Introduction

The aggregation of α-synuclein is one of the major pathogenic factors in neurodegenerative disorders, including Parkinson’s disease (PD) and Alzheimer’s disease (AD)1,2. Current knowledge has identified synucleinopathy and neuronal loss in subcortical nuclei such as the substantia nigra compacta (SNc) in PD patients3. Such a model of midbrain dopaminergic circuit in PD pathogenesis, however, cannot fully explain motor syndromes. For example, limb rigidity and tremor were believed to be affected by the primary motor cortex (M1)4, which regulates voluntary movement via local excitatory-inhibitory circuits and top-down projections. In human PD patients, cortical regions such as the cingulate cortex5 also presented a higher abundance of α-synuclein oligomers. Besides PD, cortical α-synuclein deposits are also found in human Lewy body disease6 and AD patients7. Therefore, further understanding of cortical synucleinopathy is critical for illustrating the pathology of neurodegenerative diseases.

Different theories have been proposed to explore the molecular mechanisms of neuronal degeneration under PD. The emerging model of lipid peroxidation and ferroptosis8 has drawn research interest, as the potential role of ferroptosis in PD pathology has been reported9,10. In specific, potentiated ferroptosis has been found in both drug-induced11 and transgenic PD models with α-synuclein overexpression12. The induction of ferroptosis may occur with the activation of neuroinflammation13,14, contributing to neuronal death. The activation of microglial cells also resulted in the elevated release of chemokines release, further enhancing microglia migration15. As a consequence, microglia-mediated phagocytosis of synapse and neuronal soma disrupted neural networks16,17. These existing studies, however, mainly analyzed the ferroptosis in midbrain nuclei during PD pathogenesis18, with insufficient focus on cortical regions.

In the current work, we generated a mouse synucleinopathy model by viral-mediated overexpression of the human α-synuclein (SNCA) gene into the M1. The de novo production of α-synuclein aggregates resulted in deficits of complex motor learning, in association with hyperactivation of cortical pyramidal neurons (PNs). Such overexcitation was found to be caused by the downregulation of parvalbumin-interneurons (PV-INs). Further molecular studies showed that following SNCA overexpression, iron overload and disrupted ferroptosis probably caused cortical dysfunctions. In sum, our work revealed one previously unrecognized molecular pathway by which SNCA disrupts circuitry integrity and cortical function.

Results

SNCA overexpression in mouse M1 leads to cortical hyperexcitability and motor learning impairments

We first introduced SNCA-coding gene into the bilateral M1 of adult male mice using an adeno-associated virus (AAV) vector (Fig. 1a). Three weeks later, these mice presented significantly elevated levels of α-synuclein in M1 extracts (P < 0.001; Fig. 1b, c, Supplementary Fig. 1). In specific, phosphorylated α-synuclein proteins in neurons (NeuN+) can be found across different layers of M1 (Fig. 1d) as well as in axonal fibers from M1 to the dorsal striatum (DS) (Supplementary Fig. 2a). However, the postsynaptic neurons in DS did not exhibit transmission of protein aggregations, nor did other cortical regions (Supplementary Fig. 2a), thus suggesting M1-specificity of SNCA overexpression. In behavioral phenotyping, SNCA mice displayed normal anxiety levels or novel object recognition function (P > 0.05; Supplementary Fig. 2b–d). When examining the motor functions, SNCA mice had unchanged general motor ability (P > 0.05; Fig. 1e), whilst their abilities to acquire complex motor skills were disrupted. Of note, SNCA mice presented similar latency on the accelerating rotarod compared to control groups during the initial trial, whilst their performance did not improve with repeated training sessions (P < 0.0001; Fig. 1f). In a similar manner, SNCA mice had longer duration in accomplishing vertical climbing or horizontal beam walking task (P < 0.05; Fig. 1g, h). These data suggested that SNCA accumulation in the motor cortex resulted in dysregulated motor learning functions.

Fig. 1: Human α-synuclein resulted in motor learning deficits associated with cortical hyperexcitability.
figure 1

