A total of 84 ESCC patients (44–79 years old, mean age of 62.02 ± 9.49 years) were recruited from The Second Hospital of Hebei Medical University. Among them, 42 cases were in stage I, 28 cases in stage II, and 14 cases in stage III. All patients underwent tumor resection before chemotherapy. ESCC tissues and adjacent normal tissues were collected and preserved at −80°C. All patients were pathologically diagnosed as ESCC, and no radiotherapy or chemotherapy before surgery was performed. Patients who died from diseases unrelated to ESCC during the follow-up were excluded (He et al., 2020).

Human normal esophageal epithelial cells (HEEC) and ESCC cell lines EC9706, Eca109, TE-13, TE-1 and TTN were provided by China Center for Type Culture Collection (Wuhan, China). Cells (3 × 10⁵ cells/well) with 80% confluence were subjected to transfection using Lipofectamine 2000 (Invitrogen, CA, United States). Overexpressed (oe)-VEGFA, small interfering RNA (si)-LINC00662, oe-LINC00662, miR-195-5p mimic and their corresponding negative control (NC) were all from GenePharma (Shanghai, China) (Li Z et al., 2019).

ESCC cells were placed in Roswell Park Memorial Institute (RPMI) 1640 medium containing EVs-free 10% fetal bovine serum (FBS) and cultured for 3 d. Then, the cell supernatant was collected and centrifuged to remove cell debris. The EVs were extracted using Hieff™ Quick exosome isolation kit (41201ES50, YEASEN, Shanghai, China). The supernatant and exosome separation reagent were added into the Eppendorf (EP) tube with a proportion of 2:1 overnight, and then centrifuged at 100,00 g for 1–2 h and the precipitate was resuspended in phosphate buffered saline (PBS). The EV suspension (30 μl) together with an equal volume of radioimmunoprecipitation assay (RIPA) lysis buffer was continuously dissolved in microwave (2 times, 10 s/time). Then, the EVs were purified at 4°C at 12,000 × g for 2 min, and the supernatant was retained and stored at −80°C. Protein quantification of EV was performed by bicinchoninic acid (BCA) kit (Beyotime, Nantong, China).

EVs identification: Transmission electron microscope (TEM; JEM-1010, JEOL, Tokyo, Japan) was adopted to observe EV morphology, tunable resistive pulse sensing (TRPS) to examine concentration and diameter of EVs, and Western blot analysis to test EV markers (CD81 and CD9) and non-marker protein (Calnexin and GM130) (Li Z et al., 2019).

EVs uptake was monitored using PKH26 fluorescent labeling kit (Sigma-Aldrich, CA, United States). EV sample was diluted in diluent C, and then PKH26 dye was added and kept at room temperature. After 4 min incubation avoiding light exposure, the reaction was neutralized by 5% bovine serum albumin. Then, EVs, resuspended in PBS were centrifuged at 100,000 g for 70 min, and incubated with HUVECs for 4 h. Subsequently, HUVECs (ATCC, VA, United States) were fixed with 4% paraformaldehyde, labeled with 4′,6-diamidino-2-phenylindole (0.5 μg/ml; Invitrogen) and imaged under a confocal microscope (Carl Zeiss, Oberkochen, Germany) (Zhang et al., 2021).

When HUVECs reached 50–60% confluence, they were incubated with dextran 498 (Dx498, 100 μg/ml) for 2 h to label cell membrane lysosomes, washed with PBS and cultured with complete culture medium for 18 h. Finally, HUVECs were incubated with PKH26-labeled EVs for colocalization analysis.

