In mice, restriction of loss of the mitochondrial complex I gene Ndufs4 to glutamatergic neurons confers a profound hypersensitivity to volatile anesthetics.
Astrocytes are crucial to glutamatergic synapse functioning during excitatory transmission.
In a tamoxifen-activated astrocyte-specific Ndufs4(KO) mouse, the induction EC50s for tail clamp in both isoflurane and halothane were similar between the control and astrocyte-specific Ndufs4(KO) mice at 3 weeks after 4-hydroxy tamoxifen injection. However, the emergent concentrations in both anesthetics for the astrocyte-specific Ndufs4(KO) mice were half that of the controls.
Similarly, the induction EC50s for loss of righting reflex were similar between the control and astrocyte-specific Ndufs4(KO) mice; concentrations for regain of righting reflex in both anesthetics for the astrocyte-specific Ndufs4(KO) mice were much less than the control.
Thus, mitochondrial complex I function within astrocytes is essential for normal emergence from anesthesia.
In mice, restriction of loss of the mitochondrial complex I gene Ndufs4 to glutamatergic neurons confers a profound hypersensitivity to volatile anesthetics similar to that seen with global genetic knockout of Ndufs4. Astrocytes are crucial to glutamatergic synapse functioning during excitatory transmission. Therefore, the authors examined the role of astrocytes in the anesthetic hypersensitivity of Ndufs4(KO).
A tamoxifen-activated astrocyte-specific Ndufs4(KO) mouse was constructed. The specificity of the astrocyte-specific inducible model was confirmed by using the green fluorescent protein reporter line Ai6. Approximately 120 astrocyte-specific knockout and control mice were used for the experiments. Mice were anesthetized with varying concentrations of isoflurane or halothane; loss of righting reflex and response to a tail clamp were determined and quantified as the induction and emergence EC50s. Because norepinephrine has been implicated in emergence from anesthesia and astrocytes respond to norepinephrine to release gliotransmitters, the authors measured norepinephrine levels in the brains of control and knockout Ndufs4 animals.
The induction EC50s for tail clamp in both isoflurane and halothane were similar between the control and astrocyte-specific Ndufs4(KO) mice at 3 weeks after 4-hydroxy tamoxifen injection (induction concentration, EC50(ind)—isoflurane: control = 1.27 ± 0.12, astrocyte-specific knockout = 1.21 ± 0.18, P = 0.495; halothane: control = 1.28 ± 0.05, astrocyte-specific knockout = 1.20 ± 0.05, P = 0.017). However, the emergent concentrations in both anesthetics for the astrocyte-specific Ndufs4(KO) mice were less than the controls for tail clamp; (emergence concentration, EC50(em)—isoflurane: control = 1.18 ± 0.10, astrocyte-specific knockout = 0.67 ± 0.11, P < 0.0001; halothane: control = 1.08 ± 0.09, astrocyte-specific knockout = 0.59 ± 0.12, P < 0.0001). The induction EC50s for loss of righting reflex were also similar between the control and astrocyte-specific Ndufs4(KO) mice (EC50(ind)—isoflurane: control = 1.02 ± 0.10, astrocyte-specific knockout = 0.97 ± 0.06, P = 0.264; halothane: control = 1.03 ± 0.05, astrocyte-specific knockout = 0.99 ± 0.08, P = 0.207). The emergent concentrations for loss of righting reflex in both anesthetics for the astrocyte-specific Ndufs4(KO) mice were less than the control (EC50(em)—isoflurane: control = 1.0 ± 0.07, astrocyte-specific knockout = 0.62 ± 0.12, P < 0.0001; halothane: control = 1.0 ± 0.04, astrocyte-specific KO = 0.64 ± 0.09, P < 0.0001); N ≥ 6 for control and astrocyte-specific Ndufs4(KO) mice. For all tests, similar results were seen at 7 weeks after 4-hydroxy tamoxifen injection. The total norepinephrine content of the brain in global or astrocyte-specific Ndufs4(KO) mice was unchanged compared to control mice.
