Locus Coeruleus Activation Patterns Differentially Modulate Odor Discrimination Learning and Odor Valence in Rats

ABSTRACT The locus coeruleus (LC) produces phasic and tonic firing patterns that are theorized to have distinct functional consequences. However, how different firing modes affect learning and valence encoding of sensory information are unknown. Here, we show bilateral optogenetic activation of rat LC neurons using 10-Hz phasic trains of either 300 ms or 10 s accelerated acquisition of a similar odor discrimination. Similar odor discrimination learning was impaired by noradrenergic blockade in the piriform cortex (PC). However, 10-Hz phasic light-mediated learning facilitation was prevented by a dopaminergic antagonist in the PC, or by ventral tegmental area (VTA) silencing with lidocaine, suggesting a LC–VTA–PC dopamine circuitry involvement. Ten-hertz tonic stimulation did not alter odor discrimination acquisition, and was ineffective in activating VTA DA neurons. For valence encoding, tonic stimulation at 25 Hz induced conditioned odor aversion, whereas 10-Hz phasic stimulations produced an odor preference. Both conditionings were prevented by noradrenergic blockade in the basolateral amygdala (BLA). Cholera Toxin B retro-labeling showed larger engagement of nucleus accumbens-projecting neurons in the BLA with 10-Hz phasic activation, and larger engagement of central amygdala projecting cells with 25-Hz tonic light. These outcomes argue that the LC activation patterns differentially influence both target networks and behavior.

optogenetic patterns and LC firing, described below, differences in optical activation frequencies and patterns produce distinct behavioral outcomes. Here we characterize differing outcomes of phasic and tonic LC activations in acquisition of olfactory learning, spontaneous behavior and conditioned olfactory responses. Differing optogenetic patterns and/or frequencies, at the same site, redirected LC output to support the behavioral differences we observed. We suggest how LC optogenetic input redirects LC output is a critical future question for illuminating LC function.
In the early 1980s, natural LC firing patterns were recorded in awake rats and monkeys (Foote et al. 1980). Although LC neurons in anesthetized animals are activated by noxious input, LC neurons in awake animals are sensitive to a wide range of sensory stimuli. Brain state gates these LC responses. In rats, mean LC firing in Hz was 0.02 (REM sleep); 0.69 (slow wave sleep); and 2.12 (awake). Under stressful conditions, LC neurons display tonic discharge rates of 3-6 Hz (Abercrombie and Jacobs 1987;Lechner et al. 1997). Firing rates in monkeys were similar. Sustained rates of 7-15 Hz were seen when monkeys watched a syringe of preferred juice. Sensory stimuli at 5-s intervals were associated with LC bursts up to 10 Hz with post-burst pause, or no LC firing, depending on arousal (Aston-Jones and Bloom 1981). Habituation was not observed, but see Herve-Minvielle and Sara (1995). Reduced LC excitability accompanied consummatory behavior. In the context of associative learning (Sara and Segal 1991) and odd-ball vigilance , LC responses depended on contingencies and behavior. Bursts increased to cues predictive of reward and disappeared with overtraining. Overall LC increases predicted correct decisionmaking in oddball tasks.
Although LC firing patterns are described as varying between tonic and phasic firing modes , natural LC firing appears to be aperiodic with nuanced variations in rate as seen in raster plots (Aston-Jones and Bloom 1981;Herve-Minvielle and Sara 1995;Takeuchi et al. 2016), where spike groupings in doublet and triplet patterns are common. With optogenetic tools, the functional effects of differing frequency and patterns of LC activation have been investigated. Although periodic optogenetic activation is an imperfect tool for mimicking natural LC patterns, recent investigations with this approach have provided us with rich information on LC functionality.
In awake mice, Takeuchi et al. (2016) used optogenetics to evaluate LC's role in novelty facilitation of memory consolidation. Rodents exposed to novel environments after training exhibit strengthened training memory, which was mimicked by LC activation. The LC neurons in both the familiar and novel environments exhibited aperiodic firing. The average response rate increased in the novel environment, from 0-3 to 0.6-6 Hz for the same cells. In the novel environment, 25% of LC spikes met burst criteria. Within burst firing rate averaged ∼ 18 Hz with overall firing at 2 Hz. For optogenetic purposes, cells were considered light-activated if 1/3 light pulses elicited a spike. Phasic light pulses at 25 Hz for 1 s every 5 s led to all LC spikes occurring in bursts and an overall firing increase to 4 Hz in the example provided. Consolidation of the training memory was facilitated with such an optical LC stimulation pattern.
In anesthetized rats, Vazey et al. showed that tonic activation increased EEG arousal at a threshold of 10 Hz. Tonic at 3 Hz and phasic pulses averaging 1.5 Hz (12 Hz, 3 pulses every 2 s) did not produce arousal. This permitted elucidation of differing effects of non-arousing tonic and phasic activation in the modulation of somatosensory input. Phasic input increased cortical eventrelated potential, whereas tonic did not. Both phasic and tonic input enabled subthreshold responses to become suprathresh-old, recruiting more somatosensory neurons in response to a sensory stimulus. Phasic also increased late activity in the subthreshold somatosensory neurons, mimicking those of salient painful inputs. Between light pulses, LC neurons continued with irregular firing, with overall firing unchanged. This contrasts with likely supra-arousal phasic activation in the awake rodent (Takeuchi et al. 2016) when it appeared all spikes occurred in bursts and overall firing increased. It also contrasts with Carter et al. (2010) who observed clear arousal promoting effects of 3-Hz LC activation in mice. Other studies suggest less alignment of LC spikes with optogenetic pulses in vivo. During 5-Hz tonic LC light activation in mice, average LC firing varies from 0 to 13 Hz with all cells averaging < 5 Hz (McCall et al. 2015). Tenhertz tonic light activating central amygdala (CeA) input to the LC generated the same variable LC firing, with a portion of cells showing decreased firing. However, both 5 and 10-Hz tonic light in the LC promoted anxiety and aversion. In another report, using 20-Hz pulses for 7 s, LC spiking varied from 2 to 5 Hz (Bari et al. 2020). Here, the behavioral output was improved attention and response inhibition depending on target structures.
The fascinating aspect of optogenetic experiments with phasic and tonic patterns is that they can consistently bias behavior in such different ways. Without concomitant in vivo awake recording during light activation in a behavioral task, a highly challenging undertaking, it is not possible to assume a tight relationship between optogenetic pulses and LC neuronal spiking. It is also not surprising that periodic stimuli interact in labile ways with state-dependent aperiodic LC firing. Nonetheless, different light pulse frequencies and patterns elicit distinct behavioral outcomes, even with small subject samples. Characterizing the differing outcomes and uncovering the mechanistic links between optogenetic inputs and their behavioral sequelae will deepen our understanding of how LC functions. Here we demonstrate that LC output redirection is one such mechanism, leading to the proposal that optogenetic patterns may differentially recruit LC downstream ensembles.

