Target-specific regulation of presynaptic release properties at auditory nerve terminals in the avian cochlear nucleus

Abstract

Short-term synaptic plasticity (STP) acts as a time- and firing rate-dependent filter that mediates the transmission of information across synapses. In the auditory brain stem, the divergent pathways that encode acoustic timing and intensity information express differential STP. To investigate what factors determine the plasticity expressed at different terminals, we tested whether presynaptic release probability differed in the auditory nerve projections to the two divisions of the avian cochlear nucleus, nucleus angularis (NA) and nucleus magnocellularis (NM). Estimates of release probability were made with an open-channel blocker of N-methyl-d-aspartate (NMDA) receptors. Activity-dependent blockade of NMDA receptor-mediated excitatory postsynaptic currents (EPSCs) with application of 20 μM (+)-MK801 maleate was more rapid in NM than in NA, indicating that release probability was significantly higher at terminals in NM. Paired-pulse ratio (PPR) was tightly correlated with the blockade rate at terminals in NA, suggesting that PPR was a reasonable proxy for relative release probability at these synapses. To test whether release probability was similar across convergent inputs onto NA neurons, PPRs of different nerve inputs onto the same postsynaptic NA target neuron were measured. The PPRs, as well as the plasticity during short trains, were tightly correlated across multiple inputs, further suggesting that release probability is coordinated at auditory nerve terminals in a target-specific manner. This highly specific regulation of STP in the auditory brain stem provides evidence that the synaptic dynamics are tuned to differentially transmit the auditory information in nerve activity into parallel ascending pathways.

short-term synaptic plasticity (STP) acts as a time- and firing rate-dependent filter that mediates the transmission of information across synapses and network dynamics (Abbott and Regehr 2004; Fortune and Rose 2001; Klug et al. 2012). In the avian auditory brain stem, two divisions of the cochlear nucleus, nucleus magnocellularis (NM) and nucleus angularis (NA), receive auditory nerve inputs via axons that bifurcate to send a branch to each (Carr and Boudreau 1991; Köppl 2001; Rubel and Parks 1988). The terminals of those axonal projections show numerous morphological and physiological differences depending on their target, including different STP (Carr and Soares 2002; MacLeod 2011; Trussell 1999). In NM the plasticity is dominated by short-term depression (Brenowitz and Trussell 2001a; Zhang and Trussell 1994), while in NA the plasticity is a balanced mixture of facilitation and depression (MacLeod et al. 2007). In the mammalian cochlear nucleus, a similar distinction in STP profiles can be observed at nerve terminals onto the bushy cells and those onto stellate cells, which form analogous pathways to NM and NA, respectively (Cao et al. 2008; Cao and Oertel 2010; Oleskevich and Walmsley 2002; Wang and Manis 2008; Yang and Xu-Friedman 2008). Differential synaptic dynamics therefore could contribute to the divergence of acoustic timing and intensity information into parallel pathways in the auditory brain stems of birds and mammals.

What determines the characteristics of STP at a given synaptic connection? In the avian auditory brain stem, a series of comparative studies have investigated the mechanisms that underlie STP, revealing a number of similarities and a few differences. Recovery from depression had similar biexponential times courses in both avian cochlear nuclei (Brenowitz and Trussell 2001a; MacLeod et al. 2007; MacLeod and Horiuchi 2011), suggesting common activity-dependent mechanisms that accelerate recovery rates or the presence of multiple vesicle pools with heterogeneous recovery kinetics. The recovery trajectories in the avian cochlear nuclei thus have similarities to recovery trajectories found in mammalian auditory brain stem and several other areas (Dittman et al. 2000; Hallermann et al. 2010; Wang and Kaczmarek 1998; Wang and Manis 2008; Wu and Borst 1999; Yang and Xu-Friedman 2008). Rapidly desensitizing glutamate receptors are expressed in both NA and NM (Parks 2000; Raman et al. 1994; Sugden et al. 2002). Postsynaptic desensitization of these receptors contributed to short-term depression during high-frequency trains at NM synapses (Brenowitz and Trussell 2001b) but not at NA synapses (MacLeod and Horiuchi 2011). In contrast, facilitation was more common at NA synapses under normal conditions (MacLeod et al. 2007); however, weak facilitation could be elicited at NM synapses under low-release-probability conditions (Brenowitz and Trussell 2001b). With the exception of desensitization, these studies demonstrate multiple similarities in fundamental synaptic mechanisms in the two pathways.

One key factor in determining the synaptic dynamics is initial release probability. In this study, we asked whether the presynaptic release probability differed in the two pathways, explaining the differences in the STP. Average release probability at the nerve-to-NM synapses has been estimated to be as high as ∼0.5–0.6 in some studies (Brenowitz and Trussell 2001a; Taruno et al. 2012) but also as low as 0.28 in others (Oline and Burger 2014). A high release probability would ensure a strong connection that accurately encodes timing but also results in synaptic depression. The greater facilitation and lesser depression at nerve synapses in NA suggest that release probability may be lower in this nucleus, but it has not been measured previously. To compare release probability in parallel in NA and NM, we measured the blockade of synaptic responses upon application of an open-channel blocker of N-methyl-d-aspartate (NMDA) receptors (NMDARs), (+)-MK 801 maleate (MK801). This method progressively blocks synaptic responses in an activity-dependent manner in proportion to the release probability and has been used in comparing cortical and hippocampal release probability in different pathways (Castro-Alamancos and Connors 1997; Hessler et al. 1993; Huang and Stevens 1997; Rosenmund et al. 1993) as well as recently in mammalian cochlear nucleus (Yang and Xu-Friedman 2012). We further examined how STP, and by proxy release probability, differed at converging and diverging auditory nerve terminals within the cochlear nucleus.