a Upper, the design of adeno-associated virus (AAV)- human synapsin (hSyn)-human α-synuclein (SNCA)-mCherry viral vector. Lower, Schematic diagram of generating mouse synucleinopathy model by expressing SNCA gene into bilateral primary motor cortex (M1) of adult male mice under the help of an AAV vector carrying hSyn promoter. Three weeks later, behavioral test batteries were performed. b Representative blotting bands of total synuclein and phosphorylated synuclein in M1 extracts. c Quantification for the relative expression of total (left) and phosphorylated synuclein (right). Two-sample unpaired t-test, t(8) = 6.967, P = 0.0001 for synuclein; t(8) = 3.907, P = 0.005 for p-synuclein. N = 5 mice in each group. d Upper panels: coronal section of mouse M1 after SNCA overexpression. AAV-SNCA was mostly expressed within neurons (NeuN+). Lower panels: cortical expression of synuclein and phosphorylated synuclein were observed. Scale bar, 100 μm or 20 μm (enlarged panels in lower right). e Total distance traveled in the open field arena was unaffected after SNCA overexpression. Two-sample unpaired t-test, t(14) = 0.3807, P = 0.7091. f SNCA-overexpressing mice had remarkably impaired motor performance on the accelerating rotarod. Two-way analysis of variance (ANOVA) with respect to group factor, F(1126) = 55.81, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. g SNCA group spent longer time to accomplish the vertical climbing task. Two-way ANOVA with respect to group factor, F(142) = 4.553, P = 0.0387. Sidak’s multiple comparison test of post hoc comparison at each time point. h Worsening performance on the walking beam in mutant mice. F(139) = 13.55, P = 0.0007. Sidak’s multiple comparison test of post hoc comparison at each time point. N = 8 mice in each group in (e–h). i Patch-clamp recording traces of layer 5 pyramidal neurons (L5 PNs) showed potentiated miniature excitatory postsynaptic currents (mEPSCs) in SNCA group. j SNCA group had higher amplitude of mEPSCs. Non-parametric Mann–Whitney test U = 6, P = 0.018. k Increased frequency of mEPSCs was observed under synucleinopathy. Non-parametric Mann–Whitney test U = 6, P = 0.017. l Elevated number of action potentials (AP) in SNCA group under fixed injection currents. Two-way ANOVA with respect to group factor, F(1104) = 4.530, P = 0.0357. m Higher potentials of afterhyperpolarization (AHP). Two-sample unpaired t-test, t(13) = 3.817, P = 0.0021. n Higher rheobase values in SNCA group. Two-sample unpaired t-test, t(13) = 3.026, P = 0.0097. o Decreased resting membrane potential (RMP) in SNCA group. Two-sample unpaired t-test, t(13) = 4.181, P = 0.0011. N = 8 and 7 neurons from 3 animals in mCherry and SNCA group, respectively, in (j–o). ns, no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. All data were presented as mean ± SEM.

Full size image

Cortical neuronal activity is closely associated with motor learning functions, and hyperexcitability of the cortex has been reported in neurodegenerative diseases19,20. To investigate the cortical neural network homeostasis, we employed ex vivo whole-cell patch-clamp recording on acute M1 slices. SNCA mice had potentiated miniature excitatory postsynaptic currents (mEPSCs; Fig. 1i) of layer 5 pyramidal neurons (L5 PNs). In specific, both the amplitude and frequency of mEPSCs were elevated (P < 0.05; Fig. 1j, k). Such observations of higher synaptic transmission parallelled with the potentiated membrane excitability in SNCA group, including more frequent action potential spikes (P < 0.05; Fig. 1l), plus higher values of afterhyperpolarization (AHP), Rheobase and resting membrane potential (RMP) (P < 0.01, Fig. 1m–o).

The disruption of synaptic functions indicated the dysregulation of the cortico-striatal pathway, which was closely associated with complex motor skills21. To further support this hypothesis, we performed in vivo calcium imaging of M1 by co-expressing genetically encoded calcium indicator GCaMP6s along with SNCA (Fig. 2a). Three weeks later, the calcium transient of L5 PNs was recorded on awake, head-fixed mice (Fig. 2b). SNCA mice displayed higher total integrated calcium activities (P < 0.001; Fig. 2c), which can be attributed to the higher frequency of calcium spikes (P < 0.001; Fig. 2d). To investigate the downstream projecting region of M1, we utilized the in vivo fiber photometry to record the calcium activity of cortical PN fibers within the dorsal striatum (DS) while the mouse was trained on the rotarod (Fig. 2e, f). A major calcium peak occurred when the animal fell down from the rod (Fig. 2g). The strength of this calcium episode, however, was remarkably increased in the SNCA group (P < 0.05; Fig. 2g–i). These results paralleled the hyperexcitability of cortical PNs (Fig. 2b–d) and suggested that cortical overexpression of SNCA resulted in motor learning deficits, which was correlated with hyperexcitation of the cortico-striatal pathway.

Fig. 2: Dysregulated cortico-striatal pathway under SNCA.
figure 2

a Illustrations for the in vivo two-photon microscopy assay, in which one single imaging session was performed at 3 weeks after co-expressing GCaMP6s and SNCA into M1. b Sampled field of views (FOVs) from each group, with sample traces of calcium transients, were listed for each color-coded neuron. The calcium signal was calculated as ΔF/F0. Scale bar, 50 μm. c SNCA-overexpressed mice had remarkably higher total calcium transients. Non-parametric Kolmogorov–Smirnov test, D = 0.3749, P < 0.001. d Model mice showed higher frequency of calcium spikes. D = 0.3503, P < 0.001. n = 82 and 196 cells from 4 mice of mCherry and SNCA group, respectively, in (c, d). e Experimental design of fiber photometry recording of PN-DS terminus. At 14 days after infecting M1 with GCaMP6f, an optic fiber was implanted into DS. One week later, the calcium activity within DS was recorded when the mouse was trained on the accelerating rotarod. f Fiber implantation and GCaMP6s infection sites within DS. Scale bar, 100 μm. g Heatmaps for the normalized calcium activities (in ΔF/F0) during the motor task. A major peak occurred when the mouse fell down from the rod (t = 0 s). h Averaged calcium values from 20 trials. i Quantification of peak calcium values found potentiated signal strength in SNCA group. Two-sample unpaired t-test, t(13) = 2.640, P = 0.020. N = 9 and 6 mice from mCherry and SNCA groups, respectively. *P < 0.05, ***P < 0.001. All data were presented as mean ± SEM.

