Distinct neural ensembles in the rat gustatory cortex encode salt and water tastes

Abstract

The gustatory cortex (GC) is important for perceiving the intensity of tastants but it remains unclear as to how single neurons in the region carry out this function. Previous studies have shown that taste-evoked activity from single neurons in GC can be correlated or anticorrelated with tastant concentration, yet whether one or both neural responses signal intensity is poorly characterized because animals from these studies were not trained to report the intensity of the concentration that they tasted. To address this issue, we designed a two-alternative forced choice (2-AFC) task in which freely licking rats distinguished among concentrations of NaCl and recorded from ensembles of neurons in the GC. We identified three neural ensembles that rapidly (<300 ms or ∼2 licks) processed NaCl concentration. For two ensembles, their NaCl evoked activity was anticorrelated with NaCl concentration but could be further distinguished by their response to water; in one ensemble, water evoked the greatest response while in the other ensemble the lowest tested NaCl concentration evoked the greatest response. However, the concentration sensitive activity from each of these ensembles did not show a strong association with the behaviour of the rat in the 2-AFC task, suggesting a lesser role for signalling tastant intensity. Conversely, for a third neural ensemble, its neural activity was well correlated with increases in NaCl concentration, and this relationship best matched the intensity perceived by the rat. These results suggest that this neuronal ensemble in GC whose activity monotonically increases with concentration plays an important role in signalling the intensity of the taste of NaCl.

Key points

• In the primary taste cortex, two populations of neurons have been identified whose activity either correlates or anticorrelates with tastant concentration.

• The relative contribution of each population to signalling the perceived intensity of taste is unknown.

• To resolve this issue, we recorded activity from neurons in primary taste cortex while rats made choices that depended on NaCl concentration cues to obtain water reward.

• Two neural ensemble's neural activity anticorrelated with NaCl concentration, but differed according to their response to water as one ensemble tracked the solution's osmotic pressure.

• A third neural ensemble's activity was correlated with NaCl concentration and had the strongest relationship to the rat's choices, suggesting a central role in signalling the perceived intensity of NaCl solutions.

• A fourth neural ensemble responded only to water, but some of these neurons were influenced by behavioural context because they responded differently to water presented as a cue versus reward.

Introduction

Sensory physiologists have long been interested in how neurons signal the quality and intensity of a stimulus. In the gustatory system, tastant-selective cells provide information about a particular taste's quality (e.g. salt, sugar) through the activation of specific receptors (reviewed in Yarmolinsky et al. 2009). Information about a tastant's intensity is thought to be encoded through a concentration-dependent increase in taste cell activity that in turn is faithfully transmitted through primary afferent taste fibres to gustatory neural relay areas in the hindbrain (Frank, 1973; Scott & Perroto, 1980; Danilova et al. 2002; Hellekant et al. 2010; Chen et al. 2011a). At this level of processing, there is a positive correlation between concentration and the taste-evoked neural response (i.e. the neural concentration sensitivity curve). In rats and monkeys, this correlation is preserved as tastant information ascends the gustatory processing hierarchy through the hindbrain (Ganchrow & Erickson, 1970; Scott & Erickson, 1971; Ogawa et al. 1972; Scott & Perrotto, 1980). However, once this information reaches higher order taste centres, a more complex relationship between taste-evoked activity and concentration has been observed. Indeed, taste-evoked activity recorded from neurons in the gustatory cortex (GC), and in other forebrain areas comprising the gustatory processing circuit, can be either correlated or anticorrelated with increases in the concentration of tastants (Scott & Yalowitz, 1978; Yamamoto et al. 1984, 1989; Scott et al. 1991; Stapleton et al. 2006; Chen et al. 2011b). However, the relative contribution of each response type to signalling the perceived intensity of taste is, in general, poorly characterized, and this is particularly noteworthy for the GC because it is considered important for generating this percept. In this regard, lesions of the GC in humans cause deficits in reporting intensity differences (Pritchard et al., 1999; Small et al., 2001), and functional magnetic resonance imaging (fMRI) studies have shown that changes in the blood oxygen level-dependent (BOLD) signals observed in GC are correlated with reported tastant intensity (Small et al., 2003). However, the relationship between the BOLD signal measured in fMRI studies and single neuron activity is unclear (e.g. Logethetis et al. 2002), as is whether single neurons in GC encode changes in tastant intensity through concentration-graded increases in activity, decreases in activity, or changes of both types.

The reason why the neural coding scheme for taste intensity in GC is obscured is twofold. (1) In most electrophysiological studies using awake animals, neural concentration sensitivity curves are evaluated without taking into account how the animal perceived the taste that it sampled. In these studies, different kinds of tastants and concentrations are typically employed as stimuli, but the animal is not trained to report differences (i.e. distinguish) among them; instead neural activity is often evaluated without regard to behaviour (Yamamoto et al. 1984, 1989; Stapleton et al. 2006; Chen et al. 2011b) or compared indirectly to perceived intensities reported in human studies (see Scott et al. 1991). Thus, to date, there have not been studies that have characterized neural concentration sensitivity curves in animals trained to distinguish concentrations and therefore have not, in the same animal, linked concentration-dependent neural activity with concentration-dependent behaviour. (2) In most electrophysiological studies of GC, the time scale over which neural activity is evaluated to generate concentration sensitivity curves varies considerably across studies (from 1 s to 5 s), but this time scale is an order of magnitude longer than the time it takes for animals (including humans) to perceive and respond to tastant features (<300 ms) (Halpern & Tapper, 1971; Yamamoto and Kawamura, 1981; Gutierrez et al. 2010; Samuelsen et al. 2012). Because rapid excitatory concentration-dependent responses have been reported in the periphery (Bealer, 1978; Breza et al. 2010; Breza & Contreras, 2012), taste-evoked responses that persist past 1 s may, to some degree, reflect other aspects of the taste such as reward (Katz et al. 2001; Yoshide & Katz, 2011). Therefore, the relevant time scale to consider for neural concentration sensitivity curves should be consistent with the rapid decisions animals can make about taste quality and concentration.