METHODS

Brain Slice Preparation

All animal procedures were performed in accordance with Federal guidelines on animal welfare and approved by the University of Maryland Institutional Animal Care and Use Committee. Chicken embryos incubated 17–18 days were cold anesthetized and rapidly decapitated, and the brain stem was submerged in chilled and oxygenated low-sodium artificial cerebrospinal fluid (ACSF) (in mM: 97.5 NaCl, 3 KCl, 2.5 MgCl₂, 26 NaHCO₃, 1 CaCl₂, 1.25 NaH₂PO₄, 10–12 dextrose, 3 HEPES, and 57 sucrose; 305–315 mosM, pH 7.4). Transverse slices (250 μm) were made on a vibrating tissue slicer (Leica Microsystems, Wetzler, Germany). Slices were incubated in oxygenated normal ACSF (in mM: 130 NaCl, 3 KCl, 2 MgCl₂, 26 NaHCO₃, 2 CaCl₂, 1.25 NaH₂PO₄, 10 dextrose, and 3 HEPES) for 30 min at 34°C and then held in solution at room temperature (22–24°C) until being used.

Patch-Clamp Electrophysiology

Whole cell patch-clamp recordings were made from NA and NM cells with infrared-differential interference contrast (IR-DIC) video microscopy. For recordings, slices were continuously perfused with warmed, oxygenated normal ACSF (30–33°C, 1–2 ml/min). For NMDAR excitatory postsynaptic current (EPSC) experiments, glycine was added to the ACSF (5 or 30 μM).

Initial micropipette resistances were 3–7 MΩ with a standard intracellular voltage-clamp solution [in mM: 70 cesium sulfate, 5 QX-314 (lidocaine), 1 MgCl₂, 1 Na₂ATP, 0.3 Na₂GTP, 10 phosphocreatine, 4 NaCl, 10 HEPES, 5 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA), and 32 sucrose, with 0.1% biocytin]. In most MK801 experiments, BAPTA concentration was increased to 10 mM and Na₂ATP increased to 4 mM in the intracellular solution to combat NMDAR rundown. For all experiments, excitatory synaptic potentials were isolated by inclusion of 20 μM SR95531 (gabazine) and 3 μM strychnine in the ACSF. The pharmacological manipulations used 6,7-dinitroquinoxaline-2,3-dione (DNQX, 10–20 μM; Sigma) to block AMPA receptor (AMPAR)-mediated currents and the open-channel antagonist MK801 (20 μM; Tocris Cookson) to block NMDAR-mediated currents in an activity-dependent manner. Electrophysiological recordings were made with a Multiclamp 700B patch-clamp amplifier (Molecular Devices, Sunnyvale, CA) in voltage-clamp mode with the series resistance corrected by 60–90% (corrected residual series resistance: 4–8 MΩ). Junction potential (measured approximately −9 mV) was left uncorrected. Stimulation and recordings were controlled by a PC running custom software in IGOR Pro (WaveMetrics, Lake Oswego, OR).

Auditory Nerve Fiber Stimulation

Synaptic currents were evoked with a low-impedance tungsten monopolar electrode placed in the 8th (auditory) nerve fiber tracts as they enter NA at the medial margin (MacLeod and Carr 2005). For NM recordings, the stimulation electrode was placed at the dorsal or lateral margin of the nucleus. Fiber paths were often visible under IR-DIC optical conditions. Biphasic stimulus waveforms passed through an analog, constant-current stimulus isolation unit (model A395, World Precision Instruments, Sarasota, FL) that allowed computer control of stimulus amplitude. Stimulus artifacts were digitally removed for clarity in figure presentations. Paired-pulse stimuli and trains of five pulses were evoked with a constant stimulus frequency of 50 Hz and 20-s interstimulus intervals. Single stimuli for pharmacological manipulation were evoked every 20 s for 50–150 continuous trials.

Glutamate Puff Stimulation

To measure MK801 blockade of the postsynaptic receptors with direct glutamate activation, synaptic responses were evoked by puffs delivered by Picospritzer of 1 mM glutamate in ACSF onto the postsynaptic neuron. The amplitude and duration of the puff (10–40 psi, 10–40 ms with a 1-μm tip patch pipette, 50–75 μm from the recorded cell) were calibrated to generate a similar response magnitude as the evoked synaptic responses. The stimulus protocol and analysis were identical to presynaptic stimulation experiments: puffs were delivered in control ACSF with an interval of 20 s; then DNQX was bath applied, followed by bath application of MK801.

Analysis

In the absence of DNQX, evoked EPSCs at depolarized potentials were composed of a fast peak and a slow, late component. In experiments in which MK801 was applied, but not DNQX, current values were measured at the peak and at 10 ms after the stimulus (Fig. 1). Some temporal overlap in the AMPAR and NMDAR currents likely explains why MK801 slightly decreased the peak current and incompletely blocked the late current in these experiments. In separate experiments, the MK801 blockade time course was measured using NMDAR EPSCs pharmacologically isolated with DNQX and elicited at positive voltages (+40 or +50 mV). These were measured as an area under the EPSC measured from 5 ms to 55 ms after the stimulus.

Fig. 1.

Activity dependent blockade of the slow component of the excitatory postsynaptic current (EPSC) by MK801 at auditory nerve synapses in the avian cochlear nucleus angularis (NA). A: EPSCs recorded under whole cell patch clamp in the avian brain stem slice. Holding voltage (V_hold) at +50 mV while electrically stimulating the nerve inputs resulted in a large EPSC in control ACSF. Slow component was blocked by 20 μM MK801; total blockade occurred with 10 μM DNQX + 20 μM MK801. Antagonists against ionotropic GABA and glycine receptor were present in all conditions (10 μM gabazine and 3 μM strychnine). Stimulus artifact was digitally blanked for clarity. B: schematic of experimental configuration recording in A. Transverse section, dorsolateral quadrant of the brain stem. NM, nucleus magnocellularis. I_m, membrane current. Adapted from MacLeod and Carr (2005) with permission. C: time course of MK801 blockade in experiment in A. Peak EPSC (open circles) persisted in MK801, while the late EPSC (filled circles) showed activity-dependent decrement. Timing of the current measurements is illustrated by position of open and filled circles above traces in A. AMPAR, AMPA receptor; NMDAR, NMDA receptor. D: summary data of the remaining EPSC following maximal MK801 blockade in 7 neurons (mean + SD). n.s., Not significant. **P < 10⁻⁶.

