Sensorimotor integration is known to occur at the level of motor circuits as well as in upstream interneurons that regulate motor activity. Here we show, using the crab stomatogastric nervous system (STNS) as a model, that different sensory systems affect the same set of projection neurons. However, they have qualitatively different effects on their activities (excitation vs. inhibition), and these differences contribute to the selection of motor patterns from multifunctional circuits. We compare the actions of the proprioceptive anterior gastric receptor (AGR) and the inferior ventricular (IV) neurons, which relay chemosensory information from the brain to the STNS, on modulatory commissural neurons 1 and 5 (MCN1 and MCN5) and commissural projection neuron 2 (CPN2) and their resulting actions on the gastric mill central pattern generating circuit in the stomatogastric ganglion. When stimulated, AGR and the IV neurons affect all three projection neurons but elicit distinct gastric mill rhythms. The effects of both sensory pathways on the projection neurons differ in the type of excitation provided to CPN2 and MCN5 (electrical vs. chemical) and the effect on MCN1 (direct inhibition by AGR vs. polysynaptic excitation by the IV neurons). The latter is functionally important because a restoration of MCN1 activity during the AGR rhythm made it more similar to that elicited by IV neuron stimulation. Our results thus support the hypothesis that sensory pathways activate different combinations of projection neurons to select distinct outputs from the same neuronal circuit.
Many rhythmic behaviors are controlled by multifunctional central pattern generators (CPGs). Sensory influences on these CPGs range from phasic, cycle-specific tuning of the motor output to the activation or termination of motor programs. Sensory actions commonly result from both direct synaptic actions on CPG circuit neurons and sensory influences on upstream projection neurons (Beenhakker et al. 2005; Blitz et al. 2004; Katz and Harris-Warrick 1989, 1991). Although the direct actions of sensory pathways on the motor circuits are usually well characterized, the mechanisms used to select distinct motor patterns from multifunctional networks, and in particular the involvement of higher neural centers in the sensory response, remain unclear. Here, we use the stomatogastric nervous system (STNS; Fig. 1Ai) of the crab, Cancer pagurus, to report that two sensory pathways affect the same set of upstream projection neurons, but activate different combinations of these neurons. This differential activation contributes to the selection of distinct motor programs.
The STNS contains several CPGs that control various aspects of the processing of food by the foregut (Nusbaum and Beenhakker 2002). The gastric mill circuit, which controls chewing of food, is located within the stomatogastric ganglion (STG; Mulloney and Selverston 1974; Selverston and Mulloney 1974). It is capable of generating a variety of different rhythms (Heinzel et al. 1993; Stein et al. 2006). Modulatory projection neurons that descend from the paired commissural ganglia (CoGs; Fig. 1Aii) (Nusbaum and Beenhakker 2002) excite the gastric mill CPG. Their modulatory actions seem to be the reason for the broad spectrum of observed rhythms. For example, modulatory commissural neurons 1 (MCN1) (Stein et al. 2007) and commissural projection neuron 2 (CPN2) (Norris et al. 1994) can elicit gastric mill rhythms that are specific for each of these neurons. However, sensory pathways drive the activities of projection neurons and typically activate more than one. The question of how different sensory pathways influence these neurons and whether their actions on them contribute to the selection of motor patterns is still unknown.
We study the actions of the two well-characterized sensory pathways, the anterior gastric receptor (AGR) and the inferior ventricular (IV) neurons, on three identified projection neurons and the role of these actions in determining the motor pattern. AGR has been characterized as a muscle tendon organ (Combes et al. 1995). Its soma is located in the STG and projects to the paired CoGs (Fig. 1Aiii) (Smarandache and Stein 2007). The IV neurons are located in the brain and project to both the STG and the CoGs (Fig. 1Aiii) (Christie et al. 2004; Hedrich and Stein 2008). They relay sensory information to the STNS (Hedrich and Stein 2008; Sigvardt and Mulloney 1982b), for example, from the antennae. We show that both sensory pathways affect the same set of projection neurons but that they have distinct actions on projection neuron activity that lead to different gastric mill motor patterns. Our results indicate that the combination of projection neurons activated by sensory pathways contribute to the selection of distinct motor programs from multifunctional circuits.
Adult C. pagurus L. were obtained from commercial sources (Feinfisch, Neu-Ulm, Germany). Crabs were kept in filtered, aerated artificial seawater (10–12°C). Animals were anesthetized by packing them in ice for 20–40 min before dissection. We used isolated nervous systems to perform our experiments.
Cancer pagurus physiological saline had the following composition (mM): 440 NaCl; 26 MgCl2; 13 CaCl2; 11 KCl; 10 trisma base; 5 maleic acid. In some experiments, high divalent saline (5 × Ca2+/5 × Mg2+) was applied exclusively to the CoGs to block polysynaptic connections (Blitz and Nusbaum 1999; Hedrich and Stein 2008). Furthermore, low calcium saline was used to block chemical synapses in the CoGs (Blitz and Nusbaum 1997; Smarandache and Stein 2007). Composition of the low calcium saline (mM): 440 NaCl; 39 MgCl2; 0.1 CaCl2; 11 KCl; 11.2 trisma base; 5.1 maleic acid. All solutions were kept at a constant temperature of 11–13°C and at pH 7.4–7.6.
