INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

During voluntary contractions, there is a weak tendency for pairs of motor units in the human muscle to discharge within a few milliseconds of one another. This has been observed in a number of limb and trunk muscles (e.g., Datta and Stephens 1990; Dengler et al. 1984; Dietz et al. 1976). This phenomenon has been termed "short-term synchrony" and is believed to result from common excitatory inputs that branch widely to innervate some or all motoneurons in the motoneuronal pool of a given muscle (Sears and Stagg 1976). The common inputs that could produce synchrony in tonically active motoneurons include the corticospinal pathway (reviewed by Porter 1985), the Ia afferent pathway (Mendell and Henneman 1971), and spinal interneurons (Fetz et al. 1999; Hamm et al. 1999). Many individual neurons in these pathways have inputs to most or to all motoneurons in the motoneuronal pool of a given muscle. It has also been suggested that common inhibitory inputs can generate synchronous discharge (Lytton and Sejnowski 1991; Miles et al. 1987; Moore et al. 1970), but there has been relatively little experimental investigation of this possibility (see, however, Cobb et al. 1995; Gauck and Jaeger 2000).

Motor-unit synchrony is classically detected by cross-correlation analysis of the spike trains of pairs of motoneurons (Moore et al. 1966). Synchrony is manifest by a peak at or near the time of discharge of the reference unit (time 0) in the cross-correlation histogram (CCHist). The size of CCHist peak is thought to be related to the proportion of shared input (e.g., Datta and Stephens 1990; Nordstrom et al. 1992), and its width may indicate whether or not the synchronizing input is derived from common input from branched presynaptic axons or synchronization of the discharge of presynaptic fibers (Kirkwood et al. 1982). However, the CCHist cannot distinguish between common excitatory and common inhibitory inputs. It has recently been proposed that the sign of the net common input may be determined by plotting the instantaneous frequency of one motor unit's discharge against the discharge time of the other (Türker et al. 1996). In that study, the perispike frequencygram (PSFreq) indicated a long-lasting increase in the discharge frequency in human masseter motor units and a decrease in tibialis anterior motor units that coincided with the peak in the CCHist. This was interpreted as evidence that during voluntary discharge of motor units, the sign of the net common input that drives the masseter motoneurons is excitatory, whereas the net common input to tibialis motoneurons is inhibitory.

To determine whether the PSFreq can in fact indicate the sign of the underlying process that generates synchronous discharge, we simulated the arrival of net excitatory and net inhibitory common inputs by applying trains of injected current transients to rat hypoglossal motoneurons recorded in brain stem slices. This method allowed us to compare the ability of excitatory and inhibitory inputs to generate synchronous discharge in motoneurons and to test the usefulness of the PSFreq technique for bringing out the properties of the underlying common input. We found that the amount of synchronization was affected by both the sign and the composition of the injected current waveforms used to simulate common input. Common input waveforms composed of depolarizing (excitatory) transients produced larger and narrower CCHist peaks than did those composed of hyperpolarizing (inhibitory) transients. The PSFreqs associated with net excitatory inputs generally exhibited a clear increase in discharge rate at the time of the CCHist peak, whereas those associated with inhibitory inputs generally exhibited little or no change in rate during the peak. Although, these differences can be used to estimate the sign of the underlying net synchronizing input, the PSFreq analysis cannot be used to unambiguously distinguish common excitation from common inhibition. A preliminary account of some of these results has been presented (Türker and Powers 2000).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The basic surgical and experimental procedures we used to obtain intracellular recordings from rat hypoglossal motoneurons in vitro have recently been described in detail (Poliakov et al. 1996, 1997; Sawczuk et al. 1995, 1997), so only the main features of the protocols will be summarized here.

Rat hypoglossal motoneurons were studied in 400-µm-thick brain stem slices obtained from 18- to 24-day-old Sprague-Dawley rats. Following the induction of anesthesia with an intramuscular injection of a mixture of ketamine (68 mg/kg) and xylazine (4 mg/kg), a section of brain stem was removed and glued to a Plexiglas tray filled with cooled, artificial cerebrospinal fluid in which Na⁺ had been replaced with sucrose [S-ACSF; composed of (in mM) 220 sucrose, 2 KCl, 1 · 25 NaH₂PO₄, 26 NaHCO₃, 2 MgCl₂, 2 CaCl₂, and 10 glucose]. A series of transverse slices was cut throughout the length of the hypoglossal nucleus, transferred to a holding chamber, and incubated at room temperature (19-21°C) in S-ACSF for 30 min, followed by 30 min incubation in standard ACSF (the same as S-ACSF except that sucrose was replaced with 126 mM NaCl).

For the experimental recordings, the slices were submerged in a recording chamber and perfused with ACSF at a rate of 2 ml/min. We used glass micropipettes filled with 3 M KCl (electrode resistances of 20-60 M) to obtain intracellular recordings from hypoglossal motoneurons. Motoneuron identity was based on location and on the similarity of cell properties to those reported in previous studies (Haddad et al. 1990; Sawczuk et al. 1995; Viana et al. 1993a,b).

Recording and current injection techniques

Motoneurons were initially accepted for study if they exhibited resting potentials more negative than 60 mV and action potentials with positive overshoots. We performed the complete experimental protocol only on those motoneurons capable of producing sustained, repetitive discharge in response to long (35 s), suprathreshold current steps. Following impalement, we used steps of injected current to determine the motoneuron's input resistance, rheobase, and steady-state current frequency relation (cf. Sawczuk et al. 1995). We then measured the motoneuron's response to a series of injected current waveforms consisting of suprathreshold current steps with superimposed noise and synaptic-like current transients. The waveforms were stored as sequences of digitized values and converted to a current command via a D/A converter at a rate of 10 kHz. The membrane potential was simultaneously sampled at the same rate and stored.

Stimulus waveforms

Repetitive discharge was elicited by 42-s injected current waveforms consisting of four components: 1) a 35-s suprathreshold step, 2) a 26-s "background" noise waveform starting 5 s after the onset of the step, 3) a 26-s "common input" waveform, also starting 5 s after the step onset, and 4) two series of eight 1-ms, 1-nA hyperpolarizing current pulses applied before and after the current step (Fig. 1A). The background noise waveform (top trace in Fig. 1B) was filtered Gaussian noise with a zero mean amplitude. The standard deviation of this waveform was generally 0.073 nA, and its time constant was 1 ms. The common input waveform (middle trace in Fig. 1B) was generated by summing a number of 26-s trains of brief transients designed to mimic the synaptic currents associated with repetitive discharge in a set of presynaptic fibers. The intervals between transients in a given train were drawn randomly from a normal distribution with a coefficient of variation of 0.2. The mean intervals in different trains ranged from 14 to 42 ms. The number of transients in the 26-s trains ranged from 612 to 1,815, and the timing of the transients in the different trains were independent of one another. The time course of individual transients in each train were specified by alpha functions (Rall 1967), which typically had rise times of 0.5 ms and amplitudes ranging from 0.0075 to 0.24 nA.