To resolve these issues, and investigate how neurons in the GC contribute to perceiving the intensity of a taste, we first developed a novel behavioural task in which rats distinguish among different concentrations of NaCl. The precise timing and control of NaCl cue delivery occurred during the rat's natural licking behaviour, and allowed us to evaluate rapid concentration-dependent responses from neurons over short, behaviourally relevant time scales. Furthermore, a water cue was also presented to provide a baseline comparison to NaCl activity. The present study provides the first analysis of how GC neurons rapidly respond to changes in NaCl concentration in rats that are actively reporting the intensity of a taste stimulus. This protocol allowed us to evaluate concentration sensitivity curves for neuronal ensembles, and compare this relationship to the rat's perceived intensity of NaCl that is reflected in the choices that the rat made. We found that about half of the recorded neurons rapidly responded (<300 ms) to the water and/or NaCl cue (‘cue sensitive’) and a significant population of them showed graded increases in spiking activity to increased NaCl concentration. This relationship best matched the choices made by the rat, suggesting that these neurons are strongly linked to the perceived intensity of NaCl. However, in other neuronal populations we identified NaCl-evoked activity that anticorrelated with increases in NaCl concentration. This population could be separated further into two groups according to whether the lowest concentration of NaCl (‘low best’ neurons) or water (‘water best’ neurons) evoked the greatest response. Finally, a fourth neural ensemble selectively responded to water but not NaCl, but some of these neurons responded differently depending on whether water was presented as a cue or reward. Possible roles for these ensembles are discussed.

Methods

All protocols were approved by the Duke University Institutional Animal Care and Use Committee. Many of the methods used in this paper have previously been described in detail (MacDonald et al. 2009).

Experimental subjects

Eight adult male Long–Evans rats that weighed between 325 g and 375 g served as subjects. The rats were housed individually, maintained on a 12 h light–dark cycle, and given ad libitum access to food. The rats were subjected to a daily 23.5 h water deprivation schedule. About 30 min following the end of a session, the rats were also given free access to water in their home cage for 30 min while food was available.

Behavioural apparatus

Behavioural testing was performed in a standard operant chamber (MED-PC, MED Associates, Georgia, VT, USA). A PC-compatible computer attached to an electronic interface was used to control the experimental equipment and record the behavioural data. Inside the operant chamber, a lick spout was located on the back of a recessed panel in the centre of the front wall (sample spout). On either side of the sample spout were two additional spouts (choice spouts) where the rats could receive a task-related water reward (see below). Attached to the back of the lick spout were several independent tubes through which 20 μl of the tastants were delivered. A photobeam was positioned directly in front of all of the lick spouts and was monitored by the computer to detect lick responses. A photobeam was also positioned in the middle of the central recess to detect head pokes from the rat as it approached the central lick spout. The tastants were deionized water and four daily made concentrations of NaCl (30, 47, 76, and 120 mm).

Behavioural procedure

Figure 1 illustrates a single trial from the 2-AFC task. A single trial began with five dry licks on the sample spout during which no tastants were presented, followed by the presentation of the water cue, which involved delivering two 20 μl water aliquots after two successive licks. After five additional dry licks (about 0.75 s), the NaCl cue began and entailed the presentation of an aliquot (20 μl) of NaCl that was delivered after two successive licks. The concentration of NaCl after both licks was the same during a trial and was chosen pseudorandomly from a list of four concentrations. Following the second NaCl aliquot, licks on the sample spout had no consequence until the rat initiated a licking sequence on one of the choice spouts. The low cue – 30 mm NaCl – reliably predicted that water reward would be available on only one of the choice spouts (the low choice). The high cue – 120 mm NaCl – reliably predicted that water reward would be available on the other choice spout (the high choice). Two intermediate NaCl concentrations – 47 mm and 76 mm– could also be delivered during the trial but were always unrewarded though the rats were required to make a choice in order to start the next trial. The concentration range we used was suprathreshold, and sufficiently small so that NaCl remains qualitatively salty to the rat (Yamamoto et al. 1994).

Figure 1. The 2-AFC task.

A schematic diagram of a trial in the 2-AFC task. A rat will first lick on the central sample spout (left) to receive water as a tastant and one of four NaCl solutions. After making a high/low decision as to the intensity of NaCl it will go to one of the two choice spouts (right). Licks are depicted as vertical ticks. Black circles are placed over the two licks in which the water cue is delivered, over the two licks during which time one of four noted concentrations of NaCl are delivered (the NaCl cue), and finally during the two rewarded water licks. The other licks are dry, meaning no tastants are delivered. Note that the two licks for water are delivered as reward after five dry licks. The reward is only given if the rat chooses correctly after the 30 and 120 mm NaCl cues. The time between the last lick of the NaCl cue and the time the subject leaves the sample spout is called the reaction time.

The sixth and seventh lick on the correct choice spout resulted in water delivery for the low and high cues. These two licks are referred to as the reward cue, and once completed the rat could begin the next trial by licking on the sample spout. Incorrect choices resulted in no water delivery and a 15 s time out before a trial could be started. The relationship between the correct choice spout and the low and high cue concentration was fixed for an individual rat across all sessions and counterbalanced across rats. Additional details regarding the training protocol are given in MacDonald et al. (2009). Once rats learned the 2-AFC task, they were chronically implanted with extracellular recording electrode arrays that targeted the GC (see below).

Surgical methods and testing

Recordings of extracellular activity on six rats were obtained using microwire arrays made from 16 moveable 35 μm diameter microwires. Each array was arranged in a 4 × 4 design with ∼250 μm spacing among each wire, and the array moved together with each turn of a small screw. In each rat, one array was implanted in the left hemisphere just dorsal to the granular area of the GC (1.2 mm anterior to bregma, 5.3 mm lateral to bregma, 3.80 mm ventral to the brain surface). For two rats, two driveable microwire bundles of thirty-two 25 μm diameter wires were used, and these bundles also moved together with each turn of a small screw. After 1 week of recovery, the rats were acclimated to the task (1–2 sessions) with the recording cables plugged in, and testing began when the array (or bundle) of electrodes was moved downward and estimated to have passed into the granular layer. After moving the array, if no putative neurons (see below) could be confirmed online, then the rat was not tested, the electrodes were moved ventrally (about 125 μm increments) to sample from a new area, and the neuron count was evaluated the subsequent day. For a given rat, this process stopped when the electrodes’ positions were estimated such that they could no longer be moved downward without passing beyond the region of interest.

To identify electrode placement after completion of the experiments, rats were then deeply anaesthetized with sodium pentobarbital in preparation for perfusion. They were perfused transcardially using physiological saline followed by 10% formalin. Their brains were then removed from the skulls and stored in a 10% formalin–20% sucrose solution for several days before sectioning. Using standard procedures from this lab (e.g. Katz et al. 2001; Stapleton et al. 2006), brains were cut into 50 μm sections surrounding the electrode tracks.

