We have previously suggested that the discharge characteristics of some neurons in the pontomedullary reticular formation (PMRF) are contingent on the simultaneous requirement for activity in both ipsilateral flexor muscles and contralateral extensors. To test this hypothesis we trained cats to stand on four force platforms and to perform a task in which they were required to reach forward with one forelimb or the other and depress a lever. As such the task required the cat to make a flexion movement followed by an extension in the reaching limb while maintaining postural support by increasing extensor muscle tonus in the supporting limbs. We recorded the activity of 131 neurons from the PMRF of three cats during left, ipsilateral reach. Of these, 86/131 (66%) showed a change in discharge frequency prior to the onset of activity in one of the prime flexor muscles and 43/86 (50%) showed a bimodal pattern of discharge in which activity decreased during the lever press. Among the remaining cells, 28/86 (33%) showed maintained activity throughout the reach and the lever press. Most cells showed a broadly similar pattern of discharge during reaches with the right, contralateral limb. We suggest these results support the view that a population of neurons within the PMRF contributes to the control of movement in one forelimb and the control of posture in the other forelimb as a coordinated unit. Another population of neurons contributes to the control of postural support independently of the nature of the activity in the reaching limb.
The pontomedullary reticular formation (PMRF) makes an important contribution to the regulation of posture. Lesion to this structure or to the reticulospinal pathways, in both cats and primates, leads to an inability to maintain an upright posture during normal everyday activities (Kuypers 1963; Lawrence and Kuypers 1968a,b) including locomotion (Afelt 1974; Bem et al. 1995; Brustein and Rossignol 1998; Górska et al. 1990, 1993, 1995). The deficits that are observed following such lesions are widespread and normally include a loss of control over all four limbs as well as of axial musculature.
These deficits are compatible with the known anatomy of the PMRF. Individual axons arising from reticulospinal neurons (RSNs) within the PMRF may innervate both cervical and lumbar enlargements (Matsuyama et al. 1993, 1997; Peterson et al. 1975) and may also cross the midline to innervate both sides (Matsuyama et al. 1993). Moreover, it is also known that RSNs innervate commissural interneurons that provide an indirect pathway to affect contralateral activity (Jankowska et al. 2003; Matsuyama et al. 1997, 2004, 2006). Given this widespread branching pattern, it is not surprising that both microstimulation and spike-triggered averaging (STA) studies show that the PMRF, in both cats (Drew 1991; Drew and Rossignol 1990a,b; Schepens and Drew 2006) and primates (Davidson and Buford 2004, 2006; Davidson et al. 2007), may influence flexor and extensor muscles on both sides of the body as well as head movements and neck musculature (Cowie and Robinson 1994; Drew and Rossignol 1990a,b; Isa and Sasaki 1988; Quessy and Freedman 2004). However, during locomotion, the nonspecific pattern of activation of the limb muscles is reorganized (Drew 1991; Drew and Rossignol 1984; Orlovsky 1972; Perreault et al. 1994) so that stimulation during ipsilateral swing activates primarily ipsilateral flexor muscles and contralateral extensor muscles. In contrast, stimulation during contralateral swing activates contralateral flexor muscles with more variable effects on ipsilateral extensor muscles. These results show that the PMRF may influence both flexor and extensor muscles in a phase-dependent manner.
Single-unit recordings studies in intact cats have shown that reticular neurons, including RSNs, discharge phasically during both locomotion (Drew et al. 1986; Matsuyama and Drew 2000; Prentice and Drew 2001) and reaching movements, in both intact cats (Schepens and Drew 2004, 2006) and primates (Buford and Davidson 2004; Gibson et al. 1998; Stuphorn et al. 1999; Werner et al. 1997). In our own experiments, in which the cats are standing on all four limbs, these reaching movements require that the cats produce anticipatory postural adjustments in both the reaching and the supporting limbs prior to movement onset, and that they produce substantial postural adjustments in the supporting limb during the movement (Schepens and Drew 2003). Our results showed that several populations of neurons were active during the task. A very few discharged only during the preparatory anticipatory postural adjustment (pAPA) preceding the movement but most discharged during both the pAPA and during the reaching movement (Schepens and Drew 2004, 2006). Among this latter population, a proportion discharged only during the dynamic reach but a majority discharged throughout the time that the limb was removed from the support surface, during which the cat generated anticipatory postural adjustments that accompanied the movement (aAPAs). A multiple-regression analysis suggested that the discharge activity in these latter cells correlated better with the combined activity in the flexor muscles of the reaching limb together with the extensor muscle activity in the supporting forelimb than with either independently. This result suggests that many individual cells in the PMRF may signal both movement and posture as a coordinated activity.
To test this hypothesis directly we have modified our original reaching task (Schepens and Drew 2003) to one in which cats are required to reach forward and to depress a lever. In this lever-press task, the cat has to generate flexor activity in the reaching limb as it is moved toward the lever and then an active extensor torque to depress the lever, before again activating flexor muscles to replace the limb on the supporting surface. However, throughout this period, the contralateral limbs have to sustain an increased level of extensor tonus to maintain postural support. If our hypothesis is correct, we would expect to observe a population of neurons in which discharge activity is increased during the initial part of the reach when flexor muscle activity in the reaching limb is coordinated with extensor muscle activity in the supporting limbs. We would then expect a decrease in activity during the lever press when the flexor muscle activity in the reaching limb is suppressed and the coupling between flexor activity on one side and extensor muscle activity on the other side is lost. Activity would be expected to again increase at the end of the lever press. In contrast, if the primary function of the cell is to maintain postural support in the standing limb, then we would expect no decrease in activity during the lever press. Such cells would discharge either with a pure tonic pattern of activity throughout the entire sequence of activity or with a phasic/tonic pattern of activity, as observed in our previous studies (Schepens and Drew 2004, 2006). The results suggest that both types of neurons exist; some signal a coordinated pattern of activity, whereas others signal movement and posture independently.
