Mitochondrial oxidant stress in locus coeruleus is regulated by activity and nitric oxide synthase

Summary

Loss of noradrenergic locus coeruleus (LC) neurons is a prominent feature of aging–related neurodegenerative diseases, like Parkinson’s disease (PD). The basis of this vulnerability is not understood. To explore possible physiological determinants, LC neurons were studied using electrophysiological and optical approaches in ex vivo mouse brain slices. These studies revealed that autonomous activity in LC neurons was accompanied by oscillations in dendritic Ca²⁺ concentration attributable to opening of L–type Ca²⁺ channels. This oscillation elevated mitochondrial oxidant stress and was attenuated by inhibition of nitric oxide synthase. The relationship between activity and stress was malleable, as arousal and carbon dioxide, each increased the spike rate, but differentially affected mitochondrial oxidant stress. Oxidant stress also was increased in an animal model of PD. Thus, our results point to activity–dependent Ca²⁺ entry and a resulting mitochondrial oxidant stress as factors contributing to the vulnerability of LC neurons.

Noradrenergic (NA) LC neurons innervate much of the brain, serving to maintain wakefulness as well as to modulate neural activity and plasticity during periods of arousal and stress ^1,2. Despite their importance, this small group of brainstem neurons is vulnerable to aging and aging–related neurodegenerative diseases, like PD and Alzheimer’s disease. The loss of LC neurons might contribute to many of the non–motor symptoms that accompany these diseases, including memory deficits, depression, and daytime hypersomnolence ³.

Why LC neurons are at risk is not clear. One theory of the aging–related decline in neural function is that it reflects a bioenergetic insufficiency stemming from mitochondrial dysfunction ^4,5. Neurons require mitochondrial oxidative phosphorylation to meet their bioenergetic needs. In regions like the substantia nigra pars compacta (SNc), where there is a clear aging–related decline of neuronal number and function, mitochondrial DNA deletions of the type produced by oxidant stress are significantly higher than in unaffected brain regions⁶. Thus, declining mitochondrial function is likely to be a consequence of cumulative oxidant stress ⁵. Recessive genetic mutations that increase the risk of PD boost this oxidant stress and create deficits in mitochondrial quality control that could amplify the long–term consequences of oxidant damage⁷.

Mitochondrial oxidant stress can arise from either extrinsic or intrinsic sources. Extrinsic oxidant stress can arise when non–mitochondrial processes, such as lysosomal degradation of proteins, generate reactive oxygen species (ROS) that enter mitochondria ^8,9. Intrinsic oxidant stress can arise when ROS are generated by electron leakage from the electron transport chain (ETC) ⁷. This oxidant stress can be amplified by genetic or pharmacological perturbations that alter the balance between ROS generation and clearance. In SNc dopaminergic neurons, whose loss is responsible for the cardinal motor symptoms of PD¹⁰, intrinsic mitochondrial oxidant stress has been traced to Ca²⁺ entry through L–type channels during autonomous pacemaking ¹¹. Ca²⁺ entering through these channels is weakly buffered by cytosolic proteins, allowing it to be taken up by the endoplasmic reticulum (ER) and then passed to mitochondria ¹². Mitochondrial Ca²⁺ entry de–represses enzymes of the tricarboxylic acid cycle, increasing the production of reducing equivalents for the electron transport chain and respiration¹³. However, the precise mechanism by which mitochondrial Ca²⁺ augments oxidant generation is not fully established.

The studies reported here draw strong parallels between the physiological determinants of vulnerability in SNc and LC neurons, showing that activity–dependent opening of L–type Ca²⁺ channels leads to mitochondrial oxidant stress. As in SNc dopaminergic neurons, this stress is exacerbated by deletion of DJ–1,a gene linked to an early onset form of PD, and alleviated by antagonism of L–type Ca²⁺ channels. Our work on LC neurons makes four additional points of importance. First, L–type channels with a Cav1.3 pore–forming subunit make a significant contribution to Ca²⁺ oscillations. Second, the endoplasmic reticulum plays an important role in the genesis of mitochondrial oxidant stress. Third, the relationship between activity and oxidant stress is not fixed, but subject to modulation. Fourth, activity–dependent oxidant stress depends on a mitochondrial form of nitric oxide synthase (NOS).

Results

L–type channels are engaged during neuronal spiking in LC

Initially, the LC region was mapped using immunocytochemical techniques in ex vivo brain slices so that NA neurons were reliably sampled (Fig. 1a, Supplementary Figure 1). In addition, neurons were filled with biocytin and subsequently reconstructed to verify their identity (Fig. 1a). As previously reported ¹⁴, LC NA neurons in ex vivo brain slices at physiological temperatures are spontaneously active, spiking at 1–6 spikes/sec (Fig. 1b). Their spiking rate did not change with the addition of glutamatergic and GABAergic synaptic blockers, suggesting that LC neurons were autonomous pacemakers ¹⁴. Another signature feature of LC neurons described previously ^15-18 was the presence of small spikelets (20–30 mV amplitude) following blockade of Nav1 channels with tetrodoxin (TTX) (Fig. 1b). The spikes of LC neurons were broad, being 2.5 msec in duration at half–amplitude (Fig. 1c).

Figure 1. LC neurons were autonomous pacemakers with broad action potential spikes.

(a) Biocytin–labeled neuron (red, stained with streptavidin–conjugated to alexa594) colocalised with neurons immunoreactive for the LC biomarker tyrosine hydroxylase (TH). To the right, a representative reconstruction of a biocytin–filled LC neuron is shown. (b) Representative LC recording displaying autonomous spiking activity, recordings performed in the presence of glutamate and GABA receptor synaptic blockers. Bottom panel shows a representative trace of LC recording with spikelet activity following blockade of Nav1 channels with TTX (1 μM). Under our recording conditions, more than 90% of the LC neurons recorded displayed spikelet activity (n=20 neurons, from 10 mice). (c) Representative traces showing spike width in LC neurons (n=5 neurons, from 3 mice, median=2.4 sec) compared to globus pallidus (GP) neurons (n=6 neurons, from 3 mice, median=0.28 sec). Right panel shows box–plot quantification with spike widths in LC neurons significantly wider than GP neurons (p=0.004). Data presented as whisker box plots displaying median and interquartile ranges, and analyzed using Mann-Whitney test

Previous studies using somatic recording have implicated engagement of Cav1 L–type channels during autonomous spiking in LC neurons ^14,16,17. To pursue this suggestion, LC neurons were loaded with a low concentration of Ca²⁺–sensitive dye (Fluo–4, 50 μM) and two photon laser scanning microscopy (2PLSM) was used to monitor changes in intracellular Ca²⁺ concentration at dendritic locations. Autonomous spiking was accompanied by phase–locked fluctuations in dendritic Ca²⁺ concentration (Fig. 2a). Moreover, consistent with previous studies, these dendritic Ca²⁺ oscillations were still present following blockade of Nav1 channels with TTX and were eliminated by bath application of the Cav1 channel antagonist isradipine (1 μM) (Fig. 2a)^15,16.

Figure 2. Engagement of L–type channels mediated dendritic Ca²⁺ oscillations during spiking activity in LC.

