Mechanisms contributing to the dopamine induction of crawl-like bursting in leech motoneurons

Dopamine (DA) is known to play an impressive and universal role in the control of animal movements, often by exerting prolonged changes in the patterns of activity of locomotor-related neural networks (Chandler and Goldberg, 1984; Flamm and Harris-Warrick, 1986; Kemnitz, 1997; Lapointe et al., 2009; Quinlan et al., 1997; Teyke et al., 1993). In the medicinal leech, for example, DA has been shown to activate the central pattern generator (CPG) underlying crawling behavior (Puhl and Mesce, 2008; Puhl and Mesce, 2010). DA also serves an important function in the selection of competing or incompatible behaviors, exemplified by its suppression of swimming activity in the leech (Crisp and Mesce, 2004). Thus, DA often can act as a behavioral switch, both to activate and suppress competing behaviors (Mesce and Pierce-Shimomura, 2010; Vidal-Gadea et al., 2011). This theme of dual regulation appears to be conserved evolutionarily, as additional examples can be found across multiple levels of animal complexity; for example, DA released by sensory neurons in Caenorhabditis elegans suppresses locomotion so as to prolong grazing when food is available (Sawin et al., 2000) or helps to match locomotor form to either an aquatic or terrestrial environment (Mesce and Pierce-Shimomura, 2010). In the human basal ganglia, DA plays a clinically relevant role in the focused selection of competing patterns of intended movement (Mink, 1996; Smith et al., 1998), as well as regulating total body movement (e.g. hypokinesia associated with DA deficiency).

That DA can turn on and turn off the operation of competing neural circuits poses a number of fascinating questions, including those regarding the cellular mechanisms underlying DA's actions and how individual neurons are modulated in ways that promote or suppress locomotion. The medicinal leech (Hirudo verbana) possesses a number of significant advantages for the study of locomotion: it contains many neurons that have been uniquely identified with respect to their morphology, location, characteristic physiological activity, connectivity and behavioral functions (Kristan et al., 2005). Because of these advantages and the accessibility of its nervous system for electrophysiological studies, multiple behaviors in the leech have been analyzed at the level of individual cells, including a cellular-level understanding of behavioral decision-making (Kristan et al., 2005). Because network computational functions and biochemical signaling pathways are remarkably similar to their vertebrate counterparts (Burrell and Sahley, 2001; Mullins et al., 2011), studies in the leech often have broad instructional value.

In the present study, we set out to understand how DA influences the membrane properties of neurons that might facilitate the production of crawling while suppressing swimming, a competing locomotor behavior. Because it is established that the crawl and swim networks share a number of the same participatory interneurons and motoneurons (Briggman et al., 2005), the issue of how a given neuromodulator can act to bias or facilitate a subset of neurons to perform either a crawl or swim pattern is especially compelling. Our study focuses on the larger and more experimentally accessible locomotor-related motoneurons as a starting point, and the potential roles that DA has on their physiological properties.

Adult medicinal leeches (Hirudo verbana Linnaeus 1758) weighing 2–3 g were obtained from Leeches USA (Westbury, NY, USA) or Niagara Medicinal Leeches (Cheyenne, WY, USA) and housed at 10–15°C in deionized water containing 0.5 g l^–1 Instant Ocean aquarium salts (United Pet Group, Cincinnati, OH, USA) with the pH adjusted to 7.4. Leeches were anesthetized on ice for 10–15 min prior to dissection.

All drugs were obtained from Sigma-Aldrich (St Louis, MO, USA) except 8(p-sulfophenyl)theophylline (SPT), which was obtained from Research Biochemicals (Natick, MA, USA). Drugs were dissolved at room temperature in leech saline (Nicholls and Baylor, 1968) containing (in mmol l^–1): 115 NaCl, 1.8 CaCl₂, 4 KCl, 10 Tris-maleate (pH 7.4). 3-Isobutyl-1-methylxanthine (IBMX) and riluzole were solubilized in DMSO, which was then diluted to the working concentration (0.5 mmol l^–1) in normal leech saline; the final concentration of DMSO did not exceed 0.1%. Theophylline (THP) was placed in HCl prior to dilution in saline and adjusted to pH 7.4. The minimal concentration of all solvents was similar to that used previously and shown not to cause physiological effects during drug delivery in leech salines (Crisp and Muller, 2006). During intracellular recordings from motoneurons, 2 mmol l^–1 MgCl₂ was added to the saline to reduce DA-induced oscillations, whereas in other experiments CaCl₂ was replaced with equimolar concentrations of NiCl₂ or CoCl₂.

