Free radical signalling underlies inhibition of CaV3.2 T-type calcium channels by nitrous oxide in the pain pathway


Corresponding author S. M. Todorovic: Department of Anaesthesiology, University of Virginia Health System, Mail Box 800710, Charlottesville, VA 22908-0710, USA. Email:

Non-technical summary

Nitrous oxide (N2O) has long been used as a pain reliever, but little is known of its targets in the body. We show that N2O indirectly inhibits T-type calcium channels through free radical reactions. These reactions depend on a histidine residue on the channel that binds metal ions. If we prevent this histidine from binding metals, N2O cannot inhibit T-currents. Mice that are treated with EUK-134 to remove free radicals show little pain relief after N2O administration. This report provides new information on how N2O interacts with ion channels, helping our understanding of how this popular pain reliever works.


Nitrous oxide (N2O, laughing gas) has been used as an anaesthetic and analgesic for almost two centuries, but its cellular targets remain unclear. Here, we present a molecular mechanism of nitrous oxide's selective inhibition of CaV3.2 low-voltage-activated (T-type) calcium channels in pain pathways. Using site-directed mutagenesis and metal chelators such as diethylenetriamine pentaacetic acid and deferoxamine, we reveal that a unique histidine at position 191 of CaV3.2 participates in a critical metal binding site, which may in turn interact with N2O to produce reactive oxygen species (ROS). These free radicals are then likely to oxidize H191 of CaV3.2 in a localized metal-catalysed oxidation reaction. Evidence of hydrogen peroxide and free radical intermediates is given in that N2O inhibition of CaV3.2 channels is attenuated when H2O2 is neutralized by catalase. We also use the adrenochrome test as an indicator of ROS in vitro in the presence of N2O and iron. Ensuing in vivo studies indicate that mice lacking CaV3.2 channels display decreased analgesia to N2O in response to formalin-induced inflammatory pain. Furthermore, a superoxide dismutase and catalase mimetic, EUK-134, diminished pain responses to formalin in wild-type mice, but EUK-134 and N2O analgesia were not additive. This suggests that reduced ROS levels led to decreased inflammation, but without the presence of ROS, N2O was not able to provide additional analgesia. These findings reveal a novel mechanism of interaction between N2O and ion channels, furthering our understanding of this widely used analgesic in pain processing.


dorsal root ganglia


diethylenetriamine pentaacetic acid




human embryonic kidney cells


metal-catalysed oxidation


reactive oxygen species


superoxide dismutase


T-channel, low voltage-activated calcium channel


N2O has been used as an anaesthetic and excellent analgesic for close to two centuries in clinical practice. It is also a common drug of abuse. In spite of this, the molecular mechanisms underlying its clinical effects are poorly understood, partly because of difficulties in working with this agent in vitro. Ion channels as targets for N2O have only recently been taken into consideration. The first papers describing specific interactions between N2O and N-methyl-d-aspartate (NMDA) receptors in the CNS (Jevtovic-Todorovic et al. 1998; Mennerick et al. 1998) have begun to explore this mechanism. These effects may be the basis of N2O anaesthesia since reduction of excitatory drive in CNS neurons is one of the key theories of anaesthesia. The first study of the effects of N2O on ion channels in the pain pathway showed that N2O inhibition of T-type channels (T-channels) is specific for the CaV3.2 isoform (Todorovic et al. 2001b). This isoform is abundantly expressed in the major pain processing regions of the peripheral nervous system (PNS) and central nervous system (CNS): the peripheral sensory neurons and superficial layers of the dorsal horn in the spinal cord (Talley et al. 1999). In addition, CaV3.2 channels were directly implicated in pain using both knock-out and knock-down strategies (Bourinet et al. 2005; Choi et al. 2007), strongly suggesting that at least part of N2O's potent analgesia could be mediated by blockade of CaV3.2 T-channels. However, a mechanistic link between these two phenomena has not been established.

Much of the research about the interaction of N2O and metals has been in bacterial systems, such as the clusters of copper and iron bound by histidine residues in N2O reductase and the membrane protein NosR (Brown et al. 2000; Wunsch & Zumft, 2005). While another study found evidence of free radical intermediates during N2O metabolism in human intestinal contents (Bosterling et al. 1980), no one has been able to establish an interaction between N2O and metal ions in a mammalian system.

Reactive oxygen species (ROS) have been implicated in a number of pathological conditions, including chronic and inflammatory pain (Chung, 2004). Intraplantar and intrathecal application of free radical scavengers have been shown to reduce oxidative stress, inflammation and pain (Salvemini et al. 1999; Nishiyama & Ogawa, 2005). Synthetic superoxide dismutase (SOD) and catalase mimetics, such as EUK-134, have been shown to prevent oxidative stress and have neuroprotective properties in ischaemic brain injury (Baker et al. 1998; Rong et al. 1999). By minimizing the ROS level, these agents are believed to prevent the induction of stress-related gene responses, as well as reducing molecular damage, such as lipid peroxidation and protein oxidation. Affected proteins may also include potassium channels and NMDA receptors, which are known to have altered responses after exposure to ROS (Aizenman et al. 1990; Aizenman, 1995; Ichinari et al. 1996).

Until recently, very little was known about the role of T-channels in sensory processing, specifically in nociception. However, recent data indicate that these channels may play an important role in boosting peripheral and central sensory transmission (reviewed in Todorovic & Jevtovic-Todorovic, 2006; Nelson et al. 2007b). Thus, sensory neurons provide an excellent in vitro model for studies of T-channels in nociception. Here, we use recombinant and native CaV3.2 channels expressed in sensory neurons to study the molecular mechanisms of selective inhibition of these channels by N2O, a commonly used anaesthetic and analgesic.

