Treadmill exercise modulates the medial prefrontal-amygdala neural circuit to improve the resilience against chronic restraint stress

Physical exercise reduces the risk of anxiety disorders in humans and rodent models. Prior work has focused on mechanisms such as enhanced hippocampal neurogenesis and reduced oxidative stress or neuroinflammation, but the contribution of specific neural circuits remains unclear. Brain imaging studies suggest exercise can alter connectivity among anxiety-related regions, including effects on amygdala activity. The basolateral amygdala (BLA) integrates inputs from multiple regions and exhibits CRS-induced synaptic and membrane changes. The medial prefrontal cortex (mPFC) provides dense input to BLA, and the mPFC–BLA pathway is activated by CRS. Given reports that exercise alters mPFC–amygdala connectivity, the study tests whether exercise enhances stress resilience by modulating the mPFC–BLA circuit.

The authors summarize evidence that exercise reduces anxiety and depression and is linked to increased hippocampal neurogenesis and improvements in brain homeostasis (e.g., reduced neuroinflammation and oxidative stress). Human neuroimaging shows altered connectivity with exercise involving parietal, orbitofrontal, cingulate, striatal regions, and modulation of amygdala functional connectivity by habitual physical activity. The BLA is a hub for anxiety behaviors and undergoes CRS-induced synaptic plasticity and conductance changes. Prior studies identified activation of the mPFC–BLA pathway in CRS and exercise-associated changes in mPFC–amygdala connectivity, motivating investigation of this circuit as a substrate for exercise-conferred stress resilience.

• Animals: Male C57BL/6J mice (4 weeks old) housed under standard conditions. All procedures approved by Jinan University ethics committee.
• CRS model: 3 h/day restraint in a pastry bag (20:00–23:00) for 14 consecutive days.
• Exercise paradigm: Treadmill running 1 h/day (10:00–11:00) at 10 m/min for 14 consecutive days; 10 min habituation at 5 m/min the day before training. Controls placed on a stationary treadmill for 1 h.
• Behavioral assays: Open field test (5 min; center time and total distance) and elevated plus-maze (5 min; open-arm time and total distance), video-tracked and analyzed with EthoVision.
• Viral tracing and targeting: • Retrograde labeling of BLA-projecting mPFC neurons: AAV-Retro-EGFP into BLA (5.4×10^12 GC/ml; 0.12 μL/injection; AP −1.0, ML ±3.2, DV −4.8 mm). • Chemogenetic manipulation of BLA-projecting mPFC neurons: rAAV2/2-Retro-Cre into BLA (3×10^12 GC/ml; 0.1 μL/injection) and Cre-dependent AAV2/9-DIO-hM4Di-mCherry, AAV2/9-DIO-hM3Dq-mCherry, or AAV2/9-DIO-mCherry into mPFC (3.5×10^12 GC/ml; 0.15 μL/injection; AP 2.0, ML ±0.2, DV −2.0 mm). CNO (1 mg/kg, i.p.) or saline administered daily during CRS and/or exercise protocols. • Postsynaptic labeling of BLA neurons innervated by mPFC: AAV2/1-Cre into mPFC (1.5×10^13 GC/ml; 0.1 μL/injection) plus AAV2/9-DIO-EGFP into BLA (3.5×10^12 GC/ml; 0.12 μL/injection). • Collateral and subtype mapping: AAV-Retro-EGFP into mPFC/BLA and cholera toxin subunit B (CTB) into contralateral cortex (IT labeling) or ipsilateral periaqueductal gray (PAG) (PT labeling) to classify BLA-projecting mPFC neurons as intratelencephalic (IT) or pyramidal tract (PT). Additional labeling of mPFC–VTA projections to assess specificity.
• Electrophysiology: Ex vivo whole-cell patch clamp on acute coronal slices (250 μm) of mPFC and BLA in oxygenated ACSF. Neuronal excitability measured with 0.8 s depolarizing current steps (−90 to 300 pA, 30 pA increments) at −70 mV. mEPSCs recorded with TTX (1 μM) and bicuculline (20 μM); mIPSCs with TTX (1 μM), NBQX (20 μM), D-AP5 (50 μM). Standard internal solutions used; data sampled at 10 kHz, filtered at 4 kHz; Rs 10–20 MΩ; exclusion if Rs changed >20%.
• Immunostaining: CaMKIIα immunofluorescence to identify glutamatergic identity of BLA-projecting mPFC neurons; confocal imaging and quantification.
• Statistics: Normality testing; parametric tests when appropriate. Two-sample t-tests, one-way and two-way ANOVAs with post hoc tests (Tukey, Bonferroni), Kruskal–Wallis and Kolmogorov–Smirnov for nonparametric analyses. Significance at P<0.05. Blinded analysis. Sample sizes in figure legends; no technical replicates.

