HMGA2 gene is a promising target for ovarian cancer silencing therapy



Ovarian cancer is one of the most lethal gynecological malignancies and the small success rate of routine therapeutic methods justifies efforts to develop new approaches. Evaluation of targets for effective inhibition of ovarian cancer cell growth should precipitate clinical application of gene silencing therapy. In our previous work, we showed upregulation of HMGA2 gene expression as a result of Ras-induced rat ovarian surface epithelial cell transformation. This gene codes the HMGA2 protein, a member of the high-mobility group AT-hook (HMGA) family of nonhistone chromatin proteins. Genome-wide studies revealed upregulation of the HMGA2 gene in human ovarian carcinomas. Herein we have evaluated over-expression of the HMGA2 gene, relevant to ovarian cancer, in subsets of human specimens and cell lines by in situ RNA hybridization and RT-PCR. Transient silencing of HMGA2 gene by means of siRNA inhibited proliferation of those ovarian cancer cells, which over-express this gene initially. Growth suppression was mediated by cell-cycle arrest. Stable silencing of highly expressed HMGA2 gene by shRNAi in A27/80, Ovcar-3 and OAW-42 ovarian cancer cell lines resulted in growth inhibition because of G1 arrest and increase of apoptosis as well. The tumor growth inhibition effect of HMGA2 silencing for Ovcar-3 cells was validated in vivo. Our findings revealed that the HMGA2 gene represents a promising target for gene silencing therapy in ovarian cancer. © 2008 Wiley-Liss, Inc.

Epithelial ovarian cancers account for only 4% of the total cancers in women, but the management of this cancer is a major challenge for gynecological oncologists. Because the early stages of ovarian cancer do not manifest any clinical symptoms, the majority of patients are diagnosed with an advanced disease.1 Thus, the development of new and effective methods for treatment of ovarian cancer is an evident necessity. Gene silencing therapy is thought to be one of the most attractive strategies,2 and therefore search and validation of feasible targets is important.

In our previous work, we have shown over-expression of the gene encoding nuclear factor HMGA2 as a result of Ras-induced transformation of rat ovarian surface epithelial cells.3 Genome-wide investigations have also revealed upregulation of the HMGA2 gene in primary human ovarian cancer.4, 5 The HMGA2 protein belongs to the high mobility group (HMGA) family of nuclear nonhistone phosphoproteins. These proteins have a relatively small molecular weight and contain three short basic sequences, called the AT-hook, in the N-terminal part and a highly acidic C-terminal domain. The AT-hook can bind to the minor groove of B-form DNA in AT-rich regions. As one of the major nonhistone chromosomal proteins, HMGA proteins are multifunctional and are involved in many fundamental cellular processes, including chromatin organisation, cell-cycle control, differentiation and cellular senescence.6, 7 In addition to their DNA-binding characteristics, these proteins are able to specifically interact with a number of transcription factors (NF-kB, ATF-2/c-Jun, Elf-1, Oct-2, Oct-6, SRF, NF-Y, PU-1, RAR) and participate in the formation of stereospecific multiprotein enhanceosome complexes. Because of this feature, HMGA proteins are often referred to in literature as architectural transcriptional factors.6

Members of the HMGA protein family are relatively abundant in the early embryo, where cells proliferate rapidly. HMGA1 genes are expressed mostly in regions of proliferating epithelial cells and the tissues of parenchymal organs,8 whereas the HMGA2 gene exhibits high expression levels in all mesenchymal cell condensations and in mesenhymal derivatives.9 Expression of HMGA genes undergoes suppression in differentiated cells and the HMGA2 gene is not detectable in adult human tissues where it is probably completely silenced.10, 11 Reexpression of HMGA2 gene was observed in the cells of many human malignancies such as breast12 and nonsmall lung cancers,13 pancreatic carcinoma,14 retinoblastomas,15 squamous cell carcinomas16 and myeloproliferative disorder.17 However, the precise role of HMGA2 in malignant transformation is still not clear, although it is shown to be essential in several studies. Thus, HMGA2 gene silencing prevented Ras-induced transformation of thyroid cells and resulted in growth inhibition and increased apoptosis of liposarcoma cells.18, 19 Recent studies have described mechanism of post-transcription control of HMGA2 gene expression by microRNA.20, 21 Members of let-7/miR-98 family were shown as negative regulators of HMGA2 gene activity in normal human cells, whereas aberrant function of miRNA resulted in derepression of HMGA2 in cancer.22 Alteration of let-7/HMGA2 regulatory pathway was described as a highlight of ovarian cancer progression.23 Because HMGA2 gene expression seems to be nonessential for normal cell survival, therapeutic gene silencing is thought to be relatively safe in this case. The goal of this study was to evaluate whether the HMGA2 gene can serve as a target for effective and safe silencing therapy of serous ovarian cancer.

