e.g., CG TC) were common haplotype heterozygotes (e.g., PA_/AV_ rather than AA_/PV_) as the probability of having 2 rare haplotypes is extremely low (Wooding,

Kim et al. 2004).

95

Results

Mean liking varies substantially across alcoholic beverage type, and liking of alcoholic beverages differs by TAS2R38 genotype

Previously, we reported that in a separate cohort of reportedly healthy adults (Dotson,

Wallace et al. 2012; Duffy, Davidson et al. 2004; Hayes, Wallace et al. 2011; Wang, Hinrichs et al. 2007), AVI/AVI homozygotes consumed alcohol beverages significantly more often than either PAV/AVI heterozygotes and PAV/PAV homozygotes, who did not differ from each other.

Based on this, PA/PA and PA/AV individuals were pooled together into a PA carrier group

(PA/*) for the present analysis. (Reanalyzing with three groups, AV/AV, PA/AV and PA/PA, did not substantively alter our findings.)

Across all 20 beverages in our liking survey, there were main effects of beverage

[F(19,1910) = 14.27; p <0.001] and TAS2R38 genotype [F(1,125) = 4.07; p = 0.046]. Mean liking across participants varied substantially across the different beverages, ranging from -12 to

40.6 (Figure 4-1). Additionally, Figure 4-1 implies sweeter alcohol beverages were more highly liked across the group, irrespective of genotype. With regard to genotype, mean liking for the

PA/* group was significantly lower (p =0.046) than the AV/AV group mean (9.5±2.08 SEM versus 18.2±3.79). The interaction between individual beverages and genotype was not significant [F(19,1910) = 1.01; p = 0.46].

96

Figure 4-1: Mean liking ratings ± standard errors for 20 alcoholic beverages

97

Figure 4-2: Mean liking ratings ± SEM for AV homozygotes and PA carriers (PA/PA and PA/AV individuals) for sweet and non-sweet alcoholic beverages. Ratings were measured as part of a liking survey that also included food and non-food items to generalize to the context beyond foods and beverages (see methods). The sweet group included: flavored malt beverages, sweet white wine, fruity white wine, off-dry or semi-sweet white wine, malt liquor, margaritas or daiquiris, spirits with juice or milk, spirits with energy drinks, and spirits with soda. The non-sweet group included: lager beer, pale ales, bitter beer, dry wine, scotch, vodka, and martinis with gin or vodka. See text for results of repeated measures ANOVA; p-values are unadjusted t-tests.

TAS2R38 Genotype associates with liking of non-sweet alcoholic beverages, but not for sweet alcoholic beverages

Sweetness masks bitterness centrally (Lawless 1979), so any effects of TAS2R38 variation on liking may be attenuated in sweetened drinks. To formally test this, we categorized the beverages in our survey as being sweet or not sweet. The sweet beverage group (n = 9) included: flavored malt beverages, sweet white wine, fruity red wine, off-dry or semi-sweet white wine, malt liquor, margaritas or daiquiris, spirits with juice or milk, spirits with energy drinks,

98 and spirits with soda. The not-sweet beverage group (n=7) included lager beer, pale ales, bitter beer, dry wine, scotch, vodka and martinis with gin or vodka. Because participants did not rate liking for beverages they had never had, missing ratings for the remaining 4 beverages (dry hard cider, port, shots, and brandy or cognac) exceeded 30%, so they were excluded from further analyses.

Consistent with our hypothesis, there was a significant interaction between beverage group (sweet versus non-sweet) and TAS2R38 genotype [F(1,119) = 6.54; p = 0.012]. As shown in Figure 4-2, liking differed by genotype for the non-sweet beverages. However, mean liking of sweet alcoholic beverages did not differ by genotype. Additionally, we observed main effects of beverage group [F(1,119) = 32.5; p <0.0001], as mean liking of the sweet beverages was significantly higher (p <0.0001) than non-sweet beverages (22.2±2.4 versus 8.9±2.5), irrespective of genotype (not shown). Consistent with the analysis above, we also observed a significant main effect of genotype [F(1,125) = 4.87; p = 0.029] for the 16 beverages in this analysis.

99

Figure 4-3: Mean liking ratings ± SEM for AV homozygotes and PA carriers for sweet beverages (n=9). See text for results of repeated measures ANOVA; p-values are unadjusted t-tests.

To further characterize the influence of TAS2R38 genotype on liking for individual alcoholic beverages, we split the model above into two separate ANOVAs, one for the sweet group and one for the non-sweet group. When looking at just the 9 beverages in the sweet group, the genotype by beverage interaction was no longer significant [F(8,835) = 0.97; p = 0.46], as shown in Figure 4-2. Similarly, there was no main effect of genotype [F(1,121) = 0.21; p = 0.65] on liking for the individual beverages in the sweet group. We did observe a main effect of beverage [F(8,835) = 20.15; p <0.0001], indicating liking ratings differed across the individual beverages within the sweet alcoholic beverage group.

For the non-sweet beverage group, we observed main effects of genotype [F(1,123) =

6.34; p = 0.013], and beverage [F(6,634) = 13.6; p <0.0001]. Figure 4-2 shows that the PA/* individuals generally reported lower liking ratings to the individual beverages in the non-sweet group. The genotype by beverage interaction was not significant [F(6,634) = 0.68; p = 0.67].

100

Figure 4-4: Mean liking ratings ± SEM for AV homozygotes and PA carriers for non-sweet beverages (n=7). See text for results of repeated measures ANOVA; p-values are unadjusted t- tests.

Discussion

Previously, variation in TAS2R38 has been associated with differential intake for alcoholic beverages (Dotson, Wallace et al. 2012; Duffy, Davidson et al. 2004; Hayes, Wallace et al. 2011; Wang, Hinrichs et al. 2007)); presumably this acts via liking, with greater bitterness driving lower intake (Lanier, Hayes et al.). Here, we show for the first time that remembered liking for alcoholic beverages varies by TAS2R38 genotype. Additionally, we show that genetic effects are muted or absent in sweetened alcoholic beverages. To many clinicians, addiction researchers and nutritional epidemiologists, the typical focus is on the total amount of active ingredient (ethanol) consumed. However, our data suggest not all drinks are created equal, at least

101 with respect to hypotheses about the protective effects of sensory genetics on alcohol liking and presumably intake. Thus, it may be important to consider the different sensory profiles of various alcoholic beverages in addition to measuring total intake. From a public health standpoint, this may be especially important, as protective effects arising from aversive bitterness may be attenuated by increased availability of highly sweetened ready to drink alcoholic beverages (i.e., alcopops).

In the COGA cohort, Wang et al found bitter receptor genetics predicted a measure of intake, but not DSM-IV dependence diagnoses (Wang, Hinrichs et al. 2007). This implies variation in bitter taste genes may act as a variable stage gate for other risk alleles; that is, if you never learn to drink alcohol because of an aversive taste, other metabolic genetic risk factors may be largely irrelevant (Mustavich, Miller et al. 2010). Conversely, once dependent, it seems that taste may be largely irrelevant, as the rewarding properties of alcohol and learning may override any innate taste aversions. This suggests any potential protective effects of taste genetics may transient, acting only during certain developmental windows. If true, then the public health consequences of the present work are striking, as it suggests choosing sweet alcoholic beverages can overcome innate aversions that have a biological basis, similar to the classical sucrose fading techniques used in rodent models of alcohol addiction.

