Abstract Cross-talk between the brain and cervical lymph nodes (CLNs) is crucial in brain pathologies. However, the precise roles and the mechanisms of CLNs in brain damage during subarachnoid hemorrhage (SAH) remain unclear. In this study, mandibular lymph node (part of CLNs) removal attenuates brain damage in SAH mouse models. Notably, the extravasated erythrocytes following SAH are significantly engulfed by lymphatic endothelial cells (LECs) in CLNs. Single-cell RNA sequencing reveals that the differentially expressed genes in medullary LECs are enriched in lysosomes after SAH, with a notable upregulation of Ctss (which encodes cathepsin S). Importantly, the deficiency of cathepsin S specifically in LECs, achieved through transgenic mice, or the use of a cathepsin S inhibitor, significantly reduces neuroinflammation and neurological deficits induced by SAH. These findings elucidate mechanisms of how CLNs participate in brain injury following SAH in mice. Targeting this process may offer effective therapeutic strategies to alleviate SAH-related pathologies. Subject terms: Stroke, Stroke __________________________________________________________________ Crosstalk between the brain and CLNs is critical in brain pathologies. Here, the authors show in a mouse model that extravasated erythrocytes following SAH are degraded by cathepsin S of medullary LECs in CLNs, which plays an important role in SAH pathology. Introduction Subarachnoid hemorrhage (SAH) presents a significant clinical challenge due to its devastating impacts on individuals, patient quality of life, families, and society at large. The incidence of SAH varies according to country, but it typically attacks about 2–16/100,000 people^[52]1. It can occur at a relatively young age with a high fatality rate (about 50%), and approximately 95% of survivors of SAH undergo long-term disabilities^[53]1,[54]2. Following SAH, the presence of blood and its degradation products within the subarachnoid space (SAS) causes significant brain immune responses and systemic inflammatory responses, which lead to early brain injury (EBI) and delayed cerebral ischemia^[55]3,[56]4. In stroke, the related systemic inflammatory response and immune dysregulations participate in brain injury, recovery, and stroke outcome^[57]4–[58]6. However, how the injured brain cross-talks with the peripheral immune system after SAH and whether it is critical for brain pathogenesis and outcome remain unclear. Recently, meningeal lymphatic vessels have been characterized in the meninges surrounding the brain that drain into cervical lymph nodes (CLNs, including superficial and deep CLNs), providing an important route for brain-specific antigens to access the peripheral immune system^[59]7,[60]8. It has been proposed that the brain can directly activate the peripheral immune apparatus through exposure to brain-specific antigens, thereby initiating an immune response that may precipitate additional brain damage and secondary brain injury^[61]4. Planas et al. reported that myelin and neuronal antigens are observed in CLNs in patients with stroke, and that such patients have a higher immunoreactivity to brain antigens, and that the majority of brain immunoreactive cells are CD68(+) macrophages^[62]9. Esposito et al. elucidated a brain-to-CLN pathway that participates in brain injury-induced systemic inflammation in a rat model of ischemic stroke^[63]10. They hypothesized that elevated vascular endothelial growth factor C (VEGF-C) levels in the cerebrospinal fluid (CSF) drain into CLNs, where VEGF-C activates the lymphatic endothelium^[64]10. In glioblastoma, glioblastoma-associated antigens can be captured by antigen-presenting cells (APCs), including dendritic cells (DCs), and migrate to CLNs. T cells in the CLNs are activated, which subsequently enter the brain to mediate glioblastoma cytotoxicity and promote further release of glioblastoma antigen^[65]11. Collectively, the drainage of CNS antigens into the CLNs may contribute to the brain damage following CNS injury. However, the specific mechanisms through which the injured brain activates the peripheral immune system following SAH require further elucidation. We previously reported that erythrocytes in the CSF are transported into CLNs, including mandibular LNs and deep CLNs, accumulating in the LYVE-1^+ lymphatic sinus at 4 h post SAH^[66]12. However, whether these extravasated erythrocytes play a role in activating of systemic inflammation and exacerbating the brain injury remains unclear. In this study, we demonstrate that the surgical removal of mandibular LNs, a major component of the superficial CLNs, alleviated SAH-induced neuroinflammation and neurological deficits in two mouse models, thus indicating the critical involvement of mandibular LNs in brain injury following SAH. We observed that at 24 h of SAH, extravasated erythrocytes were significantly degraded within the mandibular LNs, primarily via engulfment by lymphatic endothelial cells (LECs). Furthermore, single-cell RNA sequencing (scRNA-seq) and immunofluorescence analyses of the mandibular LNs revealed that the extravasated erythrocytes were mainly degraded by lysosome protease cathepsin S (CTSS) of medullary LECs at 24 h post SAH. The depletion of CTSS specifically in LECs, implemented using Prox1-creER^T2;Ctss^flox/flox mice, or pharmacological blockade of CTSS with an inhibitor, resulted in reduced neuroinflammation and improved neurological outcomes post-SAH. These findings indicate that the erythrocytes drained into mandibular LNs are engulfed by LYVE-1^+ LECs and degraded by CTSS, which plays an important role in triggering a CLN-mediated inflammatory response and the aggravation of brain injury following SAH. Targeting this process may be a viable therapeutic strategy for treating SAH. Results CLNs are involved in SAH-induced brain injury In response to brain injury, peripheral immune cells become rapidly activated and infiltrate brain tissue, amplifying the extent of brain damage^[67]13,[68]14. Esposito et al. proposed that the brain-to-CLN crosstalk may contribute to the initiation of systemic inflammatory responses following acute stroke^[69]10. We hypothesized that CLNs, e.g., mandibular LNs in this study, play a pivotal role in SAH pathogenesis. To test this hypothesis, we surgically excised the mandibular LNs in two distinct mouse models of SAH (Fig. [70]1a): an endovascular perforation-induced SAH model and an autologous blood injection-induced SAH model. These surgical interventions (mandibular LN removal) were conducted 7 days prior to SAH induction in C57BL/6 male mice, allowing ample time for the mice to recover from the dissection. The endovascular perforation of the circle of Willis is a commonly used mouse model for studying SAH. Following SAH modeling, the mortality rate in the sham + SAH group is about 22%, while all mice in the control and mandibular LN removal + SAH groups were alive at the end of the experiment. We found that compared with the mice in the control group, mice in the SAH group displayed remarkable neurological deficit at 24 h and 48 h post-surgery, assessed utilizing the modified Garcia scores. Compared to the sham group, the LN removal + SAH group exhibited better neurological function, displaying significantly higher scores at 48 h post-surgery (Fig. [71]1b, c). The SAH surgery also induced prominent brain edema, affecting the bilateral cerebral hemispheres, cerebrum, and brainstem. The LN removal + SAH group exhibited decreased water content in both the right hemisphere and brain stem (Fig. [72]1d). A higher percentage of circulating neutrophils indicates poor outcomes of SAH^[73]12,[74]13. Our data indicated a significant increase in the percentages of peripheral blood neutrophil percentages following SAH, which was ameliorated by mandibular LN excision (Fig. [75]1e, f and Supplementary Fig. [76]1a). SAH surgery also increased the expression of microglial and astrocytic activation markers, as well as pro-inflammatory cytokines in the brains of mice, including ionized calcium-binding adapter molecule-1 (Iba-1), glial fibrillary acidic protein (GFAP), interleukin-1β (IL-1β) and tumor necrosis factor alpha (TNF-α), all of which were decreased in the mandibular LN excision group (Fig. [77]1g, h). Fig. 1. CLNs are involved in the brain injury after SAH. [78]Fig. 1 [79]Open in a new tab a Schematic diagram illustrating the experimental design involving the excision of the mandibular lymph node (mandibular LN), a one-week recovery, followed by SAH induction and analysis 48 h later. b–h SAH was performed by endovascular perforation. The neurological deficits of mice were evaluated by modified Garcia scores at 24 (b) and 48 h (c) post-SAH. Control and Mandibular LN removal + SAH; n = 9, Sham + SAH, n = 7 mice. d Quantification of the water content of each part of the brain, including the left hemisphere, the right hemisphere, the cerebrum, and the brain stem. N = 7 mice. Representative dot plots (e) and quantification of the percentage of CD11b^+Ly6G^+ neutrophils (f) in blood. N = 6 mice per group. g, h Western blot demonstrated the expression of GFAP, Iba-1, Cleaved-IL-1β, and TNF-ɑ in the brain. N = 3 mice per group. i–p SAH was induced by autologous blood injection into the cisterna magna. The neurological functions of mice were evaluated by modified Garcia scores after 24 (i) and j 48 h post-SAH. Sham + SAH; n = 9, Mandibular LN removal + SAH; n = 7 mice. k Representative images of co-staining of TUNEL with neurons in the brain. Scale bars, 50 μm. The schematic diagram indicates the region of the