The only phenotype of the astrocyte-specific Ndufs4(KO) mouse was a specific impairment in emergence from volatile anesthetic-induced general anesthesia. The authors conclude that normal mitochondrial function within astrocytes is essential for emergence from anesthesia.
The mechanism and identities of cells in the central nervous system that contribute to the anesthetic response are not well elucidated. Inhibition of mitochondrial complex I function has been proposed as a possible molecular mechanism of action of volatile anesthetics.1,2 Both the Caenorhabditis elegans mutant gas-1 and the Drosophila mutant ND23, each defective in a single distinct mitochondrial complex I subunit, display increased sensitivity to volatile anesthetics.1,3 In addition, clinical studies show that some children with complex I defects are hypersensitive to sevoflurane.4 Similarly, a mouse model of complex I dysfunction, Ndufs4(KO), is hypersensitive to volatile anesthetics.5 Furthermore, the function of complex I is inhibited by volatile anesthetics at concentrations that match the whole animal EC50s of normal and mutant mice.1,2 Thus, mitochondrial complex I function profoundly affects anesthetic sensitivity across the animal kingdom.
In the mouse, restriction of Ndufs4 loss to glutamatergic (VGLUT2-expressing) neurons conferred the same profound hypersensitivity to volatile anesthetics as seen with global loss of the protein, with an EC50 one third that of wild-type mice.6 This finding implicates a role for glutamatergic synaptic transmission in mediating volatile anesthetic hypersensitivity. Glutamatergic synapses have been shown to consist of three cells: a presynaptic neuron, a postsynaptic neuron, and a supporting astrocyte.7 The configuration is termed the tripartite synapse,7 where the astrocytic roles include both glutamate reuptake8,9 and modulation of synaptic transmission.10 We therefore questioned whether astrocytic function contributes to the change in anesthetic sensitivity in Ndufs4(KO).
In this study, we have investigated the role of astrocytes in mediating the effect of volatile anesthetics by constructing a conditional loss of Ndufs4 in astrocytes. Because this animal loses the gene acutely during adulthood and only in astrocytes, no compensatory changes are expected during development, leading to confounding phenotypes. The resulting astrocyte-specific Ndufs4(KO) animal was tested for responses to isoflurane and halothane using two different anesthetic endpoints: loss of righting reflex and response to a tail clamp. We hypothesized that the astrocyte-specific Ndufs4(KO) would be hypersensitive to the volatile anesthetics when compared to control mice.
Materials and Methods
Generation of Astrocyte-specific Ndufs4(KO) Mice
All studies were approved by the Institutional Animal Care and Use Committee of the Seattle Children’s Research Institute. Mice with a conditional exon 2-floxed allele of the Ndufs4 gene (Ndufs4lox/lox) were a kind gift from the Palmiter laboratory at the University of Washington. The 4-hydroxy tamoxifen–inducible Cre-recombinase expressing line Pgfap (glial fibrillary acidic protein)–CreERT2 was purchased from the Jackson Laboratory (USA; Jax stock no. 012849). Pgfap-CreERT2 mice were crossed to Ndufs4Δ/+ mice following the breeding scheme in figure 1A. Offspring that were Pgfap-CreERT2/+ and heterozygous for the Ndufs4 deletion (Δ/+) were selected and crossed to Ndufs4 mice floxed at exon 2 to create Pgfap-CreERT2/+;Ndufs4Δ/lox (conditional knockouts of Ndufs4) and Pgfap-CreERT2/+;Ndufs4+/lox mice (controls). The Pgfap promoter–driven Cre-ERT2 fusion enzyme requires induction by 4-hydroxy tamoxifen for nuclear translocation and function.11 4-Hydroxy tamoxifen (Sigma, USA) was diluted to a final concentration of 10 µg/µl in autoclaved, filtered sunflower seed oil and stored at −20°C until administration. The inducible knockout and control mice were injected intraperitoneally with 4-hydroxy tamoxifen at a dose of 50 µg/g bodyweight daily for a week beginning at postnatal day 33 and tested for anesthetic behavior in isoflurane and halothane at 3 and 7 weeks after injections. Two mice of approximately 120 injected mice did not survive the injection regime and were excluded from the analyses.
