Optoactivation of Locus Ceruleus Neurons Evokes Bidirectional Changes in Thermal Nociception in Rats

Pontospinal noradrenergic neurons are thought to form part of a descending endogenous analgesic system that exerts inhibitory influences on spinal nociception. Using optogenetic targeting, we tested the hypothesis that excitation of the locus ceruleus (LC) is antinociceptive. We transduced rat LC neurons by direct injection of a lentiviral vector expressing channelrhodopsin2 under the control of the PRS promoter. Subsequent optoactivation of the LC evoked repeatable, robust, antinociceptive (+4.7°C ± 1.0, p < 0.0001) or pronociceptive (−4.4°C ± 0.7, p < 0.0001) changes in hindpaw thermal withdrawal thresholds. Post hoc anatomical characterization of the distribution of transduced somata referenced against the position of the optical fiber and subsequent further functional analysis showed that antinociceptive actions were evoked from a distinct, ventral subpopulation of LC neurons. Therefore, the LC is capable of exerting potent, discrete, bidirectional influences on thermal nociception that are produced by specific subpopulations of noradrenergic neurons. This reflects an underlying functional heterogeneity of the influence of the LC on the processing of nociceptive information.


Introduction
The degree of pain perceived in response to a given noxious stimulus is greatly influenced by the context within which the injury occurs (Melzack et al., 1982). In part, this variation in perception is due to differential engagement of endogenous analgesic mechanisms within the CNS that act to modulate pain (Ossipov et al., 2010). Since the original report that electrical stimulation of the midbrain produces profound analgesia (Reynolds, 1969), much attention has been focused on descending control systems, which can influence the spinal transmission of noxious inputs (Millan, 2002;Fields, 2004). The noradrenaline (NA)containing neurons of the locus ceruleus (LC) and the A5 and A7 cell groups in the pons have been implicated as key components of this descending control system (Jones, 1991;Pertovaara, 2006).
Although the LC sends projections throughout the neuraxis, most pain studies have concentrated upon the axonal projections to the spinal dorsal horn. These have been shown to release NA (Crawley et al., 1979;Hentall et al., 2003) that acts via ␣2adrenoceptors to inhibit both primary afferents and second order projection neurons (Reddy et al., 1980;Hammond and Yaksh, 1984;North and Yoshimura, 1984;Jones and Gebhart, 1986b;Miller and Proudfit, 1990;Kawasaki et al., 2003;Sonohata et al., 2004). Electrical or chemical stimulation of the LC has been shown to be analgesic in a number of acute pain models (Jones andGebhart, 1986a, 1986b;Miller and Proudfit, 1990;Jones, 1991;West et al., 1993). In addition, in chronic neuropathic pain models, there are plastic changes in the NAergic innervation of the spinal dorsal horn (Hughes et al., 2013) and increases in ␣2-adrenoceptor sensitivity (Ma and Eisenach, 2003;Hayashida et al., 2008).
Despite this knowledge of the analgesic potential of the pontospinal NAergic system; the use of systemic ␣2-adrenergic agonists to mimic its action in the treatment of pain is limited by a confounding side effect profile including sedation and hypotension (Pertovaara, 2006). The sedation can be reduced by longterm intrathecal infusion (Hassenbusch et al., 2002), but this is reserved for the most severe patients because it is expensive to deploy and can be associated with complications. Similarly, NA reuptake inhibitors, although moderately efficacious for chronic pain (Finnerup et al., 2010), are likewise limited by side effects that are largely a consequence of interactions with other central NAergic functions. Therefore, the potential of the descending NAergic system to suppress pain has not been successfully harnessed therapeutically.
We have demonstrated previously that pontospinal NAergic neurons can be selectively and effectively retrogradely transduced using an adenoviral vector with a catecholaminergicselective promoter (Howorth et al., 2009a) and that their inhibition, through expression of a potassium conductance, can produce thermal and inflammatory hyperalgesia (Howorth et al., 2009b). Here, we tested the hypothesis that direct activation of NAergic neurons of the LC using a selective lentiviral vector (LV)-mediated optogenetic approach produces analgesia. We show that activation of the LC can exert potent and temporally discrete antinociceptive and pronociceptive effects on the response to thermal stimulation. These contrasting actions are elicited from different subgroups of the neurons, indicating an underlying functional heterogeneity of the influence of the LC on the processing of nociceptive information.

Materials and Methods
Experimental procedures. Experiments were performed on male Wistar rats. All procedures conformed to the UK Animals (Scientific Procedures) Act of 1986 and were approved by our institutional ethical review committee. Animals were housed under a standard 12 h light/dark cycle with ad libitum access to food and water.
LV constructs and preparation. The study used two LV vectors: LV-sPRS-hChR2 (H134R)-EYFP and LV-sPRS-hChR2(H134R)-mCherry, differing only in the fluorophore fused to the human codon-optimized H134R variant of ChannelRhodopsin2 (ChR2; Fig. 1A; Boyden et al., 2005). Both LVs use the 240-bp PRSx8 promoter sequence (Hwang et al., 2001) that restricts the expression of the transgene to a subset of neurons that express the Phox2 transcription factor, which are predominantly catecholaminergic (Tiveron et al., 1996;Brunet and Pattyn, 2002). We used a transcriptional amplification strategy (Liu et al., 2006) to enhance expression of the ChR2 (sPRS). The methods for LV production were as described previously in detail (Coleman et al., 2003;Hewinson et al., 2013). In short, this is an HIV-1-based vector pseudotyped with vesicular stomatitis virus glycoprotein. The LVs were produced by cotransfection of Lenti-X293T cells (Clontech) with the shuttle vector pTYF-sPRS-hChR2(H134R)-EYFP or pTYF-sPRS-hChR2(H134R)-mCherry, along with a packaging vector, pNHP, and the envelope plasmid pHEF-VSVG. The virus was collected from the supernatant and concentrated by ultracentrifugation. The titer, measured in transducing units (TU)/ml, was assayed for each batch by establishing the infection rate of an LV produced in parallel and expressing placental alkaline phosphatase (Hewinson et al., 2013). The LV was used undiluted for microinjection at a titer of 2-3 ϫ 10 10 TU/ml depending upon the batch (4 batches were used in the course of the experiments).
