Potassium Leak Channels and Mitochondrial Complex I interact in glutamatergic interneurons of the Mouse Spinal Cord

Abstract

Background.

Volatile anesthetics induce hyperpolarizing potassium currents in spinal cord neurons that may contribute to their mechanism of action. They are induced at lower concentrations of isoflurane in noncholinergic neurons from mice carrying a loss-of-function mutation of the Ndufs4 gene, required for mitochondrial complex I function. The yeast NADH dehydrogenase enzyme, NDi1, can restore mitochondrial function in the absence of normal complex I activity and gain-of-function Ndi1 transgenic mice are resistant to volatile anesthetics. We tested whether NDi1 would reduce the hyperpolarization caused by isoflurane in neurons from Ndufs4 and wildtype mice. Since volatile anesthetic behavioral hypersensitivity in Ndufs4 is transduced uniquely by glutamatergic neurons, we also tested whether these currents were also unique to glutamatergic neurons in the Ndufs4 spinal cord.

Methods.

Spinal cord neurons from wildtype, NDi1, and Ndufs4 mice were patch-clamped to characterize isoflurane sensitive currents. Neuron types were marked using fluorescent markers for cholinergic, glutamatergic, and GABAergic neurons. Norfluoxetine was used to identify potassium channel type. Neuron type-specific Ndufs4 knockout animals were generated using type-specific Cre-recombinase with floxed Ndufs4.

Results.

Resting membrane potentials (RMP) of neurons from NDi1;Ndufs4, unlike those from Ndufs4, were not hyperpolarized by 0.6% isoflurane (Ndufs4, delta(Δ)RMP −8.2mV(−10,−6.6); p=1.3e-07; Ndi1;Ndufs4, ΔRMP −2.1mV(−7.6,+1.4); p=1,). Neurons from NDi1 animals in a wildtype background were not hyperpolarized by 1.8% isoflurane (Wildtype, ΔRMP, −5.2mV(−7.3,−3.2), p=0.00057; Ndi1, ΔRMP, +0.6mV(−1.7,3.2), p=0.68). In spinal cord slices from global Ndufs4 animals, holding currents (HC) were induced by 0.6% isoflurane in both GABAergic (ΔHC, 81.3pA(61.7,101.4), p=2.6e-05) and glutamatergic (ΔHC, 101.2pA(63.0,146.2), p=0.0076) neurons. In neuron type-specific Ndufs4 knockouts, holding currents were increased in cholinergic (ΔHC 119.5pA(82.3,156.7), p=0.00019) and trended toward increase in glutamatergic (ΔHC 85.5pA(49,126.9), p=0.064) neurons but not in GABAergic neurons.

Conclusions.

Bypassing complex I by overexpression of NDi1 eliminates increases in potassium currents induced by isoflurane in the spinal cord. The isoflurane-induced potassium currents in glutamatergic neurons represent a potential downstream mechanism of complex I inhibition in determining MAC.

Introduction

The minimum alveolar concentration (MAC) required to prevent response to tactile stimuli is a standard reference point for volatile anesthetic potency.^1,2 Studies in animal models established that volatile anesthetics act, at least partially, in the spinal cord to induce this immobility.^3–7 In mouse spinal cord slices, isoflurane induces an outward potassium current causing a hyperpolarization in neurons.^8–11 The isoflurane induced currents are completely blocked by norfluoxetine which is known to specifically inhibit outwardly rectifying two-pore domain potassium (K⁺) channels (also known as TREK channels).^12,13 Since TREK channels have been shown to be activated by isoflurane,^11,14 they are compelling targets as effectors of volatile anesthetic sensitivity.^15,16

Defects in the function of complex I of the mitochondrial electron transport chain confer hypersensitivity to volatile anesthetics in multiple organisms (nematodes, flies, mice and humans).^17–20 Ndufs4 carries a critical mutation in complex I ²¹ which alters anesthetic sensitivity in the mouse¹⁷ and also sensitizes TREK currents to volatile anesthetics.⁹ The yeast protein NDi1, an NADH dehydrogenase, can function as an electron carrier in mice, allowing electrons to bypass complex I.^22,23 Transfection of the NDi1 gene into the CNS of wild type and Ndufs4 mice caused resistance to volatile anesthetics compared to siblings without NDi1.²⁴ We showed that these changes to complex I had profound effects on neurotransmitter recycling in the hippocampus and in neuronal cultures. We reasoned that if inducible potassium currents also contribute to an anesthetic mechanism, then the presence of NDi1 would lessen the isoflurane-induced currents in both wildtype and Ndufs4 spinal cords. This would establish a connection between MAC, mitochondrial complex I inhibition, membrane hyperpolarization, and TREK channel function.

Genetic studies showed that restricting the expression of the Ndufs4 mutation exclusively to glutamatergic neurons led to full hypersensitivity to volatile anesthetics, while GABAergic or cholinergic expression did not.²⁵ Our earlier studies also indicated that the currents induced by 0.6% isoflurane in Ndufs4 were in noncholinergic cells of the spinal cord, raising the question of which specific neuron type carried the currents.⁹ We hypothesized that, in animals with global loss of Ndufs4, the induction of potassium currents at 0.6% isoflurane would also be restricted to glutamatergic neurons. We labeled specific neuronal types in global Ndufs4 and determined which type(s) carried the inducible currents at 0.6% isoflurane.¹⁷ Furthermore, since restriction of the Ndufs4 mutation to glutamatergic neurons led to full hypersensitivity to volatile anesthetics, we also determined if the isoflurane-induced currents persisted when the Ndufs4 mutation itself was limited to type-specific neurons. We then compared the distribution of induced potassium currents in global Ndufs4 mutants to neuron-type specific Ndufs4 mutants.