Preparation and electrophysiology
The dissection of the isolated STNS was carried out as described previously (Gutierrez and Grashow 2009). The STNS was pinned down in a silicone elastomer-lined (ELASTOSIL RT-601, Wacker, Munich, Germany) petri dish and continuously superfused (7–12 ml/min) with chilled saline (10–13°C). We used standard methods to perform electrophysiology (described in Bartos and Nusbaum 1997; Blitz and Nusbaum 1997; Stein et al. 2006). For extracellular recordings, a petroleum jelly–based cylindrical compartment was built around a section of the nerve to electrically isolate STNS nerves from the bath. One of two stainless steel electrode wires was placed in this compartment, the other one in the bath as a reference electrode. The differential signal was recorded, filtered, and amplified with an AM Systems amplifier (Model 1700, AM Systems, Sequim, WA).
We desheathed the STG and the CoGs for better insertion of the microelectrodes during intrasomatic recordings of motor and CoG projection neurons. Neurons were visualized with white light transmitted through a darkfield condensor (Nikon, Tokyo, Japan). Microelectrodes (15–25 MΩ) were filled with a solution containing 0.6 mM K2SO4 and 0.02 mM KCl. Intracellular recordings and current injections were carried out using an NEC 10L amplifier (NPI Tamm, Germany) or an Axoclamp 2B amplifier (Molecular Devices, Sunnyvale, CA) in bridge or single electrode discontinuous current clamp. Sampling rates in discontinuous current-clamp mode ranged from 2 to 14 kHz.
Activity patterns, synaptic interactions, and axonal projection pathways in combination with current injections were used to identify STG neurons, as described previously (Bartos and Nusbaum 1997; Blitz and Nusbaum 1997; Weimann et al. 1991). CoG projection neurons were identified by their axonal projections and their influence on the STG neurons (Bartos and Nusbaum 1997; Coleman and Nusbaum 1994; Norris et al. 1996; Nusbaum and Marder 1989a,b). Each projection neuron occurs as a single copy in each CoG. MCN1 and MCN5 project via the inferior esophageal nerve (ion) and the stomatogastric nerve (stn) to the STG (Fig. 1Aii) (Coleman and Nusbaum 1994; Norris et al. 1996). CPN2 projects through the superior esophageal nerve (son) and the stn to the STG (Norris et al. 1994).
The IV neurons were activated by stimulating the inferior ventricular nerve (ivn) extracellularly with 10 consecutive stimulus trains of 40-Hz stimulation frequency and different train durations (3.5–6 s) and intertrain intervals (1–20 s). Previous studies showed that lower ivn stimulation frequencies and shorter train durations and intertrain intervals do not elicit gastric mill rhythms in crabs (Hedrich and Stein 2008). Stimulation of the ivn resulted in a selective activation of the IV neurons (Christie et al. 2004; Hedrich and Stein 2008).
For our experiments, we used two different types of preparations: without brains and with the brain attached. In the latter preparations the STNS and the brain were connected only through the ivn. The paired circumesophageal commissures (cocs) had been transected. Both preparations showed similar responses (Hedrich and Stein 2008).
To activate AGR rhythmically, we either depolarized the AGR soma with rectangular current pulses or stimulated the dorsal gastric nerve (dgn) extracellularly. We used different train durations (2–8 s) and intertrain intervals (2–16 s). The effects of AGR on the gastric motor pattern are independent of AGR firing frequency for firing frequencies between 20 and 90 Hz (Smarandache and Stein 2007).
We used extracellular inferior esophageal nerve (ion) stimulations to test the contribution of the projection neuron MCN1 to the generation of gastric mill rhythms. Therefore we extracellularly stimulated one of the ions rhythmically with train durations and intertrain intervals that depended on the parameters used for IV neuron and AGR stimulation (details given in results). Only MCN1 was activated during the ion stimulation, because the activation threshold of MCN5 (which also projects through the ions) is higher than that of MCN1 (Bartos and Nusbaum 1997; Coleman et al. 1995).
The gastric mill rhythm was monitored by the activity of the lateral gastric (LG; 1 cell) neuron, the dorsal gastric (DG; 1 cell) neuron and the gastric mill motor neurons (GM; 4 cells). This rhythm was considered to be spontaneously active when the LG neuron (member of the CPG) produced bursts of action potentials. The cycle period of the gastric mill rhythm was defined as the duration between the onset of an LG burst and the onset of the subsequent LG burst. Burst durations were defined as the time between the first and last action potential (spike) within a burst. LG was either recorded intracellularly from its soma or extracellularly from the lateral gastric nerve (lgn). DG was either recorded intracellularly or extracellularly from the dorsal gastric nerve (dgn). The GM neurons were recorded with intrasomatic microelectrodes or extracellularly from the dgn or the medial ventricular nerve (mvn).
Activities were measured either as the number of action potentials per burst or as sliding average with a bin width of 50 ms. The intraburst firing frequency was used to assess the effect of the different sensory pathways on the recorded neuron. It was defined as the number of spikes within a burst minus one divided by the burst duration. Mean values were determined from 10 consecutive cycles of gastric mill activity. In some experiments, data were normalized to the prestimulus control. To detect changes in the activity of a gastric motor neuron, its intraburst firing frequency during sensory stimulation was measured and divided by the immediate control values before stimulation. Because the spikes of projection neurons cannot always be assigned to individual bursts, projection neuron firing frequency during sensory stimulation was measured as sliding average with a bin width of 50 ms. We normalized the data to the immediate controls before stimulation. In all cases, measurements were repeated 10 times, and the results were averaged.