Full size image

Disrupted cortical excitatory-inhibitory balance underlies SNCA-induced motor learning deficits

The excitatory-inhibitory (E/I) balance of the cortical neural network is critical for maintaining normal functions. We thus tested the role of cortical E/I balance in motor learning deficits in SNCA-overexpressed mice. The first piece of evidence came from expressing the designer receptors exclusively activated by designer drugs (DREADD) hM4Di along with AAV-SNCA into mouse M1 (Supplementary Fig. 3a). Chronic infusion of specific ligand clozapine-N-oxide (CNO) remarkably improved motor learning functions (P < 0.01; Supplementary Fig. 3b–d), suggesting the participation of cortical hyperexcitation underlying motor learning deficits of SNCA model. To dissect the circuitry mechanism, local PV-INs were further studied as they innervate cortical PNs to exert indispensable functions in motor learning22. Via immunofluorescent staining, we identified remarkably decreased expression of PV-INs across different M1 layers (P < 0.001; Fig. 3a, b), plus the loss of perineuronal nets (PNNs) which coat PV-INs using wisteria floribunda agglutinin (WFA) as the marker (P < 0.001; Fig. 3c). In addition, we found that only PV-INs showed decreased density under SNCA, but not for vasoactive intestinal peptide (VIP) or somatostatin (SST) interneurons (Supplementary Fig. 4), supporting the subtype-specific effect. To recapitulate such cellular pathology, Cre-dependent Caspase 3 was expressed into M1 of PV-IRES-Cre mice (Supplementary Fig. 5a), leading to decreased density of PV-INs (P < 0.001; Supplementary Fig. 5b, c) and motor learning deficits (P < 0.05; Supplementary Fig. 5d–f) as those in SNCA-overexpressed mice. These data supported the correlation between motor dysfunction and PV-IN loss under synuclein overexpression.

Fig. 3: Parvalbumin-interneurons mediate SNCA-associated motor learning deficits.
figure 3

a Coronal sections of mouse M1 for SNCA expression and parvalbumin-interneurons (PV-INs). Scale bar, 100 μm. b SNCA-overexpressed mice had remarkably decreased density of PV-INs. Two-sample unpaired t-test, t(10) = 5.491, P = 0.0003. N = 6 mice in each group. c Fluorescent staining for the wisteria floribunda agglutinin (WFA) as the marker of perineuronal nets around PV-INs. Scale bar, 100 μm. d Experimental design for in vivo calcium recording of PV-INs via Cre-mediated GCaMP6s expression on PV-Cre mice. e Sample traces of calcium transients of PV-INs from both groups. f Comparison of total integrated calcium signal strength showed decreased activities in the SNCA group. Non-parametric Kolmogorov–Smirnov test, D = 0.2557, P = 0.0009. g The frequency of calcium transients was also depressed in the model group. Non-parametric Kolmogorov–Smirnov test, D = 0.3480, P < 0.0001. n = 134 and 105 neurons from 4 mice of mCherry and SNCA groups, respectively, in (f, g). h Illustration for experimental procedures of rescuing assay, in which excitatory hM3Dq under Cre-dependent expression was introduced into PV-IRES-Cre mice, followed by 2-week CNO administration and motor learning phenotyping. i Representative immunofluorescent staining for cFos activity. Scale bar, 100 μm. j cFos activity was reduced by CNO treatment. Two-sample unpaired t-test, t(6) = 12.52, P < 0.0001. N = 4 mice in each group. k Chemogenetic activation of cortical PV-INs improved motor learning abilities on the rotarod. Two-way ANOVA with respect to group factor, F(1126) = 43.38, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. l PV-INs excitation conferred better performance on vertical pole climbing. Two-way ANOVA with respect to group factor, F(142) = 21.66, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. m The reaction of PV-INs also enhanced the acquisition of beam walking abilities. F(142) = 19.78, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. N = 7 and 9 mice in saline and CNO group, respectively, in (k–m). ***P < 0.001. All data were presented as mean ± SEM.

