Rats are unable to discriminate quinine from diverse bitter stimuli

Abstract

Compounds described by humans as “bitter” are sensed by a family of type 2 taste receptors (T2Rs). Previous work suggested that diverse bitter stimuli activate distinct receptors, which might allow for perceptually distinct tastes. Alternatively, it has been shown that multiple T2Rs are expressed on the same taste cell, leading to the contrary suggestion that these stimuli produce a unitary perception. Behavioral work done to address this in rodent models is limited to Spector and Kopka (Spector AC, Kopka SL. J Neurosci 22: 1937–1941, 2002), who demonstrated that rats cannot discriminate quinine from denatonium. Supporting this finding, it has been shown that quinine and denatonium activate overlapping T2Rs and neurons in both the mouse and rat nucleus of the solitary tract (NTS). However, cycloheximide and 6-n-propylthiouracil (PROP) do not appear to overlap with quinine in the NTS, suggesting that these stimuli may be discriminable from quinine and the denatonium/quinine comparison is not generalizable. Using the same procedure as Spector and Kopka, we tasked animals with discriminating a range of stimuli (denatonium, cycloheximide, PROP, and sucrose octaacetate) from quinine. We replicated and expanded the findings of Spector and Kopka; rats could not discriminate quinine from denatonium, cycloheximide, or PROP. Rats showed a very weak ability to discriminate between quinine and sucrose octaacetate. All animals succeeded in discriminating quinine from KCl, demonstrating they were capable of the task. These data suggest that rats cannot discriminate this suite of stimuli, although they appear distinct by physiological measures.

INTRODUCTION

Humans describe a class of compounds that activate the type 2 taste receptor cell family (T2Rs) as bitter. The T2R family consists of ~25 receptors (4), which are collectively capable of detecting hundreds of compounds, which we will refer to as “bitter,” although we acknowledge the percept may be different in rodents and humans. Animals and humans tend to avoid these compounds, and their avoidance is often framed as an evolutionary adaptation to avoid stimuli associated with toxicity (19, 33). These compounds are often secondary metabolites produced by plants to deter herbivory and can lack structural similarity (3, 9–11, 31) and, although they activate the T2R receptors, not all are toxic to mammals (19). It has been proposed that discriminating between bitter stimuli would allow animals to distinguish nontoxic plants from toxic ones, but it has also been proposed that bitterness, as a unitary percept, acts as a warning system to reduce the likelihood of consuming a toxin and that an animal may use other cues (e.g., olfactory) to identify safe forage.

Early work showed that multiple types of T2Rs are coexpressed in single taste receptor cells (1, 29), suggesting that even though each T2R responds to a narrow range of stimuli, a single cell could be responsive to many stimuli, implying that animals would be unable to discriminate one bitter stimulus from another. However, calcium responses in individual taste receptor cells showed that each of these cells respond to only one or two (of five) commonly used bitter tastants (quinine, cycloheximide, denatonium, sucrose octaacetate, and phenylthiocarbamide) (12). Additionally, Dahl et al. (15) showed that single neurons in the glossopharyngeal and chorda tympani nerves differed in their sensitivities to a wide range of bitter stimuli (quinine, denatonium, nicotine, strychnine, caffeine, yohimbine, and phenylthiocarbamide). Faced with these conflicting data, Spector and Kopka (36) used a behavioral paradigm that compared multiple concentrations of each stimulus to show that rats were unable to discriminate quinine HCl from denatonium benzoate when the intensity of the stimulus was rendered irrelevant to task performance, implying that the separation found at initial stimulus processing was not conserved at the behavioral level.

It was later demonstrated that quinine and denatonium produced very similar response patterns in both the mouse and rat nucleus of the solitary tract (NTS), but these responses were separable from responses to cycloheximide and sucrose octaacetate (SOA) (14, 17, 18, 25, 41). Brief-access licking assays have also supported a similarity between responses to quinine and denatonium in rats, while licking to cycloheximide is correlated weakly or not at all with licking to either quinine or denatonium (8). While these findings support the conclusion of Spector and Kopka (36) that quinine and denatonium are indiscriminable for the rat, they highlight the need to extend the stimuli tested. Here we use the same behavioral method as in Spector and Kopka (36). We tasked rats with discriminating multiple concentrations of quinine from denatonium, cycloheximide, 6-n-propylthiouracil (PROP), or SOA to help clarify the relative heterogeneity of bitter taste perception in rats.

MATERIALS AND METHODS

Subjects

Subjects were 24 adult male Long Evans rats (Charles River Breeding Laboratory, Raleigh, NC), weighing 150–250 g at study onset. The colony room was maintained at 20 ± 2°C with a 12:12-h light-dark cycle. All training and testing were performed during the lights on phase. All animal procedures were approved by University at Buffalo Animal Care and Use Committees.