To demonstrate 4-hydroxy tamoxifen–induced exon excision by the Cre-recombinase, Pgfap-CreERT2 mice were crossed to Ndufs4lox/lox mice following the breeding scheme in figure 1B. Pgfap-CreERT2/+;Ndufs4lox/+ offspring were selected from the offspring and crossed to Ndufs4lox/lox to generate Pgfap-CreERT2/+;Ndufs4lox/lox mice. After intraperitoneal injection with 4-hydroxy tamoxifen, the deletion was confirmed by polymerase chain reaction.
Generation of Astrocyte-specific Ai6 Reporter Mice
The inducible Cre-recombinase expressing line, Pgfap-CreERT2/+, was crossed to the Ai6 (ai6/ai6) reporter line,12 which produces ZsGreen, a green fluorescent protein only in Cre-recombinase expressing cells. When injected with 4-hydroxy tamoxifen, the astrocytes of the Pgfap-CreERT2/+;ai6/+ progeny express Cre-recombinase and ZsGreen. The efficiency of Cre-recombinase induction and recombination by 4-hydroxy tamoxifen was assessed by immunohistochemistry (see the Immunohistochemical Analysis section below).
The mice (six or seven for each experiment; the exact number is specified in the figure legends) were anesthetized with isoflurane or halothane and assayed for loss of righting reflex or response to a tail clamp as described by Sonner et al.13 The volatile anesthetics were delivered by inline gas lines to the mouse chamber. Step sizes were 0.1% for both volatile anesthetics in loss of righting reflex, 0.2% in tail clamp. Equilibration time was 10, 20, or 30 min at each step, for separate sets of mice (data shown for 20-min interval experiments only). For induction and emergence assays (done consecutively), the same step sizes and equilibration times were maintained. Anesthetic concentrations were analyzed by gas chromatography as described.14 The animals were kept warm on a heating pad throughout and allowed to recover for at least 24 h before testing again, with a different anesthetic or for a different endpoint. The observers were blinded to the genotype of the tested mice. The concentration for induction of anesthesia was defined as the average concentrations of gas in the last sample before loss of response and in the first sample at which response was lost. The concentration for emergence was defined as the average concentration of gas in the first sample at which the animal once again responded and the sample immediately prior. All EC50s for induction and emergence were calculated from the quantal endpoints, i.e. the averages of the induction or emergent concentrations for all the mice in a specific anesthetic. The mice were tested at 3 and 7 weeks after injection of 4-hydroxy tamoxifen and then sacrificed. All animals survived the anesthetic exposure.
The mice were euthanized using CO2, and isolated brains were fixed in 4% paraformaldehyde overnight at 4°C, cryoprotected in 30% sucrose in phosphate-buffered saline for 3 days, and embedded in optimal cutting temperature compound (Tissue-tek, Sakura, USA). The brains were then sliced at 10-µm thickness and mounted on slides. For heat-induced epitope retrieval, the slides were then boiled in sodium citrate buffer (pH = 6.0) at 100°C for 20 min using a water bath. After this treatment, Cre-recombinase–mediated ZsGreen fluorescence was quenched but recaptured with anti-green fluorescent protein antibodies. The slides were then blocked in 10% normal donkey serum in phosphate-buffered saline for 1 h at room temperature. The primary antibodies used were mouse anti-glial fibrillary acidic protein (1:100, Chemicon, USA) and goat anti-green fluorescent protein (1:500, Abcam, USA) incubated overnight. The secondary antibodies used were donkey anti-mouse IgG–Alexa Fluor 568 (1:1,000, Abcam) and donkey anti-goat IgG–Alexa Fluor 488 (1:2,000, Life Technologies, Inc., USA) and were incubated for 1 hr at room temperature.
