Results
Optimised LIFUS parameters for ischaemic stroke therapy in mice
Experiments were designed as shown in online supplemental figure 1. A LIFUS with 500 kHz centre frequency was suitable for rodent research (Ultrasound Neurostimulation System, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China). The x-y diameter of ultrasound was 4.5 mm and 5 mm, respectively, the depth of ultrasound irradiation was 4 mm, and the x-y diameter was 2 mm when normalised sound pressure was more than 50% of total irradiation energy (figure 1A–C). The sequence diagram of the ultrasound stimulation was shown. 500 kHz pulse repetition frequency (PRF), 300 ms sonication duration (SD) and 50% duty cycle (DC)23 were used across all the ultrasound experiments (figure 1D).
Figure 1Establishing optimal LIFUS parameters for neuronal repair and remodelling in MCAO mice. (A) 2D distribution of the ultrasound field in oblique and longitudinal cross-sections. The x-axis and y-axis of the ultrasound spot were ~4.5 mm and ~5 mm, respectively. (B) Depth of ultrasound irradiation was 4 mm. (C) X-Y diameter was 2 mm when normalised sound pressure was more than 50% of the total irradiation energy. (D) Schematic diagram of ultrasound, PRF was 1 kHz and 1/PRF was 1 s, SD was 300 ms, and DC was 50%. (E) Cresyl violet-stained brain sections and (F) quantification of atrophy volume at 14 days following MCAO. Dashed lines indicated brain atrophy area, (n=3–7 mice/group). (G) Representative CD31 (red) and Ki67 (green) immunostaining images in the perifocal region. (H) Quantitative analysis of capillary number, (I) surface area, (J) Ki67+, and (K) Ki67+/CD31+ signals in the perifocal region of ipsilateral hemisphere after 14 days following MCAO in mice, (n=4 mice/group. Scale bar=150 µm). (L) mNSS, (M) tail suspension, (N) grid walking and (O) rotarod test for neurobehavioural outcomes in each group (n=3 mice/group in sham groups, n=10–12 mice/group in the IS groups). US1, −2 to –3=mice treated with different dose of ultrasound. IS, ischaemic stroke mice; IS US, ischaemic stroke mice treated with US. Data are mean ± SD. LIFUS, low-intensity focused ultrasound stimulation; mNSS, modified neurological severity score; PRF, pulse repetition frequency; SD, sonication duration.
Additional combinations of pressure (MPa) were used to make Ispta different (online supplemental table S1). We combined different ultrasonic intensity (UI: 22, 101, 201 mW/cm2) and stimulation duration (1, 3, 5 min) to treat stroke mice in the late acute stage (online supplemental table S2). We measured the cerebral blood flow (CBF) after ultrasound stimulation using laser speckle imaging. The result showed that CBF increased after immediate ultrasound stimulated ipsilateral hemisphere for 3 or 5 min with 101 and 201 mW/cm2 day 7 after stroke (online supplemental figure 2A). The motor function was evaluated by modified neurological severity score (mNSS) before and up to 14 days after MCAO using different stimulation parameters, results showed that only 101 mW/cm2 improved neurobehavioural outcomes. One day of ultrasound stimulation interval and 3 or 5 min of stimulation duration achieved a better recovery in neurobehavioural outcomes compared with the MCAO mice (IS, online supplemental figure 2B–D).
To explore whether longer time of ultrasound stimulation would have better or side effect in MCAO mice, we added a group of 10-min stimulation with 101 mW/cm2 every other day. Then, we labelled ultrasound stimulation 3, 5 and 10 min with 101 mW/cm2 every other day as US1, US2 and US3, respectively (online supplemental table S3). We examined the brain atrophy volume to further determine the appropriate ultrasound parameter (figure 1E). Cresyl violet staining results showed that US3 damaged the ipsilateral hemisphere in both young and aged healthy mice compared with the control, indicated that 10 min was an overtime of stimulation. For MCAO (IS) groups, US1 showed better potential than US2 to reduce the brain atrophy volume (figure 1F). The instant change of CBF and mNSS results showed that US1 and US2 were better stimulation to improve stroke recovery. To explore the effect of LIFUS on CBF of MCAO mice, we measured the spatiotemporal changes of CBF using laser speckle imaging, and chose the perifocal regions as ROI (online supplemental figure 2E). For the sham group, there was no difference in CBF changes in immediate, 7 and 13 days, and 13 days endpoint after US1 and US2. For the IS groups, both US1 and US2 groups of mice showed better CBF recovery in the ipsilateral hemisphere of the IS mice than that of IS group, while US1 exhibited increased CBF at day 7 to the endpoint of 13 days after MCAO (online supplemental figure 2F).
