Gynecologic Oncology 132 (2014) 730–738

Contents lists available at ScienceDirect

Gynecologic Oncology

j ourna l homepage: www.e lsev ie r .com/ locate /ygyno

The role ofmiR-200a in vasculogenicmimicry and its clinical significancein ovarian cancer

Qingmin Sun a,b,⁎,1, Xi Zou c,1, Ting Zhang b, Jian Shen a, Yongxiang Yin b, Jingying Xiang b,⁎⁎a Department of Pharmacy, Affiliated Hospital of Nanjing University of Chinese Medicine, Nanjing 210029, Chinab Department of Pharmacy, Affiliated Wuxi Hospital for Maternal and Child Health Care of Nanjing Medical University, Wuxi 214002, Chinac Department of Oncology, Affiliated Hospital of Nanjing University of Chinese Medicine, Nanjing 210029, China

H I G H L I G H T S

• MiR-200a inhibits vasculogenic mimicry by directly regulating EphA2 in ovarian cancer.• Vasculogenic mimicry, miR-200a and EphA2 play key roles in the progression and prognosis of ovarian cancer patients.

⁎ Correspondence to: Q. Sun, Affiliated Hospital ofMedicine, Nanjing 210029, China. Fax: +86 25 86619821⁎⁎ Correspondence to: J. Xiang, Affiliated Wuxi HospitaCare of Nanjing Medical University, Wuxi 214002, China.

E-mail addresses: qingminsun@gmail.com (Q. Sun), w1 Contributed equally.

0090-8258/$ – see front matter © 2014 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.ygyno.2014.01.047

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 31 December 2013Accepted 27 January 2014Available online 4 February 2014

Keywords:miR-200aVasculogenic mimicryEphA2Ovarian cancer

Objective. Vasculogenic mimicry (VM) indicates that aggressive cancer cells can form de novo vascular net-works and provide a perfusion pathway for rapidly growing tumors. MiR-200a has been reported significantlyderegulated in ovarian cancer. However,miR-200a regulation of VM and its clinical significance in ovarian cancerremain not elucidated.

Methods. In this study, we identified the VM structure by CD34-PAS staining in ovarian cancer tissue. MiR-200a and protein expression was tested by quantitative RT-PCR and western blot. Bioinformatics prediction, lu-ciferase assay and intervention experiments were employed to identify the target of miR-200a.

Results.We certified the VM structure in ovarian cancer, and found that the VMpositive rate was significantlyassociated with tumor grade, stage and metastasis. Further study showed that miR-200a expression levels were

significantly lower in VM positive ovarian cancer. In addition, our results suggested that miR-200a inhibited VMby negatively regulated EphA2 expression. Consistently, the inverse correlation of miR-200a and EphA2 has alsobeen found in ovarian cancer patients. Moreover, the expression ofmiR-200a/EphA2was significantly associatedwith patient's clinicopathological parameter, such as tumor stage and metastases. Kaplan–Meier curves con-firmed that the patientswith lowmiR-200a expression and/or VMpositive had a significantly shorter overall sur-vival.

Conclusions. Our research demonstrates that VM, miR-200a and EphA2 play key roles in the progression andprognosis of ovarian cancer, and for the first time suggests that miR-200a inhibits VM by directly regulatingEphA2. Therefore, we might have identified a genetic mechanism underlying the involvement of miR-200a inovarian cancer VM.

© 2014 Elsevier Inc. All rights reserved.

Introduction

Ovarian cancer remains the major cause of gynecologic cancer mor-tality. In 2013, it was estimated that about 22,240 women were diag-nosed with ovarian cancer and 14,030 died of the disease in theUnited States [1]. And most ovarian cancer new cases were usually

Nanjing University of Chinese.l for Maternal and Child HealthFax: +86 510 82725161.xxjy89@163.com (J. Xiang).

ghts reserved.

diagnosed in advanced stages. In the last decades, although the diagno-sis and treatment have improved a lot, the 5-year relative survival rateof ovarian cancer is only 43% [1]. Given the highmortality rate of ovariancancer, it is a critical need to explore the molecular pathogenesis anddevelop the new relevant biomarker to increase specificity or sensitivityfor early diagnosis and prognosis.

Tumor cell vasculogenic mimicry (VM), was first mentionedby Maniotis et al. in 1999, which indicated that the aggressive cancercells can form de novo vascular networks and provide a perfusionpathway for rapidly growing tumors, and the critical VM-modulatinggenes are correlated with vascular (EphA2, VE-cadherin), embryonicand/or stem cell (Nodal, Notch4), and hypoxia-related (hypoxia-induc-ible factor, Twist1) signaling pathways [2,3]. In recent years, VM

Table 1Relationship between VM, miR-200a, EphA2 and clinicopathological parameters in ovarian cancer specimens.