cerebral cortex where we performed the analysis. Created in BioRender. Chen (2025) [80]https://BioRender.com/4u8isww. l Quantification of the percentage of tunnel-positive neurons in the cerebral cortex. N = 5. m–o Representative dot plots (m) and quantification of the percentages of the CD16/32^+CD206^- subsets (n), CD16/32^+CD206^+ subsets (o), and CD206^+CD16/32^− subsets (p) in the brain. Sham + SAH; n = 9, Mandibular LN removal + SAH; n = 7 mice. All data are presented as mean values ± SD; one-way ANOVA with Turkey’s multiple-comparison test or two-tailed unpaired Student’s t-test. ns not significant. Source data are provided as a Source data file. In another SAH mouse model, we administered autologous blood into the cisterna magna to further substantiate our findings. After SAH modeling, the mortality rates for the sham + SAH and mandibular LN removal + SAH groups were approximately 10% and 30%, respectively. Compared to the mice with intact LNs, the mice accompanied by mandibular LN excision also had better neurological functional recovery at 24 h and 48 h post-SAH surgery (Fig. [81]1i, j). The amount of TUNEL-positive apoptotic neurons in the cerebral cortex area is commonly used to indicate the neuronal injury and cell death following SAH^[82]15–[83]18. As reported, compared to the hippocampus, more TUNEL-positive apoptotic neurons were detected in the cerebral cortex in SAH mouse models^[84]19. We assessed the percentage of the TUNEL-positive neurons in the cerebral cortex following SAH and found that neuronal apoptosis was significantly reduced in the mice with mandibular LNs removal (Fig. [85]1k, l). Additionally, the mice undergoing mandibular LN removal presented with significantly lower percentages of pro-inflammatory CD16/32^+CD206^− microglia (Fig. [86]1m, n and Supplementary Fig. [87]1b) and a diminished proportion of CD16/32^+CD206^+ intermediate microglia (Fig. [88]1o) compared to the mice with intact CLNs following SAH. However, no differences were observed in the proportion of CD16/32^−CD206^+ anti-inflammatory microglia between groups (Fig. [89]1p). These data collectively indicate that the removal of CLNs (mandibular LNs in this study) mitigates SAH-associated brain injury. The engulfment of extravasated erythrocytes in CLNs by LECs after SAH Previous reports established that erythrocytes drained into CLNs, encompassing mandibular and deep CLNs, via meningeal lymphatics during the very early phase following SAH^[90]12. Therefore, we speculated that the mechanisms by which CLNs participate in SAH-induced brain injury may involve these extravasated erythrocytes in CLNs. We compared the presence of extravasated erythrocytes in mandibular LNs at 4 h and 24 h post-SAH in C57BL/6 male mice induced by autologous blood injection. We found that both time points displayed marked extravasated erythrocyte accumulation in the lymphatic sinus, with some co-localized with LYVE-1^+ LECs (Fig. [91]2a). When compared to 4 h post-SAH, the number of intact erythrocytes within mandibular LNs significantly decreased, with a concomitant increase in degraded erythrocytes at 24 h post-SAH (Fig. [92]2b, c). We validated these results in the SAH mouse model induced by endovascular perforation in C57BL/6 male mice. At 4 h post-SAH, the mandibular LN displayed the most prominent accumulation of extravasated blood, while the deep CLN also presented some extravasated blood; the superficial parotid LN, axillary LN, and brachial LN did not display this phenomenon (Supplementary Fig. [93]2a), which is consistent with our previous findings^[94]12. Immunofluorescence staining of the mandibular LNs also showed that extravasated erythrocytes co-localized with LYVE-1+ LECs at both 4 h and 24 h post-SAH compared with the sham group, with a predominance of intact extravasated erythrocytes at 4 h post-SAH and predominantly degraded erythrocytes at 24 h post-SAH (Supplementary Fig. [95]2b). Deep CLNs manifested similar trends to mandibular LNs but accumulated fewer extravasated erythrocytes (Supplementary Fig. [96]2c). However, superficial parotid LNs exhibited negligible evidence of extravasated erythrocytes co-localizing within LYVE-1^+ LECs (Supplementary Fig. [97]2[98]d). We also co-stained the LYVE-1^+ LECs with anti-Prospero homeobox protein 1, an essential transcription factor and classic marker of LECs, to confirm the accumulation of extravasated erythrocytes within the LECs (Supplementary Fig. [99]2e). Fig. 2. Extravasated erythrocytes in mandibular LNs were engulfed and degraded by LECs. [100]Fig. 2 [101]Open in a new tab a Representative images of mandibular LNs isolated at 4 h or 24 h post-SAH were stained for LYVE-1 (LECs, red) or TER-119 (erythrocytes, green). Intact erythrocytes (white arrows) and degraded erythrocytes (blue arrows) were located in and around LYVE-1^+ LECs. Scale bars, 10 μm. b Quantification of the number of intact erythrocytes per field and c the percentage of degraded erythrocytes. Four hours post-SAH, Sham; n = 12, SAH; n = 11 mice. Twenty-four hours post-SAH, Sham; n = 7, SAH; n = 5 mice. The extravasated erythrocytes in mandibular LNs from 24 h post SAH showed a significant reduction in the number of intact cells and an increase in the degradation ratio. d Representative images of mandibular LNs labeled by LYVE-1 (LECs, blue), TER-119 (erythrocytes, red), and CD169 (macrophages, green). Regions of interest (white square) in the left images are shown right. Scale bars, left; 50 μm, right; 20 μm. The intact erythrocytes (white arrows) were located in and around LECs, and almost did not co-localize with macrophages. The degraded erythrocytes displayed co-localization with LECs (yellow arrows) and macrophages (purple arrows). Immunofluorescence staining was performed three times independently. e Representative images of mandibular LNs from the sham and SAH groups, which were labeled by LYVE-1 (LECs, blue), TER-119 (erythrocytes, red), and CD169 (macrophages, green). Regions of interest (white square) in the left images are shown right. Extravasated erythrocytes were shown to be located in LYVE-1^+ LECs (white arrows). Scale bars, left; 20 μm, right; 10 μm. f Quantification of the number of intact erythrocytes co-localized with LECs or with macrophages. The number of intact erythrocytes co-localized with LYVE-1^+ LECs in the SAH group was significantly higher than that co-localized with macrophages. N = 7 mice per group. g Representative images and orthogonal views of LECs engulfed CFSE-labeled erythrocytes (white arrows) after 24 h of incubation in vitro. Scale bars, 20 μm. Immunofluorescence staining was performed three times independently. All data are presented as mean values ± SD; two-tailed unpaired Student’s t-test. Source data are provided as a Source data file. To explore whether the extravasated erythrocytes were degraded by macrophages in the mandibular LNs, we labeled the macrophages using anti-CD169 antibodies. Most intact extravasated erythrocytes were located in and close to LYVE-1^+ LECs, whereas the degraded erythrocytes were co-localized with both LYVE-1^+ LECs and CD169^+ macrophages (Fig. [102]2d, f). Moreover, the number of intact erythrocytes overlaid with LYVE-1^+ LECs was significantly greater than that overlaid with CD169^+ macrophages (Fig. [103]2f), indicating that the extravasated erythrocytes in mandibular LNs may be engulfed by LECs. The results were also confirmed by in vitro experiments, in which erythrocytes were labeled with 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) and incubated with LECs for 24 h. And data showed that LECs internalized the CFSE-labeled erythrocytes (Fig. [104]2g). ScRNA-seq revealed lysosomes within medullary LECs involved in SAH To further explore the cellular and molecular mechanisms underlying erythrocyte degradation by LECs, we performed scRNA-seq on LECs (CD45^−CD31^+Podoplanin^+) isolated from mandibular LNs at 24 h after autologous blood infusion-induced SAH in C57BL/6 male mice. A total of 27,718 LN cells from both the sham and SAH groups were used. Unsupervised clustering revealed 11 distinct cell populations based on known cell-type-specific marker expression (Fig. [105]3a, b). LECs were identified by robust expression of the pan-endothelial marker Pecam1, as well as the LEC markers prospero homeobox protein 1 (Prox1), podoplanin (Pdpn) and fms-related tyrosine kinase 4 (Flt4) (Supplementary Fig. [106]3a). Three subsets of LN LECs were identified: floor LECs (characterized by expression of Ccl20 and Coch, Supplementary Fig. [107]3b), ceiling LECs (characterized by expression of Ackr4 and Ackr3, Supplementary Fig. [108]3c), and medullary LECs (characterized by expression of Marco, Reln, and MRC1, Supplementary Fig. [109]3d), based on recent publications^[110]20–[111]22. Immunostaining corroborated that extravasated erythrocytes predominantly accumulated within LYVE-1^+ LECs (Fig. [112]2a and Supplementary Fig. [113]2b). Notably, scRNA-seq analysis indicated that Lyve-1 expression was primarily confined to the floor and medullary LEC subsets (Supplementary Fig. [114]3e). Fig. 3. scRNA-seq revealed that the differential genes of medullary LECs in mandibular LN are enriched in lysosomes after SAH. [115]Fig. 3 [116]Open in a new tab a t-SNE visualization of scRNA-seq of cells isolated from the mandibular LNs of sham and SAH at 24 h post-surgery. The cell clusters include endothelial cell (EC), fibroblastic reticular cell, vein EC, LEC, arterial EC, floor LEC, medullary LEC, ceiling LEC, mural cell, B cell, and granulocyte. N = 20 mice per group. b A violin plot demonstrating cells expressing