|Home | About | Journals | Submit | Contact Us | Français|
The coordination of multiple neural oscillators is key for the generation of productive locomotor movements. In the medicinal leech, we determined that activation and coordination of the segmental crawl oscillators, or unit burst generators, are dependent on signals descending from the cephalic ganglion. In nearly-intact animals, removing descending input (reversibly with a sucrose block) prevented overt crawling, but not swimming. Cephalic depolarization was sufficient for coordination. To determine whether descending signals were necessary for the generation and maintenance of posterior-directed intersegmental phase delays, we induced fictive crawling in isolated whole nerve cords using dopamine (DA) and blocked descending inputs. After blockade, we observed a significant loss of intersegmental coordination. Appropriate phase delays were also absent in DA-treated chains of ganglia. In chains, when one ganglion was removed from its neighbors, crawling in that ganglion emerged robust and stable, underscoring that these oscillators operate best with either all or none of their intersegmental inputs. To study local oscillator coupling, we induced fictive crawling (with DA) in a single oscillator within a chain. Although appropriate intersegmental phase delays were always absent, when one ganglion was treated with DA, neighboring ganglia began to show crawl-like bursting, with motoneuron spikes/burst greatest in untreated posterior ganglia. We further determined that this local excitatory drive excluded the swim-gating cell, 204. In conclusion, both long-distance descending and local inter-oscillator coupling contribute to crawling. This dual contribution helps to explain the inherent flexibility of crawling, and provides a foundation for understanding other dynamic locomotor behaviors across animal groups.
Animal locomotion occurs in many forms and encompasses a wide variety of rhythmic movements. Yet, across this array of activities, it is the coordination of multiple neural oscillators that is needed to generate fluid and productive locomotor behavior (Grillner, 2003). A core feature of this coordination is the so called ‘coupling’ of central pattern generators (CPGs) within the central nervous system (CNS), which can employ one or more key strategies: 1) Coupling that is inherent to the rhythm-generating circuitry; leading to greater stereotypy in the phase relationships between oscillators, and 2) Coupling that is extrinsic to the unit oscillators. This latter strategy is often associated with a greater flexibility in behavioral output, but at the expense of the predictability provided by the first one. Because stereotypy is a beneficial attribute that renders a behavior more amenable to a physiological analysis, many studies of locomotion have revealed neural operations that are comprised of relatively tightly coupled oscillators. For example, experimental and modeling studies of swimming in the lamprey and leech, insect flight and swimmeret beating in the crayfish have provided much of the foundation underlying what is currently understood about the coordination of oscillators whose coupling is an emergent property of the oscillators themselves (Ikeda and Wiersma, 1964; Cohen and Wallen, 1980; Robertson and Pearson, 1983; Cang and Friesen, 2002; Smarandache et al., 2009). In contrast, other locomotor behaviors exhibiting much more variability are not as well understood at the cellular level (Kiehn and Butt, 2003).
Studies of walking in insects (Büschges et al., 2008), cats (Büschges, 2005) and humans (Yang and Gorassini, 2006), and those of crawling in the medicinal leech (Cacciatore et al, 2000), suggest that these locomotor events may employ oscillator-coupling strategies to maximize flexibility. Some research has indicated that higher-order descending information may contribute to oscillator coupling for both leech crawling (Cacciatore et al., 2000; Cornford et al., 2006) and cat walking (Shimamura and Kogure, 1983; Drew et al., 1986), although more direct evidence is warranted. In the cat and human, investigations of oscillator coupling and descending control have been limited by the inherent complexities of the mammalian CNS and the locomotor behaviors themselves. Thus, to understand better how a relatively slow and variable locomotor behavior is controlled, we turned to a study of crawling behavior in the medicinal leech.
Briefly, leech crawling consists of alternating elongation and contractions of the whole body with coordinated attachment and release of the anterior and posterior suckers. Segmental movements always progress along the body in an anterior to posterior direction, but they are often dynamic and can be suspended or modulated during each elongation cycle, often with other behaviors overlaid (e.g., searching-related hyperextensions, lateral turning, and dorsal-ventral undulations). Importantly, after a pause in the elongation phase, the movement is often completed from the start of the initial pause (Cacciatore et al., 2000). This flexibility resembles that observed during walking of the stick insect (Gruhn et al., 2009), cat (Akay et al., 2006) and human (Ivanenko et al., 2009). Recently, we established that every segmental ganglion of the leech ventral nerve cord contains a complete crawl oscillator (CPG), which can be independently activated (Puhl and Mesce, 2008). Because of this anatomically favorable situation, we examined the nature of the coupling of these oscillators and determined the role of higher-order descending information to choreograph their activation and coordination.
Adult medicinal leeches (Hirudo verbana) (Siddall et al., 2007) weighing between 1.5-3.0 g were obtained from either Leeches USA (Westbury, NY, USA) or Niagara Medical Leeches (Niagara Falls, NY, USA). Leeches were maintained, at room temperature, in distilled water containing dilute Instant Ocean sea salts (1.89 g/gallon; Spectrum Brands, Atlanta, GA).
