Disrupting stroke-induced GAT-1-syntaxin1A interaction promotes functional recovery after stroke

Abstract

Although stroke is a frequent cause of permanent disability, our ability to promote stroke recovery is limited. Here, we design a small-molecule stroke recovery promoting agent that works by dissociating γ-aminobutyric acid (GABA) transporter 1 (GAT-1) from syntaxin1A (Synt1A), a soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) protein. Stroke induces an increase in GAT-1-Synt1A interaction in the subacute phase, a critical period for functional recovery. Uncoupling GAT-1-Synt1A reverses stroke-induced GAT-1 dysfunction and cortical excitability decline and enhances synaptic GABAergic inhibition and consequently cortical oscillations and network plasticity by facilitating the assembly of the SNARE complex at the synapse. Based on the molecular mechanism of GAT-1 binding to Synt1A, we design GAT-1-Synt1A blockers. Among them, ZLQ-3 exhibits the greatest potency. Intranasal use of ZLQ-3-1, a glycosylation product of ZLQ-3, substantially lessens impairments of sensorimotor and cognitive functions in rodent models. This compound, or its analogs, may serve as a promoting agent for stroke recovery.

Keywords: stroke, GAT-1, syntaxin1A, network plasticity, functional recovery

Introduction

Stroke is the third-leading cause of death and disability combined in the world.^1^ Even after transient ischemic attack or minor stroke, long-term disability also occurs.^2^ To date, apart from thrombolytic agents, no therapeutic drug has clear and reproducible efficacy on primary clinical endpoints.^3^ Although thrombolysis and/or endovascular thrombectomy in eligible patients with stroke within the acute time window can significantly improve the outcome,^4^ up to 40% of the patients with stroke remain severely disabled or died after thrombolysis,^5^ and among the patients receiving thrombectomy, many of them still experienced some degree of neurological deficits.^6^ Therefore, disability is a major challenge after the acute phase of stroke.

During the subacute phase of stroke, structural and functional remodeling occurs in the adjacent and distant areas of the infarct core.^7^^,^^8^ The remodeling enhances excitability and plasticity within bilateral neural networks and is implicated in improving motor functions.^8^^,^^9^^,^^10^^,^^11^ So far, however, pharmacological targets and therapeutic agents for enhancing the excitability and plasticity of post-stroke neural networks are still lacking. Chronically elevated ambient γ-aminobutyric acid (GABA) level during the subacute phase increases tonic inhibition by activating extrasynaptic GABA type A receptor (GABAAR) and thereby antagonizes the neuronal excitability required for stroke recovery.^8^^,^^12^^,^^13^ Alpha5-GABAAR antagonists administrated in the subacute phase can reduce the tonic inhibition and thereby increase cortical excitability and improve stroke recovery in rodents.^8^^,^^12^^,^^14^ However, the clinical trial of S44819, a selective α5-GABAAR antagonist, has failed,^15^ and in addition, GABAAR antagonist may have a risk of triggering epilepsy,^16^ suggesting that using GABAAR antagonists alone in the subacute phase may not be appropriate. Beta oscillations gated by GABAAR involve somatosensory processing and motor control and are crucial for stroke recovery.^17^^,^^18^^,^^19^ Drugs enhancing GABAergic inhibition promote stroke recovery and most commonly alleviate seizures.^16^^,^^20^ We thus speculated that drugs enhancing intrasynaptic GABAergic inhibition while reducing extrasynaptic GABAergic inhibition may be more promising for stroke recovery.

To discover this kind of drugs, we focused on GABA transporters (GATs). GATs are divided into four types, including GAT-1, BGT-1, GAT-2, and GAT-3. GAT-1 is mainly expressed in GABAergic neurons in the cortex and some in astrocytes and microglia, enhances neuronal excitability by regulating extracellular GABA reuptake,^21^ and plays a role in stroke recovery.^13^^,^^22^ The binding of GAT-1 to the cytoplasmic H3 domain of syntaxin1A (Synt1A), a soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) protein, can cause GAT-1 dysfunction.^23^ Importantly, the Synt1A H3 domain (Synt1AH3) also mediates interaction with other SNARE proteins, which is crucial for the assembly and stability of the SNARE complex required for vesicular fusion at the synapse.^23^^,^^24^ Thus, competitive bindings between GAT-1 and other SNARE proteins to Synt1AH3 may interfere in synaptic GABAergic inhibition. We hypothesized that drugs dissociating GAT-1 from Synt1A in GABAergic neurons may upregulate GAT-1 function and thereby neuronal excitability, and enhance synaptic GABAergic inhibition.

Results

Blocking stroke-induced GAT-1-Synt1A interaction promotes functional recovery

To determine whether stroke affects the interaction between GAT-1 and Synt1A, we produced a photothrombotic stroke to the motor cortex in mice, in which the ischemic infarct core and the peri-infarct zone were well defined, and detected the GAT-1-Synt1A complex in the peri-infarct cortex on days 3, 5, 7, and 10 after stroke and found that stroke significantly increased the level of the GAT-1-Synt1A complex on days 5 and 7, and the level of GAT-1 coupling with Synt1A as a fraction of total GAT-1 protein (by calculating the ratio of IB GAT-1 to input GAT-1) from day 3 to day 10 (Figures 1A–1C). We further demonstrated the stroke-induced GAT-1-Synt1A interaction on day 7 after stroke using proximity ligation assay (Figures 1D and 1E). Furthermore, stroke did not increase the expression of GAT-1 or Synt1A, but it led to a significant increase in the membrane content of Synt1A (Figures S1A–S1C), resulting in a significant increase in GAT-1-Synt1A coupling. In the in vitro cultured astrocytes and microglia, Synt1A levels were much lower than that in neurons (Figure S1D). Thus, stroke-induced binding of GAT-1 to Synt1A may mainly occur in neurons.

Figure 1 Stroke-induced GAT-1-Synt1A interaction hinders stroke recovery(A–C) Coimmunoprecipitation (coIP) showing the amounts of Synt1A-GAT-1 complex and GAT-1 in the peri-infarct cortex on indicated days after stroke. (A) IgG control of coIP experiment (left) and representative immunoblots of coIP experiment (right). Input GAT-1 and input β-actin refers to total GAT-1 and β-actin level in the sample, respectively. (B) The amount of Synt1A-GAT-1 complex. One-way ANOVA followed by post hoc Scheffe test, F(4, 20) = 18.18, n = 5. (C) The level of GAT-1 coupling with Synt1A as a fraction of total GAT-1 protein. One-way ANOVA followed by post hoc Scheffe test, F(4, 20) = 16.21, n = 5.(D) Representative images showing Synt1A-GAT-1 interaction detected by proximity ligation assay (PLA) in samples from the indicated groups.(E) PLA quantification showing the amount of Synt1A-GAT-1 complex per 50 nuclei. Two-tailed t test, t(22) = −8.23, n = 12.(F) CoIP showing Synt1A-GAT-1 complex level in the neurons after treating with Synt1A251-265 or peptides clipped from Synt1A251-265. One-way ANOVA followed by post hoc Scheffe test, F(4, 20) = 17.03, n = 5.(G) CoIP showing the effect of Tat-Synt1A251-265mut on Synt1A-GAT-1 interaction in cultured neurons. Two-tailed t test, t(10) = 0.21, n = 6.(H) CoIP showing the effect of AAV-Synt1A251-265 on stroke-induced Synt1A-GAT-1 interaction. One-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 5.85, n = 5.(I) Left, foot faults of the left forelimb in the grid-walking task. Middle, foot faults of the left hindlimb in the grid-walking task. Right, forelimb symmetry in the cylinder task. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test. Left, F(2, 31) = 312.827; middle, F(2, 31) = 77.627; right, F(2, 31) = 233.376.(J) Left, foot faults of the left forelimb in the grid-walking task. Middle, foot faults of the left hindlimb in the grid-walking task. Right, forelimb symmetry in the cylinder task. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test. Left, F(3, 37) = 231.769; middle, F(3, 37) = 22.897; right, F(3, 37) = 122.262.See also Figure S1.

To explore the potential inducer underlying stroke-induced Synt1A-GAT-1 interaction, we focused on nitric oxide (NO), not only because that sustained and significant production of NO caused by increased coupling between neuronal NO synthase (nNOS) and postsynaptic density 95 (PSD-95) is a key pathological event during the subacute phase of stroke,^25^^,^^26^ but also due to the fact that NO promotes the translocation of Synt1A and GAT-1 from the cytoplasm to the cell surface, especially the translocation of Synt1A, thereby increasing the interaction between Synt1A and GAT-1 and reducing ^3^H-GABA uptake.^27^ More importantly, we treated the cultured neurons with GSNO, an NO donor, and found that the treatment significantly increased membrane contents of GAT-1 and Synt1A, but had no effects on the total levels of GAT-1 and Synt1A (Figures S1E and S1F).

It has been reported that the function of GAT-1 is critical for stroke recovery, and GAT-1-Synt1A interaction leads to GAT-1 dysfunction.^13^^,^^22^^,^^23^ To explore how the binding of Synt1A to GAT-1 affects the reuptake of GABA by GAT-1, we expressed GAT-1 alone or co-expressed GAT-1 and Synt1A in HEK293 cells. Consistent with the previous report,^23^ co-expressing GAT-1 and Synt1A in HEK293 cells significantly reduced ^3^H-GABA uptake, compared with expressing GAT-1 alone (Figure S1G). However, co-expressing GAT-1 and Synt1A had no effect on the total level and membrane content of GAT-1, compared with expressing GAT-1 alone (Figures S1H and S1I). Thus, the binding of Synt1A affects the function of GAT-1 due to Synt1A-GAT-1 interaction rather than changes in the expression or trafficking of GAT-1.

For the binding of GAT-1 to the Synt1A, three aspartic acid residues (D40, D43, and D45) in the N-terminal tail of GAT-1 (N-tailDDD) is critical.^23^ However, how the N-tailDDD binds to the Synt1A remains unclear. High salt dramatically reduces GAT-1-Synt1A interaction,^23^ suggesting an electrostatic interaction. The residues 251–265 of the Synt1AH3 (Synt1A251-265) are rich in basic amino acid residues,^28^ offering a possibility to disrupt the electrostatic interaction. To determine whether the Synt1A251-265 mediates GAT-1-Synt1A interaction, we synthesized 4 fusion peptides that comprised a cell-penetrating peptide Tat, Tat-TKKAVKYQSKARRKK (Tat-Synt1A251-265), Tat-TKKAVK (Tat-Synt1A251-256), Tat-VKYQSK (Tat-Synt1A255-260), and Tat-KARRKK (Tat-Synt1A260-265) and found that Tat-Synt1A251-265 and Tat-Synt1A251-256 but not Tat-Synt1A255-260 and Tat-Synt1A260-265 could block GAT-1-Synt1A interaction in the cultured neurons (Figure 1F), suggesting a critical role of Synt1A251-256. Based on this finding, we generated a fusion peptide that comprised Tat and mutated Synt1A251-265 (with T251A, K252A, K253A, V255A, and K256A mutations, Tat-Synt1A251-265mut) as a negative control for the following experiments. Indeed, Tat-Synt1A251-265mut did not affect GAT-1-Synt1A interaction, compared with vehicle (Figure 1G).

To test whether dissociating GAT-1 from Synt1A promotes stroke recovery, we generated an AAV vector expressing Synt1A251-265 (AAV-Synt1A251-265) or Synt1A251-265mut (AAV-Synt1A251-265mut) as a control. For the mouse photothrombotic stroke model that specifically damages motor cortex and thereby causes motor functional deficit, we used grid-walking task and cylinder task to assess motor function after stroke. We delivered these vectors into the peri-infarct cortex of the mice with photothrombotic stroke on day 3 before stroke and performed motor functional test on day 7 before stroke and on days 4, 12, 19, 26, and 33 after stroke. On day 8 after stroke, Synt1A251-265-EGFP identified as green fluorescence was abundantly expressed in surviving neurons in the peri-infarct cortex (Figure S1J). AAV-Synt1A251-265 reversed stroke-induced GAT-1-Synt1A interaction (Figure 1H) and significantly improved motor functions compared with AAV-Synt1A251-265mut (Figure 1I), but had no effect on motor functions of sham-operated mice (Figure S1K).