phenotyping markers for each population. c Top 20 KEGG enrichment of differentially expressed genes for medullary LECs of mandibular LNs. d Gene expression heatmap of the KEGG enrichment to the lysosome of medullary LECs. e A bar plot demonstrating Ctss gene expression in mandibular LN cells of the Sham and SAH groups. f Representative images of mandibular LN staining for LYVE-1 (blue) and MACRO (green) to identify medullary LEC and TER-119 (red) to determine erythrocytes, demonstrating that the erythrocytes were primarily accumulated in LYVE-1 and MACRO double-positive medullary LECs (white arrows). Scale bar, 20 μm. Immunofluorescence staining was performed three times independently. g Representative images of mandibular LN co-labeling for ceiling LECs (ANXA2, blue), LYVE-1 (green), and erythrocytes (TER-119, red), demonstrating that the erythrocytes were not observed co-localized with ceiling LECs. White arrows, ceiling LECs. Yellow arrows, erythrocytes co-localized with LYVE-1^+ LECs. Scale bar, 20 μm. Immunofluorescence staining was performed three times independently. To investigate the transcriptional responses of both the floor and medullary LEC subsets, we performed differential expression gene (DEG) analyses comparing the sham and SAH conditions. We identified DEGs (adjusted p < 0.05) in these two subsets of LECs separately. Interestingly, kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that DEGs in medullary LEC were highly enriched in the lysosome pathway (Fig. [117]3c), whereas lysosome was not among the top 20 KEGG terms of floor LECs (Supplementary Fig. [118]3f). Among the lysosome-enriched genes in medullary LECs, KEGG analysis revealed that Ctss (encoding CTSS) expression was significantly increased after SAH compared to the sham group (Fig. [119]3d). Although CTSS can be expressed by both floor and medullary LECs, Ctss expression was up-regulated in medullary LECs after SAH (Fig. [120]3e). Moreover, a t-distributed stochastic neighbor embedding (t-SNE) plot demonstrated a larger amount of medullary LECs expressing Ctss in the SAH group than that in the sham group (Supplementary Fig. [121]3g). Co-label of medullary LEC markers of MARCO and LYVE-1 with TER-119 (erythrocytes) also confirmed that extravasated erythrocytes in the mandibular LNs primarily accumulated within the LYVE-1 and MACRO double-positive medullary LECs (Fig. [122]3f). The ceiling LECs, stained by ANXA2, did not co-localize with erythrocytes (Fig. [123]3g). LECs elevated immune responses following erythrocyte engulfment Subsequently, we performed bulk RNA sequencing on LECs incubated with erythrocytes (RBCs) for 24 h and control LECs without RBC co-incubation in vitro. A total of 83 DEGs were detected in RBC-treated LECs compared to control LECs. Enrichment analyses indicated that the gene sets were involved in immune and inflammatory responses, chemotaxis, phagocytosis, and lysosomes (Supplementary Fig. [124]4a). The RBC-treated LECs had significantly higher expression of genes encoding pro-inflammatory factors, chemokines, and interferon-inducible genes relative to control LECs, including Tnf-α, Il-1β, Nos2, Ccl3, Ccl4, Cxcl2, Cxcl3, Ccl2, Mpeg1, and Lcn2. Phagocytosis and lysosomal-associated genes, including Syt7, Irf8, Itgb2, Unc93b1, and Ctss, were also dramatically up-regulated in RBC-treated LECs (Supplementary Fig. [125]4b, c). Quantitative PCR analysis validated the higher expression levels of Ctss, Tnf-α, Il-1β, Nos2, Cxcl2, Ccl2, Ccl7, and Cxcl1 in RBC-treated LECs compared to control LECs (Supplementary Fig. [126]4d, e). Engulfed erythrocytes were degraded by the CTSS of lysosomes To elucidate the subcellular localization of engulfed erythrocytes within LECs, erythrocytes were labeled with CFSE and incubated with LECs, while lysosomes were stained with LysoTracker Red. Co-localization analysis indicated that erythrocytic particles predominantly localized within lysosomal compartments, as evidenced by the co-localization of the erythrocytic particles with LysoTracker Red^+ puncta (Fig. [127]4a). LECs with erythrocyte particle uptake exhibited larger lysosomal diameters and areas than LECs without erythrocyte particle uptake (Fig. [128]4b, c). These observations were corroborated using primary LECs expressing the classical LEC markers, including Prox-1, LYVE-1, and VEGFR-3 (Supplementary Fig. [129]5a), which displayed significantly larger lysosomes when containing erythrocyte particles (Supplementary Fig. [130]5b–d). Treatment with bafilomycin A1 blocks autophagosome-lysosome fusion and inhibits acidification and protein degradation in lysosomes^[131]23. We found reduced erythrocyte particles in the Bafilomycin A1 group (Supplementary Fig. [132]5b). Fig. 4. The engulfed erythrocytes were delivered to lysosomes and degraded by CTSS. [133]Fig. 4 [134]Open in a new tab a Representative images of LECs incubated with RBC-CFSE for 24 h in vitro. Lysosomes were labeled by LysoTracker Red, and RBCs were labeled by 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) before incubation with LECs. Scale bars, 10 μm. b Quantification of the lysosomal diameter and c the lysosomal area. The lysosomes in the RBC-CFSE group demonstrated a significant increase in the diameter and area compared to those in the control group. N = 3 biological replicates. d Representative images of mandibular LN at 24 h post-SAH staining for LYVE-1^+ LEC (green), LAMP-1^+ lysosome (white), and TER-119^+ erythrocytes (red). Regions of interest in the left images are shown right. Scale bars, left; 50 μm, right; 10 μm. e Quantification of the relative fluorescence intensity of LAMP-1^+ lysosome in LYVE-1^+ area of mandibular LNs at 24 h post SAH. Sham; n = 7, SAH; n = 5 mice. f The orthogonal view shows that the erythrocytes (TER-119, red) were located in the lysosome (LAMP-1, white) of LEC (LYVE-1, green). Scale bar, 50 μm. Immunofluorescence staining was performed four times independently. g The orthogonal view shows that CTSS (white) was mainly co-localized with LECs (LYVE-1, green), and erythrocytes (TER-119, red) were internalized by LECs with CTSS. Scale bar, 20 μm. Immunofluorescence staining was performed four times independently. All data are presented as mean values ± SD; two-tailed unpaired Student’s t-test. Source data are provided as a Source data file. We next verify whether extravasated erythrocytes underwent degradation by mandibular LN LECs via lysosome in the endovascular perforation SAH model using C57BL/6 male mice. We co-stained mandibular LNs with antibodies targeting LAMP-1, LYVE-1, and TER-119, marking lysosomes, LECs, and erythrocytes, respectively. Results demonstrated that LAMP-1^+ lysosomes were predominantly co-localized with LYVE-1^+ LECs, and TER-119^+ erythrocytes were accumulated in the LYVE-1^+ LECs (Fig. [135]4d). Moreover, the expression of LAMP-1^+ lysosomes within the LYVE-1^+ regions of mandibular LNs in the SAH group was significantly elevated compared to the sham group (Fig. [136]4e). Orthogonal view also confirmed the co-localization of erythrocytes with the lysosomal compartments in LYVE-1^+ LECs (Fig. [137]4f). CFSE-labeled erythrocytes in vitro injected into the cisterna magna showed co-localization with LYVE-1^+ LEC lysosomes at both 4 h and 24 h post-injection (Supplementary Fig. [138]5e). To investigate whether CTSS, the lysosomal protease identified through sc-RNAseq, played a role in erythrocyte degradation, we co-stained CTSS with LYVE-1^+ LECs and erythrocytes. The orthogonal view demonstrated that CTSS co-localized predominantly with LYVE-1^+ LECs, and erythrocytes were co-localized with cathepsin S^+ LECs (Fig. [139]4g and Supplementary Fig. [140]5f). Ctss knockout reduces neurological deficits and neuroinflammation To verify the role of CTSS expression in LECs mediating SAH-induced brain injury, we established a Prox1-creER^T2;Ctss^flox/flox (Prox1-creER^T2/Ctss^f/f) mouse model to specifically knock out the gene Ctss in LECs. We used both male and female mice in this set of experiments. Before the Prox1-creER^T2/Ctss^f/f mice and their Ctss^f/f littermates underwent endovascular perforation-induced SAH surgery, male and female mice were given tamoxifen (100 μg/g) intraperitoneally for seven consecutive days (Fig. [141]5a). No mice died after the administration of tamoxifen. The mortality rates of the Ctss^f/f and Prox1-creER^T2/Ctss^f/f mice post-SAH modeling were approximately 25% and 14%, respectively. And no significant difference in mortality rates was observed between the two groups. The Ctss knockdown efficiency was confirmed by immunofluorescence staining of mandibular LNs, and the expression of CTSS was significantly decreased in Prox1-creER^T2/Ctss^f/f mice (Fig. [142]5b, c). At 4 h post-SAH, no significant difference was observed in the amount of intact extravasated erythrocytes in the LYVE-1^+ area of mandibular LNs between Prox1-creER^T2/Ctss^f/f and Ctss^f/f mice (Fig. [143]5d, e). However, the number of extravasated erythrocytes was higher in Prox1-creER^T2/Ctss^f/f mice at 24 h post-SAH (Fig. [144]5d, f), indicating delayed degradation of the extravasated erythrocytes with CTSS deficiency in LEC. Ctss knockdown in LECs mitigated the neurological deficits in SAH, demonstrated by higher neurological scores in Prox1-creER^T2/Ctss^f/f mice than in their Ctss^f/f littermates at both 24 h and 48 h post SAH (Fig. [145]5g, h). A significantly lower percentage of CD169^+TNF-ɑ^+ proinflammatory macrophages was also observed in the mandibular LNs of Prox1-creER^T2/Ctss^f/f mice, although the total number of CD169^+ macrophages did not differ significantly