Electrophysiological methods

Ensembles of single neurons, distributed across the GC were recorded simultaneously. The electrical activity recorded from each microwire was monitored with a multichannel acquisition processor (Plexon Inc., Dallas, TX, USA). The discriminated spike times were transferred, along with the onset and offset times of behavioural events (e.g. lick responses, salt and water deliveries, etc.), to a database for subsequent analysis. Only single units that were 3:1 signal to noise as displayed on the oscilloscope were collected. In addition, time–voltage windows, which were combined into a real-time three-dimensional principal component algorithm, were used for spike discrimination off-line (MacDonald et al. 2009). Putative single-units were also judged on the basis of visual inspection of waveform shape, size and discriminability in reference to other single units and the baseline noise on each channel. Putative single units were further verified as single neurons by evaluating the interspike interval (ISI) histogram, such that less than 1% of its interspike interval histogram fell below 2 ms. Discriminated units that did not pass these criteria were not analysed further.

Behavioural measurements and analyses

The probability of a rat responding on the high choice spout was computed separately for each NaCl concentration, and for each session. A one-factor analysis of variance (ANOVA) with repeated-measures on factor concentration was used to test whether the mean of the rat's high choice across sessions depended on NaCl concentration. To determine whether lick rate was influenced by NaCl concentration, the three interlick intervals (ILI) after the first lick of the NaCl cue were averaged for each trial, and the mean of this measure was computed for the session. A one-way ANOVA using factor concentration was conducted to test whether the mean of this measure depended on concentration across sessions. Statistical significance was set at P < 0.05.

Water and NaCl activation of neurons

NaCl- and water-evoked activity were detected by setting three consecutive and non-overlapping 50 ms time-bins in reference to lick 1 and lick 2 for the NaCl and water cue that was delivered during a trial. In this way, lick 1 and lick 2 each marked the beginning of a 150 ms time period that was divided into three time-bins (time-bins 1–3), so that a total of six time-bins (300 ms) defined the NaCl or water cue. Furthermore, on each trial two different baseline periods were defined by setting three 50 ms time-bins in reference to a dry lick immediately before the NaCl and water cue. Therefore, both the NaCl and water period used a different baseline for comparison.

To determine whether a neuron was activated during the NaCl cue, a two-way repeated-measures ANOVA compared the trial-averaged action potential counts from each of the three time-bins for the first and second lick of the NaCl cue to the comparable time-bin during NaCl baseline period (e.g. time-bin 1 from each lick during the NaCl cue was compared to time-bin 1 during the baseline period; time-bin 2 from each lick during the NaCl cue was compared to time-bin 2 during the baseline period, etc.). The factors used for this ANOVA were pre–post (i.e. baseline or NaCl cue) and time-bin. This same test was conducted separately for the water and its baseline period. A neuron was considered activated by water or NaCl if a main effect or an interaction between the two factors was confirmed. If statistical significance was verified in the analysis of either cue, standard post hoc testing was used to determine whether the effect was driven by (1) an increase or (2) a decrease in cue activity compared to baseline. In the former case, the response to the cue of interest was considered excitatory and neurons with excitatory responses to water or NaCl were further categorized as fast (NaCl: N_F; water: W_F) or slow (NaCl: N_S; water: W_S) depending on whether the greatest change in activity occurred after the first or second lick during the cue. Similarly, the cue response was considered inhibitory and the neuron was classified as N_I and/or W_I if the response was confirmed for the NaCl and water cue, respectively. If no response was confirmed to the NaCl and/or water cue, the neuron was considered NX and WX, respectively. Note that the NaCl and water cue analyses were conducted separately, so a neuron was always categorized as N_F, N_S, N_I, or NX during the NaCl cue independently of whether the neuron was categorized as W_F, W_S, W_I, or WX during the water cue. In the Results, we refer to a ‘cue sensitive neuron’ as one that responded to the NaCl and/or water cue. A ‘NaCl selective neuron’ only responded during the NaCl cue. A ‘water selective’ neuron only responded during the water cue. We considered neurons ‘taste sensitive’ if they responded to only one of two cues, or more to one cue compared to the other if the neuron responded to both cues (see below).

For NaCl selective neurons, we further tested whether NaCl concentration had an effect on the trial-averaged number of action potentials. For each neuron, we identified the lick in which the largest NaCl evoked response was observed (i.e. the chosen lick depended on whether the neuron had a slow or fast response defined above). Then for each trial, the number of action potentials in time-bins 1–3 of the NaCl baseline period was subtracted from the number of action potentials in the comparable time-bin of the chosen lick during the NaCl cue. A two-way repeated-measures ANOVA was conducted on these trial-averaged spike counts using factors concentration and time-bin. A NaCl selective neuron was considered ‘concentration sensitive’ if a main effect of concentration or an interaction between concentration and bin was confirmed.

For each cue sensitive neuron that responded during the NaCl and water cue, we directly compared the magnitude of each response to test whether it was taste sensitive and concentration sensitive. To accomplish this, the number of action potentials during time-bins 1–3 of each baseline period was subtracted from the number of action potentials during time-bins 1–3 of only one lick during the NaCl and water cue to yield a difference score. As above, only the lick during the NaCl and water cue with the largest response was used for this analysis. A three-way ANOVA with repeated measures on factors cue (NaCl or water), concentration, and time-bins (1–3) was conducted on this trial-averaged data. If a main effect of cue, but not concentration, was confirmed, then the neuron distinguished between water and NaCl and was considered taste sensitive. If there was a main effect of cue and concentration, or an interaction between the two factors, the neuron was considered taste and concentration sensitive. The response to the low and high NaCl cue was directly compared in these cases to test which concentration evoked the larger response (Fisher's least significant difference test).

Testing whether water selective neurons were also activated by the reward cue

To compare a water selective neuron's response to the water cue to the reward cue, the same approach was taken as described above for comparing the NaCl and water cue response. Briefly, first we determined whether a water selective neuron responded to the reward cue. The reward cue was defined by setting three consecutive and non-overlapping 50 ms time-bins in reference to lick 1 and lick 2 of the reward, and a new baseline period for reward were defined by setting three 50 ms time-bins in reference to a dry lick immediately before the first lick of reward. As described above for testing whether a neuron was activated by NaCl or water, a two-way ANOVA was used to compare the trial-averaged action potential counts from each of the three time-bins for the first and second lick of the reward cue to the comparable time-bin during the reward baseline period. The factors used for this ANOVA were pre–post (i.e. baseline or reward cue) and time-bin. This analysis allowed us to determine whether a water cue sensitive neuron was also activated by the reward cue and, if so, identify the whether the greatest change in activity occurred after the first or second lick during the reward cue using standard post-hoc testing. If a water selective neuron was not activated by the reward cue, then no further testing was done. If a water selective neuron was also activated by the reward cue, then using a two-way ANOVA (factors water/reward cue, and time bin 1–3) and post hoc testing, we directly compared the magnitude of the water and reward cue response using the same approach as described above when comparing of the magnitude of the NaCl and water cue response, but without concentration as a factor. This allowed us to determine whether the response to the reward cue was the same or suppressed compared to the water cue response (no water selective neurons were activated more by the reward cue).