To measure STP, the fast DNQX-sensitive (AMPAR mediated) component of the EPSCs was used. These were measured under voltage clamp as a peak current at negative voltages (−60 mV). Peak EPSC amplitudes were measured relative to baseline using an exponential fit to account for the decay of the previous EPSC in the pair/train and reported as normalized values relative to the initial EPSC amplitude. All data are reported as means ± SD, unless otherwise specified, and statistical testing used Student's t-test. Exponential fits of the pharmacological blockade time courses used the built-in curve-fitting algorithm in IGOR Pro (WaveMetrics). All time courses were well fit by single exponentials.

Estimating release probability.

To estimate the release probability, the decrement in EPSC amplitude in successive trials during application of MK801 (μM) was modeled with the difference equation described in Hessler et al. (1993):

$$\text{R}\left(t+1\right)=\text{R}\left(t\right)\times \left(1-\text{F}\right)$$ (1)

In this equation, R(t) and R(t + 1) are the responses for the tth trial and the successive (t + 1)th trial. F is the fraction of NMDA channels blocked by MK801 given release with transmitter release in one trial, out of all available (as yet unblocked) NMDA channels, which can be calculated as

$$\text{F}=\text{Pf}\times \mathrm{Pr}\times \text{FB}$$ (2)

where FB is defined as the fraction of available NMDA channels blocked by MK801 and Pr is the probability of neurotransmitter release at the terminal. While Pf accounts for the probability that the fiber stimulated actually fires an action potential, we assume this to be unity in our experiments. In NM, Pf was almost certainly 1 in our experiments, since failures would be easily observed as large fluctuations in the currents. In NA, we carefully controlled for stimulus intensity to find a plateau region clear of observable threshold effects; however, if the Pf across the multiple fibers was on average <1, this would lead to an underestimate in the Pr.

From the difference equation Eq. 1, we can extract the time constant τ of decrement by solving

$$\text{R}\left(t\right)={\text{R}}_0\hspace{0.17em}\exp \left(-t/{\tau }_{\text{syn}}\right)$$ (3)

where R₀ is the initial response in MK801, which we normalize to 1.

Defining the time constant τ_syn in terms of F:

$${\tau }_{\text{syn}}=-1/\ln \left(1-\text{F}\right)$$ (4)

and rearranging Eqs. 3 and 2 and assuming Pf = 1, we can define the release probability in terms of τ_syn (the MK801 blockade rate of the synaptic response) and FB:

$$\mathrm{Pr}=\left[1-\exp \left(-1/{\tau }_{\text{syn}}\right)\right]/\text{FB}$$ (5)

We estimated FB by using the rate of decline in the direct glutamate response in the presence of MK801 (effectively, FB = F when Pf and Pr are equal to 1):

$$\text{FB}=1-\exp \left(-1/{\tau }_{\text{puff}}\right)$$ (6)

By measuring τ_syn and FB we can then determine Pr, the synaptic release probability.

Tonotopic analysis.

To measure effects by tonotopic position, in vitro recordings were recovered morphologically and mapped according to location. A summary of the locations for NA is shown as a composite in Fig. 3A. Each recording was assigned one of eight dorsal-ventral zones. Zones were divided by the long axis in a piecewise linear fashion and interpreted as 0 at the dorsolateral end and 100% at the ventral end, split into half-quartiles. Each tonotopic band runs perpendicularly to the long dorsoventral axis, as shown by auditory nerve labeling in owls (Köppl 2001) and chicks (Fukui and Ohmori 2003). For the NM data, tonotopic position was assigned according to previously described tonotopic divisions along the mediolateral and rostrocaudal axes (Fukui and Ohmori 2004; Rubel and Parks 1975). Frequency mapping in vivo in both NM and NA in chick suggests a linear distribution of best frequency (Rubel and Parks 1975; Warchol and Dallos 1990). The available published frequency data in chick (Warchol and Dallos 1990) were linearly mapped (based on percentiles) to the long axis of NA with a median frequency of 1,155 Hz along the dorsoventral segment only. With the addition of higher frequency bands in the dorsolateral bend, the overall median frequency in NA was 1,655 Hz (see Fig. 3A), only slightly higher than that in NM (1,500 Hz), suggesting that the nuclei were in good register in regard to best frequency along their tonotopic axes.

Fig. 3.

A: summary diagram of the tonotopic axis location of NA recordings for MK801 decay experiments. Dorsoventral axis was divided into half-quartile bands that were assigned an approximate sound frequency (see methods). Each recording was assigned a tonotopic percentile band for the analysis in B. B: decay time constant during MK801 blockade plotted against tonotopic location for NA and NM recordings. Linear regression analysis showed no significant effect of location on decay times for either nucleus (NA, n = 26, R² = 0.119, P = 0.085; NM, n = 13, R² = 0. 195, P = 0.13). Two-way ANOVA did, however, show a significant difference in decay times between the 2 nuclei (see text).