To further confirm the effect of Synt1A251-265 on post-stroke functional recovery, Tat-Synt1A251-265 was microinjected into the peri-infarct cortex of mice during days 5–11 after stroke, and 0.2 or 1.0 pmol of Tat-Synt1A251-265 reversed stroke-induced increase in the GAT-1-Synt1A complex detected on day 8 after stroke (Figures S1L–S1N). Next, we delivered Tat-Synt1A251-265 or its control Tat-Synt1A251-265mut during days 5–11 after stroke into the peri-infarct cortex of GAT-1^flox/flox^ mice transfected with AAV-Cre-EGFP or AAV-EGFP on day 3 before stroke and detected motor functions on day 7 before stroke and on days 4, 12, 19, 26, and 33 after stroke (Figure S1O). GAT-1 levels decreased by ∼85% in the AAV-Cre-infected peri-infarct tissue, suggesting an effective conditional knockout (CKO) (Figure S1P). As shown in Figure 1J, treatment with Tat-Synt1A251-265 significantly improved motor functions in the GAT-1^flox/flox^ mice transfected with AAV-EGFP but not in the GAT-1^flox/flox^ mice transfected with AAV-Cre-EGFP, compared with Tat-Synt1A251-265mut. Tat-Synt1A251-265 had no effect on motor functions of sham-operated mice (Figure S1Q). These results suggest that Tat-Synt1A251-265 can promote stroke recovery and the effect of Tat-Synt1A251-265 depends on GAT-1. To determine whether peptide Tat-Synt1A251-265 affects the function, stability, or trafficking of GAT-1 without full-length Synt1A, we incubated the cultured HEK293 cells overexpressing GAT-1 with Tat-Synt1A251-265 (100 nM) or Tat-Synt1A251-265mut (100 nM) and observed their effect on ^3^H-GABA uptake, expression, and membrane distribution of GAT-1. Tat-Synt1A251-265 did not affect ^3^H-GABA uptake, expression, and membrane distribution of GAT-1, compared with Tat-Synt1A251-265mut (Figures S1R–S1T). Together, peptide Synt1A251-265 promotes stroke recovery through dissociating GAT-1 from Synt1A rather than directly affecting GAT-1.

Uncoupling GAT-1-Synt1A enhances cortical excitability

To determine whether GAT-1-Synt1A interaction reduces the reuptake of GABA by GAT-1 in the ischemic cortex, we measured the concentration of extracellular GABA ([GABA]o) in the peri-infarct cortex of stroke mice using microdialysis method and found that the microinjection of Tat-Synt1A251-265 (0.2 pmol/d) during days 5–7 after stroke reversed stroke-induced increase in [GABA]o on day 8 after stroke (Figure 2A). Next, we incubated the cultured neurons with Tat-Synt1A251-265 (100 nM) or Tat-Synt1A251-265mut (100 nM) and observed their effect on the uptake of extracellular GABA labeled by ^3^H into neurons. Tat-Synt1A251-265 increased the uptake of ^3^H-GABA significantly, compared with Tat-Synt1A251-265mut (Figure 2B). Thus, dissociating GAT-1 from Synt1A increases the reuptake of GABA by GAT-1. Based on our data described earlier and reports that elevated ambient GABA level antagonizes neuronal excitability via tonic inhibition,^8^^,^^12^^,^^13^ we hypothesized that uncoupling GAT-1-Synt1A reduces tonic inhibition and thereby increases cortical excitability in the subacute phase of stroke (Figure 2C).

Figure 2 Dissociating Synt1A from GAT-1 reverses stroke-induced decrease in cortical excitability(A) Bar graph showing the effect of Tat-Synt1A251-265 on the concentrations of extracellular GABA in the peri-infarct cortex. One-way ANOVA followed by post hoc Scheffe test, F(2, 15) = 23.29, n = 6.(B) Bar graph showing the effect of Tat-Synt1A251-265 on GABA uptake in the cultured neurons. Two-tailed t test, t(8) = −6.72, n = 5.(C) Hypothesis: stroke-induced GAT-1-Synt1A association causes GAT-1 dysfunction and consequent tonic inhibition increase, dissociating Synt1A from GAT-1 by Synt1A251-265 reverses stroke-induced GAT-1 dysfunction and tonic inhibition.(D) Representative traces and bar graphs showing tonic inhibitory currents from indicated groups. Tonic current was the change in baseline holding current after bath-application of BMI, represented as the distance between the red and blue dashed lines. One-way ANOVA followed by post hoc Scheffe test, F(4, 51) = 8.37.(E) Normalized GCaMP6s fluorescence of pyramidal neurons of mouse before and during spontaneously rearing.(F) Quantification of fluorescence changes after spontaneously rearing. One-way ANOVA followed by post hoc Scheffe test, F(3, 23) = 34.98.(G) Representative mEPSC traces.(H) mEPSC amplitude recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 45) = 2.04.(I) mEPSC frequency recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 45) = 26.07.(J) Representative sEPSC traces.(K) sEPSC amplitude recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 36) = 0.11.(L) sEPSC frequency recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 36) = 19.9.See also Figure S2.

To test whether the modified reuptake of GABA by GAT-1 affects tonic inhibition, we delivered Tat-Synt1A251-265 (0.2 pmol/d) or its control during days 5–7 after stroke into the peri-infarct cortex of GAT-1^flox/flox^ mice transfected with AAV-Cre-EGFP or AAV-EGFP on day 3 before stroke. On day 8 after stroke, tonic GABA currents (Itonic) in the layer 5 peri-infarct pyramidal neurons of acute brain slices from Tat-Synt1A251-265-treated mice were recorded (Figure S2A). Wild-type (WT) mice subjected to stroke displayed a significantly increased Itonic, compared with WT sham mice, and Tat-Synt1A251-265 had no effect on sham-operated mice, but reversed the effect of stroke on Itonic in WT mice with stroke rather than that in GAT-1 CKO mice with stroke (Figure 2D). Thus, uncoupling GAT-1-Synt1A prevents stroke-induced tonic inhibition.

Elevated ambient GABA antagonizes neuronal excitability required for stroke recovery via tonic inhibition.^8^^,^^12^ To determine whether Tat-Synt1A251-265 enhances cortical excitability via modifying tonic inhibition, we infused AAV-CaMKII-GCaMP6s and AAV-CAG-Cre-3Flag into the motor cortex of GAT-1^flox/flox^ mice on days 16 and 3 before stroke, respectively, and then we microinjected Tat-Synt1A251-265 (0.2 pmol/d) or Tat-Synt1A251-265mut (0.2 pmol/d) into the peri-infarct cortex during days 5–11 after stroke, and local cortical domain-specific GCaMP-mediated Ca^2+^ signal was recorded on day 12 after stroke (Figures S2B–S2D). The microinjection of Tat-Synt1A251-265 partially prevented stroke-induced decline in cortical excitability in WT but not in GAT-1 CKO mice, compared with the microinjection of Tat-Synt1A251-265mut (Figures 2E and 2F). Moreover, we detected miniature excitatory postsynaptic currents (mEPSCs) in the layer 5 peri-infarct pyramidal neurons and found that stroke caused a decrease in excitatory synaptic transmission as indicated by significantly reduced mEPSC frequency, and treatment with Tat-Synt1A251-265 (0.2 pmol/d) had no effect on sham-operated mice but reversed stroke-induced decrease in mEPSC frequency on day 8 after stroke (Figures 2G–2I). Furthermore, we recorded spontaneous excitatory postsynaptic currents (sEPSCs) in the layer 5 peri-infarct pyramidal neurons. Although sham-operated mice treated with Tat-Synt1A251-265 had no effect on sEPSCs, mice with stroke treated with Tat-Synt1A251-265 showed significantly increased sEPSC frequency but not amplitude, compared with Tat-Synt1A251-265mut-treated mice (Figures 2J–2L). Together, the dissociation of GAT-1-Synt1A by Tat-Synt1A251-265 enhances cortical domain excitability in the peri-infarct cortex, depending on GAT-1 function.

Uncoupling GAT-1-Synt1A enhances synaptic GABAergic inhibition

The Synt1AH3 not only interacts with GAT-1 but also mediates protein-protein interactions required for the assembly and stability of SNARE complex at the synapse.^24^^,^^29^ The neuronal SNAREs involved in neurotransmitter release are syntaxin-1, SNAP-25, and VAMP-2.^29^^,^^30^ To determine whether GAT-1-Synt1A coupling affects the interaction between Synt1A and SNAP-25 or VAMP-2, we measured Synt1A-SNAP-25 and Synt1A-VAMP-2 complex levels in the peri-infarct cortex of mice receiving the microinjection of Tat-Synt1A251-265 (0.2 pmol/d) during days 5–7 after stroke and found that Tat-Synt1A251-265 treatment reversed stroke-induced decreases in Synt1A-SNAP-25 and Synt1A-VAMP-2 complexes on day 8 after stroke (Figure 3A). To determine whether GAT-1-Synt1A interaction affects the assembly of SNARE complex, we measured Synt1A, SNAP-25, and VAMP-2 monomer and their complex levels in synaptosomes. Stroke significantly decreased SNARE complexes of Synt1A, SNAP-25, and VAMP-2, and Tat-Synt1A251-265 (0.2 pmol/d) treatment during days 5–7 after stroke reversed stroke-induced decreases in SNARE complexes on day 8 after stroke (Figures 3B–3D), suggesting a regulation of the assembly of the SNARE complex by GAT-1-Synt1A interaction (Figure 3E). Stroke or Tat-Synt1A251-265 did not affect Synt1A, SNAP-25, or VAMP-2 monomer (Figures 3B–3D).

Figure 3 Dissociating Synt1A from GAT-1 enhances synaptic GABAergic inhibition through facilitating the assembly of SNARE complex(A) Representative immunoblots (upper) and bar graphs showing the effect of Tat-Synt1A251-265 on the amount of Synt1A-SNAP-25 (middle) and Synt1A-VAMP-2 (lower) complexes. For Synt1A-SNAP-25 coupling, one-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 30.58; for Synt1A-VAMP-2 coupling, one-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 27.86. n = 5.(B) Representative immunoblots (left) and bar graphs showing SNARE complexes (middle) and SNAP-25 monomer (right). For SNARE complexes, one-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 20.37; for SNAP-25 monomer, one-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 1.84. n = 5.(C) Representative immunoblots (left) and bar graphs showing SNARE complexes (middle) and Synt1A monomer (right). For SNARE complexes, one-way ANOVA followed by post hoc Scheffe test, F(2, 15) = 13.25; for Synt1A monomer, one-way ANOVA followed by post hoc Scheffe test, F(2, 15) = 0.8. n = 6.(D) Representative immunoblots (left) and bar graphs showing SNARE complexes (middle) and VAMP-2 monomer (right). For SNARE complexes, one-way ANOVA followed by post hoc Scheffe test, F(2, 15) = 6.31; for VAMP-2 monomer, one-way ANOVA followed by post hoc Scheffe test, F(2, 15) = 1.21. n = 6.(E) Hypothesis: dissociating Synt1A from GAT-1 reverses stroke-induced impairment of SNARE complex formation.(F) Representative mIPSC traces.(G) mIPSC amplitude recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 38) = 1.67.(H) mIPSC frequency recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 38) = 13.69.(I) Representative sIPSC traces.(J) sIPSC amplitude recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 39) = 0.9.(K) sIPSC frequency recorded in the peri-infarct pyramidal neurons from mice of indicated groups. One-way ANOVA followed by post hoc Scheffe test, F(3, 39) = 7.7.See also Figure S2.

Next, we investigated whether GAT-1-Synt1A affects interneuron-mediated inhibitory synaptic transmission. By detecting miniature inhibitory postsynaptic currents (mIPSCs) in the layer 5 peri-infarct pyramidal neurons, we found that stroke caused a decrease in inhibitory neurotransmitter release as indicated by significantly reduced mIPSC frequency, and treatment with Tat-Synt1A251-265 (0.2 pmol/d) had no effect on sham-operated mice but reversed the stroke-induced decrease in mIPSC frequency on day 8 after stroke (Figures 3F–3H). Moreover, we recorded spontaneous inhibitory postsynaptic currents (sIPSCs) in the layer 5 peri-infarct pyramidal neurons. Although Tat-Synt1A251-265 had no effect on sIPSCs of sham-operated mice, Tat-Synt1A251-265-treated mice with stroke showed significantly increased sIPSC frequency but not amplitude, compared with sham and Tat-Synt1A251-265mut-treated mice with stroke (Figures 3I–3K). Together, the enhancements of mIPSCs and sIPSCs by uncoupling GAT-1-Synt1A in the peri-infarct cortex suggest that dissociating GAT-1 from Synt1A enhances synaptic GABAergic inhibition, a therapeutic target for stroke recovery.^20^

The effects of Tat-Synt1A251-265 on mIPSCs and sIPSCs described earlier (Figures 3F–3K) suggest a presynaptic mechanism. To test whether the activity of GABAergic neurons is necessary for the effect of Tat-Synt1A251-265 on stroke recovery, we used a chemogenetic technique to manipulate GABAergic neurons in the peri-infarct cortex. We infused VGAT promoter-driven AAV expressing Gi-coupled hM4Di labeled with mCherry into the peri-infarct area of mice on day 16 before stroke. The recombinant virus vectors can silence the activity of transfected neurons in the presence of designer drug agonist clozapine N-oxide (CNO). The mice were daily given intra-peri-infarct cortex microinjection of Tat-Synt1A251-265 and intraperitoneal injection of CNO during days 5–11 after stroke, and motor functions were detected on day 3 before stroke and day 12 after stroke (Figure S2E). The hM4Di/CNO system worked well in interneurons in the peri-infarct cortex (Figures S2F and S2G). Interestingly, the effect of Tat-Synt1A251-265 on stroke recovery was canceled by silencing the activities of interneurons (Figure S2H), suggesting the importance of cortical GABAergic neuron excitability for the beneficial role of Tat-Synt1A251-265 in promoting stroke recovery.