HUVECs were allocated into groups: PBS group (HUVECs without any treatment), EV group (HUVECs co-cultured with EV from ESCC cells), si-NC-EV group (HUVECs co-cultured with EVs from si-NC-modified ESCC cells), si-LINC00662-EV group (HUVECs co-cultured with EVs from si-LINC00662-modified ESCC cells), mimic-NC-EV group (HUVECs co-cultured with EVs from miR-195-5p mimic NC-modified ESCC cells), miR-195-5p mimic-EV group (HUVECs co-cultured with EVs from miR-195-5p mimic-modified ESCC cells), si-NC group (HUVECs transfected with si-NC), si-LINC00662 group (HUVECs transfected with si-LINC00662), oe-NC group (HUVECs transfected with oe-NC), oe-LINC0066 group (HUVECs transfected with oe-LINC0066), miR-195-5p mimic-EV + oe-NC group (HUVECs co-cultured with EVs from miR-195-5p mimic-modified ESCC cells and further transfected with oe-NC), miR-195-5p mimic-EV + oe-VEGFA (HUVECs co-cultured with EVs from miR-195-5p mimic-modified ESCC cells and further transfected with oe-VEGFA).

HUVECs were added with 5 mg/ml MTT solution (20 μl, Solarbio, Beijing, China), and cell viability was measured after 12, 24, and 36 h. Using a microplate spectrophotometer (BioTek Instruments, VT, United States), the absorbance at 570 nm was measured (You et al., 2021).

HUVECs were cultured in RPMI 1640 medium supplemented with 10% FBS, during which the medium was changed every 48 h. After 14 d, visible colonies were fixed with ethanol, stained with 0.1% crystal violet and manually counted.

Matrigel (BD Biosciences) was spread over 24-well plates at 100 ml/well and polymerized. HUVECs (resuspended in FBS-free medium) were transferred to the plates at 1 × 10⁵ cells/well and cultured for 6 h. Angiogenesis quantification was judged by the number of branch points of the formed tube in at least five fields of view (Zhou et al., 2019).

HUVECs were resuspended in serum-free medium, and the resuspension (1 × 10⁵ cells/mL, 200 μl) was added to the top side of Matrigel-coated Transwell chamber (Matrigel was not used to examine cell migration). RPMI 1640 medium (600 μl) containing 20% FBS was dispersed in the basolateral chamber. After 48 h of culture, cells were fixed with 4% paraformaldehyde, stained with crystal violet and observed under an optical microscope in five fields of view (Li Z et al., 2019).

LINC00662-Wt, LINC00662-Mut, VEGFA-Wt and VEGFA-Mut were cloned and merged with psiCHECK-2 vector (Promega, WI, United States). Employing Lipofectamine 2,000, the vectors were co-transfected with mimic-NC or miR-195-5p mimic into cells and detected by the dual luciferase reporter gene detection system (Promega) (Liu and Meng, 2020).

Cells were transfected with 50 nM biotin-labeled Wt-bio-miR-195-5p and Mut-bio-miR-195-5p for 48 h, and subjected to reaction in a specific lysis buffer (Ambion, Texas, United States) for 10 min. The lysate was incubated with M-280 streptavidin magnetic beads (Sigma) pre-coated with RNase-free BSA and yeast tRNA (TRNABAK-RO, Sigma). The beads were incubated at 4°C for 3 h and eluted, and the bound RNA was purified by Trizol, and the LINC00662 enrichment was detected by reverse transcription quantitative polymerase chain reaction (RT-qPCR).

The binding of LINC00662/miR-195-5p to AGO2 protein was analyzed by Magna RIP kit (Millipore, United States). Cells were lysed with an equal volume of RIPA Lysis Buffer (P0013B, Beyotime) in an ice bath for 5 min and centrifuged at 14,000 rpm, 4°C for 10 min. The magnetic beads (50 µl) were resuspended in 100 µl RIP Wash Buffer, and added 5 µg antibody AGO2 (1:50, Abcam) or IgG (1:100, Abcam). The magnetic bead-antibody complex was resuspended in 900 µL RIP Wash Buffer and incubated with 100 μl cell lysate overnight at 4°C. The resulting complex was digested with proteinase K and RNA was extracted for PCR detection.