RESULTS

Relative Release Probability Differed Between the Two Cochlear Nuclei

To examine whether the baseline synaptic release probability at auditory nerve terminals differs in the two divisions of the cochlear nuclei in birds, NA and NM, we measured the use-dependent blockade rate of evoked excitatory synaptic currents (EPSCs) during whole cell patch-clamp recordings in chick brain stem slices containing the 8th nerve fibers. Use-dependent blockade was measured by application of the NMDAR antagonist MK801, an activity-dependent open-pore blocker. Excitatory neurotransmission at synapses in NA and NM neurons have substantial NMDAR-mediated components (MacLeod and Carr 2005; Zhang and Trussell 1994). Single EPSCs were evoked by electrical extracellular nerve stimulation every 30 s while stepping the voltage of the postsynaptic neuron to positive potentials to relieve magnesium block and reveal the combined AMPAR- and NMDAR-mediated EPSC (Fig. 1, A–C). Bath application of 20 μM MK801 blocked the late, slow component of the evoked EPSC in an activity-dependent manner (Fig. 1C; remaining late EPSC amplitude: 15.75 ± 11.02%, mean ± SD in 7 NA neurons; significant blockade relative to baseline, P = 9.5 × 10⁻⁷, Student's 1-sample t-test; Fig. 1D). The blockade time course of the amplitude of the late component was fit with a single exponential with time constants from 5.1–15.3 trials with a mean of 9.3 trials (SD ±3.7; data not shown). Application of MK801 had relatively little effect on the fast peak component (remaining peak EPSC: 86.16 ± 16.86%, not significantly different from baseline, P = 0.073). After maximal MK801 blockade of the slow component, the remaining fast component of the evoked synaptic response was blocked by DNQX, a specific AMPAR synaptic blocker (Fig. 1A). Activity dependence was assessed by following the washin of MK801 with a delay in stimulation (1–3 min). The first EPSC following the delay was on average 93.2 ± 19.5% of control, and steep declines in the EPSC amplitude only occurred with resumed stimulation. Breaks in stimulation alone did not cause declines in the EPSC (data not shown, n = 3). These data show that the MK801 blockade was activity dependent and specific.

To determine whether the blockade time course of the synaptic response, and therefore release probability, differed between NA and NM, we measured the time course of MK801 blockade in neurons in both nuclei without the AMPAR component present. The NMDAR-mediated EPSCs were pharmacologically isolated by application of 20 μM DNQX. Bath application of 20 μM MK801 blocked the NMDAR-only EPSCs in NA neurons by ∼92% (Fig. 2A; remaining EPSC current: 7.3 ± 8.1% of baseline in 29 NA neurons). The time course of the blockade for each neuron was fitted with a single-exponential function (Fig. 2B). A range of blockade time constants were observed in NA neurons, from 3.3 to 16.4 trials with a mean of 8.9 trials (SD ±3.2; Fig. 2C). The time constants of blockade were not significantly different in these experiments relative to those conducted without DNQX as described above (Student's t-test, P = 0.34).

Fig. 2.

MK801 blockade rates differed in the 2 cochlear nuclei. A: application of DNQX to EPSCs evoked at depolarized voltages eliminated the fast peak component for 1 NA neuron. Inset: an expanded view with peak indicated with an asterisk. Washin of MK801 (20 μM; in addition to DNQX) blocked the slow component in an activity-dependent manner [early (a) and late (b) blockade; see B]. B: activity-dependent blockade of the slow DNQX-insensitive component with MK801. MK801 is washed in for 1–3 min before the stimulus is restarted. Rapid blockade was fit with a single-exponential decay curve [solid line; time constant (τ) = 11.3 trials]. Same NA neuron as in A. C: histogram of MK801 blockade time courses in 29 NA neurons. Shaded bars indicate the subset of neurons (n = 11) that were recorded under higher glycine conditions identical to NM recordings (see methods). Mean blockade rate was nearly identical in the 2 conditions (5 μM glycine, 8.6 ± 3.0; 30 μm glycine, 9.4 ± 3.8, P = 0.54; mean ± SD). D: activity-dependent blockade of the slow DNQX-insensitive component by MK801 in 1 NM neuron. Similar experiment as in B except that in this example there was no delay in stimulation during washin of MK801. E: histogram of MK801 blockade time courses in NM neurons (n = 13). F: averaged time courses of blockade during synaptic stimulation while recording in NA (n = 29) or in NM (n = 13). Individual experiments were aligned on the onset of decline, normalized in amplitude, and averaged (data are means ± SE). Solid lines are the single-exponential fit of these declines that result in similar values as mean of individual fits reported in text. In separate experiments, the intrinsic blockade rate tested with direct glutamate stimulation showed no significant differences between the two nuclei (“glu puff,” see methods; Student's t-test, P = 0.35, n = 10 experiments in NA, n = 9 experiments in NM, solid and dashed lines respectively, no markers).

If release probability were higher at auditory nerve synapses in NM than in NA, blockade should occur with a faster time course. Indeed, blockade did occur more rapidly, with significantly fewer trials on average in NM than NA neurons (5.0 ± 2.7 trials, n = 13, Student's t-test vs. NA, P = 0.00057; Fig. 2, D–F). Blockade rates in NM ranged from 1.7 trials to 10.4 trials. The degree of blockade of the synaptic response was similar in the two nuclei (remaining EPSC in NM: 7.7 ± 7.7%, n = 13, no difference between NA and NM, Student's t-test, P = 0.88). These results show that while MK801 was equally effective at blocking the NMDAR EPSC in NA and NM, blockade occurred much faster at synapses in NM.

To test whether the differences in blockade rate in the two nuclei were due to differences in the intrinsic blockade rate of NMDA receptors, rather than differences in synaptic release probability, MK801 blockade experiments were repeated with direct glutamate application in place of synaptic stimulation. Glutamate puffs were applied via Picospritzer, adjusted to result in postsynaptic currents that had amplitude and duration similar to synaptic responses. These direct glutamate responses were blocked rapidly and in a stimulus-dependent manner after bath application of 20 μM MK801. No significant differences in the rates of blockade were observed between neurons recorded in NA vs. NM (blockade time constant in NA was 2.3 ± 0.72 trials, n = 10 cells; in NM 2.0 ± 0.35 trials, n = 9 cells, Student's t-test, P = 0.35; Fig. 2F). MK801 application blocked the directly applied glutamate responses as completely as it did the synaptically evoked NMDAR currents (remaining glutamate current in NA was 3.8 ± 3.7%, in NM 8.5 ± 7.0%; no difference between NA and NM, Student's t-test, P = 0.083). These results suggest that the slower blockade of synaptically evoked EPSCs must be due to differences in presynaptic release efficacy.

