Pattern-Specific Synaptic Mechanisms in a Multifunctional Network. I. Effects of Alterations in Synapse Strength

Steven P. Lieske, Jan-Marino Ramirez


Many neuronal networks are multifunctional, producing different patterns of activity in different circumstances, but the mechanisms responsible for this reconfiguration are in many cases unresolved. The mammalian respiratory network is an example of such a system. Normal respiratory activity (eupnea) is periodically interrupted by distinct large-amplitude inspirations known as sighs. Both rhythms originate from a single multifunctional neural network, and both are preserved in the in vitro transverse medullary slice of mice. Here we show that the generation of fictive sighs were more sensitive than eupnea to reductions of excitatory synapse strength caused by either the P/Q-type (α1A-containing) calcium channel antagonist ω-agatoxin TK or the non-N-methyl-d-aspartate (NMDA) glutamate receptor antagonist 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX). In contrast, the NMDA receptor antagonist MK-801, while also inhibiting eupnea, increased the occurrence of sighs. This suggests that among the glutamatergic synapses subserving eupneic rhythmogenesis, there is a specific subset—highly sensitive to agatoxin and insensitive to NMDA receptor blockade—that is essential for sighs. Blockade of N-type calcium channels with ω-conotoxin GVIA also had pattern-specific effects: eupneic activity was not affected, but sigh frequency was increased and postsigh apnea decreased. We hypothesize that N-type (α1B) calcium channels selectively coupled to calcium-activated potassium channels contribute to the generation of the postsigh apnea.


Rhythmic activity is a fundamental property of a large number of neural circuits, including not only the central pattern generators (CPGs) responsible for producing rhythmic behaviors but also the cortical and subcortical regions implicated in many phenomena that are not fundamentally periodic. For example, the control of sleep and wake states (e.g., Sanchez-Vives and McCormick 2000; Steriade et al. 1993), olfaction (e.g., Kay 2003), and some forms of learning (e.g., Seidenbecher et al. 2003) have all been shown to involve oscillations in neuronal activity.

In each of these examples, multiple rhythms are generated within the same circuit (Csicsvari et al. 2003; Kay 2003; Steriade et al. 1993), but distinct strategies have evolved for handling the interaction between them. In the mammalian thalamocortical system, switching between various sleep and wake states occurs on a relatively slow time scale as an all-or-none state change in the network as a whole. In the hippocampus, gamma oscillations occur both in the presence and in the absence of theta oscillations, with the variation in the power of the gamma rhythm between theta and nontheta epochs itself variable across different anatomical areas (Csicsvari et al. 2003). In the olfactory system, two different forms of gamma oscillation appear to arise from distinct circuitry, but switching between the two is again all-or-none (Kay 2003).

The most completely understood examples of this kind of functional reconfiguration, however, come from the study of invertebrate CPGs. The crustacean stomatogastric ganglion, for example, produces distinct gastric and pyloric rhythms simultaneously. In this system, certain network elements have been shown to switch their “allegiance” from one rhythm to the other, but the foundation of both rhythms is provided by separate populations of intrinsically rhythmogenic cells (Bartos et al. 1999; Dickinson 1995).

In the mammalian respiratory system, two distinct but interacting patterns originate from a single center in the medulla with no evidence for distinct populations of rhythmogenic cells (Lieske et al. 2000). Sighs, or “augmented breaths,” are large-amplitude inspirations that periodically interrupt the normal respiratory pattern. They are essential for maintaining lung function (Reynolds and Flom 1968; Szereda-Przestaszewska et al. 1976) and are associated with changes in arousal state (Lijowska et al. 1997; Orem and Trotter 1993; Poe et al. 1996); a deficit in sighing may play a role in the etiology of Sudden Infant Death Syndrome (Kahn et al. 1988). Both patterns are preserved in medullary slices in vitro, and both originate from a subregion of the ventral respiratory group (VRG) known as the pre-Bötzinger Complex (PBC) (Lieske et al. 2000). Recordings from individual neurons have shown that nearly every cell in this region has a discharge pattern that is in phase with both motor patterns (Lieske et al. 2000).

The genesis of both eupnea and sighs from a single population of cells poses an intriguing problem, however: it is not clear how a separate “sigh clock” could be maintained by cells that produce a burst of action potentials with every intervening eupneic inspiration. Results presented herein, and in the companion article (Lieske and Ramirez 2006), suggest that the production of these two patterns may involve glutamatergic synapses with qualitatively distinct properties.

One such difference is suggested by the observation that sighs, but not eupnea, were abolished after bath application of 4–8 μM cadmium (Lieske et al. 2000). In other pattern-generating networks, different subtypes of voltage-gated calcium channels have distinct functional roles (Büschges et al. 2000); could the selective blockade of sighs by low concentrations of cadmium be attributable to an effect at a specific calcium channel subtype? Calcium channels are differentiated on the basis of electrophysiological and pharmacological properties with greater cadmium sensitivity being observed in thosechannels activated at more depolarized potentials (L, N, P, Q, and R type) (see Dunlap et al. 1995; Fox et al. 1987; Randall and Benham 1999). The contributions of these different calcium channel subtypes to the response properties of isolated VRG neurons have been characterized in both current clamp (Onimaru et al. 1996) and voltage clamp (Elsen and Ramirez 1998), but their functional roles in the genesis of either eupnea or sighs have not been characterized nor have their respective contributions to synaptic transmission in this network. We screened selective antagonists for L-, N-, P-, and Q-type calcium channels for effects on fictive eupnea and sighs; our results show that both patterns are abolished after blockade of α1A (CaV2.1)-containing (P/Q-type) voltage-activated calcium channels by ω-agatoxin TK but that sighs are abolished at much lower toxin concentrations than is required to eliminate eupnea.