Leeches were anesthetized in ice-chilled saline prior to dissection. All dissections and electrophysiological recordings were performed at room temperature in chilled normal saline. Normal saline contained (in mM/L) 116.0 NaCl, 4.0 KCl, 1.8 CaCl2, 1.5 MgCl2, 10.0 dextrose, 10.0 Trizma preset crystals, pH 7.4 (Sigma-Aldrich, St. Louis, MO) (adapted from Nicholls and Baylor, 1968). To block chemical synaptic transmission a 20:1 Mg++/Ca++ ratio saline was used (mmol/L): 87.0 NaCl; 4.0 KCl; 1 CaCl2; 20 MgCl2; 10.0 dextrose; Trizma preset crystals, pH 7.4 (Szczupak et al., 1998). Saline containing dopamine (DA) was made fresh daily and used within 5 hours. To obtain final DA concentrations of 50-100 μM, a 1 mM DA stock was diluted with normal saline immediately prior to use (Puhl and Mesce, 2008).
The nervous system of the leech consists of a compound cephalic ganglion (i.e., the supraesophageal and subesophageal ganglia), 21 segmental ganglia (designated M1-M21), and a compound tail ganglion (Fig. 4A). Connectives joining these ganglia are comprised of two hemi-connectives and a smaller median Faivre's nerve (FN).
Each segmental ganglion contains a dorsal and ventral layer of neuronal somata. Key somata of uniquely identifiable motoneurons (MNs) active during crawling are: the ventral lateral circular excitor (CV) MNs and the dorsal longitudinal excitor (DE-3) MNs, located on the ventral and dorsal surfaces respectively (Stuart, 1970).
In this study, locomotor behaviors were monitored in both nearly-intact and isolated CNS preparations. The nearly-intact preparation allowed for observation of overt locomotion while providing direct access to the CNS (Esch et al., 2002). The term “isolated whole nerve cord” refers to a complete CNS preparation dissected from the body, while the term “chain preparation” refers to a subset of isolated ganglia lacking the cephalic and tail ganglia. To observe fictive locomotion in isolated or chain CNS preparations, motor patterns were monitored by recording the extracellular activity of DE-3, whose large axon resides in the dorsal posterior (DP) nerve roots of each ganglion. The DE-3 spike is typically the largest unit in the DP recordings (Kristan et al., 1974). We defined fictive crawling as three or more consecutive bursts of DE-3 with periods ranging from 5-25 s (Eisenhart et al., 2000; Puhl and Mesce, 2008).
For isolated whole nerve cord preparations, the entire intact CNS was removed with both left and right DP nerves from 2-4 segmental ganglia. Shorter chain preparations were made by removing 4-6 segmental ganglia, ensuring the connectives remained intact. These chains were taken from M7-M18 including both left and right DP nerves from at least 2-3 segmental ganglia. When necessary, a Vaseline well was placed around individual ganglia, isolating them for focal DA application. In some experiments, the FN was manually severed or crushed using a fine forceps, leaving both hemi-connectives intact. The hemi-connectives were severed by cutting them with fine-tipped scissors, leaving only the FN intact.
Briefly, for the nearly-intact preparation, the body tissues surrounding the cephalic ganglion through ganglion M3 were removed while keeping the remaining body intact. The nerve roots of the tail ganglion and M4 were severed. The preparation was stabilized in a two-chambered dish by anchoring the anterior edge of the body wall to a wax pad with the intact portion of the animal placed in the larger posterior chamber, leaving it free to locomote (see Fig. 2A, schematic). The fluid levels of the posterior chamber could be changed from shallow (< 0.1 cm) to deep (1.0 cm). We observed spontaneous crawling, as it was produced by the nearly-intact leech, or we elicited crawling by gently pulling on the rear of the leech with a small applicator stick to elongate its body, which was shown previously to produce crawling reliably (Esch et al., 2002). The applicator stick was gently applied to the dorsal aspect of the body to elicit swimming.
We employed the sucrose block technique of Masino and Calabrese (2002) to reversibly block signals descending from the cephalic ganglion in nearly-intact and isolated whole nerve cords. The sucrose delivery system was modified by constructing a bowtie-shaped well made of Sylgard (Dow Corning, Midland, MI), which was placed across the connectives between two ganglia (M2-M3). Neural transmission was blocked when the well's saline was replaced with an isotonic sucrose solution (260 mM) that lacked the ions necessary for neural transmission (e.g., Na+, K+, Ca2+). In our hands, we confirmed the effectiveness of the sucrose block by demonstrating that neural transmission was eliminated in sucrose and restored when the well was replenished with normal saline (Fig. 1).