View larger version (32K): [in this window] [in a new window]   Fig. 1. Injected current waveforms and motoneuron responses. A: injected current waveform (bottom trace) and motoneuron response (top trace). The current waveform was composed of hyperpolarizing pulses at the beginning and end, a 35-s suprathreshold current step, and 26 s of superimposed noise. The transparent box indicates the section of the current noise waveform that is shown in the bottom trace of B. B: the total noise component (bottom trace) was composed of filtered Gaussian noise (background current noise, top trace) and several superimposed trains of current transients specified by alpha functions (common input current, middle trace). C: the membrane response to the noise waveform and its subcomponents was calculated by convolving the injected current waveforms with the passive impulse response of the motoneuron.

The amplitude and the time course of the simulated common postsynaptic potentials (CPSPs) produced by the common input current waveforms were calculated by convolving the current waveform with an estimate of the passive impulse response of the motoneuron. The passive impulse response was estimated from the average membrane response to the series of hyperpolarizing current pulses preceding and following the injected current step (cf. Türker and Powers 1999). The CPSPs waveform therefore illustrated the net potential that was added to the membrane potential of the cell during the injection of the common current waveform (Fig. 1C).

A number of different common input waveforms were synthesized by varying the number of input trains and the amplitudes of the individual current transients. In each case, the original version of the waveform was applied to simulate net excitatory common input, and its inverse was applied to simulate net inhibitory input. Figure 2 illustrates the features of the three most commonly used waveforms. Wave 1 (dotted traces) was composed of 150 trains of 0.03-nA peak amplitude excitatory transients and 150 trains of 0.0075-nA peak amplitude inhibitory transients. (The net input was excitatory, due to the larger amplitude of the excitatory transients.) The waveform had a symmetric amplitude distribution (Fig. 2A) as did the voltage fluctuations induced by this injected current waveform (Fig. 2C). The dotted traces in Fig. 2, B and D, show samples of the current and voltage waveforms. The fluctuations in current and voltage are symmetrically distributed around the mean value (dashed line). Wave 2 (thin solid traces) was composed of 26 trains of 0.12-nA transients. Although the amplitude distribution of the current waveform is clearly skewed, that of the voltage waveform is nearly symmetric. Wave 3 (thick solid traces) was composed of six trains of 0.24-nA peak amplitude transients. Both the current and voltage noise waveforms exhibit clearly skewed amplitude distributions (Fig. 2, A and C), and the individual current transients and postsynaptic potentials (PSPs) are clearly visible (Fig. 2, B and D).

View larger version (31K): [in this window] [in a new window]   Fig. 2. Different common input waveforms. Three different common input waveforms were generally used: Wave 1 (dotted traces), consisting of a large number of small transients; Wave 2 (thin solid lines), consisting of a smaller number of larger transients; and Wave 3 (thick solid lines), consisting of a small number of the largest transients. A: amplitude distributions for the 3 current waveforms. B: 100-ms segments of the 3 current waveforms. C: representative amplitude distributions of the calculated voltage noise produced by the 3 different waveforms. D: 100-ms segments of the voltage noise. The insets in A and C show the spectral composition of the current (A) and voltage waveforms (C).

Common inhibitory waveforms were generated by inverting the excitatory waveform around its minimal, baseline value (i.e., multiplying the waveform by 1). The total input signal would have had a lower mean value, leading to a reduction in discharge rate. We compensated for this by increasing the size of the current step so that the cell discharged at the same background rate as when the excitatory waveform was applied.

The size of the transients and the number of trains in these three waveforms were chosen so that the standard deviation of each waveform was approximately 0.1 nA. As a result, when the background noise and common input waveforms were applied together (bottom trace in Fig. 1B), approximately 65% of the total variance in the signal was derived from the common input waveform. The spectral composition of the three common input waveforms were nearly identical (insets in Fig. 2, A and C), with a peak around 30 Hz reflecting the mean frequency of the transients in each train, and a decline in power at higher frequencies reflecting the time course of individual transients.

Experimental protocol

The effects of the common inputs on motoneuron discharge rate and probability were determined from a series of responses to the injected current waveforms described above. On a given series of trials, the random seed used to generate the background noise waveform was varied between each trial, whereas the common input waveform remained the same. This provided a number of pairs of 26-s epochs of repetitive discharge (Fig. 3, A and B) that could be treated like simultaneous recordings from a pair of motoneurons sharing a fixed percentage of common input. The background discharge rate of the motoneuron was determined on-line by counting the number of spikes in each epoch. On different sets of trials, the amplitude of the current step was varied to maintain three ranges of discharge rate: low (6-9 imp/s), medium (10-14), and high (15-20 imp/s).

View larger version (37K): [in this window] [in a new window]   Fig. 3. Calculation of cross-correlation histograms (CCHists) and perispike frequencygrams (PSFreqs) from pairs of epochs of repetitive discharge. A: segments of the noise waveforms applied on successive trials. The background noise component of the injected current is different in the 2 trials, but the common input component is the same. B: segments of the motoneuron responses during the 2 epochs of repetitive discharge. The times of spike occurrence in the 2 discharge records are shown in the bottom traces. C: CCHist (top panel) and PSFreqs (middle and bottom panels) compiled from 6 pairs of 26-s epochs of repetitive discharge. The bottom trace in the top panel is the CCHist and the top trace is its CUSUM. The CCHist was always compiled using the spikes in the lower frequency trains as triggers, whereas the PSFreqs were calculated using either the low-frequency (middle panel) or high-frequency (bottom panel) spikes as triggers. Each PSFreq panel shows the instantaneous frequency values (dots), a running mean of the frequency values (bottom solid lines superimposed on the instantaneous frequency values), and the cumulative sum of the change in frequency relative to the background discharge rate (top solid lines).

Data analysis

In each cell, we obtained a number of derived measures from the digitized membrane responses to the stimuli, including 1) CCHists between the times of occurrence of motoneuron spikes in two epochs (top panel of Fig. 3C); 2) PSFreqs, which show the relation between the times of occurrence of the spikes in one epoch and the discharge rate in the other epoch (middle and bottom panels of Fig. 3C); and 3) the profiles of the average common current and membrane potential associated with synchronous spikes (Fig. 9).