Animals and Ethics Statement
Tyrosine hydroxylase (TH)-CRE homozygous male breeders (Sage laboratories) were bred with Sprague-Dawley female breeder rats (Charles River) for TH-CRE heterozygous offspring that were used in this study. Rats of both sexes were housed in a 12-h light/dark cycle and had ad libitum access to food and water unless during food deprivation for experiments. During food deprivation, each rat was given 20 g of regular rat chow/day and was monitored for body-weight and health status on a weekly basis. All experimental protocols followed the guidelines of Canadian Council of Animal Care and were approved by Memorial University animal care committee.

Experimental Design and Statistical Analysis
Our study includes 4 major components, 1) establishment of in vivo LC optical activation using in vivo optrode recording and the Npas4 activity marker (Fig. 1); 2) the effects of phasic and tonic LC activations on general exploratory behavior and performance in the elevated plus maze (EPM; Fig. 2); 3) the effects of different LC activation patterns on difficult odor discrimination learning (Figs 3 and 4); and 4) the effects of different LC activation patterns on odor valence encoding (Figs 5-7). A total of 4 different light stimulation patterns were used in the of DBH (red) and EYFP (green) in the LC of a TH-CRE rat infused with an AAV8-Ef1a-DIO-eChR2 (H134R)-EYFP. (B1) An example of an in vivo recording from the LC, showing increased firing of a LC neuron to a 10-Hz, 10-s light (30-ms duration and 150-mA intensity). (B2) The waveform of the recorded cell. (C1-C3) LC firing frequency changes induced by 10-s, 10-Hz light at 30-ms duration, 150 mA (n = 6, C1), at 50-ms duration, 150 mA (n = 6, C2), or at 30-ms duration, 100 mA (n = 4, C3). (D) LC responses to light activation with a range of frequencies at 150-mA intensity (n [1/5/10/15/30 Hz] = 2/3/12/6/6). (E) An example of Npas4 staining of the LC following light stimulation. Right panel is the zoom in of the red square on the left panel. Scale bars, 50 μm. (F) Npas4 + cell counts in the control (n = 6) versus ChR2 rats (n = 7). * P < 0.05; * * P < 0.01. behavioral tests, namely: 10-Hz long phasic (10 s every 30 s); 10-Hz brief phasic (300 ms every 2 s); 10-Hz tonic; and 25-Hz tonic. For general behavioral effects, odor discrimination, and valence learning, all patterns were explored (Figs 2, 3, and 5), except 25-Hz tonic was precluded in odor discrimination learning due to increased freezing and reduced mobility induced by this pattern. Similar effects of 2 LC phasic activity patterns were observed in all behavioral experiments. Therefore, subsequent mechanistic studies used one of the 2 patterns. For studying the role of ventral tegmental area (VTA) and piriform cortex (PC) dopamine (DA) in odor discrimination learning (Fig. 4), 10-Hz brief phasic and 10-Hz tonic patterns were used. For the effect of basolateral amygdala (BLA) adrenergic receptor (AR) blockade in odor valence learning (Fig. 6), 10-Hz long phasic and 25-Hz tonic light was used. For cFos activation in BLA (Fig. 7), the 10-Hz brief phasic effect was compared with 25-Hz tonic light.
One-way repeated analyses of variance (ANOVAs) were used in Figure 1C to compare the frequency changes pre-, during-, and post-light stimulation, followed by post-hoc Tukey tests. The 2 group comparison in Figure 1F was subjected to Student's ttests (unpaired, 2-tailed). Two-way mixed ANOVAs were used to compare the effects of different patterns of light on general behavior in Figure 2A-C, followed by post-hoc Tukey tests. T-tests (unpaired, 2-tailed) were used in Figure 2D-E for closed arm and open arm times separately. Two-way mixed ANOVAs followed by linear trend analyses were used to determine statistical significance for the odor discrimination experiments, followed by post-hoc tests between the control and ChR2 groups in Figures 3 and 4B and C. One-way ANOVA followed by post-hoc Tukey tests were used in Figure 4F-H. Two-way repeated ANOVAs were used to compare the 2 odor valences in the real-time odor conditioned test (ROPT) and conditioned odor preference test (COPT) in Figures 5 and 6, followed by post-hoc Tukey tests. One-way ANOVA followed by post-hoc Tukey tests were used in Figure 7. Data are presented as Mean ± standard error of the mean (SEM) in the graphs.

Stereotaxic Surgery
Three to 10-month-old adult TH-CRE rats received bilateral virus infusions (5E +12 vg/mL) in the LC under isofluorane anesthesia in a stereotaxic frame. Each hemisphere received 2 infusions, each of 0.7 μL (fluorescent beads: virus = 2:5) at the rate of 0.5 μL/min. The cannula was lowered at a 20 • angle to avoid the transverse sinus. Infusion coordinates were 11.8-12.2-mm posterior, 1.2 and 1.4-mm bilateral, and 6.3-mm ventral with respect to bregma. At a minimum 1 month after infusion surgery, rats underwent optical fiber cannula (ferrule attached, containing optical fiber; Doric Lenses) implantation surgeries (11.8-12.2-mm posterior, 1.3-mm bilateral, and 6.3-mm ventral with respect to bregma), followed by a minimum of 2 weeks of recovery before commencing behavioral tests.

