Targeting microglial C1q alleviates blood–brain barrier disruption in the thalamus after cortical infarction

Abstract

Focal cerebral infarction induces secondary damages in non-ischemic remote brain regions, severely affecting the prognosis and life quality of stroke patients. Disruption of blood–brain barrier (BBB) contributes to secondary thalamic damage after distal middle cerebral artery occlusion (dMCAO). However, the underlying molecular mechanisms that initiate BBB disruption and aggravate secondary thalamic damage after dMCAO remain unclear. It has been well-established that the complement pathway involves in the maintaining BBB integrity in the brain. Herein, using bulk RNA-sequencing and proteomics, we found that the complement pathway was the most pronounced in the ipsilateral ventroposterior nucleus (VPN) of thalamus at 2 weeks after dMCAO. Furthermore, we demonstrated for the first time that complement C1q derived mainly from a subtype of microglia binds to CD93 on endothelial cells, leading to the upregulation of vascular cell adhesion molecule 1 (VCAM-1) and BBB disruption in ipsilateral VPN after dMCAO. Knockdown of microglial C1q could inhibit the C1q-CD93 interaction, decrease VCAM-1 expression and maintain BBB integrity in the VPN, thus mitigating secondary thalamic damage after dMCAO. Moreover, minocycline treatment exerted neuroprotective effects by downregulating both microglial C1q and endothelial VCAM-1, consequently alleviating BBB disruption and neuronal loss in ipsilateral VPN, and improving neurological outcomes after dMCAO. Taken together, this innovative study highlights the pivotal role in maintaining BBB integrity mediated by microglial C1q in ipsilateral VPN after dMCAO. This finding provides a novel therapeutic target for improving the long-term prognosis of ischemic stroke patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03822-2.

Introduction

Ischemic stroke has emerged as a leading cause of adult mortality and disability [1]. Converging evidence indicates that cerebral infarction induces neurological deficits not only within the ischemic core but also in remote regions such as thalamus, substantia nigra and corpus callosum etc. beyond the occluded vascular territory [2–5]. All of which the thalamus is a more common site of secondary damage because of its reciprocal connections with the cortex. The secondary thalamic degeneration hinders recovery of motor and somatosensory functions, and leads to cognitive impairment after stroke [6, 7]. Thus, elucidating the mechanisms underlying secondary thalamic degeneration after ischemic stroke and finding more efficient therapies would be of vital importance in improving outcomes of patients.

Accumulating evidence has demonstrated that the neurodegeneration of thalamus post stroke may be associated with blood–brain barrier (BBB) disruption, inflammatory responses, or chronic stress [8–10]. BBB is a multicellular interface comprising endothelial cells, astrocytes, microglia and pericytes with its structural integrity being vital for maintaining brain homeostasis [11]. The BBB integrity is maintained by specialized tight junctions (TJs), including zonula occludens 1 (ZO-1), occludin, claudin-5 and VE-cadherin (VE-cad). They are critical for preventing paracellular leakage [12]. Previous studies have shown that the disruption of TJs is a hallmark of BBB dysfunction in various neurological disorders, such as ischemic stroke, seizure, traumatic brain injury [13–15]. Importantly, BBB disruption is a key pathological event involving in secondary damage after ischemic stroke [16]. Kang et al. observed the downregulation of proteins representing TJs ZO-1, occludin, and VE-cad in the peri-ischemic area, and extravasation of immunoglobulin G (IgG) from the peripheral blood into the brain parenchyma at 3 days after middle cerebral artery occlusion (MCAO), indicating the disruption of BBB [17]. Although our previous research has demonstrated increased BBB permeability in the ipsilateral ventral posteromedial nucleus (VPN) of thalamus after distal MCAO (dMCAO) [8], the underlying signals that mediate BBB disruption, then aggravate secondary thalamic damage after dMCAO has not been fully clarified yet.

It has been increasingly recognized that vascular inflammation plays a crucial role as a mediator of BBB disruption. Propson et al. demonstrated that the upregulation of endothelial vascular cell adhesion molecule-1 (VCAM-1) would lead to endothelial stiffening and then promote BBB disruption in aged mice [18]. In another research, Zhang et al. found that VCAM-1 upregulation on vascular endothelial cells contributes to BBB disruption in the ischemic brain tissue after MCAO in rats [19]. These data suggest VCAM-1 upregulation may have the potential to be connected to BBB disruption and secondary thalamic damage after dMCAO.

Recently, the association between neuroinflammation mediated by the complement pathway and VCAM-1 has drawn more attention [18, 20]. Microglia, the resident innate immune cells in the central nervous system (CNS), can activate the complement pathway and release C1q, which is associated to the initiation and progression of neuroinflammation [21]. C1q is composed of C1qa, C1qb, and C1qc. In a mouse model of traumatic sensory cortex injury, using single-nucleus RNA sequencing (RNA-seq), Holden et al. revealed that microglia secrete substantial amounts of C1q, which promotes chronic inflammation and neuronal loss, leading to secondary thalamic damage [22]. Also, there was a study from Lozada et al. showed that exogenous administration of recombinant C1q protein to human umbilical vein endothelial cells (HUVECs) upregulated VCAM-1 expression. Furthermore, C1q can bind to cluster of differentiation 93 (CD93) and upregulate VCAM-1 expression in HUVECs [23]. However, much less is known about the role of the C1q-CD93 signaling pathway in modulating VCAM-1 and BBB integrity in VPN after dMCAO.

Minocycline, a highly lipophilic semi-synthetic tetracycline antibiotic, is readily accessible, rapidly effective, and capable of crossing the BBB [24]. Beyond its conventional antibacterial properties, it is moving forward to a potential treatment for ischemic stroke because of anti-neuroinflammatory effects [25–27]. Yang et al. demonstrated that minocycline mitigates cerebral tissue injury following cerebral ischemia–reperfusion, promotes neuronal survival in the ischemic region, and facilitates BBB remodeling [28]. In a meta-analysis, Malhotra et al. also found that minocycline treatment improved neurological function in acute ischemic stroke patients [29].

Given the emerging role of microglial heterogeneity, we hypothesized that a specific microglial subtype that expresses C1q contributes to secondary thalamic BBB disruption after cortical infarction. We demonstrated that microglial C1q activates endothelial cells via CD93, leading to VCAM-1 upregulation and subsequent BBB disruption. What’s novel is we reported that minocycline attenuates secondary thalamic damage post cortical infarction by downregulating C1q and VCAM-1, thus protecting BBB. This will provide novel therapeutic targets for improving the long-term prognosis of patients with ischemic stroke.

Results

BBB disruption and microglial activation occur in the VPN of ipsilateral thalamus after dMCAO

Firstly, to evaluate the effects of dMCAO on endothelial junction pathway in the ipsilateral VPN, scRNA-seq was performed. Enriched GO cellular component analysis showed that cell–cell junction including proteins of tight junction and adherens junction was decreased in the ipsilateral VPN after dMCAO (Fig. 1A). Then, the KEGG cellular processes analysis also exhibited that tight junction, gap junction, and adherens junction were reduced (Fig. 1B), suggesting that BBB integrity was impaired. To confirm whether BBB integrity was compromised in the ipsilateral VPN of thalamus after dMCAO, we then evaluated BBB permeability. Immunofluorescence revealed that the leakages of albumin and IgG were observed in the ipsilateral VPN at 2–4 weeks after dMCAO (Fig. 1C-E). The leakage of fibrinogen and an increased number of CD8a positive cells were also observed in the ipsilateral VPN at 2 weeks after dMCAO (Figure S1A-C), indicating BBB disruption initiated from 2 weeks after dMCAO. Notably, the density of Iba-1 positive cells increased in the ipsilateral VPN at 1–4 weeks after dMCAO (Fig. 1C, F), implying that the activation of microglia were earlier than the BBB disruption, which is positively correlated with BBB disruption. In addition, the gliosis and leakages of IgG were demonstrated predominantly in the ipsilateral thalamus and infarction border zone after dMCAO (Figure S1D). To further characterize BBB disruption, we examined the ultrastructural changes in the BBB by transmission electron microscope (TEM) in the ipsilateral VPN at 2 weeks after dMCAO. The results showed that the BBB exhibited endothelial cell swelling (Fig. 1H, d, e), fragmentation of the basement membrane (Fig. 1H, b, f-g), edema of astrocytic endfeet (Fig. 1H, c, h) and a reduction in TJ (Fig. 1H, a, b) in contrast to Sham rats. Additionally, the expressions of ZO-1, occludin, claudin-5, and VE-cad were significantly decreased in the ipsilateral VPN at 2–4 weeks after dMCAO relative to Sham or contralateral VPN group (Fig. 1I-L).

Fig. 1.

BBB disruption occurs in the ipsilateral VPN of thalamus after dMCAO. A GO cellular component (CC) enrichment bubble chart in ipsilateral VPN (dMCAO vs Sham). B KEGG cellular processes enrichment bubble chart in ipsilateral VPN (dMCAO vs Sham). C BBB permeability and microglial changes at 1–4 w after dMCAO, with the representative images of extravascular albumin (a1-a4, red), IgG (b1-b4, red) and Iba-1(c1-c4, green). Scale bar: 25 μm. D-E Quantitative analyses of albumin and IgG extravasation (n = 3). F The density of Iba-1⁺ cells in ipsilateral VPN. Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. G Correlation between the area of IgG and the density of Iba-1⁺ cells. H Electron microscopy of ultrastructural changes in the BBB of Sham and dMCAO groups. Black arrowhead: tight junction; Green arrowhead: fragmented basement membrane; Red arrowhead: swollen endothelial cell; Orange arrowhead: edema of astrocytic endfeet (I-L) Western blot shows ZO-1, OCLN, CLDN5, and VE-cad expression in ipsilateral VPN of Sham and dMCAO groups. Quantitative analyses of ZO-1, OCLN, CLDN5, and VE-cad level relative to GAPDH. Statistical analysis was performed using one-way ANOVA followed by Bonferroni test. Each bar represents the mean ± SD. ^*p < 0.05 vs Sham group, ^#p < 0.05 vs contra. TJ: tight junction; BM: basement membrane; AEF: astrocytic endfeet; EC: endothelial cell

Microglial C1q contributes to BBB disruption in the ipsilateral VPN of thalamus after dMCAO

To search for the potential molecules associated with BBB disruption after dMCAO, we performed RNA-seq and proteomics analyses in the ipsilateral VPN between Sham and dMCAO 2 weeks groups. In comparison to Sham group, 4131 genes and 1825 proteins were differentially expressed in the dMCAO group, including 2541 up-regulated and 1590 down-regulated genes, and 1098 up-regulated and 727 down-regulated proteins (Fig. 2A). The quantified genes and proteins were observed to exhibit uneven distribution across 9 quadrants of a scatter plot (Fig. 2B). The Venn diagram shows that DEGs and DEPs shared 63 co-enriched pathways. (Fig. 2C). Therefore, we further screened the top 10 pathways with the most significant changes through GO, KEGG, and GSEA analyses. By using GO biological process (BP) analysis, we found that these genes and proteins co-enriched in innate immune response, leukocyte cell–cell adhesion and complement activation, classical pathway (Fig. 2D). The KEGG pathway analyses were predominantly co-enriched in complement and coagulation cascades, antigen procession and presentation, and NOD-like receptor signaling pathway (Fig. 2E). Furthermore, the GSEA analysis showed that the complement and coagulation cascade is one of the pathways with the most significant changes (Fig. 2F). Based on the results of RNA-seq and proteomics, we believed that the complement pathway is one of the most pronounced pathways of above changes. In the complement pathway, the upregulations of C1qa, C1qb, and C1qc were the most significant (Fig. 2G). We then verified the expressions both gene and protein of C1qa, C1qb, and C1qc by qRT-PCR and western blot respectively. Compared with the Sham group and contralateral VPN, the mRNA and protein levels of C1qa, C1qb and C1qc were increased in the ipsilateral VPN at 1–4 weeks after dMCAO, and were most obvious at 2 weeks (Fig. 2H-K). These data indicate that the activation of C1q may be associated with the BBB disruption in the ipsilateral VPN after dMCAO.

