Medial Frontal Theta Is Entrained to Rewarded Actions

Abstract

Rodents lick to consume fluids. The reward value of ingested fluids is likely to be encoded by neuronal activity entrained to the lick cycle. Here, we investigated relationships between licking and reward signaling by the medial frontal cortex (MFC), a key cortical region for reward-guided learning and decision-making. Multielectrode recordings of spike activity and field potentials were made in male rats as they performed an incentive contrast licking task. Rats received access to higher- and lower-value sucrose rewards over alternating 30 s periods. They learned to lick persistently when higher-value rewards were available and to suppress licking when lower-value rewards were available. Spectral analysis of spikes and fields revealed evidence for reward value being encoded by the strength of phase-locking of a 6-12 Hz theta rhythm to the rats' lick cycle. Recordings during the initial acquisition of the task found that the strength of phase-locking to the lick cycle was strengthened with experience. A modification of the task, with a temporal gap of 2 s added between reward deliveries, found that the rhythmic signals persisted during periods of dry licking, a finding that suggests the MFC encodes either the value of the currently available reward or the vigor with which rats act to consume it. Finally, we found that reversible inactivations of the MFC in the opposite hemisphere eliminated the encoding of reward information. Together, our findings establish that a 6-12 Hz theta rhythm, generated by the rodent MFC, is synchronized to rewarded actions.SIGNIFICANCE STATEMENT The cellular and behavioral mechanisms of reward signaling by the medial frontal cortex (MFC) have not been resolved. We report evidence for a 6-12 Hz theta rhythm that is generated by the MFC and synchronized with ongoing consummatory actions. Previous studies of MFC reward signaling have inferred value coding upon temporally sustained activity during the period of reward consumption. Our findings suggest that MFC activity is temporally sustained due to the consumption of the rewarding fluids, and not necessarily the abstract properties of the rewarding fluid. Two other major findings were that the MFC reward signals persist beyond the period of fluid delivery and are generated by neurons within the MFC.

Keywords: licking; muscimol; prefrontal; reward; sucrose; theta.

Figure 1.

Behavioral task. A, Rats were tested in an incentive contrast procedure called the shifting values licking task (Parent et al., 2015a). They were required to lick on a spout to receive liquid sucrose rewards. Reward values shift between relatively high (20% w/v) and low (4% or 2% w/v) concentrations of sucrose every 30 s. B, Experienced rats (fourth training session) licked more for the high-value sucrose than for the low-value sucrose (paired t test; t₍₈₎ = 4.29, p = 0.0026). *p < 0.005.

Figure 2.

Neuronal recordings. A, All rats (N = 11) were implanted with a 16-channel microwire array targeting the rostral MFC in one hemisphere. A subset of rats (N = 4) had a drug cannula implanted in the same cortical area in the opposite hemisphere. B, Locations of recording sites are depicted on a horizontal section from the Paxinos and Watson (1997) atlas. All electrodes were placed within the prelimbic (PrL) and medial agranular (M2) regions. C, Validation of cross-hemispheric connections for this rostral MFC region. Cholera toxin subunit B with the AlexaFluor-488 reporter was injected in the rostral MFC of 5 rats. Green (left hemisphere) represents injection site spread. Neurons were labeled in the superficial layers in the opposite hemisphere (right).

Figure 3.

Neuronal activity in the MFC is entrained to the lick cycle. A, An example of an LFP recording shows clear fluctuations at the times of licks (tick marks above the LFP). B, Relationships between LFP signals and licking were assessed by bandpass filtering the LFPs near the licking frequency (defined by the interquartile range around the median ILI) and applying the Hilbert transform to measure the amplitude and phase of licking-related neuronal activity. Instantaneous phase was plotted using polar coordinates and analyzed with standard methods for circular statistics (Agostinelli and Lund, 2013). For details, see Materials and Methods.

Figure 4.

