Abstract

   graphic file with name ao1c03248_0008.jpg

   Endothelial cell damage is an important pathological basis for the
   deterioration of acute ischemia stroke. Our previous studies have been
   exploring the mechanism of blood–brain barrier (BBB) endothelial cell
   injury in the early stage of cerebral ischemia. Exosomes act as an
   important intercellular player in neurovascular communication. However,
   the characteristic of exosomes derived from BBB endothelial cells in
   early ischemic stroke is poorly understood. We exposed cultured brain
   microvascular endothelial cells (bEnd.3) to 3 h oxygen glucose
   deprivation (OGD) to mimic early cerebral ischemia in vitro and
   compared miRome and surface protein contents of exosomes derived from
   bEnd.3 cells by miRNA sequencing and the proximity barcoding assay
   (PBA). A total of 346 differentially miRNA (159 upregulated and 187
   downregulated) were identified via miRNA-Seq in bEnd.3 cells after
   exposure to OGD for 3 h. Moreover, Gene Ontology (GO) and KEGG pathway
   analyses showed that cell proliferation- and angiogenesis-associated
   miRNAs were significantly affected. The abnormal changes in top eight
   miRNAs were further verified by a quantitative polymerase chain
   reaction (qPCR). PBA experiments showed that the numbers of exosomes
   carrying the following proteins increased significantly under ischemia,
   including bFGF, CD146, EPHA2, ABCB5, and ITGB2. These proteins were
   related to angiogenesis, cell proliferation, and cell inflammation. The
   network analysis combining PBA data with miRNA-Seq data showed that 79
   miRNAs were related to 24 membrane proteins and predicted that there
   were surface proteins associated with a variety of miRNA molecules,
   such as ITGA9, XIAP, ADAM1, ITGA2, ITGA3, PDPN, and ITGB1. Meanwhile,
   there were miRNAs related to various surface proteins including
   miR-410-3p, miR-378b, and miR-1960. Taken together, our data
   demonstrated for the first time the changes of exosomal miRNAs and
   surface protein profiles derived from ischemic microvascular
   endothelial cells, which may provide new therapeutic targets for BBB
   protection in ischemic stroke.

1. Introduction

   Acute ischemic stroke is a serious threat to human health with high
   morbidity and high mortality.^[40]1 Tissue-type plasminogen activator
   (tPA) is the only approved drug for ischemic stroke by the US Food and
   Drug Administration (FDA). Severe cerebral edema and dangerous cerebral
   hemorrhage are the two potentially fatal neurovascular complications in
   tPA thrombolytic therapy, of which blood–brain barrier (BBB) injury is
   the common pathological basis.^[41]2 Brain capillary endothelial cells
   are one of the key structural components of the BBB, which could not
   only maintain the normal barrier function of the BBB but also sense
   ischemic injury.^[42]3 Dysfunction of the BBB endothelial cells has
   been extensively studied, which leads to the high permeability of the
   BBB, vasogenic edema, and vasospasm and thus aggravates ischemic brain
   damage.^[43]2 However, how the BBB endothelial cells sense ischemic
   injury and interact with surrounding cells consisting of the
   neurovascular unit, such as pericytes, astrocytes, and neurons, remains
   to be explored.

   Exosomes are small membrane-coated extracellular vesicles with a
   diameter of 30–100 nm secreted by a variety of cells. They contain
   various molecular constituents including protein and RNAs from maternal
   cells and can act as a medium bridging intercellular
   communication.^[44]4 Accumulating evidence has shown that exosomes can
   provide a novel way to alleviate ischemic brain damage through
   improving angiogenesis, inhibiting cell apoptosis, and decreasing
   inflammation.^[45]5 Exosomes derived from both neuronal and
   non-neuronal cells could improve neuron growth and promote axon
   regeneration of injured neurons.^[46]6 Neural stem cell-derived
   exosomes can improve nerve function damage and reduce brain edema after
   stroke.^[47]7 Astrocyte-derived exosomes inhibit neuronal apoptosis and
   glial scar formation and improve nerve plasticity and nerve
   function.^[48]8 In previous studies, ischemia-induced intercellular
   exosome communication is mostly concentrated on neurons and astrocytes;
   however, there are few reports on the characteristic profile of
   exosomes in the brain microvascular endothelial cells after early
   ischemia and exosome-mediated signal transmission between endothelial
   cells and other cells in the brain.