Fig. 2.

Multi-omics profiling reveals complement activation in the ipsilateral VPN of thalamus after dMCAO. A The bar chart of DEGs and DEPs. B Scatter diagram of correlation between the transcripts and proteins. The upper right quadrant includes molecules with both upregulated gene and protein expression; the upper left quadrant includes molecules with downregulated gene expression but upregulated protein expression; the lower left quadrant includes molecules with both downregulated gene and protein expression; the lower right quadrant includes molecules with upregulated gene expression but downregulated protein expression. C The Venn diagram showing the overlapped pathway between the RNA-seq and proteomics. D Bubble plot of the top 10 co-enriched pathway analyses of GO BP in RNA-seq and proteomics. E Bubble plot of the top 10 co-enriched pathway analyses of KEGG in RNA-seq and proteomics. F Bubble plot of the top 10 enriched pathway analyses of GSEA in RNA-seq and proteomics. The x-axis represents the omics type, the y-axis denotes different pathways, the size of the bubbles indicates the magnitude of the p-value, and the color of the bubbles reflects the distribution of the NES value. G Heat maps showing the relative expression of significantly altered proteins generated from the pathway of complement and coagulation cascades of proteomics. Each column represents an individual mouse. H The mRNA levels of C1qa, C1qb and C1qc in each group. Statistical analysis was performed using one-way ANOVA followed by Dunnett test. Each bar represents the mean ± SD. *p < 0.05 vs Sham group. I-K Western blot shows C1qa, C1qb and C1qc expression in ipsilateral VPN of the Sham and dMCAO animals. Quantitative analysis of C1qa, C1qb and C1qc level relative to GAPDH. Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. Each bar represents the mean ± SD. ^*p < 0.05 vs Sham group, ^#p < 0.05 vs contra (n = 4 in each group)

To characterize the changes of the main cell types comprehensively and the cellular origin of C1q in the ipsilateral VPN of thalamus after dMCAO, we employed droplet-based scRNA-seq (10 × Genomics) on cells isolated from VPN in Sham and dMCAO 2 weeks groups (Fig. 3A). Totally, 18,011 cells were obtained. They were reduced to 16,396 after quality control. With uniform manifold approximation and projection (UMAP), an unsupervised nonlinear dimensionality reduction algorithm, we clustered and visualized various cell types referring to their relative gene expression levels, in which 11 major cell types on the basis of specific cell markers were annotated. They are microglia (Tmem119 and Cx3cr1), neuron (Snap25 and Syt1), oligodendrocytes (Mbp and Plp1), astrocytes (Atp1b2 and Ndrg2), mast cells (Tpsb2 and Tpsab1), neutrophils (S100a8 and S100a9), T and natural killer (NK) cells (Cd3d, Cd3e and Nkg7), endothelial cells (Cdh5 and cWF), B cells (Cd79a and Ms4a1), macrophages (Mrc1), dendritic cells (DCs) (Ptcra) (Fig. 3B-D). The C1q components C1qa, C1qb, and C1qc are mainly derived from microglia (Fig. 3E). Immunofluorescent assay also demonstrated that C1q was colocalized with Iba-1, but not with NeuN and GFAP in the ipsilateral VPN at 2 weeks after dMCAO (Fig. 3F), and was partly colocalized with RECA-1-positive vessels at 1 and 2 weeks (Figure S2A). Furthermore, the co-staining of C1q and synapsin showed that most of C1q deposited at non-synaptic sites in the VPN of thalamus (Figure S1E). By reclustering microglia using Seurat, eight transcriptionally distinct cell subtypes (M1 to M8) were demonstrated (Fig. 3G). The M1-M8 subclusters are computationally derived cell clusters identified by single-cell RNA-seq. They are distinct from the conventional functional M1/M2 polarization classification. As shown in Fig. 3H-I, most microglia from the Sham sample were M3, M4 and M5. They highly expressed the homeostatic genes of microglia (P2ry12, Cx3cr1, Hexb, Csf1r and Tgfbr1). On the contrary, M1, M2, M6 and M8 subtypes accounted for the majority of microglia in the dMCAO group. More surprisingly, in the dMCAO group, M6 subtype was expressed higher levels of C1qa, C1qb and C1qc than other subtypes of microglia (Fig. 3I and S2F). M1 and M2 subtypes were highly expressed typical disease-associated microglia (DAM) markers such as Csf1, Spp1, Nlrp3 and Itgax. M8 subtype expressed classic proliferative genes such as Mki67. We further compared the differences in M6 microglial subtype and the leakage of albumin and IgG between the infarct border zone and the ipsilateral thalamic VPN. The results showed that the number of C1q⁺-Iba-1⁺ cells in the infarct border zone was lower than that in the ipsilateral thalamic VPN. Although the leakage areas of albumin and IgG showed no significant differences between the two brain regions, the leakage was mainly confined to the infarct border zone of cortex, whereas it was relatively dispersed in the ipsilateral thalamic VPN (Figure S2B-E). These data indicated unique gene expression patterns of microglial subtypes in the ipsilateral VPN of thalamus after dMCAO.

Fig. 3.

Single-cell RNA sequencing identifies microglia as the primary source of C1q in ipsilateral VPN of thalamus after dMCAO. A Scheme of scRNA-seq. B Umap of 11 major cell populations identified in the ipsilateral VPN in the Sham and dMCAO groups. Left: Integration of data from the Sham and dMCAO groups. Top middle: The Sham group. Bottom middle: The dMCAO group. Right: Stacked bar graphs of cell type proportions in the Sham and dMCAO groups (n = 4 in each group). C Cell type annotation was performed on the basis of the expression of well-established marker genes. D Feature plots showing the expression of known cell markers. E Feature plots showing the expression of Tmem119, C1qa, C1qb and C1qc. F Co-stained with C1q (red) and Iba-1 (green), GFAP (green), NeuN (green) within ipsilateral VPN in Sham and dMCAO groups. Scale bar: 25 µm. G Umap of eight cell populations identified by microglial reclustering. H Proportions of all microglial subclusters in Sham and dMCAO groups. I Feature plots showing the expression of eight microglial subclusters

To learn more information in each microglia subtype, KEGG pathway enrichment on top 100 DEGs were performed (Figure S3A-H). The results mainly pointed to enhanced complement and coagulation cascades, antigen processing and presentation, and Fc epsilon RI signaling pathway in M6 subtype (Figure S3F).

In order to explore the dynamic transition among different subtypes of microglia in the ipsilateral VPN of thalamus, we constructed a pseudotime map of the microglia state trajectory using monocle 2 (Fig. 4A-D). In our analysis, the progression trajectory of microglia was established. The progression trajectory of microglia, starting with the M3 subtype (Fig. 4A-D). This was followed by M4 and M5, then moving through an intermediate infiltrating phase represented by M1 and M2, and culminating in the terminally differentiated states of M6 and M8. Further examination of the single-cell transcriptomes of microglia along this trajectory identified 10,729 significantly altered genes, classified into three distinct expression patterns (Fig. 4E). Module 1 included genes that showed increased expression levels along M6. Pathway enrichment analysis indicated that these genes participated in the innate immune response, complement activation, classical pathway, translation, defense response, regulation of immune and DNA replication initiation. Module 2 contained the genes that showed decreased expression levels, with enrichment in actin cytoskeleton organization, cell migration, protein autophosphorylation, regulation of cell shape, protein refolding and brain development. Module 3 involved the genes upregulated in the early stages, which were associated with mRNA processing, cellular response to DNA damage stimulus, regulation of transcription, DNA-templated, activation of GTPase activity, DNA repair and endocytosis (Fig. 4E). Significantly, C1qa, C1qb, C1qc, and Apoe in Module 1, were exhibited the most prominent upregulated (Fig. 4F). Therefore, the levels of these genes progressively increased during microglial differentiation toward the M6 subtype, representing the special stage of the differentiation lineage.

Fig. 4.

Pseudotemporal trajectory analysis reveals the state transitions of microglia in ipsilateral VPN of thalamus after dMCAO. A Trajectory of eight clusters along pseudo-time in a two-dimensional state-space defined by Monocle 2. Each point corresponds to a single cell, and each color represents a microglial state. B The pseudo-time trajectory plots demonstrate the sample distribution along the trajectory. The dot color represents the group. C The developmental pseudo-time of neutrophils was inferred by Monocle analysis. The dark to bright color key indicates cell differentiation from early to late. D The pseudo-time trajectory plots demonstrate the sample distribution along the trajectory. Each dot color represents a neutrophil cluster. E The heatmap displays the significantly differential expression genes during the trajectory. Enriched top GO BP terms for gene sets from each module were represented on the left side. The blue-to-red color key indicates low to high relative expression levels. F Changes in the expression trends of C1qa, C1qb, C1qc, and Apoe along pseudo-time across different cell populations

To investigate the effects of microglial C1q on the BBB in the VPN of ipsilateral thalamus after dMCAO, AAV- shC1qa was injected into the left VPN at 28 days before dMCAO (Fig. 5A). As shown in Fig. 5B, green fluorescent protein (GFP) was colocalized with Iba-1 in AAV-shC1qa administrated rats at 2 weeks after dMCAO. The AAV-shC1qa administration reduced the expression of C1q (Fig. 5C). The expression of C1qa was most significantly downregulated when transduced with AAV-shC1qa (#2) (Fig. 5D). Thus, AAV-shC1qa (#2) was selected in the following experiments. Notably, the administration of AAV‑shC1qa did not affect cortical infarct volume after dMCAO (Figure S8A). Knockdown of microglial C1qa significantly maintain the expressions of ZO-1, occludin, claudin-5, and VE-cad in the ipsilateral VPN at 2 weeks after dMCAO (Fig. 5E-J). In addition, immunofluorescent results showed that AAV-shC1qa administration alleviated the leakage of albumin, IgG, fibrinogen, and decreased the number of CD8a positive cells after dMCAO (Fig. 5K-O).

Fig. 5.

