Halothane Depresses Glutamatergic Neurotransmission to Brain Stem Inspiratory Premotor Neurons in a Decerebrate Dog Model

This research was approved by the Medical College of Wisconsin Animal Care Committee (Milwaukee, Wisconsin) and conformed with standards set forth in the National Institutes of Health Guide for Care and Use of Laboratory Animals. Anesthesia was induced in dogs by mask with isoflurane, and the dogs were intubated with a cuffed endotracheal tube and from then on were mechanically ventilated with oxygen. Isoflurane (1.3–1.8 MAC) was applied throughout the surgical procedures and only discontinued after completion of decerebration (1 MAC isoflurane in dogs = 1.4%10). The animals were positioned in a stereotactic device (model 1530; David Kopf Instruments, Tujunga, CA) with the head ventrally flexed (30°). Bilateral neck dissections were performed. The C5 phrenic nerve rootlet was de-sheathed for recording, and bilateral vagotomy was performed to achieve peripheral deafferentation. This avoids interference of the artificial ventilation with the underlying central respiratory rhythm. Bilateral pneumothorax was performed to minimize brain stem movement and phasic inputs from chest wall mechanoreceptors. The animals were decerebrated at the midcollicular level 11and only then paralyzed (0.1 mg/kg pancuronium, followed by 0.1 mg · kg⁻¹· h⁻¹). An occipital craniotomy was performed to expose the dorsal surface of the medulla oblongata for single-neuron recording. Esophageal temperature was maintained at 38.5 ± 1°C. Mean arterial pressure was kept above 100 mmHg and did not differ more than 20% between 0 and 1 MAC. Blood pressure was supported as needed with infusions of phenylephrine (0.5–5 μg · kg⁻¹· min⁻¹). Protocols were only performed during steady state conditions for blood pressure.

Multibarrel compound glass micropipettes consisting of a recording barrel containing a 7-μm carbon filament, and three drug barrels were used to simultaneously record extracellular neuronal action potential activity before and during pressure ejection of the glutamate agonists and GABA^Areceptor antagonist onto inspiratory neurons of the cVRG. We used the selective glutamate receptor agonists AMPA (7.5 μm; Research Biochemicals, Natick, MA) and NMDA (200 μm; Research Biochemicals) and the GABA^Aantagonist bicuculline methochloride (BIC, 200 μm; Research Biochemicals), which were dissolved in an artificial cerebrospinal fluid (aCSF). 4Meniscus changes in the drug barrels were measured to determine the ejected dose rates (resolution, 2 nl). The neurons were located approximately 1.5–3 mm caudal from the obex and 2.5–4.5 mm lateral from the midline. A prior study had shown that approximately 90% of these neurons were bulbospinal, 12,i.e. , their soma was located in the brain stem and their axons were projected to motoneurons in the spinal cord as confirmed with antidromic stimulation techniques. 12Single-cell inspiratory neuronal activity, phrenic nerve activity, picoejection marker pulses, airway carbon dioxide and volatile anesthetic concentrations, systemic blood pressure, and airway pressure were recorded on a digital tape system (model 3000A; A.R. Vetter Co., Rebersburg, PA). These variables or their time averages were also continuously displayed on a computerized chart recorder (Powerlab/16SP; ADInstruments, Castle Hill, Australia). Timing pulses at the beginning and end of neural inspiration were derived from the phrenic neurogram and were used to determine the respiratory phases. The tape-recorded data were digitized and analyzed off-line. Cycle-triggered histograms (CTHs), triggered at the onset of phrenic activity, were used to quantify the neuronal discharge frequency data.