In Vivo Electrophysiology
Rats underwent the in vivo electrophysiology experiments 1-month post-infusion. Rats were anesthetized with 15% urethane at 10 mL/kg of body weight and placed in a stereotaxic frame. The surgical procedure was carried out following appropriate sterilization. A hole was drilled in the skull (12.3mm posterior and 1.3-mm left lateral to bregma) and an optrode, assembled just before the experiment (400-μm glass optical fiber; Thorlabs Inc), bundled with a 200/280-μm tungsten electrode; FHC), was lowered down at a 20 • angle to 6.1-6.9-mm ventral to brain surface until LC neurons were identified by slow spontaneous spiking and burst response to toe pinch (audiomonitor and oscilloscope response; Quinlan et al. 2018). A glass optical fiber was connected to a laser diode fiber light source (Doric Lenses) by a mono-fiberoptic patch cord (0.48 NA, 400/430 μm). Blue light of 450 nm (90 mW; Doric Lenses) for a 400-μm core and 0.48-nA optical fiber was applied. Light pattern was controlled from Doric software. Following recording, rats were perfused with 4% paraformaldehyde (PFA). Brains were extracted and stored in 4% PFA overnight and then transferred to 25% sucrose solution (in 0.1M phosphate buffer) until sectioning for Nissl staining to locate the optrode placement.
Only recordings that correctly targeted the LC were used for analysis.
Data analysis followed established procedures (Quinlan et al. 2018). Data were acquired and analyzed by SciWorks (DataWave/A-M Systems). Signal was detected at the lowest   threshold of 1.5X amplitude of the background. Autosort protocol based on 6 dimensions (peak time, peak amplitude, valley 1 amplitude, valley 2 amplitude, and 2 principle components) was used to isolate similar cellular waveforms and cluster them in a cell-specific manner. Only clusters with LC-spike characteristics (e.g., broad action potentials; Quinlan et al. 2018) were further analyzed. Frequency histograms were generated to compare the firing rates of LC cells before, during, and after each light protocol.

Light Stimulation for Behavioral Experiments
Two to 4 weeks after optical cannula implantation surgery, rats underwent behavioral tests. Bilateral photostimulation at 450 nm (20 mW/mm 2 at fiber tip) was delivered by 2 laser light sources (LDFLS_450; Doric Lenses) through mono-fiberoptic patch cords. Current equivalence of power was 150 mA. Different patterns of stimulation were controlled from Doric software. Behavioral sessions were video-recorded by ANY-Maze software (Stoelting) and analyzed offline. Subsets of experiments were analyzed by persons who were blind to the experimental conditions.

Drug Infusion
Drugs or vehicle were infused 30 min before behavioral testing through a 10-μL Hamilton syringe and infusion pump. Lidocaine (4%; Sigma) was infused in the VTA (0.3 μL/hemisphere over 3 min with an additional 1-min wait before withdrawing the syringe; Nunes et al. 2019). For PC and BLA infusions, 1 μL of drug or vehicle was infused per hemisphere over 2 min followed by a 1-min wait before withdrawing the syringe. The D1/5 receptor antagonist SCH 23390 (3.47 mM, Sigma), the α-AR antagonist phentolamine hydrochloride (10 mM; Sigma), and the β-AR antagonist alprenolol hydrochloride (120 mM; Tocris Bioscience; Shakhawat et al. 2015;Kempadoo et al. 2016) were used for PC and BLA infusions.

General Behavioral Experiments Open Field Maze
Rats explored an opaque plexiglass box (60 × 60 × 40.5 cm) with a black bottom for a 10-min daily session for 4 consecutive days while receiving no stimulation on day 1, 10-Hz long phasic stimulation on day 2, 10-Hz tonic stimulation on day 3, and 25-Hz tonic stimulation on day 4. Distance traveled, time spent rearing (both free and supported), and time spent freezing were recorded and analyzed. Freezing was counted as no body movement except breathing.

Elevated Plus Maze
Following a 15-min photo-stimulation in the home cage, rats were placed in the center of an EPM (50 × 10 cm each arm, 38cm wall on the closed arms, 11 × 11-cm central platform, 52-cm high from the ground, inside painted black) facing the open arm opposite to the experimenter. Rats spent 5 min in the EPM while the stimulation continued. Time spent in closed and open arms was recorded.

Olfactory Behavioral Experiments Odorants
Odors used in the experiments are listed in Table 1

Odor Discrimination Learning
Rats were food deprived for 4-7 days before the onset of the experiments and food deprivation continued during the course of the experiment. Five days of habituation for context (box [60 × 60 × 40.5 cm], sponge and positive reinforcement [chocolate cereal]) were conducted first. Afterwards rats performed an odor discrimination task consisting of 10 trials/day, each trial being a maximum of 3 min. Two sponges containing odor 1 (O1) and odor 2 (O2), respectively were presented randomly in 2 corners of the box. The O1 sponge had a 2-cm 2 hole at the top center, containing a retrievable chocolate cereal. To balance the smell of the chocolate cereal, a non-retrievable chocolate cereal was placed in a hidden hole in the O2 sponge. Between trials rats were confined to a "home corner" in the box with an L-shaped plexiglass barrier for 20 s while sponge positions were changed. During trials rats were allowed to explore the box and sniff the sponges. Trials ended as soon as a nose poke was made inside the hole, irrespective of the odor identity.
The response was considered correct if nose poke was in O1 sponge. Trials in which no nose-poke occurred within 3 min were excluded from analysis. Percentage of correct responses was counted as the number of correct responses over the number of total nose pokes. Light stimulation was given only during trials and terminated at nose-poke, but not during intertrial intervals. Dissimilar odor discrimination continued for a minimum of 7 days until rats reached 2 consecutive days of 80% success rate, followed by similar odor discrimination for 10 days.