• Behavior: CRS induced anxiety-like behavior without affecting locomotion; exercise prevented anxiety-like behavior. • Open field total distance: F(3,32)=2.183, P=0.1093 (ns). Center duration: F(3,32)=9.739, P=0.0001; CRS decreased center time, exercise restored it. • Elevated plus-maze total distance: F(3,32)=0.4678, P=0.7068 (ns). Open-arm time: F(3,32)=7.947, P=0.0004; CRS decreased, exercise restored.
• mPFC BLA-projecting neurons: CRS increased excitability; exercise suppressed it to near control. • Spike number across groups (BLA-projecting): F(3,710)=50.84, P<0.0001; representative traces show higher firing in CRS, reduced with exercise. • E/I balance shifted toward excitation under CRS; exercise reversed.
  • mEPSC amplitude: F(3,44)=9.794, P<0.0001; frequency: F(3,44)=5.958, P=0.0017 (CRS↑; exercise↓).
  • mIPSC amplitude: F(3,44)=6.429, P=0.0010; frequency: F(3,44)=6.388, P=0.0011 (CRS↓; exercise↑).
• mPFC non-BLA projecting PrL neurons: Opposite pattern; CRS reduced excitability, exercise increased it. • Spike number: F(3,520)=50.97, P<0.0001. • mEPSC amplitude: F(3,54)=5.125, P=0.0034 (exercise restored); frequency: F(3,54)=0.03437, P=0.9914 (ns). • mIPSC amplitude: F(3,54)=5.330, P=0.0027; frequency: F(3,54)=0.2435, P=0.8656 (ns). • Exercise in unstressed mice did not alter excitability or E/I balance in either group.
• Chemogenetics: Inhibiting BLA-projecting mPFC neurons mimicked exercise; activating them blocked exercise benefit. • Inhibition (hM4Di+CNO): Reduced firing (F(3,400)=21.04, P<0.0001); improved anxiety-like behavior without affecting locomotion: open field distance F(3,28)=1.852, P=0.1067; center time F(3,28)=8.140, P=0.0005; EPM distance F(3,28)=0.4935, P=0.6897; open-arm time F(3,28)=17.82, P<0.0001. • Activation in exercised mice (hM3Dq+CNO): Increased firing (F(4,520)=34.39, P<0.0001); reversed exercise-induced anxiolysis without affecting locomotion: open field distance F(4,35)=0.2056, P=0.9336; center time F(4,35)=9.986, P<0.0008; EPM distance F(4,35)=0.05699, P=0.9937; open-arm time F(4,35)=15.03, P<0.0001.
• Postsynaptic BLA neurons innervated by mPFC: CRS shifted E/I toward excitation; exercise normalized. • mEPSC amplitude F(3,44)=6.758, P=0.0008; frequency F(3,44)=6.786, P=0.0007. • mIPSC amplitude F(3,44)=7.274, P=0.0005; frequency F(3,44)=7.091, P=0.0005.
• Circuit identity and specificity: • BLA-projecting mPFC neurons were primarily glutamatergic (CaMKIIα+; t(10)=42.18, P<0.0001) and predominantly intratelencephalic (>70% IT; t(10)=7.201, P<0.0001) with a minor PT fraction (t(10)=14.76, P<0.0001). Limited bidirectional labeling (<20% co-labeled) supports specificity. Collateral projections to other regions were minimal. mPFC–VTA neurons showed unchanged excitability across groups (interaction F(27,620)=0.6698, P=0.8978). Overall, treadmill exercise selectively suppressed the hyperexcited mPFC–BLA pathway induced by CRS, and this suppression was necessary and sufficient for exercise-conferred anxiolysis.

The study addresses whether physical exercise improves stress resilience by modulating a defined anxiety-related circuit. CRS increased excitability and shifted the E/I balance toward excitation in BLA-projecting mPFC neurons and their downstream BLA targets. Fourteen days of treadmill exercise reversed these changes without affecting general locomotion. Chemogenetic inhibition of BLA-projecting mPFC neurons mimicked exercise-induced anxiolysis, while their activation in exercised mice reinstated anxiety-like behavior, demonstrating necessity and sufficiency of mPFC–BLA suppression for the behavioral benefit. The specificity of modulation—suppression of BLA-projecting cells but potentiation of non-BLA-projecting PrL neurons—highlights circuit-level selectivity rather than a global mPFC effect. Given BLA’s central role in anxiety and its multiple inputs, other pathways may also contribute. Potential mechanisms include cellular changes in mPFC and BLA during exercise and interactions with peripheral exercise-induced factors (e.g., SAM and clusterin) that can modulate neural circuits. These findings provide a circuitry framework linking exercise to improved anxiety resilience.

Fourteen days of treadmill exercise attenuates CRS-induced hyperexcitation in the mPFC–BLA pathway, normalizing E/I balance in both presynaptic mPFC and postsynaptic BLA neurons and alleviating anxiety-like behaviors. Chemogenetic manipulations establish that suppression of BLA-projecting mPFC neurons is both necessary and sufficient for exercise’s anxiolytic effects. The work identifies a specific cortico-amygdala circuit as a key substrate for exercise-conferred stress resilience and suggests that targeted modulation of this pathway could inform therapeutic strategies for anxiety disorders. Future studies should elucidate the molecular and synaptic mechanisms by which exercise reshapes this circuit, examine contributions of other BLA-afferent pathways, and explore interactions with peripheral exercise-induced factors.

The study does not resolve the precise molecular and synaptic mechanisms by which exercise suppresses the mPFC–BLA pathway. While focused on BLA-projecting mPFC neurons, other inputs to BLA (e.g., thalamic or hippocampal) may also contribute to anxiolysis under exercise and were not comprehensively tested. The findings are based on a CRS mouse model and ex vivo electrophysiology; generalizability to other stress paradigms, sexes, species, or in vivo circuit dynamics remains to be established.