In this study, we attempted to validate cancer-relevant over-expression of HMGA2 gene in subset of primary serous ovarian cancers versus normal ovarian tissues, as well as in a subset of cell lines established from human serous ovarian cancer versus cultured ovarian surface epithelium. We analyzed the effect of HMGA2 gene silencing on cell survival, proliferation, cell cycle distribution and anchorage independence in ovarian cancer cells. In addition, we determined the effects of the systemic administration of HMGA2-targeting construct on the growth of subcutaneous and intraperitoneal tumor xenografts in athymic nude mice.

Material and methods

In vitro labeled-RNA transcription

A specific HMGA2 (NM_003483) cDNA fragment was obtained by PCR including the T7 promoter sequence at the 5′ ends either of forward or reverse primer. For the antisense probe, the forward primer 5′-taatacgactcactatagggagaTGGGAGGAGC GAAATCTAAA-3′ including T7 promoter sequence and the reverse primer 5′-AAGCACCTTGGTCAACCATC-3′ were used. For the sense probe, the forward primer 5′-TGGGAGGAGC GAAATCTAAA-3′ and the reverse primer 5′-taatacgactcactatagg gagaAAGCACCTTGGTCAACCATC-3′ including T7 promoter sequence were used. Afterwards, labeled RNA was in vitro transcribed from the 234 bp PCR fragments using a T7 polymerase (Fermentas, St.Leon-Rot, Germany) and the Digoxygenin-RNA labeling Mix (Roche Diagnostics GmbH, Mannheim, Germany). The specific 235 bp β-actin (NM_001101) cDNA fragment was synthesized by PCR using forward primer 5′-taatacgactcactatagggagaTGAAGATCAAGATCATTGCTCC-3′ and reverse primer 5′-GCCATGCCAATCTCATCTTG-3′. Antisense labeled-RNA was in vitro transcribed and used as a positive control for in situ hybridization.

RNA in situ hybridization

The 48 serous ovarian cancer samples and 12 samples from normal ovaries were collected from patients who underwent surgery due to ovarian cancer or extirpation of the uterus due to noncancerous diseases at the N.N. Petrov Research Institute of Oncology (St.Petersburg, Russia) and at the Clinic for Gynecology, Charité (Berlin, Germany). Tissues were snap frozen after recovery, stored at −80°C and diagnosed according to standard histopathological criteria. Tissue sections were cut at 8 μm, air dried and fixed in 4% paraformaldehyde in PBS buffer for 10 min at 4°C. After 2 consecutive washes (5 min) in PBS buffer, the tissue sections were treated for 10 min in 0.25% acetic anhydride/0.1% triethanolamine and washed in PBS buffer for another 5 min. Slides were prehybridized for 4 hr at room temperature in 250 μl hybridization buffer (50% formamide, 5× SSC, 5× Denhardt's solution, 500 μg/ml salmon sperm DNA and 250 μg/ml baker's yeast tRNA). Hybridization was carried out in hybridization buffer containing 400 ng/ml digoxygenin-labeled sense or antisense RNA probe overnight at 64°C. Slides were washed twice for 5 min in 5× SSC at 64°C and once in 0.5× SSC buffer, treated with 20 μg/ml RNAse A for 30 min at 37°C in RNAse buffer (0.5 M NaCl, 10 mM Tris-HCl, pH 8.0), equilibrated in maleic acid buffer (0.1 M maleic acid, 0.15 M NaCl, pH 7.5) for 5 min, preincubated in 1% blocking reagent (Roche Diagnostics GmbH, Mannheim, Germany) dissolved in maleic acid buffer for 1 hr and incubated for another 1 hr with a sheep alkaline phosphatase-conjugated anti-digoxygenin antibody diluted as 1:500 in fresh solution of 1%blocking reagent. Subsequently, tissues were washed twice in 0.1 M maleic acid buffer for 15 min and equilibrated in detection buffer (0.1 M Tris-HCl, 0.1 M NaCl, pH 9.5). Alkaline phosphatase activity was visualized by incubation with a chromogenic substrate (5-bromo-4-chloro-3-indolyl/nitroblue tetrazolium in detection buffer) for 15–40 min. The incubation time was controlled microscopically and the reaction was stopped by rinsing in TE buffer. Finally, sections were stained with Hematoxylin RL for 1 min and covered with Mounting medium (Biomeda, Foster City, CA). Results were estimated visually via light microscopy and digital images were acquired using a Leica DC500 camera.

Gene silencing constructs

For the design and synthesis of HMGA2 specific esiRNA, we used resource (RZPDp3000D027D) from RZPD (Deutsches Ressourcenzentrum für Genomforschung GmbH, Berlin, Germany). This product represented double-stranded 311 bp DNA, which was complementary to 3,754–4,064 region of human HMGA2 mRNA (NM_003483) and tailed by strong T7 promoters at either end. In vitro transcription of dsDNA into dsRNA, generation and purification of diced esiRNA duplexes were performed using X-tremeGENE siRNA Dicer Kit (Roche Diagnostics GmbH, Mannheim, Germany).

