The retrosplenial cortical role in encoding behaviorally significant cues

Abstract

The retrosplenial cortex (RSC) has recently begun to gain widespread interest because of its anatomical connectivity with other well-known memory structures, such as the hippocampus and anterior thalamus, and its role in spatial, contextual, and episodic memory. Although much of the current work on the RSC is focused on spatial cognition, there is also an extensive literature that shows that the RSC plays a critical role in a variety of conditioning tasks that have no obvious spatial component. Many of these studies suggest that the RSC is involved in identifying and encoding behaviorally significant cues, particularly those cues that predict reinforcement or the need for a behavioral response. Consistent with this idea, recent studies have shown that RSC neurons also encode cues in spatial navigation tasks. In this article, we review these findings and suggest that the encoding of cues is an important component of the RSC contribution to many forms of learning. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Figure 1

Neuronal responses to auditory cues during discriminative approach and avoidance learning. Plot A) Tone-evoked multi-unit neuronal responses to the auditory cues are shown in the form of z-scores normalized to pre-tone baseline, with 400 msec of firing data shown in 10 msec time bins (Smith et al., 2002). Before learning, RSC neurons respond equally to auditory cues of different tonal frequencies (left). After learning, neuronal firing in response to the predictive tone (CS+, black bars) is significantly greater than to the non-predictive tone (CS−, white bars). Plot B) RSC neuronal responses are shown for subjects trained to perform the approach and avoidance tasks on alternating days (Freeman et al., 1996). Note that RSC neurons preferentially respond to the reinforcement-predictive auditory cue in both tasks, regardless of differences in tonal frequency, hedonic value (appetitive vs aversive) and response requirements (approach or avoidance).

Figure 2

RSC responses during blocked alternation. Plot A) Rats were trained to approach the east arm for reward for the first 15 trials of each daily training session and the west arm for the next 15 trials. Each day, the rats were given a 30 sec lights-out period after trial 15 to indicate that the reward location was about to shift to the west arm. However, even in well-trained subjects (>80% correct overall), the rats never learned to use the ‘lights-out’ cue to shift their responding. Instead, they invariably went to the old (east) reward location until no reward was found there and only then did they shift to the west arm. Plot B) Average behavioral performance is shown for a baseline recording session prior to learning (pretraining, PT), the first training session (Day1), the session midway through acquisition (Mid) and during asymptotic performance (Asymp). Plot C) The percentage of RSC neurons that exhibited a significant response that was selective for both the receipt of the reward and the location (east or west) at each of the same training stages as in plot B. Plots D and E) Two examples of RSC neurons that selectively responded to one reward location are shown in the form of peri-event time histograms showing 10 seconds of firing data before and after the receipt of the reward at time zero, with trial by trial rasters below each histogram. The neuron in plot D fired selectively for rewards received at on the east arm while the neuron in plot E fired on the west arm.

Figure 3

RSC responses during a cued T-maze task. Plot A) Rats were trained to approach a flashing light cue positioned over the right or left reward location (left reward trial is illustrated). The rat was placed on the stem of the maze facing away from the choice point and the light was illuminated as soon as the rat turned around and took a step forward (dashed line). Before regular training sessions began, we recorded baseline responses to a light cue positioned at the choice point (dashed circle) which was illuminated during half of the trials in a random manner. This light was only used during this pretraining session and it did not predict the reward locations, which were also randomly selected. Plot B) Rats learned this task in 6.4 training sessions on average, and reached an asymptote of 94% correct. Plot C) The percentage of RSC neurons that exhibited a significant response to the light cue are shown for each training stage (same as in plot 1B). Plot D) An example RSC neuron that responded to the onset of the light cue (time zero) is shown in the form of a peri-event time histogram along with a raster display. From Vedder, L. C., Miller, A. M. P., Harrison, M. B., & Smith, D. M. (2017). Retrosplenial Cortical Neurons Encode Navigational Cues, Trajectories and Reward Locations During Goal Directed Navigation. Cerebral Cortex, 27(7), pp 3716, 3717. Adapted with permission.

Figure 4

Other RSC responses from the cued T-maze task. Plot A) The percentage of RSC neurons that exhibited a significant response that was selective for both the receipt of the reward and the location (left or right) is shown for each training stage (same as in plot 1B). Note that these responses did not emerge immediately on the first day of training. Plot B) An example neuron with a reward-location response is shown in the form of a peri-event time histogram with a raster display. Trials with the reward on the right are shown in blue while trials with the reward on the left are shown in red. Left and right trials were randomly intermixed and they are only separated in the raster for illustration. Plot C) The percentage of RSC neurons that encode an explicit navigational cue, such as the light cue, increases dramatically on the first day of training (solid line). In the blocked alternation study where rats used a win-stay strategy (Fig 2), the reward and its location may have served as an important navigational cue and similarly rapid encoding was seen (fine dashed line). In contrast, the reward location could not be used as a cue in the light cued T-maze study because the current reward did not indicate the location for the upcoming reward which randomly switched from one trial to the next. Under these conditions, the responses did not show a sudden increase early in training (coarse dashed line). Data are expressed as the increase in the percentage of neurons at each stage of training, relative to the pre-training baseline. Plot D) An example neuron that exhibited significant responses to three separate task events is shown in the form of a peri-event time histogram aligned to receipt of the reward (time zero). Additional events are indicated by arrows (S=trial start, L=light onset and R=reward). From Vedder, L. C., Miller, A. M. P., Harrison, M. B., & Smith, D. M. (2017). Retrosplenial Cortical Neurons Encode Navigational Cues, Trajectories and Reward Locations During Goal Directed Navigation. Cerebral Cortex, 27(7), pp 3716, 3717. Adapted with permission.

Figure 5

Theoretical model of RSC interactions with the hippocampus. Early in learning, visuospatial and other sensory information arrives at the RSC where behaviorally significant cues are identified and encoded. Large scale RSC responses to these predictive cues emerge rapidly (e.g. Fig. 3) and may play an important role in establishing or orienting hippocampal representations (blue pathway). In contrast, RSC spatial and contextual representations, which do not rely on reinforcement contingencies, emerge slowly and may reflect consolidation of information from the hippocampus, where such representations are known to emerge rapidly (red pathway). These interactions could arise from relatively direct anatomical pathways via the subiculum or CA1 projections to the RSC (direct arrows) or through more indirect routes involving the anterior thalamus and entorhinal cortex.