From the blockade rates of direct glutamate puffs, we estimated the “blockade fraction”: the fraction of available receptors blocked with this concentration of antagonist per trial (see methods). This fraction was 0.353 for NA recordings and 0.393 for NM recordings. Overall, the average blockade fraction was 0.372 (corresponding to an average time constant of 2.15 trials), similar to previous estimates in neocortical experiments (Hessler et al. 1993). With these blockade fraction calculations, the release probability was estimated to be 0.35 ± 0.16 at NA synapses (mean ± SD; 95% CI = [0.292, 0.408]) and 0.51 ± 0.22 at NM synapses (95% CI =[0.399, 0.619]). These values were significantly different (P = 0.008, Student's t-test), confirming the hypothesis that the initial release probability at auditory nerve synapses onto NA neurons was significantly lower on average than the release probability at those onto NM neurons.

By mapping the location of the recordings in NA, we examined whether the measured MK801 blockade rate, and therefore release probability, was correlated with the putative tonotopic axis. The tonotopic axis in NA is oriented in the dorsoventral direction, with the highest frequencies in the dorsolateral bend (Fukui and Ohmori 2003; Köppl 2001; Warchol and Dallos 1990). NA was divided in a piecewise linear fashion into eight half-quartile bands (see methods). Each recording site that was recovered morphologically (n = 26) was assigned a tonotopic position (from 12.5% to 100%) (Fig. 3A). While there was a trend of increasing decay time from high to low frequency, the effect was not significant (R² = 0.119, P = 0.085; Fig. 3B). Similar analysis of the MK801 decay times in NM was performed, assigning location according to previously described tonotopic division along the mediolateral and rostrocaudal axes (Fukui and Ohmori 2003; Rubel and Parks 1975). NM recording spanned a similar range along this axis and also showed a nonsignificant trend in decay times (R² = 0.195, P = 0.13; Fig. 3B). To confirm that the differences between decays between NA and NM were not due to position, a two-way analysis of variance (ANOVA) of nucleus and position along the tonotopic axis was conducted. A significant main effect of nucleus was found [F(1,26) = 7.70, P = 0.010], while the main effect of position was not significant [F(7,26) = 2.35, P = 0.053]. These results suggest that differences in recording position cannot account for the differences in MK801 blockade time constants between nuclei.

Release Probability Was Correlated with Short-Term Plasticity in NA

To determine whether the MK801 blockade rate of the slow NMDAR-mediated EPSC was related to the STP of the corresponding fast EPSC at NA synapses, we compared the time course of MK801 blockade with the paired-pulse ratio (PPR, defined as EPSC2/EPSC1) at each synapse for 26 NA neurons. In these experiments, the PPR of the AMPAR-mediated EPSC was assessed before DNQX application. The AMPAR-mediated EPSC was isolated by holding the voltage close to the resting potential (−60 mV), at which the NMDAR-mediated component is negligible. After PPR assessment, first DNQX and then MK801 were applied during stimulation of single EPSCs at positive potentials. The PPR and MK801 blockade rates measured in the same neurons were correlated (R² = 0.29, P = 0.004; Fig. 4A). The rate of blockade was uncorrelated with initial amplitude of the AMPAR-mediated EPSC (R² = 0.094, P = 0.128; Fig. 4B). The covariation of the MK801 blockade rate and the PPR indicate that both phenomena are dependent in common on initial release probability.

Fig. 4.

A: time course of MK801 blockade was correlated with short-term plasticity in NA (n = 26, linear regression, R² = 0.29, P = 0.0043). Insets: EPSC traces in response to a paired-pulse stimulus from 2 different example neurons (a and b). B: time course of MK801 blockade was uncorrelated with EPSC amplitude (n = 26, linear regression, R² = 0.094, P = 0.128).

Short-term Plasticity Was Similar Across Different Inputs to the Same Target Neuron

The variation in PPR across different recordings from NA neurons is consistent with previous work using trains of stimuli that showed a range of synaptic plasticity with complex frequency relationships recorded in different neurons (MacLeod et al. 2007). The data presented here suggested that differences in initial release probability could explain differences in the synaptic dynamics. One question that remained was whether the plasticity measured using a single stimulus level was representative of the total population of inputs that innervate a given neuron. To answer this question, PPR was measured across a range of presynaptic stimulation strengths designed to elicit EPSCs composed of smaller (potentially single) or larger (multiple) numbers of presynaptic fibers (Fig. 5A). Increasing the amplitude of the electrical stimulus recruited additional fibers, increasing the amplitude of the initial EPSC (Fig. 5, B and C) (Allen and Stevens 1994; MacLeod and Carr 2005). The amplitudes and step sizes of the stimulus levels were adjusted to evoke EPSCs that allowed discrimination of at least two different level “plateaus” (Fig. 5, D and E).

Fig. 5.

Increasing the amplitude of the electrical stimulation in the auditory nerve tract providing input to NA neurons led to recruitment of additional fibers and larger EPSC amplitudes but no change in the paired-pulse ratio (PPR). A, left: schematic of stimulation paradigm illustrated with the inner circle stimulating a single fiber (level 1), while increased stimulus strength activated a larger circle stimulating the 1st fiber and 1 additional fiber (level 2). Right: paired-pulse stimulation (top) evoked a pair of EPSCs (bottom 2 traces), with larger EPSCs for level 2. Traces were averaged from multiple trials at each plateau indicated in B–E. Stimulus illustrates timing but not amplitude. B: amplitude of EPSC1 of the pair increased with stimulus level, showing transition point at threshold for second level. C: amplitude of EPSC2 of the pair also increased with stimulus level with nearly the same threshold point as EPSC1. D: mean (±SD) amplitudes for EPSCs for neuron shown in B and C. Filled circles, EPSC1; open circles, EPSC2. Level plateaus were selected to avoid threshold transitions. E: PPR (mean EPSC2/mean EPSC1 for each stimulus level point) is relatively stable over the range of stimulus levels, excluding threshold transitions.

