Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2012 May 16;32(20):6878-93.
doi: 10.1523/JNEUROSCI.5736-11.2012.

NMDA Receptors With Incomplete Mg²⁺ Block Enable Low-Frequency Transmission Through the Cerebellar Cortex

Affiliations
Free PMC article

NMDA Receptors With Incomplete Mg²⁺ Block Enable Low-Frequency Transmission Through the Cerebellar Cortex

Eric J Schwartz et al. J Neurosci. .
Free PMC article

Abstract

The cerebellar cortex coordinates movements and maintains balance by modifying motor commands as a function of sensory-motor context, which is encoded by mossy fiber (MF) activity. MFs exhibit a wide range of activity, from brief precisely timed high-frequency bursts, which encode discrete variables such as whisker stimulation, to low-frequency sustained rate-coded modulation, which encodes continuous variables such as head velocity. While high-frequency MF inputs have been shown to activate granule cells (GCs) effectively, much less is known about sustained low-frequency signaling through the GC layer, which is impeded by a hyperpolarized resting potential and strong GABA(A)-mediated tonic inhibition of GCs. Here we have exploited the intrinsic MF network of unipolar brush cells to activate GCs with sustained low-frequency asynchronous MF inputs in rat cerebellar slices. We find that low-frequency MF input modulates the intrinsic firing of Purkinje cells, and that this signal transmission through the GC layer requires synaptic activation of Mg²⁺-block-resistant NMDA receptors (NMDARs) that are likely to contain the GluN2C subunit. Slow NMDAR conductances sum temporally to contribute approximately half the MF-GC synaptic charge at hyperpolarized potentials. Simulations of synaptic integration in GCs show that the NMDAR and slow spillover-activated AMPA receptor (AMPAR) components depolarize GCs to a similar extent. Moreover, their combined depolarizing effect enables the fast quantal AMPAR component to trigger action potentials at low MF input frequencies. Our results suggest that the weak Mg²⁺ block of GluN2C-containing NMDARs enables transmission of low-frequency MF signals through the input layer of the cerebellar cortex.