The increase of cerebral blood volume is attributed to a surge of angiogenesis.24 25 We investigated angiogenesis in the perifocal region of the ipsilateral hemisphere after MCAO (figure 1G). The number of microvessels and surface CBF did not increase in the sham groups after US1 or US2 treatment (figure 1H,I). The number of Ki67+ cells and Ki67+/CD31+ microvessels increased in perifocal region of the ipsilateral hemisphere at 14 days of MCAO after US1 or US2 treatment, but they were more prominent after US1 treatment (figure 1J,K).
To further determine which was the best stimulation parameter of LIFUS for the neurobehavioural recovery in the MCAO mice, we applied a number of neurobehavioural tests including mNSS, tail suspension test, foot fault and rotarod to comprehensively evaluate the sensorimotor functions up to 14 days after MCAO (figure 1L–O). Both US1 and US2 stimulation did not induce significant change compared with that in the sham groups. For IS groups, both US1 and US2 showed better outcomes than that in IS groups, while US1 exhibited fewer motor deficits and better neurological recovery in all the four neurobehavioural tests in the MCAO mice (figure 1L–O). qPCR results also showed that mRNA level of VEGF, BDNF and endothelial nitric oxide synthase (eNOS) increased in the MCAO mice after US1 stimulation compared with the IS group (online supplemental figure 2G). Taken together, these data suggested that US1 stimulation attenuated neurological deficits and promoted functional recovery at both the histological and neurobehavioural levels. Therefore, US1 was chosen as the optimal ultrasound stimulation method in the following experiments.
LIFUS upregulated HMGB1 and downregulated CAMK2N1 in a new cluster of astrocytes
To explore the mechanism of ultrasound effect on CBF recovery and newly formed microvessels, we performed iTRAQ-based proteomic analysis to detect the mechanism in the protein level. Genes with a p value<0.05 and |fold change|≥1.5 were regarded as DEGs. Overall, 261 genes were upregulated and 848 genes were downregulated after LIFUS compared with the IS mice (online supplemental figure 3A). Volcano plot displayed 184 genes involved in angiogenesis and synapse pathway (figure 2A). We further conducted gene ontology (GO) pathway analyses to display the enrichment of 15 terms clustering from all different expression genes that we were interested in the IS US group compared with the IS group, indicating that ultrasound truly change the physiological process after ischaemic stroke significantly, and correlated to VEGF and calcium related pathways (figure 2B). Heatmap displayed changed genes, which related with angiogenesis and synapse signal pathways in the IS US group compared with the IS group (figure 2C).
Figure 2LIFUS upregulated HMGB1 and downregulated CAMK2N1 in a new cluster of astrocytes. (A) Volcano plot demonstrated fold change of protein level of HMGB1and CAMK2N1 in the IS US group compared with the IS group. (B) Bar chart of GO terms showed angiogenesis and synapse related pathways including BP (biological process, green), CC (cell component, blue), and MF (molecular function, red). (C) Heatmap showed different protein expression in the IS US group compared with IS group. (D) UMAP plot showed the expression profiles in the left ipsilateral by clustering cell types in the ipsilateral hemisphere of mouse brain. (E) Heatmap showed differentially expressed genes (DEGs) in the IS US group compared with IS group of astrocytes. (F) Bar chart of GO terms showed enriched HMGB1 and (G) CAMK2N1-related pathways of astrocytes in the IS US group. (H) Expression profiles of HMGB1 and CAMK2N1 in astrocytes organised into groups, and coloured based on gene expression patterns. (I) Bar chart showed the proportions of three subgroups in four different groups. (J) Violin plots represented the expression distributions of HMGB1 and CAMK2N1 in astrocytes organised into groups. GO, gene ontology; IS, ischaemic stroke mice; IS US, ischaemic stroke mice treated with US; US, mice treated with ultrasound; LIFUS, low-intensity focused ultrasound stimulation; UMAP, uniform manifold approximation and projection; VEGF, vascular endothelial growth factor.