(a) Left panel shows a schematic of an LC neuron illustrating of our whole–cell current clamp–recording configuration of spiking activity synchronized to 2PLSM dendritic Ca²⁺ line–scan imaging. Spiking activity is accompanied with phase–locked Ca²⁺ oscillations in distal dendrites (80–100 μm away from soma), detected in full–spike mode (control) or in the presence of TTX. Antagonism of L–type Ca²⁺ channels with 1 μM isradipine eliminates Ca²⁺ oscillations in the presence of TTX (n=4 neurons from 4 mice; p<0.001, Wilcoxon matched–pairs signed rank test). (b) Quantitative PCR analysis demonstrates mRNA expression of both Cav1.2 and Cav1.3 L–type channel subunits in LC neurons (n=6 mice). (c) Representative spiking activity traces in the presence of TTX before and after selective antagonism of Cav1.3 channels with 50 μM BPN4689, and non–selective antagonism of Cav1.2 and Cav1.3 with 1 μM isradipine. Antagonism of Cav1.3 channels with BPN4689 significantly decreased spike amplitude (p=0.017), frequency (p=0.018), and amplitude of dendritic Ca²⁺ oscillations (p=0.023) in LC neurons (n=5 neurons from 5 mice). Residual spiking activity and Ca²⁺ oscillations were attenuated completely by 1 μM isradipine suggesting that Cav1.2 channels mediated the residual activity (p<0.001). Data presented as whisker box plots displaying median and interquartile ranges, and analyzed using Mann-Whitney test.

To determine the molecular identity of the channels expressed by LC neurons, they were subjected to quantitative PCR (qPCR) analysis. LC neurons expressed mRNA for both the Cav1.3 pore–forming subunit of the L–type channel and the Cav1.2 subunit (Fig. 2b). Functional expression of Cav1channels was verified by somatic voltage clamp measurements (Supplementary Figure 2). Application of the Cav1.3 selective antagonist BPN4689 (50 μM) ¹⁹ significantly attenuated Ca²⁺ spikes and intracellular Ca²⁺ oscillations in the presence of TTX and the residual oscillatory activity was eliminated by the non–selective L–type channel antagonist isradipine (Fig. 2c).

Although antagonism of Cav1 channels eliminated the dendritic fluctuations in Ca²⁺ concentration, it did not alter pacemaking rate or regularity (Fig. 3a), arguing that, as in SNc DA neurons, Cav1 channels contribute to pacemaking but are not essential for its generation ²⁰. In agreement with this inference, in the presence of TTX (1 μM), autonomous spiking continued, but at a slower rate (Fig. 3b) – suggesting that proximally generated Na⁺ spikes were pacing dendritic Ca²⁺ spikes. To test this proposition, ectopic spikes were initiated in the somatic region by injection of brief current pulses through the patch electrode while monitoring dendritic Ca²⁺ fluctuations. These ectopic somatic spikes reset the dendritic Ca²⁺ oscillations, as predicted (Supplementary Figure 3).

Figure 3. L–type Ca²⁺ channels were not essential to sustain autonomous spiking but necessary under conditions of elevated spike rate.

(a) Representative traces showing that antagonism of L–type channels with isradipine (5 μM) does not affect spike rate in LC neurons (n=4 neurons from 4 mice, p=0.700). (b) Box–plot quantification showing blockade of Nav1 channels slows down spike rate (p=0.032), and leaving smaller amplitude spikelets attenuated by antagonism of L–type Ca²⁺ channels with isradipine (1 μM) (n=10 neurons from 10 mice, p<0.001). (c) Representative perforated–patch current clamp recordings of LC neurons following high potassium (K⁺), leading to elevated spike rate. Under control conditions (n=3 neurons from 3 mice), elevated spike rate is sustained for as long as recordings were held. In the presence of 500 nM isradipine, LC neurons failed to continue spiking after 20 minutes of high K⁺ (n=3 neurons from 3 mice), suggesting that L–type channels are necessary during periods of elevated spike rates. Data presented as whisker box plots displaying median and interquartile ranges, and analyzed using Mann-Whitney test.

Quantitative PCR analysis of LC neurons was used to identify channels that help support pacemaking in addition to Cav1 channels. This revealed robust expression of mRNA for voltage–dependent Na⁺ channels (Nav1.1, 1.2,1.6) and Na⁺ ‘leak’ channels (NALCN). Both types of Na+ channels are inward conducting, causing depolarization at sub–threshold membrane potentials. The expression of mRNA for NALCN was significantly greater in LC neurons than in SNc DA neurons, suggesting a more central role in pacemaking. To verify functional expression of NALCN channels, extracellular Na⁺ was replaced with channel impermeable N–methyl–D–glucamine in the presence of TTX and isradipine (to eliminate the contribution of Nav1 and Cav1 channels respectively); this intervention hyperpolarized LC neurons by ~20 mV, supporting the proposition that NALCN channels make an important contribution to the membrane potential near spike threshold (Supplementary Figure 4).

Although L–type channels were not required for pacemaking, they might enhance the ability of the cell to sustain spiking when driven by extrinsic inputs ²⁰. This robustness could be manifested by the ability to sustain an elevated spike rate, as this elevation would increase slow inactivation of Nav1 Na⁺ channels that normally drive pacemaking. To test this idea, spike rate was increased by elevating the extracellular [K⁺] while monitoring spike rate for long periods in the non–invasive perforated–patch configuration. In untreated LC neurons, increasing the extracellular [K⁺] to 17 mM (from 2.5 mM) led to a sustained elevation in spike rate (Fig. 3c). Antagonizing L–type Ca²⁺ channels had no effect on basal spiking rate either before or immediately after elevating extracellular K⁺ concentration; however, after approximately 20 minutes, spike rate fell in the presence of isradipine (Fig. 3c), suggesting that L–type channels help to maintain spiking in this condition.

Spiking triggers Ca²⁺–induced mitochondrial oxidant stress

In SNc DA neurons, Ca²⁺ is less efficiently buffered by cytosolic proteins than in neighboring VTA DA neurons, allowing it to diffuse in the cytoplasm more readily ²¹. In LC neurons, the situation was similar. Intrinsic Ca²⁺ buffering was estimated using the added extrinsic buffer method ²². The amplitude and decay time constant of the fluorescence signal evoked by a spike was monitored continuously after rupturing the membrane at the tip of a cell–attached patch electrode filled with Fluo–4 (200 μM). As the dye filled the cell, the relative amplitude of the fluorescence signal fell and the time constant rose (Fig 4a, Supplementary Figure 5). These parameters allowed the intrinsic buffering (K_B) to be estimated (Fig. 4a). The median K_B for LC neurons was just greater than 50 (the smaller the number the lower the intrinsic buffering). This K_B estimate was similar to that for SNc DA neurons and significantly lower than for neurons with strong Ca²⁺ buffering (e.g., cerebellar Purkinje neurons K_B=2000, ²³).

Figure 4. Endoplasmic reticulum sequestered cytosolic Ca²⁺ under conditions of low intrinsic Ca²⁺ buffering.

(a) Representative somatic Ca²⁺ transients evoked by single spikes sampled at different concentrations of Fluo4. Ca²⁺ transients displayed decreased transient amplitude and increased decay time with increasing Fluo4. The change in intracellular Ca²⁺ concentration ([ΔCa²⁺]) was calculated and then plotted against the calculated added extrinsic buffer (K_F) (middle panel), and data points were fitted using a linear regression to compute endogenous buffering capacity (K_B), calculated from the negative x–intercept. Box plot quantification revealed a median K_B in LC neurons of 58 (n=4 neurons from 4 mice). (b)Representative dendritic Ca²⁺ traces averaged from pacemaking spike–triggered Ca²⁺ transients before and after antagonism of ryanodine receptors (RyR) with 10 μM ryanodine. Antagonism of RyRs significantly decreased the amplitude of Ca²⁺ transients, consistent with Ca²⁺–induced Ca²⁺ release (CICR) occurring during autonomous spiking (n=5 neurons from 5 mice, p=0.028). Data presented as whisker box plots displaying median and interquartile ranges, and analyzed using Mann-Whitney test.

