MicroRNA-138 suppresses ovarian cancer cell invasion and metastasis by targeting SOX4 and HIF-1α


  • Yu-Ming Yeh,

    1. Institute of Molecular and Genomic Medicine, National Health Research Institute, Taiwan, Republic of China
    2. Institute of Molecular Medicine, National Tsing Hua University, Hsinchu, Taiwan, Republic of China
    Search for more papers by this author
  • Chi-Mu Chuang,

    1. Department of Obstetrics and Gynecology, Taipei Veterans General Hospital, Taiwan, Republic of China
    2. Institute of Clinical Medicine, National Yang-Ming University, Taipei, Taiwan, Republic of China
    Search for more papers by this author
  • Kuan-Chong Chao,

    1. Department of Obstetrics and Gynecology, Taipei Veterans General Hospital, Taiwan, Republic of China
    Search for more papers by this author
  • Lu-Hai Wang

    Corresponding author
    1. Institute of Molecular Medicine, National Tsing Hua University, Hsinchu, Taiwan, Republic of China
    • Institute of Molecular and Genomic Medicine, National Health Research Institute, Taiwan, Republic of China
    Search for more papers by this author

Correspondence to: Lu-Hai Wang, Institute of Molecular and Genomic Medicine, National Health Research Institute, No. 35, Keyan Road, Zhunan, Miaoli County 35053, Taiwan, Republic of China, Tel.: +886-37-246-166 #35300, Fax: +886-37-585-242, E-mail: lu-hai.wang@nhri.org.tw


Metastasis is the major factor affecting patient survival in ovarian cancer. However, its molecular mechanisms remain unclear. Our study used isogenic pairs of low- and high-invasive ovarian cancer cell lines to demonstrate the downregulation of miRNA-138 in the highly invasive cells, and its functioning as an inhibitor of cell migration and invasion. An orthotopic xenograft mouse model further demonstrated that the expression of miRNA-138 inhibited ovarian cancer metastasis to other organs. Results indicated that miR-138 directly targeted SRY-related high mobility group box 4 (SOX4) and hypoxia-inducible factor-1α (HIF-1α), and overexpression of SOX4 and HIF-1α effectively reversed the miR-138-mediated suppression of cell invasion. Epidermal growth factor receptor acted as the downstream molecule of SOX4 by way of direct transcriptional control, whereas Slug was the downstream molecule of HIF-1α by way of proteasome-mediated degradation. Analysis of human ovarian tumors further revealed downregulation of miR-138 and upregulation of SOX4 in late-stage tumors. Patients with miR-138low/SOXhigh signature are predominant in late stage and tend to have malignant phenotypes including lymph nodes metastasis, larger ascites volume and higher tumor grade. Our study demonstrates the role and clinical relevance of miR-138 in ovarian cancer cell invasion and metastasis, providing a potential therapeutic strategy for suppression of ovarian cancer metastasis by targeting SOX4 and HIF-1α pathways.

Metastatic ovarian cancer is the deadliest among gynecologic malignancies, with an estimated 15,500 deaths in the United States in 2012, as reported by the National Cancer Institute. The overall 5-year survival rate is ∼33% when diagnosed at advanced stages; it is about 90% while the cancer is still confined to the ovary (stage I).[1] At late stages, tumor cells spread beyond the pelvic cavity and commonly undergo metastasis to the mesentery, omentum and diaphragm.[2] Metastasis to the pelvic and para-aortic lymph nodes may also occur. Elucidating the mechanisms underlying cancer metastasis will, thus, make significant contributions toward combating this disease. MicroRNAs (miRNAs), a family of small noncoding single-stranded RNAs, have recently been shown to play essential roles in cancer cell invasion and metastasis.[3] One miRNA can suppress multiple gene expressions by interacting with the 3′ nontranslated regions (3′UTRs) of its target mRNAs and promoting their degradation or translational suppression resulting in the modulation of those genes' expressions and functions.

A couple of recent studies have reported the role of miRNAs in modulating ovarian cancer cell invasion and metastasis. Cowden Dahl et al. showed that epidermal growth factor receptor (EGFR)-responsive miR-125a induced a mesenchymal-to-epithelial transition (MET) in ovarian cancer cells. MiR-125a directly targets ARID3B, which is overexpressed in serous ovarian cancer.[4] Corney et al. reported that p53-transactivated miR-34 was decreased in metastatic ovarian cancer cells. Overexpression of miR-34 reduced the MET protein levels, and suppressed cancer cell migration, invasion and proliferation in p53-null SKOV-3 cells.[5] Those studies provided initial clues for miRNAs in regulating ovarian cancer metastasis.

MicroRNA-138 is downregulated in different cancers including head and neck squamous cell carcinoma (HNSCC), aggressive papillary thyroid carcinoma and lung cancer tumors in never smokers.[6-8] A recent study identified that miR-138 suppressed cancer cell migration and invasion by targeting RhoC and ROCK2, two GTPase signaling-related molecules, in tongue squamous cell carcinoma (TSCC) cell lines.[9] The role of miR-138 in ovarian cancer cell invasion and metastasis, however, has not been reported.

