Spike Width Reduction Modifies the Dynamics of Short-Term Depression at a Central Synapse in the Locust

Depression and recovery at the FETi-FlTi synapse during natural spike trains

Electrical stimulation of the nerve innervating the LR evoked prolonged spike trains in the single FETi that outlasted the duration of the stimulus (Fig. 1A). Simultaneous recordings of FETi and FlTi motoneurons showed that some FlTi motoneurons did not receive synaptic inputs after electrical stimulation of the LR (Fig. 1B). The LR-evoked FETi spike trains contained similar numbers of spikes with a similar frequency distribution to those occurring during kicking. The bursts typically lasted for 1-4 sec, with each burst consisting of 10-40 spikes at frequencies up to 70 Hz (cf. Burrows, 1995). Each FETi spike was followed by an EPSP in FlTi motoneurons transmitted by a monosynaptic connection (Fig. 1A,B). The central output synapses of FETi are intermingled with its input synapses on neuropilar processes that branch from the primary neurite (Watson and Burrows, 1981; Burrows et al., 1989). At this region, there is a transition between active and inactive membrane along the primary neurite, and spike initiation takes place (Gwilliam and Burrows, 1980) (Fig. 1C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Depression at the FETi-FlTi central synapse during synaptically evoked spike trains. A, Brief electrical stimuli applied to the LN evokes bursts of spikes in the FETi that continue after the termination of the stimuli. Spikes in FETi evoke an EPSP in antagonistic FlTi via a monosynaptic connection. Inset, A schematic of the circuit generating the bursts of spikes in FETi and their corresponding FlTi EPSPs. B, A single FETi spike evokes a monosynaptic EPSP in a FlTi. Stimulation of the LN evokes EPSPs in FETi (indicated by asterisk) that evoke spikes but does not evoke inputs to a subset of FlTi motoneurons, which receive inputs only from the FETi and not from local interneurons. C, A schematic diagram of the relative positions of input (black arrows) and output (gray arrows) synapses in FETi. Dashed lines indicate the resting potential in all traces.

During LR-evoked spike trains, the amplitude of the FlTi EPSPs depressed to ∼6% (6.43 ± 0.65%; N = 19; n = 6; mean ± SE; where N refers to the number of animals and n to the number of repeats per animal) of their initial amplitude (Figs. 1A, 2Ai). The depression at the FETi-FlTi synapse was fitted by a single exponential decay with a rate constant of 1.39 sec^-1 (Fig. 2Aii). Most of this depression occurred over the initial four to six spikes, during which the instantaneous spike frequency (ISF) of FETi increased linearly before reaching a plateau phase of ∼40Hz on which fluctuations of up to 30 Hz were superimposed (Fig. 2Ai,Aii). These fluctuations had little effect on the amplitude of the FlTi EPSP (Fig. 2Ai). At the end of this plateau phase, the FETi spike frequency slowed, allowing recovery in the EPSP amplitude within 1 sec (Fig. 2Ai).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

The dynamics of depression at the FETi-FlTi synapse during natural-like synaptically evoked spike trains. Ai, The FETi ISF and posterior fast FlTi EPSP amplitude corresponding to each spike during a burst. Aii, The mean (±SE) of 19 FFlTi motoneuron EPSPs during bursts of spikes in FETi fitted by a single exponential. The shaded area indicates the plateau region of the FETi ISF. B, The relationship between the FETi ISF and FlTi EPSP amplitude for a single FETi-FlTi pair (n = 6) could be fitted by a single exponential decay. C, The relationship between depression at the FETi-FlTi synapse and the initial FlTi EPSP amplitude. The level of depression after 15 spikes was dependent on the initial FlTi EPSP amplitude; depression was greater after larger initial EPSPs than after smaller initial EPSPs.

Frequency dependence of depression at the FETi-FlTi synapse

Figure 2B shows the relationship between ISF and the FlTi EPSP amplitude for three successive spike trains recorded from the a single FETi-FlTi pair. During the first four to six spikes (i.e., before the onset of the plateau phase), the frequency of FETi spikes accurately predicted the EPSP amplitude; spikes with similar ISFs occurred at similar times within each spike train, and they, therefore, contained information about the preceding number of spikes in the train and their interspike intervals. Once the ISF had plateaued, there was little or no further depression in the FlTi EPSP, and large fluctuations in the FETi ISF had little effect on EPSP amplitude (indicated by the flattening of the exponential decay in Fig. 2B).