Three cats (RS24–RS26, weights 4.8–7.4 kg), different from those used in our previous studies (Schepens and Drew 2003, 2004, 2006), were trained to perform a task in which they were required to reach forward to depress a lever (Fig. 1). The cats were trained to stand quietly with each paw on a force platform in response to an audible tone (0.5 s) that initiated a trial. After a fixed delay of 1.5 s the cat received a second cue of variable duration (0.5–1.5 s). The frequency of this instruction cue informed the cat whether the subsequent movement should be made with the left forelimb (frequency = 400 Hz) or the right forelimb (frequency = 4 kHz). Following the delay, the tone ceased and a shutter was raised (Fig. 1B) to give the cat access to the lever (Go signal, Fig. 1A). The cat received a food reward for depressing the lever within 2 s. The torque required to depress the lever was normally maintained at 2 Nm because this relatively low torque level ensured that the cats would work for extended periods of time. Increasing the torque magnitude resulted in a marked decrease in the number of trials that the cats were willing to perform. The lever was located between 14 and 19 cm from the leading edge of the forelimb force platforms in all three cats and the apparatus was designed with two yoked levers so that the reaches with either limb were confined mostly to the sagittal plane. The lever was adjusted to be between 5 and 9 cm higher than the force platforms (see Fig. 2A).
Once the cats were capable of performing the task with either limb with a high level of success (>80%), they were instrumented for recording. At this level of performance cats generally switched from one limb to the other on the first trial.
The methods used for preparing and executing the surgery were detailed previously (Schepens and Drew 2003, 2004). In brief, cats were prepared for surgery under general anesthesia with a mixture of 2–3% isoflurane and oxygen. They were cannulated for the application of intravenous fluid and drug administration and body temperature and heart rate were continuously monitored. Methylprednisolone (Solu-Medrol, 30 mg/kg), antibiotics (Penicillin G, 40,000 IU/kg), and buprenorphine (5 μg /kg) were administered at the onset of the surgery.
Pairs of insulated stainless steel wires, attached to a 51-pin connector, were passed subcutaneously and inserted into selected flexor and extensor muscles of all four limbs. The connector was attached to the cranium with dental acrylic. A craniotomy was made in the occipital bone, overlying the cerebellum, and a stainless steel base plate (internal diameter, 8 × 6 mm) was cemented over the craniotomy to provide access to the PMRF. Three microwires (50-μm diameter) were inserted into the spinal cord at the L2 level to allow for antidromic activation of neurons in the PMRF with axons projecting to the lumbar spinal cord (RSNs). These wires were also led to a connector on the cranium.
Following completion of the surgery, the animals were administered supplementary doses of antibiotics and buprenorphine and placed in an incubator to recover. Antibiotics were administered for a minimum of 10 days following the surgery.
Following recovery from the surgery, the cats were brought into the laboratory three to five times/wk for the recording of unit activity from the PMRF. A microelectrode was lowered through the cerebellum into the brain stem. Once the electrode passed the IVth ventricle, stimulation was applied at a rate of 0.7/s to the L2 electrode that preliminary experiments had shown to be most effective in activating RSNs. The electrode was then advanced slowly until a single neuron was identified either by its spontaneous activity or by antidromic stimulation of the L2 electrode. If a single neuron was isolated but not identified by the initial stimulation, it was tested by stimulating through the other two spinal microwires. Neurons that discharged at constant latency and collided with spontaneous action potentials (Lipski 1981) were classified as RSNs; others were classified as unidentified cells.
Once antidromic testing was completed, the activity of the neuron was recorded while the cat walked on a treadmill and stepped over an obstacle attached to the moving belt. The results from these experiments are not included in this report. The cat was then transferred to the reaching apparatus adjacent to the treadmill belt and the activity of the cell was recorded during reaches with the left, ipsilateral to the recording site, and right limbs. In general, 5 reaches were recorded with the left limb, 10 with the right limb and another 5 with the left limb. Supplementary reaches were occasionally recorded.
We recorded electromyographic (EMG) activity from between 20 and 24 muscles in each cat. These signals were digitized on-line at 1 kHz and recorded to disc together with the 24 signals from the four force platforms (three force and three moment signals from each). Neuronal activity was digitized at a frequency of 100 kHz to allow off-line discrimination of single units with custom software. Video recordings were taken of the cats during each session and were visually examined prior to analysis of each neuron. Reflective points were placed over bony landmarks on the left forelimb (Fig. 1B) to allow for analysis of limb kinematics in selected trials.
Neuronal and EMG data were analyzed using methods similar to our previous experiments (Schepens and Drew 2003, 2004). In brief, all individual trials were examined to eliminate any trial in which the vertical force in the reaching limb varied by more than 10% in the 1-s period preceding the Go signal. Any DC offset in the force traces was removed by using calibration files recorded intermittently throughout the recording session. The force and EMG traces were then filtered at 25 Hz by using a dual-pass, second-order digital Butterworth filter. We used an interactive program to measure the onset and offset of selected force and EMG events as indicated in Schepens and Drew (2003). The measured events always included the onset of activity in the cleidobrachialis (ClB) or brachialis (Br) muscles, used to synchronize averaged traces. As described previously (Schepens and Drew 2003), these two muscles, which are activated simultaneously, are involved in the initiation of the reaching movement and become active just prior to the limb being raised from the support surface. In addition, we always measured the initial increase in vertical force (FV) in the reaching limb (when possible) as well as the time that FV decreased to zero (referred to as lift). Other selected events that we measured are indicated in results.
In our initial analysis, we classified cells into a number of different categories using the same criteria as in our previous publications (Schepens and Drew 2004, 2006). Neuronal discharge was synchronized to the onset of activity in the ClB or Br, averaged, and the resulting histograms were digitally filtered at 25 Hz with the same filter as used for the EMG and force traces (see earlier text). We calculated the average discharge frequency, together with its SD, from a 500-ms period just prior to the instruction period (Fig. 1A); we refer to this as the control discharge. Neurons were classified as showing a significant change in activity if they deviated from the control level (±2SD).
In this study we place an emphasis on those neurons that showed a significant increase in activity that preceded the onset of activity in the Br or ClB. For this group of neurons we identified cells as showing a phasic, tonic, or phasic/tonic pattern of activity (see Fig. 5, A–C), using criteria similar to those of Schepens and Drew (2004). We defined discharge during the dynamic period as being the maximal discharge obtained between the Go signal and the time that the cat made contact with the lever; this is similar to our previous definition that was based on the time that the paw entered the feeding tube. We defined the static phase of the movement as extending from 1,000 to 1,500 ms after the Go signal. This corresponded to a period in which the cats had actively depressed the lever prior to returning the paw to the force platform (see Fig. 2). All measurements of discharge frequency were taken from the averaged and filtered traces.