Water Restriction

For brief-access licking, training took place during a single week, during which animals were water restricted. Following this, animals were water restricted 22 h before testing, which took place on Monday, Wednesday, and Friday of each week. Discrimination training and testing took place 5 days/wk (Monday-Friday). Water was removed from cages on Sunday so animals were water deprived for 22 h before testing on Monday, and water bottles were returned on Friday after testing. This way, animals were allowed ad libitum access to water for 48 h in between training/testing sessions. During the week, animals worked for all of their water during the test sessions.

Brief-Access Licking

Experimental design.

To determine the range of concentrations that would be isointense for each stimulus, we recorded the licking responses in male rats (n = 8/group) to varying concentrations of quinine, denatonium benzoate, cycloheximide, PROP, and SOA in a brief-access test. The first group of rats were tested on quinine, denatonium, and cycloheximide, while the second group was tested on PROP and SOA. Lick ratios for each stimulus were calculated, and curves were fit to the lick ratios of each stimulus using a three-parameter logistic function. From this function, three concentrations of each stimulus were chosen for later discrimination testing (see Brief-Access Licking, Statistical analysis).

Testing apparatus.

For the brief-access licking test, we used a chamber designed to measure the immediate licking response to taste stimuli (Davis MS80 Rig; Dilog Instruments and Systems, Tallahassee, FL). This apparatus is described in detail elsewhere (26). Briefly, it is a Plexiglas cage with a wire mesh floor. An opening at the front of the cage allowed access to one of 16 drinking tubes mounted on a sliding platform. A mechanical shutter opened and closed to allow the rat access to one of the tubes for a user-specified length of time. A computer controlled the movement of the platform, order of tube presentation, opening and closing of the shutter, duration of tube access, and interval between tube presentations. Each individual lick was detected by a contact lickometer and recorded on a computer via DavisPro collection software (Dilog Instruments and Systems).

Training procedure.

Rats were adapted to the test chamber and trained to drink from the sipper tubes for 5 days. Throughout training and testing, rats were 20 h water deprived. On the first day of training, the rat was presented with a single stationary bottle of 0.25 M sucrose for 30 min. On the second day, a single tube containing 0.25 M sucrose was presented again, and the rat was given 180 s to initiate licking. Once licking was recorded, the rat was given 30-s access to the tube. At the conclusion of either the 30-s access or the 180-s limit, the shutter was closed again for 10 s. On the third through fifth day of training, rats were given access to a range of concentrations of sucrose. The program began only when the rat initiated licking, after which the animal was given 10 s access to each tube. After the 10 s, the shutter was closed for 10 s and a new tube was presented, which remained until the rat licked.

Testing procedure.

During testing, animals were allowed to take as many trials as possible in 30 min. Rats were tested on varying concentrations of quinine (0, 0.003, 0.01, 0.03, 0.1, 0.3, and 1.3 mM) presented in repeated randomized blocks on Monday, Wednesday, and Friday in a single week. Following this, rats were tested on varying concentrations of denatonium benzoate (0, 0.01, 0.03, 0.1, 0.3, 1, 3, and 10 mM) and then cycloheximide (0, 0.1, 0.2, 0.3, 1, 2, 3 µM). A second group of rats was trained as described above and then tested on varying concentrations of PROP (0, 0.01, 0.03, 0.1, 0.3, 1, 3, and 7 mM) and SOA (0, 0.5, 1, 1.5, 2, 3, 4, and 5 mM). Once the animals began licking the tube, they were allowed 10 s of access before the shutter closed. There was a 10 s intertrial interval before rats were then given one 1 s of licking access to water, which served as a rinse, before the shutter closed again and reopened 10 s later on another tube.

Statistical analysis.

Lick ratios were calculated for brief-access licking by dividing the average number of licks at each concentration by the average number of licks to water. This correction is meant to control for motivational differences between individuals. Lick ratios for each stimulus were averaged, and curves were fit to the lick ratios of each stimulus using a three-parameter logistic function:

$$f\left(x\right)=\frac{a}{{\left[1+10\left(log10x-c\right)\right]}^{*b}}$$

in which the curve parameters a, b, and c, were defined as the asymptotic maximum performance, the slope of the curve, and the midpoint concentration between asymptotic maximum and minimum (EC₅₀), respectively. To determine isointense concentrations of each taste stimulus, we determined the taste ratio (licks/licks of water) for 0.1, 0.3, and 1.0 mM quinine and 0.352, 0.879, and 2.512 mM denatonium. These concentrations were defined as isointense by Spector and Kopka using the same method (36). The lick ratios were 0.74, 0.52, and 0.28, respectively. We then calculated which concentrations of each of the other stimuli animals responded to with the same lick ratios (see Fig. 1). The exception to this was SOA, where we saw variable responding to the stimulus (see results). In this case, only animals that licked SOA differently from water were used to calculate isointense concentrations. Curve fitting was conducted in Systat 12.

Fig. 1.

