Functional consequences of activity-dependent synaptic enhancement at a crustacean neuromuscular junction

The neuronal activity produced by the nervous system controls the muscle contractions that ultimately result in movements of these muscles. The amplitude of the muscle contractions and the actual movements evoked by the motor output are determined by the temporal pattern of the motor neuron discharge and depend considerably on the temporal characteristics of the neuromuscular junction of the receiving muscle. Here, we study the temporal dynamics of facilitation and augmentation at the neuromuscular junction of the stomach musculature of crustaceans and their impact on muscle force production. The muscles of the crab foregut are driven by the rhythmic activity of the pyloric (filtering of food) and gastric mill (chewing of food)central pattern generators in the stomatogastric nervous system (STNS). They receive excitatory innervation from motor neurons in the stomatogastric ganglion (STG). Activity-dependent facilitation and depression act presynaptically on the motor axon terminals and modulate the responses of the muscles (Jorge-Rivera et al.,1998; Katz et al.,1993). Most of the motor neurons in the STG use glutamate to affect the muscles (Hooper et al.,1986; Marder,1976), while some of them make excitatory cholinergic synapses onto the muscles they innervate (Marder,1974; Marder,1976). Stomatogastric muscles do not receive any direct inhibitory innervation (Govind et al.,1975).

The neuronal circuits in the isolated STNS (in vitro) produce a large repertoire of motor outputs under modulatory control(Marder and Calabrese, 1996; Nusbaum, 2002; Nusbaum and Beenhakker, 2002). For example, the gastric mill rhythm shows great variability in firing frequency and temporal distribution of bursts of activity, depending on the modulatory state of the STNS (Nusbaum,2002). It is assumed that an even greater plasticity of motor outputs is present in vivo. In this light, it would be interesting to know how the temporal dynamics of the neuromuscular junctions of the receiving muscles contribute to the response of the muscle during these different motor neuron discharge patterns. In the present investigation, we used the gastric mill muscle gm6 of the crab Cancer pagurus to study the effects of facilitation and augmentation (Maynard and Dando, 1974). In general, facilitation acts on timescales shorter than 1 s, while the effects of augmentation last for several seconds(Zucker and Regehr, 2002). The gm6 muscles are bilaterally symmetric and innervated by a single motor neuron,the lateral gastric motor neuron (LG). During a gastric mill rhythm, the gm6 muscles participate in the protraction of the lateral teeth in the foregut of C. pagurus. The excitatory junction potentials (EJPs) recorded in the gm6 muscles are initially small but are summated and facilitate strongly with repeated stimulation (Jorge-Rivera et al.,1998). While gastric mill rhythms show a great variability in their temporal firing characteristics(Nusbaum, 2002), the effects of these varying motor neuron discharge patterns on the membrane potential and the force production of the muscle are unknown. Therefore, we characterized the electrical response of the gm6 muscles and their force production via a range of different stimulus protocols, including gastric mill rhythm-like stimulations. For the latter, we used a variety of gastric mill rhythms, which we recorded in vitro and in vivo.

Adult crabs Cancer pagurus L. were purchased from commercial sources (Feinfisch GmbH, Neu-Ulm, Germany). Crabs were maintained in filtered,aerated artificial seawater (10–12°C). After anesthetizing the animals on ice for 20–40 min, the dissection of the STNS was done in physiological saline at ∼4°C. C. pagurus physiological saline(Pantin, 1961; Heinzel et al., 1993) had the following composition (mmol l^–1): NaCl, 440;MgCl₂, 26; CaCl₂, 13; KCl, 11; Tris base, 12; maleic acid, 5; pH 7.4–7.6. The Mg concentration in the saline was based on values measured in the hemolymph of C. pagurus(Tentori and Lockwood, 1990). The neurons in the stomatogastric nervous system of C. pagurus and their connectivity and properties are similar to those in C. borealis(Heinzel et al., 1993; Stein et al., 2005).

Dissections were carried out as described(Blitz and Nusbaum, 1997). Experiments were performed on the isolated STNS(Fig. 1) using standard electrophysiology methods as described previously(Bartos and Nusbaum, 1997; Blitz and Nusbaum, 1997). The STNS was pinned down in a silicone elastomer-lined (Elastosil RT-601, Wacker,Munich, Germany) Petri dish and superfused continuously (7–12 ml min^–1) with chilled physiological saline (10–13°C). Extracellular recordings of neuronal activity were obtained by electrically isolating individual sections of STNS nerves from the bath by building a petroleum jelly-based cylindrical compartment around a nerve section. The action potentials propagating through the nerve were recorded by placing one of two stainless steel electrode wires within this compartment. The second wire was placed in the bath as a reference electrode. The differential signal was recorded.

Two methods were used to elicit gastric mill rhythms in the isolated STNS.(1) We stimulated the dorsal posterior oesophageal nerve (dpon; Fig. 1, top left), a sensory nerve, with 10 stimuli trains (train duration = 6 s, pause between trains = 4 s) at a stimulus frequency of 15 Hz. Stimulations were performed by applying pulses of 0.5 ms duration with a Master-8 stimulator driven by a micro 1401 AD/DA board (CED, Cambridge, UK). This consistently elicited a gastric mill rhythm that outlasted the stimulation for 20–30 min(Beenhakker et al., 2004; Stein et al., 2005). (2) We tonically stimulated the inferior oesophageal nerve (ion; Fig. 1) with a stimulation frequency of 20 Hz. For these experiments, all connecting nerves between the STG and the commissural ganglia were transected(Fig. 1, top right). During this stimulation, a gastric mill rhythm(Bartos and Nusbaum, 1997; Stein et al., 2005) was elicited. The gastric mill rhythm was monitored by the activity of the lateral gastric neuron (LG; Fig. 1). LG was recorded on the lateral gastric nerve (lgn; Fig. 1). The gastric mill cycle period was defined as the interval between the onsets of two consecutive impulse bursts of LG.