The siRNA constructs were produced by in vitro transcription using the Silencer™ siRNA Construction Kit (Ambion, Austin, TX). We produced 4 duplexes for targeting HMGA2 mRNA in following sites: No. 1 (233–251): 5′-GGCACTTTCAATCT CAATC-3′, No. 2 (709–725): 5′-CAGGAAGCAGCAGCAA GAA-3′, No. 3 (1146–1164): 5′-CGCCAACGTTCGATTTCTA-3′ and No. 4 (2298–2316): 5′-GATGGTTGACCAAGGTGCT-3′. No. 3X scrambled duplex 5′-CGCCA(tgca)TC(ttag)TCTA-3′ (sense) carrying 4-base-pair overturns in the middle of sequence was synthesized the same way for a negative control of the experiment.

The short hairpin RNA was designed for targeting the same site of the mRNA molecule as No3 siRNA duplex. The design of the cloned insert was done according to published recommendations.24, 25 The sequence of the insert included targeted sense and antisense fragments in opposite directions separated by a loop sequence and tailed by mir-30 structures and EcoRI restriction sites was 5′-GAATTCGCTGTTGACAGTGAGCG (GCGTCAA CGTTCGATTTCTGCC) TAGTGAA GCCACAGATGTA (GGT AGAAATCGAACGTTGGCGC) TGCCTACTGCCTCGGAGA ATTC-3′. The insert for scrambled short hairpin RNA expression was 5′-GAATTCGCTGTTGACAGTGAGCG (GCGTCA(tgca) TC(ttag)TCTGCC) TAGTGAAGCCACAGATGTA (GGTAGA (ctaa)GA(tgca)TGGCGC) TGCCTACTGCCTCGGAGAATTC-3′. Fragments were amplified by PCR and cloned into a pSM2 expression vector, designed further as pSM-sh3 and pSM-sh3x correspondently (Open Biosystems, Huntsville, AL).

Cell lines and transfection procedures

Human serous ovarian carcinoma cell lines Skov-3, A27/80, Ovcar-3, CAOV-3 and OAW-42 were maintained in Dulbecco's modified Eagles Medium (Cambrex, Walkersville, MD) supplemented with 10% fetal calf serum, 2 mM glutamine and antibiotics. Human ovarian surface epithelial cells HOSE26 were cultivated in a 1:1 mixture of medium 199 (Sigma, Deisenhofen, Germany) and MCDB105 (LifeTechnologies, Karlsruhe, Germany) supplemented with 10% fetal calf serum, 2 mM glutamine and antibiotics.

For transient transfection, cells were seeded in 6-well plates and grown to 30% confluence. Transfection by single duplexes or a mix of diced siRNA was carried out in serum-free OptiMEM (Invitrogen GmbH, Karlsruhe, Germany) using Oligofectamine Reagent (Invitrogen GmbH, Karlsruhe, Germany). The final concentration of siRNA was 0.1 μM and the time of transfection was 4 hr; and then FCS was added into the medium to a final concentration of 10%.

For stable transfection, cells were seeded in 6-well plates and grown to 50% confluence. Transfection by shRNA expression pSM plasmid was carried out in serum-free OptiMEM (Invitrogen GmbH, Karlsruhe, Germany) using Arrest-In transfection reagent (Open Biosystems, Huntsville, AL). Stable transfectants were selected for puromycin resistance in 2.5 μg/ml concentration.

RNA extraction and RT-PCR analysis

The total RNA from cells or from tumor tissues was isolated by a Trizol-based method (Invitrogen GmbH, Karlsruhe, Germany) according to a protocol of the supplier. Analysis of HMGA2 gene expression in cultured cells was performed by One-Step Reverse Transcription-PCR System (Roche Diagnostics GmbH, Mannheim, Germany). HMGA2 specific primers (5′-TGGGAGGAGC GAAATCTAAA-3′ and 5′-AAGCACCTTGGTCAACCATC-3′) and beta actin specific primers (5′-TGAAGATCAAGAT CATTGCTCC-3′ and 5′-GCCATGCCAATCTCATCTTG-3′) were used. Analysis of HMGA2 gene expression in tumor tissue was done by Real-Time SYBR Green PCR system (Qiagen GmbH, Hilden, Germany). HMGA2 specific primers (5′-TCCCTCTAAAGCAGCTCAAAA-3′ and 5′-ACTTGTTGTGGC CATTTCCT-3′) and beta-actin specific primer (5′-TCC CTCTAAAGCAGCTCAAAA-3′ and 5′-CCAACCGCGAGAA GATGA-3′) were designed with the help of UniversalProbe Library software from Roche (