Data were recorded with Spike2 (version 5.03–6.09, CED, Cambridge, UK) and a micro 1401AD board (CED) and analyzed using the Spike2 script language. Individual scripts are published at http://www.neurobiologie.de/spike2. Final figures were prepared with CorelDraw (version 12 for Windows, Corel, Ottawa, Canada). Graphics and statistics were generated using Excel (Microsoft), Plotit (version 3.2, Scientific Programming Enterprises, Haslett, MI) or Origin (version 7.0273, OriginLab, Northampton, MA). Statistical tests for data analysis were the nonparametric U test (Mann-Whitney) for two samples or one-way ANOVA on ranks for repeated measures. Data are presented as box-and-whisker plots showing medians, lower and upper quartiles, and minimums and maximums. N refers to the number of animals; n gives the number of trials. For all statistical tests, significance with respect to control is indicated separately for each figure.
The gastric mill motor pattern is highly flexible (Heinzel et al. 1993; Stein et al. 2006) and is shaped by exteroceptive and proprioceptive sensory information, for example, by the IV neurons (Hedrich and Stein 2008) and by AGR (Smarandache and Stein 2007). In crabs, both types of feedback seem to involve the actions of modulatory projection neurons in the CoGs, for example, those of the modulatory projection neurons 1 and 5 (MCN1 and MCN5) and commissural projection neuron 2 (CPN2). If and how the two different sensory pathways affect the projection neurons and the contribution of projection neuron activity to the selection of distinct motor patterns is currently unknown.
AGR and the IV neurons elicit different gastric mill rhythms
The two sensory pathways studied here have very different functions. AGR is a proprioceptive muscle tendon organ that measures muscle tension during chewing (Combes et al. 1995; Simmers and Moulins 1988a; Smarandache et al. 2008), whereas the IV neurons are activated by different sensory signals (Hedrich and Stein 2008; Sigvardt and Mulloney 1982b). In C. pagurus, the IV neurons relay chemosensory information from the brain to the STNS and show rhythmic activity (Hedrich and Stein 2008). Because previous studies showed that both sensory pathways affect the gastric mill rhythm (Hedrich and Stein 2008; Smarandache and Stein 2007; Smarandache et al. 2008), we thus first determined whether AGR and the IV neurons elicit different types of gastric mill rhythms. For this purpose, we stimulated each sensory pathway separately in preparations without spontaneous gastric mill activity. We used stimulation protocols that corresponded in period and frequency to the activity previously obtained in intact animals (Hedrich and Stein 2008; Smarandache et al. 2008) (for details, see methods). In all tested preparations (n = 12), AGR stimulation elicited a gastric mill rhythm (Fig. 1B) that included bursting of the protractor motor neurons (GMs and LG) and the retractor motor neuron (DG) and lasted for the duration of the stimulus. All protractor motor neurons were active in time with the AGR stimulus train. Similarly, ivn stimulation elicited a gastric mill rhythm, during which the protractor motor neurons were active during the stimulation phase (Fig. 1C). The gastric mill period always corresponded to the stimulation period of AGR and IV neurons, respectively.
In contrast, the phase relationship of the gastric mill neurons were different when gastric mill rhythms elicited by AGR and IV neuron stimulation were compared (Fig. 1D). The end of the LG activity phase was significantly delayed during ivn stimulation in comparison to AGR stimulation (Fig. 1D). Both onset and end of DG activity started significantly earlier during ivn stimulation. The phasing of the gastric mill motor neurons GM was also significantly different. Their onset was delayed (Fig. 1D). The end of the GM activity was not significantly different. In contrast to the phasing, the firing frequencies of the gastric motor neurons did not differ between AGR and ivn stimulation (Fig. 1E).
Besides these phase differences, we found differences in the subthreshold oscillations of the LG neuron. The core of the gastric mill CPG consists of two cells (LG and interneuron 1) (Stein et al. 2007) with reciprocal inhibition. Interneuron 1 (Int1) receives additionally rhythmic inhibition from the fast pyloric CPG (Bartos et al. 1999). During this pyloric-timed inhibition of Int1, the LG membrane potential escapes from inhibition and depolarizes, which results in pyloric-timed disinhibitions in LG (Fig. 2, A and B). These disinhibitions occur during the LG interburst and critically affect the timing of the next LG burst and thus of the whole rhythm. They represent the main contact point for the interaction between gastric mill and pyloric rhythms (Bartos et al. 1999). Figure 2, A and B, shows intracellular LG recordings of a single gastric mill cycle during repetitive AGR and ivn stimulations, respectively. Although the amplitudes of the subthreshold oscillations remained almost unchanged during AGR stimulation experiments, they increased significantly during ivn stimulation experiments (Fig. 2C; see legend for statistical tests), indicating a fundamental difference in the mechanisms used in the timing of the two rhythms. This difference was also reflected in the beginning of the subsequent LG burst after onset of the stimulation phase. The mean delay between the start of the stimulation and the beginning of the LG burst was 0.59 ± 0.28 s (n = 17) for AGR and thus significantly shorter than for ivn stimulation (3.39 ± 0.8 s; Fig. 2D; n = 20; U test; P < 0.001).
