Background
Epithelial Ovarian Cancer (EOC) is the most lethal female reproductive tract malignancy with nearly 200,000 new cases and > 125,000 deaths attributable to the disease each year worldwide [
1]. The high fatality-to-case ratio is due, in part, to lack of effective screening modalities to detect ovarian cancer at an early stage wherein rates of cure exceed 90%. Most patients present with advanced stage disease and the cornerstone of treatment is surgical debulking followed by platinum-based chemotherapy. The other major contributor to the high fatality-to-case ratio is chemoresistant disease. In fact, while 80% of patients have a complete clinical response to their primary therapy, the majority will die from disease recurrence within 5 years. The overall worldwide 5-year survival rate of the disease is < 40% [
2], however, when detected early, the 5-year survival rate more than doubles [
3]. Unfortunately, EOC has non-specific, vague, gastrointestinal, and often ignored symptoms such as bloating, irregularity, and indigestion and there are no approved population screening methods, making early detection difficult and uncommon.
The search for reliable, specific, and sensitive serum-based biomarkers for EOC has a long history and its major highlight remains the identification of CA125 nearly 30 years ago [
4]. Although CA125 is expressed in a majority (~80%) of late stage disease, it is elevated in only a subset (~50%) of early disease, thus limiting its usefulness for early disease detection [
5]. In an attempt to overcome this limitation, proteomic-based studies have sought novel biomarkers. Examples of promising markers found through candidate and proteomic approaches include HE4, transthyretin, and CA72.4 [
6‐
8]. Nonetheless, no markers are approved for population screening or disease detection whereas only CA125, along with HE4, are approved for monitoring of recurrent disease [
9]. The history of poor performance of individual markers has led researchers to also evaluate panels or combination markers [
6,
7,
10,
11].
Next Generation Sequencing technologies, as applied to cancer genomes and transcriptomes, has allowed a relatively unbiased and more complete view of the global changes that define tumors [
12,
12‐
14]. Specifically, analyses of melanoma, pancreatic, lung, and breast cancers have revealed key pathways and genes affected in these cancers by mutations, copy number variations, and transcriptional changes [
12,
12‐
14]. Furthermore, application of this knowledge can be used to discover both personalized and global diagnostic and prognostic biomarkers [
3,
15]. We hypothesized that application of RNA-Seq technology to ovarian cancer could identify overexpression of secreted proteins that could act as novel biomarkers.
We analyzed the global gene expression patterns of a highly clinically annotated sample set of EOC representing both early and late stage tumors by RNA-Seq. Focusing specifically on transcripts that had evidence for secretion of their translated protein products, we identified IGFBP-4 to be highly expressed across all stages of EOC. IGFBP-4 is one of six IGFBP's, a family of regulators of normal and tumor cell biology [
16], whose function is to inhibit IGF-I and -II binding to their receptors, IGF1R and IGF2R [
17]. It is present in all body-fluids, secreted primarily by the liver, but also expressed by a number of organs, including the ovaries. In the ovary, it is involved in follicle selection and is upregulated
in vivo and
in vitro in response to estrogen [
17]. Tumor expression of IGF family members has been linked to breast, endometrial, colon, and skin cancers [
16]. In this study, we demonstrate that IGFBP-4 serum levels were significantly upregulated in primary and recurrent EOC patients even in a number of cases where CA125 levels were within normal limits.
Methods
Patients and Specimen Collection
EOC tumor samples were collected from Mount Sinai School of Medicine (New York, USA) and San Gerardo Hospital (Milan, Italy) patients at the time of surgery under their respective IRB-approved protocols. Samples were divided in the operating room and a portion sent for pathology confirmation and staging. Portions were flash frozen for RNA and protein analysis or used immediately for generating patient-derived cell lines. Papillary serous tumor samples were collected across all stages (five stage I/II, 11 stage III/IV, two disseminated peritoneal lesions, and two recurrent tumors). For comparison, two borderline tumors were also sequenced.
