Mitochondrial Complex I Mutations Predispose Drosophila to Isoflurane Neurotoxicity

Approval from the institutional animal care and use committee was waived. We followed ARRIVE guidelines to the degree applicable to invertebrate organisms.^15,16 

All flies were cultured at 25°C on standard cornmeal-molasses food. The ND23⁶⁰¹¹⁴ and UAS-ND23 lines are described by Loewen and Ganetzky.¹² ND23^G14097, which is homozygous lethal, contains a P-element insertion 259 base pairs downstream of the transcription start site for ND23. ND23^Del (i.e., Df(3R)Exel8162), which is homozygous lethal, deletes an ~125-kilobase pair region that contains 19 genes, including ND23. Canton S, w¹¹¹⁸, Tub-Gal4, C155-Gal4, Repo-Gal4, RAL352, and RAL774 lines were obtained from the Bloomington Drosophila Stock Center (Bloomington, Indiana). Canton S flies are referred to as wild-type throughout the article because the ND23⁶⁰¹¹⁴ mutation was generated in the Canton S genetic background. For overexpression of ND23⁶⁰¹¹⁴, flies were generated by crossing female ND23⁶⁰¹¹⁴ flies containing the UAS-ND23 transgene to male ND23^G14097 flies containing the driver transgene. Heterozygous lines were generated by crossing ND23 females to males of the other genotype.

On the day of the experiment, flies of different sex, age, and genotype were allocated to vials ensuring equal numbers for the different treatment groups. Flies were exposed to volatile general anesthetics and oxygen at specific concentrations using a device that permits simultaneous exposure of up to eight groups of flies.¹⁴  Volatile general anesthetics were administered using an Aestiva/5 anesthesia machine equipped with commercial agent-specific vaporizers (Datex–Ohmeda Inc., USA). Compressed gas cylinders (Airgas, USA) containing 100% O₂, 100% N₂, or air (21% O₂/79% N₂) provided carrier gas of the desired composition. Anesthetic exposure consisted of either 2% isoflurane or 3.5% sevoflurane for 2 h at room temperature. Isoflurane and sevoflurane were obtained from Piramal Enterprises Ltd. (India). To control for effects of circadian rhythm, the flies were anesthetized at a similar time of day (between 10:00 am and 2:00 pm).

Percent mortality was determined by counting the number of dead flies 24 h after the end of an anesthetic exposure and dividing by the total number of flies per vial. Mortality is a binary readout and was assessed by an unblinded observer. We interpret the observed mortality as due to anesthetic and/or oxygen toxicity. Each experiment was conducted in at least triplicate with 25 to 50 flies in each sample. The mortality time course was composed of independent experiments covering three time frames: 0 to 8, 8 to 16, and 16 to 24 h. All experiments followed the same basic structure. All flies were cultured at 25°C until the time of collection at 0 to 2 days posteclosion when they were separated by sex under carbon dioxide narcosis. Flies were then maintained at 29°C for the remainder of the experiment, with food changes every other day. For older heterozygous lines, food changing was paired with carbon dioxide narcosis.

We used quantitative real-time reverse transcription–polymerase chain reaction to measure the expression of select genes known to respond to oxidative stress. For each treatment, ND23⁶⁰¹¹⁴ flies were collected 30 min after the end of exposure to anesthesia. RNA isolation and processing have been described previously.¹⁷  In brief, 25 flies per treatment were transferred to 1.5-ml tubes and frozen in liquid N₂, tubes were shaken vigorously by vortexing to shear heads from bodies, and heads were isolated from other body parts by sequential use of US Standard no. 25 (710 μm) and no. 40 (425 μm) sieves. Total RNA was isolated from heads using TRI reagent (Sigma, USA). The resulting RNA was resuspended in 40 μl of diethylpyrocarbonate-treated water and subjected to DNase treatment using the TURBO DNA-free kit (Invitrogen, Thermo Fisher Scientific, USA). The resulting RNA was ammonium acetate–precipitated and resuspended in 20 μl of diethylpyrocarbonate-treated water. cDNA was generated using the iScript cDNA synthesis kit (Bio-Rad, USA). Real-time reverse transcription–polymerase chain reaction was carried out using a Bio-Rad CFX96 real-time system C1000 Touch thermal cycler (Bio-Rad). Primer sequences are provided in Supplemental Digital Content (https://links.lww.com/ALN/C454).

