Cues, context, and long-term memory: the role of the retrosplenial cortex in spatial cognition

Abstract

Spatial navigation requires memory representations of landmarks and other navigation cues. The retrosplenial cortex (RSC) is anatomically positioned between limbic areas important for memory formation, such as the hippocampus (HPC) and the anterior thalamus, and cortical regions along the dorsal stream known to contribute importantly to long-term spatial representation, such as the posterior parietal cortex. Damage to the RSC severely impairs allocentric representations of the environment, including the ability to derive navigational information from landmarks. The specific deficits seen in tests of human and rodent navigation suggest that the RSC supports allocentric representation by processing the stable features of the environment and the spatial relationships among them. In addition to spatial cognition, the RSC plays a key role in contextual and episodic memory. The RSC also contributes importantly to the acquisition and consolidation of long-term spatial and contextual memory through its interactions with the HPC. Within this framework, the RSC plays a dual role as part of the feedforward network providing sensory and mnemonic input to the HPC and as a target of the hippocampal-dependent systems consolidation of long-term memory.

Keywords: allocentric; consolidation; context; hippocampus; learning; long-term memory; navigation; retrosplenial cortex.

Figure 1

Retrosplenial connectivity. (A) The retrosplenial cortex (RSC) is centrally positioned between cortical sensory regions (blue) and limbic memory regions (green). The parietal lobe merges information arriving from early sensory areas, and shares connectivity with both the medial prefrontal cortex (mPFC) and the RSC. Reciprocal connections between the RSC, anterior thalamic nuclei (ATN), and the hippocampus (HPC) constitute a limbic memory circuit that is essential for many forms of learning and memory. (B) The RSC shows regional differences in connectivity with limbic (green) and cortical (blue) regions. The granular RSC (areas 29a-c) shows greater connectivity with limbic regions such as the subicular cortex and the antero-dorsal (AD) and antero-ventral (AV) nuclei of the ATN, while the dysgranular RSC (area 30) shows greater connectivity with cortical regions including the parahippocampal region, the posterior parietal cortex, and early visual areas. The connections of the ATN also differ by region, with the AD and AV nuclei showing greater connectivity with limbic areas and the antero-medial (AM) nucleus showing greater connectivity with neocortical regions.

Figure 2

Impaired map drawing by patient with RSC lesion. Patients with RSC damage are severely impaired at drawing maps from memory. This patient (Patient 1, Takahashi et al., 1997) was incapable of recalling the locations of buildings in his neighborhood, despite being able to recall their names (left drawing). His wife’s drawing is shown on the right for comparison. Adapted from Takahashi et al. (1997).

Figure 3

Neural coding by RSC and hippocampal neurons during navigation. (A) A schematic of the blocked alternation task from Smith et al. (2012). Rats navigated to one location for a reward during the first 15 trials of every session (east arm), and navigated to a different location (on the same maze) during the second 15 trials (west arm). (B) Rats used the reward location as a navigation cue. Line graph shows average percent correct on each trial during asymptotic performance of the blocked alternation task. Rats employed a win-stay strategy, returning to the location that was rewarded on the previous trial. After discovering that the reward location had changed on trial 16, the rats began navigating to the new reward location. (C) Examples of RSC place cells and (D) hippocampal place cells. (E) Examples of RSC location-specific reward responses. Each pair of perievent time histograms shows the spiking activity of a single neuron, with the left and right histograms showing firing during the 10 s before and after the go east and go west rewards. RSC neurons occasionally showed punctuate firing at the instant of one reward (cells 1–2). Other RSC neurons showed sustained elevated firing while the rat consumed the reward in one location, and decreased firing during the other reward (cells 3–4). (F) Navigation performance improved with training. The average percentage of trials in which the rats made the correct arm entry is shown for random foraging pretraining (RF), the first training session (ACQ1), the session half way through training (middle acquisition, MACQ), and asymptotic performance (ASYMP). (G) The proportion of RSC neurons that differentially encoded the two reward locations increased dramatically on the first day of training (when the reward locations became useful navigation cues) and then remained high throughout the rest of training. (H) The proportion of RSC neurons with place fields increased steadily with training. (I) Although the proportion of hippocampal neurons with place fields remained the same throughout training (not shown), the activity of these neurons became more distinct in the Go East and Go West conditions as the rats learned the task (average spatial correlations, Pearson’s r). Adapted from Smith et al. (2012).

Figure 4

Overlapping activation between the default network and the context association network. (A) Medial view of the default network. The labeled regions, including the RSC, are those that typically show greater BOLD activation during inter-trial rest than during task performance. (B) An activation map shows the difference between perceiving highly contextual objects (e.g., a shopping cart) and weakly contextual objects (e.g., a rope). Data were obtained by averaging together six similar experiments. The superimposed outlines of default network areas demonstrate the overlap between context processing and the default network. (C) Activity related to context processing in the regions of overlap with the default network, as manifested by percent of signal change. In each of these regions, highly contextual objects elicited either stronger positive or less negative activation compared with weakly contextual objects. Abbreviations: retrosplenial cortex (RSC), posterior cingulate cortex (PCC), parahippocampal cortex (PHC), medial temporal lobe (MTL), medial prefrontal cortex (MPFC). Adapted from Bar et al. (2007).

Figure 5

RSC and hippocampal activation during spatial navigation learning. (A) Participants received first-person tours of a virtual maze. On the left is a bird’s-eye-view diagram of the maze (not shown to participants) showing the locations of the landmarks and the route of the tour. In the upper right is an example of the first-person view of a landmark. The lower right shows an example test question. Participants indicated by button press the relative position of the small building, imagining that they were standing in front of the large building. (B) RSC BOLD activation increased along with behavioral accuracy over the course of training. (C) In contrast, hippocampal activation was specifically correlated with the slope of the learning curve (data from one participant shown). Adapted from Wolbers and Buchel (2005).

Figure 6

Working model of RSC-hippocampal interaction during learning. A model of RSC-hippocampal interactions over the course of learning. (A) Encoding in naïve animals is characterized by feedforward sensory information (blue lines) driving plasticity in both the cortex and the HPC. Plasticity in the RSC (circles and their connections) includes the early formation of cue-related activity, while plasticity in the HPC leads to the formation of a stable hippocampal memory representation (red). (B) After the learning event, interactions between the RSC and the HPC (dashed red lines) consolidate memories into a more stable form. (C) In the fully trained subject, feedforward sensory input activates the consolidated RSC memory representation, which, in turn, activates the corresponding representation in the HPC. (D) Coordinated activity in the RSC and HPC enables memory updating, whereby the detection of novelty by the HPC initiates rapid consolidation of this new information into the existing cortical memory trace. Abbreviations: retrosplenial cortex (RSC), hippocampus (HPC).