Phasic cells were defined as those in which there was a significant increase during the dynamic phase of the movement and in which the discharge level during the static period was <150% of that observed during the control period or, in cells with low discharge rates, exceeded the control level by <10 spikes/s. In contrast, cells showing a tonic component all showed a discharge frequency during the static period that was >150% of control. Phasic/tonic cells were differentiated from tonic cells in having a discharge frequency during the dynamic period that was >125% of that during the static period. In our initial analysis and classification of the cells, however, we frequently observed another pattern of discharge that was only rarely observed in our previous experiments (Schepens and Drew 2004). These cells showed a bimodal pattern of activity (see Fig. 5D) in which discharge initially increased, then showed a substantial decrease in activity and then showed a second period of increased activity. For these cells we measured the maximal discharge frequency during the dynamic and static periods from the averaged traces as reported earlier and then measured the minimum discharge frequency between these two peaks. We then defined a bimodal cell as one in which the two maxima both exceeded the minima by >150% (other examples are shown in Fig. 6). In our analysis, classification as a bimodal cell took priority over any other classification. We also reanalyzed our previous database (Schepens and Drew 2004, 2006) using this same criterion (see results). In many of the cells recorded in this previous task, the second peak was not as well defined as in this new task (see e.g., Fig. 13). We thus defined the value of the second peak as the maximal discharge occurring following the minima and ≤1,500 ms following the Go signal.
We defined the initial period of discharge activity as being related to either the Go signal (Go-related) or the onset of the ClB or Br (movement-related) based on linear regression analyses using a modified version of our previous method (Schepens and Drew 2003, 2004, 2006). We measured the onset of the cell discharge on a trial-by-trial basis and we plotted this against the onset of activity in the ClB or Br. Cells that showed a significant (P < 0.05) linear relationship were classified as movement-related. We equally plotted the difference between the onset of the flexor muscle activity and the onset of cell activity (lead time) as a function of ClB or Br onset. In this case the closer the value of the correlation coefficient (r) to 1.0, the more likely that the onset of the cell discharge is time-locked to the onset of the stimulus (Go-related) (Chapman et al. 1986; Vicario et al. 1983). Effectively, if the cell discharges at a relatively fixed latency, the lead time will vary directly with the latency of the ClB or Br onset. As a correlate of this, if cell discharge is Go-related, the mean latency of the discharge will be short and the variance of the discharge will be relatively small. To obtain an objective measure of the proportion of cells with Go-related activity, we first calculated the mean variance of that population of cells with a coefficient of determination (R2) >0.81 (r = 0.9) for the relationship between the lead time and the ClB onset during left, ipsilateral reach. We then calculated the interval of confidence (95%) of the mean variance and classified any (nonmovement-related) cells within the population with variance less than this level as Go-related. This method has the advantage over our previous method of including only cells with short latencies and low variance as Go-related (see Fig. 9). Moreover, it reduces the confound produced by the fact that there is an obligatory covariance between the values of the lead time and the latency of the ClB onset.
In selected penetrations and at the end of the experimental period, marking lesions (10–25 μA, 10 s, cathodal DC current) were made. The cats were deeply anesthetized and perfused per aorta. The brain was removed, sectioned, and stained with cresyl violet. Recording sites were interpolated and plotted on standard sagittal sections of the brain stem (Berman 1968).
The task that we used in this study required that the cats lift their limb from a platform, reach forward to a lever, press it, and return the limb to the platform. We refer to this as a lever-press task to differentiate it from our previous reaching task (Schepens and Drew 2003). As shown in Fig. 2A, the task requirements are reflected in the kinematics that showed an initial flexion of all four joints of the forelimb as the limb was raised from the support surface (lift) and advanced toward the lever. Subsequently, there was an extension at all four joints as the limb touched the lever (lever touch) and depressed it (lever max). The cat remained pressing on the lever for a variable period of time (normally until the food reward was delivered) before lifting the paw (lever end) and replacing it on the force platform. In all general aspects, these kinematics (Fig. 2C) were similar to those observed in our previous reaching task (Schepens and Drew 2003).
The EMG and resultant forces required to produce the movements were also similar in both tasks as shown in Fig. 2, B and D. The only substantial difference was that the initial change in vertical force observed in the reaching forelimb (ΔFLV) at the onset of the pAPA (Fig. 2D) was less frequently observed during the lever-press task (Fig. 2B). Nonetheless, there was always a clear increase in the activity of the lateral and long heads of triceps (TriL and Tri, respectively) prior to ClB onset in the lever-press task, as for the original reach task. There was also a clear increase in the mediolateral component of the forelimb force (FLML) in both tasks (Fig. 2, B and D). This latter increase in force shows that the cat was pushing outward with the reaching forelimb at this time. Together, these data suggest that cats produced a similar preparatory anticipatory postural adjustment (pAPA) during both tasks, but that it is less prominent during the lever-press task. Because of the lack of a clear change in ΔFLV in some trials, we define the pAPAs that precede movement in this study as extending from the onset of the period of activity in the lTriL to the onset of activity in the lClB (shaded regions in Fig. 2, B and D). In addition, the activity in the ClB was frequently depressed during the lever press, leading to a division into two clear bursts of activity, one during the reach and a second during the retraction of the arm following the lever press. This depression of the ClB activity during the lever press can be seen in Fig. 2B but is more evident in the traces of Fig. 5. A similar pattern was observed in the teres major (not illustrated).
The general pattern of the kinematics and of the EMG activity was similar in the other two cats, although the relative magnitude of the activity in the lTriL and the lTri during the lever press varied slightly. For example, cat RS25 always exhibited robust increases in activity in the lTriL and the lTri during the lever press, whereas in cat RS24 they were strong only in the lTri and in cat RS26 they were relatively small in both extensor muscles. In part, this difference reflects the strategy adopted. Cat RS25 always pressed very enthusiastically on the lever, producing much more force than necessary to depress it, whereas cat RS26 pressed more delicately. As a consequence, there was appreciably more unloading of the weight over the supporting limb in cat RS25 than that in the other two cats.
The latency of the movements was similar in all three cats, with the mean latency to ClB onset ranging from 266 to 472 ms for the left forelimb and from 289 to 412 ms for the right forelimb. The lever press was initiated at means values of 578–804 ms for the left forelimb and of 595–704 ms for the right forelimb.