Visualizing event-related changes in neural activity

The peristimulus time-histograms (PSTH) shown in Figures 2–5 were made by aligning six contiguous, non-overlapping time-bins of size 50 ms (300 ms ∼2 licks) in reference to the event in question.

Figure 2. Gustatory cortical responses to NaCl and water.

A, top, a raster plot for a single neuron, where each row shows its action potentials (black ticks) time locked to the delivery of the first lick during NaCl cue from a single trial. The licks during each trial are also indicated in each row (red ticks). Bottom, the PSTH illustrating the trial averaged firing rate (spikes s⁻¹) for the neuron in reference to the NaCl cue. This neuron is representative of a fast response to NaCl (N_F). B and C, trial by trial raster plots and trial averaged PSTHs using the same format as in A, but for representative examples of a slow (B: N_S) and inhibitory (C: N_I) response to the NaCl cue. D, the average firing rate (normalized – see Methods) to the NaCl cue across the population of N_F (black), N_S (grey), and N_I (broken line) neurons. E–G, trial-by-trial raster plots and trial averaged PSTH in the same format as in A–C except that these neurons are time locked to the first lick of the water cue, and are representative examples of neurons with a fast (E: W_F), slow (F: W_S), or inhibitory (G: W_I) response to the water cue. H, the same format as in D except that this plot shows the average population firing rate (normalized – see Methods) to the water cue for W_F (black), W_S (grey), and W_I (dashed line) neurons.

Population PSTH

A PSTH was determined for each neuron in a population with reference to the event under consideration (see above). Neural activity in each bin was standardized as a Z score according to the mean and standard deviation across all of the time bins in all trials in the analysis. Subsequently, the average PSTH across the whole population was determined.

Comparing the integrated response to each NaCl concentration

For neurons already considered concentration sensitive, the integrated response was computed by taking the sum of the standardized action potential counts across time-bins from the NaCl cue delivered on each trial and then calculating the trial average. The standardization was done with respect to the mean and standard deviation of the action potential counts across time-bins and trials analysed. A one-way ANOVA was used to test whether the trial-averaged integrated response differed according to NaCl concentration.

Results

Over many sessions (see Methods), eight rats were trained on a two-alternative forced choice (2-AFC) task for which the rat reported whether the intensity of NaCl it sampled was ‘high’ or ‘low’. Through a lick spout positioned centrally on one wall of an operant chamber (sample spout), on each trial freely licking rats were first presented with a small aliquot of water (water cue) after each of two successive licks (20 μl for each lick), then licked 5 times during which no solution was delivered (dry licks), which was followed by two consecutive licks that delivered a single NaCl concentration (NaCl cue; 20 μl for each lick; see Figure 1 for trial structure). The concentration of NaCl delivered after each lick during the NaCl cue was the same, and randomly selected for a trial from four values: 30 mm, 47 mm, 76 mm and 120 mm. Once a rat sampled the NaCl cue, it was free to make a choice by licking on only one of two additional response options, which were lick spouts positioned to its left and right (choice spouts). The lowest NaCl concentration (30 mm) was associated with only one response option (low choice) and the highest NaCl concentration (120 mm) was associated with the other response option (high choice) so that the rat only received a water reward by making a correct choice. The two intermediate NaCl concentrations (47 mm and 76 mm) were never rewarded, but the rats reliably made more low choices for 47 mm of NaCl and more high choices for 76 mm of NaCl.

Indeed, as previously reported (see Figure 5E and MacDonald et al. 2009), the probability of a high choice increased monotonically with the NaCl concentration showing that the rats evaluated the intensity of the NaCl concentration to guide their choice (one-way ANOVA; main effect of concentration, F(3,21) = 146.6, P < 0.001). The rats licked for an additional 1.17 s ± 0.17 s (mean ± SEM) after the NaCl cue was delivered, but there was no difference in this reaction time (RT; see Figure 1) measure in relation to NaCl concentration (one-way ANOVA; main effect of concentration, F(3,21) = 1.65, P= 0.21). Furthermore, during the NaCl cue we found no effect of NaCl concentration on the interlick intervals (ILIs) (one-way ANOVA; main effect of concentration, F(3,21) = 0.94, P= 0.44). That the ILI was found to be independent of concentration is important because electrophysiological recordings from GC can reflect tongue and other mouth movements (Gutierrez et al. 2010), and neural activity is influenced by the rat's licking (see below). Thus, the concentration sensitive responses we report hereafter do not result from concentration-dependent changes in the rat's licking behaviour.

Figure 5. Neural activity is correlated or anti-correlated to the NaCl concentration.

A, trial-by-trial raster plot and trial averaged PSTH for a representative NaCl selective neuron that is time locked (left to right) to water, or 30 mm, 47 mm, 76 mm, or 120 mm NaCl. This neuron shows no response to water but a correlated increase in activity with NaCl concentration. Below is the average response across the population of NaCl selective neurons (n= 16). The legend indicates the colour coding used to distinguish among the different NaCl concentrations. B, the top panel shows trial by trial raster plots and trial averaged PSTHs for a ‘water best’ neuron that was maximally activated by water, but whose NaCl evoked response shows an anticorrelated relation to an increase in NaCl concentration. Below is the average response to all cues across the population of ‘water best’ neurons (n= 4). C, the top shows trial-by-trial raster plots and trial averaged PSTHs for a ‘low best’ neuronal response that was maximally activated by 30 mm NaCl. Below is the average response to all cues across the population of ‘low best’ neurons (n= 5). D, the normalized integrated response during the NaCl cue for the three concentration sensitive neural populations. The red trace represents the average integrated response for NaCl selective neurons (see A), the blue trace shows the average integrated response for ‘water best’ neurons (see B), and the green trace shows the average integrated response for ‘low best’ neurons (see c). E, the probability (mean ± SEM) of the rat selecting the high choice spout as a function of NaCl concentration as obtained from the 2-AFC task shown in Figure 1.

GC neurons activate rapidly to both NaCl and water

Across 29 recording sessions, we isolated extracellular waveforms recorded from 274 neurons in the GC while rats performed the 2-AFC task. The final position of the electrode arrays and bundles, as well as the estimated distance covered throughout testing, confirmed that they generally spanned layers II–VI, and neurons were sampled primarily from dysgranular cortex, but also granular and dorsal agranular areas of the anterolateral insula (Shi & Cassel, 1998; MacDonald et al. 2009; see Figure S1).