Our study demonstrates that miR-138 suppresses ovarian cancer cell invasion and metastasis by targeting SRY-related high mobility group box 4 (SOX4) and hypoxia-inducible factor-1α (HIF-1α), and that miR-138low/SOX4high signature is associated with malignant phenotypes in ovarian cancer. Thus, our study identifies novel signaling pathways and molecules as potential prognostic biomarkers and targets for intervention of ovarian cancer metastasis.

Material and Methods

Cell culture

The human ovarian cancer cell lines SKOV-3 and TOV-112D were from ATCC, Manassas/VA. The A1847 line was obtained from Dr. Stuart Aaronson (Mount Sinai School of Medicine). All cell lines were cultured in dulbecco's modified eagle medium (DMEM) (GIBCO, Carlsbad/CA) supplemented with 10% fetal bovine serum (FBS; Biological Industries, Kibbutz Beit Haemek/Israel) at 37° in 5% CO2.

Human ovarian tumor samples

Tissue slides from 78 ovarian tumors were obtained from the Taipei Veterans General Hospital (TVGH) Department of Pathology under approved IRB protocol. Ovarian tumor samples were collected during debulking surgery and the pathologic data were recorded. Commercial tissue array slides (OV1005, US Biomax, Rockville/MD; CJ2, SUPER BIO CHIPS, Seoul/South Korea) were purchased as additional cohort of clinical samples. Expression profiles of miR-138, SOX4 and HIF-1α of these samples were analyzed by fluorescence in situ hybridization (FISH) and IF staining.

Cell migration/invasion assay and in vitro selection

Cell migration/invasion assay and in vitro selection were done as previously described[10] except 5 × 104 cells were seeded.

Tail-vein metastasis assay and in vivo invasion selection

Parental or invasion-selected cells (106 cells) were harvested and resuspended in 100 μl phosphate buffered saline (PBS). The cells were injected into the tail vein of 6- to 8-week-old CB17/lcr-Prkdc scid/Crl mice (BioLasco Taiwan, Taipei/Taiwan). All mice were anesthetized and dissected 4 weeks after injection. Various organs were harvested, fixed with formalin, sectioned and subjected to H&E staining. Five pictures were captured at 100× magnification from each H&E-stained slide, and the numbers of metastatic nodules per field were counted using ImageJ software. For in vivo selection of invasive cells, the unfixed metastatic nodules were harvested, trypsinized and cultured in DMEM supplemented with 10% FBS at 37°C.

Plasmid constructs

The pcDNA 6.2-GW/EmGFP-miR-negative control and pcDNA 6.2-GW/EmGFP-miR-138 were synthesized by GeneDireX, Vegas/NV. HIF-1α (dODD) cDNA (synthesized by GenScript, Piscataway/NJ) was ligated into the KpnI site of pcDNA3.1 (+). The cDNA of SOX4 was amplified using KOD FX (TOYOBO, Osaka/Japan) and the following primers: forward, 5′-CTATGGTACCTATGGTGCAGCAAACC-3′; reverse, 5′-GCGCTCTAGATCAGTAG-GTGAAAACCA-3′. The SOX4 cDNA was gel extracted and ligated between the KpnI and XbaI sites of pcDNA3.1 (+).

Transfection and generation of stable cell lines

Cells were harvested and plated onto six-well plates (Corning, Corning/NY) with 70% confluence overnight before transfection with Lipofectamin 2000 (Invitrogen, Carlsbad/CA). The miRNA precursor pre-miR™-138 (Applied Biosystems, Carlsbad/CA), anti-miR-138 (synthesized by GeneDireX, Vegas/NV) or siRNA was transfected for 48 hr under normoxic or hypoxic conditions at 37°C. For generation of miR-138 and miR-CL stable cell lines, SKOVI6iv cells were transfected with 2 μg of pcDNA 6.2-GW/EmGFP vectors expressing miR-138 or miR-scramble control, followed by 4 weeks of Blasticidin S HCl selection (20 μg/ml; Invitrogen, Carlsbad/CA). To generate double stable cell lines, the previously generated stable cell lines were further transfected with 2 μg of pcDNA 3.1(+) SOX4 or empty vector, followed by 4 weeks of G418 selection (800 μg/ml; Sigma, St. Louis/MO). Three individual clones were combined for animal experiment. A list of siRNAs is found in Supporting Information.

Orthotopic xenograft mouse model

The GFP-tagged stable cell lines were harvested and resuspended in 20 μl PBS containing 50% Matrigel (BD Biosciences, San Jose/CA). Cells were injected intrabursally into 8- to 10-week-old CB17/lcr-Prkdc scid/Crl mice (1 × 106 cells per mouse). All procedures were done under approved Institutional Animal Care and Use Committee protocols. The mice were sacrificed 4 weeks later. Incidences of metastases to various organs were measured by naked eye. Various organs harvested were homogenized with TissueLyser II (QIAGEN, Venlo/Netherlands) in Trizol reagent (Invitrogen, Carlsbad/CA). Metastases to different organs were quantified using quantitative real-time polymerase chain reaction (qRT-PCR) by determining the green fluorescent protein (GFP) expression levels normalized to that of β-actin.

