Exercise enhances motor skill learning by neurotransmitter switching in the adult midbrain

Motor skill learning is regulated by a distributed neural network including cortex, thalamus, basal ganglia, brainstem, cerebellum and spinal cord. Both neuronal and glial plasticity are essential; impairments produce motor deficits. Aerobic exercise enhances acquisition of new motor skills and serves as therapy in motor disorders, but the mechanism is not well understood. Running is ethological in mice and induces plasticity in multiple brain regions. Neurotransmitter switching—activity-dependent changes in neuronal transmitter identity—can alter behavior. The authors hypothesized that chronic running induces neurotransmitter switching in adult midbrain circuits important for motor skill learning. They tested whether one week of wheel running enhances motor learning (rotarod, balance beam) and whether this is temporally correlated with transmitter switching of cholinergic neurons in the caudal pedunculopontine nucleus (cPPN) from acetylcholine (ACh) to GABA, with projections to SN, VTA, and VL-VM, and whether manipulating this switch affects learning.

The paper situates motor learning within a large motor network (basal ganglia, cerebellum, cortex; refs 1–4). Prior work shows aerobic exercise boosts motor skill acquisition and is therapeutic in disorders like Parkinson’s disease and developmental coordination disorder (refs 5–8, 11, 15). Running induces neuroplastic changes across brain regions (refs 4, 10, 11). Neurotransmitter switching is a recognized form of plasticity in developing and adult brain, often activity-dependent and behaviorally relevant (refs 12, 18–20, 44). The pedunculopontine nucleus (PPN) has cholinergic, glutamatergic and GABAergic neurons, participates in gait, locomotion, attention and arousal, and interfaces with basal ganglia and thalamus (refs 23, 25, 40–41, 51–53). Prior studies have shown co-release or co-expression of ACh with GABA or glutamate in related regions and species (refs 33–35). These set the stage for testing whether exercise drives transmitter switching in PPN to affect motor learning.

- Animals: 8–12-week-old male mice (C57BL/6J; ChAT-IRES-Cre; PV-IRES-Cre). Standard housing on 12h:12h reverse light cycle; food/water ad libitum. Approvals per UCSD/Scripps IACUC. - Wheel running paradigm: Mice single-housed with running wheels (FastTrac or Med Associates digital). Voluntary running for 1 week; controls had wheelbases without wheels. Running behavior recorded (video, digital wheels). Episode durations/time on wheel scored; running speed calculated. - Behavioral assays: Motor skill learning assessed with accelerating rotarod (5→80 rpm over 6 min). Training: 9 trials day 1; tests: 3 trials next day; retests after 1, 2 or 4 weeks rest. Balance beam: 1 m beams (square 12 mm, 6 mm; rods 6 mm, 4 mm), three trials each on training day, tested/retested similarly. Performance metrics: speed at fall (rpm) and time to cross beam. Analyses included learning curve slopes and performance on test days. - Activity mapping: c-Fos immunostaining to assess activation in multiple regions (DG, M1, M2, STN, GPe, STR, SNc, SNr, PPN, LDT) after 1-week running. Double labeling NeuN/c-Fos; quantification per mm². - Identification of switching in cPPN: Stereological DAB counts of ChAT+ neurons; in situ hybridization for GAD1+ neurons; assessment in rostral vs caudal PPN (rPPN vs cPPN). RNAscope multiplex FISH for chat and gad1 transcripts combined with nNOS immunostaining (nNOS as biomarker for cholinergic PPN). Quantification of transcript puncta per neuron and categorical classification (classic cholinergic; co-expressing; neither; switched). - Lineage tagging of cholinergic neurons: Cre-dependent AAV-DIO-mRuby2 injected bilaterally into cPPN of ChAT-Cre mice to permanently label cholinergic neurons. After 4-week expression and 1-week run, triple IHC for mRuby2, ChAT, GABA; classification of mRuby2+ neurons into categories above. - Projections and synaptic transporter changes: Anterograde tracing via AAV-DIO-mRuby2; retrograde tracing with red/green retrobeads from SN, VTA, VL-VM to identify projecting cPPN neurons. Triple labeling of retrobeads, chat mRNA (RNAscope), and nNOS to quantify chat transcript levels in projecting neurons. Presynaptic labeling: AAV8-FLEX-tdTomato-T2A-Synaptophysin-EGFP injected into cPPN of ChAT-Cre mice; in SN, co-immunostaining for EGFP with VAChT or VGAT to quantify putative cholinergic terminal colocalization (percent and M1 coefficient). - Necessity tests for switching: (1) Override loss of ACh with AAV-hSyn-DIO-ChAT-flag-P2A-mRuby2 in cPPN of ChAT-Cre mice; verify increased ChAT expression and maintained ChAT+ counts after running; assess behavior (rotarod/beam) immediately and after 1-week rest. (2) Prevent gain of GABA with AAV8-CAG-DIO-shRNAmir-mGAD1-EGFP (shGAD1) vs scramble (shScr) in cPPN; confirm reduced GAD1+ counts and blocked GABA gain in ChAT+ neurons; assess behavior. (3) Suppress cholinergic cPPN activity with AAV-DIO-Kir2.1 vs control to test activity dependence of switching. - Time course/reversal: After 1-week running followed by 1-week rest without motor training, stereological counts of ChAT+ and GAD1+ neurons to assess reversal; align with behavioral retention. - Histology: Perfusion after last running episode; IHC for ChAT, nNOS, c-Fos, PV, VAChT, VGAT, NeuN, GABA, GFP variants, DCX, Ki67; DAB and fluorescence protocols detailed. RNAscope protocols for chat (Cat#408731-C2) and gad1 (Cat#400951). TUNEL and BrdU assays to assess apoptosis and neurogenesis. - Viral constructs/injections: Stereotaxic bilateral injections targeting cPPN (AP −4.80 mm, ML ±1.25 mm, DV −3.25 mm). Volumes/flow rates: 300 nL at 100 nL/min; vectors from Salk, Stanford, Vector Biolabs; titers 7.1×10^12–2.2×10^13 vg/mL. Retrobeads coordinates: SN, VTA, VL-VM as specified. - Statistics: Welch’s t-test, Mann–Whitney U, paired t-test, ANOVA with Tukey, Kruskal–Wallis with Dunn’s correction; Pearson correlations. Data as mean ± SEM; n detailed per figure; experiments replicated 2–4+ times depending on assay.