In summary, both sensory pathways elicited gastric mill rhythms, but these rhythms showed marked differences in the phases and durations of bursts of the involved gastric motor neurons.
AGR and the IV neurons differ in their effects on spontaneous gastric mill rhythms
Not only do AGR and IV neurons elicit gastric mill rhythms, they also modify ongoing rhythms (Hedrich and Stein 2008; Smarandache and Stein 2007). In preparations with spontaneously active gastric mill rhythms (AGR: n = 6; IV neurons: n = 19), both rhythmic AGR and ivn stimulation entrained these rhythms, i.e., in a certain range of stimulation periods, the gastric mill period and the stimulation period fell into synchrony. After the end of the stimulation, the gastric mill period returned to the period measured before the stimulation. For these experiments, we used different train durations (AGR: 2–8 s; ivn: 3.5–6 s) and intertrain intervals (AGR: 2–16 s; ivn: 1–20 s) to study the effects on ongoing gastric mill rhythms over a wide range. The periods of the control gastric mill rhythms did not differ between AGR and ivn experiments (P > 0.2; AGR: 8.11 ± 1.93 s, n = 6; ivn: 9.31 ± 2.43 s, n = 19).
Figure 3, A and B, shows the plots of the normalized gastric mill period over the normalized stimulation period for AGR (Fig. 3A) and ivn stimulation (Fig. 3B), respectively. Both parameters were normalized to the period of the spontaneous active gastric mill rhythm to compare the effects of AGR and ivn stimulations in animals with varying control periods. The plot of the AGR stimulation shows that the period of the rhythm could be decreased by ≤40% and increased by ≤60% (slope: 0.99, R2 = 0.99; n = 6; number of trials = 29; Fig. 3A). Beyond these stimulation periods, the gastric mill rhythm was not entrained (<0.6; Fig. 3A) or we observed two gastric mill cycles per stimulation cycle (>1.6; Fig. 3A) (see also Smarandache and Stein 2007). In contrast, the range of de- and increasing periods was more expanded for ivn stimulation (slope = 0.99; R2 = 0.98; n = 19; number of trials = 44; Fig. 3B). Beyond this range, the gastric mill rhythm could not be entrained.
We found further differences between both sensory pathways when we compared the intraburst spike activities (see methods) of the protractor neurons LG and GM (Fig. 3C). For this, we normalized all data to the values obtained during the spontaneous gastric mill rhythm, i.e., without sensory stimulation. The firing frequencies of the protractors LG and GM increased during AGR stimulation (LG: 1.11 ± 0.21; n = 7; GM: 1.26 ± 0.28; n = 6), whereas they decreased during ivn stimulation (LG: 0.93 ± 0.33; n = 19; GM: 0.86 ± 0.26; n = 9). They were thus significantly different from each other (U test; P < 0.05; Fig. 3C). In contrast, the mean normalized firing frequency of DG was not significantly different, although it tended to be lower during IV neuron stimulation (AGR: 1.08 ± 0.18; IV neurons: 0.98 ± 0.34; n = 16). Despite the differences in LG and GM firing frequencies burst durations, and the number of spikes per burst did not show a significant difference when AGR and IV neuron stimulation were compared.
Together, these results show that the gastric mill motor circuit shows different responses to the proprioceptor AGR and the chemosensory IV neurons, which indicates that they use different strategies to elicit and affect gastric mill rhythms. Because gastric mill rhythms are driven by the modulatory input of CoG projection neurons, we next determined whether AGR and the IV neurons influence identified projection neurons and, if so, how this influence is exerted.
AGR and IV neurons affect the same CoG projection neurons
Previous work has indicated that, in crabs, both sensory pathways act exclusively via projection neurons in the CoGs (Combes et al. 1999a; Hedrich and Stein 2008; Norris et al. 1994; Smarandache and Stein 2007). We thus recorded from a set of three identified CoG projection neurons, MCN1, MCN5, and CPN2, during either AGR or IV neuron stimulation. We found that both AGR and the IV neurons affected all three projection neurons, but that their effects on each projection neuron were different. To show the differences in the effects of AGR and the IV neurons on projection neuron activity, we normalized all data to their mean prestimulus control.
EFFECTS ON MCN5.
AGR stimulation excited MCN5 by depolarizing its membrane potential and prolonging its bursts (Fig. 4A). The latter fact was apparently caused by a lack of rhythmic inhibitory input to MCN5 (arrows). This input originates from the pyloric network in the STG (Norris et al. 1996). The normalized mean firing frequency of MCN5 increased significantly from 0.99 ± 1.32 before AGR stimulation to 2.27 ± 1.77 during AGR stimulation (Fig. 4B; 1-way ANOVA on ranks for repeated measures; P < 0.001; n = 11). After stimulation, it returned to 1.02 ± 0.92. When we hyperpolarized MCN5 below the spike threshold, each AGR spike elicited a discrete excitatory postsynaptic potential (EPSP; n = 4; Fig. 4C). The MCN5 excitation by AGR was direct, because it was still visible when we superfused the CoGs with high divalent saline (n = 3), which blocks polysynaptic connections in the STNS (Blitz and Nusbaum 1997). In contrast, saline containing low concentrations of calcium blocked the MCN5 excitation (n = 2; Fig. 4C), indicating that it relied on chemical transmitter release.
