snRNA-seq analysis of the mouse brain over the course of lipopolysaccharide challenge

This study employed both wild type and Anxa1 knockout (Anxa1−/−) C57BL/6 mice as experimental models. To validate the successful generation of the Anxa1−/− mouse model, Western blotting was performed on brain tissue samples obtained from both wild type and Anxa1−/− mice. As anticipated, the ANXA1 protein band was not detected in the Anxa1−/− mice, while it was observed in the wild type mice, thereby confirming the successful construction of Anxa1−/− mouse model. Furthermore, Real-time Polymerase Chain Reaction (PCR) was utilized to validate the absence of Anxa1 in the tails of the mice, providing further support to the findings obtained through Western blot analysis. This additional analysis further strengthens the evidence and confirms the successful establishment of the Anxa1−/− model (Supplementary Fig. 1C). Both wild type and Anxa1−/− mice were intraperitoneally injected with either Escherichia coli lipopolysaccharide (LPS; serotype 0111: B4, Sigma; 15 mg/kg body weight) or a control solution of endotoxin-free saline. During the 132-hour observation period, notable weight loss was observed in both wild type and Anxa1−/− mice following LPS administration. However, after 72 h, their weights stabilized or even increased, suggesting the occurrence of intense inflammatory responses during the initial phases of simulated endotoxemia (Supplementary Fig. 1D).

We employed both wild type and Anxa1−/− C57BL/6 mice as experimental models to investigate the intricate cellular composition and transcriptional pathways in the central nervous system (CNS) during sepsis or SAE. Through snRNA-seq and high-resolution spatial transcriptomic analyses, we examined brain tissues obtained at 12 h, 24 h, and 72 h post LPS challenge. After rigorous quality control and filtering procedures, a toral of 376,909 high-quality single nucleus cells were obtained. On average, each cell contained approximately 3028 unique molecular identifiers (UMIs) and expressed 1,431 genes (Supplementary Table 2). Through our rigorous analysis, we successfully identified nine clusters representing six distinct cell types within the CNS. These cell types encompassed neurons, microglia, astrocytes, oligodendrocytes, oligodendrocyte precursor cells, and vascular cells (Fig. 1B). Specifically, neurons were characterized by the expression of Syt1 [38], oligodendrocytes exhibited the marker Cldn1 [39], and oligodendrocyte precursor cells displayed Pdgfra expression [40,41,42]. Notably, microglia and astrocytes were further subdivided into two subclusters based on their distinctive gene expression patterns. Microglia were classified as Tmem119+ (Micro-1) [43, 44] and Gpr84+ (Micro-2) [45], while astrocytes were divided into Cldn10+ (Astro-1) and Igtp+ (Astro-2) subclusters [46]. Furthermore, vascular-associated cells were also classified into two subclusters: Col1a2+ (Vas-1) [47] and Vtn+ (Vas-2) [48] (Fig. 1B and C). Our findings are consistent with previous studies investigating cellular diversity in the mouse cortex and hippocampus [49]. Interestingly, no substantial differences in cell types were observed between Anxa1−/− and wild type mice following LPS challenge, suggesting that Anxa1 does not influence the composition of cells involved in brain inflammatory responses (Fig. 1D). However, notable changes were detected in the UMAP plots (Fig. 1E), particularly in the Astro-2 and Micro-2 populations. These subpopulations initially exhibited an increase in abundance, followed by a subsequent decrease at 24 h, and eventually returned to baseline levels by 72 h in both mouse groups. At the 12-hour time point, we observed a comparable increase in the astro-2 and micro-2 populations in both Anxa1−/− and wild type mice. This initial response suggests a shared early reaction in these populations irrespective of the Anxa1 status. However, the subsequent responses diverged significantly between the two groups. In wild type mice, the astro-2 and micro-2 populations reached a stable state at 24 h, indicating a controlled and self-regulatory response in the presence of Anxa1. Contrastingly, Anxa1−/− mice exhibited a rapid decline in both astro-2 and micro-2 populations during this same time point. This decline underscores the potential critical role of Anxa1 in sustaining these cell populations beyond the initial response phase (Fig. 1F-G). We conducted a thorough analysis of differentially expressed genes (DEGs) between Astro-2 and Astro-1 subpopulations, which uncovered significant upregulation of genes such as Gbp2, Cxcl10, Serping1, and MHC-associated genes H2-D1 and H2-T23 in Astro-2 (Supplementary Table 3). In contrast, a comparison with Micro-1 revealed the downregulation of steady-state genes Tmem119 and P2ry12, alongside upregulation of inflammatory genes including TNF, Il1a, Il1, and chemokines Ccl3, Ccl4, and Ccl5 in Micro-2 (Supplementary Table 4). These alterations in the Micro-2 subpopulation align with the well-known characteristics of microglia activated by LPS, as previously reported [50]. Furthermore, pathway analysis highlighted that the upregulated genes in both Astro-2 and Micro-2 were primarily enriched in pathways associated with viral response, biotic stimulus regulation, and innate immune response (Supplementary Fig. 1E). Moreover, to investigate the transcriptional impact of LPS stimulation on oligodendrocytes as a whole, we also conducted a differential expression analysis on a cluster of oligodendrocytes previously identified. This analysis revealed dynamic changes in their transcriptional profiles. For instance, we found that oligodendrocytes exhibit distinct gene expression patterns at different time points, with genes showing high expression at 12 h and 24 h, while displaying deceased expression or no expression at 0 h and 72 h following the challenge (Fig. 1I). Within this set of genes, we identified Serpina3n and Cdkn1a, which have established associations with specific diseases and differentiation processes. In addition, genes such as Sox4, Apod, and Klf9 were also observed. These findings suggest that oligodendrocytes undergo distinct differentiation states at 12 h and 24 h compared to other time points. These particular oligodendrocytes, referred to as Cdkn1a+Serpina3n+ OligoD, exhibited enriched pathways related to cytoplasmic translation, macromolecule biosynthetic process regulation, and viral response (Supplementary Figure S1F). Our snRNA-seq results indicate a significant increase in the proportions of Astro-2, Micro-2, and Cdkn1a+Serpina3n+ OligoD within 24 h of LPS challenge, emphasizing their pivotal roles during the initial phases of simulated endotoxemia.