We first hypothesized that the presence of NDi1 in neurons would lessen the induction of potassium currents by isoflurane both in wildtype and Ndufs4. Secondly, since glutamatergic loss of Ndufs4 conferred marked reduction of whole animal MAC, we hypothesized that isoflurane-induced hyperpolarizing currents would be limited to glutamatergic cells. While many studies have focused either on the molecular targets or neural pathways affected by volatile anesthetics, we are unaware of studies determining potential differences between spinal neuron types in mediating anesthetic responses.

Methods.

Additional Methods are found in the Supplemental Digital Content.

Animals

All animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and were approved by the Institutional Animal Care and Use Committee of Seattle Children’s Research Institute. Mice were housed at 22°C with a 12-hour light-dark cycle and maintained on a standard rodent diet. Food and water were available ad libitum. Male and female mice were used for each experiment and the total number of each gender for the data in the text are listed in the Tables and Supplemental Data file. There was no attempt to match female and male mice numbers, but each gender was used as available in approximately equal numbers in all experiments. The exceptions were NDi1 cells exposed to 1.8% isoflurane (skewed to males) and global Ndufs4 GABAergic cells (skewed to males).

Mouse strain nomenclatures

In this report we use the term wildtype to denote C57Bl/6 mice. The term mutant refers to Ndufs4(knock-out) or NDi1(knock-in)-containing mice. Global CNS loss of Ndufs4 is referred to as Ndufs4 and NDi1(knock-in) is referred to as NDi1 in accordance with Anesthesiology convention. Neuron-specific loss of Ndufs4 is referred to as “neuronal type”-specific Ndufs4. Cell-specific knockout of Ndufs4 has been previously described, as has use of spinal cord slices from these animals.^9,25

When referring to the protein product of a gene, letters are capitalized and not italicized (e.g., NDUFS4). When referring to the gene or the mutant strain, the first letter is capitalized, and the term is italicized (e.g., Ndufs4). If multiple mutations are carried in the same animal, their genes are separated by a semicolon (e.g., NDi1;Ndufs4). All mutants are in a wildtype C57Bl/6 background. The generation²¹ and functional ^9,17,26 characteristics of the Ndufs4 strain have been previously described. NDi1 was the kind gift of Navdeep S. Chandel (Northwestern University).

ARRIVE Guidelines

Arrive Guidelines are listed in the Supplemental Digital Content 1.

Spinal Cord Slices

Spinal cord slices for whole-cell patch clamp electrophysiology were prepared as described previously. ⁹ In all cases, cells of the lateral ventral horn were patched. Ventral horn spinal cord cells were visualized using differential interference contrast microscopy. Fluorescence microscopy was used to identify different neuron types in mice expressing a fluorescent tag (Ai14) in either cholinergic, GABAergic, or glutamatergic neurons, using a genetic strategy described previously.⁹ The mice express Cre-recombinase (an enzyme which cuts DNA at lox sites) under the control of a cell-specific promoter (ChAT-IRES-Cre for cholinergic neurons, Gad2-IRES-Cre for GABAergic neurons, or Vlgut2-IRES-Cre for glutamatergic neurons). Cre recombinase removes a STOP codon in front of the ROSA26 promoter in Ai14 which drives the expression of the red fluorescent protein tdTomato in the cell of interest. Electrophysiologic protocols have been described previously.^8,9 Since studies were limited by the availability of genotypes and designed labeled cells, electrophysiologic studies were not blinded. Our success rate for patch clamping cells in adult spinal cord approximated 50%.

Drug Administration to spinal cord slices

Slices of lumbar spinal cord were isolated and studied as previously described. ⁹ They were first held in the superfusate for 30 minutes without isoflurane for baseline, unexposed measurements. Isoflurane was then applied in the superfusate at equilibrated concentrations delivered by passing carbogen (a mixture of 95% O₂ and 5% CO₂) through a calibrated isoflurane vaporizer. The superfusate was sampled during isoflurane exposure, and the isoflurane concentration was determined using gas chromatography. To rule out “run down” of the preparation, recordings were also made with no isoflurane exposure for the same period that matched the sum of experimental exposure and wash.

Norfluoxetine (NF) (hydrochloride) (Cayman Chemical #15900) was diluted to a final concentration of 20 μM in recording solutions from a stock solution of 10 mM in dimethyl sulfoxide (DMSO; Sigma-Aldrich, USA), based on prior published IC₅₀=9mcM for TREK blockade.¹² Norfluoxetine was added to the bath following a previously published protocol that compared the effects of isoflurane with and without norfluoxetine.²² In short, this protocol recorded at 0% isoflurane, then isoflurane exposure followed by its washout, then application of norfluoxetine for 15 minutes, followed by resumption of isoflurane exposure, now in the presence of norfluoxetine.

Mitochondrial Function

Mitochondrial function was performed in a Clarke electrode (Oxygraph-K2; Oroborus Instruments, Innsbruck, Austria) as previously described.²⁷

Statistical analysis

All values for testing are expressed as the mean anesthetic concentration with 95% confidence intervals in parentheses (mean (95% CI)) followed by N for the number of animals/slices studied and a corresponding p-value. In the graphs, boxes refer to 95% confidence intervals. Effect sizes are the absolute changes in holding currents (pA) or membrane potential (mV) where reported. All accumulated data is available in Supplemental Table 1.