Uncoupling GAT-1-Synt1A enhances cortical oscillations, neural population activity, and plasticity

Electrical signatures of the brain have rhythmic patterns in neural activity, also known as oscillations. Beta oscillations (∼13–30 Hz) are commonly implicated in sensorimotor processing.^18^ Low gamma (30–50 Hz) oscillatory power in the peri-infarct cortex undergoes a long-lasting deficit after stroke.^31^ Synaptic GABAergic inhibition is involved in beta (12–29 Hz) and low-gamma (30–50 Hz) oscillations.^17^ To test whether GAT-1-Synt1A association affects cortical oscillations in the subacute phase of stroke, a recording electrode was implanted in the ipsilateral cortical layer 5 on day 4 after stroke, and Tat-Synt1A251-265 was infused into the peri-infarct cortex during days 5–11 after stroke (Figures S3A and S3B). We recorded local field potential (LFP) activity in vivo under stillness and movement conditions on day 12 after stroke. As shown in Figure 4A, mice with stroke displayed significantly reduced LFP power in beta (15–25 Hz) and low-gamma (30–50 Hz) oscillations, suggesting the disruption of rhythmic activity in the peri-infarct cortex by stroke. Tat-Synt1A251-265 reversed stroke-induced decreases in beta and low-gamma oscillations, suggesting that dissociating GAT-1-Synt1A enhances movement-related cortical oscillations in the subacute phase of stroke. Moreover, Tat-Synt1A251-265 significantly enhanced delta (1–3 Hz) and theta (4–12 Hz) oscillations also. All effects of Tat-Synt1A251-265 on cortical oscillations disappeared in GAT-1 CKO mice, suggesting the GAT-1-Synt1A association is necessary for the effects of Tat-Synt1A251-265. Therefore, dissociating GAT-1 from Synt1A facilitates movement-related cortical oscillation after stroke. Stroke survivors often have reduced movement vigor and impaired “speed and accuracy” tradeoff.^32^ Based on the growing evidence for the emergence of stable neural population activity patterns in the neural network with motor learning,^33^ we also recorded spiking activity of peri-infarct cortical neurons under stillness and movement conditions. We found that stroke significantly decreased movement-related spike firing rates but not those under stillness condition, and Tat-Synt1A251-265 reversed stroke-induced decrease in spike firing rates in WT but not in GAT-1 CKO mice (Figures 4B, 4C, S3C, and S3D). Movement-related LFP and spiking activity has been observed in the intact non-human primate M1 and human motor regions during reaching tasks.^34^ Thus, the enhancement of movement-related LFP power and spiking activity by dissociating GAT-1-Synt1A may play a critical role in stroke recovery.

Figure 4 Dissociating Synt1A from GAT-1 enhances cortical oscillations, neural population activity, and plasticity of projections(A) Left, percentage of change power spectrum of the peri-infarct cortex LFPs during movement. Solid lines represent the average and shaded areas indicate SEM. Right, percentage of change in LFP power of each waveband during movement. For delta, theta, beta, low-gamma, and high-gamma oscillations, one-way ANOVA followed by post hoc Scheffe test, F(3, 43) = 28.69, 38.78, 33.77, 12.38, and 2.08, respectively.(B) Raster plots showing neural population firing within the layer 5 of the peri-infarct cortex during movement. Bottom: bar graphs of firing events.(C) Percentage of change in firing events during movement. One-way ANOVA followed by post hoc Scheffe test, F(3, 18) = 12.61.(D) Right: representative traces and input-output curve peak amplitude of LFPs in the contralateral motor cortex. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test, F(3, 35) = 14.435, n = 9 slices from 6 animals for stroke + Tat-Synt1A251-265mut (GAT-1 WT) group, n = 10 slices from 6 to 7 animals for other groups. Middle: representative traces and input-output curve peak amplitude of LFPs in the ipsilateral striatum. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test, F(3, 36) = 20.519, n = 10 slices from 6 to 7 animals for each group. Right: representative traces and input-output curve peak amplitude of LFPs in the spinal cord. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test, F(3, 34) = 76.282, n = 9 slices from 6 to 7 animals for sham + Tat-Synt1A251-265mut (GAT-1 WT) and stroke + Tat-Synt1A251-265 (GAT-1 CKO) groups, n = 10 slices from 6 to 7 animals for other groups.See also Figure S3.

Glutamatergic projection neurons in the mammalian neocortex send axons to distant brain targets for processing multiple modalities of sensory information and controlling motor output, including callosal projection, corticostriatal projection, and corticospinal projection.^35^ To determine whether GAT-1-Synt1A coupling is implicated in the plasticity of these projections, we transduced the peri-infarct pyramidal neurons of GAT-1^flox/flox^ mice with an AAV encoding ChR2-eYFP fusion gene under control of the CaMKII promoter (Figure S3E). Three weeks later, ChR2-eYFP, identified as green fluorescence, was abundantly expressed in surviving neurons in the peri-infarct cortex, and ChR2-expressing peri-infarct motor cortical fibers abundantly projected to the contralateral motor cortex, the ipsilateral striatum, and the C5-C7 cervical spinal cord segment (Figure S3F). Action potentials were evoked in a ChR2-eYFP-expressing neuron in current-clamp mode by photostimuli pulses at 10 Hz, validating the functionality of expressed ChR2 (Figure S3G). Next, we recorded LFPs in the contralateral motor cortex, the ipsilateral striatum, and the C5-C7 cervical spinal cord segment by photostimulating the fibers projected from the peri-infarct cortex (Figure S3H) and found that Tat-Synt1A251-265 treatment during days 5–11 after stroke ameliorated stroke-induced reductions in LFP amplitude in the contralateral motor cortex, the ipsilateral striatum, and the spinal cord in WT mice but not in GAT-1 CKO mice (Figure 4D). Thus, dissociating GAT-1 from Synt1A facilitates the output synaptic strengthening of surviving neurons in the peri-infarct cortex.

Axonal sprouting is an attractive mechanism underlying the restoration of neuronal networks and functions after stroke.^8^^,^^36^ To determine whether GAT-1-Synt1A coupling is implicated in the structural plasticity, we microinjected the tract tracer biotinylated dextran amine (BDA) into the contralateral cortex of mice on day 34 after stroke and performed fluorescence staining of BDA in the contralateral motor cortex and the spinal cord on day 48 after stroke (Figures S3I and S3J). BDA fluorescence in the contralateral motor cortex was comparable between groups, suggesting that BDA injection was consistent (Figures S3K and S3L). Tat-Synt1A251-265 treatment during days 5–11 after stroke significantly increased the number of BDA^+^ fibers crossing the midline into the denervated spinal cord at the C7 level (Figures S3K and S3M), suggesting an increased axonal sprouting. Collectively, dissociating GAT-1 from Synt1A in the subacute phase of stroke improves functional and structural plasticity required for stroke recovery.

Design of small-molecule GAT-1-Synt1A blockers by targeting the N-terminal of GAT-1

To deeply understand how GAT-1 binds to Synt1A, we performed site-directed point mutations on Synt1A. We generated 6 lentiviral (LV) vectors that expressed WT Synt1A (LV-Synt1A) or Synt1A containing site mutations, i.e., Synt1AK252A/K253A (LV-Synt1Amut 1), Synt1AK252A/K256A (LV-Synt1Amut 2), Synt1AK253A/K256A (LV-Synt1Amut 3), Synt1AK252A/K253A/K256A (LV-Synt1Amut 4), or Synt1AT251A/K252A/K253A/V255A/K256A (LV-Synt1Amut 5). By the transfection of cultured neurons with these LV vectors, we found that LV-Synt1Amut 4 and LV-Synt1Amut 5 were more potent in reducing GAT-1-Synt1A complex than other LV vectors (Figure 5A), suggesting that residues K252, K253, and K256 in the Synt1AH3 are strictly necessary for GAT-1-Synt1A interaction.

Figure 5 Design of small-molecule GAT-1-Synt1A blockers by targeting the N-terminal of GAT-1(A) Synt1A-GAT-1 complex level in the cultured neurons after overexpressing Synt1A with site-directed point mutations at residues 251–256. One-way ANOVA followed by post hoc Scheffe test, F(5, 24) = 13.72, n = 5.(B) Synt1A-GAT-1 complex level after overexpressing the mutation of GAT-1K33A/K36A/K37A. Two-tailed t test, t(10) = 2.67, n = 6.(C) GABA uptake after overexpressing the mutation of GAT-1K33A/K36A/K37A. Two-tailed t test, t(8) = −2.62, n = 5.(D) Hypothesis: stroke induces Synt1A-GAT-1 association, which plays negative roles in the extracellular reuptake of GABA into neurons.(E) Upper, general structure of ZLQ series of compounds designed to uncouple Synt1A-GAT-1; lower, structure of ZLQ-3.(F) Molecular docking of ZLQ-3 to the N-terminal of GAT-1.See also Figure S4.

In addition, there are three lysine residues (K33, K36, and K37) in the N-terminal tail of GAT-1,^23^ and the residues 206–250 of Synt1AH3 are rich in acidic amino acid residues.^28^ To determine whether K33, K36, and K37 of GAT-1 are implicated in the binding of GAT-1 to Synt1A through electrostatic interaction, we generated an LV vector that expresses GAT-1 containing K33A, K36A, and K37A mutations (LV-GAT-1mut). We transfected cultured neurons with LV-GAT-1mut and found that, compared with WT GAT-1, LV-GAT-1mut caused a slight decrease in the GAT-1-Synt1A complex (Figure 5B) and increased the uptake of ^3^H-GABA (Figure 5C). Thus, residues K33, K36, and K37 of GAT-1 may also contribute to a certain extent to GAT-1-Synt1A interaction. Based on reports and our findings earlier,^23^^,^^28^^,^^37^ we supposed a molecular mechanism underlying how GAT-1 binds to Synt1A and how this interaction affects the function of GAT-1 (Figure 5D).

Based on the supposed molecular mechanism of binding the GAT-1 to the Synt1AH3, compounds that can bind well to the N-tailDDD of GAT-1 or to Synt1A251-256 may disrupt GAT-1-Synt1A interaction. Usually basic amino acid-rich peptides easily cross biological barriers,^38^ and targeting Synt1AH3 may affect not only GAT-1-Synt1A interaction but also protein-protein interaction required for the assembly and stability of the SNARE complex^24^^,^^39^; we thus chose the N-tailDDD of GAT-1 as target. Accordingly, we designed a general chemical structure rich in basic amino acids and unnatural amino acids and with a certain degree of hydrophobicity (Figure 5E). We synthesized a series of small-molecule compounds and detected their effects on the GAT-1-Synt1A interaction (Figure S4A). The structure-activity relationship of these compounds is shown in Figure S4B. Among them, ZLQ-3 had the most potent GAT-1-Synt1A uncoupling activity, but did not affect the stability and trafficking of GAT-1 (Figures 5E and S4B–S4D). Advantageous structural characteristics of GAT-1 are shown in Figure 5F: docking of ZLQ-3 in the N-terminal of GAT-1 indicated that a salt bridge (distance, 2.6 Å) between the P1 Arg side chain of ZLQ-3 and the side chain of Asp43, in addition to a hydrogen bond (H-bond) between the guanidine of the P1 Arg and the carboxyl group of Asp43, provided much favorable binding interactions; the side chain of the P4 Lys of ZLQ-3 also formed a salt bridge interaction with Asp40 (distance, 5.9 Å); and the guanidine group of P5 arginine of ZLQ-3 could not only form a salt bridge with the carboxyl side chain of Asp45 at a distance of 2.8 Å, but also the hydrogen of the guanidine group could form two H-bonds with the oxygen atom of the carboxyl group of Asp. The hydrophobic side chains of Dmt (P2) and Cha (P3) of ZLQ-3 could prevent these salt bridges from the interference of solution. In addition, three H-bonds also were observed in the peptide backbone of ZLQ-3 and GAT-1. The H on the phenolic hydroxyl group of Dmt (P2) could also form an H-bond with the backbone of GAT-1.