Fig. 1

Single-nucleus transcriptomic profiling of the mouse brain following peripheral LPS stimulation. (A) Schematic diagram illustrating the acquisition and analysis workflow of the spatial transcriptomics (ST) maps. Brain tissue samples from 8 wide type and 8 Anxa1 knockout C57BL6 mice were subjected to Stereo-seq, resulting in a total of 1,943,116 spots. Paired brain tissue samples of the wild type and Anxa1−/− mice were also subjected to snRNA-seq, and 376,909 qualified nuclei were identified. The processed data was subjected to dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP) and spatial cell types mapping for visualization, respectively. Finally, a combined analysis of the single-nucleus sequencing results and the spatial transcriptome sequencing results was performed. (B) UMAP plot of the snRNA-seq data. Each cluster in the plot corresponds to a specific cell type and is color-coded accordingly. The cell types included in the plot are Oligodendrocytes (OligoD), Vascular cells (Vas), Microglial cells (Microglia), Astrocytes (Astro), and Oligodendrocyte precursor cells (OPC). (C) Expression of canonical markers in each cell type cluster. (D) The UMAP plot displays cells from both the wild type and Anxa1−/− mice, with each cell color-coded based on the genotype of the respective mouse. (E) The UMAP plots demonstrate the changes in cell dynamics after administering LPS to both wild type and Anxa1−/− mice. (F-G) The relative proportions of Astro-1 and Astro-2 cells (F), as well as Micro-1 and Micro-2 cells (G), were compared at different time intervals following LPS treatment in both wild type and Anxa1−/− mice. (H) Volcano plots were generated to visualize the differentially expressed genes (DEGs) between Astro-2 and Astro-1 cells, as well as between Micro-2 and Micro-1 cells. The X-axis and Y-axis represent the log2 fold-change differences between the compared cell types and the statistical significance as the negative log10 of DEG P-values, respectively. Significantly up-regulated and down-regulated genes are indicated by orange and green dots, respectively, while non-significant genes are represented as black dots. (I) Heatmap displaying the DEGs between two groups of Oligodendrocytes. The first group consists of Oligodendrocytes from wild type and Anxa1−/− mice at 12- or 24-hour, while the second group comprises Oligodendrocytes from wild type and Anxa1−/− mice at 0- or 72-hour. Only the up-regulated genes with a log2 fold-change value greater than 1 and an adjusted p-value smaller than 0.05 are shown. Oligodendrocytes from wild type mice are colored in yellow, while those from Anxa1−/− mice are colored in red

Spatially-resolved transcriptomics of the mouse brain over the course of lipopolysaccharide challenge