No a priori power calculation was conducted in this discovery-driven project. The unit for data analysis (replicate) was one vial containing at least 20 flies. The number of flies in a vial varied because of uneven fertility among mutant and wild-type strains. Independently tested vials were considered biologic replicates. Data analysis is based on biologic replicates and visualized using Prism 6 software (GraphPad, USA).

The box plots used to describe the data include the second and third quartiles of data. The mean (+) and the median (horizontal bar) are indicated, and whiskers extend to the minimum and maximum data point. We used parametric descriptive statistics because mortality rates for male and female mutants were normally distributed, and mean and median values were similar in all experiments. Statistical significance between two data points (control and experimental condition) was tested using an unpaired equal-variance two-tail Student’s t test. Multiple groups were compared using one-way ANOVA unless otherwise indicated in the figure legend followed by the Dunnett’s post hoc test for multiple comparisons. The significance level was set at 0.05. Outliers were included in all analyses.

To examine whether the ND23⁶⁰¹¹⁴ mutation sensitizes flies to toxic effects of volatile general anesthetics, we exposed wild-type and ND23⁶⁰¹¹⁴ flies to equipotent doses of isoflurane and sevoflurane (2% and 3.5% for 2 h, i.e., 4%h and 7%h, respectively) in 21% O₂ (normoxia) and determined the percent mortality at 24 h after the exposure. We examined male and female flies at two age ranges, 2 to 4 and 10 to 13 days old, which for simplicity we refer to as young and old flies, respectively. For wild-type flies, no significant mortality was observed under any of the conditions (fig. 1A). In contrast, for old male and female ND23⁶⁰¹¹⁴ flies, isoflurane caused about 50% mortality (fig. 1B). No significant mortality was observed for young ND23⁶⁰¹¹⁴ flies exposed to isoflurane or for ND23⁶⁰¹¹⁴ flies exposed to sevoflurane at either age. These data indicate that the underlying age-associated isoflurane toxicity mechanism is not shared by sevoflurane, despite the fact that sevoflurane induces a state of anesthesia indistinguishable from that of isoflurane.¹⁴ 

To test the age dependence, the time course, and the ND23 allele dependence of isoflurane toxicity, we conducted the experiments shown in figure 2. First, to narrow the age window for the onset of the toxic effect of isoflurane, we tested 6- to 8-day-old male and female flies. We observed no mortality in this age group (fig. 2A). We conclude that sensitization to isoflurane mortality develops in a narrow time window between 8 and 10 days of age. To examine whether mortality was due to an acutely lethal dose of isoflurane or a delayed effect, we determined when 10- to 13-day-old ND23⁶⁰¹¹⁴ flies died during the 24-h period after exposure. All flies initially regained mobility after exposure to isoflurane, and most flies that died did so between 12 and 24 h after exposure (fig. 2B). Therefore, the mechanism underlying isoflurane toxicity in ND23⁶⁰¹¹⁴ flies is likely to involve cellular and molecular signaling events that take several hours to build up. Finally, to test whether the age-dependent isoflurane toxicity of ND23⁶⁰¹¹⁴ flies was due to the ND23 mutation, we repeated the mortality assay with other ND23 alleles. We tested flies that were heteroallelic for ND23⁶⁰¹¹⁴ and ND23^G14097, which contains a P-element insertion in ND23, or ND23^Del, which deletes ND23. We found that similar to 10- to 13-day-old ND23⁶⁰¹¹⁴ flies, exposure to isoflurane caused greater than 50% mortality in 10- to 15-day-old ND23⁶⁰¹¹⁴/ND23^G14097 and ND23⁶⁰¹¹⁴/ND23^Del flies (fig. 2C). Thus, mutations in ND23 are a risk factor for age-dependent isoflurane toxicity.

