Reverse engineering inhibitory circuits in hippocampal CA3

To understand the computations performed by neural circuits, it is essential to reveal the details of microcircuit organization at the level of identified neuronal subtypes. This includes determining not only which subtypes are preferentially connected, but also understanding the dynamics of their synaptic properties. In particular, synapses exhibit short‐term plasticity that can evolve over tens of milliseconds, as postsynaptic responses depress or facilitate in response to on‐going presynaptic action potentials. Importantly, the dynamics of such plasticity depends greatly on the identity of the presynaptic neuron. However, it is only recently that tools, such as transgenic mice, have become available to consistently target specific neuronal types for investigation.

In the cerebral cortex and hippocampus, great strides are being made towards identifying and studying distinct types of GABAergic inhibitory interneurons (Kepecs & Fishell, 2014). Although they represent only about 20% of the total cortical neuron population, interneurons are crucial regulators of neural network activity. In particular, they can restrict the time window of effective excitation, adjust circuit gain, and play a key role in both generating and coordinating network oscillations (Klausberger & Somogyi, 2008). Interneurons are incredibly diverse and are often distinguished by distinct combinations of electrophysiological characteristics, morphology, and molecular expression. Importantly, different subtypes target distinct membrane compartments of excitatory glutamatergic principal cells (PCs) to regulate their synaptic integration and output. One such group, basket cells, target the perisomatic region (soma and proximal dendrites) and play a crucial role in multiple network rhythms, such as gamma oscillations and sharp wave ripples. There are two major types of basket cells, generally distinguished based on the differential expression parvalbumin (PV) or cholecystokinin (CCK) (reviewed in Armstrong & Soltesz, 2012). Although they innervate similar domains of PCs, PV+ and CCK+ basket cells play very different roles in circuit function. They possess distinct electrophysiological profiles, exhibit different modes of synaptic transmission, and are sensitive to different neuromodulators. Importantly, the expression of PV or CCK alone is not sufficient to identify an interneuron as a basket cell. PV+ cells also include axon initial segment targeting axo‐axonic cells (AACs) and dendrite targeting bistratified cells; CCK+ cells include many that are dendrite targeting (Tricoire et al. 2011). Thus, the study of interneuron circuitry requires great care and multiple criteria must be applied when specifying subtypes.

In this issue of The Journal of Physiology, Kohus and colleagues (Kohus et al. 2016) admirably provide a comprehensive data set describing the connectivity and synaptic dynamics of PV+ and CCK+ cells in hippocampal CA3 using paired whole‐cell patch recordings in two transgenic mouse lines that differentially label these interneurons. Importantly, they carefully segregated PV+ cells into baskets (PV+BC) or axo‐axonic cells (AACs) based on post hoc analysis of axon targeting and immunostaining for ankyrin G, which is expressed in the axon initial segment. Furthermore, CCK+ baskets (CCK+BCs) were distinguished from CCK+ dendrite targeting interneurons (CCK+DTIs). Connectivity was investigated between all four of these interneuron subtypes as well as their outputs to PCs. Kohus et al. (2016) found that interneurons formed recurrent circuits within the same subtype (e.g. PV+BC to PV+BC, CCK+BC to CCK+BC) but rarely or never across subtypes (e.g. CCK+BC to CCK+DTI). Interestingly, although occasional connections were found from PV+BCs to AACs, no connections were found from AACs to PV+BCs or between AACs.

In order to understand how these interneurons participate during different network states and oscillations, Kohus et al. (2016) characterized the short‐term synaptic plasticity of each synaptic connection across a wide range of presynaptic action potential frequencies, as well as during firing patterns that mimic sharp wave ripples or interictal events. These data should prove invaluable for future modelling studies, and indeed the authors show that simple mathematical models can capture some aspects of their synaptic data quite well. Interestingly, although PV+ cells demonstrated synaptic depression across firing rates and connections, CCK+ interneurons generated heterogeneous synaptic responses (facilitation, depression, asynchronous transmitter release) that depended on the postsynaptic target as well as presynaptic firing rate. Furthermore, CCK+DTIs failed to evoke any transmitter release onto PCs unless they were first stimulated with pathological interictal events. It is possible that transmission from CCK+DTIs to PCs is enhanced under specific modulatory states, but this awaits future study.

Finally, Kohus et al. (2016) show that their short‐term plasticity data provide a possible mechanistic account for the behaviour of sharp wave ripples recorded in vitro. Sharp waves (fast oscillations of ∼160 Hz) are observed in the hippocampus during slow wave sleep and periods of immobility in rodents, and are thought to be important in memory formation. In a previous study (Schlingloff et al. 2014), the authors found that sharp waves require the coordinated activity of recurrently connected CA3 PCs and PV+BCs. Importantly, optogenetically stimulating PV+BCs alone could trigger sharp waves, but subsequent sharp waves could only be evoked after a recovery period of ∼300 ms. In the present study Kohus et al. (2016) demonstrate that the amplitude of synaptic currents from PV+BCs to PCs follows a similar recover time during presynaptic trains that mimic sharp wave ripples. Thus, the rate of spontaneous sharp waves is very likely paced by the short‐term synaptic dynamics of PV+BCs in CA3. This example very nicely demonstrates the importance of the data set collected by Kohus et al. and its likely utility in explaining complex network phenomena.