A cohort of 60 primary ovarian GCT (44 samples of benign and 6 samples of malignant tumor samples. Ten cases of ovarian cyst) was obtained from US Biomax. Detection probes for miR-126 (miRCURY LNA miRNA) were purchased from Exiqon (Denmark). The assays were performed according to Exiqon's standard protocol of miRNA FISH. To avoid experimental bias the samples were blinded before analysis.

Two human GCT lines, the adult GCT cell line KGN (14), the juvenile GCT cell line Cov434 (15) and normal human GC line SVOG (16) were kindly provided by Prof Zijiang Chen (Shandong University). KGN cells were cultured in DMEM/F12 media (Gibco) with 10% FBS and 1% penicillin/streptomycin. Cov434 cells were cultured in DMEM media (Gibco) with 10% FBS, 1% penicillin/streptomycin and 1% NEAA. As controls, HEK293T cells were purchased from Invitrogen and cultured in DMEM media (Gibco) with 10% fetal bovine serum and 1% penicillin/streptomycin. A 37°C/5% CO₂ humidified incubator was used for cell culture.

miR-126 mimics/inhibitor and scrambled control were obtained from GenePharma. A lipofectamine transfection reagent (RNAiMAX, Life Technologies) was used for transfection of miR-126 mimics/inhibitor.

EGFL7 short hairpin (sh)RNAs were designed using the webpage http://sirna.wi.mit.edu/home.php. The vector pLVTHM (Addgene No. 12247) was used for shRNA cloning. Lentivirus packaging was carried out according to the manufacturers' protocols. Green fluorescent protein (GFP) sorting was used to isolate successfully transfected cells (17). MDH1-PGK-EGFP plasmid is a retroviral construct for miR-126 expression (Addgene No. 11375). Quantitative RT-PCR and western blot were used to validate knockdown effect.

Total RNA was extracted with Trizol reagent (Invitrogen) according to standard protocol. Concentration and quality of all RNA samples were evaluated by Nanodrop 2000 (Thermo). MasterMix kit (Takara) and TaqMan reverse transcription kit (Life Technology, USA) were used for mRNA and miRNA reverse transcription. Universal SYBR Green Master mix (Applied Biosystems) and TaqMan specific microRNA probe (Life technology) were used for qPCR assays. All qPCR were performed by StepOnePlus real-time PCR system (Applied Biosystems). GAPDH and U6 snoRNA were used for normalization of mRNA and miRNA.

Cell-counting kit-8 was used for a quantification of proliferation.

Alexa Fluor®488 annexin V/Dead Cell Apoptosis Kit (Invitrogen, CA, USA) was used for apoptosis evaluation according to the manufacturer's guidelines.

Transwell plates with inserts were used for invasion and migration assays (Costar) Matrigel (BD) diluted to 0.1 mg/ml with DMEM media was used to coat the upper chamber for the invasion assay. KGN cells (2 × 10⁴ cell per chamber) were seeded into the upper chamber with FBS-free medium and 10% FBS medium was added into the lower wells. After 24 h, the transwell chambers were washed and stained with 1% crystal violet.

Supernatants of KGN cells were collected after different treatments. Human Estron competitive ELISA kit (Invitrogen) and Estradiol human ELISA kit (Thermofisher) were used for Estrone and Estrodial detection.

Paraffin embedded tissue slides that obtained from KGN cells-formed tumors were blocked in 5% goat serum for half hour, then incubated in primary anti-PCNA (Santa Cruz), P-Akt (Cell Signaling) and EGFL7 (Santa Cruz) antibodies at 4°C for 12 h (18).

The total RNA was collected from KGN cells that transfected with miR-126 mimics and negative control. The RNA quality was measured by NanoDrop2000 spectrophotometer. RNA-seq and subsequent transcriptome analysis were carried out by GROKEN Bioscience (China) according to standard protocol.

Cells were lysed in SDS buffer (100 mM Tris-Cl (pH 6.8), 4% SDS (sodium dodecyl sulfate), 0.2% bromophenol blue, 20% glycerol, 200 mM β-mercaptoethanol). BCA assay kit (Thermofisher) was used for protein measurement. The primary antibodies used were listed as follows: EGFL7 (Santa Cruz), PI3K (Abcam), AKT (Abcam), S6K (Cell signaling), mTOR (Abcam), p-AKT (Abcam), FOXO1 (Abcam). GAPDH (Cell signaling) and β-ACTIN (Santa Cruz) were used as loading controls.

