Expression of FUT7 and sLeX was positively associated with BLCA metastasis
It is known that fucosyltransferase VII (FUT7) is the key rate-limiting enzyme for sLeX synthesis (Fig. 1A) [27]. Using the TCGA database, we observed that FUT7 expression was elevated in various tumor types with high neoplasm histologic grades (Fig. S1A), and was associated with advanced clinical T stages (Fig. 1B), correlating with poor survival outcomes in BLCA patients (Fig. 1C). TMA analysis confirmed that FUT7 expression was significantly higher in BLCA tissues compared to normal adjacent tissues (Fig. 1D; Fig. S1B, C).
Consistent with abnormal FUT7 expression, higher level of sLeX were observed in BLCA tissues compared to normal tissues (Fig. 1E; Fig. S1B, D). The serum from BLCA patients also presented elevated level of sLeX compared to healthy controls (Fig. S1E). Moreover, high expression of FUT7 and elevated level of sLeX were closely correlated with lymph node metastasis, lymphovascular metastasis (Fig. 1F, G; Fig. S1F-H) and were indication of poor survival outcomes for BLCA patients (Fig. 1H, I).
SLeX promoted the malignant progression of BLCA
FUT7 expression and sLeX level were elevated in muscle-invasive BLCA cell lines (J82, T24 and YTS-1) compared to non-muscle-invasive BLCA cell lines (RT-4, KK47 and 5637) and human bladder epithelial cell lines (HCV29) (Fig. S2A-C). Introduction of the FUT7 gene into KK47 cells (referred to as oeFUT7) resulted in a significant increase in sLeX level (Fig. 2A), and enhanced the migration and invasion capabilities of the cells (Fig. 2B, C). Additionally, high level of sLeX was observed to stimulate cell colony formation and proliferation in KK47 cells (Fig. 2D, E). On the other hand, silencing of FUT7 expression in YTS-1 cells (referred to as shFUT7) significantly decreased sLeX level (Fig. 2A), and suppressed cell migration (Fig. 2B, C), colony formation and proliferation (Fig. 2F, G). However, no significant alteration of apoptosis was noted in either shFUT7 or oeFUT7 (Fig. S2D).
The mouse xenograft model results showed that tumorigenic rate was significantly higher in the oeFUT7 group compared to the KK47 group (Fig. S2E). In the footpad model, mice injected with oeFUT7 cells demonstrated significantly increased metastasis to the popliteal LNs compared to those injected with KK47 cells (Fig. S2F, G). The oeFUT7 group displayed larger weight and size of the footpad tumors and popliteal LNs (Fig. S2H). These in vitro and in vivo findings indicate that high level of sLeX promote bladder tumor growth and invasion.
BLCA derived-sEVs increased vascular permeability in vivo and in vitro
The metastasis of malignant carcinoma is largely driven by increased permeability of vascular endothelial cells (ECs) in tumor microenvironment. Our observation showed that the conditioned medium (CM) from YTS-1 cells significantly compromised the integrity of monolayer endothelial cell (Fig. S3A and B). Recognizing that sEVs play a crucial role as components of CM from malignant tumors and facilitate metastasis by disrupting the vascular barriers [8, 26, 28], we next isolated the sEVs from HCV29, KK47 and YTS-1 cells, and confirmed their sphere-like morphology (Fig. 3A), a narrow size distribution around 100 nm (Fig. 3B) and robust expression of sEV markers such as CD63, Alix, and Tsg101 (Fig. 3C).