Characteristics No. VM P miR-200a P EphA2 P

53 Positive Negative Low High Low High

Age (Median) n n n n n nb53 years 26 7 19 0.184 12 14 0.678 14 12 0.494≥53 years 27 12 15 14 13 12 15

Histological typeSerous 19 7 12 0.962 11 8 0.483 5 14 0.005*Endometrioid 18 6 12 9 9 8 10Mucous or others 16 6 10 6 10 13 3

Tumor sizeb10 cm 42 16 26 0.505 19 23 0.277 22 20 0.344≥10 cm 11 3 8 7 4 4 7

Tumor gradeG1 10 1 9 0.025* 3 7 0.089 6 4 0.394G2 24 7 17 10 14 13 11G3 19 11 8 13 6 7 12

FIGO stageI–II 32 7 25 0.009* 10 22 0.001* 22 10 0.000*III–IV 21 12 9 16 5 4 17

MetastasisYes 29 14 15 0.038* 18 11 0.037* 10 19 0.020*No 24 5 19 8 16 16 8

AscitesYes 36 14 22 0.502 21 15 0.049* 16 20 0.328No 17 5 12 5 12 10 7

CA 125b300 U/ml 35 11 24 0.349 19 16 0.288 19 16 0.288≥300 U/ml 18 8 10 7 11 7 11

Statistical analyses were performed by the χ2 test. *P b 0.05 was considered significant.

731Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

has been described in several types of aggressive tumors, includingcarcinomas of breast, ovary, lung, prostate, bladder, and kidney [4].Our previous research also suggested that the VM structure was foundin highly aggressive uvealmelanoma, and found that genistein could in-hibit VM formation in vivo and in vitro [5]. In addition, more and morestudies suggested that the VM acted as a potential marker for the prog-nosis in tumors [6,7]. Although theVMhas been certified in ovarian can-cer cells, few studies focused on the microRNA regulation of VM and itsclinical significance and prognosis in ovarian cancer patients.

MicroRNAs, an abundant class of small noncoding RNAs, elicittheir regulatory effects by imperfectly binding to the 3′ untranslatedregion of target mRNA [8]. Several studies showed that miRNA de-regulation is associated with the initiation and progression ofhuman cancer [9]. Recent study indicated that some miRNAs couldbe considered as potential candidates as novel biomarkers in ovariancancer [10]. Our latest research also revealed that oncogenicmiR-27aplays an important role in ovarian cancer cell growth and metastasis[11]. In addition, many studies showed that miR-200a was signifi-cantly deregulated in ovarian cancer tissues [12,13]. And recentstudy showed that EphA2 is a target gene of miR-200a in breastcancer [14]. EphA2 is a 130-kDa transmembrane protein which wasfound in human epithelial cells [15]. The activation of EphA2receptor can mediate endothelial cell network formation andVEGF-dependent endothelial cell migration and survival, suggestinga potential role as a regulator of vasculogenesis and VM [16–18].Moreover, Landen et al. revealed that EphA2 is overexpressed inmany human cancers, and treatment with EphA2-targeting agentsis an effective approach for ovarian cancer therapy [19]. However,no study reports the relationship betweenmiR-200a and VM in ovar-ian cancer.

To expand our knowledge regarding VM and biological function ofmiR-200a, the current study was designed to determine whether themiR-200a was associated with VM in ovarian cancer. Meanwhile, weexplore the mechanistic roles of miR-200a and hypothetic target geneEphA2 in VM, and investigate the significance of miR-200a and/or VMin ovarian cancer patients.

Materials and methods

Cell lines and tissue samples

The human ovarian cancer cell lines SKOV3 and A2780 weregenerously providedbyProfessor Binghua Jiang (Cancer Center, NanjingMedical University, China) and maintained in DMEM supplementedwith 10% fetal bovine serum. A total of 53 formalin-fixed paraffin-embedded (FFPE) ovarian cancer tissue samples were obtained fromthe Department of Pathology, Affiliated Wuxi Hospital for Maternaland Child Health Care of Nanjing Medical University in 2004 to 2011.The clinical and pathological characteristics of the study participantswere summarized in Table 1. All the patients in the studywere followedup for survival analysis until death or until July 2012. The follow-upduration ranged from 11 to 98 months (mean, 56.79 ± 3.23 months).All patients had never been given chemotherapy or biotherapy before,and the professional pathologist scored tumor grade and stage accord-ing to the criteria from the International Federation of Gynecologistsand Obstetricians (FIGO). The study was approved by the NanjingMed-ical University Affiliated Hospital Ethics Committee and informed con-sent was obtained from all participants.

Three-dimensional culture and CD34-PAS dual staining

Ovarian cancer cell VM was observed by 3-D culture containingMatrigel (BD Biosciences), as described previously [5]. Briefly, 24-wellculture plates were coated withMatrigel, and then allowed to polymer-ize at 37 °C for 30 min. After a wash with PBS, 2.5 × 105 cells/ml wereplated onto the surface of Matrigel at 1 ml/well, and incubated for48 h. The tube-like structures were measured using phase contrastmicroscopy and photographed. In FFPE tissues, immunohistochemicaland PAS stain was used to detect the VM. Immunohistochemistry wasfirstly conducted with endothelium marker CD34 (1:50 dilution,Zhong ShanGoldenbridge, Beijng) to identify endothelium. The sectionswith CD34 stain were washedwith runningwater andwere deoxidizedwith 0.5% periodic acid solution. Finally, these sections were treated

Fig. 1. Vasculogenic mimicry (VM) structure and miR-200a expression in ovarian cancer. (A) Representative VM channels (thick arrows) lined with tumor cells contained red blood cellsand PAS-positive substances formed a basementmembrane-like structure. (B) Typical endothelium-dependent vessels (thin arrows), thosewere positive for CD34 immunohistochemisty.All sections were stained by CD34 and PAS. (C) Relative highly aggressive ovarian cancer SKOV-3 cells displayed the presence of tubular network and channel formation in the 3Dmatrix(Matrigel) after 48 h incubation, (D) whereas it was not notable in poorly invasive A2780 cells. (E) MiR-200a expression in ovarian cancer tissue samples with or without VM.(F) Transfection efficiency in ovarian cancer SKOV3 cells. The cells were treated with miR-200a mimic, mimic-negative control, inhibitor and inhibitor-negative control for 48 h. Theexpression of miR-200a was measured by TaqMan microRNA assays. Fold change was calculated by 2−ΔCt (ΔCt = Ct miR-200a − Ct miR-U6). ***P b 0.001.

