Heterogeneity of perivascular astrocyte endfeet depending on vascular regions in the mouse brain

Summary Astrocytes interact with not only synapses but also brain blood vessels through perivascular astrocyte endfeet (PV-AEF) to form the neurovascular unit (NVU). However, PV-AEF components have not been fully identified. Here, we biochemically isolated blood vessels from mouse brain homogenates and purified PV-AEF. The purified PV-AEF were observed in different sizes, similar to PV-AEF on brain blood vessels. Mass spectrometry analysis identified 9,762 proteins in the purified PV-AEF, including cell adhesion molecules, nectin-2δ, Kirrel2, and podoplanin. Immunofluorescence microscopic analysis revealed that nectin-2δ and podoplanin were concentrated mainly in arteries/arterioles and veins/venules of the mouse brain, whereas Kirrel2 was mainly in arteries/arterioles. Nectin-2α/δ, Kirrel2, and podoplanin were preferentially observed in large sizes of the purified PV-AEF. Furthermore, Kirrel2 potentially has cell adhesion activity of cultured astrocytes. Collectively, these results indicate that PV-AEF have heterogeneity in sizes and molecular components, implying different roles of PV-AEF in NVU function depending on vascular regions.


INTRODUCTION
Protoplasmic astrocytes, the predominant glial cell type in the brain, highly ramify by radiating from the cell soma primary branches that gradually divide into numerous fine terminal processes. 1,2They establish an interconnected network through terminal processes and further interact with numerous synapses through perisynaptic astrocyte processes to form tripartite synapses and with blood vessels through perivascular astrocyte endfeet (PV-AEF) to form neurovascular unit (NVU).][4][5][6] During brain development, glutamatergic excitatory synapses are generated after astrocyte formation, whereas GABAergic inhibitory interneurons establish a functional network in the embryonic brain before astrocyte formation. 7,8In parallel with this, the blood vessel network is formed and established. 8Astrocytes start to ramify postnatally by first extending longer and less ramified processes, with spongiform process formation starting at the soma and extending centrifugally until all processes are equipped with fine terminal processes. 2,8,9At this stage, one process first approaches blood vessels to become PV-AEF to form NVU, and the remaining processes then approach many synapses of one neuron to form tripartite synapses and processes of neighboring astrocytes to establish an interconnected network. 1,2,8,9eurons and blood vessels secrete many substances to astrocytes to regulate their differentiation and ramifications, whereas astrocytes also secrete many substances to neurons to regulate synapse formation, maturation, and function and to blood vessels to regulate angiogenesis, vessel relaxation and contraction, and BBB. 4,6,10Thus, neurons, astrocytes, and blood vessels mutually regulate their functions through many soluble factors.However, it remains unknown how astrocytes ramify and extend processes to blood vessels to form PV-AEF during the development of the brain.
Astrocytes are functionally polarized with perisynaptic astrocyte processes enriched with the glutamate transporters EAAT1/GLAST and EAAT2/GLT-1 and the inward rectifier K + channel 4.1 (Kir4.1) 113][14][15][16][17][18][19] PV-AEF attach to the extracellular matrix proteins, such as laminin and agrin, in BM of blood vessels through the a-dystroglycan and b-dystroglycan complex.AQP4, TRPV4, and Kir4.1 localize by associating with b-dystroglycan through the Isolation of the purified PV-AEF with different sizes from the isolated blood vessels We next isolated PV-AEF from the isolated blood vessels (Figure 3A).The isolated blood vessels (B0) were treated with Liberase DL and DNase I (Figure S3), 34 followed by separation of PV-AEF-detached blood vessels (B1) and the crude PV-AEF by a 20-mm mesh filter.Immunofluorescence microscopic analysis revealed that the AQP4 signal in B1 was lower than that in B0 (Figure 3B), and western blot analysis showed that the AQP4 signal was mostly reduced in B1 (Figure 3C), suggesting that PV-AEF are detached from the isolated blood vessels.Indeed, in the crude PV-AEF, AQP4-positive and 4 0 ,6-diamidino-2-phenylindole (DAPI)-negative cell fragments and vesicles were observed (Figure 3D).To remove contaminated cells from the crude PV-AEF, Ficoll density gradient centrifugation was performed.Western blot analysis revealed that the AQP4 signal was observed in Fractions 4-11, whereas the lamin B1 nuclear marker signal was observed in Fraction 5 (Figure 3E).Because PV-AEF contain elongated mitochondria, 35 it was examined which fraction contained mitochondrial proteins.The VDAC1/2 mitochondria marker signal was observed in Fractions 6-10 (Figure 3E).SEM analysis also showed that BM structure was observed in B1 vessels (Figure S4A) and that Fractions 6-11 consisted of putative PV-AEF with various sizes ranging from 2.07 to 1,246.93 mm 2 (length of major axis: 1.69-56.95mm) (Figure 3F).Consistent with SEM analysis, various sizes of putative PV-AEF were also observed in transmission electron microscopic (TEM) analysis (Figures 3G and S4B).Presence of different sizes of putative PV-AEF observed in Fractions 6-11 was consistent with the different sizes of PV-AEF on brain blood vessels in situ as estimated by AQP4 immunostaining (Figure 3H) and in vivo as estimated by MLC1 immunostaining (Figure S4C).These results indicate that various sizes of PV-AEF are purified from the isolated brain blood vessels.

