Long Noncoding RNA VESTAR Regulates Lymphangiogenesis and Lymph Node Metastasis of Esophageal Squamous Cell Carcinoma by Enhancing VEGFC mRNA Stability

Esophageal squamous cell carcinoma (ESCC) is the predominant pathologic type of esophageal cancer that is known as a common and deadly malignancy worldwide, especially in China (1–3). The poor outcome in esophageal cancer is largely due to cancer metastasis, with 5-year survival rate declining from 43% for patients with localized disease to 23% and 5% for those with regional and distant metastasis, respectively (4). In particular, regional lymph node metastasis represents the key indication of tumor cell dissemination, and is a strong predictor for poor survival of patients with ESCC (5, 6).

Lymphatic vasculature acts as an important route for metastasis of tumor cells from primary sites to lymph nodes (7). Tumor lymphangiogenesis, a complex process regulated by multiple growth factors, induces the generation of new lymphatic vessels that facilitate metastasis (7, 8). VEGFC is a well identified lymphangiogenic growth factor, which promotes lymphangiogenesis and subsequent regional lymph node metastasis by activating its receptor VEGFR3 expressed on lymphatic endothelial cell (9–12). The expression of VEGFC in primary tumors correlates with lymphatic metastasis in a range of human cancers, including ESCC (12, 13). Inspiringly, blockade of VEGFC/VEGFR3 signaling axis has shown an effective suppression on tumor lymphangiogenesis and metastasis to regional lymph nodes (14, 15). Interestingly, HuR, a RNA binding protein, is reported to interact with VEGFC mRNA and increase its half-life (16). Cytoplasmic HuR is associated with high VEGFC expression, as well as lymph node metastasis or lymphangiogenesis in human cancers (17–19). Thus, revealing the molecular mechanisms that control lymphatic metastasis will provide more information for clinical treatment.

Long noncoding RNAs (lncRNA), a large class of nonprotein-coding transcripts >200 nucleotides in length, serve as key regulators of gene activity at various aspects, such as epigenetic regulation (20), mRNA translation (21), and protein modification (22). Lots of studies have provided evidence that lncRNAs are involved in the development and progression of human cancers (23, 24). Interestingly, lncRNA LNMAT1 and BLACAT2 are recently shown to regulate lymphangiogenesis and lymphatic metastasis in bladder cancer (25, 26). Nevertheless, the roles and mechanisms of lncRNAs in tumor lymphangiogenesis and lymphatic metastasis are still largely unknown.

Somatic copy number alterations (SCNA) are common genomic abnormalities throughout human cancers and contribute to tumorigenesis (27, 28). Traditionally, protein-coding genes located within SCNAs are considered as driver candidates to exert the functions of SCNAs; in fact, a large number of nonprotein-coding genes are also harbored in SCNAs and play important roles in carcinogenesis (29, 30). Our previous study revealed the landscape of genomic alterations in ESCC and identified 6 significantly amplified regions that are associated with regional lymphatic metastasis, such as 14q32.2–32.33 (30). Here, we screened 14q32.2–32.33 region and identified a lymphatic metastasis associated-lncRNA, LINC00638, which we named VESTAR (VEGFC mRNA stability associated lncRNA). VESTAR was overexpressed in ESCC, and was even exported from the nucleus to the cytoplasm via NXF1 and adaptor protein SRSF3. VESTAR was shown to play important roles in lymphangiogenesis and lymph node metastasis of ESCC. Mechanistically, VESTAR promoted the VEGFC mRNA stability by binding to both VEGFC mRNA and HuR and subsequently enhancing the interaction between them. Collectively, we reveal that VESTAR, an lncRNA harbored in SCNAs, regulates the VEGFC mRNA stability and promotes lymphangiogenesis and lymphatic metastasis in ESCC.

Human ESCC cell lines KYSE510, KYSE450, KYSE410, KYSE30, KYSE140, KYSE70, KYSE150, KYSE180, YES2, and COLO680N, which were kindly provided by Dr. Yutaka Shimada of Kyoto University, (Kyoto, Japan) were grown in RPMI 1640 medium (Gibco) supplemented with 10% FBS and 1% penicillin/streptomycin. Immortalized esophageal epithelium cell lines NE2 and NE3 were kindly provided by Prof. Enmin Li from Shantou University and were cultured in EpiLife/dKSFM (1:1) medium (Gibco) with 1% penicillin/streptomycin. Human dermal lymphatic endothelial cells (HDLEC) were purchased from ScienCell Research Laboratories and were cultured in DMEM (Gibco) supplemented with 10% FBS and 1% penicillin/streptomycin. All cells were maintained at 37°C with 5% CO₂ in a humidified incubator. Cell lines were used within five passages after thawing. The cells were authenticated using short tandem repeat profiling and were Mycoplasma-free identified by Mycoplasma Real-time PCR Detection Kit (Applied Biosystems).

