Changes in locomotor activity parameters with variations in cycle time in larval lamprey

The basic temporal pattern of locomotor activity (i.e. walking, flying, and swimming) is generated by central pattern generators (CPGs) in the spinal cord that consist of neuronal oscillators as well as motor neurons and coordinating neurons (for reviews, see Stein,1978; Grillner,1981; McClellan,1996). Oscillators project to motor neurons, which directly control contraction of muscles, while coordinating neurons couple the oscillators to control the relative timing of locomotor activity for different parts of the body. The spinal CPGs can produce the basic pattern of locomotor activity without sensory feedback, but sensory inputs are crucial for normal locomotion, particularly in response to environmental perturbations (for a review, see Grillner, 1981). Locomotor activity is initiated by a command system in the brain that activates the spinal CPGs.

Undulatory locomotor behavior (i.e. swimming) in most fish and certain aquatic amphibians is a repetitive motor act that is mediated by two basic features (Gray, 1933a, b, 1936; for reviews, see Grillner and Kashin, 1976; Williams, 1986; Roberts et al., 1997)(Fig. 1). (1) At each segmental level of the body, right—left bending is produced by right—left alternation of burst activity in axial muscles; and (2) rostral-to-caudal propagating body undulations are produced by a rostrocaudal delay of axial muscle burst activity on the same side of the body. These features result in an S-shaped body wave that propagates toward the tail and results in both lateral and backward force vectors (Gray, 1933a, b; Webb, 1984; Bowtell and Williams, 1991). The lateral force vectors in the rostral and caudal body are in opposite directions and cancel, leaving the backward components of the force that push against the water and propel the animal forward. During swimming, there is usually approximately one S-wave along the body during different speeds of swimming (reviewed by Williams,1986). A single S-wave along the body is a highly efficient manner in which to generate backward force during swimming, since multiple S-waves would require additional muscle contraction forces to produce multiple regions of sharper than normal bending of the body.

For swimming, the CPGs in the spinal cord are thought to consist of a chain of left and right segmental oscillators that are coupled by reciprocal inhibition and that produce left—right alternation (for reviews, see Grillner et al., 1995, 2000; Roberts et al., 1997). Longitudinally in the spinal cord, the oscillators are coupled by a spinal coordinating system that is important for generating rostrocaudal phase lags(for a review, see McClellan,1996).

Recent evidence from the isolated spinal cord of adult lamprey, in which locomotor activity was evoked by bath-applied pharmacological agents, suggests that intersegmental phase lags might not be constant but appear to increase with decreasing cycle times (Tegnér et al., 1997). However, it is not known if the parameters of locomotor activity are controlled in a similar fashion in larval lamprey. For example, there appear to be some minor qualitative differences in the kinematics of swimming in adult and larval lamprey (A. D. McClellan,unpublished observations). In addition, it has been suggested that the spinal locomotor circuitry in larval lamprey is immature compared to adults, due to a lack or immaturity of some types of cells(Cohen et al., 1990). In larval lamprey, it is important to determine how the parameters of swimming motor activity vary with cycle time, because this information will be necessary for formulating and testing models of spinal CPGs.

In larval lamprey, preliminary results for brain-initiated locomotor activity in brain/spinal cord in vitro preparations suggest that the slopes for intersegmental phase lag versus cycle time are not significantly different from zero (i.e. ϕ is relatively constant)(McClellan and Hagevik, 1999). In the present study, a much more comprehensive analysis was made using archival locomotor activity from several of our previous studies (see Materials and methods for list of references) to test whether in fact burst proportions (BP) and intersegmental phase lags (ϕ) are constant with changes in cycle time (T) in larval lamprey. Spontaneous or sensory-evoked locomotor muscle activity (electromyographs, EMGs) from whole animals as well as brain-initiated in vitro locomotor activity from brain/spinal cord preparations were further analyzed for larval lamprey. For each animal in the whole animal or in vitro group, the slopes of the BP and ϕ values versus T were determined by regression analysis,and then statistics were used to determine if the composite slope of each parameter versus cycle time for the group was significantly different from zero, similar to the methods previously described(Hagevik and McClellan, 1997; McClellan and Hagevik,1999).

The data analyzed in the present study were derived from larval lamprey Petromyzon marinus L. Spontaneous or sensory-evoked locomotor muscle activity (EMGs) in whole animals (Davis et al., 1993; Paggett et al.,1998) and brain-initiated locomotor activity in brain/spinal cord in vitro preparations (McClellan,1994; McClellan and Hagevik,1999) were further analyzed for the purposes of the present study(see below). Since the methods for the collection of this data are described in detail elsewhere, only the most relevant information will be repeated here.