Mass spectrometry analysis of the purified PV-AEF
We used Fractions 6-11 obtained from the Ficoll density gradient centrifugation as the purified PV-AEF and analyzed by high-resolution mass spectrometry to elucidate PV-AEF components.The isolated blood vessels (B0) and the crude PV-AEF fraction were used for the comparison of the purified PV-AEF.Four independent replicates, each of which was used from pooled tissues of different mice, were prepared.Firstly, to evaluate the intrasample variation of proteins, four samples from the isolated blood vessels (B0), the crude PV-AEF, and the purified PV-AEF were analyzed.A total of 11,062 proteins were identified in B0, and 9,432 proteins (85.3% of all B0 proteins identified) were common in all four samples (Figure S5A).A total of 10,506 proteins in the crude PV-AEF and 8,600 proteins (81.9% of all crude PV-AEF proteins identified) were common in all four samples (Figure S5B).A total of 9,762 proteins in the purified PV-AEF and 7,732 proteins (79.2% of all purified PV-AEF proteins identified) were common in all four samples (Figure S5C).The correlation coefficient of log 2 intensity of identified proteins between two samples was higher than 0.95 (Figures S5A-S5C).Moreover, uniform manifold approximation and projection (UMAP) and hierarchical clustering analysis of log 2 intensity for the 12 samples revealed a clear separation of the purified PV-AEF from the crude PV-AEF and B0 (Figures S5D and S5E).Thus, these results were convincing and used for further analysis.A total of 11,420 proteins were identified among all samples (Figure 4A; Table S1); 9,762 proteins were identified in the purified PV-AEF, in which 9,353 proteins were common in all samples, 83 proteins were specific for the purified PV-AEF, and 188 proteins were observed in the purified PV-AEF and the crude PV-AEF (Figure 4A; Table S2).It was noted that these 83 proteins that were identified only in the purified PV-AEF were not identified in B0 or the crude PV-AEF, presumably because their concentrations were too low to be detected in B0 and the crude PV-AEF by the high-resolution mass spectrometry used here, indicating that these 83 proteins are highly enriched in the purified PV-AEF.Among 9,353 proteins that were identified to be common in all samples, 8,432 proteins showed significant differences in expression levels (ANOVA test, p < 0.05) (Table S3).Heatmap and hierarchical clustering analysis showed that these 8,432 proteins with significant differential expression proteins by ANOVA test were divided into five major clusters (Figure 4B).Cluster 4 and Cluster 5 had exclusively decreased proteins in the crude PV-AEF and the purified PV-AEF, and Cluster 1 also had decreased proteins in the purified PV-AEF, whereas Cluster 3 and Cluster 2 had exclusively increased proteins in the purified PV-AEF (Figure S6A).Gene ontology (GO) analysis revealed that Cluster 5 was majorly related to extracellular components (Figure 4C), and Cluster 4 was related to mRNA metabolism and gene expression (Figure 4D).Cluster 1 was related to post-translational modifications and transport (Figure S6B).Of note, Cluster 3 and Cluster 2 were related to mitochondria activity and translation processes (Figures 4E and S6C), consistent with the previous report showing that PV-AEF are enriched with mitochondria-related proteins. 36These results indicate that PV-AEF components are efficiently identified in this method.