For each neuron, we first evaluated whether the water or NaCl cue evoked a response by testing whether the trial-averaged neural firing rate differed from baseline during either of the two licks that comprised water and NaCl cue delivery (see Methods). This analysis revealed that the activity from over 56% (155/274) of the neurons was not influenced by water or NaCl (see NX and WX in Table 1), and thus was cue insensitive. In contrast, we found that 44% (119/274) of the neurons were cue sensitive because they responded to NaCl and/or water. Consistent with several previous reports from our and other laboratories (Yamamoto et al. 1989; Katz et al. 2001; Stapleton et al. 2006; MacDonald et al. 2009; Gutierrez et al. 2010; Samuelsen et al. 2012), there was strong 6–7 Hz oscillatory structure in neural activity from the GC because many neurons were active while the rat was licking at a rate of 6–7 Hz. It is worth noting that while cue insensitive neurons did not respond to water or NaCl, and we did not consider them further in the analyses that follow, they were nevertheless typically sensitive to the rat's licking response (see Figures S2A and S2B and MacDonald et al. 2009).

Table 1.

Number of neurons in each response profile

Profile	NF	NS	NI	NX	Total
WF	23	2	0	26	51
WS	0	10	5	22	37
WI	0	0	10	2	12
WX	5	13	1	155	174
Total	28	25	16	205	274

Numbers of neurons with reference to their NaCl and water response profiles. See text for details. Note that NX and WX refer to no responses to NaCl or water respectively.

To test whether NaCl or water activated more neurons, we analysed the neural response to each cue separately. The response to the NaCl cue evoked a rapid, excitatory change in neural activity in 19% (53/274) of the neurons. However, across these neurons the timing of activation varied depending on whether the neuron responded maximally during the first or second lick of the NaCl cue. Figure 2A (top) shows a raster plot from a single neuron where on each row, the black ticks depict action potentials that are referenced to the first lick of the NaCl cue delivered on each trial, regardless of the NaCl concentration delivered. The red ticks depict the time of the rat's lick. Using the data illustrated in the raster plot, the trial-averaged firing rate of the neuron during the NaCl cue was computed and is shown as a peristimulus time histogram (PSTH, black line) immediately below the raster plot (Figure 2A bottom). In this way, it is clearly seen that this neuron increased its activity within 150 ms of the first lick of NaCl, but the response of to the subsequent lick of NaCl was relatively diminished. Ten per cent (28/274) of the population exhibited this phasic response profile, which was named a fast excitatory NaCl response (N_F). Alternatively, Figure 2B shows an example of a neuron with a second NaCl related response profile. The excitatory response from these neurons, which represent 9% of the population (25/274), developed more slowly such that activity was larger by the second NaCl lick compared to the first. We will refer to neurons with these slow excitatory responses to NaCl as N_S. These responses were also distinct from N_F responses because their activation was typically sustained while the rat continued licking before leaving to make a choice (see Figure S3A for the average response across the N_S population for 1.2 s after NaCl delivery; see also MacDonald et al. 2009). We also observed 13 neurons whose activity was inhibited by NaCl during either of the two licks (N_I; Figure 2C). Though activity in N_I neurons was rapidly suppressed by the delivery of NaCl, these neurons recovered from the inhibition to various degrees after 300 ms (see also MacDonald et al. 2009). Figure 2D (labelled ‘population response’) summarizes the normalized response to NaCl averaged across all neurons that made up each of the three, NaCl-related response profiles.

The response profiles to water were very similar to those evoked by NaCl. Figures 2E, F and G show raster plots and the trial-averaged PSTHs from representative neurons that make up the fast (W_F), slow (W_S; see Figure S3B for the average response of the W_S population for 1.2 s after water delivery), and inhibitory (W_I) water response profiles, respectively. Figure 2H shows the average population response for each water-related neuron type.

We also tested whether the proportion of W_F neurons (51/274, 19%) was greater than the proportion of N_F neurons (10%), and found this difference to be significant (χ₁= 7.16, P= 0.007). However, the proportion of neurons with slowly developing, excitatory activation to water (W_S= 37/274, 14%) did not differ from NaCl (N_S= 9%; χ₁= 2.2, P= 0.14), nor did the proportion of neurons with an inhibitory response to water (W_I= 5%—13 /274) differ from inhibitory responses to NaCl (N_I= 4%; χ₁= 0, P= 1). Table 1 summarizes these responses, and also shows the neuron count for each NaCl and water response profile combination.

GC neurons rapidly distinguish between NaCl and water

As noted, there were 119 neurons that were cue sensitive because they responded to NaCl and/or water. Consequently, we determined what proportion of these neurons distinguished between water and NaCl and were therefore taste sensitive (see Methods). The cue sensitive neurons could be grouped according to the relationship between their response to water and NaCl, and based on this were assigned exclusively to four broad categories (see Table 1). The first category is composed of neurons that had an excitatory response (slow or fast) to NaCl and water. This category constituted 14% (35/274) of the total neural population. Most of the neurons in this category (33/35) had the same response profile to NaCl and water (e.g. a neuron with a fast, excitatory response profile for NaCl also had a fast, excitatory response to water). By comparing the trial-averaged firing rate of NaCl to water, we found that 80% (28/35) of these neurons were taste sensitive. Moreover, 75% (21/28) of the taste sensitive neurons from this category were activated more strongly by water compared to NaCl averaged across all concentrations. Figure 3A shows a trial-by-trial raster plot and trial-averaged PSTH for a representative neuron that met this criterion. Though the concentration sensitivity is discussed in the subsequent section, 9 out of 28 of these taste sensitive neurons responded differently across the four tested NaCl concentrations. An effect of concentration was not confirmed for 19 of these neurons even though the NaCl and water responses were different (see Figure S2C). However, as only water and a relatively small range of NaCl concentrations were delivered during these experiments, nothing else can be said about them except that they are taste sensitive.

Figure 3. Comparison of responses to water and different NaCl concentrations.

A, raster plot and PSTH showing, respectively, trial by trial action potentials (black ticks) and trial averaged firing rate for a single neuron whose activity is time locked to the first lick of the water (H₂O) or 30 mm, 47 mm, 76 mm, or 120 mm NaCl cue delivered in the session. Note that the data from the water cue delivered before each NaCl cue are pooled irrespective of the concentration. This is a representative example of a neuron whose response to water was greater than any of the responses to NaCl. B, same format as in A but illustrates a neuron that was not activated by water but was activated by different NaCl concentrations (NaCl selective). C, same format as above but of a neuron that was inhibited by water and all NaCl concentrations. D, same format as above but of a water selective neuron that was activated by the water cue but not by the NaCl cue.

Within the 300 ms time scale used to analyse a response, the remaining 7 out of 35 neurons in this category were considered taste insensitive because no difference could be detected between the NaCl and water response (see Figure S2D for an example). Given that neural activity was suppressed while rats licked before water and NaCl's delivery, it is unlikely that these neurons exclusively encoded somatosensory information. However, no other conclusion can be made about their function except to note that they detected the presence of liquids.