Additionally, Wang and colleagues found that TAS2R38 genotype predicted the high-risk maxdrinks phenotype in African-Americans, but not in European-Americans (Wang, Hinrichs et al. 2007). Assuming the protective affects of TAS2R38 on alcohol intake are blunted by the type of beverage, then different cultural habits, access, and drinking patterns may help explain such discrepancies. For example, if European-Americans are more likely to chose wine over beer compared to African-Americans for reasons unrelated to genetics, then we would expect

TAS2R38 effects to be weaker in European-Americans, consistent with the findings of Wang et al. Moreover, it is tempting to speculate that effects of bitter receptor genetics may also differ by

102 gender due to sociocultural norms. That is, protective effects may be weaker in women, as men may experience more pressure to avoid sweet ‘girly’ drinks. Future work should revisit this question, as the present study was not powered to explore gender differences.

Here, several of the non-sweet alcoholic beverages failed to show a significant effect of

TAS2R38 genotype on liking, specifically scotch and bitter beers. As the trends were in the same direction as the other non-sweet beverages, this may simply be a matter of power, although we find this explanation unsatisfying. Previously, we found that TAS2R38 genotype did not associate with liking for blended whisky tasted in the laboratory (Hayes, Wallace et al. 2011).

Here, we asked individuals to report liking for single malt scotch. Presumably, peaty, mossy, smoky aromas are far more important than bitterness as drivers of liking. For bitter beers, we are likely observing a floor effect; bitter beers like Stouts and India Pale Ales are extremely bitter and are readily perceived as bitter by both PA carriers and AV homozygotes. Also, liking for these beer styles may be less a function of bitterness, with olfactory cues again playing a larger role

(e.g., chocolate and coffee aromas, and hoppy, floral notes, respectively). Finally, the effect for martinis was marginal. This may reflect participants including sweetened cocktails (e.g.

‘appletinis’), along with more classic dry martinis. Separating out traditional martinis from other flavored cocktails served in martini glasses in future work may help resolve this question.

Additionally, asking participants to sample beverages in the laboratory rather than providing remembered liking in future work may provide additional insight, although remembered liking is typically a reliable predictor of liking for sampled foods (e.g., (Hayes, Sullivan et al. 2010;

Sharafi, Hayes et al. 2012)).

103

Conclusions

Here we show that TAS2R38 genotype is a significant predictor of remembered liking for a variety of alcoholic beverages. These data extends previous findings by addressing a gap in the sensation-liking-intake model assumed in prior studies that associate alcohol intake with variation in bitter receptor genes. That is, we provide empirical data supporting the assumption that the relationship between taste genetics and alcohol intake is mediated through liking. Additionally, this work suggests protective effects of variation in sensory genetics may be attenuated when bitter sensations are minimized in sweetened beverages. However, these effects may also be reduced in social and physical environments that support consumption of alcoholic beverages, which was not assessed here. These data indicate that not all drinks are behaviorally equivalent even if they contain the same amount of ethanol, and that additional insight may be gained by considering the sensory profile of an alcoholic beverage in addition to quantifying them as standard drinks. More work is needed to determine how the taste effects shown here may interact with gender, culture and environment to influence drinking patterns and habits.

104

Chapter 5

Conclusions and Future work

The findings reported here give greater insight into how polymorphisms within bitter taste receptor genes can alter the ability of an individual to perceive bitterness from different food ingredients. These findings may help to improve the development of food products, pharmaceutical formulation and also to prescribe diets that decrease risk for disease. This knowledge can be utilized in future studies interested in bitter taste receptor genetics along with sensitivity to bitter compounds to further understand how genetic variation alters perception. The major experimental findings of these studies are:

1) The bitterness of AceK differs greatly across individuals. It has been demonstrated that

AceK activates hT2R31 and has several polymorphisms within the gene that are significant predictors of detection threshold. Suprathreshold bitterness ratings of AceK are explained by functional polymorphisms in TAS2R31 and unexpectedly TAS2R9. Whether AceK activates hT2R9 in vitro is unknown.

2) The bitterness elicited from the natural non-nutritive sweeteners rebaudioside A and rebaudioside D is not explained by polymorphisms that predict AceK bitterness. This would be expected as we also show the bitterness of AceK does not associate with the bitterness of RebA or RebD.

3) Functional polymorphisms in TAS2R38 have been associated with the ability to perceive bitterness for a wide range of bitter foods, including alcohol and vegetables. Frequently, greater bitterness is often associated with decreased liking and intake. Many alcoholic beverages elicit both sweet and bitter tastes. Here we found that reported liking of non-sweet alcoholic

105 beverages is associated with TAS2R38 functional polymorphisms. As hypothesized, reported liking for sweet alcoholic beverages was not predicted by TAS2R38.

The findings reported here provide greater understanding into the effects of functional polymorphisms in bitter taste receptor genes on food sensations. Additional work is needed to confirm these findings, and also answer the questions that have arisen from these data.

Functional polymorphisms in TAS2R31 explain variation in the perceived bitterness of suprathreshold AceK; however, it is unknown what moiety is activating this receptor. By comparing bitterness response or cellular activation for compounds with similar structures, this may help to identify the moiety that is directly interacting with the TAS2R31 receptor and provide insight to other compounds that activate the receptor (e.g. aloin). It is possible that by identifying the specific moiety that interacts with this receptor may make it possible to develop a sweetener derived from AceK that lacks the ability to bind to hT2R31 while retaining desired sweetness.

Similarly, it would be useful to use this method and determine the chemical configuration necessary for TAS2R9 activation, as AceK may interact with these receptors differently. While previous work suggests oflaxacin activates hT2R9 in vitro, there has not been any structure activity research for this receptor to date.

Moreover, it is unclear whether AceK activates TAS2R9 in vitro. First, it would be beneficial to measure the activation of hT2R9 in vitro to confirm if the functional polymorphism in TAS2R9 is directly responsible for the ability to perceive bitterness from AceK. On the other hand, this SNP may be in LD with the causal SNP, which may be located within the gene or in a nearby gene. I suspect that the later is the case, as a silent DNA mutation in TAS2R50 is in LD with the SNP in TAS2R9 and this SNP was also a significant predictor of AceK bitterness (not reported here). Thus, further studies are needed to confirm activation of TAS2R9 and TAS2R50 by

AceK.

106

Here, we determined that bitterness from the non-nutritive sweeteners RebA and RebD is not predicted by SNPs in TAS2R31 and TAS2R9 that do predict AceK bitterness. Nor did RebA and RebD bitterness correlate with the bitterness of AceK, which reinforces the findings that their perceived bitterness is not explained by thesepolymorphisms. Practically, individuals who reject

AceK due to its bitterness may not reject other non-nutritive sweeteners. Future studies exploring polymorphisms in other bitter taste receptors may determine the functional receptor, and the SNP responsible for the variability in the bitterness response to RebA and RebD. This study reinforces the importance of using the target compound as a reference in sensory panels, that individuals who perceive bitterness from one sweetener may not perceive bitterness from a different sweetener.

Additionally, RebD appears to be a superior non-nutritive sweetener compared to RebA.

Overall, it elicits less bitterness on average than the currently approved RebA. It may be of interest to commercialize rebaudioside D due to minimal bitterness. However, the concentration of RebD that is found in Stevia leaves is relatively low compared to stevioside and RebA. Thus, additional work is needed to investigate methods to alter the plant to produce higher levels of

RebD, as has already been implemented for the production of RebA.

Interestingly, all the work to date involves measuring bitterness of AceK in water. It is unknown whether these SNPs would predict the bitterness of AceK when presented in a food or beverage. Similarly, it would be useful to determine if SNPs in TAS2R31 and TAS2R9 impact consumer preference for a food or beverage prepared with and without AceK. These studies would help the food industry determine if product developers should consider these polymorphisms when selecting a non-nutritive sweetener for their product.

Previous studies have reported that variation in the bitter taste receptor TAS2R38 gene predicts intake of alcoholic beverages. Here, we report liking for different types of alcoholic beverages differs greatly by beverage type. Moreover, reported liking for non-sweet alcoholic

107 beverages was associated with functional polymorphisms in TAS2R38. This finding may help to better understand the relationship between bitterness, liking and alcohol intake. Additionally, this highlights the importance of separating out different types of alcoholic beverages, as not all alcoholic beverages are liked equally.

