Inhibitory and excitatory neuromodulation by hydrogen peroxide: translating energetics to information

Abstract

Historically, brain neurochemicals have been broadly classified as energetic or informational. However, increasing evidence implicates metabolic substrates and byproducts as signalling agents, which blurs the boundary between energy and information, and suggests the introduction of a new category for ‘translational’ substances that convey changes in energy state to information. One intriguing example is hydrogen peroxide (H₂O₂), which is a small, readily diffusible molecule. Produced during mitochondrial respiration, this reactive oxygen species, can mediate dynamic regulation of neuronal activity and transmitter release by activating inhibitory ATP-sensitive K⁺ (K_ATP) channels, as well as a class of excitatory non-selective cation channels, TRPM2. Studies using ex vivo guinea pig brain slices have revealed that activity-generated H₂O₂ can act via K_ATP channels to inhibit dopamine release in dorsal striatum and dopamine neuron activity in the substantia nigra pars compacta. In sharp contrast, endogenously generated H₂O₂ enhances the excitability of GABAergic projection neurons in the dorsal striatum and substantia nigra pars reticulata by activating TRPM2 channels. These studies suggest that the balance of excitation vs. inhibition produced in a given cell by metabolically generated H₂O₂ will be dictated by the relative abundance of H₂O₂-sensitive ion channel targets that receive this translational signal.

Margaret E. Rice is a Professor in the Department of Neurosurgery and the Department of Neuroscience and Physiology at New York University School of Medicine. She received her PhD in chemistry from the University of Kansas, with a focus on neurochemistry. Her NIH-funded laboratory studies factors that regulate the release of dopamine, a key transmitter in motor and reward pathways in the brain. On-going projects include studies of axonal dopamine release regulation in dorsal and ventral striatum by hydrogen peroxide and by diet and insulin. Her group is also pursuing mechanistic studies of somatodendritic dopamine release, focused on substantia nigra dopamine neurons that degenerate in Parkinson's disease. Dr Rice is a member of the NYU Parkinson and Movement Disorders Center.

Introduction

F. O. Schmitt and Fred Samson introduced the notion of ‘informational’ substances to describe a neurochemical class that is distinct from ‘energetic’ substances (Schmitt, 1984, 1985). In this classification scheme, informational substances include neurotransmitters, exemplified by glutamate and GABA, whereas energetic substances include metabolic substrates, like glucose and oxygen. Increasingly, however, the line between these broad categories has blurred. Glutamate is not only the primary excitatory transmitter in the CNS, but also an important metabolic substrate (e.g. Dienel, 2013), and glucose not only provides energy for cells, but also conveys information that modulates the activity of specific glucose-sensing neurons (Routh, 2010; Thorens, 2012). Adding further complexity to the delineation between information and energy is that oxygen consumption during mitochondrial respiration, an energetic process, generates reactive oxygen species (ROS), including superoxide (•O₂⁻) and hydrogen peroxide (H₂O₂) (Boveris & Chance, 1973; Dugan et al. 1995; Liu et al. 2002; Bao et al. 2009; Rigoulet et al. 2011; Mailloux et al. 2013) that can act as signalling molecules (Sundaresan et al. 1995; Kamsler & Segal, 2004; Rhee, 2006; Stone & Yang, 2006; Avshalumov et al. 2007; Kishida & Klann, 2007; Miller et al. 2007b; Gerich & Funke, 2009; Groeger et al. 2009; Rigoulet et al. 2011; Rice, 2011; Jeong et al. 2012; Sies, 2014).

Molecules like ROS are at the interface between energy and information, and merit a class of their own. Indeed, they could be viewed as ‘translational’ substances that provide an interpretation of cellular activity within a cell or to neighbouring cells. In the case of ROS in the CNS, increased neuronal activity drives energy demand, which drives metabolism, and thereby generates signalling molecules that can act both within individual neurons (as intracellular signals) and between neurons (as diffusible messengers) to give immediate modulatory feedback about local activity. Other molecules in this class might include ATP, adenosine and glucose.

A particularly strong candidate as a ‘translational’ substance is H₂O₂. During mitochondrial electron transport, oxygen is first reduced to the free radical •O₂⁻, up to 5% of which leaves the respiratory process (Arnaiz et al. 1999), and can be converted to H₂O₂ by the enzyme superoxide dismutase or by spontaneous dismutation (see Peuchen et al. 1997). Oxygen consumption is proportional to local activity (Kennedy et al. 1992), with the greatest demand to support ATP-dependent signalling, e.g. informational processes (Engl & Attwell, 2015). The H₂O₂ produced in this process is well positioned to serve as a dynamic reporter of neuronal activity. Other slower metabolic processes also generate H₂O₂, including NADPH oxidases (NOXs) that produce •O₂⁻, and thus H₂O₂ (Babior, 1984; Lambeth, 2004; Infanger et al. 2006; Rhee, 2006; Bedard & Krause, 2007), and monoamine oxidases (MAOs), which produce one molecule of H₂O₂ for each biogenic amine molecule metabolized (Maker et al. 1981; Azzaro et al. 1985; Cohen, 1994). These additional sources allow a range of timescales for H₂O₂ signalling that extend from a rapid, subsecond level for mitochondrial H₂O₂ generation to slower regulation by growth factor activation of NOX, for example, which can proceed over hours, days, or even longer (Miller et al. 2007b).

Among the best-studied intracellular targets for slow regulation by H₂O₂ are redox-sensitive phosphatases that are inactivated by H₂O₂ and kinases that are activated by this ROS (Klann & Thiels, 1999; Rhee et al. 2005; Woolley et al. 2013). Modulation of phosphatase and kinase activities regulates signal propagation downstream of receptors, so that H₂O₂ helps fine-tune transmission to further downstream targets, including transcription factors like p53, NF-κB and AP-1, which are also H₂O₂ sensitive (Groeger et al. 2009; Woolley et al. 2013). Through these processes, H₂O₂ plays key roles in cell growth and survival. The primary source of H₂O₂ for these regulatory processes is NOX, which is activated by growth factors and other receptor activators (Rhee et al. 2005; Miller et al. 2007b, 2010; Groeger et al. 2009; Woolley et al. 2013).