Pontine LV injection. Stereotaxic injections of LVs were made into the dorsal pons in the region of the LC. The procedures for viral vector injections have been described in detail previously (Lonergan et al., 2005;Howorth et al., 2009a). In brief, rats were anesthetized (intramuscularly or intraperitoneally) with ketamine (5 mg/100 g, Vetalar; Pharmacia) and medetomidine (30 g/100 g, Domitor; Pfizer) until loss of paw withdrawal reflex. The animal was placed in a stereotaxic frame and core temperature was maintained at 37°C using a homeothermic blanket (Harvard Apparatus). Aseptic surgical techniques were used throughout.
To transduce LC neurons directly before pontine brain slice/working heart brainstem (WHBP) preparations, LV injections were made into postnatal day 17 (P17)-P19 Wistar rats using a microcapillary pipette (calibrated in 1 l intervals; Sigma) with a tip diameter of 20 m backfilled with vector (Fig. 1A). Unilateral or bilateral injections (each 250 m apart in the dorsoventral axis) of LV (250 nl/per injection of LV over 2 min) were made at 4 sites along a single track with the following coordinates: Ϫ1 mm, L 1 mm, 4.5-5.5 mm below the brain surface. All WHBP experiments and the initial pontine slice experiments (n ϭ 3) were done with the LV-sPRS-hChR2(H134R)-EYFP; the mCherry-expressing vector was used for the remainder of the slice experiments because this allowed the neurons to be visualized without activation of ChR2.
To transduce LC neurons before cutaneous thermal stimulation, injections of LV were made in Wistar rats (250 -275 g, n ϭ 37; Charles River Laboratories) at the following coordinates: Ϫ2.1 mm, L 1.2 mm, and 3ϫ injections of LV (300 nl) at 5.5-6.0 mm deep from brain surface with a 10°rostral angulation. After LV injection, a guide cannula was implanted over the injection site (C313G; Plastics One) and secured to the skull with dental cement/skull screw (Zhang et al., 2010). The guide cannula was closed with a dummy cap until the time of the experiment between 2 and 3 weeks after transduction. Sixty-three percent of the animals used in heat ramp experiments received LV-sPRS-hChR2(H134R)-mCherry and 37% received LV-sPRS-hChR2(H134R)-EYFP. There was no difference in the proportion of animals showing antinociception versus pronociception with the different vectors.
Pontine slices. Pontine slices were prepared from juvenile Wistar rats (P24 -P40, n ϭ 13, 1-3 weeks after LV transduction), which were terminally anesthetized with halothane before decapitation. The brainstem was removed and bathed in ice-cold cutting solution (similar to the recording solution but NaCl was reduced to 85 mM and substituted with sucrose 58.4 mM). Slices of the pons were cut (250 -350 m thick coronal or longitudinal) using a vibratome (Linearslicer Pro 7; DSK) in cold (4°C) cutting solution. Slices were kept at room temperature in carbogenated recording solution (composition is provided in the "Electrophysiology" section below) for at least 1 h before recording.
Electrophysiology. Pontine slices were transferred into the recording chamber of an upright fluorescence microscope (DMLFSA; Leica Microsystems), superfused with artificial CSF containing the following (in mM): 126 NaCl, 2.5 KCl, 26 NaHCO 3 , 1.25 NaH 2 PO 4 , 2 MgCl 2 , 2 CaCl 2 , and 10 D-glucose saturated with 95% O 2 /5% CO 2 , pH 7.3, osmolality 290 mOsm/L, 35°C at a rate of 2-3 ml min Ϫ1 . Patch pipettes (resistances of 4 -6 M⍀) were filled with internal solution containing the following (in mM): 130 K-gluconate, 10 KCl, 10 Na-HEPES, 4 MgATP, 0.2 EGTA, and 0.3 Na 2 GTP. All membrane potentials were corrected for a junction potential of 13 mV. Fluorescent transduced LC neurons (EYFP/mCherry) were identified and whole-cell recordings were then obtained under gradient contrast illumination (Dodt and Zieglgänsberger, 1990). Once confirmed as fluorescent, the cells were not exposed to epilumination via the objective until after the whole-cell patch-clamp recordings were completed. Recordings were made in current-clamp or voltage-clamp mode and current pulses were injected to examine the current-voltage and current-spike frequency relationships (Axopatch 1D amplifier; Molecular Devices). The signal was filtered at 3 kHz, digitized at 10 kHz, and stored using Spike 2 software (Cambridge Electronic Design). Light was pulsed onto the cells (0.1-5 mW for 20 ms to 10 s) using a focally placed optical fiber proximate to the recorded cell (400 m in diameter) pigtailed to a 473 nm LED source (Doric).
WHBP. The WHBP was used to examine the responses of LC neurons to optoactivation in an in vivo-like preparation. Preparations were made from LV-transduced Wistar rats aged 24 -30 d (n ϭ 6) according to previously published methods (Paton, 1996;Pickering et al., 2003;McAllen et al., 2011). In brief rats, were deeply anesthetized with halothane, bisected subdiaphragmatically, exsanguinated, cooled by immersion in 5-10°C Ringer's solution containing the following (in mM): 125 NaCl, 24 NaHCO 3 , 3 KCl, 2.5 CaCl 2 , 1.25 MgSO 4 , 1.25 KH 2 PO 4 , and 10 dextrose, pH 7.35-7.4 after carbogenation, before a precollicular decerebration and cerebellectomy to expose the dorsal surface of the pons. While still cold, the phrenic nerve and descending aorta were dissected free and a bilateral pneumonectomy was performed. The preparation was then transferred to a recording chamber and retrogradely perfused with Ficoll-70 (1.25%) containing carbogenated aCSF (32°C) through a double lumen cannula inserted into the descending thoracic aorta. As the preparation warmed and brainstem perfusion was reestablished, respiratory movements returned and the preparation was paralyzed with vecuronium (2 g/ml). Preparation viability was assessed by the presence of ramping phrenic nerve discharge recorded with a suction electrode and robust responses to cardiorespiratory reflex activation.