Protein extraction and western blot analysis

For HMGA2 protein analysis, cells were washed in PBS buffer, twice freezed in liquid nitrogen and thawed and afterwards lysed by the following buffer: 10 mM Hepes (pH 7.9), 400 mM NaCl, 0.1 mM EGTA, 5% glycerol, 0.5 mM DTT and 0.5 mM PMSF. The protein extracts were separated by SDS-Page and transferred to the polyvinyl membrane. The membrane was blocked for 1 hr in 5% nonfat dry milk in TBST buffer and incubated overnight with a polyclonal antibody against HMGA2 (Eurogentec, Seraing, Belgium). Blots were developed using the ECL system (AmershamPharmacia, Buckingshamshire, UK). To confirm equal protein loading, blots were stripped (Western Blot recycling kit, Alpha Diagnostics, San Antonio, TX) and reprobed with an actin antibody (Chemicon, Temecula, CA).

Cell proliferation assays

Cell proliferation was monitored semiquantitatively using a sodium 3′ [1-(phenyl-amino-carbonyl)-3,4-tetrazolium]-bis(4-methoxy-6-nitro)-benzene sulfonic acid hydrate (XTT)-based colorimetric assay (Roche Diagnostics GmbH, Mannheim, Germany). Cells were seeded in 96-well plates (1,000 cell/well). For evaluation of the anchorage-independent cell growth, plates were preliminarily coated with poly-2-hydroxyethyl methacrylate (Sigma, Deisenhofen, Germany). Cell growth was monitored after 24, 48 and 72 hr. The formation of formazan through cleavage of the tetrazolium salt XTT in metabolically active cells was measured as the absorbance at 490 nm using a spectrophotometer.

Cell-cycle and apoptosis analysis

Assays were performed by FACScan flow cytometer (BD, Heidelberg, Germany). To analyze cell-cycle distribution, cells were trypsinized, washed with phosphate-buffered saline (PBS) and fixed with 70% ethanol overnight. The fixed cells were then pelleted, washed, resuspended in PBS containing 40 mg/ml DNase-free RNase A (Roche Diagnostics GmbH, Mannheim, Germany) and incubated for 30 min at 37°C. After treatment, cells were pelleted and resuspended in PBS containing 50 mg/ml propidum iodide (PI) (Sigma, Deisenhofen, Germany). To analyze apoptosis rate, cells were trypsinized, washed with PBS, pelleted, resuspended in 100 ml of incubation buffer (10 mM Hepes/NaOH, pH 7.4; 140 mM NaCl; 5 mM CaCl2) containing Annexin-V-Fluos and PI, and incubated for 15 min on ice as recommended in the manufacturer's protocol (Roche diagnostics GmbH). Annexin-V staining and PI incorporation were measured using a logarithmic amplification in the FL1-H channel for Fluorescein detection (515 nm) and FL3-H channel for PI fluorescence detection (623 nm).

In vivo studies

Plasmid DNA (pSM-sh3 and pSM-sh3x) for in vivo experiments was produced by transformed PirPlus DH10bpir116 cells (Open Biosystems, Huntsville, AL) propagation and purified with QIAfilter Plasmid Mega Kit (Qiagen, Hilden, Germany). DNA was complexed ex tempore with low molecular weight polyethylenimine jetPEI (Polyplus-transfection SA, Illkirch, France) or PEI F25-LMW.27 DNA and PEI were diluted separately, mixed and allowed to equilibrate 1 hr at a room temperature. Complex formation was carried out at N/P ratio 10 (JetPEI) or 55 (PEI F25-LMW), amount of polymers was calculated in relation to DNA dose. Mixture was injected intraperitoneally as a bolus of 600 μl once pro week.

Female 8-week-old nude mice were used. Institutional guidelines for animal welfare and experimental conduct were followed. Ovcar-3 ovarian carcinoma cells were injected subcutaneously into both flanks (3 × 106 in 150 μl PBS), or intraperitoneally (10 × 106 in 500 μl PBS). Treatment was begun 2 weeks after tumor cell inoculation. One group of 5 mice bearing s.c. tumor was treated with a high dose of pSM-sh3 plasmid (100 μg pro injection), and 5 mice were treated with a same amount of pSM-sh3x plasmid as a negative control. For the analysis of HMGA2 gene targeting efficiency, the experiment was repeated with a lower dose of DNA (35 μg pro injection). Mice bearing i.p. tumor (10 animals) were treated at a dose of 35 μg pro injection.

In a case of s.c. xenograft model, tumor growth was monitored weekly, and tumor sizes were estimated from the perpendicular diameters of the tumors. Mice treated with a low-dose of DNA were treated over 3 weeks, whereas the high dose treatment was performed for 4 weeks. Mice bearing i.p. tumor underwent 4 injections as well. Regimes of treatment are indicated by arrows in Figures 9a. and 9b. Animals were sacrificed 1 week after the last injection. The weight of intraperitoneal tumor tissue was measured. Samples of tumor tissues were collected for real-time PCR analysis.