Full size image

Subsequently, the calcium activity of PV-INs was recorded by in vivo 2-photon microscopy (Fig. 3d). SNCA mice displayed a decreased frequency of calcium transients of PV-INs, resulting in lower total calcium activities (Fig. 3e–g). The depressed inhibitory transmission might contribute to the overexcitation of L5 PNs, which receive GABAergic innervations from PV-INs. To provide the direct causal relationship, we then used a chemogenetic approach to selectively activate cortical PV-INs by expressing excitatory receptor hM3Dq into PV-INs along with AAV-SNCA (Fig. 3h). Chronic CNO infusion significantly elevated cortical neuronal activities (P < 0.001; Fig. 3i, j). Furthermore, behavioral phenotyping showed remarkable improvements in the accelerating rotarod test (P < 0.0001; Fig. 3k), as well as the vertical pole climbing (P < 0.0001; Fig. 3l) or horizontal beam walking task (P < 0.0001; Fig. 3m). Taken together, α-synuclein impaired cortical E/I balance via suppressing PV-INs, leading to motor learning deficits.

Microglia activation impairs cortical circuits to disrupt motor learning

Under SNCA overexpression, microglia is the primary cell type for extracellular protein aggregates clearance23, during which both morphological and molecular features of microglia are prominently altered. We thus used ionized calcium-binding adapter molecule 1 (Iba1) as the cellular marker to study the morphology of cortical microglia (Fig. 4a). In SNCA-overexpressing mice, microglia presented elevated cell density and phagocytosis toward SNCA-expressing cells (P = 0.011; Fig. 4a, b), in association with higher Iba1 and CD68 expressions (P = 0.037; Fig. 4c, d, Supplementary Fig. 6). The morphometric study of individual microglial cells further revealed enlarged soma size (P < 0.001; Fig. 4e, f) and highly branched processes (P < 0.001; Fig. 4g), suggesting the activation of cortical microglia upon SNCA overexpression.

Fig. 4: Reactive microglia mediate PV-INs depopulation to impair motor behaviors.
figure 4

a Immunofluorescent staining of microglial cells (Iba1+) and its phagocytosis toward SNCA-expressing cells (mCherry labeled) in M1 from SNCA model mice. Scale bare, 50 μm or 25 μm (enlarged views only). b The number of microglial cells in the given field was increased by SNCA overexpression. Two-sample unpaired t-test, t(7) = 3.419, P = 0.0112. N = 5 and 4 mice from mCherry and SNCA group, respectively. c Representative blotting bands of Iba1 and CD68 from M1 extracts. d Quantification of relative protein expression of Iba1 and CD68. Multiple t-test showed significant difference between groups. P = 0.038 for Iba1 and P = 0.030 for CD68. N = 4 mice in each group. e Contours of sampled microglial morphology. Scale bar, 15 μm. f Microglial cells of model mice also had larger soma size compared to control group. D = 0.9420, P < 0.0001. g SNCA model mice presented more cell processes of microglia on average. Non-parametric Kolmogorov–Smirnov test, D = 0.7429, P < 0.0001, n = 30 and 35 cells from 4 mice in each group, in (f, g). h Graphic illustration of microglial ablation assays. SNCA-overexpressed mice were fed with PLX5622-containing food chows for 2 weeks, in contrast with a control group receiving normal (ANI76) diet. Behavioral phenotyping was performed later. i Representative immunofluorescent staining images of microglial cells and PV-INs. Although PLX5622 did not change the level of α-synuclein, PV-IN population was maintained. Scale bar, 150 μm in upper panels and 50 μm in lower panels. j Oral PLX5622 administration remarkably decreased microglial cell density. Two-sample unpaired t-test, t(13) = 12.61, P < 0.001. k Ablation of microglial cells increased PV-INs density. t(13) = 9.071, P < 0.001. N = 7 and 8 mice in ANI76 and PLX5622 group, respectively, in (j, k). l PLX5622 treatment recovered mouse performance on the accelerating rotarod. Two-way ANOVA with respect to group factor, F(1117) = 34.80, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. m Microglia ablation improved the performance on vertical pole climbing task. Two-way ANOVA with respect to group factor, F(139) = 21.88, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. n PLX5622 intervention rescued impairment on the walking beam. F(1,39) = 45.28, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. N = 7 and 8 mice from ANI76 and PLX5622 group, respectively, in (l–n). *P < 0.05, **P < 0.01, ***P < 0.001. All data were presented as mean ± SEM.

Full size image

To validate the role of microglia activation in cortical dysfunction, we employed PLX5622, one potent inhibitor of colony-stimulating factor 1 receptor (CSFR1), to suppress the microglial population24. By feeding mice with PLX5622-contained food chows (Fig. 4h), we successfully eliminated the majority of cortical microglial cells (P < 0.001; Fig. 4i, j). Of note, the number of PV-INs was also recovered by microglial ablation (P < 0.001; Fig. 4i, k). Further motor learning tasks revealed the remarkable improvement of motor learning deficits of SNCA mice after microglial inhibition (P < 0.01; Fig. 4l–n). These data highlighted the role of activated microglia in synucleinopathy-related motor learning deficits.

