A common feature of arousing stimuli used as reinforcement in animal models of learning is that they promote memory formation through widespread effects in the CNS. In the marine mollusk Aplysia, sensitization is typically induced by tail-shock, an aversive reinforcer that triggers a state of defensive arousal characterized by escape locomotion and increased heart rate. Serotonin (5-HT) contributes importantly to sensitization of defensive reflexes as well as to the regulation of locomotion and heart rate. Although specific serotonergic neurons increase their firing after tail-shock, it remains unclear whether this effect is restricted to these neurons or whether tail-shock recruits a more global serotonergic system. In this study, we recorded from serotonergic neurons throughout the CNS, which were prelabeled with 5,7-dihydroxytryptamine, during an in vitro analog of sensitization training, tail-nerve shock. We found that most of the serotonergic neurons that we recorded from (80%) increased their firing rate for several minutes after nerve shock. Most serotonergic neurons in the pedal and abdominal ganglion were also excited by 5-HT and by intracellular activation of the two serotonergic neurons CB1/CC3. This interconnectivity between serotonergic neurons might contribute to spread excitation within a large proportion of the serotonergic system during sensitization training. It is also possible that serotonergic neurons could be activated by 5-HT present in the hemolymph via a neuro-humoral positive feedback mechanism. Overall, these data indicate that sensitization training activates a large proportion of Aplysia serotonergic neurons and that this form of learning occurs in a context of increased serotonergic tone.
Emotionally arousing events are readily encoded into memory. For this reason, arousing stimuli are widely used as reinforcers in animal models of memory formation. Moreover, studies of fear conditioning in mammals indicate that brain structures implicated in specific aspects of arousal, like the control of emotions or the stress response, can also contribute to memory encoding. For example, the amygdala, which is implicated in processing emotional stimuli, is critically involved in memory consolidation (Davis and Whalen 2001; LeDoux 2000; McGaugh et al. 2000; Zald 2003). In addition, the nucleus of the solitary tract, which participates in vagal motor control and cardiovascular regulation during stress, is also implicated in memory encoding (Clayton and Williams 2000; Williams and McGaugh 1993; Williams et al. 2000). While the connection between arousing stimuli and memory formation is well established, the cellular mechanisms mediating this connection are less well understood.
The marine mollusk Aplysia provides a useful model for studying the interactions between arousal and memory encoding in a simple nervous system. Sensitization, a form of nonassociative learning, is typically induced by an aversive stimulus such as tail-shock, which also induces defensive arousal, a general alarm response characterized by escape locomotion and increased heart rate (Carew et al. 1971; Dieringer et al. 1978; Pinsker et al. 1973). The neurotransmitter serotonin (5-HT) is thought to contribute importantly to both the induction of sensitization and to the triggering of the alarm response. Sensitization of defensive reflexes is blocked by a chemical lesion of the serotonergic system (Glanzman et al. 1989), and facilitation of sensory neuron (SN) to motor neuron (MN) transmission, a major cellular correlate of sensitization, is induced by 5-HT application and blocked by 5-HT receptor antagonists (Brunelli et al. 1976; Mercer et al. 1991). Moreover, 5-HT is released in the Aplysia hemolymph and in the CNS in response to noxious stimulation (Levenson et al. 1999; Marinesco and Carew 2002). 5-HT is also important for the regulation of locomotion and heart rate (Koester et al. 1974; Liebeswar et al. 1975; Mackey and Carew 1983; McPherson and Blankenship 1992; Palovcik et al. 1982; Parsons and Pinsker 1989), two important behavioral manifestations of defensive arousal in Aplysia. Although many serotonergic neurons have been identified using immunohistochemical approaches (Goldstein et al. 1984; Longley and Longley 1986; Nolen and Carew 1994; Ono and McCaman 1984), only two serotonergic neurons, the CC3/CB1 cells, have been shown to increase their firing rate in response to tail-shock (Mackey et al. 1989; Wright et al. 1995). Thus it remains unclear whether the effects of sensitization training are restricted only to the CC3/CB1 neurons or whether it recruits a more global serotonergic system.
In this study, we sought to better understand the effects of sensitization training on the Aplysia serotonergic system. We performed intracellular recordings in Aplysia serotonergic neurons that were visually identified by prelabeling them with 5,7-dihydroxytryptamine (5,7-DHT) (Jahan-Parwar et al. 1987; Kemenes et al. 1989; McPherson and Blankenship 1991a). We systematically recorded from cells throughout the serotonergic system and found that the majority of them (∼80%) increase their firing rate for several minutes in response to a brief tail-nerve shock. Interestingly, a large subset of these serotonergic neurons are also excited by 5-HT and are interconnected by excitatory synapses, suggesting that they are organized as a distributed network capable of sustained activation during the induction of sensitization. In particular, we found that the CC3/CB1 neurons excite many follower serotonergic neurons in the pedal and abdominal ganglia. These data indicate that sensitizing stimuli induce a widespread and persistent activation of the serotonergic system. In the companion paper, we examine the functional consequences of such an increase in serotonergic activity or tone (Marinesco et al. 2004).
Some of the data in this paper have been presented in abstract form (Kolkman et al. 2003).
Wild-caught adult Aplysia californica (Marinus, Long Beach, CA) weighing ∼250 g were used throughout this study. Animals were housed in tanks containing aerated artificial seawater (ASW) kept at 15°C. To achieve a brown pigmentation in serotonergic neurons, animals were injected twice with 10 mg/kg 5,7-DHT, dissolved at 5 mg/ml in ASW, at a 1-wk interval. Ascorbic acid (10 mg/kg) was added as an antioxidant to prevent 5,7-DHT from being degraded while in solution (Jahan-Parwar et al. 1987; Venton et al. 2002). Drugs were administered by intrahemocoel injections through the foot. Animals were then left undisturbed for 2 mo before experimentation (Jahan-Parwar et al. 1987). To prepare the tissue samples, animals were first anesthetized by injection of 150–200 ml of 0.35 M MgCl2 and immediately dissected for removal of the entire CNS without the buccal ganglion. The ganglia were then fixed in 0.4% glutaraldehyde for ∼30 s and desheathed in 50:50 artificial seawater (ASW): 0.35 M MgCl2, to expose the neurons. Ganglia were then kept in a recording chamber continuously perfused with ASW [which contained (in mM) 460 NaCl, 55 MgCl2, 11 CaCl2, 10 KCl, and 10 Tris]. In some experiments, the composition of ASW was changed to achieve 0 Ca2+ and high (3 times) Mg2+ concentrations [contents (in mM): 295 NaCl, 10 KCl, 176 MgCl2, and 10 Trizma] or high (3 times) Ca2+ and Mg2+ concentrations [containing (in mM) 328 NaCl, 33 CaCl2, 10 KCl, 165 MgCl2, and 10 Trizma]. In each case, ASW was buffered to a pH of 7.6 before use.