Human ESCC tissue microarrays with detailed clinical characteristics and survival times were purchased from Outdo Biotech (Shanghai, China). The tissue microarrays were deparaffinized and incubated with Proteinase K (15 μg/mL) for 8 minutes at 37°C. After dehydration, the microarrays were hybridized with a mixture of two double-(5′ and 3′)-digoxin labeled VESTAR probes (5DigN/TGGCAGAGGGAGAAGCACAT/3Dig_N and 5DigN/AACTCACAGGCAGGAGATGGCT/3Dig_N; Exiqon) for 1 hour at 50°C, followed by stringent washes with SSC buffers at hybridization temperature and incubation with anti-digoxin-AP antibody at 4°C overnight. Microarrays were then subjected to enzymatic development with NBT/BCIP substrate. The stained microarrays were counterstained with nuclear fast red and were mounted. The percentage of positive cells in nucleus or cytoplasm was graded as 0, < 5%; 1, 5%–25%; 2, 26%–50%; 3, 51%–75%; 4, >75%. Staining intensity was graded as 0 for negative; 1 for weak; 2 for moderate; 3 for strong. The multiplication of percentage of positive cells and staining intensity was used to define the nuclear value or cytoplasmic value, and nuclear value plus cytoplasmic value were used to define the expression level of VESTAR. After removing the unqualified specimens, 161 paired ESCC tissues and adjacent noncancerous tissues, as well as 8 cases of ESCC tissues were involved into the analysis.

The cytoplasmic and nuclear fractions were obtained by using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific) according to the manufacturer's instructions. TRIzol reagent was then added to the cytoplasmic and nuclear fractions, respectively, to extract the corresponding RNA.

RIP assay was done with Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore) according to the manufacturer's instructions. Anti-CRM1 antibody (ProteinTech, 27917–1-AP), anti-NXF1 antibody (Abcam, ab50609), anti-SRSF3 antibody (Abcam, ab198291), or anti-HuR antibody (ProteinTech, 11910–1-AP) was incubated with cell lysates to immunoprecipitate the protein–RNA complexes. The retrieved RNA was isolated with phenol:chloroform:isoamyl alcohol and was subjected to reverse transcription-quantitative real-time PCR (qRT-PCR). The protein presented in the complexes was detected by Western blotting. Normal IgG was used as the negative control.

Biotin-labeled VESTAR, VESTAR-1, VESTAR-2, VESTAR-3, VESTAR-4, VESTAR-5, and GAPDH were in vitro transcribed using MEGAscript T7 Transcription Kit (Life Technology) with Bio-16-UTP (Life Technology) according to the manufacturer's instructions. The DNA template of VESTAR was linear pcDNA3.1-VESTAR plasmid digested by NotI, a restriction enzyme downstream of VESTAR. The DNA templates of GAPDH and VESTAR fragments were PCR products containing T7 promoter sequence upstream. The gene-specific primers used for PCR are presented in the Supplementary Table S1.

RNA pull-down assay was then performed with these biotin-labeled RNAs. Streptavidin agarose beads (Invitrogen) were washed by RNA washing buffer (150 mmol/L KCl, 25 mmol/L Tris-HCl pH 7.4, 0.5 mmol/L DTT, 0.5% NP40) for six times. Cell extracts were preincubated with beads at room temperature for 15 minutes to remove the nonspecific interacted proteins. 5 pmol biotin-labeled RNA in RNA structure buffer (10 mmol/L Tris-HCl pH 7.0, 0.1 mol/L KCl, 10 mmol/L MgCl₂) was heated at 95°C for 2 minutes, chilled on ice for 3 minutes, and then shifted to the room temperature for 20 minutes. The folded RNA was mixed with cell extracts (600 μg proteins) in binding buffer (washing buffer supplemented with 40 μg/mL tRNA, 80 U/mL RNase inhibitor, and 1 × protease inhibitor cocktail), and incubated at room temperature for 30 minutes. The washed beads were then added to the mixture and were incubated at room temperature for another 30 minutes. The beads were washed with washing buffer for six times. For protein analysis, the beads were boiled in 1 × loading buffer for 10 minutes, while the RNA present in the complexes was isolated and analyzed by qRT-PCR.