Quantitative real-time PCR for genes and miRNAs

RNA was isolated with TRIzol (Invitrogen, Carslabd/CA) according the manufacturer's instructions and cDNA was generated using ReverTra Ace Set (PURIGO, Taipei/Taiwan) with Oligo-dT or corresponding micro-RNA RT primers, as listed in Supporting Information Table S1. The experiments were carried out with CFX96 Real-Time System (Bio-Rad, Hercules/CA), and analysis was conducted with ΔCT method. Expression of mRNA was normalized to β-actin. The KAPA SYBR FAST Universal qPCR Kit (KAPA Biosystems, Woburn/MA, KK4600) was used for gene detection. The KAPA PROBE FAST universal qPCR Kit (KK4701) and Universal ProbeLibrary Probe #21 (Roche Applied Science, Penzberg/Germany) were used for miRNA detection. Comprehensive lists of PCR primers are found in Supporting Information.

Western blot analysis

Western blot analysis was performed as previously described.[10] Briefly, cells were lysed in RIPA buffer (150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS and 50 mM Tris-HCl, pH 8.0) containing Complete and PhosStop tablets (Roche, Penzberg/Germany), separated by SDS-PAGE and immunoblotted with appropriate antibodies. Antibodies list is found in Supporting Information.

3′UTR luciferase reporter assay

Double-stranded oligonucleotides (∼60 nt) (+ strand and − strand) containing the wild-type or mutant sequence of the predicted miR-138-binding site in SOX4 or HIF-1α 3′UTR were synthesized and ligated between the XbaI and FseI restriction sites of pGL3-Control Vector (Promega, Madison/WI). The sequences of synthetic oligonucleotides are found in Supporting Information. SKOV-I6 cells were transiently transfected with the indicated molecules (0.25 μg for vectors and 25 nM for others) and pcDNA3.1/His/lacZ vector (0.25 μg). Luciferase assays were performed using the luciferase reporter assay system (Promega, Madison/WI) 48 hr after transfection. Luciferase activity was normalized by the β-galactosidase activity of the corresponding cell lysate.

Chromatin immunoprecipitation assay

The chromatin immunoprecipitation procedure was a modified version of the EZ-Magna ChIP A kit (Millipre, Billerica/MA) protocol using SOX4 antibody (Millipore, Billerica/MA, AB5803). Primer sequences are found in Supporting Information.

Fluorescence in situ hybridization, immunofluorescence and quantification

Commercial ovarian cancer tissue arrays CJ2 (SUPER BIO CHIPS, Seoul/South Korea), OV1005 (US Biomax, Rockville/MD) and human ovarian tumor slides from TVGH were used. Biotin-labeled locked nucleic acid (LNA)-modified RNA probes (synthesized by GeneDireX, Vegas/NV), directed against the full-length mature miR-138 sequence, were used for miRNA detection. Streptavidin-HRP IgG (PerkinElmer, Waltham/MA) and tyramide signal amplification reactions were used to detect the hybridized probes. Immunofluorescence (IF) was done using mouse monoclonal anti-HIF-1α (NB100–105, Norvus, Littleton, CO) and rabbit polyclonal anti-SOX4 (ab52043, Abcam, Cambridge/UK) antibodies. Alexa Fluor 594 goat anti-mouse or goat anti-rabbit IgG (H+L) (Invitrogen, Carlsbad/CA) were used as the secondary antibodies. The slides were mounted and images were captured using fluorescence microscopy. The intensities of miR-138 (green) and protein (red) signals in tumor regions were quantified using ImageJ software.


The selected invasive ovarian cancer cells exhibit an aggressive phenotype

We initially used in vitro Boyden chamber invasion selection to obtain highly invasive cancer cells. Four to six cycles of selection yielded the highly invasive ovarian cancer sublines from three corresponding parental lines: SKOV-3, TOV-112D and A1847. Cancer cell properties including cell morphology, migration/invasion ability, in vivo metastasis ability, epithelial-to-mesenchymal (EMT) markers expression and colony-forming ability were compared between the selected and parental cells (Supporting Information Figs. S1 and S2). The SKOV-I6 cells displayed loosened cell–cell adhesion and spindle-like shapes compared to the parental cells; both TOV-I4 and A1847-I4 cells exhibited more fibroblast-like phenotypes compared to the cuboid morphology of parental cells (Supporting Information Fig. S1A). The selected invasive cells showed significantly higher migration and invasion abilities than those of their corresponding parental cells (Supporting Information Figs. S1B and S1C). Examination of EMT markers in SKOV-3 and TOV-112D pairs revealed decreased E-cadherin, an epithelial marker, and increased β1-integrin, a mesenchymal marker, in the invasive cells, whereas others, including vimentin and β-catenin, remained unchanged (Supporting Information Fig. S2A). For soft agar colony formation assay, the selected invasive cells formed a significantly greater number of colonies than their parental counterparts (Supporting Information Fig. S2B). The SCID mice tail-vein xenograft model was used to assess metastatic ability. Supporting Information Figure S1D displays the representative pictures and results. All three invasive cells exhibited a higher metastatic ability. A1847-I4 cells formed ovarian, whereas SKOV-I6 and TOV-I4 cells formed lung metastases. Explanting the cancer cells from lung metastasis nodules in the SKOV-3 group produced the SKOViv cells; similarly, those from the SKOVI6 group produced the SKOV-I6iv cells (Supporting Information Fig. S3 and see Supporting Information for detailed description). The in vivo selected cells displayed significantly higher invasive abilities suggesting that the in vivo selection could impose more powerful selection for the invasive ability. Taken together, these results demonstrate that the Boyden chamber-selected invasive cells display greater oncogenic properties, especially the invasiveness and metastatic ability, than the parental cells.

