Incorporating single cell contribution into network models of ripple generation

Numerous lines of experimental evidence support the fundamental role of hippocampal sharp‐wave‐ripples (SPW‐Rs) in memory consolidation. These complexes are composed of a fast frequency oscillation (∼150 Hz) in the pyramidal layer (termed a ‘ripple’) and a large negative wave in the stratum radiatum (termed a ‘sharp‐wave’). It is accepted that the sharp‐wave is generated by excitatory inputs from the CA3 area but there is considerable debate about the mechanisms of ripple generation, demonstrated by the number of different models put forth to explain it. In general terms, all of these models are based on circuit properties, and account for the high spiking synchrony characteristic of ripples with connectivity or functional network features. On the other hand, there has been significant experimental demonstration of precise transmembrane conductance and spiking dynamics of single cells during SPW‐Rs. However, the difficulty of isolating individual synaptic contributions during in vivo experiments (such as small inhibitory potentials) and the inherent limitations of in vitro studies (drug effects, differences in the slice preparation and absence of long range connectivity) have prevented measurement of the contribution of single cells to SPW‐Rs. A precise quantification of this contribution has recently been provided by Bazelot et al. (2016), offering a framework for the integration of the subcellular, single cell, and network mechanisms into a comprehensive model of ripple generation.

One popular network model of ripple generation states that ripples are mediated by axo‐axonic electrical synapses between pyramidal neurons (Draguhn et al. 1998). Detailed computational models in combination with in vitro data support this theory, which describes how ectopic action potentials (APs) can be generated in axons of pyramidal cells and then propagate both orthodromically and antidromically to pace the ripple oscillation. A local network of pyramidal cells in the CA3 area could then synchronize by electrotonic coupling through gap junctions, generating oscillations at ripple frequency. However, recent experimental evidence obtained in vivo argues against the mechanism proposed by this model. First, a combination of intracellular and extracellular recording of APs in CA1 showed exclusive orthodromic propagation during ripples (English et al. 2014) and, second, optogenetic silencing of either CA1 pyramidal cells or interneurons abolishes ripple events, suggesting the necessity of both interneuron and pyramidal cell activation, and the interplay between them, for ripple generation and maintenance (Stark et al. 2014). Thus, although distinct mechanisms could be interacting in the generation of CA1 and CA3 ripples, it is difficult for the electrical coupling model to stand alone as the unique mechanism of ripple generation. A second class of model proposes that the recurrent properties of a local network act as a pacemaker of ripple oscillations. Reciprocal inhibition between perisomatic targeting interneurons could synchronize rhythmically the firing of pyramidal cell assemblies through inhibitory rebound spiking. Another possible mechanism of recurrent activity that would generate ripple frequency oscillations comes from the feedback inhibition of pyramidal cells by perisomatic interneurons. Also, as suggested by recent data from local optogenetic manipulation of CA1 activity (Stark et al. 2014), a combination of both mechanisms described by the second type of model could also explain ripple generation. However, in all these different models the contribution of single cell activity is not specified.

Bazelot et al. (2016) have shed light on this issue, describing how a single CA3 pyramidal neuron can influence the generation of ripples. The authors performed intracellular recordings of individual CA3 pyramidal cells and elicited single APs using precise intracellular current injection. They found that ripples followed induced action potentials after an interval of 2–6 ms. This time window between when an AP is delivered and the ripple initiation could be explained by the synaptic delays of local recurrent activity. Latencies for pyramidal cell–interneuron interactions within the CA3 region in in vitro preparations have been measured to be ≤ 3 ms, which would be enough to recruit other neurons via recurrent collaterals. In addition, the authors also found that the same pyramidal cell can trigger ripple events and also activate interneurons, which elicit inhibitory postsynaptic potentials (IPSPs) with similar probability. Furthermore, interneuron spiking is correlated in time and magnitude with IPSPs occurring in the stratum pyramidale during ripples. The coincidence of successive ripple cycles with depolarizing events onto fast‐spiking interneurons supports the fundamental role of excitation–inhibition loops as described during in vivo ripples (English et al. 2014; Stark et al. 2014).