Using evoked excitatory postsynaptic potentials (EPSPs), we also demonstrate that α1A calcium channels are essential for synaptic transmission at only a subset of excitatory synapses within the respiratory network. In the companion article (Lieske and Ramirez 2006), we describe similar results after activation of the metabotropic glutamate receptor mGluR8, well known as a presynaptic modulator of synapse strength. These results suggest that the synaptic interactions between neurons can vary with the output pattern produced. The present study is a first step in identifying the cellular mechanisms that could play an important role in the production of multiple patterns by a single network in the absence of specialization at the cellular level.


Preparation and population recordings

Rhythmic 600- to 700-μm-thick slices containing the PBC were obtained from 6- to 12-day-old CD-1 outbred mice (Charles River Laboratories, Wilmington, MA) as described in detail elsewhere (Ramirez et al. 1997; Telgkamp and Ramirez 1999). These slices were transferred to a 5-ml recording chamber, and superfused (flow rate: 10–15 ml/min) with an artificial cerebrospinal fluid (ACSF) containing (in mM) 118 NaCl, 3 KCl, 1.5 CaCl2, 1 MgCl2, 25 NaHCO3, 1 NaH2PO4, and 30 d-glucose, and equilibrated with carbogen (95% O2- 5% CO2, pH 7.4). Temperature was maintained at 29–31°C, and KCl was elevated to 8 mM over a span of 30 min before commencing recordings. Except where otherwise noted, the superfused ACSF was continuously recycled (total bath volume: 200 ml). For experiments with Ca2+ channel toxins, the total volume of the recycling bath was reduced to 20–50 ml, ≥10 min before bath-applying toxin. Some experiments were performed with two slices (from different animals) in a single perfusion chamber. All salts were obtained from Sigma (St. Louis, MO).

Population recordings were obtained using borosilicate glass electrodes with a tip diameter of ∼20–30 μM. Neural activity was recorded extracellularly, in most cases from the caudal surface of the slice in a region presumably corresponding to the rostral VRG (in which inspiratory cells predominate) (Merrill 1970; Sun et al. 1998; von Euler 1986). A few population recordings were obtained from the rostral surface of the slice in a region apparently corresponding to the PBC. Recordings were band-pass filtered between 300 Hz and 3 KHz, rectified, and integrated (time constant ∼150 ms). The resulting data were digitized at 500 Hz with a Digidata board (Axon Instruments, Foster City, CA), stored on an IBM-compatible PC using Axotape software (Axon), and analyzed off-line using Igor Pro (WaveMetrics, Lake Oswego, OR), Clampfit (Axon), and Prism (GraphPad, San Diego, CA). Only recordings with qualitatively good signal-to-noise ratios (SNR) were used (we estimate <2% of preparations were discarded). SNR was computed for 20 such preparations, estimated as the ratio between the amplitude of the burst and the SD of the recording during the interburst interval—values ranged from 15.4 to 67.4, with a mean SNR of 35.4. In one experiment using ω-conotoxin GVIA the electrode was repositioned between the control and GVIA recordings; this prevents the comparison of amplitude for that preparation but should have no effect on the time-based measures.

Stimulators of the cAMP pathway such as forskolin (Mironov et al. 1999; Shao et al. 2003), and CPT-cAMP (Lieske and Ramirez 2006) have been shown to increase fictive respiratory frequency in vitro. In some experiments, eupneic frequency was inhibited coincident with the blockade of sighs in the presence of agatoxin; in such cases, 10–20 μM forskolin was applied to increase respiratory frequency so as to confirm the selective blockade of sighs (Fig. 1, B and C).

FIG. 1.

Blockade of α1A (P/Q-type) Ca2+ channels by ω-agatoxin TK abolishes sighs at low concentrations and eupnea at high concentrations. A: integrated ventral respiratory group (VRG) population activity shows bursts of activity (upward deflections) corresponding to eupneic inspirations (smaller amplitude) and sighs (larger amplitude) in control conditions, but only eupnea in the presence of 15 nM Aga-TK. B: in another preparation, 30 nM Aga-TK markedly inhibits eupnea as well as abolishing sighs; stimulation of the network with forskolin restores the fictive eupneic rhythm while sighs remain blocked. C: higher concentrations of Aga-TK abolish both rhythms, and neither is recovered in the presence of forskolin. Inset: expanded traces corresponding to specified regions in control and in the presence of agatoxin. Note that all rhythmic activity is abolished, leaving only baseline noise indistinguishable from that present in control. Compare with Fig. 3B.