The dorsal carapace and the hypodermis above the STNS were opened. All dissections were done in physiological saline with antibiotics (Gentamicin 50μg ml^–1; Sigma) at a temperature of ∼4°C. Handmade extracellular cuff electrodes were attached to the lateral ventricular nerve(lvn) and fixed with a holder to the remaining carapace. LG activity was monitored on the lvn. After implanting the electrodes, the carapace was sealed with a plastic cover and dental cement (Protemp II, ESPE,Seefeld, Germany) and the animals were placed back in the tank. Only measurements from animals that survived more than 2 days were taken into account.

In experiments in which muscle responses were recorded, the isolated STNS was transferred to the Petri dish (Fig. 1). The gastric mill muscles gm6 were left attached to the motor nerves (lgn, lvn). The lvn was transected between the lgn branch and the STG, and all anterior parts of the STNS were removed. A petroleum jelly-based cylindrical compartment was built around the stump of the lvn, and a stimulation wire was placed within the compartment. The second stimulation wire was placed in the bath. The lvn was stimulated at the threshold for LG axon spike activation. The success of the stimulation was monitored by the intracellular gm6 muscle response (Fig. 1, right muscle). For this, gm6 muscle fibers were impaled with microelectrodes(15–25 MΩ) filled with 0.6 mol l^–1K₂SO₄ plus 0.02 mol l^–1 KCl. Intracellular muscle responses were recorded and filtered using an NPI NEC 10L amplifier (NPI, Tamm, Germany) in bridge mode. Various stimulus protocols were used. For measuring the facilitation of EJPs in the gm6 muscle, we applied paired-pulse stimuli (duration: 0.5 ms) with interstimulus intervals ranging from 100 ms to 6400 ms. For determining the facilitation during gastric mill-like activity, we applied trains of stimuli. Each train consisted of 10 stimulus pulses. Stimulation frequencies ranged from 5 Hz to 30 Hz. In some experiments, we additionally applied a test pulse after each stimulus train with a fixed delay of 500 ms. In order to detect the effects of augmentation,we repeated these trains 10 times with inter-train intervals ranging from 2 s to 32 s. Additionally, in some experiments, we applied test pulses after the end of the last stimulus train, with delays ranging from 2 s to 18 s. To test the presence of augmentation during gastric mill rhythms, stimulation events were extracted from previously recorded gastric mill rhythms (in vitro and in vivo) by using a script for Spike2 (CED, Cambridge,UK; script is available at http://www.neurobiologie.de/spike2).

In some experiments, gm6 muscle force instead of the intracellular response was recorded. Force recordings were obtained with a force transducer (Swema SG4-25, Farsta, Sweden) connected to a balanced bridge. One of the gm6 muscle insertions was pinned down in the dish while the other was attached to the force transducer using a clamp (Fig. 1, left muscle). Afterwards, the muscle was stretched to its resting length. Forces were calibrated using a regression line that was obtained by attaching weights between 0.2 g and 20 g after each experiment.

The facilitation index F was calculated in different ways,depending on the experiment. (1) In experiments with two-pulse stimulations,facilitation was calculated by the ratio of the second EJP to the first EJP.(2) In experiments with train stimulations, the ratio of the current EJP to the first EJP or the ratio of last EJP to the first EJP was used (within-train facilitation). For determining EJP amplitudes within trains of stimuli, the amplitude of the previous EJP at the time of the peak of the current EJP was subtracted from the amplitude of the current EJP to eliminate the effects of summation (Katz et al., 1993). Since EJP amplitudes were in the range of few mV, they were not corrected for non-linear summation effects. Facilitation values are thus not intended to give an accurate measurement of transmitter release. The maximum possible facilitation was estimated by fitting facilitation values with exponential decay functions and calculating the facilitation at time zero. The time course of decay of the facilitation was obtained from these exponential decay functions.

When several trains of stimuli were applied, the ratio of the current EJP to the last EJP of the first train (augmentation index A) was used to estimate the amount of augmentation. During gastric mill-like stimulations both the ratio of the first EJP of each train to the first EJP of the first train, and the ratio of the last EJP in each train to the last EJP of the first train were used. Since EJP amplitudes were not corrected for non-linear summation effects, augmentation values were only used to assess the long-lasting effect on the electrical response of the muscle. They do not represent an accurate measure for transmitter release. Similar to facilitation, the maximum possible augmentation and the time course of decay of the augmentation index was estimated by fitting augmentation index values with exponential decay functions.

Standard intracellular and extracellular recording techniques were used in this study (Stein et al.,2005). Extracellular nerve activity was recorded, filtered and amplified through an amplifier from AM Systems (Model 1700, Carlsborg, WA,USA). Data were recorded onto computer using Spike2 (CED) and a micro 1401 AD board (CED, Cambridge, UK). Data were analyzed using the Spike2 script language. Individual scripts are available at http://www.neurobiologie.de/spike2. Final figures were prepared with CorelDraw (version 12.0 for Windows) and PowerPoint (Microsoft). Graphics and statistics were generated using Excel(Microsoft) or Plotit (Scientific Programming Enterprises, version 3.2). Statistical tests used to analyze data were student's t-test and paired samples t-test. Data are presented as means ± s.d.; N, number of animals; n, number of trials. For all statistical tests, significance with respect to control is indicated on figures: *P<0.05, **P<0.01.