Extracellular recordings were made (using a custom-built AC amplifier) from neurons in the LC using glass microelectrodes filled with 1 M NaCl (resistance 8 -15 M⍀). The signal was band-pass filtered (200 Hz to 3 kHz), digitized at 10 kHz (micro1401; Cambridge Electronic Design), and stored on a PC for analysis with Spike2 software. Neurons were optoactivated by light pulses (445 nm diode laser, Omicron Phoxx, 0.1-15 mW, 10 ms to 10 s) from an optical fiber (200 m core, multimode; Thorlabs) placed over the dorsal surface of the pons or inserted up to 300 m into the tissue. At the end of the experiment, pontine tissue was fixed in 4% paraformaldehyde for 48 h before sucrose cryoprotection A C D E B Figure 1. Selective and functional expression of hChR2 in rat locus ceruleus neurons. A, Schematic of experimental approach. LV-sPRS-hChR2(H134R)-mCherry or LV-sPRS-hChR2(H134R)-EYFP is injected directly into the LC. B, Two weeks after unilateral LC injection of LV-sPRS-hChR2(H134R)-EYFP, there is robust, selective expression in NAergic neuronal somata and processes (arrowed cells on edge of LC shown magnified beneath). EYFP expression (green), DBH immunoreactivity (red), and overlaid images (right). 4 th V, Fourth ventricle. C, Voltage-clamp recordings from transduced LC neurons in brainstem slices showed inward currents to blue light, showing a characteristic fast desensitization typical of hChR2 (H134R; 473 nm light, 1-5 mW, Vhold Ϫ60 mV). D, Pulses of blue light (4.5 mW, 50 ms) reliably evoked action potential trains in the same neuron as C at a range of frequencies (up to 10 Hz). E, In some transduced neurons, pulsed illumination (50 ms, 5 mW) only produced subthreshold depolarizations, whereas constant illumination of the same intensity (5 mW) brought the neuron to threshold.
Cutaneous stimulation with heat ramps. LC-transduced (n ϭ 37) and naive Wistar rats (n ϭ 12) were anesthetized with isofluorane (1.5-4%) until loss of paw withdrawal reflex. The external jugular vein was cannulated and the animal was switched to an intravenous anesthetic (Alfaxalone, 10 mg/ml, 7.5-15 mg/h; Vetoquinol). The carotid artery and trachea were also cannulated and, for some experiments, an intrathecal catheter (32 gauge tubing [Recathco] coupled to PE 10 tubing) was inserted at the L4 -5 interspace through a 25 G needle. The animal was placed in a stereotaxic frame and EMG wires (Teflon-coated stainless steel, 75 m diameter; Advent Research Materials) were inserted into the belly of biceps femoris. The signal was AC amplified (Neurolog NL-104; Digitimer), band-pass filtered (60 Hz to 5 kHz Neurolog NL125) and digitized (at 20 kHz) via a micro1401 into Spike2 software (Cambridge Electronic Design).
Once the surgical procedures were completed, the anesthetic was titrated to a stable, lighter plane of anesthesia until a withdrawal reflex could be evoked by moderate pinch of the forepaw. Heat ramp stimuli were applied to the dorsal surface of the left hindpaw using a custombuilt thermal heat stimulator consisting of a heating lamp focused on a copper disc placed in contact with the foot with a surface mounted thermode (McMullan et al., 2004). The rate of heating of the hindpaw was A C B Figure 2. Optoactivation of locus ceruleus neurons in the working heart brainstem preparation. A, Rats (P21) received a unilateral injection of LV into the LC. After 1-3 weeks, extracellular recordings were made from LC neurons in a decerebrate (and cerebellectomized), artificially perfused preparation that allows direct access to the floor of the fourth ventricle to study the cells in a functionally connected network (Paton, 1996). PNA, Phrenic nerve activity; PP, perfusion pressure. B, Action potential trains recorded from a spontaneously active LC neuron expressing LV-sPRS-hChR2(H134R)-EYFP. Pulsed illumination ( 445 nm, 20 mW, 20 ms pulse, 3-12.5 Hz) over transduced neurons produced temporally precise activation, followed by a refractory period. At higher stimulus frequencies (Ͼ8 Hz), the neuron started to fail to follow each individual stimulus, resulting in a fall in frequency seen in the instantaneous firing frequency plots. C, The same neuron responded to continuous illumination with a prolonged increase in firing frequency (response to 9 mW shown in the top two panels). This was seen in response to lower light intensities and could produce a higher level of maintained neuronal discharge than pulsed stimulus protocols (instantaneous firing frequency plots shown overlaid for stimuli from 0.03 to 9 mW). It was also apparent that the LC neurons show accommodation to the stimulus. adjusted to be in the range of 6.5-7.5°C s Ϫ1 . The temperature at the point of hindlimb withdrawal (based on onset of EMG response) was measured as the output of the assay. Ramps were applied at 8 min intervals to minimize the risk of sensitization. Three baseline threshold measures (within 1.5°C of each other) were obtained before light stimulation was tested to ensure stable recording conditions. Light stimuli ( 445 nm, 1-30 mW, pulsed or continuous) were applied via a bare-ended fiber optic (200 m core, multimode; Thorlabs) lowered through the implanted guide cannula to a predetermined depth just above the left LC (4.5 mm deep to the brain surface). The threshold of the withdrawal response to heat ramp was rechecked after changing the fiber position to ensure that it remained stable before trials of optoactivation. The optical stimuli were initiated 60 s before application of the heat ramp and were terminated immediately after the ramp. If no response was seen (i.e., change in threshold of Ͻ1.5°C) with successive heat ramps, then the fiber was advanced in 500 m steps to a maximum depth of 6.5 mm. If no change in threshold was seen on reaching this depth, then the experiment was ended (n ϭ 11). In three of these experiments, the post hoc histology showed that the implanted guide cannula was off target and these were excluded from further analysis. In the remaining eight, there was no evidence of viral transduction (no fluorophore expression), but the fiber was correctly targeted to the LC. These preparations were then used as controls for thermal withdrawal thresholds and also blood pressure changes.