Trans-endothelial migration assay showed that a higher number of GFP-labeled KK47 cells migrated through sEVsYTS−1 treated HUVEC monolayers compared to sEVsKK47 or sEVsHCV29 treated group (Fig. 3D, E). Similarly, HUVEC monolayers treated with sEVsYTS−1 showed increased permeability to rhodamine (Fig. 3D, F) and exhibited enhanced tube formation compared to the other two groups (Fig. S3C). It is known that the permeability of endothelial layer mainly depends on cell-cell junctions [29]. We then examined the transmembrane adhesive proteins in aforementioned HUVECs. TJ-associated proteins ZO-1, occludin and claudin-5 were significantly downregulated in the CM from YTS-1 and in the sEVsYTS−1-treated groups, but not in the other groups, while adhesion junction (AJ)-associated proteins such as E-Cadherin, N-Cadherin, and β-catenin did not show a clear difference in all groups (Fig. 3G; S3D and E). In contrast, the treatment with GW4869, a widely-utilized sEVs biogenesis inhibitor, restored TJ proteins expression in CM treated-HUVECs (Fig. S3F). In vivo mouse study revealed that injection with sEVsserum from BLCA patients resulted in the disruption of vascular barrier, indicated as the leakage of FITC-dextran out of blood vessels and more fluorescence in lung section, compared to the injection with sEVsserum from healthy controls (Fig. 3H, I). Not surprisingly, sEVsYTS−1 showed higher level of sLeX than sEVsHCV29 and sEVsKK47 (Fig. 3J). Elevated level of sLeX was also presented in sEVs derived from the serum of BLCA patients (Fig. 3K).
SLeX modified-vesicles were efficiently uptaken by HUVECs and destroyed vascular endothelial barriers
To assess the impact of sLeX modification on sEVs, we isolated sEVs from KK47 and oeFUT7 cells (referred to as sEVsKK47 and sEVsoeFUT7, respectively). sEVsoeFUT7 presented a higher level of sLeX modification compared to sEVsKK47 (Fig. 4A; Fig. S4A). Moreover, sEVsoeFUT7 demonstrated more efficient internalization by HUVECs (Fig. 4B). Given that sLeX acts as a ligand for E-selectin on inflammatory endothelial cells [30], we pre-treated the HUVECs with the E-selectin inhibitor, A-205804 (Fig. S4B), resulting in reduced endocytosis of sEVsoeFUT7 (Fig. 4C).
The expression of TJ proteins, including ZO-1, occludin, and claudin-5, were diminished in HUVECs treated with CMoeFUT7 or sEVsoeFUT7 compared to those treated with CMKK47 or sEVsKK47 (Fig. 4D; Fig. S4C, D). The permeability of HUVECs increased to both rhodamine and GFP+ KK47 cells in the sEVsoeFUT7-pretreated group compared to the sEVsKK47-pretreated group (Fig. 4E, F). Conversely, TJ proteins were elevated, and permeability reduced in HUVECs treated with CMshFUT7 or sEVsshFUT7 compared to those treated with CMYTS−1 or sEVsYTS−1 (Fig. S4C; Fig. 4G-I). Tube formation was increased in HUVECs treated with CMoeFUT7 (compared to CMKK47), but abolished in HUVECs treated with CMshFUT7 (compared to CMYTS−1) (Fig. S4E). Furthermore, Spearman correlation analysis demonstrated that FUT7 expression was positively correlated with angiogenesis pathway (Fig. S4F).
The above results were further validated by the in vivo oeFUT7-injected mouse footpad model. Compared to the control group, oeFUT7-injected group showed lower density of vessel integrity, indicated by occludin and claudin-5, and higher density of neovascularization, indicated by VEGF and CD31, in both the peritumoral and intratumoral regions (Fig. 4J, K). This suggests that high sLeX expression significantly disrupts vascular integrity and induces neovascular formation to facilitate tumor metastasis.