In contrast to slow signalling mediated by NOX- and MAO-generated H₂O₂ that can regulate intracellular signalling cascades, rapid, subsecond signalling by H₂O₂ originates in mitochondria (Bao et al. 2009) and acts on cellular ion channels (Seutin et al. 1995; Krippeit-Drews et al. 1999; Avshalumov et al. 2003, 2003; Avshalumov & Rice, 2003; Bao et al. 2005; Patel et al. 2011; Lee et al. 2011, 2013). Early physiological studies showed that application of exogenous H₂O₂ can cause membrane hyperpolarization by activating a K⁺ conductance in various cell types, including CA1 hippocampal neurons (Seutin et al. 1995) and pancreatic β-cells (Krippeit-Drews et al. 1999). Our laboratory subsequently identified ATP-sensitive K⁺ (K_ATP) channels as key inhibitory targets of endogenous, as well as exogenous H₂O₂ (Avshalumov et al. 2003, 2005; Avshalumov & Rice, 2003; Patel et al. 2011). More recently, we identified a subclass of transient receptor potential (TRP) channels in basal ganglia neurons that are dynamically regulated by H₂O₂ (Bao et al. 2005; Lee et al. 2011, 2013). Work in our laboratory has focused on rapid signalling by H₂O₂ in regions of the basal ganglia, particularly the striatum and the substantia nigra pars compacta (SNc) and pars reticulata (SNr). Our cellular focus has been on dopaminergic (DAergic) neurons of the SNc that project to the dorsal striatum, forming the nigrostriatal pathway that plays critical roles in movement and motor learning (Carlsson, 2002; Redgrave et al. 2011), and on GABAergic neurons of the SNr, which are basal ganglia output neurons (Zhou & Lee, 2011).

This review describes how H₂O₂ rapidly translates a dynamic change in cellular metabolism, specifically mitochondrial oxygen consumption, to a salient signal. The targets of this signal are H₂O₂-sensitive ion channels that can influence the excitability of the neuron in which H₂O₂ is generated, as well as neighbouring cellular elements. Most data presented were obtained using ex vivo brain slices prepared from adult male guinea pigs or mice after inducing deep anaesthesia (50 mg kg⁻¹ sodium pentobarbital, i.p.). Methods include fast-scan cyclic voltammetry (FCV) to detect DA release, whole-cell recording to monitor neuronal activity, and fluorescence imaging to indicate H₂O₂ generation (e.g. Avshalumov et al. 2003, 2005, 2008; Bao et al. 2009; Lee et al. 2011, 2013; Patel & Rice, 2013).

Cellular regulation of H₂O₂

The previous section discussed sources of H₂O₂, but it is important to recognize that sinks, including metabolism and diffusion, are critical in shaping patterns of H₂O₂ signalling, as well. Thus, cellular H₂O₂ regulation must not only prevent oxidative damage from H₂O₂ elevation, but also allow intracellular and extracellular concentrations sufficient for a signalling effect to be achieved (Avshalumov et al. 2004; Murphy et al. 2011; Rice, 2011). Absolute basal and dynamic intracellular concentrations of H₂O₂ remain a matter of debate; the technical and conceptual issues involved have been reviewed elsewhere (Adimora et al. 2010; Murphy et al. 2011; Rice, 2011).

Peroxidase enzymes that metabolize H₂O₂ include glutathione (GSH) peroxidase, which is found in the cytosol and in mitochondria (Stults et al. 1977), and catalase, which is localized to intracellular peroxisomes (Cohen, 1994; Peuchen et al. 1997; Dringen et al. 2005). Dynamic H₂O₂ regulation is also provided by peroxiredoxins and thioredoxins, which act in a complicated dance that not only buffers H₂O₂ levels to prevent potentially toxic consequences of oxidative stress, but also facilitates the elevation of intracellular H₂O₂ to levels sufficient for signalling (Rhee et al. 2005; Adimora et al. 2010; Rhee & Woo, 2011; Jeong et al. 2012).

Complementing H₂O₂ regulation by cellular peroxidase enzymes are roles played by the low molecular weight antioxidants GSH and ascorbate. These antioxidants protect against possible pathological consequences of H₂O₂ elevation through their actions as scavengers of the aggressive hydroxyl radical (•OH) produced from the interaction of H₂O₂ or •O₂⁻ with trace metal ions (Cohen, 1994). Additionally, GSH is an essential cofactor for GSH peroxidase activity, and with other cellular thiols provides critical regulation of peroxiredoxins and thioredoxins through reduction of disulfide bonds involved in activation and inactivation (Stults et al. 1977; Rhee & Woo, 2011; Mailloux et al. 2013). Ascorbate, by contrast, has a higher redox potential than GSH and so cannot reduce oxidized thiols or break disulfide bonds, and does not interact with H₂O₂ directly (Rice, 2000; Avshalumov et al. 2004; Rhee & Woo, 2011). Notably, ascorbate is the primary low molecular weight antioxidant in neurons (with an intracellular concentration of 10 mm) (Rice & Russo-Menna, 1998), so that it is in a prime position to permit neuronal H₂O₂ signalling, yet prevent pathological consequences that could occur from unregulated H₂O₂ generation and •OH production (Rice, 2012).

One final point about H₂O₂ regulation is the extent to which H₂O₂ can diffuse from a site of generation to act at targets in the same or neighbouring cells. In contrast to other ROS, H₂O₂ is not a free radical and not an ion. These properties not only limit H₂O₂ reactivity and extend its lifetime (Cohen, 1994), but also increase its membrane permeability (Ramasarma, 1982; Bienert et al. 2007; Adimora et al. 2010). However, increasing evidence indicates that net cellular H₂O₂ efflux and entry is governed by cell-specific membrane permeability factors coupled with competing effects of the antioxidant network (Makino et al. 2004; Bienert et al. 2007; Adimora et al. 2010; Miller et al. 2010; Mishina et al. 2011; Bienert & Chaumont, 2014). In particular, the cellular expression of specific aquaporin isoforms, including aquaporins 3 and 8 (Miller et al. 2010; Bienert & Chaumont, 2014), has been shown to facilitate the passive diffusion of H₂O₂ across membranes, and thereby influence its efficacy as an intracellular signalling agent, as well as a diffusible messenger.