Extracellular recording. Extracellular recordings of LC neurons (ipsilateral to LV injection) were obtained during preliminary heat ramp testing. A parylene-c insulated tungsten microelectrode (5 M⍀; A-M systems) referenced against a sintered silver chloride pellet placed under the scalp was lowered into the left LC (5.2-6.5 mm deep to the brain surface). The signal was amplified (Axon Multiclamp 700A; Molecular Devices), filtered (100 Hz to 2-3 kHz), and data were collected as described for the WHBP. Neurons were considered to be LC if they fit several criteria: (1) a duration of action potential Ն1 ms, (2) spontaneous firing with a slow firing rate of 0.5-7 Hz, (3) a large amplitude characteristic action potential waveform, or (4) biphasic response to contralateral hindpaw pinch with activation followed by a transient silent period (as described previously;Cedarbaum and Aghajanian, 1978). Extracellular recordings were used to assay the response of LC cells to noxious thermal stimuli and then optoactivation. The baseline firing rate (in Hertz) of LC neurons was measured over a 5 s window (always taken 20 s before start of either heat or light stimuli). In those preparations (n ϭ 5) examining LC responses to thermal stimuli, LC firing rate (in Hertz) was recorded over a 5 s time period from the onset of the burst of activity. In preparations measuring cell responses to optoactivation, the recording electrode was coupled to an optical fiber (150 m core; Thorlabs; Abbott et al., 2009) with its recording tip 500 m ahead of the fiber tip. Note that this assembly could not be easily maneuvered down the implanted guide cannula (usually only one recording track was possible), so these recordings were not performed routinely in the heat ramp experiments. Cells were illuminated (445 nm diode laser, Omicron Phoxx, 1-30 mW, continuous 5-30 s) and the change in LC firing rate was compared with the baseline firing rate taken before onset of illumination.
Tissue fixation. Rats were killed with an overdose of pentobarbital (Euthatal, 20 mg/100 g, i.p.; Merial Animal Health) and perfused transcardially with 0.9% NaCl (1 ml/g) followed by 4% formaldehyde (Sigma) in 0.1 M phosphate buffer (PB; pH 7.4, 1 ml/g). The brain was removed and postfixed overnight before cryoprotection in 30% sucrose. Coronal tissue sections were cut at 40 m intervals using a freezing microtome and left free floating for fluorescence immunohistochemistry.
Immunohistochemistry. Tissue sections were washed 3 times in 0.1 M PB and permeabilized in 50% ethanol for 30 min before further washing. The tissue was incubated on a shaking platform with primary antibodies against dopamine ␤-hydroxylase (mouse anti-DBH, MAB308, 1:10,000 or 100 ng/ml; Millipore), or EYFP (chicken anti-EGFP, AB13970, 1:5000 or 2 g/ml; Abcam) in PB with 5% horse serum (HS) and 0.3% Triton X-100 for 12-24 h at room temperature. After further washing, sections were incubated for 4 h with appropriate secondary antibodies conjugated to fluorophores (Alexa Fluor 488 and Alexa Fluor 594 1:1000 or AMCA A B C D Figure 3. Response of LC neurons to noxious hindpaw stimuli. A, Extracellular recordings were made from LC neurons in alfaxalone-anesthetized animals to verify their location with reference to the cerebellar surface. Heat ramp stimuli were applied to the dorsal surface of the left hindpaw using a contact heat device (McMullan et al., 2004). EMG activity (green trace) from the biceps femoris was used to indicate the onset of withdrawal. Pinch stimuli were also applied to the ipsilateral and contralateral hindpaws. B, LC neurons (found at depths 5-6.5 mm) showed the characteristic spike waveform, spontaneous activity, and biphasic response (activation-inhibition) to contralateral hindpaw pinch. C, A qualitatively similar but less intense biphasic response was evoked by ipsilateral hindpaw pinch. D, Application of the heat ramp stimulus to the ipsilateral hindpaw also activated the same LC cell. The activation occurred only as the skin temperature approached the noxious range, with the increase in firing commencing just before the withdrawal response.
1:200) in PB with 2% HS and 0.3% Triton X-100. Sections were washed three times before mounting. We validated the specificity of the DBH antibody previously (Howorth et al., 2009a). The anti-EGFP antibody has been validated by Western blot and immunocytochemistry to be specific for both EGFP and EYFP (Abcam manufacturer's datasheet).
Here, it showed an overlapping distribution with EYFP-positive cell bodies and processes and amplified the signal in distal processes; no staining was seen in control brainstem tissues without viral vector injection. Negative controls were routinely run by omitting primary antibodies.
Photomicrography. Representative images were taken on a Zeiss Axioskop 2 fluorescence microscope with a Zeiss Axiocam HRC using the appropriate excitation-emission filter sets (EYFP/Alexa Fluor 488 #10, mCherry/Alexa Fluor 594 #1 and #5, and AMCA #02). The images were initially processed using Zeiss software (Axiocam 4.7 software),Zeiss and prepared for presentation using Adobe Photoshop CS5 with optimization of contrast/brightness as required and the addition of annotation.
Analysis. All data are presented as mean Ϯ SEM or median (interquartile range) as appropriate. The normality of data was assessed using the D'Agostino-Pearson test. Subsequent statistical testing was undertaken using paired and unpaired t tests, one-and two-way ANOVA (with Bonferroni's posttests), and Mann-Whitney/Kruskal-Wallis (with Dunn's posttest) tests as appropriate. Data were analyzed using Prism Version 5 software (GraphPad) and differences were considered significant at p Ͻ 0.05.