There are numerous drawbacks to using only reported liking for alcoholic beverages such as influence of past experiences, social pressures and expectations. Regardless, we were able to explain a significant amount of variance in liking. To further improve our understanding of this liking-intake relationship, additional studies that incorporate liking and intensity ratings of sampled beverages and also reported intake, in addition to TAS2R38 genotype are required. This type of study would greatly impact the field. Having individuals taste sampled alcoholic beverages would help to minimize outside factors such as remembered previous experiences.

Combining liking and intensity ratings would confirm our findings, and help to determine the validity of collecting reported liking when considering intake frequency as a potential predictor of alcoholism risk. Also, comparing liking and intensity ratings with reported intake will help to determine how liking influences intake for different alcoholic beverages.

Overall, this thesis provides new contributions regarding polymorphisms within bitter taste receptor genes and their effects on perceived bitterness of non-nutritive sweeteners and the reported liking of alcoholic beverages. These findings, while directly relevant to non-nutritive sweeteners and alcoholic beverages, also improve our understanding of how SNPs in TAS2Rs can mediate perceived bitterness. Our understanding of these relationships are still in their infancy, as there are many SNPs yet to be tested for many bitter compounds and foods. Continuing to conduct research in determining what compounds activate the 25 bitter taste receptors, identifying functional polymorphisms and exploring their impact on sensitivity, liking and intake will help to advance the field and grow our understanding of the genetic influences on taste perception.

108 References

Adler, E., Hoon, M.A., Mueller, K.L., Chandrashekar, J., Ryba, N.J.P., et al. (2000). "A novel family of mammalian taste receptors." Cell 100(6): 693-702. Allen, A.L., Mcgeary, J.E. and Hayes, J.E. (2013). "Bitterness of the non-nutritive sweetener Acesulfame Potassium varies with polymorphisms in TAS2R9 and TAS2R31." Chemical senses DOI: 10.1093/chemse/bjt017. Anderson, G.H., Foreyt, J., Sigman-Grant, M. and Allison, D.B. (2012). "The Use of Low- Calorie Sweeteners by Adults: Impact on Weight Management." The Journal of Nutrition 142(6): 1163s-1169s. Bajec, M.R. and Pickering, G.J. (2008). "Thermal taste, PROP responsiveness, and perception of oral sensations." Physiology & behavior 95(4): 581-590. Barnicot, N., Harris, H. and Kalmus, H. (1951). "Taste thresholds of further eighteen compounds and their correlation with PTC thresholds." Annals of eugenics 16(1): 119-128. Barratt-Fornell, A. and Drewnowski, A. (2002). "The taste of health: Nature's bitter gifts." Nutrition Today 37(4): 144-150. Barrett, J., Fry, B., Maller, J. and Daly, M. (2005). "Haploview: analysis and visualization of LD and haplotype maps." Bioinformatics 21(2): 263-265. Bartoshuk, L., Marino, S. and Snyder, D. (2007). "Age and hormonal effects on sweet taste and preference." Appetite 49(1): 277. Bartoshuk, L.M. (1975). "Taste mixtures: is mixture suppression related to compression?" Physiology & behavior 14(5): 643-649. Bartoshuk, L.M. (1978). "The psychophysics of taste." The American journal of clinical nutrition 31(6): 1068-1077. Bartoshuk, L.M. (1979). "Bitter taste of saccharin related to the genetic ability to taste the bitter substance 6-n-propylthiouracil." Science 205(4409): 934-935. Bartoshuk, L.M. (2000). "Comparing sensory experiences across individuals: recent psychophysical advances illuminate genetic variation in taste perception." Chemical senses 25(4): 447-460. Bartoshuk, L.M., Conner, E., Grubin, D., Karrer, T., Kochenbach, K., et al. (1993). "PROP supertasters and the perception of ethyl alcohol. ." Chemical senses 18: 526-527. Bartoshuk, L.M., Duffy, V.B., Green, B.G., Hoffman, H.J., Ko, C.W., et al. (2004). "Valid across- group comparisons with labeled scales: the gLMS versus magnitude matching." Physiology & behavior 82(1): 109-114. Bartoshuk, L.M., Duffy, V.B. and Miller, I.J. (1994). "PTC/PROP tasting: Anatomy, psychophysics, and sex effects." Physiology & behavior 56(6): 1165-1171. Basson, M.D., Dichello, S.Z. and Panzini, L. (2005). "Association between 6-n-propylthiouracil (PROP) bitterness and colonic neoplasms." Digestive diseases and sciences 50(3): 483- 489. Beck, C.I. (1974). Sweetness, character, and application of aspartic acid-based sweeteners. . "Symposium:Sweetners". G.E. Inglett, The Avi Publishing Co., Inc. Bell, K.I. and Tepper, B.J. (2006). "Short-term vegetable intake by young children classified by 6-n-propylthoiuracil bitter-taste phenotype." The American journal of clinical nutrition 84(1): 245-251. Bell, K.I. and Tepper, B.J. (2006). "Short-term vegetable intake by young children classified by 6-n-propylthoiuracil bitter-taste phenotype." American Journal of Clinical Nutrition 84(1): 245-251.

109

Bennett, S.M., Zhou, L. and Hayes, J.E. (2012). "Using Milk Fat to Reduce the Irritation and Bitter Taste of Ibuprofen." Chemosensory Perception: 1-6. Berg, H., Filipello, F., Hinreiner, E. and Webb, A. (1955). "Evaluation of thresholds and minimum difference concentrations for various constituents of wines. II. Sweetness: the effect of ethyl alcohol, organic acids and tannin." Food Technol 9(6): 138-140. Berkey, C.S., Rockett, H.R.H., Field, A.E., Gillman, M.W. and Colditz, G.A. (2012). "Sugar- Added beverages and adolescent weight change." Obesity research 12(5): 778-788. Blanco, V., Auw, J., Sims, C. and O'keefe, S. (1998). "Effect of processing on phenolics of wines." Advances in experimental medicine and biology 434: 327-340. Bleich, S.N., Wang, Y.C., Wang, Y. and Gortmaker, S.L. (2009). "Increasing consumption of sugar-sweetened beverages among US adults: 1988–1994 to 1999–2004." The American journal of clinical nutrition 89(1): 372-381. Borg, G. (1982). "A category scale with ratio properties for intermodal and interindividual comparisons." Psychophysical judgment and the process of perception: 25-34. Brandle, J., Starratt, A. and Gijzen, M. (1998). "Stevia rebaudiana: its agricultural, biological, and chemical properties." Canadian Journal of Plant Science 78(4): 527-536. Bravo, L. (2009). "Polyphenols: chemistry, dietary sources, metabolism, and nutritional significance." Nutrition Reviews 56(11): 317-333. Bufe, B., Breslin, P.A., Kuhn, C., Reed, D.R., Tharp, C.D., et al. (2005). "The molecular basis of individual differences in phenylthiocarbamide and propylthiouracil bitterness perception." Current Biology 15(4): 322-327. Bufe, B., Hofmann, T., Krautwurst, D., Raguse, J.D. and Meyerhof, W. (2002). "The human TAS2R16 receptor mediates bitter taste in response to beta-glucopyranosides." Nature genetics 32(3): 397-401. Byrnes, N.K. and Hayes, J.E. (2013). "Personality factors predict spicy food liking and intake." Food Quality and Preference 28(1): 213-221. Calo, C., Padiglia, A., Zonza, A., Corrias, L., Contu, P., et al. (2011). "Polymorphisms in TAS2R38 and the taste bud trophic factor, gustin gene co-operate in modulating PROP taste phenotype." Physiology & behavior. Cannon, D.S., Baker, T.B., Piper, M.E., Scholand, M.B., Lawrence, D.L., et al. (2005). "Associations between phenylthiocarbamide gene polymorphisms and cigarette smoking." Nicotine & tobacco research 7(6): 853-858. Cardello, A.V. (1995). "Food quality: relativity, context and consumer expectations." Food Quality and Preference 6(3): 163-170. Carey, G. (2007). The association design and a continuous phenotype. Institute for Behavioral Genetics, Univeristy of Colorado, Available from: http://psych.colorado.edu/~carey/Courses/PSYC5102/handouts/Association- Quantitative.pdf Chandrashekar, J., Hoon, M.A., Ryba, N.J.P. and Zuker, C.S. (2006). "The receptors and cells for mammalian taste." Nature 444(7117): 288-294. Chandrashekar, J., Mueller, K.L., Hoon, M.A., Adler, E., Feng, L., et al. (2000). "T2Rs function as bitter taste receptors." Cell 100(6): 703-711. Cines, B.M. and Rozin, P. (1982). "Some aspects of the liking for hot coffee and coffee flavor." Appetite 3(1): 23-34. Clark, C.C. and Lawless, H.T. (1994). "Limiting response alternatives in time-intensity scaling: an examination of the halo-dumping effect." Chemical senses 19(6): 583-594. Clauss, K. and Jensen, H. (1973). "Oxathiazinone dioxides- a new group of sweetening agents." Angewandte Chemie International Edition in English 12(11): 869-876.