The CNS was placed in a recording chamber and observed with a stereomicroscope to identify cells labeled with the 5,7-DHT brown pigmentation. Neurons were impaled with intracellular glass micropipettes (5–15 MΩ) filled with 3 M KCl. Membrane potential was measured using an Axoclamp 2B amplifier (Axon instruments, Union City, CA), digitized with an ITC-16 AD-DA computer interface (Instrutech, Great Neck, NY), and recorded on computer file with a homemade software written with Igor Pro 4.03 (Wavemetrics, Lake Oswego, OR). Intracellular recordings were performed beginning 15 min before and continuing 45 min after tail-nerve shock (2 s at 40 Hz, 5-ms pulses at 30 V). Input resistance of putative serotonergic neurons stained with 5,7-DHT was measured by injecting several 500-ms, 1-nA hyperpolarizing current pulses. Only those pulses that were not contaminated by spontaneous electrical activity were taken into account. After intracellular recordings, the same cells were re-impaled with microelectrodes filled with Neurobiotin (5%, Vector Labs, Burlingame, CA) in 1 M KCl. Neurobiotin was injected ionophoretically for 45 min to 1 h (+5-nA pulses, 500-ms duration at 1 Hz) to trace their processes in the pedal ganglion and allow immunohistochemical confirmation of their serotonergic nature.
When studying the heart-exciter cell RBhe, it was necessary to keep the entire CNS connected to the heart through the auricular branch of the pericardial nerve. For this purpose, the pericardial nerve was dissected until it reached the kidney and was no longer visible. The auricle and the piece of the kidney presumed to contain the pericardial nerve was dissected and pinned in the recording chamber together with the CNS. The auricle was then washed with flexible tubing and left undisturbed for 30 min before starting intracellular recordings. Neurons labeled with 5,7-DHT in the RB cluster were systematically impaled and stimulated by intracellular current injection. RBhe was identified by its ability to elicit heart contractions on stimulation (J. D. Koester, personal communication).
To assess synaptic transmission between tail SNs and tail MNs, we recorded intracellularly from one tail SN and one tail MN connected by a synapse as already described in a previous study (Marinesco and Carew 2002). For excitability measurements, the SN was depolarized for 300 ms every 15 min, using a current intensity sufficient to elicit one or two spikes under baseline conditions. After three stable pretests, tail-nerve was stimulated for 2 s at 40 Hz (5-ms, 30-V pulses). One minute after the shock, the number of spikes evoked in the SN by the same depolarizing pulse provided a measure of SN excitability. Short-term facilitation was evaluated by measuring the size of the SN-evoked excitatory postsynaptic potential (EPSP) in the MN (or the size of the 1st EPSP when multiple spikes were evoked in the SN) (see also Bunge et al. 1997). Input resistance was measured by current injection (−1 nA) into the recorded neuron.
The ganglia were fixed overnight with 4% formaldehyde (in 0.01M PBS with 20% sucrose) at 4°C and permeabilized with 4% Triton (Sigma, St Louis, MO) in PBS for 1 h at room temperature. Nonspecific binding was blocked by immersion in 10% goat serum (Jackson ImmunoResearch Laboratories, West Grove, PA) with 0.5% Triton for 1 h before exposure to the rabbit polyclonal anti-serotonin antibody (Diasorin, Stillwater, MN) at 1/1000 for 1.5–2.5 days at 4°C. Visualization of the primary antibody to 5-HT was performed with a goat anti-rabbit secondary antibody coupled to Cy5 (Jackson ImmunoResearch Laboratories, dilution 1/100, 3 h at room temperature). Neurobiotin was then revealed using Streptavidin coupled to Alexa Fluor 488 (Molecular Probes, Eugene, OR, 10 μg/ml, overnight at 4°C). Fluorescence images were acquired with a BioRad confocal microscope (Bio Rad, Hercules, CA). Excitation was performed at 488- and 647-nm wavelengths by a Kr/Ar mixed gas laser. We used standard T1/E2 filters with two detection filters of 522/35 nm (green pseudocolor, Figs. 3–8) and 680/32 nm (red pseudocolor). These two wavelengths corresponded to the fluorescent dyes used in our experiments, Alexa 488 and Cy 5, respectively.
Immunohistochemistry allowed us to determine whether a given cell was serotonergic. Most often, the difference between signal and background was sufficient to allow unambiguous determination of the serotonergic nature of the cell. In a few ambiguous cases, we quantified fluorescence intensity in several unlabeled areas using the National Institutes of Health image software. A given cell was considered serotonergic if its fluorescence intensity was >2 SD above mean background level.
Chronoamperometric detection of 5-HT was achieved as described previously by Marinesco and Carew (2002). Briefly, carbon fiber electrodes (7 μm diam, 300 μm length) were implanted in the neuropil of pedal or pleural ganglia. Chronoamperometry was performed with four successive voltage steps (80 mV, 40 ms; 230 mV, 15 ms; 250 mV, 40 ms; and 400 mV, 15 ms) applied between the working and reference electrodes at 2 Hz. Oxidation currents were measured using a VA10 voltammeter (NPI Electronic, Tamm, Germany) connected to a three-electrode potentiostat. Serotonin-related oxidation currents were measured as the difference between the currents measured at the end of the fourth pulse (400 mV) and the third pulse (250 mV). The first two pulses provided a control for dopamine release.
Data were presented as means ± SE. Comparisons between two data groups were performed using the Student's t-test for equal or unequal variances as determined by the F-test (significance level was P < 0.05). For comparisons among three or more data groups, we used an ANOVA followed by an LSD post hoc test (P < 0.05). Statistics software was the analysis tool-pack of Microsoft Excel 2000 and SPSS for Windows version 10.