   MiRNAs are emerging as key players in the pathogenesis of stroke and
   have been implicated in the regulation of cell death,^[49]9 cell
   metabolism,^[50]10 inflammatory,^[51]11 and angiogenic^[52]12
   processes. Growing therapeutic potential of specific miRNAs, such as
   miR-15a/16-1,^[53]12 miR-126,^[54]13 miR-223,^[55]14 and
   miR-124,^[56]15 has been identified in the treatment of stroke. Exosome
   surface membranes proteins derived from parent cells reflect the
   physiological, pathological, and functional status of the cells from
   which they are derived and can be used as disease biomarkers in the
   future. However, these exosome surface protein markers may not be
   detectable because of their low abundance. Recently, Wu et al. reported
   a new single exosome analysis technique with high sensitivity and good
   specificity that can identify the surface protein composition of a
   single exosome through antibody DNA conjugates and next-generation
   sequencing and thus expand the heterogeneity research of
   exosomes.^[57]16

   In this study, we used in vitro oxygen glucose deprivation (OGD) models
   to mimic ischemia and investigated the contents in miRNAs and membrane
   surface proteins associated with exosomes purified from conditioned
   media (CM) of ischemic bEnd.3 cells using the miRNA sequence and
   proximity barcoding assay (PBA), respectively, and analyzed the
   correlation between miRNA and surface proteins. This is a study to
   identify the characteristics of exosomes derived from brain
   microvascular endothelial cells in early ischemia, moreover, to conduct
   a preliminary integrative analysis between exosome miRNA and membrane
   surface proteins in the ischemic condition.

2. Results

2.1. Characterization of Exosomes Derived from Mice Brain Microvascular
Endothelial Cells

   To determine the properties of brain microvascular endothelial
   cell-derived exosomes under the ischemic condition, we assessed the
   characterization of exosomes isolated from bEnd.3 cells exposed to 3 h
   OGD. We used ultracentrifugation or ExoQuick-TC total exosomes kits to
   isolate exosomes from cell media of bEnd.3 cells. Exosomes were
   identified by transmission electron microscopy (TEM), nanoparticle
   tracking analysis (NTA), and western blotting. First, we demonstrated
   that bEnd.3 cells generated exosome-enriched particles with a mean
   diameter of 118.4 ± 9.4 nm in the control group versus 132.6 ± 10.1 nm
   in the OGD group ([58]Figure [59]1
   A) using a dynamic light scattering analysis system (NTA-based exosome
   detection system). The TEM image showed that exosomes were present as
   small membrane vesicles with the typical nanoparticle size range
   ([60]Figure [61]1B). In addition, the exosomes were found to express
   exosomal markers including CD9, CD63, and Alix by western blotting
   ([62]Figure [63]1C). These results indicated that the particles
   isolated from CM of bEnd.3 cells were verified to be exosomes.

Figure 1.

   [64]Figure 1
   [65]Open in a new tab

   Characteristics of exosomes derived from brain microvascular
   endothelial cells after OGD. BEnd.3 cells were subjected to 3 h OGD
   before exosome isolation. Size distribution of exosomes of the control
   group (A) and OGD-treated group (B) analyzed with the NTA. (C)
   Representative TEM images of exosomes. Scale bar = 100 nm. (D) Positive
   markers for exosomes-CD63, CD81, and Alix as detected by western blot
   analysis. OGD, oxygen glucose deprivation; NTA, nanoparticle tracking
   analysis; TEM, transmission electron microscopy; CD63, cluster of
   differentiation 63; CD81, cluster of differentiation 81; Alix,
   apoptosis-linked gene 2-interacting protein X.