The protocols were performed under hyperoxic (fraction of inspired oxygen [Fio²] > 0.8) and steady state hypercapnic conditions (arterial carbon dioxide tension [Paco²] 55–65 mmHg). The optimal level of Paco²was adjusted from animal to animal so that adequate phasic respiratory activity during the anesthetic state (1 MAC) was ensured. Great care was taken to keep the Paco²tightly controlled within each neuron protocol. One complete neuron protocol consisted of two sets of three separate picoejection runs (run 1: AMPA, run 2: NMDA, and run 3: BIC;fig. 2). One set was performed in this sequence at 0 and the other set at 1 MAC halothane (0.9%10). To maximize the yield of complete neuron protocols, we performed the current protocols with the order of the sets (0 and 1 MAC) randomized, which eliminated the need for end controls. 5,13Control ejections with the vehicle aCSF, an artificial cerebrospinal fluid in which the neurotransmitters were dissolved, were performed for each experimental setup to confirm lack of vehicle effect.

For the control period (F^con) and at each dose rate, CTHs (5–20 cycles) were used to obtain values of the average peak neuronal discharge frequency (Fⁿ) for each condition. The glutamate agonist AMPA was applied in increasing dose rates until an increase in peak Fⁿof at least 25 Hz was achieved. Typically, picoejection durations of 6–8 min with 2 or 3 dose rates were needed.

The effect of 1 MAC halothane on the postsynaptic AMPA receptor response was quantified by linear regression of Fⁿon dose rate since previous studies have shown the dose-response data for glutamate to be linear. 4,13This was confirmed for inspiratory neurons. In this regression analysis, the y-intercept was constrained to pass through the F^convalue at the zero dose rate. Thus, any change in the slope of the regression line reflected the anesthetic-induced change in the dose-response relation. To compare the dose responses at 0 and 1 MAC, the slope values were then normalized to the slope at 0 MAC (slope⁰), and the normalized difference was determined for each neuron [i.e. , Δslope = (slope¹− slope⁰)/slope⁰]. A Wilcoxon signed-rank test was performed to test if Δslope was significantly different from no change (StatView; SAS Institute, Inc., Cary, NC). To illustrate the effect of an anesthetic-induced change in slope on Fⁿand to allow comparison with our previous expiratory neuron studies, 4we also calculated the net increase in Fⁿat a D^40Hzdose rate for each dose-response relation. For this, the net increase in Fⁿwas calculated from the 1 MAC linear regression line at the dose rate that caused a 40-Hz increase in Fⁿat 0 MAC.

After recovery from AMPA, the same picoejection run was repeated with NMDA. The analyses of the NMDA data were performed using the same procedure as described for AMPA.

After recovery from NMDA, the GABA^Areceptor antagonist BIC was picoejected until complete block of GABA^Aergic inhibition occurred, i.e. , when an increase in picoejection dose rate did not result in any further increase in Fⁿ. Typically, picoejection durations of 5–10 min with several increasing dose rates were required. After the BIC run, complete postejection recovery was awaited, which required 30–45 min. Then, the randomized state of anesthesia was switched, and after a minimum equilibration time of 15 min, both steps of the protocol were repeated in the same fashion. State of anesthesia refers to either 1 MAC anesthesia or absence of anesthesia (0 MAC).

During complete GABA^Aergic block with BIC, Fⁿequals the overall excitatory drive to the neuron (F^e). Under control conditions, the prevailing GABA^Aergic inhibition reduces F^eto F^conby the inhibitory factor α, where α= (F^e− F^con)/F^e. To calculate the change in overall excitatory drive, the data were normalized to F^eat 0 MAC, which was assigned a value of 100%. A two-way repeated-measures analysis of variance (ANOVA) was used with main factors of anesthetic state (0 or 1 MAC) and neurotransmitter status (preejection control vs.  maximal BIC block; SuperANOVA; Abacus Concepts, Inc., Berkeley, CA). The values for F^eand F^conwere obtained for the 0 MAC level (F^con0, F^e0) and the 1 MAC level (F^con1, F^e1) from the experimental runs. They were then used in the calculation of the anesthetic effect on overall excitation ΔF^e(where ΔF^e=[F^e1− F^e0]/F^e0) and overall inhibition Δα (where Δα=[α¹−α⁰]/α⁰). All results are given as mean ± SD, and P < 0.05 was used to indicate significant differences unless stated otherwise.