We used a two-pulse paradigm to characterize the facilitation of the gm6 neuromuscular junction. The lateral ventricular nerve (lvn), which contains the axon of LG, was stimulated with series of trains of two pulses with different interstimulus durations and inter-train durations of 10–30 s. Starting with 50 ms, interstimulus intervals were doubled after every five trains until they reached 6400 ms. EJPs showed facilitation with short interstimulus intervals. Fig. 2A (left) shows an averaged recording of EJPs with a stimulus interval of 200 ms. The second EJP increased more than twofold in amplitude. Fig. 2A (right) shows that with increasing interstimulus intervals the amplitude of the second EJP decreased. Fitting the facilitation index F(Fig. 2B; see Materials and methods) revealed a maximum facilitation of 2.23 and a time constant of decay of 0.54±0.11 s (N=9), which indicates a quick relaxation of the amplitude of the second EJP.

During a gastric mill rhythm LG is usually firing bursts of action potentials with discharge frequencies between 20 Hz at the beginning of the burst and 5 Hz at the end of the burst(Beenhakker et al., 2004). We therefore recorded the response of the gm6 muscle to a train of 10 stimuli with stimulus frequencies of 5 Hz, 10 Hz and 20 Hz. After each train we applied a test pulse with a delay of 500 ms. During this delay, the membrane potential in all recordings returned to its baseline value. This allowed a measurement of the amplitude of this subsequent EJP without interference of summation. The average resting potential was –70.00±5.94 mV(N=10). As is obvious from Fig. 2C (left), individual EJPs during a 5 Hz train facilitated. The following test pulse also elicited an EJP that was clearly higher in amplitude than the first EJP of the train (first EJP 0.36±0.28 mV, N=10;test EJP 1.89±1.31 mV, N=10; significantly different P<0.01). No summation was seen with 5 Hz in five of the nine recordings. With 10 Hz stimuli (Fig. 2C, middle), the membrane potential did not return to its baseline after each EJP (summation). At the same time, EJPs facilitated. Facilitation was stronger than with 5 Hz stimuli, as is obvious from the response to the single test pulse after the 10 Hz train. The amplitude of this EJP exceeded the amplitude of the 5 Hz test EJP (2.52±1.36 mV, N=10;significantly different from 5 Hz test EJP, P<0.01). Similarly,with 20 Hz trains, a stronger summation and facilitation than with 5 Hz and 10 Hz stimuli was achieved (Fig. 2C, right). The amplitude of the test EJP (3.95±2.28 mV, N=10) was significantly higher than the amplitudes of the test EJPs elicited after 10 Hz stimulation (P<0.05) and 5 Hz stimulation(P<0.01).

Since we were interested in how facilitation contributed to the observed membrane potential of the muscle, we needed to separate the effects of summation from those of facilitation within a single train of stimuli. Therefore, we subtracted the amplitude of the previous EJP at the time of the peak of the current EJP from the amplitude of the current EJP(Katz et al., 1993). This calculation requires a similar time constant of decay for EJPs with different amplitudes. We verified this requirement for every recording either by using the test EJPs or by using a 5 Hz train of stimuli in which facilitation, but not summation, was present (Fig. 2D, left). We superimposed either the test EJPs or all EJPs in the 5 Hz train (Fig. 2D, middle),which showed the different amplitudes obtained during the train and also that the membrane potential always returned to its baseline after each EJP. We then normalized each EJP to its peak value and superimposed them again(Fig. 2D, right). This revealed that despite their different amplitudes, all EJPs possessed a similar shape and thus also a similar rate of decay and time constant of decay. The time course of the EJP decay was fitted by an exponential decay function. When the time constant of decay of EJPs with different amplitudes was compared within the same muscle fiber of the gm6 muscle, no significant difference was found(P>0.2, N=10 tested animals). The average time constant of decay for EJPs was 43.89±17 ms (N=10).

When a single train is analyzed with the calculations mentioned above, the amplitude of each EJP can be calculated independently from summation. In addition, the summated baseline potential can be given by subtracting the calculated EJP amplitude from the peak of the EJP. In Fig. 2E (left), the EJPs of a 20 Hz train were subdivided in baseline potential (white circles) and EJP amplitudes (black circles). Each peak within this train can thus be described by the baseline potential and the amplitude of the EJP(Fig. 2E, right). This permitted the measurement of within-train facilitation during gastric mill-like stimulations.

With all stimulus frequencies, EJP amplitudes increased with the number of stimuli within the train (the number of the elicited EJP; Fig. 2F–I). With 5 Hz stimulation, EJP amplitudes increased significantly from 0.40±0.07 mV for the first EJP to 4.30±0.82 mV for the tenth EJP (N=20; P<0.001; Fig. 2F). Similarly, with 10 Hz stimulation, the tenth EJP (5.82±0.97 mV) was significantly larger than the first EJP (0.56±0.17 mV, N=20, P<0.001, Fig. 2G). After the tenth stimulus with 20 Hz stimulation, EJP amplitudes were significantly larger than at the beginning of the stimuli (first EJP,0.61±0.22 mV, tenth EJP 6.53±1.07 mV, N=20, P<0.001; Fig. 2H).

EJPs facilitated more strongly with higher intratrain stimulus frequencies. Due to the shorter interstimulus intervals, maximum EJP amplitudes were reached earlier with higher stimulus frequencies, as shown in Fig. 2I, in which EJP amplitudes were plotted against the time of the EJP in the train. The average EJP amplitude of the tenth EJP with 20 Hz stimulation was significantly larger than with 10 Hz stimulation (for values see above, P<0.05) and the latter was significantly larger than the average EJP amplitude of the tenth EJP with 5 Hz stimulation (P<0.001).

