Estimation of postsynaptic potentials in rat hypoglossal motoneurones: insights for human work

Abstract

Classical techniques for estimating postsynaptic potentials in motoneurones include spike-triggered averages of rectified surface and multiunit electromyographic recordings (SEMG and MU-EMG), as well as the compilation of peristimulus time histograms (PSTH) based on the discharge of single motor units (SMU). These techniques rely on the probability of spike occurrence in relation to the stimulus and can be contaminated by count- and synchronization-related errors, arising from post-spike refractoriness and the discharge statistics of motoneurones. On the other hand, since these probability-based techniques are easy to use and require only inexpensive equipment, it is very likely that they will continue to be used in clinical and laboratory settings for the foreseeable future. One aim of the present study was to develop a modification of these probability-based analyses in order to provide a better estimate of the initial phase of postsynaptic potentials. An additional aim was to combine probability-based analyses with frequency-based analyses to provide a more reliable estimate of later phases of postsynaptic potentials. To achieve these aims, we have injected simple as well as complex current transients into regularly discharging hypoglossal motoneurones recorded in vitro from rat brainstem slices. We examined the discharge output of these cells using both probability- and frequency-based analyses to identify which of the two represented the profile of the postsynaptic potential more closely. This protocol was designed to obtain PSTHs of the responses of single motor units to repeated application of the same afferent input. We have also simulated multiunit responses to afferent input by replacing the times of spike occurrence in individual trials with a representation of either an intramuscular or surface-recording single motor unit waveform and summing many of these trials to obtain either a simulated SEMG or MU-EMG. We found that in a regularly discharging motoneurone, the rising phase of an EPSP moves the occurrence of spikes forward and hence induces a substantial peak in all probability-based records. This peak is followed immediately by a period of reduced activity ('silent period') due to the phase advancement of spikes that were to occur at this period. Similarly, the falling phase of an IPSP delays spikes so that they occur during the rising phase of the IPSP. During the delay, the probability-based analyses display gaps and during the occurrence of the delayed spikes they generate peaks. We found that all the probability-based analyses (SEMG, MU-EMG and PSTH) can be made useful for illustrating the underlying initial PSP by a special use of the cumulative sum (CUSUM) calculation. We have illustrated that, in most cases, the CUSUM of probability-based analyses can overcome the delay- or advance-related (i.e. the count-related) errors of the classical methods associated with the first PSP only. The probability-based records also induce secondary and tertiary peaks and troughs due to synchronization of the spikes in relation to the stimulus (i.e. the synchronization-related errors) by the first PSP to occur at fixed times from the stimulus. Special CUSUM analyses cannot overcome these synchronization-related errors. Frequency-based analysis (PSFreq) of individual and summed trials gave comparable and often better indications of the underlying PSPs than the probability-based analyses. When used in combination, these analyses compliment each other so that a more accurate estimation of the underlying PSP is possible. Since the correct identification of the connections in the central nervous system is of utmost importance in order to understand the operation of the system, we suggest that as well as the using the special CUSUM approach on probability-based records, researchers should seriously consider the use of frequency-based analyses in their indirect estimation of stimulus-induced compound synaptic potentials in human motoneurones.

Figure 2. Simulation of surface and intramuscular EMG responses

A, simulation of a SEMG recording. One-second sample of the times of spike occurrence (indicated by delta functions, top trace), the contribution of a single spike train to the SEMG recording (middle trace) and the simulated SEMG record, calculated by summing the contribution of 9 spike trains and rectifying the result (bottom trace). The contribution of an individual spike train is calculated by replacing the delta functions in the top trace with a typical single motor unit waveform as recorded with surface electrodes (inset). B, effects of a PSP on the simulated SEMG record. Top trace, PSP-triggered average of the simulated SEMG record. Middle trace, CUSUM of the simulated SEMG record. Bottom trace, PSP. C, simulation of multiunit (intramuscular) EMG recording. Same arrangement of traces as in A, except that the bottom trace represents the contributions of only three motor unit spike trains (firing at high, medium and low background rates), and the final result is not rectified, but is instead high-pass filtered at 1000 Hz. Times of threshold crossing are calculated from negative-going deflections that exceed a fixed threshold (shown by the dashed line in the bottom trace). Threshold-crossings were followed by a dead time of either 2 or 10 ms, so that the maximum frequency of threshold-crossings was either 500 or 100 Hz. D, PSFreqs and their CUSUMs calculated from three simulated multiunit trials using a frequency cut-off of either 500 (upper traces) or 100 Hz (lower traces). In B and D, the SEMG and the PSFreq are made up of 9 trials.

