This article is mentioned in:

Introduction

Surface epithelial carcinomas are the most common type of ovarian cancer and the most lethal gynecological malignancy. Epithelial ovarian cancer (EOC) comprises the majority of malignant ovarian tumors in adult women, and its high mortality is mainly due to late diagnosis (1). Serous epithelial carcinoma, the primary type of ovarian cancer, usually presents at advanced stage. Endometrioid, mucinous and clear-cell carcinomas, accounting for non-serous carcinomas, more often appear as low-stage disease (1,2). In advanced stages, the response rate to first-line chemotherapy with a platinum combination after surgical resection is approximately 80%, with 40–60% complete response. However, the median progression-free survival is only 18 months in these patients, as most relapse. The overall response rate in platinum-refractory or drug-resistant tumors is only 10–25% with subsequent relapse, resulting in a 5-year survival of only 25% (3). In addition, clinicians have long known that different subtypes of ovarian cancer respond differently to treatment and have different prognoses. High-grade serous ovarian cancers typically harbor mutations in p53. These cancers also have mutations of BRCA1 or BRCA2, as well as defective homologous recombination, the preferred mechanism of the DNA double-strand break repair pathway (3,4). EOC, a morphologically and biologically heterogeneous disease, is affected by various gene activations and alterations (2–5).

Current studies suggest that the different histological types of EOC represent distinct disease entities and exhibit varied gene expression patterns (6,7). Thus, a better understanding of the molecular basis of EOC subtypes and identification of a reliable gene expression profile for each type is necessary to uncover the fundamental mechanisms of carcinogenesis and to predict prognosis, and may provide therapeutic guidance in these diseases (8).

DNA microarray research is widely applied in determining gene expression profiles and has opened the field for analysis of the expression levels of thousands of genes (9). However, a more accurate technique is needed for a better-tailored approach to identify special targets of diagnostic and prognostic markers that predict patient response to chemotherapy and survival outcomes. We used real-time relative-quantity (RQ)-PCR assay to illustrate a gene expression signature and to distinguish expression patterns in EOC subtypes in a large patient sample panel. Based on the function of targeted genes and their expression patterns, we identified ILF3 and UBE2I as potential biomarkers for this disease.

Materials and methods

Human tissue specimens

Forty-eight pairs of ovarian specimens (tumor and adjacent normal tissues), plus an additional 35 tumors from a total of 83 patients with advanced ovarian cancer were obtained from the Cooperative Human Tissue Network (CHTN), Pediatric Division, Children's Hospital, Columbus, OH, USA. Tumor and normal samples were collected at primary surgery prior to chemotherapy, flash frozen in liquid nitrogen, and stored at −80°C until RNA/DNA extraction. All samples were evaluated by pathologists, and the 83 tumors were classified as serous carcinoma (n=51), endometrioid carcinoma (n=13), mucinous carcinoma (n=11) and clear cell carcinoma (n=8).

RNA extraction

Total RNA from each specimen and the normal control of the ovarian cancer patients was extracted and purified by the method of hot phenol/chloroform extraction as previously reported (10). Isolated RNA was purified and dissolved in DEPC water and stored at −80°C.

Oligo synthesis

Following a search of the gene database, we selected a total of 50 genes based on their functions and the expression literature in ovarian, breast and lung cancers for this investigation. Primary gene functions included transcription factor, DNA/RNA and protein binding activity, gene activation and regulation. Forward and reverse primers for each of the 50 genes were synthesized by Gene Probe Technologies, Inc. (Gaithersburg, MD, USA). Primer sequences of the 4 targeted genes used in real-time quantitative PCR assay are listed in Table I.

Table I. Primer sequences of the 4 targeted genes used in real-time quantitative PCR.

Table I.

Primer sequences of the 4 targeted genes used in real-time quantitative PCR.

