Synaptic Specializations Support Frequency-Independent Purkinje Cell Output from the Cerebellar Cortex

Abstract

The output of the cerebellar cortex is conveyed to the deep cerebellar nuclei (DCN) by Purkinje cells (PCs). Here, we characterize the properties of the PC-DCN synapse in juvenile and adult mice and find that prolonged high-frequency stimulation leads to steady-state responses that become increasingly frequency independent within the physiological firing range of PCs in older animals, resulting in a linear relationship between charge transfer and activation frequency. We used a low-affinity antagonist to show that GABA_A-receptor saturation occurs at this synapse but does not underlie frequency-invariant transmission. We propose that PC-DCN synapses have two components of release: one prominent early in trains and another specialized to maintain transmission during prolonged activation. Short-term facilitation offsets partial vesicle depletion to produce frequency-independent transmission.

Keywords: Purkinje cells; TPMPA; cerebellum; deep cerebellar nucleus; presynaptic; receptor saturation; recovery from depression; short-term facilitation.

Figure 1

Charge transfer at PC-DCN synapses is linearly dependent on presynaptic firing frequency in juveniles and adults. (A) Schematic of sagittal cerebellar slice preparation and recording configuration. (B) Average IPSCs at P14 evoked by 100 stimuli at the indicated frequencies for the same cell in 1.5 mM external calcium (Ca_e). (C) (left) Summary of normalized steady-state IPSC amplitude as a function of stimulus frequency in P13-14 animals in 1.5 mM Ca_e (n = 6) and 2 mM Ca_e (n = 7). (right) Charge transfer due to incremental evoked IPSCs (product of IPSC_SS and stimulation frequency, normalized to 10 Hz) as a function of stimulus frequency. (D) As in B but for a P25 animal. (E) As in C but for P21-25 animals (n = 8). (F) Same as in B but for a P102 animal. (G) As in C but for P80-P110 animals (n = 7). (H) Amplitudes of IPSCs as a function of stimulus frequency for 100 Hz stimulation. (I) Ratio of steady-state responses to 10 Hz trains to those of 100 Hz trains as a function of age. Each point is one cell. All experiments were performed in 1.5 mM Ca_e other than those indicated for P13-14 animals and summarized in C and I. All data points are presented as means ± SEM unless otherwise noted. See also Figure S1.

Figure 2

Recovery from depression following a stimulus train is slow and independent of the train frequency in P21-30 animals. (A) IPSCs evoked by a 100 Hz stimulus train followed by IPSCs after a delay of 0.1 s to 20 s. Traces are averages of 6-8 trials. (B) Recovery of IPSC amplitude following 100 Hz stimulation is approximated with a single exponential fit constrained to go to 1 (R² = 0.994). The average steady-state IPSC amplitude is indicated by dashed line. Expanded axes at right shows recovery time points within the first second after the train (n = 8 cells). (C) As in (B) but for 20 Hz stimulation (n = 7 cells). R² = 0.992 for the exponential fit. All data are presented as means ± SEM unless otherwise noted.

Figure 3

PC-DCN synapses in P21-30 animals cannot be described by a single pool depletion model (A) Schematic representation of a single pool model with a readily releasable pool (RRP) of vesicles with uniform properties. During the first stimulus, the fraction of the RRP released is determined by the release probability (P). Between stimuli, replenishment of the RRP is dependent on extent of depletion and on the recovery time constant (τ_R, see methods). (B) Experimental data of IPSC amplitudes evoked by 100 Hz stimulation as a function of stimulus number were well approximated by an exponential decay. Data are presented as means ± SEM, error bars are occluded by markers (n = 8 cells). (C) IPSC amplitudes determined from a single pool model for 100 Hz trains using experimentally measured τ_R (7.5 s, Fig 2B) and the indicated values of P. (D) Decay constant λ in terms of number of stimuli according to a single pool depletion model for a range of P. Dashed lines indicated the range of experimentally observed responses. (E) Steady-state IPSC size produced by the model across all values of P. Dashed lines indicated the range of experimentally observed responses.