To determine whether specific cell types control vulnerability to isoflurane toxicity, we attempted to rescue mortality in ND23 mutants using the GAL4-UAS system to express wild-type ND23 in various subsets of cells.¹⁸  We found that ubiquitous expression of UAS-ND23 (using the Tubulin-Gal4 driver) completely rescued the mortality of old male and female ND23⁶⁰¹¹⁴/ND23^G14097 flies exposed to isoflurane (fig. 3). This result provides additional evidence that isoflurane-induced mortality was due to the loss of ND23, as opposed to the loss of another gene. Neuron-specific expression of UAS-ND23 (using the C155-Gal4 driver) also resulted in almost complete rescue of the mortality from 97.6 ± 2.1 to 8.8 ± 10.7% and 99.1 ± 1.60 to 0.9 ± 1.7% in males and females, respectively. In addition, glia-specific expression of UAS-ND23 (using the Repo-Gal4 driver) partially rescued the mortality from isoflurane exposure from 100.0 ± 0.0 to 61.2 ± 27.7% and 100.0 ± 0.0 to 39.2 ± 27.7% in males and females, respectively. Taken together, these results indicate that isoflurane-induced mortality in ND23 mutants is largely due to cell nonautonomous effects of loss of ND23 activity in the nervous system. Cell nonautonomous effects may reflect the close interconnection between glial and neuronal energetic homeostasis in neurodegenerative conditions.¹⁹ 

Because volatile general anesthetics are typically administered under hyperoxic conditions, we explored whether changes in oxygen concentration influenced the vulnerability of ND23⁶⁰¹¹⁴ flies to isoflurane and sevoflurane toxicity. Male and female wild-type and ND23⁶⁰¹¹⁴ flies were maintained in room air (~21% O₂) before and after a 2-h exposure to volatile general anesthetic in 5% O₂ (hypoxia) or 75% O₂ (hyperoxia). As with normoxia (fig. 1), no significant mortality was observed in young or old wild-type flies exposed to isoflurane or sevoflurane and hyperoxia or hypoxia (fig. 4, A and C). In contrast, old male and female ND23⁶⁰¹¹⁴ flies exposed to isoflurane and hyperoxia had significantly increased mortality compared with normoxia, whereas exposure to isoflurane and hypoxia had significantly reduced mortality compared with normoxia (fig. 4, B and D). The mortality from sevoflurane and hyperoxia or hypoxia was low, but male ND23⁶⁰¹¹⁴ flies exposed to sevoflurane and hyperoxia had significantly higher mortality than equivalently treated wild-type flies (9.63 ± 8.87% and 0.0 ± 0.0% for ND23⁶⁰¹¹⁴ and wild-type flies, respectively. P = 0.0473). Therefore, oxygen is an obligate cofactor for the manifestation of isoflurane toxicity in old ND23⁶⁰¹¹⁴ flies, and the susceptibility to inhalational agent toxicity may be greater in males than females. Rescue of isoflurane-induced mortality by hypoxia and aggravation by hyperoxia in ND23⁶⁰¹¹⁴ flies is comparable with the oxygen responsiveness of disease progression in Ndufs4 mutant mice, which is also a model of Leigh syndrome.²⁰  The Ndufs4 subunit (NADH:ubiquinone oxidoreductase subunit S4) is, like ND23 (NADH:ubiquinone oxidoreductase subunit S8), part of the core of complex I of the mitochondrial electron transport chain. Our results are consistent with a potential role of therapeutic hypoxia in certain mitochondrial diseases.²¹ 

A parsimonious explanation of these data is that an excessive increase in reactive oxygen species induced by a combination of isoflurane and hyperoxia²²  tilts the balance from redox homeostasis to oxidative stress²³  in the brains of old ND23⁶⁰¹¹⁴ flies. To test whether ND23⁶⁰¹¹⁴ flies have increased susceptibility to reactive oxygen species toxicity, we fed 10- to 13-day-old wild-type and ND23⁶⁰¹¹⁴ flies paraquat, which increases superoxide generation by complex I.²⁴  For the first 24 h of the experiment, after a 2-h starvation period, flies were fed paraquat in 5% sucrose, and for the second 24 h, they were fed standard cornmeal-molasses food. At 48 h after exposure to 0.05 or 0.5 mM paraquat, ND23⁶⁰¹¹⁴ flies had a significantly higher percent mortality than wild-type flies (fig. 5), indicating that the ND23 mutation reduces resilience to oxidative stress.