HEK293T cells were co-transfected with miR-126 mimics, pmirGLO vector (Promega) tagged with EGFL7 3′UTR by using Lipofectamine 2000 (Invitrogen). The empty pmirGLO plasmid was used as negative control. Dual-Luciferase Reporter Assay System (Promega) was used for Firefly/Renilla luciferase activities.

The related in vivo experiments were approved by the Animal Research Committee at the LASEC, Chinese University of Hong Kong. All nude mice used in this experiment were obtained from LASEC. For the ectopic tumor formation model (19), Control KGN cells and miR-126 ovexpressing KGN cells were injected subcutaneously (5 × 10⁶ cells/mice) in 4–6 weeks female nude mice (8 mice for each group). All mice were sacrificed after 4 weeks. Then cancerous GC-formed tumor were isolated for following analysis.

Prism (GraphPad) was used for statistical tests. Unpaired, two-tailed Student's t-tests was used for parametric results. One-way ANOVA with Tukey's post-hoc tests was used to further examine pairwise differences. Two-sided Mann-Whitney test was performed for non-parametric results. A level of P < 0.05, was used to designate significant differences. ^*, ^**, and ^*** indicate P < 0.05, P < 0.01, and P < 0.001, respectively.

The expression level of miR-126 in clinical specimens obtained from a biobank of a GCT cohort was analyzed using fluorescence in situ hybridization (FISH). The endogenous level of miR-126 in was examined in 60 GCT patients. Our results showed that miR-126 expression was significantly lower in both malignant and benign GCT tissues than in non-cancerous ovarian cyst material (Figure 1A). To investigate the effect of miR-126 on the oncogenesis of cancerous GCs in vitro, miRNA mimics and inhibitors were transfected into two GCT patient-derived cancer cell lines, KGN cells (derived from an adult GCT patient) and COV434 cells (derived from a juvenile GCT patient) (Supplementary Figure 1). Proliferation of these two cancerous GCs was significantly repressed (Figures 1B,C) while, simultaneously, apoptosis was induced by miR-126 mimics in cancerous GCs (Figure 1E). By contrast, miR-126 didn't affect proliferation of a normal GC line, SVOG (14) (Figure 1D), suggesting that the proliferation-inhibitory effect of miR-126 was more evident in cancerous GCs. Cell migration and invasion are the two characteristics of metastasis. Using in vitro transwell assays, our results showed that miR-126 could inhibit both migration and invasion of KGN cells (Figures 1F,G). In addition, estrone and estradiol are highly secreted by GC in GCT patients. The abnormal production of estrone and estradiol promotes GCT development. ELISA results indicated that miR-126 could repress secretion of estrone and estradiol from KGN cells (Figures 1H,I). Chemoresistance of KGN cells to anti-cancer drugs was measured by apoptosis assay after treating cells with Taxol, a common anti-cancer drug. The results revealed that the Taxol-induced cells death is, at least partially, miR-126 dependent (Figure 1J). In addition, similar in vitro results were obtained using Cov434 cells (Supplementary Figure 1). These in vitro experiments suggest a tumor suppressor role for miR-126 in GCT, which might account for its low expression in the malignant form of GCT.