Figures

Figure 1.
Figure 1.
DHPG evokes asynchronous patterns of cerebellar MF-like activity that resembles in vivo activity. A, Simplified schematic of the vestibular cerebellum. Extrinsic MFs (black) convey vestibular afferent signals to GCs (blue) and UBCs (green) in the GC layer. UBC axons are called intrinsic MFs, which synapse onto the dendrites of GCs and other UBCs and therefore form a feedforward MF network. GCs integrate extrinsic and intrinsic MF activity and transmit this information to PCs via their axons, referred to as parallel fibers. As mGluR1 expression in the GC layer is restricted to UBCs, local pressure ejection of DHPG (purple) only excites UBCs. B1, A typical UBC voltage response to local pressure ejection of DHPG in the GC layer (arrow) is biphasic, consisting of a bursting phase of APs (boxed region) followed by a phase of sustained firing. B2, Enlargement of boxed region in B1 showing transition between AP bursting and sustained activity phase. B3, Enlargement of boxed region in B2 showing an individual burst of APs. Right, Plot of IEI versus AP number within the burst, and a line fit to the data. B4, Frequency of APs during sustained activity phase in UBCs. The average frequency (thick trace) and SEM (shaded region) for 1 s bins are displayed along with individual cell responses (thin traces). Circles denote time of peak frequencies. C1, GC synaptic currents recorded under voltage-clamp in response to local pressure ejection of DHPG, showing a biphasic pattern similar to that of the UBC APs in B1. C2, Enlargement of boxed region in C1 showing initial phase of EPSCs. C3, Enlargement of boxed region in C2 showing initial burst of EPSCs. Right, Plot of IEI versus EPSC number within the burst, and a line fit to the data. C4, Frequency of EPSCs during sustained activity phase in GCs. Average frequency, SEM and individual cell responses as for B4.
Figure 2.
Figure 2.
Asynchronous UBC MF activity modulates cerebellar cortex output in an NMDAR-dependent manner. A1, Extracellular PC recording showing an increase in AP rate from 30 to 43 Hz after pressure ejection of DHPG in the GC layer. Right, A single AP on expanded time scale. A2, DHPG increased the AP rate in 12 PCs (rate computed in 500 ms bins). A3, Normalized average rate (black line) and SEM (shaded region) of the traces in A2. Bottom, Average z-score of the same traces. A4, Z-score of AP rate before (black) and after application of 5 μm NBQX (blue) or 50 μm APV (red). Shaded regions denote SEM. B1, Example traces of EPSCs recorded in PCs in response to puff application of DHPG in the GC layer (arrow) in control (black) and APV (red). Below, enlargement of boxed region. B2, Duration of the synaptic responses in PCs for various drug conditions. Cont, Control; Mibe, 1 μm Mibefradil; SR, 1 μm SR-95531; Str, 500 nm Strychnine. B3, Charge integral of the synaptic responses in PCs for various drug conditions as in B2.
Figure 3.
Figure 3.
NMDARs downstream of UBCs enable GC layer signaling. A, DHPG-evoked synaptic currents recorded in GCs held at −82 or +18 mV in control conditions (black) and 50 μm APV (red). Integration of synaptic currents is illustrated by the shaded region. Traces at bottom show enlargement of boxed regions. B, Comparison of EPSC rate and duration of each burst in control and APV. C, EPSC rate (computed in 1 s bins) in control (black) and APV (red). Shaded regions denote SEM. Holding potential was −82 mV. DHPG was applied at time 0. D, Average integral of the GC synaptic current in APV (red) and APV + NBQX (blue) normalized to control conditions during the bursting and sustained activity phase at −82 or +18 mV.
Figure 4.
Figure 4.
GC NMDARs have a weak voltage dependence, and a pharmacological profile of GC layer transmission consistent with GluN2C-containing receptors. A1, DHPG-evoked GC NMDAR EPSCs isolated by bath application of 5 μm NBQX. The GC membrane potential was ramped from +28 to −92 mV with (top, black) and without (gray) evoking NMDAR currents by DHPG stimulation. Subtraction of the two current responses revealed the NMDAR current–voltage relationship (bottom, shaded region). A2, Resulting current–voltage relationship from shaded region in A1 (black) and a fit to Equation 1 (red): GNMDAR = 401.7 pS, ENMDAR = 2.4 mV, C1 = 3.70 mm, C2 = 0.033 mm, δbind = 0.34, δperm = 0.47 (T = 33°C; [Mg2+]out = 1.5 mm), where C1 = koff/kon and C2 = kperm/kon. B, Fraction of NMDARs not blocked