Averaged lick data (±SE) and concentration response functions for the five bitter tastants used in the discrimination task. A: quinine. B: denatonium. C: cycloheximide. D: 6-n-propylthiouracil (PROP). E: sucrose octaacetate (SOA). Rats (n = 8/group) were given 10-s access to varying concentrations of each stimulus to minimize postingestive effects on licking. Lick ratios were found by dividing averaged licks for each concentration by averaged licks to water. Curves were fit to the averaged data with a logistic function. Isointense concentrations for cycloheximide, PROP, and SOA were found by comparing lick ratios for 0.1, 0.3, and 1.0 mM quinine/0.352, 0.879, and 2.512 mM denatonium as shown in Spector and Kopka (36; lick ratios: 0.74, 0.52, and 0.28). Drop-down lines indicate these concentrations for each stimulus.

Discrimination

Experimental design.

For discrimination testing, 18 male rats were divided into three groups (n = 6/group). The first group was trained to discriminate quinine from KCl, after which they were taken through a series of discrimination tasks where the training solutions were replaced with cycloheximide or NaCl. The second group of rats was subjected to discrimination training, where the training solutions were quinine and cycloheximide, to determine if rats were capable of learning to discriminate these solutions. The third group was subjected to the same procedure as the second; however, the training solutions were quinine and denatonium; this was done to replicate the results of Spector and Kopka (36). Following the failure of the second and third groups to acquire quinine versus cycloheximide or quinine versus denatonium discrimination, these animals were trained to discriminate quinine from KCl. After confirmation that the animals could learn the discrimination task, all three discrimination groups were tested on quinine versus PROP and then quinine versus SOA. Lastly, all groups were assigned a discrimination pairing (group 1: quinine versus denatonium; group 2: quinine versus PROP; and group 3, quinine versus cycloheximide) and tested again.

Testing apparatus.

Discrimination training and testing were conducted in a specialized stimulus delivery system that we refer to as the gustometer. The gustometer, which has been previously described and used to assess taste sensitivity in rats (2, 5, 16, 28, 34, 36, 38), is a modified skinner box housed in a sound attenuation chamber. Animals are placed in a small, plastic, wire-bottom chamber. The front panel of the chamber has three slots where the animal can access three response balls: the sample (center) ball is a borosilicate ball that spins on a horizontal axis, and the two reinforcement (left and right) balls are stationary. Solution is dispensed onto the sample ball, from one of 14 tubes, which are driven by computerized syringe pumps. The sample ball is attached to a computer-controlled retractable arm. After the sampling period, the arm is retracted and the ball is washed with deionized (DI) water, dried with pressurized air, and repositioned. Water is dispensed from PE tubing run through the two stationary reinforcement balls, available at the left and right slots. Licks on all balls are recorded by force transducers. To limit any sounds that could be used by the animals as a nontaste cue (turret shifting, unique pump clicking, etc.), white noise is played during all training and testing sessions. A stainless steel shield is positioned between the animals and the stimulus delivery system to limit visual cues. An exhaust fan is positioned directly above the sample ball to minimize olfactory cues.

Trial structure.

Training and testing methods were based on Bales et al. (2) and Blonde et al. (5). Discrimination was conducted using a two-response operant task. The animal was required to sample stimulus solution and respond on the assigned reinforcement ball, according to the stimulus presented, e.g., if the animal is presented with quinine, it may be trained to respond to the right reinforcement ball, and if the animal is presented with KCl, it may be trained to respond to the left reinforcement ball. Within test sessions, two stimuli were presented (e.g., quinine and KCl) and the concentration of each stimulus was varied so that stimulus concentration could not be used as a cue. Animals were water restricted for all training and testing; water was only acquired via stimulus sampling and reinforcement for correct responses in the task. Sessions lasted 30 min.

At the beginning of the session, the house light would turn on and the sample ball was washed and then positioned in front of its slot. The rat could initiate a trial by licking the sample ball twice within 250 ms, after which it would dispense sample solution. During the sample phase, rats were allowed to sample either 3 s or five licks (5 µl/lick) of solution. The sample ball was then retracted, and the rat was given a specified amount of time to respond at a reinforcement ball. The time allowed to respond at the reinforcement balls was progressively shortened during training, with its final testing duration at 5 s. If the animal responded at the correct reinforcement ball, it was allowed 5 s or 25 licks of DI water. If the animal responded at the incorrect ball or took too long to respond, it was given no water and was punished with a 20-s time-out where it could not initiate a new trial. The intertrial interval was 6 s.

Training and testing procedure.

Animals were trained to associate the presence of a stimulus (e.g., quinine) on the sample ball with water delivery on one of the reinforcement balls (stimulus 1 paired side), alternatively, the presence of a second stimulus (e.g., KCl) on the sample ball was associated with water being delivered on the opposite reinforcement ball (stimulus 2 paired side). This task required a water-restricted rat to respond to the presence of the stimulus 1 by licking the reinforcement ball on the stimulus 1 paired side, and to respond to the other stimulus by licking on the stimulus 2 paired side. When trained rats respond at chance levels (50% correct) we consider the stimuli indiscriminable.

