Norepinephrine release in the cerebellum contributes to aversive learning

Traumatic events forge learned associations of sensory signals with aversive outcomes and can lead to excessive conditioned fear responses as well as anxiety and posttraumatic stress disorders. While dopamine release from ventral midbrain axons in the striatum is central to reward-based learning, the circuitry underlying aversive learning has been debated, with evidence implicating subsets of ventral midbrain dopamine neurons and locus coeruleus norepinephrine (LC-NE) neurons. NE release from LC projections has been implicated in fight-or-flight responses to stressful stimuli and stimulus-associated fear; aversive stimuli and stress trigger NE release in regions such as the hypothalamus and amygdala where it activates β-adrenergic receptors. Pharmacologically, β-adrenergic receptor antagonists or α2-adrenergic receptor agonists impair fear acquisition. LC axons project to the cerebellum, a region implicated in emotional associative learning via outputs to the amygdala, periaqueductal gray, hypothalamus, and prefrontal cortex. Stimulation of the cerebellar vermis produces freezing and bradycardia, whereas lesions of lobules IV-V inhibit autonomic and behavioral fear responses. However, whether NE signaling within the cerebellum supports fear learning had not been directly tested. The authors hypothesized that conditioned aversive learning relies on experience-dependent increases of NE release in the cerebellum when a conditioned stimulus (CS) becomes associated with an aversive unconditioned stimulus (US), and that LC-NE projections to the cerebellum are required for fear memory without affecting motor function.

Prior work shows widespread LC catecholaminergic projections to many brain regions including the cerebellum, with dense TH-immunoreactive fibers across cerebellar lobules and nuclei and highest expression in the anterior vermis and cerebellar nuclei. NE release is linked to arousal and stress, and foot-shock increases NE in the amygdala, engaging β-adrenergic receptors. Systemic manipulation of adrenergic receptors impairs fear acquisition. The cerebellar vermis has been implicated in fear-related autonomic and behavioral responses; stimulation induces freezing/bradycardia and lesions impair fear responses. Parallel fiber–Purkinje cell plasticity has been proposed as a substrate for fear memory and is modulated by NE via β-adrenergic receptor signaling, cAMP/PKA, and AMPAR trafficking. Cerebellar outputs via the fastigial nucleus to the ventrolateral periaqueductal gray modulate freezing circuits. LC projections to lateral cerebellar nuclei also influence associative learning and cognition. LC projections to amygdala and hippocampus contribute to cued and contextual fear, respectively, highlighting a broader LC role across fear circuits. These studies motivated testing whether cerebellar NE release is learned and necessary for aversive memory formation.