732 Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

with Schiff solution for 15 min. The evaluation of VMwas performed asfollows [5,20]: The CD34-PAS sections were viewed with a light micro-scope at ×400. The patterned channels lined by a PAS-positive materialbut not CD34-positive endothelial cells were defined as VM. In addition,immunohistochemistry was also used to detect the expression ofEphA2. Briefly, the FFPE tissue sections were placed on slides,deparaffinized, rehydrated through graded alcohol, and quenched in3% hydrogen peroxide and then were incubated overnight at 4 °C withpolyclonal EphA2 antibody (1:100 dilution; Santa Cruz, CA, USA). Forquantification, the immunoreactivity was calculated based on the stain-ing intensity and percentage of the positively labeled cells (b25%, 25%–75%, and N75%). All sectionswere evaluated in a blindedmanner by twoinvestigators.

Quantitative RT-PCR

Total RNA was extracted from these FFPE tissues as previouslydescribed in our recent study [11]. TaqMan microRNA assays (AppliedBiosystems Inc.) were used to quantify the relative expression of miR-200a. Small nuclear RNA U6 was treated as normalization control.EphA2 and GAPDH mRNA levels were assessed by SYBR Green

quantitative PCR. Forward (F) and reverse (R) primers were used as fol-lows: EphA2: GGGACCTGATGCAGAACATC and AGTTGGTGCGGAGCCAGT; GAPDH: TGTTCGACAGTCAGCCGC and GGTGTCTGAGCGATGTGGC.All real-time amplifications were measured in triplicate and performedwith the ABI Prism 7300 sequence detection system. The fold-change ofmiR-200a and EphA2 mRNA was calculated using the 2−ΔCT method.

Cell transfection and invasion assay

To determine the function of miR-200a on cell VM and invasion, theovarian cancer cells were transfected with miR-200a mimic, mimic-negative control, inhibitor and inhibitor-negative control (GenePharma,Shanghai, China) by Lipofectamine 2000. Cell invasion assay wasmeasured with cultrex-24-well membrane invasion chambers. Briefly,40 μl of coating solution was coated in each well. After transfec-tion, 1 × 104 cells were seeded into the top chamber of each transwell.Complete medium containing 10% FBS filled in the bottom wells. After48 h, the non-invasive cells were removed from the top of the mem-brane. Then the cells with invasion were fixed with methanol andstained by 0.1% crystal violet solution.

Fig. 2. The effect of miR-200amimic and inhibitor on VMand cell invasion of ovarian cancer cells. (A) Comparedwith the control group, themiR-200amimic could significantly inhibit theVM formation in SKOV3 cells.Moreover, themiR-200a inhibitor also could increase the number of VM intersectionwhenSKOV3 cellswere cultured in the 3Dmatrix for 72 h. (B) Transwellassay showed that ectopic expression of miR-200a significantly suppressed the invasion capacity of SKOV3 cells. Meanwhile, the data suggested that inhibited expression of miR-200acould promote the invasion of A2780 cells. *P b 0.05, **P b 0.01, ***P b 0.001.

733Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

Plasmid construction and luciferase assays

To determine the target of miR-200a, PicTar (http://pictar.mdc-berlin.de/), TargetScan (http://targetscan.org/) and miRBase (http://mirbase.org/) were firstly employed, and the result showed that theEphA2 gene was considered to be a putative target. Furthermore, thecDNA of the EphA2 3′UTR fragment and EphA2 3′UTR with mutatedseed sites was digested by Nhe I and Xho I, and then was subclonedinto the same site of the pmirGLO dual-luciferase miRNA target expres-sion vector (Promega) to construct the pmirGLO-Wt-EphA2-3′UTR andMut-EphA2-3′UTR vector. Next, SKOV3 cells were added to 24-wellplates and immediately transfected with miR-200a mimic or mimic-negative control for 24 h, and then transfectedwithwild-type ormutantreporter vectors using Lipofectamine 2000. After 24 h, cells werewashedin PBS, lysed at room temperature for 15 min in 100 μl of Passive LysisBuffer (Promega), and 20 μl of the lysate was used in a dual-luciferaseassay (Promega) in aMonolight 2010 luminometer (Analytical Lumines-cence Laboratory; San Diego, CA). Results were normalized to Renilla.Additionally, a full-length human EphA2 cDNA was obtained from Prof.Haruhiko Sugimura (Hamamatsu University School of Medicine, Japan).

Western blot analysis

The cells were lysed in 1% Triton X-100 lysis buffer. Protein was sep-arated by 10% SDS-PAGE and transferred to the PVDF membrane. Themembranes were incubated with rabbit anti-EphA2 antibody (1:200)or rabbit anti-GAPDH antibody. The primary antibodies were examinedby HRP-conjugated anti-rabbit secondary anti-body. Protein bandswere visualized using an enhanced chemiluminescence system andquantitated by Quantity One software (Bio-Rad, Hercules, CA, USA).