IV neuron stimulation also excited MCN5 (Fig. 4D). MCN5 firing frequency increased from 1.1 ± 0.53 before stimulation to 5.54 ± 1.36 during stimulation (Fig. 4E; 1-way ANOVA on ranks for repeated measures; P < 0.05; n = 7) and was thus much more pronounced than during AGR stimulation. After stimulation, MCN5 firing frequency decreased again to control levels (0.72 ± 0.62). During stimulation, MCN5 activity was rather constant and did not show any adaptation. We found that each ivn stimulus pulse elicited a PSP in MCN5, which, up to a stimulation frequency of 20 Hz, elicited a spike in MCN5. At higher stimulus frequencies, every other PSP failed to elicit a spike, resulting in an approximate maximum MCN5 firing frequency of ∼20 Hz. When we applied low calcium saline to the CoGs, the MCN5 excitation continued (Fig. 4F), indicating that the IV neurons are electrically coupled to MCN5. A comparison of the PSPs in regular saline and low calcium saline shows that no apparent chemical component in the synaptic response was present (Fig. 4F).
EFFECTS ON CPN2.
AGR stimulation excited CPN2 (Fig. 5A) and depolarized its membrane potential. The normalized CPN2 firing frequency increased rapidly from 1.0 ± 0.39 before stimulation to 2.35 ± 0.65 during AGR stimulation (Fig. 5B; 1-way ANOVA on ranks for repeated measures; P < 0.001; n = 14) and decreased again to 1.39 ± 0.44 after the stimulation. Each AGR spike either elicited a PSP or an action potential. When low calcium saline was applied to the CoGs, the excitation of CPN2 continued (Fig. 5C; n = 4), which shows that AGR is electrically coupled to CPN2.
The IV neurons also excited CPN2 (Fig. 5D). CPN2's normalized firing frequency increased from 0.98 ± 0.56 before ivn stimulation to 3.19 ± 0.96 during stimulation (Fig. 5E; 1-way ANOVA on ranks for repeated measures; P < 0.001; n = 9) and decreased again to 1.01 ± 0.67 after stimulation. In contrast to AGR stimulation, the actions of the IV neurons were mediated polysynaptically, because they were completely absent in high divalent saline (Fig. 5F).
EFFECTS ON MCN1.
AGR reduced the activity of MCN1 by hyperpolarizing its membrane potential (Fig. 6A). In the example shown in Fig. 6A, the inhibitory effect was strong enough to terminate MCN1 activity. On average, the normalized MCN1 firing frequency dropped from 1.0 ± 0.41 in control to 0.34 ± 0.30 during stimulation (Fig. 6B; 1-way ANOVA on ranks for repeated measures; P < 0.001; n = 11). After AGR stimulation, MCN1 showed a short postinhibitory rebound and returned to control levels (1.02 ± 0.56). Immediately after the onset of the AGR stimulation, individual PSPs time-locked to the AGR stimuli were obvious (Fig. 6A, inset). The amplitude of these PSPs diminished with increasing hyperpolarization until they could no longer be perceived. The hyperpolarization was still present in high divalent saline (n = 3; Fig. 6C) but was absent in low calcium saline (n = 4; Fig. 6C), which argues for a direct chemical inhibition.
IV neuron stimulation increased MCN1 firing frequency from 1 ± 0.44 to 2.11 ± 0.23 (Fig. 6, D and E; 1-way ANOVA on ranks for repeated measures; P < 0.001; n = 9). After the end of the ivn stimulation, MCN1 activity decreased again to control levels (1.22 ± 0.43). No individual PSPs were obvious in the MCN1 recording, even when MCN1 was hyperpolarized below spike threshold. Furthermore, high divalent saline blocked the actions of the IV neurons on MCN1 (Fig. 6F), showing that the connection was polysynaptic.
In summary, our results show that both AGR and IV neurons affect the same set of projection neurons, namely MCN1, MCN5, and CPN2 (Fig. 7). AGR chemically inhibits MCN1, whereas the IV neurons polysynaptically excite it. AGR excites CPN2 via an electrical synapse, whereas the excitation by the IV neurons is mediated via a polysynaptic pathway. Both sensory pathways excited MCN5: AGR chemically and the IV neurons via electrical coupling.
MCN1 is essential for eliciting a gastric mill rhythm during ivn stimulation
Could the difference in the response of MCN1 to AGR and IV neuron activity contribute to the specifics of the elicited motor patterns, i.e., does the combination of active projection neurons support the selection of motor patterns? To test this hypothesis, we either eliminated or added the influence of MCN1 during stimulation of both sensory pathways. First, we tested the contribution of MCN1 to the ivn-elicited rhythm. The IV neurons project to the CoGs via both ions and sons (Fig. 1Aiii). However, spike propagation and activation of the projection neurons can only be observed through the sons (Cazalets et al. 1990; Christie et al. 2004; Hedrich and Stein 2008). The influence of MCN1 and MCN5 to the ivn-elicited gastric mill rhythm can thus be removed either by transecting the ion and with it the axons of both projection neurons (Fig. 1Aii) or by transecting the superior esophageal nerve (son) to eliminate the actions of the IV neurons from the CoG neurons. We used a combination of both approaches. We first blocked the influence of the IV neurons on one CoG by transecting the left son (for nerves, see Fig. 1Ai). In these preparations, rhythmic ivn stimulation still elicited a gastric mill rhythm (Fig. 8 i; n = 8), indicating that the effects of the IV neurons on the projection neurons in a single CoG were sufficient to initiate a gastric mill rhythm. We eliminated the actions of MCN1 and MCN5 on the gastric mill motor circuit in the remaining CoG by transecting the right ion, leaving only the influence of CPN2 intact. In these conditions, no gastric mill rhythm could be elicited during ivn stimulation (Fig. 8ii; n = 8). LG and DG were not rhythmically active. The GM neurons still received excitatory input, but their membrane potential never crossed spike threshold.