Experimental model: Adult C57BL/6J and TH-IRES-Cre mice (>8 weeks; both sexes) were used under IACUC-approved protocols. Mice were housed on a reverse light/dark cycle with ad libitum food and water. Group sizes: TH-Cre for LC labeling (n=6), FFN270 imaging (n=7 C57BL/6J), GRABNE photometry (n=24 C57BL/6J), chemogenetics hM4Di vs YFP (15 vs 11 TH-Cre), optogenetics Arch3 vs YFP during conditioning (10 vs 3 TH-Cre), and Arch3 during recall (n=7 TH-Cre). Viral tools and targeting: For LC fiber labeling, AAV5-synP-DIO-eGFP was injected into LC of TH-Cre mice. For photometry, AAV9-hSyn-GRAB_NE1m was injected unilaterally into cerebellar vermis (CB). For chemogenetic inhibition of LC axons, AAV5-hSyn-DIO-hM4D(Gi)-mCherry was injected into LC of TH-Cre mice; for controls, AAV5-EF1a-DIO-eYFP. For optogenetic inhibition, AAV5-EF1a-DIO-eArch3.0-eYFP was injected into LC; controls received eYFP. Typical injection volume was 230 nl at 1×10^13 vg/ml. Surgery: Performed under 2% isoflurane using stereotaxic coordinates. Example coordinates: CB anterior vermis AP −6.7, ML 0.75, DV 1 mm; LC AP −5.45, ML 1.3, DV 3.8 mm (relative to bregma). For photometry, an optic fiber was implanted 0.1 mm above the CB injection site. For optogenetics, bilateral optic fibers were implanted in CB vermis. For chemogenetics, bilateral 26-gauge guide cannulas were implanted in CB vermis for microinfusion. Implants were 0.5 mm lateral to midline (bilateral) or 0.5 mm right of midline (unilateral photometry). Experiments were conducted ≥4 weeks post-surgery. Implant locations were verified postmortem. Histology and imaging: Standard perfusion with 4% PFA, 40 µm sections. Immunofluorescence with anti-GFP and anti-TH primary antibodies; secondary antibodies conjugated to 488 and 647 fluorophores. Imaging with an Olympus IX81 fluorescence microscope. To visualize NE axons and vesicles, acute cerebellar slices were incubated with FFN270 (10 µM), a fluorescent NE analog taken up by NET and loaded into vesicles via VMAT2. Two-photon microscopy imaged FFN270 puncta in en passant varicosities; NET dependence was tested by nomifensine (5 µM) co-incubation. Behavioral paradigms: Pavlovian auditory fear conditioning used a 30 s tone CS (80 dB, 3.5 kHz) co-terminating with a 1 s 0.5 mA foot-shock US. Day 1: two CS–US pairings separated by 2 min. Day 2 recall: 10 CS presentations (no shock) in a modified context with 90–120 s intertrial intervals. Freezing was quantified with ANY-MAZE (threshold: ≥2 s immobility). Open field (40×40 cm) assessed locomotion; balance beam (1.0 cm wide, 100 cm long) assessed motor coordination (hindpaw footslips; trained over 3 days). Photometry: Time-correlated single photon counting measured GRABNE fluorescence via a 473 nm pulsed laser through implanted fibers. Photon counts were converted to dF/F using a 20 s sliding window baseline, then to z-scores; traces were baseline-normalized to the 10 s pre-tone period. Area under the curve (AUC) during defined windows was computed (MATLAB TRAPZ). Foot-shock series consisted of ten 0.5 mA, 1 s shocks with 60 s ISI. Chemogenetics: CNO (0.3 µl, 300 nM) or saline was infused bilaterally into CB vermis at 0.1 µl/min via internal cannulas extending 0.5 mm beyond the guide; the injector remained in place for ≥10 min to allow diffusion. Mice were tested 30 min after removal. CNO was administered before fear conditioning or motor tests. Optogenetics: A 532 nm laser (8–10 mW at fiber tip) drove Arch3-mediated inhibition. Light was delivered bilaterally in CB vermis only during CS–US pairings on Day 1 for acquisition studies, or only during recall CS presentations on Day 2 for expression studies; light was off during intertrial intervals. Statistics: Data are mean ± SEM. Two-tailed paired/unpaired t-tests for two-group comparisons; one-way or two-way ANOVA with Tukey or Sidak post hoc tests for multiple comparisons (GraphPad Prism 8). Analyses were conducted blind to condition.

• Anatomical and functional evidence of LC-NE projections in cerebellum: TH-Cre-dependent GFP expression in LC neurons produced dense GFP-positive axons in cerebellar cortex co-localizing with TH immunolabel. FFN270 accumulated in punctate varicosities in cerebellar cortex and was blocked by the NET inhibitor nomifensine, confirming NE terminals.
• Foot-shock evokes a biphasic NE signal in cerebellum measured with GRABNE: A 1 s, 0.5 mA foot-shock produced an early peak at ~0.8 s (±0.76) followed by a trough at ~3.2 s (±0.29). AUC 0–1.4 s increased (paired t-test no shock vs shock p=0.0443; n=8), and AUC 1.4–5 s decreased (paired t-test p=0.0001; n=8). Systemic amphetamine (10 mg/kg), a NET substrate/inhibitor, had no effect on the peak (ANOVA F(1.657,11.6)=1.653, p=0.2327) but attenuated the trough by 52.9% (±12) (ANOVA F(1.962,13.73)=30.38, p<0.0001; Tukey shock vs shock+amphetamine p=0.0171; n=8), consistent with NET-mediated clearance contributing to the trough.
• Conditioning transforms a neutral tone into a cue that elicits NE release in cerebellum: During Day 1 training, tones did not change GRABNE fluorescence (AUC 0–30 s vs intertrial: ANOVA F(1,442)=7.212, p=0.662; n=6), while foot-shocks induced significant decreases during 29–34 s windows (ANOVA F(1,574)=7.869, p=0.0001). On Day 2 recall in a novel context, tones alone increased GRABNE fluorescence relative to intertrial periods (paired t-test p=0.0117; n=6) and were 4.8-fold higher than tones on Day 1 (paired t-test p=0.0258; n=6). GRABNE activity during recall negatively correlated with movement.
• LC-NE cerebellar projections are necessary for fear memory formation but not motor function: Chemogenetic inhibition of LC axons in the cerebellum during conditioning (hM4Di activated by CNO locally in CB) reduced freezing to the recall tone (two-way ANOVA: treatment F(1,18)=8.376, p=0.0097; viral expression F(1,18)=4.632, p=0.0452; interaction F(1,18)=4.589, p=0.0461; hM4Di/CNO n=7, hM4Di/vehicle n=6, YFP/CNO n=5, YFP/vehicle n=4). Open field distance and speed and beam-walk footslips were unaffected across groups, indicating no motor deficits.
• Temporally precise optogenetic inhibition during CS–US pairing impairs fear learning: Arch3-mediated inhibition of LC axons in cerebellum only during training pairings suppressed freezing at recall (one-way ANOVA F(2,10)=16.87, p=0.0006; Tukey: Arch3 laser-on vs YFP laser-on p=0.0013; Arch3 laser-on vs Arch3 laser-off p=0.002; Arch3 laser-off vs YFP laser-on p=0.62; Arch3-on n=5, Arch3-off n=5, YFP-on n=3). Optogenetic inhibition during recall tones also reduced freezing, indicating NE release is required for expression of conditioned freezing. Overall, conditioning induces a learned increase in cue-evoked NE release in cerebellum that correlates with freezing and is necessary for fear memory formation and expression, independent of motor performance.