Gene symbol	Gene ID	Forward primer	Reverse primer
TAL2	6887	5′-tcaccctccagacaaaaagc-3′	5′-ccaggtgaaggaacctggta-3′
EGF	1950	5′-aggtggctggaagcctttat-3′	5′-tgtggacagaacctccatca-3′
ILF3	3609	5′-ctggtgctgctgtgtaagga-3′	5′-agggacaatggaggctcttt-3′
UBE2I	7329	5′-caggagaggaaagcatggag-3′	5′-tcgggtgaaataatggtggt-3′

Reverse transcription-PCR

Through reverse transcription (RT), using the Super Script Preamplification System (Life Technologies, Inc.), cDNA was generated with oligo-dT primers from 5 μg of total RNA of 83 ovarian tumor tissues and 48 adjacent normal samples. RT-PCR was performed using AmpliTaq DNA polymerase, FS cells and gene-specific primers of the 50 genes in all cDNA samples. PCR amplicons were separated by 1% agarose gel electrophoresis, and visible density bands of the gene amplification in these samples were indicative of gene expression. Genes expressed in RT-PCR detection were selected for subsequent real-time quantitative PCR analysis.

Real-time quantitative PCR

Real-time RQ-PCR was performed using SYBR® Green reagent kit (ABI Cat. 4367659) according to manufacturer's recommendations. This assay was used for 20 selected genes in 30 EOC specimens for the gene expression signature profile. Amplifications were carried out on ABI PRISM 7000 Detection System and analyzed by ABI 7500 software (PE Applied Biosystems, Foster City, CA, USA).

In brief, all reactions were optimized to obtain the best amplification kinetics under the same cycling conditions (10 min at 95°C, 40 cycles of 15 sec at 95°C, and 1-min at 60°C). The composition of the reaction mixture in a final volume of 20 μl contained Power SYBR Green Master Mix 10 μl, gene-specific primers (10 μM) 1 μl and cDNA 2 μl. Negative controls containing all PCR components without template DNA (denoted NTC) were used to ensure that the reagent mix was free of contamination. Each reaction was run in triplicate. The average threshold cycle (Ct) and the comparative ΔΔCt method were automatically calculated for the expression of the gene and normalized to the mean Ct-value of 18S ribosomal. The RQ-value was calculated using the ΔΔCt method. Fold change in gene expression was calculated as 2−ΔΔCt. For each gene in this study, a Ct-value <32 was considered as positive expression and vice versa.

Statistical analysis

In real-time RQ-PCR assays, the average Ct, ΔCt, ΔΔCt and RQ were calculated by Applied Biosystems Sequence detection software (7500 Fast System SDS Software version 1.4). RQ (ΔΔCt value) was used to compare gene expression between samples. Since RQ equals 2−ΔΔCt, a higher negative ΔΔCt value, represents higher expression. SPSS 15.0 software was used to determine statistical significance. A P-value <0.05 was considered statistically significant between test means.

Results

Patient characteristics

Forty-eight pairs of ovarian specimens (tumor and adjacent normal tissues) plus an additional 35 tumors (from a total of 83 ovarian cancer patients) were assessed in this investigation. The clinicopathological characteristics of the 83 EOC patients in this study are shown in Table II. The median age of the patients at diagnosis was 59 years (range, 37–83 years). Approximately 65.1% of the patients (54 out of 83) were diagnosed with advanced stage tumors (FIGO stages III/IV), and 84.3% (70 out of 83) had moderately poorly differentiated tumors (grades 2 and 3).

Table II. Clinicopathological characteristics of the EOC cases.

Table II.

Clinicopathological characteristics of the EOC cases.

Patient no.	Age range (years)	Histological type	FIGO stage at diagnosis				Tumor grade
			I	II	III	IV	G1	G2	G3
51	40–83	Serious	3 (5.88%)	2 (3.92%)	42 (82.35%)	4 (7.84%)	3 (5.88%)	9 (17.65%)	39 (76.47%)
11	40–74	Mucinous	7 (63.64%)	3 (27.27%)	1 (45.45%)	0	5 (36.36%)	4 (18.18%)	2 (9.09%)
13	37–76	Endometrioid	4 (30.77%)	3 (7.69%)	6 (46.15%)	0	2 (15.38%)	4 (30.77%)	7 (53.85%)
8	48–77	Clear cell	6 (75%)	1 (12.50%)	1 (12.50%)	0	3 (37.50%)	1 (12.50%)	4 (50%)

[i] FIGO, International Federation of Gynecology and Obstetrics.