Disruption of ferroptosis mediates cortical PV-INs loss and motor learning dysfunctions

Although we have observed the contribution of microglial activation in cortical deficits under SNCA overexpression, the detailed molecular mechanism remained unclear. We thus employed a transcriptomic study of mouse M1 extracts to investigate the differentially regulated genes (DEGs). Bioinformatics analysis identified a total of 157 up-regulated and 115 down-regulated genes in SNCA mice (Supplementary Fig. 7). Using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, the ferroptosis pathway was found to be prominently affected (Fig. 5a). Previous knowledge recognized ferroptosis to be involved in the molecular pathology of neurodegenerative diseases8,25. The quantification of gene transcript expression level in M1 found upregulation of ferroptosis-related genes such as Phospholipase A2 group VI (Pla2g6), cystine/glutamate antiporter solute carrier family 7 member 11 (Slc7a11), prostaglandin-endoperoxide synthase 2 (Ptgs2) and transferrin receptor (Tfrc) (P < 0.05; Fig. 5b), supporting the activation of ferroptosis in SNCA group.

Fig. 5: SNCA mediates iron homeostasis and the ferroptosis pathway for PV-INs loss.
figure 5

a Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis for the differentially regulated pathways of SNCA group. b Relative gene expression levels of major factors in the ferroptosis pathway. Multiple t-tests were used. P = 0.0021, 0.0047, 0.0012, 0.0490, 0.0010, and 0.0406 for Pla2g6, Slc7a11, Ptgs2, Tfrc, Alox5, and Dmt1 gene, respectively. N = 3 mice in each group. c Graphic illustration of ferroptosis inhibition assay using ferrostatin-1 (Fer-1) administration for 2 weeks on SNCA model. d Representative immunofluorescent staining for SNCA-expression (mCherry) and PV-INs population. Scale bar, 100 μm. e Fer-1 treatment remarkably increased PV-INs density. Two-sample unpaired t-test, t(8) = 4.146, P = 0.0032. N = 5 mice per group. f Blockade of ferroptosis significantly improved mouse performance on the accelerating rotarod. Two-way ANOVA concerning group factor, F(1,90) = 41.67, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. g Fer-1 treatment enhanced the mouse performance on the vertical pole climbing task. Two-way ANOVA concerning group factor, F(130) = 25.03, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. h Under Fer-1 infusion, the ability to finish the walking beam take was improved. Two-way ANOVA concerning group factor, F(130) = 23.67, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. N = 6 mice per group in (f–h). i Experimental design of iron uptake inhibition assay using shRNA-based silencing of Tfrc gene on SNCA model. j Relative expression of Tfrc gene after shRNA expression. Two-sample unpaired t-test, t(4) = 3.449, P = 0.0261. N = 3 mice per group. k Representative immunofluorescent staining for SNCA-expression (mCherry) and PV-INs population. Scale bar, 100 μm. l Tfrc gene knockdown significantly recovered the PV-INs population. Two-sample unpaired t-test, t(10) = 4.849, P = 0.0007. m Blockade of cellular iron uptake significantly improved the mouse performance on the accelerating rotarod. Two-way ANOVA concerning group effect, F(199) = 52.80, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. n The silencing of the Tfrc gene improved the mouse performance on vertical pole climbing tasks. Two-way ANOVA concerning group factor, F(130) = 18.74, P = 0.0002. Sidak’s multiple comparison test of post hoc comparison at each time point. o Under Tfrc gene deficiency, the ability to finish the walking beam take was improved. F(130) = 31.66, P < 0.0001. Sidak’s multiple comparison test of post hoc comparison at each time point. N = 6 mice per group in (l–o). *P < 0.05, **P < 0.01, ***P < 0.001. All data were presented as mean ± SEM.

Full size image

To substantiate the role of ferroptosis in SNCA-induced cortical pathology, the potent inhibitor of ferroptosis, ferrostatin-1 (Fer-1)26 was infused into SNCA model mice (Fig. 5c). Fourteen-day drug treatment remarkably recovered normal PV population density in M1 (P = 0.003; Fig. 5d, e) and rescued impairments of complex motor skill learning (P < 0.0001; Fig. 5f–h). As Fer-1 mainly targets the excess free oxidative radicals in the presence of iron26, an alternative approach was performed via short hairpin RNA (shRNA)-mediated silencing of Tfrc gene expression selectively in PV-INs to manipulate the intracellular trafficking of iron (Fig. 5i, j). Morphometric studies revealed that PV-IN population was increased after iron transport was blocked (P < 0.001; Fig. 5k–l). Moreover, the skill of complex motor functions was remarkably improved with the deprivation of Tfrc gene in PV-INs (P < 0.001; Fig. 5m–o) as those of Fer-1 treated animals. Due to the unaffected microglial cell population in DS after SNCA overexpression (Supplementary Fig. 8a), and similar microglial density under either Fer-1 or shRNA (Tfrc) treatment (Supplementary Fig. 8b, c), it is likely that SNCA overexpression in M1 only leads to local neuroinflammatory response, and ferroptosis acts at the downstream of neuroinflammation to confer cellular toxicity to PV-INs. Taken together, our molecular, physiological, and behavioral data supported the role of ferroptosis in mediating PV-INs depopulation and cortical dysfunction, under local overexpression of SNCA.