During first phase of training, stationary training (Table 1), animals learned that fluid could be received at the sample ball and at each of the reinforcement balls. For 1 day each, animals were allowed ad libitum access to water at the left, right, or center ball, while access to the other two balls was blocked by a shutter.

Table 1.

Discrimination training

Training Phase	Days
Stationary training	3
Side training	8
Alternate training	16
Random training	6
Discrimination I	6
Discrimination II	5
Discrimination III	14

List of phases included in discrimination training and the number of days required to achieve 80% or higher performance for group 1 animals.

During the second phase of training, side training (Table 1), animals were allowed access to one of two solutions at the center ball and access to one of the reinforcement balls, while the other was blocked with a shutter. Correct response assignments were counterbalanced, such that half the rats had quinine paired to right-side responses, and the other half had quinine paired with left-side responses. Animals were given four pairings of each of the stimuli with the assigned response side. During side training, animals were allowed 180 s to respond and were not punished with a time out, although they did not receive water for incorrect responses.

After side training, animals moved to alternation training (Table 1). Animals had access to the two taste stimuli during the same session and had access to both reinforcement balls. Animals were presented with the same sample repeatedly, until they made a predetermined number of correct responses (correct responses were not required to be consecutive), and then, the other fluid was presented. Animals started with a stimulus change after eight correct responses, which was lowered across sessions (to 6, then 4, 3, 2, and 1). Animals being trained on the quinine versus KCl task were advanced to the next response criterion when they reached 80% accuracy. Animals receiving the quinine versus denatonium or quinine versus cycloheximide training were unable to acquire the task, and as such failed to reach the required accuracy (80%) for moving forward during training. Because of this, all quinine versus denatonium and quinine versus cycloheximide rats underwent 18 sessions of alternate 8, five sessions of alternate 6, and three sessions each of alternate 4, 3, 2, and 1. During the alternation phase, animals were allowed 15 s to respond, and incorrect or nonresponses resulted in a 20-s time-out.

During random training (Table 1), sample fluids were presented randomly, such that animals had a 50% chance of receiving either taste stimulus (e.g., quinine or KCl) throughout the session. Quinine versus KCl animals were moved to discrimination training after they reached/maintained 80% accuracy on the task (average of 5 days), while the quinine versus denatonium and quinine versus cycloheximide animals did not reach 80% accuracy and were moved to discrimination training after 7 days. During the first phase of discrimination (discrimination I), the number of sample solutions was increased from two to six. Where animals were originally trained on a single concentration of a given stimulus, (e.g., 0.3 mM quinine and 0.3 M KCl), two additional concentrations of each stimulus were added (e.g., 0.1 and 1.0 mM quinine and 0.1 and 1.0 M KCl). This was done to render intensity of the solution irrelevant (see Table 2). Animals were still required to respond on the assigned side for all concentrations of a given stimulus. Animals being trained on the quinine versus denatonium or quinine versus cycloheximide task maintained 50% accuracy at discrimination I and did not move forward in the training paradigm. After quinine versus KCl animals reached and maintained 80% accuracy in the task, they moved to discrimination II, where the decision time duration was shortened from 15 to 10 s and then to discrimination III, where the sample volume was decreased from 10 licks to 5 licks.

Table 2.

Discrimination stimuli

Stimulus	Concentrations
Quinine	0.1, 0.3, and 1.0 mM
Denatonium benzoate	0.35, 0.88, and 2.51 mM
Cycloheximide	0.45, 1.15, and 2.09 µM
PROP	0.60, 1.59, and 4.46 mM
SOA	1.82, 3.16, and 5.0 mM
KCl	0.1, 0.3, and 1.0 M
NaCl	0.1, 0.3, and 1.0 M

List of stimuli and concentrations used in discrimination testing. PROP, 6-n-propylthiouracil; SOA, sucrose octaacetate.

After animals had achieved 80% accuracy on the task, they were taken through a series of discrimination tasks where the training solutions (quinine and KCl) were substituted (see Table 3). Briefly, we first substituted quinine with cycloheximide, making the discrimination pairing cycloheximide versus KCl. If quinine and cycloheximide were perceptually similar, this would not be expected to disrupt performance. We then substituted cycloheximide with NaCl (NaCl versus KCl). KCl is thought to have both “salty” and “bitter” qualities, where NaCl is only salty (35) and rats can discriminate them reliably. We expected that the disruption in their performance when adding NaCl would be greater than switching between two stimuli that are thought to be only bitter. Following 5 days of NaCl versus KCl discrimination, rats switched back to quinine vs KCl for 2 days before being subjected to a “water control test” (see below) to verify that stimulus control was maintained. Rats again discriminated quinine from KCl for two sessions before they were required to discriminate quinine from cycloheximide.