between Prox1-creER^T2/Ctss^f/f and Ctss^f/f mice (Fig. [146]5i–k). Flow cytometry (gating strategy shown in Supplementary Fig. [147]6a) was used to assess the leukocyte infiltration and microglial activation in the brain. The results revealed significantly lower proportions of CD16/32^+CD206^− pro-inflammatory microglia (Fig. [148]5l,m) and CD16/32^+CD206^+ intermediate microglia (Fig. [149]5n) in Prox1-creER^T2/Ctss^f/f mice than in their Ctss^f/f littermates following SAH. However, the percentage of CD16/326^−CD206^+ anti-inflammatory microglia (Fig. [150]5o), the number of CD45^high leukocytes (Fig. [151]5p–r), and CD11b^+F4/80^+ macrophages (Fig. [152]5l) showed trends of change, but there was without statistical significance. Fig. 5. Ctss knockout alleviates the neurological deficits and neuroinflammation after SAH. [153]Fig. 5 [154]Open in a new tab a The schematic diagram demonstrates the experimental design. Prox1-CreER^T2/Ctss^f/f mice and their Ctss^f/f littermates were intraperitoneally administered tamoxifen (100 μg/g) for seven consecutive days, followed by the endovascular perforation-induced SAH surgery, and sacrificed for analysis at 48 h. Created in BioRender. Chen, J. (2025) [155]https://BioRender.com/bu6nimn. b Representative images of mandibular LNs stained with LYVE-1 and CTSS antibodies. Scale bar, 20 μm. c Quantification of the relative fluorescence intensity of cathepsin S^+ in the LYVE-1^+ area of mandibular LNs. Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 5. d Representative images of mandibular LNs staining for LYVE-1^+ LEC (blue) and TER-119^+ erythrocytes (red). Scale bar, 20 μm. e, f Quantification of the number of intact erythrocytes co-localized with LECs at 4 h (e) and 24 h (f) post-SAH. Four hours post-SAH, Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 5. Twenty-four hours post-SAH, Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 8. g, h Quantification of the modified Garcia scores at 24 h and 48 h post-SAH. Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 5. i Representative images of mandibular LNs stained with LYVE-1 (blue), CD169 (green), and TNF-ɑ (red) antibodies to display LECs, macrophages, and proinflammatory macrophages. Scale bar, 20 μm. j, k Quantification of the numbers of CD169^+ macrophages (j) and the percentages of CD169^+ TNF-ɑ^+ proinflammatory macrophages (k). Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 5. l Representative dot plots demonstrate the percentages of microglia subsets in the brain at 48 h post-SAH. m–o Quantification of the percentages of the CD16/32^+CD206^−, CD16/32^+CD206^+, and CD16/32^−CD206^+ subsets of the CD11b^+CD45^low populations in the brain at 48 h post-SAH. Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 5. p Representative dot plots present the proportions of CD11b^+F4/80^+ macrophage in CD45^high populations in the brain at 48 h post-SAH. q, r Quantification of the number of CD45^high leukocytes (q) and CD11b^+F4/80^+ macrophages (r) in the brain. Ctss^f/f; n = 6, Prox1-CreER^T2/Ctss^f/f; n = 5. All data are presented as mean values ± SD; two-tailed unpaired Student’s t-test. ns not significant. Source data are provided as a Source data file. Consistent with a previous publication^[156]24, we observed a reduction in cerebral blood flow (CBF) at 15 min post-SAH in both Prox1-creER^T2/Ctss^f/f and Ctss^f/f mice compared to baseline (Supplementary Fig. [157]6b, c). However, no significant difference was found between the two groups at either 15 min or 24 h post-SAH established by endovascular perforation (Supplementary Fig. [158]6c). We recorded the respiratory rates (Supplementary Fig. [159]6d) and blood pressure (Supplementary Fig. [160]6e) of mice at different time points; however, no statistical difference was observed between Prox1-creER^T2/Ctss^f/f and Ctss^f/f mice. Similarly, SAS blood clearance, as indicated by SAH grading scores (Supplementary Fig. [161]6f, g), was comparable between the groups. These results suggest that LEC-specific CTSS deficiency does not affect CBF, respiration, blood pressure, or blood clearance in SAS after SAH. CTSS inhibitor reduces neurological deficits and neuroinflammation We also confirmed the CTSS inhibition on SAH-associated neurological deficits and neuroinflammation. Male C57BL/6 mice were administered LY3000328 (80 mg/kg, intraperitoneally, daily, for 3 days), a small-molecule CTSS inhibitor^[162]25, following SAH induction by autologous blood injection. Mortality in the SAH + vehicle and SAH + LY3000328 groups was approximately 30% and 10%, respectively, with no deaths in the sham + vehicle group. Flow cytometry (gating strategy in Supplementary Fig. [163]7a) demonstrated that the percentages of CD45^high leukocytes were dramatically increased at 24 h post SAH surgery, which were significantly reduced by LY3000328 (Fig. [164]6a, b). Similarly, the numbers of F4/80^+ macrophages in the brain were also elevated in the SAH group, which was decreased in the LY3000328 group (Fig. [165]6c, d). However, microglial phenotypic proportions remained unchanged across all groups (Fig. [166]6e, f). Flow cytometry of CLNs (the gating strategy shown in Supplementary Fig. [167]7b) showed no difference in total F4/80^+ macrophage percentages (Fig. [168]6g, h), but a significant increase in F4/80^+TNF-α^+ pro-inflammatory macrophages following SAH, which was attenuated by LY3000328 (Fig. [169]6i, j). Fig. 6. CTSS inhibition mitigates the neurological deficits and neuroinflammation following SAH. [170]Fig. 6 [171]Open in a new tab a–f Flow cytometry analyses demonstrated the cell populations of leukocytes, macrophages, and microglia in the brain at 24 h after SAH induced by autologous blood injection. Representative dot plots (a) and quantification of the percentages of CD45^high leukocytes (b) in the brain. Representative dot plots (c), and quantification of the numbers F4/80^+ macrophages (d) in the brain. Representative dot plots (e), and quantification of the percentages of the CD16/32^+CD206^−, CD16/32^−CD206^+, as well as microglia subsets (f). Sham + vehicle; n = 10, SAH + vehicle; n = 6, SAH + LY3000328; n = 9 mice. Representative dot plots (g) and quantification of the percentages of CD45^+F4/80^+ macrophages in the mandibular LNs at 24 h post-SAH (h). Representative plots (i) and quantification of the percentages of F4/80^+ TNF-α^+ pro-inflammatory macrophages in the mandibular LNs at 24 h post SAH (j). k Survival curve of mice within 72 h. Sham + vehicle; n = 10, SAH + vehicle; n = 5, SAH + LY3000328; n = 8 mice. l Quantification of the modified Garcia scores at 72 h post SAH. Sham + vehicle; n = 10, SAH + vehicle; n = 5, SAH + LY3000328; n = 8 mice. m–o Flow cytometry analyses demonstrated the cell populations of leukocytes, macrophages, and microglia in the brain at 72 h post-SAH. Quantification of the percentages of CD45^high leukocytes (m), the numbers of F4/80^+ macrophages (n), the percentages of the CD16/32^+CD206^− subsets, the CD16/32^−CD206^+ subsets, and the CD16/32^+CD206^+ subsets (o). Sham + vehicle; n = 10, SAH + vehicle; n = 5, SAH + LY3000328; n = 8 mice. All data are presented as mean values ± SD; one-way ANOVA with Turkey’s multiple-comparison test. ns not significant. Source data are provided as a Source data file. SAH surgery also induced significant mortality and neurological deficits; although LY3000328 improved neurological function, it did not significantly enhance survival (Fig. [172]6k, l). Mortality rates were approximately 50% and 20% for the SAH + vehicle and SAH + LY3000328 groups, respectively, with no deaths in the sham + vehicle group. At 72 h post-SAH, LY3000328 treatment continued to reduce the brain leukocyte infiltration, as evidenced by lower percentages of CD45^high leukocyte and numbers of F4/80^+ macrophages (Fig. [173]6m, n and Supplementary Fig. [174]7c). However, its effect on microglial activation and polarization remained limited (Fig. [175]6o). Taken together, the results demonstrated that CTSS inhibition reduces brain leukocyte infiltration and CLN macrophage activation following SAH. Similar to results from CTSS knockout mice, the CBF decreased at 15 min post-SAH compared to baseline in both treatment groups, with no significant difference between vehicle and LY3000328-treated mice (Supplementary Fig. [176]8a, b). Respiratory rate (Supplementary Fig. [177]8c), blood pressure (Supplementary Fig. [178]8d), and blood clearance in SAS (Supplementary Fig. [179]8e, f) also showed no significant differences at 24 h post-SAH. Collectively, these results confirm that CTSS inhibition via LY3000328 has no significant effect on the CBF, respiratory rate, blood pressure, and blood clearance in SAS after SAH. Discussion SAH results in the leakage of blood into the CSF, leading to significant neuroinflammation and an activation of peripheral immune responses, which contribute to EBI and adverse outcomes in SAH patients. Previously, we reported that the extravasated erythrocytes drained to the CLNs via the meningeal lymphatics after SAH^[180]12. However, the mechanisms underlying the cross-talk between the CNS and the peripheral immune system post-SAH, as well as the role of extravasated erythrocytes in CLNs in contributing to SAH-associated brain injury, are inadequately characterized. In this study, we demonstrated that extravasated erythrocytes in the mandibular LNs, one of the superficial CLNs, could be engulfed and degraded by medullary LECs via the lysosomal protease CTSS. This process induces LEC immune response and subsequently promotes SAH-associated brain damage. Conditional depletion