2.2. Quantitative Exosome miRNA Profiling

   To study the effect of exposure to OGD on the noncoding RNA profile of
   exosomes derived from bEnd.3 cells, we conducted NGS experiments with
   triplicate biological replicates extracted from OGD-treated bEnd.3
   cells and control cells. The extracted RNAs sized between 17 and 70 nt
   were used to construct small RNA cDNA libraries. The libraries were
   sequenced on the Illumina HiSeq. 2500 platform. By comparing the
   filtered reads with the authoritative miRNA/rRNA/tRNA/snRNA/snoRNA
   database, we obtained functional categorization profiling of the miRNA
   sequences, and the miRNAs of the control group and the OGD group were
   7.33 ± 2.30 and 6.00 ± 1.00%, respectively ([66]Figure [67]2
   A). The majority of reads ≥19 nt corresponded to mature miRNAs. The
   obtained miRNA sequences were compared with publicly available data on
   the miRDeep2 and miRBase V21 database. For each miRNA sequence-based
   profile, the expression level of miRNAs was estimated by the number of
   sequence reads. After quartile normalization, we identified 159
   upregulated and 187 downregulated miRNAs (|log 2(Fold Change)| ≥ 1 and
   P < 0.05) ([68]Figure [69]2B,C). The top ten significantly
   differentially expressed miRNAs between control and OGD groups are
   listed in [70]Table [71]1. Consequently, the target gene prediction and
   enrichment analysis of differential miRNAs between control
   bEnd.3-derived and OGD-treated bEnd.3-derived exosomes were performed.
   To characterize the predominant pathways, KEGG pathway enrichment
   analysis was executed to compare the specific miRNA targets to the
   whole reference gene background. Thirty signal pathways showing the
   most statistical difference were selected in our KEGG analysis data
   ([72]Figure [73]2D). Among these 30 pathways, metabolic pathways,
   PI3K-Akt signaling pathway, mitogen-activated protein kinase (MAPK)
   signaling pathway, Wnt signaling pathway, focal adhesion, mammalian
   target of rapamycin (mTOR), forkhead box O (FoxO), endocytosis, calcium
   signaling pathway, and protein cyclic guanosine monophosphate-kinase G
   (cGMP-PKG) signaling pathways may contribute to endothelial cell
   survival, cell metabolism, and exosomal transportation.

Figure 2.

   [74]Figure 2
   [75]Open in a new tab

   Exosomal miRNA sequencing profiling for bEnd.3 cells in response to 3 h
   OGD treatment. (A) Histogram representation of the distribution of
   noncoding RNA categories in exosomes derived from the control group and
   OGD-treated group. (B) Heatmap revealing the differentially expressed
   miRNAs in three replicates of the control and OGD-treated bEnd.3 cells.
   Each row represents the relative expression levels of a single miRNA
   across all samples. Red blocks represent high expression relative to
   the control cells, and blue blocks represent low relative expression.
   (C) Scatter plot showing the differences of miRNA expression between
   the control group and OGD-treated group. Red dots indicated upregulated
   miRNAs, blue dots indicated downregulated miRNAs, and gray dots
   indicated no change. (D) Top 30 pathways of different miRNA target
   genes that were identified by KEGG pathway analysis.

Table 1. Top Ten Significantly Differentially Expressed miRNAs between
bEnd.3-Derived and OGD bEnd.3-Derived Exosomes.

      miRNA_ID     up/down log 2 (fold change)    P-value
   mmu-miR-412-5p  down         –12.5054       1.44 × 10^–63
   mmu-miR-122-5p  down          –9.1688       2.63 × 10^–57
   mmu-miR-369-3p  down         –10.1621       8.20 × 10^–51
   mmu-miR-127-3p  down          –7.5257       4.96 × 10^–44
   mmu-miR-409-3p  down          –8.4086       1.33 × 10^–43
   mmu-miR-379-5p  down          –9.9317       6.07 × 10^–43
   mmu-miR-494-3p  down          –9.4927       2.12 × 10^–40
   mmu-miR-214-3p  down         –14.8316       4.00 × 10^–40
   mmu-miR-299b-3p down         –14.7051       1.43 × 10^–38
   mmu-miR-8095    down         –14.8813       9.00 × 10^–38
   [76]Open in a new tab

   Next, we used the GO approach to analyze the enrichment of candidate
   target genes. The top ten significant terms of three GO categories,
   including biological process, cellular component, and molecular
   function, were presented ([77]Figure [78]3
   A). Based on the correlation between miRNA and mRNA, we performed
   coexpression network analysis to screen the important miRNAs involved
   in endothelial cells injury induced by OGD. The result of network
   analysis indicated that miRNAs with high degrees were miR-379-5p,
   miR-122-5p, miR-494-3p, miR-412-5p, miR-369-3p, miR-409-3p, miR-214-3p,
   and miR-127-3p ([79]Figure [80]3B). These top eight miRNAs were all
   downregulated in bEnd.3 cells subjected to OGD treatment, and of note,
   miR-412-5p and miR-379-5p were located in the core position of the
   network. Moreover, we found that miR-412-5p and miR-379-5p shared a
   common target gene Slco3a1, and miR-494-3p and miR-369-3p had a common
   target gene Elovl5 ([81]Figure [82]3B).

Figure 3.

   [83]Figure 3
   [84]Open in a new tab

   Gene Ontology and miRNA regulatory network analysis. (A) GO enrichment
   analysis of predicted targets genes to the differentially expressed
   miRNAs in bEnd.3 cells after OGD treatment. Green dots indicated
   biological process, red dots indicated cellular component, and blue
   dots indicated molecular function. (B) Interaction network of top eight
   miRNAs and target genes involved in ischemia-induced endothelial cells
   injury.