Characterization of three CAMs enriched in the purified PV-AEF
Nectin-2 is an immunoglobulin (Ig)-like CAM with three Ig-like domains and consists of two splicing isoforms, nectin-2a and nectin-2d. 20We previously reported that nectin-2a was expressed in both cultured mouse neurons and astrocytes, whereas nectin-2d was selectively expressed in the cultured mouse astrocytes and localized at the boundary between PV-AEF and BM of blood vessels in the brain. 21Furthermore, genetic ablation of nectin-2 causes degeneration of PV-AEF and neurons in the cerebral cortex. 21Therefore, we firstly analyzed whether nectin-2 was identified in mass spectrometry analysis.Nectin-2 was identified in 9,353 proteins that were common in all samples (Figure 4A) and significantly enriched in the purified PV-AEF as nectin-2 belonged to Cluster 3 (Figures 4B and S7A).However, the localization of nectin-2d in different vascular regions has not been clarified in the previous study.To identify other novel CAMs of PV-AEF, we filtered the identified protein list by ''cell adhesion'' of GO biological processes.Three proteins were identified in 83 proteins specific for the purified PV-AEF (Figure 4A): IgLon5;Ntm, Itgb6, and Rgmb (Figure S7B).Single-cell RNA transcriptome analysis revealed that these three proteins were not preferentially expressed in astrocytes as analyzed by Betsholz dataset and Allen Brain Map.Nine proteins were identified in 188 proteins that were commonly observed in the purified PV-AEF and the crude PV-AEF, but not in B0 (Figure 4A): Bmp10, Cdh20, Cib1, Ephb1, Ephb1;Ephb2, Fzd1;Fzd2;Fzd7, Grid1;Grid2, Kirrel2, and Ptprt (Figures S7C and S7D).Among them, we focused on the Ig-like CAM Kirrel2, [25][26][27] because single-cell RNA transcriptome analysis revealed that Kirrel2 expression was predominantly observed in astrocytes (Figures S7E and S7F).8][39] Kirrel2 expression of astrocytes was upregulated during mouse development (Figure S7G).However, the role of Kirrel2 in astrocytes remains unclear.
We further analyzed commonly identified 9,353 proteins (Figure 4A).Using 8,432 proteins that were significant differential expression by ANOVA test (Figure 4B; Table S3), CAMs significantly upregulated in the purified PV-AEF (p < 0.05 and fold change >3) were searched, and 57 proteins were identified (Figures 5 and S8A).Among these 57 proteins, we focused on mucin-type glycoprotein podoplanin/aggrus/gp36/ E11, [28][29][30][31] because single-cell RNA transcriptome analysis revealed that Pdpn was also preferentially expressed in astrocytes (Figures S8B  and S8C).Pdpn was expressed in astrocytes during mouse development (Figure S8D).Podoplanin plays a crucial role in the development of the alveoli, heart, and lymphatic vascular system and the biology of immune cells through the interaction with the C-type lectin receptor CLEC-2. 31Although it was reported that cells expressing podoplanin are glial fibrillary acidic protein (GFAP)-positive astrocytes in the mouse brain, 40 the detailed localization of podoplanin, especially in PV-AEF, remains unclear.
In addition, mRNA expression of these three CAMs was significantly enriched in perivascular astrocytes, compared with non-perivascular astrocytes as analyzed by McCarty dataset (Figure S8E). 24In this study, we further examined the detailed localization of nectin-2d, Kirrel2, and podoplanin, especially in PV-AEF in the different vascular regions.

In vivo localization of nectin-2d, Kirrel2, and podoplanin in brain blood vessels
We examined the localization of nectin-2d, Kirrel2, and podoplanin in the mouse brain.The nectin-2d signal was partly observed in aSMA-and VWF-positive blood vessels and aSMA-negative and VWF-positive blood vessels (Figures 6A, 6B, S9A, and S9B).On the contrary, the nectin-2d signal was hardly observed in aSMA-and VWF-negative blood vessels (Figures 6A, 6B, S9A, and S9B).The nectin-2d signal co-localized with the GFAP astrocyte marker signal (Figure 6C), indicating that nectin-2d localizes at least in PV-AEF.The Kirrel2 signal was mostly observed in aSMA-and VWF-positive blood vessels and hardly observed in aSMA-negative and VWF-positive blood vessels or in aSMA-and VWF-negative blood vessels (Figures 7A, 7B, S10A, and S10B).The Kirrel2 signal co-localized with the GFAP signal (Figure 7C), indicating that Kirrel2 localizes at least in PV-AEF.The podoplanin signal was observed mostly in aSMA-and VWF-positive blood vessels and aSMA-negative and VWF-positive blood vessels and hardly observed in aSMA-and VWF-negative blood vessels (Figures 8A, 8B, S11A, and S11B).The  S3.
podoplanin signal co-localized with the GFAP signal (Figure 8C), indicating that podoplanin localizes at least in PV-AEF.These results indicate that nectin-2d and podoplanin localize mainly in arteries/arterioles and veins/venules, but hardly in capillaries, whereas Kirrel2 localizes mainly in arteries/arterioles, but hardly in capillaries or veins/venules.On the contrary, the AQP4 signal was observed in almost all the laminin a4-positive blood vessels (Figure S12).These results suggest that PV-AEF have different components depending on the different vascular regions.Localization of nectin-2d, Kirrel2, and podoplanin in large sizes of PV-AEF We next assessed nectin-2d, Kirrel2, and podoplanin expression in the purified PV-AEF.The clear nectin-2a/d signal was observed in part of the purified PV-AEF (Figure 9A), and the faint signal was observed in another part of the purified PV-AEF (Figure 9B).As the purified PV-AEF showed various sizes (Figures 3F, 3H by AQP4 immunostaining and defined as large and small sizes of PV-AEF samples.The mean fluorescence intensity of nectin-2a/d and AQP4 in large sizes of PV-AEF was significantly higher than that of small sizes of PV-AEF (Figures 9C and 9D).The clear Kirrel2 signal was observed in part of the purified PV-AEF (Figure 10A), and the faint signal was observed in another part of the purified PV-AEF (Figure 10B).The mean fluorescence intensity of Kirrel2 in large sizes of PV-AEF was significantly higher than that of small sizes of PV-AEF (Figure 10C).The clear podoplanin signal was observed in part of the purified PV-AEF (Figure 11A), and the faint signal was observed in another part of the purified PV-AEF (Figure 11B).The mean fluorescence intensity of podoplanin in large sizes of PV-AEF was significantly higher than that of small sizes of  PV-AEF (Figure 11C).These results were consistent with the in vivo observations that nectin-2d, Kirrel2, and podoplanin localized in arteries/ arterioles and/or veins/venules (Figures 6, 7, 8, and S9-S11), where the sizes of PV-AEF were larger than those of capillaries (Figures 3H and  S4C). 18Collectively, these results indicate that nectin-2a/d, Kirrel2, and podoplanin preferentially localize in large sizes of PV-AEF.