MicroRNA-138 is downregulated in invasive cells and is a suppressor of ovarian cancer cell invasion and metastasis

To identify potential miRNAs regulating ovarian cancer cell invasion and metastasis, genome-wide miRNA array analyses between the low and high invasive cells were performed. MicroRNA-138 (miR-138), not yet reported in ovarian tumors, revealed a consistent trend of downregulation in the invasive cells (Supporting Information Table S1). Stem-loop qRT-PCR confirmed the reduced expression (Fig. 1a), suggesting a suppressive role of miR-138 in ovarian cancer cell migration and invasion. Overexpression of miR-138 decreased the cell migration ability, whereas anti-miR-138, an antagonist of miR-138, partially or fully reversed the miR-138-mediated effect (Fig. 1b). Moreover, transfecting anti-miR-138 in the parental SKOV-3 cells increased the invasive ability (Fig. 1c). MiR-138, therefore, functioned as a suppressor of ovarian cancer cell migration and invasion. Orthotopic xenograft mice model with GFP-tagged SKOV-I6iv cells stably expressing miR-CL or miR-138 vectors were used to evaluate the function of miR-138 in vivo. Cells stably expressing miR-138 and their invasive ability were examined in vitro first (Fig. 1d). Figure 1e shows the in vivo results. There was no significant difference in primary tumor weight between the miR-138-expressing and miR-CL-expressing groups (0.504 ± 0.176 g vs. 0.551 ± 0.145 g, p = 0.381). However, the miR-138-expressing group showed lower incidence of peritoneal metastasis and ascites formation than the miR-CL group (Fig. 1e). By comparing GFP level in various abdominal organs with quantitative RT-PCR, cancer cell micrometastasis can be examined. Mice injected with miR-138-expressing cells displayed greatly reduced peritoneal dissemination as reflected in significantly lower GFP level than the miR-CL group (Fig. 1e). These data indicate that miR-138 suppresses ovarian cancer cell metastasis without affecting primary tumor growth.

Figure 1.

MicroRNA-138 suppresses migration, invasion and metastasis of invasive ovarian cancer cells. (a) Stem-loop qRT-PCR of miR-138 expression from three isogenic highly invasive pairs normalized to U6. (parental vs. invasive), n = 3. (b) Migration assay was conducted for 8 hr with the selected highly invasive cells transiently transfected with 25 nM of the indicated molecules, n = 3. NC, pre-miR™ miRNA precursor negative control; miR-138, miRNA precursor pre-miR™-138; anti-miR-138, miR-138 antagonist. (c) Invasion assay for 20 hr was conducted with parental SKOV-3 cells transiently transfected with the indicated doses of miR-138 antagonist (anti-miR-138), n = 3. (d) Relative expression of miR-138 (left) and invasive ability (right) of the GFP-tagged SKOVI6iv stable cell lines expressing miRNA scramble control (miR-CL) and miR-138 (miR-138), n = 3. Invasion assay was assessed at 12 hr. (e) Quantitative RT-PCR of GFP expression level (left) and tumor incidence (right) in various organs as abundance of metastatic cells resulted from orthotopic implantation mouse model with GFP-tagged stable cell lines mentioned in (d). Expression of GFP was normalized to β-actin. Tumor incidence was examined by naked eyes. The data shown were obtained from six to eight mice. ▪, miR-CL; □, miR-138. Data were shown as mean ± SD (Student's t-test, *p < 0.05, **p < 0.005 and ***p < 0.0005). Representative pictures for migration and invasion assays are shown in Supporting Information Figure S8.