Endogenous H₂O₂ inhibits axonal and somatodendritic DA release

Regulation of dopamine release from DAergic axons or cell bodies and dendrites (somatodendritic release) can be studied readily in ex vivo brain slices using carbon-fibre microelectrodes with FCV (Rice et al. 2011; Patel & Rice, 2013). For such studies, local electrical stimulation is used to elicit an increase in extracellular DA concentration ([DA]_o), which reflects the net influence of DA release and uptake. The use of pulse-train stimulation allows the influence of concurrently released transmitters and local activity changes on [DA]_o as the train progresses. For the studies summarized here, stimulation parameters were commonly 30 pulses at 10 Hz, although trains of as few as 7 pulses at 10 Hz are sufficient to reveal modulation by endogenously generated H₂O₂ in guinea pig dorsal striatum (K. A. Moran & M. E. Rice, unpublished observations).

Using these methods, we found that exogenous H₂O₂ causes a reversible, 30% suppression of evoked [DA]_o in dorsal striatum that is not accompanied by a change in tissue DA content or signs of oxidative damage (Chen et al. 2001). We then examined a role for endogenous H₂O₂ in DA release regulation by inhibiting GSH peroxidase to amplify levels of dynamically generated H₂O₂ levels using mercaptosuccinate (MCS). Exposure to MCS causes similar reversible decreases in evoked [DA]_o in dorsal and ventral striatum to those seen with exogenous H₂O₂, again with no change in tissue DA content (Chen et al. 2002; Avshalumov et al. 2008). Significantly, DA release suppression in the dorsal striatum can be reversed by the H₂O₂ metabolizing enzyme catalase in the continued presence of MCS, confirming H₂O₂ involvement (Avshalumov et al. 2003). Suppression of evoked [DA]_o also persists when DA uptake is inhibited, indicating an effect on DA release, not uptake (Avshalumov et al. 2003).

Although exogenous H₂O₂ causes suppression of somatodendritic DA release in the SNc and in the adjacent ventral tegmental area (VTA), GSH peroxidase inhibition decreases evoked [DA]_o in SNc, but not VTA (Chen et al. 2002). This difference between SNc and VTA is potentially important, because DAergic neurons of the SNc degenerate in Parkinson's disease and in animal models of Parkinson's, whereas those in the VTA are relatively spared (Yamada et al. 1990; Fearnley & Lees, 1991; Betarbet et al. 2000).

Endogenous H₂O₂ in dorsal striatum is generated downstream of glutamatergic AMPA receptor activation

Local electrical stimulation evokes release of glutamate, GABA and other transmitters, as well as DA. In striatum, GSH peroxidase inhibition by MCS has no effect on [DA]_o evoked by a single stimulus pulse (Avshalumov et al. 2003), indicating that modulatory H₂O₂ must be generated dynamically during initial and subsequent pulses of a stimulus train to inhibit on-going DA release. The generation of this modulatory H₂O₂ proved to require activation of AMPA receptors (AMPARs), as AMPAR blockade causes a marked increase in pulse-train-evoked [DA]_o in guinea pig dorsal striatum (Avshalumov et al. 2003), indicating that physiological glutamate release inhibits axonal DA release in dorsal striatum via AMPARs. An essential role for inhibitory H₂O₂ in mediating this effect, as well as in mediating the opposing effect on DA release by GABA acting at GABA_A receptors, was subsequently shown by the loss of regulation in the presence of exogenous catalase or GSH peroxidase (Avshalumov et al. 2003).

As reviewed elsewhere, the absence of AMPARs and GABA_ARs on DAergic axons in dorsal striatum led us to postulate that the cellular sources of glutamate-dependent H₂O₂ generation were not DAergic axons, but rather other striatal neurons, including the predominant striatal neurons, GABAergic medium spiny neurons (MSNs) (Avshalumov et al. 2007, 2008; Rice et al. 2011; Rice, 2011; Patel & Rice, 2012). We tested this hypothesis using single-cell fluorescence imaging of H₂O₂, in which dihydro-dichlorofluorescein diacetate (H₂DCF-diacetate) is loaded into cells via a patch pipette used for whole-cell recording (Avshalumov et al. 2005, 2008). This dye is cleaved by intracellular esterases to form H₂DCF, which becomes fluorescent DCF when oxidized. Local stimulation in guinea pig striatal slices activates a single action potential with each stimulus pulse in recorded MSNs (Fig.1A, lower panel). Concurrent imaging of DCF fluorescence confirmed a significant increase in DCF fluorescence intensity (FI) evoked using the same pulse-train stimulation parameters as in DA release studies (Fig.1A, upper panel). Activity-dependent increases in DCF FI are enhanced when GSH peroxidase is inhibited by MCS, and abolished by the addition of exogenous catalase, confirming H₂O₂ detection (Avshalumov et al. 2008). Consistent with a requirement for glutamate-dependent AMPAR activation in the generation of modulatory H₂O₂ in dorsal striatum, antagonism of AMPARs prevents stimulus-induced action potentials and H₂O₂ generation in MSNs (Fig.1B) (Avshalumov et al. 2008). These and other data support a role for dynamically generated H₂O₂ as a diffusible messenger that is generated in striatal MSNs (and possibly other local neurons), and diffuses to adjacent DAergic axons to inhibit DA release (Avshalumov et al. 2008; Rice, 2011).

Figure 1. Modulatory H₂O₂ generation in guinea pig striatal medium spiny neurons (MSNs) requires AMPAR activation.

A, local electrical stimulation (30 pulses at 10 Hz; 100 μs, 0.5 mA) causes tetrodotoxin-sensitive action potentials and increased DCF fluorescence intensity (FI) in MSNs in ex vivo slices of guinea pig dorsal striatum. B, AMPAR blockade with GYKI-52466 (50 μm) prevents stimulus-induced spikes and H₂O₂ generation (modified from Avshalumov et al. 2008).