Ethical approval

Ethical approval was obtained for all experimental protocols from the University of Virginia Animal Care and Use Committee, Charlottesville, VA, USA. All experiments were conducted in accordance with the Guide for the Care and Use of Laboratory Animals adopted by the US National Institutes of Health. Every effort was made to minimize animal suffering and the number of animals used. For removal of tissues, animals were deeply anaesthetized with inhaled isoflurane and rapidly killed with a guillotine.

HEK-293 cells

For recordings of recombinant T-currents, we used both transiently transfected and stably expressed CaV3.1 or CaV3.2 channels in human embryonic kidney (HEK-293) cells as reported previously (Joksovic et al. 2006). HEK-293 cells (CRL-1573; American Type Culture Collection, Manassas, VA, USA) were grown in DMEM/F-12 media (Invitrogen) supplemented with 10% fetal bovine serum, penicillin G (100 mg ml−1) and streptomycin (0.1 mg ml−1). Transiently transfected cell lines were made using cDNA encoding human CaV3.1 or CaV3.2, and a plasmid encoding the CD8 antigen. Cells were then incubated with polystyrene microbeads coated with CD8 antibody (Dynabeads M-450 CD8, Dynal GmbH, Hamburg, Germany). After ∼48 h, cells with microbeads bound were selected for electrophysiology.

DRG neurons

Three- to eight-week-old Sprague–Dawley rats were briefly anaesthetized with isoflurane and decapitated. Dorsal root ganglia were removed and incubated in a mixture of collagenase H (2 mg ml−1, Roche) and Dispase I (0.17 mg ml−1, Roche) for 1 h at 37°C. Ganglia were dissociated using Pasteur pipettes and plated on glass coverslips. Neurons were used for electrophysiology 2–8 h later.

Nitrous oxide preparation

Nitrous oxide solutions were prepared by bubbling 80% N2O and 20% O2 (GT&S Welco, Richmond, VA, USA) through a bubbling stone into external recording solutions. N2O concentration in the gas mixture was constantly monitored by an infrared gas analyser (Datex-Engtstrom, Helsinki, Finland). During experiments, test solutions were maintained in tightly sealed, all-glass syringes, and replaced every 30 min. However, due to extremely low solubility of N2O in water, it is likely that actual concentrations of N2O in our in vitro experiments were much lower than 80%.

Other drug preparations

Stock solutions of all drugs were made within 48 h of use. Agents were dissolved in appropriate solvents as follows: formalin (Fisher Scientific), 4% paraformaldehyde in saline; deferoxamine (Calbiochem), 10 μm in external solution; DTPA (Sigma), 100–300 μm in external solution; H2O2 (Fisher Scientific), 0.1% by volume in deionized water for adrenochrome test or external solution for electrophysiology; catalase (Sigma), 500 U ml−1 in external solution; adrenaline (epinephrine; Sigma) in deionized water; EUK-134 (Cayman Chemical), 2 mg ml−1 in saline.


All recordings were made using standard whole-cell techniques. Electrodes were fabricated from borosilicate microcapillary tubes (Drummond Scientific, Broomall, PA, USA) and fire-polished to final resistances of 2–3 MΩ. Voltage commands and digitization of membrane voltages and currents were done with Clampex 9.2 of the pClamp software package (Axon Instruments), running on an IBM-compatible computer. Membrane voltages and currents were recorded with an Axopatch 200B patch-clamp amplifier (Axon Instruments). Data were analysed using Clampfit 9.2 (Axon Instruments) and Origin 7.0 (Microcal Software). Currents were filtered at 2 kHz. Series resistance (Rs) and membrane capacitance (Cm) values were taken directly from readings on the amplifier following electronic subtraction of the capacitative transients. We compensated for 50–80% of Rs. Cells were held (Vh) at −90 mV and depolarized to a test potential (Vt) of −30 mV every 10–20 s to evoke inward Ba2+ currents, unless otherwise noted. In most experiments, a P/N protocol was used for online leak subtraction.

Multiple independently controlled glass syringes served as reservoirs for a gravity-driven local perfusion system, where manually controlled valves allowed for switching between syringes. Solution exchange was accomplished by constant suction through a glass capillary tube at the opposite end of the recording dish. All drugs were prepared as stock solutions and freshly diluted to appropriate concentrations at the time of experiments. The external recording solution used to isolate Ba2+ currents contained (in mm): BaCl2, 2–10; TEA-Cl, 152; Hepes, 10; adjusted to pH 7.4 with TEA-OH. In a limited number of experiments we used 2 mm Ca2+ in our external recording solution. We did not notice any difference in the effects of N2O, metal chelators or H2O2 regardless of which external solution was used. The internal solution contained (in mm): caesium methane sulfonate, 110; phosphocreatine, 14; Hepes, 40; Mg-ATP, 2; Na-GTP, 0.1; adjusted to pH 7.2 with KOH.