To examine the reproducibility of the picoejection technique, we performed an additional study in a series of separate animals. The glutamatergic agonist AMPA or NMDA was picoejected in at least two increasing dose rates on cVRG inspiratory or expiratory neurons at 0 MAC (run 1). After recovery to preejection control (F^con), the picoejection was repeated with the same drug (run 2), using the same ejection parameters (ejection pressure, pulse duration, ejection frequency), simulating regular experimental conditions. Note that cVRG expiratory neurons also have both glutamate receptor subtypes, but only NMDA receptors are endogenously active. 7 

Dose-response plots were created from the increases in neuronal frequency following glutamate receptor agonist application for run 1 and run 2. We calculated the linear regression parameters for each plot where the y-intercept was constrained to pass through the F^convalue. The slopes were then normalized to the slope of run 1. In addition, we determined the dose rate that caused a 40-Hz increase in Fⁿin run 1, i.e. , D^40Hz, and calculated the net increase in Fⁿat this dose rate in run 2 using the regression line. To allow for comparison of the current study with previous publications on expiratory neurons, we also calculated the net increases at the D^40Hzdose rate using our previous method based on interpolation between data points 4and compared the results of these two methods. Differences in slope and net increases at D^40Hzbetween run 1 and 2 were tested using a multivariate ANOVA (SuperANOVA).

In the main study, experiments were performed on 28 animals yielding 21 complete neuron protocols. To assess the resolution of the method, we analyzed the reproducibility of dose responses in 15 neurons from 3 animals.

Figure 3shows a representative example of an inspiratory neuronal response to increasing dose rates of AMPA at 0 MAC and 1 MAC halothane, respectively. The highest picoejected dose rates for AMPA were 0.078 pmol/min at 0 MAC and 0.092 pmol/min at 1 MAC. Halothane, 1 MAC, decreased the slope of the linear regression line fitted through the dose-response plots from 582.1 Hz · pm⁻¹· min⁻¹to 488.5 Hz · pm⁻¹· min⁻¹, i.e. , the AMPA receptor response was decreased by 16.1% (fig. 4). This is equivalent to a reduction of the net increase from 40 Hz at 0 MAC to 33.6 Hz at 1 MAC at the D^40Hzdose rate.

The pooled normalized data from 21 neurons for AMPA-induced net increases in peak Fⁿare summarized in figure 5. Halothane, 1 MAC, significantly depressed the AMPA receptor response, determined by the slope of the linear regression, by 18.6 ± 35.7% (P < 0.05). This translates into a depression of the net increase at D^40Hzfrom 40 Hz at 0 MAC to 33.6 ± 14.3 Hz at 1 MAC.

For the same neuron (fig. 6), the slope was 18.6 Hz · pm⁻¹· min⁻¹at 0 MAC and 15.8 Hz · pm⁻¹· min⁻¹at 1 MAC at peak dose rates of 3.6 pmol/min and 4.3 pmol/min NMDA, respectively. Thus, 1 MAC halothane decreased the NMDA receptor function in this neuron by 15.1%, which was equivalent to a reduction of the net increase from 40 Hz at 0 MAC to 34 Hz at 1 MAC at the D^40Hzdose rate.

The pooled normalized net increases in peak Fⁿ, induced by NMDA, are summarized in figure 5. One value was identified as an outlier according to Dixon 14and removed from further analysis. Halothane, 1 MAC, significantly depressed NMDA receptor function, determined by the slope of the linear regression, by 22.2 ± 26.2% (n = 20;P < 0.01). This is equivalent to a depression of the net increase at D^40Hzfrom 40 Hz at 0 MAC to 31.1 ± 10.5 Hz at 1 MAC halothane.

Further analysis of the same inspiratory neuron showed that at 0 MAC, BIC increased peak Fⁿfrom 105 to 267 Hz (fig. 7). This means that under control conditions, the activity of the neuron was attenuated by tonic GABAergic inhibition to 39% of the overall excitatory drive, i.e. , the overall inhibitory factor α⁰was 0.61 (fig. 8, top). At 1 MAC halothane, BIC increased peak Fⁿfrom 75 to 220 Hz, yielding an α¹value of 0.66. Thus, 1 MAC halothane enhanced overall GABAergic inhibition in this neuron by 8%. The overall excitatory drive, F^e, measured at complete BIC block of GABAergic input, was reduced by 18%, while the control frequency F^conwas reduced by 29% by 1 MAC halothane.

