Isoflurane Reduces Excitability of Drosophila  Larval Motoneurons by Activating a Hyperpolarizing Leak Conductance

Experimental cultures were maintained at 25°C by daily transfers on standard cornmeal-molasses medium in 50-ml vials. The wild type was Canton-S , with the exception of whole cell recordings, which were performed in the transgenic line w; UAS-mCD8-GFP; D42-gal4, Cha-gal80 ,15obtained from the Levine laboratory (University of Arizona, Tucson, AZ). para^ts1  mutants from lab stocks were made congenic with Canton-S  by backcrossing for 10 generations to a Canton-S  line with markers flanking the para  locus.16The strong hypomorphic mutation para⁶⁰,  17marked with forked  bristles, was maintained over the green fluorescent protein expressing balancer FM7i .18 para^ts1/para⁶⁰forked  and their heterozygous siblings were generated by crossing Cantonized para^ts1  males with para⁶⁰forked/FM7i  virgins. Genotypes of the female progeny were distinguished on the basis of the green fluorescent protein fluorescence marking the FM7i  chromosome. mle^nap-ts1,  marked with cinnabar  eye color, was obtained from the Ganetzky laboratory (University of Wisconsin, Madison, WI) and maintained as a homozygous stock.

All experiments were conducted at room temperature (22°–23°C).

Two-electrode voltage clamp of the larval neuromuscular junction was performed as described previously.6All larvae were dissected in cold, Ca²⁺-free HL-3 saline,19and most preparations were superfused with HL-3 saline containing 1.0 mm Ca²⁺, pH 7.1. To maintain excitability of mle^nap-ts1  mutants, some experiments were performed using HL-3 with Mg²⁺reduced from 20 mm to 4 mm (“HL-3.1”).20When required, pH was manipulated with either 1 m hydrochloric acid or 1 m NaOH. Muscle 6 of the fifth abdominal segment (A5) was voltage clamped to −70 mV (AxoClamp 2B; Axon Instruments, Union City, CA), and stimuli were delivered to the cut end of the segmental nerve by a glass-tipped suction electrode, placed at the A3-A4 boundary to ensure consistent latency between preparations. Latency was measured from the onset of the stimulus artifact to the initial inflection of the synaptic currents from baseline. Stimuli, solution changes, and recordings were controlled by pClamp 8.2 (Axon Instruments), and synaptic currents were analyzed with MiniAnalysis software (Synaptosoft, Decatur, GA).

Whole cell, tight-seal recordings were performed largely as described,21with slightly modified external (118 mm NaCl, 2 mm NaOH, 2 mm KCl, 4 mm MgCl², 4 mm CaCl², 40 mm sucrose, 5 mm trehalose, 5 mm HEPES; pH 7.1; osmolality to 285–305 mmol/kg with 2 m sucrose) and internal salines (130 mm K-gluconate, 2 mm NaCl, 10 mm HEPES, 1 mm EGTA, 2 mm MgCl², 0.1 mm CaCl², 10 mm KOH; pH 7.2; osmolality to 280–300 mmol/kg with 1 m K-gluconate).

RP2 (i.e. , MNISN1s22) cell bodies were visualized by membrane-bound green fluorescent protein expression in the transgenic line w; UAS-mCD8-GFP; D42-gal4, Cha-gal80 . The number of labeled cells varied between animals, but RP2 could usually be identified in segment A4 or A5 based on its anterior position in the motoneuron cluster. The central nervous system was removed from the larva, placed on a small chip of poly-l-lysine-coated glass coverslip, and then moved to the recording chamber on a fixed stage under a compound microscope equipped with water immersion optics and green fluorescent protein filters (Olympus BX51WI; Olympus America, Center Valley, PA). The sheath was removed by applying a solution of 0.05–0.1% collagenase type XIV (Sigma Chemical Company, St. Louis, MO) via  a glass micropipette (tip ID approximately 10 μm) with gentle suction/positive pressure.

Whole cell recordings were performed using a Heka EPC-10 amplifier (Instrutech, Port Washington, NY), with recording and solution changes controlled by Pulse 8.62 software (Instrutech). Patch electrodes (5–7 MΩ) were fabricated from thick-walled capillary glass (World Precision Instruments, Sarasota, FL) on a vertical puller (Narishige PP-830; Narashige International USA, East Meadow, NY). The stage, recording chamber, and superfusion system were those used for neuromuscular recordings. Current clamp recordings were performed at the cells' intrinsic V^m, whereas the cell was held at a resting potential of −70 mV when voltage clamped. Input resistance (Rⁱⁿ) was measured in voltage clamp by imposing 50-ms voltage steps from −130 to −50 mV, in 10-mV increments, and Rⁱⁿwas determined from the slope of the current-voltage curve fit by linear regression.