Table 3.

Discrimination testing

Group 1	Group 2	Group 3
Quinine vs. KCl training	Quinine vs. cycloheximide training	Quinine vs. denatonium training
Quinine vs. KCl	Quinine vs. KCl training	Quinine vs. KCl training
Cycloheximide vs. KCl	Quinine vs. KCl	Quinine vs. KCl
NaCl vs. KCl
Quinine vs. KCl refresh
Water control test
Quinine vs. KCl refresh
Quinine vs. cycloheximide
Quinine vs. KCl refresh
Quinine vs. PROP	Quinine vs. PROP	Quinine vs. PROP
Quinine vs. KCl refresh	Quinine vs. KCl refresh	Quinine vs. KCl refresh
Quinine vs. SOA	Quinine vs. SOA	Quinine vs. SOA
Quinine vs. KCl refresh	Quinine vs. KCl refresh	Quinine vs. KCl refresh
Water control test	Water control test	Water control test
Quinine vs. KCl refresh	Quinine vs. KCl refresh	Quinine vs. KCl refresh
Quinine vs. denatonium	Quinine vs. PROP	Quinine vs. cycloheximide

List of discrimination tests each group was required to complete. PROP, 6-n-propylthiouracil; SOA, sucrose octaacetate.

After completing training and/or testing, all rats from all groups were required to discriminate quinine from PROP for 5 days and then quinine from SOA for 10 days. Lastly, following a short retraining period, the three groups of rats were again separated, and each was required to discriminate quinine from 1) denatonium, 2) PROP, or 3) cycloheximide for an extended period (10 days).

Water control test.

During discrimination testing, animals were subjected (see Table 3) to a water control test where assignments of each syringe were maintained (e.g., quinine syringes were still labeled quinine in the program), but all solutions were replaced with DI water. If animals relied on taste to inform their decisions, then, without any taste information, performance on the water control test would be at ~50%. If, however, animals used other cues (visual, auditory, etc.), then their performance would appear more accurate.

Statistical analysis.

For discrimination testing, percent correct for each animal was calculated as the number of trials with correct responses divided by the total number of trials taken within a given test session. Trials with nonresponses were not included in the calculation. Rats initiated 60–80 trials per session. Performance values were compared with chance (50%) with one-sample t tests. Statistical comparisons were conducted in Systat 12.

SOA, postdiscrimination testing.

Discrimination-trained rats were tested in the brief-access paradigm. Following Davis rig training, which proceeded as described above, rats were tested on varying concentrations of SOA (0.5, 1, 1.5, 2, 3, 4, and 5 mM). For these animals, lick ratios were calculated, and rats who had lick ratios below 0.5 at the two highest concentrations of SOA were classified as avoiders, while those with lick ratios above 0.5 were classified as indifferent. Correlational analyses were performed, comparing performance on the SOA discrimination task to SOA Davis rig licking.

RESULTS

Brief-Access Licking

Rats tested in the Davis rig decreased licking to quinine, denatonium, cycloheximide, and PROP (Fig. 1). There was high variability in licking to SOA. Because of this variability, we classified rats as SOA-avoider and SOA-indifferent rats. From the lick curves, we chose three concentrations each of cycloheximide, PROP, and SOA that elicited the same degree of avoidance as the concentrations of quinine and denatonium previously used in Spector and Kopka (36). To determine these concentrations for SOA, we used the data from SOA-avoider rats only. These concentrations, and the concentrations of quinine and denatonium used in Spector and Kopka (36), were used in further discrimination testing.

Discrimination of Quinine from Cycloheximide

Group 1 rats were trained to discriminate quinine from KCl and acquired the task with ~90% accuracy [Fig. 2A, t(5) = 71.622, P < 0.001]. Once animals successfully acquired the quinine versus KCl task, we tested them on several stimulus pairings (Fig. 2B). First, animals completed 5 days of quinine versus KCl discrimination. Second, we substituted cycloheximide for quinine. Rats showed only a short-term (1 day) and mild (~10% decrease) disruption of their performance [Fig. 2B, t(5) 37.461–78.177, P < 0.001]. Third, animals were required to discriminate NaCl from KCl across five sessions and again showed a decrease in performance, but nonetheless achieved at least 80% accuracy or better during the 5 days of discrimination [t(5) 25.435–73.172, P < 0.001]. Fourth, animals were retrained on quinine versus KCl for 2 days. Fifth, rats were subjected to a water control test to ensure that stimulus control was maintained for all rats. No rat performed at significantly above 50% accuracy [t(5) 1.688, P < 0.152]. Sixth, we gave the animals two sessions of quinine versus KCl retraining. Seventh, rats were required to discriminate quinine from cycloheximide. Across 5 days, rats consistently performed at and around 50%, indicating that they were unable to discriminate the two solutions [t(5) = 0.517–2.149, P = 0.422–1.000].