We reasoned that the “tumor suppressor” role of miR-126 in cancerous GCs might be coupled with gene(s) or signaling pathway(s) that change the cell behavior. To unlock the underlying molecular mechanism at the transcriptional level, RNA-seq was performed to compare the difference in gene expression between control and miR-126 overexpressing cancerous GCs (Supplementary Figure 2). Results from KEGG pathway analysis revealed that 11 of the top 20 miR-126-associated pathways were cancer-associated (Supplementary Figure 2). Some previously reported GCT-related pathways, such as the estrogen signaling and the PI3K-AKT pathways, were also found in the list (Supplementary Figure 2). Next, we attempted to identify the key functional targets of miR-126 in cancerous GCs. By comparing all down-regulated genes from the RNA-seq analysis and with the help of software prediction, 10 candidate genes were identified (Figure 2A). Interestingly, in line with literal reports, five of the candidate genes have been proven of playing significant functions in different types of cancer (Figure 2B). To confirm the putative functional targets of miR-126 specifically in cancerous GCs, shRNAs were employed to knock down each individual gene among the 10 candidates (Supplementary Figure 3). After validation of the knockdown efficiency in KGN cells, CCK-8 assay was performed to examine the knockdown effect on the proliferation of cancerous GCs (Supplementary Figure 3). While the knockdown of 9 genes had negligible effects on KGN cell proliferation it was significantly reduced upon EGFL7 knockdown. This observation was consistent with the result of miR-126 overexpression that led to proliferation arrest in cancerous GCs (Figures 1B,C). Interestingly, genomic analysis indicated that EGFL7 was the target gene that contained the miR-126 binding site (Figure 2D). We also performed Western blotting to confirm the RNA-seq results, concluding that miR-126 could repress EGFL7 expression in cancerous GCs at the protein level (Figure 2E). To determine whether miR-126 directly binds to the EGFL7 3′UTR and suppresses its translation (Figure 2C), a luciferase reporter assay demonstrated that miR-126 binding to the EGFL7 3′UTR could indeed decrease the reporter activity (Figure 2F). Additionally, we also demonstrated in the PI3K-AKT pathway, a miR-126 associated, oncogenic pathway in GCT, was altered following miR-126 ovexpression (Figure 2G). A series of essential proteins in PI3K-AKT pathway, including PI3K, p-AKT, TSC1, TSC2, mTOR, FOXO1, and S6K, were modulated by miR-126 (Figure 2G). In summary, through transcriptional analyses and biochemical and cellular assays, we established a functional link between EGFL7 and miR-126 in the regulation of GC tumorigenesis, potentially through the PI3K-AKT pathway.

To investigate whether miR-126 inhibits GCT growth in vivo, an ectopic tumor formation mice model was used. A KGN cell line that stably over-expressed miR-126 was generated (Supplementary Figure 1) and control or miR-126 overexpressing KGN cells were subcutaneously injected to the flank back of nude mice. After 4 weeks, the cancerous GCs formed solid tumors. The size of tumors formed by miR-126 overexpressing KGN cells was significantly smaller than that of tumors formed by untransfected control cells (Figures 3A,B,D). Quantitative PCR results confirmed that miR-126 expression was significantly induced in miR-126 overexpressing KGN cells-formed tumor (Figure 3C). Immunohistochemical analysis revealed that PCNA, a proliferation marker, was significantly decreased in miR-126 overexpressing tumor cells (Figure 3E). Consistently, we also observed suppression of EGFL7 in these tumor tissues (Figure 3E). In addition, we also investigated the expression of p-AKT, the critical node of the AKT pathway. Result from IHC showed that this protein was significantly downregulated in miR-126 overexpressing KGN tumors (total AKT didn't change), indicating that AKT pathway was altered by miR-126 in vivo. To summarize, our results demonstrated a “tumor suppressor” role for miR-126 in GCT. EGFL7 appeared as a relevant target of miR-126 in mediating the effect of the miRNA, in part by changing the AKT pathway.

We found that a stable knockdown of EGFL7 could inhibit proliferation of cancerous GCs (Supplementary Figure 3). This anti-mitotic effect was similar to that of miR-126 overexpression (Figure 4B). To investigate whether EGFL7 is a critical target for mediating miR-126 function in GC tumorigenesis, we examined EGFL7 expression in GCT tissues. Compared to ovarian cyst and benign GCT tissues, EGFL7 was highly expressed in malignant GCT tissues (Figure 4A), which is diametrically opposed to the expression of miR-126 in GCT tissues. To demonstrate a role of EGFL7, a lentivirus-based stable knockdown of EGFL7 was performed. Following the sorting of positively transduced cells (Figure 4B), the expression of EGFL7 was stably decreased, as validated by Western blot (Figure 4B). In these EGFL7 deficient KGN cells, proliferation and migration were found to be decreased (Figures 4C,D). In addition, there were more apoptotic cells in the EGFL7 stable knockdown KGN cells (Figure 4E). Similar as in miR-126 overexpressing cancerous GC, the AKT pathway was also repressed in EGFL7 KD cancerous GC (Figure 4F), further confirming that the effect of the miR-126/EGFL7 regulatory axis on cancerous GC is mediated by AKT pathway. Taken together, these results demonstrated that EGFL7 depletion imitated the effect of miR-126 overexpression indicating EGFL7 as a functional target of miR-126 in cancerous GCs.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.