Extracellular recordings of motor neuron DE-3 were obtained by differential recording of the dorsal posterior (DP) nerve. The DP nerve was carefully dissected free from muscle fibers, and a watertight silicon well was constructed around the cut end of the nerve. Differential recordings were obtained using a model 1700 differential AC amplifier (A-M Systems, Sequim, WA, USA) and filtered using a LPF202A low-pass Bessel filter (Warner Instruments, Hamden, CT, USA). Noise (60 Hz) and harmonics were eliminated using the Hum Bug Noise Eliminator (Quest Scientific Instruments, North Vancouver, BC, Canada). Raw data were digitized on a DigiData 132x series analog-to-digital converter (Molecular Devices, Sunnyvale, CA, USA) and recorded using PClamp or AxoScope version 9.2 (Molecular Devices).

Specific neuronal somata were identified by their electrophysiological signature, position in the ganglion and morphology when filled with label. Intracellular recordings were obtained with an AxoClamp 2B amplifier in bridge or discontinuous current clamp (DCC) mode and with glass microelectrodes (o.d. 1.0 mm, i.d. 0.75 mm, borosilicate glass; DAGAN Instruments, Minneapolis, MN, USA), pulled to a tip resistance of 25–30 MΩ on a P-97 microelectrode puller (Sutter Instruments, Novato, CA, USA). Electrodes were filled with 3 mol l^–1 K⁺ acetate for recordings and with 10 mmol l^–1 Alexa Fluor 568 (Invitrogen, Carlsbad, CA, USA) in 200 mmol l^–1 KCl for iontophoretic dye injection. Label was injected into neurons (after re-impalement) for 3–5 min using 500 ms square pulses of –1 nA current. When used for DCC, electrode tips were coated with dimethylpolysiloxane prior to use in order to reduce capacitance. Intracellular filtering and digitizing were as described above for extracellular signals.

Enzyme immunoassays (EIAs) for measurement of cyclic adenosine monophosphate (cAMP) concentrations were conducted using a cAMP EIA kit from Cayman Chemical (Ann Arbor, MI, USA). Briefly, chains of ganglia from the second to the twentieth midbody ganglia inclusive were treated for 30 min at room temperature with 0.5 mmol l^–1 IBMX or drug-free saline. cAMP was then extracted by incubating the leech tissue in 0.1 mol l^–1 HCl for 30 min at room temperature and analyzed for cAMP concentration according to the manufacturer's instructions. Lysates were analyzed using absorbance in 96-well plates on a plate reader.

Parameters and methods used to measure crawl-related burst cycle periods and durations were those of Puhl and Mesce (Puhl and Mesce, 2008; Puhl and Mesce, 2010) and typically involved bursts that had clearly discernable onsets. A minimum of six to 10 consecutive bursts were analyzed for each preparation and means were compared across the number (N) of preparations tested. Standard unpaired Student's t-tests and one-way ANOVAs were conducted, as well as Fisher's least significant difference (LSD) post hoc tests, using STATISTICA version 7.1 (StatSoft, Tulsa, OK, USA). Graphs were generated using SigmaPlot 2004 (Systat Software, Chicago, IL, USA). All statistical tests were two-tailed with a confidence level of 95% or greater (α=0.05). Means are reported ±s.e.m., and were calculated using Excel 2003 (Microsoft, Redmond, WA, USA).

Although DA has been convincingly demonstrated to activate fictive crawling, even in a single ganglion (Puhl and Mesce, 2008), the subcellular and biochemical mechanisms underlying such activation have yet to be explored. To begin to identify these mechanisms, we sought to determine whether crawling-related activity was dependent on cAMP second-messenger signaling. This approach was taken as the leech central nervous system (CNS) is known to possess at least one DA₁-like receptor, which has been shown to be positively coupled to adenylyl cyclase activity (Ali et al., 1998). This activity, in turn, increases the cytoplasmic concentration of cAMP, thus enhancing protein kinase A (PKA) activity and the serine/threonine phosphorylation of cellular proteins, including ion channels possibly involved in neuronal bursting. For example, PKA targets a persistent sodium current to increase bursting in the cerebral giant cells of Lymnaea (Nikitin et al., 2006), and increased cAMP levels have been linked to changes in the fictive swimming motor pattern in Tritonia (Clemens et al., 2007). Phosphodiesterases (PDEs) are known to inactivate cAMP (thus lowering PKA activity), enabling phosphatases to dephosphorylate target proteins. Thus, perturbations in cAMP via PDE inhibitors (PDIs) were conducted here to test their contribution towards crawl-related bursting in relevant neurons.