Odor Valence Tests
For the COPT, the innate preference for 2 odors were first measured in a T-maze, then one of the odors was paired with LC light activation and preference for the 2 odors were tested again in the T-maze. Rats underwent a single 30-min session of habituation in a T-maze (long arm 183 × 19 cm, neutral arm 19 × 19 cm; and 20.5-cm high wall) on day 1. On day 2, O1 and O2 sponges were placed in 2 opposing arms, with positions counter-balanced between morning and afternoon sessions. Time spent in the arms corresponding with O1 and O2 during a 10-min session was recorded. On day 3, rats were confined to the O1 arm for 10 min in the morning with light stimulation; and to the O2 arm in the afternoon without light stimulation. The odors were switched in the arms on day 4 and the same conditioning as day 3 was repeated. Rats were trained for 1 trial or 3 trials (repeating procedures as in day 4-5 three times) and data were pooled. On the testing day, arm time was measured in morning and afternoon sessions with O1 and O2 sponges positioned in the exact same manner as day 2.
For the ROPT, rats were placed in a T-maze with 2 odorized arms and allowed to explore freely. Baseline preference for 2 odors was first recorded without light stimulation. During ROPT, LC light stimulation was associated with rats' presence in one odorized arm, but not in the other. A 122-cm long arm was used. Day 1 habituation and day 2 baseline odor responses were conducted in the same manner as in COPT. On day 3, rats explored the maze freely for two 10-min sessions in the morning and afternoon. Light stimulation started upon a rat entering the O1 arm and stopped upon rat leaving the O1 arm. Odor positions were switched between morning and afternoon sessions.

CFos and Npas4 Induction Experiments
For cFos induction in Figures 4 and 6, rats were habituated to the experimental environment for 2 days. On the third and fourth days, rats were optically stimulated with either phasic or tonic patterns in their home cages, for 10 min/day, while being exposed to an odorized sponge (benzaldehyde 0.05%). Control rats were exposed to the odor only without light stimulation. Ninety min following the odor+light, or odor only stimulation on the fourth day, rats were anesthetized, perfused, and brains were collected.
For Npas4 induction to validate light activation in Figure 1, rats were habituated to the experimental environment for 1 day. Next day they were perfused 30-min following light stimulation in their home cage.

Immunohistochemistry and Histology
Rats underwent trans-cardiac perfusion with cold isotonic saline followed by 4% PFA. Brains were extracted and kept in 4% PFA. Brains were then sectioned using a vibratome (Leica VT 1000P; Leica Biosystems) in 50-μm thick coronal slices and saved in a polyvinylpyrrolidone solution. For immunohistochemistry with Npas4 and DBH, slices were washed in PBS, and then incubated with primary antibodies.
Npas4 (1:500, Thermo Fisher Scientific) primary antibody mixed in phosphate buffer saline (PBS) with 2% normal goat serum and 0.2% Triton-X was applied for 3 nights at 4 • C. Following a 3 × 10 min wash in PBS, a biotinylated anti-rabbit secondary antibody (1:1000; Vector laboratories, Burlingame, CA) was applied. After 2 h of secondary incubation and a 3 × 10 min wash, slices were incubated in an avidin-biotin complex for 1.5 h followed by a PBS wash. Slices were then placed in a solution containing 15-μL SG Gray chromogen (Vector laboratories) with 24-μL peroxide per ml of PBS. After optimum color development, slices were washed in PBS, dried overnight, dehydrated in graded ethanol, and mounted with permount.
For cFos and TH co-labeling, 50-μm free floating sections from phasic and tonic stimulated rats, belonging to similar positions in the anterior-posterior axis of the brain as determined by unaided visual observation, were chosen in an unbiased manner. The sections were washed in Tris buffer (0.1 M, pH 7.6) twice for 10 min each, followed by 10 min in Tris A (0.1% TritonX in Tris buffer), and Tris B (0.1% TritonX and 0.005% BSA in Tris buffer) before applying a blocking solution of 10% normal goat serum (Sigma-Aldrich) for 1 h. This was followed by a 10-min wash each in Tris A and Tris B before incubating in 1:2000 primary antibody solution prepared in Tris B at 4 • C (TH, EMD Millipore; cFos, Cell Signaling). After 2 nights, sections were washed for 10 min each in Tris A and Tris B and incubated in a 1:1000 secondary antibody solution prepared in Tris B at 4 • C (anti-rabbit Alexa 647, anti-mouse Alexa 488; Invitrogen/Thermo Fisher Scientific). This was followed by 10-min washes in Tris A, Tris D (0.1%Triton X and 0.005% BSA in 0.5 M Tris buffer), and Tris buffer, respectively. Finally, sections were mounted with antifade mounting medium.
Nissl staining was done by rehydrating the slides in graded ethanol, incubating in 0.5% cresyl violet for 8 min, washing in distilled water for 1 min, and then dehydrating in graded ethanol. After a 5-min xylene step, slides were coverslipped with permount.

Image Acquisition and Analysis
Images were acquired by an EVOS 5000 (Thermo Fisher Scientific), and a BX-51 (Olympus) for fluorescent and brightfield images. Images were acquired similarly for phasic and tonic stimulated rats, keeping gain and exposure time the same throughout each experiment. Images were analyzed using ImageJ software. Images underwent background subtraction before manual cell counting. For LC activation success, the number of Npas4 + cells was counted. For BLA and VTA activation, cFos + cells and double-labeled CTB cells were counted. Three to 6 images per animal were analyzed and values from both hemispheres were averaged. A subset of the images was analyzed blindly.