To block chemical synaptic activity within a single ganglion while not disrupting fibers-of-passage, a given ganglion was isolated within a Vaseline well and bathed for 15-20 min in a saline solution containing a high Mg++-low Ca++ (recipe provided above). Similar methods have been used in the leech and crayfish to silence individual segmental CPGs for locomotion (Pearce and Friesen, 1985; Tschuluun et al., 2001). To increase the general excitation of the cephalic ganglion, a 2M K+ acetate/20 mM K+ chloride solution was briefly applied; a similar method was used by Friesen and Brodfuehrer (1984) to depolarize cephalic leech neurons. The solution was washed away within several seconds of application by flow of the saline-containing superfusion system.
Extracellular unit activity was recorded from a chosen DP nerve using suction electrodes (~ 50 μm tip width). Extracellular signals were amplified using a Grass Technologies P15 AC preamplifier (West Warwick, RI) and recordings were digitized (10,000 samples/s) using a MacLab/4s data acquisition system (ADInstruments, Bella Vista, New South Wales, Australia) attached to a Microsoft Windows-compatible personal computer. Activity of DE-3 in all DP recordings served as a monitor of fictive crawling. The intracellular activities of identified MNs were recorded using a Dagan (Minneapolis, MN) IX2-700 amplifier and digitized as described above. Electrodes were filled with a solution of 2 M potassium acetate and 20 mM potassium chloride and had resistances of 40-65 MΩ. For cell labeling, intracellular pipette tips were back filled with a 2 M potassium acetate solution that contained 5% (w/v) Neurobiotin (Vector Laboratories, Burlingame, CA).
MNs were identified by their electrophysiological signatures and the position and size of their somata. Cell identities were confirmed by iontophoretically injecting 5% Neurobiotin (0.5-1 nA) into a given cell for 10-30 min. Tissues were fixed, processed, treated with 1:50 Cy3-conjugated Streptavidin (Jackson ImmunoResearch, West Grove, PA) in PBS for 24 h at 4°C and prepared for viewing according to the methods of Gilchrist et al. (1995).
To determine the temporal properties of fictive crawling, we analyzed bursting activity of DE-3. A grouping of DE-3 spikes was deemed a burst if a minimum of 10 spikes had less than 2 s between them, although most bursts contained dozens of spikes with more than 5 s of separation. Crawl periods were calculated by measuring onsets of sequential DE-3 bursts. If burst start times were not clearly discernable (< 15% of samples), burst midpoints were used to calculate the period. To determine midpoints, DE-3 spike frequency over time was calculated by counting the number of DE-3 spikes per 250 ms time bin and multiplying by 4 bins/s (spikes/s). For each preparation, a spike frequency threshold was calculated by adding one standard deviation of the DE-3 spike frequency to the minimum spike frequency observed during the analysis window. Burst start and end times were when the spike frequency crossed threshold. The midpoint of each burst was calculated by adding 0.5 times the burst duration to the burst start times. Mean crawl periods were determined by analyzing 6-10 consecutive crawl cycles. Intersegmental delays (ISDs) were determined by subtracting posterior DE-3 onsets from anterior ones and dividing by the number of segments separating the two. Mean ISDs were determined by analyzing 5-13 consecutive crawl cycles. Burst midpoints were used to calculate ISDs in lieu of burst start times if burst onsets were not clearly discernable. Intersegmental phase delays, in degrees per segment, were calculated by multiplying the ISD by 360° and dividing by the corresponding crawl period. For ISD and phase delay calculations, the most anterior DP recordings were used as the reference signal.
The Mann–Whitney–Wilcoxon (MWW) nonparametric test was used to assess whether two independent samples came from the same distribution and was performed using the free software package R (R Development Core Team, 2009). All statistical tests were two-tailed with a 95% confidence level (α = 0.05). Sample means were presented ± the standard error of the mean (SEM).
Figure 1 demonstrates the effectiveness and reversibility of the sucrose block. By placing this block below the cephalic ganglion, we determined the necessity of cephalic signals for overt crawling behavior in nearly-intact leech preparations (Fig. 2A,B). Figure 2B depicts the contraction and elongation phase of one overt crawl cycle. Spontaneous crawling behaviors were monitored when the body of the leech was situated in either shallow or deep water, and measured prior to the sucrose blockade (Pre-Block), during the block (Blocked) and after the sucrose was exchanged with normal saline (Post-Block). We observed no spontaneous overt crawling during the Block in either shallow (black bars) or deep water (gray bars) (Fig. 2C). In shallow water the number of spontaneous crawl cycles was significantly less compared to the Pre-Block and Post-Block conditions (p < 0.01) (Fig. 2C). Spontaneous crawling in deep water was limited and only observed during the Pre-Block phase (Fig. 2C). Although these data indicate that descending cephalic signals are necessary for spontaneous overt crawling, they did not establish whether their absence rendered leeches incapable of executing crawling when stimulated. To study this possibility, we attempted to elicit crawling by gently hyperelongating the posterior body with a small applicator stick (Esch et al., 2002). During the Pre-Block phase, in shallow water, crawling was elicited in all animals (n = 8) in 100% of trials (5 out of 5 attempts) (Fig. 2D). During the Blocked phase, it was not possible to elicit a single cycle of crawling in any of the animals tested (Fig. 2D). During this phase, we attempted to elicit crawling by pricking the cuticle with a metal pin just anterior to the tail sucker (n = 3). Under this noxious stimulation, 100% of preparations failed to crawl, although the stereotypic shortening response was exhibited (Kristan et al., 1982; data not shown). During the subsequent Post-Block phase, in shallow water, crawling was stimulated in 5 out of 5 attempts among all animals tested (Fig. 2D). These data provide clear evidence that descending signals from the cephalic ganglion are necessary for the expression of overt crawling in nearly-intact animals.