Brains of 40-day-old global Ndufs4(KO) and wild-type mice along with those of 60-day-old astrocyte-specific knockout mice (3 weeks after 4-hydroxy tamoxifen injection) were snap-frozen in dry ice and stored at −80°C. The 60-day old astrocyte-specific knockout animals were used to allow for complete loss of the NDUFS4 protein after 4-hydroxy tamoxifen injection at postnatal day 33.15 The brains were thawed, homogenized in PBS-containing protease inhibitor cocktail (Sigma), sonicated on ice, and centrifuged. After estimating the protein content of the supernatant, 20 µg each of the total protein extract was subjected to norepinephrine extraction and measurement following the manufacturer’s protocol (mouse/rat norepinephrine assay kit, catalogue no. NOU39-K010, Eagle Biosciences, USA).
The effective concentration for 50% of the animals tested (EC50) for volatile anesthetics was determined as described by Sonner et al.,13 using an up-and-down method. Values for EC50s were compared between the wild type and knockout strains using two-tailed two-sample t tests with unequal variance. No statistical power calculation was conducted before the study. The sample sizes were based on the sample sizes necessary to establish significance in previous studies with global and glutamatergic-specific knockouts of Ndufs4.6 For the norepinephrine assay, the norepinephrine concentrations were compared between the wild type, global Ndufs4(KO), and astrocyte-specific Ndufs4(KO) brain samples using one-way ANOVA and assessed by constructing a bar plot. We set the P-value threshold at P < 0.01 to lessen the likelihood of reporting a false-positive result. Because we compared multiple groups, a Bonferroni correction was performed to determine significance of the P values. Significance was defined as a P < 0.01 for all analyses except where subjected to a Bonferroni correction where the critical P < 0.01 was divided by the number of comparisons to derive the corrected cutoff. The measures of variability assessed by SD are indicated in the figures and legends. Outliers, if any, were always included in the analyses. All statistical analyses were performed using R version 3.4.1 and plotted using ggplot2.
Characterization of the Inducible Pgfap-CreERT2 Model
We constructed the inducible astrocyte-specific Ndufs4(KO) mouse model, with the Cre-recombinase expressed under the control of an astrocyte-specific promoter Pgfap (fig. 1A). Activation of Cre-recombinase was induced by 4-hydroxy tamoxifen injections at postnatal days 33 to 40, after gfap expression had ended in neuronal progenitors, thereby temporally limiting the expression and nuclear localization of Cre-recombinase to astrocytic cells. Knockout of Ndufs4 in the brain cells of 4-hydroxy tamoxifen–injected Ndufs4lox/lox mice was confirmed by genotyping the mice brains with and without 4-hydroxy tamoxifen injections (fig. 1B). At 3 and 10 days after injection, we found that Cre-recombinase–mediated excision of Ndufs4 was evident in the brains of the 4-hydroxy tamoxifen–injected Pgfap-CreERT2/+;Ndufs4lox/lox mice, whereas it was absent in the brains of sham-injected Pgfap-CreERT2/+;Ndufs4lox/lox mice (fig. 1B). The lox and knockout (Δ) allelic polymerase chain reaction amplification bands in the gel from the astrocyte-specific Ndufs4(KO) were excised and sequenced to confirm Cre-recombinase–mediated deletion of exon 2. This result confirmed that 4-hydroxy tamoxifen injections resulted in knockout of Ndufs4 in the brain cells. The astrocyte-specific KO mouse was similar to the controls in terms of life span, appearance, grooming characteristics, and growth pattern. Unlike the global Ndufs4(KO), the astrocyte-specific KO mice did not become progressively ataxic, lose weight, or die young.