Spatial information is critical for understanding cell-cell interactions that occur within tissues. Regrettably, this valuable information is absent from the snRNA-seq data. To gain deeper insights into the specific cell type that exerts the most significant impact on the inflammatory response at 12 h and 24 h, as well as to investigate the spatial relationships and potential mechanisms of interaction among these cell types, we employed Stereo-seq technology. This advanced method allowed us to capture in situ gene expression profiles. The Stereo-seq chips used were designed with capture probes incorporating a 25 bp coordinate identity barcode, a 10 bp molecular identity barcode, and a 22 bp poly T tail to facilitate mRNA hybridization. Brain tissues were collected from both wild-type and Anxa1−/− mice at 12 h, 24 h, and 72 h following LPS challenge. These tissues were subsequently embedded in optimal cutting temperature (OCT) compound. Utilizing this approach, we prepared serial cryosections with a thickness of 10 μm for Stereo-seq, hematoxylin and eosin (H&E) staining, and immunohistochemical (IHC) staining. Following the initial snRNA-seq analysis performed on 16 samples (including wild type and Anxa1−/− at 0 h, 12 h, 24 h, and 72 h, with two samples at each time point), we proceeded to conduct spatial transcriptome sequencing. The Stereo-seq chips utilized in this study were designed with capture spots measuring 220 nm in diameter and spaced at a distance of 500 nm from each other. To generate a comprehensive expression matrix with single-nucleus cell resolution, we employed a registration process that aligned the ssDNA image with the DNB image. We then utilized the StereoCell package for precise cell segmentation, thereby enabling the creation of a spatial gene expression matrix. To ensure the accuracy and reliability of gene annotation, we established stringent criteria of a minimum of 660 gene types and 1,300 counts per cell bin across all utilized chips. Notably, the gene capture and UMI numbers demonstrated a consistent pattern across samples obtained from both wild type and Anxa1−/− mice. This observation serves as a strong indicator of the acquisition of high-quality spatial transcriptomic data (Supplementary Figures S2A and S2B). To conduct cell annotation, we generated distinct marker gene sets for each cell cluster by utilizing dimensionality-reduced snRNA-seq data. These gene sets were further visualized, demonstrating their remarkable specificity (Supplementary Figure S2C). To ensure comprehensive analysis, we examined 20,000 spots per sample for UMAP graph visualization, successfully mitigating any potential batch effects (Supplementary Figures S2D and S2E). Additionally, we devised a marker gene signature score for each cell cluster using spatial transcriptomics data. The accuracy of this annotation was confirmed through robust cell type decomposition (RCTD) [51] (Supplementary Figures S2F and S2G). Furthermore, by comparing the RCTD scores with the marker gene signature scores, we established the consistency and validity of our annotation outcomes.

H&E staining unveiled a considerable population of ependymal cells within the third ventricle, encompassed by astrocytes, aligning with spatial transcriptomic annotations. Spatially resolved transcriptomic analysis revealed the distribution of neurons, astrocytes, and microglia throughout the entire brain slice, with neurons exhibiting the highest prevalence, in accordance with previous research (Fig. 2A) [52]. We conducted a thorough analysis of the cellular clusters within the spatial transcriptome data to ascertain their respective proportions, and observed no notable disparities in terms of cell type composition or quantities between wild type and Anxa1−/− mice. Neurons constituted the largest proportion, followed by glial cells, with vascular cells being the least abundant (Fig. 2B and Supplementary Figure S2H). We examined the ratios of Astro-2 and Micro-2 subtypes at different time points, indicating a noteworthy upregulation of both subtypes in both mouse brains at 12 h following LPS challenge. Nevertheless, while Micro-2 levels continued to be significantly elevated at 24 h in both groups, Astro-2 levels remained sustained in the wild type mice but reverted to baseline in the Anxa1−/− mice. This observed pattern was consistently supported by the results obtained from the snRNA-seq analysis. At the 72-hour mark following the challenge, both Astro-2 and Micro-2 levels reverted to their baseline levels (Fig. 2C and D). To validate and ascertain the observed patterns during simulated endotoxemia, we employed spatial visualization to analyze the temporal trends of the inflammation-related cell clusters (Fig. 2E and F). Furthermore, for the analysis of oligodendrocytes, we used a specific gene set, Cdkn1a+Serpina3n+ OligoD, derived from the snRNA-seq data, which demonstrated a high degree of specificity (Supplementary Figure S2J). Gene module scores were assessed in the spatial transcriptome data, utilizing a predefined threshold of 0.4 for the specific gene set. Clusters surpassing this threshold were identified as Cdkn1a+Serpina3n+ OligoD (Supplementary Figure S2K). Within a 24-hour timeframe, this particular population exhibited a significantly increase, amounting to approximately 50% in both mouse brains, thereby corroborating the findings obtained through snRNA-seq analysis. Moreover, the spatial visualization of these oligodendrocytes consistently demonstrated similar trends (Fig. 2G and H). The analysis of the spatial transcriptome data further confirmed that alterations in cell proportions of Astro-2, Micro-2, and Cdkn1a+Serpina3n+ OligoD were in concordance with those observed in snRNA-seq analysis. The Gene Ontology (GO) enrichment analysis revealed that these clusters were primarily associated with pathways such as “response to virus”, “regulation of neuron death” and “defense response to virus”, which are all pertinent to neuron death and defense responses, particularly evident 24 h following LPS stimulation. In contrast, Astro-2 and Micro-2 subtypes were implicated in additional pathways related to the regulation of neuron death and inflammatory responses, including “neutrophil chemotaxis” and “myeloid leukocyte migration” (Fig. 2I). Furthermore, the spatial transcriptome gene expression mapping revealed that while a significant proportion of spots in both mouse strains comprised a single nucleus cell type, there were also a notable number of spots containing multiple cell types (Supplementary Figure S2I).