H₂O₂ acts via K_ATP channels in DAergic axons

The targets at which H₂O₂ acts to regulate nigrostriatal DA release are K_ATP channels (Avshalumov et al. 2003; Avshalumov & Rice, 2003). Blocking K_ATP channels with a sulfonylurea receptor antagonist not only enhances pulse-train-evoked [DA]_o in striatum, but also prevents the usual H₂O₂-dependent inhibition of DA release by MCS, as well as the patterns of DA release regulation by AMPAR and GABA_AR activation (Avshalumov et al. 2003; Avshalumov & Rice, 2003; Rice, 2011). These data show that K_ATP channels are required for DA release modulation by H₂O₂, glutamate and GABA (Avshalumov et al. 2003; Avshalumov & Rice, 2003). Consequently, striatal DA release is decreased by K_ATP channel openers, whether evoked with single-pulse or pulse-train stimulation (Avshalumov & Rice, 2003; Patel et al. 2011). Although release suppression is seen with an opener selective for either SUR1- or SUR2-subunit-containing K_ATP channels, only a SUR1-selective opener occludes the effects of MCS, as well as the effects of AMPAR and GABA_AR antagonism, on evoked [DA]_o showing that DA release regulation by glutamate-dependent H₂O₂ requires SUR1-based K_ATP channels (Avshalumov & Rice, 2003).

A key aspect of our hypothesis that H₂O₂ is a diffusible messenger in the striatum is that K_ATP channels are located directly on DAergic axons. Our observation that single-pulse-evoked [DA]_o is suppressed by K_ATP channel openers supports direct localization, which was confirmed using immunohistochemical methods showing striatal colocalization of a K_ATP channel subunit with tyrosine hydroxylase, an enzyme required for DA synthesis (Patel et al. 2011; Patel & Rice, 2012). Together with our previous work, these data indicate that H₂O₂ is a diffusible messenger, which is generated in striatal MSNs, but acts at K_ATP channels on DA axons.

It should be emphasized at this point that H₂O₂-dependent signalling via ion channel activation is fast and transient, with a subsecond to second time scale (Patel et al. 2011; Patel & Rice, 2012). This was assessed using paired-pulse stimulation to evoke [DA]_o with pharmacological blockade of K_ATP channels or amplification of H₂O₂ levels using GSH peroxidase inhibition. Maximal H₂O₂-/K_ATP channel-dependent suppression of DA release is seen 500 ms after an initiating stimulus, with a significant influence persisting until ∼1000 ms (Patel et al. 2011).

Mitochondria are the subcellular source of H₂O₂ for dynamic striatal signalling

Studies in isolated mitochondria indicate that succinate, a substrate of mitochondrial complex II, drives H₂O₂ production by back-flow of electrons to complex I, which is prevented by partial complex I inhibition by rotenone (Votyakova & Reynolds, 2001; Liu et al. 2002; Gyulkhandanyan & Pennefather, 2004). We demonstrated that this regulation also occurs in whole cells in brain slices using a reversible H₂O₂-sensitive dye Redoxfluor-1 (RF1) (Miller et al. 2007a). Succinate caused a rapid increase in RF1 FI in striatal MSNs (Fig.2A and B), which was reversed by co-application of rotenone at a concentration that leads to partial complex I inhibition (50 nm) in the continued presence of succinate (Fig.2B) (Bao et al. 2009). We then examined pulse-train-evoked DA release in this rotenone + succinate cocktail to assess whether the subcellular source of modulatory H₂O₂ was mitochondrial respiration. Consistent with the expected effect of enhanced H₂O₂ generation, succinate alone suppressed evoked [DA]_o; suppression of DA release was prevented by catalase, confirming H₂O₂ involvement, and was reversed by rotenone (Bao et al. 2009). Most importantly, rotenone + succinate prevents the usual increase in evoked [DA]_o with AMPAR blockade (Fig.2C) and the decrease that usually accompanies GSH peroxidase inhibition by MCS (Fig.2D). By contrast, inhibition of either NADPH oxidase or MAO had no effect on pulse-train-evoked [DA]_o or on the usual suppression of DA release seen when GSH peroxidase is inhibited by MCS (Bao et al. 2005). These data show that mitochondrial respiration is the source of H₂O₂ that can then translate the significance of a glutamate-activated increase in cellular activity and mitochondrial metabolism in MSNs to a signal at DAergic axons to dynamically regulate DAergic transmission.

Figure 2. Role of mitochondrial H₂O₂ in dynamic modulation of striatal DA release.

A, superfusion of succinate (5 mm) causes an increase in fluorescence intensity (FI) of the reversible H₂O₂-sensitive dye RF1 in a medium spiny neuron (MSN) in a guinea pig striatal slice; Alexa Red was included in the pipette solution to allow morphological identification of the imaged MSN (scale bar = 20 μm). B, time course of the increase in RF1 FI in the MSN in A during succinate exposure; the increase in FI was reversed by rotenone, at a concentration (50 nm) that leads to partial inhibition of mitochondrial complex I, in the continued presence of succinate (n = 7). C, representative pulse-train-evoked [DA]_o (30 pulses, 10 Hz) in artificial cerebrospinal fluid (aCSF) alone (Control) or in the presence of rotenone + succinate (Rot + Succ), before and after the addition of an AMPAR antagonist, GYKI-52466 (GYKI, 50 μm). Bar graphs show average evoked [DA]_o normalized to the starting condition for each experiment; GYKI caused a significant increase in evoked [DA]_o (***P < 0.001 vs. same-site control; n = 5), which was prevented by the rotenone–succinate cocktail (P > 0.05 vs. same-site Rot-Succ; n = 6). D, representative pulse-train-evoked [DA]_o in aCSF alone or in the presence of Rot-Succ before and after the addition of a GSH peroxidase inhibitor, MCS (1 mm). Bar graphs show average evoked [DA]_o normalized to the starting condition for each experiment; MCS caused a significant decrease in evoked [DA]_o (**P < 0.01 vs. same-site control; n = 5), which was prevented by the rotenone-succinate cocktail (P > 0.05 vs. same-site Rot-Suc; n = 5).