Discussion

The current study revealed the previously unrecognized role of disrupted E/I balance in cortical regions following α-synuclein aggregation. Cortical inhibitory neurons are critical for maintaining normal neural functions, and the dysregulation of INs may result in neurological disorders. Of note, our work suggested potentially suppressed presynaptic inhibitory inputs towards PNs, leading to the overexcitation of cortico-striatal pathway and motor learning deficits. These results thus illustrated the changes of cortical circuits under SNCA overexpression, providing the neural mechanism for motor learning deficits. As further molecular evidence, iron overload and ferroptosis are believed to be involved in cortical E/I imbalance.

The region- and cell-type-specific expression of α-synuclein is one critical issue for illustrating the circuitry mechanism of PD. Current studies mainly focused on dopaminergic neurons whilst leaving the GABAergic cells largely unattended. Some comparative studies in mouse tissues found the co-existence between α-synuclein and vesicular glutamate transporter 1 (vGluT1) but not in cortical GABAergic neurons under normal conditions27,28. The exogenous supply of SNCA fibrils induced intracellular aggregation formation in inhibitory neurons, which did not present the formation of Lewy body structure as those in excitatory neurons28. These results highlighted the potentially differentially regulated pathway for synucleinopathy in inhibitory neurons. In our model, SNCA was expressed in both glutamatergic and GABAergic cells, the latter of which presented cell loss in the first place. It is worth noticing that the downregulation of PV-INs has been found across different neurological disorders, including autism29, PD30 and AD31. Although no direct evidence has been obtained to demonstrate the neuronal loss of PV-INs in our work, and reports from PD patients did not reveal the neuronal degeneration in cortical regions32, a few models have been discussed to explain the vulnerability of GABAergic neurons under proteinopathy, including the relatively higher energy requirement33, or the preference of microglial phagocytosis targeting the extracellular matrix of PV-INs34.

GABAergic INs play indispensable roles in maintaining the structural and functional integrity of cortical circuits. PV-INs are the major subtype of inhibitory cells forming peri-somatic innervations of PNs to exert rapid and powerful inhibition35, providing critical inputs for maintaining the E/I balance of the cortical network36. From functional perspectives, cortical PV-INs have been found to regulate social behaviors37, sensory integration38,39 and spatial working memory40. The disruption of normal PV-IN function resulted in mental disorders such as schizophrenia41. In motor cortex, PV-INs work with other INs to shape task-dependent ramping signals preceding motor movements42 and help to regulate the plasticity of other inhibitory boutons during rotarod motor learning43. Our data illustrated the participation of PV-INs in tuning the calcium activity of PNs, as cell-specific ablation and chemogenetic excitation effectively recapitulated or reversed SNCA-induced hyperexcitability of local L5 PNs, respectively. These results highlighted the importance of cortical E/I balance in executing normal motor learning tasks. Since inhibitory transmission can be regulated by noninvasive neuromodulation approaches such as transcranial magnetic stimulation (TMS)44, strategies might be developed to target cortical inhibitory network for the alleviation of Parkinsonism.

In searching for the molecular mechanism governing the gradual loss of PV-INs upon SNCA overexpression, we proposed that ferroptosis signaling might be related to PV-INs loss. In the current work, we found prominent changes in ferroptosis-related genes, including Pla2g6, Slc7a11, Ptgs2, and Tfrc. To further demonstrate the participation of this ferroptosis pathway, the pharmaceutical inhibitor Fer-1 was employed to rescue PV-IN degeneration and motor learning deficits. Moreover, the involvement of iron homeostasis was supported via depressing Tfrc gene expression, in order to inhibit cellular iron uptake for relieving iron overload. The excess iron content may further disrupt the catabolism of lipid peroxide, leading to the activation of the ferroptosis cascade in cortical neurons, as supported by a recent study using the spinal injury model45. Since ferroptosis in the striatum might be related to intracerebral hemorrhage46 and PD47, the targeting of ferroptosis may provide a potential strategy for PD treatment.

The activation of the ferroptosis pathway is closely related to PD pathogenesis. Amounts of studies have supported the participation of iron overload and the disruption of critical regulators of the ferroptosis pathway during PD pathogenesis. For example, in a homozygous PD mouse model (α-Synuclein A53T mutant), dopaminergic neuronal death is caused by iron overload during the early disease stage48. The iron overload also induced the transcriptomic dynamic shift of PD patients’ brains toward ferroptosis, as supported by the human-induced pluripotent stem cell-derived microglial cell culture assay49. In a drug-induced PD model, iron accumulation and ferroptosis were also reported, in addition to the rescued phenotype by Fer-1 treatment50 similar to our work. Moreover, a recent study using dopaminergic cell culture model further supported the protective role of Ferritin against neurotoxicity51. These in vitro and in vivo data all agreed on the relationship between iron overload in conjunction with ferroptosis and PD susceptibility and proposed the potential value of targeting the ferroptosis pathway for neuroprotection. Although these abovementioned works mainly studied midbrain dopaminergic neurons, here we presented amounts of data showing the critical role of this pathway in cortical neuronal functions. More importantly, by specifically modulating Tfrc gene expression in PV-INs, we reported the necessary role of ferroptosis in cortical INs under synucleinopathy.