   To validate the inspection result, we performed a quantitative
   real-time polymerase chain reaction (qRT-PCR), and the data showed that
   these miRNAs were indeed downregulated in the OGD group compared with
   the control group ([85]Figure [86]4
   ). These results indicated that the miRNAs enriched in the exosomes of
   brain microvascular endothelial cells may participate in the modulation
   of cell survival, proliferation, and angiogenesis under ischemic
   conditions.

Figure 4.

   [87]Figure 4
   [88]Open in a new tab

   qRT-PCR validation of exosomal miRNA sequencing data. Endothelial cells
   were treated with OGD for 3 h and total RNA was extracted for qRT-PCR
   to analyze the expression of miR-379-5p (A), miR-122-5p (B), miR-494-3p
   (C), miR-412-5p (D), miR-369-3p (E), miR-409-3p (F), miR-214-3p (G),
   and miR-127-3p (H). The results were normalized to the U6 values for
   each sample, and fold changes were shown as means ± standard deviation
   (SD) in three independent experiments (P < 0.05). *P < 0.05 from
   one-way analysis of variance (ANOVA) followed by Tukey’s multiple
   comparison test for pairwise comparisons.

2.3. Profiling of Surface Proteins of Exosomes by PBA

   Special membrane proteins on the surface of exosomes can be used as
   specific markers for the separation of exosomes; however, their
   potential significance as a disease biomarker is ignorable in view of
   the limitations of single exosomal feature detection. Therefore, we
   applied PBA to investigate the profiling of cell surface proteins of
   bEnd.3 cell-derived exosomes and examined whether exosomes from
   OGD-treated cells have distinct surface protein patterns compared with
   nonischemic cells. To test this, we performed dimensional reduction of
   individual exosomes from two groups based on their surface protein
   compositions using T-distributed Stochastic Neighbor Embedding (t-SNE)
   analysis.^[89]17 The results showed that OGD treatment significantly
   changed the distribution of exosomal surface proteins compared with the
   nonischemic control group ([90]Figure [91]5
   A). As shown in [92]Figure [93]5B, there were 11751 exosomes expressing
   only one surface protein in the control group, while there were 25820
   such exosomes (∼1.2-fold increase) in the OGD group. We also compared
   the numbers of the exosomes expressing two surface proteins and three
   surface proteins between the OGD-treated group and the control group,
   and the results are shown in [94]Figure [95]5B. There were 6518
   exosomes expressing two proteins and 858 exosomes expressing three
   proteins recorded in the OGD-treated group compared with 1872
   two-protein exosomes and 223 three-protein exosomes recorded in the
   control group. These data indicated that OGD significantly increased
   the number of exosomes secreted by bEND3 cells. By comparing the
   surface protein profiles of the two groups, the pyramid diagrams of the
   U1 pattern (one-protein exosome), U2 pattern (two-protein exosome), and
   U3 pattern (three-protein exosome) are presented, as shown in
   [96]Figure [97]5C–E. For one-protein exosomes, significant differences
   were observed in 21 proteins between the control group and OGD group,
   including UPA, TROP2, SURVIVIN, N-CADHERIN, MUC4, MUC1, MCP-1, LMP1,
   ITGB1, ITGB2, ITGAX, ITGAM, ITGAE, ITGA2B, ITGA2, IL-8, EGF, CD146,
   ANNEXIN, ALDH1, and ABCB5 ([98]Figure [99]5C). As shown in [100]Figure
   [101]5D, there were 28 combinations in the U2 pattern, for which the
   numbers of two-protein exosomes in the OGD group were significantly
   higher than those in the control group. There were five combinations in
   the U3 pattern showing significant difference between the OGD group and
   the control group, CD146, FGFBASIC, ITGB2 (P = 0.01952); CD146,
   FGFBASIC, EPHA2 (P = 0.02313); CD146, CD340, CD166 (P = 0.03024);
   ABCB5, FGFBASIC, ITGB2 (P = 0.03024); and ABCB5, FGFBASIC, EPHA2 (P =
   0.00011) ([102]Figure [103]5E). Of note, the exosomes with more surface
   proteins identified were better distinguished, and these differential
   surface proteins were related to cell proliferation and angiogenesis,
   which might act as exosomal markers for endothelial cells challenged by
   ischemia.

Figure 5.