Adhesion activity of Kirrel2 in cultured astrocytes
Lastly, we evaluated the cell-cell adhesion activity of Kirrel2 in astrocytes because it has not been reported that Kirrel2 has cell adhesion activity in astrocytes.Primary astrocytes were prepared from dissociated neurospheres stimulated with 1% FBS. 21,41More than 95% of the cells expressed astrocyte markers, such as AQP4 and EAAT1 (Figure S13A), consistent with previous reports. 21,41,42As primary astrocytes formed the intricate structure of terminal processes, and endogenous expression levels of Kirrel2 were low, both membrane-anchored GAP-43 fused to GFP (GFP-mem) and Kirrel2 proteins were overexpressed in primary astrocytes (Figure S13B).The intense Kirrel2 signal was observed at the boundary between neighboring astrocytes (Figure 12A).The Kirrel2 signal co-localized with another CAM N-cadherin signal (Figure 12A).We next examined the cell adhesion activity of Kirrel2 using a cell aggregation assay.Aggregates of Kirrel2-and GFP-mem-transfected primary astrocytes were larger than those of control cells (Figures 12B and 12C).Collectively, these results indicate that Kirrel2 potentially has cell adhesion activity of cultured astrocytes at least in vitro.

DISCUSSION
In this study, we biochemically isolated brain blood vessels from mouse brain homogenates and isolated PV-AEF from the isolated blood vessels.Electron microscopic analysis revealed that the purified PV-AEF consisted of various sizes of cell fragments and vesicles.We performed mass spectrometry analysis on the purified PV-AEF and identified 9,762 proteins in the purified PV-AEF, in which mitochondria-related proteins were enriched, compared with the isolated blood vessels, whereas extracellular matrix proteins and transcription-related proteins were not.9][30][31] We identified Kirrel2 and podoplanin as novel PV-AEF molecules.Nectin-2d and podoplanin localized mainly in PV-AEF of arteries/arterioles and veins/ venules, but hardly in those of capillaries, whereas Kirrel2 localized mainly in PV-AEF of arteries/arterioles, but hardly in those of capillaries or veins/venules.We also found that nectin-2a/d, Kirrel2, and podoplanin were preferentially observed in large sizes of PV-AEF.These results revealed that PV-AEF had various sizes and different molecular components, implying different roles of PV-AEF in NVU depending on different vascular regions.Recently, another isolation method of PV-AEF from that shown in this study was reported and identified 516 proteins, including numerous critical electron transport chain proteins and BBB-related proteins. 36In this method, brain blood vessels were dissociated into a single-cell suspension by treatment with 100 mg/mL Liberase DL and 20 U/mL DNase I for 60 min, and PV-AEF were isolated using magnetic activation cell sorting (MACS) system with an anti-astrocyte cell surface antigen-2 antibody (Ab).Proteolytic digestion at these high concentrations and long incubation periods for PV-AEF dissociation might affect the degradation of cell surface proteins, and it is not well understood whether astrocyte cell surface antigen-2 is expressed in all PV-AEF, although the MACS system is a valuable way to isolate PV-AEF.To improve these two weaknesses of this method as much as possible, we dissociated PV-AEF from the isolated blood vessels with Liberase DL and DNase I at low concentrations (37.5 mg/mL Liberase DL and 2 U/mL DNase I) for 15 min incubation and isolated PV-AEF biochemically using Ficoll density gradient centrifugation.The present method is helpful for the isolation of various types and sizes of PV-AEF.Indeed, the known components of PV-AEF, such as AQP4, dystroglycan, syntrophin, dystrophin, dystrobrevin, glucose transporter 1, TRPV4, integrin a6, sideroflexin-5, 36 MLC1, [16][17][18] and GlialCAM, 19 were included in common 9,353 proteins.The purified PV-AEF contained different sizes of cell fragments and vesicles, consistent with the different sizes of PV-AEF on brain blood vessels in situ and in vivo.
Brain blood vessels are mostly wrapped with PV-AEF. 2,18Consistently, the AQP4 astrocyte marker signal was observed in almost all the laminin a4-positive blood vessels in vivo.It was recently shown that mature oligodendrocytes in gray matter directly associate with the vascular BM. 43 In this report, 17% of oligodendrocytes mainly contact with capillaries, but not with arterioles or venules, in the mouse cerebral cortex, hippocampus, and cerebellum.In this study, we did not observe the MBP oligodendrocyte marker signal on the isolated blood vessels, but our mass spectrometry analysis showed that oligodendrocyte marker proteins, such as Mag, Mog, and Cntn2, were detected in the purified PV-AEF fraction.Although the source of these oligodendrocyte proteins identified in the purified PV-AEF fraction was unclear, they may be derived from blood-vessel-associated oligodendrocytes.Further investigations are needed for making this conclusion.
Diphtheria-toxin-receptor-mediated ablation targeted to Mlc1-or Slc1a3-expressing astrocytes induces plasma protein leakage and downregulation of CAMs in VECs, leading to BBB breakdown. 44,45[6]8 Single astrocyte contacts at least one blood vessel, and most single astrocytes contact three blood vessels in the somatosensory cortex. 46In addition, some astrocytes are tightly associated with blood vessels through their somata, called juxtavascular astrocytes. 47However, the molecular components and morphology of PV-AEF in different vascular regions and brain regions have not been well understood.Recent vascular single-cell RNA transcriptome analysis revealed that the transcriptional profiles of VECs and mural cells are different depending on vascular regions in mice, 22 raising the possibility that the expression profiles of PV-AEF have regional differences depending on vascular regions.Consistently, we observed the different regional localization of nectin-2d, Kirrel2, and podoplanin in mouse brain blood vessels.In addition, we observed the different sizes of PV-AEF on brain blood vessels in situ and in vivo, and the sizes of PV-AEF in large blood vessels seem to be larger than those in small blood vessels.To support this observation, the expression levels of Kirrel2, which were observed in arteries/arterioles, and those of nectin-2d and podoplanin, which were observed in arteries/arterioles and veins/venules, were higher in large sizes of the purified PV-AEF.These present results are in good agreement with the earlier observations in the mouse brain that the sizes of PV-AEF increase with vessel diameter and that this size variation is most pronounced in arteries/arterioles rather than veins/venules, although the heterogeneity of PV-AEF in molecular components was not shown. 