Expression of the HMGA2 gene is upregulated in serous ovarian cancer tissues and cultured cells but is undetectable in normal ovarian surface epithelium

We evaluated the expression level of HMGA2 genes in 48 samples of serous ovarian cancer tissue using in situ RNA hybridization. High or moderate expression levels of the HMGA2 gene were detected in 31 of 48 (65%) samples of serous ovarian cancer (Table I). A positive signal was distributed on areas of cancer cell growth and was almost absent in sites of connective tissue (Figs. 1a and 1b). HMGA2 gene expression was not detected in any of 12 samples of normal ovarian tissue (Table I). Since ovarian adenocarcinomas arise from the epithelium, we have studied this region in detail. The surface of ovarian epithelium as well as the epithelial lining inclusion cysts were negative for HMGA2 gene expression (Figs. 1c and 1d).

Figure 1.

In situ RNA hybridization. Samples of (a) solid and (b) papillary serous ovarian carcinoma, normal ovarian samples including (c) surface epithelium and (d) large inclusion cysts lined with epithelium. HMGA2 mRNA expression was detectable in cancer specimens (a1, b1), but not in normal ovaries (c1, d1) by antisense HMGA2 specific labeled RNA strand. No signal was detected in tissues after hybridization with the sense strand (a2, b2, c2, d2). Positive signal was detected in all tissues samples after hybridization with beta-actin specific antisense RNA strand (a3, b3, c3, d3). Magnification is ×200. [Color figure can be viewed in the online issue, which is available at]

Table I. Analysis of HMGA2 Expression in 48 Samples of Serous Ovarian Cancer and 12 Samples of Normal Ovaries
Type of tissue (N)Number of samples
NegativeWeakly positiveStrongly positive
  1. Staining intensity was evaluated visually and graded as negative, weakly positive and strongly positive.

Serous ovarian carcinoma, solid (12)336
Serous ovarian carcinoma, papillary (14)455
Serous ovarian carcinoma, glandular (22)1048
Total (48)171219
Normal ovarian tissue (12)1200

Next, we measured the levels of HMGA2 mRNA and protein in ovarian cancer cell lines and in immortalized human ovarian surface epithelial cells using RT-PCR and Western blot (Figs. 2a and 2b). Expression of HMGA2 mRNA and protein was detected in Ovcar-3, Skov-3, A27/80 and OAW-42 ovarian cancer cells; the highest expression was obtained in the A27/80 cell line in both mRNA and protein levels. Immortalized human ovarian surface epithelial cells (HOSE) and CAOV-3 cancer cells had undetectable levels of HMGA2 mRNA/protein expression.

Figure 2.

HMGA2 gene/protein expression. (a) RT-PCR; control of RNA amount is performed in parallel reaction with beta-actin specific primers. (b) Western blot analysis; loading control is performed with actin specific antibodies.

The effect of HMGA2 gene transient silencing depends on targeting sequence and initial level of expression

To choose the optimal sequence for effective HMGA2 silencing, we used the Ovcar-3 cells because these cells can be reproducibly transfected with high efficiency. Four siRNA duplexes targeting HMGA2 mRNA in different site were performed. Cocktail of various 21–23-base siRNA duplexes targeting together region 3,754–4,064 bp of HMGA2 mRNA molecule was generated by in vitro dicing reaction. This siRNA mixture (esiRNA) served as a robust positive control for the gene knockdown effect. As shown in Figure 3a, level of HMGA2 mRNA was completely reduced by esiRNA and by siRNA duplex No. 3, was partially reduced by duplex No. 4 and was not changed by duplexes No. 1 and No. 2.

Figure 3.

Transient HMGA2 gene silencing. (a) RT-PCR analysis of silencing efficiency in Ovcar-3 cells 72 hr after transfection by means of siRNA duplexes No. 1, No. 2, No. 3, No. 4 and cocktail of different 21–23-base esiRNA (esiRNA) duplexes. Control RT-PCR is performed for cells treated by Oligofectomine transfection reagents (Mock) for monoduplex as well as for polyduplex transfection, re-spectively. Parallel RT-PCR reactions were performed with beta-actin specific primers. (b) RT-PCR analysis of transient silencing of HMGA2 using duplex No.3 in A27/80, Ovcar-3 and OAW-42 cell lines at 0, 24, 49, 72 and 96 hr after transfection. Level of beta-actin mRNA expression in transfected cells were controlled in parallel reactions. Transfection with scrambled duplex No. 3 did not change the level of HMGA2 mRNA expression. (c) Western blot analysis of HMGA2 protein amount was done in A27/80, Ovcar-3 and OAW-42 cells 72 hr after transient transfection with duplex No. 3. Cells treated with scrambled duplex were used as control. To control antibody specifity, we used protein extract from COS-7 cells transfected with HMGA2 expressing plasmid. Loading control was performed with beta-actin specific antibody.