Labeling of serotonergic neurons in vivo using 5,7-DHT
5,7-DHT, a well-known neuronal toxin capable of lesioning serotonergic and noradrenergic neurons in mammals (Baumgarten et al. 1973; Pujol et al. 1978), is usually much less effective in mollusks (Osborne and Cottrell 1974). Except when very large concentrations are used (Glanzman et al. 1989), serotonergic neurons in Lymnea stagnalis, Helix pomatia, and two species of Aplysia (californica and brasiliana) survive treatment with this drug, while developing a dark brown pigmentation within 4–8 wk (Jahan-Parwar et al. 1987; Kemenes et al. 1989; McPherson and Blankenship 1991a, 1992; Rozsa et al. 1986). This pigmentation develops in ∼85% of neurons immunoreactive to 5-HT in Lymnea (Kemenes et al. 1989). We used this approach to undertake a systematic study of the serotonergic system in Aplysia and its role in the encoding of an aversive reinforcer like tail-shock. In the first phase of the study, we sought to determine if serotonergic neurons in 5,7-DHT-treated animals retained their ability to release 5-HT and to exert their known modulatory actions and the degree of overlap between 5,7-DHT labeling and 5-HT immunoreactivity in the CNS.
About 2 mo after administration of 5,7-DHT, many neurons in the CNS showed a dark brown pigmentation (Fig. 1A), as described previously (Jahan-Parwar et al. 1987; McPherson and Blankenship 1991a, 1992). A previous study comparing firing properties of homologous neurons in 5,7-DHT-treated and control animals concluded that 5,7-DHT treatment did not affect the physiological properties of serotonergic neurons (Jahan-Parwar et al. 1987). We confirmed this finding using three different approaches: 1) one of the best-known neuromodulatory actions of serotonergic neurons in Aplysia is increased SN excitability and facilitation of SN-MN transmission. Tail or tail-nerve stimulation increases SN excitability and facilitates SN-MN synapses at least in part through the release of 5-HT (Brunelli et al. 1976; Glanzman et al. 1989; Mackey et al. 1989; Marinesco and Carew 2002). We found that animals treated with 5,7-DHT showed the typical increase in SN excitability and SN-MN synaptic efficacy after tail-nerve shock (Fig. 1B). A depolarizing pulse sufficient to evoke a single action potential in a SN produced an average of 5.7 spikes 1 min after ipsilateral tail-nerve shock (n = 4). Similarly SN-MN synapses were facilitated to 292% of their baseline level 1 min after ipsilateral tail-nerve shock (n = 4).
Second, 5-HT release evoked by tail-nerve shock in the pleural and pedal ganglia could be clearly detected by carbon-fiber electrodes (Fig. 1C). 5-HT release averaged 38.3 ± 5.5 nM in the pleural ganglion and 58 ± 12.7 nM in the pedal ganglion (n = 4), consistent with previous data obtained in control animals (Marinesco and Carew 2002). It is unlikely that the electrochemical signal reflects 5,7-DHT released as a false transmitter because our electrodes did not detect 500 nM 5,7-DHT, a concentration much higher than the endogenous 5-HT released in response to tail-nerve shock (data not shown).
Finally, the serotonergic heart exciter neuron RBhe was recorded several times in this study (see following text). To identify RBhe among all 5,7-DHT-stained neurons within the RB cluster of the abdominal ganglion, we relied on its ability to evoke heart contractions in response to intracellular activation, an effect that depends largely on 5-HT (Liebeswar et al. 1975). Visual observation indicated that intracellular activation of 5,7-DHT-stained RBhe neurons produced heart contractions apparently similar to those produced in control animals (Mayeri et al. 1974). Collectively, these and previous data (Jahan-Parwar et al. 1987) strongly support the view that the physiological properties of serotonergic neurons are intact 2 mo after 5,7-DHT treatment. In a previous study reporting deleterious effect of 5,7-DHT on serotonergic neurons in Aplysia, the dose administered to the animals was 200 times higher than in this study (2 g/kg vs. 10 mg/kg) (Glanzman et al. 1989).
We then compared 5,7-DHT staining with 5-HT immunoreactivity in ganglia that had been processed for both labeling methods (Fig. 2). The two labeling techniques did not cross-react because the commercial antibody that we used does not recognize 5,7-DHT (Marchant et al. 1997). We consistently detected a total of ∼100 serotonin-containing neurons in the cerebral, pedal and abdominal ganglia. In the cerebral ganglion, 5-HT immunoreactivity and 5,7-DHT staining were restricted to the dorsal surface. The metacerebral cells (MCCs) were consistently labeled with both methods, together with a variable number of zero to six neurons in their immediate vicinity, in the G clusters (CG). Two bilateral pairs of neurons were also detected in the C clusters: the well-known CC3/CB1 cells (Mackey et al. 1989; Wright et al. 1995; Xin et al. 2001), and two previously unidentified cells, that we named CC8 (Fig. 2, A and B). A few small-diameter serotonin-containing cells were occasionally detected in the C cluster, but their labeling with 5,7-DHT was quite variable. In the pedal ganglion, 25–40 neurons stained with 5,7-DHT and 5-HT antibodies formed a large group of cells that wrapped around the pedal commissure, in a semi-circular shape (Fig. 2C). Because these neurons shared similar morphological and electrophysiological properties, we named this group the pedal serotonergic (PS) cluster. Finally, the abdominal ganglion contained 15–30 neurons positive for both staining methods, most of them on the dorsal surface: 15–25 in the RB cluster and occasionally, ≤5 neurons on the left hemiganglion (Fig. 2, A and D). These neurons (on the left side) have been described by other authors but have not been identified and incorporated into conventional nomenclature (Hawkins 1989; Kistler et al. 1985). In this paper, we refer to these neurons as belonging to the “left” abdominal hemiganglion (Table 1). Rarely, a few neurons were labeled on the ventral surface of the abdominal ganglion but could not be systematically studied. These numbers are in general agreement with previous anatomical studies using immunohistochemistry (Goldstein et al. 1984; Longley and Longley 1986; Nolen and Carew 1994; Ono and McCaman 1984) or the glyoxylic acid histofluorescence technique (Tritt et al. 1983). However, larger numbers of serotonin-positive neurons have been detected when using sections of ganglia (Hawkins 1989; Kistler et al. 1985). Indeed, a significant number of small-diameter neurons appear to lie underneath the surface of the ganglion and form a deep layer of cell bodies that are difficult to observe in intact ganglia. Thus our examination may have underestimated the total number of serotonergic neurons in the Aplysia CNS.