110

Craig, W.J. (1997). "Phytochemicals: guardians of our health." Journal of the American Dietetic Association 97(10): S199-S204. Crammer, B., Ikan, R. (1987). Progress in the Chemistry and Properties of the Rebaudiosides. In Developments in Sweeteners. T.H. Grenby. London, Elsevier Applied Science: 45−64. Davidson, T.L., Martin, A.A., Clark, K. and Swithers, S.E. (2011). "Intake of high-intensity sweeteners alters the ability of sweet taste to signal caloric consequences: Implications for the learned control of energy and body weight regulation." Quarterly Journal of Experimental Psychology 64(7): 1430-1441. Davis, E.A. (1995). "Functionality of sugars: physicochemical interactions in foods." The American journal of clinical nutrition 62(1): 170S-177S. De La Hunty, A., Gibson, S. and Ashwell, M. (2006). "A review of the effectiveness of aspartame in helping with weight control." Nutrition Bulletin 31(2): 115-128. De Snoo, K. (1937). "Das trikende Kind im Uterus." Monatsschrift fur Geburtshilfe und Gynakologie 105: 88-97. Delwiche, J. (1996). "Are there ‘basic’tastes?" Trends in Food Science & Technology 7(12): 411- 415. Delwiche, J.F. and Warnock, A.R. (2008). Perception and Acceptance of Sweeteners. Sweetness and Sweeteners. D.K. Weerasinghe and DuBois, G.E. Washington, Amer Chemical Soc. 979: 285-295. Di Lorenzo, P.M., Kiefer, S.W., Rice, A.G. and Garcia, J. (1986). "Neural and behavioral responsivity to ethyl alcohol as a tastant." Alcohol 3(1): 55-61. Dinehart, M., Hayes, J., Bartoshuk, L., Lanier, S. and Duffy, V. (2006). "Bitter taste markers explain variability in vegetable sweetness, bitterness, and intake." Physiology & behavior 87(2): 304-313. Dionne, R.A., Bartoshuk, L., Mogil, J. and Witter, J. (2005). "Individual responder analyses for pain: does one pain scale fit all?" Trends in pharmacological sciences 26(3): 125-130. Dotson, C.D., Wallace, M.R., Bartoshuk, L.M. and Logan, H.L. (2012). "Variation in the Gene TAS2R13 is Associated with Differences in Alcohol Consumption in Patients with Head and Neck Cancer." Chemical senses 37(8): 737-744. Dotson, C.D., Zhang, L., Xu, H., Shin, Y.K., Vigues, S., et al. (2008). "Bitter taste receptors influence glucose homeostasis." PloS one 3(12): e3974. Drewnowski, A. (2001). "The science and complexity of bitter taste." Nutrition Reviews 59(6): 163-169. Drewnowski, A. and Gomez-Carneros, C. (2000). "Bitter taste, phytonutrients, and the consumer: a review." The American journal of clinical nutrition 72(6): 1424-1435. Drewnowski, A., Henderson, S. and Shore, A. (1997). "Taste responses to naringin, a flavonoid, and the acceptance of grapefruit juice are related to genetic sensitivity to 6-n- propylthiouracil." The American journal of clinical nutrition 66(2): 391-397. Drewnowski, A., Henderson, S.A. and Shore, A.B. (1997). "Genetic sensitivity to 6-n- propylthiouracil (PROP) and hedonic responses to bitter and sweet tastes." Chemical senses 22(1): 27-37. Dubois, G.E. and Prakash, I. (2012). "Non-caloric sweeteners, sweetness modulators, and sweetener enhancers." Annual Review of Food Science and Technology 3: 353-380. Dubois, G.E., Walters, D.E., Schiffman, S.S., Warwick, Z.S., Booth, B.J., et al. (1991). Concentration-response relationships of sweeteners. Discovery, Molecular Design and Chemoreception, American Chemical Society, Washington, D.C. Duffy, V.B. (2007). "Variation in oral sensation: implications for diet and health." Current opinion in gastroenterology 23(2): 171-177.

111

Duffy, V.B. and Anderson, G. (1998). "Position of the American Dietetic Association: use of nutritive and nonnutritive sweeteners." Journal of the American Dietetic Association 98(5): 580-587. Duffy, V.B. and Bartoshuk, L.M. (2000). "Food acceptance and genetic variation in taste." Journal of the American Dietetic Association 100(6): 647-655. Duffy, V.B., Davidson, A.C., Kidd, J.R., Kidd, K.K., Speed, W.C., et al. (2004). "Bitter Receptor Gene (TAS2R38), 6-n-Propylthiouracil (PROP) Bitterness and Alcohol Intake." Alcoholism: Clinical and Experimental Research 28(11): 1629-1637. Duffy, V.B., Hayes, J.E., Davidson, A.C., Kidd, J.R., Kidd, K.K., et al. (2010). "Vegetable intake in college-aged adults is explained by oral sensory phenotypes and TAS2R38 genotype." Chemosensory Perception: 1-12. Duffy, V.B., Hayes, J.E., Sullivan, B.S. and Faghri, P. (2009). "Surveying Food and Beverage Liking." Annals of the New York Academy of Sciences 1170(1): 558-568. Duffy, V.B., Peterson, J.M. and Bartoshuk, L.M. (2004). "Associations between taste genetics, oral sensation and alcohol intake." Physiology & behavior 82(2-3): 435-445. Ellis, J.W. (1995). "Overview of sweeteners." Journal of chemical education 72(8): 671. Engen, T. and Kling, R. (1972). "Psychophysics: 1." Discrimination and detection. In. Feeney, E. (2011). "The impact of bitter perception and genotypic variation of TAS2R38 on food choice." Nutrition Bulletin 36(1): 20-33. Fischer, R., Griffin, F., England, S. and Garn, S. (1961). "Taste thresholds and food dislikes." Nature 191: 1328. Forshee, R.A. and Storey, M.L. (2003). "Total beverage consumption and beverage choices among children and adolescents." International journal of food sciences and nutrition 54(4): 297-307. Fox, A.L. (1932). "The relationship between chemical constitution and taste." Proceedings of the National Academy of Sciences of the United States of America 18(1): 115. Glanville, E. and Kaplan, A. (1965). "Food preference and sensitivity of taste for bitter compounds." Glanz, K., Basil, M., Maibach, E., Goldberg, J. and Snyder, D. (1998). "Why Americans Eat What They Do: Taste, Nutrition, Cost, Convenience, and Weight Control Concerns as Influences on Food Consumption." Journal of the American Dietetic Association 98(10): 1118-1126. Glendinning, J.I. (1994). "Is the bitter rejection response always adaptive?" Physiology & behavior 56(6): 1217-1227. Go, Y., Satta, Y., Takenaka, O. and Takahata, N. (2005). "Lineage-specific loss of function of bitter taste receptor genes in humans and nonhuman primates." Genetics 170(1): 313-326. Gorovic, N., Afzal, S., Tj√∏Nneland, A., Overvad, K., Vogel, U., et al. (2011). "Genetic variation in the h TAS2R38 taste receptor and brassica vegetable intake." Scandinavian Journal of Clinical & Laboratory Investigation 71(4): 274-279. Green (1993). "Derivation and evaluation of a semantic scale of oral sensation magnitude with apparent ratio properties." Chemical senses 18: 683- 702. Green, B.G., Dalton, P., Cowart, B., Shaffer, G., Rankin, K., et al. (1996). "Evaluating the Labeled Magnitude Scale‚ for measuring sensations of taste and smell." Chemical senses 21(3): 323-334. Green, B.G. and Hayes, J.E. (2003). "Capsaicin as a probe of the relationship between bitter taste and chemesthesis." Physiology & behavior 79(4): 811-821. Green, B.G. and Hayes, J.E. (2004). "Individual Differences in Perception of Bitterness from Capsaicin, Piperine and Zingerone." Chemical senses 29(1): 53-60.