Knockdown of microglial C1qa alleviates BBB disruption in the VPN of ipsilateral thalamus at 2 w after dMCAO. A The AAV vector maps and study design/timeline of left VPN of thalamus with AAV vectors carrying the CX3CR1 promoter. B Representative photomicrographs show the co-localization of GFP (green) and Iba-1 (red) in ipsilateral VPN at 2 w after dMCAO from animals with AAV-shC1qa injection. Scale bar: 25 μm. C Representative images show the co-localization of GFP (green) and C1q (red) in ipsilateral VPN at 2 w after dMCAO from animals with AAV injection. Scale bar: 25 μm. D Representative western blot showing the expression of C1qa in different AAV-shC1qa treatment sequence in Sham groups. Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. Each bar represents the mean ± SD. *p < 0.05 vs AAV-shCon group at 2 w after dMCAO (n = 4 in each group). E On day 14 post operation, the expressions of C1qa, ZO-1, OCLN, CLDN5, and VE-cad in ipsilateral VPN from Sham and dMCAO animals with AAV-shC1qa injection were showed through representative western blot. F-J On day 14 post operation, quantitative analyses of C1qa, ZO-1, OCLN, CLDN5, and VE-cad level relative to GAPDH in the ipsilateral VPN from Sham and dMCAO animals with AAV-shC1qa injection were conducted (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. (K) On day 14 post operation, representative images of extravascular albumin, IgG or fibrinogen deposits and CD8a positive cells in ipsilateral VPN from Sham and dMCAO animals with AAV-shC1qa injection were displayed. Scale bar: 25 μm. (L-O) On day 14 post operation, quantitative analyses of the area of albumin, IgG and fibrinogen leakage and the number of CD8a positive cells in the ipsilateral VPN from Sham and dMCAO animals with AAV-shC1qa injection were conducted (n = 6 in each group). Statistical analysis was performed using Kruskal–Wallis H test. Each bar represents the mean ± SD. *p < 0.05 vs Sham + AAV-shCon group, ^#p < 0.05 vs dMCAO + AAV-shCon group, ^&p < 0.05 vs Sham + AAV-shC1qa group

To provide the evidence that knockdown of microglial C1qa was involved in the neuroprotection against secondary thalamic damage after dMCAO, we further examined the alteration in neuronal loss, glial activation and neurological outcome. As shown in Figure S4A-D, no obvious neuronal damage or glial activation was found in Sham rats either with AAV-shCon or AAV-shC1qa. The number of NeuN positive cells was decreased, with smaller cell bodies after dMCAO. Moreover, the densities of GFAP and Iba-1 positive cells were increased, accompanied by enlarged cell bodies and shortened branches. The knockdown of microglial C1qa significantly ameliorated neuronal loss, and decreased densities of GFAP and Iba-1 positive cells in the ipsilateral VPN after dMCAO. The neurobehavioral assessments were conducted to assess the neurological function of rats. Rats in AAV-shC1qa group exhibited lower scores in the Bederson and the beam-walking tests than those in the AAV-shCon group after dMCAO (Figure S4E-F). The asymmetry scores were significantly shorter in AAV-shC1qa administration than in AAV-shCon group after dMCAO (Figure S4G). Then, we used MWM to assess the cognitive function of rats. After 5 days spatial orientation training, all rats improved their capacity to find the platform. However, the path length and escape latency in dMCAO rats were longer than those of Sham. There was no significant difference in the mean swimming speed among all the observed groups (Figure S4H). Notably, AAV-shC1qa administration significantly shortened the path length from the 1 st to 5th day of training and escape latency after dMCAO (Figure S4I-J). Moreover, in the probe phase, rats treated with AAV-shC1qa spent longer time in the target quadrant occupancy after dMCAO (Figure S4K-L).

Microglial C1q activates vascular endothelial cells via CD93 in the ipsilateral VPN of thalamus after dMCAO

Given that microglial C1q is a secretory protein, we performed ligand-receptor analysis by CellChat on scRNA-seq data to identify potential C1q receptor. Herein we principally focus on the interaction between microglia and endothelial cells. The results showed that the interaction between C1qa and CD93 was significantly enhanced in the ipsilateral VPN after dMCAO (Fig. 6A). Additionally, scRNA-seq revealed that CD93 is located in von Willebrand factor (vWF)-positive endothelial cells (Fig. 6B). Immunofluorescence showed that CD93 was colocalized with laminin and upregulated in the ipsilateral VPN after dMCAO (Fig. 6C-D). The expression of CD93 showed significant upregulation in the ipsilateral VPN at 2–3 weeks after dMCAO compared to the Sham group (Fig. 6E). Co-immunoprecipitation further confirmed that C1qa-CD93 interaction was strengthened (Fig. 6F) and that C1q was partly colocalized with CD93 in the ipsilateral VPN after dMCAO (Fig. 6G).

Fig. 6.

The interaction between microglia and endothelial cells and the activation of endothelial cells in the ipsilateral VPN of thalamus after dMCAO. A List of the molecular ligand-receptor pairing between microglia and endothelial cells from Sham and dMCAO groups. B Feature plots showing the expression of vWF and CD93. C Double-staining of Laminin (green) with CD93 (red) in the ipsilateral VPN in Sham and dMCAO groups. Scale bar: 25 μm. D Quantitative analyses of the expression of CD93 in Sham and dMCAO groups (n = 6 in each group). Statistical analysis was performed using two-tailed t-test. E Representative western blot showing the expression of CD93 in VPN at 1–4 w after dMCAO. Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. F Immunoprecipitation assay shows the levels of C1qa and CD93 in ipsilateral VPN in Sham and dMCAO groups (n = 4 in each group). Statistical analysis was performed using two-tailed t-test. G Double-staining of C1q (red) with CD93 (green) in the ipsilateral VPN in Sham and dMCAO groups. Scale bar: 25 μm. H Representative western blot showing the expressions of VCAM-1 and ICAM-1 in VPN at 1–4 w after dMCAO. Quantitative analyses of VCAM-1 and ICAM-1 level relative to GAPDH (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. Each bar represents the mean ± SD. *p < 0.05 vs Sham group, ^#p < 0.05 vs contra. I Double-staining of VCAM-1 (green) with RECA-1 (red) in the ipsilateral VPN in Sham and dMCAO groups. Scale bar: 25 μm. J Colocalization analyses of VCAM-1 (green) and RECA-1 (red) in Sham and dMCAO groups

To further explore the effect of C1q binding to CD93 on the activation of endothelial cells, we detected the expressions of VCAM-1 and intercellular cell adhesion molecule-1 (ICAM-1) in the VPN after dMCAO. As shown in Fig. 6H, the expressions of VCAM-1 were upregulated at 1–4 weeks after dMCAO, which was peaked at 2 weeks. However, there was no statistically significant difference in ICAM-1 expression among Sham, ipsilateral and contralateral groups after dMCAO. Immunofluorescence of VPN vasculature revealed that VCAM-1 was colocalized with RECA-1, where VCAM-1 expression was increased after dMCAO (Fig. 6 I-J).

To confirm the role of the C1q-CD93 axis in mediating vascular endothelial cells activation in the ipsilateral VPN after dMCAO, microglial C1qa were knocked down with AAV-shC1qa. Immunofluorescent results showed that the colocalization of VCAM-1 and RECA-1 was decreased after AAV-shC1qa administration in dMCAO rats (Fig. 7A). Knockdown of microglial C1qa could inhibit the C1qa-CD93 interaction in the ipsilateral VPN after dMCAO (Fig. 7B) and downregulate the expression of VCAM-1 (Fig. 7C). To further validate that C1q can upregulate VCAM-1 expression in endothelial cells, an indirect co-culture system of BV2 and rat brain endothelial cells (BECs) was established. As shown in Fig. 7D, compared with control and vehicle groups, the levels of C1qa in BV2 cells significantly increased with LPS at the concentrations of 0.01, 0.1, 1, and 10 μg/mL, repectively. Since LPS at 10 μg/mL obviously upregulated the expression of VCAM-1 in BECs (Fig. 7E), the lowest concentration (0.01 μg/mL) was selected for the following experiments. After treating BECs with the cell supernatant of BV2 or recombinant C1qa, we found that VCAM-1 expression was significant increased. Instead, it was decreased in the presence of C1q NAb (Fig. 7F). We further administered a C1q NAb in vivo. As shown in Figure S5A-E, the administration of C1q NAb maintained the expressions of ZO-1, occludin, claudin-5, and VE-cad in the ipsilateral VPN at 2 weeks after dMCAO. In addition, immunofluorescent results showed that C1q NAb administration alleviated the leakage of albumin, IgG, fibrinogen, and decreased the number of CD8a positive cells after dMCAO (Figure S5F-J). To provide the evidence that the administration of C1q NAb was involved in the neuroprotection against secondary thalamic damage after dMCAO, we further examined the alteration in neuronal loss, glial activation and neurological outcome. As shown in Figure S6A-D, no obvious neuronal damage or glial activation was found in Sham rats either with C1q NAb or vehicle. The number of NeuN positive cells was decreased, with smaller cell bodies after dMCAO. Moreover, the densities of GFAP and Iba-1 positive cells were increased, accompanied by enlarged cell bodies and shortened branches. C1q NAb significantly ameliorated neuronal loss, and decreased densities of GFAP and Iba-1 positive cells in the ipsilateral VPN after dMCAO. The neurobehavioral assessments were conducted to assess the neurological function of rats. Rats in C1q NAb group exhibited lower scores in the Bederson, the beam-walkingand and asymmetry tests than those in the vehicle group after dMCAO (Figure S6E-G). MWM was used to assess the cognitive function of rats. After 5 days spatial orientation training, all rats improved their capacity to find the platform. However, the path length and escape latency in dMCAO rats were longer than those of Sham. There was no significant difference in the mean swimming speed among all the observed groups (Figure S6H). Notably, C1q NAb administration significantly shortened the path length from the 1 st to 5th day of training and escape latency after dMCAO (Figure S6I-J). Moreover, in the probe phase, rats treated with NAb spent longer time in the target quadrant occupancy after dMCAO (Figure S6K-L).

Fig. 7.

Microglial C1q activates endothelial cells via CD93 in the ipsilateral VPN of thalamus after dMCAO. A Representative images show the co-localization of RECA-1 (red) and VCAM-1 (grey) in ipsilateral VPN from Sham and dMCAO animals with AAV-shC1qa injection. Scale bar: 25 μm. B Immunoprecipitation assay shows the levels of C1qa and CD93 in ipsilateral VPN in Sham and dMCAO animals with AAV-shC1qa injection (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Bonferroni test. C Representative western blot showing the expressions of VCAM-1 in ipsilateral VPN in Sham and dMCAO animals with AAV-shC1qa injection (n = 4 in each group). Quantitative analysis of VCAM-1 level relative to GAPDH. Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. *p < 0.05 vs Sham + AAV-shCon group, ^#p < 0.05 vs dMCAO + AAV-shCon group. D Representative western blot showing the expression of C1qa in BV2 cells with LPS treatment (n = 4 in each group). Quantitative analysis of C1qa level relative to GAPDH. E Representative western blot showing the expression of VCAM-1 in brain endothelial cells with LPS treatment (n = 4 in each group). Quantitative analysis of VCAM-1 level relative to GAPDH (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. Each bar represents the mean ± SD. *p < 0.05 vs Veh group. F Representative western blot showing the expression of VCAM-1 in each group (n = 4 in each group). Quantitative analysis of VCAM-1 level relative to GAPDH. Each bar represents the mean ± SD. *p < 0.05 vs Veh group, ^#p < 0.05 vs BV2 or r-C1q groups. G Flow chart illustrates drug administration and experimental design. Rats received intracerebroventricular injection of DCBY02 or Veh at 3 days prior to dMCAO and every 2 days thereafter until the end of experiment. H–L Representative western blot showing the expressions of VCAM-1, ZO-1, OCLN, CLDN5, and VE-cad in ipsilateral VPN, respectively. Quantitative analyses of VCAM-1, ZO-1, OCLN, CLDN5, and VE-cad level relative to GAPDH from Sham and dMCAO animals with DCBY02 or Veh injection (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Bonferroni test. Each bar represents the mean ± SD. *p < 0.05 vs Sham + Veh group, ^#p < 0.05 vs dMCAO + Veh group

Next, to determine the effects of CD93 on endothelial cells activation and BBB disruption in the ipsilateral VPN following dMCAO, we applied a CD93 inhibitor DCBY02 to block CD93 (Fig. 7G). As shown in Fig. 7H-L, the expression of VCAM-1 was downregulated and ZO-1, occludin, claudin-5, and VE-cad were increased in the dMCAO group after inhibition of CD93.