Validation of Light Activation of LC Neurons
Three weeks following LC AAV infusion (for targeting see Supplementary Fig. S1), we observed LC ChR2 uptake marked by the fluorescence reporter EYFP (Fig. 1A) or mCherry. All rats examined expressed the reporter bilaterally in the LC. The percentage of LC neurons (DBH expressing cells) transfected with ChR2 (co-expressing EYFP or mCherry) were 73.7 ± 12.8% (n = 8), and 92.3 ± 2.7% ChR2 cells were DBH + . We conducted in vivo optrode LC recordings. Figure 1B1 and B2 shows an LC neuron activated by a 10-s, 10-Hz light train (30-ms pulses, laser intensity 150 mA). Figure 1C1-C3 shows LC activation by 10-Hz trains with 2 light intensities and pulse widths. The LC firing frequency increases were significantly induced by the 10-s, 10-Hz light at 30-ms duration, 150 mA (F 2,10 = 8.69, P = 0.006, n = 6; Fig. 1C1), or 10 Hz at 50-ms duration, 150 mA (F 2,10 = 25.42, P < 0.001, n = 6; Fig. 1C2), but not at 30-ms duration, 100 mA (F 2,6 = 2.48, P = 0.16, n = 4; Fig. 1C3). The 30-ms, 150-mA light that effectively activated LC firing was used subsequently in recording (Fig. 1D) and behavioral experiments. Increasing light frequency elevated LC firing in ∼linear fashion in the frequency range < 30 Hz (Fig. 1D), with a firing output up to 15 Hz. Although average LC firing rates with 5-and 10-Hz light approximate activation rates, we found LC neurons were not individually driven at those rates in vivo. Similar outcomes were observed in mouse in vivo where average LC firing with light activation was made up of LC neuron responses of widely varying frequencies (McCall et al. 2015). A caveat is that LC output from optrode recording is not an accurate estimate of the average effects throughout the LC, as the electrode tip is in close proximity to the tip of the optical fiber where light intensity is maximal. Therefore, average LC output during the light stimulation is likely lower than the output reported here. The immediate early gene Npas4 revealed light-induced expression in LC (t = 4.53, P < 0.001, n [ChR2/Control] = 6/7; Fig. 1E and F). Recordings with the optrode placements outside the LC did not respond to the light (an example see Supplementary Fig. S2).

Different Patterns of LC Activation Show Distinct General Behavioral Effects
First we investigated whether phasic and tonic LC activations have differential effects on locomotion, exploratory behavior, and stress. We included 4 patterns including 2 phasic patterns: 10-Hz long phasic (10 s every 30 s), and 10-Hz brief phasic (300 ms every 2 s). These patterns are consistent with recent studies in terms of frequency range and duration (Carter et al. 2010;Kempadoo et al. 2016;Vazey et al. 2018). The 10-Hz brief phasic pattern mimics physiological LC firing in response to environmental stimuli (Aston-Jones and Bloom 1981; Nakamura et al. 1987). The 2 tonic patterns are 10-Hz tonic and 25-Hz tonic, corresponding to the LC output in the range of 10-15 Hz in our in vivo recording (Fig. 1D).
The effects of these LC activation patterns on open field distance traveled, duration of rearing, and freezing were tested ( Fig. 2A-C). In the open field experiments, rats underwent a series of tests in a fixed order: baseline without light activation, 10-Hz long phasic, 10-Hz tonic, and 25-Hz tonic. Data generated by different light patterns were normalized to the same baseline parameters for comparisons. A subset of rats underwent baseline measurements followed by 10-Hz brief phasic light activation (see Supplementary Fig. S3). Ten-hertz phasic and tonic stimulated rats showed increased rearing while 25-Hz rats showed less mobility and increased freezing.
Anxiety during tonic and phasic 10-Hz light stimulations was measured in an EPM ( Fig. 2D and E), and no differences were observed in the light-activated and control groups.

LC Phasic Patterns Enhance Similar Odor Discrimination Learning
Rats were trained to associate a food pellet with one odor from an odor pair (Fig. 3A). After learning the simple odor discrimination (SOD) with a dissimilar odor pair in the absence of light activation (Fig. 3B-D), rats then learned the DOD with a similar odor pair with LC activation (Fig. 3E-G). Three light patterns were used during DOD training: 10-Hz long phasic (Fig. 3B and E), 10-Hz brief phasic (Fig. 3C and F) and 10-Hz tonic (Fig. 3D and G). The 25-Hz tonic stimulation that induced significant freezing was not included in this learning paradigm. While both phasic patterns facilitated DOD learning, 10-Hz tonic light activation had no effect on DOD learning.
We then tested whether LC phasic activation also improves SOD. Our result showed that SOD learning was not affected by phasic 10-Hz LC activation (n [ChR2/Control] = 6/7; Fig 3H). There was no Group X Day interaction (F 6, 66 = 0.755, P = 0.608), or Group effect (F 1,11 = 0.527, P = 0.483). Latency to nose poke showed reduction with training, and was not different between the control and ChR2 groups (see Supplementary Fig. S4). This underscores norepinephrine (NE's) role in difficult discriminations that require pattern separation (Shakhawat et al. 2014(Shakhawat et al. , 2015. Enhancement of subtle tactile discriminations by tonic 5-Hz LC activation (Rodenkirch et al. 2019) has also been reported in rats, a frequency not explored here.