ω-Agatoxin TK, ω-agatoxin IVA, ω-conotoxin GVIA, and ω-conotoxin MVIIC were obtained from Sigma or from Bachem (King of Prussia, PA). Nifedipine and MK-801 were obtained from Sigma. 6-Cyano-7-nitroquinoxalene-2,3-dione (CNQX), 6,7-dinitroquinoxaline-2,3-dione (DNQX), (RS)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP), and forskolin were obtained from Tocris (Ellisville, MO).


Current-clamp recordings were obtained in the whole cell patch configuration from inspiratory neurons located within 300 μM of the rostral surface of the slice in a region presumably corresponding to the PBC. Recordings included cells with discharge patterns spanning the various classification schemes that have been proposed for inspiratory cells (Rekling et al. 1996; Schwarzacher et al. 1995; Thoby-Brisson and Ramirez 2001; Zheng et al. 1991). Population recordings were obtained from the contralateral PBC for reference, as described in the preceding text, except always from the rostral surface of the slice. Patch electrodes were manufactured from filamented borosilicate glass (Clark GC150F-10, Warner Instruments, Hamden, CT) and filled with a solution containing (in mM) 140 K-gluconic acid, 1 CaCl2, 10 EGTA, 2 MgCl2, 4 Na2ATP, and 10 HEPES. A liquid junction potential of 12 mV was measured for this solution in ACSF, and taken into account. Intracellular recordings were low-pass filtered at 1.8 kHz and sampled at 4.2 kHz.

Once the whole cell patch configuration was obtained, The ACSF was replaced with a fresh solution containing only 3 mM K+, so that the low-amplitude evoked EPSPs could be recorded without a confounding population rhythm. In a few cases, the population rhythm persisted in 3 mM K+; in these cases, blockade of NMDA receptors with CPP sufficed to eliminate population-level activity. When necessary, constant current was applied to hyperpolarize the recorded cell to a level such that tonic firing activity also ceased (typically <1 nA for an Em of −70 to −80 mV). Stimuli were delivered to the contralateral VRG via the same borosilicate glass electrode from which the population rhythm had been recorded, using an Iso-flex stimulus isolator (AMPI) in constant-voltage mode. Individual stimulation pulses were 0.2–0.3 ms in duration at the lowest amplitude that consistently evoked an EPSP in the recorded cell (typically 200 mV). Stimuli were delivered at 0.5 Hz, and 20 stimulations were averaged for each data point. Stimulation experiments were conducted in a total of 68 inspiratory neurons of which 56 showed a clear EPSP (5 had an inhibitory postsynaptic potential, IPSP, 5 had both an EPSP and an IPSP, and in 2 no PSP was observed; details in results). Of the 56, 21 were used in experiments described herein, and 7 in the companion article (Lieske and Ramirez 2006); 28 could not be used beyond the initial characterization for one or more of the following reasons: EPSP amplitude deteriorated in control conditions or the cell was lost altogether (n = 14), preliminary experiments were performed that could not be included statistically due to differences in drug concentration or experimental protocol (n = 12) or the EPSP peak appeared polysynaptic or could not be adequately distinguished from the stimulus artifact or a retrograde action potential (AP, n = 5). It was nonetheless possible to characterize the EPSP in control conditions for 14 of these of 28, as well as all 28 experiments incorporated either here or in the companion article (total n = 42 for the EPSP description). Evoked EPSPs were recorded for 10 min prior to initiating any experiment, and only those preparations in which EPSP amplitude appeared stable were used. Four preparations were used as a time control, to verify subsequent stability when this initial criterion was met. EPSP amplitude averaged over the period corresponding to 15 to 20 min after the initiation of an experimental manipulation was 100.7 ± 15.5 (SD) % of baseline.

Data analysis

Statistical values are given as means ± SE unless otherwise specified. Significance was assessed with a paired t-test (burst width, sigh interval, postsigh apnea), the Wilcoxon matched pairs test for paired data which could not be assumed to be normally distributed (irregularity), a one-sample t-test for ratio data (burst amplitude), or the Wilcoxon signed-rank test for ratio data that failed normality tests or could not be assumed to be distributed normally (frequency, sigh frequency). In all cases, significance was defined as P < 0.05. “Burst duration” refers to the full width at half-maximal amplitude. Cycle periods were defined from the beginning of one fictive inspiration to the beginning of the next. The irregularity score was computed for each cycle as Sk = 100% × Abs(Pk –Pk-1)/Pk-1, where Sk is the score of the kth cycle, Pk is its period, and Pk-1 is the period of the preceding cycle (Barthe and Clarac 1997)—this represents the breath-by-breath percent change in cycle length. Median irregularity scores were used for comparisons because the distributions were often highly skewed, and a few outliers contributed greatly to the means. Postsigh apneas (a brief pause in the eupneic rhythm following each sigh) were excluded from calculations of eupneic frequency and irregularity. Mean burst frequencies were computed as a time-weighted average, mathematically equivalent to the reciprocal of the mean cycle period.