Statistical analysis

The difference between the two groups was determined by usingtwo-tailed Student's t test. The association between VM formation,miR-200a/EphA2 expression and clinicopathologic features was ana-lyzed using χ2 test. Correlation between the expression levels ofEphA2mRNAandmiR-200awas analyzed using the Pearson correlationcoefficient. Survival curves were plotted by the Kaplan–Meier analysis,and the log-rank test was performed for significant differences. Data

Fig. 3. (A) The region represents the alignment of potential miR-200a binding sites inthe 3′UTR of human EphA2 mRNA as predicted by TargetScan, miRBase and Pictar.(B) Wt-EphA2-3′UTR represents the plasmid with the full length EphA2 3′UTR andMut-EphA2-3′UTR means the plasmid with the same UTR mutated in the miR-200apairing sequences. Luciferase reporter assay showed the suppression of miR-200aon Wt-EphA2-3′UTR luciferase activity in SKOV3 cells after co-transfection for 48 h.Bars indicate the firefly luciferase activity normalized to Renilla. *P b 0.05.

734 Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

represents mean ± SD. A P value of b0.05 was considered statisticallysignificant. All the analyses were carried out with the SPSS 15.0.

Results

Vasculogenic mimicry formation in ovarian cancer

As summarized in Table 1, we found that 19 (35.85%) samples havetheVMstructure by CD34-PAS dual staining (Fig. 1A, B).Meanwhile, thepatterned networks were also observed in ovarian cancer SKOV-3 cellscultured in the 3D matrix, whereas this was not notable for A2780cells (Fig. 1C, D). To investigate the clinical significance of VM in ovariancancer, we combined clinicopathological parameters for further analy-sis. The result showed that the VM positive rate was significantlydifferent with respect to tumor grade (P = 0.025). In addition, VMchannels were most frequently observed in high FIGO stage (22.64%VS 13.21%, P = 0.009), and the VM positive rate correlated withtumor metastasis significantly (26.42% VS 9.43%, P = 0.038). Therewas no difference in other demographic characteristics including histo-logical type, tumor size and ascites (Table 1).

MiR-200a is downregulated in VM positive ovarian cancer tissue

To evaluate the role of miR-200a in ovarian cancer VM, the ex-pression of miR-200a in 53 ovarian cancer tissue samples was ana-lyzed by real-time PCR. After performed, patients were divided intotwo groups based on VM. We found that the expression levels ofmiR-200a in VM positive ovarian cancer were significantly lowerthan those seen in VM negative ovarian tissues (P = 0.011)(Fig. 1E), which demonstrated that the expression of miR-200aplays an important role in ovarian cancer VM. Moreover, the patientswere subdivided into two groups bymedian expression ofmiR-200a:high miR-200a expressers (2−△Ct ≥ 12.623) and low miR-200a ex-pressers(2−△C b 12.623). As presented in Table 1, the low miR-200a expressers were 18.87% (10 of 53) of FIGO stage I–II tumorsand 30.19% (16 of 53) of FIGO stage III–IV tumors. This differencewas statistically significant (P = 0.001). In addition, it is observedthat the low expression of miR-200a was significantly associated

with tumor metastasis (P = 0.037). Interestingly, we also foundthat the low expression of miR-200a correlated significantly withascites (P = 0.049). However, we found no significant associationbetween the expression of miR-200a and histological type, tumorsize or tumor grade, respectively (Table 1). These results indicatedthat miR-200a plays a key role in the progression of ovarian cancer.

MiR-200a inhibits vasculogenic mimicry and invasion of ovarian cancer

Considering the importance of miR-200a in ovarian cancer, we in-vestigate whether miR-200a influences the VM formation in SKOV3cells. By gain-of-loss function experiments (Fig. 1F), we found that VMwas significantly affected when SKOV3 cells were transfected withmiR-200a mimic or inhibitor (Fig. 2A). To determine the function ofmiR-200a on cell invasion, ovarian cancer cells (SKOV3, A2780) weretransfected with miR-200a mimic or inhibitor, and then evaluated intranswell assay. As shown in Fig. 2B, the invasion result displayed an in-hibitory effect of miR-200a in transfected SKOV3 cells compared withcontrol cells, but only had little effect on miR-200a inhibitor (data wasnot shown).Moreover, in A2780 cells,we found thatmiR-200a inhibitorcould promote the invasion of cells (Fig. 2B). But there was no availableresult about miR-200a mimic in A2780 cells because of A2780 cells'poor invasion. These results suggested that ectopic expression of miR-200a repressed the invasive potential of ovarian cancer cells.

MiR-200a downregulates EphA2 expression by targeting its 3′UTR

In order to further explore the molecular mechanism by whichmiR-200a regulates VM and invasion of ovarian cancer, we employedalgorithm programs to predict the putative target genes. Amongthese genes, EphA2 was of special interest, since it has been pro-posed that EphA2 is an important target gene for miR-200a in breastcancer [14]. However, there is no further fluorescent report assay toapprove the direct relationship between miR-200a and EphA2 in anycells. Moreover, it has been indicated that EphA2 is critical to VM [2].Therefore, we assumed that the increased expression of miR-200aresulted in the low levels of EphA2. As shown in Fig. 3A, there is aconserved target site for miR-200a in the positions 754–761 ofEphA2 3′UTR. In addition, we confirmed the binding of miR-200a tothe 3′UTR of human EphA2 using a luciferase reporter assay. Theresults showed that miR-200a significantly suppressed the luciferaseactivity in SKOV3 cells co-transfected with miR-200a and Wt-EphA2-3′UTR. However, the fluorescence intensity was unaffectedwhen Mut-EphA2-3′UTR reporter vector and miR-200a were co-transfected (Fig. 3B). These results showed that miR-200a could di-rectly bind to EphA2 and could regulate its expression in post-transcriptional levels.