To determine the extent of colocalization between 5,7-DHT labeling and 5-HT immunoreactivity, we systematically counted cells that were stained by each method in five animals (7 pedal, 5 abdominal, and 5 cerebral ganglia, Fig. 2, B–D). A total of 82.3% of cells showing brown pigmentation were also labeled with the anti-5-HT antibody (195/239 cells), showing that 5,7-DHT efficiently labeled serotonergic neurons in the CNS. The total number of cells showing unequivocal brown pigmentation was lower than the total number of serotonergic cells identified with immunohistochemistry. In particular, only two neurons were detected in the cerebral C cluster using 5,7-DHT, whereas four to five serotonergic cells have been detected with immunohistochemistry (McPherson and Katz 2001; Nolen and Carew 1994). However, the majority of serotonergic cells were stained by both methods (195/284 cells, 69%), which is consistent with data obtained in the Lymnea CNS (Kemenes et al. 1989).
These data indicate that in Aplysia, 5,7-DHT treatment specifically labels serotonergic neurons in vivo without affecting their physiological properties. Thus using this method, it was possible to record from serotonergic neurons systematically in the living CNS and to determine their response to an in vitro analog of sensitization training, tail-nerve stimulation. However, to control for occasional false positives (labeled with 5,7-DHT but not immunoreactive to 5-HT), we always injected the recorded cell with the neuronal tracer Neurobiotin to subsequently confirm its serotonergic nature with immunohistochemistry. A few experiments (15–20% of total recordings) were discarded because the recorded cell turned out to be nonserotonergic.
Identification of novel serotonergic neurons
A pair of bilateral symmetrical neurons consistently stained for both 5,7-DHT and 5-HT immunohistochemistry in the C cluster of the cerebral ganglion (Fig. 3A). To our knowledge, these neurons have not been previously identified. According to common nomenclature (Frazier et al. 1967; Jahan-Parwar and Fredman 1976), we named them CC8s. Each CC8 cell projected contralaterally to the anterior tentacular nerve (AT) and in a few cases to the upper labial nerve (ULAB, Fig. 3A). Their main neurites overlapped over a distance of 2–3 mm as they traveled to the contralateral side of the ganglion. Intracellular recordings revealed a very powerful electrical coupling that made the cells fire in complete synchrony (Fig. 3, B and C). The function of these neurons remains unknown although they appeared to increase their firing rate in response to tail-nerve stimulation (see following text). It is possible that they release 5-HT within the anterior tentacles and/or lips and thereby modify chemosensory processing, a major function associated with these structures.
Activation of the serotonergic system after tail-nerve stimulation
To examine the response of the serotonergic system to sensitizing stimuli, we systematically recorded from serotonergic neurons stained by both 5,7-DHT and 5-HT immunohistochemistry before and after tail-nerve stimulation, an in vitro analog of sensitization. We gathered a total of 103 recordings in 65 different serotonergic neurons, where we measured their firing rate before and after ipsi- and/or contralateral tail-nerve shock. A given cell was determined to be “activated” by tail-nerve stimulation when its firing rate exceeded its baseline firing rate +2.5 SD, during ≥1 min within the 5 min after the stimulation. We also measured the input resistance of the cell before and after the experiment.
The vast majority of recorded serotonergic neurons were activated after tail-nerve shock: 74% of the neurons increased their firing rate after ipsilateral tail-nerve stimulation (45/58 cells) and 78% after contralateral shock (32/41 cells, see Table 1, Bold). Most activated cells increased their firing rate within the first minute after stimulation (51/54 cells that showed an increased firing rate). Moreover, the input resistance was increased in 80% of the recorded serotonergic neurons (35/44 cells, see Table 1, italic).
Most identified serotonergic cell types increased their firing rate in response to tail-nerve shock. These include Parapodia-opener-phase-like neurons, RBhe, CC3/CB1 neurons, CC8 neurons, and metacerebral cells.
Parapodia-opener-phase-like (POP-like) neurons
In swimming Aplysia species, POP neurons located in the pedal ganglion modulate the neuromuscular junctions in parapodial swim muscles, enhancing the power and timing of parapodial movements during swimming (McPherson and Blankenship 1991a; Parsons and Pinsker 1989). In A. californica, a crawling species, homologous neurons, named POP-like cells enhance contractions of foot and body-wall muscles during escape locomotion (McPherson and Blankenship 1992). These neurons are located in the PS cluster and were labeled with both 5,7-DHT treatment and 5-HT immunohistochemistry. Moreover, these cells are easily identified by their bursting firing pattern, which is believed to synchronize with pedal motor neurons during escape locomotion (McPherson and Blankenship 1992). About half of the neurons that we recorded from within the PS cluster were POP-like neurons (11/25 cells). They responded to both ipsi- and contralateral tail-nerve shock with a bursting pattern that lasted ≥5 min (Fig. 4, B and C; Table 1). The majority of these neurons already showed a bursting firing pattern at rest (see Fig. 3A in companion paper) but inter- and intra-burst frequency as well as burst duration increased after shock. In four POP-like neurons, a neurite could be traced down to peripheral nerves P8 (body-wall nerve) and/or P9 (tail-nerve) confirming the view that these neurons modulate foot and body-wall muscles (McPherson and Blankenship 1992).
The heart exciter cell RBhe is usually located in the mediorostral part of the RB cluster of the abdominal ganglion (Koester et al. 1974; Mayeri et al. 1974). In eight preparations, the CNS was kept connected to the heart through the genital/pericardial nerve in the recording chamber (see methods). RBhe was consistently labeled by both 5,7-DHT treatment and 5-HT immunohistochemistry (Fig. 5) and was identified when repetitive firing by intracellular stimulation elicited one or several heart contractions. Although RBhe is usually considered unique (Koester et al. 1974), in two preparations, we found two distinct serotonergic cells exciting the heart. In these preparations, one cell had a large diameter (∼200 μm) and was located rostral to the RB cluster, whereas the other had a smaller diameter and was located deep into the groove separating the two hemiganglia. Although we do not know whether these two cells can be considered equivalent (it is unclear, for example, whether the smaller cell increases heart contractions through mono- or polysynaptic pathways), it is possible that more than one RBhe-like cell may exist in this region of the abdominal ganglion. RBhe was tonically active at rest, with firing frequencies ranging from 15 to 25 spikes/min (except in 1 experiment where baseline firing rate was 136 spikes/min, see Table 1). It increased its firing rate in response to ipsi- and contralateral tail-nerve shock in all four preparations that we studied (Fig. 5, B and C; Table 1). The increase in firing rate was variable, lasting between 2 and 40 min. Moreover, input resistance was increased at the end of the experiments (45 min after shock) in all four RBhe neurons tested (Table 1). These data are consistent with recent studies by Koester et al. (2000).
