Anesthetics Have Different Effects on the Electrocorticographic Spectra of Wild-type and Mitochondrial Mutant Mice

THE molecular mechanisms of action of volatile anesthetics are not fully elucidated.^1–3  Ketamine, like the volatile anesthetics, is a complete general anesthetic. Ketamine is an N-methyl-d-aspartate (NMDA) receptor antagonist,^4,5  but also affects other ion channels.^6,7  Moreover, competitive inhibitors with greater affinity for the NMDA receptor than ketamine do not cause the anesthetized state.⁸  Thus, like the volatile anesthetics, the mechanism of action for ketamine is not understood.

Disruption of a specific step of mitochondrial electron transfer, complex I, causes hypersensitivity to volatile anesthetics in humans, nematodes, and mice,^9–11  suggesting an evolutionarily conserved connection between neuronal metabolism and sensitivity to volatile anesthetics. However, the mechanisms linking mitochondrial function to anesthetic sensitivity are unclear. Mice carrying the knockout of the nuclear gene Ndufs4, which encodes an 18-kDa protein of complex I of the mitochondrial electron transport chain, have a striking threefold increase in sensitivity to volatile anesthetics. This is the largest change in volatile anesthetic anesthetic sensitivity measured in a mammal.¹¹  Surprisingly, Ndufs4 homozygous knockout mice are also resistant to ketamine, requiring ~50% higher doses to achieve loss of the righting reflex.¹¹ 

Electroencephalography monitors electrical activity within the brain as oscillations that reflect voltage changes from currents and local field potentials generated by neurons.¹²  There is a growing body of evidence to suggest that different anesthetic agents alter or disrupt the oscillations produced by the brain, and unique patterns of the electrocorticogram can be associated with the anesthetized state caused by specific anesthetics.^13–15  An electrocorticogram power spectrum depicts the power distributions of individual frequency components. Similar analyses of electroencephalogram spectra have yielded systems-level insights into the phenomenon of general anesthesia.^16,17  Characteristic spectral patterns have been associated with different planes of anesthesia as well as different types of anesthetic agents.^13,14  While the spectral changes under various anesthetics have been explored, it is unclear how disruption of neuronal aerobic metabolism affects field oscillatory behavior detected with the electrocorticogram. Given that volatile anesthetics directly inhibit complex I^18–20  and that the Ndufs4(KO) decreases complex I activity in neurons by ~50%,¹²  we hypothesized that the volatile anesthetics would induce similar changes in the electrocorticogram spectra in control and knockout animals at equipotent concentrations. We studied the electrocorticogram patterns of knockout and control mice exposed to halothane and isoflurane.

Restricting the Ndufs4 mutation to glutamatergic neurons gives the full increase in sensitivity to volatile anesthetics.^11,21  Here we determined whether the resistance of Ndufs4(KO) to ketamine was also dependent on glutamatergic-specific knockout of the gene. Finally, we determined whether the electrocorticogram patterns in knockout and control mice were similar when anesthetized by ketamine. We hypothesized that ketamine would induce similar changes in the electrocorticogram spectra in control and knockout animals at equipotent doses.

The goal of these studies were to establish if the anesthetized state in the Ndufs4(KO) is similar to that of control mice. If differences exist between the two genotypes, they may help determine which aspects of the electrocorticogram spectra are important for anesthesia.

The study was carried out in strict accordance with the recommendations in the Guide of the Care and Use of Laboratory Animals of the National Institutes of Health (Bethesda, Maryland). All animal experiments were performed with the approval of the Institutional Animal Care and Use Committee of the Seattle Children’s Research Institute (Seattle, Washington). All surgeries were performed humanely with all efforts to minimize pain and suffering.

Mice were maintained on a standard rodent diet with a 12-h dark-light cycle at 22°C. Water and food were available ad libitum. Both male and female mice were used for all experiments. Mice heterozygous for Ndufs4 null allele (Ndufs4^-/+) in a C57BL/6 genetic background were crossed to produce wild-type (Ndufs4^+/+), heterozygous (Ndufs4^-/+), and knockout (Ndufs4^-/-, KO, or Ndufs4[KO]) animals. The offspring genotype was determined by polymerase chain reaction. Only heterozygous (Ndufs4^-/+) mice were used for the control group.