Given the relatively low intrinsic buffering capacity of LC neurons, Ca²⁺ entering the cytoplasm could reach the ER, potentially evoking Ca²⁺–induced Ca²⁺ release (CICR)¹². To test this possibility, RYRs were antagonized by application of ryanodine (10 μM). Doing so decreased the dendritic Ca²⁺ transients evoked by pacemaking, demonstrating that RYR–dependent CICR was contributing to the observed Ca²⁺ oscillation (Fig. 4b).

In SNc DA neurons, Ca²⁺ entry through plasma membrane L–type channels leads to increased mitochondrial oxidant stress. To determine if the situation was the same in LC neurons, ex vivo slices were prepared from transgenic mice expressing a mitochondrially targeted ratiometric redox probe mito–roGFP ^11,24 under control of the cytomegalovirus (CMV) promoter (Fig.5a). LC neurons exhibited a basal mitochondrial oxidant stress that was typically 40–50% of the dynamic range of the probe (Fig. 5a). Antagonizing plasma membrane L–type channels with isradipine dose–dependently diminished this stress without altering pacemaking (Fig. 5b–c). Blocking Ca²⁺ entry into mitochondria with the mitochondrial uniporter (MCU) blocker RU360 (10 μM), also virtually eliminated oxidant stress (Fig. 5b–c) without changing pacemaking rate (data not shown).

Figure 5. L–type channels triggered mitochondrial oxidant stress in LC.

(a) Mito–roGFP positive LC neuron immunoreactive to tyrosine hydroxylase (TH). (b) Mito-roGFP calibration with relative oxidation of 0.4. Isradipine attenuated oxidation in a dose-dependent manner (n=5 neurons control, 5 mice, n=4 neurons 200 nM isradipine, 4 mice, p=0.028; n=5 neurons 1 μM isradipine, 4 mice, p=0.002). Ryanodine (n=6 neurons, 5 mice, p=0.001) and Ru360 (n=4 neurons,4 mice, p=0.007) also attenuated mitochondrial oxidation. (c) Ca²⁺ entry through L–channels is pumped into ER via the sarcoendoplasmic reticulum Ca²⁺ ATPase (SERCA). Mitochondrial–associated–membranes (MAMs) allow Ca²⁺release from inositol–1,4,5–trisphosphate receptors (IP₃Rs) and RyRs into the mitochondria through MCU. (d)Mitochondrial–membrane potential (MMP) oscillations of LC and SNc neurons using TMRM. (e) UCPs antagonized with genipin (100 μM) attenuated the amplitude (p=0.031) and frequency (p=0.016) of MMP oscillations. (f) Isradipine decreased the amplitude (p=0.047) and frequency (p=0.017) of MMP oscillations. Panel e–f , n=6 neurons, 3 mice per antagonist, Wilcoxon–matched pairs signed rank test. (g) Loss of DJ–1 function (DJ–1^−/−) exacerbated mitochondrial oxidant stress in DJ–1^−/− (red) compared to WT (p=0.029, n=4 WT neurons, 3 mice, n=4 DJ–1^−/− neurons, 3 mice). Neurons with 1 μM isradipine (green) for 1 hr attenuated oxidant stress in DJ–1^−/− (p=0.032, n=5 DJ–1^−/− neurons with isradipine, 3 mice), Data presented as whisker box plots with median and interquartile ranges, and analyzed using Mann-Whitney test compared to control.

Mitochondrial Ca²⁺ entry is widely thought to occur at junctions between the ER and mitochondrion where the local Ca²⁺ concentration can rise into the micromolar range ²⁵. Both RYRs and inositol trisphosphate receptors (IP₃Rs) populate these junctions (referred to as mitochondria–associated membrane or MAMs)²⁶. Consistent with the idea that mitochondrial Ca²⁺ entry was occurring at MAMs, antagonizing RYRs profoundly lowered mitochondrial oxidant stress, diminishing it by roughly the same amount as blocking the MCU with RU360 (Fig. 5b–c). These results suggest that plasma membrane Ca²⁺ entry through L–type channels is sequestered by the ER and then passed through RYRs to mitochondria, leading to oxidant stress (Fig. 5b–c).

Mitochondrial oxidant stress was accompanied by fluctuations in the mitochondrial membrane potential as measured with the cationic dye tetramethyl rhodamine methylester (TMRM). These potential fluctuations had a frequency in the range of 10–30 milliHz and amplitudes of about 10% of the TMRM dynamic range (Fig. 5d), suggesting that they were attributable to opening and closing of an ion channel in the inner mitochondrial membrane. The frequency and amplitude of fluctuations were diminished by genipin (100 μM), an inhibitor of uncoupling proteins (UCPs) – ion channels that are capable of depolarizing the inner mitochondrial membrane ²⁷ (Fig. 5e). The potential fluctuations also were attenuated in frequency and amplitude by diminishing Ca²⁺ entry through plasma membrane L–type channels (Fig. 5f).

Oxidant stress is elevated in LC neurons lacking DJ-1

Loss of function of DJ–1 protein is associated with autosomal recessive early onset forms of PD in humans⁷, and has been described as a redox–sensitive protein that participates in orchestrating oxidant defense ²⁸. Previous studies have shown that deletion of DJ–1 exacerbated mitochondrial oxidant stress in SNc and DMV neurons^11,29. If mitochondrial oxidant stress is a driver of LC degeneration, it should be increased in mice lacking functional DJ–1. To test this hypothesis, stereotaxic injection of an AAV viral vector was used to deliver a CMV–mito–roGFP expression construct to LC neurons in DJ–1 knockout (DJ–1^−/−) and wild–type mice. Ten to fourteen days later, brain slices were prepared and oxidant stress assayed as described above. These experiments were done at room temperature, since previous studies showed that probe oxidation was nearly complete at physiological temperatures¹¹. In LC neurons from DJ–1^−/−, mitochondrial oxidant stress was significantly greater than in wild–type mice (Fig. 5g). Moreover, antagonism of L–type channels significantly attenuated mitochondrial oxidant stress in DJ–1^−/−neurons (Fig. 5g).

Extrinsic signals differentially regulate oxidant stress

The activity of LC neurons varies with behavioral state ¹, raising the possibility that oxidant stress varies in parallel. For example, during sleep the activity of LC NA neurons is suppressed by GABAergic input ³⁰, which could lessen mitochondrial stress. Indeed, application of the GABA_A receptor agonist muscimol hyperpolarized LC neurons, stopped spiking and dramatically diminished mitochondrial oxidant stress (Fig. 6a–b). Conversely, during arousal the discharge rate of LC NA neurons increases, which could augment mitochondrial oxidant stress. The increase in LC discharge rate during arousal is attributed to the release of orexin in the LC by hypothalamic neurons ^31-33. Bath application of orexin increased the discharge rate of LC neurons in our preparation (as predicted) (Fig. 6a). Yet surprisingly, orexin attenuated, rather than increased, mitochondrial oxidant stress (Fig. 6b). One possible explanation of this change is that orexin altered the pacemaking mechanism to rely more upon Na⁺ channels and less upon the opening of L–type Ca²⁺ channels³⁴. To test this hypothesis, two experiments were performed. First, if orexin inhibited L–type channels, then spiking in the presence of TTX (which depends upon L–type channels) should be impaired. Indeed, orexin virtually eliminated spiking in the presence of TTX (Fig. 6c). Second, inhibition of L–type channels should decrease dendritic Ca²⁺ transients during normal pacemaking. As predicted, orexin significantly diminished these transients (Fig. 6d). Thus, mitochondrial oxidant stress in LC neurons is state dependent, falling both during GABAergic suppression of activity and during orexin–mediated acceleration of activity.