VCAM-1 is indispensable for BBB disruption in the ipsilateral VPN of thalamus after dMCAO

To investigate the role of VCAM-1 in the BBB disruption in the ipsilateral VPN of thalamus after dMCAO, we applied AAV-shVCAM-1 to result in endothelial VCAM-1 downregulation in vivo study (Fig. 8A). As shown in Fig. 8B, GFP was colocalized with RECA-1 in AAV-shVCAM-1 administrated rats at 14 days after dMCAO. The AAV-shVCAM-1 administration reduced the colocalization of VCAM-1 and RECA-1 (Fig. 8C). Administration of AAV‑shVCAM-1 did not affect cortical infarct volume after dMCAO (Figure S8A). Compared with AAV-shCon group, the expression of VCAM-1 was downregulated, while the expressions ZO-1, occludin, claudin-5, and VE-cad were increased in the AAV-shVCAM-1 administrated group after dMCAO (Fig. 8D-H). Furthermore, immunofluorescent results uncovered that AAV-shVCAM-1 administration alleviated the leakage of albumin, IgG, fibrinogen, and decreased the number of CD8a positive cells after dMCAO (Fig. 8I-M).

Fig. 8.

Knockdown of endothelial cells VCAM-1 alleviates BBB disruption in the ipsilateral VPN of thalamus at 2 w after dMCAO. A The AAV9-X1.1 vector maps and study design/timeline of left VPN of thalamus with AAV vectors carrying the Tie-1 promoter. B Representative photomicrographs show the co-localization of GFP (green) and RECA-1 (red) in ipsilateral VPN at 2 w after dMCAO from animals with AAV-shVCAM-1 injection. Scale bar: 25 μm. C Representative images show the co-localization of GFP (green), RECA-1 (red) and VCAM-1 (grey) in ipsilateral VPN at 2 w after dMCAO from animals with AAV-shVCAM-1 injection. Scale bar: 25 μm. D-H On day 14 post operation, the expressions of VCAM-1, ZO-1, OCLN, CLDN5, and VE-cad in ipsilateral VPN were showed respectively through representative western blot. Quantitative analyses of VCAM-1, ZO-1, OCLN, CLDN5, and VE-cad level relative to GAPDH from Sham and dMCAO animals with AAV-shVCAM-1 or AAV-shCon injection (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. I On day 14 post operation, representative images of extravascular albumin, IgG or fibrinogen deposits and CD8a positive cells in ipsilateral VPN from Sham and dMCAO animals with AAV-shVCAM-1 or AAV-shCon injection were displayed. Scale bar: 25 μm. J-M On day 14 post operation, quantitative analyses of the area of albumin, IgG and fibrinogen leakage and the number of CD8a positive cells in the ipsilateral VPN from Sham and dMCAO animals with AAV-shVCAM-1 or AAV-shCon injection were conducted (n = 6 in each group). Statistical analysis was performed using Kruskal–Wallis H test. Each bar represents the mean ± SD. *p < 0.05 vs Sham + AAV-shCon group, ^#p < 0.05 vs dMCAO + AAV-shCon group, ^&p < 0.05 vs Sham + AAV-shC1qa group

To know that the knockdown of VCAM-1 was involved in the neuroprotection against secondary thalamic damage after dMCAO, we further examined the alteration in neuronal loss, glial activation and neurological outcome. As shown in Figure S7A-D, the number of NeuN positive cells was decreased, with smaller cell bodies after dMCAO. Moreover, the densities of GFAP and Iba-1 positive cells were increased, accompanied by enlarged cell bodies and shortened branches. The knockdown of VCAM-1 significantly ameliorated neuronal loss, and decreased densities of GFAP and Iba-1 positive cells in the ipsilateral VPN after dMCAO. The neurobehavioral assessments were conducted to assess the neurological function of rats. Rats in AAV-shVCAM-1 group exhibited lower scores in the Bederson, the beam-walking and asymmetry tests than those in the AAV-shCon group after dMCAO (Figure S7E-G). Then, we used MWM to assess the cognitive function of rats. After 5 days spatial orientation training, all rats improved their capacity to find the platform. However, the path length and escape latency in dMCAO rats were longer than those of Sham. There was no significant difference in the mean swimming speed among all the observed groups (Figure S7H). Notably, AAV-shVCAM-1 administration significantly shortened the path length from the 1 st to 5th day of training and escape latency after dMCAO (Figure S7I-J). Moreover, in the probe phase, rats treated with AAV-shVCAM-1 spent longer time in the target quadrant occupancy after dMCAO (Figure S7K-L).

Minocycline alleviates BBB disruption by inhibiting microglial C1q-mediated VCAM-1 upregulation in the ipsilateral VPN of thalamus after dMCAO

To determine whether minocycline alleviates BBB disruption by inhibiting microglial C1q-mediated vascular endothelial VCAM-1 upregulation in the ipsilateral VPN after dMCAO, the molecular docking analysis was conducted to verify the stability of minocycline-C1q interaction. It was showed that minocycline could be successfully docked into C1q with a binding energy of −7.74 kcal·mol⁻¹. In addition, minocycline forms four hydrogen bond interactions (green dashed lines) with the amino acid residues LEU-213, LEU-214, ILE-219, and PHE-220, and two hydrophobic interactions (yellow dashed lines) with GLN-90 in the C1q, indicating a very strong interaction (Fig. 9A). The free energy landscape (FEL) plot of the minocycline-C1q complex shows a single, concentrated minimum energy cluster, suggesting robust stability of the complex (Fig. 9B). RMSD of the minocycline-C1q complex remained consistently stable at minimal fluctuation (Figure S9A). The Rg value of the minocycline-C1q complex remained constant at approximately 1.95 ~ 2.00 nm, exhibiting negligible fluctuation throughout the simulation (Figure S9B). The number of hydrogen bonds between C1q and minocycline remained stable at 2–6 throughout the process. The hydrogen bond curve showed smooth fluctuations, indicating that strong hydrogen bond interactions were formed (Figure S9C). These pieces of evidence supported the stability and functional integrity of the minocycline-C1q complex. Then we administered minocycline via intraperitoneal injection within 30 min after dMCAO (Fig. 9C). The effects of minocycline with different dose on microglial C1q after dMCAO were investigated. As shown in Fig. 9D, the inhibitory effect of 40 mg/kg dose on C1qa was most significant. Moreover, in comparison to vehicle, minocycline at a dose of 10 mg/kg had no significant effect on the microglial activation in the ipsilateral VPN after dMCAO, whereas the inhibitory effect was observed at 20 mg/kg, and it became more pronounced at 40 and 60 mg/kg (Figure S9D-E). Additionally, both dose at 40 and 60 mg/kg of minocycline treatment could downregulate expression of C1qa, but there was no significant difference between the two groups (Figure S9F). Also, minocycline at a dose of 60 mg/kg could downregulate expressions of TNF-α and IL-1β, whereas it (at 40 mg/kg) only downregulates TNF-α (Figure S9G, H). The administration of minocycline did not affect cortical infarct volume after dMCAO (Figure S8A). Compared to vehicle treatment after dMCAO, minocycline could downregulate the levels both C1qa and VCAM-1, and maintain the expression of ZO-1, occludin, claudin-5, and VE-cad (Fig. 9E-K). Immunofluorescent assay also revealed that minocycline significantly reduced the co-labelling of C1q and Iba-1 in ipsilateral VPN after dMCAO (Fig. 9L-M). Similarly, the colocalization of VCAM-1 and RECA-1 was decreased after minocycline treatment in dMCAO rats (Fig. 9N). Furthermore, minocycline administration could increase the number of NeuN positive cells, and decrease the densities of GFAP and Iba-1 positive cells with smaller cell bodies and relatively elongated processes in the ipsilateral VPN after dMCAO (Fig. 9O-R).

Fig. 9.

Minocycline alleviates BBB disruption by downregulating microglial C1q-mediated VCAM-1 upregulation in ipsilateral VPN after dMCAO. A Binding conformation of minocycline to C1q. Left: overall view of minocycline in the binding position of C1q. Middle: three-dimensional (3D) close-up view of the binding conformation of minocycline on the C1q structure. Right: the enlarged image of the binding conformation of minocycline on the C1q structure. B Free energy distribution profile of the C1q-minocycline complex. C Experimental design/timeline of minocycline injection. D Representative western blot showing the expression of C1qa in different dose of minocycline treatment (+ : 10 mg/kg; + + : 20 mg/kg; + + + : 40 mg/kg, n = 3 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. Each bar represents the mean ± SD. *p < 0.05 vs Sham + Veh group, ^#p < 0.05 vs dMCAO + Veh group. E Representative western blot showing the expressions of C1qa, VCAM-1, ZO-1, OCLN, CLDN5, and VE-cad in ipsilateral VPN, respectively. F-K Quantitative analyses of C1qa, VCAM-1, ZO-1, OCLN, CLDN5, and VE-cad level relative to GAPDH from Sham and dMCAO animals with MC or Veh administration (n = 4 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. L Representative images show the co-localization of Iba-1 (green), C1q (red) and DAPI (blue) in ipsilateral VPN from dMCAO animals with MC or Veh administration. Scale bar: 25 μm. M Quantitative analysis of Iba-1⁺-C1q⁺ cells from Sham and dMCAO animals with MC or Veh administration (n = 6 in each group). Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. N Representative images show the co-localization of VCAM-1 (green) and RECA-1 (red) in ipsilateral VPN from dMCAO animals with MC administration. Scale bar: 25 μm. O Representative microphotographs of immunostaining of NeuN, GFAP, and Iba-1 in ipsilateral VPN. Scale bar: 75 µm. (P-R) Quantitative analyses of NeuN positive cells (P), and GFAP (Q) and Iba-1 (R) positive densities were showed from (O) in ipsilateral VPN from Sham and dMCAO animals with MC or Veh administration (n = 6 in each group). Statistical analysis was performed using one-way ANOVA followed by Bonferroni test. Each bar represents the mean ± SD. *p < 0.05 vs Sham + Veh group, ^#p < 0.05 vs dMCAO + Veh group, ^&p < 0.05 vs Sham + MC group

To demonstrate that minocycline treatment could preserve the integrity of BBB in the ipsilateral VPN of thalamus and improve neurological function after dMCAO, immunofluorescent analyses for the leakage of albumin, IgG, fibrinogen and infiltration of CD8a positive cells as well as neurological behaviour evaluation have been performed. Our results revealed that minocycline administration alleviated the leakage of albumin, IgG, fibrinogen, and decreased the number of CD8a positive cells in the ipsilateral VPN after dMCAO (Fig. 10A-E), suggesting that minocycline preserves BBB integrity. Next, the neurobehavioral assessments were conducted. Rats in minocycline group exhibited lower scores in the Bederson, beam-walking and asymmetry tests than those in the vehicle group after dMCAO (Fig. 10F-H). Then, MWM was used to assess the cognitive function of rats. After 5 days spatial orientation training, all rats improved their capacity to find the platform. However, the path length and escape latency in dMCAO rats were longer than those of Sham. There was no significant difference in the mean swimming speed among all the observed groups (Fig. 10I). Notably, minocycline administration significantly shortened the path length from the 1 st to 5th day of training and escape latency after dMCAO (Fig. 10J-K). Moreover, in the probe phase, rats treated with minocycline spent longer time in the target quadrant occupancy after dMCAO (Fig. 10L-M).

Fig. 10.