   [104]Figure 5
   [105]Open in a new tab

   Surface protein profiling of individual exosomes derived from bEnd.3
   cells after OGD. (A) t-SNE plots of all samples splitted and
   autoclustered by flowSOM (each sample down-sampled to the size of
   10 000), which visualized the marker intensity of individual exosomes
   with one identified protein type. (B) Comparison of exosome numbers in
   two groups. U1, exosomes with one protein; U2, exosomes with two
   proteins; U3, exosomes with three proteins. Pyramid diagram showing the
   numbers of one-protein exosomes (C), two-protein exosomes (D), and
   three-protein exosomes (E). Red columns indicate the control group, and
   blue columns indicate the OGD group. Filter out records with 0 exosome.
   Representative P values were marked in the right column.

2.4. Correlation between Exosomes miRNAs and Surface Proteins

   The exosomes have been shown to regulate cell growth, cell migration,
   and cell-to-cell communication through the cargos they carry including
   proteins, mRNA, miRNA, and other signal molecules; however, it remains
   unknown whether there is a potential correlation between miRNAs and
   surface proteins in exosomes. To investigate the relationship between
   miRNAs and membrane proteins of exosomes derived from OGD-treated
   endothelial cells, we performed integrated analyses. As shown in
   [106]Figure [107]6
   , there were 79 miRNAs closely related to 24 membrane proteins.
   Combining with miRNA sequencing data analysis (especially the top eight
   miRNAs), we found that miR-379-5p was associated with ITGAX and
   miR-494-3p and miR-409-3p were related to ADAM10. There were several
   other surface proteins associated with a variety of miRNA molecules,
   such as ITGA9, XIAP, ADAM1, ITGA2, ITGA3, PDPN, and ITGB1 in
   OGD-treated endothelial cells. Meanwhile, several miRNAs were relevant
   to surface proteins including miR-410-3p, miR-378b, and miR-1960. These
   results clearly indicated that OGD treatment enhanced the correlation
   between miRNA expression and surface proteins of exosomes derived from
   brain microvascular cells, in which miRNAs might regulate the
   distribution of exosomal surface proteins, while the surface proteins
   possibly contributed to the miRNA sorting process.

Figure 6.

   [108]Figure 6
   [109]Open in a new tab

   Integrated miRome and surface proteomic analyses of the exosomes
   secreted by OGD-treated endothelial cells. The predicted
   miRNA–gene–surface protein network was shown for OGD-treated bEnd.3
   cells. Red octagons indicated upregulated miRNAs, green octagons
   indicated downregulated miRNAs, blue squares indicated surface
   proteins, and blue triangles indicated genes.

3. Discussion

   The BBB is composed of endothelial cells, pericytes, astrocytes, tight
   junctions, and the basement membrane.^[110]18 Accumulating evidence
   shows that injury of BBB endothelial cells is a major cause of
   postischemic reperfusion-associated hemorrhagic complications.^[111]19
   Our previous study has shown that endothelial cell injury plays an
   important role in BBB dysfunction in the early stage of
   ischemia.^[112]20 However, the exosomes derived from endothelial cells
   involved in the cellular response to OGD have not been fully addressed.
   In this study, we performed the first investigation of mapping the
   miRNA expression profile and surface protein profile of the exosomes
   derived from ischemic brain microvascular endothelial cells using
   integrated miRNome and surface membrane proteomic analyses. The major
   findings are as follows: (1) OGD promotes the release of exosomes in
   endothelial cells, and significant differences of exosomal miRNome and
   surface proteomics are observed between OGD-treated cells and the
   control cells; (2) the exosomal miRNAs and surface proteins identified
   in OGD-treated endothelial cells are closely associated with
   angiogenesis, cell proliferation, and cell inflammation; and (3) there
   is a clear correlation between miRNAs and surface proteins identified
   in the exosomes isolated from ischemic brain microvascular endothelial
   cells. These data indicate that the exosomal miRNAs and surface
   proteins may serve as new biomarkers for brain microvascular
   endothelial cell damage in ischemic stroke.