The second category consisted of 18 NaCl selective neurons (7% of the total population) that had an excitatory response to NaCl but not to water (see Figure 3B, which shows a trial-by-trial raster plot and trial-averaged PSTH for a representative neuron in this category). The majority of these neurons (89%; 16/18) were concentration sensitive.

A third category of responses contained 18 neurons that were inhibited by either cue. Here a neuron was considered to be taste insensitive if both NaCl and water suppressed its activity. This was the case for more than half of the neurons in this category (10/18), and a raster plot and trial-averaged PSTH for one such neuron is shown in Figure 3C.

Lastly, we identified a fourth category that comprised 48 neurons (18% of the population) that were activated by water but not by NaCl. An example from this category is shown in Figure 3D, which illustrates a trial-by-trial raster plot and trial-averaged PSTH. Neurons within this category are referred to as water selective, though the full extent of water selectivity is unknown because only NaCl and water were used in these experiments. Nevertheless, none of the NaCl concentrations reliably evoked a response from these neurons. To further examine the selectivity of the water response, we compared the magnitude of the neural response when water was used as a cue to the neural response evoked when it was used as a reward (see Methods). To the extent that the water response is purely sensory (i.e. it reflects the physicochemical properties of water), then the response to water as both a cue and reward should be the same. To this end, we found that 22 of the 48 water selective neurons (8% of the population) had equivalent responses to water as a cue and reward, which is consistent with a sensory-driven, water selective response (see Discussion). Figure 4A illustrates a representative example of this type of water response. Shown is a raster plot illustrating action potentials for all trials in which water was delivered as a cue (left) or reward (right), and the trial-averaged PSTHs are below each plot. Conversely, for 26 water selective neurons, water evoked a different response when used as a reward; that is the water response was either weaker when delivered as a reward (10 neurons; see Figure 4B), or there was no excitatory response to water presented as a reward (16 neurons; see Figure 4C). Therefore, while neurons in this fourth category were selectively activated by water as a cue, the underlying cause of this response appears to differ across the population.

Figure 4. Comparison of responses to water presented as a cue and reward.

A, top, raster plot showing black ticks representing action potentials recorded from a neuron that is time locked to the delivery of water as a cue (left) or as a reward (right) on every trial. This representative neuron responded equivalently to water in both contexts. In the population, the percentage of these types responses was 9%. Bottom, the trial averaged PSTH. B, similar to the format shown in A except that this neuron responds to water in both contexts, but with the water reward exhibiting a smaller response. The percentage of these neurons in the population is 4%. C, shown is a representative neuron that responds only to water as a cue and is inhibited when it is given as a reward. The percentage of these neurons in the population is 6%. For all three panels, the licks during each trial on the sample and choice spouts are also indicated on each row with red and blue ticks, respectively.

In summary, all of the cue sensitive neurons could be assigned to four broad categories based on the relationship between a neuron's response to NaCl and water. Table 2 summarizes the number of neurons in each category and the numbers in brackets indicate how many neurons was concentration sensitive.

Table 2.

Cue sensitive neurons in relation to their NaCl and water response

Category	n	Taste sensitive	Water > NaCl	Water < NaCl	Water = NaCl
I. Both cues evoke an excitatory response	35	28	21 (9)	7	7
II. ‘NaCl selective’ only NaCl evokes an excitatory response	18	18	0	18 (16)	0
III. ‘Inhibitory’ NaCl or water evokes an inhibitory response	18	9	5	3	10
IV. ‘Water selective’ only water evokes an excitatory response	48	48	48	0	0
Total	119	110	69	25	17

Numbers of neurons placed in only one of four categories based on of their relationship between the NaCl and water response. Note that the numbers in brackets refer to the number of concentration-sensitive neurons.

Taste sensitive neurons show a correlated and anticorrelated relationship to the NaCl concentration

Next we investigated whether taste sensitive neurons could differentially encode changes in NaCl concentration and, if so, whether their activity correlated or anticorrelated with increases in NaCl concentration. Because only neurons in the first two categories discussed above were concentration sensitive, we focused our analysis on those responses without regard to the fast or slow notation illustrated in Figure 2. There were 25 neurons (9% of the total population) whose trial-averaged response depended on NaCl concentration. Sixteen of them were NaCl selective because there was no response to water (either as a cue or reward), and the highest NaCl concentration evoked the largest response. A representative response of this type is shown in Figure 5A (top), and the trial-by-trial raster plots and trial-averaged PSTHs are separated according to concentration. Below this example (labelled ‘population response’) is a PSTH showing normalized activity referenced to the first lick of the NaCl cue, but averaged across all 16 neurons within this category and colour-coded depending on the concentration that was delivered. This figure shows that, as a population, the peak response to each concentration of NaCl occurred rapidly (<150 ms) and dissipated soon after the second lick of the NaCl cue; in addition, higher concentrations of NaCl evoked more activity (see next section).

Two other concentration sensitive responses were identified, and each was activated more strongly by the low cue compared to the high cue. However, neurons showing this response could be distinguished with respect to their response to water. For four neurons, the response to water was greater than the NaCl response (Figure 5B, top), and the population response to NaCl for these ‘water best’ neurons showed a rapid response during the NaCl cue that diminished by the second lick (Figure 5B, bottom). Importantly, all of the ‘water best’ neurons showed an equivalent response to water as a cue and reward. For the five remaining concentration sensitive neurons, the low cue (30 mm NaCl) evoked the greatest response while water was least effective in activating the neuron (Figure 5C, top). All of these ‘low best’ neurons showed a reduced response to water when it was used as a reward compared to its response to water as a cue. Moreover, all of the ‘low best’ neurons were N_S and the slow response to NaCl is shown clearly in the population response (Figure 5C, bottom). Thus, the temporal profile of ‘low best’ neurons clearly differed from NaCl selective and ‘water best’ neurons (compare population responses in the bottom panel of Figures 5A–C).