This cell type is present as a bilateral pair of neurons located in the B (Mackey et al. 1989; Wright et al. 1995) or the C cluster (Xin et al. 2001) of the cerebral ganglion and projects bilaterally to pleural, pedal, and abdominal ganglia (Fig. 6A) (Mackey et al. 1989; Wright et al. 1995; Xin et al. 2001). CC3/CB1 has been reported to produce presynaptic facilitation in siphon SN-MN synapses in the abdominal ganglion and is therefore considered to be involved in sensitization (Mackey et al. 1989). In our recordings, CC3/CB1 was tonically active at rest, with firing frequencies ranging from 5 to 60 spikes/min. In two preparations, CC3/CB1 firing rate and input resistance was increased after ipsi- and contralateral tail-nerve shock (Fig. 6, B and C; Table 1). In one preparation, CC3/CB1 increased its firing rate after contralateral but not ipsilateral stimulation (Table 1). These data confirm earlier results by other authors showing that CC3/CB1 was activated by tail- or tail-nerve shock (Mackey et al. 1989; Wright et al. 1995).
This newly identified pair of electrically connected cells (see above) was recorded in four preparations. CC8 was tonically active at rest, firing at higher frequencies than most Aplysia serotonergic neurons (between 35 and 125 spikes/min, Fig. 7). In three preparations, CC8 firing frequency and input resistance were increased after ipsi- and contralateral tail-nerve stimulation (Fig. 7, B and C; Table 1). In one preparation, CC8 firing rate was increased after contralateral but not ipsilateral stimulation (Table 1). In one experiment, we could record intracellularly from both CC8 neurons. They fired in complete synchrony during the entire recording, both increasing their firing rate after tail-nerve shock.
The only serotonergic cell type that did not respond to tail-nerve stimulation was the metacerebral cells (Fig. 8). These cells were usually silent in the isolated CNS (8/11 recordings) or fired at a low frequency (0.1–1 spike/min in 3 preparations), confirming earlier studies (Gerschenfeld et al. 1978). In four preparations, the recorded metacerebral cell did not fire after tail-nerve stimulation despite a slight depolarization in two cases (Fig. 8). Only one recording showed an increase in firing rate after tail-nerve stimulation (see Table 1).
Having characterized the response profile of several classes of serotonergic cells, we next quantified and analyzed the effects of tail-nerve shock on the firing rate and input resistance of serotonergic neurons. Among all 103 recordings, 68 were performed for ≥10 min before and 40 min after shock and were included in the analysis. Metacerebral cell recordings were excluded from the data set because these neurons did not show any consistent sign of activation after tail-nerve shock. Firing frequencies showed a distribution that was nonnormal and was shifted to the right 1 min after tail-nerve stimulation (Fig. 9A). Thus the mean firing rate of serotonergic neurons was significantly increased after tail-nerve stimulation (Fig. 9B; Wilcoxon signed-rank test, z = −6.7, P < 0.01). The firing rate of individual neurons was standardized to its baseline firing rate during the 10 min preceding tail-nerve shock (z scores, Fig. 9C). Serotonergic neurons in the pedal and abdominal ganglia showed a similar response to tail-nerve shock characterized by a pronounced increase in firing rate lasting ∼5 min (Fig. 9D). Serotonergic neurons in the cerebral ganglion also increased their firing rate after tail-nerve shock, although CC8 and CG neurons seemed to stay activated for a longer time (Fig. 9E). The increase in firing rate typically occurred within a few seconds, reached its maximum during the first minute after tail-nerve shock, and slowly returned to its baseline level (Fig. 9, C and D). In only four recordings (1 CC3 neuron, 1 CC8 neuron, 1 neuron in the RB cluster, and 1 in the left abdominal cluster), the maximum firing rate was reached ≥2 min after tail-nerve shock. To further analyze the duration of the effects of tail-nerve shock on firing rate, we defined the duration of activation as the time between shock and the first 1-min period during which the firing rate of the cell went below its mean value during the 10 min preceding tail-nerve shock (baseline preshock value). When considering pedal and abdominal serotonergic neurons together, duration of activation was normally distributed around 4.2 min, indicating a common pattern of activation among these neurons (Fig. 10A). However, the duration of activation was variable in the cerebral CG, CC3/CB1 and CC8 cells (Fig. 10B). Finally, input resistances showed a normal distribution and increased from 13.8 ± 1.1 to 17.9 ± 1.3 MΩ by the end of the experiment (P < 0.01, Fig. 11). Overall, these data indicate that a large percentage of serotonergic neurons (excepting the metacerebral cells) are activated by tail-nerve shock, suggesting a general role in the neuronal encoding of the noxious stimuli used to produce sensitization.
Aplysia serotonergic neurons are excited by 5-HT
We next investigated the mechanisms responsible for the increase in firing evoked by tail-nerve shock in the Aplysia serotonergic system. In mammals, the firing rate of serotonergic neurons in the dorsal raphe nucleus is regulated by an autoinhibitory feedback mechanism mediated by 5-HT1a receptors (review in Barnes and Sharp 1999; Pineyro and Blier 1999). Thus application of 5-HT or 5-HT1a agonists onto serotonergic neurons typically results in the inhibition of their firing rate, an effect thought to participate in the regulation of the serotonergic tone during behavior (Blier and de Montigny 1987; however, see Johnson et al. 2002). However, in a number of invertebrate species including Pleurobranchaea californica and Aplysia brasiliana, some serotonergic neurons can be excited by 5-HT (Parsons and Pinsker 1989; Sudlow and Gillette 1995). We investigated the effects of 5-HT on the serotonergic system in Aplysia by determining the effects of bath application of 5-HT (10 μM) on the firing rate of identified serotonergic neurons.