To investigate the molecular pathways mediating the differential mortality of 10- to 13-day-old ND23⁶⁰¹¹⁴ flies to isoflurane and hyperoxia relative to sevoflurane and hyperoxia, we examined the expression of genes that are transcriptionally induced by oxidative stress. We examined the expression of GstD1 and GstD2 (encoding glutathione S-transferase D1 and D2, respectively, homologous to mammalian glutathione S-transferases), Jafrac1 and Jafrac2 (encoding thioredoxin peroxidase 1 and 2, respectively, homologous to mammalian peroxiredoxins 1 and 4),²⁵ SOD1 and SOD2 (encoding superoxide dismutase 1 and 2, respectively, homologous to mammalian superoxide dismutases), Cat (encoding catalase, homologous to mammalian catalases), and IRC (encoding immune-regulated catalase, homologous to mammalian COX2). We used real-time reverse transcription–polymerase chain reaction to measure mRNA amounts in the heads of 10- to 13-day-old ND23⁶⁰¹¹⁴ flies 30 min after a 2-h exposure to isoflurane or sevoflurane in hyperoxia or hypoxia. We examined this time point, which is before significant mortality (fig. 2B), because gene expression changes relevant to death should occur before its onset and because others have found that changes in gene expression become detectable as early as 30 min after initiation of isoflurane exposure.²⁶ 

For almost all of the genes, the transcript amount was altered by exposure to either agent under both hyperoxic and hypoxic conditions, and the fold change was smaller in females than males (fig. 6, A and B). Moreover, in males, isoflurane and hyperoxia or hypoxia increased GstD2 mRNA 5.6 ± 0.6- and 4.8 ± 0.3-fold, respectively (fig. 6A), whereas sevoflurane elicited a substantially larger increase in expression: 22.1 ± 0.4- and 25.2 ± 0.6-fold, respectively. Comparable differences between isoflurane and sevoflurane were observed in females (fig. 6B). Likewise, in both males and females, sevoflurane and hyperoxia or hypoxia led to a greater reduction in Jafrac2 expression than isoflurane and hyperoxia or hypoxia (fig. 6). In contrast, in both males and females, isoflurane or sevoflurane and hyperoxia or hypoxia caused equivalent increases in Sod1 and Sod2 expression. Unexpectedly, oxygen concentration did not affect the expression of any of the genes. These data suggest that both isoflurane and sevoflurane activate reactive oxygen species–protective pathways, but they do so to different extents, which may contribute to the differences in mortality.

To examine the dose dependence of ND23 for isoflurane toxicity, we repeated the mortality assay with heterozygous ND23⁶⁰¹¹⁴ flies that have approximately half the amount of ND23 protein and a slightly longer lifespan compared with wild-type flies.¹²  At 15–22 days old, ND23⁶⁰¹¹⁴/wild-type flies (derived by crossing female ND23⁶⁰¹¹⁴ flies to male wild-type flies) were not vulnerable to mortality from exposure to isoflurane and hyperoxia (2 h of 2% isoflurane in 75% O₂; fig. 7). Because aging is associated with deteriorating mitochondrial function,²⁷  we also tested older ND23⁶⁰¹¹⁴/wild-type flies. We did not observe substantial mortality for ND23⁶⁰¹¹⁴/wild-type flies until they were more than 30 days old. At 33 to 39 days old, exposure to isoflurane and hyperoxia resulted in 54.05 ± 19.59% mortality in mixed-sex ND23⁶⁰¹¹⁴/wild-type flies. Thus, aging sensitizes old heterozygous ND23 mutant flies to toxicity from isoflurane and hyperoxia.