Cell-specific lines were generated and genotyped as previously described^11,21,22  and were compared to their siblings heterozygous for Ndufs4^lox and for the Cre-recombinase driver. No animals were excluded from analysis. The numbers of animals scored for each anesthetic and genotype are given in the legends. The offspring genotype was determined by polymerase chain reaction. Cre-recombinase expression was localized to central nervous system cell types as previously described.²³  Animals were genotyped using tail DNA and tested for absence of ectopic recombination at the end of the experiment using central nervous system DNA. Loss of righting reflex was determined as previously described.¹¹ 

Mice underwent survival surgery at ages postgestational days 28 to 32 to implant electrocorticogram and electromyogram electrodes under isoflurane anesthesia with additional subcutaneous bupivacaine for analgesia as described previously.²⁴  Using aseptic technique, a midline incision was made anterior to posterior to expose the cranium. Fine (diameter: 0.127 mm bare; 0.178 mm coated) silver wires were placed through cranial holes created with a fine cutting needle and fixed in place with cyano-acrylate adhesive. Electrocorticogram electrodes were placed at visually identified locations—left and right frontal cortices, approximately 1 mm anterior to the bregma and 3 mm lateral to the sagittal suture. Electromyogram electrodes were placed in back muscles. A reference electrode was placed at the midline cerebellum, a ground electrode was placed subcutaneously over the back, and the skin was closed with sutures. Electrode impedances were typically less than 10 kΩ. One milligram per kilogram of 0.25% bupivicaine in sterile saline was injected subcutaneously immediately after surgery. The mice were recovered for 1 h and then transferred to their cages. Carprofen MediGel (Clear H20, USA) was put in the animal cages for 48 h.

Twenty-four to 48 h after electrocorticogram electrodes placement, the mice were exposed to anesthetic while the electrocorticogram was monitored as previously described.^24,25  Briefly, all biopotential signals were acquired at a 1-KHz sampling rate and 100 × gain. Electrocorticogram signals were processed offline with a 1- to 70-Hz bandpass filter and a 60-Hz trap filter to reduce line noise if necessary. The electromyogram signals were processed offline with a 3-Hz high-pass filter. Electrocorticogram patterns were evaluated before recording to ensure complete recovery from anesthesia. A digital video/electrocorticogram/electromyogram recording system (LabChart, ADInstruments, Australia) was used to record electromyogram/electrocorticograms from freely behaving awake or anesthetized mice. All bioelectrical signals were acquired at a 1-KHz sampling rate and 100× gain. Electrocorticogram signals were processed offline with a 1- to 70-Hz bandpass filter and a 60-Hz trap filter to reduce line noise if necessary. Mice were exposed to stepwise increases in volatile anesthetic concentration (8 to 10 min in 0.2 to 0.3% increments) to reach their EC₅₀ for loss of response to tail clamp (0.4% for knockout and 1.2% for controls) as well as ~1.5% × EC₅₀ (0.6% for knockouts and 2% for controls). Sensitivity to the second volatile anesthetic was assessed 24 h later in an identical manner. Isoflurane and halothane were done in alternate order in sequential mice. For all anesthetic electrocorticograms (volatile anesthetic and ketamine), the baseline electrocorticogram for each anesthetic was determined by an unexposed electrocorticogram done immediately before the anesthetic exposure. Blinding of the observer to mouse genotype was not possible as the mutant and control mice are visibly distinct. Since electrocorticogram data were obtained independently of genotype, blinding was not necessary. Based on previous experience with the complete knockout, an N = 5 was felt to be sufficient to obtain significance for biologically relevant changes.

For ketamine studies, age-matched wild-type and Ndufs4(KO) mice underwent electrocorticogram implantation as per the volatile anesthetics. Twenty-four to 48 h after electrocorticogram electrodes placement, the mice were injected with intraperitoneal racemic ketamine (Ketaved, Vedco, USA) at their ED95s (determined previously as 100 mg/kg for wild-type, and 150 mg/kg for knockout¹¹ ). Animals underwent surgery at postgestational days 28 to 30 and had electrocorticogram recorded in awake and ketamine anesthetized states at their respective ED95 dose. Both controls and knockouts were euthanized after experiment completion of ketamine or volatile anesthetic experiments. As for volatile anesthetics, based on previous experience with the complete knockout, an N = 5 was felt to be sufficient to obtain significance.