18In this previous report, the sizes of PV-AEF in the mouse brain were analyzed in vivo using an anti-MLC1 Ab, because MLC1 localizes at the boundary between neighboring PV-AEF in mouse, rat, and human brain tissues. 17,18The role of this size variation of PV-AEF remains unclear, but computational modeling simulations suggest that it is involved in relatively constant flux between perivascular and interstitial compartments. 18It was previously shown that proteins are synthesized in PV-AEF, sustaining their structural and functional polarization. 34Therefore, it could be speculated that each endfoot process of highly ramified astrocytes has specific functions and morphology depending on vascular regions.In addition, different brain regions have different vascular permeability and disease susceptibility in the mouse brain, 48 suggesting that there may be differences in PV-AEF depending on brain regions.Thus, the isolation of PV-AEF or astrocytes by MACS system with an anti-nectin-2d, anti-Kirrel2, or anti-podoplanin Ab and comparison of the expression profiles might provide further insight into the different roles of PV-AEF and heterogeneity of astrocytes.The precise molecular components of PV-AEF are essential for our understanding of the regulatory mechanisms of BBB in NVU by astrocytes.Furthermore, protein expression profiles of transporters and receptors in brain microvessels in human are different from those in mice, 49 indicating that BBB properties are different among species. 49,50Comparison of PV-AEF components across species is necessary for elucidating the species differences of BBB.Mutations of PV-AEF molecules, such as MLC1 and GlialCAM, cause a human developmental disease, called megalencephalic leukoencephalopathy with subcortical cyst, in which gliovascular functions are dysregulated. 51,524][5][6] We focused here on three CAMs, nectin-2d, Kirrel2, and podoplanin, among identified proteins in the purified PV-AEF.Expression of these three CAMs is also enriched in perivascular astrocytes. 24Nectin-2d, also known as CD112 or PVR-related 2 (PVRL2/PRR2), is a Ca 2+ -independent CAM. 20We previously showed that nectin-2d localizes at the boundary between PV-AEF and BM of blood vessels in the mouse brain and that genetic ablation of nectin-2 causes degeneration of PV-AEF. 21At one month of age of nectin-2deficient mice, the electron density of the cytoplasm in PV-AEF is increased, and the plasma membrane of PV-AEF protrudes toward the abluminal side.At six months of age, nectin-2-deficient mice show the atrophy of white and gray matters and the enlargement of the lateral ventricles, although the detailed molecular mechanisms of these phenotypes remain unclear.The present study revealed that nectin-2d localized mainly in PV-AEF of arteries/arterioles and veins/venules.Because BM components differ depending on vascular regions, 22,53 these different molecular components of BM may regulate the localization of nectin-2d.Genome-wide association study revealed that the human NECTIN-2 gene is associated with Alzheimer's disease. 20,54Comparison of PV-AEF components between nectin-2-deficient and wild-type mice may provide insight into the mechanisms for the degeneration of PV-AEF and pathological aspect of Alzheimer's disease.
8][39] In the developing mouse cerebellum and spinal cord, Kirrel2 is expressed in early postmitotic neural precursors and GABAergic progenitor-selective cells and regulates neuronal differentiation. 55,56However, the role of Kirrel2 in astrocytes remains unclear.We showed here that Kirrel2 localized mainly in PV-AEF of arteries/arterioles, but hardly in those of capillaries or veins/venules, in the mouse brain, and that Kirrel2 potentially had cell adhesion activity of cultured astrocytes at least in vitro.Because electron microscopic 3D reconstitution analysis using the rat brain revealed that PV-AEF cover most of the surface of brain blood vessels and attach to each other, 35 Kirrel2 may localize at the boundary between neighboring PV-AEF of arteries/arterioles and play a role in adhesion between neighboring PV-AEF in vivo.Consistently, Kirrel2 is also expressed in the glomerular podocytes, which cover the glomerular capillaries in the kidney. 57Kirrel2 localizes at the boundary between neighboring podocyte foot processes, called slit diaphragms, which contribute to the glomerular filtration barrier. 57Considering these previous observations, it could be speculated that Kirrel2 localizing at the boundary between neighboring PV-AEF of arteries/arterioles may contribute to the regulation of the BBB permeability and influx and efflux transport.Of note, in cerebral amyloid angiopathy, deposition of amyloid, such as amyloid b protein, is observed in the tunica media and adventitia of the arteries and arterioles in the human cerebral cortex, which may lead to neurodegenerative diseases through vascular dysfunction. 58,59It would be interesting if Kirrel2 localizing at the boundary between neighboring PV-AEF of arteries/arterioles is involved in cerebral amyloid angiopathy.Further studies are needed to clarify the role of Kirrel2 at the boundary between neighboring PV-AEF of arteries/arterioles.
Podoplanin is a mucin-type glycoprotein and has multiple physiological functions associated with lymphangiogenesis, platelet aggregation, immune response, and oncogenesis. 31Podoplanin contributes to the formation of membrane-actin structures through the interaction with ezrin/radixin/moesin proteins to activate RhoA signaling. 31,60Interestingly, podoplanin plays an important role in maintaining the normal morphology of podocyte foot processes and glomerular permeability in the kidney. 31,61In puromycin aminonucleoside nephrosis, a rat model of human minimal change nephropathy, the extensive flattening of podocyte foot processes and severe proteinuria are observed in association with the high reduction of podoplanin expression. 28Treatment of anti-podoplanin neutralization Ab in rat leads to proteinuria in association with extensive flattening of foot processes. 61It could be speculated that podoplanin localizing in PV-AEF of arteries/arterioles and veins/venules regulates the different morphology of PV-AEF, which controls the BBB function in NVU, although we did not reveal here whether it localized at the boundary between PV-AEF and BM, between neighboring PV-AEF, or both, or at the intracellular organellar membrane of PV-AEF.Recently, podoplanin expression correlates with microglial activation, including cell mobility and phagocytosis, after traumatic brain injury in a mouse model, 62 raising the possibility that upregulation of podoplanin by some pathological conditions may affect astrocyte morphology and mobility, leading to the induction of activated astrocytes.Because half of Pdpn-deficient mice die in the first week, 31 a conditional knockout technology is useful for clarifying the role of podoplanin in the morphology of PV-AEF and regulation of BBB.

