Spatially-modulated neurons in the hippocampus and associated cortical areas tend to have well-defined spatial firing fields, spiking either at a particular point in space (place cells) or at the vertices of a triangular lattice covering the environment (grid cells). As the animal explores the enclosure and the hippocampus is engaged by theta oscillations, activity of hippocampal place cells is specific to animal’s position and drops to zero when the animal moves out of the cell’s spatial firing field. On the other hand, when the hippocampus is not in theta mode (e.g. during quiet rest or slow-wave sleep), place cells sometimes fire independently of animal’s current position. These place cell sequences, manifesting as brief sharp wave-ripple (SWR) oscillatory events (2), reflect the sequences of place fields crossed during prior exploration or learning periods and are referred to as ‘replay’ of recent experience. A substantial body of work established that hippocampal replay is important for consolidation of recently formed memories, as well as for planning and reward learning. In addition, several studies have shown that hippocampal replay is associated with reactivation in other parts of the brain such the striatum (3), neocortex (4), as well as reward-associated neuromodulatory nuclei (5). Thus, it is tempting to speculate that the hippocampus binds different components of a memory into a brain-wide neural ensemble through SWR-associated replay. To further investigate the extend to which extrahippocampal structures participate in memory reactivation, O’Neill and colleagues trained rats on a spatial task and performed simultaneous unit recordings in the hippocampus and medial entorhinal cortex (MEC) – the brain area that his highly interconnected with the hippocampal formation and contains spatially-tuned neurons such as grid cells, border cells and head-direction cells.
The authors focused their analysis on the superficial layers of MEC (sMEC) – the main input to the hippocampal formation. They simultaneously recorded neurons in MEC and CA1 while rats performed a delayed non-match to sample task on a continuous T-maze. The task consisted of two phases. In the sample phase rats were forced to make a particular turn at the choice point (left or right, with one choice blocked), which then led them to the reward location. They were then given a short delay and proceeded to the ‘choice’ phase, where this time they could choose to turn either left or right at the same choice point, with the rewarded choice being the opposite of the forced turn made during the sample phase (hence ‘non-match to sample’). Their first important observation was that although sMEC neurons maintained firing fields across training trials, they varied their firing rates depending not only on the type of trial (sample or choice) but also on the future trajectory of the rat (left or right turn). Such prospective ‘splitter’ coding (or rate remapping) has been previously described in place cells and indicates that sMEC cells code for more than just the metric of space.
Recording a large number of neurons simultaneously enabled the authors to investigate the population dynamics of sMEC cells during the task. They observed that on a population level sMEC cells exhibit occasional periods of highly coherent spiking, with many cells firing in a sequence outside of their spatial firing fields. Authors found that these short (~100 ms) events represented continuous trajectories along the track. These replayed routes often started close to the rat’s current location, extended in both forward and reverse directions and were mainly observed in the delay compartment as well as near the reward sites, with notable near-absence at the choice point. Surprisingly, awake replay of trajectories in sMEC and HPC occurred at different times and sMEC replay was not time-locked to hippocampal SWRs, indicating that MEC cells engage in reactivation independently of the hippocampus. Authors applied the same analysis to the post-training sleep epochs and made the same observation – CA1 and sMEC reactivation events were not correlated in time.
Replay in the MEC was also observed by another research group, who focused their analysis on deep MEC layers (dMEC) that constitute the main hippocampal output (6). Interestingly, Ólafsdóttir and colleagues did find an association between dMEC and HPC replay, with dMEC reactivation occurring mainly during forward replay and lagging behind the hippocampus by about 10 ms, consistent with dMEC being downstream of HPC. In order to reconcile these apparently discrepant results, one should address both the possible theoretical implications as well as methodological differences.
The intriguing possibility emerging from these two studies is that deep and superficial MEC layers may engage differently with the hippocampus. It is possible that sMEC and dMEC are parts of two independent replay networks: dMEC relays the reactivation initiated in the HPC to the rest of the brain while sMEC engages in reactivation that is independent of the HPC-associated replay network. This possibility could be investigated in a challenging follow-up study involving simultaneous recordings in both deep and superficial layers of the MEC.