112

Green, B.G., Lim, J., Osterhoff, F., Blacher, K. and Nachtigal, D. (2010). "Taste mixture interactions: surpression, additivity, and the predominance of sweetnenss." Physiology & behavior 101(5): 731-737. Guthrie, J.F. and Morton, J.F. (2000). "Food sources of added sweeteners in the diets of Americans." Journal of the American Dietetic Association 100(1): 43-51. Hall, M.J., Bartoshuk, L.M., Cain, W.S. and Stevens, J.C. (1975). "PTC taste blindness and the taste of caffeine." Nature. Hamer, D.H. (2000). "Beware the chopsticks gene." Molecular psychiatry 5(1): 11-13. Hanger, L., Lotz, A. and Lepeniotis, S. (1996). "Descriptive profiles of selected high intensity sweeteners (HIS), HIS blends, and sucrose." Journal of food science 61(2): 456-459. Hansen, J.L., Reed, D.R., Wright, M.J., Martin, N.G. and Breslin, P.a.S. (2006). "Heritability and genetic covariation of sensitivity to PROP, SOA, quinine HCl, and caffeine." Chemical senses 31(5): 403-413. Harrington, H.L., Kennedy, K.E., Sharafi, M., Hayes, J.E. and Duffy, V.B. (2012). "Otitis media exposure associates with dietary preference and adiposity: A community-based observational study of at-risk preschoolers." Physiology & behavior. Harris, H. and Kalmus, H. (1949). "The measurement of taste sensitivity to phenylthiourea (PTC)." Annals of Human Genetics 15(1): 24-31. Harwood, M.L., Ziegler, G.R. and Hayes, J.E. (2012). "Rejection thresholds in chocolate milk: Evidence for segmentation." Food Quality and Preference 26(1): 128-133. Hayes, J.E. (2008). "Transdisciplinary perspectives on sweetness." Chemosensory Perception 1(1): 48-57. Hayes, J.E., Allen, A.L. and Bennett, S.M. (2013). "Direct comparison of the generalized Visual Analog Scale (gVAS) and general Labeled Magnitude Scale (gLMS)." Food Quality and Preference. Hayes, J.E., Bartoshuk, L.M., Kidd, J.R. and Duffy, V.B. (2008). "Supertasting and PROP Bitterness Depends on More Than the TAS2R38 Gene." Chemical senses 33(3): 255-265. Hayes, J.E. and Duffy, V.B. (2007). "Revisiting sugar-fat mixtures: sweetness and creaminess vary with phenotypic markers of oral sensation." Chemical senses 32(3): 225-236. Hayes, J.E. and Keast, R.S.J. (2011). "Two decades of supertasting: Where do we stand?" Physiology & behavior 104(5): 1072-1074. Hayes, J.E. and Pickering, G.J. (2012). "Wine Expertise Predicts Taste Phenotype." American journal of enology and viticulture 63(1): 80-84. Hayes, J.E., Sullivan, B.S. and Duffy, V.B. (2010). "Explaining variability in sodium intake through oral sensory phenotype, salt sensation and liking." Physiology & behavior 100(4): 369-380. Hayes, J.E., Wallace, M.R., Knopik, V.S., Herbstman, D.M., Bartoshuk, L.M., et al. (2011). "Allelic variation in TAS2R bitter receptor genes associates with variation in sensations from and ingestive behaviors toward common bitter beverages in adults." Chem Senses 36(3): 311-319. Helgren, F.J. (1957). Sweetening compositions and method of producing the same. U.S. Patent No. 2,803,551. U.S. Patent and Trademark Office. Washington, DC. Helgren, F.J., Lynch, M.J. and Kirchmeyer, F. (1955). "A taste panel study of the saccharin “off- taste"." Journal of the American Pharmaceutical Association 44(6): 353-355. Heller, K.E., Burt, B.A. and Eklund, S.A. (2001). "Sugared soda consumption and dental caries in the United States." Journal of Dental Research 80(10): 1949-1953. Hellfritsch, C., Brockhoff, A., Staehler, F., Meyerhof, W. and Hofmann, T. (2012). "Human Psychometric and Taste Receptor Responses to Steviol Glycosides." Journal of Agricultural and Food Chemistry 60(27): 6782-6793.

113

Henkin, R. and Gillis, W. (1977). "Divergent taste responsiveness to fruit of the tree Antidesma bunius." Nature 265: 536-537. Hill, J.O. and Prentice, A.M. (1995). "Sugar and body weight regulation." The American journal of clinical nutrition 62(1): 264S-273S. Hinrichs, A.L., Wang, J.C., Bufe, B., Kwon, J.M., Budde, J., et al. (2006). "Functional Variant in a Bitter-Taste Receptor (hTAS2R16) Influences Risk of Alcohol Dependence." The American Journal of Human Genetics 78(1): 103-111. Hofmann, T. (2009). "Identification of the key bitter compounds in our daily diet is a prerequisite for the understanding of the hTAS2R gene polymorphisms affecting food choice." Annals of the New York Academy of Sciences 1170(1): 116-125. Horne, J., Lawless, H.T., Speirs, W. and Sposato, D. (2002). "Bitter taste of saccharin and acesulfame-K." Chemical senses 27(1): 31-38. Howard, B.V. and Wylie-Rosett, J. (2002). "Sugar and cardiovascular disease." Circulation 106(4): 523-527. Hu, F.B., Manson, J.a.E., Stampfer, M.J., Colditz, G., Liu, S., et al. (2001). "Diet, lifestyle, and the risk of type 2 diabetes mellitus in women." New England Journal of Medicine 345(11): 790-797. Ific (2011). Consumer Attitudes Toward Food Safety, Nutrition & Health. Food & Health Survey, International Food Information Council Foundation. Intranuovo, L.R. and Powers, A.S. (1998). "The Perceived Bitterness of Beer and 6-n- Propylthiouracil (PROP) Taste Sensitivity." Annals of the New York Academy of Sciences 855(1): 813-815. Jecfa (2010). Steviol glycosides. Compendium of Food Additive Specifications. 73th Meeting. Rome, FAO JECFA Monographs 10; FAO: 17-22. Joshipura, K.J., Hu, F.B., Manson, J.a.E., Stampfer, M.J., Rimm, E.B., et al. (2001). "The effect of fruit and vegetable intake on risk for coronary heart disease." Ann Intern Med 134(12): 1106-1114. Kamerud, J.K. and Delwiche, J.F. (2007). "Individual differences in perceived bitterness predict liking of sweeteners." Chemical senses 32(9): 803-810. Keast (2003). "A Psychophysical Investigation of Binary Bitter compound interactions." Keast, R.S.J. (2008). "Modification of the bitterness of caffeine." Food Quality and Preference 19(5): 465-472. Keast, R.S.J. and Roper, J. (2007). "A complex relationship among chemical concentration, detection threshold, and suprathreshold intensity of bitter compounds." Chemical senses 32(3): 245-253. Keller, K.L., Steinmann, L., Nurse, R.J. and Tepper, B.J. (2002). "Genetic taste sensitivity to 6-n- propylthiouracil influences food preference and reported intake in preschool children." Appetite 38(1): 3-12. Kim, U., Jorgenson, E., Coon, H., Leppert, M., Risch, N., et al. (2003). "Positional cloning of the human quantitative trait locus underlying taste sensitivity to phenylthiocarbamide." Science's STKE 299(5610): 1221. Kim, U., Wooding, S., Ricci, D., Jorde, L.B. and Drayna, D. (2005). "Worldwide haplotype diversity and coding sequence variation at human bitter taste receptor loci." Human mutation 26(3): 199-204. Kuhn, C., Bufe, B., Winnig, M., Hofmann, T., Frank, O., et al. (2004). "Bitter taste receptors for saccharin and acesulfame K." The Journal of neuroscience 24(45): 10260-10265. Lanier, S.A., Hayes, J.E. and Duffy, V.B. (2005). "Sweet and bitter tastes of alcoholic beverages mediate alcohol intake in of-age undergraduates." Physiology & behavior 83(5): 821-831.