Minocycline alleviates BBB disruption and improves neurological functions at 2 w after dMCAO. A On day 14 post operation, representative images of extravascular albumin, IgG and fibrinogen deposits, and CD8a positive cells in ipsilateral VPN were displayed from Sham and dMCAO animals with MC or Veh administration. Scale bar: 25 μm. B-E On day 14 post operation, quantitative analyses of the areas of albumin, IgG and fibrinogen leakage and the number of CD8a positive cells in the ipsilateral VPN from Sham and dMCAO animals with MC or Veh administration. (n = 6 in each group). Statistical analysis was performed using Kruskal–Wallis H test. Each bar represents the mean ± SD. *p < 0.05 vs Sham + Veh group, ^#p < 0.05 vs dMCAO + Veh group, ^&p < 0.05 vs Sham + MC group. (F–H) Neurological functions were measured on day 1 before dMCAO and on day 1, 7, 14 after dMCAO with MC or Veh administration (n = 10 in each group). Quantitative analyses of Bederson’s scores (F), beam-walking scores (G), and asymmetry scores in adhesive removal test (H). Statistical analysis was performed using Kruskal–Wallis H test. Data are expressed as mean ± SD. *p < 0.05 dMCAO + Veh vs Sham + Veh; ^#p < 0.05 dMCAO + MC vs dMCAO + Veh group; ^&p < 0.05 dMCAO + MC vs Sham + MC. I-M Morris water maze test was conducted. The swimming speed (I), the path length (J) and escape time (K) were recorded during the training period. Statistical analysis was performed using one-way ANOVA followed by Tamhane’s T2 test. Representative path tracings in each quadrant during the training period (learning) and the probe trial (memory) (L). The percentage of time spent in the target quadrant to total time (30 s) was recorded at 14 days after dMCAO (M). Statistical analysis was performed using one-way ANOVA followed by Bonferroni test. Data are expressed as the mean ± SD. *p < 0.05 vs Sham + Veh group; ^#p < 0.05 vs dMCAO + Veh group (n = 6 in each group)

Discussion

The activation of different microglial subtypes at different stages may have different effects on the ipsilateral thalamus after cortical infarction. In this study, with scRNA-seq analysis we identified eight distinct microglia subclusters in the ipsilateral VPN of thalamus after dMCAO. We further found that C1q derived from M6 triggers the upregulation of VCAM-1 by binding to CD93 in endothelial cells at 2 weeks after dMCAO, resulting in BBB disruption in the ipsilateral VPN. Targeting the C1q-CD93 signaling pathway or knockdown VCAM-1 in endothelial cells attenuated VCAM-1 expression and maintained BBB integrity, consequently mitigating secondary thalamic damage after dMCAO. Additionally, minocycline exerted neuroprotective effects by inhibiting the expressions of both microglial C1q and VCAM-1 in the ipsilateral VPN, thereby alleviating BBB disruption after dMCAO. These results reveal previously unknown interactions between microglia and endothelial cells within the ipsilateral VPN and support the therapeutic targeting of C1q in ischemic stroke.

The BBB disruption leads to the extravasation of immune cells, inflammatory cytokines and toxic blood proteins from periphery into brain. This impairs neurovascular function and leads to persistent brain damage [30, 31]. Previously, we had demonstrated that the presence of BBB disruption as evidenced by the leakage of albumin in the ipsilateral VPN of thalamus after dMCAO [8]. In the present study, we confirmed that increased BBB permeability, decreased TJs proteins and structural damage of BBB in the ipsilateral VPN occured, which began from week 2 after dMCAO. Just as Jones et al. reported, two weeks post-stroke is a period of intense neuroinflammation at the site of secondary neurodegeneration following cortical infaction [10]. Therefore, week 2 may represent a critical time window for intervention. It marks the convergence of emerging BBB disruption and intensified neuroinflammation with the exacerbation of other injury-related processes, such as neuronal death, gliosis, and β-amyloid deposition [32–34]. In present study, BBB disruption in the ipsilateral VPN of the thalamus begins at 2 w after dMCAO, and the activation of microglia are more pronounced. The microglial activation was earlier than the BBB disruption, indicating that it correlated with the BBB disruption and that BBB disruption isn’t the initial trigger of neurodegeneration, but the result. Notably, toxic substances, leukocytes, and pathogens from the circulation would enter the brain parenchyma following BBB disruption, further activate and amplify microglial inflammatory responses, thereby promoting progression of neurodegeneration [35, 36].

Previous studies have established microglia as a primary cellular source of C1q in CNS [37]. Through transcriptomics and proteomic analyses we showed that DEPs were predominantly enriched in complement and coagulation cascades in the ipsilateral VPN after dMCAO. In the complement pathway, the upregulations of C1qa, C1qb, and C1qc were most obvious. Further, we confirmed that the mRNA and protein expressions of C1qa, C1qb and C1qc were increased significantly in the ipsilateral VPN at 1–4 weeks, and peaked at 2 weeks after dMCAO. There have been reports that C1q localizes to synapses during development and is required for synaptic elimination [38, 39]. Microglia mediate activity-dependent synaptic pruning via the C3-CR3 pathway [40]. Previously we verified that the disruption of neural circuits between the cortex and thalamus after dMCAO [7]. Reduced neuronal activity following deafferentation may trigger the activation of microglia post cortical infarction. Dejanovic et al. reported that the percentage of C1q-labeled synapses in the hippocampal CA1 in Tau-P301S mice was higher when compared to Sham group. Most of C1q is deposited at non-synaptic sites on neurons and possibly other cells, indicating that non-synaptic structures on other cells might be the targets of microglial engulfment or complement-mediated damage [41]. Similarly, our results revealed that most of C1q deposited at non-synaptic sites, in the VPN of thalamus after dMCAO.

The upregulation of C1q has been found to exacerbate BBB disruption in the cerebral cortex following intracerebral hemorrhage [42]. Kitic et al. reported that C1q was upregulated in the striatum and deposited on endothelial cells, leading to BBB disruption, enhanced peripheral immune cell infiltration, and exacerbated tissue damage in neuromyelitis optica disease [43]. The microglia-derived C1q was verified to drive chronic inflammation and neuronal loss in the thalamus in a traumatic cortical injury model [22]. Consistent with these findings, our results demonstrate that C1q was mainly derived from microglia. The knockdown of microglial C1q significantly alleviates BBB disruption and neuronal loss and improves the neurological outcomes.

Accumulating evidence supports that C1q can initiate multiple intracellular signaling cascades, including p38 mitogen-activated protein kinase (MAPK), extracellular regulated protein kinases (ERK) 1/2, and phosphatidylinositol-3-kinase (PI3K)/AKT [44]. In HT1080 cells, C1q drives the expressions of matrix metalloproteinases (MMP) 2 and 9 via p38 MAPK/ERK1/2 pathways [45]. MMP 2/9 can directly degrade key tight junction components of the BBB [46–48]. The activation of above signaling pathways decreased expression of VE-cadherin, which would impair endothelial barrier integrity [49, 50]. In this study, the simultaneous recovery of ZO-1, occludin, claudin-5, and VE-cadherin by C1q inhibition may result from a general preservation of BBB integrity in the ipsilateral VPN after dMCAO. What needs to be pointed out is that our results are insufficient to establish that C1q directly regulates tight and adherens junction complexes. To verify the mechanistic link between C1q and tight and adherens junction complexes, extensive experiments should be performed in future study.

For a long time microglia were categorized into pro-inflammatory (M1) and anti-inflammatory (M2) phenotypes. Nevertheless, this binary classification fails to capture their remarkable regional specialization and functional complexity in vivo [51]. Elucidating the functional heterogeneity of microglia is crucial for developing targeted therapies for neurological diseases. Li et al. identified distinct ischemia core- and penumbra-associated microglial subpopulations in the ischemic cortex at 3 days after MCAO [52]. Marschallinger et al. discovered lipid droplet-accumulating microglia in the hippocampus of aged mice [53]. However, the precise subtypes of microglia and the expression of C1q in different subtypes of microglia after focal cerebral ischemia remain poorly understood. Herein, we identified a specific microglial subcluster named M6. It was highly enriched in C1qa, C1qb, and C1qc transcripts, suggesting its critical role as a regulator of BBB integrity in ipsilateral VPN after dMCAO. Further analysis revealed that M6 subcluster was associated with antigen processing and presentation, Fc epsilon RI signaling, and C-type lectin receptor signaling pathways. These processes are intrinsically linked to complement activation [54–56].

In this study, there were significant transcriptional differences between M6 type microglia and disease-associated microglia (DAM) as well as conventional M1/M2 type microglia. These differences were manifested by the low expression of genes such as SPP1, ItgaX, Nlrp3, Csf1, and CD206 [57], while the high expression of genes related to the complement pathway. The latter characteristics distinguish M6 type microglia from other established phenotypes in the microglial response spectrum, which may play a specialized role in complement-mediated BBB disruption in the ipsilateral VPN after dMCAO,

Given that C1q was mainly derived from M6 microglia subcluster in the ipsilateral VPN after dMCAO, to understand the factors governing M6 fate selection is of importance. Evidence suggests that microglial fate selection is largely dictated by specific gene expression programs [58]. Zhao et al. found that the overexpression of microglial CYP46A1 reduced the formation of stroke-associated foamy microglial clusters in the ischemic cortex after MCAO [59]. In our study, pseudotime trajectory analysis positioned the M6 subtype as a terminal state of microglial activation in the ipsilateral VPN. This subcluster showed a robust transcriptional signature of complement activation and innate immune response. This suggests that complement gene activation may actively promote the differentiation of microglia into the M6 subcluster. Intriguingly, we noted that the M6 subtype differs from the terminally inflammatory microglia (TIM) identified in a mouse model of Alzheimer's disease, which exhibits a less inflammatory, functionally impaired, exhausted-like state [60]. One plausible explanation for this phenomenon is that TIM may represent a terminal state of activated microglia [61], whereas M6 microglia may be in a transitional pro-inflammatory and pathogenic state. This interpretation is supported by the observation that transitional microglia in Alzheimer's disease mice highly expressed the complement C1qa and C1qb [62]. We consider that driving microglia toward unique M6 subcluster after dMCAO may arise in response to the distinctive deafferented microenvironment within the thalamus or to damage-associated signals released by degenerating neurons or stressed glial cells in the VPN. Since the time point for intervention in this study is at 2 weeks after dMCAO, it is necessary to determine whether M6 subcluster represents a transition to a TIM-like state or the maintenance as a distinct terminal state, future targeted lineage tracing experiments on the M6 subcluster will help to understand the fate of these cells.

The interaction between microglia and endothelial cells is critical for BBB integrity in pathological conditions. It is well established that C1q can upregulate VCAM-1 through interacting with the 100- to 126-kDa C1q-binding protein on the endothelial cells. This causes endothelial damage and ultimately leads to vascular hyperpermeability [23]. By bioinformatic analysis we predicted a strong interaction between microglial C1q and the CD93. CD93, type I transmembrane glycoprotein. CD93 is widely expressed in endothelial cells and has been implicated in regulating vascular endothelial adhesive function [63, 64]. In the vascular tissues of human cerebral cavernous malformations, increased CD93 expression leads to VCAM-1 upregulation in endothelial cells and vascular injury [65]. In line with these findings, we demonstrated that CD93 was mainly expressed in endothelial cells and obviously upregulated. Also, the interaction between C1qa and CD93 was enhanced, and the expression of VCAM-1 upregulated in the ipsilateral VPN after dMCAO. Importantly, the inhibition of the C1q-CD93 interaction with AAV-shC1qa or C1q NAb would downregulate VCAM-1 expression on endothelial cells in the ipsilateral VPN after dMCAO. Consistent with in vivo results, C1q was found to upregulate VCAM-1 expression in a co-culture system of BECs with LPS treatment. Oppositely, the expression of VCAM-1 was significantly decreased after C1q ablation. Furthermore, with DCBY02 to block CD93, the expression of VCAM-1 was downregulated and BBB disruption alleviated.