The sequential dose response curve for sigh frequency in CNQX (Fig. 5C) involved experiments that took place over several hours of total duration. In control experiments, sigh frequency increased over this duration even in the absence of any pharmacological manipulation (Fig. 5C, inset). This increase was well fit by a linear function with slope = 0.1233 min-1/h (slope significantly nonzero: P < 0.0001, n = 6). This effect was controlled for in the CNQX group by computing sigh frequency as a percent of the predicted frequency given the passage of time, according to the formula %predicted = fEXP/(fCTRL + t × k) where fEXP is the observed sigh frequency for a given dose, fCTRL is the sigh frequency for the same preparation in control conditions (at the start of the experiment), t is the length of time passed since the control sigh frequency was measured, and the constant k is the slope described above. The corrected data were fit with the variable-slope sigmoid equation: %predicted = 100/(1 + ([CNQX]/EC50)σ) where σ is the Hill slope factor (fit from the data along with the EC50) and [CNQX] is the concentration of CNQX. For both eupnea and sighs, the quality of the fit was significantly better than with the standard (slope factor held constant, =1) sigmoid (P < 0.05, by F test).


Effects of Ca2+ channel blockers on eupnea and sighs

As described previously, fictive sighs were biphasic in shape, larger in amplitude than eupneic bursts, and followed by a brief pause in the eupneic rhythm, consistent with the definition of sighs in vivo (Cherniack et al. 1981; Glogowska et al. 1972; Orem and Trotter 1993; Takeda and Matsumoto 1998). The P/Q-type calcium channel toxin ω-Agatoxin TK (Aga-TK) selectively abolished sighs without blocking eupnea (n = 15). In our initial experiments, very low concentrations of Aga-TK sufficed, thus in nine preparations in which the concentration of Aga-TK was increased stepwise, the median concentration for the blockade of sighs was 33 nM (Fig. 1A). In later experiments (using a different batch of agatoxin), larger concentrations (60–120 nM) were required, but despite this variability, the toxin was consistently selective for sighs over eupnea as the fictive eupneic rhythm was preserved in every case after the blockade of sighs.

The eupneic rhythm was not unaffected: both frequency and burst duration were significantly decreased, whereas the irregularity increased (Table 1). In some cases where the eupneic frequency was markedly depressed, stimulation of the network with forskolin restored the eupneic rhythm without restoring sighs (Fig. 1B). Stimulation of the cAMP pathway has been shown elsewhere to increase eupneic frequency (Mironov et al. 1999; Shao et al. 2003); in the companion article, we show that sigh frequency is also markedly increased (Lieske and Ramirez 2006). At higher concentrations of Aga-TK (500 nM), both eupnea and sighs were permanently blocked (Fig. 1C, n = 4).

View this table:

Effects of selective calcium channel antagonists on fictive eupnea

This effect was specific for α1A-containing channels. Bath application of the N-type calcium channel toxin ω-conotoxin GVIA (0.5–1 μM), rather than blocking sighs, led to a significant and dramatic increase in sigh frequency, and a concomitant decrease in the duration of the postsigh apnea (Fig. 2, n = 4). There were no significant effects on the fictive eupneic rhythm (Table 1). The L-type calcium channel blocker nifedipine (4–8 μM) resulted in a significant decrease in the duration of the eupneic burst (Table 1) but had no other significant effect on any examined parameter of either eupnea or sighs (Table 1, Fig. 2).

FIG. 2.

Effects of N-type (α1B) and L-type (α1C/D) specific antagonists. A: integrated VRG population activity shows a dramatic increase in sigh frequency and a decrease in the duration of the postsigh apnea in the presence of 1 μM GVIA. These effects are quantified in C and D, effects on eupnea in Table 1. B: both eupnea and sighs persist in the presence of 4 μM nifedipine. The increase in sigh frequency and decrease in eupneic frequency suggested in this example were not apparent in the averaged data (C; Table 1). C: sigh frequency in nifedipine (n = 6) and in GVIA (n = 4), computed as a percentage of that observed in control. D: postsigh apnea in control conditions and in nifedipine or GVIA, normalized to eupneic cycle length. Means ± SE; Values different from control: *P < 0.05, **P < 0.01.

To test whether the channel blocked by Aga-TK had a P- or Q-type pharmacology, we applied the Q-type preferring ω-conotoxin MVIIC (McDonough et al. 1996; Randall and Tsien 1995; Sather et al. 1993) in the bath at concentrations of 0.1, 0.3, and 1 μM. At 0.1 μM, MVIIC led to a specific blockade of sighs, which developed over an interval of 20–25 min (Fig. 3A). There was also a significant reduction in eupneic frequency (Table 1), but no change in the shape of the eupneic burst. At 0.3 μM, the fictive eupneic rhythm was also abolished, but disorganized population-level activity persisted (Fig. 3A). A complete elimination of all population-level activity was observed with 1 μM MVIIC (Fig. 3B), similar to the effects of high concentrations of Aga-TK (Fig. 1C), and consistent with the blockade of both P and Q-type channels at this concentration.

FIG. 3.