As in our previous reaching task (Schepens and Drew 2003), there was a strong relationship between the onset of activity in the lClB and the time of paw lift from the platform (Fig. 3A), as is to be expected if the former is responsible for the latter. There was also a linear relationship between the onset of the second period of activity in the lTriL, occurring prior to the lever press, and the depression of the lever (Fig. 3B). Again this is expected because we presume that this burst of activity in the lTriL is a prime contributor to the lever press. In our previous work (Schepens and Drew 2003), we showed that the initial period of activity in the iTriL was always time locked to the Go signal. This relationship was less clear in the current task (Fig. 3C), perhaps reflecting the weaker change in ΔFlV in this lever-press task (see Fig. 2B). Nonetheless, the initial decrease in activity in the supporting forelimb (rTriL; Fig. 3D) was tightly time locked to the Go signal, suggesting a similar temporal dissociation of the pAPA and the movement as in our original task (Schepens and Drew 2003). Similar relationships to those illustrated in Fig. 3 were also observed during movements of the right limb and in the other two cats.
We recorded the discharge activity of 131 neurons from three cats (see Table 1) during the performance of at least four reaches with the left limb that fulfilled the criteria for inclusion. Most neurons (127/131) were also recorded during at least four reaches with the right limb. Neurons were recorded throughout the PMRF ranging from P11.4 to P2.8. As illustrated in Fig. 4A, most of the cells were located in the nucleus reticularis gigantocellularis (NRGc) or in the nucleus reticularis pontis caudalis (NRPc). Neurons were recorded from the NRGc in all three cats and from the NRPc in two cats (RS24 and RS26). Although there was overlap in the recording sites in all three cats, recordings in RS24 were made relatively more rostrally than in the other two cats. In cats RS25 and RS26, 15/21 (71%) and 25/47 (53%) of the neurons, respectively, were positively identified as RSNs. All neurons in cat RS24 were classified as unidentified because of a malfunction with the stimulating electrodes in the lumbar spinal cord. The conduction velocity of the RSNs in cats RS25 and RS26 ranged from 34.4 to 136.5 m s−1 (mean = 105.1 ± 24.6 m s−1) (Fig. 4B).
The number of neurons with the different types of discharge activity that we identified in methods is presented in Table 1 and examples of these discharge patterns are shown in Fig. 5. Among those cells showing increased activity prior to the onset of activity in the ClB during the left (ipsilateral) reach, the same three major categories of neurons that we previously identified and characterized—phasic, tonic, and phasic/tonic cells—could be identified (Table 1). These phasic, tonic, and phasic/tonic cells made up 50% (43/86) of the population of neurons, showing an increased discharge prior to the onset of activity in the iClB. Eleven (11/17, 65%) of these cells in RS25 and RS26 were identified as RSNs. An initial analysis of these cells suggested that their characteristics were identical in all major features, including discharge frequency and temporal relationship to the onset and offset of activity in the ClB, to those previously documented (Schepens and Drew 2004, 2006). As in those previous studies, we suggest that many individual cells have multiple functions, including a contribution to the pAPA preceding movement, a contribution to the flexor muscle activity during the reach, and a contribution to the extensor muscles of the contralateral limb responsible for postural support. The characteristics of these phasic, tonic, and phasic/tonic cells will not be further detailed.
The other 50% of the cells discharging prior to ClB onset (Tables 1 and 2) showed a bimodal pattern of activity (Fig. 5D). These cells increased their activity during the initial reach, were suppressed during the lever press and became active again when the paw was transported to the support surface. As such, they are compatible with our hypothesis that the discharge activity in a population of cells would be contingent on a coordinated pattern of activity in ipsilateral flexor muscles and contralateral extensor muscles. We detail the characteristics of the activity of this new population of cells in the text that follows.
Note that although most cells showed a generally similar and nonreciprocal pattern of activity during the right reach, some cells did change their classification during the right reach. Most frequently, as detailed in Schepens and Drew (2006), some cells with a tonic component were classified as tonic during one condition and phasic/tonic during the other (and vice versa). Similarly, in this study we found cells that were classified as bimodal in one condition and either tonic or phasic/tonic in the other (and vice versa). This serves to emphasize our previous statement (Schepens and Drew 2004) that the pattern of activity in these cells forms a continuum. In the text that follows, unless otherwise specified, we will refer to the class of neurons according to its discharge pattern during the left, ipsilateral reach.
General characteristics. As detailed in methods, bimodal cells were defined by two peaks of discharge that were both >150% of the intervening minima. Forty-three (43/86, 50%, Tables 1 and 2) of the cells that showed increased activity prior to the onset of the movement were classified as bimodal. In cats RS25 and RS26, 14/23 (61%) of these bimodal cells were identified as RSNs. Figure 6 shows three examples of bimodal cells as determined on the basis of their discharge patterns during the left, ipsilateral reach. These cells included some having a substantial suppression of discharge activity during the lever press (Fig. 6, A and B), whereas others had a more modest change (Fig. 6C). There was no evident difference in the characteristics of these bimodal cells recorded in each of the three cats. As indicated earlier, the discharge of these cells is compatible with our initial hypothesis that the activity patterns of some cells in the PMRF are dependent on a coordinated pattern of activity involving flexor muscle activity in one limb and extensor muscle activity in the supporting limb or limbs. Although all cells were classified on the basis of their discharge activity during the left, ipsilateral reach, a majority, 28/43 (65%) showed a similar bimodal discharge activity during the right reach (Fig. 6, right column). Four additional neurons showed a bimodal discharge during the right reach but not during the left reach.
Both the relative timing of the different components of the discharge activity and the magnitude of the changes were similar for left and right reaches, as illustrated in Fig. 7. For example, the time of the averaged peak discharge activity during both the left and right reaches was centered on the onset of activity in the ClB muscle (time = 0, Fig. 7, A and B), with almost half of the cells (20/43, 47%) attaining their peak discharge prior to ClB onset during left reach. A t-test showed the values for the left and right reaches were not significantly different (P = 0.48). The depth of the modulation was also similar for both left and right reaches, with respect to both the initial peak of activity (Peak_1, Fig. 7, C and D, P = 0.10) and to the second peak (Peak_2, P = 0.13, not illustrated). Moreover, the relationship between the two peaks was quite symmetrical. This can be seen in Fig. 7, E and F, where the discharge frequency during the first peak of activity was significantly linearly related to that observed during the second peak of activity for both the left and the right reaches. The similarity in the discharge activity during the left and right reaches is likewise demonstrated by the significant linear relationship of the scatterplots illustrated in Fig. 7, G and H. Together these analyses emphasize the symmetry of the neuronal responses during reaches of either limb.
Temporal relationships: go-related discharge.