Reproducible gene expression identified in the studied samples

Initially, we selected 50 genes based on their functions and the expression literature. Using reverse transcription-polymerase chain reaction assay (RT-PCR), 39 of the 50 genes were expressed by showing gel electrophoresis density; the remaining 11 genes were not expressed (data not shown). Of the 39 expressed genes, 20 demonstrated dominant and reproducible expression among the samples. These 20 genes were further evaluated in 30 EOC specimens by real-time quantitative PCR. The functions of these 20 genes are summarized in Table III.

Table III. Function of 20 dominant and reproducible genes expressed in the studied ovarian specimens.

Table III.

Function of 20 dominant and reproducible genes expressed in the studied ovarian specimens.

Gene symbol	Gene name	Function
RAD52	RAD52 homolog (S. cerevisiae)	DNA double-strand break repair and homologous recombination
RPUSD2	RNA pseudouridylate synthase domain containing 2	Pseudouridine synthase activity
SEH1L	SEH1-like (S. cerevisiae)	Intracellular protein transport across a membrane
SLC25A5	Solute carrier familly 25 member 5	Adenine transmembrane transporter
PLSCR1	Phospholipid scramblase 1	Phospholipid scramblase activity
INPPL1	Inositol polyphosphate phosphatase-like 1	Inositol or phosphatidylinositol phosphatase activity
TXNRD1	Thioredoxin reductase 1	Protein disulfide oxidoreductase activity, thioredoxin-disulfide reductase activity
SSBP1	Single-stranded DNA binding protein 1	Housekeeping gene involved in mitochondrial biogenesis
FAT	Fat tumor suppressor homolog 1 (Drosophila)	Calcium ion binding, protein binding
SMARCD2	SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily d, member 2	Transcription coactivator activity protein binding
MAP2K2	Mitogen-activated protein kinase 2	Protein serine/threonine kinase activity, protein tyrosine kinase activity, transferase activity
HNRNPA3	Heterogeneous nuclear ribonucleoprotein A3	Nuclear mRNA splicing, via spliceosome
MSX2	msh homeobox 2	Transcription factor activity
TAL2	T-cell acute lymphocytic leukemia 2	Transcription regulator activity
EGF	Epidermal growth factor	Epidermal growth factor receptor activating ligand activity, calcium ion binding
ZNF71	Zinc finger protein 71	Metal ion binding, zinc ion binding
ILF3	Interleukin enhancer binding factor 3	Double-stranded RNA binding, transcription regulation
UBE2I	Ubiquitin-conjugating enzyme E2I (UBC9 homolog, yeast)	Post-translational protein modification, ubiquitin-dependent protein catabolic process
INSR	Insulin receptor	Insulin receptor activity, phosphoinositide 3-kinase binding receptor signaling protein tyrosine kinase activity
NP220	Zinc finger protein 638	Double-stranded DNA binding, metal ion binding, zinc ion binding

Expression patterns of TAL2, EGF, ILF3 and UBE2I in normal and EOC specimens

Using real-time RQ-PCR we further identified four genes (TAL2, EGF, ILF3 and UBE2I) among the 20 selected genes for extensive study. These genes demonstrated distinct expression patterns among the EOC subtypes. In brief, TAL2 was expressed in 80 of the 83 EOC and all 48 normal samples. EGF was expressed in 69 of the 83 EOC and 45 of the 48 normal samples. Seventy-one of the 83 EOC samples exhibited ILF3 expression, and 54 of the 83 EOC exhibited UBE2I expression. In comparison, ILF3 was expressed in 28 of the 48 and UBE2I was expressed in 12 of the 48 normal samples. The differences in ILF3 and UBE2I expression between the tumor and normal tissues were statistically significant (P<0.05) (Table IV). As shown in Fig. 1, 4 genes were expressed in all types of EOC and normal tissues. However, ILF3 expression in EOC was 4.8-fold higher than in the normal samples; UBE2I expression in EOC was >2-fold higher compared to that of the normal samples.