Oxidation of adrenaline to adrenochrome

Detection of adrenochrome production was performed using a spectrophotometer (ThermoSpectronic, Rochester, NY, USA). Optical absorbance at 485 nm was measured every 5 min over a 30 min period. Reagents were placed in a 2 ml sealed glass cuvette and immediately inserted into the spectrophotometer for data collection. All measurements were corrected against blanks containing 20 mm epiHCl (Sigma) in 0.1 m HCl and 50 mm Hepes/KOH buffer (pH 10). To the blanks, we added the required amounts of FeCl3, H2O2 or 80% N2O. 80% N2O was bubbled into Hepes/KOH buffer for 30 min prior to experiments.

Data analysis and statistical procedures

Peak currents and exponential fits to currents were calculated using Clampfit, while curve fitting was performed using Origin 7 software. For general statistical evaluations, we used one- and two-way ANOVAs with values <0.05 judged to be statistically significant, and paired or unpaired Student's t tests as appropriate. Average data are presented as mean ± standard errors of the mean (s.e.m.).

Behavioural studies

Adult mice of either sex were used for all behavioural studies. CaV3.2 knock-out mice (CaV3.2−/−) and their age-matched wild-type littermates (CaV3.2+/+) were studied at 16 to 60 weeks of age. The initial breeding pair for our colony of CaV3.2−/− mice was obtained through the courtesy of Dr Kevin Campbell (University of Iowa, Iowa City, IA, USA). These mice were repeatedly backcrossed for at least six generations against wild-type (C57BL/65J) mice. Animals were genotyped using tail biopsy. Mutant mice were normal in appearance, weight and overt behaviour. They demonstrated a brisk righting reflex and in a previous study showed no differences in motor abilities under control conditions (Choi et al. 2007).

Behavioural tests for inflammatory pain after formalin injection were performed in a clear Plexiglass chamber prefilled with 20% N2O, 20% O2 and 60% N2 at a flow rate of 6 l min−1. The mouse was first placed in the prefilled chamber to accommodate for 30 min. The mouse was then rapidly removed from the chamber, injected with 20 μl of 4% formalin or saline into the plantar side of the right paw and returned to the test chamber. The time in seconds spent licking and biting the paw was measured for an hour and recorded for every 5 min interval. Afterwards, the mouse's temperature was measured to ensure normothermia. Control mice were tested under the same protocol, with normal room air flow instead of the N2O mixture.

To test the effects of EUK-134 (EUK; Cayman Chemical), 5 mg of EUK was dissolved in 2.5 ml of saline by vortexing for 2.5 h. A mouse was injected intraperitoneally (i.p.) with fresh EUK solution (25 mg kg−1). Thirty minutes after the EUK injection, the mouse was placed inside the chamber to equilibrate and familiarize itself with the environment. One hour after the EUK injection, the mouse was injected with formalin, and the rest of the experiment was performed as described above. Control mice received equal volumes of saline in an i.p. injection.

The effects of EUK on inflammation were quantified after i.p. injection of EUK. After 1 h following i.p. saline or EUK injections, control paw volume was measured by insertion into a beaker of water on a weight scale (Fereidoni et al. 2000). During this time, the mouse is mechanically restrained and fastened to an immobile structure. Afterwards, formalin was injected, and paw volume measured again after 1 h. The difference of the two volumes was then calculated.


N2O is more effective at reducing inflammatory pain in wild-type mice than mice lacking CaV3.2 channels (CaV3.2−/−)

Previous studies have shown that CaV3.2−/− mice have attenuated pain responses (Choi et al. 2007). Thus, to explore the interaction of N2O and CaV3.2 channels in vivo, we used injections of formalin into the hind paws of mice as a commonly used model for inflammatory pain. The amount of time animals spend licking and biting the injected paw in the first 5 min results from direct activation of peripheral nociceptors (phase 1, P1). In contrast, responses in the 10–60 min period following injection reflects the central sensitization of pain responses at the level of the spinal cord (phase 2, P2). For these in vivo experiments, we used sub-anaesthetic concentrations of N2O to avoid unwanted sedation. As shown in Fig. 1A, 20% N2O significantly reduced licking and biting of the affected paw in all measures of wild-type animals (total time, P1, P2). In contrast, CaV3.2−/− mice demonstrated greatly reduced, though still significant, analgesia to N2O in phase 1 while total time and phase 2 were not significantly affected by N2O (Fig. 1B). As expected, CaV3.2−/− mice had significantly decreased basal pain responses overall to injections of formalin (Choi et al. 2007). In another control experiment, we injected morphine (i.p., 5 mg kg−1) 30 min before formalin injection and assessed its effect on nociceptive responses. Figure 1C indicates that morphine had significant effects in all pain measures in wild-type mice similar to N2O (Fig. 1A). However, in contrast to N2O, morphine also significantly decreased all nociceptive responses in CaV3.2−/− mice (Fig. 1D). This strongly suggests that reduced N2O analgesia in CaV3.2−/− mice is not due to decreased basal pain responses. Thus, inhibition of CaV3.2 may at least partly be responsible for analgesic effects of N2O in this inflammatory pain model.

Figure 1.

