The Timing of Excitatory and Inhibitory Synapses Rules the Cerebellar Computation

The timing of synaptic inputs is critical for various neurophysiological processes, including synaptic integration and plasticity. The importance of temporal correlations is exemplified in the cerebellum, which receives afferents conveying sensory information that is used for rapid motor control (Person, 2019). The largest number of synaptic inputs is provided by mossy fibers that convey vestibular, proprioceptive, and sensorimotor information. Mossy fibers provide indirect excitatory input (feedforward excitation) to Purkinje cells via a synaptic relay on glutamatergic granule cells, which innervate Purkinje cells via axons called parallel fibers. Importantly, mossy fibers also produce feedforward inhibition via a second synaptic relay produced by granule cells on molecular layer interneurons, which, in turn, are connected by GABAergic synapses to Purkinje cells (Hull and Regehr, 2022). Thus, mossy fibers excite granule cells, which in turn generate synaptic excitation and feedforward inhibition of Purkinje cells. The Purkinje cells constitute the sole synaptic output of the cerebellar cortex.

Anatomically, the synapses of mossy fibers on granule cells are organized in rosette shapes. Each mossy fiber can innervate granule cells across multiple rosettes, thereby enabling the activation of feedforward excitation, inhibition, or both (Quy et al., 2011; Valera et al., 2016). Notably, temporal variability of Purkinje cells spikes has been reported to depend on the activation mode of these feedforward circuits. Direct electrical stimulation of granule cells results in minimal temporal variability in Purkinje cell spikes (Brunel et al., 2004). In contrast, the activation of mossy fibers, a condition closer to physiological activation, produces higher temporal variability in Purkinje cell spikes (Brown and Raman, 2018). This contrast underscores the need to explore how mossy fibers recruit the different feedforward circuit motifs that result in high variability in patterns of Purkinje cell spiking.

A recent article by Binda et al. (2023) explored how feedforward excitatory and inhibitory circuits, activated by mossy fibers, generate temporal variability of synaptic input to Purkinje cells and how the timing of synaptic inputs affects synaptic integration in Purkinje cells. To activate subsets of mossy fibers, the authors expressed Channelrhodopsin-2 in the dorsal column nuclei, where mossy fibers originate, and then optically stimulated mossy fiber terminals in either single rosettes, where mossy fibers synapse with granule cells, or at the surface, which activated multiple rosettes. They also activated granule cells directly, via electrical stimulation. During these different stimulation protocols, evoked excitatory postsynaptic currents (EPSCs) and inhibitory postsynaptic currents (IPSCs) were recorded in voltage clamp from Purkinje cells in cerebellar brain slices. From these recordings, the authors measured the onset time and amplitude of EPSCs and IPSCs. The onset latency between an IPSC and an EPSC was used to estimate the EPSC–IPSC delay, which varied depending on the type of stimulation used. Classical electrical stimulation of granule cells produced very low variability in EPSC–IPSC delay. In contrast, optogenetic stimulation of single rosettes produced EPSC–IPSC delays with slightly higher variability than that evoked by electrical stimulation. The variability was even higher with surface stimulation. The authors posited that increased temporal variability reflects activation of multiple feedforward circuits. They hypothesized that rosette activation initiates an antidromic action potential through the mossy fiber, continuing until axonal branches. This allows orthodromic propagation to reach different rosettes. Thus, the activation of feedforward excitation and inhibition originating from different groups of granule cells is a potential explanation for the variability in EPSC–IPSC delay (Binda et al., 2023; Fig. 2E).

Next, the authors categorized EPSC–IPSC delays into three groups: Group 1 corresponded to delay characteristic of classical feedforward inhibition and feedforward excitation—they had short latencies with narrow variability, and EPSCs always preceded IPSCs. Group 2 included delays in which IPSCs preceded EPSCs, as well as delays in which EPSCs preceded IPSCs, but were longer or shorter than delays in Group 1. Group 3 encompassed cases where only IPSCs occurred. The probability of activation for these groups varied depending on whether single rosette stimulation or surface stimulation was applied. Single rosette stimulation had the highest probability of activating group 1, while surface stimulation had a similar probability of activating each group, with a slight preference for activating group 2. This supports the idea that stimulation of multiple rosette results in higher variability in synaptic delays.