MicroRNA-138 suppressed ovarian cancer cell invasion and metastasis by targeting SOX4 and HIF-1α

To elucidate the miR-138 signaling pathways, we performed miRNA target gene prediction with PicTar, TargetScan and Miranda databases. The SOX4 and HIF-1α exhibited miR-138-binding sequences in their 3′UTR regions (Supporting Information Fig. S4). Quantitative RT-PCR and Western blotting analyses demonstrated that miR-138 downregulated SOX4 and HIF-1α mRNA and protein expression in the invasive cells (Figs. 2a2c). SOX4 and HIF-1α protein expression were upregulated in parental SKOV-3, TOV-112D and A1847 cells when expressing anti-miR-138 (Fig. 2d). Analyses using 3′UTR luciferase reporter plasmids containing the wild-type or mutant miR-138-binding sequences of SOX4 and HIF-1α 3′UTRs showed that miR-138 inhibited the luciferase activities of the SOX4 and the HIF-1α 3′UTR constructs, and that coexpression of anti-miR-138 partly or fully reversed this inhibition in SKOV-I6 cells (Fig. 2e). The mutation of the miR-138 sites in the 3′UTRs abrogated the repressions indicating that miR-138 directly targeted SOX4 and HIF-1α (Fig. 2e and Supporting Information Fig. S4). To assess the role of SOX4 and HIF-1α in ovarian cancer cell invasion, we used SOX4 or HIF-1α siRNAs and performed invasion assays. Under normoxic conditions, HIF-1α protein expression is extremely low; therefore, all HIF-1α-related assays were performed under hypoxic conditions. Knockdown of SOX4 or HIF-1α in SKOV-I6 cells markedly decreased invasive abilities (Figs. 3a and 3b), suggesting that they might act as downstream effectors of the miR-138-mediated effect. To test this hypothesis, functional rescue assay was performed by transiently expressing SOX4 or HIF-1α in combination with miR-138 in SKOV-I6 cells. Coexpression of SOX4 rescued the miR-138-mediated effect (Fig. 3c). Similarly, coexpression of HIF-1α (dODD), a mutant HIF-1α nonresponsive to normoxia-induced degradation,[11] rescued the miR-138-mediated inhibition of cell invasion under hypoxic conditions (Fig. 3d). These data suggest that the downregulation of SOX4 and HIF-1α is responsible, at least in part, for the miR-138-mediated suppression of ovarian cancer cell invasion.

Figure 2.

SOX4 and HIF-1α are the direct targets of miR-138. (a and b) Relative mRNA expression levels of SOX4 and HIF-1α were analyzed in the invasive cells transiently transfected with 25 nM of miRNA negative control (NC) or pre-miR-138 (miR-138) using qRT-PCR under (a) normoxic and (b) hypoxic conditions with β-actin as the internal control, n = 3. (c and d) Western blot analysis showing SOX4 and HIF-1α expression in (c) the invasive cells transiently transfected with 25 nM of indicated molecules (d) SKOV-3, TOV-112D and A1847 parental cells transiently transfected with indicated dose of miR-138 antagonist (anti-miR-138). Representative experiment of three repetitions. (e) 3′UTR luciferase reporter assays (top, SOX4; bottom, HIF-1α) were performed in SKOV-I6 cells transiently transfected with luciferase reporter vectors with wild-type (WT) or mutant (Mut) miR-138-binding sequences (also see Supporting Information Fig. S4). The luciferase reporter vectors (0.25 μg), β-galactosidase vector (0.25 μg) or 25 nM of indicated molecules were transfected in SKOV-I6 for 48 hr. The luciferase activity was normalized by β-galactosidase signal (O.D. 660 nm) and was arbitrarily assigned as 1 in the miRNA negative control groups, n = 3. Data were shown as mean ± SD (Student's t-test, *p < 0.05, **p < 0.005 and ***p < 0.0005).

Figure 3.

The SOX4 and HIF-1α transcriptional factors are responsible for miR-138-mediated suppression of cell invasion and metastasis. (a and b) Top panel: Immunoblotting of (a) SOX4 protein or (b) HIF-1α protein from SKOV-I6 cells transiently transfected with 25 nM of siRNA scramble control (siCL) or different siRNAs under (a) normoxic or (b) hypoxic conditions. Actin was used as the loading control. Histogram: Numbers of invaded SKOV-I6 cells transiently transfected with 25 nM of (a) SOX4 siRNA under normoxic conditions or (b) HIF-1α siRNA under hypoxic conditions, n = 3. (c and d) MiRNA negative control (NC) or pre-miR-138 (miR-138) transiently transfected (25 nM) SKOV-I6 cells were cotransfected with 0.5 μg of HIF-1α (dODD), a nondegradable form of HIF-1α, SOX4 or pcDNA 3.1(+) empty vector (Vector) for 48 hr before invasion assay under normoxic (SOX4) or hypoxic (HIF-1α) conditions for 16 hr, n = 3. (e) Quantitative RT-PCR of GFP expression level (left) and tumor incidence (right) as abundance of metastatic cancer cells in various organs following 28 days of orthotopic implantation with the indicated GFP-tagged double stable cell lines. Generation of the indicated double stable cell lines is described in Material and Methods section. The data were calculated from four mice each. Data were shown as mean ± SD (Student's t-test, *p < 0.05, **p < 0.005 and ***p < 0.0005). Representative pictures for invasion assays are shown in Supporting Information Figure S8.