The lentiviruses containing shRNA against VESTAR (shVESTAR, the same sequence as siRNA-2 of VESTAR) or control shRNA (shNC) were provided by HANBIO. COLO680N cells were infected with lentiviruses in the presence of 5 μg/mL polybrene and were selected with 2 μg/mL puromycin to generate the stable knockdown cell line.

Nude mice (NU/NU, 4 to 6 weeks old, male) were purchased from Charles River (Beijing, China). The mice were divided into two groups (n = 6 mice/group). The 1 × 10⁶ COLO680N cells stably expressing control shRNA (shNC) or shRNA targeting VESTAR (shVESTAR) were inoculated into the foot-pads of nude mice. When the tumor grew to 200 mm³, the mice were sacrificed, and the xenograft tumors and popliteal lymph nodes were resected and embedded in paraffin, followed by IHC analysis with anti-CD31 antibody (Abcam, ab28364), anti-LYVE1 antibody (Abcam, ab14917), or anti-cytokeratin antibody (ZSGB-BIO, ZM-0069). Animal experiments were approved by the Institutional Animal Care and Use Committee of Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College.

The IHC was performed on the paraffin sections from mouse lymph node tissue and xenograft tumor. Slides were deparaffinized in xylene and rehydrated through gradient alcohol. Antigen retrieval was performed with microwave for 15 minutes in citrate buffer (pH 6.0). Endogenous peroxidase activity was inhibited with 3% hydrogen peroxide for 15 minutes at room temperature. Slides were then blocked with 5% normal goat serum, and subsequently incubated with the indicated antibody at 4°C overnight. After incubation with goat anti-mouse/rabbit IgG HRP-polymer (ZSGB-BIO, China) for 20 minutes at room temperature, the slides were visualized with 3,3-diaminobenzidine solution (ZSGB-BIO, China), counterstained with hematoxylin, dehydrated, and mounted.

ESCC cells transfected with the indicated siRNAs or plasmids were cultured for 48 hours. The culture medium was replaced with fresh medium containing 1 μg/mL actinomycin D (Sigma). Then, cells were collected at the indicated time points: 0, 1, 3, and 5 hours, followed by RNA extraction and qRT-PCR to analyze the mRNA levels of VEGFC.

Statistical analysis was performed by SPSS 19 software or GraphPad Prism 5 software. Nonparametric Mann–Whitney U test or two-tailed unpaired/paired Student t test was used to analyze the statistical difference of two groups. The correlation between DNA copy number and RNA expression was analyzed by Pearson correlation. Kaplan–Meier analysis and log-rank test were employed to evaluate the association of VESTAR expression or subcellular localization with patient overall survival, and the cut-off value of VESTAR expression was obtained by X-Tile software (31). The χ² test was used to analyze the association between VESTAR subcellular localization and clinicopathologic parameters. Data are presented as the mean ± SD. P value <0.05 was considered statistically significant.

Additional Materials and Methods are provided in the Supplementary Materials and Methods.

In our previous study, 14q32.2–32.33 region was shown to be frequently amplified in ESCC, which was significantly associated with regional lymph node involvement (30). Of note, besides protein-coding genes, a dozen of lncRNAs annotated by NCBI RefSeq are located in this amplified region. To identify the lymphatic metastasis-associated lncRNAs in 14q32.2–32.33 region, we used two filter criteria: first, the abundance of lncRNAs should be high; second, the DNA copy number alteration should be in accordance with RNA expression levels, as the SCNA-harbored genes with consistent DNA copy number and RNA expression are likely to be drivers of cancer (32). According to the Cancer Cell Line Encyclopedia (CCLE) data (33), VESTAR, MEG3, MEG9, and DIO3OS showed relatively higher expression levels than other lncRNAs in esophageal cancer cell lines (Supplementary Fig. S1A); while, of these, only VESTAR exhibited a significantly positive correlation between its DNA copy number and RNA expression level (Supplementary Fig. S1B–S1E). We also observed the positive correlation of VESTAR DNA and RNA levels in ESCC cell lines detected by qRT-PCR (Supplementary Fig. S1F). In addition, we found that VESTAR was upregulated in several ESCC cell lines compared with the immortalized esophageal epithelium cell lines NE2 and NE3 (Supplementary Fig. S1G).