Depression at the FETi-FlTi synapse is dependent on the initial EPSP amplitude

Synaptic depression is often generated by presynaptic mechanisms that are affected by changes in the initial release probability (Zucker, 1989). If depression at the FETi-FlTi synapse is produced by a presynaptic mechanism, the final level of depression may be dependent on the amplitude of the first EPSP in a train. Larger initial EPSPs may indicate that more of the available synaptic vesicles have been released. The amplitude of the first FlTi EPSP varied from 11.1-20.5 mV (N = 18; n = 6), a similar range to FlTi EPSPs evoked by antidromic spikes in FETi (Burrows et al., 1989; Parker, 1995). During bursts of spikes with larger initial EPSP amplitudes, the final level of depression was greater (p < 0.05; t test) (Fig. 2C). The relationship was, however, weak (r² = 0.34) possibly because of variation in the spiking pattern during the burst. During antidromic spiking at this synapse, the final level of depression was also dependent on the initial EPSP amplitude (Parker, 1995).

Comparison with depression during antidromic trains of spikes in FETi

Despite the differences in the waveform of antidromic and synaptically generated spikes in FETi (Gwilliam and Burrows, 1979), they evoked EPSPs with similar amplitudes and time courses in FlTi motoneurons (Fig. 3A). During antidromic stimulation at frequencies up to 70 Hz, the maximum spike frequency that occurred during LR-evoked spike trains, the level of depression at the FETi-FlTi synapse did not drop below 20% (20.36 ± 0.33%; N = 19; n = 10; mean ± SE) of the initial amplitude (Fig. 3B). The relationship between the antidromic spike frequency and the steady-state EPSP amplitude, at which there is no further reduction in EPSP amplitude, was fitted by a single exponential with a rate constant of 6.21 sec^-1. Depression during antidromic spike trains never exceeded depression during synaptically generated spike trains, even when the frequency of the antidromic spikes was greater than the maximum frequency of synaptically generated spikes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Depression at the FETi-FlTi synapse during antidromically generated spike trains. A, EPSPs evoked in FlTi motoneurons have similar amplitudes and durations after antidromic spikes (left) and synaptically generated spikes (right) in FETi. B, Steady-state EPSP amplitudes after depression at the FETi-FlTi synapse during antidromic spike trains in FETi calculated after the first 10 spikes (mean ± SD; N = 12; n = 10) fitted with a single exponential. Inset, Mean steady-state EPSP amplitudes at 0.5 and 5 Hz from a single FlTi motoneuron (n = 10).

Modeling short-term depression at the FETi-FlTi synapse

A model was used to compare the dynamics of depression at the FETi-FlTi synapse during synaptically and antidromically generated spike trains. The model produced an accurate fit of the depression at the FETi-FlTi synapse during a synaptically driven FETi spike burst (N = 1; n = 6) (Fig. 4A), shown by plotting the observed against the predicted EPSP amplitudes (Fig. 4B). The mean error in the fit of the model to LR-evoked spike trains was 0.16 ± 0.06 (N = 19; n = 6; for details of the calculation of the mean error, see Materials and Methods). To ensure that a more accurate fit could not be obtained from the model using additional parameters, a facilitation time constant was added. This produced only a small improvement in the fit (0.004 ± 0.0004) (Fig. 4C,D), suggesting that any underlying facilitating mechanisms make only a small contribution.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Modeling depression at the FETi-FlTi synapse during LR-evoked spike bursts. A, Averaged FlTi motoneuron responses during FETi spike bursts (n = 6) were fitted by a simple three-parameter model of synaptic depression (U = 0.61; A = 159.59; τ_Rec = 505.32). The timing of each spike in a train was measured from the first spike in the train, and the observed (black) and predicted (gray) FlTi EPSPs are plotted for each spike. B, Plotting the model against the data show the fit of the model to the data. Points lying on the diagonal indicate a perfect fit. C, The addition of a facilitating parameter to the model did not improve the fit of the model to the data (U = 0.62; A = 158.07; τ_Rec = 504.02; τ_Facil = 13.18). D, As in B. E, The parameters derived from fitting the model to the data enabled the model to accurately predict the responses of the same FlTi motoneuron to single FETi spike bursts not used to calculate the original model parameters. F, As in B.