Isoflurane (Baxter Healthcare, Deerfield, IL) was dissolved into all external salines (HL-3, HL-3.1, patch clamp solution) by vortexing for 1–2 min on the day of the experiment. A concentrated stock solution of tetrodotoxin (Sigma Chemical Company, St. Louis, MO) in distilled water was diluted to the working concentrations at the time of the experiment.

Anesthetic solutions were constantly perfused from sealed reservoirs to an open recording chamber via  a peristaltic pump.6For neuromuscular experiments, the axon was stimulated every 10 s for 25 min, with 5 min for baseline, 10 min of isoflurane application, and 10 min wash. The values presented correspond to the average of stimuli 71–80, normalized to that of stimuli 1–10, for each preparation. Because the synapse shows slight activity-dependent depression,6,23the values for excitatory junctional current (EJC) amplitude were lower than, and those for latency were higher than, 1.0 in untreated preparations.

To confirm the concentration of isoflurane for each experiment, 50–200 μl saline was removed from the recording chamber and placed in a headspace vial. The concentration of isoflurane in the sample was determined by headspace analysis, using either a Shimadzu GC-9A gas chromatograph (Shimadzu Scientific Instruments, Columbia, MD6) or an Agilent 6850 gas chromatograph with a model 7683 autoinjector (Agilent Technologies, Wilmington, DE). Results from the two chromatographs were identical. Experiments were binned on the basis of anesthetic concentration ±0.025 mm, such that, e.g. , 0.2 mm = 0.176 to 0.225 mm.6 

The neuromuscular junction was modeled with the Neuron Simulation Environment version 5.9.24The model comprised an axon with Hodgkin-Huxley-style conductances that terminated in a bouton with voltage-gated Ca²⁺channels (fig. 1A).

The anatomical parameters for Drosophila  larval axons and boutons were based on reported values,25,26with a 1.5-μm diameter, 5-mm-long axon that terminated in a 5-μm-long, 4-μm-diameter type 1b bouton.

The biophysical properties of the presynaptic axon were based on a combination of measured and estimated values (tables 1 and 2). External ionic concentrations were based on HL-3 saline containing 1.0 mm Ca²⁺, and equilibrium potentials were calculated using estimated intracellular ionic concentrations and the Nernst equation. Axial resistivity was 150 Ω cm, membrane capacitance was 1 μF/cm², and external resistivity was considered negligible. Membrane currents were described by I^ion =g^ion  (V − E^ion ). Conductances for voltage-gated channels were modeled using modifications of the equations described by Hodgkin and Huxley27(table 1). In the absence of data for larval axons, the parameters for voltage-gated Na⁺and K⁺conductances (g^NaV  and g^KV ) were derived from those of the squid giant axon,27with the polarity of membrane potential reversed to reflect present convention, and time constants adjusted to a recording temperature of 22°C using Q¹⁰= 3.24The dominant presynaptic Ca²⁺conductance in Drosophila  motoneuron terminals is mediated by channels resembling N-type Ca²⁺channels (g^CaN ),28,29modeled using the parameters for the N-type Ca²⁺conductance described by Benison et al. ,30†without modification.

The axon was divided into 51 segments, while the bouton was modeled as a single segment, isopotential with the axon. Because the primary goal of the simulations was to examine the relation between action potential shape and Ca²⁺influx, it was deemed undesirable to add the complexity of a more realistic terminal arbor into the analysis. A leak with E^ion= V^m(here called g^pas ) was distributed in all compartments, while g^NaV , g^KV , and leak K⁺and Cl⁻conductances (g^KL  and g^ClL ; table 2) were distributed uniformly in the axon. g^CaN  was restricted to the bouton.29The starting values of g^NaV , g^KV , and g^pas  were adjusted to generate a baseline conduction velocity of 40 cm/s². To examine the effects of g^KL  and g^ClL  in isolation, the baseline value for each leak was zero. Where it has been measured at invertebrate neuromuscular synapses,31neurotransmitter release is proportional to [Ca²⁺]ⁱ⁴so glutamate release was taken to be proportional to the fourth power of total Ca²⁺influx (Q^Ca), measured as the area of I^Ca.