MiR-200a inhibits vasculogenic mimicry by negatively regulating EphA2

To further confirm the function of miR-200a as an EphA2 regulator,we next examined levels of EphA2 by transfecting the ovarian cancercell (SKOV3, A2780) with miR-200a and anti-miR-200a. As shown inFig. 4A, we found that EphA2 mRNA levels were reduced to 66% afterSKOV3 cell transfection of miR-200a mimic for 24 h, suggesting thatmiR-200a regulates endogenous EphA2 mRNA levels partly throughthe mechanism of mRNA degradation. Similarly, the definitive connec-tion between EphA2 and miR-200a was also observed in protein levels(Fig. 4B). Meanwhile, poorly invasive A2780 cells carrying low levelsof EphA2 were transfected with miR-200a mimic or inhibitor. Com-paredwith the control, EphA2 protein expression levels were decreased~50% after transfected with miR-200a mimic for 48 h, but there was nosignificant effect on mRNA levels (Fig. 4C, D). Furthermore, in an at-tempt to test whether EphA2 as the direct functional mediator of miR-200a inhibits VM, we co-transfected miR-200a mimic and EphA2 ex-pression vector to ovarian cancer SKOV-3 cells. The result showed that

Fig. 4.MiR-200a negatively regulates EphA2 mRNA and protein expression. (A) In SKOV3 cells, miR-200a negatively regulated EphA2 mRNA levels after transfection of miR-200a mimicfor 24 h. However, there was no significant effect on transfection of miR-200a inhibitor. (B) In protein levels, miR-200a could significantly regulate EphA2 expression after transfection ofmiR-200amimic or inhibitor for 48 h. (C, D)Western blotting results showed thatmiR-200a significantly inhibited EphA2 protein expression levels after transfection ofmiR-200amimic inA2780 cells for 48 h. *P b 0.05, **P b 0.01.

735Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

reconstitution of EphA2 overrode the effect ofmiR-200a on inhibition ofVM (Supplementary Fig. S1A). Therefore, combining the above results,our finding indicated that VM formation was dependent on the expres-sion of EphA2.Moreover, our results also showed that overexpression ofEphA2 abrogated the reduction of invasion caused by ectopic expres-sion of miR-200a in ovarian cancer SKOV3 cells, and the invasion abilitywas significantly increased when A2780 cells were transfected withmiR-200a inhibitor plus EphA2 expression vector (SupplementaryFig. S1B and C). Collectively, these results indicated that miR-200ainhibited VM and invasion by negatively regulating the expression ofEphA2.

EphA2 expression in clinical specimens and its reverse correlation withmiR-200a

Next, we attempted to determine whether EphA2 expression levelsare associated with ovarian cancer patients' clinicopathological param-eters. By real time PCR assay, we demonstrated that overexpression ofEphA2 was significantly associated with ovarian cancer FIGO stage(Table 1, P b 0.001). We also observed that EphA2 expression was in-creased in ovarian cancer with metastases (P = 0.020). To examine

the clinical relevance of our findings, we analyzed the correlation be-tween EphA2 levels and miR-200a expression in the same patients. Asexpected, we found that EphA2 mRNA expression inversely correlatedwith the expression of miR-200a (r=−0.375, P= 0.006; Supplemen-tary Fig. S2A). Meanwhile, representative samples with different levelsof miR-200a were chosen to perform immunohistochemisty of EphA2.As shown in Supplementary Fig. S2B and C, the inverse correlationwas also confirmed in protein levels. It suggests that EphA2was a directand functional target of miR-200a.

MiR-200a, EphA2 and VM as predictors in ovarian cancer patient survival

At last, we sought to test the prognostic value of miR-200a, EphA2levels and VM in 53 patients with ovarian cancer. Log-rank test andKaplan–Meier curves were employed to estimate the association. Asshown in Fig. 5A, the patients with low miR-200a expression had asignificantly poor overall survival compared to patients with highmiR-200a expression (P = 0.025, log rank = 5.031). Although it is notnotable, the expression of EphA2 mRNA also showed a correlationwith patient overall survival (Fig. 5B; P = 0.053, log rank = 3.736). Inaddition, we compared the survival rate between the VM and non-VM

Fig. 5. Kaplan–Meier analysis for overall survival in 53 patients with ovarian cancer according to miR-200a expression (A), EphA2 levels (B), VM (C) and miR-200a/VM (D). The survivalcurves were analyzed by the log rank test.

736 Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

groups and found that the patients with VMhad a shorter survival com-pared to patients without VM (Fig. 5C; P = 0.019, log rank = 5.484).The mean survival time was 72 months for patients with VM and88 months for patients who had no VM. Moreover, patients with lowlevels of miR-200a in the VM positive group showed shorter survivaltime than others (Fig. 5D; P = 0.036, log rank = 4.399).