The temporal relationship between different features of the discharge pattern and the selected behavioral events is illustrated in Fig. 8 for one cell from cat RS24. This neuron showed a typical bimodal period of discharge during both left and right reaches with an initial peak of activity occurring just prior to (Fig. 8A) or just subsequent to (Fig. 8B) the onset of the period of activity in the ClB. Linear regression analysis of the relationship between the onset of the increase in cell discharge and the onset of the activity in the ClB during the left reach (see insets) showed that there was only a weak relationship between these two variables (Fig. 8C, top left, ○). In contrast, a similar plot using the lead time (see methods) of the cell discharge with respect to ClB onset showed a highly correlated relationship, with R2 = 0.92 (Fig. 8C, top left, •). As detailed previously (Schepens and Drew 2003, 2004), this suggests that the onset of the cell discharge is better related to the Go signal than to the onset of the movement. The mean latency of the discharge in this neuron with respect to the Go signal was 79 ± 29.5 ms.
We were able to measure the latency of the initial change in discharge activity with respect to the Go signal for 38/43 bimodal cells during the left reach and for 29/32 bimodal cells during the right reach. Of the 38 bimodal cells for which the latency could be measured during reaching with the left limb, 15 had reach-related relationships, as determined on the basis of the linear regressions, and 20 had Go-related relationships, as determined on the basis of the variance of the latency (see methods). For 29 cells amenable to this analysis for reaching with the right limb, 9 had Go-related onsets and 12 had reach-related onsets. We believe that the apparent difference between the number of cells identified as Go-related during left reach (n = 20) and during right reach (n = 9) is, at least in part, attributed to the fewer number of trials during right reach, thus reducing the likelihood of detecting a significant relationship. Box plots (Fig. 9A) of the mean variance (top) and the mean latency (bottom) of each of those bimodal cells identified as Go-related (left plot, n = 32) and movement-related (right plot, n = 24) showed a clear separation between the two populations. The averaged mean latency of the Go-related cells (55.1 ± 17.5 ms) was statistically different from that of the movement-related cells (217.9 ± 167 ms, P < 0.001). Moreover, the small overall variance of the Go-related populations of cells attests to the fact that their onset varies little with respect to the time of the Go-stimulus.
In Fig. 9B we plot the cell latencies and the lead time for all those trials during the left, ipsilateral reach for all cells (including bimodal, phasic, tonic, and phasic/tonic) classified as Go-related. This serves to emphasize the consistent nature of the latency of the discharge activity of these Go-related cells, regardless of classification. A similar relationship was found for the discharge activity during the right, contralateral reach (not illustrated); there was no difference in mean latency for left and right reaches (P = 0.36). These analyses show that these bimodal cells could potentially contribute to the pAPA preceding movement onset, as well as the movement itself, in the same manner as we have suggested previously (Schepens and Drew 2004). The slopes of those 20 bimodal cells identified as Go-related during left reach are illustrated in Fig. 10A.
Temporal relationships: movement-related discharge.
Analysis of the termination of the initial period of activity in the cell during the left reach for the cell illustrated in Fig. 8 demonstrated a significant relationship with the end of the initial phasic period of activity in the lClB (Fig. 8C, top right). There was equally a strong relationship between the end of the period of activity in the cell and the onset of the second period of activity in the lTriL (Fig. 8C, bottom left) as well as the onset of the lever press (not illustrated). This is to be expected given the strong interrelationships between these variables (see Fig. 3B). There was also a weak relationship between the later period of activity in the ClB and the onset of the second peak of discharge activity (Fig. 8C, bottom right). Similar relationships between cell discharge and these same variables were observed during the right reach (Fig. 8D). Note that we were not able to detect the second burst of activity in the rTriL in this example and rTri was not recorded in this cat.
Significant relationships between these same variables were observed for a large proportion of the bimodal cells recorded from all three cats. For all cells showing a bimodal discharge, we measured (when possible) the same variables as illustrated in Fig. 8, C and D and performed linear regression analyses. A majority of these bimodal cells showed significant relationships between the onset of the end of the period of the activity in the lClB and the end of the first period of cell discharge (29/40, Fig. 10B) and between the onset of the second period of activity in the triceps and the end of the first period of cell discharge (23/42, Fig. 10C). As explained earlier, there is a close relationship between these two variables. Similarly, because of the correlated nature of the relationships between the onset of the second period of activity in the Tri and the onset of the lever press, essentially the same relationships as in Fig. 10C were obtained for the relationships with the lever press, whether using the end of the first period of cell discharge or the onset of the second period. There was also a weaker relationship between the onset of the second period of activity in the cells and in the ClB (Fig. 10D). In most of these examples, the slopes of the relationships were close to 1.0 and the intercepts were clustered around 0. Similar relationships were also observed during right reach (not illustrated).
Together, these results suggest the initial pause in cell discharge is strongly related either to the offset of the activity in the ClB or the associated onset of the activity in the Tri or TriL responsible for the lever press. The onset of the second period of activity occurs at approximately the same time as the EMG activity, related to the second part of the movement, begins.
Receptive fields and microstimulation.
Receptive fields were tested for 33/43 of the cells identified as being bimodal during the left reach and could be identified, at least partially, for 26/33. Of these, 19/26 included input from at least one of the forelimbs, with many of these cells discharging to light touch of all four limbs. The other 8/26 cells had a receptive field that included the neck and head. Microstimulation (11 pulses each of 0.2 ms; frequency, 330 Hz; strength, 25 μA), with the cat standing quietly, was applied at the site of recording in 13 cases. In all these, the microstimulation evoked a pattern involving activation of ipsilateral (left) forelimb flexors, right extensors, and left neck muscles (see Drew and Rossignol 1990a,b). In all cases, these effects were sufficiently strong to evoke increased FV under the right forelimb and decreased FV under the left forelimb (not illustrated).
Given that a large percentage of neurons showed a decrease in their discharge during the production of the active extensor torque required to depress the lever, we examined our database to determine whether the discharge activity of any of the other cells recorded was compatible with a contribution to this part of the behavior. We found that 13/131 cells showed a phasic increase of activity within 100 ms of the lever press. Of these, we were able to analyze temporal relationships, on a trial-by-trial basis, for 12/13. One example of such a cell, discharging reciprocally with the bimodal cells, is illustrated in Fig. 11A (filled histogram). This cell exhibited a decrease in activity prior to the onset of the lClB and then a brief phasic period of activity beginning just after the end of activity in the lClB. Linear regression analysis showed that the onset of the period of increased activity showed a weak and nonsignificant relationship with the end of the period of activity in the lClB (Fig. 11B, top) but showed a strong and significant relationship to the onset of the lever press (Fig. 11B, middle). There was no significant relationship with the onset of the second burst of activity in the lTri (not illustrated), although there was a significant relationship between the end of the phasic period of discharge and the end of the period of activity in the second burst of the lTri (Fig. 11B, bottom).