Vesicular ITGA3 was identified as a target protein modified by sLeX
It is known that sLeX are terminal epitopes of N/O-linked glycans on proteins and of glycolipids. Using proteomic analysis, we identified 3,654 differentially expressed proteins including 299 downregulated and 483 upregulated proteins, in response to altered sLeX modification in KK47 Vs. oeFUT7 cells (Fig. 5A; Fig. S5A; Table S2). The upregulated proteins were primarily enriched in cellular processes such as mitotic vasculogenesis, tight junction assembly, cell-cell adhesion, and integrin binding (Fig. 5B; Table S3), indicating that sLeX modification was closely associated with cell-cell adhesions. Further antibody based enrichment and mass spectrum (MS)-based glycoproteomic analysis identified integrin α3 (referred to as ITGA3) as a sLeX-bearing glycoprotein (Fig. 5C, D; Fig. S5B; Table S4).
Immunoprecipitation (IP) confirmed the enhanced sLeX modification on ITGA3 in oeFUT7 cells and sEVsoeFUT7 (Fig. 5E, F). A correlation was observed between higher ITGA3 expression and higher sLeX modification level (Fig. 5A; Fig. S5C, D). Immunofluorescence (IF) staining also revealed a marked raise of both sLeX and ITGA3 in oeFUT7 cells, but not in KK47 cells (Fig. S5E). Spearman correlation analysis further illuminated a positive correlation between FUT7 and ITGA3 expression (Fig. S5F), suggesting that sLeX modification regulates ITGA3 expression at the protein level. Inhibiting protein synthesis with cycloheximide (CHX) resulted in the slower degradation of ITGA3 in oeFUT7 relative to KK47 cells (Fig. S5G, H). Further study revealed that sLeX modification stabilizes ITGA3 by inhibiting its degradation via the lysosomal pathway (Fig. S5I, J).
ITGA3 promoted vascular permeability and angiogenesis
The IHC results of The Human Protein Altas (HPA) database showed that BLCA patients with high expression of ITGA3 account for approximately 75% of all BLCA patients (Fig. 6A; Fig. S6A; Table S5). ITGA3 in serum and serum-sEVs were clearly enhanced from BLCA patients than those from healthy individuals (Fig. 6B, C). Using TMA analysis, we found significant increases in ITGA3 expression in BLCA tissues than paired normal tissues (Fig. 6D, E; Fig. S6B, C). ITGA3 expression was also associated with lymphovascular metastasis, lymph node metastasis and poor prognosis in BLCA patients (Fig. 6F-H). Similarly, Spearman correlation analysis showed that the expression of ITGA3 was positively correlated with the angiogenesis in BLCA (Fig. 6I). Based on these findings, we hypothesize that sEVs enter HUVECs and disrupt vascular endothelial barriers through the highly sLeX-modified ITGA3.
As mentioned above, GO analysis indicated the correlation of ITGA3 expression with tight junction and angiogenesis. Spearman correlation analysis revealed that ITGA3 expression was positively correlated with VEGFA (vascular endothelial growth factor A), whereas negatively correlated with TJ-related proteins, including claudin-5, occludin, ESAM and JAM3 (Fig. 7A). When ITGA3 gene was stably overexpressed in HUVECs (referred to as oeITGA3), the proteins involved in the MAPK pathway in endothelial permeability [31, 32], including phosphorylated c-Jun and p38 were increased in oeITGA3 cells, along with reduced TJ protein expression and increased endothelial permeability (Fig. 7B, C). Consistent with these findings, incubation with CMoeITGA3 promoted tube formation of HUVECs (Fig. S6D). These results demonstrate that ITGA3 can disrupt vascular endothelial barriers and induce angiogenesis by activating MAPK signaling.
Silencing ITGA3 reversed sLeX induced tumorigenesis and metastasis
After endocytosis, NHS-biotin-labelled vesicular ITGA3 of oeFUT7 can be recycled by HUVECs (Fig. 7C, D). The silence of ITGA3 in oeFUT7 cells (referred to as oe-shITGA3) resulted in the decreased ITGA3 expression in sEVs (Fig. S6E, F), but did not affect the endocytosis of sEVs into endothelial cells (Fig. S6G). The permeability of HUVECs was dramatically reduced when incubated with CMoe−shITGA3 or sEVsoe−shITGA3 compared with CMoeFUT7 or sEVsoeFUT7 (Fig. S6H, I; Fig. 7G, H), accompanied by the restoration of the expression of ZO-1, occludin and claudin-5 (Fig. 7F). The MAPK-associated proteins were activated in sEVsoeFUT7-treated HUVECs, but deactivated in sEVsoe−shITGA3-incubated HUVECs (Fig. 7F). Additionally, less tube formation was observed in HUVECs incubated with CMoe−shITGA3 than CMoeFUT7 (Fig. S6I).