Briefly, pairs of fine copper wires (56 μm diameter) were implanted in contact with body musculature at approximately 20% body length (BL),40% BL and 60% BL (Fig. 2Ai; electrode configurations were either 1-2-3-4 or 1-2-3-5, see Fig. 1B). Subsequently,locomotor behavior and motor activity occurred either spontaneously or were elicited by brief stimulation of the oral hood or tail. During locomotion,muscle activity (EMGs) was recorded and stored on videotape (Neurodata DR886,11 kHz sampling rate per channel), and episodes of locomotor activity were selected for analysis during relatively constant velocity swimming along a straight line (Fig. 2Aii).

In vitro brain/spinal cord preparations from lamprey were prepared as described in detail elsewhere(McClellan, 1994; Hagevik and McClellan, 1994, 1999; McClellan and Hagevik, 1999). The preparations (Fig. 2Bi)were pinned dorsal side up in a recording chamber containing lamprey Ringer's solution (McClellan, 1990)maintained at 6-9°C. During recordings, the Ringer's solution contained 15 mg l^-1 D-tubocurarine chloride (Sigma; St Louis, MO, USA) to block possible muscle contractions in any remaining muscle along the notochord. It is unlikely that curare significantly altered the in vitro rhythms,since brain-initiated in vitro locomotor activity is virtually identical with or without 150 mg l^-1 curare in the bath, or ten times the concentration used for the present study (N=3; P. Hinton and A. D. McClellan, unpublished data). Suction electrodes were placed in contact with ventral roots in the rostral (approx. 20% BL), middle(approx. 40% BL) and caudal (approx. 60% BL) spinal cord,which correspond approximately to segments 13, 35 and 61, respectively. In vitro spinal locomotor activity (Fig. 2Bii) was initiated by chemical microstimulation in brain locomotor areas, as described previously in detail(McClellan, 1994; Hagevik and McClellan, 1994; McClellan and Hagevik,1999).

Selected episodes of muscle activity during locomotor behavior were played out on a thermal array recorder (Gould TA2000) at 50 mm s^-1. Similar procedures were used for in vitro locomotor activity, except that this activity was integrated (τ=50 ms) because the signal-to-noise ratio was not always high enough to reliably determine the onsets and offsets of locomotor bursts (see figure 2B,C in McClellan, 1988). The onsets and offsets of locomotor burst activity were digitized with interactive software and a digitizing tablet (GTCO 1117A). Data points were imported into a spreadsheet program (Lotus 1-2-3) for performing calculations of cycle time(in ms), burst proportion and phase lag for each locomotor cycle (see Fig. 1B) as well as for obtaining graphs.

For each animal, the calculated parameters of locomotor activity were imported into a statistical analysis program (InStat), and linear regression analysis was performed for cycle times versus each of the following parameters: BP1, BP2, BP3, BP4, BP5, ϕ₁₅, ϕ₂₃and ϕ₃₄ (Fig. 1B; for example, BP1 refers to burst proportions calculated from activity recorded at electrode 1, while ϕ₁₅ refers to phase lags calculated from burst activity recorded at electrodes 1 and 5) (see McClellan and Hagevik, 1997). Linear regression analysis was used to determine the slopes and y-intercepts of a best-fit line through the data points and whether the points could be sufficiently described by linear analysis. For each locomotor activity parameter (i.e. EMG or in vitro activity), the mean slopes for all of the animals within the group were averaged to determine the composite mean ± S.D. for that particular parameter(Table 1).

First, for lamprey in the whole animal or in vitro group, the slopes of the locomotor activity parameters versus cycle time were used to create distribution histograms (Figs 3, 4). These histograms were then evaluated to determine if the slopes were distributed approximately uniformly or if there were other trends (e.g. bimodal distributions). Second, for lamprey in the whole animal or in vitro group, the proportion of positive and negative slopes for each locomotor activity parameter versus cycle time was analyzed using the Sign test, as previously described (McClellan and Hagevik,1999), to determine if the composite slope of the parameter for the group was significantly different from zero (i.e. P≤0.05) (see Table 1).