In studies of 30- to 39-day-old flies, isoflurane and hyperoxia did not increase mortality in wild-type flies beyond natural attrition (fig. 8A); however, it significantly increased the mortality of female ND23⁶⁰¹¹⁴/wild-type from 3.9 ± 3.1 to 27.4 ± 15.9% (fig. 8B). Mortality was also increased in ND23⁶⁰¹¹⁴/wild-type male flies, but only the increase in isoflurane and normoxia from 25.7 ± 8.6 to 53.5 ± 24.5% was significant, probably because of the high rate of natural mortality. Similar results were observed with another ND23 allele, ND23^Del/wild-type flies (fig. 8C). Because of their short lifespan, ND23^Del/wild-type flies were tested at 21 to 26 days old. Finally, we found that isoflurane and hyperoxia increased the mortality of 30- to 39-day-old ND23^G14097/wild-type flies, but the difference did not reach the significance cutoff (Supplemental Digital Content, https://links.lww.com/ALN/C455). This was unexpected because ND23^G14097 is a stronger allele than ND23⁶⁰¹¹⁴ in other contexts.¹²  Thus, advanced age renders heterozygous carriers of some ND23 alleles susceptible to toxicity from exposure to isoflurane and hyperoxia.

To test whether genetic background influences mortality from isoflurane and hyperoxia, we crossed ND23⁶⁰¹¹⁴ flies to w¹¹¹⁸ flies (a standard laboratory line) and to RAL774 and RAL352 flies, which are part of the Drosophila Genetic Reference Panel collection of inbred lines from a natural population.²⁸  RAL774 and RAL352 were selected for analysis because they are behaviorally more sensitive to isoflurane than w¹¹¹⁸ and wild-type and less sensitive than ND23⁶⁰¹¹⁴.¹⁴  Exposure to isoflurane in hyperoxia significantly increased the mortality of ND23⁶⁰¹¹⁴/wild-type flies (P < 0.0001) and ND23⁶⁰¹¹⁴/w¹¹¹⁸ flies (P < 0.05). Interestingly, ND23⁶⁰¹¹⁴/RAL774 flies (P = 0.667) and ND23⁶⁰¹¹⁴/RAL352 flies (P = 0.671; fig. 9) were not significantly affected. We conclude that polymorphisms in the genetic background modulate the toxicity of isoflurane and hyperoxia in aged heterozygous ND23 flies and that the tested RAL lines contain genetic polymorphisms in their background that apparently counteract the increased mortality that would otherwise be expected under the conditions tested. The nature of these polymorphisms could be explored with a genome-wide association study analysis of the Drosophila Genetic Reference Panel collection.

It is not known how age, sex, and environmental factors modulate the penetrance of the pathophysiology underlying Leigh syndrome,⁹  as they do for other more common diseases caused by complex I mutations, such as Leber’s hereditary optic neuropathy.¹⁰  Obtaining information to answer these questions would thus be an important advance for the mitochondrial disease community and their medical teams. We used an invertebrate model of Leigh syndrome and of asymptomatic carriers to explore interactions with age, sex, volatile general anesthetics, and oxygen.

Young, 2- to 4-day-old ND23⁶⁰¹¹⁴ flies are equally hypersensitive to isoflurane and sevoflurane, compared with six inbred strains, including the wild-type.¹⁴  Notably, the relative potencies of isoflurane and sevoflurane (expressed as the ratio of EC₅₀ isoflurane/EC₅₀ sevoflurane) are nearly identical (0.42 and 0.40 for ND23⁶⁰¹¹⁴ and wild-type flies, respectively), indicating a proportional increase in the sensitivity of behavioral circuits to both agents without toxicity in animals lacking morphological signs of neurodegeneration.¹²  Similarly, studies in mammals also report increased sensitivity to vapor anesthetics in young (23- to 27-day-old) mice harboring mutations in complex I at an age when the animals do not yet exhibit signs of neurodegeneration.²⁹  Whether advancing age uncovers toxic phenotypes in rodents is unknown. Furthermore, in the developing chemotactic system of worms, isoflurane neurotoxicity is linked to inhibition of mitochondrial function.³⁰ 

In contrast with the similar behavioral sensitivity of ND23 mutant flies to isoflurane and sevoflurane, ND23 mutants exposed to isoflurane but not to sevoflurane abruptly developed a mortality phenotype between 8 and 10 days of age (figs. 1 and 2A). In this context, it is important to note that we do not expect the toxic phenotype (i.e. mortality) in the fly model to translate as mortality in higher animals. Hyperoxia (99% O₂), for example, dramatically shortens lifespan in fruit flies^31,32  but causes only impaired cognition without lethality in rats.³³  Analogously, mortality in flies serves as an indicator of interference with brain mitochondrial homeostasis that may present with cognitive (“neurotoxic”) phenotypes in higher animals.