After steady state exposure to equipotent doses of anesthetics, average power spectral analyses were performed by fast fourier transform, and results were compared between anesthetic and awake states. Each electrocorticogram tracing was inspected for signal quality and noise contamination in IGOR Pro 6 (WaveMetrics, Inc., USA). The ratio of averaged power spectral density in the anesthetized and awake animal was calculated for each animal at the appropriate anesthetic concentration, then pooled for analysis. For the graphs of power versus hertz, values were pooled in 1-Hz bins: 0.1 to 1 Hz represented at 1 Hz, 1 to 2 Hz at 2 Hz, and continuing in that pattern until 58 to 59 Hz represented at 59 Hz. Sixty-hertz values were omitted due to a high background signal.

Data are reported as mean ± SD. Comparisons of spectral ratio data across five electrocorticogram frequency ranges (delta 1 to 4 Hz, theta 4 to 8 Hz, alpha 8 to 13 Hz, beta 13 to 30 Hz, and gamma 31 to 59 Hz) were made after each group was tested with a two-way ANOVA. Since there were five groups to be tested with ANOVA (the five different frequency ranges, delta, theta, alpha, beta, and gamma), we applied a Bonferroni correction to a P value threshold for ANOVA statistical significance of 0.01, changing the threshold for an acceptable ANOVA significance to 0.01/4 = 0.0025. If the ANOVA significance reached that value, we tested the individual treatment groups in pairs comparing the anesthetic-treated groups to the untreated group of the same genotype. Comparisons of power in each frequency between the paired groups were done using a paired two-sided Student’s t test assuming unequal variance. Since in volatile anesthetic experiments there were four concentration groups for the controls and three for the knockouts, we again used a Bonferroni correction to our basic limit of P = 0.01. The correction was 0.01/3 (0.003) for the controls and 0.01/2 (0.005) for the knockouts. For the ketamine studies, there were only a treated and untreated group at each frequency range. We therefore used a Bonferroni correction for the number of frequency ranges to use a P < 0.01/4 = 0.0025 as a cutoff for significance for a two-way ANOVA. The only comparisons made for all tests, other than the baseline comparisons (fig. 2), were within a single genotype but at different anesthetic concentrations. While in all studies, we did not compare power densities between different genotypes, such a comparison was done at baseline to establish the similarities of the two genotypes to each other. Since that led to an implied comparison between the genotypes in all further studies, the use of a two-way ANOVA is the most conservative approach to determine significance within groups. For all experiments, N = 5 mice except the ketamine control, N = 7. Results of all ANOVA tests and t tests when indicated are shown in the Supplemental Digital Content (https://links.lww.com/ALN/B759); P values are the first number, with ANOVA tests in parentheses.

Dose response curves for ketamine were constructed as follows using matched controls for each cell-specific Ndufs4(KO) line. For each curve shown in figure 5, blue squares and blue solid lines reflect data points for the respective cell-specific Ndufs4(KO) strains. Similarly, red squares and red solid lines reflect the data points for control mice used for those specific experiments. Green and magenta smoothed lines are the best fit lines fit to the data points. The curve fits for the loss of righting reflex dose response curves were constructed as an interpolation of a standard sigmoidal curve using a four-parameter logistic equation of the form Y = Bottom + (Top – Bottom) / {1 + 10 ^ [(LogEC50 – X) * HillSlope]}. Magenta and green dotted lines reflect the 95% CIs for the interpolated fits.

During peer review, at the reviewer’s request, sample sizes were increased from N = 4 to N = 5 for the isoflurane and halothane exposure groups. However, no further adjustments were made for the repeated analysis of the enhanced sample sizes other than an altered Bonferroni correction due to the larger sample sizes.