Fig. 2.

Averaged performance data (±SE) for rats (n = 6) originally trained on quinine vs. KCl (group 1) discrimination. To conserve space, only the first (F) and last (L) day of each phase were shown, starting with the alternation (Alt) phase, criterion 8. All rats were included in all sessions, but each training phase did not necessarily require more than one day to reach satisfactory performance (80%). Performance values were compared with chance (50%) with one-sample t-tests. A: all animals were able to acquire the quinine (Q) vs. KCl discrimination task (DT). B: following quinine vs. KCl discrimination training, animals were required to discriminate several stimulus pairs, starting with the training stimuli (1; quinine vs. KCl) and then substituting cycloheximide (2; Cyc) for quinine (cycloheximide vs. KCl). This was followed with NaCl vs. KCl (3), quinine vs. KCl (4), a water control test (5), quinine vs. KCl (6), and finally, quinine vs. cycloheximide (7). Substitution of cycloheximide for quinine resulted in a minor disruption of performance, where substitution of NaCl for cycloheximide resulted in a larger, more extended decrease in performance. During the water control task (WCT), animals performed at chance (50%). Finally, animals were unable to discriminate quinine from cycloheximide.

Acquisition of Discrimination Task with Quinine and Cycloheximide or Quinine and Denatonium

Group 2 was tasked with discriminating quinine from cycloheximide (Fig. 3A). Although during the early stages of training (alternation), animals appeared able to complete the task, once the stimuli were presented randomly, the animals were unable to perform the task, suggesting that the early success was due to nontaste, or perhaps intensity, cues. Performance at the final stage of training (discrimination III) was at ~50% accuracy [t(5) = 1.314, P = 0.246]. To replicate the findings in Spector and Kopka (36), we also trained a group of animals to discriminate quinine from denatonium, designated group 3 (Fig. 3A). Consistent with their findings, we were unable to train animals to discriminate the two tastants. Animals performed with ~50% accuracy on the last day of training [t(5) = 1.305, P = 0.249].

Fig. 3.

Averaged performance data (±SE) for rats (n = 6/group) originally trained on quinine vs. cycloheximide (group 2) or quinine vs. denatonium (group 3) discrimination. To conserve space, only the first (F) and last (L) day of each phase were shown, starting with the alternation (Alt) phase, criterion 8. Performance values were compared with chance (50%) with one-sample t tests. A: as the alternation criterion was lowered, performance tended to decrease. At the first discrimination (DT) phase, all animals performed at chance (50%), meaning they were unable to discriminate quinine from their assigned bitter stimulus (cycloheximide or denatonium). B: both groups of animals were then trained to discriminate quinine from KCl. All rats were included in all sessions, but each training phase did not necessarily require more than 1 day to reach satisfactory performance (80%). All animals were able to acquire the quinine vs. KCl discrimination task.

Acquisition of the Discrimination Task with Quinine and KCl

To show that group 2 and 3 rats could acquire the discrimination task when the training stimuli were discriminable, we trained both groups to discriminate quinine from KCl (Fig. 3B). Animals successfully learned to discriminate with no apparent deficits. They performed at ~90% on the task by the end of training [group 2: t(5) = 73.076, P < 0.001; group 3: t(5) = 111.273; P < 0.001].

Discrimination of Quinine from PROP

After all groups of rats were consistently able to perform the discrimination task (quinine vs. KCl), they were tasked with discriminating quinine from PROP. Across 5 days, the animals were unable to discriminate the stimuli [t(17) = 1.634–2.208, P = 0.206–0.604].

Discrimination of Quinine from Cycloheximide, PROP, or Denatonium with Extended Exposure

To ensure that performance did not improve with extended experience comparing bitter tastants, we first reacquainted all groups with the quinine versus KCl task and then required them to discriminate quinine from denatonium (Fig. 4A, group 1), PROP (Fig. 4B, group 2), or cycloheximide (Fig. 4C, group 3) for 10 days. All animals maintained their inability to discriminate throughout the testing [t(17) = 0.339–2.498, P = 0.230–1.000].

Fig. 4.

Averaged performance data (±SE) for rats from all groups (n = 6/group), once satisfactorily trained to discriminate quinine from KCl. Performance values were compared with chance (50%) with one-sample t tests. To test the effects of extended experience with the discrimination task, all animals were subjected to quinine vs. KCl refresher training, after which they were divided into their original groups and assigned a bitter stimulus discrimination pair. A: group 1: quinine vs. denatonium. B: group 2: quinine vs. 6-n-propylthiouracil (PROP). C: group 3: quinine vs. cycloheximide. Across 10 discrimination sessions, animals showed no ability to discriminate any stimulus pair.

