Distinct roles of dentate gyrus and medial entorhinal cortex inputs for phase precession and temporal correlations in the hippocampal CA3 area

Abstract

SummaryThe hippocampal CA3 subregion is a densely connected recurrent circuit that supports memory consolidation and retrieval by generating and storing sequential neuronal activity patterns that reflect recent experience. While theta phase precession is thought to be critical for generating sequential activity during memory encoding, the circuit mechanisms that support this computation across hippocampal subregions are unknown. By analyzing CA3 network activity in the absence of each of its theta modulated excitatory inputs, we show necessary and unique contributions of the dentate gyrus (DG) and the medial entorhinal cortex (MEC) to phase precession. DG inputs are essential for generating the preferential spiking of CA3 cells during late theta phases and for organizing the temporal order of neuronal firing, while MEC inputs modulate the general precision of phase precession. A computational model that accounts for the empirical findings suggests that DG inputs affect the phase and MEC inputs affect the amplitude of inhibitory subnetworks. Our results thus identify a novel and unique functional role of the DG for the generation of sequence coding in the CA3 recurrent circuit.

A theory of hippocampal theta correlations accounting for extrinsic and intrinsic sequences

“…One prediction of the DG-loop model, consistent with findings from DG lesion experiments ( Ahmadi et al, 2022 ), is that DG would contribute to the temporal organization of spike sequences in CA3. To verify this hypothesis also in the model, we implemented a lesion of DG by disabling activity in the DG layer.…”

“…Theta compression (Dragoi and Buzsáki, 2006) measures how much theta phase encodes a certain interval in space and is quantified by magnitude of the linear-circular regression slopes (a, in radians per maximum pair distance) of the lag-distance plots. Theta compression is lower for the DG lesion case (a = 0.54 for both θ DG = 0 ° and θ DG = 180 °), as compared to the control case (a = 2.23 for θ DG = 0 ° and a = −0.77 for θ DG = 180 °) reproducing the finding (Ahmadi et al, 2022) that spatial encoding via theta sequences crucially depends on intact DG and predicting that loss of DG inputs is compensated for by the increase of release probability in the spared afferent synapses from the MEC.…”

“…To explain the expression of intrinsic sequences in CA3, we propose them to be generated by the interaction of two networks, CA3 and DG ( Figure 3A ). DG is a good candidate region to be involved in phase precession, since lesions of it were shown to reduce prospective spiking ( Sasaki et al, 2018 ) and to lower the onset phase of phase precession ( Ahmadi et al, 2022 ). In our model, the neurons in DG receive excitatory synaptic inputs from CA3 place cells (putatively via hilar mossy cells) and project back to the CA3 cells with place field centers at a different location ( Equation 3 ) to induce propagation of intrinsic sequences along a specific spatial direction.…”

“…This indicates that the excitatory recurrent pathway from CA3 via DG exists and might allow activity reverberation between two layers. Moreover, lesions of the DG layer were shown to eliminate the coordinated temporal structure of CA3 activity and to be instrumental to sequence organization (Figure 4 and (Ahmadi et al (2022))).…”

“…Here we present a network model that accounts for both types of correlations by separating their generation into two anatomically distinct layers: CA3 and dentate gyrus (DG). Extrinsic sequences are generated in the CA3 layer by short-term synaptic plasticity mechanisms, while the intrinsic sequences are evoked by the CA3-DG feedback loop with fixed asymmetrical weights, as inspired by the experimental evidence that lesions of DG abolish non-local activation of CA3 place cells ( Sasaki et al, 2018 ) and CA3 theta correlations ( Ahmadi et al, 2022 ).…”