Limitations of the study
In this study, because we dissociated PV-AEF from the isolated blood vessels with Liberase DL and DNase I, proteolytic digestion might affect the degradation of PV-AEF components.Because we isolated PV-AEF biochemically using Ficoll density gradient centrifugation, cell fragments and vesicles derived from blood-vessel-associated cells, including fibroblasts, oligodendrocytes, oligodendrocyte precursor cells, and microglia, could not be completely separated from PV-AEF.The heterogeneity of PV-AEF in different brain regions remains to be pursued.We analyzed the PV-AEF components of C57BL/6J mouse, but not other species.Comparative studies on PV-AEF components among various species should be performed in future because BBB properties are different among species.It also remains unclear whether the heterogeneity of PV-AEF is observed in other species, including humans.In this study, we showed that Kirrel2 potentially had cell adhesion activity of cultured astrocytes at least in vitro.However, the direct evidence for the adhesion between neighboring PV-AEF in vivo remains to be determined.We attempted to perform the cell aggregation assay using the purified PV-AEF, but the purified PV-AEF did not aggregate with each other in our experimental conditions, although they may contain many CAMs.The exact reason for the inability of the purified PV-AEF to aggregate with each other is not known but may be because the adhesion activity of many CAMs other than cadherins is very weak.Furthermore, the function and molecular mechanisms of the identified molecules enriched in PV-AEF remain to be pursued, because it is crucially overnight, washed and then immunostained with secondary Abs and 1 mg/ml DAPI (Nacalai Tesque) in the PBS(-) at room temperature for 2 h.The samples were mounted in FluorSave reagent (Merck Millipore).Images were captured with a confocal laser-scanning microscope (LSM 510 META, CarlZeiss) or a fluorescence microscope (BZ-X710, Keyence).Abs used in this study were shown in Table S4.For biotinylation of anti-VWF Ab, Ab-10 Rapid Biotin Labeling Kit (Dojindo) was used according to the manufacturer's protocol.The sizes of PV-AEF and the fluorescence intensity were measured by Fiji software (version 2.3.0,ImageJ2). 63For quantification of the ratio of PV-AEF protein signal-positive blood vessels, multiple hippocampus and cerebral cortex images were obtained and the number of brain blood vessels was counted manually, and their percentages were calculated.