In sum, SNCA disrupted E/I balance of the motor cortex, mainly due to the downregulation of PV-INs, which leads to the hyperexcitation of PNs and motor learning deficits. The loss of PV-INs was potentially related to microglial activation and elevated ferroptosis. Our work provides more mechanisms for neuropathology under synucleinopathy and implies potential molecular and circuitry targets for future interventional strategies.

Methods

Experimental animals

C57BL/6J mice (5 weeks old, male) were purchased from Guangdong Medical Laboratory Animal Center. PV-Cre mice (5 weeks old, male) were obtained from Jackson Laboratory and were bred by Shanghai Model Organisms Center, Inc. Unless otherwise specified, all mice were group housed (4–5 individuals per standard cage) with food and water ad libitum. The animal facility was kept with a 12 h light/dark cycle (light 8:00–20:00, dark 20:00–8:00), and the temperature was maintained at 24 ± 2 °C, with the relative humidity at 50 ± 10%. All animal experimental protocols have been approved by the Ethics Committee of Experimental Animals of Jinan University in accordance with Institutional Animal Care and Use Committee guidelines for animal research.

For PLX5622 treatment, 0.12% PLX5622 was prepared in standard AIN-76A rodent diet, which was provided by Dyets Inc. Mice were provided PLX5622-containing chow starting at 7 days after virus infection. For Fer-1 treatment, mice received intraperitoneal injection of Fer-1 (MedChemExpress, US) at 5 mg/kg for 14 consecutive days.

Stereotaxic injection and two-photon imaging

Mice were anesthetized with 1.25% Avertin. After local sterilization and incision of head skins, the primary motor cortex (M1, AP: +0.5 mm anterior to the Bregma; ML: ±1.2 mm lateral; and DV: −0.6 mm depth) was located under a stereotaxic instrument (RWD Life Science, China). After drilling a hole with the high-speed microdrill (WPI, USA), 100 nl of AAV serotype 2/9 (viral titer, 5 × 1012 genome copies per ml, see Supplementary Table 1) was injected using a glass micropipette connected to an ultra-micro injection pump (Nanoliter 2020, WPI, USA). The glass electrode was retained within the brain tissue for 8 min before retraction. After the virus injection, the scalp was sutured, and the mouse was placed back in the cage for 1-week recovery. Analgesics (buprenorphine at 0.1 mg/kg) was used when necessary.

One day before the imaging session, a transcranial imaging window was prepared. In brief, the head skin and the skull over the primary motor cortex were carefully removed by a microdrill. The window was covered by a glass coverslip using Tissue Adhesive (3M, USA). The imaging session was recorded at 2 Hz using a 920 nm excitation laser with a water-immersed objective (20×; ZEISS, Germany) under an LSM780 two-photon microscope (ZEISS, Germany).

Acquired images were analyzed by ImageJ (NIH, US). The fluorescent value was quantified from the designed region of identifiable soma. The △F/F0 was calculated as (F − F0)/F0, where the F0 was the minimum of F values as the basal level. A calcium spike was defined when the △F/F0 was higher than three-fold standard deviations (SDs).

Behavioral assays

Open field

The test apparatus contained a transparent field at 50 × 50 × 50 cm, with an arbitrary defined central area (25 × 25 cm). During the 5-min single test session, the Ethovison XT software was used to track and analyze the mouse movement trajectory. The anxiety level was evaluated using the time in the central area.

Novel object recognition task

The experiment procedure followed our previous works52. In brief, the animal’s preference toward a novel object was evaluated by the exploration time during 10 min test session, which was carried at 15 min or 24 h after the initial learning session (10 min, including 2 identical objects). The Ethovison XT software was used to record and analyze the movement trajectories.

Accelerating rotarod test

The Rotarod system (UGO Basile, Italy) was set at an initial minimum speed of 2 rpm to reach the maximum speed of 80 rpm within 5 min. The whole test was performed in 3 consecutive days, and on each day, there were 3 test trials. Before the formal test, an acclimation session was performed using 2 rpm velocity for 1 min. The time the mouse fell down from the rod was recorded and was averaged across 3 trials.

Vertical climbing test

The experiment apparatus consists of a vertically placed rod (70 cm height, 1 cm diameter). At the beginning of each session, the mouse was placed on the top and was allowed to climb down to the bottom. There were 3 trials on each day, with 5-min interval between each trial. The whole test session lasted for 3 consecutive days.

Horizontal beam walking test

The experiment was performed using a wooden plate with 3 cm width and 70 cm length, with an escape box at one end of the beam. The whole apparatus was placed 100 cm above the ground. The mouse was placed at the end of the beam, and the time it took to reach the escape box was recorded. A total of 3 test trials were performed each day, with 5-min intervals. The whole test session lasted for 3 consecutive days.