In this companion study, we determined if the removal of descending signals also prevented swimming. Although a number of previous reports had shown that swimming can occur in the absence of the cephalic ganglion, they involved surgical manipulations that were irreversible (Brodfuehrer and Friesen, 1986a; Cornford et al., 2006). Thus, we analyzed spontaneous and evoked overt swimming behavior using the reversible sucrose block. We conducted our analysis in shallow and deep water using the same animals analyzed in the previous section. Because no leeches swam in shallow water, data only for the deep water condition were plotted (Fig. 3).
Spontaneous swimming was observed in all animals tested in all three conditions (Fig. 3A-C; n = 8). The number of swim episodes (bouts) observed during the blockade of cephalic signals was significantly different from the Pre- and Post-Block phases (p < 0.01) (Fig. 3A), although no significant differences in the total number of swim cycles was found between the three phases (p > 0.5; Fig. 3B). It is noteworthy that the number of swim cycles within an individual swim episode was significantly larger (p < 0.05) during the Blocked phase (127.9 ± 57 cycles per episode) as compared to the Pre- and Post-Block phases (Fig. 3C; 13.9 ± 3.8 and 24 ± 5.9 cycles per episode, respectively).
To examine evoked swimming, we gently stroked the dorsal surface of the leech's body. In nearly all attempts (out of 5), swimming was elicited during the Pre-Block, Blocked and Post-Block phases (Fig. 3D). These data demonstrate that crawling and swimming are differentially dependent on descending cephalic signals, and are consistent with a previous report indicating that the cephalic ganglion may play a predominantly suppressive role in controlling swimming behavior (Brodfuehrer and Friesen, 1986a).
Based on the results of the previous section we hypothesized that descending inputs provide multiple levels of regulation: 1) a system-wide activation and maintenance of the independent segmental crawl oscillators and 2) the setting of the intersegmental phase relationships during each crawl cycle. To understand the specific contributions of descending inputs for oscillator coordination, we needed first to bias the CNS in the direction of crawling while simultaneously suppressing swimming so that these questions could be addressed. This strategy was best achieved by applying 50-100 μM DA (Crisp and Mesce, 2004; Puhl and Mesce, 2008) to isolated whole nerve cords and monitoring fictive crawling via 2 or more DP nerves separated by a given number of segments (Fig. 4A). We again used the sucrose block to reversibly remove cephalic descending signals; in addition, in some preparations we focally stimulated the cephalic ganglion to initiate fictive crawling.
In Figure 4B, we show activity of the crawl-monitor cell, DE-3, across 5 ganglia (DP nerves 10, 12 and 14) in the absence (Block ON) and then presence (Block OFF) of cephalic descending inputs. The isolated whole nerve cord was biased to crawl by applying 50 μM DA, a concentration at the threshold for inducing crawling (Puhl and Mesce, 2008). At the beginning of the recordings shown (Fig 4B, far left arrow), the cephalic ganglion was depolarized with K+ while the sucrose block was in effect; note the lack of any changes in recorded activity. When the block was subsequently removed (Fig 4B, Block OFF), robust and coordinated fictive crawling emerged across all three segments. In a complementary experiment, we stimulated the cephalic ganglion focally with K+ (Fig. 4C, arrow) to demonstrate that cephalic activation is sufficient to induce coordinated fictive crawling when descending inputs are present (note: DE-3 bursts showed no intersegmental coordination prior to cephalic stimulation). This stable and robust metachronal crawling was subsequently abolished in the presence of the block (Fig. 4D).