Isoflurane-induced mortality of ND23 mutants is not the result of an acute anesthetic overdose because: (1) mortality increased over many hours after initial recovery from the behavioral effect of anesthesia (fig. 2B), (2) no death occurred in flies up to 8 days of age despite behavioral hypersensitivity (fig. 2A),¹⁴  and (3) ND23⁶⁰¹¹⁴ flies were equally hypersensitive to isoflurane and sevoflurane, but only isoflurane caused high mortality. We interpret these results as indicating that anesthesia and mortality are dissociable: molecular targets mediating the conventional behavioral effects of vapor anesthetics (“anesthesia”) are largely shared between isoflurane and sevoflurane, but targets contributing to mortality in mutant flies (and possibly to adverse neurocognitive outcomes in higher animals) may differ. The finding that isoflurane is more toxic than sevoflurane in cultured neuroblastoma cells injured by oxygen-glucose deprivation supports this notion.³⁴  However, only a few studies have directly compared toxic effects of isoflurane and sevoflurane administered in vivo. No differences were observed in the induction of apoptotic cell death by isoflurane and sevoflurane in the brains of neonatal mice,³⁵  but with a different experimental paradigm that used behavioral outcomes, isoflurane resulted in a more extensive memory impairment phenotype than sevoflurane.³⁶  Moreover, a clear difference between isoflurane and sevoflurane was found in their propensity to activate members of the transient receptor potential superfamily TRPA1 and TRPV1, resulting in an in vivo inflammatory response after exposure to isoflurane but not to sevoflurane,³⁷  thereby demonstrating that biologically significant differences exist between these interchangeably used agents.

Our finding that isoflurane-induced mortality of ND23⁶⁰¹¹⁴ flies was rescued by expression of wild-type ND23 in neurons or in glia indicates that mortality is driven by cell nonautonomous consequences of ND23 activity in the brain (fig. 3). Notably, selective expression of ND23 in neurons rescued mortality to a greater extent than expression in glia. Given that glial cells constitute only 10% of cells in the fly brain, the disproportionally high degree of rescue observed with glia-specific expression of ND23 may be due to a higher rate of reactive oxygen species production by glia and/or to the energy support of neurons by glia.³⁸  Lopez-Fabuel et al.³⁹  found that complex I in neurons is mainly assembled into supercomplexes that produce reactive oxygen species at a lower rate than free complex I, which is more abundant in astrocytes. Interestingly, mice carrying a mutation of complex I only in astrocytes exhibit a subtle behavioral phenotype under isoflurane anesthesia.⁴⁰  Thus, isoflurane-induced production of reactive oxygen species in neurons and glia may mediate toxicity in ND23 mutants.