Figure 6. Modulation of spiking and mitochondrial oxidant stress.

(a) Muscimol (10 μM, MUS) silenced pacemaking whereas orexin (500 nM) nearly doubled spike rate compared to control (n=4 neurons from 2 mice with MUS, p=0.002; n=8 neurons orexin from 8 mice, p=0.022,). (b) Muscimol and orexin decreased relative oxidation in LC neurons (n=5 control neurons from 5 mice; n=4 neurons with MUS from 3 mice, p=0.003; n=6 orexin-treated neurons from 3 mice, p=0.010). (c) Orexin attenuated spikelet amplitude in LC neurons (n=4 neurons from 4 mice, p=0.038). (d) Dendritic Ca²⁺ traces averaged from pacemaking–triggered Ca²⁺ transients before and after bath application of orexin. Orexin decreased the amplitude of dendritic Ca²⁺ transients despite an increase in spike rate (n=4 neurons from 4 mice, p=0.014). (e) Pacemaking firing in LC neurons recorded in cell–attached mode under control conditions of 95% oxygen and 5% carbon dioxide, pH 7.4, or hypercapnia, induced by bubbling ACSF with a gas mixture of 85% oxygen and 15% carbon dioxide. From the 3 neurons (from 3 mice) recorded before and after hypercapnia, all three neurons increased spike rate under hypercapnic conditions (p=0.038). (f) LC neurons exposed to 1 hour of hypercapnia (n=8 neurons from 3 mice, p=0.023) displayed increased mitochondrial oxidant stress when compared to control neurons (n=5 neurons from 3 mice). Data presented as whisker box plots displaying median and interquartile ranges, and analyzed using Mann-Whitney test comparing to respective controls.

Hypercapnia also increases the spiking rate of LC neurons ^15,16, which may contribute to the modulation of breathing. Moreover, hypercapnia–induced acceleration in the spiking of LC neurons is dependent upon L–type Ca²⁺ channels ^15,16, suggesting hypercapnia might increase mitochondrial oxidant stress. To test this hypothesis, LC neurons in brain slices were bathed in a solution bubbled with gas containing 15% carbon dioxide (CO₂) to induce hypercapnia. As shown previously, hypercapnia increased the spiking rate of LC neurons monitored in cell–attached recording mode (Fig. 6e). To determine if mitochondrial oxidant stress was affected, slices were bathed in the hypercapnic solution for 1 hour and then mitochondrial redox status was assessed using 2PLSM. As predicted, hypercapnia significantly increased mitochondrial oxidant stress relative to slices maintained in the normal 5% CO₂ environment (Fig. 6f).

Oxidant stress was dependent upon nitric oxide synthase

How does Ca²⁺ entry into mitochondria elevate oxidant stress? There is no consensus on this point. Mitochondrial Ca²⁺ influx de–represses matrix dehydrogenases of the tricarboxylic acid (TCA) cycle, leading to increased generation of electron donors for the ETC¹³. In physiological circumstances, increasing the proximal flux of electrons into the ETC could increase the production of superoxide and other reactive oxygen species (ROS) ³⁵. However, there are other ways in which mitochondrial Ca²⁺ entry might increase ROS production. For example, sustained Ca²⁺ influx into mitochondria could stimulate a mitochondrial form of NOS ^36,37. NO produced by NOS activation competes with O₂ at cytochrome oxidase, thereby changing the redox status at earlier points along the ETC and increasing the production of superoxide ³⁷. To determine if these mechanisms were at work in LC neurons, cells were loaded with the NO–sensitive probe 4–amino–5–methylamino– 2,7–difluorescein (DAF–FM) with a local perfusion pipette;then neurons were imaged using 2PLSM. In unperturbed LC neurons, DAF–FM fluorescence was readily detectable (Fig. 7a, Supplementary Figure 6). The NO signal was decreased by pre–incubation with the NOS inhibitor L–N^G–nitroarginine methyl ester (L–NAME, 100 μM) (Fig. 7a). Antagonizing plasma membrane L–type channels also decreased the DAF–FM signal (Fig. 7b). Although the source of NO could not be determined directly because the dye was largely cytosolic, blocking Ca²⁺ entry into mitochondria with RU–360 decreased the cytosolic NO signal (Fig. 7b), arguing that mitochondria are a site of origin. Conversely, blocking Ca²⁺ efflux from the mitochondria by antagonizing the mitochondrial Na⁺–Ca²⁺ exchange (NCX) with 5 μM CGP37157^26,38, led to an increased cytosolic NO signal (Fig. 7b), arguing that increasing intramitochondrial [Ca²⁺] stimulates NO production.

Figure 7. Nitric oxide synthase contributed to oxidant stress.

(a) LC neurons labeled with the nitric oxide (NO) fluorescent probe DAF–FM with 100 μM L–NAME relative to control (n=3 mice). L–NAME attenuated NO production shown as decreased DAF-FM fluorescence. (b) NO production was decreased with isradipine (1 μM) or Ru360 (10 μM) (n=7 neurons isradipine, p=0.002; n=8 neurons Ru360, p=0.002, data collected from 3 mice). Antagonism of mitochondrial Na⁺–Ca²⁺ exchange (NCX) with 5 μM CGP37157 led to enhanced production of NO (n=6 neurons from 3 mice with CGP37157, p=0.001). Schematic displays the signals tracing Cav1 channel-mediated Ca²⁺ entry to increased intramitochondrial Ca²⁺. Ca²⁺-induced stimulation of NO impairs ETC and leads to generation of mitochondrial oxidant stress. (c) NOS inhibition in SNc and LC neurons decrease mitochondrial oxidant stress (n=7 control SNc neurons from 5 mice, n=6 L–NAME–treated SNc neurons from 5 mice, p<0.001; n=6 L–NAME–treated LC neurons from 4 mice, p=0.007). Control relative oxidation values of LC neurons were replotted from fig. 5b for comparison purposes. A different NOS inhibitor, L–NNA (10 μM) also decreased mitochondrial oxidation in LC (n=6 L–NNA treated neurons from 4 mice, p=0.036). (d) L–NAME decreased the amplitude (p=0.029)and frequency (p=0.014)of mitochondrial membrane potential oscillations in LC neurons measured with TMRM (n=4 neurons from 4 mice). Data presented as whisker box plots displaying median and interquartile ranges, and analyzed using Mann-Whitney test relative to controls.

To determine if NO production increased mitochondrial oxidant stress, slices were incubated with L–NAME (100 μM) prior to determining the redox status of mitochondria. Previous work has shown that NO does not directly alter mito–roGFP fluorescence, so any change in these experiments is attributable to altered dithiol formation ³⁹. Pre–incubation with L–NAME significantly diminished mitochondrial oxidant stress in LC neurons (Fig. 7c). L–N^G–nitroarginine (L–NNA, 10 μM), a different NOS inhibitor also decreased mitochondrial oxidant stress (Fig. 7c). Exposure of SNc DA neurons to L–NAME had a very similar effect (Fig. 7c), suggesting that NOS activity was a significant source of oxidant stress in both cell types. To provide an additional test of the inference that NOS activity was increasing mitochondrial oxidant stress, TMRM was used to monitor mitochondrial membrane potential. As predicted, L–NAME significantly diminished the amplitude and frequency of fluctuations in mitochondrial potential (Fig. 7d).