Figure 1. Expression patterns of TAL2, EGF, ILF3 and UBE2I in normal and tumor tissues. The mean RQ-value of TAL2 in EOC (0.344) vs. the value in normal sample (3.118) was 0.11. The mean RQ-value of EGF in EOC (0.426) vs. the value in normal samples (3.174) was 0.134. The mean RQ-value of ILF3 in EOC (1.968) vs. the value in normal sample (0.409) was 4.812 (P<0.05). The mean RQ-value of UBE2I in EOC (3.757) vs. the value in normal sample (1.771) was 2.121 (P<0.05).

Table IV. Comparison of the rate of expression between tumor and normal samples (Chi-square test).

Table IV.

Comparison of the rate of expression between tumor and normal samples (Chi-square test).

	TAL2	EGF	ILF3	UBE2I
EOC tumors (83)	96.4% (80/83)	83.1% (69/83)	85.5% (71/83)	65.1% (54/83)
Normal samples (48)	100.0% (48/48)	93.8% (45/48)	58.3% (28/48)	25.0% (12/48)
P-value	0.468	0.081	<0.05	<0.05

The expression patterns of TAL2, EGF, ILF3 and UBE2I in the histological types of EOC

Fig. 2 shows the mean RQ-value of TAL2, EGF, ILF3 and UBE2I in histological types of EOC. TAL2, EGF, ILF3 and UBE2I were expressed in all EOC subtypes. Interestingly, the mean RQ value of ILF3 in the serous patients was 2.671 compared to 0.6 in clear cell, 0.256 in endometrioid and 0.336 in mucinous carcinomas. The mean RQ value of UBE21 in serous patients was 4.979 compared to 0.224 in clear cell, 0.243 in endometrioid and 0.46 in mucinous carcinomas. In other words, ILF3 and UBE2I showed extremely high expression in serous carcinomas.

Figure 2. Expression patterns of TAL2, EGF, ILF3 and UBE2I in histological types of EOC samples. ILF3 and UBE2I showed extremely high expression in serous carcinomas. The mean RQ-value of ILF3 in serous patients was 2.671 compared to 0.6 in clear cell, 0.256 in endometrioid and 0.336 in mucinous carcinomas. The mean RQ-value of UBE2I in serous patients was 4.979 compared to 0.224 in clear cell, 0.243 in endometrioid and 0.46 in mucinous carcinomas.

Association of expression patterns with FIGO stage and tumor grade

The expression patterns of TAL2, EGF, ILF3 and UBE2I were correlated with FIGO stage and tumor grade among 83 EOC tissues using real-time quantitative PCR analysis (Table V). The mean RQ-values of TAL2 and EGF in early stage or low grade and in advanced disease were overall low. In contrast, the mean RQ for ILF3 in advanced stage vs. early stage was 3.8; in poorly differentiated (G2/G3) vs. well-differentiated (G1) this value was 9.34; for UBE2I, the RQ-value was 5.76 and 3.85. The overexpression of ILF3 and UBE2I in advanced stage and advanced grade indicates that these two genes may play an important role in tumor progression and pathological differentiation of this disease.

Table V. Association of expression patterns with FIGO stage and tumor grade.

Table V.

Association of expression patterns with FIGO stage and tumor grade.

Mean RQ-value	TAL2	EGF	ILF3	UBE2I
Early stage (I–II)	0.424	0.514	0.671	0.864
Advanced stage (III–IV)	0.301	0.402	2.550	4.975
Advanced stage vs. early stage (fold)	0.710	0.782	3.800	5.758
G1	0.517	0.582	0.241	0.885
G2–G3	0.310	0.400	2.251	3.406
G2–G3 vs. G1 (fold)	0.600	0.687	9.340	3.849

Discussion

TAL2, one of three transcription factors of the basic helix-loop-helix (bHLH) family, was found at junctions of chromosomal translocations associated with T-cell acute lymphoblastic leukaemia (T-ALL). This gene was activated in the chromosomal translocation t(7;9) (q35;q34) in a subset of T-ALL patients causing overexpression in the T-cell lineage (11,12). Thus, TAL2 is identified as an oncogenic transcription factor of T-ALL (13). TAL2 expression in adult testes represents the gene's role in the developing midbrain, diencephalon and anterior pons (11). Bucher et al reported that TAL2 normally plays a pivotal role in brain development, and that without this gene, mice cannot survive to maturity (12). In the present study, we report, for the first time, TAL2 expression in normal human ovarian and ovarian tumor tissues.