- One week of voluntary wheel running enhanced motor skill acquisition in adult mice on accelerating rotarod and narrow balance beams. Runner mice showed significantly higher speeds at fall and faster beam crossing during training and tests compared with controls (e.g., Fig. 1c–f; n≈19–20/group; p<0.05 to p<0.001). Enhancement persisted at least 2 weeks but diminished by 4 weeks. Immediate training after running yielded enhancement; delaying training by 1 week of rest abolished enhancement (Fig. 1h–j), indicating a time dependence of the enabling effect on learning. - Running increased c-Fos expression in the pedunculopontine nucleus (PPN), particularly in its caudal subregion (cPPN) within cholinergic neurons, consistent with activity-dependent plasticity (Fig. 2b–f). - Neurotransmitter switching in cPPN: After 1-week running, stereological DAB counts showed a significant decrease in ChAT+ neuron number and an increase in GAD1+ neuron number specifically in cPPN, with no apoptosis or neurogenesis detected and no change in glutamatergic (vGluT2) counts (Fig. 2g–k; Supplementary Fig. 3). The switch reversed after 1-week rest (Fig. 2k), temporally correlating with loss of enhanced learning when training was delayed. - Lineage tagging demonstrated within-neuron switching: Among mRuby2-tagged cholinergic cPPN neurons (ChAT-Cre), control mice were 65% ChAT-only, 29% co-expressing ChAT+GABA, 2% GABA-only, 4% neither (n=844). Runners shifted to 24% ChAT-only, 31% co-expressing, 33% GABA-only, 12% neither (n=738) (Fig. 3b–c), indicating loss of ACh and gain of GABA in a subset of previously cholinergic neurons. The increase in "neither" suggests loss of ChAT precedes gain of GABA. - Transcript-level changes mirrored protein-level switching: RNAscope in nNOS+ cPPN neurons showed significant decrease in chat transcript puncta and increase in gad1 puncta per neuron in runners (Fig. 4d–g). Categorization of nNOS+ cells paralleled IHC: controls 45% classic, 44% co-expressing, 6% neither, 5% gad1-only; runners 18%, 43%, 12%, 27% respectively (Fig. 4e). Total nNOS+ neuron counts were unchanged (Fig. 4c), indicating phenotype change rather than cell loss/gain. - Projected targets of switching neurons: Anterograde and retrograde tracing established cPPN cholinergic projections to SN, VTA, and VL-VM (Fig. 5a–c). In runners, projecting nNOS+ neurons to each target had reduced chat transcript puncta; low-puncta neurons (<8) increased particularly for SN projections (SN: 11%→42%; VTA: 8%→11%; VL-VM: 6%→19%) (Fig. 5e–f), suggesting prominent involvement of SN-targeting neurons. - Presynaptic transporter changes at SN terminals: In runners, putative cPPN cholinergic terminals (Synaptophysin-EGFP+) in SN showed decreased VAChT colocalization and increased VGAT colocalization (percent terminals and M1 coefficients), consistent with functional ACh→GABA switching at synapses (Fig. 5h–j). - Necessity of switching for behavioral enhancement: Overexpressing ChAT in cPPN cholinergic neurons (AAV-DIO-ChAT) prevented the running-induced decrease in ChAT+ counts and blocked enhancement of rotarod and beam learning both immediately after running and after 1-week rest (Fig. 6c–g). Similarly, knocking down GAD1 (AAV-DIO-shGAD1) prevented running-induced gain in GABA in ChAT+ neurons and blocked behavioral enhancement (Fig. 7b–f). Neither manipulation affected running behavior itself. Suppressing cholinergic cPPN neuron activity with Kir2.1 prevented switching (Supplementary Fig. 2g–h), indicating activity-dependence. - The degree of switching correlated with behavioral enhancement: Numbers of ChAT+ vs GAD1+ neurons were inversely correlated and directly related to improvement in motor learning, whereas total running amount did not correlate with acquisition (Supplementary Fig. 4).