Altogether, we found 9/12 cells that showed a significant relationship to the onset of the lever press during the left reach. Seven (7/12) showed a similar increase during the right reach. The 9 cells with a significant relationship to the onset of the lever press included those classified as showing an initial decrease in discharge frequency prior to the onset of the ClB (“Decrease” in Table 1, n = 5; i.e., including the example in Fig. 11), those showing an increase subsequent to ClB onset (“Late Increase,” n = 2), and those with a tonic component (“Tonic,” n = 2).
SPIKE-TRIGGERED AVERAGING (STA).
We applied the STA technique to all neurons for which >1,000 action potentials were available during either the left or right reaches to all cells that we recorded. This provided a total of 84/131 averages during left reach and 81/127 averages during right reach; all averages were performed with ≥20 muscles for each cell. Significant postspike responses (±2SD of control activity) were observed in only 20 muscles from 12/165 neurons. This small number of positive results most likely reflects the fewer number of triggers available for the averaging in these experiments compared with our previous work (Schepens and Drew 2006). In almost all cases (18/20), the STA evoked postspike depression (PSD). However, because of the paucity of postspike responses, further quantification of the responses is not justified.
There was a large overlap between the regions in the PMRF from which different cell types were recorded (Fig. 12). Bimodal cells were recorded widely within both the NRGc and the NRPc. In contrast, cells discharging with a phasic or a phasic/tonic pattern of discharge were restricted mostly to the NRGc. There was no obvious difference in the regions from which phasic versus phasic/tonic cells were recorded.
Reanalysis of previous results
In our previous study (Schepens and Drew 2004) we recorded 82 neurons that showed an increase in their discharge frequency prior to the onset of activity in the ClB during left reach. Of these, 73/82 could be identified as discharging in one of our three major classifications (phasic, tonic, phasic/tonic). Of the other 9/82, 7/9 showed a bimodal discharge pattern but were not analyzed further because they formed only a small percentage of the cells. However, because of the results obtained during the present lever-press task, we reanalyzed our previous data using the same criteria as in the present study, with bimodal cells taking preference over other classifications. The results of this reanalysis showed that (excluding 2 unclassified neurons), 13/80 (16%) of the cells were classified as bimodal during left, ipsilateral reach (Table 2). However, when the same criteria were applied to the activity of the neurons during the right reach, fully 27/66 (41%) were classified as bimodal. These latter cells were mostly classified in our previous study (Schepens and Drew 2006) as phasic/tonic but with a decrease in discharge following the end of the activity in the ClB. One example of such a cell is shown in Fig. 13. This cell showed a phasic/tonic discharge during the left reach (Fig. 13A), with little evidence of any decrease in discharge following the initial burst of activity in the lClB compared with the bimodal cells illustrated in Fig. 6. However, during the right reach there was a complete cessation of discharge after the initial peak, as the limb was extended forward into the feeding tube (see Fig. 2C). There was then a resumption of the discharge activity throughout the time that the right limb was removed from the support surface (Fig. 13B). During this period, there was concomitant activity in the flexor muscles of the right limb (e.g., ClB, Fig. 13B) and the extensor muscles of the supporting limb (e.g., TriL, Fig. 13B). As such, this cell closely resembles the patterns of activity illustrated in Fig. 6, with the prolonged period of discharge following the minimum of activity being simply explained by the increased time that the limb was held in flexion before being replaced on the support surface in our original task compared with the lever press.
Quantitatively, those responses classified as bimodal during movements of the right limb in the reaching task resembled those observed during the lever-press task. This is shown by Fig. 13C, illustrating the ratio of the discharge frequency during Peak_1 with that observed during the minimum, as well as by Fig. 13D, which shows that the relative discharge frequency of the two peaks was similar. It should be noted that even following reanalysis of the data, there was still a substantial proportion of cells (32%: Table 2) that sustained their discharge throughout the entire period that the contralateral limb was off the ground.
The results in this study indicate that the pattern of modulation observed in many neurons in the PMRF is influenced by both movement and posture. Our results provide evidence for two major populations of PMRF neurons for the control of posture and limb movement. One population appears to control elements related to both posture and movement. These neurons are distinguished by an arrest of their discharge during the lever press, despite the maintained need for postural support in the other three limbs. We interpret this as indicating that the discharge was influenced by the type of movement (flexion vs. extension) that was being made. In contrast, a second substantial population of neurons continued to discharge despite the lever press. We interpret this as indicating that these neurons were more influenced by the postural requirements of the task. We suggest that these results show that there are populations of cells whose discharge activity best reflects the nature of the movement being made and others that have a preferential role in regulating postural support during the movement.
The task that we used in this study is a modification of that used previously in our laboratory to study the discharge characteristics of reticular neurons in the PMRF (Schepens and Drew 2004, 2006). The primary modification in this task was to require the cats to make an active extension of the limb to depress a lever against a torque of 2–3 Nm. This task was designed to allow us to determine whether the discharge of a majority of reticular neurons was linked to the combined presence of flexor activity in the reaching limb together with increased extensor activity in the supporting limb, as suggested by a previous multiple-regression analysis (Schepens and Drew 2006), or whether discharge was more dependent on one or the other.
A major goal in designing this task was to ensure that the characteristics of the movement resembled as closely as possible those observed in the previous task. The EMG and kinematic patterns illustrated in Fig. 2 for activity during both tasks, from the same cat, show that we mostly succeeded in this. The trajectory of the limb, the joint angles, and the EMG patterns are mostly comparable between tasks. The more variable onset of the postural responses in the current task may be explained by the more symbolic relationship between the Go signal and the food reward, perhaps resulting in lower motivational levels. In addition, the lower location of the lever compared with the feeding tube in Schepens and Drew (2003) may have required smaller postural responses.