We subsequently analyzed the expression of VESTAR in ESCC tissues using the data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO; ref. 34), and found that VESTAR was highly expressed in ESCC tissues compared with noncancerous tissues (Fig. 1A–C). Furthermore, we performed RNA ISH assay and validated the overexpression of VESTAR in a cohort of 161 paired ESCC tissues and adjacent noncancerous tissues (Fig. 1D and E). Kaplan–Meier analysis demonstrated that high VESTAR expression was associated with poor overall survival in all patients with ESCC (n = 169; Fig. 1F), as well as in patients with TNM stages I/II (n = 87; Supplementary Fig. S2A). Moreover, analyses of TCGA database revealed that VESTAR was also elevated in several other types of digestive system cancers, including liver hepatocellular carcinoma, stomach adenocarcinoma, and colon adenocarcinoma (Supplementary Fig. S2B–S2D).

We then performed 5′ and 3′ rapid amplification of cDNA ends (RACE) assay and observed that VESTAR was polyadenylated and spliced, and may have four transcripts in ESCC cells (Supplementary Fig. S3A and S3B). Northern blot and RT-PCR assays revealed that transcript 2 with a full length of 2,908 nucleotides presented in ESCC cells (Supplementary Fig. S3C and S3D; Supplementary Table S2), which also existed in NE2 cells but at a lower level (Supplementary Fig. S3C). Thus, transcript 2 was chosen for further study. The Coding Potential Calculator (35) and Coding-Potential Assessment Tool (36) predicted that VESTAR was a noncoding RNA (Supplementary Fig. S3E), and in vitro translation assay further showed that there were no obvious translated products of VESTAR (Supplementary Fig. S3F).

To our surprise, ISH results showed that, besides the altered expression levels, VESTAR location was also changed (Fig. 1D). VESTAR was largely localized to the nucleus in 158 out of 161 cases of adjacent noncancerous tissues, whereas it was located to both nucleus and cytoplasm in 84 out of 169 cases of ESCC tissues (Fig. 2A). Consistently, we observed more cytoplasmic VESTAR in KYSE180 and COLO680N cells compared with NE2 cells, by using subcellular fractionation assay combined with qRT-PCR (Fig. 2B and Supplementary Fig. S3G). According to VESTAR subcellular localization, we divided 169 ESCC cases into two groups: the VESTAR-export group (n = 84) and the VESTAR-non-export group (n = 85). VESTAR expression in export group was higher than that in nonexport group (Fig. 2C), implying that the overexpressed VESTAR might be exported from the nucleus to the cytoplasm in ESCC.

Statistical analysis revealed that VESTAR export was correlated with lymph node metastasis and TNM stages (Supplementary Table S3). Compared with the lymph node–negative ESCC tissues, the lymph node–positive ESCC tissues exhibited a high rate of VESTAR export (Fig. 2D). Furthermore, the cytoplasmic levels of VESTAR were upregulated in lymph node–positive ESCC tissues compared with lymph node–negative ESCC tissues (Fig. 2E). Although VESTAR export had no correlation with overall survival in all patients with ESCC, combination of VESTAR expression and export better predicted the overall survival than alone in patients with TNM stages I/II (n = 87; Fig. 2F; Supplementary Fig. S2A), and was even more predictive of prognosis than positive lymph node (Fig. 2G). However, compared with the lymph node–positive rate of 31.03% (9/29) in stage IIB, the lymph node–positive rate was only 14.29% (3/21) in VESTAR high and export group of patients with TNM stages I/II, indicating that for the stages I/II patients, combination of VESTAR high and export is not more predictive of spread than stage IIB. Taken together, these data demonstrate that VESTAR is exported from the nucleus to the cytoplasm in ESCC, which is associated with lymph node metastasis.

We then set out to explore how VESTAR is exported from the nuclear to the cytoplasm. On the basis of the splicing and polyadenylation of VESTAR in ESCC cells as mentioned above, we speculated that the mechanisms underlying VESTAR export may be similar to that of mRNAs. There are two main traditional mRNA export pathways: the CRM1- and NXF1-dependent pathways (37). To determine whether these two pathways contribute to VESTAR export, we used siRNAs to knock down CRM1 and NXF1 in COLO680N cells (Supplementary Fig. S4A and S4B), and subsequently detected the VESTAR abundance in nuclear and cytoplasmic fractions via subcellular fractionation assay combined with qRT-PCR. As shown in Fig. 3A and B, the subcellular localization of VESTAR was unchanged upon CRM1 knockdown, but VESTAR was aggregated in the nucleus after NXF1 depletion. To make the results more convincing, we used IFNα1 mRNA and NANOG mRNA as positive controls, and used GAPDH mRNA and ACTIN mRNA as negative controls, for CRM1 and NXF1, respectively (Fig. 3A and B). It has been reported that IFNα1 mRNA was colocalized with CRM1 but not with NXF1 in the nucleus of HeLa cells, and the dysfunction of CRM1 inhibited the export of IFNα1 mRNA (38), while the NANOG mRNA bound to NXF1, and its export was associated with NXF1 (39, 40). Furthermore, analysis of published RNA-seq data of GSE139151 from MCF7 cells with NXF1 depletion (41) showed the consistence with our results (Supplementary Fig. S4C), collectively indicating that NXF1 is associated with VESTAR export.