To reveal a temporal pattern of HMGA2 gene silencing, we transfected A27/80, Ovcar-3 and OAW-42 cells with siRNA duplex No. 3 and confirmed gene silencing efficiency by RT-PCR at 24, 48, 72 and 96 hr after transfection (Fig. 3b). Complete silencing was observed at 48 hr after transfection with No. 3 siRNA duplex in all cell lines, and after 96 hr, the level of HMGA2 mRNA returned to the initial level. We confirmed silencing efficiency at the protein level by Western blot analysis at 72 hr after transfection (Fig. 3c).

We analyzed the effect of the transient HMGA2 gene silencing on the growth of cultured serous ovarian cancer and HOSE cells (Fig. 4, upper panel). Two siRNA duplexes (No. 3 and No. 4) were used in these assays and have shown comparable effects. Gene silencing did not change the adherent proliferation rate of HOSE as well as Caov-3, which both do not express the HMGA2 gene detectably. Proliferation of Ovcar-3, Skov-3, A27/80 and OAW-42 was suppressed to a variable extent. Growth suppressive effect was achieved due to retention of the cell cycle in G1 phase, as it was assayed by FACS analysis 48 hr after transfection (Table II). We did not find an increase of apoptosis in any studied cell line as a result of HMGA2 gene transient silencing (data not shown).

Figure 4.

Cancer cell growth inhibition due to transient HMGA2 gene knockdown. Measurements were performed in triplicate for the cells treated with duplex No. 3 (black bars) or duplex No. 4 (grey bars) as well as for control cells treated with scrambled duplex 72 hr after transfection. Since studied cell lines exhibit differences in growth capacity, results are represented as percentage of growth rates of control cells. Inhibition of anchorage and anchorage-independent cell proliferation are shown on upper and lower panels, respectively.

Table II. Flow Cytometric Analysis of Cell-Cycle Distribution
  1. Assay was performed 48 hr after transient transfection with HMGA2-targeting siRNA No.3 or scrambled siRNA No3x as a control. Numbers indicate percentage of cells in cell cycle phases.


Advanced ovarian cancer often manifests malignant ascites, consisting of single or aggregated tumor cells, so-called spheroids. These cells contribute to the peritoneal dissemination of ovarian cancer and show evidently anchorage independency.28 Since HMGA2 gene expression was identified as a marker of advanced cancer,23 we evaluated the role of HMGA2 in anchorage-independent cells proliferation. Silencing of HMGA2 gene retarded the proliferation of Ovcar-3, Skov-3, A27/80 and OAW-42 to a comparable extent (Fig. 4, lower panel). These data indicated that HMGA2, being over-expressed in ovarian cancer cells, is involved in the control of anchorage and anchorage-independent proliferation as well. However, the adherent proliferation of Ovcar-3 and Skov-3 was retarded strongly as compared to other cell lines; whereas the inhibition of anchorage-independent proliferation was observed at a similar extent for all cells expressing HMGA2. This discrepancy can reflect a diversity of the role of the HMGA2 protein in maintenance of malignant phenotype in different types of cancer cells.29

Stable silencing of the HMGA2 gene in A27/80, OAW-42 and OVCAR-3 cell lines inhibits growth, induces G1 arrest and increases apoptosis

To study the effect of HMGA2 gene silencing in detail, we selected 3 ovarian cancer cell lines which have relative high levels of HMGA2 gene expression and are most sensitive to transient silencing of HMGA2 gene. We have established pools of A27/80(sh3), OVCAR-3(sh3) and OAW-42(sh3) cells stably transfected with plasmid pSM-sh3 expressing a HMGA2 gene specific small hairpin RNA, which corresponds to duplex No. 3 (Material and Methods). Cells transfected with a plasmid expressing a scrambled hairpin were generated as control. HMGA2 gene silencing was confirmed by RT-PCR and Western blot analysis (Fig. 5).

Figure 5.

Stable HMGA2 gene silencing. Cells transfected with scrambled hairpin expression plasmid were used as control. (a) RT-PCR, control of RNA amount is performed in parallel reaction with beta-actin specific primers. (b) Western blot analysis, loading control is performed with actin specific antibodies.

Stable silencing of the HMGA2 gene resulted in growth inhibition (Fig. 6). Adherent growth was suppressed predominately in the Ovcar-3 cells, whereas anchorage-independent proliferation was suppressed more strongly in A27/80 and OAW-42 cells. These observations confirmed the results of the transient silencing experiment. Cell-cycle distribution analysis showed an increase in the percentage of cells in G1 and a reduction of those in S-phase in treated cells when compared with control (Table III). G1 arrest was mostly exhibited in Ovcar-3(shHMGA2) cells as shown in the histogram (Fig. 7). This data is in agreement with results obtained after transient HMGA2 gene silencing. Simultaneous FACS detection of the apoptotic marker Annexin-V and propidium iodide incorporation revealed an increase in apoptosis and cell death induced by HMGA2 gene silencing (Table III). Dotplot diagrams (Fig. 8) show an increase in the percentage of cells in early and late phase of apoptosis in OAW-42(shHMGA2) cells and an increase in the number of dead cells due to necrosis in A27/80(shHMGA2) cells.