Total RNA in cells and tissues was extracted by Trizol (Takara, Dalian, China) and expression of miR-195-5p was tested by qRT-PCR with the All-in-One miRNA qRT-PCR Detection Kit (GeneCopoeia) and All-in-One miRNA qPCR Primer (GeneCopoeia). Expression of LINC00662 and VEGFA was determined by qRT-PCR RevertAid First Strand cDNA Synthesis Kit (Fermentas) and SYBR Premix Ex Taq (TaKaRa). Primer sequences (Supplementary Table S1) were synthesized by Genepharma. Target expression was analyzed using ABI 7500 (ABI, CA, United States), and standardized by U6 or glyceraldehyde phosphate dehydrogenase. Determination of gene expression was dependent on 2^−ΔΔCt method (Li Z et al., 2019).

Total protein was separated with RIPA lysis buffer (Beyotime) and quantified by BCA kit (Pierce, MA, United States). After separation by 10% sodium dodecyl sulphate polyacrylamide gel electrophoresis (Beyotime), the protein sample was electroblotted onto nitrocellulose membrane (HTS112 M), covered with 5% skim milk and kept overnight with primary antibodies VEGFA (1:1,000), CD81 (1 μg/ml), CD9 (1:1,000), Calnexin (1 μg/ml), GM130 (1 μg/ml) and GAPDH (1:1,000, all from Abcam). Next, horseradish peroxidase-conjugated secondary antibodies goat anti-mouse IgG (1:5,000) and goat anti-rabbit IgG (1:5,000, both from Abcam) were incubated with the membrane, after which signals were detected using SuperSig-nal-West Pico chemiluminescent substrate (Thermo Fisher Scientific, MA, United States). Data analysis was carried out with Image Lab™ 3.0 (Bio-Rad, CA, United States) (Sui et al., 2021).

Eca109 cells (2 × 106) were injected subcutaneously into the hind limb of BALB/c nude mice (6 weeks old, athymic) for the establishment of the ESCC xenograft model. When the nude mice generated the tumors (100 mm³), 10 μg EVs from ESCC cells that had been transfected with si-NC, si-LINC00662, mimic-NC, miR-195-5p mimic, PBS, and exosome release inhibitor (GW4869, 1 mg/kg) were subsequently injected into the center of tumor sites every 2 days. Nude mice were euthanized 4 weeks later, and the subcutaneous tumors were obtained to calculate tumor volume (maximum diameter × vertical diameter²/2).

GraphPad Prism 6.0 (GraphPad Software Inc.) and SPSS 20.0 (IBM, NY, United States) were used for data analysis. Values were expressed as mean ± standard deviation and compared by t-test (two groups) or analysis of variance (multiple groups). p < 0.05 was statistically significant.

LINC00662, miR-195-5p, and VEGFA levels in tissue specimens of 84 patients with ESCC were determined by RT-qPCR, as the results reflecting that LINC00662 and VEGFA were upregulated and miR-195-5p was downregulated in ESCC tissues (Figures 1A–C).

ESCC patients were grouped according to TNM staging, and RT-qPCR analysis found higher LINC00662 expression in patients in the late stage than in the early stage (Figure 1D). LINC00662 expression was compared in HEEC and ESCC cell lines (EC9706, Eca109, TE13, TE1 and TTN) by RT-qPCR, and the outcomes indicated that LINC00662 was abundantly expressed in ESCC cell lines (Figure 1E). Because Eca109 had the highest LINC00662 level, it was selected for subsequent experiments.

To clarify the effect of LINC00662 on Eca109 angiogenesis in vitro, LINC00662 was overexpressed or specifically interfered. RT-qPCR test results showed that oe-LINC00662 effectively promoted LINC00662 expression, while si-LINC00662 down-regulated LINC00662 expression in HUVECs (Figure 2A). As reflected by MTT assay, colony formation assay, Transwell assay and angiogenesis test, downregulated LINC00662 depressed proliferation, invasion, migration and angiogenesis but up-regulated LINC00662 promoted these phenotypes of HUVECs (Figures 2B–F). Collectively, LINC00662 downregulation inhibits tumor angiogenesis.