The study addresses whether aversive learning involves plastic changes in norepinephrine release within the cerebellum. Using GRABNE photometry, the authors show that an initially neutral auditory cue acquires the capacity to elevate cerebellar NE release after CS–US pairing, and the magnitude of cue-evoked NE correlates with freezing. Chemogenetic and optogenetic inhibition of LC-NE axons in the cerebellum during acquisition or recall reduces conditioned freezing without affecting locomotion or motor coordination, indicating the LC–cerebellum pathway is necessary for fear memory formation and expression rather than motor output per se. Mechanistically, NE likely promotes LTP at parallel fiber–Purkinje cell synapses via β-adrenergic receptor activation and cAMP-PKA signaling that enhances AMPA receptor phosphorylation and insertion, strengthening PF–PC efficacy. Enhanced PF–PC output may refine Purkinje cell signaling to deep cerebellar nuclei, particularly the fastigial nucleus, which projects to the ventrolateral periaqueductal gray to engage downstream freezing circuits in the brainstem and spinal cord. The data align with prior evidence that cerebellar vermis outputs modulate freezing-related neurons and that LC projections to amygdala and hippocampus contribute to cued and contextual fear, respectively. The lack of motor deficits with LC–CB inhibition suggests that phasic NE signaling during salient CS–US events, rather than tonic NE levels relevant for locomotion and balance, underpins the observed effects on fear learning. Clinically, the findings resonate with reports of increased cerebellar vermis activation and connectivity in PTSD and suggest that β-adrenergic blockade may mitigate excessive fear responses by modulating cerebellar NE contributions.

This work identifies a locus coeruleus to cerebellum norepinephrine pathway as a key substrate for aversive learning. Foot-shock evokes cerebellar NE release, and conditioning endows a neutral cue with the ability to trigger NE release in cerebellum. Inhibiting LC-NE axons in cerebellum during CS–US pairing or during recall suppresses conditioned freezing without motor impairments, demonstrating necessity for fear memory formation and expression. The results support a model in which NE-dependent enhancement of PF–PC synaptic efficacy refines cerebellar output to brainstem fear circuits. Future research should define the precise timing and dynamics of LC activity and NE signaling required for learning within the CS–US window, dissect receptor and cell-type specific mechanisms within cerebellar microcircuits, map contributions of vermis versus lateral hemispheres, and integrate cerebellar NE signaling with amygdalar and hippocampal LC targets during different phases of fear learning and memory.

• Temporal precision of LC activity and NE signaling necessary for learning within the 30 s CS window is unresolved; the exact duration and timing of enhanced signaling remain to be determined.
• Local CNO delivery may not fully suppress NE signaling in distal cerebellar regions; regional specificity could spare circuits involved in locomotion/balance.
• Opto/chemogenetic manipulations target LC axons broadly in CB vermis; contributions of lateral hemispheres and specific cell types are not isolated.
• Behavioral assessments focused on cued fear; contextual fear and generalization were not systematically tested here.
• Group sizes for some conditions (e.g., optogenetic controls) were modest, which may limit power for detecting subtle effects.
• Translation to humans is inferential; while clinical observations align (e.g., PTSD vermis activation), direct human mechanistic evidence is lacking.