To elevate cAMP levels, we applied the classic PDIs THP and IBMX while recording from previously identified motoneurons (i.e. DE-3) known to exhibit slow bursting during spontaneous and DA-evoked crawling (Puhl and Mesce, 2008). We observed that 4.4 mmol l^–1 THP applied to individual isolated ganglia induced crawl-like bursting in DE-3 (Fig. 1A; 13 of 15 preparations). However, THP is not only an inhibiter of cytoplasmic PDE activity, but also an adenosine receptor blocker. To control for whether THP might induce crawl-like bursting by blocking adenosine receptors, we used the polar adenosine receptor blocker SPT (4.4 mmol l^–1 bath application). [Because of its polarity, SPT mimics the extracellular actions of THP (e.g. adenosine receptor antagonism) but not its intracellular activity (e.g. PDE inhibition).] Blockade of adenosine receptors with SPT was not found to induce rhythmic bursting activity in DE-3 (N=3; Fig. 1A). Similar to the effects of THP, IBMX (500μmol l^–1) produced crawl-like rhythmic bursting in DE-3 (21 of 22 preparations; Fig. 2A). The cycle periods and burst durations of DE-3 induced by PDIs and DA were compared using statistical methods (Fig. 1B). Burst durations and cycle periods were measured and analyzed from a randomly selected subset of ganglia treated with THP (N=10) and IBMX (N=10). DE-3 activity during unevoked (i.e. non-pharmacological or sensory induced) fictive crawling was also included in our analyses. Data for the DA and spontaneous fictive crawl studies were originally obtained by Puhl and Mesce (Puhl and Mesce, 2008); the raw data were reanalyzed here for the quantitative comparisons made. An ANOVA was employed to compare burst durations and cycle periods across all treatments (Fig. 1B). There was no statistically significant difference across treatments with respect to cycle period (F_4,38=0.71, P=0.5), but the effect of treatment on burst duration was statistically significant (F_4,38=5.07, P<0.01). Post hoc Fisher's LSD tests showed no statistically significant differences between burst durations obtained from single ganglia treated with DA (N=10), THP (N=10) or IBMX (N=10), or from intact (i.e. whole) leech nerve cords (N=10) that exhibited spontaneous fictive crawling in normal saline (NS) in the absence of DA (CNS in NS; N=10; Fig. 1B). Only the IK channel blocker clotrimazole (CLO; discussed in more detail later) resulted in a significantly longer burst duration (N=3) after post hoc analysis across the five treatment groups (P<0.05; Fig. 1B).

Fig. 1C shows that THP applied to a chain of ganglia also induced periodic bursting in DE-3 (four preparations). As with DA applied to chains lacking the cephalic ganglion, normal intersegmental phase delays were absent (Puhl and Mesce, 2010).

To determine the degree of cAMP elevation caused by IBMX specifically and at the concentration we used to induce bursting, we relied on an ELISA. Based on an ELISA whereby isolated nerve cords (N=3) were incubated for 30 min with IBMX, we observed a 128% increase in cAMP (from 12.9±4.9 to 29.4±0.96 pmol ml^–1 extract). This measure demonstrated directly that cAMP levels were indeed amplified.

Because DA and PDIs produced similar bursting rhythms, we suspected that they may induce rhythmicity by a similar mechanism or pathway. If, however, DA and PDIs act through distinct and incompatible pathways, co-application of DA and a PDI should disrupt or degrade rhythmicity. Thus we measured 80 bursts from five IBMX-treated ganglia and 57 bursts from four ganglia treated with both DA and IBMX. For statistical comparisons, mean cycle periods from each ganglion were compared because of the unequal numbers of bursts measured per ganglion. We found that IBMX treatment caused a mean cycle period of 20.8±3.1 s (N=5), whereas IBMX+DA treatment caused a mean cycle period of 21.9±8.7 s (N=4); a Student's t-test yielded no statistically significant difference between these two treatment groups (t₇=–0.13, P=0.9). Whether DA and IBMX converge on a common pathway to induce rhythmicity or act through distinct mechanisms that can be co-activated without interfering with one another must await further study.

It is noteworthy that although the DP nerve recording shown in Fig. 1A showed no units firing (showing that THP can induce bursting in quiescent preparations), other preparations exhibited a mean (pre-treatment) DE-3 tonic discharge rate of 5.4±1.4 Hz (N=7; rates averaged over 10 min intervals). Such spontaneous activity is common and has been described elsewhere (Garcia-Perez et al., 2007). We cannot yet account for why a few of our isolated ganglia showed very low levels of spontaneous DP nerve activity while others did not.

Because PDIs can activate DE-3 bursting, we reasoned that adenylyl cyclase inhibition might suppress spontaneous neural activity in the DP nerve. To test this idea, we used MDL-12330a, a known inhibitor of adenylyl cyclase that significantly lowers basal cAMP concentrations in leech nervous tissues (Hunt and Evans, 1980; Biondi et al., 1990). We observed that after a 10–30 min application of 0.1–0.5 mmol l^–1 MDL-12330a, only a few DE-3 spikes per min were observed in isolated ganglia. Furthermore, no bursting was ever observed in either DE-3 or in any other units within the DP nerve from isolated ganglia (N=4) or chains of ganglia (M2–M19; N=3).