A white precipitate was separated from the culture medium of Eca109 cells by differential centrifugation. Under the TEM, the white precipitate was round or elliptical vesicles with a double-layer membrane structure and a diameter of 40–100 nm (Figure 3A); TRPS showed that the mode of particle diameter was 90 nm and the average diameter was 130 nm (Figure 3B); Western blot detected that CD81 and CD9 were both positively expressed whilst calnexin and GM130 were not expressed in these vesicles (Figure 3C), confirming that EVs were extracted successfully.

PKH26-labeled EVs were co-cultured with HUVECs. Observed under a fluorescence microscope, EVs could enter HUVECs, and over time, the more EVs taken up by HUVECs (Figure 3D). Fluorescence colocalization experiment showed that HUVECs incubated with Dx498 were co-cultured with PKH26-labeled EVs, and EVs and lysosomes were colocalized (Figure 3E).

After co-culture with Eca109 cells-derived EVs, RT-qPCR detection found that LINC00662 expression was increased in HUVECs, while LINC00662 expression was decreased after co-culture with EVs extracted from Eca109 cells transfected with si-LINC00662 (Figures 4A,B). Subsequently, analysis of cell functions revealed that Eca109 cells-derived EVs augmented proliferation, invasion, migration and angiogenesis of HUVECs, while downregulated LINC00662 reversed the effects of Eca109 cells-derived EVs (Figures 4C–G). In conclusion, Eca109 cells-derived EVs transfer of silenced LINC00662 blocks angiogenesis.

RT-qPCR test measured that co-culture with EVa from miR-195-5p mimic-treated Eca109 cells augmented miR-195-5p expression in HUVECs (Figure 5A). As a result of miR-195-5p upregulation, Eca109 cells-derived EVs-induced aggressive behaviors and angiogenesis of HUVECs were all constrained (Figures 5B–F). In short, EVa transfer of upregulated miR-195-5p exerts anti-angiogenic effect.

On the bioinformatics website, a binding relationship was predicted between LINC00662 and miR-195-5p (Figure 6A), as well as miR-195-5p and VEGFA mRNA (Figure 6B). Using dual luciferase reporter assay, it was estimated that miR-195-5p mimic reduced the luciferase activity of LINC00662-Wt and that of VEGFA-Wt (Figures 6C,D). RNA-pull down assay showed that Mut-bio-miR-195-5p had no significant effect on LINC00662 enrichment and Wt-bio-miR-195-5p could increase the amount of LINC00662 enrichment (Figure 6E). RIP experiment suggested an elevation in Ago2-bound LINC00662, indicating that LINC00662 could directly bind with miR-195-5p (Figure 6F).

Alternations of gene expression were observed: EVs delivery of silenced LINC00662 elevated miR-195-5p expression in HUVECs (Figure 6G) while that of upregulated miR-195-5p lowered VEGFA levels (Figures 6H,I).

miR-195-5p/VEGFA axis-induced function for HUVECs was further evaluated. After co-culture with EVs carrying upregulated miR-195-5p, HUVECs were transfected with oe-VEGFA, which successfully elevated VEGFA expression (Figures 7A,B). Overexpression of VEGFA reversed the effects of EVs carrying upregulated miR-195-5p on proliferation, invasion, migration and angiogenesis of HUVECs (Figures 7C–G).

To further determine the angiogenic role of EVs in vivo, an ESCC tumor-bearing nude mice model was established and treated with EVs. In mice xenograft model, after treatment with EVs delivering depleted LINC00662 or restored miR-195-5p, tumors grew slowly, accompanied by smaller tumor volume and lighter tumor weight (Figures 8A–C). Furthermore, the effect of EV-specific inhibitor GW4869 on the growth of nude mice was observed, reflecting that GW4869 treatment had a better efficacy to repress tumor growth (Figures 8D–F). Simply, LINC00662 suppression or miR-195-5p overexpression inhibits the growth of tumor.