To test whether ZLQ-3 facilitates stroke recovery, we delivered it or vehicle during days 5–11 after stroke into the peri-infarct cortex of GAT-1^flox/flox^ mice transfected with AAV-Cre-EGFP or AAV-EGFP on day 3 before stroke and performed motor functional test on day 7 before stroke and on days 4, 12, 19, 26, and 33 after stroke (Figures S4E and S4F). The microinjection of ZLQ-3 reversed stroke-induced GAT-1-Synt1A interaction and significantly attenuated the impairment of motor functions caused by stroke in WT but not in GAT-1 CKO mice (Figures S4G and S4H), indicating that targeting GAT-1-Synt1A is necessary for the effect of ZLQ-3 on stroke recovery.

ZLQ-3-1, a glycosylated derivative of ZLQ-3, promotes stroke recovery

It has been known that glycosylation of peptides can reduce their susceptibility to enzymatic degradation and increases membrane penetration in the blood-brain barrier.^40^ We thus performed a glycosylation modification on ZLQ-3 and obtained a ZLQ-3 derivative, ZLQ-3-1 (Figure 6A). In the in vitro cultured neurons, ZLQ-3-1 significantly increased ^3^H-GABA uptake (Figure 6B), suggesting an upregulated GAT-1 function. ZLQ-3-1 at dose of 2.5, 5, 10, or 20 mg/kg was intraperitoneally injected during days 5–11 after stroke, and motor functional test was performed on day 7 before stroke and day 12 after stroke (Figure S5A). As shown in Figures S5B–S5D, ZLQ-3-1 attenuated the impairment of motor functions dose dependently.

Figure 6 Small-molecule Synt1A-GAT-1 blockers promote stroke recovery(A) Structure of ZLQ-3-1.(B) GABA uptake after ZLQ-3-1 treatment in the cultured neurons. Two-tailed t test, t(8) = −4.02, n = 5.(C) Experimental design for (D–K).(D) The time course of concentrations of ZLQ-3-1 in the brain tissue.(E) Left: coIP showing Synt1A-GAT-1 complexes in the peri-infarct cortex. One-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 26.21. Right: coIP showing Synt1A-GAT-1 complexes in the hippocampus. One-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 54.09.(F) The concentrations of extracellular GABA in cultured neurons. One-way ANOVA followed by post hoc Scheffe test, F(2, 12) = 23.49, n = 5.(G) Left: foot faults of the left forelimb in the grid-walking task. Middle: forelimb symmetry in the cylinder task. Right: forelimb sticky-tape ratio in modified sticky-tape test. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test. Left: F(2, 36) = 514.555. Middle: F(2, 36) = 242.291. Right: F(2, 36) = 131.507.(H) Escape latency to the platform during the training trails in a Morris water maze. Two-way repeated-measures ANOVA followed by post hoc Bonferroni test, F(2, 36) = 19.383.(I) Target (platform) entries in the probe test. One-way ANOVA followed by post hoc Scheffe test, F(2, 36) = 9.01.(J) Time spent in target quadrant in the probe test. One-way ANOVA followed by post hoc Scheffe test, F(2, 36) = 15.58.(K) Mean swimming speed of rats. One-way ANOVA followed by post hoc Scheffe test, F(2, 36) = 0.04.See also Figures S5 and S6.

Intranasal delivery of peptides is promising in treating many neurological disorders, because it can enter the brain directly and rapidly and limit side effects associated with peripheral administration.^41^ Moreover, because the treatment of patients with stroke during the repair phase is often done outside the hospital, administration of drugs through injection may be limited. We thus carried out several experiments to test the effects of ZLQ-3-1 when intranasally delivered. In the mice subjected to photothrombotic stroke, ZLQ-3-1 intranasally given during days 5–11 after stroke reversed stroke-induced GAT-1-Synt1A interaction, but had no effect on the expression of GAT-1 after stroke (Figures S5G and S5H). To test whether the binding of ZLQ-3-1 to GAT-1 affects the function of GAT-1, we incubated the cultured HEK293 cells overexpressing GAT-1 with ZLQ-3-1 and found that ZLQ-3-1 did not directly affect the reuptake of GABA by GAT-1 (Figure S5I). To investigate whether ZLQ-3 glycosylation modification shows better effects in vivo, ZLQ-3-1 or ZLQ-3 at dose of 5 mg/kg was intranasally delivered during days 5–11 after stroke and motor functional test was performed on day 7 before stroke and day 12 after stroke. As shown in Figure S5J, although both ZLQ-3-1 and ZLQ-3 attenuated the impairment of motor functions, ZLQ-3-1-treated mice displayed a better motor functional recovery than ZLQ-3-treated mice did, suggesting the importance of glycosylation modification. Moreover, treatment with ZLQ-3-1 during days 5–11 after stroke showed long-lasting motor functional recovery effects in WT but not in GAT-1 CKO mice (Figure S5K). Next, we detected the effects of ZLQ-3-1 on stroke recovery in the rats subjected to transient middle cerebral artery occlusion (tMCAO), a classical stroke model that leads to both motor functional deficits and cognitive impairments. Thus, we used grid-walking task, cylinder task, and modified sticky-tape test to assess motor functions and Morris water maze task to assess cognitive functions after tMCAO. ZLQ-3-1 was intranasally given during days 5–11 after stroke; foot faults in the grid-walking task, forelimb symmetry in the cylinder task, and forelimb sticky-tape ratio in modified sticky-tape test were performed on day 7 before stroke and on days 4, 12, 19, and 26 after stroke, and Morris water maze task was performed during days 13–18 after stroke (Figure 6C). The intranasal delivery of ZLQ-3-1 had a rapid and long-lasting brain exposure of ZLQ-3-1 (Figure 6D) and reversed stroke-induced GAT-1-Synt1A interaction in the peri-infarct cortex and hippocampus (Figure 6E) and increase in [GABA]o in the peri-infarct cortex (Figure 6F). More importantly, the intranasal delivery of ZLQ-3-1 significantly improved somatosensory and motor functions, as indicated by decreased foot faults, forelimb symmetry index, and increased forelimb sticky-tape ratio (Figure 6G), and reversed stroke-induced learning and memory deficits, as indicated by significantly decreased latency (time to reach the platform) and increased target entries and time spent in target quadrant (Figures 6H–6J). The intranasal ZLQ-3-1 administration did not affect swimming speed (Figure 6K). Furthermore, the intranasal administration of ZLQ-3-1 reversed stroke-induced increases in extracellular GABA and tonic inhibition and increased inhibitory synaptic transmission, and prevented stroke-induced decline in cortical excitability in WT but not in GAT-1 CKO mice (Figures S6A–S6G). In addition, ZLQ-3-1 did not change infarct volume (Figures S6H and S6I).

Discussion

The subacute phase of stroke is a critical period for the damaged brain repairing. During this phase, chronically elevated ambient GABA level activates extrasynaptic GABAARs and thereby antagonizes the neuronal excitability required for stroke recovery.^8^^,^^12^^,^^13^ Over the past decade or so, the α5-GABAAR has received great attention due to its extrasynaptic location.^8^^,^^12^^,^^15^ However, the clinical trial of α5-GABAAR antagonist in stroke recovery failed.^15^ In fact, α5-GABAAR is also localized at GABAergic synapses in the hippocampus and cerebral cortex.^42^ Moreover, the subtype selectivity of current α5-GABAAR antagonist is relatively weak, such as S44819, a representative selective α5-GABAAR antagonist (IC50: α5 vs. α1, 585 nM vs. 2,378 nM).^43^ It is particularly worth emphasizing that a substantial decrease of synaptic GABAergic inhibition occurs due to reduced expression of GABAARs, and enhancing synaptic GABAergic inhibition promotes stroke recovery.^20^^,^^44^^,^^45^^,^^46^ Thus, both blocking extrasynaptic GABAARs and activating synaptic GABAARs are beneficial for stroke recovery. We discovered that dissociating GAT-1 from Synt1A in the subacute phase of stroke reduced extrasynaptic GABA inhibition by increasing the reuptake of GABA into neurons while potentiating synaptic GABAergic inhibition by facilitating the assembly of the SNARE complex. Therefore, targeting GAT-1-Synt1A has the effect of killing two birds with one stone. ZLQ-3, a small-molecule GAT-1-Synt1A blocker we designed by exploring the molecular mechanism of the binding of GAT-1 to Synt1A, promoted stroke recovery when used in the subacute phase. To increase brain exposure and patient compliance, we designed ZLQ-3-1, a ZLQ-3 derivative, by glycosylation of ZLQ-3. Intranasal administration of ZLQ-3-1 in the subacute phase had high brain exposure, substantially improved somatosensory and motor functions, and reversed stroke-induced learning and memory deficits. Moreover, ZLQ-3-1 did not affect locomotor activity and conditioned place preference (Figures S6J–S6M), suggesting it has no effect on general activity and addiction.

It is an interesting question how stroke induces the interaction between GAT-1 and Synt1A. Our data and other studies suggest that stroke-induced NO production may be a key inducer; the causality is supported by three (1) our previous studies showed that sustained and significant production of NO caused by increased coupling between nNOS and PSD-95 is a key pathological event during the subacute phase of stroke,^25^^,^^26^ (2) it was previously reported that NO promotes the translocation of Synt1A and GAT-1 from the cytoplasm to the cell surface, especially the translocation of Synt1A, thereby increasing the interaction between Synt1A and GAT-1 and reducing ^3^H-GABA uptake,^27^ and more importantly, (3) we treated the cultured neurons with GSNO, an NO donor, and found that the treatment significantly increased membrane contents of GAT-1 and Synt1A, but had no effects on the total levels of GAT-1 and Synt1A (Figures S1E and S1F).

Potentiation of GAT-1 function can reduce extracellular GABA and thereby promote stroke recovery through enhancing neuronal excitability.^13^^,^^21^^,^^22^ Unfortunately, no GAT-1 activator is currently available. The function of GAT-1 is diminished by GAT-1-Synt1A interaction.^47^ We discovered that the GAT-1-Synt1A blocker is a GAT-1 activator and increases cortical excitability by reducing extracellular GABA in the subacute phase of stroke. GAT-1 is located at GABAergic axon terminals.^48^ In GABAergic neurons, the assembly of SNARE complex mediating neurotransmitter release should regulate synaptic GABAergic signaling directly.^24^^,^^29^ Yes, we found that the GAT-1-Synt1A blocker facilitates the assembly of the SNARE complex and thereby GABAergic synaptic transmission as indicated by increased mIPSCs and sIPSCs. GABAergic inhibition is related to beta and low-gamma oscillations implicated in sensorimotor processing and stroke recovery.^17^^,^^18^^,^^19^^,^^31^ Dissociating GAT-1 from Synt1A reversed stroke-induced decreases in the movement-related beta and low-gamma oscillations and neural population activity. The SNARE-mediated neurotransmitter release is activity dependent.^49^ Consistently, we showed that the activity of GABAergic neurons is critical for the effects of GAT-1-Synt1A blocker on stroke recovery. Theta oscillations (4–8 Hz) and delta (1–3 Hz)-modulated cortical alpha oscillations (9–12 Hz) are implicated in learning and memory.^50^^,^^51^ Dissociating GAT-1 from Synt1A significantly increases theta and delta oscillations. Moreover, enhanced plasticity leads to the sprouting of new axons, the formation of new synapses, and the remapping of sensory-motor functions during the subacute phase of stroke and is associated with motor recovery.^8^ Dissociating GAT-1 from Synt1A also enhances the plasticity of projections from the peri-infarct cortex to the contralateral motor cortex, the ipsilateral striatum, and the spinal cord and increases axonal sprouting. Therefore, enhanced network excitability, synaptic GABAergic inhibition, and plasticity collectively interpret the effects of GAT-1-Synt1A blockers on stroke recovery.

The GAT-1-Synt1A interaction is mediated by residues 30–54 in the N-tail of GAT-1, in which, residues D40, D43, and D45 are crucial.^23^ However, how these residues bind to Synt1A has yet to be determined. We discovered that residues K252, K253, and K256 in the Synt1AH3 are necessary for the binding of GAT-1. Moreover, residues K33, K36, and K37 in the N-terminal tail of GAT-1 also participate in the regulation of GAT-1-Synt1A association, possibly targeting to the acidic amino acid residues of Synt1A. The role of residues K33, K36, and K37 may be auxiliary, and the interaction between D40, D43, and D45 in GAT-1 and K252, K253, and K256 in the Synt1AH3 may induce the binding of K33, K36, and K37 of GAT-1 to Synt1A, which needs to be demonstrated in the future. It has been assumed that the IL-4 domain of GAT-1 serves as a barrier for transport; the binding of residues D40, D43, and D45 in the N-tail of GAT-1 to arginine residues in the IL-4 removes this barrier and thereby GAT-1 begins to work.^37^ If so, simply dissociating the N-tail from IL-4 by occupying residues D40, D43, and D45 will cause GAT-1 dysfunction. However, the fact that occupying the N-tail by Synt1AH3 did not cause GAT-1 dysfunction does not support this assumption.^23^ Based on these reports and our data,^23^^,^^37^ we modified this assumption (Figure 5D): the IL-4 of GAT-1 serves as a barrier for GABA reuptake; electrostatic repulsion between residues K33, K36, and K37 in the N-tail and arginine residues in the IL-4 of GAT-1 can push the IL-4 barrier away and thereby GAT-1 functions. When peptide Synt1AH3 binds to the N-tail of GAT-1, residues D40, D43, and D45 of GAT-1 are occupied by residues K252, K253, and K256 in the Synt1AH3; free K33, K36, and K37 in the N-tail of GAT-1 will push IL-4 barrier away, which requires more effective experiments to confirm.