To further explore the mechanism at single-cell level, cell lineage analysis by comprehensive single-cell RNA-sequencing was performed to gain information of the transcriptional profile in different treated groups. Cluster analysis using a uniform manifold approximation and projection for dimension reduction (UMAP) revealed the difference in global gene expression profiles of cell types in four different groups, and identified clusters of cells with unique genetic signature (figure 2D, online supplemental figure 3B). Gene expression profiles of healthy and injured region with or without ultrasound stimulation were shown by UMAP, suggesting that the expression profiles after LIFUS both in physiological and pathological conditions were different (online supplemental figure 3C).
Since previous studies showed that astrocyte mediated the effects of ultrasonic neuromodulation, we then concentrated on the subclusters of astrocytes in four groups of astrocytes. Heatmap showed 17 differentially expressed genes overlapped with genes of iTRAQ proteomic analysis related to angiogenesis and synapse pathways in astrocytes of the IS US mice compared with that in the IS mice (figure 2E). We found that most of the other significantly differential genes were phenotypic genes such as Pecam1 or pathway genes such as Pik3c2a/Atp5f1d. We aimed to find the initial genes that directly response to ultrasound as the target genes therefore, then focused on HMGB1 and CAMK2N1. HMGB1 was the top differential expressed gene related to angiogenesis, and CAMK2N1 was the top differential gene related to synaptogenesis, suggesting that these two genes were involved in the promotion of angiogenesis and synaptogenesis in MCAO mice with LIFUS. Bar chart of GO terms showed enriched HMGB1 and CAMK2N1-related pathways in astrocytes from IS US group, including angiogenesis-related, calcium-related and synapse-related pathways which we focused on (figure 2F,G). The secondary profiling of astrocytic subclusters yielded three subtypes with distinct functional cell identities (online supplemental figure 3D). Feature plot showed HMGB1 and CAMK2N1 expression in secondary profiles of four different groups (figure 2H). Among the three subclusters, the proportion of subcluster 2 (C2) astrocytes in the IS US mice was~70%, more than IS group (figure 2I), suggesting that IS US promote the C2 positive new subcluster. Heatmap showed expressed genes that HMGB1 increased and CAMK2N1 decreased in subcluster 2 compared with other subclusters (1 and 3) in astrocytes, which was consistent with iTRAQ proteomics analysis (online supplemental figure 3E). The distributions of HMGB1 and CAMK2N1 expression in astrocytes, microglia and endothelial cells were displayed by violin plots (figure 2J, online supplemental figure 3F). Heatmap of top differentially expressed genes in astrocytes related to angiogenesis and synapse pathway were shown in the US group compared with sham group, while there has no change of HMGB1 and CAMK2N1 expression (online supplemental figure 3G). GO terms displayed enriched pathways in all cell types in US group compared with the sham group, and showed that enriched C2 promote angiogenesis, neuroregeneration, immune and inflammatory related pathways after stroke, and other cell types could be involved in angiogenesis and synaptogenesis via different signal molecules in physiological condition (online supplemental figure 3H).
Astrocytic HMGB1 inhibition decreased angiogenesis-related factor expression and reversed neurobehaviour recovery after LIFUS in MCAO mice
To investigate whether LIFUS increased angiogenesis by upregulating HMGB1 in astrocytes of MCAO mice, we knocked down HMGB1 in whole-brain cells by pAAV-U6-shRNA (HMGB1)-CMV-WPRE virus, and in astrocytes by pAAV-GfaABC1D-3xFLAG-miR30shRNA (HMGB1)-WPRE virus. pAAV-U6-shRNA-CMV-WPRE virus was used as a control.
FISH was used to colocalise targeted gene with astrocytes in RNA level. FISH results showed that HMGB1 significantly expressed in astrocytes compared with the corresponding controls at 14 days after MCAO (figure 3A). Expression of HMGB1 increased after US treatment compared with the IS group of mice and decreased in the IS US sh (HMGB1 inhibition in the whole brain) and IS US gf-sh groups (HMGB1 inhibition in astrocytes) in mRNA level (figure 3B), which demonstrated that viral inhibition was effective. By quantifying the number of HMGB1+/GFAP+ astrocytes in the perifocal region at 14 days after MCAO by FISH, we found that 64% of HMGB1+ cells were astrocytes and 36% HMGB1+ cells were other type of cells (figure 3C).