Most serotonergic neurons responded to 5-HT with an increase in firing rate. The effect was most prominent in abdominal and pedal serotonergic neurons (Fig. 12, A and B). They depolarized and increased their firing rate within the first minute of 5-HT application. Firing rate increased from 19.1 ± 5.3 to 66.1 ± 7.8 spikes/min in the abdominal ganglion (P < 0.01, n = 8) and from 16.6 ± 2.8 to 234.5 ± 25 spikes/min in the pedal ganglion (P < 0.01, n = 10). When 5-HT was washed out of the bath, firing rate slowly returned to baseline levels within 3–4 min (Fig. 12, A and B). When 5-HT was applied again in 0 Ca2+, high-Mg2+ ASW, to block neurotransmitter release in the ganglion, abdominal, and pedal serotonergic neurons still depolarized and fired, suggesting that 5-HT acted directly on serotonergic neurons (an example of this effect is shown in Fig. 3C of the companion paper that follows).
The effects of 5-HT in the cerebral ganglion were mixed. Metacerebral, CC3/CB1, and CC8 neurons did not significantly change their firing rate in response to 5-HT (n = 6, 5, and 6 respectively; Fig. 12C). However, we did observe a significant increase in firing rate in CG serotonergic neurons. Their firing rate increased from 6.7 ± 2.8 to 116 ± 22.8 spikes/min and returned to baseline during washout (n = 5, P = 0.01; Fig. 12C). Finally, 5-HT was ineffective in all cerebral serotonergic cell types when it was perfused in 0 Ca2+, high-Mg2+ ASW, suggesting that 5-HT effects, if any, were mediated by interneurons.
These data indicate that the 5-HT1a-mediated autoinhibitory feedback mechanism found in mammalian serotonergic neurons is not present in Aplysia. On the contrary, many serotonergic neurons in Aplysia are excited by 5-HT. These data suggest the existence of a positive feedback mechanism that could contribute to the widespread excitation in serotonergic neurons during sensitization training. Specifically, it is possible that excitatory connections between serotonergic neurons contribute to spreading excitation throughout the serotonergic system after tail-nerve shock.
Excitatory connections between identified serotonergic neurons
We next examined whether serotonergic neurons were synaptically connected with one another, in particular those that were activated by exogenous 5-HT. To this end, we systematically recorded from pairs of serotonergic neurons labeled by 5,7-DHT throughout the CNS. A burst of action potentials was evoked in one neuron by intracellular current injection while we recorded the membrane potential of the other. We never observed functional connections between pairs of neurons in the pedal or abdominal ganglia or between individual neurons in each pedal ganglion (n = 125 pairs). To obtain a generalized activation of a population of serotonergic neurons in one pedal ganglion, we placed two pedal ganglia connected by the pedal commissure in a split chamber where each pedal ganglion was perfused independently. We then perfused 5-HT (10 μM) onto one pedal ganglion to induce repetitive firing in virtually all serotonergic neurons (see preceding text), while recording from one serotonergic neuron in the contralateral pedal ganglion. Increased firing induced by 5-HT in one pedal ganglion did not produce any noticeable effects in contralateral serotonergic neurons (n = 4).
In contrast to the lack of detectable interconnectivity in the preceding text, we did observe excitatory connections between CC3/CB1 cells and pedal and abdominal serotonergic neurons (Fig. 13). A 1- to 2-s burst of action potentials evoked by current injection into CC3/CB1 increased the firing rate of pedal and abdominal serotonergic neurons and produced a slow EPSP when the postsynaptic cells were hyperpolarized to −70 mV (Fig. 13, A and B). The excitatory effect of CC3/CB1 stimulation on pedal serotonergic cells was rather rare in the pedal ganglia (4/66 recordings, Fig. 13B). Interestingly, all four slow EPSPs that we observed in the pedal ganglion occurred in response to stimulation of the contralateral CC3/CB1 (4/39 recordings) but not the ipsilateral one (0/27 recordings), which is surprising, given that each CC3/CB1 is thought to project to both pleural and pedal ganglia (Xin et al. 2001). However, earlier data by Wright et al. (1995) suggested that although each CC3/CB1 neuron projects to both pleural ganglia, only one, presumably the contralateral one, reaches the pedal ganglion. Further studies will be required to elucidate the precise projection pattern of the CC3/CB1 neurons. In the abdominal ganglion, slow EPSPs evoked by intracellular activation of CC3/CB1 were observed much more frequently, in >90% of the recorded serotonergic neurons (47/50 recordings, Fig. 13A). Among these recordings, 46 were obtained from serotonergic cells in the RB cluster (43/46 connected to CC3/CB1), and 4 from the left hemiganglion (4/4 connected to CC3/CB1). The majority of these abdominal cells received bilateral input from both CC3/CB1 neurons (14/16 recordings). Finally, RBhe also received excitatory input from both CC3/CB1 neurons (4/5 recordings, Fig. 13A), confirming Xin et al. (2001).
To determine whether the excitatory connections between CC3/CB1 and serotonergic cells in the pedal and abdominal ganglia were mono- or polysynaptic, we perfused the CNS with ASW containing three times the normal concentration of Mg2+ and Ca2+ ions to reduce the likelihood of interneuronal firing. Due to the increased firing threshold induced by the high divalent solution, the current required to evoke a burst of action potentials in CC3/CB1 was much higher than in normal ASW, so it was difficult to evoke the same number of spikes as in normal ASW. However, we were still able to observe slow EPSPs in a majority of serotonergic neurons in the abdominal ganglion (16/27 recordings, Fig. 13C), suggesting that at least some of the connections between CC3/CB1 and these neurons may have a monosynaptic component. The presence of such slow EPSPs is consistent with the involvement of G-protein-coupled receptors such as those activating a cyclic AMP-gated sodium current in the marine mollusk P. californica (Sudlow and Gillette 1995). Because CC3/CB1 neurons contain 5-HT and pedal/abdominal serotonergic cells are excited by 5-HT, it is likely that this monosynaptic component reflects a serotonergic excitatory connection. However, we cannot exclude the possibility that transmission between CC3/CB1 and other serotonergic neurons is mediated by interneurons or by another transmitter co-released with 5-HT.
Collectively these data indicate that excitatory, presumably serotonergic, connections exist between identified serotonergic neurons in the Aplysia CNS, mainly between CC3/CB1 and other serotonergic neurons in the pedal and abdominal ganglia. CC3/CB1 might therefore contribute to the generalized increase in firing in the serotonergic system after tail-nerve stimulation by spreading excitation to other serotonergic neurons.