Figure 6.

Cancer cell growth inhibition due to stable HMGA2 gene knockdown. Cells transfected with scrambled hairpin expressing plasmid were used as control. Growth of shHMGA2-transfected cells is shown by thick line and growth of control cells by thin line. Adherent growth (2D) is shown by continuous line and anchorage-independent growth (3D) is by dotted line. Measurements were done in triplicate 24, 48 and 72 hr after cell seeding.

Figure 7.

Cell cycle inhibition due to stable HMGA2 gene knockdown. Flow cytometric analysis of Ocvar-3 cells after staining with propidium iodide. Cells transfected with scrambled hairpin expressing plasmid were used as control. Increase of percentage of cells in G0/G1 phase indicates cell cycle arrest due to stable HMGA2 gene silencing.

Figure 8.

Apoptosis induction due to stable HMGA2 gene knockdown. Flow cytometric analysis of A27/80 and OAW-42 cells after staining with Annexin-V-Fluos and propidium iodide. Cells transfected with scrambled hairpin expressing plasmid were used as negative control. Nonmanipulated cells exposed to puromycin at 2.5 μg/ml for 1 day were used as positive control of apoptosis. Circled regions indicate shift due to early (LR quadrant) and late (UR quadrant) apoptosis in OAW-42 cells and due to necrotic death (UL quadrant) in A27/80 cells.

Table III. Flow Cytometric Analysis of Cell-Cycle Distribution and Apoptosis Induction
pSM-sh3x (−)pSM-sh3pSM-sh3x (−)pSM-sh3pSM-sh3x (−)pSM-sh3
  1. Cells A27/80(sh3), Ovcar-3(sh3), OAW-42(sh3) stably transfected with pSM-sh3 plasmid and counterparts transfected with pSM-sh3X (control) were cultured under conventional conditions 24 hr prior to experiments. Results indicate percentage of cells being in conditions as indicated.

G0/G1 phase23.5734.0217.3628.9625.3632.91
S phase50.8142.1357.2553.7363.7957.36
G2/M phase25.6223.8525.3917.3110.859.73
 sh3x (−)sh3sh3x (−)sh3sh3x (−)sh3
Early apoptotic3.715.643.116.073.3311.6
Late apoptotic4.284.244.895.276.8325.6

Systemic administration of a HMGA2-targeting construct resulted in suppression of s.c. tumor xenografts in athymic nude mice

To validate the antitumoral effect of HMGA2 gene knockdown in vivo, we chose a serous ovarian cancer cell line, Ovcar-3, which showed a distinct growth inhibition effect in previous in vitro experiments. Subcutaneous and intraperitoneal tumor xenografts in athymic nude mice were generated. Two week after injection mice developed clearly visible s.c. tumors. Treatment began 2 weeks after tumor cell inoculation and was carried out once per week. Mice bearing s.c. tumors were treated at dose of either 35 or 100 μg DNA per injection. Mice bearing i.p. tumor were treated at dose of a 35 μg DNA per injection, since the intraperitoneal mode of delivery can be considered as local and requiring lower amounts of DNA.

None of the animals died during the experiment and no signs of acute toxicity were observed. Difference in tumor size between mice treated with pSM-sh3 versus mice treated with pSM-sh3x plasmids became visible one week after the first injection and increased gradually independently of dose applied (Figs. 9a and 9b). Mice treated with low dose received therapy 3 times and were sacrificed 1 week after last injection. Real-time PCR analysis revealed decreased level of HMGA2 mRNA in tumor tissue of mice treated with pSM-sh3 plasmid versus control animals (Fig. 9b). These results confirmed that the difference in tumor size (see Fig. 9c for representative example) was due to HMGA2 gene silencing. Mice treated with higher dose were injected one more time and sacrificed 1 week later. In this group, s.c. tumors in mice treated with pSM-sh3 plasmid almost disappeared at this time in contrast to the control group.

Figure 9.