The responses of the motoneuron to a series of common input current waveforms were used to compile CCHists and PSFreqs (Fig. 3C). CCHists and PSFreqs were typically compiled from three repetitions of the injected common input current waveform at various background discharge rates ranging from 6 Hz to about 20 Hz. Generally we calculated PSFreqs and CCHists between one set of epochs with approximately the same background discharge rate and another set with a different background discharge rate. However, when possible, we elicited two sets of epochs with approximately the same background discharge rate so that we could correlate one discharge rate against another of the same rate. Typically there were about 2,000-3,000 reference spikes (triggers) used in each CCHist and PSFreq. CCHists were constructed to cover ±100 ms around the time of occurrence of the reference spike (1-ms binwidth). PSFreqs were compiled by determining the frequency of each interspike interval in one epoch (target spikes) and correlating it to the time of occurrence of a spike in another epoch (reference spikes), also covering time lags of ±100 ms around the time of occurrence of the reference spike.

The bin counts in the CCHists were converted to probabilities of spike occurrence by dividing by the number of triggers. Cumulative sums (CUSUMs) (Ellaway 1978) were calculated from the CCHists by subtracting the mean bin count over portions of the histogram away from the central peak (lags less than 40 ms and more than 40 ms) from the CCHists, and integrating the result (top trace in top panel of Fig. 3C). The area of the CCHist peak was calculated from the difference between the maximum and minimum CUSUM values occurring at lags of between 10 and +10 ms. The duration of the CCHist peak was estimated from the difference in the lags at which the CUSUM minimum and maximum occurred. The strength of synchronization was quantified by the area of the CCHist peak above the baseline, normalized to the number of triggers (E, extra counts per trigger) (cf. Datta and Stephens 1990). To facilitate comparison to other studies using this measure, spikes from the train with the lower discharge rate were used as triggers. The significance of the CCHist peak was assessed by calculating a z-score based on the difference between the mean counts in the peak and in a baseline region of the histogram as described by Garnett and Stephens (1980).

Two different waveforms were calculated from the original PSFreqs. First, a running mean of frequency values was calculated by first sorting the frequency values by time lag and then calculating the average frequency over a window of 10 consecutive values. To determine the total duration of the change in the discharge rate in one spike train that was associated with spike occurrences in the other train, a CUSUM of frequency values (top traces in middle and bottom panels of Fig. 3C) was calculated in an analogous fashion to that described above for the CCHist. This PSFreq CUSUM was calculated by subtracting the mean prestimulus discharge rate from the PSFreq and integrating the remainder. As in the case of the CCHist, the minimum and maximum of the PSFreq CUSUM were used to determine the duration of the common input's influence on discharge rate.

The profile of the current leading to synchronous spikes was estimated from a spike-triggered average of the common input current waveform, using the synchronous discharges of a pair of spike trains as the trigger events. Spikes that occurred within the range of lags indicated by the duration of the CCHist peak were treated as synchronous discharges (usually about ±5 ms). A similar spike-triggered average was calculated using the CPSP waveform as the input. The resultant wave illustrated the variation in membrane potential caused by the common input around the time of synchronous spikes (see Average voltage changes associated with synchronous spikes and Fig. 9).

Statistical analysis

An ANOVA was used to compare the effects of different common input waveforms on CCHist and PSFreq features. A t-test was used for pairwise comparisons, and the Bonferroni-Dunn correction was used to adjust the alpha level in the case of multiple comparisons (i.e., the minimum significance level of 0.05 was divided by the number of comparisons).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We obtained several 35-s epochs of repetitive discharge from 21 rat hypoglossal motoneurons. In each cell, at least six epochs of repetitive discharge were obtained in response to different noise waveforms and repeated applications of a given common input waveform. Usually, however, 30-50 epochs were recorded from a given cell, and various common input current waveforms were tried at various background discharge rates. The main body of the results was derived from 313 cross-correlations (and PSFreqs) made using three different common input waveforms and their inverses (see METHODS). A further 39 cross-correlations were performed using epochs in which the common input was a train of low-frequency, large PSPs (see next page and Figs. 5 and 6). When the combination of background noise and common input waveforms was superimposed on the injected current steps, these stimuli typically (>70% of the trials) induced repetitive discharge that closely resembled that recorded during voluntary activation of human motor units (Kranz and Baumgartner 1973; Person and Kudina 1972), i.e., mean rates of 6-20 imp/s and coefficients of variation of 0.05-0.2.

Control experiments

To ensure that our common input waveform was the only source of synchronization in our experiments, we examined the effects of current steps + Gaussian noise in three cells. A different seed was used to generate the noise on each trial, so that there was no common component across trials. By comparing a number of sets of trials in these cells we generated nine control CCHists. The amount of synchronization under these conditions was not statistically different from zero in any of these nine control CCHists [E values ranging from 0.05 to 0.01 (0.01 ± 0.02, mean ± SD, n = 9)].

Although it might appear that a depolarizing current step could itself be a source of synchrony, this is not in fact the case. The reason is that the precise timing of neuronal spikes is strongly influenced by current transients rather than the mean level of injected current. This has been demonstrated in neocortical neurons (Mainen and Sejnowski 1995) and also confirmed in the present experiments (Fig. 4). If the initial period of discharge prior to the onset of the noise is compared across several trials, the timing of the initial 3-4 spikes following the onset of the current step can be quite similar across trials, but the timing of spikes becomes quite variable after 0.5-1 s, even in the absence of noise. This is illustrated clearly in the bottom panel of Fig. 4A, which shows a raster diagram of spike times for all 24 trials. Figure 4B shows cross-correlation histograms generated between sets of these 2-s records when all of the spikes are used (top) and when the initial 0.5 s of discharge is excluded (bottom). The significant peak in the cross-correlogram disappears if the initial 0.5 s of discharge is excluded, indicating that by the time the noise is applied, there should be no contribution of the mean current level to discharge synchrony between different epochs.

View larger version (45K): [in this window] [in a new window]   Fig. 4. Current step vs. synchronous discharge. Twenty-four trials were selected in which the number of spikes occurring prior to the noise onset was between 26 and 30 (during the 1st 2 s of current step application, cf. Fig. 1A). Four of these are superimposed in A. Note that, although the 1st 3-4 spikes occur at similar times, the timing of subsequent spikes is not so precise. This point is illustrated more clearly in the bottom panel of A, which shows a raster diagram of spike times for all 24 trials. B shows CCHists generated between sets of these 2-s records when all of the spikes are used (top) and when the initial 0.5 s of discharge is excluded (bottom). The significant peak in the cross-correlogram disappears if the initial 0.5 s of discharge is excluded, indicating that by the time the noise is applied, there should be no contribution of the mean current level to discharge synchrony between different epochs.