The NDUFS4 antibody did not resolve astrocyte-specific loss of the protein in a background in which all other cells in the central nervous system (CNS) were fully expressing it. Therefore, to determine the specificity of the 4-hydroxy tamoxifen–induced Cre-lox system, we crossed the 4-hydroxy tamoxifen–sensitive Pgfap-CreERT2 driver line with the ai6/ai6 reporter mice,12 which harbors a Cre-inducible green fluorescent protein (ZsGreen) expression cassette. After genotyping the progeny, Pgfap-CreERT2/+;ai6/+ mice and control ai6/+ sibling mice were injected daily with 4-hydroxy tamoxifen for a week from postnatal days 33 to 40. When injected with 4-hydroxy tamoxifen, the glial fibrillary acidic protein expressing astrocytes of Pgfap-CreERT2/+;ai6/+ mice should express induced ZsGreen. Three weeks after the last day of injections, the brains were harvested and immunostained (fig. 2; supplemental fig. 1, http://links.lww.com/ALN/B820). Specific expression of Cre-recombinase within astrocytes of the Cre/+;ai6/+ mice was confirmed by the yellow appearance of the colocalized glial fibrillary acidic protein (labeled red) and ZsGreen (labeled green) under confocal microscopy (fig. 2). The control ai6/+ brain slices did not coexpress ZsGreen, and hence the astrocytes appeared red.
The half-life for the NDUFS4 protein in mouse brain has been shown to be 17 days; its half-life in astrocytes is not known.15 To allow for possible tissue-specific differences in protein half-life, we tested the mice at 3 and 7 weeks after knocking out the gene; the phenotype remained stable over time. We subjected the mice to 10-, 20-, and 30-min intervals of anesthetic exposure at each step to rule out a pharmacokinetic effect. All data are shown for 20 min of exposure of mice to each concentration of the volatile anesthetic. The astrocyte-specific Ndufs4(KO) (Pgfap-creERT2/+;Ndufs4Δ/lox) and control sibling mice were first tested for tail clamp response at 3 and 7 weeks after injection of 4-hydroxy tamoxifen, with increasing concentrations of isoflurane or halothane. The induction concentrations for the loss of tail flick were not different between the astrocyte-specific Ndufs4(KO) and control animals in both isoflurane and halothane (induction concentration, EC50(ind)—isoflurane: control = 1.27 ± 0.12, astrocyte-specific Ndufs4(KO) = 1.21 ± 0.18, P = 0.495; halothane: control = 1.28 ± 0.05, astrocyte-specific Ndufs4(KO) = 1.20 ± 0.05, P = 0.017 at 3 weeks; isoflurane: control =1.24 ± 0.07, astrocyte-specific Ndufs4(KO) = 1.18 ± 0.13, P = 0.305; halothane: control = 1.19 ± 0.04, astrocyte-specific Ndufs4(KO) = 1.22 ± 0.07, P = 0.396 at 7 weeks; table 1). Control animals emerged from the anesthetized states at concentrations similar to their induction concentrations. However, the conditional astrocyte-specific Ndufs4(KO)s emerged from the anesthetic state at significantly lower concentrations than did the controls (emergence concentration, EC50(em)—isoflurane: control =1.18 ± 0.10, astrocyte-specific Ndufs4(KO) = 0.67 ± 0.11, P < 0.0001; halothane: control = 1.08 ± 0.09, astrocyte-specific Ndufs4(KO) = 0.59 ± 0.12, P < 0.0001 at 3 weeks; isoflurane: control = 1.20 ± 0.05, astrocyte-specific Ndufs4(KO) = 0.67 ± 0.10, P < 0.0001; halothane: control = 1.15 ± 0.10, astrocyte-specific Ndufs4(KO) = 0.66 ± 0.10, P < 0.0001 at 7 weeks; table 1; fig. 3; supplemental fig. 2, http://links.lww.com/ALN/B821). Although the difference between the induction/emergence concentrations was not as large as seen with the astrocyte-specific Ndufs4(KO), there was a statistically significant change between the induction versus emergence in control animals in halothane, with 1.28% versus 1.08%, respectively, at 3 weeks (P = 5.695 × 10−4). The difference was not seen at 7 weeks (1.19% vs. 1.15%, P = 0.277).
Loss of Righting Reflex.