In addition, Kv2.1 potassium channels bind to Synt1A and mediate cell death-promoting cytoplasmic K^+^ loss in cortical neurons.^52^ We showed here that the GAT-1-Synt1A blocker did not affect Kv2.1-Synt1A interaction in the peri-infarct cortex (Figure S6N).

In sum, we discovered that dissociating GAT-1-Synt1A enhances network excitability through increasing GABA reuptake, while it facilitates synaptic GABAergic inhibition and cortical oscillations, and therefore promotes stroke recovery, indicating that GAT-1-Synt1A is a promising target for developing stroke recovery-promoting agents. Importantly, the GAT-1-Synt1A blocker ZLQ-3-1 we designed, when used in the subacute phase, substantially promotes stroke recovery (Figure S6O).

Limitations of the study

Although stroke may promote GAT-1-Synt1A coupling via increased NO production during the subacute phase, it cannot be ruled out that other mechanisms may contribute to increased GAT-1-Synt1A association. Our study showed that ZLQ-3-1 promotes stroke recovery through reducing stroke-induced GAT-1-Synt1A interaction in rodents. However, the brain tissue of rodents and humans differs greatly; therefore, the efficacy of ZLQ-3-1 on stroke recovery needs to be further confirmed in non-human primate stroke models. In addition, although we update the molecular mechanism underlying the interaction between GAT-1 and Synt1A, more effective experiments, i.e., structure biology, are needed to further confirm how GAT-1 binds to Synt1A.

Resource availability

Lead contact

Further information and requests for sources should be directed to, and will be fulfilled by the lead contact, Dr. Dongya Zhu (dyzhu@njmu.edu.cn).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement.

Data and code availability

• Original western blot images are available at Mendeley Data: https://doi.org/10.17632/bgnb8gxdz3.1.
• •This paper does not report original code.
• Any additional information required to reanalyze the data in this paper is available from the lead contact upon request.

Acknowledgments

This research was supported by grants from National Natural Science Foundation of China (82090042, 82171368, 82373728, and 82171293), the National Key Research and Development Program of China (2021YFA1101803), the Natural Science Foundation of Jiangsu Province (BK20211255 and BK20231266), and by the Collaborative Innovation Center for Cardiovascular Disease Translational Medicine.

Author contributions

D.-Y.Z. and Y.-H.L. conceived the ideas. D.-Y.Z., Y.-H.L., and Y.-J.Q. prepared the article. Y.-J.Q. and T.-Y.L. designed and synthesized compounds. Y.-H.L., F.W., L.L., X.-M.X., M.-Y.C., Y.-L.H., C.-J.Z., H.-Y.W., and L.C. performed the experiments and acquired the data. F.G. assisted with in vivo electrophysiology. L.-J.Z. performed SNARE complex experiment. C.-X.L. contributed unpublished analytic tools. D.-Y.Z. supervised the research.

Declaration of interests

D.-Y.Z., Y.-H.L., Y.-J.Q., and T.-Y.L. are inventors on the following PCT on ZLQ series compounds filed on 15/08/2024, application no. PCT/CN2024/112442. Y.-H.L., Y.-J.Q., and T.-Y.L. are inventors on the following Chinese patent on ZLQ series compounds publication no. CN117384253A.

STAR★Methods

Key resources table

Experimental model and subject details

Animals

GAT-1^flox/flox^ mice (C57BL6/J background), exon 6 with loxP recombination sites, were generated by CRISPR genome editing technology (CRISPR-Cas9 system) at Shanghai Model Organisms (Shanghai, China). Wild-type male adult (6–7 weeks) C57BL6/J mice were purchased from GemPharmatech Co. Ltd (Nanjing, China). Male adult (9–10 weeks) Sprague-Dawley rats were purchased from Slac Laboratory Animal (Shanghai, China). Animals were maintained under standard laboratory conditions (20 ± 2°C, 60% humidity, 12 h light/dark cycle, food and water ad libitum) at the Model Animal Research Center of Nanjing Medical University (Nanjing, China). Every effort was made to minimize the number of animals used and their suffering. All procedures concerning animal care and treatment were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee of Nanjing Medical University (IACUC-1812012) and complied with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines.

Cell cultures

Primary cortical neurons were isolated from E16 mice using the method described previously.^22^ Primary neurons were cultured on dishes coated with poly-L-ornithine in Neurobasal medium (Invitrogen) containing 2% B27 supplement. Primary astrocytes were prepared from postnatal day 1 mice. Briefly, primary astrocytes were cultured on poly-L-ornithine coated dishes in DMEM/F12 medium containing 10% FBS (Gibco). Primary microglia were collected from mixed cultures harvested from neonatal mouse cortex. Briefly, mature microglia were identified as semi-adhesive cells and collected by gentle shaking, centrifuging, resuspending in MEM medium containing 3% FBS (Gibco).

HEK293 cells were cultured in MEM medium containing 10% FBS (Gibco) and transfected with LV-GAT-1-EGFP alone or co-transfected with LV-GAT-1-EGFP and LV-Synt1A-EGFP. Cells were used for further experiments 5 days after viral infection.

All cultures were maintained in an incubator with a humidified atmosphere of 95% air and 5% CO2 at 37°C.

Method details

Drugs and peptides

Clozapine-N-oxide (CNO, cat #4936) and bicuculline methiodide (cat #0109) are products of Tocris Bioscience, GSNO (cat #E4443) is a product of Selleck, γ-aminobutyric acid (GABA, cat #03835) is purchased from Sigma-Aldrich, and ^3^H-GABA (cat #3087709) is a product of PerkinElmer. Tat-Synt1A251-256, Tat-Synt1A255-260, Tat-Synt1A260-265, Tat-Synt1A251-265, and Tat-Synt1A251-265mut (with K251A, K252A, K253A, K253A and K256A mutations) were produced by GL Biochem Co., Ltd (Shanghai, China). All drugs used in electrophysiological experiments were supplied by Tocris Bioscience or Sigma-Aldrich.

Photothrombotic stroke model

Focal cortical ischemia was induced in male adult mice through photothrombosis of the motor cortex using a well-established procedure.^13^ The mouse was anesthetized with 2% isoflurane (RWD Life Science, cat #R510-22) and fixed in a stereotaxic apparatus (DKI 902, David Kopf Instruments). The skull was exposed after disinfection. A cold light source (Z-LITE-Z, World Precision Instruments) attached to a custom-made opaque template, giving a 2 mm diameter spot, was positioned 1.5 mm lateral from the bregma. Five minutes after an i.p. injection of rose bengal (100 mg/kg; Sigma-Aldrich, cat #330000), the brain was exposed to 12000 lux illumination through the intact skull for 15 min. Body temperature was maintained at 37 ± 0.5°C with a thermostatically controlled infrared lamp throughout the surgery. Following illumination, the skin was surgically glued and the mice were monitored until recovery from anesthesia. Control mice received the same procedure of surgery and the same dose of rose bengal but without illumination.

Middle cerebral artery occlusion (MCAO) stroke model

Transient focal ischemia was induced in male adult rats using intraluminal filament method as described previously.^13^^,^^53^^,^^54^ Rats were anesthetized with 2% isoflurane (RWD Life Science, cat #R510-22). A 4-0 surgical nylon monofilament with rounded tip was introduced from the external carotid artery into the internal carotid artery and advanced to the branching point of the middle cerebral artery (MCA). A laser Doppler probe (Moor Instruments) was used to monitor regional cerebral blood flow during MCAO. Only rats with a blood flow reduction to less than 25% of baseline were included. Two hours after MCA occlusion, the monofilament suture was gently withdrawn for reperfusion. Sham-operated rats receive the same procedure with MCAO rats except that the occluding filament was inserted only 7 mm above the carotid bifurcation. Body temperature was maintained at 37 ± 0.5°C with a thermostatically controlled infrared lamp throughout the surgery. Following surgery, the rats were monitored until recovery from anesthesia.

Recombinant virus production and their stereotaxic microinjection

The recombinant lentivirus (LV) that expressed EGFP (LV-EGFP), wild-type Synt1A (LV-Synt1A-EGFP) or Synt1A containing site mutations, including Synt1AK252A/K253A (LV-Synt1Amut 1-EGFP), Synt1AK252A/K256A (LV-Synt1Amut 2-EGFP), Synt1AK253A/K256A (LV-Synt1Amut 3-EGFP), Synt1AK252A/K253A/K256A (LV-Synt1Amut 4-EGFP) and Synt1AT251A/K252A/K253A/V255A/K256A (LV-Synt1Amut 5-EGFP), and the LV that expressed wild-type GAT-1 (LV-GAT-1-EGFP) or GAT-1 containing K33A, K36A and K37A mutations (LV-GAT-1mut-EGFP), were purchased from GeneChem Co., Ltd (Shanghai, China). The titer of lentivirus was 1.5×10^9^ virus particles/mL. The adeno-associated virus (AAV) that expressed Synt1A251-265 (AAV-Synt1A251-265-EGFP, 1×10^13^ virus particles/mL) or mutated Synt1A251-265 (with T251A, K252A, K253A, V255A and K256A mutations, AAV-Synt1A251-265mut-EGFP, 1×10^13^ virus particles/mL) were produced by GeneChem Co., Ltd (Shanghai, China). AAV-VGAT-hM4D (Gi)-mCherry (3×10^12^ virus particles/mL) and AAV-CaMKIIα-GCaMP6s (5×10^12^ virus particles/mL) were purchased from BrainVTA (Wuhan, China). AAV-CAG-3Flag (2×10^13^ virus particles/mL), AAV-CAG-Cre-3Flag (2×10^13^ virus particles/mL) and AAV-CaMKIIα-hChR2 (E123A)-eYFP (5×10^12^ virus particles/mL) was purchased from GeneChem Co., Ltd (Shanghai, China). AAV-CAG-EGFP (7×10^12^ virus particles/mL) and AAV-CAG-EGFP-T2A-Cre (7×10^12^ virus particles/mL) were purchased from Obio Technology (Shanghai, China). Viruses were stored at −80°C freezer until the day of infusion.

Mice were anesthetized with isoflurane (2%, RWD Life Science, cat #R510-22) before stereotaxic injection of virus solution. After making an incision to the midline of the scalp, a small craniotomy was made using a micro-drill with 0.5-mm burr. Glass capillaries with the tip size of 30 μm in outer diameter were loaded with virus and injected at the following coordinates (anteroposterior, 0 mm; mediolateral, −1.5 mm; dorsoventral, −1.3 mm; relative to the bregma). We injected 500 nL of virus-containing solution into the motor cortex at 1 nL/s and the needle was not removed until 10 min after the end of infusion to assure even distribution of the virus. The injection sites of virus solutions were examined after experiments by the expression of GFP, YFP, or mCherry.

Cannula implantation and drug microinjection

Immediately after photothrombotic stroke was induced, stainless-steel guide cannulae (26-gauge, 3.5 mm, RWD Life Science) were implanted into the core of the infarction, and the tips of cannulae were at the following coordinates (anteroposterior, 0 mm; mediolateral, −1.5 mm; dorsoventral, −1.0 mm; relative to the bregma). The cannulae were fixed to the skull with adhesive luting cement and acrylic dental cement. Following surgery, stainless-steel obturators were inserted into the guide cannulae to avoid obstruction until microinjection was made. During drug infusion, mice were briefly head restrained, while stainless-steel obturators were removed and injection cannulae (33-gauge, 4.0 mm, RWD Life Science) were inserted into the guide cannulae. Injection cannulae protruded 0.5 mm from the tips of guide cannulae, thus penetrating into the peri-infarct cortex. A total volume of 2 μL drug solutions was slowly infused at a flow rate of 0.2 μL/min. Following infusion, the injection cannulae were left in place for an additional 5 min to allow the solution to diffuse into the peri-infarct cortex. The stainless-steel obturators were subsequently reinserted into the guide cannulae.