Cartography of major serotonergic projections
To better characterize the projection sites receiving serotonergic input during sensitization, we analyzed the Neurobiotin labelings, used to confirm the serotonergic nature of the recorded neurons, to trace their projections. We compiled data from a total of 153 experiments. In 68 preparations, Neurobiotin staining revealed long-distance projections to other ganglia or peripheral targets. In other preparations, Neurobiotin staining could not be observed outside the ganglion where it was injected. Lack of a signal in a nerve could result from two primary sources: we injected a local interneuron or we injected a projection neuron but the dye did not travel far enough to be detected in a nerve. To distinguish between these two possibilities, we estimated the number of times our injections did not travel far enough into the neurites to be detected in nerves by counting the total number of cases in which Neurobiotin could be traced outside the ganglion after being injected in a known projection neuron. Specifically, we examined RBhe (projecting to the genital/pericardial nerve, 4 of 7 injections), CC3/CB1 (projecting to pleural-cerebral nerves, 9/11 injections), and CC8 (projecting to the anterior tentacular nerve and/or to the upper labial nerve, 13/16 injections). Failure to detect a signal in a nerve occurred at a rate of ∼24% and could account for about half of the cases where staining was not observed outside the ganglion. Thus the remaining half likely reflects local interneurons.
Serotonergic neurons in the pedal ganglia projected to three major nerves: P8 (body wall, n = 5), P9 (tail, n = 8), and the pedal commissure (PC, n = 3, Fig. 14). We never observed projections to the pleural ganglion or into the cerebral-pedal connective. Because activating serotonergic neurons in one pedal ganglion did not produce observable changes in contralateral serotonergic neurons (see preceding text), it is possible that these projections through the pedal commissure do not release 5-HT into the contralateral pedal ganglion but travel to further peripheral targets.
In the abdominal ganglion, serotonergic neurons also projected to the periphery through the siphon nerve (n = 3), the genital/pericardial nerve (n = 12) and the branchial nerve (n = 15, Fig. 14). We did not observe any projections to the rest of CNS, via pleural-abdominal connectives. In the cerebral ganglion, the following pattern was observed (Fig. 14): 1) CC8 projected to the contralateral AT and/or ULAB (n = 14), 2) serotonergic neurons in the CG cluster projected mainly to AT and ULAB (n = 6), and in one preparation, into another nerve that we could not identify, 3) the metacerebral cells projected to the cerebral-buccal connective (n = 3), as previously described by other authors (Kistler and Schwartz 1982; Weiss et al. 1978), and 4) CC3/CB1 projected to both cerebral-pleural connective (n = 9), as described by other authors (Mackey et al. 1989; McPherson and Blankenship 1991b; Wright et al. 1995; Xin et al. 2001). The ability of CC3s (CB1s) to produce slow monosynaptic EPSPs in the abdominal and pedal ganglia, together with data from previous studies (Mackey et al. 1989; McPherson and Blankenship 1991b; Wright et al. 1995; Xin et al. 2001), indicates that these neurons project to the pleural, pedal, and abdominal ganglia (Fig. 14).
Although specific serotonergic projections might have been overlooked in our Neurobiotin injections, these data indicate that serotonergic neurons project to a wide variety of central and peripheral targets. Therefore the generalized increase in firing rate throughout the Aplysia serotonergic system after tail-nerve stimulation probably results in 5-HT release in central ganglia as well as peripheral organs such as tentacles, lips, tail, body wall, siphon, heart, and gill (and/or gill ganglion).
The principal goal of this study was to examine the functional properties of the Aplysia serotonergic system in response to sensitizing stimuli. Toward that end, we first visually identified serotonergic neurons in living ganglia using 5,7-DHT injections (Jahan-Parwar et al. 1987; Kemenes et al. 1989; McPherson and Blankenship 1991a). This labeling technique did not change the physiological properties of these neurons and tremendously increased the probability of recording from serotonergic neurons in the isolated CNS. Although 5-HT immunoreactivity does not provide proof that stained neurons actually synthesize 5-HT, a recent study in adult Aplysia CNS found that patterns of immunoreactivity to 5-HT and 5-HTP, the 5-HT precursor presumably reflecting 5-HT synthesis, were very similar (Fickbohm et al. 2001). Therefore in Aplysia, storage of 5-HT in nonserotonergic neurons seems unlikely, indicating that studies of the Aplysia serotonergic system can be achieved using this approach. We then used this labeling technique to systematically record from serotonergic neurons in the isolated CNS after tail-nerve stimulation. Tail-nerve shock provides an in vitro analog of sensitization because it triggers 5-HT release in the CNS (Marinesco and Carew 2002) and induces SN-MN plasticity similar to that observed in sensitized animals (Cleary et al. 1998; Scholz and Byrne 1987; Zhang et al. 1994). We found a persistent generalized increase in the firing rate of recorded serotonergic neurons throughout the CNS after tail-nerve stimulation. We also determined major serotonergic projection areas and found that serotonergic neurons project to many central ganglia as well as peripheral targets such as lips, tentacles, tail, body-wall, siphon, heart, and gill or gill ganglion. It is likely that these areas receive increased serotonergic input during sensitization training.
Perhaps the most striking observation of our study is the widespread nature of tail-nerve shock-elicited activation of serotonergic neurons. Several mechanisms could contribute to the increased firing rate of serotonergic neurons after tail-nerve stimulation. First, most serotonergic neurons showed increased input resistance by the end of the recording session; this would make them more excitable and contribute to increasing their firing rate.
A change in input resistance cannot be the only contributing factor to increased firing, however, because it persisted after firing rate had already returned back to its baseline level. A likely additional mechanism is suggested by the presence of numerous excitatory connections between CC3/CB1 and other serotonergic neurons within the pedal and abdominal ganglia. At least part of the excitation evoked by CC3/CB1 stimulation probably relies on monosynaptic connections. Because postsynaptic serotonergic neurons in these ganglia are directly excited by 5-HT, and CC3/CB1 contains 5-HT, it is likely that these connections use 5-HT as their neurotransmitter. The existence of these excitatory connections enables CC3/CB1 to spread excitation to postsynaptic serotonergic neurons in abdominal and pedal ganglia and therefore might contribute to the generalized increase in firing rate throughout the serotonergic system (see Fig. 14). Excitatory connections between cerebral serotonergic neurons and other serotonergic neurons have been described in P. californica, Clione limacine, and Tritonia (Jing and Gillette 2000; Katz et al. 2001; Norekian and Satterlie 1996) and could underlie a similar network in other marine mollusks.