Inhibition of s.c. tumor growth due to systemic HMGA2 gene targeting. Systemic treatment of mice with the PEI-complexed pSM-sh3 plasmid expressing HMGA2-specific hairpin was performed at a high dose (100 μg/injection) (a) and low (35 μg/injection) (b) dose of DNA. Mice of the control group were treated with the same amount of pSM-sh3x plasmid expressing scrambled hairpin. shRNA expression vectors were i.p. injected weekly (indicated by arrows). Animals treated with a low dose were sacrificed 5 weeks after tumor cell inoculation (see d for a representative example) and tumor tissues were removed for analysis. Effect of HMGA2 gene silencing was analysed by real-time PCR (c). Levels of HMGA2 gene expression are calculated as % of beta actin expression and shown as a mean ± SD of 5 animals. [Color figure can be viewed in the online issue, which is available at]

Mice bearing i.p. tumor were treated 4 times and sacrificed 1 week after last injection. Mice treated with control pSM-sh3x plasmid had a hemorrhagic ascites and multiple peritoneal tumor nodules. In contrast, the ascites was almost negligible and amounts of the tumor tissue in the abdominal cavity were much less in animals treated with pSM-sh3 plasmid (Fig. 10a). Amounts of tumor tissue were quantitated by weighing and are shown in Figure 10c. Real time PCR analysis of tumor tissue revealed an efficiency of HMGA2 gene silencing even more prominent as compared to the s.c. tumor model.

Figure 10.

Inhibition of i.p. tumor growth due to systemic HMGA2 gene targeting. Systemic treatment of mice with the PEI-complexed pSM-sh3 plasmid expressing HMGA2-specific hairpin was performed at a dose of 35 μg DNA per injection. Mice of the control group were treated with the same amount of pSM-sh3x plasmid expressing scrambled hairpin. shRNA expression vectors were i.p. injected 4 times and sacrificed 6 weeks after tumor cell inoculation. Amounts and sizes of intraperitoneal tumor nodules were found be much less in the specific treated group (see a for representative examples). The effect of HMGA2 gene silencing was analysed by real-time PCR (b), results are shown as % of beta actin expression ± SD of 5 animals. The panel (c) shows differences in intraperitoneal tumor burden between treated and control groups at the end of the experiment. [Color figure can be viewed in the online issue, which is available at]


Cancer relevant over-expression is a first crucial feature of a targeted gene. Recently, results of more than 30 genome-wide profiling studies were published. Although these data represent attractive sources for primary target selection, there are reasons for biases in the observed results. First, the admixture of noncancer components in tissue specimens and expression shifts due to culturing cells in vitro are general disadvantages in array technology in cancer research. Second, expression profiling studies of ovarian cancer require a normal specimen, which consists of a single cell layer covering the ovary and is difficult to obtain in any quantity. Specimens commonly used as a normal reference for ovarian cancer genome-wide studies (brushing OSE, whole ovary samples, short-term cultures, SV40- or telomerase-immortalized OSE cells) show crucial differences in their gene expression pattern.30 These data suggests that the selection of normal specimen in microarray studies may strongly influence the results. We analyzed the data from previous studies carried out by high-throughput methods (microarray and SSH) and verified them by means of in situ RNA hybridization. We revealed high or moderate expression of HMGA2 gene in 65% samples of serous ovarian cancer. We did not detect any expression activity of HMGA2 gene in normal surface ovarian epithelium as well as in morphologically altered epithelium lining inclusion cysts of the ovary suggested as a site from which ovarian cancer likely arises. This result indicated that upregulation of the HMGA2 gene is a frequent feature of ovarian cancer.

The necessity of high expression for malignant phenotype maintenance is a second feature of targeted genes. Members of the HMGA gene family were shown be implicated in the development of different types of tumors. The role of HMGA genes in cancer development is not still understood, but is likely related to their physiological function as architectural transcription factors. The effect of ectopic over-expression of the HMGA genes in vitro depends on the cell type. Thus, forced expression of the HMGA1 gene induced deregulation of the S-phase cell cycle and apoptosis in rat thyroid cells, but promoted neoplastic transformation in human breast epithelial cells.29, 31 The rat fibroblasts or human lymphoid cells were transformed as a result of HMGA2 gene ectopic over-expression, whereas murine fibroblasts did not undergo malignant transformation.32 The effect of HMGA gene silencing was shown to be also variable. Proliferation of human retinoblastoma and liposarcoma cells was suppressed by HMGA2 gene silencing, whereas chicken DT40 B-cell lymphoma cells showed neither growth attenuation nor expression profiling changes in a similar experiment.15, 19, 33 Results of our study revealed that HMGA2 silencing suppresses ovarian cancer cell proliferation in cases of high initial level of its expression (A27/80, OAW-42, Ovcar-3). Extrapolation of this observation to the clinic leads to an evident conclusion: diagnostic expression profiling of a tumor should precede the gene targeting therapy.

Indifference to silencing for normal cells is a third feature of optimal targets. The expression of HMGA2 gene in adult tissues is undetectable and, probably, strongly silent.11, 12 These data suggest safety of systemic HMGA2 gene silencing, while our results indicated indifferent effect of this gene knockdown in human ovarian surface epithelial cells. Taking together these data show that a gene therapy based on the suppression of the HMGA2 gene could represent a new and promising approach for the treatment of serous ovarian cancer.