The model was tested further by producing predictions to novel spike trains (individual spike trains) (Fig. 4E,F). Errors between the model predictions of novel LR-evoked spike trains and the observed EPSP amplitudes were greater than the errors in the fit of the model to specific data sets, although this increase in the error was small (0.05 ± 0.02) and fell within the variation in the fitted model spike trains (Fig. 4E,F, compare with Fig. 6B,D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Modeling depression in multiple types of FlTi motoneurons. A, The model accurately average response of an intermediate FlTi motoneuron during FETi spike bursts but with different parameters to the PFFlTi motoneuron shown in Figure 4 (U = 0.72; A = 97.04; τ_Rec = 729.24). Inset, FETi makes monosynaptic connections to a population of nine FlTi motoneurons, up to three of which were recorded sequentially in a single preparation. B, Plotting the model against the data show the fit of the model to the data. Points lying on the diagonal indicate a perfect fit. C, The fit of the model to a second FFlTi motoneuron (anterior position) recorded in the same preparation (U = 0.67; A = 109.11; τ_Rec = 508.87). D, As in B.

For constant frequency antidromic spike trains, the model consistently predicted a faster rate of depression and a smaller EPSP amplitude for a given spike frequency than was observed (Fig. 5A,B). There was a significant increase in the mean error of model predictions for antidromic spike trains at frequencies above 10 Hz (0.99 ± 0.14; p < 0.001; N = 17; n = 10). To test whether the model could accurately fit the antidromic data, both with and without the additional facilitation parameter, it was allowed to iteratively fit the antidromic data (Fig. 5C-F). The model was then able to fit the antidromic spike trains, although its parameters were significantly different from those used to fit the LR-evoked spike trains (Fig. 5C,D). The addition of a facilitation parameter, as for the synaptically evoked spike trains, did not affect the fit of the model (Fig. 5E,F).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Model predictions do not fit observed antidromic spike trains. A, Parameters derived from fitting the model to FlTi responses during FETi spike bursts could not predict the responses of the same FlTi motoneurons to constant frequency antidromic spike trains at 20 Hz. B, Plotting the model against the data show the fit of the model to the data. Points lying on the diagonal indicate a perfect fit. C, The model could produce fits to constant frequency antidromic spike trains (U = 0.25; A = 394.31; τ_Rec = 1420.70). The gray bar indicates a 1-sec period of recovery. D, As in B. E, The model with an additional facilitating time constant produced more accurate fits to the constant frequency antidromic spike trains (U = 0.43; A = 219.90; τ_Rec = 1057.71; τ_Facil = 226.94). F, As in B.

The model was able to fit the depression at the FETi-FlTi synapse for all types of FlTi motoneurons observed during simultaneous recordings and during sequential recordings, including both fast and intermediate FlTi motoneurons. The model parameters were unique for each FlTi motoneuron, although all of a particular class (e.g., fast FlTi motoneurons) had parameters clustered around similar values. Figure 6A-D shows two FlTi motoneurons recorded sequentially from the same preparation as the FlTi motoneuron in Figure 4. Two (Figs. 4, 6C,D) were fast FlTi motoneurons whereas the third (Fig. 6A,B) was an intermediate motoneuron. This suggests that, although there is some variation between FlTi motoneurons, the dynamics of depression at the FETi-FlTi synapse occurs within a restricted range.