CaV3.2−/− mice experience decreased N2O analgesia but similar morphine analgesia after formalin injection into hindpaws
A, WT littermates experience significant decreases in licking/biting time in the presence of 20% N2O (n= 16) compared to air controls (n= 12): total: air, 480.1 ± 18.4 s; N2O, 351.4 ±18.3 s; phase 1: air, 114.9 ± 5.0 s; N2O, 41.6 ± 4.6 s; phase 2: air, 365.3 ± 16.6 s; N2O, 309.8 ± 18.5 s. B, CaV3.2−/− mice show decreased licking/biting time in the presence of 20% N2O (n= 15) compared to air controls (n= 12) in phase 1, but not phase 2 or overall: total: air, 293.9 ± 14.7 s; N2O, 266.3 ± 15.8 s; phase 1: air, 64.6 ± 2.8 s; N2O, 37.1 ± 3.2 s; phase 2: air, 229.3 ± 13.9 s; N2O, 229.2 ±16.6 s. C, WT littermates experience significant decreases in licking/biting time in the presence of morphine compared to air controls (n= 12): total: air, 480.1 ± 18.4 s; morphine, 345.0 ± 15.9 s; phase 1: air, 114.9 ± 5.0 s; morphine, 55.8 ± 4.3 s; phase 2: air, 365.3 ± 16.6 s; morphine, 289.2 ± 13.3 s. D, CaV3.2−/− mice show decreased licking/biting time in the presence of morphine compared to air controls (n= 12) in phase 1, phase 2 and overall: total: air, 293.9 ± 14.7 s; morphine, 221.2 ± 10.2 s; phase 1: air, 64.6 ± 2.8 s; morphine, 33.7 ± 2.0 s; phase 2: air, 229.3 ± 13.9 s; morphine, 187.4 ± 9.1 s. Note that basal pain responses (filled bars in A) are largely decreased in CaV3.2−/− mice than in WT littermates (filled bars in B). *Significant difference (P < 0.05).

N2O-mediated inhibition of CaV3.2 is dependent on H191

Previously, we reported that inhibition of T-currents by N2O is selective for the CaV3.2 isoform compared to the CaV3.1 (Todorovic et al. 2001b) and CaV3.3 isoforms (Joksovic et al. 2005). Interestingly, the specificity of N2O for CaV3.2 T-channels mirrors that of trace metals such as nickel (Kang et al. 2006) and zinc (Nelson et al. 2007b), as well as redox agents such as ascorbate (Nelson et al. 2007a) and l-cysteine (Todorovic et al. 2001a; Joksovic et al. 2006; Nelson et al. 2007b), which all converge to a unique H191 residue in the IS3–4 linker of CaV3.2. Thus, we tested the possibility that the selective inhibition of CaV3.2 by N2O is also due to interaction with H191. To examine the means of this specificity, we tested the effects of 80% N2O on point mutations of this histidine residue (Fig. 2A). The results are summarized with representative traces (left panels of Fig. 2B) and averaged values from multiple experiments are presented in Fig. 2C. Mutation of H191 to a glutamine residue (CaV3.2/H191Q) largely abolishes N2O inhibition of the channel. In contrast, the reverse mutation of a corresponding glutamine to histidine in a N2O-insensitive isoform, CaV3.1, produces a gain of N2O sensitivity. In this CaV3.1/Q172H mutation, 80% N2O is able to inhibit 30% of calcium current, compared to only 4% of wild-type CaV3.1 currents (data not shown). A third mutation was created to determine the specific role of histidine in N2O inhibition by replacing it with another metal-binding amino acid, cysteine. Interestingly, N2O produced greatly attenuated inhibition of CaV3.2/H191C (14% inhibition), suggesting a need for the histidine amino acid residue at this position for its maximal effect. In contrast, these point mutations had no effect on isoflurane's isoform non-specific inhibition of T-type channels, verifying the integrity of the mutant channel structures (representative traces are depicted on the right side of Fig. 2B and average histograms are shown in Fig. 2D). Similarly, we have shown that these point mutations do not alter basic kinetic properties of CaV3.2 currents (Nelson et al. 2007a,b).

Figure 2.

Histidine 191 is required for N2O-mediated inhibition of CaV3.2 channels
A, schematic diagram of histidine 191 in the CaV3.2 channel. B, traces from representative HEK cells before (black) and after (grey) application of 80% N2O (left) or 300 μm isoflurane (right). C, histogram showing inhibition by 80% N2O on wild-type CaV3.2 (n= 3), 35.8 ± 3.2%; H191Q (n= 7), 3.3 ± 2.7%; Q172H (n= 5) 29.9 ± 5.4%; H191C (n= 7), 14.2 ± 2.5%. Air controls were pooled from all mutations (n= 8), 3.4 ± 0.8%. D, histogram showing inhibition by 300 μm isoflurane on wild-type CaV3.2 (n= 6), 48.0 ± 8.5%; H191Q (n= 7), 54.3 ± 2.5; Q172H (n= 6), 54.6 ± 2.4; H191C (n= 6), 60.2 ± 4.9. Wild-type CaV3.2 data are pooled from HEK and DRG cells. Vertical lines represent s.e.m., * indicates significant difference from WT CaV3.2 (P < 0.05), N.S., not significant. All values in isoflurane experiments were not statistically different between WT and mutant channels. Inset on the right bottom side of this figure indicates voltage-clamp protocol used in these experiments.

Metal chelators attenuate N2O-mediated inhibition of CaV3.2 currents

Previously, we used specific chelators to demonstrate that H191 of CaV3.2 may bind endogenous trace metal ions such as zinc (Nelson et al. 2007a) and copper (Nelson et al. 2007b). However, the most abundant trace metal ion in the body is iron (Freitas, 1999) and very little is known about interactions of iron with ion channels, T-channels in particular. Thus, we explored the possibility that endogenous iron may tonically inhibit CaV3.2 currents via H191. When a high-affinity iron-specific chelator, deferoxamine (DFX, 10 μm), was applied, CaV3.2 currents increased to over 150% of control, probably demonstrating removal of tonic block (Fig. 3A). The left panel of Fig. 3A shows traces from representative cells, and the right panel shows time courses from the same HEK cell. Furthermore, this effect is isoform specific, as CaV3.1 currents do not significantly increase in the presence of DFX (Fig. 3B, left). Upon closer examination, we found that mutant H191Q CaV3.2 currents also do not significantly increase when exposed to DFX, supporting a role of histidine 191 in binding the iron (average histograms depicted in Fig. 3B, right).