Synchronizing effects of single trains of large excitatory and inhibitory postsynaptic potentials (EPSPs and IPSPs, respectively)

The common input waveforms were composed of a number of high-frequency trains of individual transients (see METHODS), and this complicated the interpretation of the features in the CCHists and PSFreqs associated with these inputs. In particular, the synchronizing effect of individual transients may be affected by the presence of the other transients in the stimulus. To aid in the interpretation of these data, we also applied low-frequency trains (interstimulus intervals of 200-600 ms) of large synaptic current transients instead of the common input waveform in some of the cells. This allowed us to compare the effects of single trains of transients on firing rate and discharge probability in individual trials with their ability to synchronize discharge between pairs of trials. As previously described (Türker and Powers 1999), the excitatory transients induced large peaks in the peristimulus time histograms (PSTHs) during the rising phase and troughs during the falling phase of the EPSP on individual trains (Fig. 5, left). They also induced secondary and tertiary peaks and troughs at latencies longer than the EPSP duration. As previously described, these longer latency features reflect the properties of the autocorrelogram of the "postsynaptic" neuron (Fig. 5, left) (see also Gustafsson and McCrea 1984; Moore et al. 1970; Türker and Powers 1999). When epochs that received the same trains of EPSPs were correlated, there was a peak in the CCHist around the time 0 and an increase in the discharge rate of the target spikes around the time of occurrence of the reference (trigger) spike (Fig. 5, right column). The increase in discharge rate was most prominent when the low-frequency motoneuron spike train was used as the trigger source (Low vs. High).

View larger version (36K): [in this window] [in a new window]   Fig. 5. Effects of large, low-frequency excitatory postsynaptic potentials (EPSPs; separated from each other by at least 200 ms) on spike timing and discharge rate in single trials and synchronous discharge in pairs of trials. The top 2 panels in the left column show the peristimulus histograms produced when the EPSP train is applied during a low background frequency epoch of discharge (PSTH-Low) and a high background frequency epoch of discharge (PSTH-High). The top traces in each panel show the cumulative change in spike probability over the background value (CUSUM). The bottom 2 panels in left column show the peristimulus frequencygrams when the EPSPs are applied during a low discharge frequency (PSFreq-Low) or a high discharge frequency epoch (PSFreq-High). In all graphs, the timing of the EPSPs was used as the trigger. The right column shows the cross-correlation histogram (CCHist) and perispike frequencygrams (PSFreqs) compiled from pairs of discharge epochs in which the same train of EPSPs was applied in the presence of different background noise current waveforms. The arrangement of the traces is the same as in Fig. 3. The spike train that is used as the trigger is indicated in each graph.

In contrast, inhibitory current transients were much less effective in inducing synchronous discharge and changes in firing rate in pairs of motoneuron spike trains (Fig. 6). Individual IPSPs were associated with a trough in the PSTH during the initial, hyperpolarizing phase of the IPSP, followed by an increased number of counts during the repolarizing phase of the IPSP. As in the case of the EPSPs, however, there were further peaks and troughs that occurred long after the termination of the underlying IPSP (Fig. 6, left column) (see also Türker and Powers 1999). The PSFreq plots of the discharge rate in individual trials showed a decrease in discharge rate during the repolarizing phase of the IPSP. When epochs that received the same IPSP trains were correlated, there was a relatively small, but statistically significant peak in the CCHist around the time 0 and decrease in the discharge rate of the source spikes around the time of occurrence of the trigger spike (Fig. 6, right column). As was the case for common EPSPs, the correlated changes in discharge frequency between pairs of trials were most prominent when the low-frequency spike train was used as the trigger source.

View larger version (38K): [in this window] [in a new window]   Fig. 6. Effects of large, low-frequency inhibitory postsynaptic potentials (IPSPs) on spike timing and discharge rate in single trials and synchronous discharge in pairs of trials. The top 2 panels in the left column show the peristimulus histograms produced when the IPSP train is applied during a low background frequency epoch of discharge (PSTH-Low) and a high background frequency epoch of discharge (PSTH-High). The top traces in each panel show the cumulative change in spike probability over the background value (CUSUM). The bottom 2 panels in left column show the peristimulus frequencygrams when the IPSPs are applied during a low discharge frequency (PSFreq-Low) or a high discharge frequency epoch (PSFreq-High). In all graphs, the timing of the IPSPs was used as the trigger. The right column shows the cross-correlation histogram (CCHist) and perispike frequencygrams (PSFreqs) compiled from pairs of discharge epochs in which the same train of IPSPs was applied in the presence of different background noise current waveforms. The spike train that is used as the trigger is indicated in each graph.

Effects of different common input waveforms on synchronization

Using any of the three common waveforms, both the net excitatory and inhibitory waveforms induced significant (z > 1.96) CCHists peaks in all comparisons (n = 313). The significant peaks reflected the fact that the proportion of common input was relatively large (65% of the total signal variance) in the present experiments.

NET EXCITATORY WAVEFORMS. The three different common input waveforms that were used differed in their ability to produce synchronous spikes and correlated changes in motoneuron firing rate. Figure 7 illustrates typical features of the CCHists and PSFreqs associated with different common input waveforms. The top row illustrates CCHists produced by a common input waveform composed of a large number (300) of trains of small (0.03 and 0.0075 nA peak amplitude) transients (Wave 1, left), one with a smaller number of larger transients (26 trains, 0.12-nA peak amplitude: Wave 2, middle), and one with a small number of large transients (6 trains, 0.24-nA peak amplitude: Wave 3, right). There were significant differences in the areas of the CCHist peaks produced by the different synchronizing waveforms (ANOVA, F = 5.880, P = 0.0039). Wave 3 produced the largest peak areas (E = 0.197 ± 0.057, range = 0.094-0.336, n = 54), whereas Wave 1 and Wave 2 produced smaller peak areas (Wave 1: E = 0.153 ± 0.010, range = 0.072-0.330, n = 15; Wave 2: E = 0.158 ± 0.052, range = 0.088-255, n = 29). The Wave 3 CCHist peak areas were significantly larger than those associated with either of the other two waves (P < 0.0167).

View larger version (31K): [in this window] [in a new window]   Fig. 7. CCHists and PSFreqs produced by 3 different net excitatory common input waveforms (Wave 1, Wave 2, and Wave 3). The top row shows CCHists (bottom traces) and CUSUMs obtained in between 2 sets of spike trains in which either Wave 1 (left), Wave 2 (middle), or Wave 3 (right) was the common input. Each panel in the middle and bottom rows shows the instantaneous frequency values (PSFreq, dots), a running mean of the frequency values (bottom solid lines superimposed on the instantaneous frequency values) and the cumulative sum of the change in frequency relative to the background discharge rate (top solid lines) for the same sets of trials used to calculate the CCHists and their CUSUMs in the top row. In the middle row the lower frequency spike train is used as the trigger for calculation of the PSFreqs and associated traces, whereas in the bottom row the higher frequency spike train is used as the trigger.