Immunofluorescent staining

Mice were treated with 1.25% Avertin for anesthesia, followed by cardiac perfusion using 0.9% saline and 4% paraformaldehyde solution (PFA). The whole brain tissue was carefully removed and was fixed overnight using 4% PFA. The brain tissue was subsequently immersed in a 30% sucrose solution for 1–2 days of dehydration. The whole brain tissue was then cut into 40 μm using a sliding microtome (Lecia, Germany). After PBS washing and blocking, the brain slices were incubated with primary antibody at 4 °C for 48 h, followed by secondary antibody (see Supplementary Table 2 for antibody information). Images were captured with a confocal microscope (ZEISS, Germany).

Western blotting

Tissue lysates were extracted from the primary motor cortex, using the RIPA buffer (Thermo Scientific, US) containing protease and phosphatase inhibitors. After quantification using Pierce Rapid Gold BCA kit (Thermo Scientific, US), 25 μg protein samples were electrophoresed on 8% SDS-polyacrylamide gel and were transferred to the 0.2 μm PVDF membranes under 280 mA for 120 min. The membrane was washed by the TBST and blocked by 5% BSA and was incubated with primary antibody at 4 °C for 8 h. The membrane was then washed and incubated in a secondary antibody for 2 h (Supplementary Table 2). Images were captured by a protein imaging system (Bio-Rad, US) and the integrated gray values were calculated by ImageJ (NIH, US).

RT-qPCR

Total RNA was extracted from the primary motor cortex with Trizol (Invitrogen, US) reagent. Total RNA was reversely transcribed using Prescript RT reagent Kit with gDNA Eraser (Takara, Japan). Real-time PCR was performed using SYBR Green reagent (Takara, Japan) and specific primers (see Supplementary Table 3).

Whole-cell patch-clamp recording

The mouse was deeply anesthetized with isoflurane, and coronal sections (250 μm thick) containing the primary motor cortex were prepared using a vibratome (VT1200S; Leica Microsystems, Germany) in ice-cold artificial cerebrospinal fluid (aCSF: 119 mM NaCl; 2.5 mM CaCl2,2.5 mM KCl, 1.3 mM MgCl2, 1 mM NaH2PO4, 11 mM glucose, 26.2 mM NaHCO3, 1.3 mM MgCl2, and 290 mM mOsm, pH 7.4). The brain slices were recovered for 1 h at room temperature in aCSF. Slices were then transferred to the recording chamber with continuous perfusion of oxygenated aCSF containing 1 μM tetrodotoxin.

The whole-cell recording was performed on cortical L5 PNs using an epifluorescence microscope (Olympus, Japan). A glass pipette (WPI, US) was filled with internal solutions (135 mM CsMeSO3, 8 mM NaCl, 10 mM Na2-phosphocreatine, 0.25 mM EGTA, 2.17 mM Mg–adenosine 5′-triphosphate, 0.34 mM Na3–guanosine 5′-triphosphate, and 10 mM HEPES (pH 7.4) with CsOH). The miniature excitatory postsynaptic currents (mEPSCs) were recorded using 1 μM TTX and 20 μM bicuculline. The miniature inhibitory postsynaptic currents (mIPSCs) were recorded using 1 μM TTX, 20 μM NBQX and 50 μM D-AP5. All recordings were performed using a Multiclamp 700B amplifier (Molecular Devices, US). Traces were low-pass-filtered at 2 kHz and digitized at 10 kHz. The pipette resistance ranged from 4 to 6 MΩ. During the recording, series resistance (Rs) was maintained between 10–20 MΩ. If Rs changed >20%, the data were excluded. Offline data analysis was performed using Clampfit 10.0 software (Molecular Devices, US).

Fiber photometry

In brief, at 2 weeks after AAV2/9- GCaMP6s virus injection, an optical fiber [OD = 230 mm, numerical aperture = 0.37] was inserted towards the dorsolateral striatum (AP: +0.6 mm, ML: ±2.4 mm, DV: −0.3 mm). The fiber was fixed with a skull penetrating screw and dental acrylic. Mice were then individually housed for 1-week recovery. The fluorescence signals were detected through the optic fiber using a fiber photometry system (Thinker Biotech, China). The 410 nm signal and 470 nm signal were recorded at the same time, while the 410 nm channel was used to correct for false positive signals from animal movement and bleaching. The values of fluorescence change (ΔF/F) were calculated as (F − F0) / F0, where F0 was the averaged baseline fluorescence signal.

Statistics and reproducibility

GraphPad Prism 9.5 (GraphPad, US) was used for statistical analysis and data plotting. All data were plotted as mean ± SEM. Unpaired Student’s t-test was used for comparing the means between two groups in two-tailed manners. For 2 independent variables, 2-way analysis of variance (ANOVA) and Sidak post hoc comparison were adopted. A significant level was defined when P < 0.05. The sample size of each experiment was reported in the relevant figure captions. No replicate was performed in most of the studies, except in those of qPCR, in which 3 replicates were performed for each sample.