H₂O₂ regulates SNc DA neuron activity via K_ATP channels

DCF imaging in SNc DAergic neurons in guinea pig midbrain slices revealed tonic and activity-dependent H₂O₂ generation in these spontaneously active cells (Avshalumov et al. 2005). Notably, tonically generated H₂O₂ has a significant effect on DA cell excitability: depletion of intracellular H₂O₂ by including catalase in the pipette solution or blockade of K_ATP channels causes a significant increase in spontaneous firing rate in all DAergic neurons tested. Moreover, catalase in the pipette has no effect when K_ATP channels are blocked, demonstrating that tonically generated H₂O₂ regulates DAergic cell activity via K_ATP channels (Avshalumov et al. 2005). Notably, the pipette backfill solution typically contains 3 mm ATP, which should close K_ATP channels (Häusser et al. 1991), suggesting that H₂O₂ may act by decreasing channel sensitivity to ATP; the mechanism of regulation has yet to be elucidated, however (for discussion, see Patel & Rice 2012).

Conversely, elevation of H₂O₂ by exogenous application or GSH peroxidase inhibition causes neuronal hyperpolarization in ∼50% of SNc DAergic neurons examined in guinea pig midbrain. Neurons in this sensitive population, which we call ‘responders’, are also hyperpolarized by an SUR1-subunit-selective K_ATP channel opener whereas those that are insensitive to H₂O₂ (‘non-responders’) are hyperpolarized by an SUR2-subunit-selective opener (Fig.3) (Avshalumov et al. 2005). These data indicate that SUR1 expression in SNc DA neurons conveys sensitivity to elevated H₂O₂, as seen for striatal DA axons (Avshalumov & Rice, 2003; Patel et al. 2011). Overall, these data show that H₂O₂ plays an auto-regulatory role in SNc DA neurons via the activation of inhibitory K_ATP channels.

Figure 3. SUR1-containing K_ATP channels convey enhanced sensitivity to H₂O₂ elevation in guinea pig SNc DAergic neurons.

A, exogenous H₂O₂ (1.5 mm) (or MCS, 1 mm) causes hyperpolarization in ‘responders’; this is mimicked by SUR1-selective K_ATP-channel opener, diazoxide (60 μm), whereas a SUR2-selective opener, cromakalim (60 μm), has no effect (n = 6). B, a second population of SNc DAergic neurons does not respond to H₂O₂, MCS, or diazoxide, but hyperpolarizes with cromakalim (n = 6) (modified from Avshalumov et al. 2005).

Activation of TRPM2 channels by H₂O₂ in GABAergic neurons in the basal ganglia

Increasing evidence points to an additional ion channel target for H₂O₂, which is a subclass of TRP channels that provides regulation complementary to that provided by K_ATP channels. Although a variety of TRP channels are expressed in the brain, one subclass, TRPM2 (TRP melastatin 2), is uniquely sensitive to activation by H₂O₂ (Fleig & Penner, 2004). Activation of these channels by H₂O₂ leads to an increase in neuronal excitability (Bao et al. 2005; Lee et al. 2011, 2013), rather than the decrease that accompanies activation of K_ATP channels. How H₂O₂ activates TRPM2 channels is somewhat better understood than its mechanism at K_ATP channels, but not without debate. Although there is evidence for direct activation of TRPM2 channels by H₂O₂ (Wehage et al. 2002), other data argue against this (Tóth & Csanády, 2010). Instead, activation may be mediated by H₂O₂-dependent elevation of ADP ribose or a synergistic action of H₂O₂ and ADP ribose (Perraud et al. 2005; Lange et al. 2008).

The first evidence for H₂O₂-dependent regulation of neuronal excitability by TRP channels emerged in the course of studies to investigate the role of mitochondria as a source of modulatory H₂O₂ (Bao et al. 2005, 2009). In these studies we found that partial mitochondrial complex I inhibition by nanomolar concentrations of rotenone leads to unregulated generation of H₂O₂, indicated by single-cell DCF imaging in MSNs in guinea pig dorsal striatum (Fig.4A) (Bao et al. 2005; Avshalumov et al. 2007). Unsurprisingly, this increase in H₂O₂ leads to suppression of evoked DA release, which is prevented by catalase or by K_ATP channel blockade (Fig.4B), demonstrating H₂O₂ and K_ATP channel involvement. What was more surprising at the time, however, was that simultaneous current-clamp recording in MSNs showed a depolarization and an increase in excitability of those neurons, with a time course that paralleled the increase in DCF FI (Fig.4C) (Bao et al. 2005). Changes in MSN membrane properties were prevented by catalase, confirming H₂O₂ dependence. Moreover, the increase in excitability was also prevented by flufenamic acid (FFA), a non-selective TRP channel antagonist, implicating H₂O₂-dependent TRP channel activation in these GABAergic projection neurons (Bao et al. 2005). Although FFA can inhibit several TRP subtypes, as well as other ion channels (Guinamard et al. 2013), recognized targets include H₂O₂-sensitive TRPM2 channels (Hill et al. 2004). As discussed further below, TRPM2 channels have been identified as the target for H₂O₂-dependent modulation of SNr GABAergic projection neurons (Lee et al. 2013). Given that TRPM2 channels are expressed in striatal MSNs, it is likely that they also contribute to H₂O₂ modulation of MSNs (Hill et al. 2006). Blocking K_ATP channels led to enhanced MSN depolarization during rotenone exposure, indicating that concurrently activated K_ATP channels in these cells serves to counterbalance the predominant TRP-dependent effects of H₂O₂ elevation.

Figure 4. Unregulated H₂O₂ generation during partial mitochondrial complex I inhibition by rotenone inhibits striatal DA release via K_ATP and excites MSNs via TRP channels.