To determine whether the altered concentration for emergence was specific to a behavioral endpoint and possibly a CNS region, we tested a second endpoint, loss of righting reflex. The anesthetic concentration requirement for loss of righting reflex (induction) was similar between the control and astrocyte-specific KO mice in isoflurane and halothane (induction concentration, EC50(ind)—isoflurane: control = 1.02 ± 0.10, astrocyte-specific Ndufs4(KO) = 0.97 ± 0.06, P = 0.264; halothane: control = 1.03 ± 0.05, astrocyte-specific Ndufs4(KO) = 0.99 ± 0.08, P = 0.207 at 3 weeks; isoflurane: control = 0.92 ± 0.07, astrocyte-specific Ndufs4(KO) = 0.98 ± 0.05, P = 0.128; halothane: control = 0.98 ± 0.08, astrocyte-specific Ndufs4(KO) = 0.93 ± 0.13, P = 0.410 at 7 weeks; table 2). The control animals regained the righting reflex (emergence) at close to the induction concentration, whereas the astrocyte-specific Ndufs4(KO)s recovered their righting reflex at significantly lower concentrations than the controls (emergence concentration, EC50(em)—isoflurane: control = 1.0 ± 0.07, astrocyte-specific Ndufs4(KO) = 0.62 ± 0.12, P < 0.0001; halothane: control = 1.0 ± 0.04, astrocyte-specific Ndufs4(KO) = 0.64 ± 0.09, P < 0.0001 at 3 weeks; isoflurane: control = 0.89 ± 0.06, astrocyte-specific Ndufs4(KO) = 0.68 ± 0.08, P = 0.0005; halothane: control = 0.94 ± 0.04, astrocyte-specific Ndufs4(KO) = 0.55 ± 0.07, P < 0.0001 at 7 weeks; table 2; fig. 4; supplemental fig. 2, http://links.lww.com/ALN/B821). The differences between induction and emergence seen for both loss of righting reflex and tail clamp were similar for halothane and isoflurane and were not altered by prolonged exposure times (30 min) at each concentration of the anesthetic.
Because norepinephrine is known to activate cortical astrocytes,16 we tested whether the levels of norepinephrine in the global Ndufs4(KO) are reduced when compared with the astrocyte-specific Ndufs4(KO) or control mice. There was no significant difference in the total norepinephrine content in whole brain extracts between the global or astrocyte-specific Ndufs4 KOs, compared with wild-type mice (fig. 5).
Astrocytes affect synaptic transmission by multiple mechanisms. During synaptic transmission, glutamate is removed from the synaptic cleft by astrocytes, actively converted to glutamine by the astrocyte-specific glutamine synthetase, and supplied back to neurons through glial transporters SN1 and SN2.8,9 In addition, synaptic glutamate release, as well as neuromodulators like norepinephrine, evoke a calcium spike in astrocytes that can be transmitted between adjacent neurons and astrocytes.17,18 This calcium signaling causes the release of neural modulators, called gliotransmitters, from astrocytes into the synapse. Gliotransmitters include adenosine triphosphate (ATP), GABA, glutamate, and d-serine, all of which modulate local neuronal function and synaptic transmission10,19,20 at the tripartite synapse (fig. 6). It is surprising that the acute loss of NDUFS4 in astrocytes specifically lowers the concentrations at which the mutant emerges from anesthesia, without altering the induction concentrations, for both responses to tail clamp and loss of righting reflex assays. We interpret the data to indicate that the mutant mitochondria in astrocytes are inhibited by volatile anesthetics at lower concentrations than are control mitochondria. This inhibition results in decreased release of excitatory gliotransmitters and/or inhibited glutamate metabolism (fig. 6). Defective gliotransmitter release or glutamate metabolism in mutant astrocytes results in a failure of reactivation of neuronal function until volatile anesthetic concentrations are lowered sufficiently for astrocytic mitochondrial function to recover. Two important caveats must be considered: (1) we have not tested the effects of nonastrocytic nonneuronal cells, which may play a role; and (2) although we saw no evidence of compensatory changes in this acute knockout model, the existence of such changes cannot be completely ruled out.