Figure 3.

Metal chelators significantly reduce inhibition of CaV3.2 currents by N2O
A, left, traces from a representative HEK cell transfected with WT CaV3.2 channels before (black) and after (grey) application of 10 μm DFX. Right, time course of DFX application on the same cell. Horizontal bar indicates time of treatment. B, left, traces from an HEK cell transfected with CaV3.1 channels before (black) and after (grey) DFX. Right, histogram showing the effects of DFX on wild-type and mutated channels, relative to control: CaV3.2 (n= 5), 152.6 ± 14.2%; CaV3.1 (n= 5), 103.6 ± 6.1%; H191Q (n= 4), 106.8 ± 3.2%. Vertical lines represent s.e.m. *Significant difference from CaV3.2, P < 0.05. C, left, histogram comparing the effects of 300 μm DTPA alone and in the presence of 80% N2O on recombinant CaV3.2 channels (n= 7), relative to control currents: DTPA alone, 151.9 ± 10.5%; DTPA and N2O, 147.4 ± 10.8%. N.S., not significant. Vertical lines represent s.e.m. Right, time course of DTPA and N2O application from a representative cell. Horizontal bars indicate times of treatment.

Based on our experiments with H191Q, we theorized that an endogenous metal ion, bound to CaV3.2 channels, is necessary for the effects of N2O on the channel. However, since H191 is a high-affinity metal binding site of CaV3.2, capable of interacting with other endogenous metals such as zinc (Nelson et al. 2007a) and copper (Nelson et al. 2007b), we used the non-specific chelator diethylenetriamine pentaacetic acid (DTPA) to test this hypothesis. As shown in the average histograms (left panel, Fig. 3C) and representative time course (right panel), our hypothesis was confirmed when DTPA (300 μm) alone produced on average about 50% increased amplitude of basal T-current and greatly reduced ability of N2O to inhibit CaV3.2 currents (only about 4% inhibition of the new baseline, P > 0.05). In contrast, after DTPA washout, N2O inhibited 35% of basal T-current amplitude in the same cells (right panel, Fig. 3C).

Free radical moieties are needed for N2O-mediated inhibition of CaV3.2 currents

Iron, like copper, is a redox-reactive transition metal, capable of promoting free radical reactions. Thus, we explored the possibility that ROS play a role in N2O inhibition. ROS are extremely reactive compounds, capable of oxidizing proteins such as ion channels and altering their function (Stadtman & Levine, 2003; Gao et al. 2007). In the presence of iron (Fe2+ or Fe3+), hydrogen peroxide (H2O2) is transformed into the hydroxyl radical (·OH) through the Fenton reaction:

display math
display math

We reasoned that if N2O inhibits T-currents via metal-catalysed oxidation (MCO), its effects should be mimicked and occluded by an agent known to generate ROS. Accordingly, 0.1% H2O2 partially inhibited CaV3.2 currents and the inhibitory effects of H2O2 and N2O are not additive. This suggests that they share a common pathway of interaction with CaV3.2 (Fig. 4A and B).

Figure 4.

ROS are required for N2O-mediated inhibition of CaV3.2 channels
A, histogram comparing the effects of 0.1% H2O2 alone and in the presence of 80% N2O in the same DRG cells (n= 4–7), relative to control currents: H2O2 alone, 76.26 ± 1.8%; H2O2 and N2O, 74.3 ± 2.4%; wash, 108.0 ± 8.2%; N2O alone, 73.4 ± 1.1%. B, time course of H2O2 and N2O application in representative DRG cell. Horizontal bars indicate times of treatment. C, histogram showing the effects of catalase (CAT) and N2O relative to control levels in the same HEK cells (n= 5): CAT alone, 109.7 ± 1.5%; CAT and N2O, 92.9 ± 4.4%; wash, 94.0 ± 3.4%; N2O alone, 51.1 ± 14.8%. D, time course of peak T-current amplitude from a representative HEK cell where bars indicate times of drug application: control, 500 U ml−1 CAT alone, CAT and 80% N2O, wash, and 80% N2O alone. *Significant difference from control basal current (P < 0.05). **Significant difference from control basal current (P < 0.01). N.S., not significant.

To confirm this hypothesis, we performed the reverse experiment. To reduce the number of ROS available, we took advantage of an endogenous enzyme, catalase. Catalase catalyses the decomposition of H2O2 to water and oxygen, reducing the subsequent production of free hydroxyl radicals. When catalase (500 U ml−1) was applied in the bath alone, it caused a small increase (∼10%) in basal T-current amplitude (Fig. 4C and D). Importantly, when co-applied with catalase, 80% N2O was only able to inhibit 15% of the new CaV3.2 current baseline. In contrast, after catalase washout, N2O inhibited 40% of basal T-current amplitude in the same cells (Fig. 4C and D). These experiments further emphasize the key role that ROS have in the ability of N2O to affect CaV3.2 channels.