LC-VTA-PC DA Circuitry Mediates the Facilitating Effect in Similar Odor Discrimination Learning
Increased extracellular DA in the hippocampus following LC stimulation has been shown to be critical in spatial learning (Kempadoo et al. 2016) and novelty-mediated memory consolidation (Takeuchi et al. 2016). We next tested the potential involvement of NE and DA in the PC upon LC 10-Hz brief phasic light activation in DOD (Fig. 4A). The DOD was prevented when a mixture of α1-AR antagonist phentolamine and β-AR antagonist alprenolol were infused in the PC before training, however, the D1/5 receptor antagonist SCH 23390 infusion selectively abolished LC phasic light induced learning facilitation (n [ChR2 vehicle/Control vehicle/ChR2 AR block/ChR2 DR block] = 9/6/6/6); Fig. 4B), but not DOD acquisition. There was a significant Day effect (F 9, 207 = 15.98, P < 0.001), a Group X Day interaction (F 27, 207 = 5.196, P < 0.001), and a Group effect (F 3, 23 = 17.73, P < 0.001). A significant difference in learning was observed between the D1/5 antagonist group and ChR2 vehicle group (t = 8.624, P < 0.001), and the AR antagonist group (t = 8.091, P < 0.001), and between the AR antagonist group and the ChR2 vehicle group (t = 17.488, P < 0.001). However, the D1/5 antagonist group performed similarly to the non-ChR2 control vehicle group (t = 1.421, P = 0.748).
These results argue that while NE in the PC is essential for pattern separation-dependent odor discrimination to occur, DA release in the PC during LC phasic light contributes a learning facilitating effect. There are 2 possible scenarios: either LC axon terminals co-release DA or increase extracellular DA through other mechanisms upon LC phasic activation (Kempadoo et al. 2016;Takeuchi et al. 2016), or VTA releases DA into the PC (Datiche and Cattarelli 1996;Aransay et al. 2015) upon LC activation. To determine the source of DA released during odor discrimination learning, we infused lidocaine into the VTA to silence the VTA during DOD. Lidocaine infusion prevented the learning facilitation effects of the LC phasic light (n [lidocaine/vehicle] = 6/6; Fig. 4C), but not the acquisition of DOD learning. There was a significant Day effect (F 9, 90 = 41.46, P < 0.001), a Group X Day interaction (F 9, 90 = 5.60, P < 0.001), and a Group effect (F 1, 10 = 90.94, P < 0.001). Higher correct response rates of the ChR2 vehicle group relative to the ChR2 lidocaine group on days 4-7 were observed (P < 0.01). However, the performance of the 2 groups on the last 3 days (8-10) was comparable (P > 0.05).
Why does phasic LC activation promote DOD while tonic activation at the same frequency does not? In other words, does LC phasic activation engage DA neurons in the VTA more effectively than tonic activation? We next studied neuronal activation patterns in the VTA induced by phasic versus tonic LC activations. We measured cFos expression in the VTA following odor exposure only (no-light control), 10-Hz brief phasic or 10-Hz tonic LC activation paired with an odor (n [control/tonic/phasic] = 5/5/5; Fig. 4D). The phasic light increased overall cFos activation in the VTA, as well as the portion of activated TH + cells (Fig. 4E-G). Despite the similar numbers of TH + cells in the 3 groups (F 2,12 = 0.062, P = 0.940), there were different proportions of cFos + cells (F 2,12 = 11.633, P = 0.002) and TH + /cFos + cells (F 2,12 = 10.436, P = 0.002; Fig. 4E and F). The phasic pattern activated significantly more cFos + cells in the VTA than the nolight control (t = 4.649, P = 0.017) and the tonic pattern (t = 6.647, P = 0.001), and generated a larger number of TH + /cFos + cells compared with the control (t = 4.598, P = 0.017) and the tonic activation (t = 6.230, P = 0.002). The percentage of cFos + cells in the total TH + population is significantly higher with the phasic pattern than with the no-light control and the tonic pattern (F 2,12 = 33.984, P < 0.001; Fig. 4G). The percentage of TH + /cFos + cells over total cFos + cells is also higher in the phasic group (F 2,12 = 4.938, P = 0.027; Fig. 4H). These results argue that the 10-Hz LC phasic pattern engages VTA DA neurons whereas the 10-Hz tonic pattern does not. This may explain the learning facilitation effect of PC DA with phasic, but not tonic, LC activation.

LC Phasic and Tonic Patterns Promote Differential Odor Valence Learning
Besides differentially modulating discrimination learning by different LC activation patterns, tonic LC activation is involved in stress and aversive learning (Hirschberg et al. 2017;McCall et al. 2017;Llorca-Torralba et al. 2019). However, the effect of phasic LC activation on adult valence encoding is unknown. We next tested the intriguing possibility that tonic and phasic LC activations could differentially mediate valence learning. Tonic LC optical stimulation has been associated with real-time place aversion and conditioned place aversion (McCall et al. 2015(McCall et al. , 2017. Here, we employed similar tests with odors using ROPT and COPT (Fig 5A).