of Ctss specifically in LECs, achieved using transgenic mouse models, or systemic inhibition of CTSS by its inhibitor effectively mitigated the neuroinflammation and neurological deficits caused by SAH. The highly specialized microenvironment within the LNs allows them to collect lymph, antigens, and immune cells from tissue, and plays a crucial role in immunity^[181]26,[182]27. Antigens and soluble molecules travel to the LNs along interstitial fluid through afferent lymphatic vessels. Brain antigens and metabolites may directly drain into CLNs via the meningeal lymphatics or be captured by DCs before such drainage, where these brain antigens can activate immune cells that subsequently migrate to the vasculature system^[183]10,[184]28. In ischemic stroke, meningeal lymphatic vessels facilitate the drainage of injury-induced VEGF-C from the CSF to superficial CLNs, leading to the activation of peripheral immune response and, in turn, aggravating the brain injury^[185]10. Therefore, the immune response in the CLNs is crucial for both initiating and sustaining the inflammatory processes affecting the brain. We reported previously that extravasated erythrocytes within the CSF could be drained to CLNs (both mandibular LNs and deep CLNs) at 4 h post SAH. Ablation of the meningeal lymphatics prior to SAH surgery obstructed this drainage, resulting in the accumulation of extravasated erythrocytes in the brain. These accumulated extravasated erythrocytes are detrimental to SAH by activating resident immune cells such as microglia and astrocytes within the brain^[186]12. In the current study, with the meningeal lymphatics intact during the creation of the SAH model, extravasated erythrocytes in the CSF were successfully drained into CLNs, where they activated the peripheral immune responses to promote brain damage associated with SAH. Consequently, CLN removal could be beneficial by reducing systemic inflammation and brain injury. Therefore, brain lymphatic vessels and CLNs likely play a distinct role in the pathogenesis of SAH, warranting further investigation. Recently, vascular endothelial cells (VECs) were reported to function as amateur APCs, engulfing IgG-opsonized myelin debris following spinal cord injury and experimental autoimmune encephalomyelitis and subsequently delivering it to lysosomes^[187]29. These processes promote the macrophage recruitment and fibrosis^[188]29. Moreover, VECs have been shown to phagocytose aged and phosphatidylserine-exposing erythrocytes, which are important for erythrocyte clearance in an atherothrombotic and tumor environment^[189]30,[190]31. Although LECs share many features with VECs, however, their phagocytic capacity has been less thoroughly explored. Transcriptomic analyses revealed that LN LECs display high phagocytic and endocytic capacities, expressing C-type lectin, scavenger, and Fc receptors^[191]27,[192]32. LECs are capable of capturing both exogenous and endogenous antigens by phagocytosis^[193]33. Fluorescently labeled ovalbumen, viral particles, and antigens injected into mice were captured by LYVE-1^+ LECs in the draining LN^[194]34–[195]36. Additionally, genes encoding several scavenger receptors involved in receptor-mediated endocytosis are upregulated in LN LECs^[196]32. These findings collectively suggest that LN LECs possess substantial phagocytic capacities. In our study, we observed that extravasated erythrocytes drained into mandibular LNs and were significantly degraded at 24 h post-SAH, with most intact and degraded erythrocytes co-localizing with LYVE-1^+ LECs. We confirmed these findings using two widely employed murine models of SAH. In vitro experiments demonstrated that CFSE-labeled erythrocytes were engulfed by LECs and subsequently delivered to lysosomes for degradation. The presence of CFSE-labeled erythrocytic particles within lysosomes and the corresponding increase in lysosome size confirmed successful engulfment. Furthermore, the PI3K inhibitor wortmannin, which blocks the phagocytosis ability of macrophages^[197]37, significantly decreased erythrocyte engulfment by macrophages but had a limited impact on LECs (Supplementary Fig. [198]9). These data indicate that the mechanisms by which the LECs and macrophages phagocytose erythrocytes may be different; however, the details still need further elucidation. CTSS, a member of the cathepsin family, is present in endo-lysosomal compartments and may be produced under both steady-state and pathological conditions^[199]38. DCs and macrophages express high levels of CTSS, which is important for the formation of major histocompatibility complex class II molecules and collagen degradation in diseases, respectively^[200]39,[201]40. LN LECs also exhibit a strong ability to secrete CTSS, whereas LECs from lymphatic vessels express it at minimal levels^[202]32. Our scRNA-seq results demonstrated that the DEGs of medullary LECs were highly enriched in lysosomes, and among these genes, Ctss (encoding CTSS) was significantly increased in the SAH group at 24 h post-surgery. In vitro, bulk RNA-seq data also showed that the expression of Ctss significantly increased after incubation with RBCs, and the results were verified by quantitative PCR. Moreover, immunofluorescence staining validated that the lysosomes and CTSS mainly co-localized with LYVE-1^+ LECs in the mandibular LNs, respectively. Erythrocytes were observed to be enveloped by the lysosomes and CTSS within LYVE-1^+ LECs. We generated Prox1-CreER^T2;Ctss^flox/flox mice to specifically knock out the CTSS in LECs and then established an endovascular perforation-SAH model. Results showed that the depletion of Ctss specifically in LECs attenuates SAH-associated brain damage, indicating the critical role of LEC-derived cathepsin-S in SAH pathogenesis. LY3000328, an inhibitor of CTSS that has been utilized in a clinical trial^[203]25, also dramatically decreased the neurological deficits and brain leukocyte infiltration in mice following SAH. When LECs are exposed to antigens, they secrete cellular chemokines that attract a series of innate and adaptive immune cells into LNs to broaden and maintain ongoing immune responses^[204]41. LEC-produced CCL21 is essential for the recruitment of DCs into the LNs and the promotion of LN expansion^[205]42. LN LECs not only receive signals from immune cells, but also provide signals to the adaptive immune system to regulate peripheral immune tolerance and protective immune responses^[206]43. In our study, we found that LECs, after engulfing erythrocytes, upregulated the expression of genes including those related to cytokines, inflammatory factors, and interferon-inducible genes, which play important roles in the immune response. Malhotra et al. also reported that LN LECs increase interferon-inducible gene expression, including genes that encode Mpeg1, Lcn2, and Irf7, following inflammatory stimulation^[207]44; these genes were also upregulated in our study. As revealed by an scRNA-seq study, multiple transcriptomic differences exist among LEC sub-populations, e.g., cells lining the floor or the ceiling of the subcapsular sinus, the medullary sinus, and the cortical sinus^[208]45. In our scRNA-seq, three LEC subsets in mandibular LNs included ceiling, floor, and medullary LECs. Based on immunofluorescence staining, extravasated erythrocytes predominantly co-localized with LYVE-1^+ LECs, and as shown by scRNA-seq data, the gene Lyve1 was primarily expressed by floor and medullary LECs. The DEGs in medullary LECs were highly enriched in lysosomes, and among lysosomal genes, Ctss levels were significantly upregulated at 24 h post-SAH. However, lysosome was not among the top 20 KEGG terms in floor LECs, and Ctss expression did not increase in this subset. Co-labeling of medullary LEC markers MARCO and LYVE-1 with TER-119 (for erythrocytes) also confirmed the endocytosis of extravasated erythrocytes by medullary LECs. These results indicate that medullary LECs in mandibular LNs are the primary subset to engulf and degrade the extravasated erythrocytes after SAH. The importance of CLNs in brain pathology has been highlighted in recent studies^[209]10,[210]11,[211]46; however, how CLNs participate in SAH pathology when extravasated erythrocytes in CSF have been drained into CLNs remains largely unclear. In the current study, we demonstrated that extravasated erythrocytes are engulfed by LECs in the mandibular LNs at 24 h post-SAH and degraded by the lysosomal protease CTSS from LECs. Removal of mandibular LNs, specifically deficiency of CTSS in LECs using transgenic mice, and systemic blockage of CTSS by its inhibitor decreased SAH-induced brain damage. Our study provides some insights into the CLNs-brain crosstalk in early SAH; however, several limitations and questions remain to be addressed in the future. First, to the best of our knowledge, deep CLNs also play an important role in the drainage of the brain^[212]7,[213]8. We observed robust accumulation of extravasated erythrocytes in mandibular LNs compared to deep CLNs, and we only removed the mandibular LNs in this study. A recent study also highlighted the significance of superficial CLNs in ischemic stroke^[214]10. Future research should elucidate how and to what extent deep CLNs participate in SAH pathology. Second, our scRNA-seq and immunofluorescence results showed that medullary LECs in the mandibular LN may be the primary LEC subset involved in the degradation of extravasated erythrocytes via CTSS. However, whether other LEC subsets, such as ceiling and floor LECs, respond to extravasated erythrocytes after SAH still needs