The response to tail clamp during volatile anesthetic exposure in rodents is mediated by the spinal cord with supraspinal modification.21,22 Previous studies have implicated the spinal cord in mediating nociception and the tail clamp reflex.21,23 Although the spinal cord astrocytes have been implicated in modulating nociception,24 their role has been mostly studied in relation to chronic neuropathic pain and allodynia. Inhibition of the astrocytic glutamate–glutamine shuttle attenuates nociceptive neuronal responsiveness in response to inflammatory and nociceptive stimuli in the medulla.25 Alterations of these pathways caused by mitochondrial dysfunction may play a part in causing the hysteresis that we observed.
In contrast with tail clamp, the loss of righting reflex is likely mediated by interactions between the brainstem26 and thalamocortical pathways.27–29 These two behavioral endpoints, primarily determined by different regions of the CNS, model different human anesthetic responses (minimal alveolar concentration vs. loss of consciousness). Because the significant difference between induction and emergence concentrations (increased anesthetic hysteresis) was found with both of the behavioral endpoints, the emergence mechanism involved is unlikely to be dependent on one specific neuronal pathway. Instead, an astrocyte-specific mechanism is necessary for emergence with both endpoints. In addition, the structurally different anesthetics isoflurane and halothane showed that similar reductions in the concentration needed to be attained before the animals could emerge from anesthesia, suggesting that dependence of emergence on astrocytes may be a general feature of volatile anesthetic response.
Control animals and global Ndufs4(KO) do not display marked hysteresis presumably because both neurons and astrocytes have equal mitochondrial functions in each genotype. Mice with Ndufs4 loss restricted to glutamatergic neurons do not display overt hysteresis either, potentially because the concentration necessary for induction is so low as to obscure the role of astrocytes in emergence. The loss of NDUFS4 solely in astrocytes reveals the critical role for this cell type in restoring consciousness.
Although the response to the tail clamp was abrupt and often strong, the return to normal activity was qualitatively different in the astrocyte-specific Ndufs4(KO). Unlike the control mice, most astrocyte-specific Ndufs4(KO) mice displayed ataxia during the arousal period. Comparatively, the transition to the fully awake state after regaining of righting reflex was without overt ataxia, presumably because of the induction concentrations for tail clamp being higher than for loss of righting reflex. Extending our observations to the clinical setting is difficult, but our findings suggest that there could be a link between postoperative emergence trajectory and changes in glial metabolism in some individuals.
Thrane et al.30 showed that anesthetic agents (ketamine/xylazine, isoflurane, urethane) depress widespread astrocytic IP3R2 (inositol 1,4,5-triphosphate type 2 receptor)–dependent Ca2+ transients associated with arousal in mice. One model of arousal argues that there is an evolutionarily conserved single arousal switch governing both loss and recovery of consciousness.31–33 However, anesthesia-induced loss and recovery of consciousness have also been suggested to be two distinct states clearly defined by separate anesthetic concentration response curves.34,35 Kelz’s group34,35 first described a difference between the concentrations needed for induction and emergence from the state of general anesthesia in Drosophila and in mice. In their pioneering work, they characterized this as a hysteresis, e.g., arousal was in some way state-dependent. They termed this phenomenon “neural inertia.”34 Our results suggest that the neural inertia may be mediated by astrocytes and indicates that the restoration of consciousness is an energetically demanding process, inhibited by volatile anesthetics independent from the mechanisms causing initial loss of consciousness. The difference in emergence versus induction concentrations of isoflurane for the astrocyte-specific Ndufs4(KO) was similar to that described by Kelz’s group34 for mice lacking dopamine β-hydroxylase. However, loss of righting reflex for wild-type mice in halothane was reported by Friedman et al.34 to reduce the emergence/induction ratio to less than half. In the present report, control mice did not display statistically significant hysteresis under halothane or isoflurane for loss of righting reflex. Whether this disparity was due to a different stepwise decrease in halothane concentration (0.04% vs. our 0.1%), the manner in which loss of righting reflex was measured, the genetic background of the mouse strain, or some other technical difference is not clear.