What is the significance of variability in EPSC–IPSC delays? The amplitude of EPSCs and IPSCs is important for determining the output of the postsynaptic neuron (i.e., the probability of action potentials). The authors observed that the amplitude of EPSCs and IPSCs increased proportionally with stimulus intensity, either from single or rosette stimulation, maintaining a stable ratio of excitatory to inhibitory input. Further analysis of the EPSC–IPSC delays and total synaptic charge revealed that long delays promote excitation (Binda et al., 2023; Fig. 2H). Thus, with a stable ratio between excitatory and inhibitory synapses, the timing between EPSCs and IPSCs determines the output of Purkinje cells.

Binda et al. (2023) further tested this idea using a computational simulation, which enables the study and testing of a mathematical model across a wide range of conditions. They used a previously published mathematical model (Grangeray-Vilmint et al., 2018) that estimates Purkinje cell spike rates based on synaptic integration of inputs from granule cells and molecular layer interneurons, as well as presynaptic short-term dynamics. Binda et al. (2023) updated the model by incorporating the delays and variability they had observed when excitatory and inhibitory synapses was recruited by the feedforward circuit motifs resulting from mossy fiber rosette stimulation. The model revealed a positive correlation between the spike rate of Purkinje cells and the delays between excitatory and inhibitory synapses in the range of delays from −5 ms (inhibition preceding excitation) to +2.5 ms (excitation preceded the inhibition). This correlation was prominent when the stimulation of feedforward circuits was set within a frequency range of 30–75 Hz, corresponding to the physiological baseline range of Purkinje cell spike frequency. The correlation decreased significantly when stimulation frequency exceeded 100 Hz, a spike frequency that is easily reached by Purkinje cells under different physiological conditions, such as during locomotor activity (Jelitai et al., 2016). These results suggest that synaptic delays could control the spike rate of Purkinje cells under certain physiological conditions, such as during periods of rest.

In summary, Binda et al. (2023) delved into how the mossy fiber pathway activates feedforward circuits, which generate different timings for the arrival of excitatory and inhibitory synaptic input to Purkinje cells. Their results suggest that mossy fiber activation may recruit several groups of granule cells, involving multiple feedforward circuits. This recruitment may underlie the greater temporal variability in the occurrence of Purkinje cell spikes occurring after mossy fiber stimulation compared with electrical stimulation of granule cells. The authors demonstrate that delays between excitatory and inhibitory synapses play a pivotal role in the synaptic integration of Purkinje cells. The timing of synaptic delays determines whether mossy fiber activation induces inhibition or excitation in Purkinje cells, thereby either facilitating or inhibiting the occurrence of spikes.

The consequences of the temporal delay in the excitatory and inhibitory synapses on Purkinje cells could interact with other physiological processes, such as spike time-dependent plasticity (STDP). The temporal coupling of postsynaptic potential with the spikes of Purkinje cells results in long-term plasticity when this coupling occurs primarily between −40 and +40 ms (Bell et al., 1997). Interestingly, Binda et al. (2023) show that the spike rate of Purkinje cells is particularly sensitive to delays between excitatory and inhibitory synapses when the stimulus intervals are between ∼15 and 35 ms, falling within the coupling timing of STDP. Therefore, we can hypothesize that the capability of Purkinje cells to activate STDP may be either allowed or regulated by the feedforward circuits activated by mossy fibers, promoting or inhibiting the occurrence of spikes.

The main conclusion of this study, based on the simulated spike output ratio of Purkinje cells, relied on a computational model. However, it is important to note that all computational models are built on assumptions and generalizations of certain biological processes. This specific model involves physiologically validated parameters for the short-term dynamics of excitatory and inhibitory synapses in granule cells and molecular layer interneurons, respectively (Grangeray-Vilmint et al., 2018). However, other local circuits, including Golgi cells, unipolar brush cells, and candelabrum cells, may also play a role in the activation of mossy fibers, generating additional sources of variability in the latency of excitatory and inhibitory synaptic input to Purkinje cells (Hull and Regehr, 2022). Therefore, it would be valuable to complement the results obtained from the computation simulations with a biological approach. Potential experiments could involve directly recording the action potentials of Purkinje cells under various synaptic delays evoked by single rosette or surface stimulation. The origin of synaptic inputs that generate the postsynaptic potentials in Purkinje cells may be diverse and could involve granule cells or climbing fibers and potentially contribute to synaptic plasticity that underlies learned behaviors. Therefore, the delay between excitatory and inhibitory synapses addressed in Binda et al. (2023) becomes more significant when considered in the context of the physiological spiking and plasticity processes operating in Purkinje cells in the cerebellar cortex.