Typically, the EPSC1 and EPSC2 amplitudes were constant over a plateau. As the stimulus amplitude increased there was a transition point to the next plateau level when an additional fiber(s) were recruited (Fig. 5D). Plotting the PPR for individual step levels showed a constant PPR across the plateaus (Fig. 5E). To get a good estimate of the PPR, trials were pooled within a level plateau, excluding transitions between levels (5–18 trials per pool). In 28 neurons we could determine at least two distinct levels, and in 5 cases we could also determine a third distinct level. To determine whether the PPRs at different fibers were correlated, we calculated the linear regression of the PPR at higher stimulus levels (PPR2, PPR at level 2; PPR3, PPR at level 3) vs. the lowest level (PPR1, PPR at level 1; Fig. 6A). These data were highly correlated (R² = 0.809, P < 10⁻¹¹; n = 33 PPR2 or PPR3 measurements from 28 neurons). Thus changing the stimulus amplitude did not alter the PPR.

Fig. 6.

Correlated short-term synaptic plasticity across the range of stimulus strength levels: PPRs. A: summary scatterplot of the PPR measured as the plateau level 1 (x-axis) and the PPR measured at the next higher stimulus plateau (y-axis). For a subset of neurons the PPR at a third level could be measured, and these were also plotted against the PPR at the corresponding level 1 (level 3). The points clustered along the diagonal (thin line), and linear regression (thick line) showed a strong correlation in the PPR across stimulus levels (n = 33 levels measured in 28 neurons, R² = 0.797, P < 0.0001). B: PPR showed no dependence on the amplitude of EPSC1 at any level (n = 61 levels, R² = 0.052, P = 0.066). C: to estimate the amplitude of the synaptic recruitment at higher levels (independent EPSClev2), the raw amplitudes of the EPSCs at level 1 were subtracted from the raw amplitudes of the EPSCs at level 2. To estimate the amplitude of fibers recruited to level 3, the raw amplitudes of the EPSCs at level 2 were subtracted from the raw amplitudes of the EPSCs at level 3 (not shown). D: PPR measurements based on independent EPSC estimates of level 2 (or 3) inputs were also clustered around the diagonal and correlated with the PPR measured for level 1 EPSCs (n = 33 levels, R² = 0.292, P = 0.0012). E: similar plot as B but for the subtracted EPSC amplitudes showed that PPR for each level had a weak negative correlation with EPSC amplitudes (n = 61, R² = 0.062, P = 0.043).

Because the group of fibers activated by the stronger stimuli must include the fiber(s) activated at lower stimulus levels, by definition, some correlation would be expected in the PPR between lower and higher levels. To determine whether the plasticity at independent fibers was correlated, the amplitude solely due to the additionally recruited fibers at the higher level (e.g., level 2) was estimated by subtraction of the EPSC at the lower stimulus level from that at the next higher stimulus level (Fig. 6C). The PPRs were recalculated with these subtracted “independent” amplitudes. When the regression between the independent PPR2 (or PPR3) and PPR1 was calculated (Fig. 6D), the correlation was also high (R² = 0.3187, P = 0.00076; n = 28 neurons at 33 levels), although reduced relative to the raw values. The raw PPR and the independent PPR were, at best, only weakly related to the EPSC1 amplitudes at any level (Fig. 6, B and E; R² = 0.036, P = 0.066 for raw PPR; R² = 0.062, P = 0.043 for independent PPR).

Many of the responses elicited by extracellular stimulation likely represent the summed response of multiple fibers rather than single inputs, and so the plasticity observed could still be a mixture of heterogeneous fibers. To avoid potential mixed plasticity effects, we reanalyzed a subset of the data restricted to those recordings in which the initial EPSCs had amplitudes of <200 pA that were likely to be minimal responses representing single fibers (MacLeod and Carr 2005). The similarities in the PPRs at different stimulation levels persisted, with correlations as high as for the whole population (R² = 0.750 for raw PPR, R² = 0.557 for independent PPR, n = 13; data not shown). Although we cannot completely exclude the possibility that these EPSCs reflected input from multiple fibers, the analysis of minimal responses suggests that the similarities in STP found in the larger data set were not simply due to comparing averaged traces of multiple, varied plasticity types that regress to a similar, but misleading, mean.

The PPR analysis implied that different inputs onto target neurons expressed similar initial release probability. To determine whether similarities in PPRs translated into similarities in synaptic plasticity over trains of stimuli, we repeated the level experiments with 5-pulse trains at 50 Hz instead of paired pulses (Fig. 7; n = 33 NA neurons). Two examples of neurons stimulated at multiple stimulation strengths are shown in Fig. 7, A and B. The unscaled and scaled trains of EPSCs (top and bottom trace overlays, respectively, in Fig. 7, A and B) showed strong qualitative similarities during responses evoked at lower and higher stimulus strengths. To quantify the train plasticity, we measured the ratio of the EPSC in response to the fifth stimulus relative to the initial EPSC amplitude (EPSC5/EPSC1). These train ratios at different stimulus levels were strongly correlated using raw EPSC amplitudes (R² = 0.499, P < 10⁻⁵, n = 36 levels across 33 NA neurons). By subtracting the lowest-level responses from higher-level responses, the train ratios were also measured for independent EPSC amplitudes, similar to the PPR analysis described above; independent ratios were also correlated between the higher and lower levels (R² = 0.161, P = 0.0152). The PPRs (EPSC2/EPSC1) measured within the train showed similar relationships across stimulation levels as measured for paired-pulse stimuli alone, indicating that the populations of neurons in the two analyses were similar (data not shown; raw PPR correlations: R² = 0.616, P < 10⁻⁴; independent PPR correlations: R² = 0.185, P = 0.00874). Thus the correlations in plasticity across convergent fibers were significant for short trains as well as paired-pulse stimulation.

Fig. 7.