Selective expression of ChR2 in noradrenergic LC neurons
Direct stereotaxic injection of LV-sPRS-hChR2(H134R)-EYFP or LV-sPRS-hChR2 (H134R)-mCherry to the LC produced selective expression of the fluorophore-ChR2 construct within LC neurons ( Fig.  1 A, B). These transduced neurons showed characteristic membrane-delimited fluorescence with strong signal from the dendrites and axonal processes, as expected of the ion channel-fluorophore construct. The expression was limited to neurons that were identified as NAergic on the basis of DBH immunocytochemistry (Fig.  1B). The strength of expression increased, as judged from the fluorescence intensity and the extent of visible processes over a period of 3 weeks.

Functional expression of ChR2 in LC neurons in vitro
To confirm the functional expression of hChR2 in the LC, we transduced neurons by stereotaxic injection of LV 1-3 weeks before preparing pontine slices. Successful transduction of the neurons was confirmed by the presence of visible fluorescent somata within the boundaries of the LC in the slice. Current-clamp recordings of the transduced fluorescent neurons (n ϭ 15) showed typical LC electrophysiology, indicative of good cellular health, including spontaneous action potential discharge, large action potentials, and a pronounced slow AHP. Optoactivation evoked inward currents in voltage clamp (Fig. 1C) that reversed close to zero and showed a rapid phase of desensitization, characteristic of ChR2. Current-clamp recordings showed that light pulses evoked time-locked depolarizations in all cells and action potential discharge (Fig. 1D). This pattern of optoexcitation was not seen in neurons that were not fluorescent or in transduced neurons in response to incident white light illumination. The charging of the membrane in these LC neurons typically took 20 -50 ms before the threshold for spike discharge was reached. The LC neurons showed accommodation during a train of pulses with failures of spike discharge to limit the frequency of repetitive firing (typically to Ͻ10 Hz). Both pulsed and continuous illumination were effective stimuli for some cells (n ϭ 4/6 tested with both stimuli). However, in several of these LC neurons (n ϭ 2/6), only the application of constant light (or prolonged pulses of duration Ͼ100 ms) could bring the neuron to threshold, whereas pulsed illumination (with durations of Յ50 ms) only produced subthreshold depolarizations (Fig. 1E). Furthermore, continuous illumination consistently increased spike discharge at lower light intensities than pulsed illumination. In all neurons, the optoexcitation was followed by a period of neuronal quiescence; the magnitude of this silent period was related to the Continuous illumination (30 mW) over the transduced LC produced repeatable, titratable increases in neuronal discharge, followed by a period of neuronal silence that is characteristic of LC neurons. Bii, Pooled data showing optoactivation of LC neurons produced robust increases in cell activity (n ϭ 5; *p Ͻ 0.05, paired t test). duration and intensity of prior stimulation, as is characteristic of LC neurons (Aghajanian et al., 1977).

Functional expression of ChR2 in LC neurons in the WHBP
The WHBP was used to examine the responses of LC neurons to optoactivation in an in vivo-like preparation to test whether their activity could be driven in a more intact setting with competing ongoing synaptic drives and compensatory mechanisms (Fig. 2). Extracellular recordings in the WHBP showed that pulsed illumination over the transduced LC could produce temporally locked activation of neurons (n ϭ 6; Fig. 2B). These extracellularly recorded LC neurons faithfully generated action potentials in response to pulsed optoactivation (at 1-2 Hz with pulses of 20 -30 ms duration, 20 mW). However, for each cell, failures of optoactivation were observed at higher stimulus frequencies of between 2 and 10 Hz that limited the maximum discharge frequency. Constant illumination of the same neurons produced robust increases in firing frequency (Fig. 2C) at relatively lower light intensities. Furthermore, the peak firing frequency that could be attained was higher than that seen with pulsed stimulation (reaching a peak of 29 Ϯ 14 Hz, n ϭ 5), although, in every case, the neurons exhibited a degree of accommodation to these stimuli. As in slices, the periods of optoexcitation were followed by a refractory period (7.9 Ϯ 2.1 s) when spontaneous action potential discharge was inhibited.

LC optoactivation strongly modulates thermal nociception in vivo
The withdrawal response to a thermal heat ramp stimulus applied to the hindpaw of anesthetized adult rats was used to assay the effect of LC optoactivation on nociception (Fig. 3). Extracellular recordings of LC neurons were made in vivo to confirm the coordinates for subsequent optoactivation experiments (n ϭ 3 rats) and to characterize their responses to heat ramp stimuli (n ϭ 5 rats). LC neurons were found from 5.0 to 6.5 mm ventral to the cerebellar surface and were recognizable by their characteristic spike shape, spontaneous activity, and biphasic response to paw pinch, which was characteristically stronger in response to stimuli applied to the contralateral paw (Cedarbaum and Aghajanian, 1978;Sugiyama et al., 2012). Application of heat ramp stimuli to the ipsilateral hindpaw triggered a phasic burst of firing as the thermal stimulus approached the noxious range and triggered a withdrawal response in all LC neurons tested (from a baseline mean rate of 2.5 Ϯ 0.7 to 7.2 Ϯ 1.9 Hz, p ϭ 0.048, paired t test, n ϭ 6 cells; Fig. 3). This increase in LC firing was immediately followed by a characteristic refractory period before returning to baseline firing levels after the stimulus.