Behavioral assessment

Grid-walking task

Grid-walking task was performed using a well-established procedure.^13^ Each animal was placed individually on top of a 50 cm high (for mice) or 60 cm high (for rats) wire grid and allowed to freely walk for 5 min. A wire grid (length, 32 cm; width, 20 cm) with 12 mm square wire meshes was used for mouse grid-walking task. A wire grid (length, 60 cm; width, 60 cm) with 3 cm square wire meshes was used for rat grid-walking task. Stepping errors (foot faults) were captured using a video camera (C1000e, Logitech) positioned beneath the wire grid and analyzed offline in slow motion (1/5th real time speed) by a rater blinded to conditions. If a step did not provide support and the foot went through the grid or the animal was resting with the grid at the level of the wrist, it was considered a foot fault. Per cent foot faults were calculated number of foot faults/number of total steps × 100. Differences between animals and trials in the degree of locomotion were excluded by calculating the ratio between foot faults and total steps taken.

Spontaneous forelimb task (cylinder task)

Cylinder task was performed using a well-established procedure.^13^ Each animal was placed individually in a clear plexiglass cylinder (for mice, 15 cm in height with a diameter of 10 cm; for rats, 30 cm in height with a diameter of 20 cm) and allowed to freely explore for 5 min. The animal would support its weight with either one or both of its forelimbs on the side of the cylinder wall during spontaneous vertical exploration. The free exploration of the animal was captured using a video camera (C1000e, Logitech) positioned in front of a cylinder and analyzed offline in slow motion (1/5th real time speed) by an investigator blinded to group allocation. The time (s) during each rear that the animal spent on either its right, left or bilateral forelimbs, was calculated. Only rears in which both forelimbs could be clearly seen were timed. The percentage of time spent on each limb was calculated and these data were used to derive an asymmetry index as (per cent ipsilateral use) - (per cent contralateral use).

Modified sticky-tape test

Modified sticky-tape test was performed to assess somatosensory dysfunction following cerebral ischemia in rats.^55^ The rats were handled for 30 min/day to reduce the effects of stress on task performance on three consecutive days before formal testing. The forepaw was wrapped with a paper tape (length, 3.0 cm; width, 1.0 cm) and the time attending to the tape was collected. Each behavioral test was composed of five trials, each trial run consisting of 30 s. After each trial, the tape was removed and rats received a 5 min resting time. The order of placement of the tape (right or left) was alternated between each animal and each trial. Ratio of interest of each trial was calculated the number of seconds the animal attended to the stimulus/30. The best two trials obtained were averaged to obtain the ratio of interest on the testing day. Modified sticky-tape ratio was calculated the ratio of interest in left forepaw/the ratio of interest in right forepaw.

Morris water maze

Morris water maze was performed to evaluate spatial cognitive functions as described previously.^13^^,^^56^ The maze (diameter, 180 cm) filled with opacified water was drained every day. Rats were allowed to habituate the water maze 1 day before the experiment. During the learning phase, rats were allowed to freely swim for up to 60 s to escape to the hidden platform, which was fixed 1.2 cm beneath the water surface. Rats were allowed to remain or placed on the platform (if the rat could not find the platform within 60 s) for 10 s with prominent spatial cues displayed around the maze. One block of four trials was performed on each day for 5 consecutive days, and the data presented are the average of the four trials. In the memory phase at 24 h after the last learning trail, the platform was removed and the rats were tested for memory retention in a probe trial. The swimming activity of each rat was monitored using a video camera mounted overhead and was automatically recorded with a video tracking system (Digbehv4.1.7).

Conditioned place preference (CPP)

The CPP test was recorded in a 2-chamber apparatus (A and B) and analyzed using the Topscan software (Clever Sys, Inc.). The CPP procedure consisted of three the preconditioning test (pretest), conditioning (CPP training), and postconditioning test (test). During CPP training, for ZLQ-3-1 group, mouse received ZLQ-3-1 injection in chamber A in the morning and received vehicle injection in chamber B in the afternoon, and was allowed free to move around for 45 min in the paired chamber; for vehicle group, except for the mouse receiving only vehicle injection in the morning and afternoon, the other experimental procedures are the same as that in ZLQ-3-1 group. During the pretest and test, mice were allowed to freely access two chambers for 15 min. After each CPP pretest, training and test, 75% alcohol was used to eliminate the residual smell of mice. The CPP score was calculated by the time spent in drug-paired chamber minus that in vehicle-paired chamber.

Open field test (OFT)

The OFT was performed to measure the activity of mice. Each mouse was placed individually in a plastic box (30 × 30 × 50 cm) and allowed to freely explore for 5 min. Their behaviors were recorded and analyzed with Topscan software (Clever Sys, Inc.). Total distance moved in the open field and speed showed the locomotor activities.

Fiber photometry

Following the microinjection of AAV-CaMKⅡ-GCaMP6s into the peri-infarct cortex at the following coordinates (anteroposterior, 0 mm; mediolateral, −2.0 mm; dorsoventral, −1.3 mm; relative to the bregma), an optical fibre (230 μm OD, 0.37 NA) coupled to a ceramic ferrule was implanted at the same coordinate and secured to the skull using dental cement. Calcium transients in the peri-infarct cortex were measured using commercial fiber photometry equipment (Thinker Tech Nanjing Biotech Ltd, Nanjing, China). Fluorescence signals were obtained by stimulating cells expressing GCaMP6s in the peri-infarct cortex with a 488 nm LED (20 μW at fiber tip). Fiber-photometric recording data were exported as MATLAB files, and fiber photometry-related behavioral data were analyzed using MATLAB. The fluorescence change values (ΔF/F) were calculated (F-F0)/(F0-Voffset), where F0 is the baseline fluorescence signal averaged over a 2 s time-window prior to a trigger event and Voffset is the fluorescence signal recorded before the cannula was connected to the optical fiber. ΔF/F values were presented as heatmaps or average plots with a shaded area indicating the SEM.

On testing days, each mouse was placed individually in a clear plexiglass cylinder (15 cm in height with a diameter of 10 cm) and allowed to freely explore for 10 min. The average ΔF/F during the mouse spontaneously rear to a standing position was calculated.

Cell membrane protein fractions extraction

The preparation of cell membrane protein fractions was performed using Transmembrane Protein Extraction Kit (Millipore, cat #71772-3) according to the manufacturer’s protocol. Briefly, cells or homogenized tissues were permeablized using Extraction Buffer 1 then centrifugated at 1000 ×g for 5 min to separate cytosol and membrane protein fractions. Phase partitioning resulted in the hydrophilic cytosol proteins layering at the top and the hydrophobic membrane proteins at the bottom. Integral membrane proteins were extracted by resuspending pellets (membrane proteins) in Extraction Buffer 2B. The isolated membrane protein fractions were used directly in Western blot. β-Actin, a cytosol protein, was measured to test the purity of the isolated membrane proteins. No β-Actin was detected, excluding the possibility of contamination by cytosol proteins.

Coimmunoprecipitation (Co-IP)

Cells were washed twice in PBS, and then lysed in pre-cooling RIPA buffer, containing 50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM EDTA-Na, 1% NP-40, 0.02% sodium azide, 0.1% SDS, 0.5% sodium deoxycholate, 1% PMSF, 1‰ aprotinin, 1‰ leupeptin, and 0.5‰ pepstatin A. Tissues were rapidly dissected over the ice box and then lysed in ice-cold RIPA buffer. After lysis for 30 min, samples were centrifugated at 12,000 ×g for 15 min at 4°C. The supernatant was collected and preincubated for 1 h at 4°C with protein G-magnetic beads (Millipore, cat #LSKMAGAG10). The target supernatant with proteins that adhered nonspecifically to the beads removed was collected for following Co-IP experiment. The capture antibody (mouse antibody to syntaxin1A, 100, cat #110111, RRID: AB_887848, Synaptic System) or normal mouse IgG (1:100, cat #12–371, RRID: AB_145840, Sigma-Aldrich) were added into the target supernatant together with protein G-magnetic beads (Millipore, cat #LSKMAGAG10) for incubation at 4°C overnight. Immune complexes were isolated by magnetism, washed 5 times with ice-cold PBS. Proteins were analyzed using western blot. Primary antibodies used rabbit antibody to GAT-1 (1:4000, cat #274102, RRID: AB_2620000, Synaptic System), rabbit antibody to syntaxin1A (1:4000, cat #110302, RRID: AB_887846, Synaptic System), rabbit antibody to VAMP-2 (1:2000, cat #104008, RRID: AB_2802152, Synaptic System), rabbit antibody to SNAP-25 (1:2000, cat #111002, RRID: AB_887790, Synaptic System), and rabbit antibody to Kv2.1 (1:2000, cat #231002, RRID: AB_2131650, Synaptic System).

Isolation of vesicles and immunoblotting

Synaptosomes were prepared as described previously with some modifications.^57^^,^^58^ Mice were decapitated, and the peri-infarct tissue was rapidly removed and homogenized in preparation buffer, containing 320 mM sucrose, 1 mM NaHCO3, 1 mM MgCl2, 0.5 mM CaCl2, using a homogenizer at 4°C. The lysate was centrifuged at 500 ×g for 2 min to collect the supernatant. The supernatant was then centrifuged at 10100 ×g for 10 min to collect the pellet. The pellet was resuspended in 0.32 M sucrose, and then layered onto 0.8 M sucrose. The synaptosomes were collected from the 0.8 M sucrose fraction after centrifuging at 9100 ×g for 15 min, and then centrifuged at 12000 ×g for 15 min and resuspended in SDS-lysis buffer. The protease inhibitors aprotinin, leupeptin, pepstatin and PMSF were added to all solutions. Synaptosome samples were processed for SDS-PAGE, transferred, probed with antibodies, and visualized with enhanced chemiluminescence as described below. The non-boiled SNARE complex divided by the boiled monomer represents the assembly of the SNARE complex. The SNARE monomer proteins were calculated by the non-boiled SNARE monomer proteins/the boiled SNARE monomer proteins.

Western blot analysis

Western blot analysis was performed as described in detail previously.^59^ Tissues were rapidly dissected over the ice box as previously described.^60^ The primary antibodies were as rabbit antibody to GAT-1 (1:4000, cat #274102, RRID: AB_2620000, Synaptic System), mouse antibody to syntaxin1A (1:4000, cat #110111, RRID: AB_887848, Synaptic System), rabbit antibody to VAMP-2 (1:2000, cat #104008, RRID: AB_2802152, Synaptic System), rabbit antibody to SNAP-25 (1:2000, cat #111002, RRID: AB_887790, Synaptic System), rabbit antibody to Na^+^-K^+^-ATPase (1:1000, cat #3010, RRID: AB_2060983, Cell Signaling Technology). Mouse anti-β-Actin (1:4000; Sigma-Aldrich, cat #A1978, RRID: AB_476692) or mouse anti-GAPDH (1:4000; Kangchen Biotech, cat #KC-5G4, RRID: AB_2493106) was used as internal reference. Secondary antibodies including HRP-linked goat-anti-rabbit and goat-anti-mouse antibodies were used for detection by enhanced chemiluminescence (Millipore, cat #WBKLS0500).

Immunostaining, imaging and analysis

The details of immunofluorescence for brain section have been described in detail previously.^26^ Mice were transcardially perfused with ice-cold saline followed by 4% PFA under deep anesthesia (2%, RWD Life Science, cat #R510-22). Brains were removed, post-fixed overnight and then coronal sections (40 μm) were cut on a vibratome (VT1200s, Leica). For CaMKⅡ immunostaining, sections were subjected to antigen retrieval with sodium citrate (pH 7.0) at 60°C for 2 h. The primary antibodies used mouse anti-CaMKⅡ (1:50; CST, cat #50049, RRID: AB_2721906). Secondary antibodies were as goat anti-mouse Alexa 488 (1:400; Jackson ImmunoResearch Laboratories, cat #115-545-003, RRID: AB_2338840). For the biocytin-injected slices streptavidin-Cy3 (1:1000, Sigma-Aldrich, cat #S6402) was added to the secondary antibody solution. Finally, DAPI (Sigma-Aldrich, cat #MBD0020) was used to counterstain the slices to label the nuclei. Images were captured with a confocal laser-scanning microscope (LSM700, Carl Zeiss) at identical settings for each of conditions.