114

Lawless, H.P., Richard Kroll, Beverley (2010). "A comparison of the labeled magnitude (LAM) scale, and 11-point category scale and the traditional 9-point hedonic scale." Food Quality and Preference 21(1): 4-12. Lawless, H.T. (1979). "Evidence for neural inhibition in bittersweet taste mixtures." Journal of comparative and physiological psychology 93(3): 538. Lawless, H.T. (1996). Flavor. Cognitive Ecology. M.P. Friedman, Carterette, E.C. San Diego, CA, Academic Press: 325-374. Lemon, C.H., Brasser, S.M. and Smith, D.V. (2004). "Alcohol activates a sucrose-responsive gustatory neural pathway." Journal of neurophysiology 92(1): 536-544. Lewis, W.H. (1992). "Early uses of Stevia rebaudiana (Asteraceae) leaves as a sweetener in Paraguay." Economic botany 46(3): 336-337. Lobstein, T., Baur, L. and Uauy, R. (2004). "Obesity in children and young people: a crisis in public health." Obesity reviews 5(s1): 4-85. Lotsch, J., Geisslinger, G. and Tegeder, I. (2009). "Genetic modulation of the pharmacological treatment of pain." Pharmacology & therapeutics 124(2): 168-184. Lucas, L., Riddell, L., Liem, G., Whitelock, S. and Keast, R. (2011). "The influence of sodium on liking and consumption of salty food." Journal of food science 76(1): S72-S76. Ludwig, D.S., Peterson, K.E. and Gortmaker, S.L. (2001). "Relation between consumption of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis." The Lancet 357(9255): 505-508. Malik, V.S., Schulze, M.B. and Hu, F.B. (2006). "Intake of sugar-sweetened beverages and weight gain: a systematic review." The American journal of clinical nutrition 84(2): 274- 288. Margolskee, R.F. (2002). "Molecular mechanisms of bitter and sweet taste transduction." Journal of Biological Chemistry 277(1): 1-4. Marks, L.E., Stevens, J.C., Bartoshuk, L.M., Gent, J.F., Rifkin, B., et al. (1988). "Magnitude- matching: the measurement of taste and smell." Chemical senses 13(1): 63. Marshall, T.A., Levy, S.M., Broffitt, B., Warren, J.J., Eichenberger-Gilmore, J.M., et al. (2003). "Dental caries and beverage consumption in young children." Pediatrics 112(3): e184- e191. Martin, S. and Pangborn, R.M. (1970). "Taste interaction of ethyl alcohol with sweet, salty, sour and bitter compounds." Journal of the Science of Food and Agriculture 21(12): 653-655. Matsunami, H., Montmayeur, J.P. and Buck, L.B. (2000). "A family of candidate taste receptors in human and mouse." Nature 404(6778): 601-604. Mattes, R.D. (1994). "Influences on acceptance of bitter foods and beverages." Physiology & behavior 56(6): 1229-1236. Mattes, R.D. and Dimeglio, D. (2001). "Ethanol perception and ingestion." Physiology & behavior 72(1-2): 217-229. Mattes, R.D. and Popkin, B.M. (2009). "Nonnutritive sweetener consumption in humans: effects on appetite and food intake and their putative mechanisms." American Journal of Clinical Nutrition 89(1): 1-14. Mazur, R.H. (1976). "Aspartame-a sweet surprise." Journal of Toxicology and Environmental Health, Part A Current Issues 2(1): 243-249. Mcburney, D. and Collings, V. (1977). "Introduction to sensation/perception." Mennella, J., Pepino, M.Y., Duke, F. and Reed, D. (2010). "Age modifies the genotype- phenotype relationship for the bitter receptor TAS2R38." BMC genetics 11(1): 60. Mennella, J.A. and Beauchamp, G.K. (1998). "Early flavor experiences: research update." Nutrition Reviews 56(7): 205-211.

115

Mennella, J.A., Pepino, M.Y., Duke, F.F. and Reed, D.R. (2011). "Psychophysical dissection of genotype effects on human bitter perception." Chemical senses 36(2): 161-167. Mennella, J.A., Pepino, M.Y. and Reed, D.R. (2005). "Genetic and environmental determinants of bitter perception and sweet preferences." Pediatrics 115(2): e216-e222. Meyerhof, W. (2005). "Elucidation of mammalian bitter taste." Reviews of physiology, biochemistry and pharmacology: 37-72. Meyerhof, W., Batram, C., Kuhn, C., Brockhoff, A., Chudoba, E., et al. (2010). "The molecular receptive ranges of human TAS2R bitter taste receptors." Chemical senses 35(2): 157- 170. Miller, I.J. and Reedy, F.E. (1990). "Variations in Human Taste Bud Density and Taste Intensity Perception." Physiology & behavior 47(6): 1213-1219. Mintel (2010). Sugar and Sweeteners- US- December, Mintel Reports database. Mintel (2012). Attitudes toward Beverage Sweeteners - US - March Mintel Reports database. Mintel (2012). Beer - US - December, Mintel Reports database. Mintel (2012). Carbonated Soft Drinks - US - February, Mintel Reports database. Mintel (2012). Wine - US - October Mintel Reports database. Moir, M. (2000). "Hops: a millennium review." Journal of the American Society of Brewing Chemists 58(4): 131-146. Moskowitz, H.R. and Sidel, J.L. (1971). "Magnitude and hedonic scales of food acceptability." Journal of food science 36(4): 677-680. Mueller, K.L., Hoon, M.A., Erlenbach, I., Chandrashekar, J., Zuker, C.S., et al. (2005). "The receptors and coding logic for bitter taste." Nature 434(7030): 225-229. Mustavich, L.F., Miller, P., Kidd, K.K. and Zhao, H. (2010). "Using a Pharmacokinetic Model to Relate an Individual's Susceptibility to Alcohol Dependence to Genotypes." Human Heredity 70(3): 177-193. Niaaa. (2005). "What Is a Standard Drink." A Pocket Guide for Alcoholc Screening and Brief Intervention, from http://pubs.niaaa.nih.gov/publications/Practitioner/pocketguide/pocket_guide2.htm. Nielsen, S.J., Siega-Riz, A.M. and Popkin, B.M. (2002). "Trends in energy intake in US between 1977 and 1996: similar shifts seen across age groups." Obesity 10(5): 370-378. Ogden, C.L., Carroll, M.D., Kit, B.K. and Flegal, K.M. (2012). Prevalence of Obesity in the United States, 2009-2010. , NCHS Data Brief 82: 1-8. Ohta, M., Sasa, S., Inoue, A., Tamai, T., Fujita, I., et al. (2010). "Characterization of novel steviol glycosides from leaves of Stevia rebaudiana Morita." Journal of Applied Glycoscience 57. Olsen, N.J. and Heitmann, B.L. (2009). "Intake of calorically sweetened beverages and obesity." Obesity reviews 10(1): 68-75. Ooi, S.X., Lee, P.L., Law, H.Y. and Say, Y.H. (2010). "Bitter receptor gene (TAS2R38) P49A genotypes and their associations with aversion to vegetables and sweet/fat foods in Malaysian subjects." Asia Pacific Journal of Clinical Nutrition 19(4): 491-498. Pereira, M.A. (2006). "The possible role of sugar-sweetened beverages in obesity etiology: a review of the evidence." International Journal of Obesity 30: S28-S36. Phillips, K. (1989). Stevia: steps in developing a new sweetener. Developments in sweeteners. T.H. Grenby, Elsevier Applied Science. 3: 1-43. Prescott, J., Soo, J., Campbell, H. and Roberts, C. (2004). "Responses of PROP taster groups to variations in sensory qualities within foods and beverages." Physiology & behavior 82(2- 3): 459-469.