Fig. 2

Spatially-resolved transcriptomic analysis following peripheral LPS stimulation. (A) Hematoxylin and eosin (H&E) staining, single-stranded DNA (ssDNA) picture, and cell type mapping of the ST slide from a wild type mouse at 72-hour. The spatial expression signal in the slide was detected using Stereo-seq technology, and the raw spatial expression matrix was transformed into spots based on the ssDNA picture. The cell types within the spots were then identified using the robust cell type decomposition (RCTD) method. In the right panel, the H&E staining and cell type labels around the third ventricle are emphasized. (B) The number of spots in the spatial transcriptome (ST) data with the eight assigned cell type labels. Spots that are not categorized into any of the eight cell type labels, namely Astro-1, Astro-2, Micro-1, Micro-2, Neuron, OligoD, OPC, and Vas-1, are labeled as “other”. (C-D) The relative proportions of Astro-1 and Astro-2 spots (C), as well as Micro-1 and Micro-2 spots (D), were compared at different time intervals following LPS treatment in both wild type and Anxa1−/− mice. (E-F) Spatial map of Astro-1 or Astro-2 spots (E), as well as Micro-1 or Micro-2 spots (F) at different time intervals following LPS treatment in both wild type and Anxa1−/− mice. (G) Proportions of Cdkn1a+ Serpina3n+ OligoD spots relative to all Oligodendrocyte spots across the ST data. (H) Spatial map of Cdkn1a+ Serpina3n+ OligoD spots across the ST data. (I) The enriched GO BP-terms for the DEGs between Astro-2 and Astro-1, Micro-2 and Micro-1, as well as Cdkn1a+ Serpina3n+ OligoD and OligoD in the ST data. The top six GO BP-terms for the up-regulated genes in each category-Astro-2, Micro-2, and Cdkn1a+ Serpina3n+ OligoD are presented

The integrated findings from our snRNA-seq and spatial transcriptomics analyses demonstrate a significant upregulation of Astro-2, Micro-2, and Cdkn1a+Serpina3n+ OligoD in both wild-type and Anxa1−/− mice at 12 h following LPS challenge. By the 24-hour timepoint, Micro-2 and Cdkn1a+Serpina3n+ OligoD remain elevated, while Astro-2 exhibits upregulation in wild-type mice but returns to baseline levels in Anxa1−/− mice. The spatial visualization of the transcriptome data further confirms that within 24 h post-challenge, the proportions of these three cell types, namely Astro-2, Micro-2, and Cdkn1a+Serpina3n+ OligoD, experience a significant increase, thereby playing key roles in neuron death and defense response pathways in both experimental groups.

snRNA-seq combined spatial transcriptome analysis reveals special pathological niche in mouse brain during sepsis