In isolated whole nerve cord preparations receiving the sucrose block, DA-induced fictive crawl-like bursts were either intermittent or ceased altogether [i.e., stopped crawling (SC) (n = 5)] or exhibited altered intersegmental coordination (n = 13) (Fig. 5A). The removal of cephalic inputs resulted in a spectrum of changes whereby ISDs changed but cycle periods remained fairly constant (Fig. 5B, n = 9/13) or periods changed within and between segments (Fig. 5C, n = 4/13), all of which abolished the metachronal intersegmental phase relationships. During experiments in which recorded crawl bursts could readily be discerned across all recorded segments (e.g., Fig. 5B), intersegmental phase delays were calculated and differed significantly in Pre-Blocked and Blocked conditions (Fig. 5D). Statistical significance between intersegmental phase delays of Pre-Block and Blocked conditions was determined for each preparation. In preparations that showed more pronounced alterations (Fig. 5C), changes in intersegmental coordination were assessed by measuring the period of DE-3 crawl bursts within and across segments. Figure 5E shows changes in the intra- and intersegmental periods for 3 different preparations (i.-iii.) prior to and during the sucrose block. During Pre-Block conditions, all 3 nerve cords revealed isochronicity (i.e., a fairly constant period), and all crawl oscillators within a preparation exhibited a similar crawl period (Fig. 5E, note flat and overlapping lines). Figure 5E (i) illustrates that even when one crawl period transiently deviated from its set point (in Pre-Block), all 3 segmental crawl oscillators shared the same deviation. Upon removal of descending signals, however, the stability of these periods was lost within and across the segmental oscillators (Fig. 5E, Blocked). This block-induced variability in crawl period precludes the formation of appropriate phase delays and intersegmental coordination. In addition, we calculated the correlation of ISDs and period length for 80 crawl cycles (n = 8 preparations) in control leeches and found a coefficient of correlation to be 0.73. This result indicated that the ISD increases with cycle period. For overt crawling, Stern-Tomlinson et al. (1986) reported that ISDs also increase as crawl periods lengthen. Even if one were to argue that the loss of cephalic inputs translates into a decline in excitatory drive and slower crawl period, theoretically ISDs should still be able to adjust to changes in crawl period, but as shown in Figure 5C the ISDs differ over successive cycles, indicating a disruption in coordination.
Finally, we tested whether removal of ascending information from the compound tail ganglion affects the intersegmental coordination of crawling. We used the sucrose block to suppress signals from the tail ganglion. We induced sustained fictive crawling by treating whole nerve cords with 75-100 μM DA and observed that blockade of ascending signals had no significant effects on the ISDs and periods of fictive crawling (data not shown; n = 4).
In our previous study (Puhl and Mesce, 2008), we demonstrated that a single segmental ganglion can express robust and stable fictive crawling when treated with DA. Crawl periods within and across ganglia were similar (Puhl and Mesce, 2008). Based on this information, one might predict that a DA-treated chain of only a few ganglia would reveal more stable crawling; furthermore, as isolated chains they would provide yet another test of the need for cephalic descending signals to coordinate the iterated oscillators.
We observed that a DA-treated chain of 5 ganglia, lacking the cephalic ganglion, never exhibited coordinated crawling with appropriate intersegmental phase delays (Fig. 6A) (n = 5). Although each segmental crawl oscillator gave rise to DE-3 bursting in each segment (Fig. 6A), the intra- and intersegmental timing events never mirrored those in the isolated ganglion or whole nerve cord. Thus, the crawl oscillators appear to be most stable either in complete isolation or in the presence of intact descending inputs. One hypothesis is that any intermediate variants in input perturb the crawl system as a whole. An alternative hypothesis is that the crawl oscillators in chains lack sufficient descending cephalic drive needed to activate DA-biased crawling. To test these ideas further, we removed the end ganglion (in the chain shown in Figure 6A) so that it was freed from all intersegmental inputs (Fig 6B, left diagram). As predicted by our first hypothesis, we found that a strong crawl rhythm emerged (Fig. 6B); it became even more robust and stable over time in isolation (Fig. 6C).
At this juncture, it is clear that the cephalic ganglion plays a vital role in coordinating the segmental crawl oscillators, thus generating the posterior-directed metachronal waves of motor activity during each crawl cycle. Furthermore, DA-induced crawling is most unstable when only a sub-set of intersegmental signals is present, but Figure 6A also hints at the idea that the crawl oscillators influence the cycling frequency of nearby crawl oscillators. To begin to identify shorter-distance intersegmental signals, we asked whether individual crawl oscillators are capable of influencing or driving crawl-like activity in neighboring ganglia.