The study identifies activity-dependent neurotransmitter switching—from cholinergic to GABAergic phenotype in cPPN neurons—as a mechanism enabling exercise-induced enhancement of motor skill learning. The switch occurs in adult brain after one week of voluntary running, involves transcript-level regulation of transmitter synthetic enzymes (chat down, gad1 up), extends to presynaptic terminals (reduced VAChT, increased VGAT), and targets key motor-learning nodes (SN, VTA, VL-VM). Blocking either component of the switch (preventing ChAT loss or GAD1 gain) abolishes the behavioral benefit without affecting the ability to run, demonstrating necessity for enhanced acquisition of demanding motor skills (rotarod, narrow beam). The reversal of the switch after rest, while learned performance persists, suggests the switch enables acquisition but is not required for maintenance, implying downstream plasticity consolidates the skill. Potential molecular mediators include activity-dependent transcription factor phosphorylation and microRNAs previously implicated in transmitter switching. Within basal ganglia-brainstem loops, converting excitatory cholinergic input from cPPN to inhibitory GABAergic input onto inhibitory SN neurons could modulate feedback to optimize motor coordination and learning, consistent with reduced c-Fos in SN. The findings suggest that transmitter identity plasticity can dynamically rewire adult motor circuits in response to sustained exercise, offering a mechanistic link to observed clinical benefits of training in movement disorders.

One week of sustained voluntary running enhances motor skill learning in adult mice by inducing an activity-dependent neurotransmitter switch in caudal PPN neurons from acetylcholine to GABA. Switching is observed at protein and transcript levels, extends to presynaptic terminals, and occurs in neurons projecting to SN, VTA, and VL-VM. Overriding either the loss of ChAT or the gain of GAD1 prevents the exercise-induced learning enhancement, establishing the switch as necessary for this benefit. These results reveal neurotransmitter switching as a key mechanism through which sustained exercise facilitates motor learning and propose it as a therapeutic target for movement disorders. Future work should define upstream molecular pathways governing the switch, determine sufficiency for enhancing learning, map additional target circuits, assess electrophysiological synaptic consequences, explore sex and age generalization, and evaluate translational strategies to modulate transmitter identity in disease contexts.

- Generalizability: Experiments used only adult male mice; sex and age differences were not assessed. - Behavioral scope: Enhancement shown for high-demand rotarod and narrow beams; broader motor and cognitive domains were not tested. - Mechanistic depth: While necessary elements (ChAT loss, GAD1 gain) were established, sufficiency of switching to enhance learning without running was not tested, and upstream molecular regulators were not identified. - Electrophysiology: No direct electrophysiological recordings demonstrated synaptic sign change or functional inhibition in targets; inference is based on transporter colocalization and transcript/protein markers. - Detection thresholds: Immunostaining and RNAscope quantification rely on method-specific thresholds; co-expression percentages differ from some literature, potentially affecting estimates of switching incidence. - Regional focus: Emphasis on cPPN; other PPN subregions and additional projections were not exhaustively characterized. - Time course: Detailed dynamics of switch onset and decay beyond 1-week windows were not mapped.