clarification. Third, we have only elucidated the importance of LEC-CTSS and observed activation of CLN-derived macrophages in SAH; however, the downstream mechanisms and how CLN immune responses affect brain damage remain unclear. The CLN-derived macrophages and other immune cells penetrating into the brain after SAH are complex issues. Studies reporting macrophages and immune cells in SAH provide some insights for future study^[215]47–[216]49. More rigorous studies should be designed to examine the details of CLN immune responses in relation to the severity and timing of brain damage in SAH. Fourth, while we emphasized the endocytosis and degradation of extravasated erythrocytes by LECs in this study, we also found that degraded erythrocytes co-localized with both LYVE-1^+ LECs and CD169^+ macrophages. Although our experiments suggest that endocytosis of extravasated erythrocytes by macrophages might not play a critical role in our SAH models, we still cannot exclude the possibility of their contribution to SAH pathogenesis. However, as revealed in this study, at least LECs in mandibular LNs play a significant role in erythrocyte degradation via CTSS, which is critical in SAH-induced brain damage. Further studies should clarify the differences between macrophages and LECs in terms of phagocytosis and degradation, and the extent of their participation in brain pathologies during diseases. Finally, although we present strong experimental data in this study and in a previous publication^[217]12 demonstrating that extravasated erythrocytes in CLNs were drained from CSF via meningeal lymphatics after SAH, we cannot exclude the possibility that some erythrocytes extravasated via existing arteries in CLNs. In summary, our study demonstrates that CLNs, e.g., mandibular LNs in this study, are involved in the neuroinflammatory response following SAH. Extravasated erythrocytes that drain into the mandibular LNs are engulfed by the medullary subset of LECs and subsequently degraded via lysosomal protease CTSS. LECs increased immune response following erythrocyte engulfment, worsening SAH-associated damage. Notably, targeted depletion of CTSS specifically in LECs using Prox1-creER^T2/Ctss^f/f mice, as well as systemic inhibition of CTSS, significantly attenuated SAH-induced neurological damage. These findings reveal a mechanism by which SAH triggers the peripheral immune signaling through CLNs and subsequently promotes brain damage, and identify a potential therapeutic strategy to ameliorate brain injury associated with this condition. Methods Ethical regulation All experimental procedures were approved by the Longhua Hospital-Animal Ethics Committee (no. LHERAW-25054) and were conducted under the Guiding Principles for the Care and Use of Laboratory Animals Approved by Animal Regulations of the National Science and Technology Committee of China. Animals C57BL/6 male mice (6–8-weeks-old) were purchased from the Shanghai Model Organisms Center or Shanghai Jiesijie Laboratory Animal Co., Ltd. Ctss^flox/flox mice were generated by the Shanghai Model Organisms Center. Prox1-Cre-ER^T2 mice (Stock Number: 022075) were obtained from Jackson Laboratories (Bar Harbor, ME, USA). Both Ctss^flox/flox and Prox1-Cre-ER^T2 mice were in a C57BL/6J background. To generate Prox1-CreER^T2;Ctss^flox/flox mice, Ctss^flox/flox mice were crossed with Prox1-Cre-ER^T2 mice to obtain Prox1-Cre-ER^T2; Ctss^flox/+ mice, which were then mated with Ctss^flox/flox mice. These processes were conducted by the Shanghai Model Organisms Center. Both sexes (male and female) of Prox1-CreER^T2;Ctss^flox/flox mice and their Ctss^flox/flox littermates were used in the experiment. All mice were housed in specific pathogen-free facilities with a temperature of 22 ± 2 °C, humidity of 50% and 12-h day and night cycles, and free to access food and water. All mice were anesthetized with isoflurane (RWD Life Science Co., Ltd, China) followed by surgery, perfusion, or tissue harvest. Antibodies Rabbit anti-LYVE-1 antibody (1:1000, ab14917, Abcam), rat anti-CD11b fluorescein isothiocyanate (FITC)-conjugated antibody (1:100, 11-0112-82, eBioscience), rat anti-CD16/32 allophycocyanin (APC)-conjugated antibody (1:100, 558636, BD Bioscience), rat anti-CD206 R-phycoerythrin (PE)-conjugated antibody (1:100, 12-2061-80, eBioscience), rat anti-CD45 BUV395 antibody (1:100, 565967, BD Pharmingen), rat anti-F4/80 BV421-conjugated antibody (1:100, 123137, Biolegend), rat anti-F4/80 PerCP-Cy5.5-conjugated antibody (1:100, 45-4801-82, eBioscience), rat anti-CD45 FITC-conjugated antibody (1:100, 11-0451-82, eBioscience), rat anti-CD11b APC-conjugated antibody (1:100, 17-0112-82, eBioscience), rat anti-Ly6G PE-conjugated antibody (1:100, 12-9668-82, eBioscience), rat anti-TNF-α PE-conjugated antibody (1:100, 12-7321-82, eBioscience), rat anti-CD31 PE-conjugated antibody (1:10, 553373, BD Pharmingen), rat anti-Pdpn PE-conjugated antibody (1:10, 566390, BD Pharmingen), rat anti-TER-119 PE-conjugated antibody (1:100, 12-5921-81, eBioscience) rabbit anti-Iba-1 antibody (1:1000, 019-19741, woko), rat anti-CD169 antibody (1:100, 142402, Biolegend), rat anti-TER-119 antibody (1:100, 14-5921-85, eBioscience), rat anti-MARCO antibody (1:200, GTX39744, Genetex), rabbit anti-ANXA2 antibody (1:200, ab178677, Abcam), rabbit anti-LAMP-1 antibody (1:1000, ab24170, Abcam), rabbit anti-CTSS antibody (1:1000, NBP3-25398, Novus Biologicals), rat anti-LYVE-1 antibody (ALY7) (1:200, 14-0443-82, eBioscience), rabbit anti-Prox1 (1:100, 11-002 P, AngioBio), rabbit anti-NeuN antibody (1:1000, ab177487, abcam), rabbit anti-TNF-ɑ antibody (1:1000, 11948, Cell Signaling Technology), rabbit anti-GFAP antibody (1:1000, ab7260, abcam), rabbit anti-Cleaved-IL-1β antibody (1:1000, 63124, Cell Signaling Technology), mouse anti-GAPDH antibody (1:1000, 60004-1-Ig, proteintech), HRP-linked anti-rabbit antibody (1:1000, 7074, Cell Signaling Technology), HRP-linked anti-mouse antibody (1:1000, 7076, Cell Signaling Technology), Alexa Fluor 555-conjugated goat anti-rat antibody (1:500, 4417S, Cell Signaling Technology), Alexa Fluor 488/555-conjugated goat anti-rabbit antibody (1:500, 4412S/4413S, Cell Signaling Technology), Alexa Fluor 405-conjugated goat anti-Rabbit IgG (H + L), and cross-adsorbed secondary antibody (1:200, A-31556, Invitrogen). Alexa Fluor 647-conjugated goat anti-rabbit IgG (H + L), F(ab’)2 Fragment (1:500, 4414, Cell Signaling Technology). Generation of Prox1-CreER^T2/Ctss^f/f mice Ctss^flox/flox mice were generated on a C57BL/6J mouse background by Shanghai Model Organisms Center, Inc. (Shanghai, China) using the CRISPR/Cas9 system. Cas9 mRNA was transcribed in vitro using the mMESSAGE mMACHINE T7 Ultra Kit (Ambion, TX, USA), following the manufacturer’s instructions. A Ctss donor vector containing flox sites flanking exon 2 of the Ctss gene was constructed. Single-guide RNA (sgRNA) sites targeting introns 1 and 2 were transcribed in vitro. SgRNA target sites for intron 1 were 5’-TTCCAGAAAATTTCCAAGTGGGG-3’. SgRNA sites for intron 2 were 5’-CATCCTACCAAGGATGCGGTGGG-3’. The donor vector with two sgRNAs and Cas9 mRNA was microinjected into C57BL/6J fertilized eggs. F0 generation mice positive for homologous recombination were identified by long-range PCR. The primers (P1-P4) used for genotyping the correct homology recombination were P1: 5’-TGTCCCTAGCAGCATGACAC-3’ and P2: 5’-GATAGCCGCAGGAGGCTTAG-3’ for the correct 5’ homology arm recombination, and P3: 5’-GGGTTAGCAGGCAAAGGTACT-3’ and P4: 5’-TGTGTTACCGACACTGGAGG-3’ for the correct 3’ homology arm recombination. The PCR products were further confirmed by sequencing. The genotype of F1 generation Ctss flox heterozygous mice was identified by long-range PCR. Ctss flox heterozygous mice were crossed with Prox1-Cre-ER^T2 mice to obtain Prox1-Cre-ER^T2;Ctss^flox/+ mice. Both male and female Prox1-CreER^T2;Ctss^flox/flox, as well as Ctss^flox/flox mice were used in experiments, as no significant sex-based differences were observed in terms of phenotype or expression of the genes of interest. Mandibular lymph node excision C57BL/6 male mice (6–8-weeks- old) were anesthetized with 5% isoflurane (RWD Life Science Co., Ltd, China) and maintained under 2% isoflurane. The anterior neck area of the mice was shaved and sterilized. Skin incisions were made, and bilateral mandibular LNs were surgically excised. The control group mice underwent the same procedures without mandibular LNs removal. The mice were sutured and allowed to recover from the surgery for 7 days before SAH induction. Induction of SAH To establish the SAH model, C57BL/6 (6–8-weeks-old) male mice were anesthetized with isoflurane (RWD Life Science Co., Ltd, China) and secured in a stereotactic frame (RWD Life Science Co., Ltd, China). The hair of the mice of the posterior neck was shaved, and the skin was incised, then the posterior neck muscles were separated to access the cisterna magna. In the SAH group, 60 µL of autologous blood withdrawn from the right femoral artery was injected into the cisterna magna of mice, whereas mice in the control group received 60 µL of saline. The needle was retained in place for 2 min to prevent leakage. The SAH mouse model was also induced by endovascular perforation^[218]50. C57BL/6 male mice (6–8-weeks-old), as well as both male and female Prox1-CreER^T2;Ctss^flox/flox and Ctss^flox/flox mice (7 weeks old), were anesthetized with isoflurane (RWD Life Science Co., Ltd, China) and placed under the microscope (OLYMPUS, Japan). Then the left carotid artery was exposed, and the external and internal carotid arteries were separated. A small incision