There were no statistically significant differences in average power densities of individual frequency ranges or five frequency bands between the two genotypes at baseline (figs. 1 and 2). In the control mice (N = 5), the average power densities were 2.04 µV²/Hz in the delta frequency band, 1.17 µV²/Hz in the theta frequency band, 0.26 µV²/Hz the alpha frequency band, 0.68 µV²/Hz in the beta frequency band, and 0.020 µV²/Hz in the gamma frequency band. In Ndufs4(KO) mice (N = 5), the average power densities were 1.37 µV²/Hz (P = 0.08 compared with the control value) in the delta frequency band, 0.54 µV²/Hz (P = 0.11) in theta frequency band, 0.16 µV²/Hz (P = 0.06) in the alpha frequency band, 0.054 µV²/Hz (P = 0.14) in the beta frequency band, and 0.021 µV²/Hz (P = 0.73) in the gamma frequency band (fig. 2C). None of the differences between knockout and control reached the Bonferroni corrected P value limit of 0.0025 (Supplemental Digital Content, https://links.lww.com/ALN/B759).

Since the Ndufs4(KO) mice are anesthetized at 0.6% isoflurane and 0.6% halothane while controls are not, we first characterized the differences in their respective electrocorticograms at those concentrations of anesthetic. We then compared the electrocorticograms in the two genotypes at anesthetic concentrations normalized to their EC50s or 1.5 × EC50s. To make the comparisons between genotypes, for the knockout we used the EC50 as 0.4% (both isoflurane and halothane) and 1.5 × EC50 as 0.6% (for both volatile anesthetics). For the control we used the EC50 as 1.2% (both volatile anesthetics) and 1.5 × EC50 as 1.8% (for both volatile anesthetics).

In the control mice, the average power densities in all frequency bands at 0.6% isoflurane were not statistically significant from those of the unexposed animals (fig. 3, A and B; blue, black curves, Supplemental Digital Content, https://links.lww.com/ALN/B759). The average power densities decreased significantly at the EC50 for isoflurane (fig. 3, A and B; blue, gray curves, 1.2% isoflurane) when compared to baseline in delta (2.45 µV²/Hz and 0.50 µV²/Hz; P < 0.0001), theta (1.41 µV²/Hz and 0.16 µV²/Hz; P < 0.0001), alpha (0.23 µV²/Hz and 0.05 µV²/Hz; P < 0.0001), beta (0.066 µV²/Hz and 0.016 µV²/Hz; P < 0.0001), and gamma (0.020 µV²/Hz and 0.005 µV²/Hz; P < 0.0001) frequency bands (fig. 3, E and F; blue, green bars). At 1.5 × EC50 (fig. 3, A and B; blue, green curves; 1.8% isoflurane)_, the average power densities decreased further across all frequency bands when compared to baseline: delta (0.13 µV²/Hz; P < 0.0001), theta (0.073; P < 0.0001), alpha (0.021 µV²/Hz; P < 0.0001), beta (0.007 µV²/Hz; P < 0.0001), and gamma (0.004 µV²/Hz; P < 0 .0001); fig. 3, G and H; blue, green bars).

At the EC50 for Ndufs4(KO) in isoflurane (fig. 3, C and D; red, gray curves; 0.4% isoflurane), the average power densities were unchanged in the delta (1.08 µV²/Hz and 1.38 µV²/Hz; P = 0.64), theta (0.36 µV²/Hz and 0.26 µV²/Hz; P = 0.48), and alpha (0.09 µV²/Hz and 0.069 µV²/Hz; P = 0.14) frequency bands but decreased in the beta (0.041 µV²/Hz and 0.023 µV²/Hz; P < 0.0001) and gamma (0.020 µV²/Hz and 0.0068 µV²/Hz; P < 0.0001) frequency bands (fig. 3, E and F; red, purple bars). At 1.5 × EC50 (fig. 3, C and D; red, green curves; 0.6% isoflurane), the mutant had significant changes only in the beta (0.016 µV²/Hz; P < 0.0001) and gamma (0.0054 µV²/Hz; P < 0.0001) frequency bands compared to baseline (fig. 3, G and H; red, purple bars). Power in the delta, theta, and alpha ranges were not changed from baseline at either concentration of isoflurane.

In summary, comparing the patterns of changes in figure 3 (see also table S1, Supplemental Digital Content, https://links.lww.com/ALN/B759), at their respective EC50s and 1.5 × EC50s for isoflurane, the decrease in total power across all frequencies was greater in the control than in the mutant. The differences between genotypes were noted especially in the lower frequencies.