To simulate the axons of para  and mle  mutants, ḡ^NaV  was reduced by 50%. This was the phenotype described biochemically for mle^nap-ts1  32and the primary electrophysiologic defect in para^ts2 ,33which have the identical molecular change as para^ts1 .34 

Anesthetic action on voltage-gated channels can potentially be modeled as effects on many aspects of kinetics and voltage dependence. However, at clinically relevant concentrations, the most consistently reported effects of anesthetics on neuronal voltage-gated conductances are alterations in open probability, single channel conductance, and/or steady state inactivation.7,10,35Similarly, anesthetics have been shown to alter open probability for non-voltage-gated channels.36The primary effect of these changes on single action potentials at the macroscopic level is altered channel availability, and therefore, anesthetic action was modeled as altered maximum conductance (ḡ ). Conductances were altered in 10 steps, starting at baseline and ending at failure of action potential conduction.

Data are presented as mean ± SEM. In most cases, groups were compared statistically using one-way analysis of variance (ANOVA), followed by the Dunnett test when more than two groups were compared. When assumptions regarding homoskedacity were violated, as indicated by the Levene test, groups were compared using the Kruskal-Wallis test, followed by repeated Mann-Whitney U tests using the Bonferroni correction for repeated comparisons. All tests were performed using SPSS software (SPSS Inc., Chicago, IL).

Linear regression was performed in origin 5.0 (OriginLab Corp, Northampton, MA).

At the neuromuscular junction, isoflurane reduces presynaptic excitability and neurotransmitter release.6This is manifested as increased latency between nerve shock and the onset of synaptic currents (EJCs), and decreased amplitude (fig. 2A). Sandstrom6described the following properties for isoflurane-mediated presynaptic inhibition. First, increased latency results from slowing of action potentials in motoneuron axons. Second, EJCs were restored to control amplitude by direct electrotonic stimulation of the synapse, indicating that isoflurane was acting on the size and shape of the action potential, rather than affecting Ca²⁺channels or the vesicle release machinery. Finally, the EC⁵⁰for the synaptic effect (0.17 ± 0.01 mm) was in the same range as that for larval immobility (0.21 ± 0.01 mm). Larvae were unaffected at concentrations below 0.1 mm and immobilized above 0.3 mm.

Based on these observations, the strongest candidate anesthetic targets would be axonal ion channels, including the voltage-gated Na (Na^V) channel, encoded by a single, X-linked gene, paralytic  (para ). Indeed, viable mutants with reduced para  function are hypersensitive to volatile anesthetics in assays of adult fly locomotion.8,9When I applied isoflurane to neuromuscular junctions of para^ts1/Y  mutant males and measured EJC latency (which reflects conduction velocity), the mutants were significantly more sensitive to isoflurane than Canton-S  male congeners (fig. 2B). Latency in untreated para  preparations was not different from Canton-S , but the responses to 0.1 and 0.15 mm isoflurane were significantly larger in the mutants (P < 0.02, one-way ANOVA).

To map enhancement to the para  locus more definitively and to examine allele specificity, I tested additional mutants affecting Na^Vfunction. For these tests, I took advantage of the fact that 0.1 mm isoflurane does not significant affect wild-type preparations (figs. 2B and C; P > 0.05, one-way ANOVA), but significantly increases latency in para^ts1  (P < 0.001, one-way ANOVA). Therefore, enhancement was defined as a significant increase in latency by 0.1 mm isoflurane. I used para⁶⁰ , which is stronger than para^ts1  but not a null allele,17to further examine the dependence of the isoflurane phenotype on the para  locus. In para⁶⁰/para^ts1  females, 0.1 mm isoflurane caused significantly increased latency (fig. 2C; P < 0.05, Mann-Whitney U test). Siblings heterozygous for para^ts1  and the FM7i  balancer chromosome (which is wild type for para  function), were not significantly affected by isoflurane (fig. 2C; P > 0.05, one-way ANOVA). Therefore, enhancement by para^ts1  maps to the para  locus, affects both sexes, and is recessive.