was made on the left external carotid artery (ECA), through which a blunt 5-0 nylon suture was inserted and advanced into the internal carotid artery, further advancing the filament when resistance was encountered, which ultimately punctured the circle of Willis. Sham-operated mice underwent the same procedures except for the perforation of the circle of Willis. After surgery, mice were sutured and placed on a 37 °C thermo plate (TOKAI HIT, Japan) for recovery, the mice that died before the end of the experiment were excluded from data analysis. Tamoxifen intervention Both sexes (male and female) of Prox1-CreER^T2/Ctss^f/f mice (5-weeks-old) and their Ctss^f/f littermates received intraperitoneal injections of tamoxifen (T5648, Sigma) in a dose of 100 μg/g body weight daily for seven consecutive days. When Prox1-CreER^T2/Ctss^f/f mice and their Ctss^f/f littermates were seven weeks old, they underwent SAH induction via endovascular perforation. Neurological function assessment The neurological function of mice was assessed using the Modified Garcia scores^[219]51. The components of the scale contain six parts with a total of 18 points, including spontaneous activity, spontaneous movement of all limbs, movement of forelimbs, reaction to touch on both sides of the trunk, and response to vibrissae touch. The first four items are scored from 0 to 3 points, and the last two items from 1 to 3 points. Higher scores indicate better neurological function. SAH grading SAH grading was used to evaluate the blood clearance in the SAS at 24 h post-SAH and was conducted blindly by independent investigators. The basal brain was divided into six segments, with a total score of 18 points. Each segment was scored from 0 to 3 points according to the amount of subarachnoid blood clot in the segment. Grade 0: no subarachnoid blood; Grade 1: minimal subarachnoid blood; Grade 2: moderate blood clot with recognizable arteries; and Grade 3: blood clot obliterating all arteries within the segment^[220]52. CBF To monitor the CBF following SAH induction, a laser speckle blood flow imaging system (Tow-INT Tech Co., Ltd, China) was employed. Mice were anesthetized with isoflurane (5% for induction, 2% for maintenance), secured in a stereotaxic frame (RWD Life Science Co., Ltd., China), and their scalps were shaved. The skull was exposed along the midline and thinned using a handheld micro drill (RWD Life Science Co., Ltd., China). Mineral oil (M8040, Solarbio) was applied to prevent dryness of the skull. CBF measurements were recorded at baseline, and at 15 min and 24 h post-SAH. Measurements from the ipsilateral cortex (excluding the transverse and sagittal sinuses) were expressed as a percentage relative to the baseline values. Respiratory rates Respiratory rate was assessed using a whole-body plethysmograph (Tow-INT Tech Co., Ltd., China)^[221]53. The room for respiratory rate assessment was kept quiet, and mice were placed in the chambers for acclimation for 30 min before recording. When the respiratory wave in the detector is stabilized, respiration was recorded for over a 20-min period, and the breath frequencies were determined by the ResMass 1.4.2. software (Tow-INT Tech Co., Ltd., China). Measurements were obtained before SAH surgery, at 2 h and 24 h post-SAH. The average breaths per minute across the recording interval were calculated for each mouse. Blood pressure Blood pressure was measured via the non-invasive tail-cuff method (NIBP, Yuyan Instrument Co., Ltd., China). Mice were placed in a restraining chamber, and a cuff maintained at 37 °C was applied to the tail for 30 min before examination. The cuff inflated to detect pressure fluctuations. Measurements were obtained once readings stabilized. Each mouse was recorded at least three times, with a systolic pressure variation < 10 mmHg between measurements. The mean blood pressures of each mouse were calculated from three recordings. Flow cytometry Peripheral blood was collected and centrifuged at 450×g for 15 min at 4 °C to remove plasma. Red blood cells were eliminated by lysis buffer (ACK, C3702, Beyotime) for 5 min with gentle shaking at 4 °C, followed by centrifugation and a wash with FACS buffer (phosphate-buffered saline (PBS), 2% fetal bovine serum, 5 mM EDTA). Cells were then collected for the subsequent antibody staining. Brains of mice collected after transcardial perfusion were gently minced and digested by collagenase A (10103578001, Sigma Aldrich). The cell suspension was filtered through a 70-μm nylon mesh cell strainer (BD Biosciences) and added to 3 mL of 100% stock isotonic percoll (SIP) (17089109, GE). Subsequently, 2 mL volume of 70% SIP was carefully layered beneath the cell suspension, and the tubes were centrifuged at 500×g for 30 min without brake. The cells were collected from the interphase of the gradient, washed with PBS, and centrifuged. The collected cells were stained with the corresponding antibodies. Mandibular LNs were gently pressed and filtered to prepare a single cell suspension, followed by centrifugation to collect the cell pellet. The cells were stained using Fixable Viability Dye eFluor^TM 780 (65-0865-18, eBioscience) and the corresponding cellular markers. The control labeled cells were used to determine the gates, voltages, and compensations in multivariate flow cytometry. Samples were analyzed by Longzoe Shanghai Biotechnology Co., Ltd. using BD Fortessa X20 or Attune TxT Acoustic Focusing Cytometer (Invitrogen), and data were processed with FlowJo V10 software. scRNA-seq Mandibular LNs dissociation and cell purification At 24 h post-SAH or sham operation, mandibular LNs were harvested from 20 mice per group. The LNs were minced with 25 G needles and digested at 37 °C in an enzyme solution containing DNase (100 μg/mL), dispase II (800 μg/mL), and collagenase P (200 μg/mL). The mixture was gently pipetted, and after 8 min of digestion, the supernatant was collected and added to RPMI/2 mM EDTA/1% FCS. Fresh enzyme media was then added to the remaining tissue. Once the tissue was completely digested, the cell suspension was collected and filtered through 70-μm nylon mesh cell strainers (BD Bioscience). After centrifuging, the cells were stained by CD45-microbeads (130-052-301, Miltenyi Biotec) for 15 min at 4 °C and passed through LD columns (Miltenyi Biotec) in a magnetic separator (Miltenyi Biotec) to deplete CD45^+ cells. The CD45 negative cells were collected, centrifuged, and incubated with anti-CD31 PE-conjugated and anti-PDPN PE-conjugated antibodies for 10 min. After washing with FCS buffer, the cells were incubated with anti-PE microbeads (130-048-801, Miltenyi Biotec) for 15 min at 4 °C and subsequently passed through LS columns (Miltenyi Biotec) in the magnetic separator (Miltenyi Biotec). The PE-positive cell population was collected for scRNA-seq. scRNA-seq The sorted cells were loaded onto a 10× Genomics Chromium platform to generate Gel Beads-in-Emulsion (GEM). During cDNA synthesis, each cDNA molecule was tagged on the 5’ end with a unique molecular identifier (UMI) and cell label. Briefly, 10× beads were then subjected to second-strand cDNA synthesis, adapter ligation, and universal amplification. The library construction was performed according to the manufacturer’s instructions ([222]CG000206 RevD). Sequencing libraries were quantified using a High Sensitivity DNA Chip (Agilent) on a Bioanalyzer 2100 and the Qubit High Sensitivity DNA Assay (Thermo Fisher Scientific). The libraries were sequenced on NovaSeq6000 (Illumina) using 2 × 150 chemistry. ScRNA-seq data processing Reads were processed using the Cell Ranger 5.0 pipeline. FASTQ files generated from Illumina sequencing output were aligned to the mouse genome mm10 using the STAR algorithm. Gene-barcode matrices were generated for each individual sample after filtering and which contain the barcoded cells and gene expression counts. The Seurat (v3.0.2) R was used for quality control and downstream analyses of our scRNA-seq data. Cells expressing fewer than 200 or greater than 6000 unique genes, or with more than 10% mitochondrial genes, were excluded. The data was normalized by the Seurat package for extracting variable genes, which were identified while controlling for the strong relationship between variability and average expression. Data was integrated from different samples after identifying ‘anchors’ between datasets using FindIntegrationAnchors and IntegrateData in the Seurat package. Dimensionality reduction and clustering Principal component analysis (PCA) was conducted, and the top 20 principal components were selected after scaling the data. The clusters were visualized on a 2D map produced using t-SNE. Identification of cell types and subtypes by nonlinear dimensional reduction. Shared nearest neighbor (SNN) clustering optimized with the Louvain Method. For each cluster, DEGs comparing the remaining clusters were identified using the Wilcoxon Rank-Sum Test, and SingleR and known marker genes were used to characterize the cell type. Enrichment analysis DEGs were identified using Seurat’s FindMarkers function with all default parameters. Gene Ontology (GO), KEGG, and GSEA pathway enrichment analysis of DEGs were conducted by the clusterProfiler (v 3.14.0) package. The Wilcoxon rank sum test was applied for comparing the data, and the Bonferroni method was used for P-value correction. In the comparison between two groups, genes with an absolute log2 fold change (|log2FC|) of more than 0.5 and adjusted P < 0.05 were considered to have significant differences. Drug administration LY3000328 (HY-15533, MedChemExpress) was