Since VCAM-1 is strongly linked to the homeostasis of endothelial cells, it is crucial for maintaining the integrity of the BBB. Under pathological conditions, endothelial cells are in a quiescent state and maintain the stability of the BBB through TJs. When activated, they show upregulated expression of VCAM-1, which would compromise the stability of the BBB [19, 66]. A study from Gao et al. confirmed that inhibiting the expression of VCAM-1 in endothelial cells could alleviate BBB disruption in the ischemic cortex after MCAO [67]. Targeted knockdown of VCAM-1 in endothelial cells can effectively reduce infarct volumes at 6 days after MCAO in mice [68]. Consistent with the above studies, our results showed that the knockdown of VCAM-1 in endothelial cells could decrease BBB permeability and maintain the expression of TJ proteins and adherens junction protein in the ipsilateral VPN and improve neurological functions after dMCAO. In short, our data support that C1q-CD93-VCAM-1 axis mediates the activation of endothelial cells. By blocking this signal axis, the activation of endothelial cells was inhibited. In this context, the recovery of tight junction and adherens junction proteins may serve as an indicator of attenuated endothelial activation and improved BBB integrity. Nevertheless, the existing data could not confirm C1q signaling directly regulates endothelial junctions. More additional mechanistic experiments would be necessary in future study.

There have been reports that NF-κB and p38 MAPK-related may act as potential downstream pathways of CD93 activation in diverse inflammatory contexts [69, 70]. Importantly, both pathways are known to transcriptionally upregulate VCAM-1 expression [71, 72]. The upregulation of VCAM-1 on vascular endothelial cells facilitates the infiltration of leukocytes including neutrophils, monocytes and lymphocytes following ischemic stroke [67, 73]. For example, VCAM-1 can bind to very late antigen (VLA) 4 on T cells, promoting T cell adhesion, activation, and infiltration [74]. Activated T cells upregulate MMP 2 and MMP-9 in cerebral endothelial cells by secreting IL-17A. These MMPs can degrade tight junction proteins and increase BBB permeability [75–77]. Furthermore, neutrophil infiltration into the ischemic region leads to excessive secretion of MMP-9 and proinflammatory cytokines such as IL-1β, which further exacerbates BBB disruption [78, 79]. In this study, marked T cell infiltration was found in the ipsilateral VPN after dMCAO.

Although thrombolysis and endovascular treatment have significantly enhanced the functional outcomes of acute ischemic stroke patients [80–82], these treatments are only available for a limited number of patients because of the short time window. A major challenge remains in the prevention and treatment of delayed secondary damage in brain regions [83]. Therefore, alleviating secondary neurodegeneration can be a promising target of neuroprotection beyond the therapeutic time window for acute stroke. More encouragingly, minocycline has been recognized for its neuroprotective effects through inhibition of microglia-mediated inflammation and BBB disruption in stroke [84, 85]. In a model with basal ganglia hemorrhage, Zheng et al. found that minocycline treatment can inhibit the expression of C1q in microglia, alleviate basal ganglia damage, and improve neurological functions of mice [86]. Using molecular docking, we revealed that minocycline can interact with microglial C1q. We also verified that minocycline administration downregulated the expression of C1qa, inhibited the activation of microglia and downregulate the expression of TNF-α and IL-1β. Besides, minocycline could downregulate VCAM-1 expression, increase TJs protein levels and maintain BBB integrity, thus conferring neuroprotection. Therefore, it is reasonable to believe that minocycline affords neuroprotection against thalamic inflammation after cortical infarction via two potential ways. On the one hand, minocycline selectively binds to the ligand of C1q and may alter its structural configuration and further blocks C1q-mediated BBB disruption. On the other hand, minocycline inhibits the expression of C1q through suppressing microglial activation and inflammatory cytokine secretion. Cheng et al. demonstrated that long-term minocycline treatment plays neuroprotective roles in inhibiting inflammatory phenotype of microglia after ischemic stroke [87]. Recently. Lu et al. reported that early administration of minocycline in patients with ischemic stroke can improve long-term neurological function, without exerting a significant effect on neurological function in the acute phase [88]. Based on the above findings, we consider that early treatment with minocycline for the patients with cerebral infarction, if combined with recombinant tissue-type plasminogen activator (rtPA) or thrombectomy, may be not only to re-canalize the blood vessel, but also to improve the long-term prognosis by neuroprotection. It is possible that minocycline becomes an important candidate for the treatment of secondary degeneration after cortical infarction. However, further studies are needed to elucidate when minocycline treatment should be initiated after cortical infarction, how long it used, and which administration route is optimal for alleviating secondary thalamic damage.

In conclusion, this study explored the role of targeting microglial C1q in alleviating BBB disruption in the thalamus after cortical infarction. For the first time, we uncovered that microglial C1q derived from M6 subtype binds to endothelial CD93, upregulates VCAM-1, and leads to BBB disruption in the ipsilateral VPN after dMCAO. The inhibition of the interaction between C1q and CD93 could maintain BBB integrity and protect against secondary thalamic neuronal injury after dMCAO. What is of crucial clinical significance is that minocycline treatment protected the vascular endothelium, attenuated BBB disruption and neuronal loss in the ipsilateral VPN, thereby improving neurological outcomes of rats. These observations would shed light on microglial C1q as an important intervention target and led to possibilities for specific ischemic stroke therapies.

Materials and methods

Animals

The animal experiments were conducted on adult male SD rats weighing 280–320 g (10–12 weeks) (from Southern Medical University, Guangzhou, China). Rats were housed in a temperature-controlled (22 ± 1℃) and 12-h light/dark cycle environment, having free access to food and water. All efforts had been made to minimize the suffering and the number of animals.

In this study, 487 rats were used, 5 died during the surgery. One in the sham-operated (Sham) and 1 in the dMCAO groups died during the dMCAO procedure. One rat died after intracerebroventricular injection, and 2 died during anesthesia. Sixteen rats that presented without cortical infarcts after surgery were excluded.

Distal middle cerebral artery occlusion model

A permanent dMCAO was performed with a unipolar electrocoagulation as previously reported [89]. Briefly, animal anesthesia was induced with isoflurane (3–4%) in a chamber and maintained with isoflurane (1.5–2.5%) in air, delivered through a nose mask (SurgiVet, Waukesha, WI, United States) during the surgical procedure. The distal striatal branch of left MCA was exposed and then occluded by a unipolar electrocoagulation device. Sham rats received the same surgical procedures except for left dMCA electrocoagulation. After surgery, neurological functional evaluations of rats were conducted. All animals exhibited normal feeding and drinking behaviors after operation.

Single cell RNA sequencing

Samples from the ipsilateral VPN of Sham and dMCAO 2 weeks were collected (n = 4 in each group) for single cell RNA-seq (scRNA-seq). The brains were removed at 2 weeks after dMCAO. Cell viability was assessed using an automated cell counter. Samples exhibiting > 90% viability were used for scRNA-seq library preparation. Reverse transcription and library construction were performed on qualified samples. Subsequent bioinformatics analyses of sequencing data were also conducted by OE Biotech Co., Ltd (Shanghai, China). To eliminate the potential batch effects between samples (Sham and dMCAO groups) and to ensure the accurate integration of datasets, the mutual nearest neighbors (MNN) algorithm [90] was applied using the batchelor package (version 1.6.3) in R programming. The data quality control (QC) was performed firstly using Cell Ranger, the official software of 10 × Genomics. Based on the preliminary quality control results from Cell Ranger, further quality control was conducted on the experimental data to exclude data derived from multicellular aggregates, doublets, or unbound cells. The quality control criteria are as follows: cells were retained only if they met the thresholds of number of detected genes > 200, number of UMIs > 1000, log10GenesPerUMI > 0.7, mitochondrial UMI percentage < 15%, and red blood cell gene percentage < 5%. Potential doublets were further removed using DoubletFinder software. In this study, the resolution settings used during clustering and annotation were 0.4

Evaluation of BBB disruption

Since Evans blue (BB) can bind to albumin (67 kDa) in the circulation, EB assay was used to reflect BBB disruption in this study. For EB assay, EB dye (2%, 2 mL/kg; Sigma-Aldrich) was administered intravenously via the tail vein at 1, 2, 3, 4 weeks after dMCAO, respectively. After 1 h of circulation, the mice were anesthetized with 4% isoflurane and then perfused with 0.9% normal saline (NS), followed by 4% cold paraformaldehyde (PFA) in phosphate-buffered saline (PBS, 0.01 M, pH 7.4). The brains were removed quickly and post-fixed in 4% PFA for 12 h, and then cryoprotected overnight with 10%, 20%, and 30% sucrose in the same fixative. Coronal tissue blocks were cut into 30 µm-thick sections on a freezing microtome (Leica, Wetzlar, Hessen, Germany). The area of albumin leakage in the ipsilateral VPN of thalamus was analyzed by a confocal laser microscope (SP8, Leica Microsystems, Wetzlar, Hessen, Germany). For immunoglobulin G (IgG) (150 kDa) staining, the brain sections were stained with Cy3-conjugated goat anti-mouse IgG antibody (1:200; Merck, Cat# AP124C).

Transmission electron microscopy

Rats were anesthetized with 4% isoflurane and transcardially perfused with 0.9% NS followed by ice-cold 4% PFA (PH 7.4), containing 0.2% glutaraladehyde, in 0.01 M PBS. Brain was removed and then washed in 0.01 M PBS. Frontal 50 μm sections of whole brain were cut on a vibratome (VT1200, Leica, Germany). The brain sections were rinsed with PBS, and immersed for 0.5 h with 3% H₂O₂ to block the endogenous peroxidase activity. The brain sections were postfixed in 1% osmium tetroxide in 0.01 M PBS, dehydrated in a graded series of ethanol and propylene oxide, and flat-embedded in Epon. Ultrathin sections were cut with a diamond knife and then counterstained with lead citrate and uranyl acetate. Ultrathin sections were mounted on grids, examined and photographed using TEM (Tecnai G2 Spirit, FEI). Ten electron micrographs of VPN synapse were taken per section with × 6800 magnification.

Western blot

Rats were sacrificed at 1, 2, 3, and 4 weeks after operation, respectively. The brain tissue was cut into 2-mm coronal slices using a brain matrix (Shuolinyuan Technology Co., Ltd., Beijing, China) and the VPN regions of bilateral thalamus were quickly divided under a stereomicroscope. Proteins of the VPN were extracted and western blotting procedure was performed as previously described [8]. To determine protein concentration, bicinchoninic acid (BCA) method was recommended by the manufacturer (Beyotime). The proteins of each sample were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred to polyvinylidene fluoride (PVDF) membranes (MilliporeSigma, Burlington, MA, USA). The primary antibodies used in these studies include ZO-1 (1:2000; Proteintech, Cat# 21773–1-AP), occludin (1:2000; Proteintech, Cat# 27260–1-AP), VE-cadherin (1:2000; Immunoway, Cat# YT4869), C1qa (1:500; Santz crus, Cat# sc-58920), C1qb (1:1000; Signalway Antibody, Cat# 40660), C1qc (1:1000; Signalway Antibody, Cat# 34608), CD93 antibody (1:1000; Santz crus, Cat# sc-365172), VCAM-1 (1:2000; Abcam, Cat# ab134047), ICAM-1 (1:2000; Immunoway, Cat# YT2269) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH, 1:10000; Proteintech, Cat# 60004-I-Ig). Densitometric analysis for the quantification of the bands was performed with Image J (NIH, Bethesda, MD, United States). Relative optical densities of protein bands were calibrated with GAPDH and normalized to those in Sham rats.