Subsequently, we undertook functional rescue experiments with miR-138 and SOX4 double stable expressing cells in orthotopic xenograft mice model to test if miR-138/SOX4 signaling pathway functioned alike in vivo. Mice injected with miR-138+SOX4 double stable expressing SKOVI6iv cells showed more severe peritoneal dissemination than those injected with miR-138+control vector cells, suggesting that SOX4 effectively reversed miR-138-mediated effects in vivo (Fig. 3e). Supporting Information Figure S5 displays the relative expression of miR-138, SOX4 and HIF-1α in different organs from the orthotopic xenograft mice. As expected, most of miR-138-overexpressing metastatic tissues displayed lower SOX4 and HIF-1α expression, further confirmed our hypothesis.

Epithelial growth factor receptor and Slug are downstream mediators of SOX4 and HIF-1α, respectively

Possible downstream target gene(s) of transcriptional factors SOX4 and HIF-1α in miR-138-mediated signaling were further examined. Several previous studies have suggested that EGFR signaling pathway is able to promote cancer cell invasion.[12-14] The downstream targets of SOX4 are reported to include LCK, SEMA3C, EGFR and Tenascin C in breast and prostate cancers.[15, 16] Those observations implied that EGFR might be a downstream mediator of SOX4 in the miR-138 signaling pathway. EGFR mRNA and protein expression were examined following transient transfection of miR-138 or SOX4 siRNA in SKOV-I6 cells. Both EGFR mRNA and protein expression were reduced (Figs. 4a and 4b), suggesting that EGFR was a regulatory target gene of SOX4. Furthermore, we identified three potential SOX4-responsive elements (SREs) with the A/T A/T CAAAG motif[17] on EGFR promoter region and designed corresponding primers for chromatin immunoprecipitation (ChIP) assay. Results indicated a direct SOX4 binding at the predicted −720 bp binding element, whereas no binding at the other two sites of EGFR promoter (Fig. 4c). These data indicate that SOX4 binds to a specific promoter element of EGFR and directly activates its transcription.

Figure 4.

EGFR and Slug act as the downstream mediators of SOX4 and HIF-1α. (a and d) Relative mRNA expression of (a) SOX4/EGFR and (d) HIF-1α/Slug, detected by qRT-PCR in SKOV-I6 cells transiently transfected with 25 nM of the indicated molecules. For HIF-1α experiments, cells were incubated under hypoxia for 24 hr before RNA extraction, n = 3. (b and e) Immunoblotting of SOX4/EGFR and HIF-1α/Slug protein from SKOV-I6 cells transiently transfected with 25 nM of the indicated molecules. For HIF-1α experiments, cells were incubated under hypoxia for 24 hr before protein extraction. (c) Top: Schematic representation of EGFR promoter region with three potential SOX4-responsive elements (SRE). Bottom: Chromatin immunoprecipitation assay was performed using antibodies against SOX4 (α-SOX4) and three sets of PCR primers (−720, −1310 and −2357) in SKOV-I6 cells. Immunoprecipitation using rabbit IgG (α-IgG) and beads only (Beads) provided as negative controls. Total genomic DNA input (Input) was shown as the positive control. Data are representative of three independent experiments. (f) Protein expression level of HIF-1α and Slug was examined in SKOV-I6 cells transiently transfected with 25 nM of the indicated molecules for 24 hr by Western blot. After transfection, the cells were cultured with or without MG132 (5 μM) for 18 hr under hypoxic conditions. Data are representative of three independent experiments. (g and h) Functional link of SOX4/EGFR (left) and HIF-1α/Slug (right) pathways in SKOV-I6 cell invasion. EGFR siRNAs, Slug siRNAs or scramble siRNA control (siCL) transiently transfected SKOV-I6 cells were cotransfected with 0.5 μg of HIF-1α (dODD), SOX4 or pcDNA 3.1(+) empty vector (Vector) for 48 hr before invasion assay under normoxic (SOX4) or hypoxic (HIF-1α) conditions for 12 hr, n = 3. Data were shown as mean ± SD (Student's t-test, *p < 0.05, **p < 0.005 and ***p < 0.0005).

Several studies have linked HIF-1α- and EMT-related molecules, such as Twist, Snail and Slug in promoting cancer cell invasion.[18-20] Among the three molecules, miR-138 expression decreased only Slug in ovarian cancer cells under hypoxic conditions. The increase of Slug protein expression under hypoxia was attenuated by miR-138 (Supporting Information Fig. S6). Slug might, therefore, be a downstream mediator of miR-138/HIF-1α signaling pathway in ovarian cancer. We found that under hypoxic conditions, siHIF-1α and miR-138 downregulated Slug protein, but not its mRNA levels (Figs. 4d and 4e). Furthermore, treatment by MG132, a proteasome inhibitor, abrogated both miR-138- and siHIF-1α-inhibited Slug protein expression (Fig. 4f), suggesting that HIF-1α regulated Slug protein levels through proteasomal degradation.