   The exosomes are involved in the transmission of information between
   BBB cell components and the regulation of BBB integrity by delivering
   functional proteins, mRNA transcription products, and miRNAs to
   recipient cells.^[113]21 There is evidence that exosomes can act as
   sensors for neuroinflammation.^[114]22 Gao et al. found that
   endothelial colony-forming cell (ECFC)-generated exosomes played a
   positive role in BBB integrity in mice after traumatic brain
   injury.^[115]23 Exosomes are also secreted and transferred between glia
   and neurons in the nervous system, exerting extensive effects on nerve
   activation, development, and regeneration.^[116]24 More recently,
   miR-132 was reported to be transported from neurons to cerebral
   vascular endothelial cells by releasing exosomes and regulating the
   integrity of cerebral blood vessels.^[117]25 In this study, we used the
   nanoparticle tracking method to investigate whether OGD affects exosome
   size and concentration. Although we did not find significant
   differences in the size and concentration of exosomes between control
   and OGD-treated endothelial cells, the numbers of exosomes were
   significantly increased after OGD treatment, which is consistent with a
   previous report in which the vascular endothelial cells showed
   increased exosome secretion in response to hypoxia.^[118]26 Using
   quantitative approaches to analyze exosomal miRNA contents, we further
   demonstrated that the exosomes derived from OGD-treated bEnd.3 cells
   were clearly distinguishable from the exosomes isolated from control
   cells. We further analyzed the eight miRNAs that were most
   significantly affected by OGD treatment and found that these miRNAs
   regulate a variety of functions, including cell proliferation and
   migration (miRNA-122-5p and miR-409-3p),^[119]27 vascular structure
   formation (miR-412-5p),^[120]28 nerve fiber regeneration
   (miR-379-5p),^[121]29 cell apoptosis (miR-494-3p),^[122]30 and cell
   adhesion (miR-127-3p).^[123]31

   The exosomes from different sources can be distinguished by identifying
   their surface protein combinations. Haqqani et al. found that the
   exosomes derived from human brain microvascular endothelial cells
   contained at least 1179 proteins, including 60 marker proteins, such as
   Alix, TSG101, CD81, and CD9.^[124]32 Currently, the exosomal proteins
   are analyzed as a whole, while few studies have been carried out to
   analyze the characteristics of the surface proteins in a single
   exosome, in which the heterogeneity of exosomes would be largely
   ignored. In our study, the type and number of surface proteins on every
   single exosome were resolved by the PBA technique, which helped to
   classify exosomes into different subgroups. By comparing the surface
   protein profiles identified in the exosomes between control and
   OGD-exposed exosomes, we observed significant increases in a bunch of
   surface proteins in the exosomes derived from OGD-treated endothelial
   cells, such as FGFBASIC, CD146, EPHA2, ABCB5, and ITGB2. These proteins
   have been shown to promote angiogenesis, such as FGFBASIC,^[125]33
   CD146,^[126]34 SURVIVIN,^[127]35 TROP2,^[128]36 and EGF,^[129]37 cell
   proliferation, such as EPHA2^[130]38 and N-CADHERIN,^[131]39 or
   inflammation regulation, such as MCP-1^[132]40 and IL-8.^[133]41 Our
   data that OGD treatment induces an increase in the numbers of exosomes
   carrying proangiogenic surface proteins, such as CD146, FGFBASIC, and
   ITGB2 or CD146, FGFBASIC, and EPHA2 suggest that OGD may activate the
   endothelial cells a compensatory mechanism to promote the establishment
   of microcirculation through cell-to-cell communication in endothelial
   cells in response to ischemic stimuli. Our data also raise an important
   possibility that the specific surface protein combinations identified
   in exosomes may serve as biomarkers for evaluating the ischemic status
   of brain microvascular endothelial cells.

   Emerging evidence has shown that the miRNAs are not randomly integrated
   into exosomes but rather through a sorting mechanism to guide specific
   miRNAs into exosomes.^[134]42 The sorting mechanism for exosomal miRNA
   is associated with proteins that act as an assembly protein to
   participate in exosomal miRNA composition, such as AGO2 and
   GW182.^[135]43 In other words, there are certain proteins that
   deliberatively control the sorting process of exosomal miRNA. Here, our
   data show that OGD enhances the association of a variety of miRNA
   molecules with surface proteins identified in the exosomes derived
   brain endothelial cells, such as miR-410-3p (associated with ITGA2 and
   ADAM1), miR-378b (associated with ITGA2 and CD36), and miR-1960
   (associated with ITGB8 and ITGA9). However, in view of the limitations
   of the PBA technique, which can only be qualitative but not
   quantitative, it is uncertain that the presence of membrane surface
   proteins might promote the aggregation of miRNAs or it needs to reach a
   certain expression threshold to achieve this effect, which needs to be
   further investigated in future studies. In turn, it also remains to be
   determined whether the exosomal miRNAs contribute to the localization
   of specific proteins to the exosomal membrane in future studies.