We used % anesthetic for comparison to prior reports. Temperature corrected aqueous concentrations of isoflurane for comparison in holding current results are as follows; 0.6% isoflurane, 0.25mM (~2XEC50 for Ndufs4); 1.8% isoflurane, 0.74mM (~2XEC50 for wildtype). In labels, “Iso” always refers to isoflurane. For normally distributed data we used a paired or unpaired, one tailed t-test when comparing two groups. For data groups exhibiting a non-normal distribution in either group, a Wilcoxon test (paired or unpaired as appropriate) was used for statistical comparisons. The statistical test used is listed in each panel above the data points along with the resulting p-value. p-values were determined in R and calculated for each paired group individually.

p-values refer to CB57Bl/6 (wild type), Ndufs4 or NDi1 values exposed vs. unexposed to isoflurane determined in paired experiments unless explicitly stated otherwise. Significance level was selected as a p-value less than 0.05. Post-hoc values were subjected to a Bonferroni correction for multiple comparisons to adjust significance for multiple comparisons. For comparisons to C57Bl/6 and Ndufs4 genotypes, when there were three groups of cells (cholinergic, GABAergic, and glutamatergic neurons) the correction led to a corrected p-value for significance of 0.025. Our primary reported measurements were resting membrane potentials or holding currents from neurons in the spinal cord. In addition to the p-values, to consider alterations in neuronal activity we also interpreted that at least a 3mV change in resting membrane potentials or a 50pA change in holding current values was necessary to be of functional significance. We did not discuss changes of less than these values even if the p-value was significant (see, for example, Figure 2B, panel 3).

Figure 2. Holding currents in specific neuronal types in wildtype and Ndufs4 spinal cord neurons exposed to isoflurane.

A. Effects of 0.6% isoflurane on specific neuronal types in wildtype and Ndufs4 spinal cord neurons. Plotted are the differences in holding currents between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Non-cholinergic indicates non-fluorescent neurons when cholinergic neurons were labeled and are included because those data led to the current comparison of glutamatergic and GABAergic neurons. Spinal cord neurons from wildtype animals are in grey, from Ndufs4 animals are in blue. Red symbols indicate data that was also presented earlier and are used here for comparison to the newer data.⁹ p-values for the difference between the effect on wildtype neurons and Ndufs4 neurons are given at the top of each panel. B. Paired absolute effects of 0.6% isoflurane on specific neuronal types in wildtype and Ndufs4 spinal cord neurons. Plotted are the paired absolute values for holding currents between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Spinal cord neurons from wildtype animals are in grey, from Ndufs4 animals are in blue. p-values are for the differences between the unexposed and exposed (isoflurane) holding current for wildtype neurons and Ndufs4 neurons and are given at the top of each panel. C. Paired absolute effects of 1.8% isoflurane on specific neuronal types in wildtype spinal cord neurons. Plotted are the paired absolute values for holding currents between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Only wildtype neurons were exposed to 1.8% isoflurane. Holding currents for wildtype at 0.6% are repeated here to facilitate comparison to 1.8% isoflurane. p-values are for the differences between the unexposed and exposed (isoflurane) holding current for wildtype neurons and are given at the top of each panel. Abbreviations: pA (picoAmperes)

We measured a maximum of one cell per spinal cord slice and one slice per animal. However, the failure rate of obtaining any successful patch clamped cell was about 50%. Thus, our N refers to successful animals, slices, and cells. For holding current studies and resting membrane potentials, means and 95% confidence intervals (M(+/− C.I) are given in the Results section and shown by box plots in the Figures. p-values for comparisons of changes in holding currents are noted in the text. All raw data are available upon request. We used a difference of 10% from the baseline for WT and mutant slices, a standard deviation of 0.1, an alpha of 0.05 and a desired power of 0.8 to determine adequate sample size (generally 5). In most cases, our sample sizes exceeded these values.

Results.

NDi1

NDi1 animals are resistant to volatile anesthetics relative to wildtype; likewise, NDi1;Ndufs4 animals are resistant relative to Ndufs4.²⁴ The presence of NDi1 also maintains synaptic ATP (adenosine triphosphate) levels in the presence of isoflurane.²⁴ We exposed NDi1 to isoflurane and found its function to be unaffected by isoflurane at a maximum dose of 2.4% (~1mM) (Supplemental Figure S1). If potassium currents and resting membrane potentials in spinal cord neurons are important in determining MAC, and are dependent on mitochondrial complex I function, then we reasoned that isoflurane induced changes in holding currents and membrane potentials may be less sensitive to isoflurane if NDi1 is present. We tested whether NDi1 lessened isoflurane’s effects on holding currents and membrane hyperpolarization in ventral horn neurons in spinal cord slices from Ndufs4 and wildtype animals. In NDi1;Ndufs4, NDi1 eliminated the increased holding currents caused by 0.6% isoflurane in Ndufs4 (Ndufs4, ΔHC, 126.1pA(99.4,154.0);N=21, p=9.5e-07; Ndi1;Ndufs4, ΔHC, 24.2pA(−1.1,49.5);N=5, p=0.168) (Figure 1A, Table 1). Hyperpolarization of Ndufs4 neurons caused by 0.6% isoflurane was also eliminated by the presence of NDi1 (Ndufs4, ΔRMP, −8.2mV(−10,−6.6);N=17, p=1.3e-07; Ndi1;Ndufs4, ΔRMP, −2.1mV(−7.6,+1.4);N=5, p=1) (Figure 1B). In a wildtype background, NDi1 lessened, but did not eliminate, the increase in holding currents induced at 1.8% isoflurane (wildtype, ΔHC, 80pA(58.7,100.6);N=13, p=1.29e-08; Ndi1, ΔHC, 55.7(33.6,87.7);N=7, p=0.018) (Figure 1C, Table 1). However, similar to the effect observed in NDi1;Ndufs4, NDi1 eliminated the hyperpolarization caused by isoflurane in a wildtype background at 1.8% isoflurane (wildtype, ΔRMP, −5.2mV(−7.3,−3.2), N=13, p=0.00057; Ndi1, ΔRMP, +0.6mV(−1.7,3.2),N=6, p=0.68) (Figure 1D).

Figure 1. Holding currents and resting membrane potentials for spinal cord neurons in slices from four genotypes.