There are now several reports, in both cats and primates, that neurons in both the PMRF and the mesencephalic reticular formation show increased discharge rates during reaching (Buford and Davidson 2004; Gibson et al. 1998; Schepens and Drew 2004, 2006; Stuphorn et al. 1999; Werner et al. 1997). In the primate studies, the question of whether the discharge is best related to movement or postural requirements has not been a major issue because the animals were seated and movement was not destabilizing. In the experiments from our own laboratory, performed in standing cats, postural support is essential for production of the movement. However, because movement and posture are so tightly coupled, it has proven difficult to distinguish them. Indeed, our multiple-regression analysis in a previous study (Schepens and Drew 2006) suggested that, at least for some cells, discharge might contribute to both the movement and the posture. This is in agreement with microstimulation and STA studies, suggesting that neurons in the PMRF activate flexor muscles on the ipsilateral side together with extensor muscles of the contralateral side (Davidson and Buford 2004, 2006; Davidson et al. 2007; Drew 1991; Drew and Rossignol 1990a,b; Isa and Sasaki 1988). As such, it is consistent with a system designed to modulate movement in one limb and the postural responses in the supporting limb(s) as a coordinated unit. The current task attempted to determine the extent to which these two elements are coupled by requiring the cat to perform a lever press, requiring exertion of an extensor torque in the ipsilateral limb while maintaining postural support in the contralateral limb.
NEURONAL ACTIVITY DURING THE LEVER-PRESS TASK.
Bimodal cells. One of the major findings in these experiments was the presence of a large proportion of bimodal cells. This population of bimodal neurons constituted 50% of those neurons that discharged in advance of the left, ipsilateral reach, compared with 16% that showed bimodal activity during ipsilateral reach in our previous task (reanalysis in Table 2). This is unlikely to be the result of a sampling bias because comparison of the location of the recording sites in the two studies showed a large overlap in the regions explored in the different experiments. Rather, it is most likely that the prevalence of the bimodal pattern in the current task is directly related to the task requirement to generate an extensor torque to depress the lever. Many of these cells showed an initial increase in activity in which the onset of the activity was Go-related and the offset was related to the termination of the initial period of activity in the ClB. As such, this period of discharge activity closely resembles that described previously for the phasic and phasic/tonic cells (Schepens and Drew 2004, 2006). However, in contrast to the phasic/tonic cells, this bimodal population showed a substantial depression of their activity during the lever press, despite the requirement for maintained postural support.
We interpret this as support for our suggestion that a population of reticular neurons are responsible for contributing both to the flexor muscle activity in the moving, ipsilateral limb and to the extensor muscle activity in the contralateral, supporting forelimb as a coordinated unit. It is to be expected that discharge activity in such a population of neurons would have to be suppressed if the moving limb changes from flexion to extension, as during the lever press. The fact that we found a wide range of bimodal activity, varying from cells that showed a minimal decrease in activity during the lever press to those that showed an almost complete cessation of their discharge (see Fig. 6), suggests that there may be a range of cells with varying contributions to the behavior. The discharge activity in some cells appears to be very strongly influenced by the coordinated pattern of activity in the two limbs, whereas others are more strongly influenced by either the requirement for flexion in the reaching limb or the requirement for extensor activity in the supporting limbs.
An alternative explanation for the bimodal activity is that these cells signal only the postural support in the contralateral limb and that the decrease in cell discharge is related to the small decrease in the vertical force that is observed at this time. Although we cannot entirely rule out this possibility, it should be emphasized that in many cases the decrease in cell discharge was substantial but the decrease in FLV in the supporting limb was minimal (e.g., Fig. 8). It is hard to relate such major changes in cell discharge activity to such small changes in the level of postural support, although the possibility cannot be completely discounted.
Bimodal patterns of activity were seen almost as frequently during the lever press with the limb contralateral to the recording site (32/82, 39%) as with the ipsilateral limb (43/86, 50%). This is in agreement with the general results obtained during our previous study (Schepens and Drew 2006) that most cells recorded in the PMRF showed broadly similar, nonreciprocal patterns of activity during both left and right reaches. This is compatible with our view that the influence of these cells on muscle activity patterns during movement may be modified (gated) according to context (see following text).
Other cell types.
The bimodal cells constituted only 50% of those cells discharging prior to movements of the ipsilateral limb and 39% of those discharging prior to movements of the contralateral limb (Table 2). The other neurons consisted of phasic, tonic, and phasic/tonic cells as in our previous studies (Schepens and Drew 2004, 2006). The activity in the phasic cells, discharging only prior to and during the reach, would not have been affected by the modifications in task in this study. As such, it is not surprising that we recorded approximately equal proportions of these cells in each task (Table 2). In contrast, all those cells with a tonic component are candidates for possible modification. The fact that a substantial proportion of cells continued to show a maintained tonic discharge throughout the lever press therefore suggests that these cells contribute primarily to the postural support required throughout the behavior, independently of the type of activity (flexion or extension) in the reaching limb. However, this finding has to be reconciled with the results from our multiple-regression analysis in our previous work (Schepens and Drew 2006) that suggested that many phasic/tonic cells likely contribute to both flexor muscle activity in the reaching limb and extensor muscle activity in the supporting limb, especially during the dynamic part of the reach. One possibility is that the population of phasic/tonic cells recorded in this task form a subpopulation that do, indeed, contribute only to the extensor activity in the supporting limb(s). Another possibility is that the signal to flexor-related circuits in the spinal cord may be gated out (reduced) during the lever-press activity. Such a gating signal may be provided by the population of neurons that increased their discharge activity during the extensor torque. Such cells may participate in the production of the extensor tonus in the reaching limb and, simultaneously, modulate (reduce) the activity of the interneurons involved in the flexion activity. Alternatively, or in a complementary fashion, the gating of the spinal activity may be a function of descending signals from the cortex that specify the movement to be made (Drew et al. 2004; Schepens and Drew 2006).
Considerations and possible mechanisms.