THP and other PKA activators, such as DB-cAMP, have previously been shown to be capable of activating fictive swimming (Hashemzadeh-Gargari and Friesen, 1989). In this earlier study, however, fictive crawling had not yet been characterized; thus it was not discussed. Fictive swimming is defined as at least three consecutive DE-3 bursts (typically recorded from the DP nerve) with a cycle period of 0.5–2.0 s (Ort et al., 1974). The earlier experiments performed by Hashemzadeh-Gargari and Friesen (Hashemzadeh-Gargari and Friesen, 1989) used chains of ganglia from M2 to M19. In the present study, only one of seven preparations (M2–M19) exhibited fictive swimming when treated with 4–5 mmol l^–1 THP and only one of eight swam spontaneously in 50–200μmol l^–1 DB-cAMP; we also observed a few swim episodes in two of four chains comprised of seven ganglia treated with THP (data not shown). No swim episodes were observed from isolated single ganglia treated with THP (N=17) or IBMX (N=15), although swim induction in single ganglia is infrequent even in response to potent swim-inducing stimuli (Hocker et al., 2000).

Because elements of the crawl CPG (specifically, the elongation network) are thought to be a component of the swim CPG (Esch et al., 2002; Briggman et al., 2005), it is not surprising that some swimming was observed when the crawl CPG was activated. In contrast, when DA is used to activate the crawl CPG, swimming is never observed because DA is a potent inhibitor of the swim CPG (Crisp and Mesce, 2004). Thus we suggest that THP mimics DA in activating the crawl CPG, but not in suppressing the swim CPG.

As shown in Fig. 1, THP dissolved in normal leech saline (containing 1.8 mmol l^–1 Ca²⁺) produced crawl-like DE-3 bursting. Similar bursting, however, has previously been observed in Ca²⁺-free saline (Angstadt and Friesen, 1991). To determine whether THP-induced bursting requires Ca²⁺ influx, we dissolved THP in saline in which we replaced the Ca²⁺ with equimolar nickel (N=3) or equimolar cobalt (N=2). We observed that THP-induced DE-3 bursting in the absence of Ca²⁺ was not statistically different from THP-induced bursting in the presence of Ca²⁺ with respect to cycle period (t₆=–0.41, P>0.05) or burst duration (t₆=–0.86, P>0.05). For example, THP-induced bursting had a cycle period of 20.4±2.4 s (N=5) in 1.8 mmol l^–1 Ca²⁺ compared with 22.3±4.2 s (N=3) in 1.8 mmol l^–1 Ni²⁺. Burst durations appeared to be slightly longer in 1.8 mmol l^–1 Ni²⁺ (7.4±2.7 s) than in 1.8 mmol l^–1 Ca²⁺ (5.5±0.8 s), but this difference was not statistically significant (t₆=–0.86, P>0.05). THP-induced bursts were not observed when 2 mmol l^–1 MgCl₂ (in addition to 1.8 mmol l^–1 Ca²⁺) was included in the saline (N=6) or when Ca²⁺ was elevated to 8 mmol l^–1 (N=5). It is unclear whether these actions are due to interference with Ca²⁺-dependent mechanisms or are a general effect of divalent cations on cellular excitability (Frankenhaeuser and Hodgkin, 1957).

CLO, briefly mentioned earlier, is a selective blocker of intermediate conductance Ca²⁺-dependent K⁺ (IK) channels (Engbers et al., 2012; Ishii et al., 1997; Joiner et al., 1997), which are known to be inhibited when phosphorylated by PKA (Neylon et al., 2004). We observed that 25μmol l^–1 CLO produced bursting in the DP nerve with a cycle period (20.0±2.2 s) similar to that observed with DA, THP and IBMX (N=3; Fig. 1B). This cycle period was also similar to that previously reported using Ca²⁺ channel blockers (23.2±2.7 s) (Angstadt and Friesen, 1991). However, the mean burst durations observed in CLO were significantly longer (13.5±1.7 s, P<0.05) than bursts observed in DA (5.4±0.6 s), THP (5.5±0.8 s) or IBMX (5.1±0.5 s; Fig. 1B), and longer than previously reported for Ca²⁺ channel blockers (6.6±0.52 s) (Angstadt and Friesen, 1991). Thus, the duty cycle observed in CLO (66%) was considerably longer than that in DA (35%) (Puhl and Mesce, 2008), THP (37%) or IBMX (28%). CLO has also been reported to inhibit Na⁺-K⁺ pump activity (Bartolommei et al., 2008); treatments that inhibit this pump, such as oubain (Angstadt and Friesen, 1991) and low extracellular K⁺ (Angstadt et al., 1998), have been shown to increase the duration of bursts induced by Ca²⁺ channel blockers in leech neurons.