LC Phasic and Tonic Patterns Engage Differential BLA Circuitry in Odor Valence Learning
In mice, it has been shown that LC tonic activity promotes aversive behavior through BLA β-adrenoceptors (McCall et al. 2017). The BLA is a critical site for valence associative learning (O'Neill et al. 2018). The BLA has been found to contain functionally distinct neuronal populations projecting to either negative-(CeA) or positive-(NAc) valence encoding circuitry (Kim et al. 2016;Namburi et al. 2016). It is plausible that the LC-BLA projection is involved in the odor valence encoding observed here with differential LC light activation patterns.
Taken together, while BLA NE mediates conditioned valence learning dependent on differential patterns of LC activation, this circuitry is not involved in real-time preference in ROPT. Phasic light mediated enhanced exploration (Fig. 2B) may explain the acute ROPT effect of 10-Hz phasic light. This is consistent with a previous report that tonic LC mediation of real-time place aversion is not associated with the LC-BLA projections (McCall et al. 2017).
After establishing the requirement of BLA NE in LC light mediated valence learning in COPT, we next tested whether tonic and phasic activation of the LC biases activation of the BLA ensembles projecting to BLA-CeA aversive and BLA-NAc reward circuitry respectively. We infused retro-tracing dyes linked to CTB in the CeA and NAc, and examined the overlap of CeA or NAc projecting neurons with cFos + cells in the BLA activated by odor only (no-light control), 10-Hz brief phasic or 25-Hz tonic LC lights ( Fig. 7A and B). The CTB labeled CeA (F 2,7 = 2.028, P = 0.202) and NAc (F 2,7 = 0.340, P = 0.723) projecting cell numbers were comparable in the 3 groups (n (control/tonic/phasic) = 4/3/3; Fig. 7C). Intriguingly, although the 2 LC light patterns activated similar numbers of cFos + cells in the BLA compared with the control (F 2,7 = 0.888, P = 0.453; Fig. 7C), the distribution patterns of cFos + cells were dramatically different (see example images in Fig 7B). The proportion of CeA + cells that were cFos + cells was significantly higher in the 25-Hz tonic group (F 2,7 = 14.232, P = 0.003; Fig. 7D) compared with either the nonlight control (t = 4.539, P = 0.008) or the 10-Hz brief phasic light (t = 4.796, P = 0.006). On the other hand, the proportion of NAc + cells in cFos + cells was significantly higher in the 10-Hz phasic light group (F 2,7 = 10.648, P = 0.008; Fig. 7E) compared with either non-light controls (t = 4.033, P = 0.015) or the 25-Hz tonic light (t = 4.052, P = 0.015). The differential distributions of cFos + cells in different groups are displayed in the pie charts (Fig. 7F). In nolight controls, equal amount of cFos + cells (11%) were NAc and CeA projecting cells, whereas 37% were NAc projecting and 5% were CeA projecting with the 10-Hz brief phasic light, and 54% were CeA projecting and 9% were NAc projecting with the 10-Hz tonic light. A small portion of projecting cells (4.4%) expressed both CTBs, however, the activation of the double-CTB-labeled cells was very low (0.72%). Additionally, phasic and tonic light activations in the absence of an odor did not lead to a difference in cFos activation patterns (see Supplementary Fig. S5), consistent with a modulatory role of NE. Phasic and tonic modes of LC activations re-distributed neuronal ensembles activated by an odorant in the BLA. Selective activation of one valence encoding ensemble may inhibit the ensemble of the opposite valence in the BLA (Namburi et al. 2015). Taken together, the 10-Hz LC phasic pattern preferentially activates NAc projecting neurons in the BLA, whereas the 25-Hz tonic LC activation preferentially engages CeA projecting neurons.
VTA projections to the NAc facilitate BLA-NAc circuitry in promoting a reward-seeking response to sensory cues (Yun et al. 2004;Ambroggi et al. 2008). The 10-Hz phasic, but not 10-Hz tonic, LC activation, efficiently recruited VTA dopaminergic neurons in our study (Fig. 4D-H) suggesting that 10-Hz phasic LC activation likely engages the VTA-NAc pathway more efficiently compared with a tonic pattern. This is indeed the case when we compared the cfos activation in the VTA by 10-Hz phasic light versus 25-Hz tonic light (see Supplementary Fig. S6). More cFos + cells were observed with 10-Hz phasic LC activation and a bigger portion of NAc projecting neurons were involved.