The question remained as to whether the enhancement of isoflurane by para  mutants was due to molecular lesions in the channel or a more general reduction of Na^Vfunction. To distinguish between these alternatives, I tested an allele of the maleless  (mle ) gene, mle^nap-ts1 , which has similar temperature-sensitive paralysis to para^ts1 ,37and alters the expression of Na^Vchannels without altering their structure.32To accommodate the reduced excitability of mle^nap-ts1  mutants, recordings were performed in modified saline (HL-3.1), reported to better preserve axonal excitability.20HL-3.1 saline did not affect the response of wild type or para  to 0.1 mm isoflurane (fig. 2D), in that Canton-S  showed no significant response (P > 0.05, one-way ANOVA) and para^ts1  mutants were still hypersensitive (P < 0.05, one-way ANOVA). Like para  mutants, mle^nap-ts1  preparations were hypersensitive to 0.1 mm isoflurane (fig. 2D; P < 0.05, Mann-Whitney U test). Therefore, the enhancement of isoflurane's effects by para  and mle  mutants reflects a general reduction in Na^Vfunction, rather than a specific change in the kinetics or anesthetic binding of the Para channel itself.

Although it demonstrated that para  mutations act on anesthetic-sensitive neurons, this experiment did not address whether isoflurane acts by altering Na^Vchannel function or whether it sensitizes axons to the effects of other isoflurane-modulated conductances. Because conduction velocity and neurotransmitter release depend on many factors that vary nonlinearly in space and time, I turned to computer simulation to test potential biophysical mechanisms.

To examine how selected conductances affect measurable parameters of presynaptic function, I generated a minimal model of the axon and nerve terminal (fig. 1A; details in Materials and Methods), modeling the effect of anesthetic as altered maximum conductance (ḡ ). Although other conductances were modeled ( appendix 1), results of the following are presented here because of their potential roles in anesthetic mechanisms: voltage-gated Na⁺and K⁺(Na^Vand K^V), and non-voltage-gated, or leak, K⁺and Cl⁻currents (K^Land Cl^L).

In the simulations, reduction of ḡ^NaV  resulted in spikes with decreased conduction velocity, amplitude, and width before ultimately resulting in conduction failure (B1 in fig. 1, top trace, and  appendix 1). These smaller spikes produced less depolarization of the terminal and therefore activated less Ca²⁺influx into the bouton (B1 in fig. 1, bottom trace). Activation of a K⁺leak current produced a superficially similar effect, but slowed the action potential more and reduced I^Caless dramatically before failure (B2 in fig. 1).

Direct comparison of the results of manipulating different conductances (e.g. , B1 and B2 in fig. 1) was problematic because the independent variable changed with each conductance tested. Because the experimental data are expressed in terms of latency and amplitude of synaptic currents, the results of the simulations were plotted as the relation between spike latency and Ca²⁺influx raised to the fourth power (Q^Ca4), which is proportional to neurotransmitter release (fig. 1C).

Expressed in this way, alteration of each conductance produced a distinct effect (fig. 1C). g^KL , g^NaV , and g^KV  reduced latency and Ca²⁺influx, but with increasingly steep slopes, indicating that the voltage-gated conductances have relatively stronger effects on spike size and shape than conduction velocity. Activation of g^ClL  hyperpolarized the terminal more strongly than g^KL  and resulted in larger Ca²⁺influx even as it slowed action potentials.

To test the predictions of the model regarding g^NaV , I applied low concentrations of tetrodotoxin (0.1–0.3 μm), a specific and potent inhibitor of Na^Vchannels, to the neuromuscular junction. Consistent with the model, tetrodotoxin caused significant, concentration-dependent reduction in EJC amplitude (fig. 3A, open circles; P < 0.01, one-way ANOVA), while having a nonsignificant effect on EJC latency (fig. 3A, filled squares; P > 0.05, one-way ANOVA).

Isoflurane had a very different effect, increasing latency significantly (fig. 3B, filled squares; P < 0.05, Kruskal-Wallis), while having a smaller but significant effect on EJC amplitude (fig. 3B, open circles; P < 0.05, one-way ANOVA).

Replotting the above latency and amplitude against one another highlighted the differences between the effects of tetrodotoxin and isoflurane (fig. 3C). The latency-amplitude relation for tetrodotoxin had a steep downward slope with increasing drug concentration (fig. 3C, filled squares). The close match between the experimental result of reducing g^NaV  with tetrodotoxin and the predicted effect of reducing ḡ^NaV  in simulations (fig. 3C, dashed line) provided strong support for the assumptions underlying the model.