To identify which of the two transected projection neurons was responsible for the loss of rhythmic gastric mill activity, we selectively restored the activity of MCN1 by rhythmically stimulating one ion simultaneously with the IV neurons (i.e., during ivn stimulation). We used a stimulation frequency of 20 Hz, which corresponded approximately to the MCN1 frequency during ivn stimulation (MCN1 mean firing frequency during ivn stimulation: 18.17 ± 1.43 Hz, n = 9). Because the activation threshold of MCN5 is higher than that of MCN1 to ion stimulation, only MCN1 was activated by the ion stimulus (Bartos and Nusbaum 1997; Coleman et al. 1995). We found that restoring MCN1 also reinstated the gastric mill rhythm during ivn stimulation (Fig. 8iii; n = 8). In contrast, exclusive rhythmic ion stimulation with the given stimulus parameters, but without additional ivn stimulation was not sufficient to elicit a gastric mill rhythm (n = 8; data not shown). Rather, LG started to fire tonically with low firing frequencies. Thus the combined excitation of the projection neurons MCN1 and CPN2 was essential for eliciting the gastric mill rhythm, and a loss of MCN1 activity prevented the generation of the rhythm during ivn stimulation.
When we stimulated AGR after eliminating its effects on the left CoG by transecting the left son and with it the AGR axon to that CoG, stimulation still elicited a gastric mill rhythm (Fig. 9Ai; for nerves, see Fig. 1Aiii). In contrast to the IV neurons, a gastric mill rhythm could still be elicited even when the ion (and thus the axons of MCN1 and MCN5) of the right CoG was transected (Fig. 9Aii; n = 10). This finding supports the hypothesis that MCN1 was less important in eliciting the rhythm than during IV neurons stimulation. It also matches the fact that AGR inhibits MCN1, whereas the IV neurons excite it. Could this difference be contributing to the differences in the elicited gastric mill rhythms? If so, one should expect that adding MCN1 activity during AGR stimulation will make the rhythms more similar. To test this, we added ivn-like MCN1 activity during AGR stimulation (Fig. 9Aiii; n = 10) by stimulating the ion extracellularly. We selectively activated MCN1 in addition to the AGR stimulation such that its activity matched that during IV neuron stimulation both in frequency (20-Hz stimulus frequency) and timing, i.e., MCN1 was only activated during the AGR stimulus phase (Fig. 9Aiii). This had a pronounced effect on the phasing of the GM neurons (see phase diagram in Fig. 9B): their activity now started after LG activity, similar to an ivn-elicited gastric mill rhythm. In contrast, the phases of DG did not change after adding MCN1 activity.
Similar to the comparison of exclusive AGR and ivn stimulation, the firing frequencies of the gastric mill motor neurons LG and GM was not significantly different from ivn stimulation when MCN1 was added on top of the AGR stimulation [LG: AGR + MCN1 stimulation: 7.57 ± 2.07 Hz (n = 10); IV neuron stimulation: 7.83 ± 3.82 Hz (n = 31); GM: AGR + MCN1 stimulation: 4.15 ± 1.78 Hz (n = 10); IV neuron stimulation: 6.39 ± 2.42 Hz (n = 13)], indicating that it is rather the timing of the pattern that is altered by the two sensory pathways than is its strength.
When we looked at the mean delay between the start of the AGR stimulation and the beginning of the LG burst, we again found that additional MCN1 stimulation caused the AGR rhythm to become more similar to the ivn-elicited rhythm: the delay during exclusive AGR stimulation and AGR stimulation with the ion transected was on average <1 s (AGR stimulation, 0.78 ± 0.31 s; n = 10; AGR with ion transected, 0.97 ± 0.27 s; n = 10) and thus significantly shorter than that of the ivn-elicited rhythm (3.39 ± 0.8 s; n = 20; U test; P < 0.05). When MCN1 was activated in addition to AGR, the delay increased to 1.79 ± 0.39 s (n = 10) and was thus no longer different from ivn stimulation (U test; P > 0.2).
In summary, replacing MCN1 inhibition during AGR stimulation by excitation made the AGR rhythm more similar to the ivn-elicited rhythm, although marked differences still remained. Thus the differences in the gastric mill rhythms that are elicited by AGR and the IV neurons seem to partly depend on the combination of active projection neurons (for details, see discussion).