m(V) from the fit in A2 (red curve). Blue curve shows fit of Equation 1 to GC NMDAR currents evoked by direct MF stimulation: GNMDAR = 367.9 pS, C1 = 2.07 mm, C2 = 0.015 mm, δbind = 0.35, δperm = 0.53 (T = 35°C; [Mg2+]out = 1.0 mm; ENMDAR constrained to 0 mV; data from Rothman et al., 2009). Gray dashed curve shows a simultaneous fit of Equation 1 to the P7–P9 wild-type GC data of Takahashi et al. (1996) using both 0.1 and 1.0 mm [Mg2+]out data: C1 = 1.97 mm, C2 = 0.0035 mm, δbind = 0.43, δperm = 0.48, average GNMDAR = 952.6 pS, ENMDAR = 10.28 mV (T = 24.5°C). For comparison, m(−80 mV) = 7.8, 7.0, and 1.4% for the 3 respective curves. C, Normalized charge integral of the PC synaptic response after application of SR and Str, zinc (Zn), Ro25-6981 (Ro-25), and APV. Tricine (10 μm) was included in all conditions to buffer free zinc.
Figure 5.
Figure 5.
Amplitude and kinetics of AMPAR and NMDAR components of UBC MF-GC EPSCs. A1, GC EPSC recorded at −82 mV during DHPG stimulation obtained by averaging synaptic events isolated by >100 ms (black trace). The EPSC integrals in control and APV are shown in gray. Subtraction of the integral in APV from the integral in control conditions gives the kinetics of charge due to NMDARs, which is fit with a single exponential (red trace). A2, Amplitude (left) and 20–80% rise time constant (right) of EPSCs isolated by >100 ms. Filled circles represent the average value. A3, Percentage of the total charge carried by NMDARs (left). Decay time constant of the NMDAR component (right). B, Top, EPSC burst recorded at −82 mV. Bottom, NMDAR bursts isolated with NBQX at +18 and −82 mV and a single exponential fit (red trace). Right, Decay time constant of NMDAR component determined from isolated bursts.
Figure 6.
Figure 6.
Short-term plasticity at UBC MF-GC synapses. A1, Example of a DHPG-evoked EPSC burst recorded in a GC held at −82 mV, with denoted EPSC number. A2, Amplitude of EPSCs during the burst for individual GCs shows short-term depression and a wide range of variability. A3, Short-term depression of EPSCs during the burst plotted as normalized average amplitude (black) and for individual cells (colors). B, Linear fits to the relationship between EPSC amplitude and frequency (1 s bins) for the sustained component of DHPG-evoked activity recorded in individual cells (colors).
Figure 7.
Figure 7.
Input-output relationship of GCs during low-frequency UBC MF input. A1, GC current-clamp recording of EPSPs and APs evoked by DHPG stimulation of UBC MF input (arrow). A2, Enlargement of boxed region in A1. Inset is enlargement of boxed region, showing the AP at the peak of a burst. Calibration: boxed region, 10 ms. B, Input-output relationship (ratio of APs per EPSC) is plotted versus membrane potential (Vm) with linear fit (red). Gray points indicate APs evoked during the bursting phase. C1, GC AP rate versus EPSP rate (500 ms bins). Red line is a linear fit. C2, Average GC model input-output curves for single MF inputs with physiological levels of NMDAR conductance (φ = 1.0, 1.2, 1.4; black, orange, and green respectively; 242 GC models). The curves match the GC data in C1.
Figure 8.
Figure 8.
Conductance-based integrate-and-fire models mimic heterogeneity in GC membrane properties and MF synaptic input. A1, A simple RmCm model fit (red; Eq. 3) to the subthreshold voltage trajectory of a GC (black) during injection of a synaptic conductance train (purple) injected via dynamic clamp. The injected conductance train consisted of the sum of 4 different MF inputs, for a sum total rate (Total MF) of 24 Hz (left) and 170 Hz (right). Similar fits were computed for 37 other GCs (data not shown). Data from Rothman et al. (2009). A2, Relationship between membrane conductance (Gm; estimated from RmCm model fits as in A1, red circles; black circle denotes GC in A1) and membrane capacitance (Cm measured in voltage clamp) for 38 GCs, together with a linear fit (red line; Gm = Cm · 318.6 pS/pF; R2 = 0.89, p < 0.001). Gray symbols show passive properties for 204 additional GCs, estimated using their experimentally measured Cm and the linear relation and the residual distribution (differences between the line fit and data points, which were well described by a normal distribution between −0.5 and 0.5 nS; data not shown) to generate appropriate Gm values. Histograms show distributions of Cm and Gm for the full GC population in bins of 0.4 pF and 0.1 nS, with average values (red dashed lines; 3.22 ± 1.02 pF and 1.06 ± 0.37 nS). B, Membrane reversal potential (Em) measured from 242 GCs in the presence of inhibitory blockers. Dashed line denotes average (−79.9 ± 5.3 mV). C, Time course of model AMPAR conductance (AMPAR all, blue; Eq. 5; peak normalized to one) and underlying spillover (dashed orange) and direct (solid orange) components with a peak ratio of 0.34. The 10–90% rise times of the direct, spillover and summed components were 0.17, 0.57, and 0.17 ms, respectively, as in the work of DiGregorio et al. (2002). The decay of “AMPAR all” was fit by a double-exponential function with fast and slow components (A1 = 74%, τ1 = 1.15, A2 = 24%, τ2 = 9.71) that closely matched those of the isolated AMPAR currents in this study. The model NMDAR conductance waveform (green, NMDAR; Eq. 5; peak normalized to one) had a 10–90% rise time of 1.4 ms, which matches that of NMDAR EPSCs recorded at MF-GC synapses (n = 4; data from Rothman et al., 2009), and a single-exponential decay time constant of 24.7 ms, which matches that of the isolated NMDAR currents in this study (Fig. 5B). D, Distribution of mean peak AMPAR conductances (GAMPAR) for single MF synapses. Dashed line denotes average (0.63 ± 0.48 nS, n = 79). Data from Sargent et al. (2005), their Figure 1C. E1, Fit of the plasticity model (blue; Eq. 6) to a 30 Hz MF-GC AMPAR conductance train (black). The fitting routine minimized the sum of squared residuals of the two conductance trains, not the peak amplitudes. Dashed lines denote time-averaged conductance values for the data (black) and fit (blue) measured at t >100 ms from the onset of synaptic input. Data from Rothman et al. (2009). E2, Time-averaged AMPAR conductance (dashed lines in E1) as a function of the input rate of a single MF, for real GCs (black circles; n = 4; SEM; Rothman et al., 2009) and the synaptic model (blue circles). Conductance trains were normalized so the amplitude of the first synaptic input equaled one. For comparison, model gAMPAR with no plasticity is shown as the dashed line. F1 and F2, Same as E1 and E2, but for the unblocked NMDAR conductance (green). Data in F1 were low-passed filtered at 1 kHz for presentation.
Figure 9.
Figure 9.
Incomplete Mg2+ block of GluN2C-containing NMDARs enhances AP firing in GC models during low-frequency MF input. A1, Voltage traces from a GC model with mid-range membrane properties (Cm = 3.0 pF, Gm = 1.1 nS, Em = −80 mV) during excitation with 4 asynchronous MF inputs with total input rate of 90 Hz (top) and 240 Hz (bottom) under different input conditions: AMPAR + NMDAR (black; φ = 1.0), AMPAR all (blue; φ = 0), NMDAR only (green), AMPAR spillover only (orange), AMPAR spillover + NMDAR (red). Resting potential was −77.9 mV (solid gray line). AP threshold was −40 mV (dashed gray line); APs truncated to −15 mV. A2, Steady-state input-output relationships (GC firing rate vs total MF input rate) for 242 GC models with different Cm, Gm, Em, and synaptic properties (Fig. 8). To construct the input-output relationships, each GC was driven by 4 statistically independent random MF synaptic conductance trains, with the same mean frequency, and this frequency was then varied between 7.5 and 60 Hz. The blue traces show GC input-output relations for AMPAR only MF inputs, while MF inputs with both AMPAR and NMDAR components are shown in black. The voltage dependence of the NMDAR conductance was set to that for mature GCs (B2 inset, solid line) that express GluN2C. φ denotes the ratio of GNMDAR to GAMPAR, such that for AMPAR only, φ = 0 and for AMPAR plus NMDAR components, φ = 1.0. These ratios were the same across all 4 MF inputs. A3, Average input-output relations of the 242 GC models in A2 (blue and black) plus simulations for other φ in the range 0–1.4. Error bars denote SEM. Top graph shows the percentage of the total synaptic current injected into the GC models mediated by NMDARs. B1, B2, and B3, The same as A1, A2, and A3, except the NMDAR voltage-dependent block was that of immature GCs (B2 inset, dashed line) that express GluN2A/B. B2 inset, Fraction of NMDARs not blocked as a function of voltage (m(V); Eq. 2). Solid line is from a fit to the GC data in Figure 4A2. Dashed line is from a fit to the P7–P9 wild-type GC data of Takahashi et al., 1996, shown in Figure 4B.

Similar articles

See all similar articles

Cited by 20 articles

See all "Cited by" articles

Publication types

MeSH terms

Feedback