116

Prescott, J. and Stevenson, R.J. (1996). "Psychophysical responses to single and multiple presentations of the oral irritant zingerone: relationship to frequency of chili consumption." Physiology & behavior 60(2): 617-624. Prescott, J. and Swain-Campbell, N. (2000). "Responses to repeated oral irritation by capsaicin, cinnamaldehyde and ethanol in PROP tasters and non-tasters." Chemical senses 25(3): 239. Prescott, J. and Tepper, B.J. (2004). Genetic variation in taste sensitivity, CRC. Pronin, A.N., Tang, H., Connor, J. and Keung, W. (2004). "Identification of ligands for two human bitter T2R receptors." Chemical senses 29(7): 583-593. Pronin, A.N., Xu, H., Tang, H., Zhang, L., Li, Q., et al. (2007). "Specific alleles of bitter receptor genes influence human sensitivity to the bitterness of aloin and saccharin." Current Biology 17(16): 1403-1408. Raben, A., Vasilaras, T.H., Moller, A.C. and Astrup, A. (2002). "Sucrose compared with artificial sweeteners: different effects on ad libitum food intake and body weight after 10 wk of supplementation in overweight subjects." The American journal of clinical nutrition 76(4): 721-729. Reed, D.R., Zhu, G., Breslin, P.a.S., Duke, F.F., Henders, A.K., et al. (2010). "The perception of quinine taste intensity is associated with common genetic variants in a bitter receptor cluster on chromosome 12." Human molecular genetics 19(21): 4278-4285. Reedy, F., Bartoshuk, L.M., Miller, I., Duffy, V., Lucchina, L., et al. (1993). "Relationships among papillae, taste pores, and 6-n-propylthiouracil (PROP) suprathreshold taste sensitivity." Chem. Senses 18: 618-619. Riboli, E. and Norat, T. (2003). "Epidemiologic evidence of the protective effect of fruit and vegetables on cancer risk." The American journal of clinical nutrition 78(3): 559S-569S. Rifkin, B. and Bartoshuk, L.M. (1980). "Taste synergism between monosodium glutamate and disodium 5'-guanylate." Physiology & behavior 24(6): 1169-1172. Robichaud, J.L. and Noble, A.C. (1990). "Astringency and bitterness of selected phenolics in wine." Journal of the Science of Food and Agriculture 53(3): 343-353. Roudnitzky, N., Bufe, B., Thalmann, S., Kuhn, C., Gunn, H.C., et al. (2011). "Genomic, genetic and functional dissection of bitter taste responses to artificial sweeteners." Human molecular genetics 20(17): 3437-3449. Rousseff, R.L., Ed. (1990). Bitterness in foods and beverages. Developments in food science. Amsterdam, Elsevier Roy, G.M. (1990). "The applications and future implications of bitterness reduction and inhibition in food products." Critical Reviews in Food Science & Nutrition 29(2): 59-71. Rozin, P. and Vollmecke, T.A. (1986). "Food likes and dislikes." Annual review of nutrition 6(1): 433-456. Sacerdote, C., Guarrera, S., Smith, G.D., Grioni, S., Krogh, V., et al. (2007). "Lactase persistence and bitter taste response: instrumental variables and mendelian randomization in epidemiologic studies of dietary factors and cancer risk." American journal of epidemiology 166(5): 576-581. Sandell, M.A. and Breslin, P.A. (2006). "Variability in a taste-receptor gene determines whether we taste toxins in food." Current Biology 16(18): 792. Sandrou, D. and Arvanitoyannis, I. (2000). "Low-fat/calorie foods: Current state and perspectives." Critical reviews in food science and nutrition 40(5): 427-447. Schiffman, S.S., Booth, B., Carr, B., Losee, M., Sattely-Miller, E., et al. (1995). "Investigation of synergism in binary mixtures of sweeteners." Brain research bulletin 38(2): 105-120. Schiffman, S.S., Booth, B.J., Losee, M.L., Pecore, S.D. and Warwick, Z.S. (1995). "Bitterness of sweeteners as a function of concentration." Brain research bulletin 36(5): 505-513.