In saline lacking Ca²⁺ ions, many leech neurons show bursting activity that is dependent on sodium influx (Angstadt and Choo, 1996). To test the hypothesis that PDIs activate bursting that is driven by sodium influx, a series of experiments were conducted in which THP or IBMX were co-applied with riluzole, a sodium channel blocker. Riluzole is known to selectively block the persistent, non-inactivating component of the sodium current while sparing the transient component that is required for normal action potential generation (Kononenko et al., 2004; Niespodziany et al., 2004; Urbani and Belluzzi, 2000). In the leech, riluzole has been shown to impair sodium-dependent PIR but does not interfere with spike generation in motoneurons (Angstadt et al., 2011). We found that 200μmol l^–1 riluzole was sufficient to block bursting produced by either 4.4 mmol l^–1 THP (N=3) or 0.5 mmol l^–1 IBMX (N=4). For example, Fig. 2A shows an experiment in which DE-3, in a single ganglion, was induced to burst with IBMX (15 min application). Perfusion of saline containing 200μmol l^–1 riluzole (with 0.5 mmol l^–1 IBMX) for several minutes halted bursting, although these preparations were still able to fire impulses (Fig. 2B). Removing riluzole, by extensively perfusing the ganglion with saline containing 0.5 mmol l^–1 IBMX only, restored bursting (Fig. 2C). Similarly, a 50% reduction of extracellular sodium ion concentration by substitution with equimolar N-methyl-D-glucamine suppressed THP-induced bursting (N=3; data not shown). In contrast, 25μmol l^–1 nifedipine, a blocker of voltage-gated Ca²⁺ channels, did not noticeably affect THP-induced bursting (N=2; data not shown). Together, these data are consistent with the hypothesis that PDIs activate a bursting mechanism in which sodium influx, rather than Ca²⁺, drives burst generation.

In response to DA, motoneurons DI-1, VI-2 and CV rhythmically burst during the elongation phase of crawling, whereas motoneurons DE-3, VE-4 and AE fire out-of-phase during the contraction phase (Puhl and Mesce, 2008). To determine whether DA acts through cAMP-mediated cascades, to promote crawl-like bursting across a wide variety of crawl-related motoneurons, we examined the cells just mentioned in the presence of THP or IBMX. Furthermore, we asked whether the PDIs could substitute, in part, for the network-organizing actions of DA. Because only DE-3 can be effectively monitored via spikes in the DP nerve, intracellular recordings were made from the other motoneurons. In IBMX or THP (0.5 and 4.4 mmol l^–1, respectively), motoneurons DI-1, VI-2, VE-4 and AE fired not only in bursts but also during their correct phases. Cell CV was observed to burst; however, it fired incorrectly and was in phase with DE-3 (N=4; Fig. 3). Because the somata (i.e. recording site) of these motoneurons are electrically inexcitable, their recorded impulses appear small in amplitude. DI-1 and VI-2 are inhibitory motoneurons that become inhibited when the excitatory motoneurons DE-3 and VE-4 are excited during the contraction phase. It is noteworthy that other identified neurons known not to burst during DA-induced or spontaneous crawling were not observed to burst in IBMX or THP; these included the serotonergic Rz cell (N=8), the touch (T) mechanoreceptor cell (N=4) and the nociceptive N cell (N=4). Each of these neurons, in contrast, fires bursts of impulses spontaneously in Ca²⁺-free saline (Angstadt and Friesen, 1991), indicating that the PDIs are more selective in their actions.

To identify the range of ionic currents modulated by DA via cAMP-dependent mechanisms, we needed first to characterize how the membrane properties of the crawl-related neurons change in the context of DA modulation. To address this issue, we exposed ganglia to a 30 min bath application of 75μmol l^–1 DA, which is known to induce fictive crawling (Puhl and Mesce, 2008). We used a DCC and monitored membrane potential trajectories for 3 s following the termination of a 1 s single depolarizing and single hyperpolarizing current pulse. All neurons were held at –40 mV with current injection during the DCC experiments. Neurons requiring more than ±0.25 nA to maintain a membrane potential of –40 mV were considered damaged and data from such cells were not analyzed.

For CV, we found an increase in its afterhyperpolarization (AHP; the area below baseline during 500 ms beginning 50 ms after the end of current injection) that was statistically significant (F_1,17=6.15, P<0.05, N=9 control, N=10 in DA). DA increased the mean area below baseline during the AHP from –577±123 mV ms in NS (N=9) to –1862±284 mV ms in DA (N=10). We also observed a decrease in the duration of its post-inhibitory rebound (PIR) (Fig. 4A,C), which was not associated with a statistically significant reduction of the area under the curve. The area under the curve of the PIR is a measure of peak membrane potential and duration of sustained change during 500 ms beginning 50 ms after the end of current injection. CV showed a nonsignificant decrease in the mean area under the curve from 3199±530 mV ms in NS (N=11) to 2519±517 mV ms in DA (N=11). The similar area measurement was likely due to a more pronounced peak and shorter time course in DA as compared with the more long-lasting PIR observed in NS. The enhanced peak was associated with a statistically significant increase in firing rate during the PIR in CV, from 8.3±1.1 to 17.2±3.3 spikes in an interval of 500 ms following termination of a 1 s current pulse of –2 nA (t₅=–2.61, P<0.05).