To test whether higher stimulus frequencies were capable of eliciting even higher EJP amplitudes, we additionally used 30 Hz stimulus frequencies(interstimulus interval 33 ms, N=10). With 30 Hz stimulation, EJPs also increased significantly from 0.56±0.16 mV (first EJP) to 6.60±0.94 mV (tenth EJP, N=10, P<0.001), but the maximum amplitudes of the tenth EJP were not different from those obtained with 20 Hz stimulation (P>0.2, N=10).

As a measure for within-train facilitation we used the ratio of the amplitude of the tenth EJP to the amplitude of the first EJP in the train(facilitation index F). The average facilitation index at stimulus frequencies of 5 Hz (F=10.87±1.83, N=20) was significantly smaller than at 10 Hz (F=14.19±2.92, N=20, P<0.001). With 20 Hz stimulation, F was significantly larger than with lower frequencies(F=15.88±2.95, N=20, P<0.05). With 30 Hz stimulations, however, no further increase in F in comparison to the 20 Hz stimulations was obtained (F=16.06±2.80, N=10, P>0.2), indicating a maximum possible facilitation factor of around 16 during a single stimulus train.

The two-pulse protocol experiments indicated that the facilitation of the EJPs in the gm6 muscle decayed with a time constant <1 s. Facilitation contributed to the gm6 muscle membrane potential during trains of stimulation with interpulse intervals in the range of the normally occurring firing of the LG motor neuron. During a gastric mill rhythm, however, LG is active in series of bursts with various firing frequencies and interburst intervals, depending on the neuromodulatory state of the nervous system(Nusbaum, 2002). To investigate the effects of such motor neuron activity on the muscle membrane potential, we stimulated LG with 10 consecutive trains of stimuli with various stimulus frequencies (5 Hz, 10 Hz, 20 Hz) and an inter-train interval of 4 s. We found that EJP amplitudes within single trains increased with the number of stimulus trains that were applied (Fig. 3A), although the 4 s interval that was used clearly exceeded the time constant of decay of the facilitation described in the paired-pulse experiments. Also, the most depolarized peak of the membrane potential during the train increased with the number of trains. This augmentation of EJP amplitudes was most obviously seen for the test EJP, which was elicited 500 ms after the end of each stimulus train. We compared the amplitudes of the test EJPs of the first and the tenth stimulus trains at the different stimulus frequencies. With 5 Hz intratrain stimulus frequency, the amplitude of the test EJP increased significantly from 2.16±1.39 mV after the first train to 3.44±2.12 mV after the tenth train (P<0.01, N=11). With 10 Hz stimulus frequency, the amplitude of the test EJP after the tenth stimulus train (4.38±2.33 mV) was significantly increased in comparison to the test EJP after the first stimulus train(2.79±1.47 mV, P<0.01, N=11). Similarly, with 20 Hz intratrain stimulus frequency, the test EJP after the tenth train(6.17±3.23 mV) was significantly larger than the test EJP after the first train (3.81±2.18 mV, P<0.01, N=11).

To test the time course of EJP augmentation, we stimulated LG with 10 consecutive trains of stimuli with a fixed stimulus frequency of 20 Hz and varying train intervals (2 s, 4 s, 8 s, 16 s, 32 s). In Fig. 3B we show that the amplitudes of the test EJPs increased with repeated stimulus trains. Here, EJP amplitudes were normalized to the amplitude of the test EJP after the first train. The augmentation of the test EJPs depended on the interval between trains. With shorter intervals, EJP amplitudes reached higher values after 10 stimulus trains than with longer inter-train intervals(Fig. 3B). The analysis of the EJP amplitudes (Table 1)revealed that stimuli with inter-train intervals of up to 16 s showed a significant enhancement of the test EJP after the tenth stimulus train in comparison to the test EJP after the first stimulus train. With inter-train intervals of 32 s, EJP amplitudes did not show a significant increase. With intervals shorter than 8 s, each successive shorter interval elicited significantly higher EJP amplitudes (Table 1). EJP amplitudes with intervals of 8 s and 16 s, as well as 16 s and 32 s were not significantly different. However, EJPs with inter-train intervals of 8 s were significantly larger than with 32 s intervals(Table 1).

Like the test EJPs, all other EJPs within the stimulus train showed augmentation and reached higher amplitudes with repeated stimulus trains. This is shown in Fig. 3C, in which the amplitudes of the first, fifth and tenth EJP in each successive stimulus train were normalized to the amplitude of the tenth EJP of the first train (20 Hz stimulation, 4 s inter-train interval). EJPs showed augmentation and reached maximum amplitudes after the fifth train. We tested the dependence of this augmentation on inter-train intervals by measuring the facilitation of the tenth EJP in each stimulus train with different inter-train intervals(Fig. 3D, 20 Hz stimulation). Data were normalized to the last EJP of the first train. Higher augmentation was reached with decreasing inter-train intervals. Up to inter-train intervals of 16 s the amplitude of the tenth EJP of the tenth stimulus train was significantly higher than that of the first stimulus train(Table 2). We compared the maximum augmentation (A) that was obtained at different inter-train intervals by calculating the ratio of the last EJP in the last (tenth) train and the last EJP of the first train with 20 Hz stimuli(Fig. 3E). Fitting the maximum augmentation with an exponential decay function revealed a time constant of decay of the augmentation of 4.47±0.75 s (9<N<11) that clearly exceeded the time constant of facilitation obtained with two-pulse stimulations. The maximum augmentation as revealed by the fit was 2.90(9<N<11).