A. Holding currents, Effects of NDi1 on Ndufs4 at 0.6% Isoflurane. For comparison, the left panel (grey boxes) shows the increase caused by 0.6% isoflurane in wildtype neurons. The induced current increased from 86pA to 113.7pA (Table 1), (effect size = 27.7pA). The middle panel (blue boxes) shows the increase caused by 0.6% isoflurane in Ndufs4 neurons, an increase from 86.3pA to 212.4pA (effect size = 126.7pA). The right panel (tan boxes) shows the effect of NDi1 on the Ndufs4 response, an increase from 131.7pA to 155.9pA (effect size = 24.3pA). In this and all subsequent panels, p-values (and the statistical test used for each group) compare unexposed to exposed and are listed at the top of the panel. B. Resting membrane potentials, Effects of NDi1 on Ndufs4 at 0.6% Isoflurane. The left panel (grey boxes) shows the increase caused by 0.6% isoflurane in wildtype neurons. The membrane potential decreased from −65.6mV to −68.9mV (Table 1), (effect size = −3.3mV). The middle panel (blue boxes) shows the decrease caused by 0.6% isoflurane in Ndufs4 neurons, a decrease from −65.0mV to −73.1mV (effect size = −8.1mV). The right panel (tan boxes) shows the effect of NDi1 on the Ndufs4 response, a decrease from −69.0mV to −71.1mV (effect size = −2.1mV). C. Holding currents, Effects of NDi1 on wildtype at 1.8% Isoflurane. The left panel (grey boxes) shows the increase caused by 1.8% isoflurane in wildtype neurons. The induced current increased from 192.1pA to 272.1pA (Table 1), (effect size = 80pA). The right panel (tan boxes) shows the effect of NDi1 on the wildtype response, an increase from 58.1pA to 115.2pA (effect size = 57.1pA). While the effect sizes were similar, the difference in holding current in the presence of NDi1 did not reach statistical significance. D. Resting membrane potentials, Effects of NDi1 on wildtype at 1.8% Isoflurane. The left panel (grey boxes) shows the decrease caused by 1.8% isoflurane in wildtype neurons. The membrane potential decreased from −68.8mV to −74.1mV (Table 1), (effect size = −5.3mV). The right panel (tan boxes) shows the effect of NDi1 on the wildtype response, an increase from −73.2mV to −72.6mV (effect size = +0.6mV). Abbreviations: mV (milliVolts), pA (picoAmperes)

Table 1.

Holding currents (no highlight) and resting membrane potentials (grey highlight) for spinal cord neurons in slices from four genotypes. Values are the mean holding current to hold the membrane potential at −60mV at 30°C and the corresponding membrane potentials. 95% confidence intervals are shown in parentheses following mean values. The N for each measurement follows parentheses. Males/Females (m/f) are listed in parenthesis after N. Unexposed data is paired to the isoflurane exposure in the immediately following cell. P-values are shown in the Figure 1 for various comparisons;

	Unexposed	0.6% Isoflurane	Unexposed	1.8% Isoflurane
Wildtype Holding current (pA)	86.0(21.1,170.6) N=9(3m/6f)	113.7(37.9,206.2) N=9(3m/6f)	192.1(107.7,284.6) N=13(8m/5f)	272.1(177.5,361.8) * N=13(8m/5f)
Wildtype Resting membrane potential (mV)	-65.6(−69.7,−60.8) N=6(3m/3f)	-68.9(−71.3,−67.1) N=6(3m/3f)	-68.8(−72.0,−65.7) N=13(8m/5f)	-74.1(−76.8,−71.4) * N=13(8m/5f)
NDi1 Holding current (pA)	ND	ND	58.1(36.4,84.2) N=6(5m/1f)	115.2(83.4,150.1) * N=6(5m/1f)
NDi1 Resting membrane potential (mV)	ND	ND	-73.2(−78.9,−67.4) N=6(5m/1f)	-72.6(−76.7,−68.6) N=6(5m/1f)
Ndufs4 Holding current (pA)	86.3(47.1,134.2) N=21(8m/13f)	212.4(153.1,273.2) * N=21(8m/13f)	ND	ND
Ndufs4 Resting membrane potential (mV)	-65.0(−67.3,−62.6) N=17(6m/11f)	-73.1(−75.0,−71.1) * N=17(6m/11f)	ND	ND
Ndi1;Ndufs4 Holding current (pA)	131.7(47.1,216.2) N=5(3m/2f)	155.9(89.1,222.7) N=5(3m/2f)	ND	ND
Ndi1;Ndufs4 Resting membrane potential (mV)	-69.0(−74.6,−60.1) N=5(3m/2f)	-71.1(−74.4,−67.8) N=5(3m/2f)	ND	ND

^*, p<0.025 with Bonferroni correction.

Cell-type Specific Holding Currents in the Global Ndufs4 Knockout

GABAergic and glutamatergic cells of the spinal cord were separately labeled with the fluorescent marker, Ai14. Global loss of NDUFS4 increased holding currents at 0.6% isoflurane in both GABAergic (Ndufs4, ΔHC, 81.3pA(61.7,101.4), N=17, p=2.6e-05) and glutamatergic (Ndufs4, ΔHC, 101.2pA(63.0,146.2),N=10, p=0.0076) neurons (Figure 2A, Right Two Panels). This isoflurane concentration represents the whole animal EC₉₅ for Ndufs4 but does not immobilize wildtype animals. For comparison, we include previously published results⁹ of labeled spinal cord cholinergic neurons (red symbols, Figure 2A, Left Two Panels) which were not changed by isoflurane. No significant changes in holding currents were seen in wildtype neurons of any type at 0.6% isoflurane (Figure 2A, grey boxes, Table 3). Paired comparisons of the same absolute holding currents from each individual cell also showed consistent increases in currents in both GABAergic (Ndufs4, ΔHC, 81.3pA(61.7,101.4), N=17, p=7.6e-07) and glutamatergic (Ndufs4, ΔHC, 101.2pA(63.0,146.2),N=10, p=0.0014) cells of the global Ndufs4 mutant exposed to 0.6% isoflurane (Figure 2B, Table 2). In spinal cord slices from wildtype animals, 1.8% isoflurane increased absolute holding currents in cholinergic (ΔHC,96.6pA(60.8,135.6);N=15;p=0.0003), GABAergic (ΔHC,114.3pA(82.9,150.1);N=12;p=0.0005), and glutamatergic (ΔHC,57.6pA(30.6,86.6);N=13;p=0.0017) neurons (Figure 2C, Table 2).