The pooled data for 21 neurons show that 1 MAC halothane reduced F^conby 20.6 ± 18.0% (P < 0.001) and overall excitation by 15.4 ± 20.2% (P < 0.01;fig. 8, bottom). Overall inhibition α was 0.60 ± 0.1 at 0 MAC and 0.62 ± 0.1 at 1 MAC (Δα=+5.2 ± 14.9%;P = 0.19).

Figure 9shows a representative example of two consecutive dose responses to NMDA application in the same neuron. Linear regression analysis yielded a slope of 20.1 Hz · pm⁻¹· min⁻¹for picoejection run 1 and of 19.6 Hz · pm⁻¹· min⁻¹for run 2 (fig. 9, top). When the net increases at D^40Hzwere calculated from the linear regression line, the net increase in run 2 was 39 Hz, i.e. , 2.5% lower (fig. 9, middle). When the net increase at D^40Hzwas interpolated from data points on the original dose-response plot, the increase in run 2 was 40.6 Hz, i.e. , 1.5% higher than in run 1 (fig. 9, bottom).

One value was identified as an outlier 14and removed from further analysis. The pooled data from six inspiratory and eight expiratory neurons shows that there was no significant difference in the slope of the linear regression between picoejection run 1 (100%) and run 2 (99.5 ± 27.5%;P = 0.95). This was equivalent to a net increase of 39.8 ± 11.0 Hz for run 2 at D^40Hz(individual data in fig. 10). When the net increase at D^40Hzwas interpolated between the single dose-response values, the net increase in run 2 was 37.6 ± 8.3 Hz (P = 0.30). There was no difference between the values calculated from the linear regression line and from those interpolated between single dose-response values. The multivariate ANOVA showed that the order of the runs did not have a significant influence on the values. Also, there was no difference in results between inspiratory and expiratory neurons.

This in vivo  preparation provides the first conclusive evidence that halothane depressed the activity of inspiratory premotor neurons by reduction of excitatory glutamatergic neurotransmission. Overall inhibitory neurotransmission remained unchanged. A significant part of the reduction of excitatory neurotransmission appeared to be mediated by anesthetic depression of the postsynaptic glutamate receptor response.

In related studies on canine expiratory premotor neurons, the halothane-induced depression in postsynaptic NMDA receptor response (−3.3 ± 38.4%) 4did not reach statistical significance. Similar results were found for sevoflurane. 13It was concluded that excitatory drive to expiratory neurons was primarily reduced at a presynaptic site. This was in agreement with data obtained in rat hippocampal slice preparations, where the magnitude of glutamatergic excitatory postsynaptic currents, elicited by stimulation of the presynaptic Schaffer collaterals, was depressed by clinical concentrations of halothane. 15However, the kinetics of these currents and the receptor response to exogenous NMDA or AMPA were not changed. 15In addition, paired pulse facilitation, i.e. , the ratio of the amplitude of the second to the first evoked excitatory postsynaptic currents, was increased by halothane, 16which also suggested that presynaptic glutamate release was depressed by the anesthetic. Furthermore, a reduction of presynaptic glutamate release by volatile anesthetics has been demonstrated in synaptosomes 17and in rat cortical brain slices. 18 