4. Conclusions

   In conclusion, our present study has demonstrated that ischemia in
   vitro promotes the release of exosomes and the changes of exosomal
   miRNA and surface protein profiles in brain microvascular endothelial
   cells, and these changes may represent a compensatory mechanism to
   promote angiogenesis, cell proliferation, and inflammatory regulation.
   Moreover, there is a clear correlation between these exosomal miRNAs
   and surface proteins. Screening and identifying the types and
   characteristics of the exosomes secreted by ischemic endothelial cells
   may provide a new way to evaluate the functional status of
   cerebrovascular endothelial cells in ischemic stroke.

5. Experimental Section

5.1. Cell Culture

   Mouse brain microvascular endothelial cells (bEnd.3, American Type
   Culture Collection, Manassas, VA) were cultured in 75 cm^2 flasks. The
   bEnd.3 cells were grown in Dulbecco’s modified Eagle’s medium (DMEM)
   (Gibco, Carlsbad, CA) containing 10% fetal bovine serum (FBS) (Gibco,
   Carlsbad, CA) and 1% penicillin/streptomycin at 37 °C in a humidified
   incubator gassed with 5% CO[2]/95% air. The cells were subcultured and
   allowed to grow to confluence before exposure to OGD treatment.

5.2. Oxygen Glucose Deprivation Model

   To mimic the ischemic condition in vitro, the bEnd.3 cells were exposed
   to OGD as described previously.^[136]20 In brief, bEnd.3 were seeded on
   Nunc cell culture dishes, then incubated in glucose-free medium (Gibco,
   Carlsbad, CA), and subsequently moved into a humidified airtight
   chamber (Billups-Rothberg Inc., Del Mar, CA) equilibrated with 95% N[2]
   and 5% CO[2] for 15 min. The chamber was then sealed and kept at 37 °C
   for 2 h. The oxygen concentration was controlled below 0.2% by an
   oxygen analyzer (Sable Systems, Las Vegas, NV). The medium of OGD
   treatment was FBS-free, and the medium of the control group was
   exosome-depleted FBS (Obio Technology Corp. Ltd., Shanghai, CN). After
   OGD treatment, the conditioned medium was collected to isolate exosomes
   for further analysis.

5.3. Exosome Isolation from Cell Culture Supernatants

   Exosomes were isolated from bEnd.3 culture supernatants by
   ultracentrifugation or the Total Exosome Isolation (from cell culture
   media) kit (Thermo Fisher Scientific, Waltham, MA) according to the
   manufacturer’s protocol. In brief, conditioned media (CM) were
   collected from 80 to 90% confluent bEnd.3 in sterile conditions and
   filtered using a filter unit (Millipore, Billerica, MA) with a 0.22 μm
   membrane to remove intact cells and debris. The CM was then
   ultracentrifuged at 120 000g for 2 h at 4 °C using an ultrahigh-speed
   centrifuge (Beckman, Brea, CA). The resulting pellets were washed with
   cold phosphate-buffered saline (PBS) and then ultracentrifuged one more
   time as above. Finally, the collected pellets were resuspended in PBS
   and immediately stored at −80 °C until subsequent usage.

5.4. Nanoparticle Tracking Analysis

   Nanoparticle tracking analysis (NTA) was performed by a NanoSight LM 10
   instrument (Malvern Instruments Ltd., U.K.) following the
   manufacturer’s instructions. Briefly, 1 mL of sample was drawn and then
   slowly injected into the sample pool, \the laser module was put back on
   the base, and then the probe of the OMEGA thermometer was put into the
   copper hole of the laser module panel. The focal lengths of the
   appropriate Screen Gain and Camera Level were adjusted. The
   measurements were performed by selecting the appropriate measurement
   method and measurement parameters. NTA acquisition settings were
   optimized between sample readings.

5.5. Transmission Electron Microscopy

   The collected exosome pellet was diluted in 30 μL of PBS and kept at
   −20 °C until transmission electron microscopy (TEM) analysis. Each 10
   μL droplet of aspirate samples was placed on the copper mesh, allowed
   to settle for 1 min, and then rinsed by PBS. Phosphotungstic acid (10
   μL) on the copper mesh was absorbed and precipitated for 1 min. The
   floating liquid was absorbed by filter paper and kept at room
   temperature for 2 min. The preparation was examined using a
   transmission electron microscope (JEM-1200EX, Japan Electronics
   Corporation, Japan).