N2O causes the generation of ROS in the presence of transition metal ions

Next, we tested the hypothesis that N2O may generate ROS in the presence of redox-reactive endogenous trace metals such as iron using the oxidation of adrenaline to adrenochrome as a model of protein oxidation. The adrenochrome reaction has been widely employed as a standard assay for ROS production (Valerino & McCormack, 1971; Misra & Fridovich, 1972; Loschen et al. 1974; Ahmed et al. 1993). Adrenaline is a colorless liquid, but its primary oxidation product, adrenochrome, is a red–orange colour (Fig. 5A). Changes in the peak absorbance (485 nm) of adrenochrome indicate successful oxidation of adrenaline, and thus, the occurrence of ROS. In Fig. 5B, we show that 0.1% H2O2 is unable to produce free radicals efficiently without the presence of trace metals, and that the rate of adrenochrome production is H2O2-dose dependent. We also found significant oxidation of adrenaline in the presence of 80% N2O and 10 μm FeCl3, but not N2O alone (Fig. 5C). It is important to note that the rates of adrenochrome production in H2O2+ FeCl3 and N2O + FeCl3 are very similar for the first 10 min. Because of the gaseous properties of N2O, it soon comes out of solution and is unable to participate in any further aqueous reactions.

Figure 5.

N2O causes the production of ROS in the presence of iron
A, the oxidation of adrenaline to adrenochrome by free oxygen radicals is a colour-changing reaction. B, the production of adrenochrome from adrenaline is H2O2 dose dependent, and occurs only in the presence of transition metals, such as iron (n= 3–4). C, adrenaline oxidation in the presence of 80% N2O is similar to the effect seen with 0.1% H2O2 (n= 3–4).

EUK-134 reduces inflammatory pain, but EUK and N2O analgesia are not additive

Next, we examined the hypothesis that free radical signalling underlies the analgesic effects of N2O in vivo using the formalin test. EUK-134 is a synthetic manganese-containing complex with SOD and catalase activity that has been shown to reduce oxidative stress in vivo and have neuroprotective properties in ischaemic brain injury (Baker et al. 1998; Rong et al. 1999). We reasoned that an SOD/catalase mimetic would significantly reduce the number of ROS present at an injury site and thus decrease inflammation and pain responses. As expected, we found that systemic i.p. injections of EUK-134 (25 mg kg−1) in wild-type littermate mice decreased inflammation by 22 ± 4% (P < 0.01), measured by the volume of formalin-induced paw oedema (n= 8 mice) in comparison to wild-type mice injected with i.p. saline (n= 8 mice, data not shown). Importantly, Fig. 6 shows that systemic injections of EUK-134 i.p. in an air-filled test chamber (open bars) largely diminished pain responses to formalin when compared to age-matched wild-type littermate mice injected with saline i.p. (black bars) as demonstrated by the reduced time spent licking/biting of the affected paw in all measures of inflammatory pain (P1, P2 and total time). Furthermore, Fig. 6 further illustrates that EUK-134 prevented N2O from providing additional analgesia to inflammatory pain (grey bars); consequently EUK-134 and N2O analgesia were not additive. In contrast, EUK-134 did not affect the morphine-induced analgesia (horizontal-hatched bars). Thus, these in vivo experiments support the idea that CaV3.2 T-channels and ROS are necessary for N2O analgesia in this inflammatory pain protocol.

Figure 6.

Free radical scavenger EUK-134 attenuates the analgesic effects of N2O in formalin test of inflammatory pain in wild-type mice
EUK-134 significantly reduced time spent licking/biting the injected foot in all measures (total, P1, P2), but no additional analgesia was seen when EUK-134 and N2O were co-applied. In contrast, morphine co-applied with EUK-134 produced additional analgesia in all pain measures. Total: Saline–Air, 468.6 ± 21.7 s; EUK-134–Air, 313.1 ± 17.0 s, EUK-134–N2O, 325.6 ± 14.9 s, EUK-134–Morphine, 230.3 ± 21.7 s. P1: Saline–Air, 66.4 ± 6.2 s; EUK-134–Air, 46.4 ± 4.6 s, EUK-134–N2O, 37.2 ± 5.4 s, EUK-134–Morphine, 31.6 ± 3.6 s. P2: Saline–Air, 468.6 ± 21.8 s; EUK-134–Air, 266.6 ± 19.0 s, EUK-134–N2O, 288.33 ± 18.2 s, EUK-134–Morphine, 198.7 ± 20.1 s. Vertical lines represent s.e.m. *Significant difference from Saline–Air control (P < 0.02). *,#Significant difference from EUK-134–Air (P < 0.05).

In a separate set of experiments, we injected the same dose of EUK-134 (25 mg kg−1i.p.) in age-matched CaV3.2−/− mice. In contrast to wild-type littermates (Fig. 6), injections of EUK-134 in CaV3.2−/− mice did not significantly affect nociceptive responses in P1, P2 or total times of licking/biting of affected paws as follows, total: Air, 293.8 ± 14.6 s; EUK-134–Air, 287.5 ± 65.0 s; P1: Air, 64.6 ± 5.8 s; EUK-134–Air, 86.5 ± 17.4 s; P2: Air, 229.3 ± 13.8 s; EUK-134–Air, 201.0 ± 62.2 s; P > 0.05 for all measures, n= 15 mice for Air and n= 6 mice for EUK-134–Air groups.


Nitrous oxide is frequently used in clinical settings as a potent analgesic, but little is known of its molecular mechanism of action. Here, we present evidence that N2O-based inhibition of T-channels in vitro and in vivo is mediated by free radical signalling.