The Q-type preferring ω-conotoxin MVIIC abolishes sighs at low concentrations and eupnea at higher concentrations. A: integrated VRG population activity shows a complete blockade of both eupnea and sighs in the presence of 1 μM MVIIC. B: in another preparation, 100 nM MVIIC leads to an inhibition of eupnea and a blockade of sighs, developing over ∼25 min. In 300 nM MVIIC, eupnea is also abolished. Inset: expanded traces corresponding to specified regions in control and in the presence of MVIIC. Note that disorganized population-level activity persists in MVIIC (2), which is distinct from the baseline noise present in control (1). Compare with Fig. 1C.

Cellular level effects of Ca2+ channel blockade

When inspiratory cells are functionally decoupled from the network by glutamate receptor antagonists, sigh bursts are no longer observed within the individual recorded cells (Lieske et al. 2000). Indeed, sighs are abolished by concentrations of glutamate receptor antagonists much lower than those necessary to block population-level fictive eupnea (see following text). This precludes the possibility of directly studying the blockade of sighs in isolated cells. Similarly, at agatoxin concentrations sufficient to abolish sighs at the population level, no endogenous synaptic activity that appeared to correspond to sighing remained. At concentrations of agatoxin insufficient to abolish sighs, there was no apparent effect on the shape or amplitude of the intracellular burst potential during sighs (Fig. 4, A–C) and at most a modest effect on inter-sigh interval (Fig. 4D). Effects of agatoxin on the membrane potential are described in detail in the following text.

FIG. 4.

Cellular level effects of gradated concentrations of agatoxin-TK. A: integrated VRG population activity (bottom), intracellular potential in current clamp (top) during eupneic and sigh bursts in control conditions. Right: expanded intracellular trace. B: in 30 nM Agatoxin-TK, there is little effect on the intracellular or population-level sigh burst. C: in 50 nM Agatoxin-TK, sighs are abolished. D: effects of gradated agatoxin concentrations on sigh interval in a representative preparation. At 30 nM, there was a single sigh after a long pause, after which sighs ceased altogether.

Effects of glutamate receptor antagonists on fictive eupnea and sighs

Because P/Q-type Ca2+ channels play an essential role in synaptic transmission at a variety of synapses (Regehr and Mintz 1994; Takahashi and Momiyama 1993; Turner et al. 1992), we hypothesized that a reduction in synapse strength induced postsynaptically, by glutamate receptor antagonists, would lead to a selective inhibition of sighs similar to that observed with the calcium channel toxins. The non-NMDA receptor antagonists CNQX and DNQX are used routinely in the in vitro respiratory network to abolish population-level rhythmic activity (20–40 μM; Fig. 5A) (see also Thoby-Brisson and Ramirez 2001). At lower concentrations, however, where blockade of AMPA receptors should be incomplete (1–2 μM), sighs were abolished selectively (Fig. 5B, n = 7). In addition to abolishing sighs, 1 μM CNQX/DNQX also led to significant decreases in the frequency and amplitude of the eupneic rhythm, and a significant increase in eupneic irregularity (Table 2). Similar results were also obtained with the AMPA-selective antagonist GYKI 52466 (20–40 μM abolished sighs selectively, n = 3; 100 μM abolished both eupnea and sighs, n = 5; data not shown), whereas the kainate-selective antagonist NS-102 was without significant effect (12.5–40 μM, n = 5; data not shown).

FIG. 5.

Low and high concentrations of AMPA receptor antagonists block sighs and eupnea, respectively. A: integrated VRG population activity is abolished in 20 μM 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX). B: 1 μM CNQX abolishes sighs but not eupnea. C: sequential dose-response curve for sigh frequency (•) and eupneic frequency (▴) in CNQX. Because sigh frequency increases at a constant rate in the absence of any pharmacological manipulation (inset), sigh frequencies are presented as a percentage of the frequency predicted from the length of time passed (see methods). ○, an estimate of the maximum frequency possible for an interval in which no sighs were observed.

View this table:

Effects of glutamate receptor antagonists on fictive eupnea

The observation that sigh frequency was reduced when not blocked outright suggested that the mechanism of sigh blockade was not all-or-none but that even lower concentrations might reduce the occurrence of sighs without abolishing them altogether. To test this hypothesis, a sequential dose response curve was computed for sigh frequency over concentrations of CNQX from 10 nM to 1 μM (Fig. 5C). This curve was best fit by a variable-slope sigmoid with IC50 = 0.423 ± 0.056 μM and Hill slope = −2.213 ± 0.506 (P = 0.0009 vs. standard sigmoid by F test). For comparison, the curve for eupnea frequency had IC50 = 1.11 ± 0.08 and Hill slope = −2.24 ± 1.40, thus the selectivity for the effects of CNQX on sighs over eupnea was ∼2.6-fold.

The effects of NMDA receptor blockade on the respiratory rhythm were also examined, but contrary to what was observed with non-NMDA receptor antagonists, sighs were not abolished in the presence of the NMDA receptor open-channel blocker MK-801 (20 μM). Indeed, the opposite effect was observed, as sigh frequency was significantly increased (P = 0.0178, n = 7, Fig. 6, A and B). Effects on eupnea, however, were similar to the effects of low concentrations of non-NMDA receptor antagonists: the frequency and duration of the eupneic rhythm were reduced, and the irregularity increased (Fig. 6A, Table 2).