The relation for isoflurane was much more shallow (fig. 3C, open circles), with increasing concentrations of isoflurane affecting latency more strongly than amplitude. This resembled the effect of the simulated leak K⁺current (fig. 3C, solid line) more than any of the other conductances tested in the model. Because the slopes of the latency-amplitude relations for tetrodotoxin and isoflurane are so dissimilar, it is highly unlikely that isoflurane acts by inhibiting Na^Vchannels in this preparation. Instead, the data are consistent with isoflurane activating a hyperpolarizing leak conductance.

However, para  and mle  mutant axons were undeniably hypersensitive to isoflurane (figs. 2B and C). To gain insight into the underlying mode of action, I modeled mutant axons as having ḡ^NaV  reduced by 50% (see Materials and Methods). The model indicated that reduction of Na^Vchannel function would enhance the sensitivity of the model axons to further reductions in excitability, such as activation of g^KL  (fig. 4). Therefore, enhancement by para  and mle  seems to be a result of generalized reduction of excitability in the mutants.

The above experiments provide indirect evidence for the activation of g^KL  by isoflurane. To look at leak channel activity more directly, I performed whole cell recordings from larval abdominal motoneuron cell bodies. All recordings were performed on the dorsal common exciter, RP2, because its axon is sensitive to isoflurane in extracellular recordings.6The RP2 cell body can be located and identified based on its position in the dorsal cluster of each neuromere, and its distinctive physiology, in that a square current pulse elicits a delayed train of spikes (fig. 5A).21 

In current clamp, application of 0.4 mm isoflurane caused a significant and reversible hyperpolarization from an average V^mof −48.9 ± 2.6 mV to −55.6 ± 1.2 mV (fig. 5B; P < 0.05, one-way ANOVA). As at the neuromuscular junction, responses to application and removal of the drug occurred rapidly upon exchange of solutions in the chamber. The V^mat the soma was more positive than that used for the simulations of the axon (−70 mV), consistent with the relatively hyperpolarized V^mof nerve terminals.38 

Under voltage clamp, isoflurane caused a decrease in input resistance (Rⁱⁿ; figs. 5C and D). Control Rⁱⁿwas 2.54 ± 0.18 GΩ, consistent with values reported previously.21Application of 0.3–0.4 mm isoflurane produced a reversible and statistically significant reduction in Rⁱⁿ(figs. 5C and D; P < 0.05, one-way ANOVA). Subtraction of the pretreatment current-voltage curves from those recorded in isoflurane yielded an isoflurane-activated conductance of 97.5 pS that reversed at −72.7 ± 2.2 mV (fig. 5E), values consistent with the hyperpolarization observed in current clamp mode (fig. 5B).

The best-characterized anesthetic-sensitive leak channels are K⁺channels in the tandem-pore, or K^2Pfamily. Most K²P channels are inhibited by acidic extracellular pH,11,39so I attempted to block isoflurane's effect on the neuromuscular junction with extracellular acidification.

At the neuromuscular junction, low pH reduces neurotransmitter release and alters glutamate receptor kinetics (Badre et al.  40and David Sandstrom, Ph.D., Bethesda, Maryland, unpublished electrophysiologic recordings of the Drosophila  neuromuscular junction acquired between April 12, 2005, and August 25, 2005). However, the effect of 0.3 mm isoflurane on EJC latency and amplitude was not altered at pH 5.75 compared with pH 7.1 (figs. 6A and B; P < 0.05, one-way ANOVA). Therefore, isoflurane seems to be acting via  a novel, pH-insensitive leak channel.

Although the precise mechanisms of volatile general anesthetic action continue to be elaborated, there is increasing evidence that presynaptic actions are common and important.3I reported previously that isoflurane acts presynaptically at the Drosophila  larval neuromuscular junction, reducing conduction velocity and neurotransmitter release by decreasing excitability of motoneuron axons. Multiple lines of evidence in the current study indicate that this is the first example of presynaptic inhibition by an anesthetic-modulated leak conductance.

The current study focused primarily on an insect neuromuscular junction, but its goal was to identify general cellular and molecular mechanisms of anesthetic action. This synapse has been used to successfully identify molecular and cellular mechanisms of excitability and synaptic transmission relevant to most metazoans, so it is likely that the effects described here will generalize to other synapses across phyla. Further, because the sensitivity of flies and larvae to a wide range of volatile anesthetics is nearly identical to those of mammals,6,41,42there are likely to be many shared targets.