For muscle activity (EMGs) during locomotion in whole animals, distribution histograms indicated that the slopes for the parameters of locomotor activity versus cycle time (range: 130-826 ms) were approximately uniformly distributed (Fig. 3), without obvious gaps or groupings in the distributions, as would be characteristic for bimodal or multimodal distributions. The slopes for burst proportion versus cycle time were centered around negative values(Fig. 3A), while the slopes for intersegmental phase lag versus cycle time were mostly centered approximately at zero (Fig. 3B).

As was the case for muscle activity from whole animals, for in vitro locomotor activity, the distribution histograms of the slopes for the parameters of locomotor activity versus cycle time (range:409-3416 ms) were approximately uniformly distributed(Fig. 4). In addition, the distributions of slopes for burst proportion versus cycle time were centered around negative values (Fig. 4A), while those for intersegmental phase lag versuscycle time were mostly centered approximately around zero(Fig. 4B).

Since the distribution histograms for the slopes of locomotor activity parameters versus cycle time were approximately uniformly distributed, use of the Sign test to evaluate this data was justified (see Materials and methods). Thus, for all lamprey in the whole animal or in vitro group, the Sign test was applied to the proportion of positive and negative slopes for each locomotor activity parameter versus cycle time to determine if the composite slope was significantly different than zero. In contrast, had the distribution histograms been multimodal, with groupings of slope values, the Sign test probably would have been inappropriate. For example, if a given distribution histogram for a parameter had been bimodal, with half of the slopes centered around a positive value and half centered around a negative value, the Sign test would indicate that the slopes are not significantly different from zero, even though this obviously is not true.

In individual whole animals, analysis of locomotor muscle activity (i.e. EMGs) indicated that there was a clear tendency for burst proportions (range of mean values for all animals ≈0.12-0.44) to be larger for shorter cycle times than for longer cycle times (see Fig. 5 and legend). For all whole animals, regression analysis (lines in Fig. 6Ai,Bi) indicated that most of the slopes for burst proportion versus cycle time had values that were negative (Fig. 6Ci). Furthermore, the mean slopes of BP versus T for each parameter (i.e. BP1-BP5) were negative and ranged from -2.5×10^-4 to-5.0×10^-4 (Table 1). Analysis of the proportion of positive and negative slopes using the Sign test indicated that the slopes for all of the BP values versus cycle time were significantly less than zero(P≤0.05, Table 1). Thus, for EMGs in larval whole animals, burst proportions appear to increase significantly with decreasing cycle times (i.e. a negative slope).

In individual whole animals, intersegmental phase lags (range of mean values for all animals ≈0.0039-0.0110) for locomotor muscle activity did not appear to vary appreciably with changes in cycle time (see Fig. 5 and legend). For all whole animals, regression analysis (lines in Fig. 6Aii,Bii) indicated that the individual slopes for phase lag versus cycle time could be positive or negative (Fig. 6Cii). The mean slopes of ϕ versus T for each parameter (i.e. ϕ23, ϕ34 and ϕ15) ranged from-0.17×10^-6 to -1.8×10^-6(Table 1), and analysis of the proportion of positive and negative slopes using the Sign test indicated that these slopes were not significantly less than zero (P>0.05, Table 1). Thus, phase lags for muscle activity in larval whole animals do not appear to change significantly with variations in cycle time.

For brain-initiated locomotor activity recorded from in vitrobrain/spinal cord preparations, the composite mean slopes for burst proportion(range of mean values for all animals ≈0.28-0.52) versus cycle time had negative values that ranged from -0.23×10^-4 to-1.5×10^-4 (Table 1). However, application of the Sign test to the proportion of positive and negative slopes for burst proportions versus cycle time(i.e. BP1—BP5) indicated that these slopes were not significantly different from zero (Table 1).

For intersegmental phase lags (range of mean values for all animals≈0.0013-0.0055), the composite mean slopes versus cycle time ranged from -0.97×10^-6 to -2.9×10^-6(Table 1). Applying the Sign test to the proportion of positive and negative slopes for ϕ versus T indicated that these slopes were not significantly different from zero(Table 1), similar to our previously reported preliminary results(McClellan and Hagevik,1999).

For locomotor muscle activity in larval lamprey, the composite mean slopes for burst proportion versus cycle time were negative(Fig. 4A), and the Sign test indicated that these slopes were significantly less than zero(Table 1). In in vitropreparations, although the mean slopes of burst proportion versuscycle time were also negative (Fig. 4A), they were not significantly less than zero(Table 1).