Discussion

Three conclusions can be drawn from the studies presented. First, LC NA neurons have a robust pacemaking phenotype that leads to basal mitochondrial oxidant stress that is exacerbated in a genetic model of PD. Second, although mitochondrial stress generally tracked spike activity and Ca²⁺ entry through L–type channels, this relationship was subject to modulation. Third, the activity–dependent oxidant stress stems in large part from stimulation of a mitochondrial form of NOS. These conclusions provide a framework for understanding how age and disease might lead to the loss of LC NA neurons, as well as how this loss might be averted.

Spiking elevates oxidant stress in LC neurons

Previous work on LC neurons has not provided a mechanistic foundation for understanding their sensitivity to aging and aging–related neurodegenerative diseases, like PD and AD. A number of studies have shown that LC neurons are spontaneously active and this activity is autonomously generated ^14,15. Our results were consistent with these studies and extend them in several important respects.

Using 2PLSM Ca²⁺ imaging approaches, it was revealed that autonomous pacemaking was accompanied by large fluctuations in dendritic Ca²⁺ concentration attributable to opening of a mixture of L–type channels with Cav1.3 and Cav1.2 pore–forming subunits. As in SNc DA neurons, the low–threshold Cav1.3 L–type channels in LC neurons were active throughout the interspike interval, leading to a sustained Ca²⁺ influx. Although Ca²⁺ current through L–type channels contributed to the inward currents driving LC pacemaking, they were not necessary to maintain it, as low micromolar DHP concentrations virtually eliminated dendritic fluctuations in intracellular Ca²⁺ concentration, but had no effect on pacemaking rate and regularity.

Although not necessary for pacemaking, L–type channels increased the robustness of pacemaking. In fact, dendritic L–type Ca²⁺ channels sustained pacemaking after blocking Nav1 channels with TTX, albeit at a lower rate. This contrasts with the situation in SNc DA neurons where L–type channels are capable of only sustaining a sub–threshold dendritic oscillation after block of Nav1 channels ⁴⁰. For this type of Ca²⁺–dependent dendritic pacemaking to work, Ca²⁺must be weakly buffered by cytosolic proteins so as to allow free [Ca²⁺] to control the onset of repolarization by Ca²⁺ activated K⁺ channels ⁴¹. As a consequence, much of the entering Ca²⁺ must be slowly sequestered by intracellular organelles. The ER is widely viewed as a key mediator of intracellular Ca²⁺ buffering on this time scale ⁴². Although ER Ca²⁺ content was not monitored, it was found that blocking ER RYRs diminished cytosolic Ca²⁺ transients during pacemaking, suggesting that the ER was playing an important role in intracellular Ca²⁺ dynamics. In many cell types, ER buffering leads to Ca²⁺ being shuttled to mitochondria at specialized junctions, so–called MAMs ²⁶. RYRs are localized on the ER side of MAMs and MCUs are found on the mitochondrial side, creating a mechanism by which Ca²⁺ can move from the ER to mitochondria. This same arrangement appears to be present in LC neurons, where Ca²⁺ shuttled into mitochondrial was sufficient to create an oxidant stress as measured with a matrix–targeted redox–sensitive variant of GFP.

Given the negative consequences of sustained Ca²⁺ entry, why do LC NA neurons express such high levels of the Cav1 channel? One possibility is that Ca²⁺ drives the synthetic machinery necessary for proper functioning of LC neurons ⁴³. Another possibility is that the pacemaking function of LC NA neurons is essential for organismal fitness and survival. During waking states, even a transient loss of spiking in LC NA neurons results in impaired brain function, which could be catastrophic (if, for example, one were being pursued by a predator). This would create evolutionary pressure to put in place redundant pacemaking mechanisms that minimize the likelihood of failure even with sustained excitation. In slices, LC NA neurons were able to continue spiking at elevated rates for as long as the recording could be sustained. However, in the presence of L–type channel antagonists, spiking was less robust, failing within about 20 minutes after increasing extracellular K⁺. This difference is attributable to the fact that, unlike Cav1 Ca²⁺ channels, Nav1 Na⁺ channels are subject to slow, voltage–dependent inactivation with sustained depolarization. So, it is easy to imagine that there would be strong evolutionary pressure to creating a ‘backup’ dendritic oscillator that was not going to fail under the same conditions – an oscillator like that involving Cav1 Ca²⁺ channels. The cost of this design choice is the sustained elevation in mitochondrial oxidant stress. However, this cost is one that could be tolerated, as there are biochemical systems in place to mitigate oxidant stress and delay its negative consequences. Indeed, the impact on the function of the LC NA system takes five or six decades to become manifest, which is well beyond the life expectancy of animals with this design, including humans until recently.

Mitochondrial stress was state–dependent

The engagement of the dendritic oscillator was modulated by extrinsic inputs. During sleep, the autonomous activity of LC NA neurons is actively suppressed by GABAergic input ^30,44. This suppression could be mimicked in ex vivo brain slices by application of the GABA_A receptor agonist muscimol. As expected, suppression of pacemaking diminished Ca²⁺ entry and mitochondrial oxidant stress. In contrast, arousal leads to an elevation in the spiking rate of LC neurons, increasing the release of NA in brain regions controlling behavioral activation⁴⁵. One of the drivers of this type of arousal is orexin, which is released by hypothalamic neurons innervating the LC³¹. As anticipated, exogenous orexin significantly increased the discharge rate of LC neurons in brain slices. What was not anticipated was that this increase in discharge rate would lead to lowered mitochondrial oxidant stress. This happened because orexin negatively modulated L–type channel opening while increasing Na⁺ currents, presumably those through NALCN channels. This observation is consistent with early work showing that the effects of orexin on LC spiking was independent of Ca²⁺³⁴. It is not clear at this point how this modulation was brought about, but L–type Ca²⁺ channels are well known to be regulated by G–protein coupled receptors, like those mediating the effects of orexin.

Hypercapnia also increased the spiking rate of LC neurons. Augmented LC spiking is thought modulate brainstem respiratory circuits, promoting increased respiratory rate ⁴⁶. However, unlike the response to orexin, the excitatory effects of hypercapnia depend upon L–type Ca²⁺ channels in LC neurons ^15,16. Accelerating the discharge rate of LC neurons without suppressing the opening of L–type Ca²⁺ channels should lead to increased Ca²⁺ entry and mitochondrial Ca²⁺ loading. Indeed, hypercapnia increased mitochondrial oxidant stress in LC neurons. These results suggest that persistent respiratory insufficiency could have a negative impact on LC and contribute to neurodegenerative decline.