A, DCF fluorescence intensity (FI) in a MSN in a guinea pig striatal slice under control conditions (t = 0) and during exposure to rotenone (50 nm); scale bar is 20 μm. B, average DA release records after single-pulse stimulation elicited at 5 min intervals under control conditions and after 30 min exposure to rotenone (50 nm; n = 7) compared with release in the presence of glibenclamide (Glib; 3 μm) and glibenclamide plus rotenone (n = 5). Data are normalized, with maximum [DA]_o under control conditions for each slice taken as 100%. C, average time course of H₂O₂ generation (DCF FI) recorded in single MSNs during rotenone exposure (n = 13; ***P < 0.001 rotenone vs. basal FI; ANOVA) compared to the time course of rotenone-induced changes in membrane potential (Memb pot'l) in these same MSNs (n = 10; *P < 0.05; **P < 0.01 rotenone vs. control; ANOVA) (modified from Bao et al. 2005).

We then turned our attention to spontaneously active GABAergic projection neurons in the SNr (Lee et al. 2011, 2013). Consistent with opposing regulation of cellular activity by H₂O₂ acting at TRP vs. K_ATP channels, exogenous catalase causes a decrease in firing rate of SNr GABAergic neurons in guinea pig midbrain slices, indicating maintenance of excitability by basal levels of H₂O₂ (Lee et al. 2011). In sharp contrast to the inhibitory effect of H₂O₂ elevation on SNc DAergic neurons (Fig.3) (Avshalumov et al. 2005), elevation of H₂O₂ levels causes an increase in the firing rate of SNr GABAergic neurons, whether through amplification of endogenous levels by GSH peroxidase inhibition (Fig.5A, B and E) or exposure to exogenous H₂O₂ (Lee et al. 2011). Implicating H₂O₂-sensitive TRPM2 channels in this process, these increases in SNr neuron firing rates are blocked by FFA (Fig.5C and E), as well as when an antibody to the C-terminus of TRPM2 channels is included in the pipette solution (Fig.5D and F). In addition to these studies of functional TRPM2 expression, immunohistochemistry (Fig.5F) and in situ hybridization studies provide anatomical confirmation of TRPM2 channels in guinea pig SNr GABAergic neurons (Lee et al. 2013). Companion immunohistochemical studies of guinea pig midbrain also demonstrated expression of TRPM2 channels in SNc DAergic neurons (Fig.6) (Lee et al. 2013), consistent with other studies showing a functional role for these ion channels in SNc DAergic neurons in rats (Chung et al. 2011), as well as evidence for the presence of TRPM2 in DAergic neurons in mice (Mrejeru et al. 2011).

Figure 5. TRPM2 channel blockade reverses H₂O₂-dependent increases in firing rate in guinea pig SNr GABAergic neurons.

A, spontaneous activity of a SNr GABAergic neuron under control conditions. B, amplification of endogenous H₂O₂ levels by GSH peroxidase inhibition with MCS increases spontaneous firing rate. C, in the presence of a non-specific TRP channel blocker, flufenamic acid (FFA), H₂O₂ causes a decrease, rather than increase in SNr GABAergic neuron firing rate. D, the effect of elevated H₂O₂ is also converted to a suppression of spontaneous activity when TRPM2 channels are blocked by including an antibody to the C-terminus of TRMP2 channels (α-TRPM2-C) in the pipette solution. E, summary of the influence of elevated H₂O₂ on spontaneous SNr neuron firing rate. Increases in firing rate by H₂O₂ are reversed and suppressed below control levels by FFA or by α-TRPM2-C (*P < 0.05; **P < 0.01; ***P < 0.001) (modified from Lee et al. 2011, 2013). F, immunohistochemical labelling of guinea pig SNr neurons with parvalbumin (PV), a marker for GABAergic neurons, and an antibody to the N- or C-terminus of TRPM2, confirming TRPM2 expression in these cells. Staining is eliminated following pre-adsorption of the primary antibody with its immunogenic peptide (top row, third panel from left); scale bar, 10 μm. (modified from Lee et al. 2013).

Figure 6. TRPM2 channel expression in guinea pig SNc DAergic neurons.

A, Immunohistochemical labelling of guinea pig SNc shows little somatic staining with an antibody to parvalbumin (PV), a marker for GABAergic neurons, but B, abundant staining of most SNc neurons with an antibody to tyrosine hydroxylase (TH), confirming their identity as DAergic neurons. C, TH-immunopositive neurons are also immunopositive for TRPM2, indicated using an antibody to the C-terminus of TRPM2 (TRPM2-C). D, Merged image of PV, TH, and TRPM2-C. Arrows point to TRPM2 in DAergic dendrites; scale bar, 20 μm (modified from Lee et al. 2013).

In addition to regulating the spontaneous firing rate of SN neurons, TRPM2 channels are required for NMDA-induced burst firing in SNr GABAergic neurons (Lee et al. 2013). Notably, we found that H₂O₂ modulates NMDA-induced burst firing in these cells, leading to an increase in burst duration and a decrease in burst frequency (Fig.7). These findings reveal another modulatory role for H₂O₂ that could be especially relevant in Parkinson's disease, in which increased burst firing in SNr neurons and increases in ROS both occur (see Lee et al. 2013).

Figure 7. H₂O₂ modulates NMDA-induced burst firing in SNr GABAergic neurons.

A, activity of a SNr GABAergic neuron exhibiting NMDA-induced (30 μm) burst firing before and after addition of exogenous H₂O₂ (1.5 mm). B, both exogenous H₂O₂ and amplified endogenous H₂O₂ induced by inhibiting GSH peroxidase with MCS (1 mm) caused an increase in burst duration and a decrease in burst frequency (*P < 0.05) (modified from Lee et al. 2013).