As a second measure for the time course of decay of the augmentation, we used the decrement in amplitude of EJPs that were elicited at different delays after the end of a series of 10 stimulus trains (N=14, 20 Hz stimulation). EJPs were elicited at delays of 2, 4, 6, 8, 10, 12, 14, 16 and 18 s after the last stimulus train (Fig. 3F). EJP amplitudes increased significantly in amplitude with delays shorter than 18 s (Fig. 3G, Table 3) in comparison to single EJPs elicited prior to train stimulation (arrow in Fig. 3G). Increasing delays caused significantly smaller EJP amplitudes(Fig. 3G, Table 3). Using an exponential function to fit these data revealed a time constant of decay of 3.21±0.19 s (N=14), which was significantly larger than the time constant of decay obtained for facilitation (P<0.01).

While the within-train facilitation of EJPs depended on intratrain stimulus frequency (Fig. 2I),augmentation (which built up over the course of several stimulus trains) was independent of intratrain stimulus frequency. No significant difference was found when comparing the augmentation of the last EJPs in the tenth stimulus train at different intratrain stimulus frequencies and 4 s inter-train intervals. Stimulations with 5 Hz revealed an augmentation index of 1.69±0.26 (N=11), which was not different from the augmentation index obtained with 10 Hz (1.72±0.26, N=11) and 20 Hz (1.78±0.36, N=11, P>0.2 for all frequencies).

The long-lasting effect on each EJP was best seen when all EJPs were normalized to the last EJP of the first train(Fig. 3H) and plotted over the EJP number within each train. Not only did the EJPs facilitate within each train, they also showed augmentation in subsequent stimulus trains. EJPs within the stimulus train reached a maximum augmentation already after the third subsequent stimulus train (20 Hz, 4 s inter-train interval).

The long-lasting effect on the gm6 muscle not only influenced the amplitudes of the EJPs within the train, but also the time course of the build-up of the facilitation within the trains(Fig. 3I, 20 Hz stimulation, 4 s inter-train interval). When the within-train facilitation index was calculated (EJP amplitudes were normalized to the last EJP of each particular stimulus train) and plotted over the EJP number within these trains, it was obvious that the rise of the facilitation index within the first train had a different time course than all subsequent trains. This indicated that successive stimulus trains led to a quicker rise of the within-train facilitation.

Overall, these results show that besides the fast facilitation there was augmentation affecting the electrical response of the gm6 muscle. This long-lasting effect depended on inter-train interval, but not on intratrain stimulus frequency. It quickened the increase of EJP amplitudes within a stimulus train.

Gastric mill rhythms have mostly been characterized in vitro, that is, in the isolated STNS (Nusbaum,2002). The activity of different descending neuromodulatory projection neurons elicits a variety of gastric mill rhythms with different motor neuron discharge patterns. The patterns differ in the period of the rhythm, the activity of the motor neurons (like LG) and in interburst interval(the time inbetween two bursts of the motor neurons). In contrast to the in vitro situation, the characteristics of spontaneously occurring or feeding-induced gastric mill rhythms in vivo are unclear.

Facilitation and augmentation should contribute to the electrical response of the gm6 muscle during gastric mill rhythms in vitro and in vivo. We tested this hypothesis by recording different types of gastric mill rhythms in the isolated STNS (in vitro) and from freely moving animals in the tank (in vivo). We then replayed standardized versions of these rhythms in order to stimulate LG while we recorded from the gm6 muscle. `Standardized' here refers to the fact that in these stimulations an average firing pattern with averaged burst duration and interburst interval was used instead of the somewhat varying original recordings.

We elicited two different types of gastric mill rhythms. (1) With stimulation of the dorsal posterior oesophageal nerve (dpon; Beenhakker et al., 2004; Stein et al., 2005), a sensory nerve, we started a gastric mill rhythm that depended on the combined activity of two projection neurons, the modulatory projection neuron 1 (MCN1) and commissural projection neuron 2 (CPN2) (N=15). An example recording of this gastric mill rhythm is shown in Fig. 4A (top) as an extracellular recording of the lateral gastric nerve lgn. (2) After transecting the commissural ganglia and thus all descending projection neurons in these ganglia, we selectively stimulated MCN1 with an extracellular stimulation of the inferior oesophageal nerve (ion; Bartos and Nusbaum, 1997). This elicited a gastric mill rhythm (N=20, Fig. 4A, bottom, 20 Hz stimulation frequency) that relied on MCN1 activity only and which differed in its characteristics from the dpon elicited gastric mill rhythm (see also Beenhakker et al., 2004and Bartos and Nusbaum,1997).

We recorded the lvn in intact animals (N=20). The activity of LG was monitored continuously for the following days. The obtained gastric mill rhythms showed a great variability in period, burst duration and interburst interval. Fig. 4Bshows three recordings that represent the obtained spectrum of gastric mill rhythms. The weakest and slowest gastric mill rhythms are exemplified by the top recording. LG usually fired with discharge frequencies (intraburst firing frequencies) of around 10 Hz; the period of the rhythm clearly exceeded 10 s and LG burst durations were around 1 s. The middle trace shows a recording of a gastric mill rhythm that was stronger than the weak gastric mill rhythms in terms of LG firing frequency (about 15 Hz). These intermediate rhythms had periods between 4 and 10 s and LG burst durations between 2 and 4 s. The bottom recording shows a gastric mill rhythm representing the strongest rhythms obtained in vivo, with LG intraburst firing frequencies around 15 Hz and short interburst intervals (below 2 s).

For measuring the response of the gm6 neuromuscular junction to this spectrum of rhythms, we selected five rhythms, two recorded in in vitro preparations, and three recorded in vivo. The two selected in vitro rhythms represented gastric mill rhythms obtained after dpon and during ion stimulation, respectively. The other three rhythms represented the weak, intermediate and strong in vivorhythms. We averaged LG interspike intervals and burst durations as well as interburst intervals for 10 cycles of these rhythms and created stimulus trains from these averages (Table 4). We then used 10 subsequent average stimulus trains to activate LG while we recorded the gm6 muscle response.