In the control mice, the average power densities in all frequency bands at 0.6% halothane were not statistically significant from those of the unexposed animals (fig. 4, A and B; blue, black curves, Supplemental Digital Content, https://links.lww.com/ALN/B759). At the EC50 for halothane (fig. 4, A and B; blue, gray curves, 1.2% halothane), the average power densities decreased significantly in all frequency bands: delta (1.62 µV²/Hz and 0.74 µV²/Hz; P = 0.0008), theta (0.93 µV²/Hz and 0.29 µV²/Hz; P < 0.0001), alpha (0.28 µV²/Hz and 0.075 µV²/Hz; P < 0.0001), beta (0.087 µV²/Hz and 0.027 µV²/Hz; P < 0.0001), and gamma (0.020 µV²/Hz and 0.009 µV²/Hz; P < 0.0001; fig. 4, E and F; blue, green bars). At 1.5 × EC50 (fig. 4, A and B; blue, green curves; 1.8% halothane), when compared to baseline, the average power densities decreased further across all frequency bands: delta (0.23 µV²/Hz; P < 0.0001), theta (0.066 µV²/Hz; P < 0.0001), alpha (0.020 µV²/Hz; P < 0.0001), beta (0.009 µV²/Hz; P < 0.0001), and gamma (0.0043 µV²/Hz; P < 0.0001; fig. 4, G and H; blue, green bars).

In Ndufs4(KO), at the EC50 for halothane (fig. 4, C and D; red, gray curves; 0.4% halothane), the average power densities were decreased significantly only in the alpha (0.18 µV²/Hz and 0.08 µV²/Hz; P = 0.002), beta (0.06 µV²/Hz and 0.03 µV²/Hz; P < 0.0001), and gamma (0.022 µV²/Hz and 0.014 µV²/Hz; P < 0.0001) frequency bands (fig. 4, E and F; red, purple bars). At 1.5 × EC₅₀ (fig. 4, C and D; red, green curves; 0.6% halothane), when compared to baseline, there was a significant change in the alpha (0.062 µV²/Hz; P = 0.002), beta (0.024 µV²/Hz; P < 0.0001), and gamma (0.006 µV²/Hz; P < 0.0001) frequency bands (fig. 4, G and H; red, purple bars).

In summary, at their respective EC50s for halothane, both control and knockout had similar decreases in power across the higher frequency bands. At their respective 1.5 × EC50s for halothane, the decrease in total power across all frequencies was greater in the control than in the mutant as seen in figure 4 (see also table S1, Supplemental Digital Content, https://links.lww.com/ALN/B759). As with isoflurane, the differences between genotypes were noted, especially in the lower frequencies.

Previous studies showed that the volatile anesthetic hypersensitivity of Ndufs4(KO) was dependent on cell-specific loss of the gene in glutamatergic neurons.²¹  We first determined if ketamine sensitivity also was dependent on cell-specific Ndufs4(KO). Using loss of the righting reflex as the endpoint, we compared the sensitivity to ketamine of control mice with that of mice with Ndufs4 knocked out selectively in γ-aminobutyric acid (GABA)–mediated neurons (GABA-[KO]), vesicular glutamate transporter 2 (VGLUT2)-positive glutamatergic neurons (VGLUT2-[KO]), or cholinergic neurons (CHAT-[KO]). VGLUT2 knockout mice (ED50 125 ± 2 mg/kg) were markedly resistant to ketamine compared to control mice (ED50 75 ± 1.5 mg/kg; P < 0.001), similar to the total knockout mice (ED50 100 ± 2 mg/kg). GABA-mediated–specific Ndufs4(KO) (ED50 70 ± 5 mg/kg; P = 0.86) and cholinergic-specific Ndufs4(KO) mice (ED50 90 ± 4 mg/kg; P = 0.76) were not resistant to ketamine compared to the control mice (fig. 5). All mice appeared behaviorally normal until approaching the doses needed for loss of the righting reflex.

Similar to our reasoning for the volatile anesthetics, we then compared the electrocorticograms for wild-type and global knockout mice at their respective ED95s for ketamine. Mice were injected with intraperitoneal racemic ketamine (Ketaved) at their ED95s (100 mg/kg for wild-type, and 150 mg/kg for Ndufs4(KO)¹¹).