Correlated short-term synaptic plasticity across the range of stimulus strength levels: 5-pulse trains. A: in 1 NA neuron, 50-Hz trains evoked at 3 different levels all showed depression. Top: unscaled overlay of 3 average synaptic current traces evoked at different levels. Bottom: same traces but scaled to the initial EPSC amplitude: amplitude of the final scaled EPSCs in the trains for lowest level/weakest stimulus (1), middle (2), and highest level/strongest stimulus (3). B: second example NA neuron, in which 50-Hz trains evoked at 2 different levels both showed mild facilitation: unscaled (top) and scaled (bottom) traces; amplitudes of the final scaled EPSCs in the trains for lower level/weakest stimulus (1) and higher level/strongest stimulus (2). C: train steady-state ratios (SSR = EPSC5/EPSC1) measured for raw EPSC amplitudes at multiple stimulus levels were strongly correlated (n = 36 levels across 33 NA neurons, R² = 0.499, P < 0.0001). D: train ratios (EPSC5/EPSC1) measured for independent EPSC amplitudes are also correlated, although more weakly (n = 36, R² = 0.161, P = 0.015).

DISCUSSION

In this study, we found significant differences in the average release probability at nerve terminals providing divergent input to the two divisions of the avian cochlear nucleus. Synaptic inputs to NA neurons have lower baseline release probabilities than those to NM. This difference is likely a primary contributor to the differences in STP observed in the two pathways. Additionally, we found that the STP expressed at nerve fiber synapses within NA is target dependent, suggesting that the release properties of convergent fibers are coregulated.

Divergent Release Probabilities in Timing and Intensity Pathways

These experiments demonstrated that the probability of neurotransmitter release at the synapses between the auditory nerve terminals and their targets differ in the two cochlear nuclei, with lower release probabilities in NA than in NM. Relative release probability was assessed in experiments that applied the activity-dependent open-channel blocker MK801, such that slower blockade indicated lower release probabilities. The average release probabilities at synapses in NA (0.35) and in NM (0.51) were calculated from the experimentally determined receptor blockade rate and the receptor blockade fraction determined during direct glutamate application. Our approach of determining release probability in late embryonic chicks resulted in a value for NM that is in good agreement with two previous estimates using a different methodology. In those experiments, a technique that measures synaptic vesicle pool depletion to estimate synaptic transmission parameters was used (Bollmann 2000; Schneggenburger et al. 1999; Wu and Borst 1999). One recent study estimated vesicle release probability at nerve inputs to NM to be 0.54 in hatchling chicks [postnatal day (P)1–5, 2 mM external Ca²⁺ concentration; Taruno et al. 2012]. An earlier study similarly estimated release probability to be 0.53 in hatchling chicks (P2-3) and 0.68 in late embryonic chicks [embyronic day (E)18] (Brenowitz and Trussell 2001a). The slightly higher release probability in Brenowitz and Trussell (2001a) than reported here may be due to differences in external calcium concentrations. A more recent study investigating the tonotopic distribution of synaptic properties in NM estimated an average release probability that was substantially lower than what we report (Oline and Burger 2014). Neither that study nor the data we present here showed a significant effect of tonotopic position on release probability per se, although Oline and Burger suggest that an inverse correlation of release probability with vesicle pool size (which is tonotopically distributed) could account for differences in STP along this dimension.

In this study we compared release probability in parallel in the two nuclei and show that synapses in NA had lower average release probability than in NM, which is consistent with the differences in STP previously reported. Synaptic responses to trains of nerve stimulation show greater short-term synaptic facilitation and less depression in NA than in NM (Brenowitz and Trussell 2001a; MacLeod et al. 2007; MacLeod and Horiuchi 2011; Zhang and Trussell 1994). The “mixed” plasticity in NA allows a linear transmission of spike rate that helps encode intensity information. The argument that the release probability assessed with NMDAR-mediated EPSC blockade explains the synaptic dynamics assessed with AMPAR-mediated EPSCs rests on the assumption that the release probabilities underlying the two synaptic responses are identical. In support of this assumption, we found that MK801 blockade of the NMDAR-mediated response was highly correlated with the PPR assessed with AMPAR-mediated EPSCs, suggesting that each independently reflects the same underlying baseline release statistics.

Is there an alternative explanation for the slower blockade of the synaptically evoked NMDAR responses with MK801 in NA vs. NM? If postsynaptic NMDARs in the two nuclei differed in subunit composition or had different modulation states, for example, then MK801 could possibly bind or block differentially. The glutamate puff experiments showed that the receptors took up MK801 equally in the two nuclei, however, eliminating this source of variation. On the other hand, glutamate puffs would be expected to elicit responses by both synaptic and extrasynaptic channels, and thus it is possible that normal synaptic glutamate release may not have had the same access to extrasynaptic receptors at the terminals in NM and NA because of differences in receptor localization and terminal morphology. Calyceal synapses onto NM neurons cover the cell body with multiple active zones, and extrasynaptic receptors could potentially be activated by neurotransmitter spillover from multiple sites. In contrast, bouton terminals in NA are probably shielded from spillover effects, or may have more glial contact that results in more efficient reuptake of glutamate. If extrasynaptic receptors at synapses in NA were less accessible for binding by glutamate and MK801, that could slow blockade. However, recruitment of extrasynaptic NMDARs during synaptic stimulation seems unlikely given the low stimulation frequency (0.05 Hz) in this study (Harris and Pettit 2008), suggesting that the differences in blockade rate reflect true differences in presynaptic release properties.

Release Probability Is Primary Determinant of STP in NA

Release probability is one factor that influences the dynamics of synaptic transmission. In previous studies several other mechanisms have been investigated to determine whether they account for differences in STP across NA and NM, such as activity-dependent recovery from depression, postsynaptic receptor expression, and presynaptic calcium channel expression. Synaptic recordings from both nuclei showed recovery from synaptic depression via an activity-dependent mechanism that had a biexponential time course after high-frequency stimulation (Brenowitz and Trussell 2001a; MacLeod and Horiuchi 2011). Similar effects at other synapses have been explained with calcium-based models of activity-dependent vesicle recycling, which may apply equally well to synapses in NA and NM (Dittman et al. 2000; Gersdorff and Borst 2002; Hosoi et al. 2007; Sakaba and Neher 2001; Wang and Kaczmarek 1998). In contrast, postsynaptic receptor desensitization may differ between the two cochlear nuclei. Postsynaptic AMPAR desensitization contributed to depression at NM synapses (Brenowitz and Trussell 2001b) but made no significant contribution during train stimulation at NA synapses (MacLeod and Horiuchi 2011) despite the presence of glutamate receptors with fast gating and desensitization kinetics in both nuclei (Parks 2000; Raman et al. 1994; Sugden et al. 2002). There also do not appear to be any differences in the types of calcium receptors underlying excitatory neurotransmitter release in NA and NM, based on their pharmacological profile (Ahn and MacLeod 2013; cf. Sivaramakrishnan and Laurent 1995). Together with the present study, these results suggest that release probability may be the key determinant of short-term synaptic dynamics in the cochlear nuclei.