Figure 4

Short-term synaptic facilitation is present at the PC-DCN synapse in P21-30 animals. (A–C) Stimulus trains were used to evoke synaptic responses in 0.5 mM external calcium (Ca_e). (A) IPSCs evoked by 100 Hz stimulation. (B) Summary of average IPSC amplitudes evoked by 100 Hz trains (n = 6 cells). (C) Facilitation (IPSC_max) as a function of stimulation frequency (n = 6 cells). (D–F) In experiments conducted at 1.5 mM Ca_e, the DCN inputs were stimulated 100 times at 10 Hz, followed by 50 stimuli at another frequency. (D) IPSCs evoked by 10 Hz stimulation followed by 100 Hz stimulation. Expanded panel at right shows IPSCs just before and after the change in frequency. (E) Average IPSC size for protocol shown in (D) with the percent change in IPSC size (ΔIPSC/IPSC_SS) measured as indicated (n = 6 cells). (F) Change in IPSC size stepping from 10 Hz to various frequencies. A conditioning train of 10 Hz was applied to reach steady-state, followed by a step increase to 10-100 Hz (n = 6 cells). All data are presented as means ± SEM. See also Figure S2.

Figure 5

Relief of receptor saturation and desensitization does not alter linear charge transfer in P21-30 animals. (A) IPSCs before and after washin of the high affinity GABA_A receptor antagonist SR-95531 (300 nM) in a single cell. Graph at right shows average IPSC amplitudes as a function of stimulus number for a 100 Hz train in the absence (open symbols) and presence of SR-95531 (filled symbols, n = 6 cells). (B) Similar to (A), but using the low affinity GABA_A receptor antagonist TPMPA (2 mM, n = 5 cells) (C) Average IPSC size evoked by trains of the indicated frequencies in the presence of 2 mM TPMPA (n = 7 cells). (D) Summary of normalized steady-state IPSC amplitude as a function of stimulus frequency in control conditions (as in Fig. 1, n = 8 cells) and in the presence of 2 mM TPMPA (n = 7 cells). (E) Average charge transfer as a function of stimulus frequency, normalized to 10 Hz. All data are presented as means ± SEM, error bars are occluded in some cases by markers. See also Figure S3, S4.

Figure 6

A two-pool model and facilitation are sufficient to explain the frequency-independence of PC-DCN synapses in P21-30 animals. (A) Average IPSCs (top) in response to a 10 Hz train for experimental data in 2 mM TPMPA and two modeled vesicle pools (combined in purple). Readily releasable pool size (RRP, middle) and release probability (P, bottom) of each pool during the train. (B) Same as (A), but with 100 Hz trains. (C) Average steady-state IPSC size observed from data in 2 mM TPMPA (n = 7 cells) and predicted by models. (D) Plot of charge transfer as a function of stimulation frequency for experimental data and models. Charge was calculated as the product of steady-state IPSC size and stimulation frequency and normalized to charge at 10 Hz. All data are presented as means ± SEM, error bars are occluded in some cases by markers. See also Table S1, Figure S5, S6.

Figure 7

Input of a single PC to the DCN during conditioned eyeblink trains in P21-30 animals. (A) Average firing rate (top) and individual trials (bottom) of a PC in an awake mouse that underwent delay eyeblink conditioning from Ohmae & Medina, 2015. In these trials the conditioning light stimulus alone was applied without a periocular airpuff that was present at t = 0 in acquisition trials. (B) Top: IPSCs in a DCN neuron evoked by stimulating trial 14 followed a 50 Hz train of 25 stimuli to allow the synapse to reach steady state (not shown). Bottom: Average IPSC size for the stimulation pattern shown above (n = 4 cells). Responses are normalized to initial IPSC amplitude following a prolonged silent period. (C) Top: Average IPSC size for the 16 PC trials shown in (A, bottom), in 10 ms bins. Bottom: Average charge transfer (product of IPSC size (C, top) shown above and PC firing rate in (A, top)). Responses are normalized to the charge transfer for 10 Hz stimulation. (D) Average IPSC size from all trials for each interstimulus interval (ISI). (E) Total charge transfer as a function of instantaneous firing frequency (product of IPSC size and inverse of interstimulus intervals shown in (D)). All data are presented as means ± SEM, error bars occluded in some cases by markers. See also Table S1, Figure S7.