In the control mice, the average power densities at the ketamine ED95 (100 mg/kg; fig. 6, A and B; blue, red curves) were not statistically significant from those of the unexposed animals in the delta (2.40 µV²/Hz and 2.56 µV²/Hz; P = 0.89), theta (0.47 µV²/Hz and 0.44 µV²/Hz; P = 0.91), alpha (0.11 µV²/Hz and 0.07 µV²/Hz; P = 0.05), beta (0.031 µV²/Hz and 0.016 µV²/Hz; P = 0.06), and gamma (0.023 µV²/Hz and 0.013 µV²/Hz; P = 0.075) frequency bands (fig. 6, E and F; blue, green bars).

In Ndufs4(KO), the average power densities at the ketamine ED95 (150 mg/kg; fig. 6, C and D; blue, black curves) were not statistically significant from those of the unexposed animals in the delta (1.25 µV²/Hz and 1.78 µV²/Hz; P = 0.19), theta (0.39 µV²/Hz and 0.36 µV²/Hz; P = 0.89), and alpha ranges (0.12 µV²/Hz and 0.09 µV²/Hz; P = 0.37) (fig. 6, E and F; red, purple bars). The average power densities were decreased significantly from baseline only in the beta (0.040 µV²/Hz and 0.020 µV²/Hz; P < 0.0001) and gamma (0.035 µV²/Hz and 0.015 µV²/Hz; P < 0.0001) frequency bands (fig. 6, E and F; red, purple bars).

In summary, at their respective ED95s, control and knockout had similar responses in the delta, theta, and alpha frequency ranges but differed in the upper frequencies (beta and gamma ranges). Ketamine resistance tracks with loss of NDUFS4 in glutamatergic neurons.

These data demonstrate significantly different neurophysiologic signatures of equipotent doses of volatile anesthetics between wild-type and Ndufs4(KO) mice, an animal with known mitochondrial dysfunction. The differences are discussed in the section Volatile Anesthetics for each type of anesthetic.

We hypothesized that when the concentrations of volatile anesthetics reach their respective EC50s, both mutant and control electrocorticograms would be similarly affected in frequency ranges necessary for maintenance of consciousness. At their EC₅₀s for each volatile anesthetic, control animals had significantly lower power densities across all frequency bands than they did at baseline. These broad decreases were more pronounced at 1.5 × EC50 for each volatile anesthetic. In contrast, the mutant animals maintained their power densities close to baseline values in the low power frequency range (delta, theta) in isoflurane at both the EC₅₀ and 1.5 × EC50. In halothane, the control again had broad decreases in power in all frequency ranges at the EC50. In the knockout, the decreases only reached significance in the alpha, beta, and gamma frequency ranges. In addition, only the control mice had further decreases as the concentrations were increased to 1.5 × EC50. Therefore, there are fundamental differences in the response of the two genotypes to concentrations of anesthetic at doses matched for potency.

Since mutant mice have very low EC50s for both volatile anesthetics,¹¹  we hypothesized that we would see changes in the electrocorticograms of mutant mice at lower concentrations of volatile anesthetics than in control mice. These changes would be interpreted as the marker of the anesthetized state in the mutant but not the control. At 0.6% isoflurane or halothane concentrations, no significant differences were noted in any frequency ranges in the electrocorticograms from control mice. In the mutant, power in alpha, beta, and gamma frequency ranges was decreased at both 0.6% isoflurane and halothane. However, power in delta and theta ranges was maintained in both anesthetics despite animals being anesthetized. In humans, others have noted that, especially in anterior regions of the brain, power in the delta frequencies is maintained or even increased during onset of volatile anesthetic action.^26–28  While we did not do an exhaustive characterization of the sensitivity of the lower frequencies to anesthetic concentrations and we only placed single leads bilaterally, our results are in general agreement with those of previous studies. In fact, the knockout maintained power in the lower frequencies even at anesthetizing concentrations. Losses of power in the lower frequencies are not necessary for the anesthetized state. These results are in agreement with previous studies that arousal has been correlated with increased power in higher frequency ranges.^29–31 

Previous studies in humans have shown that in volatile anesthetics at or near EC₅₀s, power is shifted from the higher frequencies (beta and gamma) to the lower ranges (delta, theta, and alpha).^14,32  We did not see such a shift; rather, we saw a broad decrease in power at all frequencies. However, the loss was more significant in the upper ranges such that a spectral presentation may have shown increased relative power in the theta or alpha ranges. It has also been noted that as consciousness is lost, there is a loss of functional connectivity between different regions of the brain.³³  With our experimental setup, we were unable to determine whether such connectivity was affected.