Table 3.

Values are the changes(Δ) in mean holding current between unexposed and exposed (0.6% isoflurane) neurons to hold the membrane potential at −60mV at 30°C. (Note that, for ease of comparison of different neuron types, this table reports changes in holding currents between exposed and unexposed cells. Prior tables showed absolute values.) 95% confidence intervals are shown in parentheses following mean values. N for each measurement follows parentheses. Males/Females (m/f) are listed in parenthesis after N. P-values are shown in the Figures 2 and 5 for various comparisons;

Cell type/ Genotype	Wildtype	Global CNS Ndufs4	Neuron-specific Ndufs4
cholinergic	23.1(5.4,42.8) N=8(4m/4f)	40.5(8.6,68.1) N=7(5m/2f)	119.6(82.3,156.7) * N=21(5m/16f)
GABAergic	18.4(8.2,31.0) N=10(6m/4f)	81.3(61.7,101.4) * N=17(16m/1f)	18.3(5.0,38.1) N=7(3m/4f)
glutamatergic	24.1(8.5,40.6) N=10(4m/6f)	101.2(63.0,146.2) * N=10(3m/7f)	85.5(49.0,126.9) * N=22(12m/10f)

^*, p<0.025 with Bonferroni correction.

Table 2.

Holding Currents for three types of spinal cord neurons in slices from two genotypes, wildtype and Ndufs4. Values are the mean holding current to hold the membrane potential at −60mV at 30°C. 95% confidence intervals are shown in parentheses following mean values. The N for each measurement follows parentheses. Males/Females (m/f) are listed in parenthesis after N. Unexposed data is paired to the isoflurane exposure in the immediately following cell. P-values are shown in the Figure 2 for various comparisons;

Cell type/Concentration	0% Isoflurane	0.6% Isoflurane	0% Isoflurane	1.8% Isoflurane
Wildtype, cholinergic	102.2(44.3,170.7) N=8(4m/4f)	125.3(69.8,193.0) N=8(4m/4f)	204.8(151.9,257.6) N=15(10m/5f)	301.4(230.0,375.1) * N=15(10m/5f)
Wildtype, GABAergic	17.4(5.9,29.4) N=10(6m/4f)	35.9(26.3,45.4) * N=10(6m/4f)	25.3(11.8,42.5) N=12(8m/4f)	139.6(103.0,176.8) * N=12(8m/4f)
Wildtype, glutamatergic	66.5(29.0,111.5) N=10(4m/6f)	90.6(48.1,138.7) * N=10(4m/6f)	102.8(63.1,152.6) N=13(9m/4f)	160.4(110.6,214.7) * N=13(9m/4f)
Ndufs4, cholinergic	60.3(23.4,98.2) N=7(5m/2f)	100.8(69.9,140.2) N=7(5m/2f)	ND	ND
Ndufs4, GABAergic	37.1(11.8,59.6) N=17(16m/1f)	118.4(83.6,155.5) * N=17(16m/1f)	ND	ND
Ndufs4, glutamatergic	81.9(61.0,105.1) N=10(3m/7f)	183.1(139.1,225.0) * N=10(3m/7f)	ND	ND

^*, p<0.025 with Bonferroni correction.

We tested whether norfluoxetine blocked isoflurane induced increases in holding current in Ndufs4 neurons. We found that the isoflurane-induced increases in both GABAergic (-NF, Δ81.3pA(61.7,101.4)N=17; +NF,Δ13.4pA(−13.0,50.9)N=6; p=0.004) and glutamatergic (-NF, Δ101.2pA(63.0,146.2)N=10; +NF, Δ10.2(−4.2,27.9)N=7; p=0.002) neurons were inhibited by norfluoxetine, consistent with the currents being transduced by TREK channels (Figure 3). Cell-specific currents in wildtype neurons were not tested for sensitivity to norfluoxetine since earlier results indicated that all such currents were norfluoxetine sensitive.⁹

Figure 3. Norfluoxetine inhibits the current in glutamatergic and GABAergic cells.

Plotted are the differences in holding currents between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Non-cholinergic indicates non-fluorescent neurons when cholinergic neurons were labeled and are included because those data led to the current comparison of glutamatergic and GABAergic neurons. Spinal cord neurons from wildtype animals are in grey, from Ndufs4 animals are in blue. p-values are for the differences between the isoflurane effect with and without norfluoxetine (NF) and are given at the top of each panel. Abbreviations: NF (Norfluoxetine), pA (picoAmperes)

Cell-type Specific Membrane Properties in the Global Ndufs4 Knockout

Changes in applied current to hold the transmembrane potential at −60mV in response to isoflurane infer the effects of the anesthetic on membrane resting potential. We also directly measured resting membrane potential in wildtype and Ndufs4 slices. At 0.6% isoflurane, the only significant change was a decrease in resting membrane potential (−67.9mV (−72.5,−63.0) to −74.5mV (−78.0,−71.0)N=17, p=0.0008) in GABAergic neurons in slices from Ndufs4 (Figure 4A). At 1.8% isoflurane in wildtype slices, the only significant change seen was a decrease in resting membrane potential in cholinergic neurons (Figure 4B) ((−70.2mV (−73.1,−67.1) to −75.5mV (−77.2,−73.8),N=15, p=0.0004). This correlates well with the significant increase in holding current in the same cells at 1.8% isoflurane (Figure 2C). However, the increases in holding currents seen in GABAergic and glutamatergic cells of wild type animals at 1.8% (Figure 2C) were not corroborated by significant changes in resting membrane potentials (Figure 4B).