To delve deeper into the spatial distribution of inflammatory responses, we conducted a meticulous spatial co-localization analysis of 682 up-regulated genes at 12 h and 24 h after LPS challenge, using the spatial transcriptomic data. This comprehensive analysis revealed a specialized gene module consisting of 130 genes, which demonstrated consistent expression patterns in both wild-type and Anxa1−/− mice (Fig. 3A and Supplementary Figure S3A). Analysis of gene expression profiles in brain slices collected at 0 h, 12 h, 24 h, and 72 h after the challenge substantiated the augmented expression of this particular gene module at the 12-hour and 24-hour time points (Supplementary Figure S3B). Furthermore, the snRNA-seq data revealed that these genes exhibited a high level of expression in the Astro-2 and Micro-2 clusters, further reinforcing their involvement in mediating inflammatory responses at the spatial transcriptomic level. Importantly, these findings indicate that there were no significant transcriptional changes observed in the other cell clusters following LPS stimulation (Fig. 3B). GO analysis provided valuable insights into the functional properties of this gene module, indicating its involvement in crucial inflammatory pathways including “response to interferon-beta”, “response to type II interferon”, “response to virus”, and “cytokine-mediated signaling pathway”. These findings suggest the regulatory role of this module in inflammatory responses within the mouse brain, particularly 24 h after the stimulation (Fig. 3C). Subsequent analysis aimed to examine the distribution patterns of this module across different cell types within brain tissue slices. Notably, Astro-2, Micro-2, and Cdkn1a+Serpina3n+ OligoD demonstrated elevated module scores compared to non-inflammatory subpopulations at both the 12-hour and 24-hour time points following the stimulation. Significantly, the Vas-1 cluster showed pronounced upregulation and displayed the highest scores within the identified gene module (Fig. 3D). Spatial visualization analysis conducted at 12 h and 24 h post stimulation substantiated these findings, illustrating that the regions with the highest scores were primarily localized around blood vessels. This observation implies a vascular-centric distribution pattern of this module within the brain tissue (Fig. 3E). A parallel trend was observed in Anxa1−/− mice, providing evidence that Anxa1 plays a negligible role in the regulation of inflammatory responses within these specific cell types during endotoxemia simulation (Supplementary Figure S3C and Figure S3D). In summary, our integrated analysis of snRNA-seq and Stereo-seq revealed that identified gene module showcased heightened expression not only in the Astro-2 and Micro-2 clusters but also manifested augmented scores within the Vas-1 cell cluster. These findings collectively indicate a notable inflammatory activity within these regions in both wild type and Anxa1−/− mice at the 12-hour and 24-hour time points following LPS stimulation.

Fig. 3

Diversity and spatial distribution of microglia, astrocytes, and Oligodendrocytes following peripheral LPS stimulation. (A) Heatmap illustrating the gene co-localization module (Co-locM) in the ST data of wild type mice from different time periods after LPS treatment. (B) Visualization of Co-locM module score in the UMAP plot of the snRNA-seq data. (C) The enriched GO BP-terms in the genes of the Co-locM module. The top ten GO BP-terms are shown. (D) Distribution of the Co-locM module score in spots assigned with different cell type labels in the ST data of wild type mice from different time periods. The boxplot for each spot type is color-coded based on its label type. T tests (unpaired samples, two-tailed) were conducted between different spot types, and the P-value or the maximum P-value of a test set was reported. Significance levels are indicated as follows: *** P < 0.001. (E) Spatial distribution of the Co-locM module score and the spatial map of Vas-1 in the ST data obtained from wild type mice at 12- and 24-hour time points. (F) Distribution of the Co-locM module score in all spots within the ST data. The ST data is color-coded based on the mice genotype. Spots with a score above 0.5 were assigned the Co-locM label. (G) Distribution of Co-locM spot proportion across 46 spot groups. The spots were selected based on the presence of any of the nine labels (Astro-1, Astro-2, Micro-1, Micro-2, Neuron, OligoD, OPC, Vas-1, and Cdkn1a+ Serpina3n+ OligoD) and grouped according to the combination of labels present in each ST dataset. The Co-locM spot proportion in each spot group was calculated and plotted across eight ST datasets from mice at 12- and 24-hour time points. The top five spot types with the highest Co-locM spot proportion are highlighted. (H) Distribution of Co-locM spot proportion in the top five spot groups. Spot Group 1 (S1) consists of spots labeled with Micro-2, Astro-2, and Vas-1; Spot Group 2 (S2) comprises spots labeled with OligoD, Micro-2, Astro-2, and Cdkn1a+ Serpina3n+ OligoD; Spot Group 3 (S3) includes spots labeled with Astro-2 and Vas-1; Spot Group 4 (S4) consists of spots labeled with OligoD, Astro-2, and Cdkn1a+ Serpina3n+ OligoD; Spot Group 5 (S5) encompasses spots labeled with Micro-2 and Vas-1. The Co-locM proportion between S1 and S2, S3, S4, and S5 is compared using a T-test (paired samples, one-tailed). Significance levels are indicated as follows: *P < 0.05, ***P < 0.001. (I-J) Spatial maps of Astro.2, Micro-2, and Vas-1 spots in twFo different regions from wild type mice. The first region, as a Co-locM hot spot (I), is from a 12-hour mouse and represents a region where all three cell types are co-localized. The second region is a paired region from a 0-hour mouse (J). The maps illustrate the distribution and co-localization of these three cell types in each region. (K) DEG analysis showing up- and down-regulated genes across Astro-2, Micro-2 and Vas-1 clusters in snRNA-seq data of wild type mice. An adjusted P value < 0.01 is indicated in red, while an adjusted P value ≥ 0.01 is indicated in black. DEG, differentially expressed genes. (L) The GO terms obtained by performing GO enrichment analysis on the top 30 genes of Astro-2, Micro-2 and Vas-1 clusters based on the log2 FC value in the snRNA-seq data of wild type mice