In Caenorhabditis elegans, suppression of complex I function by isoflurane increases production of reactive oxygen species in wild-type and mutant mitochondria,⁴  whereas complex I mutations confer increased sensitivity to oxidizing agents such as paraquat.⁴¹  Further data from rodents⁴²  and from a mammalian cell culture model⁴³  suggest a role for reactive oxygen species in the toxicity of isoflurane. Our data regarding the effects of oxygen concentration on isoflurane toxicity (figs. 1 and 4), the increased sensitivity of ND23 mutants to paraquat compared with wild-type flies (fig. 5), are also consistent with a role for reactive oxygen species in mortality. Additional support for the reactive oxygen species hypothesis of isoflurane toxicity is provided by our gene expression analysis showing that isoflurane and sevoflurane elicit qualitatively similar but quantitatively different patterns of transcriptional changes in some gene target of reactive oxygen species response pathways (GstD2 and Jafrac2; fig. 6). GstD2 expression in flies increases in response to hyperoxic stress⁴⁴  and to stress caused by blue light that results in neurodegeneration.⁴⁵  Jafrac2 has both antioxidant and signaling activity that regulates intracellular hydrogen peroxide levels.²⁵  Suppression of Jafrac2 expression results in shortened lifespan in the face of oxidative stress with hydrogen peroxide and paraquat.²⁵  Reactive oxygen species are natural byproducts of electron transfer in mitochondria functioning as signaling molecules responsive to environmental cues.^46,47  Mitochondrial dysfunction and reactive oxygen species–induced damage accumulate with aging^32,48,49  and may sensitize flies to the acute oxidative stress–like insult from isoflurane. Therefore, although it is evident that exposure to both isoflurane and sevoflurane triggers responses in oxidative stress pathways, our data do not allow us to infer which changes are protective (e.g., strong upregulation of GstD2), which are signs of a failure to adapt (e.g., weak upregulation of GstD2), or how these changes in gene expression relate to mortality. A toxicity mechanism involving reactive oxygen species pathways is also consistent with the age dependence of isoflurane-induced mortality in ND23 mutant flies (figs. 1, 2A, 4, and 7).

About 5.9 in 100,000 individuals are at risk of nuclear DNA–related mitochondrial diseases,⁸  the majority of which are due to recessive mutations. We found that age renders heterozygous carriers of ND23 mutations vulnerable to the stress of isoflurane under normoxic and hyperoxic conditions (figs. 7 and 8). Polymorphisms in the nuclear genome, introduced either by other alleles of ND23 (fig. 8) or by crosses into other fly lines (fig. 9), modified the genetic environment of the ND23 mutation and resulted in varying degrees of vulnerability to the toxicity of isoflurane and hyperoxia. These results suggest that heterozygous mutations in complex I genes that underlie Leigh syndrome, the most common mitochondrial encephalomyelopathy,⁵  can increase the risk for deleterious manifestations under the combined stresses of advanced age, anesthesia, and hyperoxia with the degree of risk dependent on genetic background. Importantly, our data from heterozygous flies indicate that under conditions of advanced age and environmental stress (e.g., surgery under anesthesia with isoflurane and hyperoxia), an otherwise silent reduction in complex I function may translate into adverse outcomes. In the clinical scenario, the association of cardiopulmonary bypass with a high incidence of temporary central nervous system dysfunction⁵⁰  and with mitochondrial dysfunction and oxidative damage⁵¹  may not be accidental.

In summary, our data raise testable predictions. Other mutations affecting complex I and mutations affecting other components of oxidative metabolism should result in a toxic phenotype under anesthesia. Furthermore, strategies aimed at mitigating oxidative stress should reduce anesthetic toxicity. A limitation of our data is that we have not yet established a direct causal relationship between oxidative stress and mortality. Other tests such as genome-wide transcriptome analysis are necessary to provide a more comprehensive overview of pathways involved in isoflurane toxicity. A better understanding of these pathways will reveal genetic risk factors and therapeutic approaches to anesthetic neurotoxicity. Our experiments demonstrate the value of Drosophila for understanding the multifaceted interactions of anesthetics, aging, genetic background, and mitochondrial function.

The authors thank the anonymous reviewers whose comments and suggestions have substantially improved the article. The authors thank Carin Loewen, Ph.D., Department of Genetics, University of Wisconsin-Madison, Madison, Wisconsin for generating fly stocks and for advice throughout the study, and we also thank Grace Boekhoff-Falk, Ph.D., Department of Cell and Regenerative Biology, University of Wisconsin–Madison, Madison, Wisconsin, for helpful discussions throughout the project and Mary Roth, B.A., Department of Anesthesiology, University of Wisconsin-Madison, Madison, Wisconsin for excellent assistance with proofreading.

Supported by a seed grant from the Department of Anesthesiology, University of Wisconsin–Madison, Madison, Wisconsin (to Dr. Perouansky) and National Institutes of Health grant No. R01GM134107 (Bethesda, Maryland; to Drs. Perouansky and Wassarman).

The authors declare no competing interests.