NET INHIBITORY WAVEFORMS. When the inverse forms of the common input waveforms were applied to simulate net inhibitory input, the background discharge rate declined in all cases. This was compensated by an increase in the size of the current step so that the cell fired at a similar rate to the time when the excitatory common input waveform was tested. This correction allowed the comparison of the effects of the two waveforms. With the net inhibitory common input waveform, the central peaks in the CCHists (Fig. 8, top row) were smaller and wider than the peaks generated by the common net excitatory inputs. The CCHist peak widths associated with net inhibitory input ranged from 7 to 13 ms (mean, 9.9 ± 1.5), and these were significantly wider (t = 7.33, P < 0.0001) than those associated with net excitatory inputs (mean = 8.2 ± 1.4 ms, range = 5-12). There was a significant relation between the composition of the net inhibitory waveforms and the area of the CCHist peaks (ANOVA, F = 4.96, P = 0.0103), but it was in the opposite direction to that seen for the net excitatory inputs. Peak areas were largest for Wave 1 (E = 0.115 ± 0.037, range = 0.064-0.173, n = 11), smaller for Wave 2 (0.104 ± 0.043, range = 0.051-0.222, n = 21) and smallest for Wave 3 (0.079 ± 0.029, range = 0.035-0.161, n = 22). The difference between the peak areas associated with Waves 1 and 3 was statistically significant (t-test, P < 0.0167). This rather odd result may reflect the importance of PSP arrival rate over the PSP size when using inhibitory inputs.

View larger version (32K): [in this window] [in a new window]   Fig. 8. CCHists and PSFreqs and their CUSUMs produced by 3 different net inhibitory common input waveforms (Wave 1, Wave 2, and Wave 3). The arrangement of the panels is the same as that of Fig. 7. The top row shows the CCHists (bottom traces) and their CUSUMs (top traces) produced when the inverse of either Wave 1 (left), Wave 2 (middle), or Wave 3 (right) is used as a common input. The bottom 2 rows show the PSFreqs (dots), the running mean of running mean of the frequency values (bottom solid lines superimposed on the instantaneous frequency values) and the cumulative sum of the change in frequency relative to the background discharge rate (top solid lines) for the same sets of trials used to calculate the CCHists. In the middle row the lower frequency spike train is used as the trigger for calculation of the PSFreqs and associated traces, whereas in the bottom row the higher frequency spike train is used as the trigger.

Relative timing of changes in discharge probability and rate

The qualitative features of the PSFreqs associated with net excitatory and net inhibitory inputs were often quite similar. For example, the PSFreqs associated with Wave 3 and its inverse that are illustrated in the middle panels of the right column of Figs. 7 and 8 both show a decrease in discharge rate preceding the CCHist peak, followed by an increase in discharge rate around the peak. The different effects of net excitatory and net inhibitory common inputs on correlated changes in discharge rate are most apparent when the measurements of changes in discharge rate are restricted to the duration of the CCHist peak, and the lower frequency spike train is used as the trigger. The top two sets of traces in Fig. 9 illustrate this point by plotting the CUSUMs and cumulative PSFreqs illustrated for Wave 3 in Figs. 7 and 8 on an expanded time scale. This figure illustrates that a net excitatory input produces a clear increase in discharge rate (F) during the time of the correlogram peak (i.e., the rising phase of the CUSUM, indicated by the vertical dashed lines). Although there is a slight decrease in discharge rate during the initial part of the CCHist peak when the lower frequency spike train is used as a trigger, the average change in discharge rate during the peak is clearly positive. In contrast, during net inhibitory input, the decrease in discharge rate at the beginning of the CCHist peak is more prominent, and the average discharge rate during the CCHist peak is lower than the background rate. When Wave 3 and its inverse were used as the common input waveforms and the lower frequency train was used as a trigger, the effects of net excitatory and inhibitory inputs on discharge rate during the CCHist peak were significantly different (t = 5.08, P < 0.001). With net excitatory common input, the average increase in discharge rate during the CCHist peak was 0.39 ± 0.35 imp/s (range = 0.24-1.49, n = 38), whereas the net inhibitory input produced little or no average change in discharge rate (0.04 ± 0.13, range = 0.16-0.43, n = 28). Similar trends were observed for the other two input waveforms, but the differences between net excitation and inhibition were smaller and not statistically significant.

View larger version (20K): [in this window] [in a new window]   Fig. 9. Temporal relation of correlated changes in firing rate, spike probability and the underlying membrane potential fluctuations associated with common excitation (left) and common inhibition (right). The top two traces show expanded views of the central portions of the CUSUMs of the PSFreq (thin traces, F) and CCHists (thick line, spikes) of the same traces shown in the right panels of the top and middle rows of Figs. 7 (Wave 3 as common excitation) and 8 (Wave 3 as common inhibition). The duration of the CCHist peak is indicated by the pair of dashed vertical lines in each panel that mark the beginning and end of the rising phase in the central portion of the CCHist CUSUM. The bottom solid trace in each panel shows the average common voltage fluctuations associated with synchronous spikes calculated by averaging the common PSPs around the times of synchronous spikes as described in METHODS. The superimposed dotted lines show analogous potentials calculated as described in Kirkwood and Sears (1978).

The PSFreq records associated with common excitation versus common inhibition were somewhat more distinct when the range of discharge rates used for the trigger and source units were restricted. When the lowest discharge rate units (5-10 Hz) were used as the trigger and the highest discharge rate units (13-22 Hz) used as the source, there was relatively little overlap in the frequency changes recorded during the CCHist peak between excitatory and inhibitory net common synaptic input, regardless of the common input wave used. Of the 22 cross-correlations obtained using the above criterion and common excitatory postsynaptic currents (PSCs) as the input current, in only four cases was there a decrease in the discharge rate during the CCHist peak. In all other cases there were clear increases ranging between 2 and 10% of the background discharge rates. On the other hand, in cross-correlations that are obtained using inhibitory PSCs and that used the above criterion (i.e., low against high rate), a reduction in the background discharge rate during the CCHist peak was observed in 8 of 25 cases. In the 17 cases that showed increases in the discharge rate during the peak, only 3 cases showed increases >2% of the background discharge rate. Nonetheless, the differences between excitatory and inhibitory effects on discharge rate failed to reach statistical significance (t = 1.93, P = 0.06).