Given the existence of diverse cell types within certain spatial transcriptome spots (Supplementary Figure S2I), it became imperative to explore the precise distribution pattern of a 130-gene module within Vas-1 cell spots. To this end, we implemented a threshold value of 0.5, wherein any spot score surpassing this threshold denoted the presence of expression for the forementioned 130 genes (Fig. 3F). To substantiate the distribution pattern within spatial transcriptome spots comprising multiple cell types, a comprehensive analysis of the data was undertaken. Notably, it was observed that the gene module under investigation exhibited pronounced enrichment in a specific set of 46 mixed cell type spots (Fig. 3G). Furthermore, the application of paired t-tests provided compelling evidence of the module’s substantial enrichment (P < 0.05) in spots containing mixed Astro-2/Micro-2/Vas-1 cells, particularly in brain slices derived from both wild type and Anxa1−/− mice, at 12 h and 24 h subsequent to LPS stimulation (Fig. 3H). To facilitate spatial visualization, our attention was directed towards the hippocampus, lateral ventricle, and third ventricle regions. Intriguingly, a closer examination revealed a higher concentration of Astro-2 and Micro-2 cells in the vicinity of blood vessels at the 12-hour mark following stimulation in wild type mice, which aligns consistently with previous findings (Fig. 3I and J). Furthermore, an in-depth analysis of the snRNA-seq data pertaining to the genes upregulated in Astro-2, Micro-2, and Vas-1 cell types (Fig. 3K) was followed by a subsequent GO enrichment analysis. Remarkably, the results of this analysis indicated notable associations with “viral response” and “response to biotic stimulus” (Fig. 3L). The observed enrichments serve to validate the findings of the Co-locM module obtained from spatial transcriptomic analysis, thereby highlighting the coherence and agreement between different data modalities. Notably, the examination of brain slices derived from Anxa1−/− mice exhibited analogous patterns of cell distribution surrounding Vas-1 cells at both 12-hour and 24-hour intervals post LPS stimulation, with no significant differences compared to wild type mice group. This suggests that the knockout of Anxa1 did not induce any alterations in the distribution of inflammatory cells around Vas-1 cells during the simulated sepsis condition (Supplementary Figures S3E, S3F). Moreover, a subsequent GO enrichment analysis conducted on the snRNA-seq data further reinforces these findings by reaffirming the involvement of these genes in the “viral response” process, in accordance with the spatial transcriptomic analysis (Supplementary Figures S3G, S3H).

By employing co-localization analysis, differential expression analysis and GO enrichment analysis, we were able to discern that a module comprising 130 genes primarily associated with inflammation response exhibited heightened expression in Astro-2, Micro-2 and Vas-1 subtypes (specifically, the V1A2M2 colocalization structure) at both 12-hour and 24-hour time points subsequent to LPS stimulation, in both the wild type and Anxa1−/− mouse brain. Intriguingly, an exploration of the 130-gene module across various published databases revealed its presence in mouse models infected with parasites or viruses, while no occurrence was observed in models stimulated with LPS. This indicates a potential protective role of the V1A2M2 complex in response to infection scenarios (Supplementary Figure S6).