Spatial distribution of entrainment to the lick cycle. A, Spatial plot of phase tuning using the test statistic from Rao's spacing test of uniformity showed no obvious topography of lick entrainment in the MFC. Individual electrode locations were plotted according to their location in reference to bregma (N = 159 electrodes). Recording sites were depicted as circles colored by the strength of their Rao test statistic (U). Color bar represents values of U from the 5th-95th percentile range over all recording sites. Values above the black bar (near 135) were not uniform (p < 0.05). B, Polar plots represent phase tuning examples from four spatial extremes of the graph in A. The most rostral/lateral (top left; U = 134.48, p > 0.05), rostral/medial (top right; U = 152.30, p < 0.001), caudal/medial (bottom right; U = 153.51, p < 0.001), and caudal/lateral (bottom left; U = 147.44, p < 0.001) electrodes were chosen. There was no drastic difference among the four locations with regard to phase tuning. C, Group summaries of the mean phase angle at the time of licking from all 11 rats reveal significant phase tuning toward 0 degrees (i.e., peak or trough of the rhythm). These results were compared with phase angles measured from surrogate data (shuffled ILIs), which did not show evidence for significant phase entrainment.

Figure 5.

Time-frequency analysis of lick-entrained LFP data. ERSP (top) and ITC are shown for a typical LFP recording aligned to the time of licking in the behavioral task. White horizontal dashed line indicates the median licking frequency. White vertical dashed lines indicate the median ILIs. ERSP and ITC measures were computed using observed licks (left) and surrogate data (middle), created by shuffling ILIs. A, Persistent elevated ERSP was notable at very low frequencies (∼2 Hz, or delta) for both the observed (top left) and shuffled (top middle) events (i.e., was not entrained to the lick cycle). Subtraction of the shuffled ERSP matrix from the observed ERSP matrix revealed elevated power at the licking frequency (horizontal dashed line). B, ITC was apparent near the licking frequency over a period of 2 lick cycles for the observed licks (lower left), but not the shuffled licks (bottom middle). Subtraction of the shuffled ITC matrix from the observed ITC matrix revealed elevated power at the licking frequency (horizontal dashed line).

Figure 6.

Entrainment was stable over the 30 min test sessions. Sessions were split into 10 blocks with equal numbers of licks, and peak ERSP and ITC were measured in the theta frequency range (6–12 Hz) over the ILI before and after each lick. A, Group average of peak ITC showed no evidence for a change in this measure over the datasets. Similar results were obtained for ERSP (data not shown). B, Traces for peak ITC from each of the 11 rats. C, Grand average of ERSP and ITC for all LFPs in the first and last block. Together, these results suggest that entrainment of MFC LFPs to the lick cycle was not sensitive to cross-session factors, such as satiety.

Figure 7.

MFC theta entrainment to licking develops with experience. A, Recordings were made in a subset of 3 rats as they learned the behavioral task. The rats showed increased licking for the high-value sucrose compared with the low-value sucrose after the first training session and the relative difference in licking increased over the first four training sessions. B, Neuronal entrainment to the lick cycle developed with experience in the task. For example, ERPs increased in size and apparent rhythmicity between the first and fourth training session. Blue represents higher-value 20% sucrose. Green represents lower-value 2% sucrose. C, Increased entrainment to the lick cycle was also apparent in ITC, which was not apparent in Session 1 and specific to licks that delivered high-value sucrose in Session 4. White vertical lines indicate average ILIs across the session. White horizontal dashed line indicates average licking frequency across the session. Magenta ticks in the color bars indicate average ERSP or ITC at the median licking frequency. D, To capture differences in ITC values for the two types of licks across all recordings, we used a value index, defined as ((ITC-HI − ITC-LO)/ITC-HI). The index was based on the peak ITC values in a temporal window ranging from 1 ILI before lick onset up to 50 ms after the lick and for all frequencies between 4 and 12 Hz (“theta”). As shown in the parallel line plot, in which each line indicates an LFP recording from a distinct electrode, this index was larger in Session 4 compared with Session 1 (paired t test: t₍₃₉₎ = −12.085, p < 10⁻⁶). *p < 0.05, **p < 10⁻⁶.

Figure 8.