dissolved according to the manufacturer’s instructions. The mice in the LY3000328 group were administered intraperitoneally daily in a dose of 80 mg/kg for either 1 or 3 days. Control mice received an equal volume of vehicle. Tissue processing Brains, mandibular LNs, deep CLNs, and superficial parotid LNs were collected following transcardial perfusion and fixed in 10% formalin solution for 24 h. Before embedding in OCT, the tissues were dehydrated by a gradient of sucrose solutions, including 10%, 20%, and 30%. Sections of 7 µm thickness were prepared using a cryostat (Leica, CM3050S). Immunofluorescence The sections were processed with blocking by 0.3% Triton X-100 with 5% BSA for 1 h at room temperature, followed by incubation with primary antibodies diluted in 5% bovine serum albumin (BSA) overnight at 4 °C. After washing with PBS, the sections were incubated with fluorescent-conjugated secondary antibodies diluted in PBS for 2 h at room temperature. Finally, slides were mounted with mounting medium. Images of the sections were acquired using an Olympus VS200 microscope or a Nikon N-SIM confocal microscope. TUNEL staining Brain sections were first incubated with the primary antibody overnight at 4 °C, followed by the corresponding secondary antibody incubation. TUNEL staining was then performed according to the manufacturer’s instructions (C1088, Beyotime). Finally, slices were mounted, and images were obtained by an Olympus VS200 microscope (Japan). Erythrocyte isolation and labeling Whole blood was collected from the right femoral artery or abdominal aorta of C57BL/6 male mice (6–8-weeks-old), then diluted with PBS containing 2% fetal bovine serum, followed by centrifugation at 800×g for 10 min without braking. Erythrocytes were collected from the bottom of tubes and resuspended at 10^6 cells/mL. The 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE, 20 μM/mL, 65-0850-84, eBioscience) was added into the erythrocyte suspension and incubated for 10 min at 37 °C. After washing with 2% FBS–PBS, the erythrocyte suspension was centrifuged at 500×g for 5 min, then the labeled erythrocytes were resuspended either (about 10^6 cells in 60 µL) for injection into the cisterna magna of C57BL/6 male mice or at 1 × 10^6 cells/mL for co-culture with LECs for 24 h of incubation. LEC culture The mouse LEC line used in this study was originally generated by Dr. S. Ran (Southern Illinois University School of Medicine, Springfield, IL) from benign lymphangiomas induced by Freund’s adjuvant^[223]54 and used in the study of LEC for decades^[224]55. The primary mouse LECs were purchased from Procell Life Science & Technology Co., Ltd. (CP-M023) and cultured in LEC complete culture medium (CM-M023, Procell Life Science & Technology)^[225]56. To acquire primary LECs, ICR male mice (3–4-weeks-old) were killed and sterilized, and the lymphatic vessels (thoracic duct) were isolated and digested by collagenase cocktail overnight at 4 °C. The digestion was stopped with LEC complete medium, and the suspension was filtered through 70 μm nylon mesh cell strainers (BD Bioscience). After centrifugation at 450×g for 5 min, cells were resuspended in LEC complete medium and seeded in a T25 culture flask. LECs are identified by the expression of classical LEC markers, including Prox-1, Lyve-1, and VEGFR3. LEC lines, or primary LECs, were seeded in a plate for 12 h, and then an RBC suspension, or RBC-CFSE in medium, was added for 24 h. After the removal of supernatant, cells were collected for use. LysoTracker red staining and analysis LEC lines or primary LECs were incubated with or without RBC-CFSE for 24 h. After washing away the non-engulfed erythrocytes, cells were stained with LysoTracker Red (L7528, Invitrogen) and Hoechst (62249, Thermo Scientific) for 15 min at 37 °C. After washing with PBS, cells were immediately imaged using a Leica laser scanning confocal microscope with a 60 × 1.4 NA oil immersion objective. LysoTracker Red positive puncta were analyzed using Image J 6.0, and their diameter and area were measured. Bone marrow-derived macrophage (BMDM) cultures The 7–8-week-old C57BL/6 male mice were killed and sterilized in 70% ethanol. The legs were harvested, and the bone marrow in the tibia and femur was flushed into a culture dish. Then the bone marrow was centrifuged at 450×g for 5 min, and red blood cells were removed by lysis buffer (C3702, Beyotime). The cell pellet was resuspended in DMEM medium supplemented with 20 ng/mL macrophage colony-stimulating factor (MCSF) and filtered with 40 μm nylon mesh cell strainers (BD Bioscience). The cells were seeded on a dish (Corning), and fresh DMEM medium containing 20 ng/mL MCSF was added on the 2nd and 4th days. After 6–7 days, the bone marrow cells were fully differentiated into macrophages. Bulk RNA-seq LEC line was incubated with RBCs for 24 h, after which total RNA was extracted using the TRIzol reagent (9109, Takara), following the manufacturer’s instructions. RNA purity and quantification were evaluated using the NanoDrop 2000 spectrophotometer (Thermo Scientific, USA), and integrity was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA). The libraries were constructed by the TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The transcriptome sequencing and analysis were performed by OE Biotech Co., Ltd. (Shanghai, China). The libraries were sequenced on an Illumina HiSeq X Ten platform, and 150 bp paired-end reads were generated. Raw data (raw reads) of FASTQ format were first processed using Trimmomatic^[226]57, and clean reads were acquired by removing the low-quality reads. The read counts of each gene were quantified with HTSeq^[227]58. The DESeq (2014) R package was used to analyze differential expression^[228]59. A P-value < 0.05 and fold change > 2 were considered as significantly different expression. Hierarchical cluster analysis of DEGs was performed to demonstrate the expression pattern of genes in different groups and samples. GO and KEGG^[229]60 pathway enrichment analyses of DEGs were performed, respectively, using R based on the hypergeometric distribution. Western blot Brains were harvested on ice following transcardial perfusion, and cortex was isolated and lysed using RIPA buffer (P0013C, Beyotime) containing protease and phosphatase inhibitors (P1045, Beyotime). After centrifugation, the protein was collected and loaded onto 12–15% Tris/tricine SDS gels, followed by transfer to polyvinylidene fluoride membrane (IPVH00010, Millipore). The membranes were blocked with 5% BSA/TBST for 1 h at room temperature and then incubated overnight at 4 °C with the corresponding primary antibody. After washing three times, membranes were incubated with horseradish peroxidase-conjugated IgG for 1 h at room temperature. Protein bands were detected by using ECL Western blot detection reagents (WBKLS0500, Millipore). Band density was quantified with ImageJ software. Protein expression levels were analyzed from three brain samples per group. qRT-PCR Total RNA from LECs was extracted using the EZ-press RNA purification kit (EZBioscience) and homogenized. The cDNA was synthesized using the Reverse Transcription System (G490, Abm). qRT-PCR was conducted using SYBR Green qPCR Master Mix (G3322, ServiceBio). The primers were designed and synthesized by BGI (China), which are is listed in Supplementary Table [230]1. Phagocytosis analysis RBCs labeled with CFSE (1 × 10^6 cells/mL) and wortmannin (200 nM/mL) were added to the BMDMs or LEC lines for 24 h of incubation. Following incubation, cells were fixed with 2% paraformaldehyde, and the cytoskeleton was stained with Phalloidin (40734ES75, Yeasen Biotechnology Co., Ltd.) according to the manufacturer’s instructions. Images were acquired with an Olympus VS200 microscope (Japan). Image analysis The images of lymph node sections and cells on glass coverslips were acquired using an Olympus VS120 microscope (20× objective) or an N-SIM Nikon confocal microscope (20× objective; resolution 1024 × 1024 pixels; z-step 2 µm). Exposure settings and brightness/contrast adjustments were applied uniformly across all images. Images were analyzed with the ImageJ (NIH) software. One left mandibular lymph node per mouse was analyzed to assess extravasated erythrocyte engulfment by LECs. Four fields from four sections of each lymph node were used to quantify the number of erythrocytes that colocalised with LECs. Two sections of the brain were stained, and eight fields were quantified. The mean value of each brain from a total of eight fields and each mandibular LN from four fields was used to make a plot graph, respectively. Raw data were collected using the Microsoft Excel 2007 software. Statistical analysis Data were expressed as means ± SD, with differences between mean values tested by two-tailed unpaired Student’s t-test for the comparison of two groups and one-way ANOVA for the comparison of multiple groups using GraphPad Prism 6 software. No samples or animals were excluded from the data analysis, except for the mice that were dead before the end of the experiment. Each value in the statistical graph represents an individual biological sample, where technical replicates have been averaged. P-values < 0.05 were considered statistically significant. Reporting summary Further information on research design is available in the [231]Nature Portfolio Reporting Summary linked to this article. Supplementary information [232]Supplementary Information^ (2.3MB, pdf) [233]Reporting Summary^ (1.1MB, pdf) [234]Transparent Peer Review file^ (241.1KB, pdf) Source data [235]Source Data^ (27MB, xls) Acknowledgements