Cell-cell interaction analysis revealed potential ligand-receptor interactions within this niche involving Timp1 or Ackr1

To gain insights into cellular interactions, we conducted an analysis of co-localization scores for over 2,000 ligand-receptor gene pairs and identified 19 pairs activated pairs in wild-type brains and 15 pairs in Anxa1−/− mice at both the 12-hour and 24-hour time points following LPS challenge, with 11 pairs exhibiting activation in both groups. Notably, among the identified pairs, we observed the presence of three pairs involving Timp1 and Cd14, five pairs with Ackr1, and two pairs with Csf1r and Sdc4 (Fig. 4A). Intriguingly, an in-depth analysis of the snRNA-seq data uncovered high expression levels of Timp1 in the Astro-2 and Vas-1 clusters, while its receptor exhibited a more widespread expression pattern across multiple cell types. This intriguing expression pattern implies a potential regulatory interaction between Astro-2 and Vas-1 cells that may exert an influence on other cellular entities. Likewise, we observed a predominant expression of Ackr1 in neurons and Vas-1 cells, accompanied by the noteworthy presence of its ligands in Astro-2 and Micro-2 subtypes, suggesting the existence of bidirectional interactions. Furthermore, our analysis revealed the expression of Cd14 and its receptors within the Micro-2 subtype, indicating the potential involvement of self-regulatory mechanisms within this specific cluster. In addition, the expression of Csf1 and Cxcl10, along with their respective receptors, was detected in both Astro-2 and Micro-2 subtypes, implying the possibility of facilitating mutual regulation or autoregulation processes. Moreover, we observed a predominant expression of Csf1 in the Astro-2 subtype and Cdkn1a+Serpina3n+ OligoD cell population, Intriguingly, the receptor for Csf1, namely Csf1r, exhibited specific localization in Micro-2 subtype. This finding suggests a potential regulatory relationship, wherein Astro-2 cells and Cdkn1a+Serpina3n+ OligoD population may govern the activity of Micro-2 cells through the Csf1-Csf1r ligand-receptor pathway. Additionally, our analysis revealed the specific presence of Cxcl10 and Sdc4 in both Astro-2 and Micro-2 subtypes, further implicating their potential involvement in either mutual regulation or autoregulation mechanisms. (Supplementary Figure S4A). Following the initial analysis, spatial co-expression analysis was performed to validate the co-expression patterns of these ligand-receptor genes in both wild type and Anxa1−/− mice at 12-hour and 24-hour time points post-challenge. Notably, certain gene pairs, including Col4a1-Cd93 and Pdgfa-Pdgfra, maintained their co-expression status until 72-hour time point in wild type mice. Conversely, in Anxa1−/− mice, the co-expression of some gene pairs diminished before the 24-hour mark. Collectively, no significant disparity in gene co-expression was observed at the 12-hour post-challenge time point between the two mouse groups (Fig. 4B). Subsequently, we assessed the differential gene expression levels of these ligand-receptor pairs and observed that majority of genes in the brains of both wild-type and Anxa1−/− mice exhibited upregulation at both the 12-hour and 24-hour time points following LPS challenge. To confirm the robustness of these findings, we cross-validated our results by analyzing the GSE153369 dataset, which yielded consistent outcomes (Fig. 4C). Furthermore, a GO enrichment analysis on the upregulated genes shed light on their involvement in immune cell migration pathways, including leukocyte, myeloid leukocyte, granulocyte migration, and chemotaxis (Fig. 4D). Consequently, we further investigated the spatial distribution of these genes at the 12-hour and 24-hour time points, specifically focusing on the expression patterns within clusters such as Astro-2, Micro-2, and Vas-1 (collectively referred to as V1A2M2). Notably, hub genes such as Timp1 and Ackr1 exhibited highly expression levels within these clusters. Similarly, the Tgm2-Sdc4 pair demonstrated elevated expression within the same clusters. These findings imply that the intricate interactions among these cell types within the V1A2M2 structure are potentially facilitated through the involvement of Timp1 and Ackr1 genes (Fig. 4E).