However, the perceived differences in findings could be also explained by differences in experimental approach. Ólafsdóttir and colleagues centered their analysis on the hippocampal replay events and looked at grid cell activity time-locked to these sequences – an approach that was necessitated by small number of simultaneously recorded grid cells. They then identified putative dMEC replay events by matching activity of single grid cells to the trajectories replayed in the HPC. In contrast, O’Neill and colleagues detected MEC replay independently of HPC reactivation and were able to detect replay events that would have been missed by the other approach. Interestingly, a small proportion of sMEC replay events (~6%) detected by O’Neill et al were temporarily aligned with hippocampal SWRs and authors also report a weak but significant coherence between replay events in sMEC and HPC. In fact, both papers reported increased coherence between HPC replay and MEC cell firing during HPC sleep-replay events, and it would be of interest to perform the HPC-centric analysis, used by Ólafsdóttir et al, selectively on these SWR-associated sMEC activity bursts as it is indeed possible that it may result in similar conclusions.
Hippocampal replay is such a captivating phenomenon largely because it is emerges based on animal’s recent experience and thus can be interpreted as reactivation of stored memory traces. Therefore, a considerable caveat of the O’Neill et al study (also raised in the corresponding ‘News and Views’ (7)) is that the authors did not investigate whether observed MEC sequences emerge only after the learning experience and are not ‘pre-wired’ into the network. Indeed, coherent sequential activity that is independent of experience has been described in several other associated areas including the hippocampus (8) (‘pre-play’) as well as the head-direction network (9). The experimental protocol employed in the study involved both pre- and post-training sleep periods, and it would be of interest to see whether sMEC replay events become more prevalent after the rats experience the maze. Overall, the discussed article introduces a paradigm-shifting possibility that offline reactivation is not always orchestrated by the hippocampus and can often happen without hippocampal engagement, however radical this idea may seem to those with a hippocampocentric view of the memory systems. Future studies are now needed to investigate whether reactivation in other brain areas is correlated with replay in MEC rather than HPC. Still, it is imperative to assess whether sMEC replay is indeed related to memory, or does it simply reflect the non-random architecture of the entorhinal circuitry.
1. O’Neill, J., Boccara, C. N., Stella, F., Schoenenberger, P. & Csicsvari, J. Superficial layers of the medial entorhinal cortex replay independently of the hippocampus. Science 355, 184–188 (2017).
2. Buzsáki, G. Hippocampal sharp wave-ripple: A cognitive biomarker for episodic memory and planning. Hippocampus 25, 1073–1188 (2015).
3. Pennartz, C. M. A. et al. The ventral striatum in off-line processing: ensemble reactivation during sleep and modulation by hippocampal ripples. J. Neurosci. 24, 6446–6456 (2004).
4. Peyrache, A., Khamassi, M., Benchenane, K., Wiener, S. I. & Battaglia, F. P. Replay of rule-learning related neural patterns in the prefrontal cortex during sleep. Nat. Neurosci. 12, 919–926 (2009).
5. Valdés, J. L., McNaughton, B. L. & Fellous, J.-M. Offline reactivation of experience-dependent neuronal firing patterns in the rat ventral tegmental area. J. Neurophysiol. 114, 1183–1195 (2015).
6. Ólafsdóttir, H. F., Carpenter, F. & Barry, C. Coordinated grid and place cell replay during rest. Nat. Neurosci. 19, 792–794 (2016).
7. Gardner, R. J. & Moser, M.-B. Multiple mechanisms for memory replay? Science 355, 131–132 (2017).
8. Dragoi, G. & Tonegawa, S. Preplay of future place cell sequences by hippocampal cellular assemblies. Nature 469, 397–401 (2011).
9. Peyrache, A., Lacroix, M. M., Petersen, P. C. & Buzsáki, G. Internally organized mechanisms of the head direction sense. Nat. Neurosci. 18, 569–575 (2015).