Figure 4. Resting membrane potentials in specific neuronal types in wildtype and Ndufs4 spinal cord neurons exposed to isoflurane.

A. Paired absolute effects of 0.6% isoflurane on specific neuronal types in wildtype and Ndufs4 spinal cord neurons. Plotted are the paired absolute values for resting membrane potentials between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Spinal cord neurons from wildtype animals are in grey, from Ndufs4 animals are in blue. p-values are for the differences between the unexposed and exposed (isoflurane) resting membrane potentials for wildtype neurons and Ndufs4 neurons and are given at the top of each panel. B. Paired absolute effects of 1.8% isoflurane on specific neuronal types in wildtype spinal cord neurons. Plotted are the paired absolute values for resting membrane potentials between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Only wildtype neurons were exposed to 1.8% isoflurane. Resting membrane potentials for wildtype at 0.6% are repeated here to facilitate comparison to 1.8% isoflurane. p-values are for the differences between the unexposed and exposed (isoflurane) resting membrane potentials for wildtype neurons and are given at the top of each panel. In both figures, p-values are for the differences between the unexposed and exposed (isoflurane) resting membrane potential for neurons and are given at the top of each panel. After Bonferroni corrections, p<0.025 was considered significant. Abbreviations: mV (milliVolts)

As a final measure of the effects of isoflurane on neuronal membranes, we tested whether isoflurane changed neuronal membrane resistance. In spinal cord neurons from both wildtype and Ndufs4 animals, 0.6% isoflurane decreased membrane resistance only in GABAergic neurons (Supplemental Figure S2A). However, in wildtype GABAergic neurons membrane potential was maintained at 0.6% isoflurane despite the decrease in resistance, implying that GABAergic cells compensate for increased outward potassium currents. In wildtype spinal cord neurons at 1.8% isoflurane membrane resistance again decreased only in GABAergic neurons (Supplemental Figure S2B).

Cell-type Specific Knockout of Ndufs4

We earlier reported that glutamatergic knockout of Ndufs4 increased volatile anesthetic sensitivity identically to that seen in the global knockout.^17,25 We therefore tested whether we could detect an isoflurane-induced rise in holding currents in of specific types of spinal cord neurons. In contrast to cholinergic neurons from the global Ndufs4 knockout, 0.6% isoflurane increased holding currents in neurons from cholinergic-specific knockouts of Ndufs4 (ΔHC 119.5(82.3,156.7),N=21, p=0.00019) in addition to a trend to increase in glutamatergic-specific knockouts (ΔHC, 85.5(49,126.9),N=22, p=0.064) (Figure 5A, Table 3) that matched the magnitude of change in those cells in global Ndufs4 knockouts. Neurons from GABAergic-specific knockout of Ndufs4 did not show an increase in holding current (ΔHC 18.3(5,38.1),N=7, p=0.74), unlike GABAergic neurons from a global Ndufs4 knockout (Figures 2A, 5A). The increase in holding current was blocked by norfluoxetine in glutamatergic neurons but not in cholinergic neurons (Figure 5B). In neurons with neuron type-specific loss of Ndufs4, there was a significant decrease in membrane potential with cholinergic-specific (ΔRMP,−5.3mV(−7.3,−3.2),N=21, p=0.00017) and glutamatergic-specific (ΔRMP,−3.6mV(−5.7,−2.1),N=22, p=0.0013) loss of Ndufs4 (Figure 5C). GABAergic-specific loss of Ndufs4 did not change resting membrane potential (ΔRMP,−0.9mV(−3.7,+1.8),N=7, p=0.56), in concordance with the lack of change in holding current (Figure 5A).

Figure 5. Holding currents in spinal cord neurons exposed to isoflurane from cell-specific knockout of Ndufs4 compared to CNS global knockout of Ndufs4.

A. Effects of 0.6% isoflurane on specific neuronal types in neuron-specific Ndufs4 spinal cord neurons. Plotted are the differences in holding currents between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Spinal cord neurons from wildtype animals are in grey, from global CNS knockout of Ndufs4 animals are in blue, cell-specific knockout of Ndufs4 are in red (cholinergic), green (GABAergic) or purple (glutamatergic). Some cholinergic data that was also presented earlier and are used here for comparison to the newer data.⁹ p-values for the difference between the effect on wildtype neurons and Ndufs4 neurons are given at the top of each panel. B. Norfluoxetine inhibits the current in glutamatergic but not cholinergic cells. Plotted are the differences in holding currents between isoflurane unexposed and exposed neurons in the cell-specific Ndufs4 knockouts. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Color code is same as in A. p-values are for the differences between the isoflurane effect with and without norfluoxetine and are given at the top of each panel. C. Paired absolute effects of 0.6% isoflurane on specific neuronal types in wildtype and Ndufs4 spinal cord neurons. Plotted are the paired absolute values for holding currents between isoflurane unexposed and exposed neurons. The neuronal types, each containing the fluorescent marker Ai14 (see Methods) are shown in the labels for each panel. Color code is same as in A. p-values are for the differences between the unexposed and exposed (isoflurane) holding current for wildtype neurons and Ndufs4 neurons and are given at the top of each panel. Abbreviations: mV (milliVolts), pA (picoAmperes), ChAT (choline acetyltransferase), Gad2 (glutamic acid decarboxylase), Vglut2 (vesicular glutamate transporter)

Synaptic Input

Holding current responses to isoflurane in cholinergic and GABAergic neurons differed between global Ndufs4 knockout animals and animals with neuron-specific loss of Ndufs4. We first considered whether the baseline holding currents, without isoflurane, might be significantly different between global and neuron-specific loss of Ndufs4. Baseline holding currents were not significantly different in GABAergic or glutamatergic neurons whether there was global or neuron-specific loss of Ndufs4 (Supplemental Figure S3). However, compared to global loss of Ndufs4, cholinergic neuron-specific loss of Ndufs4 did cause an increase in baseline holding current. As noted in Figure 5B, this cholinergic neuron-specific holding current was further increased by isoflurane and not inhibited by norfluoxetine, indicating a TREK-independent mechanism for the effect of isoflurane on these cells.