Fig. 4

Crosstalk among microglia, astrocytes, and vascular components within the pathological microstructure. (A) Network diagram of ligand-receptor pairs that are specifically activated in wild type or Anxa1−/− mice at 12- and 24-hour time points (referred to as responsive-LRs). Each node in the diagram represents either a ligand or receptor, with the node from which the arrow originates representing the ligand and the node to which the arrow points representing the receptor. (B) Heatmap displaying the -log2 (p-value) of the activity of responsive ligand-receptor (LR) pairs in wild type and knockout (Anxa1−/−) mice at different time points. LR pairs were identified as responsive if they showed significant changes in activity compared to baseline levels. The heatmap uses a color-coding scheme to indicate whether a responsive LR is specific to wild type (blue) or Anxa1−/− (red) mice. The heatmap provides a visual representation of the changes in activity of specific LR pairs in wild type and Anxa1−/− mice over time. (C) Fold change values of genes composed of responsive ligand-receptor (LR) pairs. Fold change values were calculated in wild type and Anxa1−/− mice at 12- and 24-hour time points compared to 0- and 72-hour time points. Additionally, fold change values were calculated in a publicly available bulk RNA-seq dataset comparing LPS-treated wild type mice to PBS-treated wild type mice. The genes included in the analysis are those that are composed of the responsive LR pairs identified in the previous analysis. The figure provides a visual representation of the changes in gene expression in response to LPS treatment in wild type and Anxa1−/− mice, as well as in a separate dataset. (D) The enriched GO BP-terms in the genes composed of responsive-LRs in wild type mice. Top ten GO BP-terms are shown. (E) Distribution of ligand-receptor (LR) co-localization spot percent in different spot groups. The spots were selected based on the presence of any of the nine labels (Astro-1, Astro-2, Micro-1, Micro-2, Neuron, OligoD, OPC, Vas-1, and Cdkn1a+ Serpina3n+ OligoD) and grouped according to the combination of labels present in each ST dataset. LR co-localization spot percent in each spot group was calculated and plotted across eight ST datasets from mice at 12- and 24-hour time points. The top five spot types with the highest percentages are shown in the figure and labeled as S1, S2, S3, S4, and S5. The LR co-localization spot percent between S1 and S2, S3, S4, and S5 is compared by T-test (paired samples, one-tailed); statistical significance is indicated by asterisks (*P < 0.05; ***P < 0.001). In each subgraph, S1 represents the spot group composed of Micro-2, Astro-2, and Vas-1 labels. The figure provides insights into the co-localization patterns of LR pairs within specific cell types and across different ST datasets. (F-G) The GO BP-terms that are enriched in the up-regulated genes between Vas-1 cells (spots) with and without the ligand-receptor pair(LR) (F) and between Astro-2 cells (spots) with and without the LR (G). A cell or a spot with a LR pair means that the two genes composed the LR are co-expressed in the cell or spot. The analysis was performed to identify the biological processes that are associated with the up-regulated genes between cells with and without LR pairs. The top six enriched GO BP-terms are shown for both ST data and snRNA-seq data. The figure provides insights into the biological processes that are affected by the presence or absence of LR pairs in Vas-1 and Astro-2 cells

Our analysis of snRNA-seq data revealed that Ackr1 exhibited predominant expression (Supplementary Figure S4A). However, further investigation through spatial transcriptomic analysis unveiled its presence within the V1A2M2 region. To validate its expression in Vas-1, we performed an examination of Ackr1 across 46 mixed cell type groups. Our findings confirmed its expression in Vas-1, as illustrated in the fourth box plot of Supplementary Figure S4B. Additionally, to explore the potential role of the hub gene Timp1 and its associated receptors in the V1A2M2 region, we conducted a GO analysis focusing on the upregulated genes present in Vas-1 and Astro-2 cells expressing Timp1. Notably, we identified Cd63, Itgb1, and Lrp1 as receptor genes associated with Timp1 in this context. The analysis revealed a significant enrichment of pathways related to wound healing and amoeboid cell migration in both the snRNA-seq and Stereo-seq datasets, suggesting that Vas-1 and Astro-2 cell populations may play a crucial role in inflammatory repair processes and potentially facilitate the migration of Micro-2 cells towards Vas-1 through Timp1-mediated signaling (Fig. 4F and G). This study hypothesizes that Vas-1 and Astro-2 populations, characterized by their expression of the Timp1 receptor gene, may play a role in the regulation of inflammatory repair. Furthermore, considering the well-established evidence of amoeboid-type movement exhibited by activated microglia in previous investigations [53,54,55,56], it is plausible to suggest that the expression of Timp1 in Vas-1 and Astro-2 cells potentially facilitates the migration of Micro-2 cells. Spatial transcriptomic analysis revealed that the expression of ligand-receptor genes within the V1A2M2 region exhibited a peak at 12 h following LPS stimulation, followed by a decrease by the 24-hour time point, and eventual disappearance by 72 h. Notably, no significant disparities