Recent publications describe an essential role of histidine 191 for three other CaV3.2-selective agents: nickel (Kang et al. 2006), ascorbate (Nelson et al. 2007a) and zinc (Nelson et al. 2007b). Accordingly, we began with H191 in our search for the means of N2O inhibition of CaV3.2 channels. Mutation of this histidine residue to glutamine (H191Q) in CaV3.2 abolished N2O inhibition, while the reverse mutation (Q172H) in N2O-insensitive CaV3.1 produced a gain of N2O susceptibility.

The H191-dependent increase in CaV3.2 currents with DFX suggests removal of a tonically bound iron ion, whose presence inhibits the channel. This increased current was significantly reduced in the H191Q mutant, which further emphasizes that this histidine plays a key role in the metal binding site. Since the potentiating actions of DFX were fast and reversible (Fig. 3), it is unlikely that the observed increase in T-current was due to a side effect of DFX as a potent hypoxia-inducing factor. This also makes recruitment or incorporation of new T-channels an improbable mechanism of action of DFX in our experiments, in contrast to the chronic DFX application seen in Carabelli et al. (2007). To confirm that the observed changes were due to the chelating ability of DFX, we tested another metal chelator, DTPA. This non-specific lower-affinity chelator also increased CaV3.2 currents, but most importantly, DTPA also greatly diminished N2O inhibition of CaV3.2 channels, indicating a need for a bound metal for N2O effectiveness. In the Fenton reaction, iron and H2O2 combine to produce the hydroxyl radical ·OH. This highly unstable complex then reacts with a small number of amino acid residues in the immediate vicinity of the metal binding site (Stadtman & Oliver, 1991) in MCO. Targets for MCO may include histidine, cysteine, lysine, arginine, proline and tyrosine residues due to their participation in metal binding sites (Stadtman & Berlett, 1991; Stadtman & Levine, 2003). However, histidine is one of the most vulnerable amino acids to oxidation reactions (Uchida, 2003). Accordingly, we found that N2O had significantly smaller inhibitory effect in the H191Q mutant than the H191C mutant of CaV3.2. This could be due to the requirement for at least one additional nearby free cysteine residue to be cross-linked during an oxidation reaction involving cysteine residues.

Application of H2O2 produced T-current inhibition similar, but not additive to that of N2O, supporting our hypothesis that H2O2 is an intermediate in the reactions between N2O and CaV3.2 channels. Indeed, decomposition of H2O2 by catalase interrupts this process. Furthermore, quantification of the oxidation of adrenaline to adrenochrome allowed us to more directly assess the role of N2O in producing ROS. Again, the action of N2O paralleled that of H2O2.

MCO of the same histidine residue that we examine here (H191) in T-channels was recently implicated as the mechanism of ascorbate selective inhibition of CaV3.2 current (Nelson et al. 2007a). Ascorbate has well-known pro-oxidant properties; it is able to generate ROS by reducing transition metals via a one-electron transfer mechanism (Stadtman, 1991). Here, we propose a similar, but novel role for N2O in a molecular mechanism of inhibition of CaV3.2, as presented in our model showing a two-electron oxidation of iron and consequent reduction of N2O to N2. Fig. 7A shows a CaV3.2 channel in its normal state, passing inward calcium current, and a metal ion (in this case Fe2+) bound to H191. Upon encountering N2O (Fig. 7B), one of two sets of reactions can occur: a one-electron transfer that requires an N2O intermediate and Fenton chemistry (eqn (1)), or a two-electron transfer with an Fe4+O intermediate (eqn (2)). Both pathways could result in the appearance of hydroxyl radicals and oxidation of Fe2+ to Fe4+ as follows:

display math(1a)
display math(1b)
display math(1c)
display math(2a)
display math(2b)

• OH is so unstable that it reacts locally, oxidizing H191 to 2-oxo-histidine (Fig. 7B). We hypothesize that this results in decreased channel open probability, similar to that reported in NMDA receptors after histidine modification (Donnelly & Pallotta, 1995). Indeed, Bartels and colleagues (2009) have recently used single channel recordings to demonstrate that nitrous oxide inhibits recombinant CaV3.2 currents by decreasing single channel open time duration and probability of channel opening. To complete the cycle, both histidine and iron may be reduced to their original state by elements in the endogenous electron donor system (e.g. NADH or NADPH), and CaV3.2 resumes its normal passage of calcium current (Fig. 7C). Further extensive chemical studies are needed to precisely delineate proposed pathways of N2O interaction with this critical histidine residue of CaV3.2. In addition to these proposed pathways, we were able to demonstrate that similar mechanisms may exist in vivo and at least partly mediate N2O analgesia in inflammatory pain.

Figure 7.

Proposed mechanism of MCO of CaV3.2 channels in the presence of N2O and iron
A, at rest Fe2+ (or possibly another redox-sensitive transition metal such as Cu2+) is bound to H191 and several additional, unidentified residues that comprise the high-affinity metal binding site on the external surface of Cav3.2. B, N2O interacts with bound iron to produce H2O2, an unstable intermediate ROS such as •OH, and iron is oxidized to Fe4+. Consequently, the C-2 position of the imidazole ring of H191 is modified by •OH, most probably resulting in 2-oxo-H191. This modification causes conformational changes of CaV3.2 and ultimately reduced T-currents via a presumed allosteric mechanism. C, Fe4+ is reduced back to Fe2+ by the electron donor system, allowing the cycle to begin again.