Figure 1. Analysis of effects of simulated postsynaptic potentials (PSPs) on motoneurone firing probability and firing rate

A, top traces: peristimulus time histogram (PSTH, second trace) and its CUSUM (top trace) calculated from the response of a tonically discharging hypoglossal motoneurone to repeated presentation of the PSP shown in the bottom trace. The CUSUM has been normalized by the number of applied PSPs. The fourth trace is the peristimulus frequencygram (PSFreq), which plots the instantaneous firing rate of the motoneurones as a function of time. The third trace from the top is the PSFreq-CUSUM, which has also been normalized by the number of applied PSPs. Horizontal segments of the PSFreq-CUSUM corresponding to gaps in the discharge record are indicated by dashed lines in this and subsequent figures. The lines superimposed on the CUSUMs in this and all subsequent figures represent a zero CUSUM value (dashed horizontal lines), along with the maximum deviations in the CUSUM at negative lags (Error box, continuous horizontal lines above and below the zero CUSUM line; see text for further details). The vertical dashed lines delineate the rising phases of the PSTH-CUSUM (labelled 1 and 2). B, PSTH-CUSUMs (top three traces) and PSFreq-CUSUMs (next three traces) in response to the PSP shown in the bottom trace for three different background-firing rates. All traces except the PSPs are calculated based on three trials each containing 66 PSPs (total of 198).

Figure 3. PSTH and PSFreq-CUSUMs calculated from motoneurone responses to various PSPs

PSTH- (top traces) and PSFreq-CUSUMs (middle traces) calculated from responses to PSPs (bottom traces) for the highest motoneurone background discharge rate. A and B, responses to single EPSPs and IPSPs. C and D, responses to double EPSPs and IPSPs. Except for the PSPs, each trace is made up from three trials each containing 66 PSPs (total of 198). The vertical dashed lines delineate the rising phases of the PSTH-CUSUM (labelled 1, 2 and 3). Error boxes are indicated with continuous horizontal lines in this and other figures (Methods).

Figure 4. PSTH and PSFreq-CUSUMs calculated from responses to various mixed PSPs

PSTH- (top traces) and PSFreq-CUSUMs (middle traces) calculated from responses to mixed PSPs (bottom traces) for the highest motoneurone background discharge rate. A, response to an EPSP followed by equal-amplitude IPSP. B, response to an IPSP followed by equal-amplitude EPSP. C, response to an EPSP followed by an IPSP of half its amplitude. D, response to an IPSP followed by an EPSP of half its amplitude. Except for the PSPs, each trace is made up from three trials each containing 66 PSPs (total of 198). The vertical dashed lines delineate the rising phases of the PSTH-CUSUMs (labelled 1, 2 and 3).

Figure 5. Effects of PSPs on simulated multiunit records

The traces in each panel in this figure, from top to bottom, are: a PSP-triggered average of the simulated SEMG record, the CUSUM of the average SEMG record, the CUSUM of a MU-PSFreq with a 500 Hz cut-off, the CUSUM of a MU-PSFreq with a 100 Hz cut-off, and the PSP. A and B, responses to a single EPSP and IPSP, respectively. C and D, responses to double EPSPs and IPSPs, respectively. The SEMGs and the PSFreqs are made up of 9 trials. Vertical dashed lines delineate the rising phases of the SEMG-CUSUMs. Only 1 in A and B; 1 and 2 in C and D.