At the larval neuromuscular junction, isoflurane affects both synaptic transmission and axonal conduction. Based on the simulations, this is the expected result for any reduction in axonal excitability, because altering a given ionic conductance will affect both the shape and speed of the action potential. Because changes in action potential size and shape are ultimately translated into altered neurotransmitter release,5this result is consistent with the prevailing view that anesthetics act predominantly on synaptic transmission.

Unlike synaptic transmission, however, axonal conduction has received less attention as a process relevant to anesthesia.3This is in part because axonal conduction seems to be insensitive to anesthetics in some preparations.43However, it is affected measurably in others.44,45Even in the same tissue, the axons of different cell types can differ in their sensitivity. For example, in the CA1 region of the hippocampus, isoflurane can increase or decrease conduction velocity of presynaptic neurons, or have no effect, depending on whether the stimulus is applied to the stratum radiatum, stratum oriens, or Schaffer collaterals, respectively.46,47Superimposed upon this, different volatile anesthetics often have varying effects on conduction velocity in the same axons.46–48Regardless of the variation in effects, there are clear examples of volatile anesthetics affecting conduction velocity significantly.

The data presented here suggest that conduction velocity will be affected most strongly when hyperpolarizing leak currents are activated. At present, the effects of anesthetics on leak currents have mostly been measured from neuronal cell bodies or in heterologous expression systems, in which conduction velocity is not or cannot be measured.36,49,50Therefore, further study is likely to uncover additional examples of anesthetics significantly slowing axonal conduction. Precise timing of neural events is critical for sensory processing,51synaptic plasticity,52and binding of anatomically dispersed neural ensembles.53Therefore, even small changes in conduction velocity have the potential to profoundly alter memory and awareness, which are precisely the aspects of neural function targeted by general anesthetics.

To better understand the mechanisms of isoflurane action on conduction velocity and glutamate release at the larval neuromuscular junction, I developed a model of the axon and presynaptic terminal. Although the model is somewhat generic in terms of the anatomy and biophysics of the presynaptic elements, it captures the essence of the interactions between the elements that vary nonlinearly in space, time, and voltage.

For example, the model predicted a steep latency-amplitude relation for reduction of Na^Vfunction. Concentration-response experiments with tetrodotoxin confirmed this prediction, with partial blockage of the channels having a strong effect on neurotransmitter release relative to conduction velocity. Although there are possible differences in detail between the actions of tetrodotoxin and volatile anesthetics on Na^Vchannels,4the effect on macroscopic currents is not expected to be dramatically different, and other agents, such as local anesthetics, affect possible targets such as K^2Pleak channels.54Because the muscle is innervated by only two motoneurons, the graded reduction of EJC amplitude by tetrodotoxin cannot be due to presynaptic spike failure and must result from reduced action potential amplitude.

The model predicted a shallow latency-amplitude relation for activation of g^KL , which matched the effects of isoflurane. The presence of an isoflurane-activated leak in whole cell recordings of motoneuron cell bodies supported this result. Numerous other channels types are modulated by anesthetics in other preparations, but they produce effects different from those I observed. For example, although volatile anesthetics inhibit K^Vchannels in some preparations,7the resulting spike broadening strongly increases neurotransmitter release,55and simulations show that inhibiting K^Vchannels produced a small acceleration of conduction velocity ( appendix 1). Isoflurane-mediated inhibition of Ca²⁺channels56is contraindicated at the larval neuromuscular junction by experiments in which EJC amplitude is rescued by direct electrotonic activation of the synapse,6and by simulations showing that reduction of g^CaN  reduces Ca²⁺influx without affecting conduction velocity ( appendix 1). Similarly, genetic experiments have implicated components of the vesicle release machinery in anesthesia,57but alteration of neurotransmitter release is not expected to alter conduction velocity. Anesthetics probably act on multiple targets and neuroanatomical loci,1and any single preparation is not likely to demonstrate all possible effects. The current study introduces a novel way in which anesthetic-activated leak can alter presynaptic function.