DA significantly decreased the duration of the PIR in AE (F_1,7=8.96, P<0.05, N=5 control, N=4 in DA). The mean area under the curve decreased from 1624±328 mV ms in NS (N=10) to 499±61 mV ms in DA (N=10). The effect on its AHP was less pronounced (Fig. 4B,D) and not statistically significant. The area under baseline during the AHP decreased from 1067±81 mV ms in NS (N=10) to 873±286 mV ms in DA (N=10). The PIR in leech motoneurons appears to have an early component and a late component (Angstadt et al., 2011); it is the late component that appears to be diminished following DA treatment in AE. It has been previously reported that DA causes a statistically significant increase in the peak amplitude of the PIR in DE-3, but not the area under the curve (Vallecorsa et al., 2007).

An increase in overall cellular excitability is one way that neuromodulators can exert their influence over CPG activation (Clemens and Katz, 2003; Dai et al., 1998; Fedirchuk and Dai, 2004; Katz, 1998; Katz and Frost, 1997). Based on firing frequency, as an indicator of cell excitability, we found that DA reduced neuronal excitability, but only in the context of current injection. We again used a DCC protocol but one that stepped cells from –2 to +2 nA in steps of +0.5 nA. In normal saline, AE fired impulses tonically at a rate of 5.8±1.5 Hz (N=4), close to its resting firing rate when treated with 75μmol l^–1 DA (6.6±0.7 Hz, N=5). However, when treated with DA, a +1 nA current pulse increased the spike firing frequency of AE only to 12.6±2.1 Hz (N=5) as compared with 18±2.3 Hz (N=4) in normal saline (Fig. 5A). This difference was even greater in CV, where again the resting firing rates were similar in saline and DA (9.0±1.8 Hz control, N=8 versus 6.0±1.6 Hz in DA, N=8; Fig. 5B). A +1 nA current injection increased the firing frequency of CV only to 23.0±3.0 Hz in DA compared with 41.2±3.8 Hz in normal saline. These results are consistent with prior observations that DA exerts a generally inhibitory influence on leech neurons (Ali et al., 1998; Crisp and Mesce, 2004; Sargent, 1975), but helps to enhance their patterned activity.

We next asked whether mimicking DA₁ receptor activation with THP can cause the full range of DA-induced physiological changes in the motoneurons just presented. In contrast to the direct effects of DA, bath application of 4.4 mmol l^–1 THP for 20–40 min did not noticeably enhance the PIR of CV (N=4; data not shown), nor did it mimic the effect of DA on the excitability of CV (N=4; Fig. 5C) or AE (N=6; Fig. 5D). For example, injection of a +1 nA positive current resulted in a firing rate of 18.0±2.3 Hz in normal saline and 20.0±2.2 Hz in THP. In addition, although DA only caused a slight increase in the resting discharge rate of the AE cell (from 4.0±1.5 Hz in normal saline to 6.6±0.8 Hz in DA), THP also caused a much greater increase in resting discharge rate (to 15.2±3.8 Hz). Similarly, the resting discharge rate for CV was 9.0±1.8 Hz in normal saline but 17.8±3.7 Hz in THP (N=4). In contrast, THP diminished or eliminated the PIR in the AE cell (N=6; data not shown). ANOVAs were employed to analyze the effects of DA, THP and NS on the resting discharge rates and slopes of the F–I (steady-state firing rate as a function of injected current) curves shown in Fig. 5C,D. There was a significant effect of treatment on cell CV (F_4,34=5.11, P<0.01), and Fisher's LSD post hoc tests revealed that DA caused a significant reduction in slope compared with NS (P<0.05); there was no significant difference between the effect of THP and NS on slope. THP caused a significant increase in resting discharge rate compared with DA (P<0.05) and NS (P<0.01). Similarly, there was a significant effect of treatment on AE (F_4,22=8.13, P<0.001) and post hoc tests revealed that DA caused a significant reduction in slope compared with NS (P<0.05). THP was not different from either DA or NS with respect to slope, but was significantly different from DA (P<0.001) and NS (P<0.001) with respect to resting discharge rate. These results indicate that DA is likely affecting crawl-related targets by modulating a number of DA receptor-mediated pathways beyond a DA₁-like type.

Here we have demonstrated several cellular mechanisms that may be involved in the DA induction of fictive crawl-related bursting in leech ganglia. Specifically, we have shown that elevating cAMP by inhibiting PDE activity induces robust DE-3 motoneuron bursting that resembles DA-activated fictive crawling. This bursting activity relies on a sodium rather than a Ca²⁺ influx. Furthermore, bursting with a similar cycle period can be induced by pharmacologically inhibiting a Ca²⁺-dependent K⁺ current, specifically, an IK-like current. We have also shown, through intracellular recordings, that DA modulates the intrinsic membrane properties of a variety of crawl-related motoneurons in ways that may facilitate participation in patterned bursting. Not all the actions of DA, however, can be mimicked by inhibiting PDE activities, suggesting the involvement of non DA₁-like receptors not positively coupled to cAMP.