The decrease in AMPA and NMDA receptor response in the current study stands in contrast to our results in expiratory neurons. 4To exclude that relevant changes in postsynaptic NMDA receptor function had been missed in expiratory neurons, we performed an additional study that assessed the resolution and accuracy of our method. Fourteen protocols from inspiratory and expiratory neurons showed that the dose responses to glutamate agonists were highly reproducible. We used linear regression analysis since it is based on all available dose-response values at one anesthetic level and can thus be expected to be the better representative of the “true” dose response. The average difference between the slopes in run 1 and 2 at D^40Hzwas 0.5 ± 27.5% Hz. With a population SD of 27.5%, based on the dose-response reproducibility study, and a type I error of 0.05, our method can discern a difference of approximately 20% with 90% power, i.e. , a type II error of 0.10. Part of the variability of our method may be related to the limits of discrimination of the drug ejection rate, which is determined from the meniscus changes in the drug barrel of the micropipette. In addition, there is a breath-to-breath variability in neuron activity that is intrinsic to the system (coefficient of variation, 8–10%8). Based on our reproducibility study, we conclude that the significant reduction of the AMPA and NMDA receptor response (−18.6 ± 35.7 and −22.2 ± 26.2%) in inspiratory neurons by halothane represents a true anesthesia-induced finding, which is not due to the variability of the method.

An anesthesia-induced reduction in the postsynaptic glutamate receptor response contrasts with many studies that show that clinical concentrations of volatile anesthetics affect the presynaptic component of excitatory neurotransmission but not postsynaptic NMDA or AMPA receptors. 15,19,20However, Dildy-Mayfield et al.  21expressed different AMPA receptor subtypes in Xenopus  oocytes and showed that the effect of halothane depended on the respective subtype. Halothane had an inhibitory effect on the GluR3 receptor subtype and enhanced GluR6 receptor activity. Furthermore, two studies both performed in rat hippocampal slices reported anesthetic effects on postsynaptic glutamatergic neurotransmission that differed between non-NMDAergic and NMDAergic neurotransmission. Narimatsu et al.  22reported that the halothane concentration necessary to depress population spikes by 50% was significantly smaller when non-NMDA receptors were selectively antagonized than when the NMDA receptors were selectively antagonized and concluded that NMDA receptors were more sensitive to halothane. Nishikawa and MacIver 16found no difference for halothane but described that 1 MAC-equivalent isoflurane preferentially depressed field excitatory postsynaptic potentials mediated by NMDA receptors. In addition, paired pulse facilitation was increased by both anesthetic agents, which suggested that presynaptic mechanisms were also involved in the field excitatory postsynaptic potential depression.

Part of the decrease in overall excitatory drive to inspiratory neurons in our study may have occurred at a presynaptic level. However, an anesthetic-induced reduction of the postsynaptic glutamate receptor response to AMPA and NMDA that clearly exceeded the method-inherent variability was found in a sizeable subgroup of the inspiratory neurons studied (fig. 10). The discrepancy between our results and studies that did not find an anesthetic effect on glutamate receptors may be due to fundamental differences between in vitro  and in vivo  preparations. For example, Takeda et al.  23exposed decerebrate cats to short 90-s bouts of 2% halothane while recording intracellular potentials in inspiratory and expiratory bulbar neurons. They found that regardless of the type of neuron, the anesthetic could cause either membrane depolarization or hyperpolarization, typically combined with a decrease or complete cessation of neuronal firing. In addition, the direction of the anesthetic-induced change in membrane potential, produced by either halothane or intravenous thiopental, was consistent for each neuron. The authors concluded that a reduction of the synaptic input to the neuron, recognizable through an increase in membrane resistance and/or a decrease in respiratory-phase related membrane fluctuations, changed the membrane potential and excitability of the neuron. Dependent on a neuron's specific location within the network, inhibitory and excitatory inputs to the neuron vary so that an anesthetic may alter the excitability of apparently similar neurons in different ways. This “network factor” could account for some of the variability in the effect that 1 MAC halothane showed between individual neurons in our study. In our apparently homogenous population of inspiratory premotor neurons, baseline frequency (F^con) was strongly depressed in some neurons, while others seemed barely affected. It is possible that an anesthetic-induced change in postsynaptic neuron excitability could lead to a reduced response to glutamate agonists in the absence of a direct effect of the anesthetic on glutamate receptor function. Alternatively, the anesthetic may cause a direct depression of the NMDA and AMPA receptor function per se . These two issues cannot be resolved with our methodology.