Using popliteal LN metastasis model, we found that the high incidence of tumorigenic and LN metastasis events in oeFUT7 injected mice was significantly reversed by ITGA3 knockdown (Fig. S6K-M). ITGA3-knockdown significantly increased the expression of occludin and claudin-5, and decreased the density of VEGF and CD31-indicated neovascular in both peritumoral and intratumoral regions of primary footpad tumor (Fig. 7I, J). These findings collectively demonstrate that the silence of ITGA3 reversed the tumor growth, metastasis, and angiogenesis induced by sLeX-modified sEVs, and restored the paracancer endothelium barrier in vivo.
SLeX-modified ITGA3 induced occludin translocation into nuclear
As illustrated in Fig. 5D, asparagine 265 (Asn265) was identified as the primary sLeX-modified site of ITGA3. The site-specific mutant replacing Asn265 by Ala on ITGA3 (referred to as N265A-HA) showed the sharply reduced level of sLeX modification (Fig. 8A). The inhibition of TJ-related proteins was observed in ITGA3-HA HUVECs, accompanied by activation of the MAPK signaling pathway. However, this suppressive effect was significantly reversed in N265A-HA cells, concomitant with the inactivation of the MAPK signaling pathway (Fig. 8B, C). Similar to wild-type HUVECs, the permeability of HUVECs to rhodamine decreased in the N265A-HA group compared to the ITGA3-HA group (Fig. 8D). Additionally, we noted a reduction in the ability of tube formation in the N265A-HA group compared to the ITGA3-HA group (Fig. S7A).
Mechanically, Co-immunoprecipitation (Co-IP) coupled with tandem LC–MS/MS analysis compared the differential proteins interacting with ITGA3 in ITGA3-HA and N265A-HA cells. Of the 12 proteins presenting inhibited interaction with ITGA3 in N265A-HA group, PP1A was screened out; notably, PP1A has been shown to regulate the tight junctions between endothelial cells by dephosphorylating occludin at Ser/Thr residues [33] (Fig. 8E, F). Co-IP assays showed that sLeX-modified ITGA3 can assist occludin to interact with PP1A and result in dephosphorylation of occludin at Ser/Thr residues, whereas this effect was reversed in N265A-HA (Fig. 8G, H). Consistent with the observation that phosphorylation state of occludin at Ser/Thr residues can affect its subcellular localization [33–36], we found that more occludin entered the cytoplasm and nuclear in ITGA3-HA, while occludin predominantly remained at the cell membrane to maintain TJ between endothelial cells in N265A-HA (Fig. 8I, J).
Further analysis, integrated with Co-IP and LC–MS/MS, revealed that occludin interacted with three karyopherins, namely Importin α1, Exportin 1 and Exportin 2, which function as adapter proteins in nuclear protein import or export processes (Fig. 8K). As occludin contains a nuclear localization signal (NLS) (Fig. 8L), we confirmed significant interaction between occludin and Importin α1 in the ITGA3-HA group, whereas this interaction was absent in the N265A-HA group (Fig. 8M). In contrast to its predominant membrane localization in N265A-HA, occludin was primarily distributed in the nuclear region through interaction with Importin α1 in the ITGA3-HA group (Fig. 8N), suggesting that the sLeX-modified ITGA3 facilitates the nuclear translocation of occludin, possibly in an Importin α1-dependent manner.