For the in vivo optoactivation experiments, animals were transduced with hChR2 by injection of vector 3-4 weeks earlier and had a guide cannula implanted above the LC to direct the optical fiber. Lowering a combined optical fiber-tungsten electrode down the cannula (Fig. 4A) allowed the effect of optoactivation to be seen in vivo. LC cells were driven strongly by continuous illumination (30 mW, firing rate of LC neurons increased from 2.4 Ϯ 0.7 to 20.4 Ϯ 6.1 Hz, p ϭ 0.043, paired t test, n ϭ 5 cells; Fig. 4B), which was consistent with our observations in the WHBP. In each case, this was followed by a period of relative quiescence once the illumination ceased, but the responses were fully recapitulated on repeat optoactivation. Therefore, based on our in vitro and in situ experiences with pulsed optoactivation of LV-transduced LC neurons (showing limited frequency response), we elected to standardize on a continuous illumination protocol for the subsequent in vivo experiments because this allowed us to more closely mimic the stimulation protocol (50 -100 Hz applied for periods of up to 1 min) before noxious stimulation used in previous studies of LC antinociceptive effects (Jones andGebhart, 1986a, 1986b;Miller and Proudfit, 1990;Jones, 1991;West et al., 1993). A C B Figure 5. Optoactivation of LC produces a graded fall in arterial blood pressure. A, Temporally discrete depressor effect on illumination of the LC (6 -30 mW, blue box). Arrow indicates blood pressure response to application of the noxious thermal heat ramp stimulus to the hindpaw. B, Mean arterial blood pressure dropped by Ͼ20 mmHg during LC optoactivation in LV-transduced animals compared with controls (no viral transduction). ***p Ͻ 0.001 rmANOVA and Bonferroni's posttest). Ci, With the fiber 4.5 mm deep to the brain surface, illumination (30 mW) does not cause any change in blood pressure and the thermal withdrawal threshold is unchanged at 53°C. Cii, Advancing the fiber by 500 m, we now evoke a depressor response, but there is still no change in the withdrawal threshold. Ciii, Advancingthefibertoitsfinalposition(5.5mmdeep),weseealargerdepressorresponsenowaccompaniedbyareductioninthethreshold for withdrawal. Note that with each heat ramp stimulus, there is a corresponding transient pressor effect.
Optoactivation of the LC produced graded (according to stimulus intensity and duration), reproducible falls in mean arterial pressure (30 mW, 25 Ϯ 5 mmHg, p Ͻ 0.0001, repeatedmeasures ANOVA [rmANOVA] with Bonferroni's posttest, n ϭ 11; Fig. 5 A, B), which was not seen in control animals ( p Ͼ 0.05, n ϭ 7). The depressor response was also related to optical fiber depth such that its magnitude increased as the LC was approached (Fig.  5C). A similar depressor effect has been seen on microinjection of glutamatergic or cholinergic agonists (Jones and Gebhart, 1986b;Sved and Felsten, 1987) and is mediated by sympathoinhibition. These studies contrasted with electrical stimulation of the LC that was found to produce a pressor effect thought to be due to an offtarget action on fibers of passage. This depressor effect of LC optoactivation therefore supports the efficacy/selectivity of our stimulation methodology.
LC activation exerted a strong antinociceptive effect (seen in 55% of rats, n ϭ 11/20) with repeatable, substantial, increases in thermal withdrawal thresholds (ϩ4.7 Ϯ 1.0°C, p Ͻ 0.0001, rmANOVA with Bonferroni's posttest, n ϭ 11/20; Fig.  6A-C). These changes were seen to reverse back to baseline within 10 min. No change in threshold was seen in control rats in which the same laser stimulus was applied to the LC (n ϭ 8; Fig. 6C). After repeated optoactivation of the LC (typically Ͼ3ϫ), a prolonged increase in the withdrawal threshold was sometimes observed that did not recover over the time course of the experiment (Ͼ2 h).
In contrast, 45% of the animals showed pronociceptive effects on optoactivation of the LC with reversible decreases in thermal threshold (Ϫ4.4 Ϯ 0.7°C, p Ͻ 0.0001, rmANOVA with Bonferroni's posttest, n ϭ 9/20; Fig. 7). In individual experiments, only an increase or a decrease in threshold was seen after LC optoactivation at the A B C D E Figure 6. LC optoactivation is antinociceptive. A, Withdrawal responses to the thermal stimulus with a burst of EMG activity recorded from biceps femoris corresponding to a withdrawal threshold of ϳ46°C under baseline conditions (ϫ2). Optoactivation of the LC (30 mW for 60 s before heat ramp, blue box) increased the withdrawal threshold temperature by 5°C, which fully reversed back to control 8 min later. B, Plot of change in withdrawal threshold with repeated optoactivation (blue bars) from another single experiment with a return back to baseline within 10 min. C, Pooled data showing LC optoactivation to robustly increase withdrawal thresholds by Ͼ4°C (degrees change from baseline), which is consistent with an antinociceptive action (n ϭ 11) compared with controls (no viral transduction, n ϭ 8). D, Intrathecal clonidine (750 ng), an ␣2-agonist, significantly increased withdrawal thresholds ( p Ͻ 0.001, paired t test, n ϭ 7), which remained elevated for Ͼ30 min. However, this increase in withdrawal threshold was reversed back to baseline in the group of animals given atipamezole (50 g, i.t, n ϭ 3, p Ͻ 0.01, unpaired t test). E, In contrast, intrathecal atipamezole (100 g) did not prevent the antinociceptive effects of LC optoactivation (n ϭ 4). rmANOVA with Bonferroni's posttest, ***p Ͻ 0.001, **p Ͻ 0.01, *p Ͻ 0.05. initial fiber site. We reviewed the experimental parameters for the antinociceptive versus pronociceptive actions (n ϭ 11 and 9, respectively) and there was no difference in anesthetic infusion rate (alfaxalone 11.7 Ϯ 0.8 vs 10.0 Ϯ 1.0 mg/h, p ϭ 0.17, unpaired t test), which was titrated against the paw withdrawal response; in hindpaw skin surface temperature (27.7 Ϯ 0.5 vs 27.8 Ϯ 0.6°C respectively, p ϭ 0.9, unpaired t test); or in baseline thermal withdrawal thresholds (49.2 Ϯ 1.1 vs 51.5 Ϯ 0.3°C, respectively, p ϭ 0.13, unpaired t test). Similarly there was no difference in the magnitude of the depressor responses to LC activation in preparations showing antinociceptive (27 Ϯ 4 mmHg, n ϭ 7) or pronociceptive actions (22 Ϯ 9 mmHg, n ϭ 4, p ϭ 0.78, unpaired t test). The depressor effect was not associated with significant changes in hindpaw skin temperature (27.9 Ϯ 0.4 vs 28.2 Ϯ 0.4°C, p ϭ 0.12, paired t test, n ϭ 9). Comparable studies using heat ramps have shown that similar changes in arterial pressure associated with PAG stimulation were not responsible for the antinociceptive effects (McMullan and Lumb, 2006). Post hoc histological analysis suggested that different subpopulations of LC neurons were labeled in those animals showing antinociceptive effects (with a ventral predominance extending down to the subceruleus) compared with those showing pronociceptive effects (largely restricted to a more dorsal location in the body of the LC proper; Fig. 8A). Similarly analysis of the distribution of the successful fiber stimulation sites also demonstrated that the pronociceptive actions tended to be evoked from a more dorsal position (5.25 Ϯ 0.2 vs 5.6 Ϯ 0.1 mm, respectively, p ϭ 0.04, unpaired t test; Fig. 8B). Having noted the relationship of LC-evoked response to the placement of the optical fiber, we changed our experimental protocol and intentionally advanced the fiber (by 500 m) beyond its initial stimulating position (n ϭ 3) and observed a switch from pronociception (Ϫ3.3 Ϯ 0.1°C, p Ͻ 0.05, rmANOVA with Bonferroni's posttest, n ϭ 3) to antinociception (ϩ2.9 Ϯ 0.3, p Ͻ 0.01) as the fiber was moved to a deeper position (Fig. 8C). In one experiment, this was found to be reversible back to antinociception on withdrawal of the fiber back to its original position.