In whole animals, a negative slope for burst proportion versuscycle time that is relatively constant (e.g. ≈-2.0×10^-4)over a modest range of cycle times indicates that as cycle times decrease, for example from 1000 ms to 500 ms, and swimming speed increases, burst proportion(i.e. the relative burst duration) will increase by about 0.1 (e.g. from 0.3 to 0.4; see Fig. 6Ci). Thus,for faster swimming speeds, locomotor bursts will occupy a larger fraction of the cycle and generate proportionately more force per cycle. For faster swimming speeds, this increase in force generation may be necessary to overcome the additional viscous resistance of moving the body rapidly through water.

In whole animals, a negative slope for burst proportion versuscycle time could be due, in part, to sensory feedback. However, this may also be a property of the spinal CPGs, since in in vitro preparations, the composite mean slopes for burst proportion versus cycle time,although not significantly different from zero, were centered around negative values (Fig. 4A). For example,during locomotor activity in the lamprey, spinal motoneurons receive alternating depolarizing and hyperpolarizing synaptic potentials in a quasi-sinusoidal fashion (Russell and Wallén, 1983; Wallén et al., 1985). As cycle times become shorter, the amplitude of this quasi-sinusoidal synaptic drive to motoneurons would be expected to increase in order to recruit additional neurons (Davis and Murphey,1969; see figure 4 in Grillner and Kashin, 1976). If the thresholds for action potentials are constant, at shorter cycle times an increase in the amplitude of the sinusoidal synaptic input to motoneurons could contribute, in part, to an increase in burst proportion.

Why have changes in burst proportion with variations in cycle time not been observed in previous studies in the lamprey? First, in our previous studies with larval lamprey, variations of the parameters of locomotor activity with changes in cycle time were not analyzed statistically. In addition, in these previous studies (e.g. McClellan 1990), the data for all animals were pooled together, and this tends to obscure possible changes versus cycle time because each animal has a slightly different mean BP and mean T. Second, in contrast to the present study that examined `free swimming' in larval lamprey,in a previous study that involved adult lamprey swimming in a `swim-mill',changes in BP with T were not seen(Wallén and Williams,1984). These differences might be due to minor kinematic differences of swimming in larval and adult animals. However, there may be some differences between swimming in a swim-mill, in which water is forced to flow past a stationary animal, and `free swimming', in which an animal must actively generate forces against the water to mediate forward progression. For example, running on a treadmill and running over ground, although qualitatively similar, exhibit several kinematic(Schache et al., 2001; Frishberg, 1983) and muscle activity differences (Murray et al.,1985; Wank et al.,1998).

In larval lamprey, during the initial phase of burrowing, when an animal attempts to penetrate the viscous substrate (e.g. sand), mean cycle times are significantly shorter and burst proportions are significantly larger than during swimming (Paggett et al.,1998). Thus, in the present study, for whole animals that swam at the shortest mean cycle times (approximately 130-250 ms), some of the increase in burst proportions could be in response to the resistance encountered during rapid swimming through the water. However, even for animals that swam at longer mean cycle times (e.g. 400-750 ms), most of the slopes for burst proportion versus cycle time were negative (e.g. Fig. 6Ci).

For fish swimming in a `swim mill', there is some evidence to suggest that burst proportions can increase with decreasing cycle times. For example,during swimming in trout (figure 4A,C in Grillner and Kashin, 1976),for a mean cycle time of approx. 300 ms the mean burst proportion was approx. 0.320, while for a mean cycle time of approx. 135 ms during faster swimming,the mean burst proportion increased to approx. 0.490.

For locomotor muscle activity and in vitro locomotor activity in larval lamprey, the composite slopes for intersegmental phase lag versus cycle time were centered around zero (Figs 3B, 4B), and none of these composite slopes were significantly different from zero(Table 1). Thus, it appears that intersegmental phase lags do not change significantly with variations in cycle time, similar to our previous results(McClellan and Hagevik,1999).

A constant intersegmental phase lag indicates that as cycle times decrease and swimming speeds increase, the rostral-to-caudal delay of ipsilateral burst activity will decrease proportionally to the duration of the cycle (see Grillner and Kashin, 1976; Williams, 1986). Therefore, as cycle times change, the relative timing of burst activity will remain approximately constant, and a single S-wave along the body will be retained,which is a highly efficient manner for swimming.