We suggest that DA activates crawling through at least two pathways, only one of which is mimicked by THP. In one pathway, an extrinsic oscillator or timing network is activated that is capable of generating crawl-like bursting in leech motoneurons. This timing mechanism, under experimental conditions, is capable of driving the crawl or swim CPG, but in the presence of DA will induce only crawling, as DA is known to be a potent inhibitor of swimming (Crisp and Mesce, 2004). This timing mechanism may be a separate module of the CPG from that responsible for generating the phase relationships among the motoneurons. Occasionally, in recordings of fictive walking and scratching in cats, some motoneuron pools fail to fire during their appropriate phase in the rhythm, but the cycle period remains unaffected by this deletion, suggesting a dissociation of the timing mechanism from the motoneuron patterning network (Lafreniere-Roula and McCrea, 2005; Rybak et al., 2006). These observations suggest that some CPGs may possess independent timing elements that govern the overall temporal structure of the motor pattern (e.g. cycle period) and patterning elements that govern motor neuron activity within the cycle (e.g. phase relationships).

The second action of DA, which is not readily mimicked by THP, consists of the modification of excitability and afterpotentials of motoneurons. Although this latter function may be related to a general inhibitory function of DA that helps to suppress incompatible motor behaviors (i.e. swimming), modulation of dynamic membrane properties may help motoneurons to maintain firing patterns that follow a rhythmic excitatory drive with fidelity.

Leech neurons that are normally quiescent at rest can be induced to burst through several pharmacological manipulations. For example, in all leech neurons examined to date, a persistent sodium current will drive sodium plateau potentials, resulting in rhythmic bursting unless this inward current is opposed by Ca²⁺-dependent K⁺ currents. This rhythmic bursting activity is evident when leech neurons are treated with Ca²⁺ channel blockers (Angstadt and Friesen, 1991). Ca²⁺-dependent K⁺ and persistent sodium currents have opposing effects on excitability (Burrell and Crisp, 2008; Wu et al., 2005) and contribute to oscillations and bursting in various ways across different neuronal types (Boehmer et al., 2000; Gutfreund et al., 1995; Hu et al., 2002). Our finding that riluzole (a blocker of the persistent sodium current) impairs PDI-induced bursting, in contrast to Ca²⁺channel blockers such as cobalt, supports the hypothesis that DA acts through a cAMP-dependent pathway to activate a bursting mechanism involving sodium-based plateau potentials. A persistent sodium channel has been found to be a target of PKA in Lymnaea (Nikitin et al., 2006), suggesting a potential biochemical connection between cAMP concentrations and sodium plateau generation.

Although enhancing persistent sodium currents or blocking Ca²⁺-dependent K⁺ currents seem like plausible means by which PDIs may activate elements of the crawl oscillator, the effects of DA on CV and AE would be better explained by opposite effects on these currents. Recently, blocking persistent sodium currents with riluzole has been shown to reduce the late component of the PIR and block bursting in cell DE-3 (Angstadt et al., 2011), unlike DA, which attenuates the PIR and promotes bursting. Furthermore, although reduced sodium salines decrease PIR in DE-3, the PIR is eliminated only when sodium is reduced and Ca²⁺ is replaced with nickel (Angstadt et al., 2005) or in solutions containing high concentrations (10 mmol l^–1) of nickel (Angstadt et al., 2011). These observations suggest that a portion of the PIR is Ca²⁺-dependent and that the PIR is mediated by both persistent sodium and nonspecific cation currents.

If DA activates oscillations by unmasking or potentiating a persistent sodium current, one might also expect an increase in excitability and PIR in cells modulated by this mechanism. However, even if DA enhances the persistent sodium current in specific motoneurons, the ability of such a change to increase excitability may be limited by the cell's input resistance, the extent of its electrical coupling and the expression of other currents contributing to the whole-cell I–V (current as a function of voltage) curve (Zhao et al., 2010). Furthermore, DA may diminish PIR in DE-3 cells through combined mechanisms, such as enhancing a Ca²⁺-dependent K⁺ conductance, inhibiting the riluzole-sensitive persistent sodium current or inhibiting a Ca²⁺-dependent nonspecific cation current (Angstadt et al., 2011). Any of these mechanisms could contribute to a decrease in excitability.