To address this question, we recorded DP nerve activity in chains of 4-6 ganglia from either the middle (M7-M14) or posterior (M15-M20) body regions, and exploited the ability of DA to activate a single crawl oscillator; we then measured the influence of one oscillator on its segmental neighbors. Figure 7A shows an experiment in which a single ganglion (M13, shaded in gray) was induced by DA to express fictive crawling. The DP nerves of three ganglia were monitored simultaneously: the DA-treated ganglion (denoted G/DA) and the anterior (G/-1) and posterior (G/+1) adjacent ganglia (Fig 7A, diagram at left). Prior to DA treatment, rhythmic DE-3 bursting was absent in all three DP recordings (data not shown). Shortly after DA was applied normal fictive crawling developed in the treated ganglion with a period of approximately 22 s (Fig. 7A, shaded middle trace). In G/-1 (not treated with DA), we observed distinct DE-3 bursts with regular cycle periods of approximately 23 s, resembling fictive crawling (Fig. 7A, top trace). The intra-burst firing rate of DE-3, however, was lower than what would typically be expected in ganglia induced to crawl with DA (Puhl and Mesce, 2008). In G/+1, DE-3 was much more active, exhibiting a rhythmic modulation of its spiking frequency with a period of ca. 22 s (Fig. 7A, bottom trace), but individual pseudo-bursts were not distinct. It is noteworthy that we did not observe normal intersegmental phase delays of DE-3 bursting in any of the DA-treated chain preparations we examined (n = 60). In some preparations, however, the burst frequencies of DA-treated and non-treated ganglia were similar.
Although the DE-3 bursting in G/-1 and G/+1 resembled fictive crawling, several parameters needed to be fulfilled to substantiate this idea quantitatively. One key criterion to fulfill was that the circular (CV; elongation phase) and longitudinal (DE-3; contraction phase) MNs burst ~180° out of phase, as they do during DA-induced fictive (Puhl and Mesce, 2008) and overt crawling (Gray et al., 1938; Stern-Tomlinson et al., 1986). Figure 7B shows a representative example of DE-3 (bottom trace) and CV (intracellular, middle trace) in G/+1, showing that the spiking activity of CV decreased and the cell was hyperpolarized when DE-3 firing was elevated (note: this ganglion was not treated with DA). We also observed a similar anti-phasic relationship of CV and DE-3 in G/-1 (Fig. 7C). Note, however, that in G/-1 the CV hyperpolarization is weak (Fig. 7C), suggesting that local crawl drive may be promoted less in the anterior direction and more in the posterior one.
Figure 7D provides information about the total numbers and percentages of chain preparations in which DE-3 bursts were elicited in DA-treated and neighboring un-treated ganglia; as mentioned crawl-like bursts were more reliably observed in posterior directions. Figure 7E shows the percentage of preparations showing anti-phasic CV and DE-3 activity, and G/+1 again shows this feature more reliably as compared to G/-1. Furthermore, although the DE-3 crawl burst periods were indistinguishable across all segments and conditions (Fig. 7F), the number of DE-3 spikes/burst were significantly elevated in G/+1 and fewer in G/-1 (Fig. 7G). All values were within normal range of fictive crawling (Eisenhart et al., 2000; Puhl and Mesce, 2008).
To determine the potential extent of crawl-like drive to more distant ganglia, we examined DP nerve activity anteriorly in G/-2, and posteriorly in G/+2 and G/+3. Figure 8A shows a record of DE-3 crawl-like bursting in G/+2 (bottom trace). Although many of the bursts across the 3 segments appear co-active, they lack appropriate phase delays. The percentage of preparations exhibiting crawl-like DE-3 bursting is shown in Figure 8B. No bursts resembling true fictive crawling were observed in G/-2 (Fig. 8B).
To test whether the crawl drive from one segmental oscillator to another was based completely on a concatenated series of inputs from one segment to the next, we induced crawling in one ganglion (G/DA) and blocked chemical synaptic activity in the adjacent posterior ganglion with low Ca++-high Mg ++ saline (blocked G/+1) to determine if crawl-like activity could still be communicated to G/+2 (n = 4). Figure 8C shows the control chain of ganglia prior to treatment exhibiting spontaneous (tonic) activity in DE-3 across all 3 adjacent ganglia. In Figure 8D, fictive crawling was induced in G/DA as shown (top trace). At this time, all neural activity was abolished in G/+1 (middle trace) in response to the low Ca++-high Mg ++ solution (Fig 8D). Although no neural activity was recorded from the intervening ganglion (G/+1), the DA-induced crawling in G/DA was able to generate crawl-like bursting in G/+2 (bottom trace), indicating that local intersegmental neurons must span a distance of 2 ganglia or more from their source.
As a first step toward identifying potential local crawl coupling and/or gating cells, we transected subcomponents of the intersegmental connectives to identify the anatomical route that carries the crawl-like drive. Three bundles of neuronal fibers comprise the connectives that link the individual segmental ganglia to each other: paired hemi-connectives and a singular (medial) Faivre's nerve (FN). The hemi-connectives carry the vast majority of interganglionic fibers. The FN has been shown previously to carry an important gating cell for swimming, cell 204 (Weeks and Kristan, 1978), which conceivably could be multifunctional and help gate crawling between adjacent ganglia
To determine which of these structures carries the fibers mediating the local crawl-like drive, we severed either the FN (Fig. 9A, far left diagram) or the hemi-connectives (Fig. 9B, far left diagram) between G/DA and G/+1. A 5-ganglia chain was used for both experiments and fictive crawling in a given ganglion was activated with DA (G/DA in 100 μM). The recordings in Figure 9A demonstrate that after severing the FN, crawl-like drive (bottom trace) was not disrupted (4 out of 5 preparations). In contrast, when the hemi-connectives were severed in 5 different preparations, no crawl-like activity was recorded in G/+1 (bottom trace) (e.g., Fig. 9B).
