Abstract
Manuel Valero, Robert G. Averkin, Ivan Fernandez-Lamo, Juan Aguilar, Diego Lopez-Pigozzi, Jorge R. Brotons-Mas, Elena Cid, Gabor Tamas, Liset Menendez de la Prida. Neuron 2017;94:1234–1247.e7.
Memory traces are reactivated selectively during sharp-wave ripples. The mechanisms of selective reactivation, and how degraded reactivation affects memory, are poorly understood. We evaluated hippocampal single-cell activity during physiological and pathological sharp-wave ripples using juxtacellular and intracellular recordings in normal and epileptic rats with different memory abilities. CA1 pyramidal cells participate selectively during physiological events but fired together during epileptic fast ripples. We found that firing selectivity was dominated by an event-and cell-specific synaptic drive, modulated in single cells by changes in the excitatory/inhibitory ratio measured intracellularly. This mechanism collapses during pathological fast ripples to exacerbate and randomize neuronal firing. Acute administration of a use- and cell-type-dependent sodium channel blocker reduced neuronal collapse and randomness and improved recall in epileptic rats. We propose that cell-specific synaptic inputs govern firing selectivity of CA1 pyramidal cells during sharp-wave ripples.
Commentary
Memory consolidation is the neural process of stabilizing a memory for long-term storage. During consolidation, a fundamental change occurs in the neural representation of the remembered event. Think about your day. In a few seconds, you can mentally run through a full day's events even though they happened across hours. Therefore, there is a mechanism in your brain that transforms the neural code between those two timescales. Such a mechanism is necessary because stabilizing memory involves synaptic plasticity, which depends on neurons being coactive in millisecond time windows. This temporal compression is thought to involve a specific neuronal activity pattern in the hippocampus called sharp wave ripple (SWR) (1). Observed as fast and transient oscillations in local field potential (LFP) recordings, SWR caught the attention of neuroscientists because of what individual neurons are doing during them. When rats are exploring an environment, neurons in the hippocampus fire in sequences. Neurons in the hippocampus, so-called ‘place cells,’ are active depending on the spatial location of the animal—the sequential activation of place cells thus happens in real time according the position of the rat (2). Later, when rats are sleeping, those same sequences (correlating to a segment of the path that the rat took) repeat during SWR, but now at compressed time scales (~100 ms versus several seconds) (3).
SWR are of particular interest in epilepsy research for two reasons. One, patients with temporal lobe epilepsy have memory impairment. Two, in epilepsy, the brain generates bouts of high frequency oscillations that are clearly pathological (pHFOs) (4). In temporal lobe epilepsy, things get especially complicated because the same network that usually generates SWR can also generate pHFOs, which begs the question, what is going on with memory consolidation processes under these conditions. Studying SWR in the context of epilepsy presents the unique opportunity for simultaneously learning more about how SWR work and about the mechanisms for disrupted memory processing in epilepsy patients.
With these questions in mind, Valero et al. set out to understand how SWR processes are working (or not) in rats with epilepsy. They were interested in the mechanisms central to normal consolidation processes, so they eliminated recording sessions in which they observed interictal spikes. Therefore, the fast oscillations studied here were not the type that can be associated with interictal spiking. The first question was whether SWR in animals with epilepsy were different from those recorded in control. SWR in animals with epilepsy had more spectral power in high-frequency bands, which they quantified using a parameter called ‘fast ripple index’. Furthermore, they had higher spectral entropy (disorganization). Next, they found that rats with epilepsy had deficits in object recognition tasks. These tasks require that animals recognize objects from previous experiences and therefore should depend on proper consolidation of prior experience (5). They found that the prevalence of ripples with power in the high frequency bands (high ‘fast ripple index’) was the best parameter for explaining the variance in behavior (sessions with worse performance had inter-trial sleep sessions with SWR with higher fast ripple indexes). This pointed to a link between disrupted SWR and memory consolidation problems.
They next studied what was going on at the single neuron level during SWR, which would give a framework for thinking about why these faster, more disorganized SWR would be causing memory consolidation deficits. They found that in animals with epilepsy, individual neurons fired at higher rates during individual SWR and also participated in a higher proportion of SWR. This latter result suggested that neurons were less discriminating, which the authors proved with an analysis based on the idea that the exact identities of neurons active during a particular SWR lead to unique LFP patterns. In the analysis, SWRs were grouped into families based on their similarity to each other—the hypothesis being that they are similar (belong to the same family) because during those SWRs the active neural ensembles are also the same. Indeed, in control animals neurons fired faithfully within the same family of SWR, but in epilepsy neurons were promiscuous. At the population level, this would mean that neural sequences are getting jumbled and reactivations would therefore no longer have a meaningful connection to prior experience.
In a last set of impressive experiments, using in vivo sharp recording, the authors go after the synaptic mechanisms that support this difference in firing selectivity. In control animals during SWR, the excitatory drive coming from area CA3 was quickly counter balanced by a strong inhibitory drive—the balance between these two is what determines whether the CA1 neuron fires during a particular SWR. Interestingly, they found evidence that the circuit component that was scaled between different SWR families, and therefore determined the selectivity, was the inhibitory component. We can imagine the network dynamics during SWR as CA3 giving CA1 a big push. The local inhibitory dynamics would determine which CA1 neurons get activated. How the inhibitory circuits confer this selectivity is unknown, though there have been hints that individual inter-neurons may be dynamically tied to selective ensembles during learning (6). With this framework in mind, Valero et al. set out to manipulate the synaptic balance during SWR and test their ideas. In control animals, they reasoned that if they reduced inhibition in a single CA1 neuron (by intracellular dialysis of picrotoxin) it would lose SWR selectivity, which is exactly what they observed. Importantly, with this single-cell manipulation, they were able to exclude other possibilities. If, for example, the selectivity was conferred by certain CA3 neurons recruiting certain CA1 neurons, the picrotoxin would not have been expected to affect selectivity. Next, they tested whether they could recover selectivity in animals with epilepsy. The hope was that in epilepsy the inhibitory circuit may still be capable of conferring selectivity, but is overwhelmed by a strong excitatory push from CA3. They aimed to reduce the magnitude of the push from CA3 by using a use-dependent sodium channel blocker (carbamazepine) that shows sparing in inhibitory circuits (7). Excitingly, carbamazepine rescued both selective activation of CA1 neurons during SWR and behavioral performance.
As mentioned previously, studying neurologic diseases often affords researchers the unique opportunity to learn more about the disease and about how the brain works in general. This study by Valero et al. is an example of the power of such a study. This study gives a framework for therapeutic intervention that could ameliorate memory impairments in epilepsy patients and uncovers a remarkable mechanism for SWR selectivity involving inhibitory microcircuits.