The pattern of changes seen with volatile anesthetics in humans is similar to that of propofol and has been interpreted to indicate that the predominant effect of volatile agents is at the GABA receptor. We would caution that these similarities may indicate that similar neuronal networks are affected but not that identical or even similar molecular targets are involved. Our results are consistent with the possibility of a primary mitochondrial target for volatile anesthetics since both mutant and control lose power in the upper ranges despite being anesthetized at widely disparate concentrations; however, these results do not rule out other potential molecular targets for volatile agents.

Similar to previous responses with volatile anesthetics, loss of Ndufs4 in glutamatergic neurons is sufficient to cause resistance to ketamine. During sedation with ketamine, increases in gamma and beta frequency power spectra have been identified previously and thought to be due to anti-NMDA–mediated disinhibition of pyramidal neurons^15,34  as well as blocking fast-spiking cortical interneurons.³⁵  In this manner, the reported electroencephalogram patterns after low-dose ketamine are distinct from those seen with anesthetizing doses of volatile anesthetics and GABA type A receptor agonists. Electrocorticogram or electroencephalogram patterns during higher dosing with ketamine causing the anesthetized state have not been well described. We measured no significant changes in the electrocorticograms of control mice when comparing baseline to anesthetizing doses of ketamine. However, the mutant electrocorticogram decreased in power in the beta and gamma frequency ranges. Therefore, maintenance of power in the upper frequencies may not be necessary for the anesthetized state since the mutant is anesthetized in spite of a decrease in beta and gamma frequency oscillatory behavior.

Ketamine is associated with increased cerebral glucose utilization at subanesthetic doses³⁶  and a decrease in inhibitory tone with decreased GABA release.³⁷  The relative resistance of the mutant to ketamine may result from the decrease in power in the higher frequencies resulting from circuit disruptions in cortical and subcortical sites.³⁸  Although untested in this study, is it possible that the intercortical connectivity that produces ketamine anesthesia is not possible when pyramidal neurons in the mitochondrial mice are unable to mount a sufficiently excitatory response. However, it is important to note that there is no evidence that ketamine is an effective inhibitor of mitochondrial function. Therefore, the effect of Ndufs4(KO) may be an indirect one on an energetic state necessary for ketamine-induced sedation rather than on a ketamine target such as the NMDA receptor or potassium/sodium hyperpolarization-activated cyclic nucleotide-gated channel 1 channels.

In this study, we identified the characteristic changes in electrocorticogram patterns induced by two volatile anesthetics (isoflurane and halothane) as well as ketamine. Depression of electrocorticogram power is relatively spared in the lower frequencies for the mutant compared to the control at equipotent concentrations of volatile anesthetics. The decreases in power in the higher frequencies in the mutant at low concentrations of volatile anesthetics indicates that these changes may be sufficient to cause the anesthetized state. In contrast, when exposed to equipotent doses of ketamine, electrocorticogram power in the higher frequencies is maintained in controls but decreases in the mutant. These data suggest that increased power in the higher frequencies is not necessary for the anesthetizing effects of ketamine.

The authors thank Beatrice Predoi, M.D., Center for Integrative Brain Research, Seattle Children’s Research Institute, Seattle, Washington, for assistance with mouse care and genotyping.

Supported by the Department of Anesthesiology and Pain Medicine Bonica Scholar Program, University of Washington (Seattle, Washington; to Dr. Carspecken); the Northwest Mitochondrial Research Guild (Seattle, Washington; to Dr. Chanprasert); National Institutes of Health (Baltimore, Maryland) grant No. R01GM105696 and the Northwest Mitochondrial Research Guild (to Drs. Sedensky and Morgan); and National Institutes of Health/National Institute of Neurological Disorders and Stroke (Bethesda, Maryland) grant No. R01NS102796, the Citizens United for Research in Epilepsy (Chicago, Illinois) Epilepsy Research Grant, and Ellenbogen Chair, University of Washington Neurosurgery Research Funds (to Dr. Kalume).

The authors declare no competing interests.