FIG. 6.

N-methyl-d-aspartate (NMDA) receptor blockade inhibits eupnea but does not abolish sighs. A: integrated VRG population activity shows a clear inhibition of fictive eupnea in the presence of the NMDA receptor open-channel blocker MK-801 but does not inhibit sighs. B: sigh frequency in MK-801 (n = 7), computed as a percentage of that observed in control. C: postsigh apnea in control conditions and in MK-801, normalized to eupneic cycle length. Means ± SE. Values significantly different from control (**P < 0.01).

Effects of GVIA and Aga-TK on excitatory synaptic transmission

Evoked EPSPs were recorded after stimulation in the contralateral VRG as a means to directly examine the effects of Aga-TK and GVIA on synaptic transmission. EPSPs were recorded in a total of 56 inspiratory neurons, while hyperpolarizing responses (inhibitory postsynaptic potentials, IPSPs) were observed in 5, and both an EPSP and an IPSP in 5. In only one cell could no synaptic response be evoked (in 1 other cell an antidromic AP was evoked at relatively low stimulus intensities, and larger stimuli were not tested). Of these 68 experiments, midline stimulation was performed in 17, and stimulation of the contralateral VRG in 51. EPSP amplitudes did not differ between the two stimulation paradigms, so the data were pooled. EPSPs evoked at the midline averaged 2.35 ± 0.31 mV in amplitude and followed the stimulus by 7.4 ± 0.3 ms (n = 10), whereas those evoked in the contralateral VRG averaged 2.35 ± 0.17 mV and followed the stimulus by 8.9 ± 0.4 ms (Fig. 7; n = 32); these amplitudes are similar to those reported by other authors with dual intracellular recordings in a similar preparation (Rekling et al. 2000). EPSP decay time constants were 33.5 ± 2.6 and 37.9 ± 3.8 ms for contralateral and midline stimuli, respectively. CNQX (1 μM) reduced EPSP amplitude by ∼75% (Fig. 8; n = 5), and 20 μM CNQX abolished the EPSP altogether (not shown, n = 2).

FIG. 7.

Excitatory postsynaptic potentials (EPSPs) and antidromic action potentials evoked by stimulation of the contralateral VRG. A, top: integrated VRG population activity; bottom: simultaneous intracellular recording from a typical inspiratory cell located within the contralateral VRG. B: brief (0.2 ms) stimulation via the (contralateral) population electrode, in this case a negative voltage pulse, elicits an EPSP in the recorded neuron. Same cell as in A and C. C: larger stimulus amplitude, in this case with polarity reversed as well, elicits an antidromic action potential in the recorded neuron (↓). This response was all-or-none, and at intermediate stimulus amplitudes, not every stimulation yielded an AP (▴). Inset: expanded time scale allows clear differentiation of stimulus artifact and evoked AP. Note also that evoked action potentials (Aps) peaked significantly earlier than evoked EPSPs and are thus assumed to represent the antidromic conduction of an action potential directly evoked by the stimulus in the axon of the recorded cell.

FIG. 8.

CNQX sensitivity of evoked EPSPs. A: 1 μM CNQX reduced EPSP amplitude by ∼75% on average (n = 5). Stimulation to contralateral VRG as described in text. B: representative EPSP traces in control conditions (top 3 traces) and 1 μM CNQX (bottom 3 traces).

Twelve neurons were also tested for the presence of APs induced directly by stimulation (at higher stimulus amplitudes than were used to evoke EPSPs) and conducted back to the cell body antidromically; this was observed in 6 of the 12 cells tested (Fig. 7C).

We opted to study crossing connections for reasons of experimental accessibility; these synapses are not expected to play a specialized role in sigh production nor are they known to differ in any systematic way from synapses at ipsilateral connections. Indeed, both EPSPs and antidromic APs were recorded from inspiratory cells with a variety of discharge patterns, spanning the various classification schemes of inspiratory cell “subtypes” that have been proposed (Rekling et al. 1996; Schwarzacher et al. 1995; Thoby-Brisson and Ramirez 2001; Zheng et al. 1991). EPSPs were also observed in two of two expiratory neurons (not shown); these were not further characterized.

Bath application of GVIA (0.5 μM) reduced EPSP amplitude in each of six preparations, by ∼40% on average (Fig. 9, A and B; n = 6). In contrast, Aga-TK (120 nM) had variable effects, ranging from a nearly complete elimination of the EPSP to essentially no effect at all. On average, Aga-TK reduced the magnitude of the EPSP by ∼50% (Fig. 9, C and D; n = 5). These results suggest that the calcium influx driving synaptic transmission at these synapses includes a constant but relatively minor contribution from α1B (N-type) channels with the remainder deriving from varying contributions of α1A (P/Q-type) and α1E (R-type). There was no change in membrane potential following GVIA application (−70.3 ± 3.1 vs. −70.3 ± 3.1 mV control, P = 0.9895, n = 6), while a modest depolarization was observed after application of Aga-TK (−68.6 ± 1.2 vs. −70.3 ± 1.8 mV control, P = 0.0472, n = 5).

FIG. 9.