Discussion
This work aimed to further our understanding of whether and how the LC produces activation mode-specific behavioral responses. We assessed how LC activation patterns differentially modulate general behavior, odor discrimination learning and valence encoding.
Both tonic and phasic 10-Hz activation promoted exploration as indexed by increased rearing. Consistent with this outcome, exploratory rearing was previously enhanced with NE or a βadrenoceptor agonist infused in hippocampus (Flicker and Geyer 1982;Geyer and Masten 1989). Increased immobility, anxiety and aversion, often occurring with 5-10-Hz tonic activation in mice (McCall et al. 2015), were only seen here with 25 Hz (with an output of ∼ 15 Hz at the site of the optical stimulation). Possibly optogenetic fiber size in relation to LC extent may support more extensive LC activation in mouse. Spatially extensive recruitment is likely to be stress-encoding. This hypothesis is consistent with the outcome of chemogenetic LC activation (Llorca-Torralba et al. 2019) which seems to invariably induce anxiety and promote place aversions.
The roles of LC-NE in different aspects of olfaction and olfactory learning have been characterized extensively (see reviews Fletcher and Chen 2010;Linster and Escanilla 2019). Here, we compare phasic and tonic LC activation on DOD learning using similar odor pairs in adult rats. Discrimination of similar odors requires NE in the PC (Shakhawat et al. 2015) and the olfactory bulb (OB; Doucette et al. 2007;Mandairon et al. 2008). Increased NE lowers thresholds for odor discrimination (Escanilla et al. 2010) and is associated with higher signal-to-noise ratios (de Almeida et al. 2015). Norepinephrine enhances principle neuron excitability and regulates inhibition in both the OB (Trombley and Shepherd 1992;Hayar et al. 2001;Nai et al. 2009;Nai et al. 2010;Pandipati et al. 2010;Lethbridge et al. 2012) and PC (Gellman and Aghajanian 1993;Brosh et al. 2006;Ghosh et al. 2015) in a dose and receptor subtype-dependent manner, which may contribute to the improvement in signal-to-noise ratio. Norepinephrine enhancement of LTP processes (Yuan et al. 2000;Yuan 2009;Morrison et al. 2013) is also likely to enhance discrimination learning. In adult rats, impaired odor pattern separation is associated with AR blockade (Shakhawat et al. 2015), whereas a short burst of electrical stimulation of the LC sharpens odor representation in the PC (Bouret and Sara 2002). Here we show that phasic, but not tonic LC activation, facilitates odor discrimination learning.
Phasic 10-Hz optogenetic patterns significantly enhanced the rate of acquisition of similar odor discrimination learning. While PC adrenoreceptors are required for similar odor discrimination (see also Shakhawat et al. 2015), it was D1/D5 receptor activation that supported enhancement of acquisition. Increased extracellular DA following LC stimulation to the dorsal hippocampus has been shown to mediate novelty induced learning (Takeuchi et al. 2016) and enhance spatial learning (Kempadoo et al. 2016). In contrast, the source of PC DA in our study was the VTA as shown by prevention of enhancement with VTA lidocaine. Phasic LC activation recruited VTA DA neurons indexed by cfos, whereas tonic activation at the same frequency did not enhance acquisition and did not activate VTA DA neurons. Thus, LC output was directed to VTA by the 10-Hz phasic, but not the 10-Hz tonic optogenetic pattern. A recent paper (Soares-Cunha et al. 2019) reported that NAc neurons signal both reward and aversion depending on the optical stimulation patterns. Brief phasic pattern induces reward, whereas prolonged tonic-like stimulation leads to aversion. Interestingly, only the brief phasic stimulation is associated with increased VTA DA tone, similar to what we observed in our study.
Pauses in the phasic pattern are likely to be critical features of the mechanistic effects of phasic patterns. The LC firing pauses may reset target receptor desensitization and have been identified as required for spatial encoding consolidation in slow wave sleep (Swift et al. 2018). Preventing receptor desensitization may enhance both cell recruitment by an input and plasticity during learning. Pauses are also required for DA's role as a prediction error and teaching signal (Chang et al. 2018).
We also compare the 2 LC activation patterns on odor valence. Tonic LC patterns promote aversions and anxiety in mice (McCall et al. 2015(McCall et al. , 2017 and rats (Hirschberg et al. 2017;Llorca-Torralba et al. 2019). A conditioned place aversion and increased anxietylike behavior have been demonstrated by increasing tonic firing of prefrontal cortex-projecting (Hirschberg et al. 2017) and BLA-projecting LC neurons (McCall et al. 2017;Llorca-Torralba et al. 2019). While phasic LC activation is regarded as learningpromoting, our result suggests that phasic LC activity also carries positive valence.
By prior pairing of an odorant with either 10-Hz phasic or 25-Hz tonic activation, a conditioned odor preference or conditioned odor aversion was acquired respectively. Prior studies suggested aversive place conditioning depended on β-adrenoceptors in the BLA (McCall et al. 2017). Here adrenoceptor involvement was again confirmed for conditioned odor aversion learning. The BLA adrenoceptor activation was also required for conditioned odor preference learning. However, the subsets of basolateral neurons recruited by the phasic and tonic patterns of LC activation were significantly different. Positive valence induced by 10-Hz phasic LC activation stimulated the BLA neurons projecting to NAc, while tonic LC activation at 25 Hz stimulated BLA neurons projecting to CeA. The BLA-NAc pathway has been previously associated with reward, whereas BLA output to CeA is linked to aversive effects (Namburi et al. 2015;Kim et al. 2016). Although LC-BLA engagement in stress and aversive learning has been well documented (McCall et al. 2017;Uematsu et al. 2017), our result provides evidence that phasic activation of the LC leads to positive valence. In the 1970s, phasic electrical stimulation of LC was identified as having positive valence (Ritter and Stein 1973), but this association was criticized as electrical stimulation lacks specificity and a causal relationship was not established (Wise 1978). The present experiments address these deficiencies and reveal a positive valence associated with phasic LC activation. A caveat is that different anatomical portions or different numerical portions of neurons were activated in these experiments. However, it is unlikely that difference in neuron recruitment is a factor given that our rats were randomly assigned to each group and variations in LC activation would be random among groups. This outcome is consistent with the role for LC-hippocampal fiber activation in promoting reward remapping among place cells (Kaufman et al. 2020).
The mechanisms that underlie the differential activations in downstream structures by LC light patterns remain to be determined. The adrenoceptor heterogeneity in the target structures and differential NE release triggered by LC activation patterns may explain the different outcomes. A brain wide MRI study of the changes engendered by chemogenetic LC suggests such activation increases connectivity in salience and amygdala networks that is related to the density of α1 and β1 adrenoceptors (Zerbi et al. 2019). It will be interesting to see how phasic patterns influence such brain connectivity patterns. Other differences in target area receptor-recruitment as a function of activation frequency have also been described. Koga et al. (2020) compared 1 min of 5-Hz or 20-Hz light trains on LC fibers entering anterior cingulate cortex. They found both 5 and 20 Hz enhanced the frequency of presynaptic glutamate release. Although 5 Hz engaged β-adrenoceptor activation, 20-Hz trains further recruited α1-adrenoceptors to increase inward current.
An increase in presynaptic glutamate release frequency is consistent with the predictions of the "glutamate amplifies noradrenergic effects" (GANE) hypothesis (Mather et al. 2016). GANE proposes that LC effects depend on glutamate network activity in target structures. Koga's outcome supports a presynaptic effect of NE on glutamate release events. Enhancement of local glutamatergic activity will yield varying behavioral outcomes depending on competition among active glutamate circuits. Indeed, all behavioral outcomes with LC activation depend on the nature of activity in the target structures. Glutamate released onto VTA-DA neurons is enhanced by pre-synaptic α1-adrenoceptors (Velasquez-Martinez et al. 2012), which likely occur in our study where phasic LC light facilitates VTA neuron activation during reward-based odor discrimination training. A recent study suggests that LC activation induced DA increase in the dorsal hippocampus facilitates glutamatergic plasticity via pre-synaptic NMDARs (Sonneborn and Greene 2021). Nonetheless, the requirement for a tight pairing of input and LC activation is not necessary for the effects reported here, as well as the hippocampaldependent learning (Kempadoo et al. 2016;Takeuchi et al. 2016). The LC timing relative to input is critical, however, for attentional effects ) and spike-timing dependent plasticity (Seol et al. 2007;Huang et al. 2013).
Although there are many variables to consider in explaining differential behavioral effects of phasic and tonic optogentic patterns, the present data suggest 2 possibilities. One is differential downstream recruitment (different structures or different ensembles within the same structure) due to pattern-dependent LC NE release and adrenoceptor heterogeneity in the target structures. Another hypothesis is that different patterns may recruit different subpopulation of LC neurons that have discrete projections. Anatomical heterogeneity within the LC is now thought to support distinct LC behavioral roles (Schwarz and Luo 2015;Uematsu et al. 2015;Chandler et al. 2019). These ensembles are distributed rather broadly through the LC, but may be differentially sensitive to optogenetic patterns. We suggest that if the different pattern-dependent outcomes illuminated in the present experiments can be integrated with the structural and functional evidence for LC and downstream ensembles (Schwarz and Luo 2015;Uematsu et al. 2015Uematsu et al. , 2017Chandler et al. 2019), we will have novel insights into LC operation with broad implications for both basic and clinical brain science.

Supplementary Material
Supplementary material can be found at Cerebral Cortex Communications online.