Discussion

Our researchpresents an excitingnovelfindingonmiR-200a expres-sion associated with ovarian cancer VM. Here we described the identifi-cation and characterization of VM, miR-200a and EphA2 expressionin ovarian cancer, as well as clinicopathological parameters and ovariancancer survival. Additionally, we indicated a correlation between miR-200a and target gene EphA2, and found that miR-200a inhibited VMby negatively regulating EphA2. Therefore, we might have identified agenetic mechanism underlying the involvement of miR-200a in ovariancancer VM.

Accumulating evidence has suggested that several types of aggres-sive tumors depended on VM to form the vascular networks to supporttumor cell growth and metastasis [4,21]. In the present study, we firstexamined VM in 53 FFPE ovarian cancer tissue samples, of which

35.85% samples present the VM structure. Similarly, the highly aggres-sive ovarian cancer SKOV3 cells have also been shown to generatepatterned vascular channels when seeded in the 3D matrix. In the pre-vious study, it has also been shown that ovarian cancer cells haveplasticity to form VM [22]. Importantly, our study further found thatVM was significantly different with respect to tumor grade, and amore VM positive rate was observed in high FIGO stage of ovarian can-cer. Moreover, a significant association was reported between VM andtumor metastasis (Table 1). Recent studies also confirmed that VMwas involved in the lymph node metastasis, clinical stage and differen-tiation of malignant tumors including non-small cell lung cancer andhepatocellular carcinoma [6,21]. In addition, it has also been shownthat VM was associated with poor prognosis in colorectal cancer, glio-mas as well as ovarian cancer [23–25]. Our findings are consistentwith the above-mentioned results; we confirmed that patients withVM positive ovarian cancer had a shorter survival time. Although oursamples are relatively small, combined with previous observations, itis biologically plausible that VM formation is related to clinical stage,metastasis and survival, and plays important roles in the developmentand prognosis of ovarian cancer.

Numerous studies have reported that the endothelial cell relatedgene coding for VE-cadherin, EphA2, laminin5γ2, and VEGFR-2 are

737Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

critical for VM formation [3,26]. However, few researches focused onthe relationship between miRNA and VM. Here, we attempted to ex-plore the possibility of using the newly identified family of genes,miRNA for the regulation of VM. After literature review, miR-200a wasof special interest, because many studies found that miR-200a was sig-nificantly deregulated in ovarian cancer tissues [12,13]. Moreover, it hasbeen described that loss of expression of miR-200a could play a criticalrole in the repression of E-cadherin in ovarian cancer stem cells [27]. Tocertify the importance of miR-200a in ovarian cancer, we compared theexpression levels in the VM group and the VM negative group. Interest-ingly, we found that the expression levels ofmiR-200awere significant-ly lower in the VM positive group. The result revealed that miR-200amight play a role in ovarian cancer VM. Given the characteristic ofmiR-200a, we then investigated the relative contribution of miR-200ain ovarian cancer VM and invasion. As shown in Fig. 2, we found that ec-topic expression of miR-200a not only inhibited VM, but also displayedthe repressive effect on invasion. This finding is similar to the results ob-served by Wu et al.; they also demonstrated that miR-200a inhibitedovarian cancer stem cell migration and invasion by targeting ZEB2 [27].

Combining the clinicopathological parameters, we further foundthat relatively low miR-200a expressers are more frequently observedin III–IV stages and metastasis tumors. These results indicated thatmiR-200a might contribute to ovarian cancer VM, invasion, and metas-tasis. In addition, Hu et al. revealed that overexpression of miR-200awas associated with a better prognosis in patients with advanced ovar-ian cancer [28]. In agreement with the finding, our Kaplan–Meiersurvival curves also showed that lowmiR-200a expression had a signif-icantly poor survival time.

Based on the definition of miRNA, it can negatively regulate post-transcriptional gene expression by directly cleaving target mRNA or byinhibiting their translation [8]. Therefore, we explored the target geneof miR-200a by bioinformatics. To date, several miR-200a target geneshave been confirmed in different cancer types, such as Keap1/YAP1 inbreast cancer, ZEB2 in ovarian cancer, and ZEB1/ZEB2 in gastric adeno-carcinoma [27,29–31]. Our study found that EphA2 was assumed asthe target gene of miR-200a by the online miRNA target prediction pro-grams. Moreover, our fluorescent report assay revealed the relationshipbetweenmiR-200a and EphA2. The result suggested that miR-200a canbind directly to the EphA2 3′UTR. Furthermore, gain–loss of function as-says confirmed the post-transcriptional regulation of EphA2 by miR-200a in ovarian cancer cell lines. Pearson correlation analysis also re-vealed that EphA2 levels inversely correlated with miR-200a levels. Inaddition, the inhibitory effect by miR-200a on VM or invasion was re-versed partially during transfection with the EphA2 expression vector.Collectively, our study clearly demonstrated that EphA2 is a key down-stream regulator of miR-200a.

According to previous reports, EphA2 acted in a coordinatedmanneras a key regulatory element in the process of VM [32,33]. In our presentstudy, we found that the levels of EphA2 were overexpressed in aggres-sive ovarian cancer cells SKOV-3 which have the ability of VM; this re-sult consists with the previous study [34]. These findings suggestedthe importance of EphA2 in VM formation. Similarly, Lin et al. provedthat EphA2 overexpression is associated with angiogenesis in ovariancancer [35]. Moreover, it has been suggested that EphA2 overexpressionwas significantly associated with high tumor grade, advanced stage ofdisease in ovarian carcinoma [36]. In this study, our findings were sim-ilar to the abovementioned results, that patient with overexpression ofEphA2 had a significant association with FIGO stage and metastases ofovarian cancer. Additionally, several studies have shown that highEphA2 expression was associated significantly with shorter survival[37,38]. Our Kaplan–Meier curves also showed the tendency thatEphA2 overexpression was correlated with poor prognosis for ovariancancer patients although it is not notable. It may be that the samplesize of the present study was relatively small. Nevertheless, our obser-vation provides valuable evidence that EphA2 is emerging as a predictorof survival in ovarian cancer.