The most obvious difference between the electrical responses of the gm6 muscle to stimulation with the different gastric mill rhythms was the peak amplitude of the depolarization that was reached during the single trains(Fig. 4C,D). The strong in vivo gastric mill rhythm elicited the strongest response in the gm6 muscle. The weakest response was elicited by the weak in vivo gastric mill rhythm. The in vitro gastric mill rhythms elicited intermediate responses, indicating that both in vitro rhythms were well within the range of the in vivo rhythms. Summation and strong facilitation was seen in all recordings. While mainly facilitation determined the peak amplitude during stimulations with the weak in vivo rhythm, both summation and facilitation contributed to the maximum depolarization in all other stimulations.

Augmentation was present in all recordings. In Fig. 4C,D the first and the tenth stimulus trains of the different gastric mill stimulations are shown. The long-lasting effect on the membrane potential was obvious in all recordings and is indicated for the first EJP and the peak amplitude (dotted lines). The peak amplitude increased from the first to the tenth train in all rhythms.

To quantify and to better compare the effects of augmentation on EJP amplitude during the different stimulations, we measured the amplitudes of the first and the last EJPs of the first and the tenth stimulus train(Table 5). Additionally, we calculated the augmentation index A of these EJPs over the course of the stimulation. (1) ion elicited in vitro gastric mill rhythm. The amplitudes of the first EJP (Fig. 5A) and of the last EJP (Fig. 5B) of the last stimulus train were significantly increased in comparison to those of the first stimulus train (N=6, P<0.01). The progression of the augmentation index A during successive train stimulation is shown in Fig. 5C for the first EJP and in Fig. 5E for the last EJP. The first EJP reached a maximum A of 5.41±1.49. Maximum A of the last EJP was 1.37±0.07. (2) dpon elicited in vitro gastric mill rhythm. The first EJP of the tenth stimulus train was significantly larger than the first EJP of the first stimulus train(Fig. 5A, P<0.01, N=6). A similar result was found for the last EJP(Fig. 5B, P<0.01, N=6). The first EJP showed a maximum A of 15.98±8.20(Fig. 5C) and the last EJP of 1.29±0.08 (N=6; Fig. 5E). (3) Weak in vivo stimulation. The amplitude of the first EJP did not increase significantly with successive train stimulation(Fig. 5A, N=6, P>0.2), indicating that augmentation did not affect the first EJP during this stimulation protocol. In contrast, the last EJP of the tenth stimulus train had a significantly higher amplitude in comparison to the first train (Fig. 5B, N=6, P<0.01). The first EJP reached a maximum A of 1.56±0.33(Fig. 5D), the last EJP of 1.49±0.21 (Fig. 5F). (4)Intermediate in vivo stimulation. Both first EJP and last EJP of the tenth stimulus train possessed a significantly larger amplitude than those of the first stimulus train (Fig. 5A,B, N=6; first EJP, P<0.01; last EJP, P<0.05). Maximum A of the first EJP was 15.62±8.25(Fig. 5D). Maximum A of the last EJP was 1.32±0.19 (Fig. 5F). (5) Strong in vivo stimulation. The amplitude of the first and the last EJP increased significantly from the first stimulus train to the tenth (Fig. 5A,B, N=6, P<0.01 for the first EJP and P<0.05 for the last EJP). Maximum A values were already reached after the third(Fig. 5D) and the second stimulus train (Fig. 5F),respectively. The first EJP reached an A of 32.87±22.30 (N=6),the last EJP of 1.35±0.42 (N=6).

Augmentation also affected the time course of EJP facilitation within individual stimulus trains. We compared the within-train facilitation of the first and the tenth stimulus trains with the different gastric mill-like stimuli by normalizing EJP amplitudes to the last EJP of each individual train. The strongest effect could be seen with the strong in vivostimulation (Fig. 6A). While EJPs nicely facilitated in the first stimulus train, they did not show facilitation in the last stimulus train. This was due to the fact that EJPs had already facilitated and augmentation caused them to maintain their high amplitude. In contrast, with weak in vivo stimulations, no significant difference was found in the intratrain facilitation of the first and tenth stimulus train (Fig. 6B). Here, EJPs continued to facilitate even in the tenth stimulus train, because the first EJP of each train was not significantly increased by augmentation (compare to Fig. 5A,C). With intermediate in vivo stimulations(Fig. 6C), EJPs in the tenth stimulus train showed within-train facilitation, but not as strongly as with weak in vivo stimulation. The first EJP of the tenth train already had a larger amplitude than the first EJP of the first stimulus train. The dpon (Fig. 6D) and ion (Fig. 6E) in vitro stimulations affected the intratrain facilitation in a way that was in between the weak and intermediate in vivo stimulations. Both stimulations caused EJPs to facilitate even in the tenth stimulus train,although the first EJP of the tenth stimulus train had already increased in amplitude due to augmentation.

In summary, the comparison of the different stimulations revealed that, in accordance with the above findings, augmentation increased more strongly when gastric mill-like stimulations with shorter inter-train intervals were applied. Both within-train facilitation and augmentation contributed to the electrical response of the gm6 muscle during gastric mill rhythms in vitro and in vivo. The long-lasting effect of augmentation permitted higher EJP amplitudes in all stimulus trains.