NXF1 is a major mRNA export receptor, which binds to target mRNA via RNA-binding domain and exports it to cytoplasm through the nuclear pore complex (37, 42). To evaluate whether NXF1 binds to VESTAR, we performed RIP assay and found that VESTAR could be retained by an anti-NXF1 antibody but not by an anti-CRM1 antibody, compared with IgG control (Supplementary Fig. S4D and S4E; Fig. 3C and D). RNA pull-down assay using in vitro transcribed biotin-labeled RNAs validated the interaction between VESTAR and NXF1, as NXF1 was clearly detected in the VESTAR-associated complexes compared with that in GAPDH control group (Fig. 3E).

Efficient RNA export by NXF1 requires multiple adaptors, such as transcription export (TREX) complex components and SR proteins, which link the target transcript to NXF1 (42–44). SRSF3, a member of the SR protein family, is reported to be an important NXF1 adaptor to interact with and recruit NXF1 to mRNA (40). According to the TCGA data and GEO data (34), the mRNA levels of NXF1 and SRSF3 were higher in ESCC tissues than those in noncancerous tissues (Supplementary Fig. S4F–S4I). Subcellular fractionation and qRT-PCR assays showed that knockdown of SRSF3 using siRNAs in COLO680N cells increased the levels of VESTAR in nucleus (Supplementary Fig. S4J; Fig. 3F). RIP and RNA pull-down assays confirmed that SRSF3 was able to interact with VESTAR (Supplementary Fig. S4K; Fig. 3G and H). Importantly, SRSF3 depletion reduced the interaction of NXF1 and VESTAR as determined by RIP assay (Fig. 3I). Collectively, these results demonstrate that NXF1 contributes to VESTAR export, in which SRSF3 acts as an adaptor protein to facilitate the binding of NXF1 to VESTAR.

To investigate the biological functions of VESTAR in ESCC cells, we first employed siRNAs to knock down VESTAR in KYSE180 and COLO680N cells, which had high VESTAR expression (Supplementary Fig. S1G; Fig. 4A). As shown in Fig. 4B–E, knockdown of VESTAR led to a decrease in ESCC cell proliferation, colony formation, migration, and invasion. Then, we selected KYSE30 and KYSE450 cells, with low VESTAR expression, for ectopic expression of VESTAR (Supplementary Fig. S1G and S5A). Ectopic expression of VESTAR enhanced the migration and invasion abilities of ESCC cells (Supplementary Fig. S5B and S5C). We detected the subcellular localization of VESTAR in VESTAR-overexpressed ESCC cells, and observed that most of VESTAR was located to the cytoplasm (Supplementary Fig. S5D). However, knockdown of VESTAR had no effect on the cell proliferation and migration of immortalized esophageal epithelium cell line NE2, which showed more nuclear VESTAR (Supplementary Fig. S3G and S6A–C). These results indicate that cytoplasmic VESTAR may contribute to the malignant characters of ESCC cells.

Correlation of both 14q32.2–32.33 amplification and VESTAR export with lymphatic metastasis in ESCC prompted us to investigate whether VESTAR plays a role in ESCC lymphangiogenesis. Transwell migration assay showed that the migration ability of HDLECs was decreased in conditioned medium from VESTAR-depleted ESCC cells, but was increased in conditioned medium of VESTAR-overexpressed ESCC cells, compared with the corresponding control counterparts (Fig. 5A; Supplementary Fig. S7A). To exclude the proliferation effect, we used mitomycin C pretreated HDLECs to perform the Transwell migration assay again and obtained consistent results (Supplementary Fig. S7B and S7C). Furthermore, tube formation assay revealed that VESTAR depletion inhibited the ability of ESCC cells to induce the tube formation of HDLECs (Fig. 5B). However, VESTAR overexpression alone almost showed no effect on the tube formation of HDLECs, suggesting that other factors might be required for VESTAR-induced ESCC lymphangiogenesis. Collectively, these data indicate that VESTAR induces ESCC lymphangiogenesis in vitro.