Amplitude of intracellularly recorded evoked EPSPs reduced by calcium channel toxins. A: blockade of α1B (N-type) calcium channels reduced EPSP amplitude by ∼40% on average. Grouped data from contralateral (n = 3) and midline (n = 3) stimulations. Values for individual preparations represent the amplitude of a mean EPSP obtained by averaging the response to 20 stimulations at each time point. Stimulations were 0.2 ms in duration and delivered at a rate of 0.5 Hz. B: individual experiments contributing to average shown in A. C: blockade of P-type α1A calcium channels with 120 nM Aga-TK reduced EPSP amplitude by ∼50% on average. Stimulation to contralateral VRG, n = 5. D: individual experiments contributing to average shown in C. Note the much greater preparation-to-preparation variability than observed with GVIA (B).


Presynaptic Ca2+ channels in the respiratory network

Here we demonstrate that both eupnea and sighs persisted even at very high concentrations of ω-conotoxin GVIA, suggesting that α1B (N-type) calcium channels are not essential for synaptic transmission in this network. At the synaptic level, GVIA reduced the amplitude of evoked EPSPs consistently but only by ∼40%. EPSP amplitude has been shown to be proportional to ([Ca2+]int)k, with the value of the exponent k generally between 3 and 4 (Borst and Sakmann 1996; Mintz et al. 1995), thus a reduction in EPSP amplitude of 40% corresponds to a reduction in calcium influx of 12–16%.

The reduction in EPSP amplitude obtained with 120 nM Aga-TK, by contrast, varied considerably between cells, ranging from 8 to ≥92%. The remaining calcium influx is presumably carried by a combination of R-type (α1E) channels, and α1A channels with a lower sensitivity to agatoxin, either Q-type channels (McDonough et al. 1996; Randall and Tsien 1995; Teramoto et al. 1995) or α1A channels with properties intermediate between the classical P- and Q- phenotypes (Forsythe et al. 1998; Stea et al. 1994; Tottene et al. 1996).

Population level effects: P or Q type?

In parallel to these cellular studies, we also investigated the effects of calcium channel blockers at the population level. Here we have demonstrated that sighs were abolished selectively at an MVIIC concentration of 100 nM, suggesting a Q-type pharmacology (McDonough et al. 1996; Randall and Tsien 1995; Sather et al. 1993). On the other hand, the blockade of sighs in some preparations at Aga-TK concentrations as low as 15 nM is more consistent with a P-type pharmacology (Mintz et al. 1992; Randall and Tsien 1995; Teramoto et al. 1995). It therefore seems most appropriate to consider the channels responsible as “P/Q type,” at the population level as well with an indeterminate—and possibly variable—pharmacological phenotype.

Ca2+ channels and sigh rhythmogenesis: pre- or postsynaptic?

It is clear from the population-level results that the calcium channel essential for sighs is of the α1A variety, and equally clear from the evoked EPSP experiments that blockade of this channel reduces synaptic transmission in at least some glutamatergic synapses. To begin to assess whether these two effects might be causally related, we investigated whether other means of reducing synaptic transmission would also suppress sighs. If the reduction in EPSP amplitude as observed at the cellular level is responsible for the abolition of sighs as observed at the population level, postsynaptic inhibition of transmission should have the same effect, and indeed, bath application of CNQX consistently abolished sighs at lower concentrations than those at which eupnea was abolished.

This hypothesis is also consistent with the observation that sighs were not inhibited by NMDA receptor blockade. The observed effects on eupnea indicate that NMDA receptors do play a role in the coupling of inspiratory cells, thus the lack of effect on sighs is suggestive of the presence of distinct subpopulations of synapses. EPSPs evoked by stimulation of the contralateral VRG were not as sensitive to agatoxin as the population results would seem to predict, suggesting that the synapse essential for sighs is not made by crossing axons, not made by the neurons activated at the lowest stimulus amplitudes, not recordable somatically, or otherwise not accessible by the techniques used herein.

The cellular-level (postsynaptic) effects of P/Q-type Ca2+ channel blockade appear to be minimal. First, it was not possible to examine cellular-level effects specifically during sighs as sighs were abolished (Fig. 4). Second, regarding cellular effects not specific to sighs, voltage-clamp experiments have previously been performed in the mouse slice preparation and found that agatoxin-sensitive currents comprised a variable but small subset of the HVA calcium current somatically recordable (Elsen and Ramirez 1998). The cells that we recorded did depolarize slightly after application of toxin (1.7 ± 0.6 mV), but this seems unlikely to be meaningful as the expected result of blocking any P/Q-type channels that are active at baseline would be a hyperpolarization. In addition, metabotropic glutamate receptor activation, which is known to inhibit P/Q-type calcium channels in other systems, also abolishes sighs but results in no significant change in membrane potential (Lieske and Ramirez 2006).

Some changes in cellular-level properties have been described in a subset of inspiratory and preinspiratory neurons in the en bloc neonatal rat brain stem-spinal cord preparation (Onimaru et al. 1996), but the particular neuronal types in which these effects were observed have not been described in the preparation used here. We therefore conclude that these neurons seem unlikely to be responsible for the genesis of sighs.