Mitochondrial Ca²⁺ entry stimulated NOS and oxidant stress

Ca²⁺ that enters the mitochondrial matrix stimulates oxidative phosphorylation by de–repressing dehydrogenases, thereby helping to meet bioenergetic needs ³⁷. Our results suggest that another consequence of mitochondrial Ca²⁺ entry is stimulation of NOS. NO production, as measured with cytosolic DAF–FM, was robust in pacemaking LC neurons. Not only was NO production sensitive to Ca²⁺ entry through plasma membrane L–type channels, it was sensitive to antagonism of the mitochondrial uniporter with RU360. Moreover, while blocking entry decreased NO production, disrupting removal of Ca²⁺ from the matrix by NCX increased NO levels. The ability of L–NAME to block NO production suggests the involvement of an enzymatic process, again implicating NOS. Studies in other tissues have shown that a Ca²⁺–dependent form of NOS is localized to mitochondria ³⁶, providing a clear precedent for the observations in LC neurons.

What was more unexpected was the observation that inhibition of NOS decreased the oxidation status of matrix proteins and attenuated the fluctuations in mitochondrial potential driven by that stress. This suggests that NOS contributes to oxidant stress. How? NO generated in proximity to the electron transport chain can compete with O₂ at the binuclear center of cytochrome oxidase, leading to an increase in the reduction state of more proximal electron carriers and to an increase in the generation of superoxide ³⁷.

But what end does this signaling pathway serve? One possibility is that NO generated in response to mitochondrial matrix Ca²⁺ entry helps to prevent Ca²⁺ overload; that is, it serves as a negative feedback regulator of mitochondrial calcium uptake. By inhibiting cytochrome oxidase, NO can slow electron transport, leading to a decrease in mitochondrial membrane potential and a slowing of mitochondrial Ca²⁺ influx. By regulating matrix [Ca²⁺], this mechanism would prevent activation of the mitochondrial permeability transition pore (mPTP), which could inflict lethal injury on the cell ⁷. This model is consistent with the demonstration that TCA enzymes are more sensitive to Ca²⁺ than is mitochondrial NOS ¹³. Thus, modest mitochondrial Ca²⁺ entry can maintain oxidative phosphorylation necessary for neuronal survival⁴⁷, without activating NOS. However, when mitochondrial Ca²⁺ influx increases, NOS–mediated NO production would lead to an inhibition cytochrome oxidase, a slowing the ETC and a diminished driving force for Ca²⁺ entry ³⁷. This ‘braking’ mechanism also provides a compelling rationale for the expression of NOS in the mitochondria, as this localization confers sensitivity to matrix Ca²⁺ that would not be detected by NOS expressed in the cytosol. An unfortunate by–product of this mechanism is the production of superoxide by the ETC ^7,47. However, in most neurons, the engagement of the NOS brake (and the production of superoxide) should be infrequent. LC NA neurons (and SNc DA neurons) are exceptions, where large and sustained mitochondrial Ca²⁺ influx leads to the continuous activation of this protective brake. During a typical human lifespan, the cumulative oxidant damage produced by this mechanism could contribute to mitochondrial DNA defects that impair the ability to generate ATP, leading to bioenergetic failure and cell death ^5,6.

As appealing as the model is, there is an alternative explanation for our DAF–FM data that needs to be considered. For DAF–FM fluorescence to increase with NO exposure, its amide group must be oxidized⁴⁸. Could the dependence of the DAF–FM signal on mitochondrial Ca²⁺ entry arise from ‘enabling’ mitochondrial ROS generation and not mitochondrial localization of NOS? This seems highly unlikely as an extra–mitochondrial NOS localization would leave unexplained the ability of NOS–inhibitors to diminish mitochondrial matrix oxidant stress or oscillations in the inner mitochondrial membrane potential.

Therapeutic strategy in aging–related neurodegeneration

Many of the neurons that succumb to aging and aging–related degenerative diseases have essential physiological functions that require an ability to maintain autonomous spiking under normal and stressful states. In at least three of these cell types – DMV cholinergic neurons, SNc DA neurons and LC NA neurons –the sustained engagement of L–type Ca²⁺ channels enhances the reliability of autonomous spiking, at the cost of increased mitochondrial oxidant stress ^11,29. Our data now confirms that in a mouse model of PD, LC neurons also display exacerbated mitochondrial stress, the latter attenuated in this mouse model by using Cav1 channel antagonists. The engagement of L–type channels during autonomous spiking might be considerably more widespread than previously recognized. Use of Cav1 channel antagonists to lower blood pressure that cross the blood brain barrier have no obvious deleterious neurological consequences for humans and have been linked to reduced risk of developing PD ⁴⁹ and to slowing of progression. These drugs also might have utility in slowing the progression of AD ⁵⁰ and the non–motor symptoms of PD, particularly those potentially dependent upon the loss of LC neurons.

Methods

Brain slice preparation

Acute pontine horizontal slices (220–μm thick) were prepared from wild–type male C57BL/6 mice (Charles River) or DJ–1^−/−mice (backcrossed to C57BL/6) between postnatal ages 21 and 32 days, unless specified otherwise. The handling of mice and all procedures performed on them were approved by the institutional Animal Care and Use Committee and were in accordance with the National Institutes of Health Guide to the Care and Use of Laboratory Animals and Society for Neuroscience guidelines. Mice were anesthetized with a ketamine/xylazine mixture, followed by a transcardial perfusion with ice–cold, oxygenated artificial CSF (ACSF) containing the following (in mm): 125 NaCl, 2.5 KCl, 25 NaHCO₃, 1.25 NaH₂PO₄, 2 CaCl₂, 1 MgCl₂, and 25 dextrose, pH 7.3 (osmolality 315–320 mOsm/L). After perfusion, mice were decapitated and brains removed rapidly, followed by sectioning in ice–cold oxygenated ACSF using a vibratome (VT1000S; Leica Microsystems). Pontine slices were incubated in ACSF at 32–34°C for 30 min before electrophysiological recordings. Slices were then be transferred to a small–volume (<0.5 ml) recording chamber that mounted on a fixed–stage, upright microscope (BX51; Olympus America) equipped with infrared differential interference contrast (0.9 numerical aperture (NA) with de Sénarmont compensation (Olympus). Electrophysiological recordings were performed at 33–34°C, unless specified otherwise. The recording chamber were superfused with carbogen–saturated ACSF with a flow rate of 2–3 ml/min running through an in–line heater (SH–27B with TC–324B controller; Warner Instruments). Neuronal somata and proximal dendrites were visualized by video microscopy at high magnification (60×, 0.9 NA water–immersion objective; Olympus). Brain slices were utilized within 3 hours from slice preparation.

Electrophysiology

Cell–attached and whole–cell patch–clamp recordings were performed in locus ceruleus (LC) neurons located near the floor of the fourth ventricle. LC neurons were recognized as a cluster of neurons with large somatic compartments. Recording patch electrodes (resistance of 3–6 MΩ) were prepared with a Sutter Instruments horizontal puller using borosilicate glass with filament. For current–clamp recordings, patch electrodes were filled with internal solution containing the following (in mM): 135 K–MeSO₄, 5 KCl, 5 HEPES, 0.05 EGTA, 10 phosphocreatine–di(tris), 2 ATP–Mg, and 0.5 GTP–Na, the pH adjusted to 7.3 (osmolality adjusted to 290–300 mOsm/L). The liquid junction potential in our recording ACSF using this internal solution is 7 mV and not corrected for. Conventional tight–seal (>2 GΩ) whole–cell patch–clamp and cell–attached recordings were made on visually identified LC neurons based on (1) size, (2) somatodendritic morphology, and (3) regular spiking between 1 and 6 Hz. Signals were filtered at 1–4 kHz and digitized at 5–20 kHz with a Digidata 1400 (Molecular Devices). For current–clamp recordings, the amplifier bridge circuit were adjusted to compensate for electrode resistance and monitored. For perforated–patch current clamp experiments, internal solution was supplemented with amphotericin–B (Sigma), prepared fresh and added in the internal solution to a final concentration of 180 μg/ml. Alexa568 was added in the internal solution to confirm recordings were done under perforated–patch mode. Electrode capacitance was compensated and if series resistance increased >20% during recording, the data were discarded.