Discrimination of Quinine from SOA

Following a single day of quinine versus KCl discrimination, animals were then tested on quinine versus SOA discrimination (Fig. 5A). Performance on the task was variable and appeared to slightly improve within the first 5 days [t(17) = 1.403–4.740, P = 0.002–1.000], although the improvement tapered off from days 5 to 10. On days 5 to 10, rats performed with 60–65% accuracy [t(17) = 5.774–7.917, P < 0.001]. To confirm that this partial discrimination was not a loss of stimulus control, we subjected the rats to a water control test. All rats performed at ~50% [t(17) = 1.398, P = 0.180], indicating that the rats had not lost stimulus control during the quinine versus SOA task.

Fig. 5.

Averaged performance data (±SE), averaged lick data (±SE), and concentration response functions for rats from all groups (n = 18). Performance values were compared with chance (50%) with one-sample t tests, and compared with brief-access licking to sucrose octaacetate (SOA) via linear regression analysis. A: once satisfactorily trained to discriminate quinine from KCl, all animals were tested on quinine vs. SOA discrimination across 10 days. Rats exhibited a weak ability to discriminate quinine from SOA, as their performance values were different significantly different from 50%. B: to determine whether this ability was related to the variability in brief-access licking to SOA, we recorded Davis rig lick responses to SOA, and classified rats as SOA-avoider or SOA-indifferent rats. Discrimination performance in A is plotted according to this classification. SOA-avoider and SOA-indifferent rats did not perform differently on the discrimination task.

Following SOA discrimination, all rats were tested in the Davis rig on SOA. Rats were classified as SOA-avoider and SOA-indifferent rats to visualize possible phenotypic variance (Fig. 5B). Despite the variability in performance on both brief-access licking and discrimination of SOA, the measures were not significantly correlated (r = 0.210, P = 0.404); SOA-avoider and SOA-indifferent rats do not show any difference in discrimination ability [t(16) = 0.265, P = 0.794, Fig. 5A].

DISCUSSION

Taste research classifies stimuli according to large, general categories (bitter, sweet, salty, etc.), which are often assumed to share similar or identical percepts. Despite this, it is well known that taste qualities are not necessarily monoguesic; rats are able to discriminate NaCl from KCl (36), or maltose from sucrose (37), and there is a long-standing debate on the discriminability of various bitter stimuli. The source of this debate is a large body of conflicting evidence regarding stimulus distinctiveness at the molecular, neural, and behavioral levels. Although all approaches are required for a full understanding of bitter coding, only behavioral evidence includes the whole of gustatory processing, whereas receptor activation and nerve/brain signaling are upstream events.

Previously, only a single major study has directly compared the discriminability of bitter stimuli in the rat (36), reporting that quinine and denatonium are indiscriminable. However, evidence published following their original report suggested that quinine and denatonium might be perceptually indistinguishable due to their highly similar molecular (31) and neural signals (14, 17, 18, 41), and strong behavioral covariation (8), leaving the possibility that other bitter stimuli could still generate different perceptions. Here we used the same psychophysical method as in Spector and Kopka (36) to verify that rats are unable to discriminate quinine from a widely varying set of bitter stimuli (denatonium, cycloheximide, PROP, and SOA). With the caveat that we have tested a small number of compounds in a omnivorous laboratory animal, rats were unable to discriminate quinine from other bitter stimuli (denatonium, cycloheximide, or PROP) and showed no indication that there were any perceptible differences in taste quality, as performance did not improve from chance (50%).

Denatonium discrimination served to replicate the findings of Spector and Kopka (36). Rats, who were consistently able to complete the discrimination task with other stimuli (quinine vs. KCl), were unable to discriminate quinine from denatonium after 10 days. We attempted to train a different group of rats (group 3) to acquire the discrimination task with quinine and denatonium as the training stimuli, which they were also unable to do. With these two approaches, we support the original finding that quinine and denatonium are indiscriminable even when the animals are highly motivated to complete the task.

Cycloheximide discrimination was assessed in the same way as denatonium. These animals had extensive exposure to quinine and cycloheximide (58 days) but were still unable to acquire the discrimination task using these stimuli. We chose cycloheximide to extend the findings of Spector and Kopka (36) because cycloheximide does not elicit the same neuronal and behavioral responses as quinine. Single neuron responses and response patterns in the NTS (14, 17) are not similar between cycloheximide and quinine, and rats’ brief-access lick responses to quinine do not predict their responses to cycloheximide (8). Despite these dissimilarities in other measures, rats were unable to discriminate quinine from cycloheximide, supporting original supposition of Spector and Kopka (36) that although bitter stimuli may exhibit upstream signaling differences, this does not necessarily predict the capacity to distinguish the stimuli in a behavioral paradigm.