In adult lamprey, intersegmental phase lags of locomotor activity have been reported to be relatively constant with changes in cycle time for both whole animals swimming in a `swim mill' and isolated spinal cords(Wallén and Williams,1984). In contrast, in a more recent study with isolated spinal cords from adult lamprey, in which locomotor activity was evoked by bath-applied pharmacological agents, data suggest that intersegmental phase lags might not be constant but appear to increase with decreasing cycle times(Tegnér et al., 1997). There are at least two possible explanations for the opposing results found in the present study in larval lamprey in which phase lags were relatively constant. First, perhaps there are some minor differences in the operation of spinal locomotor networks in larval and adult lamprey (see Cohen et al., 1990). Second,for whole animal and in vitro brain/spinal cord preparations in the present study, spinal locomotor networks were activated by descending systems in the brain. By contrast, it is not known if application of pharmacological agents to the isolated spinal cord activates locomotor networks with the full complement of mechanisms that occur during descending activation from the brain.

In theory, in the lamprey, rostrocaudal phase lags could be due to at least three mechanisms (for a review, see McClellan, 1996): (a) short distance coupling between oscillators in adjacent regions of the spinal cord;(b) long distance coupling between oscillators in more separated regions of cord; and (c) gradients in oscillator frequency along the spinal cord. Several lines of evidence suggest that short distance coupling is much stronger in the descending direction than ascending coupling(Hagevik and McClellan, 1994;reviewed in McClellan, 1996)and is the main mechanism that contributes to rostrocaudal phase lags(McClellan and Hagevik, 1999; Hagevik and McClellan, 1999; Matsushima and Grillner, 1992;also see Mellon et al., 1995; Buchanan et al., 1995).

In late embryonic Xenopus laevis, swimming is produced by single spikes per cycle in motoneurons in each segment (for a review, see Roberts et al., 1997). During changes in cycle time, there is a constant delay, instead of a constant phase lag, between ipsilateral rostral and caudal spikes(Tunstall and Roberts, 1991). Just a short time later, in young tadpoles, the single spikes per cycle are replaced by bursts of action potentials, and the coordinating system then maintains a constant intersegmental phase lag during changes in cycle time(for a review, see Tunstall and Sillar,1993). It has been suggested that a head-to-tail gradient of excitation and inhibition within the spinal locomotor networks determines the normal anterior-to-posterior propagation of swimming motor activity(Tunstall and Sillar, 1993; Tunstall and Roberts, 1994;also see Skinner and Mulloney,1998).

In crayfish, the paired swimmerets make coordinated periodic power stroke/return stroke movements, and each pair appears to be controlled by a separate oscillator (Ikeda and Wiersma,1964). The rostrocaudal phase lag between the movements of each pair of appendages is approximately 25%. Coupling between the swimmeret oscillators is thought to be asymmetrical and dominated by ascending coupling,and the posterior segment is thought to lead each cycle of activity(Braun and Mulloney, 1995; Mulloney, 1997; for a review,see Skinner and Mulloney,1998).

In the leech, swimming is produced by alternating contractions of ventral and dorsal musculature in each body segment. The oscillators in specific ganglia that generate the swimming pattern are thought to be coupled symmetrically (Friesen and Hocker,2001) with a U-shaped excitability gradient along the ventral nerve cord (Hocker et al.,2000; for a review, see Friesen and Cang, 2001). Furthermore, unlike the lamprey spinal locomotor networks, in the leech both short-distance and long-distance coupling appear to be important for coordination of swimming because the phase lags for swimming activity in the isolated nerve cord are dependent on the number of segments (Pearce and Friesen,1985). Finally, phase lags for swimming in the leech appear to be partially dependent on sensory inputs (Cang and Friesen, 2000), unlike that in the lamprey(Wallén and Williams,1984).

For swimming motor activity in larval lamprey, the slopes for burst proportion versus cycle time were negative, and in whole animals,these slopes were significantly less than zero. These results suggest that, as cycle times decrease and swimming speed increases, the portion of the cycle occupied by a burst will increase, presumably to generate additional force to compensate for the increase in viscous resistance of moving the body rapidly through water. By contrast, intersegmental phase lags were relatively constant during changes in cycle time, presumably to retain the ergonomically efficient S-shaped body wave during different speeds of swimming.

We thank Dr Richard Madsen for consultation about the statistics. Supported by NIH grant NS29043, NSF grant IBN 9817905, and American Paralysis Association grant MA1-9605 awarded to A.D.M. Support for M.R.B. was provided in part by the LS UROP at the University of Missouri.