Interestingly, serotonergic neurons in the Aplysia CNS do not show the classical 5-HT1a-mediated autoinhibition present in mammals (Barnes and Sharp 1999; Blier and de Montigny 1987; Pineyro and Blier 1999). On the contrary, we found that serotonergic neurons in the pedal and abdominal ganglia are excited by exposure to exogenous 5-HT. Mammalian 5-HT1a receptors are Gi-coupled and are believed to have derived from an ancestral 5-HT1 superfamily after the differentiation of vertebrates from invertebrates (Tierney 2001; Vernier et al. 1995). Although the Aplysia CNS expresses at least two Gi-coupled 5-HT receptors (Angers et al. 1998; Barbas et al. 2002), it is possible that it specifically lacks 5-HT1a-like autoreceptors capable of inhibiting firing rate. Instead, 5-HT-mediated excitation of serotonergic neurons could rely on a cyclic AMP-gated sodium current similar to the one described in the marine mollusk P. californica (Sudlow and Gillette 1995).
The fact that Aplysia serotonergic neurons are excited by 5-HT could enable them to be engaged in a positive feedback loop where increased firing would lead to higher 5-HT levels, which would in turn further excite serotonergic neurons. We did not find experimental evidence for neuronal positive feedback. For example, CC3 (CB1) cells provide excitation to serotonergic neurons in the pedal and abdominal ganglia but do not seem to receive excitatory feedback from them. Moreover, we never observed excitatory connections between pairs of pedal or abdominal neurons. It is possible, however, that serotonergic neurons receive positive feedback through 5-HT released into the hemolymph (arrows in Fig. 14). Levenson et al. (1999) detected 5-HT in the hemolymph of Aplysia and showed that its concentration was increased by ∼25% during and after sensitization training using multiple tail-shocks. 5-HT in the hemolymph can be synthesized and released by hemocytes (Ottaviani et al. 1992), which are part of the Aplysia immune system, and possibly by nerve endings innervating blood vessels (Alevizos et al.1989; Skelton and Koester 1992). 5-HT is also metabolized by the kidney and hemocytes (Goldman and Schwartz 1977), which regulate its concentration in the hemolymph and could thus prevent it from reaching excessive levels. Such a regulatory mechanism could help preventing “runaway” in the system. Finally, some 5-HT receptors responsible for this effect might be located at least in part on neuronal cell bodies as revealed by ionophoretic applications of 5 HT onto POP cells in A. brasiliana (Parsons and Pinsker 1989); this is consistent with the relatively fast kinetics of 5-HT excitation in our experiments. Somatic 5-HT receptors would be readily accessible to circulating 5-HT (Marinesco and Carew 2002).
Our results suggest that a large subset of serotonergic neurons, consisting of CC3/CB1 cells and virtually all serotonergic neurons in pedal and abdominal ganglia (Fig. 14, red circles), appear interconnected through either direct excitatory connections between serotonergic neurons or by a positive feedback mechanism possibly exerted by circulating 5-HT. Thus these neurons can be considered a distributed network capable of synchronous and persistent activation in response to sensitizing stimuli. This notion is consistent with the general view that sensitization training in Aplysia is accompanied by an increase in serotonergic tone (Mackey et al. 1989; Levenson et al. 1999; Marinesco and Carew 2002).
A major function associated with 5-HT release during sensitization training is to facilitate SN-MN synapses and increase SN excitability, two cellular correlates contributing to increase reflex strength. We should note that both short-term sensitization and short-term synaptic facilitation significantly outlast the increase in firing rate of most serotonergic neurons we recorded (20–30 vs. 4–5 min, on average). Therefore whatever the role played by serotonergic neurons in sensitization, it is likely to be involved in the induction of sensitization rather than in its maintenance. Surprisingly, the time course of the increase in serotonergic neurons' firing rate after tail-nerve shock did not match the time course of measurements of actual extracellular 5-HT concentration reported here (Fig. 1C) or in previous studies (Bristol et al. 2004; Marinesco and Carew 2002). 5-HT release evoked by tail-nerve shock in the vicinity of tail SN somata or synapses typically last 30–40 s, whereas the serotonergic neurons show increased firing rates for several minutes. We do not yet understand the reasons for this discrepancy. It may result, at least in part, from the highly synchronous high-frequency burst of activity in serotonergic cells immediately after tail-nerve shock, which causes a 5-HT release profile that is detected by our measurement techniques. Also, carbon-fiber electrodes used to measure the extracellular 5-HT concentration are mostly sensitive to 5-HT spillover outside the immediate synaptic release sites (Bunin and Wightman 1998; Marinesco and Carew 2002). Because spillover is controlled by several factors like diffusion and reuptake, its relationship to the actual firing rate of serotonergic neurons might not be simply linear.
Finally, optical recordings of electrical activity within neuronal networks in mollusks have revealed that even simple behaviors can rely on the activation of hundreds of neurons. In particular, siphon-induced gill-withdrawal and respiratory pumping in Aplysia activate a distributed set of more than a hundred neurons within the abdominal ganglion (Wu et al. 1994; Zecevic et al. 1989). Although the effects of tail- or tail-nerve shock have not been analyzed by optical techniques, such inputs likely also activate a large neuronal population. Indeed, several nonserotonergic neurons increase their firing rate in response to tail-shock, such as the cholinergic interneuron L16 (Wright and Carew 1995) and LPL16, a neuron synthesizing the inhibitory peptide Phe-Met-Arg-Phe-amide (FMRFamide, Mackey et al. 1987). Therefore the widespread activation of the serotonergic system after tail-nerve shock is probably part of an even broader response involving hundreds of neurons throughout the CNS. To understand the function of this increased serotonergic activity, it is necessary to isolate its effects from those of the other nonserotonergic mechanisms triggered by tail-nerve shock. This general question is addressed in the companion paper that follows (Marinesco et al. 2004).
This work was supported by National Institute on Mental Health Grant RO1 MH-14-1083 to T. J. Carew.
We are grateful to A. Bristol, C. Sherff, P. Katz, and J. Koester for helpful comments on an earlier version of this manuscript. J. Koester also provided valuable technical information on identification of the serotonergic neuron RBhe. We thank N. Wickremasinghe for helpful technical assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by the American Physiological Society