It is important to note that astrocytes modulate both excitatory and inhibitory neuronal firing and are likely themselves regulated by metabolic status.36–38 Astrocyte-mediated inhibition of neuronal firing can be dysregulated by astrocytic metabolic dysfunction and has been shown to result in seizure-like events in mouse cortex, counteracted by etomidate or propofol.38 These drugs mediate their effects by accentuating GABAergic neurotransmission. Volatile anesthetics primarily mediate their effects by inhibiting excitatory neurotransmission37,38 ; our results are more consistent with astrocytic effects on excitatory synapses. The use of volatile anesthetic sensitivity as an endpoint in our studies appears to isolate the anesthetic effects on excitatory rather than inhibitory synapses. Because we previously did not see a major role for loss of NDUFS4 in GABAergic neurons in mediating anesthetic sensitivity,39 any effects of astrocytes on inhibitory neurotransmission is unlikely to be observed with our anesthetic endpoints. In the absence of anesthetic, the suboptimal mitochondrial function in mutant astrocytes appears to be sufficient to avoid the excitatory effects seen by others with more profound defective astrocyte function.36,38 The emergence defect of the astrocyte-specific Ndufs4(KO) reflects an inability of the astrocytic metabolic mutant to reinitiate excitatory firing postinduction probably because of a persistent mitochondrial inhibition in the astrocyte at relatively low volatile anesthetic concentrations.
The simplest explanation of our results is that lack of ATP in astrocytes directly alters gliotransmission, resulting in downstream inability to regain excitatory synaptic function in the presence of low concentrations of volatile anesthetic. However, whether the mitochondrial dysfunction alters ATP levels or disrupts either Ca2+ homeostasis or glutamate recycling is not known. Although astrocytes are classically thought to rely on aerobic glycolysis40,41 during synaptic transmission, the astrocytic somas contain the same volume of mitochondria as neuronal cell bodies,42 implying important mitochondrial functions in astrocytes.43–45 During resting states, astrocytic oxidative metabolism accounts for 30% of the cortical oxygen consumption, while occupying 20 to 25% of the total volume.46,47 The recognition that astrocytes are equivalent to neurons in their oxidative capacities signals a shift in the classical perception of astrocytes as primarily glycolytic, with broad impacts for the functional roles of the cells during resting and active states.
We also explored whether adrenergic signaling is affected by the Ndufs4(KO). Norepinephrine and acetyl choline are important extracellular activating signals to cortical astrocytes48,49 and have been associated with increasing astrocytic Ca2+ transients16,50,51 and inducing the release of gliotransmitters.19,52 Selective activation of the locus coeruleus neurons in rats under deep isoflurane anesthesia was shown to reduce burst suppression and shift electroencephalogram power from δ to θ, indicative of cortical arousal, and accelerate behavioral emergence.53 We observed that the total norepinephrine content is unaltered between the global and astrocyte-specific Ndufs4 knockouts and wild-type mice; it is still possible that the release or regional distribution of norepinephrine might be affected, rather than its production.
In summary, we have discovered an unexpected phenotype from acute loss of Ndufs4 in astrocytes. Loss of Ndufs4 in astrocytes did not change the volatile anesthetic concentrations required to induce anesthesia. However, the concentrations at which emergence occurred were significantly lower in the astrocyte-specific Ndufs4(KO) indicating a hysteresis in anesthetic sensitivity, produced by astrocytes alone.
The authors thank Beatrice Predoi, M.D., Hailey Worstman, B.S., Jake Nealon, and Julia Stokes, B.S., for technical support; and Ernst-Bernhard Kayser, Ph.D., Christian Woods, B.S., and Pavel Zimin, Ph.D., for meaningful discussions and critical advice.
Supported by National Institutes of Health grant No. R01GM105696 (to Dr. Sedensky) and by the Northwest Mitochondrial Research Guild, Lynnwood, Washington.
The authors declare no competing interests.