Convergent Afferents Show Target-Dependent Synaptic Plasticity

In birds and mammals, the auditory nerve bifurcates and targets divergent branches to different divisions of the cochlear nucleus. Differences in the synaptic plasticity at these afferent terminals imply that plasticity is target nucleus dependent. Within NA, there was substantial variation in plasticity across different neurons. The convergence of multiple afferent inputs onto these neurons allowed the testing of a corollary hypothesis, that convergent fibers would have similar plasticity, implying that the release probability, a presynaptic property, would depend on the specific postsynaptic target. Supporting this hypothesis, we found a strong correlation in PPR measured across stimulus levels. The correlation was strongest for the PPR but was also strong for short trains of stimuli.

Target cell-dependent specificity of synaptic plasticity has been described in numerous brain regions, and evidence has been mounting in support of the specific regulation of presynaptic properties depending on the identity and molecular signature of the postsynaptic target (Blackman et al. 2013). In the mammalian ventral cochlear nucleus, a recent study measured short-term depression and MK801 blockade at nerve inputs to bushy cells (Yang and Xu-Friedman 2012). Both measurements showed that convergent fibers on the same postsynaptic target were more similar in their release properties and plasticity than divergent fibers. Target-specific regulation of STP has also been documented in neocortical, hippocampal, and cerebellar preparations (Bao et al. 2010; Koester and Johnston 2005; Markram et al. 1998; Reyes et al. 1998; Scanziani et al. 1998; Sun et al. 2005; Sylwestrak and Ghosh 2012). The mechanisms by which the postsynaptic neuron “directs” the presynaptic function are not yet known but could involve transsynaptic signaling via cell-cell adhesion molecules, retrograde signaling of a diffusible molecule, or specifically localized expression of ionotropic or metabotropic channels that coregulate release probability (Blackman et al. 2013; Davis and Müller 2015; Engel and Jonas 2005; Pelkey et al. 2006; Shigemoto et al. 1996; Sun et al. 2009; Sylwestrak and Ghosh 2012; Tomioka et al. 2014; Vitureira et al. 2011). The finding reported here that release probability across fibers is specified in a target-dependent manner suggests that release probability, and short-term synaptic dynamics, is important for the auditory information transmission at the nerve-to-cochlear nucleus connection.

NMDAR Rundown Observed in NM

During recordings in NM, we frequently observed significant rundown of the NMDAR-mediated response independent of MK801 application (data not shown). Rundown of the NMDAR-mediated responses has been attributed to several forms of NMDAR desensitization (McBain and Mayer 1994; Rosenmund et al. 1995; Rosenmund and Westbrook 1993). Some of these effects can be mitigated by increasing the concentrations of glycine in the external solution, while a calcium-dependent form of inactivation can be mitigated by increased calcium buffering in the intracellular solution (Dingledine et al. 1999; Legendre et al. 1993; McBain and Mayer 1994). Concerned that rundown could bias MK801 blockade rates, we modified our experimental conditions to mitigate these effects by increasing the glycine in the external solution (from 5 μM to 30 μM), increasing the intracellular calcium buffer BAPTA (10 mM), and increasing intracellular ATP supplementation with Na₂ATP (4 mM). Rundown was less frequently observed in NA recordings, and the modified protocol did not have any significant effect on the time course of blockade in these neurons compared with the unmodified conditions (see Fig. 2). While it is not clear why rundown was more apparent in NM than NA, the somatic localization of the receptors in the adendritic NM neurons may lead them to be more exposed to intracellular dialysis effects during whole cell patch-clamp recording. Alternatively, these forms of rundown are subunit specific, prominent with NR2A-containing receptors, but not significant with NR2B-containing receptors (Dingledine et al. 1999). There is conflicting evidence related to the subunit composition of NMDARs in NM, with mRNA levels indicating a developmental shift from NR2B to NR2A (by E18) (Tang and Carr 2007) but physiological experiments suggesting no functional change in the active receptors (Lu and Trussell 2007). If NMDARs contained higher levels of NR2B in NM than NA, then that might explain the relative lack of desensitization compared with NM. In our experiments, however, any differences there might be in the receptor subunit composition had no apparent effect on the MK801 blockade during direct glutamate stimulation. Furthermore, differences in NMDAR rundown are unlikely to contribute to differences in the measured MK801 blockade because of the mitigation measures taken and the slow time course of the rundown. Thus the differences in the MK801 blockade rate are likely a real indicator of presynaptic release probability differences in the two pathways.

Conclusions

Release probability appears to be a major factor in differentiating the STP in the parallel pathways of the avian auditory brain stem. Multiple convergent fibers providing synaptic inputs to target neurons show similar plasticity characteristics during paired pulses and short trains. Because the STP was highly dependent on the presynaptic release probability, the data strongly suggest that release probability across fibers is coregulated in a target-dependent manner.

GRANTS

This work was supported by National Institute on Deafness and Other Communication Disorders Grant DC-10000.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: J.A. and K.M.M. conception and design of research; J.A. and K.M.M. performed experiments; J.A. and K.M.M. analyzed data; J.A. and K.M.M. edited and revised manuscript; J.A. and K.M.M. approved final version of manuscript; K.M.M. interpreted results of experiments; K.M.M. prepared figures; K.M.M. drafted manuscript.