In conclusion, we demonstrate that VM, miR-200a and EphA2 playkey roles in the progression and prognosis of ovarian cancer. Moreover,we are the first to provide evidence that miR-200a inhibits VM by di-rectly regulating EphA2 expression in ovarian cancer. These findingshelp us to better understand the function of miR-200a and its targetgene EphA2 in ovarian cancer cells and provide evidences that miR-200a might be a candidate for treatment of ovarian cancer. Therefore,we expect that ourfindingswill provide novel insights into the develop-ment of new biomarkers and new therapeutic miRNA approaches forovarian cancer.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.ygyno.2014.01.047.

Conflict of interest statement

The authors declare no conflict of interest.

Acknowledgments

We thank Prof. Haruhiko Sugimura for his generous gift of the ex-pression construct of full-length EphA2 (pAlterMax-EphA2), and Prof.Binghua Jiang for providing human ovarian cancer cells. This studywas supported by the National Natural Science Foundation of China(Nos. 81001444, 81202043) and the Science & Technology Develop-ment Project of Nanjing, China (Nos. 201201055, 2012YX001).

References

[1] Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer J Clin2013;63:11–30.

[2] Maniotis AJ, Folberg R, Hess A, Seftor EA, Gardner LM, Pe'er J, et al. Vascular channelformation by humanmelanoma cells in vivo and in vitro: vasculogenic mimicry. AmJ Pathol 1999;155:739–52.

[3] Kirschmann DA, Seftor EA, Hardy KM, Seftor RE, Hendrix MJ. Molecular pathways:vasculogenic mimicry in tumor cells: diagnostic and therapeutic implications. ClinCancer Res 2012;18:2726–32.

[4] Seftor RE, Hess AR, Seftor EA, Kirschmann DA, Hardy KM, Margaryan NV, et al.Tumor cell vasculogenic mimicry: from controversy to therapeutic promise. Am JPathol 2012;181:1115–25.

[5] Cong R, Sun Q, Yang L, Gu H, Zeng Y, Wang B. Effect of genistein on vasculogenicmimicry formation by human uveal melanoma cells. J Exp Clin Cancer Res2009;28:124.

[6] Wu S, Yu L, Wang D, Zhou L, Cheng Z, Chai D, et al. Aberrant expression of CD133 innon-small cell lung cancer and its relationship to vasculogenic mimicry. BMC Cancer2012;12:535.

[7] Liu R, Yang K,Meng C, Zhang Z, Xu Y. Vasculogenicmimicry is amarker of poor prog-nosis in prostate cancer. Cancer Biol Ther 2012;13:527–33.

[8] Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell2004;116:281–97.

[9] Esquela-Kerscher A, Slack FJ. Oncomirs—microRNAs with a role in cancer. Nat RevCancer 2006;6:259–69.

[10] Chung YW, Bae HS, Song JY, Lee JK, Lee NW, Kim T, et al. Detection of microRNA asnovel biomarkers of epithelial ovarian cancer from the serum of ovarian cancer pa-tient. Int J Gynecol Cancer 2013;23:673–9.

[11] Xu L, Xiang J, Shen J, Zou X, Zhai S, Yin Y, et al. Oncogenic MicroRNA-27a is a targetfor genistein in ovarian cancer cells. Anticancer Agents Med Chem2013;13:1126–32.

[12] Iorio MV, Visone R, Di Leva G, Donati V, Petrocca F, Casalini P, et al. MicroRNA signa-tures in human ovarian cancer. Cancer Res 2007;67:8699–707.

[13] Chao A, Lin CY, Lee YS, Tsai CL, Wei PC, Hsueh S, et al. Regulation of ovarian cancerprogression by microRNA-187 through targeting disabled homolog-2. Oncogene2012;31:764–75.

[14] Aydogdu E, Katchy A, Tsouko E, Lin CY, Haldosen LA, Helguero L, et al. MicroRNA-regulated gene networks during mammary cell differentiation are associated withbreast cancer. Carcinogenesis 2012;33:1502–11.

[15] Walker-Daniels J, Hess AR, HendrixMJ, Kinch MS. Differential regulation of EphA2 innormal and malignant cells. Am J Pathol 2003;162:1037–42.

[16] Cheng N, Brantley DM, Liu H, Lin Q, Enriquez M, Gale N, et al. Blockade of EphA re-ceptor tyrosine kinase activation inhibits vascular endothelial cell growth factor-induced angiogenesis. Mol Cancer Res 2002;1:2–11.

[17] Brantley-Sieders DM, Fang WB, Hicks DJ, Zhuang G, Shyr Y, Chen J. Impaired tumormicroenvironment in EphA2-deficient mice inhibits tumor angiogenesis and meta-static progression. FASEB J 2005;19:1884–6.