Though N2O is often thought of as an inert gas, interactions between N2O, metals and free radicals have been documented. For example, transition metals and N2O may form a metal oxide and N2 (Banks et al. 1968; Raiche & Belbruno, 1987; Delabie et al. 2001). Similar reactions between N2O and alkali metals have also been studied (Tishchenko et al. 2004). Koppenol (1991) found the generation of hydroxyl radicals from N2O to be energetically unfavourable, agreeing with our observation that N2O alone is unable to significantly oxidize adrenaline. However, Koppenol's calculations do not take into account the presence of redox-reactive transition metal ions, which could lower activation energies. Also, even if conditions remain unfavourable, even a small amount of highly reactive ROS could have significant effects in vivo. Indeed, Bosterling et al. (1980) found evidence of free radical intermediates during N2O metabolism by intestinal bacteria. ROS have been shown to be intermediates in the metabolic reduction of other nitrogen compounds (Mason & Holtzman, 1975), while irradiation of aqueous solutions of N2O produced ROS that proved toxic to Escherichia coli (Brustad & Wold, 1976). Interactions between N2O and metals are also critical clinically. Serious side-effects of exposure to N2O such as megaloblastic erythropoiesis (pernicious anaemia) and degeneration of the spinal cord are caused by oxidation of cobalamin in vitamin B12 (Hadzic et al. 1995). This oxidation renders vitamin B12 useless as a co-enzyme to methionine synthase, disrupting several pathways (Lumb et al. 1980, 1981). In fact, previous studies have shown that free radical scavengers such as dimethylthiourea and SOD are able to significantly reduce N2O-mediated inactivation of methionine synthase (Koblin & Tomerson, 1990; Nair et al. 1995).

Each of these studies supports our hypothesis that N2O indirectly inhibits T-currents by oxidizing the channel-bound metal and producing ROS, which then participate in MCO of the CaV3.2 T-channel. Our chelator experiments provide evidence of the dependence of N2O on channel-bound redox-reactive metal ions such as iron or possibly copper (Nelson et al. 2007b), while application of catalase demonstrated the necessity of H2O2 and hydroxyl radicals. Furthermore, our occlusion experiment with H2O2 and N2O strongly suggests that there is a common pathway of their inhibition of T-currents. The oxidation of adrenaline to adrenochrome allowed us to model the generation of ROS by N2O and iron in a cell-free system. Lastly, we show that intraperitoneal application of an SOD/catalase mimetic, EUK-134, is able to reduce pain and inflammation in the formalin test, and consequently abolish analgesic effects of N2O in this model. Importantly, the potent analgesic effect of sub-anaesthetic concentrations of N2O in the formalin test was almost completely absent in CaV3.2−/− mice. Thus, we present novel evidence that N2O inhibits T-channels in the pain pathway via ROS in a cell-free solution, individual neurons and the whole animal.

In addition, NMDA receptors are prominent contributors in pain transmission in the dorsal horn of the spinal cord, an important pain processing region. Importantly, the synergistic action of T-channels, NMDA receptors and neurokinin receptors in lamina I of the dorsal horn is required for sensitization to inflammatory stimuli in the formalin test (Ikeda et al. 2006). Since CaV3.2 is expressed in both peripheral sensory neurons and the dorsal horn of the spinal cord (Talley et al. 1999), it is likely that N2O inhibition of T-channels via ROS works together with inhibition of NMDA receptors to reduce inflammatory pain.

Though N2O has been in use for centuries, little is known of its mechanism of action. The data presented here provide much needed information to understanding the actions of N2O. This information also has wider applications: MCO of proteins in the presence of trace metals such as iron or copper is becoming widely recognized as a key step in many pathological conditions. These include inflammatory diseases such as rheumatoid arthritis, atherosclerosis and ischaemia reperfusion damage (Stadtman & Oliver, 1991; Wang et al. 2004). Furthermore, N2O is implicated in neurocognitive decline after general anaesthesia due to neurotoxicity (Jevtovic-Todorovic et al. 1998, 2001, 2003). Future studies are necessary to investigate these possibilities. Further research will also be needed to determine whether other ion channels (e.g. NR1/NR2A NMDA receptors) that are known to contain high-affinity metal binding sites (Herin & Aizenman, 2004) show inhibition by N2O (Jevtovic-Todorovic et al. 1998; Mennerick et al. 1998; Georgiev et al. 2008) and undergo modification by free radicals (Aizenman et al. 1990; Gao et al. 2007) are subject to similar regulation.


Author contributions

P.O., D.B., R.S., J.L., E.L., M.R.D and M.T.N contributed to data collection and analysis. P.O. and S.M.T. contributed to the conception and design of experiments, and the drafting of the article as well as revising it critically for important intellectual content. All authors have approved the final version of the manuscript. All experiments were performed in the laboratories of the Department of Anaesthesiology at The University of Virginia, Charlottesville, VA.


Our research is supported by NIH R0-1 grant GM070726 and GM075229 (to S.M.T.), NIH fellowship F31NS059190 (to P.O.) and by the InJe Research and Scholarship Foundation in 2009 (J.L.). We thank Dr Kevin Campbell for providing the initial pair of CaV3.2 KO mice and Dr Wenhao Xu for providing the services of the UVA Gene Targeting and Transgenic Facility. We would also like to thank Drs Edward Perez-Reyes, Vesna Jevtovic-Todorovic, Douglas F. Covey and W. Dean Harman for their advice and support.