To further characterize each type of concentration sensitive neuron, we constructed concentration sensitivity curves using the population response. To accomplish this, for each neuron in all three groups we summed the NaCl evoked standardized activity across each of the two licks to give an integrated response for each NaCl concentration (seem Methods). Figure 5D shows this average integrated response for each NaCl concentration and separated according to the type of concentration sensitive response. Using a one-way ANOVA, we tested whether the average integrated response depended on the concentration of NaCl, and expectedly this was the case (all P values < 0.05). Moreover, a standard post hoc contrast analysis confirmed a significant linear relationship for all three populations (all P values < 0.05), which is consistent with a graded linear relationship between neural activity and NaCl concentration. However, a follow-up analysis suggests that the ‘low best’ neuronal response was not strongly graded with respect to concentration. For these neurons, if the integrated response to lowest concentration was dropped (i.e. its best response), and the ANOVA conducted on data from the remaining three higher concentrations, there was no significant effect of concentration (F(2,8) = 1.60, P= 0.26) and a contrast analysis showed no linear relationship (F(1,4) = 1.61, P= 0.27). Conversely, if the response to the lowest and highest concentration were excluded for the ‘water best’ and NaCl selective neurons, respectively (i.e. their best response), the effect of concentration on the integrated response remained significant (F(2,6) = 12.83, P= 0.007 and F(2,30) = 3.20, P= 0.05 respectively) and the linear relationship remained (F(1,3) = 22.23, P= 0.02 and F(1,15) = 9.11, P= 0.01 respectively). Thus, the concentration sensitivity for ‘low best’ neurons is mainly due to the effectiveness of the low cue in activating the neuron, while the ‘water best’ and NaCl selective neurons showed a more graded concentration dependent response.

In order to determine which of the three types of concentration-sensitive neurons could best contribute to the rat's perception of the intensity of NaCl solutions, we measured the correlation between each concentration sensitivity curve and the rat's perceived intensity, which is reflected in the choices the rats make in response to each NaCl concentration. Figure 5E shows the probability of a rat making a high choice response as a function of NaCl concentration, and shows that as the concentration of NaCl increased, the rat was more likely to perceive the concentration as being similar to the high cue. For each of the three types, we then correlated the mean integrated response with the rat's choice behaviour shown in Figure 5D. Notably, for only the NaCl selective neurons, the correlation between their integrated response and the rat's choice was highly significant (t(3) = 6.9, P= 0.006), and this simple linear model also explained 99% of the variability (R²= 0.99). The correlation did not reach significance for ‘water best’ neurons (t(3) = 2.1, P= 0.13) or for ‘low best’ neurons (t(3) = 1.89, P= 0.15), and these models explained less variability (R²= 0.76 and 0.71, respectively) compared to NaCl selective neurons. Collectively, these results suggest that, compared to the other two populations of neurons, the rapid response of NaCl selective neurons track the rat's perceived intensity to a greater degree than neurons whose responses were anticorrelated with increases in NaCl concentration.

Discussion

Neural correlates of perceived intensity in GC

By combining a novel taste guided 2-AFC task with electrophysiological recordings in the rat GC, we identified three ensembles of single neuron responses that depended on NaCl concentration. One type was selective to NaCl and exhibited excitatory responses that increased monotonically with the NaCl concentration. These neurons were unresponsive to water. This response profile is consistent with sodium selective neural and functional imaging responses that are observed in the periphery, brainstem, thalamus, and GC of rats, humans and non-human primates (Borg et al. 1967; Ganchrow & Erickson, 1970; Scott & Erickson, 1971; Ogawa et al. 1972; Scott & Perrotto, 1980; Yamamoto et al. 1984, 1989; Scott et al. 1991; Small et al. 2003; Stapleton et al. 2006; Accolla et al. 2007; Breza et al. 2010; Chandrashekar et al. 2010; Chen et al. 2011a). However, we note that the time scale during which we assessed concentration coding (∼300 ms) is about an order of magnitude shorter than the time scale used in many other studies, and suggests that these longer time scales may contain information in addition to the tastant's intensity (Katz et al. 2001; Yoshida & Katz, 2011).

The two other neural types responded rapidly to the NaCl cue and showed a greater response to 30 mm NaCl compared to 120 mm NaCl. These two types differed according to whether the greatest neural response was to 30 mm NaCl (‘low best’) or the water cue (‘water best’). Because ‘water best’ neurons were more strongly and equivalently activated by water as a cue or reward, and the NaCl response was anticorrelated with increasing concentrations of NaCl, the excitatory response profile of ‘low best’ neurons is consistent with detecting osmotic pressure gradients. Indeed, the osmotic pressure on the lingual surface is expected to be highest with water, and less with larger concentrations of solute (until isotonicity is reached). However, because we only used NaCl in our experiment, the strength of this claim should only be considered in this context. If ‘low best’ neurons encode osmotic pressure gradients, then a strong prediction of such a function is that ‘low best’ neurons will show the same response to equivalent molarities of solute, regardless of the specific tastant presented. Osmosensors have been found in mammalian taste cells (Lyall et al. 1999; Cameron et al. 2010) and in the mammalian taste system the transduction likely occurs through aquaporin channels (Gilbertson, 2002, 2006; Watson et al. 2007). Moreover, recordings from the superior laryngeal branch of the vagus nerve, which innervates the posterior of the oral cavity, show responses that are strongest to water and anticorrelated with increases NaCl concentration over the NaCl range we tested (Hanamori, 2001). Since the taste solutions delivered to freely licking rats swathe their entire oral cavity, vagal stimulation may contribute to the observed responses.

The population of ‘low best’ neurons that responded maximally to 30 mm NaCl was interesting because their response to the remaining NaCl concentrations and the water cue was the same. While these neurons could distinguish the salient 30 mm of NaCl from the other cues used in the experiment, they would appear to be less useful in signalling differences among the remaining cues that were presented and likely play a smaller role in signalling the intensity of NaCl. Interestingly, this neuronal subpopulation showed a suppressed response to water when it was used as a reward relative to when it was used as a cue. Thus, the response to water depended on the underlying behavioural context as established by the 2-AFC task. We speculate that the ‘low best’ response may result from training animals on the 2-AFC task; the low NaCl cue is made salient on account of its relationship to a particular choice and the ensuing reward. Future studies should test whether such a response profile emerges over the course of learning.

Only NaCl selective and ‘water best’ neurons showed graded responses to increases in concentration; as a result, they exhibited concentration sensitivity that is consistent with a role in evaluating the intensity of NaCl. Because rats were trained to distinguish among all four concentrations, and reported the perceived intensity of NaCl with the choice they made, we were afforded the opportunity to evaluate which of these two concentration sensitive response types is best associated with perceiving the intensity of NaCl. To this end, the response of NaCl selective neurons to each concentration was highly correlated with the intensity perceived by the rats, as inferred from the choices they made. On the other hand, the strength of the correlation between the ‘water best’ neuron's concentration sensitivity and the rat's behaviour did not reach significance. This result suggests a more primary role of NaCl selective neurons in GC in generating a perceived intensity of NaCl. It is important to highlight that because intermediate NaCl concentrations (47 mm and 76 mm of NaCl) were unrewarded, yet a linear relationship between neural activity and NaCl concentration increases was confirmed in NaCl selective and ‘water best’ neurons, it's unlikely that their responses are purely reward-driven. Furthermore, because the relationship was graded, the function of these neurons was not simply to categorize each of the NaCl cues as low or high.