What ultimately matters for a behaving animal is the force produced by its muscles. The contribution of augmentation to EJP amplitude suggested that this long-lasting effect would also have an influence on gm6 muscle force production. We tested this hypothesis by measuring the force produced by muscle gm6 with a force transducer. We used the intermediate in vivogastric mill rhythm to stimulate LG and compared the force produced during the first train of stimuli with the successive ones. The force response of muscle gm6 to such stimuli is shown in an original recording in Fig. 7A. The peak force increased with successive stimulus trains. The maximum force produced during the tenth stimulus train was significantly higher than during the first train(Fig. 7B, P<0.05, N=6). After a pause of 100 s between stimulus trains, peak force amplitude decreased again (Fig. 7A, right). The slope of the force development during the tenth stimulus train was also significantly higher than during the first stimulus train (Fig. 7C, P<0.05, N=6). We conclude that the long-lasting effect on EJP amplitude also contributed to the muscle force produced.

We tested the contribution of facilitation and augmentation to the response of the gm6 muscle during a variety of gastric mill-like stimulations. Action potential-evoked EJPs showed facilitation with a time constant of decay of 0.54 s. Facilitation was present in trains of stimuli with stimulus frequencies between 5 Hz and 30 Hz. The increase in EJP amplitude depended on stimulus frequency. In addition to facilitation, augmentation enhanced EJP amplitudes when train stimulations were repeated with moderate stimulus frequencies and inter-train intervals of 16 s or less. With shorter intervals,the effects of augmentation increased. Stimulations with gastric mill rhythms recorded previously in vitro and in vivo showed that both facilitation and augmentation contributed to the muscle response to a realistic motoneuronal input. The effects of augmentation were reflected by an increase of muscle force during gastric mill-like stimulations.

Short-term synaptic plasticity is a feature commonly found for chemical synapses (Colino et al., 2002; Zucker, 1999; Zucker and Regehr, 2002). Several forms of activity-dependent synaptic enhancement and depression have been demonstrated. Depending on time course and duration, synaptic enhancement can be subdivided into facilitation, augmentation and post-tetanic potentiation. Facilitation acts on time scales below 1 s. The effects of augmentation last from a few seconds to 30 s, while those of post-tetanic potentiation are present from several seconds to a few minutes. With continuous activation of the synapse, all three types of synaptic enhancement act together and increase the response of the postsynaptic cell(Fisher et al., 1997; Zucker, 1999). Transient elevation in the Ca²⁺ concentration in the presynaptic terminal following the arrival of action potentials is involved in the induction of these types of synaptic enhancement, yet in different ways. The fast facilitation appears to be mainly induced by the accumulation of free Ca²⁺ (Bykhovskaia et al.,2004) or by a rapid, high concentration of residual Ca²⁺ near a fast site of exocytosis(Fisher et al., 1997; Zucker, 1999). Post-tetanic potentiation, however, seems to be due to an increase in the readily releasable vesicle pool caused by Ca²⁺ effects with slower kinetics, such as the release of mitochondrial Ca²⁺(Bykhovskaia et al., 2004; Fisher et al., 1997; Zucker, 1999; Zucker and Regehr, 2002). Ca²⁺-independent mechanisms have been proposed(Nussinovitch and Rahamimoff,1988). Further evidence that post-tetanic potentiation is mechanistically different from facilitation comes from studies in Drosophila, in which Shaker mutants displayed facilitation but lacked post-tetanic potentiation (Delgado et al.,1994). Augmentation is attributed to the combined effect of fast and slow Ca²⁺ actions in the presynaptic terminal. It appears to be regulated by two plasma membrane extrusion pumps: a Ca²⁺-ATPase and a Na⁺/Ca²⁺ exchange(Zucker and Regehr, 2002).

Our results show that facilitation and augmentation effectively change the electrical response of the gm6 muscle and also the force produced by it. While facilitation mainly affects EJP amplitude and the force produced during a single burst of LG activity, augmentation has longer-lasting effects. It quickens and strengthens muscle force production. Moreover, augmentation also affected within-train facilitation of EJP amplitudes. The relative impact of facilitation and augmentation on the muscle membrane potential depended on the motor neuron discharge pattern, i.e. the type of stimulation used. The central pattern generators in the STG produce a great variety of motor outputs under modulatory control (Marder and Calabrese,1996; Nusbaum,2002; Nusbaum and Beenhakker,2002). For the gastric mill rhythm, both firing frequency and temporal distribution of bursts of activity depend on the modulatory state of the STNS (Nusbaum, 2002). Behavioral observations (Heinzel et al.,1993) indicate that the variability of motor patterns produced by the STG in the isolated nervous system are behaviorally relevant. In intact animals, however, it is hard to predict the variability of the motor patterns from the movement of the teeth, as well as it is almost impossible to deduct the movements of the teeth from a given motor pattern. This is because the crustacean foregut shows a very complex organization of hinges, levers and fulcrums (Maynard and Dando,1974), which makes it difficult to predict the teeth movements produced by a given muscle group (Heinzel et al., 1993). The nonlinear transfer functions between motor patterns and movements, which include different mechanisms of neuronal and muscular plasticity, are only partly understood. Additionally, the presence of modulatory substances in the hemolymph or their release from neuronal terminals can alter muscle properties(Jorge-Rivera et al., 1998; Meyrand and Marder, 1991; Meyrand and Moulins, 1986)which, in turn, may modify the movements evoked by a given motor pattern.