Epidermal growth factor (EGF) has a profound effect on the differentiation of ovarian surface epithelial cells by enhancing motility and inducing secretion of pro-MMP-2 and MMP-9 resulting in localized stimulation (14). EGF likely contributes to ovarian surface epithelium (OSE) rapid post-ovulatory proliferation and to the epithelio-mesenchymal conversion trapped in the ruptured follicle. Failure of such function may lead to the formation of epithelial inclusion cysts, which are known to be the preferential sites of malignant transformation by generating a microenvironment enriched of growth factors, cytokines and hormones to the entrapped OSE (15). There are a few reports regarding EGF expression in tumors. Stromberg and colleagues reported that 10 of 10 (100%) borderline tumors and 10 of 14 (71%) epithelial ovarian tumors expressed EGF (16). Niikura et al also observed EGF expression in 18 of 25 (72%) studied epithelial ovarian tumors (17). Our study of a large sample panel (83 ovarian patients) demonstrated that EGF was expressed in both normal and the four histological tumor subtypes, which is consistent with other reports.

ILF3 (a protein known as NFAR or NF90), not previously found in ovarian cancer, was originally identified as a component of a dimeric transcription regulator and has a regulatory function in vitro. More recent studies suggest that ILF3 plays a role in transcriptional and post-transcriptional regulation (18). NFAR or NF90 are ubiquitously expressed in the nucleus of many cell types and tissues. They interact with PKR and PKR-mediated signaling, and may be involved in the mRNA processing in cells (19). Vumbaca et al have demonstrated that the DRBP76/NF90 isoform facilitates the expression of vascular endothelial growth factor (VEGF) by promoting VEGF mRNA loading onto polysomes and translation under hypoxic conditions, thus promoting breast cancer growth and angiogenesis in vivo (20). VEGF, the key angiogenic factor expressed under restricted nutrient and oxygen conditions in most solid tumors, is up-regulated in ovarian tumors and promotes tumor cell growth, migration and survival (21,22). Thus, ILF3 may play a role in tumorigenesis of EOC via regulation of VEGF expression. In our study, we observed that ILF3 was overexpressed in serous carcinoma. Serous carcinoma has the highest percentage of advanced stage and poorly differentiated cases among these four subtypes. We also found that ILF3 exhibited a higher expression trend in advanced stage or poorly differentiated EOC, compared to early stage or well-differentiated EOC.

UBE2I (Ubc9) is important for genome integrity, particularly during mitosis and overall cell survival (23). Recently, sumoylation, small ubiquitin-related modifier (SUMO) conjugation, has been identified as another type of protein modification. Ubc9, an essential E2-conjugating enzyme for sumoylation, seems to play a central role in sumoylation-mediated cellular pathways (24). In addition, several important DNA repair enzymes are subject to sumoylation, which appears to be involved in DNA damage/repair. Moreover, Ubc9/SUMO are recently reported to have a fundamental effect in tumorigenesis and tumor progression (23,25). Many oncoproteins and tumor supressors including PML, MDM2, c-MYB, c-JUN and TP53 are involved in SUMO (23–25). Our findings demonstrated that UBE2I mRNA was up-regulated in serous carcinoma of the ovary which is consistent with a report by Mo et al (24). The molecular mechanism by which Ubc9 promotes tumor growth and interacts with other factors is as yet unclear.

In conclusion, we provide the first evidence of TAL2 and ILF3 expression in normal human ovary and epithelial ovarian cancer. We demonstrated that UBE2I and ILF3 expression was higher in advanced stage or poorly differentiated tumors, compared to early stage or well-differentiated EOC. Our results indicate that ILF3 and UBE2I may play an important role in tumorigenesis of EOC. Further investigations to elucidate the molecular mechanisms involved in ILF3 and UBE2I activiation in EOC development are warranted.

Acknowledgements

This study was supported by NIH Grant RO3 CA107979-01 and the WVU Mary Babb Randolph Cancer Center, Molecular Medicine Core Facility, with editorial assistance by Michael D. Mueller.

References