To provide further functional link of SOX4/EGFR and HIF-1α/Slug pathways in ovarian cancer cell invasion, we silenced downstream mediators, EGFR and Slug, by siRNAs in the presence of their respective upstream regulators, SOX4 and HIF-1α, and performed cell invasion assay. Results showed that SOX4- and HIF-1α-mediated enhancement of invasion was abrogated by knocking down EGFR and Slug, respectively (Figs. 4g and 4h). In conclusion, the miR-138 represses ovarian cancer cell invasion by targeting SOX4 and HIF-1α, which in turn regulate EGFR and Slug, respectively; both molecules have been reported to play important roles in regulating ovarian cancer metastasis.[21-24]

MicroRNA-138low/SOX4high signature is a potential prognosis marker for ovarian cancer

Next, the clinical relevance of the miR-138 signaling pathways in human ovarian cancer was evaluated. This was done by analysis of the expression profiles of miR-138 and its target genes using FISH assay with LNA-modified miR-138 antisense probe and IF staining for SOX4 and HIF-1α, using TVGH human ovarian tumor specimens and commercial tissue array slides. The fluorescent intensities from the tumor regions were measured quantitatively. Analysis of FISH and IF staining revealed that miR-138 and SOX4 had opposite expression levels in early- and late-stage cancer, whereas level of HIF-1α did not change (p = 0.298, data not shown). Figure 5a displays the quantification and representative fluorescence images of miRNA-138 and SOX4. The late-stages tumors had decreased miR-138 and increased SOX4 expression. Summarized clinical data from 78 TVGH samples demonstrated a more malignant phenotype in miR-138 low-expression and SOX4 high-expression patients (Table 1). By further comparing clinical status between miR-138low/SOX4high and miR-138high/SOX4low signature-bearing samples, we found that patient cases with miR-138low/SOX4high signature have more ascites, poor differentiation histology and have higher frequency in advanced diseases. The tendency of harboring pelvic and para-aortic lymph node metastasis is also increased in patients with miR-138low/SOX4high expression pattern, demonstrating that miR-138low/SOX4high signature may be a poor prognosis marker for ovarian cancer (Figs. 5b and 5c).

Table 1. Clinical features between different expression status of miR-138 and SOX4 in 78 ovarian cancer samples from Taipei Veterans General Hospital
Low, n = 52High, n = 26p valueLow, n = 38High, n = 40p value
  1. a

    Fisher's exact test.

  2. b

    Student's t-test.

  3. c

    Some data are not available.

  4. Abbreviation: LN, lymph node.

Age (years)      
<654223 3233 
III/IV3410 1627 
Moderate2217 1920 
Poor244 1413 
Ave. ascitesc (ml)1075.5318.90.043b554.21111.50.11b
Pelvic LN metastasisc (yes/no)16/324/210.168a8/2711/250.594a
Para-aortic LN metastasisc (yes/no)12/362/210.125a4/3110/260.135a
Serous178 1212 
Endometroid1010 137 
Clear cell105 79 
Mucinous32 41 
Others80 17 
Figure 5.

MiR-138low/SOX4high signature represents malignant phenotypes in human ovarian cancer samples. (a) Statistic results of miR-138 and SOX4 expression in early (stage I/II)- and late (stage III/IV)-stage ovarian cancer samples, as detected by FISH and IF staining using TVGH samples (top) and commercial ovarian cancer tissue arrays (middle). Fluorescein (miR-138), Alexa Fluor 594 (SOX4 or HIF-1α) and DAPI (nuclear staining) fluorescent images of each sample were captured using fluorescence microscopy with the same exposure time (1/10 sec). The cutoff values for high and low expression were determined using the mean ± SD values of each marker. The statistical significance was determined using χ2 test. Representative images for FISH and IF staining in different stages are shown (bottom). (b) Clinical features between TVGH samples with miR-138low/SOX4high (n = 24) and miR-138high/SOX4low (n = 11) signatures were compared including distribution among early and late stages (left), tumor differentiation status (middle) and ascites volume (right). Data were shown as mean ± SE. Fisher's exact test and Student's t-test were used for statistical analysis. (c) Number of cases harbored pelvic (left) or para-aortic (right) lymph node metastasis between patients with miR-138low/SOX4high and miR-138high/SOX4low signatures. (d) Working hypothesis for miR-138-mediated suppression of ovarian cancer cell migration/invasion and metastasis.


MicroRNA-138 has been known to be downregulated in head and neck, thyroid and lung cancers.[6-9, 25] A number of studies have implicated the functions of miR-138 in sensitizing cells to therapeutic agents, promoting apoptosis and inhibiting migration/invasion.[6, 25-27] In our study, we found miR-138 has significant effects on cell migration, invasion and metastasis, but not on proliferation and paclitaxel resistance (data not shown). Our study for the first time demonstrates the functional role of miR-138 and its mediated signaling pathways in ovarian cancer cell invasion and metastasis.