The goal of this investigation was to study the cellular and synaptic mechanisms contributing to the integration of sensory information arising from two different sensory systems that affect the same motor network. For this, we used the crab stomatogastric nervous system with its well-known sensory pathways (Beenhakker and Nusbaum 2004; Blitz et al. 2004; Hedrich and Stein 2008; Smarandache and Stein 2007), modulatory system, and motor circuitry (Katz 1995; Marder and Calabrese 1996; Nusbaum and Beenhakker 2002). We showed that, first, proprioceptive and exteroceptive sensory information is carried to a multifunctional motor circuit in the STNS via the same three projection neurons (although maybe not exclusively), and that, second, the two sensory pathways have qualitatively different effects on these neurons. Finally, we showed that the combination of projection neurons activated by these pathways contributes to the selection of motor patterns.
It is a known fact that different sensory systems can converge onto the same pool of interneurons—a mechanism thought to facilitate the selection of appropriate motor responses from multifunctional circuits (Barriere et al. 2008; Blitz et al. 2004; Hess and Büschges 1997, 1999; Rossignol et al. 2006). In the crab, for instance, the proprioceptive gastro-pyloric receptor cells (GPRs) and the VCN (ventral cardiac neuron) mechanoreceptors connect to the same projection neurons. They nevertheless elicit distinct motor patterns (Blitz et al. 2004). Here, we showed that the muscle tendon organ AGR and the IV neurons in the brain converge onto the same three CoG projection neurons but affect them differently: The IV neurons reach the gastric mill network through a coactivation of three projection neurons (MCN1, CPN2, and MCN5), whereas AGR excites two projection neurons (CPN2 and MCN5) and inhibits another (MCN1). AGR and the IV neurons have been shown to influence descending projection neurons in other crustacean species, for example, in the lobster Homarus gammarus (Combes et al. 1999a,b; Meyrand et al. 1991, 1994; Simmers and Moulins 1988a). The activities of these projection neurons are crucial for the type of rhythm elicited by both sensory pathways (Combes et al. 1999b; Meyrand et al. 1994; Simmers and Moulins 1988b).
In C. pagurus, AGR and the IV neurons elicit distinct motor patterns, which is most obvious in the phasing of the pro- and retractors and in the pyloric-timed subthreshold oscillations in the CPG neuron LG, which determine the timing of the LG burst and thus of the rhythm: during IV neuron stimulation, their amplitude increases steadily until the next LG burst is elicited (Fig. 2B), a situation similar to rhythms elicited by selective stimulation of MCN1 (Bartos et al. 1999; Coleman et al. 1995). In these rhythms, MCN1 excitation of LG increases steadily during the LG interburst, which is reflected in the amplitude of the subthreshold LG oscillations. In contrast, when AGR was stimulated, the oscillations did not increase in amplitude (Fig. 2A), indicating that MCN1 excitation plays a minor role in eliciting the LG burst. The fact that the delay of the LG burst after the onset of the stimulus was shorter during AGR stimulation may reflect this difference in the underlying mechanisms.
AGR deactivates MCN1 via a direct chemical inhibition while the IV neurons polysynaptically activate MCN1. This difference in MCN1 is functionally important, because restoring MCN1 activity during the AGR rhythm made this rhythm more similar to that elicited with ivn stimulation. Nevertheless, there were still distinct differences left between both rhythms. These differences could, in principle, result from 1) the activity of additional projection neurons, 2) differences in projection neuron firing frequencies, 3) direct synapses between the sensory neurons and the motor circuit, 4) a modulation of spike propagation or synaptic efficiency via presynaptic actions, and 5) modulatory actions on projection neurons and/or the motor circuits.
1) An activation of additional projection neurons seems likely, although it has been shown that gastric mill rhythms can depend on the activities of only a few projection neurons (Beenhakker et al. 2004). We found that the combinatory actions of MCN1 and CPN2 are necessary to elicit the IV rhythm, whereas the AGR rhythm does not rely on MCN1. However, a contribution of other projection neurons to the characteristics of the elicited rhythms, such as phase relationships and firing frequencies, has thus far not been tested. It is currently unknown how many of the 20 CoG projection neurons (Coleman et al. 1992) are active and perturbed by sensory pathways. Determining the combination of active projection neurons during different behavioral states or sensory influences will challenge classical electrophysiological recording techniques and possibly require the use of multiunit recording techniques such as the optical recording of projection neuron activity with calcium- or voltage-sensitive dyes.
2) Differences in projection neuron firing frequencies are also likely to play a role in rhythm selection as shown by Blitz et al. (2004). Here, we found that CPN2 showed higher firing frequencies when AGR was active in comparison to IV neuron stimulation, most likely because of its strong electrical synapse with AGR. MCN5, on the other hand, was strongly activated by the IV neurons via an electrical synapse. Because CPN2 inhibits DG (Norris et al. 1994) and MCN5 excites it, it is reasonable to assume that the firing frequencies of CPN2 and MCN5 have a strong impact on the timing of the protraction phase and hence also on its phasing. In fact, we found that, during AGR stimulation, the protraction phase was significantly delayed. However, this can ultimately only be tested by experimentally altering projection neuron firing frequencies.