RNA sequencing analysis

RNA-seq was performed by Shanghai OE Biotech Co., Ltd (China). Samples from the ipsilateral VPN of Sham and dMCAO 2 weeks were collected (n = 3 in each group) and immediately delivered to the company for library preparation and sequencing. The cDNA libraries were constructed following the manufacturer’s standard protocols. Clean reads were aligned to the Rattus norvegicus reference genome (assembly: NCBI_GCF_000001895.5). For functional enrichment analysis, all differentially expressed genes (DEGs) were annotated based on Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. Subsequent data processing and visualization were conducted using the OE cloud platform (https://cloud.oebiotech.cn/#/home).

Proteomics analysis

Proteomics analysis was performed to identify differentially expressed proteins (DEPs) in the VPN between the Sham and dMCAO 2 weeks groups (n = 3 in each group). The analysis was conducted by Shanghai OE Biotech Co., Ltd. (Shanghai, China). Briefly, proteins were extracted from VPN tissues of both groups, and their concentrations were quantified. The proteins were then digested using Trypsin-TPCK. The protein samples following enzymatic hydrolysis were marked. The labeled samples were analyzed by liquid chromatography (LC)-mass spectrometry (MS)/MS, and the raw data were acquired through database searching. DEPs were identified based on predefined criteria, and functional enrichment analysis was performed using the KEGG, GO databases, and Wikipathways databases.

Quantitative real-time polymerase chain reaction

Total RNA was isolated from the VPN of thalamus using Trizol reagent (Invitrogen, Carlsbad, CA). Quantitative real-time polymerase chain reaction (qRT-PCR) was performed. For each sample, 1 μg of RNA was reverse transcribed into cDNA in a final volume of 20 μL, containing 1 μL primer mix, 1 μL enzyme mix, and 4 μL buffer (Takara, Shiga, Japan). The primers designed by Sangon Biotech are as follows: C1qa: 5’-ATGGAGACCTCTCAGGGATG-3’ (forward), 5’-ATACCAGTCCGGATGCCAGC-3’ (reverse);

• C1qb: 5’-CAGGGATAAAGGGGGAGAAA-3’ (forward), 5’-TCTGTGTAGCCCCGTAGTCC −3’ (reverse);

• C1qc: 5’-CATCCTTGCCTAGACCATTC-3’ (forward), 5’-GTACCAGAAGGCATTGGTTA-3’ (reverse);

GAPDH: 5’-ACCACAGTCCATGCCATCAC-3’ (forward), 5’-TCCATGGCGAACTGGTGGCG-3’ (reverse). Then, PCR was performed in an ultimate volume of 20 μL with 2 μL of RT product for cDNA amplification. Annealing temperature was 60 ℃. Quantitative PCRs were conducted by LightCycler Fast-Start DNA Master SYBR Green 1 kit and on a LightCycler 1.5 PCR machine (Roche Light Cycler 480, Germany). Each sample was run in triplicates. All thalamic samples were quantified against the identical standard curve, and the levels of C1qa, C1qb and C1qc were normalized to GAPDH. Data were analyzed by using the comparative Ct method (2^−ΔΔCt). Results were expressed as fold changes compared to Sham.

Immunohistochemistry

The rats from each group were perfused intracardially with 0.9% NS and 4% cold PBS. All brain tissues were removed quickly and postfixed in 10%, 20%, and 30% sucrose in the same fixative overnight at 4 °C for cytoprotection. After post fixation, the brains were frozen at −20 °C and sliced into coronal 30-μm-thick sections with cryotome (Leica, Wetzlar, Hessen, Germany). The coronal sections that included the VPN were used. Single-labeled immunohistochemistry was detected by the avidin-biotinperoxidase complex (ABC) method. Briefly, the sections were first soaked with 3% hydrogen peroxide for 30 min, followed by 5% normal serum for 1 h at room temperature, and then incubated with mouse monoclonal antibody against neuronal nuclei antigen (NeuN, 1:5000; Millipore, Cat# MAB377), mouse monoclonal antibody against glial fibrillary acid protein (GFAP, 1:5000; Millipore, Cat# MAB360), rabbit monoclonal antibody against ionized calcium binding adaptor molecule-1 (Iba-1, 1:5000; Wako, Cat# 019–19741) overnight at 4 °C. Then, the slides were washed with PBS for six times. Afterwards, the slides were incubated with biotinylated secondary IgG antibody for 2 h at ambient temperature. After being washed with PBS for three times, the sections were subsequently soaked with ABC for 30 min at ordinary temperature. Immunopositive cells in which the reaction product was present within a clear and regular-shaped cytoplasmic or nuclear border were quantified under a light microscope with × 660 magnification. The total number of immunopositive cells in the thalamus was counted within three nonrepeated random fields (0.037 mm²/field × 3 = 0.111 mm² in total).

Double-fluorescent or triple-fluorescent immunohistochemistry was performed as described previously [8]. It was conducted to demonstrate cell types and the exact position where C1q or VCAM-1 was expressed. NeuN, glial fibrillary acidic protein (GFAP), and ionized calcium binding adaptor molecule-1 (Iba-1) were used to identify nuclei of neurons, astrocytes, and microglia, respectively. The co-staining of C1q and Iba-1 and the leakage areas of albumin and IgG were examined in infarction border zone. The primary antibodies used in these studies included rabbit anti-fibrinogen (1:4000; Proteintech, Cat# 15841–1-AP), anti-mouse CD8 alpha (1:200; Abcam, Cat# ab33786), rabbit anti-synapsin antibody (1:500; HUABIO, Cat# ET1606-56), rabbit anti-rat endothelial cell antigen-1 (RECA-1; 1:200, Abcam, Cat# ab9774), rabbit anti-C1q antibody (1:200; Abcam, Cat# ab182451), rabbit anti-VCAM-1 antibody (1:200; Abcam, Cat# ab134047), rabbit anti-Laminin antibody (1:200; Merck, Cat# L9393), mouse anti-CD93 antibody (1:200; Santz crus, Cat# sc-365172), mouse anti-NeuN (1:1000) monoclonal antibody, mouse anti-GFAP (1:1000) monoclonal antibody (1:5000), mouse anti-Iba-1 (1:500; Synaptic Systems, Cat# 234011) monoclonal antibody (1:1000). The secondary antibody included Cy3-conjugated goat anti-mouse IgG antibody (1:200), Cy3-conjugated goat anti-rabbit IgG antibody (1:200; Merck, Cat# AP132C), FITC-conjugated goat anti-mouse IgG antibody (1:200; Merck, Cat# AP124F), and FITC-conjugated goat anti-rabbit IgG antibody (1:200; Merck, Cat# AP132F), 647-conjugated goat anti-rabbit IgG antibody (1:200; Invitrogen, Cat# A21244). After being incubated with IgG antibody, sections were washed with PBS and mounted with mounting medium containing 4′, 6-diamidino-2-phenylindole (DAPI, Solarbio, Cat# S2110). Slides were analyzed with a confocal laser microscope (SP8, Leica Microsystems, Wetzlar, Hessen, Germany).

Immunostaining images were analyzed using ImageJ. For each animal, three non-overlapping fields were randomly selected in the VPN of thalamus. The extent of BBB disruption was evaluated by measuring the percentage of area occupied by the positive area of albumin (67 kDa), IgG (150 kDa) or fibrinogen (340 kDa), and the number of CD8a positive cells. In addition, the mean fluorescence intensities (MFI) of Iba-1 and CD93 were measured. To ensure accuracy, background fluorescence was subtracted by measuring the intensity in a non-stained region of the image and normalizing the MFI to this background value. Data were expressed as relative fluorescence intensity compared with the sham group. The number of CD8a⁺ or Iba‑1⁺‑C1q⁺ cells was manually counted by an observer blinded to the experimental groups.

Immunoprecipitation

Immunoprecipitation procedure was performed as previously described [8]. An amount of 120 μg of protein was incubated with mouse anti-CD93 monoclonal antibody (1:50; Santz crus, Cat# sc-365172) overnight at 4 °C. The the protein/antibody complex was added to packed protein G agarose beads (Millipore, Cat# IP05) in next day. Following 4 h of incubation at 4 ℃, the complex was washed five times with PBS containing 1% Tween. Immunocomplexes were collected by centrifugation and eluted by boiling in loading buffer. The eluted protein samples were subjected to western blot with antibodies including rat anti-C1qa (1:500) and mouse anti-CD93 (1:1000; Santz crus, Cat# sc-365172). Densitometric analysis was performed to quantify the relative precipitated protein bands, which were calibrated with the bands of CD93 and normalized to those in Sham rats.

Adeno-associated virus construction and administration

Adeno-associated virus (AAV) vectors expressing short hairpin RNA (shRNA) targeting rat C1qa (GenBank: NM_298566) or VCAM-1 (GenBank: NM_25361), along with their respective negative controls (CON728 and CON305), were constructed by GeneChem (Shanghai, China). The shRNA cassettes were driven by cell type-specific promoters (CX3CR1 for microglia and Tie1 for endothelial cells), yielding AAV-shC1qa (titer: 9.9 × 10¹² vg/mL) and AAV-shVCAM-1 (titer: 5.58 × 10¹³vg/mL). The microglia-targeting AAV vector was AAV2, which contains inverted terminal repeats (ITRs), the microglia-specific mCx3cr1 promoter, enhanced green fluorescent protein (EGFP), a multiple cloning site (MCS), woodchuck hepatitis virus posttranscriptional regulatory element (WPRE), and a polyadenylation (poly(A)) signal. The cerebrovascular endothelial cell-targeting adeno-associated virus (AAV) vector was AAV9‑X1.1, which consists of left inverted terminal repeat (L‑ITR), right inverted terminal repeat (R‑ITR), the endothelial cell‑specific tyrosine kinase with immunoglobulin-like and EGF-like domains 1 (Tie1) promoter, EGFP, a MCS, a U6 promoter, a poly(A) signal, an origin of replication (ori), an ampicillin resistance gene (Ampᴿ), and an f1 ori. For intracranial delivery, 2 µL (0.4 μL virus diluted by 1.6 μL PBS) of AAV-shC1qa was stereotaxically injected into the VPN (3.6 mm posterior, 3.0 mm lateral to bregma, 6.0 mm depth) at 0.2 µL/min using a Hamilton syringe. For systemic delivery, 150 µL (60 μL virus diluted by 90 μL PBS) of AAV-shVCAM-1 was injected via the tail vein. All animals recovered for 28 days before subsequent experiments.

BV2 cells culture and treatment

BV2 microglial cells were cultured in Dulbecco’s modification of Eagle’s medium (DMEM) (Gibco, Cat# C11995500BT) with 10% fetal bovine serum (Gibco, Cat# 10,270–106) and 100 U/mL penicillin/streptomycin (Gibco, Cat# 15,070,063). After 24 h of incubation, the cells were treated for an additional 24 h with either 1% dimethyl sulfoxide (DMSO; vehicle control) or lipopolysaccharide (LPS; MedchemExpress, Cat# HY-D1056) at concentrations of 0.01, 0.1, 1, or 10 ng/mL.