In this study of leech crawling, we demonstrated that the compound cephalic ganglion plays a critical role in both the activation and coordination of the segmental crawl oscillators; furthermore, we demonstrated that the individual locomotor oscillators provide excitatory drive to each other (Fig. 10 summary diagram). Although a number of computational models have been proposed to explain how long-distance projections might contribute to locomotion (Cohen and Kiemel, 1993; Cacciatore et al., 2000), to our knowledge, ours is the first empirical study to separate the contributions of higher-order and local drive, and intersegmental coordination.
The combination of descending and local control is likely tied to the flexibility inherent to leech crawling and, in some respects, parallels the adaptability observed for more complex locomotor behaviors like walking (Georgopoulos and Grillner, 1989; Lavoie et al., 1995). The dynamic nature of crawling is also based on oscillator independence (Puhl and Mesce, 2008). For productive locomotion, however, the coordination of these unit oscillators must be constrained so that an anterior-to-posterior wave of intersegmental activity is generated (Eisenhart et al., 2000; Puhl and Mesce, 2008). Because leeches do not crawl backwards nor exhibit intermittent positive and negative intersegmental phase shifts, the neural architecture of the animal must be organized to ensure that posterior-directed movements are executed. Our data indicate that information descending from the cephalic ganglion is necessary for the inter-oscillator phase delays that propagate in the anterior-to-posterior direction (Fig. 5). In addition, cephalic descending inputs are sufficient to organize appropriate intersegmental coordination when the individual crawl oscillators are biased to crawl by DA (Fig. 4). This higher-order coordination, likely in combination with local inter-oscillator coupling (Figs. 7 and and8),8), acts to stabilize the intersegmental crawl periods and direct the rearward progression of intersegmental time delays during each phase (contraction or elongation) of the crawl cycle.
In other animal systems, studies have suggested that descending information is involved in oscillator coordination during locomotion. In walking cats (Shimamura and Kogure, 1983; Drew et al., 1986) and rodents (Gordon and Whelan, 2008), for example, descending signals from the midbrain and brainstem have been shown to contribute to interlimb coordination. Drew and Rossignol (1984) demonstrated that microstimulation of the medullary reticular formation in the thalamic cat, in addition to activating locomotion (Mori et al., 1978; Garcia-Rill and Skinner, 1987), can lead to phase-dependent changes in leg muscle activity that is incorporated into ongoing walking cycles.
Locomotor behaviors possessing greater stereotypy often appear less dependent on descending information for oscillator coupling. Although reticulospinal neurons in the lamprey, for example, can drive fictive swimming (McClellan and Grillner, 1984), coordinated swimming can be generated in isolated spinal networks if activated chemically or electrically (Cohen and Wallen, 1980). Such work underscores the importance of local oscillator-to-oscillator coupling.
Figures 7 and and88 demonstrate that individual crawl oscillators activated by DA can promote crawl-like activity in nearby oscillators. We further determined that crawl-like activity originating in a given ganglion extends to at least two posterior ganglia (Fig. 8D). These data suggest that this crawl-like drive is not limited to oscillator-to-oscillator interactions that are concatenated, but must also involve interneuronal signaling spanning multiple segments. These results parallel those in other systems in which well-defined segmental oscillators communicate via intersegmental interneurons projecting through multiple segments (Pearce and Friesen 1985, Tschuluun et al., 2001, Smarandache et al., 2009). Because this drive is stronger in the posterior direction, as evidenced by the increased number of spikes/DE-3 burst [proportional to muscle contraction (Mason and Kristan, 1982)] (Fig. 7G), the anterior-to-posterior propagation of crawl activity is facilitated. Importantly, this inter-oscillator signaling is not sufficient to produce normal intersegmental phase delays, thus local coupling provides only a supporting role for coordination, a feature that contrasts with the systems just discussed ((Pearce and Friesen 1985, Tschuluun et al., 2001, Smarandache et al., 2009).
Asymmetric oscillator coupling is a critical feature governing the directional propagation of motor activity in many locomotor systems; however, most studies have investigated this coupling in tightly-coupled systems, often lacking the flexibility inherent in leech crawling. Swimming behavior in the lamprey and leech, and swimmeret beating in the crayfish, are representative of more tightly coupled oscillators; arising as an emergent property of the locomotor CPGs (reviewed by: Skinner and Mulloney, 1998; Hill et al., 2003). Locomotor behaviors that require more adaptability, such as crawling or walking (Juvin et al., 2005; Borgmann et al., 2009), may preclude the configuration of oscillators that are tightly coupled and whose coordination is predominantly ‘built in’ to the pattern-generating elements.