5.6. Western Blotting

   Total proteins were extracted from the cultured cells using RIPA lysis
   buffer. The protein concentration was detected by the BCA Protein Assay
   reagent kit (Thermo Fisher Scientific, Waltham, MA). The cell protein
   samples were separated by sodium dodecyl sulfate-polyacrylamide gel
   electrophoresis (SDS-PAGE) in 10% polyacrylamide gels and transferred
   to polyvinylidene difluoride membranes. Then, nonspecific binding was
   blocked by 5% nonfat milk in trisbuffered saline with Tween-20 (P9416;
   Sigma-Aldrich) for 1 h at room temperature. Membranes were incubated
   with primary antibodies, including anti-CD63 (ab108950; Abcam),
   anti-Alix (ab117600; Abcam), and anti-CD81 (ab79559; Abcam), for
   overnight at 4 °C. Membranes were incubated with horseradish peroxidase
   (HRP)-conjugated secondary antibodies for 1 h at 25 °C before
   incubating with the enhanced chemiluminescence (ECL) reagent
   (Thermo-Pierce, Rockford, IL). The protein levels were quantitated
   after normalization to β-actin.

5.7. Exosome miRNA Sequencing and Construction of Libraries

   Total RNA from exosomes was used for miRNA sequencing. Library
   preparation and miRNA sequencing were performed by Ribobio (Guangzhou,
   China). Briefly, the total RNA of the sample or the purified sRNA
   fragments that were connected the 3′ end and the 5′ end successively
   were then reverse-transcribed to cDNA for PCR amplification. Afterward,
   the target fragment library was recovered by gel cutting, the qualified
   library was sequenced using the Illumina HiSeq. 2500 platform, and only
   small RNAs ranging from 18 to 30 nt were used for library construction.

5.8. Real-Time Quantitative PCR Analysis

   To validate the miRNA sequencing data, quantitative real-time
   polymerase chain reaction (qRT-PCR) analysis was performed using RNA
   obtained from the control group (n = 3) and the OGD group (n = 3).
   Total RNA was purified from bEnd.3 cells with Trizol reagent
   (Invitrogen). Real-time PCR was performed using the TaqMan miRNA assay
   kit protocol on an ABI7500 fast real-time PCR machine. The expression
   levels of miRNAs were measured in terms of the threshold cycle value
   and normalized to the U6 level as an internal control. The specific
   primers were purchased from GeneCopoeia Biotechnology (Guangzhou,
   China). The primer information is listed as follows: miR-379-5p (Cat
   Number: MmiRQP0476), miR-122-5p (Cat Number: MmiRQP0056), miR-494-3p
   (Cat Number: MmiRQP0536), miR-412-5p (Cat Number: MmiRQP3186),
   miR-369-3p (Cat Number: MmiRQP0454), miR-409-3p (Cat Number:
   MmiRQP0484), miR-214-3p (Cat Number: MmiRQP0320), and miR-127-3p (Cat
   Number: MmiRQP0111).

5.9. Proximity Barcoding Assay

   Profiling surface proteins on individual exosomes were detected by the
   proximity barcoding assay (PBA) as described previously.^[137]16
   Briefly, 240 cell surface protein antibodies conjugated with DNA
   sequence coding tags (antibody tags) were used to recognize all exosome
   surface proteins in CM samples. The same single exosome-specific DNA
   coding sequence (exosome tag) was added to all antibody tags on the
   same exosome through the proximity coding reaction. All encoded
   sequences are amplified by PCR to obtain the exosome tag–antibody tag
   DNA sequence, and then, the sample tag–exosome tag–antibody tag DNA
   sequence was obtained using the sample tag (sample tag) primer to
   continue the amplification. According to the Illumina sequencing
   technology requirements, the DNA fragment library was established, DNA
   testing was performed with Illumina NextSeq. 500 and Illumina NextSeq
   FC-404-2005 kits, and the read DNA sequence information was the sample
   tag–exosome tag–antibody tag DNA sequence. The composition of
   individual exosome surface proteins was identified and quantified;
   thereby, the classification of all individual exosomes was analyzed.

5.10. Statistical Analysis

   The data of statistical analysis was analyzed using SPSS 20.0 software.
   All values were presented as mean ± standard error (SEM). Differences
   between groups were compared by either an unpaired Student’s t-test or
   one-way ANOVA followed by Tukey’s posthoc test for P values. The
   numbers of exosomes with given combinations of proteins were calculated
   using the Fisher test at a confidence level of 0.05, and related
   operations were implemented using R script. The t-SNE algorithm used
   was included in the R package Rtsne. The network of miRNAs and surface
   proteins was constructed using Cytoscape software (version 3.8.0). P <
   0.05 was considered to be statistically significant.

Acknowledgments