Figure 3Inhibiting astrocytic HMGB1 attenuated angiogenesis and reversed neurobehavioural outcomes after LIFUS in MCAO mice. (A) Representative in situ hybridisation images of HMGB1 (green) signals and GFAP+ astrocytes (red) in the IS scramble, IS US scramble, IS US sh (HMGB1), and IS US gf-sh (HMGB1) mice. Scale bar=75 µm. (B) Corresponding semiquantification of HMGB1 mRNA expression level in different groups and (C) percentile of astrocytes (n = 3 mice/group). (D) Western blotting (E) and quantification of HMGB1, VEGFA and FGF2, protein levels (from left to right, normalised to corresponding sham) in ipsilateral mice brain, n=4 mice/group for all groups. (F) mNSS, (G) tail suspension, (H) grid walking and (I) rotarod test of neurobehavioural outcomes in each group (n=3 mice/group in sham groups, n=12 mice/group in the IS groups). IS, ischaemic stroke mice; IS US, ischaemic stroke mice treated with US; US, mice treated with ultrasound. Data are mean ± SD. LIFUS, low-intensity focused ultrasound stimulation.
We then analysed the expression of HMGB1, VEGFA and FGF2 in perifocal regions in MCAO mice. VEGFA and FGF2 expression showed an increasing trend in the sham group after LIFUS. The expression of HMGB1, VEGFA and FGF2 significantly increased in LIFUS treated groups compared with the IS group at 14 days after MCAO, which was reversed by HMGB1 inhibition in IS US sh and IS US gf-sh groups of mice (figure 3D,E). The expression of other proteins and neurobehavioural outcomes was not affected by HMGB1 inhibition in the sham mice (online supplemental figure 4A-F). For IS groups, ultrasound-treated groups showed better neurobehavioural outcomes including mNSS, tail suspension test, foot fault and rotarod than that in the IS group at 14 days after MCAO (figure 3F–I), and reversed by the inhibition of HMGB1 in the whole brain and astrocytes.
LIFUS-upregulated HMGB1 in astrocytes promoted CBF and lectin+ microvessels after MCAO
To determine whether LIFUS-induced angiogenesis was correlated with upregulated HMGB1 in astrocytes, we further measured the spatiotemporal changes of CBF and brain microvasculature in ipsilateral perifocal region (figure 4). Surface CBF increased in the IS US group compared with the IS group of mice (figure 4A,B). We also qualified the ratio of the ipsilateral to the contralateral of ROI; the results also showed that CBF increased at day 7 and day 13 after MCAO (figure 4). Increased CBF induced by LIFUS were reversed by inhibiting HMGB1 in the whole brain cells as well as astrocytes (figure 4A,B).
Figure 4LIFUS upregulated astrocytic HMGB1 via increased CBF and lectin+ microvessels in MCAO mice. (A) Representative images showed immediate CBF changes and endpoint CBF of ROI at 7 days and 13 days by laser speckle imaging in the sham, sham US, IS scramble, IS US scramble, IS US sh (HMGB1), IS US gf-sh (HMGB1) groups. (B) Quantification of CBF normalised to sham. Start row was the quantification of CBF at 7 days and 13 days, respectively. End row was the quantification of immediate CBF changes followed ultrasound at 7 days and 13 days respectively, (n=6 mice/group in the sham groups, n=10 mice/group in the IS groups). (C) Representative lectin (red) and Ki67 (green) immunostaining images, scale bar=150 µm. (D) Quantitative analysis of capillary number, (E) surface area showing angiogenesis in perifocal region. (F) Quantification of the Ki67+ cells and (G) Ki67+/lectin+ signals to exhibit newly formed endothelial cells and microvessels, (n=3 mice/group in sham groups, n=4 mice/group in the IS groups). Sham groups indicated sham group and US group. IS groups indicated IS scramble group, IS US scramble group, IS US sh(HMGB1) group and IS US gf-sh (HMGB1) group. Data are mean ± SD. IS, ischaemic stroke mice; IS US, ischaemic stroke mice treated with US; US, mice treated with ultrasound; LIFUS, low-intensity focused ultrasound stimulation.