Average voltage changes associated with synchronous spikes

The different effects of net excitatory and inhibitory common inputs on synchronization were associated with differences in the average voltage waveform associated with synchronous spikes. As described in METHODS, synchronous spikes between pairs of motoneuron spike trains were used as triggers to average the membrane potential fluctuations associated with the common input waveform (CPSPs, Fig. 1C). The bottom solid traces in Fig. 9 show examples of the average membrane potential fluctuations associated with the net excitatory (left) and net inhibitory (right) common inputs that produced the changes in spike timing and rate shown in the top and middle traces. As is the case for the CCHist peaks, the average voltage waveforms associated with net inhibitory input are smaller and wider than those associated with net excitatory inputs. The amplitude of the average voltage waveforms was taken as the difference between the peak value and the mean level of the voltage fluctuations (horizontal dashed lines), and its width was measured as amount of time the averaged waveform remained above this mean level. The mean amplitudes of the voltage waveforms associated with net excitatory inputs were 3.17 ± 1.26 mV (range = 1.09-8.36, n = 80), which was significantly larger (t = 6.32, P < 0.0001) than that associated with net inhibitory inputs (mean = 1.98 ± 0.72, range = 0.98-5.79, n = 55). The widths of the averaged voltage waveforms associated with net inhibitory inputs were significantly greater than those associated with excitatory inputs (t = 6.48, P < 0.0001; Excitatory widths = 10.2 ± 0.8 ms, Inhibitory widths = 11.7 ± 1.8 ms).

A comparison of the average voltage waveforms and the CCHist CUSUMs illustrated in Fig. 9 reveals that the onset of the CCHist peaks (as indicated by the onset of the rising phase in the CUSUMs) occurs slightly after the time that the average synchronizing voltage waveform first exceeds the mean level (horizontal dashed line). The average voltage waveforms associated with net excitatory inputs have a very fast rate of rise and hence are more tightly coupled to the initiation of the CCHist peaks. In contrast, the average synchronizing voltage waveforms associated with inhibitory inputs rise slowly over and above the average voltage level, and hence the initiation of the peak of the CCHist occurs about 2 ms later.

Our calculation of the average voltage waveforms associated with synchronous spikes took advantage of the fact that we could directly estimate the common input applied during collection of the spike trains. However, these estimates turn out to be remarkably similar to those obtained using the method of Kirkwood and Sears (1978), who estimated synchronizing input by averaging the total membrane noise in one motoneuron using spikes in another motoneuron as triggers (average common excitatory or ACE potential). The dotted bottom traces in Fig. 9 show ACE potentials, calculated as described by Kirkwood and Sears (1978). The ACE potentials are qualitatively similar to the synchronizing potentials shown by the solid traces, but are slightly (10-15%) smaller in amplitude, reflecting the fact that a portion of the total membrane noise is produced by a random noise component that is not common to the two spike trains.

Predicted effects of common excitation and inhibition on synchrony

The differences between the average synchronizing waveforms associated with common excitatory and inhibitory inputs reflect the fact that EPSPs and IPSPs have distinct effects on spike timing. As illustrated by the PSTHs produced by large EPSPs and IPSPs (Figs. 5 and 6), EPSPs tend to produce a sharp increase in spike probability during their rising phase, whereas IPSPs produce a decrease in spike probability during their initial phase, delaying spike occurrence over a rather broad area during their repolarizing phase. As suggested by Kirkwood and Sears (1978), an estimate of the synchronizing effect of a train of common PSPs on two spike trains can be obtained by convolving the PSTHs calculated between the common PSPs and each of the spike trains. Figure 10A shows the CCHists of Figs. 5 and 6 on an expanded time scale (thin lines) together with the CCHists predicted by convolving the PSTHs (see figure legend for details of the calculation). When large PSPs are applied at low rate, the synchronizing effect of both excitatory and inhibitory inputs is well predicted from their effects on spike timing in each of the correlated units. As is the case for the measured CCHists, the area of the peak in the predicted CCHist for common excitation is considerably larger than that of the CCHist peak predicted for common inhibition (5.9 times).

View larger version (25K): [in this window] [in a new window]   Fig. 10. Predicted and observed synchrony using different common input waveforms. A: predicted and measured CCHists for low-frequency, large EPSPs (left) and IPSPs (right). The thin lines show the CCHists of Fig. 5 (left) and Fig. 6 (right) on an expanded time scale. The thick lines show the CCHists predicted by convolving the PSTHs calculated from the PSPs to the low-frequency and to the high-frequency unit in each case. The predictions were based on the theory described in the appendix of Kirkwood and Sears (1978). The PSTHs from the common PSP to the low-frequency unit (top left panels in Figs. 5 and 6) were scaled by the ratio of the PSP arrival rate to the motoneuron discharge rate, and then reversed in time. This result, which represents the probability of a common PSP given a spike in the low-frequency unit, was then convolved with the PSTH to the higher frequency unit to give a prediction of the probability of a spike in the higher frequency unit given a spike in the lower frequency unit. B: predicted and measured CCHists for high-frequency, small EPSPs (left) and IPSPs (right). The top two rows show the PSTHs from the common PSPs to the higher and lower frequency units, and the bottom row shows the measured CCHists (thin lines) along with the predictions based on convolving the PSTHs (bold lines).

Qualitatively similar effects are obtained when using high-frequency, small PSPs as common inputs. The top set of traces in Fig. 10B show the effects of the common PSPs used in Wave 2 on the firing probability of motoneurons firing at high (PSTH high) and low (PSTH low) background rates for both excitation (left) and inhibition (right). As in the case of large, low-frequency PSPs, the PSTH peak produced by EPSPs is narrower and larger than the troughs produced by IPSPs. The thick traces in the bottom panels of Fig. 10B show the predicted effects of these common EPSPs and IPSPs on synchronization. In this example, there is also a close match between the measured (thin traces) and predicted CCHists. As in the case of large PSPs, the predicted area of the CCHist peak produced by common excitation is larger than that for inhibition, but only by about 50%. A similar difference was obtained for the predicted synchronizing effect of common EPSPs and IPSPs used in Wave 1 (i.e., excitation was 1.5 times more effective than inhibition), whereas the difference was somewhat larger for Wave 3 (excitation 1.8 times inhibition). However, the predicted CCHists did not match the measured CCHists for Waves 1 and 3. In the case of Wave 1 (many small PSPs), the predicted synchronizing effect of common input was larger than that observed, whereas for Wave 3 (fewer, larger PSPs) the predicted effect was smaller than that observed. These discrepancies may reflect that as is the case the synchronizing effect of common PSPs is not linearly related to their arrival rate, as has been reported for their effects on motoneuron firing rate (Powers and Binder 1996).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was designed to address two issues concerning the relationship between the type of synaptic inputs shared by a pair of motoneurons and the occurrence of synchronous spikes and concurrent changes in discharge rate. The first issue is the relationship between the degree of synchrony in the discharge of a pair of motoneurons and the composition of their shared synaptic input. The second issue is whether the sign of the net common input can be revealed by examination of changes in the discharge rate of one unit that are associated with the occurrence of spikes in the other unit. We elicited a series of epochs of repetitive discharge by superimposing noisy current waveforms on a long, suprathreshold step of injected current. In a given series, the input current consisted of repeated applications of a "common input" injected current waveform combined with different "background noise" current waveforms. This allowed us to treat a pair of epochs of repetitive discharge as if they represented simultaneous recordings of the discharge of a pair of motoneurons sharing a fixed proportion of their input. The common input waveform was composed of trains of transients chosen to mimic the currents associated with the activation of excitatory or inhibitory synapses and could consist of a large number of relatively small transients or a smaller number of relatively large transients.