Discussion
The LC is proposed to play an important role in an endogenous descending pain control circuit (Jones, 1991;Millan, 2002;Pertovaara, 2006). Given that this NAergic circuit component appears to be functionally underactive in chronic neuropathic pain (Howorth et al., 2009b;Alba-Delgado et al., 2012;Hughes et al., 2013), we have been interested in exploring the nature of this deficit and developing strategies to augment its activity. In this study, we have tested whether the expression of hChR2 in NAergic neurons (Carter et al., 2010;Zhang et al., 2010) will permit selective, specific, temporally discrete activation of the LC and have assessed its role in the control of thermal nociception.
We assayed the effect of LC activation on the nocifensive withdrawal response to a hindpaw thermal heat stimulus (McMullan et al., 2004). Consistent with a role in endogenous analgesia, we show that activation of the LC can produce a substantial, tempo-rally discrete antinociceptive effect. We also describe a novel acute pronociceptive effect of LC activation using similar stimulation parameters and propose that distinct subpopulations of neurons in the LC drive these opposing actions. These findings demonstrate bidirectional modulation of nociception by the LC-a characteristic that has long been known for other brainstem descending control areas such as the rostral ventromedial medulla (RVM;Heinricher et al., 2009;Ossipov et al., 2010) and the periaqueductal gray (PAG; Basbaum and Fields, 1978;Lovick and Bandler, 2005). Bidirectional modulation of nociception has also been reported previously for the A7 NAergic cell group after microinjection of opioids (Holden et al., 1999) or GABA A antagonists (Nuseir and Proudfit, 2000).
The LC extends widespread projections and is the main source of NA for much of the CNS (Dahlstrom and Fuxe, 1964;Foote and Morrison, 1987;Aston-Jones, 2004). It is a relatively small structure with ϳ3000 neurons in the rat (Loughlin et al., 1986a) and, although these cells are considered homogenous with respect to their NAergic neurochemistry, they can be divided into several different subgroups based on cell morphology and efferent targets (Loughlin et al., 1986b;Berridge and Waterhouse, 2003;Aston-Jones, 2004). With respect to pain control, it is the descending projections to the spinal dorsal horn that have received the most attention (Jones, 1991;Millan, 2002;Pertovaara, 2006) and these are predominantly located in the ventral pole and in the subceruleus (Westlund et al., 1983;Howorth et al., 2009a;Bruinstroop et al., 2012). Previous studies have shown that focal stimulation (electrical and/or chemical) of the LC region can produce antinociceptive effects (Margalit and Segal, 1979;Jones and Gebhart, 1986b;Jones, 1991;West et al., 1993). We have now shown directly that LC neuronal activation is antinociceptive through the use of an optogenetic approach that specifically targets the NAergic neurons and removes the potential confound of stimulation of axons of passage or indirect activation of neighboring neurons via chemical spread. Therefore, our findings reinforce the notion that the LC is part of the endogenous antinociceptive control system.
The LC-stimulation induced antinociception was mimicked by intrathecal administration of the ␣2-agonist clonidine, but was not blocked by intrathecal administration of the ␣2-A B Figure 7. Optoactivation of LC neurons can be pronociceptive. Optoactivation of the LC (30 mW, 60 s blue box) decreased thermal withdrawal thresholds in some preparations, suggesting a pronociceptive action (n ϭ 9). A, Plot of change in withdrawal threshold from a single experiment. The drop in withdrawal threshold was repeatable and typically reversed back toward baseline within 10 min. B, Group data (n ϭ 9) showing that withdrawal thresholds were significantly decreased by LC optoactivation (degrees change from baseline). rmANOVA with Bonferroni's posttest, ***p Ͻ 0.001, **p Ͻ 0.01, *p Ͻ 0.05.