In the in vivo  network, inspiratory and expiratory premotor neurons form two functionally distinct, antagonistic groups. Since inspiratory premotor neurons of the cVRG directly activate phrenic motoneurons, 1any anesthetic effect on premotor neurons will have a strong impact on the excitatory drive to the inspiratory musculature. We have previously shown that both inspiratory 24and expiratory 7premotor neurons can be excited by exogenous AMPA and NMDA application. However, in expiratory neurons, AMPA receptors are not endogenously activated under physiologic conditions. 7This difference in endogenous activation of AMPA and NMDA receptors suggests that in expiratory neurons, AMPA and NMDA receptors may be localized in anatomically separate postsynaptic locations. For example, a distinct but overlapping distribution of AMPA and NMDA receptors has been shown with immunolabeling in rat VRG neurons. 25 

Comparison of the current results with the effects of halothane in cVRG expiratory neurons suggests that the depression of the glutamate receptor response is unique for cVRG inspiratory neurons. One possible explanation is that the glutamate receptors on inspiratory and expiratory neurons belong to different subtypes with different sensitivities to halothane. 21Another possibility is that anesthetic-induced changes are due to an indirect effect on neuronal excitability rather than on receptor function per se .

In the current study, overall inhibition did not change from 0 to 1 MAC halothane. However, it is highly unlikely that in inspiratory premotor neurons, GABA^Areceptors were not affected by the anesthetic. Multiple studies consistently demonstrate an increase in GABA^Areceptor function with halothane. 5,26We recently showed, for expiratory neurons in vivo , that 1 MAC halothane enhanced overall inhibition by 14.1 ± 17.9%, 5which was due to a large 75% increase in GABA^Areceptor function, while presynaptic inhibitory drive appeared reduced. 5A reduction in presynaptic inhibitory drive is consistent with data by Nishikawa and McIver 27that showed in a rat hippocampal slice preparation that the inhibition of inhibitory interneurons was increased by 1 MAC halothane. We therefore hypothesize that in inspiratory neurons, GABA^Areceptor function is also enhanced by the anesthetic but that a concomitant depression of presynaptic inhibitory drive negates the increase in postsynaptic receptor function. As a result, overall inhibition, which can be understood as the product of presynaptic and postsynaptic inhibition, 5did not change significantly.

We have previously discussed the many advantages of a decerebrate preparation, including that the prevailing physiologic neurotransmitter levels remain similar to those in the intact animal and that single receptor function is not confounded by background anesthesia. 4The importance of the latter cannot be overemphasized. Only recently, Hara and Harris 28expressed various receptors on Xenopus  oocytes and showed that the commonly used and supposedly innocuous background anesthetic urethane depressed AMPA and NMDA receptor currents and enhanced currents through GABA^Aand glycine receptors at a clinical dose (i.e. , ED⁵⁰).

Our preparation did not require a baseline anesthetic. Still, traces of the induction anesthetic isoflurane or the experimental anesthetic halothane may have been present in the brain tissue even though the end-expiratory gas analyzer did not detect them. While it is theoretically possible that residual subanesthetic concentrations of halothane may have contributed to the variability of the data, we can discern highly significant differences between the two anesthetic states, e.g. , for F^eand F^con. Also, there was no statistical difference between protocols that were randomized to start at 1 MAC and those that started at 0 MAC (data not shown), and in previous studies that required 0-to-1-to-0 MAC end control runs, the 0 MAC end control did not differ from the original 0 MAC run. We conclude that the equilibration time before the 0 MAC runs routinely reduced the anesthetic to a concentration at which it would have no more than a minimal effect.

In conclusion, 1 MAC halothane reduced the activity of inspiratory premotor neurons by a reduction of glutamatergic excitation. A significant part of his reduction appeared to be due to a depression of the postsynaptic receptor response to glutamate agonists. The overall GABAergic inhibitory drive to the neurons was not changed.

The authors thank Jack Tomlinson (Biologic Laboratory Technician, Zablocki VA Medical Center, Milwaukee, Wisconsin) for excellent technical assistance.