The first is the extent to which the discharge activity that we detail is related primarily to control of the forelimb musculature that we used to make our temporal and quantitative (Drew and Schepens 2004, 2006) analyses. We concentrated on examining the correlations with the forelimb muscles in this study primarily because the task required a voluntary movement of these limbs. Moreover, in support of our view that these cells contribute to regulating activity in the forelimbs, we found good temporal correlations with forelimb muscle activity—many cells had receptive fields including the forelimbs and microstimulation evoked sufficiently strong activity in forelimb muscles to modify the ground reaction forces under the forelimbs. However, as detailed in the introduction, we recognize that many cells in the PMRF have extensive branching patterns and may influence muscle activity in multiple limbs as well as nuchal or axial musculature. As such, it must be considered that the correlations explored in this study provide only a partial representation of the full expression of the descending signal from these cells and many of these cells undoubtedly influenced other muscle groups. For example, those cells identified as RSNs must have influenced the activity of hindlimb muscles in some manner, most likely as a coordinated pattern of postural support as we have discussed elsewhere (Schepens and Drew 2006). Similarly, these cells will be influenced by head movements and will modify axial musculature. Indeed, it is probable that many of the cells recorded here will contribute to the expression of a complete behavioral pattern that includes the coordinated activity of all four limbs as well as head and trunk orientation. Others will probably have more specific actions as discussed elsewhere (Prentice and Drew 2001; Schepens and Drew 2004, 2006).
The second issue is why the cells discharge in a broadly similar manner during the left and right reaches. Although discharge activity is clearly not identical during reach of either limb, it is equally clearly not reciprocal in the same way as is the pattern of postural activity (see e.g., Fig. 5). Although we have discussed these issues previously (Schepens and Drew 2006), they bear repeating in the light of the results reported in this study demonstrating that the nonreciprocal nature of the discharges extends even to the decrease in activity produced by extension of each limb. This suggests that the bilateral, nonreciprocal nature of the discharge may be an integral part of the reticulospinal control system. The fact that so many cells show a broadly similar, nonreciprocal pattern of activity during reaching with either limb may be interpreted in two broad manners. One is that cell activity regulates activity in muscles that show similar patterns of activity during reaches with either limb. We cannot discount this completely but that such activity is seen in almost the entire population of cells and that a majority of these cells had receptive fields including one or more limbs make this unlikely, at least as the sole interpretation. The alternative explanation is that there is gating of the signal at the level of the spinal cord (Schepens and Drew 2006) so that activity in a given cell activates muscles on one side or the other side of the body according to the limb being used. This is similar to what has been described and suggested during locomotion (Drew et al. 2004) and is supported by the results from our previous STA study (Schepens et al. 2006). Unfortunately, we were not able to confirm these suggestions by using STA in this study because of the small number of trigger spikes available to us.
COMPARISON WITH OUR PREVIOUS RESULTS.
Neurons showing a bimodal pattern of activity during ipsilateral reach were only rarely observed in our previous study (Table 2), despite the fact that the limb was extended into the feeding tube (Fig. 2C). The fact that relatively fewer cells with a tonic component were found in this lever-press task than in our previous task (Table 2) suggests that the bimodal discharge pattern was favored in this task. Our suggestion, as stated earlier, is that the depression in cell discharge that produces the bimodal pattern is a result of the active production of the extensor torque needed to depress the lever. In this situation, despite the need for continued postural support in the contralateral limb, the cells do not discharge because of the decoupling between the flexor muscle activity of the ipsilateral limb and the extensor muscle activity in the contralateral supporting limb. In contrast, during contralateral reach, cells discharged similarly in the two situations and more equal proportions of each type of cell were recorded in both tasks. Thus even in our previous reaching task, these cells would not have been able to contribute to postural support in the ipsilateral limb during this period of decreased discharge activity. We thus suggest this population of cells contributes only to the flexor muscle activity in the contralateral limb and makes no contribution to postural support in the ipsilateral limb. In this situation, postural support in the ipsilateral (left) limb must be provided by the population of tonic and phasic/tonic neurons that remain active throughout this period.
Taken together, our results suggest that ipsilateral flexion and contralateral extension are regulated by at least three populations of cells, two populations that contribute primarily to the regulation of the flexor and extensor muscles independently and a third population that controls them as an integrated activity. In contrast, we suggest that contralateral flexion and ipsilateral extension are always controlled by at least two independent populations of cells and that there may be no population that controls them as an integrated unit. This supports our previous arguments (Schepens and Drew 2006), based in part on the results from our STA study, that the nature of the contribution from the PMRF to the control of the limbs during left (ipsilateral) and right (contralateral) reaches, although bilateral, must be asymmetrical.
Spatial organization of the different cell types
As we reported earlier (Schepens and Drew 2004), cells identified as phasic and phasic/tonic were distributed widely within the NRGc with little indication of any differential spatial organization. In addition, results from this study show that bimodal cells are equally widely distributed within the NRGc. In contrast, however, Fig. 12 shows that only bimodal cells were widely distributed within the NRPc. Although most of the cells in the PMRF were recorded in a single cat (RS24), bimodal cells from RS26 were also found in the same region. Moreover, all cell types were recorded in cat RS24 (Table 1) and there is no a priori reason why only bimodal cells would be located within the NRPc. Instead, this would seem to indicate that neurons in the NRPc may contribute preferentially to the coordination of movement and posture as a coordinated activity. Confirmation of this suggestion will require further study given the relatively small number of recordings in the NRPc (no recordings in previously used cats RS22 or RS23 were made in the NRPc; see Schepens and Drew 2004).
These results extend the findings that we have previously made on the contribution of the PMRF to the control of posture and movement by specifically demonstrating that some cells contribute to the control of movement and posture as a coordinated unit. This is particularly shown by the presence of a large proportion of bimodal cells during the ipsilateral reach. The fact that these cells show increased activity prior to the onset of the reach but decreased activity during the lever press suggests that such cells may contribute to the control of the flexor components of the movement and the associated postural support as a coordinated unit. In contrast, the presence of a substantial number of neurons that continued to discharge tonically throughout the extensor activity associated with the lever press suggests the presence of a population of cells contributing primarily to the postural activity in the supporting limb independently of the specific nature of the activity in the ipsilateral limb. The fact that similar activity is observed during right, contralateral reach supports our previous suggestions that the descending reticulospinal signal may be gated, with the final expression being contingent on the level of excitability of spinal interneuronal circuits. We suggest that the excitability in these circuits is determined by the same descending command signals responsible for the reaching movement.
This work was supported by an operating grant from the Canadian Institutes of Health Research and an infrastructure grant from the Fonds de la Recherche en Santé du Québec. B. Schepens was supported by a Jasper Fellowship and a Human Frontier Science Program Scholarship; P. Stapley was supported by a Jasper Fellowship.
We thank M. Bourdeau, N. De Sylva, P. Drapeau, C. Gauthier, J. Lavoie, F. Lebel, and J. Soucy for technical assistance in the performance and analysis of these experiments and Drs. C. Chapman, N. Krouchev, and S. Rossignol for helpful comments on this manuscript.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 by the American Physiological Society