antagonist atipamezole-indeed, this dose appeared to enhance the antinociceptive action. This is counter to previous reports showing that intrathecal ␣2-antagonists block LC-stimulusevoked analgesia (Jones and Gebhart, 1986b;Miller and Proudfit, 1990;West et al., 1993). We used a substantial dose of atipam-ezole (100 g) that was double that needed to completely reverse the effects of intrathecal clonidine. One previous LC stimulation study reported analgesic actions that were resistant to an intrathecal ␣2-antagonist, idazoxan (West et al., 1993). It may be relevant to note that atipamezole and idazoxan have a much higher ␣2:␣1 selectivity ratio than yohimbine (Pertovaara et al., 2005), which was used by many of the early studies (Jones and Gebhart, 1986b;Miller and Proudfit, 1990;West et al., 1993). It remains to be determined whether the antinociceptive actions of LC activation may be mediated by another mechanism, such as enhancement of inhibitory transmission by ␣1receptors at a spinal level (Baba et al., 2000;Gassner et al., 2009;Funai et al., 2014). It may also be relevant that the LC neurons are capable of releasing other transmitter substances such as galanin, neuropeptide Y, and ATP in addition to noradrenaline (Holets et al., 1988); spinal galanin in particular has been associated with antinociceptive actions (Hua et al., 2004). Alternatively, the LC may have a supraspinal action via a relay in the RVM (Aston- Jones, 2004) to inhibit the spinal nociception through opioid release (Margalit and Segal, 1979;Miller and Proudfit, 1990) or perhaps on ascending nociceptive pathways, for example, at a thalamic level (Voisin et al., 2005). Counter to our expectation, we also demonstrated a potent LC-mediated pronociceptive action that reduced the thermal withdrawal threshold. Although the LC is predominantly associated with pain inhibition, there is some evidence to suggest that it may also play a role in the facilitation of chronic pain. Brightwell and Taylor (2009) showed that the LC facilitates the maintenance of neuropathic pain. Similarly Martins et al. (2010) reported a facilitatory effect of the LC on neuropathic pain mediated via the medullary dorsal reticular nucleus. In addition, the second phase behavioral responses to formalin injection were found to be reduced after neurotoxic LC lesions (Martin et al., 1999). In contrast, we reported previously that selective inhibition of the pontospinal LC enhances the response to formalin (Howorth et al., 2009b), which raises the possibility of anatomically distinct LC effects on nociceptive processing.
By mapping of the distribution of the transduced neurons, we demonstrated that a larger proportion of the fluorescent neurons were located in the ventral LC and subceruleus in those experiments showing antinociceptive responses. Similarly, we found that the optic fiber placement was deeper in experiments showing antinociceptive responses. We  (Paxinos and Watson, 2005). Each color represents a different animal, each dot corresponds to a neuron in a 40 m section, and dark core dots are outside the LC. Animals showing antinociception had LV-transduced LC neurons with a ventral predominance clearly extending to the subceruleus, whereas transduction was mainly restricted to midcaudal LC proper in those showing pronociceptive effects. B, Graph showing the fiber stimulation site (depth from cerebellar surface) for the pronociceptive actions, which were more dorsal than those sites with antinociceptive actions (left and right, respectively, *p Ͻ 0.05, unpaired t test). C, By changing our recording protocol to allow the fiber position to be advanced beyond the first response site (by 500 m), we observed a switch in the response from pronociception to antinociception (n ϭ 3). rmANOVA followed by Bonferroni's posttest, **p Ͻ 0.01, *p Ͻ 0.05. subsequently showed both pronociceptive and antinociceptive actions in the same experiment by varying fiber depth. This suggests that the pronociceptive and antinociceptive actions are mediated by distinct subpopulations of LC neurons, with the antinociceptive effect originating from neurons in the ventral region. The balance of pronociceptive versus antinociceptive actions within any given experiment was thus a function of the distribution of the transduced neurons within the LC and of the fiber position relative to the LC. These data support the principle that the antinociception may be via a spinally mediated mechanism, because the ventral LC and subceruleus are known to project to the dorsal horn of the spinal cord (Westlund et al., 1983;Loughlin et al., 1986a;Howorth et al., 2009a;Bruinstroop et al., 2012), although it is also apparent that some of the spinally projecting LC neurons also innervate supratentorial structures (Guyenet, 1980;Leanza et al., 1989;Howorth et al., 2009a), forming a neural substrate for a multilevel influence upon pain processing. The detailed dissection of these LC modules will require more specific transduction methods that allow selective targeting of these subpopulations (Howorth et al., 2009a;Osakada et al., 2011) or activation of their terminals by focal spinal illumination.
We show here that the LC neurons are activated by noxious heat stimulation, which is consistent with a role in the modulation of nociception. Although the LC has long been known to respond to noxious mechanical stimulation (Cedarbaum and Aghajanian, 1976) and to prolonged thermal stimuli Hajó s et al., 1986), we have extended this finding using the heat ramp withdrawal response assay. With this graded stimulus, we show that LC activation occurs just as the temperature reaches the noxious range (i.e., as it becomes salient) immediately before the withdrawal response; therefore, conceptually, the LC neurons could act to attenuate the onward transmission of the nociceptive information. However, this could be part of an alerting or arousal response to the stimulus as part of a facilitatory action, as has been reported for "on cells" in the RVM (Fields et al., 1991).
The LC is implicated in the regulation of attention, sleepwake cycles, mood, and motivation mediated by interlinked, coordinated effects across distinct neural systems (Berridge and Waterhouse, 2003;Sara, 2009). Within this framework, we hypothesize that the pronociceptive action may be mediated by a subset of LC neurons also responsible for promoting wakefulness and attention (Carter et al., 2010) as part of a system to focus cognitive resources. In contrast, the antinociceptive actions may be mediated by a distinct subset that are engaged in a different behavioral context such as the "fight or flight" response to attenuate the incoming nociceptive signal to deploy defensive or escape behaviors (Heinricher et al., 2009). Bidirectional actions of the LC have been documented for other sensory modalities, in which this is thought to improve acuity (for review, see Berridge and Waterhouse, 2003;Sara, 2009;Sara and Bouret, 2012) as part of a salience detection mechanism. Given that a noxious stimulus is by definition an event with high salience (Legrain et al., 2011), it is perhaps not surprising that such bimodal modulation of nociceptive responses could be demonstrated by direct activation of the LC. This dual role of the LC may account in part for the limited efficacy of NA reuptake inhibition in the treatment of chronic pain (Finnerup et al., 2010), in which the net effect may be the sum of opposing actions, and also suggests that therapeutically targeting the whole LC for neuromodulation is unlikely to yield "pure" antinociceptive actions.