The opposing actions of DA₁ and DA₂ receptors were characterized in the mollusc Lymnaea almost 30 years ago (Stoof et al., 1984), and newer cell-molecular and pharmacological studies have suggested the presence of at least three classes of DA receptors in invertebrates (Mustard et al., 2005). In the lobster stomatogastric nervous system, DA has been shown to be an important modulator altering motor pattern expression (Ayali and Harris-Warrick, 1999; Mullins et al., 2011; Vidal-Gadea et al., 2011). The actions of DA are often mediated through multiple and sometimes opposing modulatory effects on neurons within a given circuit. For example, DA enhances an A-type K⁺ conductance in the pyloric dilator neuron (Kloppenburg et al., 1999), but depresses this same conductance in the lateral pyloric neuron (Harris-Warrick et al., 1995). These opposing actions can be explained by the expression of DA₁ (but not DA₂) receptors in the lateral pyloric neuron and DA₂ (but not DA₁) receptors in the pyloric dilator neuron (Zhang et al., 2010).

Although previous evidence suggests the presence of a DA₁-like receptor in the leech (Ali et al., 1998; Colombaioni and Brunelli, 1988; Salzet et al., 1998), the pharmacological properties of this receptor and other subtypes may be distinct from other species. The leech likely has a second class of DA receptors not positively coupled to cAMP. For example, DA but not cAMP inhibit ion transport across leech skin (Milde et al., 2001). Similarly, not all the effects of DA on the crawl circuit can be mimicked by PDIs, as evidenced by the improper phase relationships of DE-3 and CV and the inability of THP to reduce motoneuron excitability. Preliminary studies show that apomorphine, a possible DA₂-like agonist in leech, can induce crawl-like bursting with CV in phase with DE-3 (J. G. Puhl and K.A.M., personal observation). The DA₂-receptor family contains receptors that are negatively coupled to adenylyl cyclase, and although no DA₂-like receptor has yet been characterized pharmacologically in Hirudo, genes for several DA₂-receptor families have been identified in the leech Helobdella robusta (EMBL-EBI accession numbers IPR 001620, IPR 001922 and IPR 002185).

Modulation of crawl motoneurons may reflect two parallel actions of DA. DA modulates membrane properties known to contribute to rhythm generation. After DA treatment, CV shows an enhanced AHP and a higher firing frequency during the peak PIR. Interestingly, no statistically significant increase in AHP was observed in either DE-3 or AE, although DE-3 showed an enhanced peak in its PIR (Vallecorsa et al., 2007). DE-3 and CV innervate the longitudinal and circular muscles, respectively, both of which are body wall muscles that are directly responsible for locomotor movements. In contrast, AE innervates the annular erector musculature, and the role of annular erection in crawling is less clear, though it may aid in traction. Possibly for this reason, AE bursts late in contraction (Eisenhart et al., 2000), giving it a somewhat unique phase relationship with the other motoneurons that may be reflected in its unique pattern of modulatory changes.

It is not yet clear how a DA-induced decrease in motoneuron excitability would facilitate the expression of rhythmic bursting during fictive crawling. Quite to the contrary, it seems that an increase in excitability would be more likely to promote bursting in leech motoneurons. A decreased excitability may, however, contribute to a generalized inhibitory effect of DA that could help to suppress the expression of incompatible behaviors, especially swimming (Crisp and Mesce, 2004). Furthermore, the duration and firing frequencies exhibited by locomotor motoneurons affect both the rate of muscle contraction and the time to peak contraction. Thus, a decrease in excitability may well facilitate the ability of each body segment to transition smoothly from any given body length, enabling the metachronal waves of crawl activity to progress gracefully along the animal's body as it repeatedly elongates and shortens. One overarching lesson gleaned from our studies is that DA has the ability to activate a mosaic of cellular actions. Answers to how each of these pieces complement the other to form a coherent locomotor behavior will certainly benefit from future characterizations of leech-specific DA receptor subtypes.

 
• AE

  excitor of the subcutaneous annulus erector muscle

 
• AHP

  afterhyperpolarization

 
• cAMP

  cyclic adenosine monophosphate

 
• CLO

  clotrimazole

 
• CPG

  central pattern generator

 
• CV

  excitor of the ventrolateral circular muscle

 
• DA

  dopamine

 
• DCC

  discontinuous current clamp

 
• DE-3

  dorsal longitudinal muscle excitor

 
• DI-1

  dorsal longitudinal muscle inhibitor

 
• DP

  dorsal posterior (nerve)

 
• EIA

  enzyme immunoassay

 
• IBMX

  3-isobutyl-1-methylxanthine

 
• NS

  normal leech saline

 
• PDE

  phosphodiesterase

 
• PDI

  phosphodiesterase inhibitor

 
• PIR

  post-inhibitory rebound

 
• SPT

  8-(p-sulfophenyl)theophylline

 
• THP

  theophylline

 
• VE-4

  ventral longitudinal muscle excitor

 
• VI-2

  ventral longitudinal muscle inhibitor

We thank Drs William B. Kristan, Jr and James Angstadt for helpful discussions and insights throughout our study, and Dr Josh Puhl, Ms Windy Lynch, Mr Bilal Alkatout and Ms Nicole Marvin for technical assistance. We are also most grateful to the expert anonymous reviewers whose helpful suggestions greatly strengthened the final version of our manuscript.