We found that the current step and the background noise did not generate any significant synchronous discharge. Synchronization was observed only when a common input was introduced and the degree of synchrony produced by a common input waveform was strongly dependent on its composition. Both net excitatory and inhibitory common inputs generated synchronous discharge in motoneurons, but common excitatory inputs were much more effective. Excitatory common inputs composed of small numbers of large transients were more effective in synchronizing motoneuron discharge than those composed of large numbers of small inputs, as previously predicted on theoretical grounds (Segundo et al. 1968). Finally, we found that examination of correlated changes in discharge rate could not reliably bring out the sign of the net common input to motoneurons. However, in some restricted conditions, correlated changes in discharge rate with certain characteristics could be taken as indicative of net common excitation or inhibition.

Common excitation and inhibition as sources of synchronous discharge

Short-term synchronization of motor-unit discharge is often attributed to shared input from excitatory afferent fibers that branch to contact both members of a pair of motoneurons (e.g., Datta and Stephens 1990; Kirkwood and Sears 1978, 1991; Sears and Stagg 1976), together with potential contributions from di- or oligosynaptic common inputs (cf. Vaughan and Kirkwood 1997). Although theoretical and simulation studies have suggested that both common excitatory and inhibitory inputs have the capacity to synchronize neuron discharge (Aertsen and Gerstein 1985; Lisiecki and Voigt 1995; Lytton and Sejnowski 1991; Miles et al. 1987; Moore et al. 1970), there has been relatively little experimental investigation of synchronization resulting from common inhibitory inputs (Cobb et al. 1995; Gauck and Jaeger 2000). This study is the first to systematically compare synchrony induced by common excitatory and inhibitory inputs.

We found that common net inhibitory inputs typically produced smaller and broader crosscorrelogram peaks than did common excitatory inputs. This difference is likely to arise from the fact that individual EPSPs trigger spikes mainly during the brief rising phase of the EPSP, whereas IPSPs delay spikes to occur over a broader portion of their repolarizing phase (Türker and Powers 1999). The different effects of simulated EPSPs and IPSPS on synchronization can be seen most clearly under conditions in which we used large common current transients that were separated from each other by at least 200 ms. The PSTH associated with the large EPSPs was characterized by a sharp peak during the rising phase of the EPSP (Fig. 5), whereas IPSPs induced a smaller and broader peak during their repolarizing phase (Fig. 6).

The use of large PSPs also clearly demonstrated that correlated changes in discharge rate are most apparent when spikes in the low-frequency train are used as triggers to examine correlated changes in the discharge rate of the faster discharging train. This may reflect the fact that when motoneurons are firing at high background discharge rates, their rate follows the profile of the PSP better than when they are firing at lower rates (Türker and Powers 1999). It is also possible that spikes at low discharge rate are better triggers since they are more likely to fire when an extra EPSP is introduced to the driving current. This is similar to the suggestion that at very low discharge rates, motoneurons may be more sensitive to the arrival of extra synaptic inputs (Matthews 1996).

Although common excitatory input can easily be illustrated by the use of PSFreq, illustration of the inhibition was not as clear. This is because of the fact that phase advancing of spikes by EPSPs brings the spikes to the initial (rising) phase of the EPSP no matter what the background discharge rate is. Hence the phase advancement in both cells is time locked to the initial phase of the common EPSP. On the other hand, the phase delay that occurs as a result of an IPSP moves the spikes to a point that is directly proportional to the background discharge rate. This can be seen in the PSFreq record in Fig. 6. In this example, when the low discharge rate unit is used as the trigger, the reduction in the rate of the source unit occurred before the trigger (Fig. 6, right) indicating the earlier occurrence of the delayed spikes at the high rate unit (Fig. 6, left). On the other hand, when the high rate unit was used as the trigger, the reduction in the discharge rate of the source unit occurred after the trigger indicating the later occurrence of the delayed spikes at the low rate unit. Therefore unless both units are discharging at about the same frequency, the delayed spikes will not line up in time and hence the correlation between them will be weaker. However, if the discharge rates of the epochs were similar, the probability of units firing at about the same time after a synchronized discharge should increase, due to the similarity in the autocorrelograms of the two discharge trains (Moore et al. 1970). Therefore each time the spikes discharged synchronously due to the underlying common input or by chance alone, the trains stayed "synchronized" for 2-4 more spikes depending on the discharge rate and regularity of the epochs. This caused several peaks in the CCHist and hence made estimation of the true size and sign of the synchronizing voltage impossible.

The same qualitative features of synchronous changes in spike timing and discharge rate that were produced by low-frequency trains of large current transients were also seen with common input waveforms composed of many smaller transients. CCHist peaks produced by net excitatory inputs were narrower and sharper than those produced by net inhibitory inputs. In the literature, narrow sharp CCHist peaks are associated with the last-order (branched axon hypothesis of Sears and Stagg 1976) common excitatory input, whereas broad peaks are thought to reflect presynaptic synchronization (Kirkwood et al. 1982). The present study suggests that these broad peaks could also be induced by common last-order inhibitory neurons whose axons branch to contact a pair of motoneurons.

These differences in the shape and amplitude of the CCHist peaks were associated with differences in the membrane potential fluctuations produced by the common inputs around the time of synchronous spikes. To indicate the profile of the common PSPs driving synchronous discharge, we averaged the membrane potential changes produced by the common current waveform using synchronous discharges as the triggers (METHODS). The synchronizing excitatory CPSP indicated an immediate and rapid increase in the potential toward the threshold, whereas the synchronizing inhibitory CPSP showed a slower and gradual rise to the threshold. The size of the synchronizing excitatory CPSP was significantly greater than the synchronizing inhibitory CPSP. These differences may reflect the fact that common EPSPs typically generate synchronous discharge only during their relatively brief rising phase, whereas IPSPs generate synchronous discharges along a broader portion of their repolarizing phase.