One prediction made by the model is that if a presynaptic neuron spikes above a certain frequency, known as the limiting frequency, the amplitude of the EPSP evoked by each spike will decrease in proportion to the frequency (1/f). Therefore, above the limiting frequency, the mean postsynaptic depolarization from the resting potential saturates (i.e., as the spike frequency increases, the EPSP frequency increases but the individual EPSP amplitude decreases). Above the limiting frequency, increases in spike rate do not increase the mean level of depolarization in the postsynaptic neuron. Hence, the limiting frequency sets the frequency above which synapses are unable to transmit information about the presynaptic spike rate (Tsodyks and Markram, 1997). Before reaching this stationary level of mean postsynaptic depolarization, synapses can still transiently affect the level of postsynaptic depolarization. At the FETi-FlTi synapse, the limiting frequency was below 5 Hz, suggesting that it does not transmit information about the FETi spike rate over the majority of the natural spike frequency range and is more suited to transmitting information about the onset of spiking in FETi (Fig. 7A). Coupled with high values of U (0.52-0.89), this suggests that the FETi-FlTi synapse primarily transmits temporal information about the onset of spiking in FETi (Fig. 7B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Predictions from the model suggest a behavioral role for depression at the FETi-FlTi synapse. A, Model predictions of steady-state EPSP amplitude fit a 1/f relationship for frequencies above 5 Hz (see Results for details). B, Model predictions of the steady-state EPSP amplitude for all the FlTi motoneurons analyzed. The gray lines indicate the upper and lower limits of the model predictions determined by the parameters for individual FlTi motoneurons, and the mean is shown by the bottom black line. The top black is the mean steady-state amplitude for a typical FlTi motoneuron during antidromic spike trains.

Presynaptic mechanisms

Changes in the height and width of the FETi spike were measured during synaptically evoked spike trains to determine whether changes in these parameters could account for the increased levels of depression relative to those during antidromic spike trains (Fig. 8A). The FETi spike height was reduced to ∼79% of its initial height (78.8 ± 4.3%; N = 12) and was accompanied by a reduction in spike width to ∼55% of the initial width (54.7 ± 7.6%; N = 12) (Fig. 8A,B). These changes in the FETi spike waveform occurred with the same time course as the initial depression of the FlTi EPSP (Fig. 8A). Changes in the FETi spike height and width correlate with changes in both the ISF (Fig. 8C) and FlTi EPSP amplitude (Fig. 8D), decreasing spike height and width correlating with depression in the FlTi EPSP and increasing spike height and width correlating with recovery of the EPSP amplitude.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Changes in the presynaptic waveform during natural-like bursts of spikes in FETi. A, The spike height and width varied during synaptically evoked bursts of spikes in FETi. B, Three FETi spikes aligned at their onset show clear changes in both spike height and width. C, Correlation of the FETi spike height and width with the frequency of spiking during bursts of spikes as shown in A. D, Correlation of the FlTi EPSP amplitude with the FETi spike height and width during synaptically evoked bursts of spikes.

In antidromically induced trains of spikes, there was a smaller reduction in the FETi spike height to ∼82% (82.1 ± 3.2%; N = 12) of the initial height at 70 Hz (Fig. 9A). However, spike height continued to decrease as the ISF of the antidromic spike trains increased, although there was no accompanying drop in the FlTi EPSP (Fig. 9B, shaded area). In contrast, there was almost no change (95 ± 3.4%; N = 12) in the spike width during antidromic spike trains at frequencies of up to 70 Hz (Fig. 9A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Changes in the presynaptic waveform during antidromically generated spike trains in FETi. A, The steady-state spike height (black diamonds, calculated after the initial 10 spikes) and steady-state spike width (open circles) plotted against the antidromic spike frequency. Inset, The steady-state spike heights at 0.5 and 20 Hz. B, Relationship between the steady-state EPSP amplitude and the steady-state spike amplitude. The shaded region shows the plateau in the FlTi EPSP amplitude despite a continued drop in the FETi spike amplitude.