Primary microvascular endothelial cells cell extraction and treatment

Primary thalamic microvascular endothelial cells were isolated from rat brains using a density gradient centrifugation method [91]. Thalamic tissues were minced into 1 mm³ fragments and digested in a solution containing 0.1% type II collagenase and 30 U/mL DNase I at 37 °C for 1.5 h. The mixture was then centrifuged at 1000 g for 20 min at 4 °C. After discarding the supernatant, 20% bovine serum albumin (BSA) was added and homogenized gently by pipetting, followed by centrifugation at 1000 g for 20 min. The pellet was resuspended in 33% Percoll solution, mixed thoroughly and centrifuged at 1000 g for 15 min. The microvascular segments, appearing as a white and yellow layer above the erythrocyte layer, were collected, washed twice with DMEM, and centrifuged at 500 g for 5 min. Finally, the microvascular segments were resuspended in 5 mL of culture medium and then cultured in an incubator at 37 °C with 5% CO₂. The medium was DMEM supplemented with 10% FBS (Gibco, Cat# 10,270–106), 30 μg/mL EC growth supplement (Sciencell, Cat# 1052) and 100 U/mL penicillin/streptomycin (Gibco, Cat# 15,070,063). Endothelial cells were treated with 1% DMSO, LPS, LPS-treated BV2 cell culture medium, C1q recombinant protein (Medchemexpress, Cat# HY-NP0194C), or C1q neutralizing antibody (GeneTex, Cat# GTX54404), separately. After 24 h, proteins were harvested for subsequent analysis.

Pharmacologic interventions

The implantation of the intracerebroventricular injection cannula into the right lateral ventricle (1.5 mm lateral, 1.0 mm posterior, and 3.6 mm dorsal with respect to the bregma) was performed stereotaxically after isoflurane anesthesia. The cannula was affixed to the skull with stainlesssteel screws and cranioplastic cement. DCBY02 (Medchemexpress, Cat# HY-P990572), an inhibitor of CD93, or vehicle (5% DMSO) was administered at 3 days before dMCAO. DCBY02 was administered once daily for 14 consecutive days. All animals displayed normal feeding and drinking behaviors postoperatively.

To evaluate the neuroprotective effects of minocycline on thalamic injury after dMCAO, rats were treated with the minocycline intraperitoneally (Medchemexpress, Cat# HY-17412) or vehicle (5% DMSO) at 30 min after dMCAO. Minocycline was administered once daily for 14 consecutive days. Based on previous studies [92], we explored the neuroprotective effects of minocycline at doses of 10 mg/kg/d, 20 mg/kg/d, and 40 mg/kg/d, respectively. Finally, the dose at 40 mg/kg/d was chosen.

C1q neutralization in vivo

To evaluate the effect of C1q on thalamic injury after dMCAO, 4 µL of C1q neutralizing antibody (GeneTex, Cat# GTX54404) or vehicle control (mouse anti-IgG antibody; Proteintech, Cat# B900620) was injected into the ipsilateral VPN (3.6 mm posterior, 3.0 mm lateral to bregma, 6.0 mm depth) of thalamus using a Hamilton Flexifil Syringe over 10 min, followed by a 5 min delay prior to withdrawal of syringe. The concentration of C1q neutralizing antibody (100 μg/mL) was based on previous reports [93, 94]. All animals displayed normal feeding and drinking behaviors postoperatively.

Molecular docking

Molecular structures of minocycline were acquired through the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). Crystal structures of C1q were accessed via the Protein Data Bank (http://www.rcsb.org/ pdb/home/home.do). Molecular docking was conducted with Auto Dock Vina 1.2.5, and protein–ligand optimization involved PyMOL2.3.0. At least three molecular docking simulations were conducted to calculate the mean and standard deviation, thereby minimizing potential bias. The binding affinities were estimated to assess binding energies.

Molecular dynamics simulation

The simulation was performed via Gromacs 2024.4 software. Amber14sb was chosen as the protein force field, Gaff2 as the ligand force field. The TIP4P water model was chosen to add solvents to the protein–ligand system and establish a water box with a periodic boundary of 1.2 nm. Sodium and chloride ions were added to balance the charge in the system, so as to restore the real experimental environment as much as possible. The root mean square deviation (RMSD), root mean square fluctuation (RMSF) and radius of gyration (Rg) from the molecular dynamical simulation trajectories of the C1q-minocycline complex were analysis. The average binding free energy between the protein and ligand was calculated with the MM/GBSA method.

Neurobehavioral assessments

Neurobehavioral assessments of rats were performed via neurological deficit scoring, adhesive removal test, beam walking test, and morris water maze (MWM) by the investigators who were blinded to the experimental groups. Behavioral tests were conducted at 1, 7, 14 days after dMCAO. All rats were acclimated to the testing environment before surgery. All tests were almost conducted at the same time period each day to minimize variability in rat performance. The parameters for each trial were recorded by a video tracking system (SMART, Polyvalent video-tracking system, 35B73-C6C, PANLAB, Spain).

Neurological deficit scoring

Bederson scores were used to evaluate Neurological impairment, with forelimbs of normal rats extended towards the floor and flexion of the forelimb contralateral to the lesioned hemisphere occurring in ischemic rats. The behavioral scores were: 1, forelimbs curved with no other abnormality; 2, reduced resistance to lateral push towards the paralytic side; 3, circling to the paralyzed side when the flat ground is free to move; 4, spinning along the longitudinal axis of the body when the tail is raised.

Adhesive removal test

The adhesive removal test was performed by placing a 200 mm² piece of adhesive paper sequentially onto the ipsilateral and contralateral forepaw. The performance of rats was assessed by measuring the time required to sense and remove the adhesives. The time to remove the adhesives from the forelimbs was recorded in three replicates, with each trial separated by at least 3 min.

Beam walking test

The beam-walking test was used to assess motor balance function using a 122 cm long, 2.5 cm wide beam. Scoring was based on the following criteria: 0, falling from the beam directly; 1, unable to walk on the beam but remaining seated across it; 2, falling from the beam while walking; 3, walking along the beam, but the injured hindlimb not functioning in forward movement; 4, crossing the beam with more than 50% footslips; 5, crossing the beam with occasional footslips; 6, crossing the beam without footslips. All rats underwent three training trials per day for three days before dMCAO.

Morris water maze

Morris water maze (MWM) was conducted to assess cognitive function in rats. The aquatic maze consists of a circular black pool (210 cm in diameter, 60 cm in height; Taimeng Technology Co., Ltd, Chengdu, China) filled with opaque water (20 cm in depth, 23 ± 1℃). The pool was divided into four equal-sized quadrants (Q1, Q2, Q3, and Q4), in which a transparent platform (escape platform, 10 cm in diameter) was placed 2.0 cm below the water surface in Q4.

Rats were put into the pool facing toward the wall and every starting point was used in a different order each day. Starting positions were designated as Q1, Q2, Q3 or Q4. All rats were trained to find the hidden platform within 120 s and then stayed on the platform for 15 s. If the rats failed to find the platform, it was guided to get there, and the escape latency was recorded as 120 s. During the training period, the platform remained in the same location. All rats were subjected to 4 consecutive trials each day at 10 min intervals for 5 consecutive days from the 9rd days after dMCAO. The escape latency, path length and swimming speed for each trial were recorded.

Twenty-four hours after the end of a training period, a 30 s probe trial without the platform was conducted to assess the long-term memory. All rats were placed in water maze at a fastened starting point (the most distance to the platform position used during the training period). The time in the target quadrant occupancy was also recorded at the 14 days after dMCAO.

Statistical analysis

Statistical analysis was conducted with Statistical Package for Social Sciences Software for Windows, version 25.0 (SPSS, Inc., Chicago, USA). The Shapiro–Wilk test was used to assess the normal distribution of the data. For normally distributed data, a two-tailed t-test was applied for comparisons between two groups, while multiple group comparisons were performed using one-way ANOVA with subsequent Bonferroni correction for multiple pairwise comparisons. When the VPN of thalamus normally distributed and the variances are unequal, multiple comparisons were conducted using one-way ANOVA followed by Tamhane’s T2 or Dunnett test for multiple pairwise comparisons. For data with abnormal distribution and unequal variances, nonparametric tests were employed. The Mann–Whitney U test was utilized for comparisons between two groups, while the Kruskal–Wallis H test was applied for multiple group comparisons.

Abbreviations

Apoe

Apolipoprotein E

Atp1b2

ATPase Na⁺/K⁺ transporting subunit beta 2

BBB

Blood–brain barrier

BCA

Bicinchoninic acid assay

BSA

Bovine serum albumin

C1qa

Complement C1q A chain

C1qb

Complement C1q B chain

C1qc

Complement C1q C chain

CD93

Cluster of differentiation 93

CLDN5

Claudin-5

CNS

Central nervous system

Csf1

Colony stimulating factor 1

Csf1r

Colony stimulating factor 1 receptor

Cx3cr1

C-X3-C motif chemokine receptor 1

DEGs

Differentially expressed genes

DEPs

Differentially expressed proteins

dMCAO

Distal middle cerebral artery occlusion

EB

Evens blue

ERK

Extracellular regulated protein kinases

GAPDH

Glyceraldehyde-phosphate dehydrogenase

GFAP

Glial fibrillary acidic protein

GFP

Green fluorescent protein

GO

Gene Ontology

Hexb

Hexosaminidase B

Iba-1

Ionized calcium binding adaptor molecule-1

ICAM-1

Intercellular cell adhesion molecule-1

IgG

Immunoglobulin G

Itgax

Integrin, alpha X

KEGG

Kyoto Encyclopedia of Genes and Genomes

Lpl

Lipoprotein lipase

LPS

Lipopolysaccharides

Lyz2

Lysozyme-like 2

MAPK

Mitogen-activated protein kinase

Mbp

Myelin basic protein

MC

Minocycline

Mrc1

Mannose receptor C-type 1

Ms4a1

Membrane spanning 4-domain subfamily a member 1

Ndrg2

N-myc downstream regulated gene 2

NeuN

Neuronal nuclei

Nlrp3

Nucleotide-binding oligomerization domain, leucine-rich repeat and pyrin domain- containing 3

OCLN

Occludin

P2ry12

Purinergic receptor P2Y, G-protein coupled, 12

PBS

Phosphate buffered saline

PFA

Paraformaldehyde

PI3K

Phosphatidylinositol-3-kinase

Ptcra

Pre T cell antigen receptor alpha

PVDF

Polyvinylidene fluride

RECA-1

Rat endothelial cell antigen-1

S100a8

S100 calcium binding protein A8 S100

S100a9

S100 calcium binding protein A9 S100

Snap25

Synaptosome associated protein 25

SPP 1

Secreted phosphoprotein 1

Syt1

Synaptotagmin 1

TEM

Transmission electron microscope

Tgfbr1

Transforming growth factor beta receptor 1

Tmem119

Transmembrane protein 119

Tpsb2

Tryptase beta 2

Tyrobp

TYRO protein tyrosine kinase binding protein

umap

Uniform manifold approximation and projection

VCAM-1

Vascular cell adhesion molecule-1

VE-Cad

Vascular endothelial-cadherin

VLA-4

Very late appearing antigen-4

VPN

Ventral posterior nucleus

vWF

von Willebrand Factor

ZO-1

Zona Occludens 1

Authors’ contributions

E.X., L.P. and M.C. conceived the study, designed the experiments and assembled all the figures. L.P. performed the experiments with the assistance of M.C., X.Z., X.W., M.G. Z.Q. and W. S. This article was written by E.X., L.P. and M.C. All authors read and approved the final version of this article.

Funding

This work is supported by the National Natural Science Foundation of China (Grant No. 82471309).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All animal procedures and treatments were conducted in accordance with Animal Research: Reporting in vivo experiments guidelines and were approved and monitored by the Animal Care and Use Committee of Guangzhou Medical University (Guangzhou, China) (permission numbers: A2024-019).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.