[18] Ojima T, Takagi H, Suzuma K, Oh H, Suzuma I, Ohashi H, et al. EphrinA1 inhibits vas-cular endothelial growth factor-induced intracellular signaling and suppresses reti-nal neovascularization and blood-retinal barrier breakdown. Am J Pathol2006;168:331–9.

[19] Landen CN, Kinch MS, Sood AK. EphA2 as a target for ovarian cancer therapy. ExpertOpin Ther Targets 2005;9:1179–87.

738 Q. Sun et al. / Gynecologic Oncology 132 (2014) 730–738

[20] Itzhaki O, Greenberg E, Shalmon B, Kubi A, Treves AJ, Shapira-Frommer R, et al. Nic-otinamide inhibits vasculogenic mimicry, an alternative vascularization pathway ob-served in highly aggressive melanoma. PLoS One 2013;8:e57160.

[21] Sun B, Zhang S, Zhang D, Du J, Guo H, Zhao X, et al. Vasculogenic mimicry is associ-ated with high tumor grade, invasion and metastasis, and short survival in patientswith hepatocellular carcinoma. Oncol Rep 2006;16:693–8.

[22] SuM, Feng YJ, Yao LQ, ChengMJ, Xu CJ, Huang Y, et al. Plasticity of ovarian cancer cellSKOV3ip and vasculogenic mimicry in vivo. Int J Gynecol Cancer 2008;18:476–86.

[23] Liu XM, Zhang QP, Mu YG, Zhang XH, Sai K, Pang JC, et al. Clinical significance ofvasculogenic mimicry in human gliomas. J Neurooncol 2011;105:173–9.

[24] Baeten CI, Hillen F, Pauwels P, de Bruine AP, Baeten CG. Prognostic role ofvasculogenic mimicry in colorectal cancer. Dis Colon Rectum 2009;52:2028–35.

[25] Gao Y, Zhao XL, Gu Q, Wang JY, Zhang SW, Zhang DF, et al. Correlation ofvasculogenic mimicry with clinicopathologic features and prognosis of ovarian car-cinoma. Zhonghua Bing Li Xue Za Zhi 2009;38:585–9.

[26] Yao X, Ping Y, Liu Y, Chen K, Yoshimura T, LiuM, et al. Vascular endothelial growth fac-tor receptor 2 (VEGFR-2) plays a key role in vasculogenic mimicry formation, neovas-cularization and tumor initiation by glioma stem-like cells. PLoS One 2013;8:e57188.

[27] Wu Q, Guo R, Lin M, Zhou B, Wang Y. MicroRNA-200a inhibits CD133/1+ ovariancancer stem cells migration and invasion by targeting E-cadherin repressor ZEB2.Gynecol Oncol 2011;122:149–54.

[28] Hu X, Macdonald DM, Huettner PC, Feng Z, El Naqa IM, Schwarz JK, et al. A miR-200microRNA cluster as prognostic marker in advanced ovarian cancer. Gynecol Oncol2009;114:457–64.

[29] Eades G, Yang M, Yao Y, Zhang Y, Zhou Q. miR-200a regulates Nrf2 activation bytargeting Keap1 mRNA in breast cancer cells. J Biol Chem 2011;286:40725–33.

[30] Cong N, Du P, Zhang A, Shen F, Su J, Pu P, et al. Downregulated microRNA-200apromotes EMT and tumor growth through the wnt/beta-catenin pathway bytargeting the E-cadherin repressors ZEB1/ZEB2 in gastric adenocarcinoma. OncolRep 2013;29:1579–87.

[31] Yu SJ, Hu JY, Kuang XY, Luo JM, Hou YF, Di GH, et al. MicroRNA-200a promotesanoikis resistance and metastasis by targeting YAP1 in human breast cancer. ClinCancer Res 2013;19:1389–99.

[32] Hess AR, Seftor EA, Gruman LM, Kinch MS, Seftor RE, Hendrix MJ. VE-cadherinregulates EphA2 in aggressive melanoma cells through a novel signaling pathway:implications for vasculogenic mimicry. Cancer Biol Ther 2006;5:228–33.

[33] Wu N, Zhao X, Liu M, Liu H, Yao W, Zhang Y, et al. Role of microRNA-26b in gliomadevelopment and its mediated regulation on EphA2. PLoS One 2011;6:e16264.

[34] Lu C, ShahzadMM,Wang H, Landen CN, Kim SW, Allen J, et al. EphA2 overexpressionpromotes ovarian cancer growth. Cancer Biol Ther 2008;7:1098–103.

[35] Lin YG, Han LY, Kamat AA, Merritt WM, Landen CN, Deavers MT, et al. EphA2overexpression is associated with angiogenesis in ovarian cancer. Cancer2007;109:332–40.

[36] Thaker PH, Deavers M, Celestino J, Thornton A, Fletcher MS, Landen CN, et al. EphA2expression is associated with aggressive features in ovarian carcinoma. Clin CancerRes 2004;10:5145–50.

[37] Merritt WM, Thaker PH, Landen Jr CN, Deavers MT, Fletcher MS, Lin YG, et al. Anal-ysis of EphA2 expression and mutant p53 in ovarian carcinoma. Cancer Biol Ther2006;5:1357–60.

[38] Merritt WM, Kamat AA, Hwang JY, Bottsford-Miller J, Lu C, Lin YG, et al. Clinical andbiological impact of EphA2 overexpression and angiogenesis in endometrial cancer.Cancer Biol Ther 2010;10:1306–14.