Figure 1 .
Figure 1. Isolation of blood vessels from the mouse brain (A) Outline procedure for the isolation of blood vessels from mouse brain homogenate.S3 and S5 were brain-blood-vessel-enriched fractions.Representative bright-field images of S3 and S5 fraction (right).S3 fraction contained large sizes of blood vessels compared with S5 fraction.Scale bars, 100 mm.(B) Immunoblot analysis of each fraction with the indicated Abs; 5 mg of total protein was loaded in each lane.CD31, VEC marker; Laminin, BM marker; aSMA, smooth muscle cell marker; and PDGFRb, pericyte marker.As for laminin, laminin b1 and g1 chains were detected.These images are representative of three independent experiments.See also Figure S1.

Figure 2 .
Figure 2. Characterization of the isolated brain blood vessels (A and B) Immunofluorescence images of the isolated blood vessels (mixture of S3 and S5 fractions) immunostained with the indicated Abs.Scale bars, 50 mm in (A) and 10 mm in (B).A, artery/arteriole; Cap, capillary; and V, vein/venule.(C) SEM analysis of the isolated blood vessels.These images are representative of two capillaries.Scale bars, 1 mm.These images are representative of three independent experiments.See also Figure S2.

Figure 3 .
Figure 3. Isolation of the purified PV-AEF (A) Outline procedure for the isolation of the purified PV-AEF.B0, the isolated blood vessels (mixture of S3 and S5 fractions).B1, trapped vessels that detached PV-AEF.(B) Immunofluorescence images of B0 and B1 fractions immunostained with the indicated Abs.Scale bars, 50 mm.(C) Immunoblot analysis of B0 and B1 fractions with the indicated Abs.As for laminin, laminin b1 and g1 chains were detected.(D) Immunofluorescence image of flow-through fraction.Arrowheads, PV-AEF identified by AQP4-positive and DAPI-negative cell fragments and vesicles.*, contaminated cells.Scale bar, 20 mm.(E) Density gradient fractionation of flow-through fraction.Equal volumes of each fraction were loaded on SDS-PAGE gels, and membranes were blotted with the indicated Abs.Input, flow-through fraction.In this experiment, AQP4 oligomer signal was observed.(F) SEM analysis of the purified PV-AEF.Scale bar, 20 mm.The area of the purified PV-AEF was shown in the violin plot (n = 270) (right).The red dot indicates average (84.1 mm 2 ), and blue vertical lines indicate quartiles (median, 34.6 mm 2 ).(G) TEM analysis of the purified PV-AEF.Scale bar, 1 mm.(H) Immunofluorescence images of the isolated mouse brain vessels immunostained with the anti-AQP4 Ab. a, b: enlargement of area outlined at left.a: large size of blood vessel (⌀, 12 mm); b: small size of blood vessels (⌀, 5 mm).Scale bars, 20 mm.These images are representative of three independent experiments.See also Figures S3 and S4.

Figure 4 .
Figure 4. Mass spectrometry analysis of the purified PV-AEF (A) A Venn diagram of the numbers of identified proteins of each sample.B0, mixture of S3 and S5 fractions.(B) Heatmap and hierarchical clustering of significant differential expression of 8,432 proteins (ANOVA test).The color scale reflects the average of log 2 -fold changes in protein expression, compared with B0 (n = 4).The complete linkage method was used for hierarchical clustering.Crude, the crude PV-AEF; Purified, the purified PV-AEF.(C-E) Gene ontology enrichment analysis of biological processes for the indicated clusters.The top 10 enriched GO Biological Processes 2021 are shown.See also Figures S5-S7, Tables S1, S2, and S3.

Figure 5 .
Figure 5. Mass spectrometry analysis of 57 CAMs Heatmap shows significantly upregulated 57 CAMs.The color scale reflects the average of log 2 -fold changes in protein expression compared with B0 (n = 4).See also FigureS8and TableS3.