3) Although no direct actions of AGR on the STG motor circuits have been shown (Combes et al. 1999b; Elson et al. 1994; Simmers and Moulins 1988a,b; Smarandache and Stein 2007), most other sensory pathways directly affect these circuits, in addition to their actions on the projection neurons. In the spiny lobster, Panulirus interruptus, for instance, the IV neurons exert mainly direct effects on the STG circuits (Sigvardt and Mulloney 1982a,b) and apparently determine the timing of the gastric mill motor pattern via direct synapses onto the GM and Int1 neurons, the two principle components of the gastric mill CPG in this species. In Homarus gammarus, the homologous PS neurons dismantle the STG and esophageal rhythms and replace them with a single, conjoint motor pattern (Meyrand et al. 1991, 1994). This effect is partly caused by the inhibition of those STG neurons that do not participate in the conjoint motor pattern. It thus seems that the direct effects of the IV neurons are more important in lobsters and partly occlude the effects of the CoG projection neurons on pattern selection. In the crab, direct synaptic actions have a much weaker influence on the gastric mill rhythm (Christie et al. 2004) such that the rhythm elicited by the IV neurons should be mainly driven by their actions on the CoG projection neurons. One way to test the contribution of the direct connections between IV neurons and gastric mill motor circuit would be to block the IV neuron transmitter actions (mediated mostly by histamine; Christie et al. 2004) selectively in the STG during IV neuron stimulation.
4) Sensory pathways can also interfere with projection neuron output or activation via a combination of short- and long-lasting presynaptic actions (Barriere et al. 2008; Beenhakker et al. 2007). For example, one proprioceptive system is able to promote its impact on a target motor network by biasing the access of a different sensory system to the same circuit (Barriere et al. 2008). To test for such interactions, at the least the actions of successive and simultaneous AGR and IV neurons stimulation on projection neuron and motor circuit activity will have to be compared. Furthermore, changes in synaptic efficacy, as, for example, present during facilitation or depression, could contribute to the remaining differences between the AGR and IV neuron rhythms. However, only the AGR inhibition of MCN1 showed obvious changes in the postsynaptic response: the amplitude of the unitary PSPs diminished with time, but also with increasing hyperpolarization. This may have been because of synaptic depression or postsynaptic effects (i.e., the reversal potential was reached).
5) Long-lasting modulatory actions caused by the sensory stimulation may also contribute to observed differences in motor pattern. Our results did not show any obvious modulatory actions of either AGR or the IV neurons. For example, in none of our experiments did we see a long-lasting excitation of the projection or motor neurons that was comparable to those elicited by GPR or the VCN (Blitz et al. 2004; Katz and Harris-Warrick 1989). However, our experiments were not designed to detect modulatory actions of AGR and the IV neurons. For this, a more thorough study of the postsynaptic response of the projection neurons to AGR and IV neuron stimulation is necessary.
In conclusion, sensory pathways can affect the same pool of modulatory interneurons but activate different combinations of them and thereby elicit distinct motor patterns from the same CPG. Several mechanisms of how higher-order interneurons can sculpt or activate motor patterns were described previously. In Aplysia, the motor patterns for egestion and ingestion are elicited by differential activity of two higher-order interneurons. Whereas egestion is elicited by activating one of these interneurons, ingestion requires the co-activation of both (Morgan et al. 2002). In the leech, the decision between crawling and swimming depends on sensory input in hierarchical lower levels of the CNS. The activation of locomotion is triggered by a single command-like neuron, but the actual decision on whether crawling or swimming is performed depends on the combinatorial code of decision neurons, which are driven by sensory input (Esch and Kristan 2002; Esch et al. 2002).
In vertebrates, the mesencephalic locomotor region (MLR) is a region where sensory feedback from skin mechanoreceptors and olfactory afferents (Dubuc et al. 2008) and projections from different parts of the brain (forebrain, basal ganglia, neocortex, hypothalamic system; Takakusaki 2008) are integrated. The MLR neurons channel the information toward the reticulospinal cells that drive the locomotor CPGs in the spinal cord. Electrical stimulation of the MLR of the salamander can produce two distinct locomotor patterns: the CPG for stepping is activated with low threshold stimulation. With increasing stimulation amplitude, the CPG for swimming is activated, and consequently, another pool of muscles is recruited (Cabelguen et al. 2003; Chevallier et al. 2008), indicating that the MLR neurons are involved in motor pattern selection, depending on the input they receive.
The convergence of two sensory pathways onto the same subset of projection neurons shown in this work and the resulting motor pattern selection adds to the previous list and further supports the hypothesis that a combinational code of active and inactive interneurons is crucial to elicit distinct motor patterns. We show that the combination of different modulatory interneurons provides the opportunity to make a system more flexible and that different sensory systems do not require different projection neurons to elicit a distinct activity pattern from the same set of motor neurons. The fact that the diverse sensory activation of one neuron can be a reason for a distinct motor output is an indication for the high flexibility prevailed in nervous systems. Future studies of sensory systems and projection neurons affecting motor pattern generation and selection may thus involve measurements in semi-intact preparations or intact animals to provide insight into the interaction of different sensory pathways.
This work was supported by Deutsche Forschungsgemeinschaft STE 937/2-1 and 2-2.
We thank H. Wolf, U. Seifert, and B. Mulloney for helpful comments; M. Nusbaum and D. Blitz for help with the experiments; and N. Daur for providing some of the data.
Present addresses: U.B.S. Hedrich, Departments of Neurology and Applied Physiology, Ulm University, Zentrum für Klinische Forschung, D-89081 Ulm, Germany; C. R. Smarandache, Department of Neurobiology, Physiology, and Behavior, University of California, Davis, CA 95616-8519.
- Copyright © 2009 the American Physiological Society