Proximity ligation assay (PLA)

PLA was performed with Duolink in situ detection kit (Sigma) according to the manufacturer’s instructions. Mice were transcardially perfused with ice-cold saline followed by 4% PFA under deep anesthesia (2%, RWD Life Science, cat #R510-22). Brains were removed, post-fixed overnight and then coronal sections (40 μm) were cut on a vibratome (VT1200s, Leica). The brain sections were incubated with Duolink blocking solution at 37°C for 60 min. After blocking, the sections were incubated with primary antibodies overnight at 4°C at the following for GAT-1 (rabbit, 500, cat #274102, RRID: AB_2620000, Synaptic System); for syntaxin1A (mouse, 200, cat #110111, RRID: AB_887848, Synaptic System). After primary antibody incubation, PLA anti-rabbit minus (Sigma, cat #DUO92005) and PLA anti-mouse plus (Sigma, cat #DUO92001) probes were added and incubated at 37°C for 1 h. Two proximate hybridized oligonucleotides were joined into a closed circle under the catalysis effect of ligase. The DNA was then amplified at 37°C for 100 min. DAPI (Sigma-Aldrich, cat #MBD0020) was used to counterstain the slices to label the nuclei. Images were captured with a confocal laser-scanning microscope (LSM700, Carl Zeiss) at identical settings for each of conditions.

In vivo electrophysiological recording

Seven days after stroke, mice were anesthetized with isoflurane (2%, RWD Life Science, cat #R510-22) and placed in a stereotaxic apparatus (DKI 902, David Kopf Instruments). A craniotomy was performed to expose the peri-infarct sensory-motor cortex on the ipsilateral side of the brain. A 16-channel nichrome electrode array connected to a stereotaxic frame was inserted into the peri-infarct sensory-motor cortex, and the tip of the electrode was at the following coordinates (anteroposterior, 0 mm; mediolateral, −0.65 mm; dorsovental, −1.3 mm; relative to bregma). Five days after electrode implantation, multi-channel signals in vivo electrophysiological were recorded by combining the signals of the mouse in the open field. Electrophysiological recording signals of chronically implanted mouse were acquired by attaching the connector of microelectrode to head-stage preamplifier and then connected to Cerebus Neural Signal Processing System (Blackrock Microsystems, Inc., Salt Lake City, UT). Signals from each electrode were amplified by 1000 gain, sampled at 2 kHz, low-pass filtered with 250 Hz cutoff for local field potential (LFP) offline analysis, and sampled at 30 kHz, band-pass filtered between 250 and 5000 Hz for spike offline analysis, respectively.

LFP and spike analysis

Analysis of LFP and spike data was done using custom MATLAB (Mathworks Inc.) scripts. Briefly, signals sampled at 2 kHz were imported into MATLAB and LFP log power was calculated using the power spectral density output from the spectrogram function. The following frequency bands calculated under a fast Fourier transformation were delta band (1–3 Hz), theta band (4–12 Hz), beta band (15–25 Hz), low-gamma band (30–50 Hz), and high-gamma band (55–85 Hz). % Change in LFP power during movement was calculated ((LFP power during movement - LFP power during stillness)/LFP power during stillness) × 100. On the other hand, signals sampled at 30 kHz were imported into MATLAB and the 3 times standard deviation (SD) value of the filtered signal was set as the threshold to detect spikes multiunit activity. We then sorted the detected multi-unit activity using Valley Seeking clustering to achieve single-unit firing. Only clearly identifiable units with good waveforms and a high signal-to-noise ratio were used. The spike firing rate was calculated and the raster was drawn by timestamps of the sorted spikes. We also draw the spike firing time histograms computed by counting the number of spikes with sequential time bins (1 s). % Change in firing events during movement was calculated ((firing events during movement - firing events during stillness)/firing events during stillness) × 100.

Microdialysis

Microdialysis in the peri-infarct cortex was performed using a well-established procedure.^61^ The mouse was anesthetized with 2% isoflurane (RWD Life Science, cat #R510-22) and fixed in a stereotaxic apparatus (DKI 902, David Kopf Instruments). After exposing the skull and drilling a small craniotomy, a CMA7 microdialysis probe (CMA Microdialysis) was implanted into the peri-infarct cortex, and the tip of the probe was at the following coordinates (anteroposterior, 0 mm; mediolateral, −0.65 mm; dorsovental, −1.5 mm; relative to bregma). Probes were held in place snugly and prevented from rotating by using a tab. The probe was connected to a microperfusion pump (WPI) and perfused with a solution containing 10 mM glucose, 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 1.3 mM MgCl2, 1.3 mM NaH2PO4, 25 mM NaHCO3, 1.3 mM sodium ascorbate and 0.6 mM sodium pyruvate at a speed of 1 μL/min. Perfusates from the outlet end of the tubing were collected over 20 min intervals for 2 h. The second dialysate was used for analysis.

GABA measurements

GABA contents were measured using the GABA ELISA kit (Abcam; cat #AB287793) according to the manufacturer’s instructions. Briefly, the standards and samples were added into respective wells, and reaction was triggered by adding the activator solution. After incubation at 37°C for 15 min in dark, stop solution was added into respective wells. After adding stop slution for 10 min, absorbance O.D. at 450 nm was measured using microplate reader (Molecular Devices).

^3^H-GABA uptake assay

Cultured neurons were rinsed twice in 37°C HBSS and allowed to equilibrate in the final wash for 10 min. Neurons were then exchanged with drug-containing HBSS. After 1 h drug preincubation, ^3^H-GABA and unlabeled GABA was added to the final concentration of 10 nM and 30 μM to initiate the assay respectively. After incubation at 37°C for 1 h, GABA uptake assay was terminated by rapid three times ice-cold washes with 500 μL of HBSS, followed by solubilization in 200 μL of ice-cold 0.1 M NaOH. Aliquots were processed for scintillation counting using liquid scintillator (PerkinElmer) and protein quantity measurement using BCA kit (ThermoFisher; cat #A55860). The GABA uptake activity was calculated as fmol/min/mg protein. Data were from at least three separate experiments.

Design and synthesis of ZLQ series compounds

Three aspartic acid residues (D40, D43, D45) in the N-terminal tail of GAT-1 (GAT-1-N-tailDDD) is critical for the binding of GAT-1 to the Synt1AH3,^23^ and residues K252, K253 and K256 in the Synt1AH3 are strictly necessary for GAT-1-Synt1A interaction (Figure 5A), suggesting electrostatic interactions between Synt1AH3 and GAT-1-N-tailDDD. If GAT-1-Synt1A blockers are rich in basic amino acids, they may form salt bridge interactions with the GAT-1-N-tailDDD and then disrupt the GAT-1-Synt1A interaction. Therefore, we tried to introduce 2–3 basic amino acids into the GAT-1-Synt1A blockers. Because the effect of TKKAVK on GAT-1-Synt1A was much better than KARRKK (Figure 1F), two adjacent more hydrophobic amino acids were inserted into the basic amino acid sequence. Moreover, instability of peptides due to the in vivo rapid degradation by proteases is a major problem limiting their use. The incorporation of unnatural amino acids, including d-amino acids and modified amino acids, into peptides can improve their stability to proteolysis and alter the biological half-life.^62^^,^^63^^,^^64^ Based on above, we designed ZLQ series compounds (Figure 5E).

ZLQ series of compounds were synthesized by standard Fmoc-strategy solid-phase peptide synthesis method.^65^ The Rink amide resin (100–200 mesh, 0.40–0.80 mmol/g, 1% DVB), the coupling reagents (HBTU, HOBt and DIPEA), and Fmoc-protected amino acids (Fmoc-Arg(Pbf)-OH, Fmoc-D-Arg(Pbf)-OH, Fmoc-Lys(Boc)-OH, Fmoc-D-Lys(Boc)-OH, Fmoc-Dmt-OH (Fmoc-2,6-dimethyl-L-tyrosine), Fmoc-Cha-OH, Fmoc-D-Cha-OH, Fmoc-Ala-OH, Fmoc-D-Ala-OH, Fmoc-Val-OH, Fmoc-D-Val-OH, Fmoc-Phe-OH, Fmoc-Tyr(tBu)-OH), were purchased from GL Biochem CO., Ltd (Shanghai, China). The peptide sequence was assembled on the Rink Amide resin using Fmoc-D-Arg(Pbf)-OH, Fmoc-D-Lys(Boc)-OH, Fmoc-Cha-OH, Fmoc-Dmt-OH and Fmoc-D-Arg(Pbf)-OH with coupling and deprotection cycles with HBTU, DIPEA, HOBt, and 20% (v/v) piperidine in dry DMF, respectively. After complete peptide assembly, the deprotection of the side chains and final cleavage from the solid support were performed by treatment with TFA/TIPS (v/v = 19/1, 10 mL of the mixture/g of resin) for 2 h at 25°C, and the cleavage procedure was repeated twice. The crude ZLQ-3 was analyzed by HPLC and purified by semi-preparative RP-HPLC. After semi-preparation, the product was collected, freeze-dried and replaced with HCl to obtain ZLQ-3. ZLQ-3-1 was synthesized using the intermediate of ZLQ-3, which was D-Arg(Pbf)-Dmt-Cha-D-Lys(Boc)-D-Arg(Pbf)-Amide resin. The dried ZLQ-3 intermediate was added into a round-bottom flask containing 4 equivalents of D-glucose, with methanol as the solvent, to achieve a concentration of the ZLQ-3 intermediate of 0.18 mmol/mL. The mixture was heated under reflux overnight. After the reaction was complete, the resins were transferred to a peptide solid-phase synthesizer and washed sequentially with methanol and DMF. After drying, it was deprotected and cleaved three times with TFA/anisole (v/v = 95/5) for 2 h each time, precipitated with ether, and then filtered and dried to obtain the crude product of ZLQ-3-1. Other ZLQ series of compounds were also synthesized using the method of ZLQ-3 described above. These compounds were purified to greater than 98% purity, and characterized by mass-spectral analysis.

The mass spectral and HPLC data for the compounds ZLQ-1-19 and ZLQ-3-1 were as HPLC analysis HPLC Waters 2487–2695; 1.0 mg of the compound diluted to 1.0 mL with the mobile phase; injection 5.0 μL; unitary C18 column (4.6 × 250 mm, 5 μm); 30°C; UV/Visible detector, detection 220 nm, 1.0 mL min^−1^;

ZLQ-1: D-Arg-Dmt-Cha-Lys-D-Arg-NH2, MS (ESI^+^): calcd for C38H68N13O6 802.54 [M + H]^+^, found 803.06. HPLC: tR = 7.27 min mobile 0–20 min, 18–38% solvent A (0.1% TFA in CH3CN).

ZLQ-2: D-Arg-Dmt-D-Lys-Phe-NH2, MS (ESI^+^): calcd for C32H50N9O5 640.39 [M + H]^+^, found 640.74; HPLC: tR = 11.40 min mobile 0–20 min, 10–35% solvent A (0.1% TFA in CH3CN).

ZLQ-3: D-Arg-Dmt-Cha-D-Lys-D-Arg-NH2, MS (ESI^+^): calcd for C38H68N13O6 802.54 [M + H]^+^, found 803.35; HPLC: tR = 7.21 min mobile 0–20 min, 20–45% solvent A (0.1% TFA in CH3CN).

ZLQ-4: D-Arg-D-Arg-Dmt-Cha-Lys-NH2, MS (ESI^+^): calcd for C38H68N13O6 802.54 [M + H]^+^, found 802.56; HPLC: tR = 16.26 min mobile 0–20 min, 15–40% solvent A (0.1% TFA in CH3CN).

ZLQ-5: D-Arg-D-Arg-Cha-Dmt-Lys-NH2, MS (ESI^+^): calcd for C38H68N13O6 802.54 [M + H]^+^, found 803.18; HPLC: tR = 7.33 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-6: D-Arg-D-Arg-Ala-Val-D-Arg-NH2, MS (ESI^+^): calcd for C26H54N15O5 656.44 [M + H]^+^, found 656.69; HPLC: tR = 10.95 min mobile 0–20 min, 2–27% solvent A (0.1% TFA in CH3CN).

ZLQ-7: D-Arg-D-Arg-Ala-Val-D-Lys-NH2, MS (ESI^+^): calcd for C26H54N13O5 628.44 [M + H]^+^, found 628.72; HPLC: tR = 7.78 min mobile 0–20 min, 2–27% solvent A (0.1% TFA in CH3CN).

ZLQ-8: D-Arg-D-Arg-Ala-Val-Lys-NH2, MS (ESI^+^): calcd for C26H54N13O5 628.44 [M + H]^+^, found 628.84; HPLC: tR = 6.09 min mobile 0–20 min, 6–31% solvent A (0.1% TFA in CH3CN).

ZLQ-9: Arg-Arg-D-Ala-D-Val-Arg-NH2, MS (ESI^+^): calcd forC26H54N15O5 656.44 [M + H]^+^, 656.88; HPLC: tR = 4.82 min mobile 0–20 min, 10–70% solvent A (0.1% TFA in CH3CN).