It should be noted that in the model the only difference between g^KL  and g^ClL  is the predicted equilibrium potential for K⁺and Cl⁻. The predicted E^K(−72.8 mV) is only slightly negative to estimated axonal V^m(−70 mV). Therefore, the slowing of the action potential must largely be due to the reduction of axonal membrane resistivity, which retards the axial spread of depolarizing current. Activation of g^ClL(E^Cl=−85 mV) strongly hyperpolarizes the axon, which increases the driving force on Na⁺and Ca²⁺and removes resting inactivation of Na^Vand Ca^Nchannels, which increases spike amplitude and Ca²⁺influx. This occurs in parallel with the slowing of spikes due to increased axonal leakiness. Because the equilibrium potentials for K⁺and Cl⁻are not currently known, the results presented here indicate the presence of a leak conductance that reverses slightly below axonal V^m, but more experiments will be required to determine whether isoflurane activates a conductance carrying K⁺or Cl⁻, or multiple leak conductances carrying a combination of the two.

The combination of simulation, electrophysiology, and genetics was especially useful in understanding how mutations in para  cause sensitization to anesthetics. It has long been known that mutations in para  sensitize flies to volatile anesthetics,8,9but it has remained controversial whether Na^Vchannels encoded by para  mediate the effects of anesthetics in Drosophila  or whether reduction of para  function sensitizes the system to the effects of other targets.42In the current study, I demonstrated that para  mutations sensitize axons to isoflurane but that Na^Vchannels do not mediate isoflurane's effects on motoneurons. Instead, the model indicates that any reduction of para  function will render mutant axons hypersensitive to the alteration of conductances that reduce excitability, such as g^KL .

The model is an oversimplification of the anatomy and biophysics of the presynaptic elements, but the data are currently too limited to warrant more elaborate simulations. The information available for axonal and synaptic conductances derives primarily from either anatomical localization studies or the effects of manipulation of gene function (see references in Materials and Methods), which reveal the presence of molecules in the presynaptic terminal, but not their density or levels of activity. In its current form, the model complements the many approaches used at the neuromuscular junction to explore the effects of drugs and mutations on synaptic physiology, and more complex geometric or biophysical features can be added as data become available.

The activation of hyperpolarizing leak currents by volatile anesthetics has been reported in diverse vertebrate and invertebrate preparations.49,50,58–60Candidates include the K^2P11and adenosine triphosphate-sensitive K⁺(K^ATP)61channel families and tonically active γ-aminobutyric acid type A receptors.12However, the larval neuromuscular junction is insensitive to γ-aminobutyric acid agonists and antagonists62and to high concentrations of the K^ATPactivators levcromakalim and diazoxide ( appendix 2). This leaves K^2Pchannels as likely targets at present.

The K^2Pfamily contains the best-studied anesthetic-activated leak channels, with the mammalian [tandem of pore domains in a weak inward rectifying potassium channel]-related acid-sensing potassium channel (TASK), [tandem of pore domains in a weak inward rectifying potassium channel]-related spinal cord channel (TRESK), and [tandem of pore domains in a weak inward rectifying potassium channel]-related potassium (TREK) subfamilies being activated by volatile anesthetics.11,36Furthermore, loss of TREK-1 function reduces anesthetic sensitivity in mice,63providing behavioral context for the results from reduced preparations. Two of the 11 Drosophila  K^2Pfamily members, TASK-6 and TASK-7, resemble their mammalian homologs in that they are inhibited by extracellular acidification, and are the only Drosophila  K^2Pchannels that contain a six amino acid sequence associated with anesthetic sensitivity in TASK channels.39,64However, neither of the Drosophila  TASKs are sensitive to the volatile anesthetic halothane in heterologous expression systems.39Further, the current study showed that the larval isoflurane response is unaffected at pH 5.75, which blocks known pH-sensitive K²P channels.39,50The K^2Pchannel family remains of great interest, but isoflurane must act on an as-yet-undescribed domain in a pH-insensitive family member. The combined use of genetic, molecular, and electrophysiologic approaches available in Drosophila  will greatly facilitate the ongoing task of identifying isoflurane target(s).