One computational model of lamprey swimming (Cohen and Kiemel, 1993) suggests that inter-oscillator coupling can be achieved when stronger, shorter-distance, inter-oscillator signals are combined with long-distance projecting signals. This model can generate predictable inter-oscillator phase delays and without the need for precisely timed information stemming from putative long-distance projecting signals. If the inter-oscillator signals are stronger in the posterior direction, propagation will travel only in that direction, as we see in leech crawling. Our study also substantiated the role of a descending long-distance projecting signal.
Similarly, a model specifically for leech crawling predicted the need for a cephalic signal that projects through the length of the CNS (Cacciatore et al., 2000). Also proposed was the posterior-directed excitatory drive, between segments, as substantiated here (see Figs. 7,,88,,10).10). However, several important predictions of this model were not confirmed here. For example, Cacciatore et al. (2000) proposed that only a subset of ganglia contain the rhythm-generating elements for crawling. As we have shown here and in our previous report (Puhl and Mesce, 2008), a complete crawl pattern-generating circuit exists within every segmental ganglion.
These DA-biased crawl oscillators were found to be most stable either in complete isolation (Fig. 6) or when cephalic descending inputs were present (Fig. 4). Dopamine-treated collections of ganglia lacking cephalic inputs exhibited degraded crawl-like activity, supporting the idea that conflicting or inappropriate intersegmental signaling emerges when cephalic influences are absent.
Although we did not specifically indentify the cephalic neurons that activate crawling and coordinate the intersegmental phase delays, several candidates could fulfill such functions. Most notable are the R3b cephalic projection neurons. In particular, cells R3b-1 and R3b-3 can activate or reset the crawl rhythm, and they show rhythmic bursting that is phase-locked with spontaneous overt crawling in nearly-intact leeches (Esch et al., 2002; Mesce et al., 2008). All 3 pairs have been shown to project uninterrupted through the nerve cord. Thus, the long-distance-projecting neurons incorporated into the model of Cohen and Kiemel (1993) do have a biological correlate in the leech, and warrant further study in terms of intersegmental coordination.
The necessity of cephalic input for crawling sharply contrasts with swimming, the other major form of locomotion in the leech. In the nearly-intact animal, swimming was expressed spontaneously (Fig. 3A-C) in the absence of cephalic information. In fact, individual episodes of swimming were extended in the absence of descending input (Fig. 3C). Using the reversible sucrose block, our data reconfirm previous conclusions that the cephalic ganglion plays a largely inhibitory role over the decision to swim (Brodfuehrer and Friesen, 1986a; Brodfuehrer and Burns, 1995; Cornford et al., 2006). Once activated, the dynamics of swimming are fairly stereotypical and can continue with strong intersegmental coordination independent of cephalic input (Kristan and Calabrese, 1976).
Although the leech does possess gating cells for swimming, they are distributed among the segmental ganglia housing the swim oscillators. The most prominent of these swim gating cells is cell 204, which projects through Faivre's nerve (Weeks and Kristan, 1978). By severing this nerve, we showed that cell 204 is not needed to gate crawling (Fig. 9A). The segmental distribution of cell 204, in combination with a series of tightly coupled oscillators, may help to explain why swimming is less dependent on higher-order descending control. Because the swim circuitry has already been studied in exquisite detail (reviewed in: Brodfuehrer et al., 1995; Kristan et al., 2005), future studies of crawling may reveal functional architectures optimized for variable (slow) vs stereotypic (fast) forms of locomotion, features likely conserved across animal groups.
Unlike the ‘trigger’ neurons for swimming (Brodfuehrer and Friesen, 1986b), crawl-activating cells in the cephalic ganglion express gating properties; their activity must be sustained for crawl cycles to be repeated (Esch et al., 2002). In addition to providing the coordinating signal, on-going cephalic activity might also route sensory and efferent signals vital for postural or reflexive adjustments during crawling (Mesce et al., 2008). Although we did not focus on the effects of sensory systems to entrain or reset the crawl oscillators, our preliminary data indicate that mechanosensory inputs can alter their timing (Puhl and Mesce, unpublished observations). In the stick insect, sensory signals from stepping front legs can entrain the ongoing stepping of middle and back legs; however, local load signals in middle and hind legs can override this synchronization (Borgmann et al., 2009). Regardless of the signal source (i.e., sensory or cephalic signals), important parallels emerge: the intersegmental activation and coordination of segmental oscillators in tandem with local modification enable an animal to overlay an array of movements or searching-related behaviors while locomoting; all performed without missing a step.
We are grateful to Kathleen Klukas for her expert technical assistance and preparation of illustrations. We also thank Roger Moon for advice on data analyses and William B. Kristan for his always enlightening discussions. This work was supported by grants from the NSF (IOS-0523959 and 0924155 to K.A.M.) and the NIH (Individual NRSA F31NS060332 to J.G.P).