Figure 7 .
Figure 7. Localization of Kirrel2 in the mouse hippocampus (A) Localization of Kirrel2 in stratum lacunosum-moleculare of the hippocampus.a, artery/arteriole; b, vein/venule; and c, capillary.Arrowheads, arteries/ arterioles; arrows, veins/venules; *, capillaries.Scale bars, 20 mm.(B) The ratio of Kirrel2-positive blood vessels in the hippocampus is shown as a bar graph.(C) Co-localization of Kirrel2 and GFAP at artery/arteriole in stratum lacunosum-moleculare of the hippocampus.Scale bars, 20 mm.Line scan of fluorescence intensity of the white line (a-b).AU, arbitrary units.These images are representative of three independent experiments.See also Figures S10 and S12.

Figure 8 .
Figure 8.Localization of podoplanin in the mouse hippocampus (A) Localization of podoplanin (Pod) in stratum lacunosum-moleculare of the hippocampus.a, arteries/arterioles; b, veins/venules; and c, capillaries.Arrowheads, arteries/arterioles; arrows, veins/venules; *, capillaries.Scale bars, 20 mm.(B) The ratio of podoplanin-positive blood vessels in the hippocampus is shown as a bar graph.(C) Co-localization of podoplanin and GFAP at artery/arteriole in stratum lacunosum-moleculare of the hippocampus.Scale bar, 20 mm.Line scan of fluorescence intensity of the white line (a-b).AU, arbitrary units.These images are representative of three independent experiments.See also Figures S11 and S12.
Figure 8.Localization of podoplanin in the mouse hippocampus (A) Localization of podoplanin (Pod) in stratum lacunosum-moleculare of the hippocampus.a, arteries/arterioles; b, veins/venules; and c, capillaries.Arrowheads, arteries/arterioles; arrows, veins/venules; *, capillaries.Scale bars, 20 mm.(B) The ratio of podoplanin-positive blood vessels in the hippocampus is shown as a bar graph.(C) Co-localization of podoplanin and GFAP at artery/arteriole in stratum lacunosum-moleculare of the hippocampus.Scale bar, 20 mm.Line scan of fluorescence intensity of the white line (a-b).AU, arbitrary units.These images are representative of three independent experiments.See also Figures S11 and S12.

Figure 9 .
Figure 9. Localization of nectin-2a/d in the purified PV-AEF (A and B) Immunofluorescence images of the purified PV-AEF immunostained with the indicated Abs.The anti-nectin-2 Ab, which recognizes nectin-2a/d, was used.The purified PV-AEF were identified by AQP4-positive and DAPI-negative cell fragments and vesicles.The areas encircled by the cyan dotted line indicate the areas of the purified PV-AEF.Scale bars, 5 mm.(C and D) Mean fluorescence intensity of nectin-2 and AQP4 was shown in the bar graphs.The area of PV-AEF was measured by the AQP4 signal and divided into two groups according to the median value.Data are presented as mean and SD (n = 28, 28), and dot shows each data.The Mann-Whitney U test was performed.These images are representative of three independent experiments.

Figure 10 .
Figure 10.Localization of Kirrel2 in the purified PV-AEF (A and B) Immunofluorescence images of the purified PV-AEF immunostained with the indicated Abs.The purified PV-AEF were identified by AQP4-positive and DAPI-negative cell fragments and vesicles.The areas encircled by the cyan dotted line indicate the areas of the purified PV-AEF.Scale bars, 5 mm.(C) Mean fluorescence intensity of Kirrel2 was shown in the bar graphs.The area of PV-AEF was measured by the AQP4 signal and divided into two groups according to the median value.Data are presented as mean and SD (n = 28, 28), and dot shows each data.The Mann-Whitney U test was performed.These images are representative of three independent experiments.

Figure 11 .
Figure 11.Localization of podoplanin in the purified PV-AEF (A and B) Immunofluorescence images of the purified PV-AEF immunostained with the indicated Abs.The purified PV-AEF were identified by AQP4-positive and DAPI-negative cell fragments and vesicles.The areas encircled by the cyan dotted line indicate the areas of the purified PV-AEF.Scale bars, 5 mm.(C) Mean fluorescence intensity of podoplanin was shown in the bar graphs.The area of PV-AEF was measured by the AQP4 signal and divided into two groups according to the median value.Data are presented as mean and SD (n = 31, 30), and dot shows each data.The Mann-Whitney U test was performed.These images are representative of three independent experiments.

Figure 12 .
Figure 12.Localization of Kirrel2 in cultured astrocytes (A) Immunofluorescence images of astrocytes transfected with Kirrel2 and GFP-mem immunostained with the indicated Abs.N-Cad, N-cadherin.Arrowheads, boundary between neighboring cultured astrocytes.Scale bars, 10 mm.(B and C) Cell aggregation assay of astrocytes transfected with Kirrel2 and GFP-mem.Scale bars, 100 mm.The aggregation index was calculated by dividing the number of GFP-positive aggregated cells by the total number of GFP-positive cells.Data are presented as mean and SD (n = 5), and dot shows each data.The Mann-Whitney U test was performed.These images are representative of three independent experiments.See also Figure S13.