The FETi spike width during the synaptically evoked spike trains correlated strongly with the membrane potential of FETi immediately preceding a spike, increasing membrane potential leading to narrower spikes (Fig. 10A). To test whether changes in the width of synaptically evoked spikes had a causal effect on the FlTi EPSP, hyperpolarizing current was injected into FETi before the first spike of a LR-evoked spike train, increasing the spike width. The increased spike width and height produced a concomitant increase in the FlTi EPSP amplitude, implying a causal relationship between the presynaptic waveform and the FlTi EPSP (Fig. 10B). Reduction in the FETi spike width with increasing depolarization suggests that synaptic inputs may be responsible for this spike width modulation. Additionally, reduction in spike width could also be induced during antidromic spike trains by evoking synaptic inputs to FETi, providing evidence that synaptic inputs are sufficient to account for the differences in spike width between LR-evoked and antidromic spikes (Fig. 10C). This reduction in antidromic spike width was accompanied by a concomitant reduction in the amplitude of the FlTi EPSP (Fig. 10D), suggesting that the reduction in FETi spike width evoked by synaptic inputs is sufficient to generate the enhanced level of depression at the FETi-FlTi synapse during LR-evoked spike trains.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Spike width reduction during LR-evoked spike bursts may contribute to depression at the FETi-FlTi synapse. A, Changes in spike width correlate with changes in the depolarization of FETi immediately before the onset of the spike. B, Changes in the FETi spike height induced by injecting hyperpolarizing current produced a concomitant increase in the amplitude of the FlTi PSP. C, Three successive spikes during constant frequency antidromic spike trains synaptic inputs to FETi reduce the width of the antidromic spike. The second spike, which coincided with a synaptic input, has a reduced spike width. This trace is shifted (gray trace) to enable comparison with the other antidromic spikes. D, Antidromic spikes coinciding with synaptic inputs evoke smaller amplitude FlTi EPSPs.

How do synaptic inputs affect the efficacy of transmission at the FETi-FlTi synapse?

The presence of depolarizing synaptic inputs during LR-evoked spike trains correlates strongly with a 40% reduction in the FETi spike width. During constant frequency antidromic spike trains, there is only a 5% reduction in spike width under physiological conditions, suggesting that reduction in spike width may be responsible for the increased level of depression during LR-evoked spike trains. This is supported by two observations: (1) antidromic spike width is reduced in the presence of depolarizing synaptic inputs; and (2) FlTi EPSP amplitude is increased when FETi spike width is increased by hyperpolarizing current injection. Changes in spike width may affect the presynaptic calcium entry, thereby affecting the amount of neurotransmitter released at a synapse; small increases in spike width (∼20%) at both cerebellar synapses and in the squid produce large (twofold) increases in the postsynaptic current (Augustine, 1990; Jackson et al., 1991; Sabatini and Regehr, 1997).

Several possible mechanisms could explain the spike width reduction induced by synaptic inputs in FETi. Electron microscopy studies of the FETi-FlTi synapse have revealed that the input and output synapses in FETi are in close proximity in a region of neuropil that lacks voltage-gated Na⁺ channels (Gwilliam and Burrows, 1980; Burrows et al., 1989). Additionally, the spike initiation zone occurs in the region of the primary neurite that bears arborizations containing input synapses, potentially enabling them to modulate spike properties (Gwilliam and Burrows, 1980). This suggests that synaptic inputs reduce the FETi membrane resistance, reducing the membrane time constant throughout the neuropil and in the primary neurite close to the spike initiation zone. Consequently, the width of a spike propagating into this low resistance membrane would be reduced before reaching the output synapses activating voltage-gated Ca²⁺ channels for shorter periods of time, leading to smaller FlTi EPSPs. In addition, increasing numbers of depolarizing inputs evoked by LR stimulation would increase the FETi spike frequency and further reduce the spike width. This suggests that at the FETi-FlTi synapse presynaptic inputs reduce the spike waveform, thereby contributing to the dynamics of short-term depression.

Presynaptic inputs may affect synaptic output via changes in the spike waveform in both vertebrate and invertebrate neural networks (Nusbaum et al., 1997; Rudomin et al., 1998). For example, at afferent terminals in arthropods, shunting is primarily responsible for inducing changes in the presynaptic waveform that correlate with reduced postsynaptic responses (Cattaert et al., 1992; Burrows and Matheson, 1994; Cattaert and El Manira, 1999). Additionally, neurons in which graded membrane potentials reach output synapses, the presynaptic membrane potential may affect the PSP amplitude. For example, at inhibitory interneuron synapses (heart interneuron-heart excitor motoneuron) in the leech, presynaptic membrane potential affects the amplitude of PSPs (Nicholls and Wallace, 1978). The interaction of presynaptic inputs with the dynamics of short-term depression provides an additional level of plasticity for information processing in neural networks. Potentially presynaptic inputs could tune the dynamics of short-term depression or facilitation by modulating the shape of the spike waveform, enabling information to be transmitted not only by spike rate or interspike interval but also by spike width or spike height.