117

Schiffman, S.S., Reilly, D.A. and Clark, T.B. (1979). "Qualitative differences among sweeteners." Physiology & behavior 23(1): 1-9. Schulze, M.B., Manson, J.A., Ludwig, D.S., Colditz, G.A., Stampfer, M.J., et al. (2004). "Sugar- sweetened beverages, weight gain, and incidence of type 2 diabetes in young and middle- aged women." JAMA: the journal of the American Medical Association 292(8): 927-934. Scinska, A., Koros, E., Habrat, B., Kukwa, A., Kostowski, W., et al. (2000). "Bitter and sweet components of ethanol taste in humans." Drug and alcohol dependence 60(2): 199-206. Sharafi, M., Hayes, J.E. and Duffy, V.B. (2012). "Masking Vegetable Bitterness to Improve Palatability Depends on Vegetable Type and Taste Phenotype." Chemosensory Perception 6(1): 8-19. Small, D.M. and Prescott, J. (2005). "Odor/taste integration and the perception of flavor." Experimental Brain Research 166(3): 345-357. Smith, T.J. and Yang, C.S. (1994). Effects of food phytochemicals on xenobiotic metabolism and tumorigenesis. ACS Symposium Series, ACS Publications. Snyder, D.J., Prescott, J. and Bartoshuk, L. (2006). Modern Psychophysics and the Assesment of Human Oral Sensation. Taste and Smell. H. T.D. and A.B., W.-L., S. Karger AG. 63: 221-241. Song, W.O., Wang, Y., Chung, C.E., Song, B., Lee, W., et al. (2012). "Is obesity development associated with dietary sugar intake in the US?" Nutrition 28(11-12): 1137-1141. Soranzo, N., Bufe, B., Sabeti, P.C., Wilson, J.F., Weale, M.E., et al. (2005). "Positive selection on a high-sensitivity allele of the human bitter-taste receptor TAS2R16." Current Biology 15(14): 1257-1265. Steiner, J.E., Glaser, D., Hawilo, M.E. and Berridge, K.C. (2001). "Comparative expression of hedonic impact: affective reactions to taste by human infants and other primates." Neuroscience & Biobehavioral Reviews 25(1): 53-74. Steinmetz, K.A. and Potter, J.D. (1991). "Vegetables, fruit, and cancer. II. Mechanisms." Cancer Causes and Control 2(6): 427-442. Stellman, S.D. and Garfinkel, L. (1986). "Artificial sweetener use and one-year weight change among women." Preventive medicine 15(2): 195-202. Stephens, M. and Donnelly, P. (2003). "A comparison of bayesian methods for haplotype reconstruction from population genotype data." The American Journal of Human Genetics 73(5): 1162-1169. Stephens, M., Smith, N.J. and Donnelly, P. (2001). "A new statistical method for haplotype reconstruction from population data." The American Journal of Human Genetics 68(4): 978-989. Stevens, J. and Marks, L. (1980). "Cross-modality matching functions generated by magnitude estimation." Attention, Perception, & Psychophysics 27(5): 379-389. Stevens, S. (1971). "Issues in psychophysical measurement." Psychological Review; Psychological Review 78(5): 426. Stevens, S.S. (1946). On the theory of scales of measurement, Bobbs-Merrill, College Division. Suomela, J.P., Sandell, M., Vaarno, J., Lehtonen, H.M., Tahvonen, R., et al. (2012). "Children’s Hedonic Response to Berry Products: Effect of Chemical Composition of Berries and hTAS2R38 Genotype on Liking." Food chemistry. Swann, L. (2005). "Shaping sweet success for reduced-sugar foods." Prepared Foods 174(4): 85- 86. Sylvetsky, A.C., Welsh, J.A., Brown, R.J. and Vos, M.B. (2012). "Low-calorie sweetener consumption is increasing in the United States." American Journal of Clinical Nutrition 96(3): 640-646.

118

Tan, J., Abrol, R., Trzaskowski, B. and Goddard Iii, W.A. (2012). "3D Structure Prediction of TAS2R38 Bitter Receptors Bound to Agonists Phenylthiocarbamide (PTC) and 6-n- Propylthiouracil (PROP)." Journal of chemical information and modeling 52(7): 1875- 1885. Tepper, B.J. (1998). "6-n-Propylthiouracil: a genetic marker for taste, with implications for food preference and dietary habits." American journal of human genetics 63(5): 1271. Tepper, B.J. (2008). "Nutritional implications of genetic taste variation: the role of PROP sensitivity and other taste phenotypes." Annu. Rev. Nutr. 28: 367-388. Tepper, B.J., Christensen, C.M. and Cao, J. (2001). "Development of brief methods to classify individuals by PROP taster status." Physiology & behavior 73(4): 571-577. Tepper, B.J., Keller, K.L. and Ullrich, N.V. (2003). Genetic variation in taste and preferences for bitter and pungent foods: implications for chronic disease risk. ACS Symposium Series ACS Publications. Tepper, B.J., Koelliker, Y., Zhao, L., Ullrich, N.V., Lanzara, C., et al. (2008). "Variation in the bitter-taste receptor gene TAS2R38, and adiposity in a genetically isolated population in Southern Italy." Obesity 16(10): 2289-2295. Tepper, B.J. and Ullrich, N.V. (2002). "Influence of genetic taste sensitivity to 6-n- propylthiouracil (PROP), dietary restraint and disinhibition on body mass index in middle-aged women." Physiology & behavior 75(3): 305-312. Thorngate, J.H. (1997). "The physiology of human sensory response to wine: A review." American journal of enology and viticulture 48(3): 271-279. Tordoff, M.G. and Alleva, A.M. (1990). "Effect of drinking soda sweetened with aspartame or high-fructose corn syrup on food intake and body weight." The American journal of clinical nutrition 51(6): 963-969. Vetsch, W. (1985). "Aspartame: technical considerations and predicted use." Food chemistry 16(3): 245-258. Von Rymon Lipinski, G.W. (1985). "The new intense sweetener Acesulfame K." Food chemistry 16(3-4): 259-269. Wang, J.C., Hinrichs, A.L., Bertelsen, S., Stock, H., Budde, J.P., et al. (2007). "Functional Variants in TAS2R38 and TAS2R16 Influence Alcohol Consumption in High-Risk Families of African-American Origin." Alcoholism: Clinical and Experimental Research 31(2): 209-215. Wang, M.F., Yamamoto, S., Chung, H.M., Chung, S.Y., Miyatani, S., et al. (1995). "Antihypercholesterolemic effect of undigested fraction of soybean protein in young female volunteers." Journal of nutritional science and vitaminology 41(2): 187. Weir, B. (2008). "Linkage disequilibrium and association mapping." Annu. Rev. Genomics Hum. Genet. 9: 129-142. Wetherill, G. and Levitt, H. (1965). "Sequential estimation of points on a psychometric function." British Journal of Mathematical and Statistical Psychology 18(1): 1-10. Who (2003). Diet, nutrition and the prevention of chronic diseases: report of a Joint WHO/FAO Expert Consultation. W.H. Organization. Geneva, Switzerland, World Health Organization: 54-71. Wiener, A.S., M. Levit, A. Niv., M. . (2011). "BitterBD: a database of bitter compounds." from http://bitterdb.agri.huji.ac.il/bitterdb/. Wölwer-Rieck, U. (2012). "The leaves of Stevia rebaudiana (Bertoni), their constituents and the analyses thereof: a review." Journal of Agricultural and Food Chemistry 60(4): 886-895. Wong, W.W., Smith, E., Stuff, J.E., Hachey, D.L., Heird, W.C., et al. (1998). "Cholesterol- lowering effect of soy protein in normocholesterolemic and hypercholesterolemic men." The American journal of clinical nutrition 68(6): 1385S-1389S.

119

Wooding, S., Bufe, B., Grassi, C., Howard, M.T., Stone, A.C., et al. (2006). "Independent evolution of bitter-taste sensitivity in humans and chimpanzees." Nature 440(7086): 930- 934. Wooding, S., Gunn, H., Ramos, P., Thalmann, S., Xing, C., et al. (2010). "Genetics and bitter taste responses to goitrin, a plant toxin found in vegetables." Chemical senses 35(8): 685- 692. Wooding, S., Kim, U., Bamshad, M.J., Larsen, J., Jorde, L.B., et al. (2004). "Natural Selection and Molecular Evolution in PTC, a Bitter-Taste Receptor Gene." The American Journal of Human Genetics 74(4): 637-646. Yang, Q. (2010). "Gain weight by "going diet?" Artificial sweeteners and the neurobiology of sugar cravings: Neuroscience 2010." The Yale journal of biology and medicine 83(2): 101. Zhao, L.Q. and Tepper, B.J. (2007). "Perception and acceptance of selected high-intensity sweeteners and blends in model soft drinks by propylthiouracil (PROP) non-tasters and super-tasters." Food Quality and Preference 18(3): 531-540. Zuniga, J.R., Davis, S.H., Englehardt, R.A., Miller, I.J., Schiffrman, S.S., et al. (1993). "Taste performance on the anterior human tongue varles with fungiform taste bud density." Chemical senses 18(5): 449-460.

120 Appendix A

Supplemental data: Chapter 2

Figure A-1: Effect of the Ala227Val TAS2R31 polymorphism on the bitterness and sweetness of AceK and bitterness of PROP. The bitterness of AceK was significantly different across genotype; no effect was observed for AceK bitterness or PROP bitterness (p-values provided in text). Adjectives refer to semantic labels on the general Labeled Magnitude Scale (gLMS; see text for details). BD refers to ‘barely detectable’