Coherence between spikes and licks reflects reward information. A, MUA was entrained to the lick cycle. Blue represents high-value licks. Green represents low-value licks. Rasters were sorted by the latency to the last lick before the lick at time 0, with the shortest preceding intervals at the top of the raster. The high-value licks were subsampled for this plot so that neural activity could be compared for the same number of total licks (at time 0). Perievent histograms (bin: 1 ms, 10-point Gaussian smoothing), below the raster plots, denote the probability of spiking around the times of the licks. B, Group summary for spike probability at times of higher- and lower-value licks. Blue lines indicate higher spike probability for the higher-value sucrose. Green lines indicate higher spike probability for the lower-value sucrose. Spike probability was higher at the times of the higher-value licks compared with times of the lower-value licks (paired t test: t₍₄₃₎ = 3.78, p < 0.001). A total of 33 of the 44 MUA recordings showed higher spike probabilities for the higher-value licks. C, Spike-field coherence found that all 44 MUA recordings were entrained to the LFP fluctuations that encoded reward information. Power spectra are shown in the top row for example LFP and MUA recordings. Peak power was near the licking frequency (black dashed line) for the LFP. The main peak for the spike train was in the low β range (12–15 Hz), and a second peak was at the licking frequency. Coherence between these signals (bottom left plot) was found at the licking frequency (5.96 Hz), at a level approximately twice the 95% CI. The phase between the spikes and fields at the licking frequency (bottom right plot) was near −π, suggesting that the spikes and fields had an antiphase relationship. *p < 0.001.

Figure 9.

Reward context, not reinforcement per se, drives reward signaling. A, The shifting values licking task was modified to include a 2 s period between reward deliveries. This period allowed for nonreinforced licks (dry licks at the spout) to be recorded within the 30 s states of high or low-value sucrose availability. B, Group summary (N = 6) of licks per unit time (total licks emitted in each context divided by time spend in each context). This measure revealed that rats licked less in the nonreinforced lower-value blocks compared with the other blocks. C, Peak ERSP values for reinforced versus nonreinforced licks during the high-value blocks. Lines are colored by their direction (increase or decrease in power). There was no difference in power for reinforced versus nonreinforced licks (F_(1,359) = 2.52, p = 0.11). D, Peak ITC values for reinforced versus nonreinforced licks during the high-value blocks. The majority of LFPs showed increased phase-locking to nonreinforced licks (blue lines), whereas electrodes from 2 rats show a slight decrease in phase-locking for nonreinforced licks (green lines). Overall group summaries show an increase in phase-locking for the nonreinforced licks (F_(1,359) = 31.94, p < 10⁻⁶). E, F, Example of time-frequency analysis of an LFP from a rat that showed decreased ERSP and ITC (magenta box) when the rat licked in the lower-value context. ITC was higher near the licking frequency when the higher-value reward was available, regardless of whether the licks were reinforced or not. Horizontal white lines indicate the within-session licking frequencies. Vertical white lines indicate the ILIs for each session. Magenta ticks in the color bars represent average ERSP or ITC at the median licking frequency. *p < 0.05, **p < 10⁻⁶.

Figure 10.

Reward signaling depends on neuronal activity in the MFC. Rats were tested with an electrode array in one hemisphere and an infusion cannula in the other, which was used to infuse either PBS or muscimol. A, ERPs from the saline (blue line) and muscimol (yellow line) sessions showed a similar overall time course around the licks. B, All electrodes showed a decrease in peak ERSP at the licking frequency for higher-value licks during the muscimol sessions compared with the PBS sessions (F_(1,123) = 96.09, p < 10⁻⁵). C, Likewise, there was a reduction in peak ITC at the licking frequency for 28 of 32 electrodes (F_(1,123) = 18.17, p = 3 × 10⁻⁵). D, E, Example of time-frequency analysis. Effects were specific to licks for the high-value reward. Horizontal white lines indicate the within-session licking frequencies. Vertical white lines indicate the ILIs for each session. Magenta ticks in the color bars represent average ERSP or ITC at the median licking frequency. *p < 10⁻⁵.