Species similarities and differences

Most studies of H₂O₂ as a dynamic neuromodulator have been conducted using ex vivo brain slices from guinea pigs. Initial experiments suggested species independence of the basic inhibitory effect of H₂O₂ elevation on pulse-train-evoked DA release in dorsal striatum, with a similar reversible suppression of pulse-train-evoked [DA]_o in ex vivo striatal slices from rat, guinea pig and marmoset when GSH peroxidase was inhibited by MCS (Rice et al. 2002; Avshalumov et al. 2003). The Sombers group has further demonstrated H₂O₂-dependent suppression of DA release in rat striatum in vivo (Spanos et al. 2013). Recent studies have confirmed DA release suppression when endogenous H₂O₂ is elevated in mouse striatal slices, as well (B. O'Neill, R. Asri, J. C. Patel & M. E. Rice, unpublished observations). However, additional evidence suggests that dynamic regulation of DA release by glutamate, GABA and H₂O₂ may differ in mouse dorsal striatum with possible H₂O₂-independent regulation of DA release by these transmitters.

As noted above, studies in ex vivo slices from young adult guinea pig brain show greater sensitivity of SUR1- vs. SUR2-based K_ATP channels to activation by H₂O₂ in striatal DAergic axons (Avshalumov & Rice, 2003; Patel et al. 2011) as well as SNc DAergic neurons (Avshalumov et al. 2005). These findings are consistent with the greater metabolic sensitivity of SUR1- vs. SUR2-expressing SNc DAergic neurons in slices from neonatal mice (Liss et al. 1999), although adult mice appear to express only SUR1-based K_ATP channels (Liss et al. 2005). Rat SNc DAergic neurons also show K_ATP channel-dependent hyperpolarization with exogenous H₂O₂ application, with equal sensitivity of all DAergic cells, at least under the conditions tested (Geracitano et al. 2005). Interestingly, in these same studies, hypoxia-induced K_ATP channel activation was reversed by H₂O₂, which served as a source of molecular oxygen, as shown previously (Walton & Fulton, 1983).

Another species difference in H₂O₂-dependent modulation of neuronal activity is that H₂O₂ elevation in mouse midbrain slices leads to inhibition of SNr GABAergic neurons via predominant K_ATP channel activation, as opposed to the TRPM2-dependent excitation seen with H₂O₂ elevation in guinea pig SNr neurons (Lee et al. 2011). Differences in the functional activation of H₂O₂-dependent K_ATP and TRPM2 channels between guinea pigs and mice suggest divergent roles for this regulatory process across species. The need for neuronal regulation by a metabolic signal like H₂O₂ might depend on unique behavioural demands across species that require differential patterns of ion channel expression. Other possible factors include species differences in H₂O₂ generation or metabolism. For example, H₂O₂ metabolism by the glial antioxidant network differs between species, with stronger control in guinea pig (or human) vs. mouse (or rat), because of higher glia-to-neuron ratio of guinea pig brain (Avshalumov et al. 2004). Regardless of these differences, however, responsiveness to H₂O₂ across diverse species supports the idea that this molecule is an important regulator of neuronal function.

Opposing effects of H₂O₂ via K_ATP and TRPM2 channels in specific neuron populations

A seemingly paradoxical factor in H₂O₂-mediated signalling is that many target neurons express both K_ATP and TRPM2 channels that have opposing effects when activated by H₂O₂. For example, in addition to TRPM2 channels (Lee et al. 2013), GABAergic SNr neurons also express K_ATP channels (Schwanstecher & Panten, 1993; Stanford & Lacey, 1996; Lutas et al. 2014) that lead to H₂O₂-activated hyperpolarization of guinea pig SNr GABAergic neurons when TRPM2 channels are blocked (Fig.5C–E; Lee et al. 2011). Conversely, an inward current is observed in SNc DAergic neurons in response to H₂O₂ as opposed to the usual outward current and hyperpolarization when K_ATP and other potassium channels are blocked (Avshalumov et al. 2005; Geracitano et al. 2005; Chung et al. 2011). This raises the interesting possibility that the effect of H₂O₂ on target neurons is defined by the abundance and/or relative activity of these opposing channels and suggests that ATP and H₂O₂ are involved in a dynamic interplay regulating neuronal excitability, as discussed further below (Fig.8).

Figure 8. Net influence of mitochondrially generated H₂O₂ on DA release and on the excitability of guinea pig basal ganglia neurons mediated by K_ATP and TRPM2 channels.

Axonal DA release in striatum and somatodendritic DA release in SNc are suppressed by the net inhibitory effect of elevated H₂O₂ acting at K_ATP channels in DAergic neurons. In striatum, H₂O₂ generated in MSNs acts as a diffusible messenger (light blue arrows) to inhibit DA release by activating K_ATP channels on neighbouring DAergic axons. The source of dynamically generated H₂O₂ is mitochondrial respiration. The net effect of H₂O₂ generated within spontaneously active SNc DAergic neurons (dark blue arrows) is also inhibitory via K_ATP channel activation. Although striatal MSNs, which are GABAergic projection neurons, and SNr GABAergic projection neurons express both K_ATP and TRPM2 channels, the predominant effect of H₂O₂ on these cells is excitatory via TRPM2 channels. It should be noted that glutamatergic drive is required for action potential-dependent generation of H₂O₂ in MSNs, although partial mitochondrial inhibition by rotenone can lead to unregulated H₂O₂ generation. By contrast, sufficient H₂O₂ is generated tonically in spontaneously active SNr and SNc neurons to modulate firing rate independent of synaptic input.

There is a rich literature linking K_ATP channel activity to metabolic state in basal ganglia structures, with ATP identified as the primary signalling molecule. Factors that influence ATP (and ATP/ADP), such as oxygenation, influence K_ATP channel status, with anoxic conditions leading to K_ATP channel opening (Amoroso et al. 1990; Murphy & Greenfield, 1992; Jiang et al. 1994; Guatteo et al. 1998). Glucose concentration can also influence K_ATP channel activity with lower glucose concentrations favouring K_ATP channel opening, presumably through effects on intracellular ATP concentration (Amoroso et al. 1990; During et al. 1995; Marinelli et al. 2000). However, we have also found a relationship between glucose concentration and mitochondrial H₂O₂ production with lower glucose concentrations promoting mitochondrial activity and H₂O₂ production in striatal MSNs (L. Bao, C. R. Lee & M. E. Rice, unpublished observations), further linking metabolic activity to signalling by H₂O₂.