Role of AMPA and NMDA receptors

The dose-dependent, graded inhibition of sigh frequency by glutamate antagonists suggests a mechanism for sigh rhythmogenesis in which sigh frequency is determined at least in part by synaptic coupling as has been suggested for bursts in the CA3 region of the hippocampus (Staley et al. 2001). This is in contrast to pacemaker-based models of rhythmogenesis in which decreasing the synaptic coupling between pacemakers is predicted to increase bursting frequency, until ultimately synchronization is lost and population bursting ceases altogether (e.g., Butera et al. 1999). An alternative possibility is that the sigh-specific synapse is involved in the contribution of tonic drive rather than phasic synaptic coupling—this is discussed in the companion article (Lieske and Ramirez 2006)

We did not determine a dose-response relationship for the reduction of EPSP amplitude by CNQX, but the observed reduction of 75% at a concentration of 1 μM suggests an IC50 in the range of 0.3–0.4 μM, assuming a Hill slope of 1–1.3 as reported elsewhere (Stein et al. 1992). These EPSPs were thus relatively sensitive to CNQX, as IC50 values computed for EPSP inhibition in other brain areas are typically 2–4 μM (Andreasen et al. 1989; Gean and Chang 1992; Hwa and Avoli 1992; Randle et al. 1992).

The IC50 value obtained for eupneic frequency in CNQX was 1.11 μM, whereas that for sigh frequency was 0.423 μM. It must be emphasized that this does not imply the presence of glutamate receptors with different IC50 values but only that sigh rhythmogenesis is much more sensitive to a reduction in EPSP amplitude. The differential network-level response could also derive from synapses with different numbers of receptors, postsynaptic differences in electrotonic properties, or differences in circuitry (e.g., presence or absence of positive feedback) to name just a few possibilities.

Existing data addressing the role of NMDA receptors in respiratory rhythm generation are somewhat contradictory: bath application of NMDAR antagonists to in vitro slice or en bloc preparations from rats did not alter eupneic frequency (Funk et al. 1993; Greer et al. 1991), but application of NMDA itself led to a dose-dependent frequency increase (Greer et al. 1991). We observed significant reductions in the frequency and regularity of the eupneic rhythm after the blockade of NMDA receptors. Effects of NMDAR blockade on burst duration and amplitude have not previously been addressed in vitro; we found that NMDA receptor blockade led to a significant decrease in burst duration but no significant change in amplitude, whereas AMPA receptor blockade led to a significant reduction in amplitude but not in burst duration, consistent with the idea that these receptors might play different roles in rhythm generation (Krolo et al. 1999).

Role of N- and L-type Ca2+ channels

GVIA led to a dramatic increase in sigh frequency accompanied by a similarly dramatic decrease in postsigh apnea. In other systems, calcium channels have been found to associate directly with calcium-activated potassium currents in a subtype-specific manner (Marrion and Tavalin 1998). We propose that a similar coupling of N-type calcium channels to IK(Ca) is at work here, such that a decrease in calcium influx via N-type channels during the sigh would result in a less pronounced hyperpolarization after the sigh burst, thereby decreasing both the postsigh apnea and the inter-sigh interval. Preliminary observations using Charybdotoxin to block intermediate and/or high-conductance K(Ca) channels is consistent with this hypothesis: eupneic frequency was increased by a factor of 1.84 on average, whereas sigh frequency was increased by a factor of 4.90 on average, and the postsigh apnea decreased from 2.49 eupneic cycles to 1.72 eupneic cycles on average (n = 3).

L-type Ca2+ currents are present in inspiratory cells (Elsen and Ramirez 1998) and have been proposed to play a role in the shaping of respiratory neural discharge (Rybak et al. 1997). Blockade of L-type channels with nifedipine did not abolish either eupnea or sighs, indicating that they are not essential for either pattern. There was a significant reduction in eupneic burst duration, suggesting that while the active membrane properties conferred by L-type currents are not essential for rhythmogenesis, they might contribute to the strength of the burst.


Our results demonstrate that a partial blockade of α1A calcium channels abolishes sighs but not eupnea. Similar results were obtained with antagonists of non-NMDA glutamate receptors, suggesting that it is in their presynaptic role that these channels are essential. Experiments looking at EPSPs directly found a range of responsiveness to blockade of α1A channels, ranging from nearly complete elimination to virtually no response at all. These results suggest that a variety of presynaptic calcium channels are present in the respiratory network of mice with some synapses employing predominantly α1A channels with a P-type pharmacology, whereas others take a larger contribution from Q-type α1A channels or α1E (R-type) channels. Our finding is consistent with the hypothesis that among the glutamatergic synapses subserving eupneic rhythmogenesis, those sensitive to Aga-TK comprise a distinct subset, which lack NMDA receptors postsynaptically, and play a unique and essential role in the production of sighs. However, it must be emphasized that further cellular investigations will be necessary to identify those respiratory neurons that possess these predicted synaptic properties. This identification could then lead to a better understanding of how distinct excitatory synapse types may contribute to the emergence of the large-amplitude sigh burst at a network level.


This work was supported by National Institutes of Health Grants 60120 and 5 T32 GM-07281.


  • 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.


View Abstract