ZLQ-10: Arg-Arg-Ala-Val-Arg-NH2, MS (ESI^+^): calcd forC26H54N15O5 656.44 [M + H]^+^, found 656.75; HPLC: tR = 10.82 min mobile 0–20 min, 2–27% solvent A (0.1% TFA in CH3CN).

ZLQ-11: D-Arg-Dmt-Cha-D-Lys-NH2, MS (ESI^+^): calcd for C32H56N9O5 646.44 [M + H]^+^, found 646.37; HPLC: tR = 9.25 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-12: D-Arg-Dmt-Cha-D-Arg-NH2, MS (ESI^+^): calcd for C32H56N11O5 674.45 [M + H]^+^, found 674.45; HPLC: tR = 9.92 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-13: D-Arg-Dmt-Cha-Lys-NH2, MS (ESI^+^): calcd for C32H56N9O5 646.44 [M + H]^+^, found 646.46; HPLC: tR = 9.34 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-14: D-Arg-Dmt-Phe-D-Lys-D-Arg-NH2, MS (ESI^+^): calcd for C38H62N13O6 796.49 [M + H]^+^, found 796.66; HPLC: tR = 10.21 min mobile 0–20 min, 12–37% solvent A (0.1% TFA in CH3CN).

ZLQ-15: D-Arg-Tyr-Cha-Lys-NH2, MS (ESI^+^): calcd for C30H52N9O5618.41 [M + H]^+^, found 618.37; HPLC: tR = 7.14 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-16: D-Arg-Phe-Cha-Lys-NH2, MS (ESI^+^): calcd for C30H52N9O4602.41 [M + H]^+^, found 602.34; HPLC: tR = 10.06 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-17: Arg-Dmt-Cha-Lys-NH2, MS (ESI^+^): calcd for C32H56N9O5 646.44 [M + H]^+^, found 646.49; HPLC: tR = 7.45 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-18: Arg-Dmt-D-Cha-Lys-NH2 MS (ESI^+^): calcd for C32H56N9O5 646.44 [M + H]^+^, found 646.36; HPLC: tR = 6.89 min mobile 0–20 min, 18–43% solvent A (0.1% TFA in CH3CN).

ZLQ-19: D-Arg-Dmt-Phe-Lys-NH2, MS (ESI^+^): calcd for C32H50N9O5 640.39 [M + H]^+^, found 640.41; HPLC: tR = 8.07 min mobile 0–20 min, 15–40% solvent A (0.1% TFA in CH3CN).

ZLQ-3-1: Glucose-D-Arg-Dmt-Cha-D-Lys-D-Arg-NH2 964, MS (ESI^+^): calcd for C44H78N13O11 964.59 [M + H]^+^, 964.81. HPLC: tR = 8.47 min mobile 0–20 min, 10–100% solvent A (0.1% TFA in CH3CN).

Molecular docking

The structure of mouse GAT-1 (AF-P31648-F1) was acquired from the AlphaFold Protein Structure Database (https://www.alphafold.ebi.ac.uk/). Prepared ZLQ series of compounds were docked into the α-fold structure of mouse GAT-1 using Maestro 13.5 (Schrödinger software 2023). Residues involving Asp40, Asp43 and Asp45 were selected to center the docking box. The OPLS4 force field was used in all calculations. The images were created in Discovery Studio (version 3.5, Accelrys).

ZLQ-3-1 measurements

The concentrations of ZLQ-3-1 in the brain tissues were analyzed using electron spray ionization (ESI) operating at positive ion and multiple reactions monitoring (MRM) mode. Brain samples were homogenized with methanol (containing 0.5 M NaF, 0.5 M sodium ascorbate, 0.5 M HCl) at ratio of 5 (v/w). An aliquot of 20 μL of sample was added with 100 μL methanol containing 5 ng/mL verapamil for protein precipitation. After centrifugation at 13000 rpm for 5 min, a total of 60 μL of supernatant was added with 60 μL water, and an aliquot of 20 μL of the mixture was injected into the LC-MS/MS system. The solution was analyzed by HPLC-MS using a 5500+ Triple Quad LC/MS mass spectrometer (SCIEX Technologies). Analyte separation was achieved using an Xselect HSS T3 column (Waters, 3.5μm, 2.1 × 50 mm). Mobile phases A and B were composed of 1% formic acid and 5 mM ammonium acetate in HPLC grade water and methanol, respectively. The gradient elution profile was chosen as 0.8 min: 20% B (0.7 mL/min); 1.3 min: 98% B (0.7 mL/min); 2.00 min: 98% B (0.7 mL/min); 2.01 min: 20% B (0.7 mL/min). The specific ions (m/z) monitored were ZLQ-3-1 (m/z 483.0 → m/z 423.0), Verapamil (m/z 455.2 → m/z 165.1). The retention time for ZLQ-3-1 and Verapamil was 1.45 min and 1.65 min, respectively. All data were collected and analyzed using Analyst Workstation Software (SCIEX Technologies).

Nissl staining

Infarct size was measured using Nissl staining according to the manufacturer’s instructions (Beyotime Biotechnology, cat #C0117). Coronal slices (40 μm) were cut using a vibrating blade microtome (VT1200s, Leica). The sections were immersed in the Nissl staining solution at 40°C for 10 min. Images were captured with a Zeiss Axio microscope.

Slice preparation

Under deep anesthetization with 2% isoflurane (RWD Life Science, cat #R510-22), mice were decapitated. Brains and spinal cords were rapidly dissected out and cut into 350 μm slices in ice-cold solution containing 110 mM choline chloride, 20 mM glucose, 2.5 mM KCl, 0.5 mM CaCl2, 7 mM MgCl2, 1.3 mM NaH2PO4, 25 mM NaHCO3, 1.3 mM sodium ascorbate, and 0.6 mM sodium pyruvate (pH 7.4) using a vibratome (VT1200S, Leica). Slices were transferred to an interface-style chamber and recovered in ACSF containing 10 mM glucose, 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 1.3 mM MgCl2, 1.3 mM NaH2PO4, 25 mM NaHCO3, 1.3 mM sodium ascorbate and 0.6 mM sodium pyruvate at 34°C for at least 1 h before recording. All external solutions were constantly saturated with 95% O2/5% CO2.

Whole-cell recordings

Slices were visualized using an upright microscope (Olympus X51W, Nomasky) equipped with a 40 × water-immersion lens and recorded using a patch clamp amplifier (Axonpatch-700B, Axon Instruments). Whole-cell recordings were made in fluorescent or pyramidal neurons from peri-infarct cortex using patch-clamp electrodes (tip resistance, 6–8 MΩ). Signals were low-pass filtered at 2 kHz and sampled at 10 kHz using a digitizer (Digidata 1440A, Axon Instruments). Access resistance was monitored throughout the experiment. Only neurons in which the access resistance <25 MΩ and access resistance changes <20% from initial values were included in our analysis. Data were collected with pClamp 10.3 and analysis using Clamfit 10.3 (Molecular Devices). To label the putative pyramidal neurons, biocytin (0.2%, Sigma-Aldrich, cat #B4261) was added to the internal solution. After recording, slices were fixed and processed for post hoc immunohistochemistry. All experiments were performed in a blind fashion.

For tonic inhibitory current (Itonic) and spontaneous inhibitory postsynaptic current (sIPSC) recordings, patch-clamp electrodes were filled with an intracellular solution containing 120 mM CsMeSO4, 10 mM CsCl, 5 mM TEA-Cl, 1.5 mM MgCl2, 10 mM HEPES, 0.1 mM EGTA, 2 mM Na-ATP, 0.5 mM Na-GTP and 5 mM QX-314, pH adjusted to 7.25–7.30 by CsOH (275–285 mOsmol). Itonic and sIPSC were recorded by voltage-clamp at holding potential of +10 mV. GABA was added to ACSF to a final concentration of 5 μM to replenish the extracellular GABA concentration reduced by the high-flow perfusion of the slices. Itonic was the change in baseline holding current (Ihold) after bath-application of Bicuculline methiodide (BMI, 100 μM, Tocris Bioscience), a GABAARs blocker. Because the membrane capacitance of each cell may be different, we cannot obtain accurate tonic inhibition of each cell through Itonic. Therefore, we defined tonic inhibition by the extrasynaptic tonic current density of each cell. The tonic current density was calculated as amplitude of Itonic (pA)/membrane capacitance (pF). The amplitude and frequency of sIPSCs before bath-application of BMI were analyzed using MiniAnalysis Program 6.0 (Synaptosoft Inc.).

For action potential (AP) recordings, patch-clamp electrodes were filled with an intracellular solution containing 70 mM potassium gluconate, 70 mM KCl, 2 mM NaCl, 2 mM MgCl2, 10 mM HEPES, 1 mM EGTA, 2 mM MgATP, and 0.3 mM Na2GTP with the pH adjusted to 7.25–7.30 by KOH (275–285 mOsmol). AP firing was measured by current-clamp recording.

For spontaneous excitatory postsynaptic current (sEPSC), miniature excitatory postsynaptic current (mEPSC), and miniature inhibitory postsynaptic current (mIPSC) recordings, patch-clamp electrodes were filled with an internal pipette solution, containing 132.5 mM cesium gluconate, 17.5 mM CsCl, 2 mM MgCl2, 0.5 mM EGTA, 10 mM HEPES, 4 mM ATP, 5 mM QX-314, pH adjusted to 7.25–7.30 by CsOH (275–285 mOsmol). sEPSCs were isolated by adding BMI (20 μM) to ACSF to block GABAAR-mediated currents. Tetrodotoxin (0.5 μM) was added to external solution to block Na^+^ currents for mEPSC and mIPSC recordings. mEPSCs were recorded by adding BMI (20 μM) to ACSF to block GABAAR-mediated currents. mIPSCs were isolated by adding CNQX (20 μM) and AP5 (100 μM) to block AMPAR- and NMDAR-mediated currents, respectively. The recordings were longer than 5 min. Data were analyzed using Mini Analysis Program 6.0 (Synaptosoft Inc.). More than 100 events from each neuron were collected at a fixed sampling interval to generate cumulative probabilities.

Local field potential (LFP) recordings

For LFPs recordings, microelectrodes (1–2 MΩ) were filled with ACSF as internal pipette solution. An LED was used to deliver 470 nm blue light to photostimulate ChR2-expressing axons projected from the peri-infarct motor cortex in the contralateral motor cortex, the ipsilateral striatum and the C5-C7 cervical spinal cord segment. The slices were stimulated by a single 20 ms light pulse repeated every 4 s for 20 repetitions.

Tract-tracer injections and quantification of axonal sprouting

Tract tracing was used to assess post-stroke axonal sprouting as described previously.^13^^,^^66^ On day 34 after stroke, 2 μL of 10% biotinylated dextran amine (BDA; Molecular Probes, cat #D1956) were injected into two sites of the contralateral cortex at the following coordinates (anteroposterior: 0.6 mm, 1.2 mm, −1.5 mm; 0 mm, 1.8, −1.7). Fourteen days after injection, brains and spinal cords were harvested after transcardial perfusion with ice-cold saline followed by 4% PFA. Coronal brain and transverse cervical spinal cord sections were cut at 40 μm thickness. BDA^+^ axons were labeled by incubation with streptavidin-Cy3 (1:500; Jackson Immuno-Research Laboratories, cat #016-160-084, RRID: AB_2337244). The number of midline-crossing BDA^+^ fibers (from intact to lesioned side) was counted by an investigator blinded to group allocations. Three sections (80 μm apart) at the medullary pyramid level from each mouse were assessed in this manner.

Quantification and statistical analysis

Data are presented as mean ± standard error of the mean (SEM). Comparisons among multiple groups in behavioral assessments were made with two-way repeated-measures ANOVA followed by post hoc Bonferroni test, and other comparisons among multiple groups were made with one-way ANOVA followed by post hoc Scheffe test. Comparisons between two groups were made with two-tailed Student’s t test. A threshold level of significance was set at p < 0.05. The sample size was obtained with power analysis and sample size (PASS) software using a significance level of α = 0.05 with 90% power to detect statistical differences. For animal studies, the sample size was predetermined by our prior experiments. All experiments were conducted blinded and the investigator responsible for data analyses was blinded to which samples/animals represents control and treatment groups. Statistical details of experiments were described in method details or Figure Legends.

Published: October 17, 2024

Footnotes

Contributor Information

Yu-Hui Lin, Email: yuhuilin@njmu.edu.cn.

Ya-Juan Qin, Email: yjqin@njmu.edu.cn.

Dong-Ya Zhu, Email: dyzhu@njmu.edu.cn.

Supplemental information

References

Associated Data

Supplementary Materials

Data Availability Statement

• Original western blot images are available at Mendeley Data: https://doi.org/10.17632/bgnb8gxdz3.1.
• •This paper does not report original code.
• Any additional information required to reanalyze the data in this paper is available from the lead contact upon request.