Cortico-hippocampal interactions as a neural substrate for the retrieval of contextual memories in choice behavior

Abstract

Spatial working memory, the retention and use of behaviorally relevant spatial cues on a timescale of seconds, depends on complex, finely tuned interactions between hippocampus and the cortical regions anterior cingulate (ACC) and retrosplenial cortices (RSC), together hereby named medial mesocortex (MMC). In this circuit, the processing of hippocampal contextual information is hypothesized to follow a directional stream, from hippocampus to cortex, providing depolarizing drive to MMC neurons. The functional circuitry underlying these interactions and the necessity of such interactions for spatial working memory have not been established. Using retrograde and anterograde tracings, we reported the existence of a HIPP-MMC monosynaptic connection, and we characterized its topographic organization along the MMC. ACC is mainly targeted by the stratum pyramidale of dorso-intermediate HIPP (diHIPP), whereas RSC is targeted by pyramidal and non-pyramidal strata of diHIPP. In RSC, the hippocampal projection includes long-range GABAergic cells located at the border between stratum radiatum and stratum lacunosum-moleculare. Glutamatergic axons arising from diHIPP show sparse distribution and do not show preference for specific layers in the ACC. Contrarily, the glutamatergic axons arising from diHIPP project heavily to the superficial layers of RSC, particularly to layers 3 and 4, whereas the long-range GABAergic cells targeting RSC project mainly to layer 1. Using optogenetics, in vitro electrophysiology and sequential pharmacology, we showed that such hippocampal projections establish bona fide synapses throughout MMC cortical layers, and their differential targeting of ACC and RSC translates into a functional dichotomy at the microcircuit level. Specifically, the diffuse and excitatory hippocampal inputs to ACC evoke stronger potentials around layer 5, known for harboring large pyramids projecting descending axons to the basal ganglia, whereas the excitatory and inhibitory hippocampal inputs to RSC evoke stronger potentials in superficial layers (L1-3), where RSC sends and receives most corticocortical connections. By using in vivo multi site recordings, we further showed that the spontaneous activity patterns in the HIPP and MMC of the awake behaving rat follow what would be expected from the above described connectivity. First, epochs of increased spiking from HIPP are accompanied by short term increases in MMC areas, with increased levels generally preceding and following the trigger point, which is indicative of complex time--dependent cross--talk between these regions. Second, such increases are somewhat clearer in the anteriormost regions of MMC, implying that the presence of inhibitory in parallel with excitatory HIPP inputs to RSC modulates the cortical response in vivo in ways yet unexplored. Our data also showed that MMC spiking responses to HIPP have an oscillatory component, favoring frequencies known to play a significant role in hippocampal--cortical functions, and the strength of the oscillatory alignment to the HIPP rhythms increases as we move caudally along the MMC divisions, with the posteriormost RSC regions significantly more engaged to the hippocampal oscillations, under general wakefulness conditions. Our findings established the functional circuitry supporting HIPP--MMC interactions, and uncovered an underlying gradient of hippocampal inputs to the MMC. The intimate connection between RSC and HIPP, whereupon RSC receives inputs from all HIPP layers, excitatory and inhibitory, and shows increased hippocampal entrainment, is consistent with the known functional similarity of RSC and HIPP. ACC, on the other hand, receives diffuse, sparse and exclusively excitatory input from HIPP and the stronger potentials are evoked in layer 5, known to project to the basal ganglia, consistent with its role in behavior control.