As a first step to characterizing motor neuron activity in vivoand the transfer function for behaviorally relevant movements, we show here for the gastric mill motor neuron LG that there is indeed a broad range of activities produced in intact animals. Since many synapses onto gastric mill muscles show considerable activity-dependent plasticity, such as facilitation and depression (Govind et al.,1975; Hooper et al.,1986; Jorge-Rivera and Marder,1997; Jorge-Rivera et al.,1998; Meyrand and Marder,1991), we used the observed rhythms to characterize the response of the gm6 muscle. In addition, we used LG activities obtained from in vitro preparations, in which either the modulatory projection neuron 1(MCN1) was selectively activated (ion stimulation; Bartos and Nusbaum, 1997) or gastric mill rhythms were activated by stimulation of the sensory dorsal posterior oesophageal nerve (dpon stimulation; Beenhakker et al., 2004; Stein et al., 2005). These in vitro activities were well within the range observed in intact animals. In all tested gastric mill-like stimulations, facilitation and augmentation contributed to the membrane potential of the muscle fiber. Augmentation bridged the time gap between stimulus trains, because its time constant of decay was within the order of interburst intervals obtained during gastric mill rhythms. With the exception of the weak in vivo rhythm,the facilitation index within stimulus trains increased faster in all stimulations when augmentation was present, that is, after several successive stimulation trains. Most obviously, the amplitude of the first EJP was enhanced by augmentation. This indicates that during ongoing gastric mill rhythms, augmentation leads to an already high amplitude of the first and all following EJPs in each burst of motoneuronal activity. In accordance with this we found that not only the peak force produced by the gm6 muscle, but also the slope of the force development during a stimulus train, was increased by augmentation (Fig. 7). Without the long-lasting effect of augmentation, EJP amplitudes and thus most likely muscle force, would be reset by the pause after each motoneuronal burst. During the weak in vivo rhythm, within-train facilitation was not affected by augmentation. Nevertheless, the amplitude of the last EJP was enhanced after several train stimulations, indicating that during such a rhythm it is mainly peak force, but not force development, that would be affected. Our data thus indicate that the transfer function between motor pattern and muscle contraction (and also therefore teeth movement) depends on the degree of augmentation of the neuromuscular junction during a given motor pattern.

Synaptic integration has been studied extensively in crustaceans,particularly at the neuromuscular junction(Atwood, 1976; Atwood and Wojtowicz, 1986; Bittner, 1968; Bykhovskaia et al., 2004; Dudel and Kuffler, 1961; Jorge-Rivera and Marder, 1997; Jorge-Rivera et al., 1998; Katz et al., 1993; Morris and Hooper, 1997; Morris and Hooper, 1998; Morris and Hooper, 2001; Msghina et al., 1998). STG motor neurons provide the large gastric mill muscle fibers with a number of spatially separated synaptic sites, which ensures that the entire muscle fiber receives depolarizing inputs and contracts as a whole. The bilaterally symmetric gm6 muscles are protractor muscles of the lateral teeth in the gastric mill chamber of the crab foregut. They are involved in closing the teeth that are used to masticate food after ingestion. The LG motor neuron,which innervates the gm6 muscles, is part of the gastric mill central pattern generator. LG is capable of generating a very broad range of activities in vitro and in vivo (Table 4). Each burst of LG leads to a contraction of the gm6 muscles and thus to a protraction of the lateral teeth. Facilitation(Jorge-Rivera et al., 1998)and augmentation (present study) both affect gm6 muscle force production and thus, presumably, also the protraction movement of the teeth.

Short-term synaptic enhancement depends on neurotransmitter release probability. Evidence exists that synapses with a low probability of transmitter release (and thus with initially small unitary EJP amplitudes)display facilitation and augmentation, whereas synapses with higher release probability (and large unitary EJP amplitudes) show less facilitation or even depression (Govind et al.,1975; Jorge-Rivera et al.,1998; Msghina et al.,1998; Thomson,2000). This has consequences for signal processing in the postsynaptic cell or, in case of the neuromuscular junction, for the force production of the muscle. If the amplitude of the initial unitary EJP is small, the muscle membrane potential reached during a burst of motoneuronal activity will mainly depend on facilitation and thus also on the firing frequency and the duration of the motor neuron activity(Jorge-Rivera et al., 1998). The effects of summation, however, will be small. In contrast, if the initial EJP is large, the effects of facilitation on the membrane potential will be small, but summation will have a considerable effect. The peak membrane potential during a burst of motoneuronal activity will thus appear earlier(Katz et al., 1993) and depend less on the duration of the motor neuron activity(Morris and Hooper, 1997). For example, the EJPs elicited in the gm8 muscles, the synergists of the gm6 muscle in closing the lateral teeth, are initially small but facilitate during repetitive stimulation (Katz et al.,1993). This leads to a slowly rising force produced by these muscles. The gm9 muscles, which are involved in closing the cardiopyloric valve, receive innervation from the same motor neuron as the gm8 muscles, but differ in their electrical response. EJPs are initially large and depress with repetitive stimulation. Force production reaches a peak very quickly and then declines (Katz et al.,1993).

The gm6 muscles display facilitation and augmentation. Unitary EJPs are small, i.e. at the beginning of a series of stimulus trains (or at the beginning of a gastric mill rhythm) gm6 muscle EJPs initially possess small amplitudes and thus respond in a similar way to the gm8 muscles. They show strong facilitation, but slow force development. After several repetitions of the stimulus train (or during a gastric mill rhythm with repetitive bursts of LG activity), the early EJPs in each stimulus train are larger, due to the augmentation. Consequently, EJPs do not facilitate as strongly as during the first stimulus train, but force development is faster (see Fig. 7) and peak force will be less dependent on train (burst) duration. Depending on the type of gastric mill rhythm present during feeding, this will lead to a stronger and quicker protraction of the lateral teeth. Augmentation thus has behavioral consequences and directly affects the feeding behavior of the crab. We conclude that augmentation of EJPs specifically tunes the muscle response to support force production during rhythmic motor patterns and that it is important to consider the effects of augmentation when characterizing the contraction properties of a muscle.

We would like to thank Harald Wolf, Ursula Seifert and Jessica Ausborn for helpful comments on the paper. This work was supported by DFG grants STE937/2-1 and 2-2.