Interestingly, the authors could differentiate distinct IPSP patterns. The distance within which different IPSP patterns could be recorded matched with the spatial extent of the arborization of perisomatic targeting interneurons (Gulyas et al. 2010), pointing to their causal relation. Further data with paired recordings including different subtypes of interneurons would help to distinguish the role attributed to distinct types of inhibitory cells during ripple generation and maintenance (English et al. 2014; Stark et al. 2014).

Together, the main findings of Bazelot et al. (2016) suggest a continuum, rather than two separated modes, between IPSPs and ripple generation. It has been shown that during ripples both inhibitory and excitatory activity compete to control spiking (English et al. 2014). Considering the results of Bazelot et al. within this framework, the excitation–inhibition balance would lead to either only IPSP generation or ripple generation, depending upon which of them predominate. Hence, only when the excitatory tone is high enough can a ripple could be generated.

Characteristics that facilitate the dominance of local excitation over inhibition are a high excitability of principal cells, intrinsic bursting properties and strong recurrent collateral excitatory connections. These features have been attributed mainly to the CA3 (Gulyas et al. 2010) and CA2 regions of the hippocampus. The peculiarities of these regions endow them with a strong potential for eliciting synchronous population activity. It has been also demonstrated that the firing of single CA3 pyramidal cells can induce a population burst. In addition, experimental evidence has shown that some CA3 pyramidal cells have a significantly higher number of postsynaptic interneuronal targets compared with CA1. This strong, divergent excitation of inhibitory cells would potentiate the development and spreading of the ripples and/or IPSP activity. In this context and within the time window described by Bazelot et al. (∼2–6 ms), it is feasible that a single pyramidal cell with enough inhibitory postsynaptic targets in the CA3 region could synchronize the discharge of a population of around 5–10 local interneurons. These interneurons could then sustain recurrent interactions that generate a ripple oscillation if the global excitation–inhibition balance favours the former.

Bazelot et al. (2016) also compared ripples initiated by a single CA3 stimulated pyramidal cell with spontaneous events. They found that the distances from which they could record these distinct patterns were significantly different. Induced ripples appeared first close to the stimulation electrode, suggesting that they involve mechanisms of a restricted, small network. In contrast, spontaneous events could be detected at longer distances, suggesting that these ripples may occur based on similar processes but simultaneously at different sites. Although possible mechanisms for the internal generation of ripples within the CA3 region have been discussed, this last result highlights the role played by network synchrony in ripple generation. The occurrence of simultaneous events at multiple sites brings up the possibility that an external pacemaker, which can trigger the generation of a ripple (possibly also mimicked by the discharge of a single CA3 pyramidal cell in the induced ripple case), can also coexist with the local initiation. This external pacemaker could be an input from distant CA3 ensembles, entorhinal cortex (either directly or via relay in the dentate gyrus) or the highly excitable CA2 region.

The report of Bazelot et al. (2016) establishes a framework through which single cell contribution can be introduced into network models of ripple generation. Their results refine the hypotheses of recurrent interactions as a necessary feature for ripple generation. Although their study focuses on the generation of CA3 ripples, the mechanism of CA1 ripple generation should be considered separately. Recent studies at systems level with intracellular, extracellular and precise optogenetic manipulations point to similar mechanisms regarding the network inhibition–excitation balance for the generation of CA1 ripples (English et al. 2014; Stark et al. 2014). But importantly, CA1 lacks the recurrent collaterals and burst spiking properties which characterize the CA3 region, implying a possible lower impact of single cells for CA1 ripple generation. Thus, more physiological correlates from the combination of in vivo and in vitro studies, together with electron microscopy data, are still needed to distinguish between the different proposals of recurrent interaction models. Possibly, various modes and different combinations of the proposed pyramidal cell–interneuron interactions can interact to generate a ripple.

Additional information

Competing interests

None declared.