2PLSM Ca²⁺ imaging

For dendritic Ca²⁺ measurements, LC neurons were loaded with internal solutions supplemented with Alexa Fluor 594 (20 μM) and Fluo–4 (50 μM) through the patch pipette. All experiments were performed at 32–34°C. Imaging took place after 15–20 min of dye loading. Images were acquired with an Olympus LUMPFL 60×/1.0 NA water–dipping objective lens. The two–photon excitation source was a Chameleon–ultra1 laser system (690 to 1040 nm) from Coherent Laser Group. Optical signals were acquired using 820 nm excitation (80–MHz pulse repetition frequency and ~250 fs pulse duration) to simultaneously excite Alexa and Fluo–4 dyes. Laser power attenuation was achieved with two Pockels’ cell electro–optic modulators (models M350–80–02–BK and M350–50–02–BK; Con Optics) controlled by PrairieView and WinFluor, repsectively. The two cells are aligned in series to provide: enhanced modulation range for fine control of the excitation dose (0.1% steps over five decades), to limit maximum power, and to serve as a rapid shutter during line scan acquisitions. The laser–scanned images were acquired with a Prairie Technologies Ultima system. The fluorescence emission was collected by non–de–scanned photomultiplier tubes (PMTs). The green (490–560 nm) and red (580–630 nm) fluorescence were each detected by a multi–alkali–cathode PMT. The system digitized the current from detected photons to 12 bits. The laser light transmitted through the sample was collected by the condenser lens and sent through a Dodt contrast tube (Luigs and Neumann)to another PMT (R3896, Hamamatsu) to provide a bright–field transmission image (Prairie USB transmission detector) in registration with the fluorescent images. Fluorescence measurements were taken in a sample plane along dendritic segments (80–100 μm from the soma). Line scan signals were acquired at 6 ms per line and 512 pixels per line with 0.18 μm pixels and 10 μs pixel dwell time.

Endogenous Ca²⁺ buffering capacity

Buffering capacity experiments were adapted to 2PLSM based on protocols described previously by other groups using 200 μM Fluo–4 (K_d = 300 nM, R_f ≈ 85–100)^22,29. LC neurons were current clamped and pacemaking was silenced to –60 mV. Somatic fluorescence was imaged as a function of time upon breaking into the cells while inducing spikes to evoke Ca²⁺ transients. Ca²⁺ signal was detected using 2PLSM as described in the previous section (see above). For data analysis, we utilized equations described by Neher and Augustine⁵¹. In order to measure changes in intracellular Ca²⁺ (Δ[Ca²⁺] = [Ca²⁺]_peak – [Ca²⁺]_o), we calculated first δf = (f – f_o)/f_o from Ca²⁺ transients evoked by the induced spike (2 ms pulse, 1 nA). Δ[Ca²⁺] was calculated from equation (1):

$$\frac{\mathrm{\Delta }[{\mathrm{Ca}}^{2+}]}{{\mathrm{K}}_{\mathrm{d}}}=\frac{{\mathrm{f}}_{\max }}{{\mathrm{f}}_{\mathrm{o}}}(1-{\mathrm{R}}_{\mathrm{f}}^{-1})\frac{(\mathrm{\delta f})}{[({\mathrm{\delta f}}_{\max }-\mathrm{\delta f}){\mathrm{\delta f}}_{\max }]}$$ (1)

The value of δf_max was calculated by a depolarizing current step that achieves a Ca²⁺ plateau to saturate Fluo4 dye. From this value, we calculated resting [Ca²⁺]_o from equation (2):

$${[{\mathrm{Ca}}^{2+}]}_{\mathrm{o}}=\frac{(1-{\mathrm{R}}_{\mathrm{f}}^{-1})}{{\mathrm{\delta f}}_{\max }}-{\mathrm{R}}_{\mathrm{f}}^{-1}$$ (2)

The estimated value of R_f ≈ 85–100 for Fluo4 was obtained from Maravall et al. ²², making R_f^–1 value small and therefore, it was disregarded from equations (1) and (2). The dynamic range of the dye (f_max/f_o) was estimated as 1+δf_max. After determining Δ[Ca²⁺] and related parameters, we then calculated added buffering capacity (K_F) using equation (3):

$${\mathrm{K}}_{\mathrm{F}}=\frac{({\mathrm{K}}_{\mathrm{d}}[\mathrm{Fluo}-4])}{({\mathrm{K}}_{\mathrm{d}}+{[{\mathrm{Ca}}^{2+}]}_{\mathrm{o}})({\mathrm{K}}_{\mathrm{d}}+{[{\mathrm{Ca}}^{2+}]}_{\mathrm{peak}})}$$ (3)

The inverse of Δ[Ca²⁺] was then plotted as a function of added buffering capacity and with a linear regression, the estimated endogenous buffering capacity (K_B) was determined from the linear fit intersecting the negative x–axis.

2PLSM mito–roGFP imaging

Pontine slices from wild–type, roGFP transgenic mice were incubated in ACSF at physiological temperatures 32–34°C followed by a recovery period at room temperature. Experiments were performed at physiological temperatures with the exception of the experiments using the mouse model of PD. For stereotaxic injections of recombinant adenoassociated virus (rAAV) to drive expression of mito–roGFP under the regulation of the CMV promoter, C57BL6 male mice (>P21 postnatal age) were anaesthetized with isoflurane (induction 2.5%, maintenance 1.5%) delivered continuously with 100% medical O₂. Throughout the surgery, mice were placed under a heating pad to avoid hypothermia. Mice were injected subcutaneously with Metacam (0.1mg/ml) and placed in a Kopf stereotaxic frame apparatus with head position to obtain a flat skull between bregma and lambda. A craniotomy was performed using micro–drill to expose brain tissue and recombinant adeno–associated virus (AAV) virus was injected (200 nanoliters) with a glass micropipette (VWR) directly into the LC region using the following stereotaxic coordinates: 5.45 mm posterior to bregma; 1.28 mm lateral to midline; and 3.65 mm deep from the cortical surface. The mitochondrial redox sensitive probe, pFB–AAV–CMV–MTS–roGFP, was generated by cloning EcoR1 –XhoI fragment from TH–SK–MTS–roGFP–pA plasmid into EcoR1–SalI sites of pFB–AAV–CMV– SV40 vector (Virovek). pFB–AAV–CMV–MTS–roGFP–pA DNA was packaged into AAV9–CMV–MTS–roGFP–SV40 by Virovek. The AAVs with 2.4E+13vg/ml titers were injected into LC region to express MTS–roGFP. Following ten to fourteen days post–surgery, brain slices were prepared as described. Optical imaging of roGFP signals acquired using a 920 nm excitation beam in a fixed plane of focus with a pixel size between 0.18–0.21 μm and a 10 μs pixel dwell time. This wavelength was chosen by determining the two–photon excitation spectrum of mito–roGFP in cultured dopaminergic neurons from TH–mito–roGFP mice¹¹. Laser was a Chameleon–Ultra1, 690 nm to 1040 nm, 80MHz ~250 fs sample pulse duration, Coherent Laser Group. The roGFP fluorescence was detected (490–560 nm) by a non–de–scanned GaAsP PMT (H7422PA, Hamamatsu) and a laser scanned Dodt contrast transmission PMT (R3896, Hamamatsu) detector system (Prairie Technologies) that provided a bright–field transmission image. Sixty frames of the roGFP signal were collected in one optical plane at a rate of 3–4 frames per second. Records with drifting baseline (due to photo–bleaching or photo–oxidation of roGFP) were discarded. At the end of all experiments, the maximum and minimum fluorescence of mito–roGFP was determined by application of 2 mM dithiothreitol (DTT) to fully reduce the mitochondria, and then 200 μM aldrithiol (ald) to fully oxidize mitochondria. Relative oxidation was calculated with the equation ‘1–((F–F_ald)/F_DTT–F_ald))’.