These results raised the question of whether synaptic input varied in a way that might explain the different responses between global Ndufs4 knockout and neuron type-specific Ndufs4 knockout. We examined spontaneous excitatory postsynaptic currents (sEPSCs) and inhibitory postsynaptic currents (sIPSCs) in both global and neuron type-specific loss of Ndufs4. Baseline sEPSCs did not differ between wildtype and Ndufs4(KO) neurons in any neuron type. The frequencies of sEPSCs were unchanged between unexposed and 0.6% isoflurane in all neuron types from wildtype and global Ndufs4 knockout neurons (Supplemental Figure S4). In neuron-specific Ndufs4 knockouts at 0.6% isoflurane, there was an increase of sEPSC frequencies only in glutamatergic cells (Supplemental Figure S4). While statistically significant, we do not believe that this increase is biologically significant.

No biologically significant differences were obtained for isoflurane effects on inhibitory postsynaptic current frequencies in any neuron type for either wildtype or Ndufs4 cells (Supplemental Figure S5). Similarly, no statistically significant differences in sEPSC or sIPSC amplitudes were seen in any neuron type (Supplemental Figures S6,S7) although sEPSC amplitudes approached a significant increase in wildtype GABAergic cells. Decay times for inhibitory postsynaptic currents were increased by isoflurane both in wildtype and Ndufs4 GABAergic and glutamatergic neurons but not in cholinergic neurons (not shown). These increases occurred at both 0.6% isoflurane and 1.8% isoflurane.

Discussion.

We previously showed that presence of a single protein, NDi1, increases MAC for both isoflurane and halothane in both wildtype and Ndufs4 mice.²⁴ NDi1 bypasses native mitochondrial complex I and is completely resistant to inhibition by up to 2.4% isoflurane (Supplemental Figure S1). We hypothesized that the presence of NDi1 in neurons would lessen the induction of potassium currents by isoflurane both in wildtype and Ndufs4. We found that the presence of NDi1 eliminates the isoflurane-dependent induction of outward potassium currents in both genotypes and completely rescues the resulting hyperpolarization. This effect is seen even though the native, anesthetic sensitive complex I is also present in the spinal cord in wildtype mice and the hypersensitive complex I remains present in Ndufs4 mice. This finding represents the most significant discovery of this investigation since it indicates the powerful effect of retained electron flow (via NDi1) through the mitochondrial respiratory chain in the face of isoflurane inhibition of normal (wildtype) or hypersensitive (Ndufs4) complex I.

Volatile anesthetic inhibition of mitochondrial complex I inhibition has already been shown to play a major role in determining anesthetic sensitivity in multiple organisms.^{17–20,24,28} Our studies of hippocampal neurons in culture demonstrated that complex I inhibition by isoflurane decreased presynaptic ATP which, in turn, caused a defect in endocytosis of synaptic vesicles.²⁴ However, mitochondrial inhibition could affect a wide range of cellular functions that would have many downstream effects. We previously showed that enhancement of TREK channel conductance appeared to be one such targeted function.⁹ This finding was of particular interest since prior studies indicated that TREK channel activity may be linked to anesthetic sensitivity.¹⁰

Multiple reports have shown that volatile anesthetics enhance an outward potassium leak current in neurons. Such currents are predicted to cause a neuronal hyperpolarization and neuronal quiescence; thus, they are consistent with an anesthetic effect of depressing neuronal function. In neurons from wildtype mice and rats, these currents are attributed to potassium K2P leak channels known as TREK channels.^9,11,14 Our previous work was consistent with those findings and indicated that isoflurane-induced currents in both wildtype and Ndufs4 neurons were carried by TREK channels. However, we were surprised to find that those currents were enhanced in the spinal cord neurons from Ndufs4 mutants at lower concentrations of isoflurane than in those from wildtype mice⁹ identifying a novel link between complex I dysfunction and neuronal hyperpolarization by isoflurane.

The mechanism by which complex I inhibition affects TREK channel conductance is not directly addressed in this study. Phosphorylation of TREK channels has previously been proposed to both increase²⁹ and decrease³⁰ leak currents. Our data are most consistent with, but do not prove, the latter. We suggest that mitochondrial inhibition leads to decreased ATP levels which, in turn, decreases TREK channel phosphorylation and increases TREK channel conductance. Proof of this model is under investigation. Regardless of the molecular changes ultimately affecting membrane potential, the dependence of potassium channel conductance on complex I inhibition by isoflurane extends our understanding of the mechanisms by which mitochondrial function affects volatile anesthetic sensitivity of the whole animal.

The second major point from this study concerns neuron-type specific loss of Ndufs4. Since glutamatergic loss of Ndufs4 conferred marked reduction of whole animal MAC,²⁵ we hypothesized that isoflurane induced hyperpolarizing currents would be limited to glutamatergic cells. However, our model was incorrect; both GABAergic and glutamatergic spinal cord neurons have increased holding currents in the presence of low concentrations (0.6%) of isoflurane with global loss of Ndufs4. In slices from the global Ndufs4 knockout, the currents in both GABAergic and glutamatergic neurons were inhibited by norfluoxetine implicating TREK channels as causative in both neuron types. Interestingly, the other membrane properties that correlate with changes in holding current (membrane resistance and transmembrane potential) were only significantly altered in GABAergic neurons (although glutamatergic neurons approached significance), again indicating that the precise molecular changes differ between GABAergic and glutamatergic neurons (Figure 4B). These complicated and puzzling results were unanticipated and demonstrate the complexity of interactions within the mammalian CNS.