Rats exhibited some ability to discriminate quinine from SOA, though their performance was weak. We do not know what allows rats to partially discriminate the stimuli. SOA activates some of the same T2Rs as quinine, in both mouse and human models (27, 31), but NTS representation of SOA is unlike that of quinine (41). As our results with cycloheximide and PROP demonstrate, differences in upstream processing do not necessarily predict the output of the whole gustatory system. It is additionally worth noting that SOA is unlike our other stimuli in that rats were variably responsive to its orosensory properties. It has been previously shown in mice that orosensory responsivity to SOA is variably expressed (7) and under the control of the soa gene (6, 13, 22), expression of which appears to affect sensitivity to other aversive stimuli, including quinine (32, 39, 40), although no work has explored a similar variability in rats. However, rats’ brief-access licking to SOA was not correlated with discrimination performance, indicating that the acceptability of SOA, or any perceived difference in its intensity (e.g., SOA-avoider or SOA-indifferent status), did not underlie the ability discriminate SOA from quinine. We were surprised to find these measures unrelated, as the simplest explanation would have been that SOA-indifferent rats did not find SOA bitter, but this was not the case. SOA-avoider and SOA-indifferent rats were equally capable of partial discrimination.

It is important to note that this discrimination is weak (65%, compared with well over 80% for other stimulus discriminations) and that it took several days to materialize. The poor distinguishability may suggest that SOA generates another taste in addition to bitterness, which is has not been reported to our knowledge, or that the timing of the bitterness onset is distinct. Animals may have focused on this hypothetical cue as they improved in performance across the first 5 days of discrimination testing. Human psychophysical work has demonstrated that although sensitivity to bitterness is highly variable between individuals (20), discrimination between bitter stimuli can rely on cues such as the length of time they are able to detect the after taste or time to maximal intensity (23, 24, 30, 43). However, when tested in a forced-choice discrimination paradigm, where discrimination is required to learn the task, humans are unable to learn to discriminate between caffeine, quinine, and tetralone (21).

Animals in this study were unable to discriminate quinine from cycloheximide, or PROP, which we chose because they lack similarity in neural representation, and perhaps receptor activation, although the latter work was done primarily in mice. While quinine shares activation of Tas2r105 with denatonium, cycloheximide, and PROP, SOA does not activate this receptor (27). However, quinine and SOA share another receptor, Tas2r144 (which is also activated by denatonium but not cycloheximide or PROP). We also would like to note that breadth of tuning tends to increase with stimulus intensity (42) and much of the previous work on bitter stimulus differentiation has been done at vastly different concentrations [e.g., Dahl et al. (15) used 10 mM quinine while Geran and Travers (18) used 3 mM], making direct comparisons difficult. We used concentrations of cycloheximide, PROP, quinine and denatonium lower than those used by Geran and Travers (18) when they described the pattern of activation for “bitter best” neurons in the NTS. They found that neurons in the cycloheximide cluster responded much more weakly (26% as large) to quinine than to cycloheximide and were distinct in the analysis. Using a lower concentration is predicted to allow even narrower tuning. Despite our efforts to choose concentrations and compounds that one would predict could be discriminated, the rats remained unable to do so.

We are confident that in the context of our extensive controls and in light of our SOA responding that rats would perform with increased accuracy if they were able to perceive any relevant differences in taste quality. Animals were regularly tested with the water control procedure to demonstrate that they were not using any cues unrelated to the taste to discriminate between treatments and retested with the quinine KCl discrimination regularly to confirm they still understood and could perform the task. We also had an additional procedure in group 1 that confirmed that the quinine versus KCl discrimination performance was likely due to the bitterness of quinine being contrasted with the saltiness of KCl. After animals were trained to discriminate quinine from KCl, we then substituted quinine (an ionic bitter) for cycloheximide (a nonionic bitter) and the animals were unperturbed in their performance. Following this, we then switched the KCl for NaCl, and the animals had to relearn the task. This finding suggests that the salient cue in KCl that the animals used to perform the task was “saltiness.” Additionally, we used both ionic and nonionic bitter stimuli in our comparison and found that this classification changed neither the animal’s ability to discriminate the bitter nor their ability to discriminate them from the ionic component of KCl.

Perspectives and Significance

Our results suggest that bitter stimuli that were predicted to be distinct based on physiological recordings were not discriminable. These results serve as an important reminder that the upstream selectivity in bitter sensation does not necessarily facilitate perceptible differences in potential bitter compounds, and it is important to view molecular and neurophysiological data through the lens of behavior, if only to maintain an accurate perspective on the gustatory system and its output. Determining the perceptual distinctiveness of bitter taste is important to understanding bitter coding.

GRANTS

This research was supported by National Institute of Deafness and Other Communications Disorders Grant R01-DC-016869.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.-M.T. conceived and designed research; L.E.M. and K.E.K. performed experiments; L.E.M. analyzed data; L.E.M. and A.-M.T. interpreted results of experiments; L.E.M. prepared figures; L.E.M. drafted manuscript; L.E.M. and A.-M.T. edited and revised manuscript; L.E.M., K.E.K., and A.-M.T. approved final version of manuscript.