We further investigated functional angiogenesis by injecting tomato lectin. Lectin+ microvessels increased in the IS US group compared with IS group of mice which was reversed after HMGB1 inhibition (figure 4C). The quantification of lectin+ capillary number and surface area increased in the US groups compared with the IS, IS US sh and IS US gf-sh groups (figure 4D,E). Ki67+ cells colabelled with lectin+ microvessels emerged in the perifocal regions, indicating newly formed microvessels. Ki67+ cells and Ki67+/lectin+ microvessels increased in the perifocal region of the IS US group of mice compared with the other 3 groups of mice at day 14 following MCAO (figure 4F,G). scRNA results showed that HMGB1 mRNA has no difference in sham groups with or without US treatment (online supplemental figure 3G), suggesting that ischaemic mice easily respond to LIFUS, and finally led to improve outcomes.
Astrocytic CAMK2N1 overexpression reversed the synapses increase after LIFUS in MCAO mice
To determine whether LIFUS promoted synaptogenesis via decreasing CAMK2N1 in astrocytes after MCAO, we overexpressed CAMK2N1 in the brain cells by pAAV-CAG-P2A-CAMK2N1-3xFLAG-WPRE and in astrocyte by pAAV-GfaABC1D-P2A-CAMK2N1-3xFLAG-WPRE. pAAV-P2A-3xFLAG-WPRE was used as a control.
We investigated the CAMK2N1 localisation and expression by FISH and Western blot (figure 5). FISH results showed that CAMK2N1 expressed in astrocytes at 14 days after MCAO (figure 5A). CAMK2N1 expression was downregulated in the IS US group compared with the IS group at 14 days after MCAO, and successfully increased in the IS US sh and the IS US gf-sh groups compared with the IS US control (figure 5B). Quantification of the number of CAMK2N1+/GFAP+ astrocytes showed that 80% of CAMK2N1+ cells were astrocytes and 20% of CAMK2N1+ cells were other cells in the perifocal regions at 14 days after MCAO (figure 5C).
Figure 5Astrocytic CAMK2N1 overexpression reversed the synapses increase after LIFUS in MCAO mice. (A) Representative images of CAMK2N1 (green) signals and GFAP+ astrocytes (red) in IS scramble, IS US scramble, IS US CAMK2N1, and IS US gf-CAMK2N1 mice. Scale bar = 75 µm. (B) Corresponding quantification of CAMK2N1 mRNA expression level in different groups and (C) percentile of astrocytes, (n= 3 mice/group). (D) Western blotting and (E) quantification of CAMK2N1, p-CAMK2, BDNF, GluR1, VGLUT1, VGLUT2, VGAT, Gephyrin, SynI, Homer1, respectively, from left to right and up to down, relative to β-actin and normalised to corresponding sham in ipsilateral hemisphere of mouse brain, (n=4 mice/group). Sham groups indicated sham scramble group, US group, US CAMK2N1 group and US gf-CAMK2N1 group. IS groups indicated IS scramble group, IS US scramble group, IS US sh (HMGB1) group and IS US gf-sh (HMGB1) group. Data are mean ± SD. IS, ischaemic stroke mice; IS US, ischaemic stroke mice treated with US; US, mice treated with ultrasound; LIFUS, low-intensity focused ultrasound stimulation.
Western blot analysis showed that CAMK2N1 was downregulated in the IS US groups compared with the IS group, which was reversed in the IS US CAMK2N1 (CAMK2N1 overexpression in the whole brain) and the IS US gf-CAMK2N1 (CAMK2N1 overexpression in astrocytes) groups (figure 5D,E). Meanwhile, IS US groups exhibited higher expression of Phospho-Ca2+/calmodulin-dependent protein kinase II (p-CAMK2), which was also reversed after CAMK2N1 overexpression (figure 5E). In addition, LIFUS increased BDNF, excitatory synapse-related protein glutamate receptor 1 (GluR1), VGLUT1 and VGLUT2 expression. Inhibitory synapse protein (VGAT and Gephyrin), and total synapse protein (Synapsin I and Homer1) had similar expression pattern (figure 5E). At the same time, CAMK2N1 overexpression did not affect other protein expression or neurobehaviour in the sham mice (online supplemental figure 4G–L).