To investigate the effects of VESTAR on ESCC lymphangiogenesis and lymph node metastasis in vivo, we established COLO680N cells stably expressing shRNA targeting VESTAR (Supplementary Fig. S8). Then, COLO680N control cells and VESTAR-knockdown cells were inoculated into the foot-pads of nude mice, and the primary tumors and popliteal lymph nodes were obtained when the tumor grew to 200 mm³. Tumor tissues were immunostained with anti-LYVE1 antibody and anti-CD31 antibody to detect microlymphatic vessels and exclude the LYVE1⁺ macrophage. Compared with control counterparts, the tumors formed by VESTAR-knockdown cells exhibited the decreased levels of microlymphatic vessels (Fig. 5C). Furthermore, the volumes of lymph node in VESTAR-knockdown group were smaller than those in control group (Fig. 5D). IHC for cytokeratin further confirmed that the lymph nodes of VESTAR-knockdown group showed less tumor cells than those of control group (Fig. 5E). The ratio of metastatic lymph node was 83.3% (5/6) in control group, but 33.3% (2/6) in VESTAR-knockdown group (Fig. 5F). Therefore, all of above results suggest that VESTAR plays important roles in ESCC lymphangiogenesis and lymph node invasion.

It is well established that VEGFC promotes the spread of tumor cells to regional lymph nodes by inducing lymphangiogenesis (9, 10, 12). Hence, we speculated that VEGFC might be involved in VESTAR-induced lymphangiogenesis. Importantly, the expression of VEGFC, at both mRNA and secreted protein levels, was decreased in VESTAR-depleted ESCC cells, but was increased in VESTAR-overexpressed ESCC cells, as determined by qRT-PCR and ELISA (Fig. 6A and B).

In general, mRNA level is affected by transcription activity and mRNA stabilization. The transcription activity was first assessed by examination of VEGFC pre-mRNA levels, and the results showed that the pre-mRNA levels of VEGFC were unchanged upon VESTAR overexpression or depletion (Supplementary Fig. S9A). Luciferase reporter assay further revealed that VESTAR overexpression had no effect on the luciferase activity driven by VEGFC promoter (Supplementary Fig. S9B). These data suggest that VESTAR could not influence the transcription activity of VEGFC. We then examined the decay of VEGFC mRNA following treatment with actinomycin D. As shown in Fig. 6C, depletion of VESTAR accelerated, whereas overexpression of VESTAR slowed down, the degradation of VEGFC mRNA, indicating that VESTAR stabilizes the VEGFC mRNA.

LncRNAs have been evidenced to regulate mRNA stability by targeting to mRNA via RNA–RNA interaction (45, 46). To investigate whether VESTAR binds to VEGFC mRNA and acts in the same way, we conducted RNA pull-down assay using in vitro transcribed biotin-labeled RNAs and, as expected, observed an enrichment of VEGFC mRNA, but not its pre-mRNA, in VESTAR group compared with GAPDH control group (Fig. 6D; Supplementary Fig. S9C).

To identify the specific binding region of VESTAR to VEGFC mRNA, we fragmented VESTAR prior to doing RNA pull-down assay (Fig. 6E). As shown in Fig. 6F, fragment VESTAR-1 (nucleotides 1 to 580) exhibited the highest binding affinity to VEGFC mRNA than other ones. We then transfected the VESTAR-1 to KYSE30 cells and examined the decay of VEGFC mRNA following treatment with actinomycin D. Indeed, VESTAR-1 attenuated the degradation of VEGFC mRNA (Fig. 6G).

Using the NCBI BLAST analysis, we predicted five binding sites of VESTAR-1 to VEGFC mRNA (Supplementary Fig. S9D). To validate the direct interaction between them, a point mutant of VESTAR-1, namely VESTAR-1-Mut, was obtained by mutation of all five binding sites in VESTAR-1. As expected, RNA pull-down assay revealed that mutation of all binding sites in VESTAR-1 greatly eliminated its interaction with VEGFC mRNA (Fig. 6H). Furthermore, VESTAR-1, but not its mutant VESTAR-1-Mut, inhibited the decrease of VEGFC mRNA caused by VESTAR depletion (siRNA specific to VESTAR but not VESTAR-1) in COLO680N cells (Fig. 6I). Taken together, these results indicate that VESTAR could stabilize VEGFC mRNA by directly binding to VEGFC mRNA with its 5′ region.