Less is known about the physiological processes that regulate TRPM2 channels in SNc and SNr neurons. However, several metabolic and activity-dependent mechanisms have been implicated, including activity-dependent increases in intracellular calcium, as well as conditions that elevate H₂O₂ and other ROS (Lee & Tepper, 2007; Freestone et al. 2009; Chung et al. 2011; Mrejeru et al. 2011; Lee et al. 2013).

Overall, H₂O₂ can have an excitatory or inhibitory effect on an individual neuron that is apparently based on the relative activities of K_ATP and TRPM2 channels in that cell. This working model is supported by evidence from our studies of H₂O₂-dependent modulation of DA release and basal ganglia neuron excitability in ex vivo guinea pig brain slices, reviewed here (Fig.8). Specifically, the net effect of H₂O₂ elevation on striatal DA release is inhibitory via K_ATP channels (Avshalumov & Rice, 2003; Bao et al. 2005; Patel et al. 2011). Similarly, the predominant influence of K_ATP channels on SNc DAergic neuron excitability is also reflected in the net hyperpolarization seen with H₂O₂ elevation in these midbrain projection neurons (Avshalumov et al. 2005) (Fig.8). In contrast, the net influence of H₂O₂ is to increase the excitability of GABAergic striatal MSNs (Bao et al. 2005) and GABAergic SNr neurons (Lee et al. 2011, 2013), which would enhance the output of these inhibitory projection neurons, presumably by the predominant influence of TRPM2 channels in both neuron populations (Fig.8). The source of modulatory H₂O₂ that can influence transmitter release and neuronal activity on a subsecond time scale is presumed to be mitochondrial respiration, as shown for striatal DA release regulation by H₂O₂-dependent activation of K_ATP channels (Bao et al. 2009; Patel et al. 2011). Notably, basal H₂O₂ levels in midbrain SN DAergic and GABAergic neurons generated during spontaneous firing activity in these cells is sufficient to provide a modulatory tone that is mildly inhibitory in SNc DAergic neurons (Avshalumov et al. 2005) and mildly excitatory in SNr GABAergic cells (Lee et al. 2011). In the striatum, however, dynamically generated H₂O₂ in MSNs requires activation of AMPARs by glutamatergic input and consequent action potential generation (Avshalumov et al. 2003, 2008).

Conclusions

The findings summarized here reveal an exquisite interaction between mitochondrial respiration and neuronal excitability, bridged by H₂O₂, which acts as a ‘translational’ substance that communicates the increase in metabolism to neuronal membranes via activation of K_ATP and TRPM2 channels. In its translational role, H₂O₂ generated within a given neuron can mediate autoinhibition and/or autoexcitation, but can act as a diffusible messenger to influence the activity of neighbouring cells (Fig.8). Actions of H₂O₂ at K_ATP and TRPM2 channels indicate that the net effect of H₂O₂ on a given cell or transmitter release site will reflect the balance of activity between H₂O₂-sensitive target channels expressed and thereby provide cell-type-specific patterns of modulation. These patterns of regulation have implications not only for normal regulation of basal ganglia transmitters and neuronal activity, but also for pathological conditions like Parkinson's disease, in which oxidative stress has been identified as a potential underlying factor in SNc DAergic neuron degeneration (e.g. Obeso et al. 2010). It should be noted that even without DAergic neuron loss, elevated levels of H₂O₂ would be expected to cause a net decrease in DAergic transmission. Increased H₂O₂ generation (or impaired metabolism) could lead to suppression of axonal DA release via K_ATP channel activation, which would be compounded by H₂O₂-dependent inhibition of SNc DAergic neuron excitability, resulting in functional DA denervation of target regions, like dorsal striatum (Bao et al. 2005; Avshalumov et al. 2005) (Fig.8). At the same time, the increased excitability of SNr GABAergic output neurons via H₂O₂ and TRPM2 channels (as seen in guinea pig SNr) would further exaggerate motor inhibition (Fig.8).

Two final points about H₂O₂-dependent regulation of neuronal signalling are that: (1) K_ATP and TRPM2 channels are expressed by many neurons in addition to those discussed here, so that modulation by H₂O₂ is likely to be widespread; and (2) additional targets for H₂O₂-dependent regulation are emerging, including GABA receptors that mediate inhibitory synaptic transmission (Accardi et al. 2014; Penna et al. 2014). In this light, it is also likely that dynamic cellular modulation by H₂O₂ is not limited to the CNS. For example, there are established functional roles for K_ATP channels in pancreatic β-cells, cardiac myocytes and muscle (McTaggart et al. 2010; Flagg et al. 2010; Coetzee, 2013), with emerging evidence for TRPM2 channels in β-cell function, as well (Uchida & Tominaga, 2014). Moreover, K_ATP channels in β-cells and cardiac myocytes are sensitive to exogenous H₂O₂ (Ichinari et al. 1996; Tokube et al. 1998; Krippeit-Drews et al. 1999), supporting the idea that metabolically generated H₂O₂ may be poised to provide modulatory signals in excitable cells throughout the body.

Glossary

AMPAR

AMPA receptor

DA

dopamine

DCF

dichlorofluorescein

FFA

flufenamic acid

FI

fluorescence intensity

GSH

glutathione

K_ATP channel

ATP-sensitive K⁺ channel

MAO

monoamine oxidase

MCS

mercaptosuccinate

MSN

medium spiny neuron

NOX

NADPH oxidase

RF1

Redoxfluor-1

ROS

reactive oxygen species

SNc

substantia nigra pars compacta

SNr

substantia nigra pars reticulata

TRP channel

transient receptor potential channel

Additional information

Competing interests

The authors declare no competing financial interest.

Funding

The authors gratefully acknowledge support from the National Institutes of Health, grants R01 NS036362 (M.E.R.), F32 NS063656 (C.R.L.) and T32 DA007254 (B.O'N.), and from the Olympia and Attilio Ricciardi Research Fund.

Author's present address

C. R. Lee: Department of Cell Biology and Neuroscience, Rutgers, The State University of New Jersey, 604 Allison Road, Piscataway, NJ 08854, USA.