We then compared inducible currents in neuronal types affected in global Ndufs4 mutants (described in the previous paragraph) to those in which Ndufs4 is mutant only in specific neuronal types. This was of interest since only knockout of Ndufs4 in glutamatergic neurons caused a change in whole animal sensitivity.²⁵ The changes seen with neuron type-specific knockout of Ndufs4 were also unanticipated, and even more so when compared to the same neuron types from global Ndufs4 knockouts. We expected that restricting loss of Ndufs4 to glutamatergic neurons would specifically increase isoflurane induced holding currents, while loss in other neuronal types would not. However, both cholinergic-specific and glutamatergic-specific knockout of Ndufs4 led to holding current increases. In the global knockout, glutamatergic and GABAergic cells displayed this phenotype. Notably, the enhanced glutamatergic-specific increase in holding currents induced by isoflurane in global knockouts was maintained with glutamatergic-specific knockout of Ndufs4. The changes seen in cholinergic-specific neurons, while striking, do not correlate with behavioral changes²⁵ and are not likely to be part of the anesthetic phenotype.

In contrast, glutamatergic-specific knockout of Ndufs4 does change the anesthetic sensitivity of the animal,²⁵ and it also changes resting membrane potential (Figure 5C). These data are consistent with altered potassium currents playing a role in anesthetic sensitivity in the Ndufs4 mouse. Potassium currents are increased in all three types of wildtype neurons with 1.8% isoflurane, leading to a decreased membrane potential which may play a role in the anesthetic sensitivity of wildtype animals. Since only glutamatergic-specific knockout of Ndufs4 also alters whole animal anesthetic sensitivity, the glutamatergic cell-specific currents (with increased hyperpolarization) remain candidates for part of the effects of volatile anesthetics causing loss of responsiveness to tactile stimulation (MAC) in Ndufs4 and wildtype mice.^8,10

The increased currents in glutamatergic neurons, whether with global or cell-specific loss of Ndufs4, is consistent with this phenomenon being a cell autonomous effect. The effects of the spinal cord Ndufs4 mutation on potassium currents in GABAergic and cholinergic cells are not cell autonomous since the results depend on the genetic background of other neurons. We interpret these results to indicate that, if the potassium currents play a role, it most likely is in glutamatergic neurons.

There are limitations to our results of cell-specific loss of Ndufs4 and their interpretations. First, we did not test glycinergic-specific or glial-specific loss of Ndufs4. We also did not set up the study to differentiate between sexes for responses, but did test cells from both sexes. We did not detect any sex-specific differences, nor did we have study arms that would have made such differences likely. Finally, we have not determined what causes the currents to be different in the cell-specific Ndufs4 knockout setting compared to global knockout of Ndufs4. While it is tempting to invoke network effects leading to these changes, we did not see significant synaptic changes in our electrophysiologic studies to indicate a specific mechanism. However, the roles of other neuron types (glycinergic neurons, glial cells) and other effectors of transmembrane potential will need to be considered, ideally in a behavioral context like locomotion or the righting reflex.

There are also limitations in interpreting the results from expression of NDi1. Since NDi1 rescues the inhibition of complex I by isoflurane, electron flow through the respiratory chain and ATP production will both be improved. However, this will also allow for NAD⁺/NADH ratios to be normalized which could also affect neuronal activity. In previous studies, we showed that this effect did not rescue a failure of synaptic activity caused by isoflurane, but we have not shown whether NAD⁺/NADH ratios affect isoflurane-induced increases of potassium currents. In addition, the presence of NDi1 may have other, off target and unpredicted effects that affect TREK channels. Clarification of these possibilities will require continued studies.

The simplest interpretation of the results is that, since NDi1 eliminates the increase in holding currents in the spinal cord, inhibition of mitochondrial complex I is a necessary step in inducing the increase in potassium flux by an undetermined mechanism. As shown previously^8,9, norfluoxetine blocks these currents; thus, they are normally carried by TREK channels. However, in the absence of TREK channels, other potassium channels may perform this function.⁸ The behavioral data indicate that the primary neuron type in which these induced currents may play a role in contributing to MAC are the spinal cord glutamatergic neurons. The induced currents may occur in other spinal cord neuron types, but they do not correlate with behavioral anesthetic sensitivity. In conclusion, our results link mitochondrial complex I inhibition and inducible potassium channels to MAC in whole animals.

Funding.

MMS, PGM and JMR were supported in part by NIH grant R01GM105696. EBK was supported in part by the Northwest Mitochondrial Research Guild. PGM, MS, CW, MH, and BP were supported in part by NIH grant R35GM139556. CRR was supported in part by NIH Grant NHLBI T32 HL076139–11.

Supplementary Material

_1

1. Supplemental Figure S1. Oxidative phosphorylation by NDi1 in isoflurane.

_2

2. Supplemental Figure S2. Effects of isoflurane on input resistance.

_3

3. Supplemental Figure S3. Unexposed baseline holding currents.

_4

4. Supplemental Figure S4. Effects of isoflurane on excitatory frequencies.

_5

5. Supplemental Figure S5. Effects of isoflurane on inhibitory currents.

_6

6. Supplemental Figure S6. Effects of isoflurane on excitatory amplitudes.

_7

7. Supplemental Figure S7. Effects of isoflurane on inhibitory amplitudes.

Supplemental Digital Content

8. Supplemental Materials. ARRIVE Guidelines.

9. Supplemental Excel File. Spinal Cord Statistics.

Abbreviations:

mV

milliVolts

pA

picoAmperes

ATP

Adenosine triphosphate

ChAT

choline acetyltransferase

Gad2

glutamic acid decarboxylase

Vglut2

vesicular glutamate transporter

NF

Norfluoxetine

TREK

two-pore domain potassium (K⁺) channels