This difference in the way that common EPSPs and IPSPs lead to synchronous discharge may underlie their different effects on discharge rate. As was the case for large, low-frequency EPSPs, common input waveforms composed of excitatory transients were associated with increases in discharge rate during the CCHist peak, whereas net inhibitory inputs were associated with a decrease in rate preceding the peak and little or no change in discharge rate during the peak. The ability of the PSFreq to distinguish net excitatory from net inhibitory common input is considered in more detail below.

Distinguishing net excitatory from net inhibitory common input

The cross-correlation technique, which compares the timing of spikes in one neuron with those of another, is the most well-established method for measuring synchronous discharge (Kirkwood 1979). Using this technique, it has been claimed that, in the absence of any tendency toward synchronization of the two neurons, the cross-correlation histogram will be flat and that, if a tendency exists, a central peak will appear in the record (Sears and Stagg 1976). The cross-correlogram approach has been widely used as a tool to investigate the strength of motor-unit synchronization during voluntary contraction in humans. Motor-unit synchronization is consistently observed during steady isometric contractions in a variety of muscles (Baker et al. 1992; Datta et al. 1991; De Luca et al. 1993; Kirkwood 1979).

The cross-correlation technique, however, cannot identify the sign of the shared synaptic input that is responsible for synchronous discharge. It was recently proposed that the PSFreq, which plots the instantaneous frequency of one motor unit's discharge against the discharge time of the other, can be used to distinguish net excitatory from net inhibitory common input (Türker et al. 1996). In that study, the PSFreqs indicated long-lasting increases in the discharge frequency in masseter motor units and decreases in tibialis anterior motor units. This was interpreted as indicating that during voluntary discharge of motor units, the sign of the net common input that drives the masseter motoneurons is excitatory, whereas that to tibialis motoneurons is inhibitory.

The basis for this interpretation is that synaptic potentials can be expected to influence not only the timing of spikes but also the instantaneous discharge rate. For example, if an EPSP occurs when the membrane potential of a tonically discharging motoneuron is near its firing threshold, it can shorten the interspike interval (ISI) by bringing the membrane potential to threshold (Kudina 1980; Miles et al. 1989; Reyes and Fetz 1993) and hence increase its discharge frequency (Türker and Cheng 1994). Similarly, if an IPSP occurs when the membrane potential of a tonically firing motoneuron is near its firing threshold, it can delay the threshold crossing of the membrane potential. This delay in the threshold crossing lengthens the ISI (Kudina 1980; Miles et al. 1989) and hence reduces its firing frequency (Türker and Cheng 1994). This suggests that synchronous spikes may often be associated with increases in discharge rate in the case of common EPSPs, whereas simultaneously arriving IPSPs should lead to decreases in discharge rate.

Our results indicate that, although common EPSPs generally lead to the expected increase in discharge rate, common IPSPs are not invariably associated with a clear decrease in discharge rate. The PSFreq analysis provided the clearest distinction between net excitatory and net inhibitory inputs when the common input was composed of relatively few, large current transients (Wave 3). In this case, a clear increase in firing rate was seen during the CCHist peak in response to excitatory common input, whereas inhibitory input typically produced a small, slow decline or no change in firing rate during the CCHist peak, followed by a slow increase in discharge rate (see Fig. 9, right column). In contrast, PSFreqs associated with common excitation and common inhibition were similar in the case of the common input waveform composed of a large number of small transients (Wave 1).

Our results indicate that in general PSFreq records cannot be used to distinguish common excitation from common inhibition. However, under some restricted conditions the PSFreq records with certain features can be taken as indicative of net common excitation or net common inhibition. If the discharge rate of the trigger unit is between 5 and 10 Hz and that of the source unit is over 13 Hz, then an average increase in discharge rate during the CCHist peak that is over 2% of the background rate is likely to indicate common excitation. In contrast, decreases in average discharge rate are more likely to indicate common inhibition.

It is not known whether the motor-unit synchrony observed under physiological conditions reflects the influence of relatively few, large common PSPs or a large number of relatively small PSPs. The latter situation is generally assumed to underlie synchronization (e.g., Kirkwood and Sears 1978, 1991), based on the experimental evidence that the amplitudes of unitary EPSPs from single Ia fibers onto alpha motoneurons are typically on the order of 0.1 mV (Mendell and Henneman 1971). However, most of the recordings of single-fiber PSPs have been obtained in anesthetized animals, and general anesthetics have been shown to reduce PSP size (Kullmann et al. 1989). The transmitter release capacity of Ia afferent terminals may be significantly larger than is suggested by a 0.1-mV PSP size (Walmsley and Nicol 1991), and Ia EPSPs of 1-2 mV can be observed under some conditions (Burke 1967). Further, the presence of active conductances may increase the size of both EPSPs and IPSPs at voltages near spike threshold (e.g., Stuart 1999; Stuart and Sakmann 1995). Alternatively, large PSPs could arise from synchronization of presynaptic discharge.

In the present study, the estimated PSP amplitudes produced by the individual current transients used ranged from about 0.2-0.5 mV (depending on the impulse response of the motoneuron) for the transients in Wave 1 to 0.8-2 mV for those used in Wave 3. Although the smaller PSPs are likely to be within the range of single fiber PSPs obtained in motoneurons under physiological conditions, the largest PSPs are not. Nonetheless, many of the qualitative differences between the effects of common excitation and inhibition obtained with large PSPs were also obtained with small PSPs. In addition, the features of the CCHists obtained using largest PSPs are likely to be relevant for the interpretation of synchrony among neurons with relatively large single fiber PSPs, such as cortical pyramidal cells (Thomson et al. 1988).

Another potential difference between the synchronous discharge produced by our simulated common input and that arising under physiological conditions arises from the fact that in our experiments, identical common inputs were used to generate synchronous discharge. In human cross-correlation studies, the branching presynaptic fibers providing common input to a pair of motoneurons are likely to produce different amplitude PSPs in different motoneurons since both the number and location of synaptic terminals and the input resistance are likely to be different in the two cells. Therefore this would become an added complexity when analyzing human motoneuron synchronization data. However, the qualitative features of the present results would still be expected to apply to the common inputs underlying the synchronization of human motor units.

Conclusions and implications

We have studied the properties of common inputs that generate synchronous discharge in motoneurons. We have shown that both effective excitatory and inhibitory CPSPs can generate significant synchronous discharge. The synchronous discharge was more prominent when effective excitatory CPSPs were used compared with the effective inhibitory CPSPs of the same amplitude characteristics. Under some conditions, the PSFreq analysis method was able to reveal functional coupling between motoneurons that could not be detected solely by cross-correlating the discharge times of simultaneously active motoneurons. However, the present results suggest that, except under some restricted conditions, the PSFreq analysis cannot