Mechanisms underlying depression and recovery at the FETi-FlTi synapse

Depletion of the releasable transmitter store explains many of the features of short-term depression at several synapses (Liley and North, 1953; Elmqvist and Quastel, 1965; Rosenthal, 1969; Kusano and Landau, 1975), however, at other synapses, changes in the presynaptic spike waveform (Dunlap and Fischbach, 1978; Klein and Kandel, 1980; Augustine, 1990; Sabatini and Regehr, 1997), inactivation of the presynaptic calcium current (Forsythe et al., 1998; Patil et al., 1998), or the inhibitory action of neuromodulators on presynaptic calcium channels (Yawo and Chuhma, 1993; Scanziani et al., 1997) may all contribute to the observed depression.

At the FETi-FlTi synapse, the depletion model accounts for several features of depression during antidromic spike trains, although it cannot account for the inability to produce depression below 20% despite a fivefold increase in spike frequency (Parker, 1995). In contrast, there is no such limit on the steady-state level of depression during LR-evoked spike trains. However, depression during LR-evoked spike trains is unlikely to fit the depletion model because during depression there are changes in the presynaptic waveform. The reduction in both FETi spike width and height during LR-evoked spike trains produces a concomitant reduction in the FlTi EPSP amplitude, possibly because of a reduction in the levels of presynaptic calcium (Augustine, 1990; Jackson et al., 1991; Sabatini and Regehr, 1997).

Depression at the FETi-FlTi synapse occurs during antidromic spike trains in the absence of presynaptic waveform changes, suggesting that during LR-evoked spike trains several mechanisms combine to produce the observed depression.

Behavioral significance of depression at the FETi-FlTi synapse

The central synaptic connection between FETi and the FlTi motoneurons is thought to be a specialization for the generation of the jumping and kicking motor programs; FETi is usually activated only during kicking or jumping at the same time as the antagonistic FlTi motoneurons are active (Burrows, 1996), whereas during walking antagonistic tibial muscles are not coactivated and spikes are rarely observed in FETi (Burns and Usherwood, 1979). Although the FETi-FlTi synapse was originally postulated to increase FlTi muscle tension during the co-contraction phase of the kick (Heitler and Burrows, 1977), the rapid depression of this synaptic connection during antidromic spike trains suggested that inputs from this synapse could not be solely responsible for maintaining FlTi muscle tension (Burrows et al., 1989; Parker, 1995). Yet, the residual FlTi EPSP amplitude of 20% during antidromic spike trains of up to 100 Hz suggests that even after maximum depression inputs from FETi would still contribute to the FlTi spike frequency. This clearly is not the case during LR-evoked spike trains, during which the FlTi EPSP depresses rapidly to just 5% of its initial amplitude. The increase in both the rate and magnitude of the depression at the FETi-FlTi synapse are reflected in the model. The U parameter and the limiting frequency determine whether the FlTi postsynaptic response reflects the rate of spiking or the onset of spiking in FETi. U is the fraction of neurotransmitter used after a spike, hence, high values of U will cause the synapse to depress rapidly to its steady-state amplitude (Tsodyks and Markram, 1997). The limiting frequency sets the upper limit to the frequency range over which the synapse can transmit information about the presynaptic spike rate; above the limiting frequency, synapses mainly transmit temporal information. The high U parameter at the FETi-FlTi synapse was consistent between synaptic connections in different individuals (0.52-0.89). Coupled with the high spike frequencies during LR-evoked spike trains (up to 70 Hz) and the low-limiting frequency (∼5 Hz), this suggests that the FETi-FlTi synapse may be crucial for signaling the onset of FETi motoneuron activity at the start of co-contraction rather than contributing to the spike rate of the FlTi motoneurons during co-contraction.

Depression of the FETi-FlTi synapse may allow FETi activity to contribute to FlTi activity at the onset of co-contraction but enable uncoupling of their activity later in co-contraction, allowing other inputs to determine the pattern of FlTi activity. It may also enable FETi to continue to spike during the period of FlTi motoneuron inhibition that precedes tibial release without exciting the FlTi motoneurons, thus preventing contraction of the FlTi muscle that would reduce the velocity of the tibia once released.