We finally aimed to explore whether the binding of VESTAR to VEGFC mRNA would affect the interaction of VEGFC mRNA with its protein partners. HuR has been reported to interact with VEGFC mRNA and increase its half-life after sunitinib treatment (16); moreover, cytoplasmic expression of HuR is positively related to VEGFC expression in human cancers (17–19). Consistently, siRNA-mediated knockdown of HuR in COLO680N cells led to an accelerated decay of VEGFC mRNA (Supplementary Fig. S10A and S10B; Fig. 7A). Moreover, RIP assay displayed that HuR could bind to VEGFC mRNA in COLO680N cells (Supplementary Fig. S10C; Fig. 7B).

To determine whether VESTAR affects the binding of HuR to VEGFC mRNA, we performed RIP assay and observed that the interaction between HuR and VEGFC mRNA was attenuated after knockdown of VESTAR (Fig. 7C). Interestingly, VESTAR was also able to interact with HuR as determined by RIP assay (Fig. 7B). RNA pull-down assay further confirmed this interaction and showed that the 3′ end of VESTAR was responsible for the binding with HuR (Fig. 7D).

To test whether HuR is involved in VESTAR-mediated regulation of VEGFC mRNA expression, we simultaneously knocked down VESTAR and HuR in COLO680N cells, and found that simultaneous knockdown did not cause a synergistic effect on the decrease of VEGFC mRNA (Fig. 7E). Furthermore, knockdown of HuR inhibited the upregulation of VEGFC mRNA induced by VESTAR overexpression in KYSE450 cells (Fig. 7F). Taken together, these data demonstrate that VESTAR enhances the interaction between HuR and VEGFC mRNA, and that the regulation of VESTAR on VEGFC mRNA is mainly dependent on HuR.

To investigate whether VEGFC is essential for the VESTAR-induced lymphangiogenesis, Transwell migration and tube formation assays were performed. Ectopic expression of VEGFC rescued the ability of VESTAR-depleted ESCC cells to induce migration and tube formation of HDLECs (Fig. 7G and H). To further confirm that the lymphangiogenesis effect of VESTAR is VEGFC/VEGFR3 signaling axis-dependent, we employed SAR131675, a VEGFR3 inhibitor, in our study. SAR131675 repressed the promotion effect of VESTAR overexpression on HDLECs migration (Supplementary Fig S11A). Furthermore, knockdown of VESTAR had no effect on the migration and tube formation of HDLECs, upon treatment with SAR131675 (Supplementary Fig. S11B and S11C). These results indicate that VEGFC is required for VESTAR-induced lymphangiogenesis.

Tumor lymphangiogenesis, largely driven by VEGFC/VEGFR3 signaling axis, is important for lymphatic metastasis (7–10) and acts as a predictor of prognosis in ESCC (47). Protein factors, such as TBL1XR1 and TRIM3, have been reported to function as regulators of ESCC lymphangiogenesis and lymphatic metastasis through affecting the VEGFC expression (48, 49), but little is known about the roles of lncRNAs in those processes. Herein, VESTAR, an lncRNA located in 14q32.2–32.33 region that we previously reported to be amplified and associated with regional lymph node metastasis in ESCC (30), was identified to be involved in lymphangiogenesis and lymphatic metastasis of ESCC via stabilizing VEGFC mRNA by direct interaction. To our knowledge, this is the first report of lncRNAs participating in lymphangiogenesis and lymphatic metastasis of ESCC, which may provide new insights into mechanisms underlying lymph node metastasis.

Although a series of genes are located in SCNAs, only a few of them are potential cancer-associated genes (50). Based on the concordance of DNA copy number and RNA expression levels, VESTAR, an lncRNA harbored in amplified 14q32.2–32.33 region, was selected as the target gene. There may be four transcript variants of VESTAR in ESCC cells as determined by RACE assay, but the northern blot assay showed that the most abundant transcript variant was the 2908-nucleotide transcript (transcript 2). VESTAR transcript 2 was also the most abundant transcript variant in NE2 cells, indicating that this transcript variant of VESTAR is not specifically expressed in ESCC. VESTAR was overexpressed in ESCC, and was unexpectedly exported from the nucleus to the cytoplasm in some ESCC cases as determined by ISH assay. Interestingly, VESTAR export, but not its overexpression, was correlated with lymph node metastasis of ESCC, indicating that the cytoplasmic VESTAR has effect on ESCC lymphatic metastasis. These results raised the possibility that, besides the altered expression level, the change of gene's subcellular localization may also imply important information and contribute to cancer progression.