LIFUS-driven astrocytic CAMK2N1 downregulation promoted electrical signals and increased dendritic spine density after MCAO
To investigate morphological changes in neuron after increased synaptic protein levels, we performed Golgi-Cox staining to visualise neuronal dendritic spines in the peri-focal region (figure 6A). The number of total spines on secondary dendrites was shown and calculated (figure 6B). Results showed that the number of total spines was significantly increase in the IS US group compared with the IS group, which was reversed in the IS US CAMK2N1 and the IS US gf-CAMK2N1 groups compared with the IS US groups (figure 6C), suggesting that inhibition of downregulated CAMK2N1 in whole brain or astrocyte was beneficial for dendritic spines increase after ischaemic stroke.
Figure 6Astrocytic CAMK2N1 downregulated by LIFUS promoted electrical signals and increased dendritic spine density after LIFUS in MCAO mice. (A) Presentative Golgi staining images and quantitative analysis. Low magnification of Golgi staining images of neurons in the perifocal region of ipsilateral hemisphere. Scale bar=75 µm. (B) Representative images of dendritic spines and (C) a bar graph showed the number of total spines in the sham, US, US CAMK2N1, US gf-CAMK2N1, IS, IS US, IS US CAMK2N1, IS US gf-CAMK2N1 mice at 14 days after MCAO, (n=4 mice/group). (D) Average Ca2+ transients (detla F/F) and (E) variance of GCaMP6s signals were displayed at day 7 and day 13 after MCAO. (F) and (G) Heatmap displayed variance of the calcium activity of neurons during a 5-min records after ultrasound stimulation and CAMK2N1 overexpression respectively (n=3 mice/group). (H, I) Electromyography (EMG) records showed average EMG amplitude (detla A/A) and (J, K) variance heatmap during a 5-min records, (n=3 mice/group). (L) mNSS, (M) tail suspension, (N) grid walking (O) and rotarod test showed that neurobehavioural outcomes in different groups, (n=3 mice/group in the sham groups, n=12 mice/group in the IS groups). Ctrl (black line), Ctrl US (blue line), Ctrl US CAMK2N1 (yellow line), and Ctrl US gf-CAMK2N1 (red line) mice. Data are mean ± SD. IS, ischaemic stroke mice; IS US, ischaemic stroke mice treated with US; US, mice treated with ultrasound; LIFUS, low-intensity focused ultrasound stimulation.
Since GluR1 is a calcium-permeable neurotransmitter receptor and plays a key role in synaptic plasticity, we also applied fibre photometry records to detect Ca2+ changes after LIFUS. The results showed that frequency change of the calcium signals slightly increased in the US group (figure 6D), while it was significantly higher in the IS US groups than that in the IS, IS US CAMK2N1 and IS US gf-CAMK2N1 groups during 3-min LIFUS (figure 6E). Heatmap displayed variance of three mice in one group, suggesting that there was a relatively strong increase of frequency after LIFUS, and this enhancement could be reversed by CAMK2N1 overexpression in whole brain cells and astrocytes (figure 6F,G). Furthermore, we also evaluated nerve-to-muscle signal transmission of motor neurons by electromyography (EMG) amplitude. Results showed that the amplitude increased during LIFUS both in the US and the IS US groups at day 7 and day 13 after MCAO, which was consistent with Ca2+ level changes in the brain. LIFUS-induced enhancements were inhibited in US CAMK2N1 and US gf-CAMK2N1 groups (figure 6H–K). Calcium signals frequency and EMG amplitude of original waves were stable before LIFUS in the different groups (online supplemental figure 5). These results suggested that LIFUS induced calcium changes and increased neuronal activities. Neurobehavioural outcomes including the mNSS, tail suspension test, grid walking and rotarod test were better in the IS US group than that in the IS US CAMK2N1 and IS US gf-CAMK2N1 groups of mice (figure 6L–O). The recovery was reversed by CAMK2N1 overexpression in the whole brain and astrocytes.