2PLSM imaging of mitochondrial membrane potential

Brain slices from wild–type were incubated in 2–4 μM TMRM for 30–60 minutes at 32–34°C; excess dye was washed out with a TMRM–free ACSF solution. Imaging experiments were performed with TMRM–free ACSF solution at physiological temperatures (32–34°C). Fluorescence (580–630 nm) was collected by a GaAsP PMT (H7422PA, Hamamatsu) as part of the Prairie Technologies laser scanning microscope. Fluorescence measurements in a region of interest (ROI) were monitored to ensure stationarity of the signal; samples with a drifting baseline (due to photo–bleaching of dye and/or wash–out) were discarded. Time–series scanning (1000 frames) in a fixed plane were performed with a 10 μs dwell time at a rate of 2.5–3 frames per second using the 2PLSM system described above. Four to five ROIs in the cell body and one ROI in the nucleus were monitored and changes in TMRM fluorescence were plotted as a function of time. Flickering frequency is defined as the number of transitions in 100 sec epochs. The change in mitochondrial membrane potential (V) during flickering was estimated from the fluorescence in an ROI using a Nernst equation: V= (RT/zF)ln(βF_m/F_n) where R is the gas constant, T is temperature, F is Faraday’s constant, z=1, F_m is the fluorescence in the mitochondrial ROI, F_n is the fluorescence of the nucleus in the same optical plane and β is a scaling factor. The scaling factor was calculated by assuming that the mitochondrial membrane potential was –150 mV when F_maximal. For each flickering event, the percent change in V was calculated by taking the difference in the estimates before and then during the drop in fluorescence; the fluorescence during the drops was averaged.

2PLSM nitric oxide (NO) imaging

The NO probe 4–amino–5–methylamino– 2,7–difluorescein (DAF–FM) (Life Technologies) stock was dissolved in DMSO as instructed by the manufacturer. Stock solution was always prepared fresh on the day of the experiment. Following brain slice preparation, DAF–FM was perfused directly into the LC region via a patch pipette filled with 10 μM of the probe, at physiological temperatures (32–34°C) and waited 5–10 minutes prior to 2PLSM imaging. Optical imaging of DAF–FM signals acquired using a 920 nm excitation beam, in a fixed plane of focus with a pixel size between 0.18–0.21 μm and a 10 μs pixel dwell time. DAF–FM fluorescence (490–560 nm) was detected by a GaAsP PMT (H7422PA, Hamamatsu) and a Dodt contrast detector system that provided a bright–field transmission image (Prairie Technologies). For experiments using nitric oxide signaling inhibitor, brain slices were pre–incubated with inhibitors/antagonists for 2 hours, and then perfused with DAF–FM. For pharmacological experiments using isradipine, Ru360, and CGP37157, brain slices were hemi–dissected one sister slice containing LC corresponding to vehicle treatment, and the other sister slice corresponding to respective pharmacological manipulation, using same laser power and PMT settings. DAF–FM fluorescence was normalized by Dodt contrast signal to control for variations in laser power (as detected in the condenser path), and termed relative DAF–FM fluorescence in the box–plots.

Quantitative polymerase chain reaction (qPCR)

SNc and LC brain regions were micro–dissected from wild–type and total RNA was extracted using Trizol reagent (Life Technologies); cDNA was generated by reverse transcription (RT) (Quanta Biosciences) from both knockout and wild–type samples. cDNA from each sample was analyzed using quantitative PCR (qPCR) ³. qPCR experiments were performed using Sybr–Green and sense and antisense primers (Integrated DNA Technologies); primers specific to Cav1, Nav1, and NALCN channels were designed to generate PCR products between 120–200 base pairs. qPCR cycling parameters were: 1 cycle at 95°C for 3 minutes, followed by 40 cycles each consisting of 15 seconds at 94°C, 1 minute at 60°C, 30 seconds at 72°C, followed Sybr Green fluorescence reading. After the completion of 40 cycles, each sample was subjected to a melting temperature (T_m) curve analysis. Expression level was estimated using the comparative C_t approximation method to calculate relative abundance of mRNA. Samples were run in triplicate for accuracy and normalized to expression levels of seven housekeeping genes in order to obtain the delta C_t. Relative ion channel mRNA abundance was calculated by the equation 2^−ΔCt .

Immunostaining

Wild–type were perfused transcardially with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.3. Brains were removed, thin brain sections (50 μm–thick) were obtained using a Leica vibratome VT1200S. Sections were blocked in 5% normal goat serum followed by incubation with a monoclonal antibody against tyrosine hydroxylase (TH) (16–18 hrs at 4°C, working dilution of 1:1000, Immunostar) and a polyclonal antibody against dopamine–β–hydroxylase (DBH) (16–18 hrs at 4°C working dilution 1:1000, Immunostar). Staining was visualized with the respective secondary antibody conjugated to Alexa 594 or 405 (2 hrs incubation at room temperature, working dilution 1:1000) (Life Technologies). For biocytin experiments, LC neurons were filled with biocytin and stained with Streptavidin conjugated with Alexa 594. Images were acquired with a Fluoview confocal microscope (Olympus) and stored at 12–bit image depth at a resolution of 1024×1024 pixels.

Pharmacological reagents and channel ligands

Reagents were purchased from Sigma except for isradipine, L–NNA, DTT, ryanodine, orexin, muscimol, and CGP37157 were purchased from R&D Systems; TMRM, Alexa568, Alexa594, and Fluo4 obtained from Life Technologies; genipin was obtained from Wako Reagents, and Ru360 obtained from EMD Biosciences. L–NAME was purchased from Sigma–Aldrich. Drugs stock solutions were prepared in deionized water, DMSO, ethanol, or methanol as instructed by manufacturer on the day of experiment. Stocks were diluted to final concentrations in ACSF to achieve a final solvent concentration of <0.01% v/v.

Statistical analysis

Imaging data collected was analyzed with Igor Pro 6.0 (Wavemetrics), Matlab (Mathworks), or GraphPad Prism Version 5.0 (GraphPad Software). The stimulation, display, and analysis software for the two–photon imaging data was analyzed using a custom–written shareware package, WinFluor, PicViewer and PowerCAL kindly provided by John Dempster (Strathclyde University, Glasgow, Scotland; UK). Data were summarized using box–plots showing median values for small sample sizes and the interquatile range to describe the distribution of the data. Sample “n” represents the number of neurons collected from brain slices from at least three mice, one brain slice per mouse. Sample size was determined based on initial effect size, and as reported in previous publications¹¹. No randomization of samples was employed. Statistical analysis was performed with GraphPad Prism Version 5.0 (GraphPad Software) using non–parametric tests. Probability (p) threshold for statistical significance was p<0.05.