CRM1- and NXF1-dependent pathways are the two main manners responsible for mRNA nuclear export (37). LncRNAs, like mRNAs, are generally transcribed by RNA polymerase II, spliced, and added a cap and a poly (A) tail (51), implying that such lncRNAs may share the same export pathway with mRNAs. Indeed, for example, nuclear DNA-encoded lncRNA RMRP is exported to the cytoplasm in a CRM1-dependent manner by binding to HuR and is accumulated in mitochondria by interaction with GRSF1 (52). In our study, we found that the nuclear export of VESTAR in ESCC cells was regulated by NXF1- but not CRM1-dependent manner, in which SRSF3 severs as an adaptor protein to promote the interaction between NXF1 and VESTAR. However, the mechanisms by which VESTAR is retained in the nucleus of normal tissues still remain unclear. Interestingly, a research in yeast showed that transcription elongation factors Paf1 determines the differential nuclear export of mRNAs and the nuclear retention of most lncRNAs (53).

Consistent with the facts that 14q32.2–32.33 amplification and VESTAR export correlated with regional lymph node metastasis of ESCC, we found that VESTAR played important roles in ESCC lymphangiogenesis and lymph node metastasis. VEGFC, the best-characterized factor related to lymphangiogenesis, is reported to be associated with lymphangiogenesis and lymph node metastasis in ESCC (13). Here, VESTAR regulated VEGFC expression at the post-transcriptional level via increasing its mRNA stability, which is different from the transcriptional regulation by proteins, such as TBL1XR1 and AKIP1, in ESCC (49, 54), and the epigenetic regulation by lncRNA BLACAT2 in bladder cancer (26).

LncRNAs could modulate the mRNA stability through multiple ways, one of which is to bind to the target mRNA directly (45, 46). In this study, we demonstrated that VESTAR could directly bind to VEGFC mRNA through its 5′ region. Furthermore, VESTAR was also shown to interact with HuR, a known positive regulator for VEGFC mRNA stability (16), by its 3′end to enhance the interplay between VEGFC mRNA and HuR. Interestingly, the 5′ region of VESTAR alone was able to increase the VEGFC mRNA levels, indicating that besides HuR, there are other factors involved in the VESTAR-mediated VEGFC mRNA stability.

In summary, we here propose an lncRNA working model (Fig. 7I), in which a nuclear lncRNA, VESTAR, is exported to the cytoplasm in ESCC and promotes the ESCC lymphangiogenesis and lymphatic metastasis by directly interacting with VEGFC mRNA to increase its stability. Our findings demonstrate the significance of subcellular localization of lncRNAs during cancer progression, as well as the lncRNA-guided regulation of VEGFC mRNA stability with direct RNA-RNA interaction.

No disclosures were reported.

Y. Wang: Data curation, formal analysis, validation, investigation, methodology, writing–original draft. W. Zhang: Formal analysis, investigation, methodology. W. Liu: Validation, investigation. L. Huang: Validation, investigation. Y. Wang: Resources, methodology. D. Li: Resources, methodology. G. Wang: Data curation, software. Z. Zhao: Resources, methodology. X. Chi: Investigation. Y. Xue: Investigation. Y. Song: Resources, funding acquisition. X. Liu: Conceptualization, resources, supervision, writing–original draft, writing–review and editing. Q. Zhan: Conceptualization, resources, supervision, funding acquisition, writing–review and editing.

This work was supported by the National Natural Science Foundation of China (81830086 and 81988101 to Q. Zhan); Beijing Municipal Commission of Health and Family Planning Project (PXM2018_026279_000005 to Q. Zhan); Beijing Tianjin Hebei basic research cooperation project [19JCZDJC64500(Z) to Q. Zhan]; Guangdong Basic and Applied Basic Research Foundation (2019B030302012 to Q. Zhan); CAMS Innovation Fund for Medical Sciences (2019-I2M-5–081 to Q. Zhan); and CAMS Innovation Fund for Medical Sciences (CIFMS; 2016-I2M-1–001 to Y. Song). We thank Guo Cheng and Dekang Lv for help on bioinformatics analysis. The authors thank Liyan Shui, Fei Wang, and Xintao Mao for help on Western blotting and animal experiment. The results shown here are partly based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.