The transcriptional regulatory mechanism for most miRNAs remains unclear. Based on miRBase database,[28] miR-138 is an intergenic miRNA and located at two regions on human genome, Chr3: 44155704–44155802[+] and Chr16: 5689430–56892513[+]. By using PROMO 3.0,[29] an online transcriptional factor binding sites database, we tried to find potential transcriptional factors that regulate miR-138 transcription. Within the potential promoter regions (2000 bp upstream and downstream of miR-138), we found that the transcriptional factor glucocorticoid receptor-α (GR-α) has the most predictive binding sites on both miR-138 potential promoters. The GR, residing in cytosol, is activated by its ligand and translocated into nucleus as a transcriptional factor.[30] Recently, Law et al.[31] demonstrated that glucocorticoid dexamethasone inhibited triple-negative breast cancer cells invasion through inducing E-cadherin localization to the plasma membrane, suggesting a suppressive role of cancer cell invasion for glucocorticoids. Thus, it is plausible that glucocorticoids regulate miR-138 expression through GR-α to suppress cancer cell invasion. Further study is needed to test this hypothesis.

The transcription factor SOX4 is critical for development in vertebrates and is overexpressed in lung, bladder, prostate, endometrial and liver cancers.[15, 32-34] Our data indicate the increased SOX4 expression in advanced ovarian cancers, consistent with a previous study in breast cancer.[35] Our results identify SOX4 as a direct target of miR-138, and suggest it as a prognostic marker and potential therapeutic target for ovarian cancer. The −720 SOX4-binding site on EGFR promoter identified in our study is not identical to that in the intron 1 of EGFR reported previously,[16] suggesting the possibility that SOX4 regulates EGFR transcription by way of multiple cis-binding elements. Further study is needed to determine if those sites act cooperatively or not. As a recognized protooncogene, EGFR is known to be overexpressed in 4–22% of ovarian cancers, with its expression correlating with poor prognosis.[36, 37] The functional linkage of SOX4 to EGFR provides an explanation for EGFR overexpression in advanced ovarian cancer, where SOX4 is also elevated.

When induced under hypoxia conditions, HIF-1α promotes tumor metastasis by inducing systematic angiogenesis and cellular EMT.[38] Several studies have reported that HIF-1α can directly promote EMT-related genes expression,[18-20, 39] providing a possible explanation for hypoxia-induced EMT in cancer cells. Our data, however, imply an indirect regulatory mechanism between HIF-1α and Slug, namely, via proteasome-mediated degradation for hypoxia-induced EMT. Vernon and LaBonne showed that the F-box protein Ppa regulated Slug protein stability in developing embryos.[40] The Ppa protein bound to and promoted the ubiquitin-mediated proteasomal degradation of Slug. In another study, the suppression of the interaction between FBXL14, the F-box E3 ubiquitin ligase and Snail led to the inhibition of ubiquitin-mediated proteasome degradation of Snail under hypoxia.[41] It is, therefore, possible that some F-box protein or E3 ligase-related molecules are involved in the miR-138/HIF-1α/Slug regulatory mechanisms in ovarian cancer cells.

In the presence of proteasome inhibitor MG132, the miR-138-mediated inhibition of both HIF-1a and Slug was reversed despite of our results showing that HIF-1a mRNA was downregulated by miR-138. The likely explanation is that blockage of proteasome pathway of protein degradation allowed accumulation of HIF-1a translated from even a reduced level of its mRNA. In other words, inhibition of protein degradation overcame the effect of miR-138-mediated decrease of HIF-1a mRNA.

Lubec et al. demonstrated that SOX4 was upregulated when they compared the transcriptional factor profiles between the brains of normoxic and perinatally asphyxiated rats.[42] By comparing the transcriptional profiles of cultured neural progenitor cells isolated from the subventricular zone tissue between mice with or without middle cerebral artery occlusion surgery, Liu et al. demonstrated the upregulation of SOX4 in those cells with ischemic stroke, suggesting SOX4 might be a hypoxia-inducible gene during neurogenesis.[43] However, there was no consistent trend of SOX4 expression in various cancer cell lines under hypoxia,[44] suggesting a much more complex relationship between SOX4 and HIF-1α in cancer biology. There was neither significant difference in SOX4 protein expression in SKOV-3 cells under normoxia and hypoxia for 24 hr (data not shown), nor a correlation between SOX4 and HIF-1α in human ovarian tumor samples (p > 0.05), suggesting an independent relationship between SOX4 and HIF-1α in miR-138-mediated signaling pathways.

In summary, our study demonstrates that miR-138 regulates ovarian cancer cell invasion and metastasis by targeting SOX4 and HIF-1α oncogenic transcriptional factors. The corresponding downstream mediators include EGFR and Slug, which are well-known progression factors and prognostic markers in many cancers (Fig. 5d). The miR-138 signaling pathways could, thus, play important roles in ovarian cancer progression and metastasis. The expressions of miR-138 and SOX4 in late-stage ovarian cancer support this notion. These molecules and pathways, especially the miR-138/SOX4 signature, could represent prognosis markers and potential targets for therapeutic intervention.


The authors thank pathology core lab of National Health Research Institutes for all of the H&E staining. They thank Dr. Tien, Industrial Technology Research Institute, for the microRNA array analysis.