Fast, slow, and inhibitory responses to water and NaCl by neurons in the GC

For both the water and NaCl cue, two distinct excitatory responses were observed that were distinguished by the time the maximum response was reached. The fast, phasic response (N_F and W_F) peaked rapidly after the first lick of the cue, and the response during the second lick was relatively suppressed (Figures 2A and E), a change due to desensitization of the ENaC receptor (Gilbertson & Zhang, 1998). A slow excitatory response (N_S and W_S) is one that was largest after the second lick (Figures 2B and F). The precise contribution of each excitatory response in the context of this task is unknown. Both of these responses were observed in the same sessions, and throughout the extent of the GC from which we recorded. One conclusion that can be made on the basis of these data is that neurons that responded to both water and NaCl tended to have the same temporal response profile (fast or slow responses to both water and NaCl). Moreover, as a population, the slow response to NaCl tended to persist (>1 s) after the presentation of the NaCl cue while the rat was licking on the lick spout (see Figures S3A and MacDonald et al. 2009). The slow water response also persisted long past water cue delivery up until NaCl cue delivery later in the trial (Figure S3B). The interpretation of fast and slow excitatory responses in the context of our task is further complicated by the fact that both types were intermixed among taste and concentration sensitive neurons so it is unclear whether both or only one are necessary for encoding intensity. Our analysis was constrained to the ∼300 ms time window that contained the first and second cue-related licks, although the rats continued to lick for about 1.2 s before leaving to make a choice. Interestingly, although the rat's RTs were independent of NaCl concentration (see Results), the latency of neural responses to NaCl in the geniculate ganglion (Breza et al. 2010) and GC (Accolla et al. 2007– through using intrinsic imaging methods) decrease with increasing concentration indicating that the limiting step in the rat's reaction time is not encoding the tastant. In a previous report (MacDonald et al. 2009), we characterized neural activity while the rat was licking immediately before it left to make its choice. Based on the preponderance of neural activity in GC during this time period, and on other analyses, we posited that such late activity might be involved in preparing for and/or guiding the rat's behaviour made on the basis of a taste cue. In line with this explanation, we suggest that late responses may be more related to the behaviour of the animal that depends on the tastant just evaluated, rather than evaluating the taste cue in and of itself. Indeed, for trained animals, and depending on the amount of tastant delivered and its concentration, a single lick (of a particular volume and concentration) appears to be all that is necessary for an animal to distinguish among tastant qualities (Hapern & Tapper, 1971; Gutierrez et al. 2010; see also discussion in Breza et al. 2010). Nevertheless, future studies will need to tease apart the behavioural relevance of these response profiles.

A third type of GC response is one that was inhibited in the presence of both salt and water (Figure 3C). From these experiments alone we cannot tell whether these neurons would be inhibited by other tastants, although inhibitory responses to a variety of tastants were previously observed in the GC for animals that freely lick (Yamamoto et al. 1989; Stapleton et al. 2006; Gutierrez et al. 2010) as well as during passive delivery (Katz et al. 2001; Samuelsen et al. 2012). More than half of the neurons with inhibitory responses did so following the delivery of both NaCl and water. Inhibitory responses to tastants in freely licking rats have also been observed in the orbitofrontal cortex (Gutiérrez et al. 2006) and also in the nucleus accumbens shell where they have been attributed to act as a ‘gate’ to the lateral hypothalamus permitting the animal to begin feeding (Krause et al. 2010).

The water response in GC

Our results show that a water response is prevalent in GC; 18% of the neural population responded to water and not to NaCl. Functional and intrinsic imaging of the human and rat GC, respectively, revealed spatially selective areas that are responsive to water (de Araujo et al. 2003; Accolla et al. 2007). Moreover, water responses in GC neurons have been reported in single unit studies in rats and non-human primates (e.g. Yamamoto et al. 1989; Scott et al. 1991; Katz et al. 2001; Ito et al. 2001), and there is good reason to consider this on account of different gustatory qualities evoked by water after adaptation to other tastes (Bartoshuk, 1972; McBurney & Bartoshuk, 1972; see also Rosen et al. 2010). As only water and NaCl were used as tastants in our experiment, we could not fully evaluate the selectivity of the water response. One possibility is that the water selective neurons we observed may have responded to other tastants had they been tested in our experiments. In this regard, a previous study that used freely licking rats and similar short time scale to analyse neural activity reported nearly 20% of the neurons from which they recorded were water responsive. In these experiments, several tastants were used to evaluate tuning breadth and rats were not trained to distinguish among the stimuli that they tasted (Stapleton et al. 2006).

Collectively, these results suggest that water should be considered as a taste. However, it is worth highlighting that as water was delivered in two different behavioural contexts during our task (i.e. as both a cue and a reward), we could compare the water response in both contexts. A pure water response – one that reflects the physicochemical properties of water (Rosen et al. 2010) – should be the same as when sensed as a cue and reward. Nearly 46% of the water selective neurons responded had the same response to water as a cue and reward, which is consistent with a pure water response. Alternatively, the remaining neurons showed a suppressed response (21% of the water selective neurons) or no response (33% of the water selective neurons) when water was delivered as a reward. Although the orientation of the rat's head while obtaining water as a cue or reward may explain some of these differences (the rewarded choice lick spouts were positioned on either side of the rat), this explanation seems unlikely for all cases because nearly a third of the water selective neurons showed a strong excitatory response to water as a cue but no response at all to water when presented as a reward (Figure 4C). Rather, these results lend support to the idea that tastant processing in the GC may be influenced by the behavioural context that underlies the taste (Yoshida & Katz, 2011; Samuelsen et al. 2012).

Summary

For the first time, we characterized neural activity in the GC of rats making behavioural choices that were guided by the concentration of NaCl that they tasted. We confirmed an assortment of taste-related responses, and distinguished three neural types that were sensitive to changes in NaCl concentration. Two neural types were activated more by the lowest NaCl concentration compared to the highest tested NaCl concentration, and could be distinguished by their response to the water cue. The third neural type showed its greatest response to the highest concentration of NaCl tested with no response to water. Importantly, neurons with this response profile showed the greatest relationship to the choices made by the rat that were guided by the concentration of NaCl, suggesting a central role for these neurons in signalling the perceived intensity of NaCl. Taken together, our results show that neurons in GC show an assortment of rapid, taste-related responses that play a role in taste concentration processing, and in doing so may contribute to a general function of mediating taste-guided behaviour (Braun, 1990).

Glossary

GC

gustatory cortex

ILI

interlick interval

Author contributions

C.J.M. and S.A.S. conceived and designed the experiments; C.J.M. and S.A.S. collected, analysed, and interpreted the data; C.J.M., W.H.M. and S.A.S. drafted the article.