The author thanks Howard Nash, M.D., Ph.D. (Senior Investigator, National Institute of Mental Health, National Institutes of Health, Bethesda, Maryland), for space, material, and intellectual support; Benjamin White, Ph.D. (Investigator, National Institute of Mental Health, National Institutes of Health), for advice and the use of equipment; James C. Choi, Ph.D. (Postdoctoral Fellow, Children's Hospital, Boston, Massachusetts), Leslie C. Griffith, M.D, Ph.D. (Professor, Department of Biology and Volen Center for Complex Systems, Brandeis University, Waltham, Massachusetts), Jonathan Z. Simon, Ph.D. (Assistant Professor, Department of Electrical and Computer Engineering, University of Maryland, College Park, Maryland), Jeffrey S. Diamond, Ph.D. (Senior Investigator), and Annalisa Scimemi, Ph.D. (Fellow, National Institute of Neurologic Disorders and Stroke, National Institutes of Health), for advice during the project; David Ide, B.A. (Research Assistant, Research Services Branch, National Institute of Mental Health, National Institutes of Health), for the design and construction of equipment; Barry Ganetzky Ph.D. (Professor, Department of Genetics, University of Wisconsin, Madison Wisconsin), Rene Luedeman, M.S. (Research Specialist), and Richard B. Levine, Ph.D. (Professor, Arizona Research Laboratories, Division of Neurobiology, Tucson, Arizona), for fly stocks; and Barry A. Trimmer, Ph.D. (Professor, Department of Biology, Tufts University, Medford, Massachusetts), and Ling-Gang Wu, M.D., Ph.D. (Senior Investigator, National Institute of Neurologic Disorders and Stroke, National Institutes of Health), for comments on the manuscript.

Figure 7shows examples of the effects of altering selected conductances on the velocity and shape of action potentials, as well as the changes in I^Cathat result from altered spike shape. Reducing excitability by changing any of the conductances slowed spikes. Manipulating g^NaV  and g^KV  increased latency by approximately 44% (A1 and A2 in fig. 7), whereas increasing g^KL  or g^ClL  slowed spikes by 77–79% (A3 and A4 in fig. 7).

Altering each conductance had distinct effects on spike size and shape (figs. 7B and 7C). Progressively decreasing g^NaV  reduced both the amplitude (B1 in fig. 7) and width at half amplitude (t^1/2; C1 in fig. 7). Increasing g^KV  had stronger effects, with spike amplitude and t^1/2decreasing relatively steeply before action potential failure (B2 and C2 in fig. 7). The effects of increasing g^KL  were more subtle, in that amplitude and t^1/2decreased by only a few percent each (B3 and C3 in fig. 7). Curiously, g^ClL  activation resulted in increases in both the amplitude and the t^1/2of the action potential (B4 and C4 in fig. 7), presumably because the g^Cl -induced hyperpolarization from −70.2 to −78.4 increased the driving force on Na⁺and decreased resting inactivation of Na⁺channels.

Ca²⁺influx (Q^Ca) was raised to the fourth power to simulate neurotransmitter release,31which therefore amplified the effect of isoflurane on spike size and shape (fig. 7D). Alteration of three of four simulated conductances (g^NaV , g^KV , and g^KL ) caused a decrease in Q^Ca4, with differences in the steepness of the slopes (D1-D3 in fig. 7). As expected from the effects of g^ClL  activation on spike amplitude and half-width, Q^Ca4became larger as conductance was increased (D4 in fig. 7).

In addition to the latency-amplitude relations shown in fig. 7C, I simulated the effects of isoflurane on a number of conductances that were less likely to mimic isoflurane. Among these were inhibition of K^Vand Ca^Nconductances (fig. 7E). Inhibition of g^KV  caused a slight increase in conduction velocity and, as expected from experimental data,55a dramatic increase in Ca²⁺influx (circles with plus signs). Inhibition of g^CaN  reduced Ca²⁺influx, without affecting conduction velocity (open diamonds).

Adenosine triphosphate-sensitive K⁺channels are expressed in the Drosophila  larval ventral nerve cord,69and some mammalian K^ATPisoforms are sensitive to isoflurane.61In mammals, K^ATPchannels comprise octamers of four inward rectifier K⁺channels and four sulfonylurea receptors (SURs). I challenged the larval neuromuscular junction with the K^ATPchannel openers levcromakalim, which activates SUR2 containing K^ATPcomplexes, and diazoxide, which can activate complexes containing SUR2 or SUR1.70Both drugs (Tocris Life Sciences, St. Louis, MO) were dissolved in dimethyl sulfoxide, diluted to the working concentration in physiologic saline, and applied in the manner described for other drugs in Materials and Methods. Axonal conduction velocity was unchanged by 75 μm levcromakalim (Levcro) or by diazoxide at 50 or 100 μm (fig. 8; P > 0.05, one-way ANOVA). Although there is some evidence that the pharmacology of Drosophila  SURs differs from that of mammals,71the insensitivity of the synapse to two K^ATPchannel openers with different specificities does not support a role for K^ATPchannels in the response to isoflurane.