Blood samples were collected in gold top tubes (BD Biosciences, Franklin Lakes, New Jersey) directly prior to surgery, directly prior to chemotherapy, or at clinical office visits (controls), allowed to clot and centrifuged at 2600 rpm for 10 minutes to separate serum. Serum samples were aliquoted to minimize freeze thaw cycles and stored at -130°C. Patient characteristics including age, ethnicity, stage/grade of tumor are provided in Additional File
1: Table S1. Control samples were collected at the time of routine office visits.
For disease surveillance studies, blood samples were serially collected during each chemotherapy infusion, and in subsequent office visits thereafter. IGFBP-4 levels were measured for each visit, or averaged over the entire postsurgical period to give a composite value. Additional File
2: Table S2 highlights patient characteristics and treatment regimens. Disease recurrence status was assessed by a combination of positive CT/PET scans, CA125 levels, and/or positive operative laparoscopy.
RNA was extracted from frozen tissue using QIAzol according to the manufacturer's instructions (Qiagen, Valencia, California). Briefly, tissue was homogenized in QIAzol on ice. Chloroform was added, mixed and centrifuged to allow for separation and removal of the aqueous layer. RNA was precipitated in isopropanol overnight at -20°C. The suspension was centrifuged to pellet the RNA, washed with 75% ethanol and then resuspended in RNAase-free water. RNA integrity numbers (RINs) were determined (Agilent Bioanalyzer, Agilent Technologies, Santa Clara, California) and only RNA with a RIN score of ≥ 8.0 was submitted for next-generation sequencing.
RNA-Seq
Epithelial ovarian cancer transcriptomes were prepared for paired-end sequencing using the Illumina GAII platform by the manufacturer's protocols and with a second size selection step to reduce ligation artifacts. Reads were aligned using the software program ELAND32 (provided with the Illumina sequencing platform). Expression levels were quantified by running ERANGE v. 3.0.2. [
18]. For each gene, ERANGE reported the number of mapped reads per kilobase of exon per million mapped reads (RPKM).
Quantitative Real-time Reverse Transcription PCR (qRT-PCR)
RNA-Seq data was confirmed by qRT-PCR. One microgram of RNA was reverse transcribed using the BioRad Iscript system (BioRad, Hercules, California). qRT-PCR was performed on an ABI PRISM 7900HT sequence detection system (Applied Biosystems, Carlsbad, California). Cycle number values were normalized against two housekeeping genes, B2M and GAPDH. Data shown are the averages of three separate experiments, each performed in triplicate. The IGFBP-4 primers used were IGFBP-4 Fwd: 5'- AGGTCCTTCCTTTAGGTCTG-3' and IGFBP-4 Rev: 5'- GGAAGACTTGAAGCACAGAG-3'.
ELISA
Patient serum IGFBP-4 levels were analyzed in duplicate using Active IGFBP-4 ELISA (Diagnostic Systems Laboratories, Inc, a Beckman Coulter Company, Webster, Texas) according to the manufacturer's protocol. Standards and internal controls were assayed on each plate for calibration and consistency. Colorimetric absorbance was detected using a microplate reader at 450 nm with a background correction at 620 nm. A standard curve was generated for each plate and sample IGFBP-4 concentrations determined.
Statistical Analysis
Statistical differences were determined using the Student's t test or ANOVA with Bonferroni correction and post-hoc test. ROC analysis was performed using SPSS software (IBM, Chicago, Illinois).
Discussion
Using whole transcriptome analysis across all stages of EOC, we initially identified IGFBP-4 as a secreted protein highly expressed in all tumors. We then confirmed and quantitated IGFBP-4 overexpression in patient serum samples. To our knowledge, this is the first report demonstrating increased serum IGFPB-4 expression in ovarian cancer patients.
Despite large and significant differences between mean IGFBP-4 levels in cases and controls, ROC analysis revealed limited sensitivity at the specificities required for a simple single marker ovarian cancer screening test (Figure
2C, D). This is due to the level of overlap between the controls and cases which makes differentiating the two groups more complex. Above all, the low occurrence rate of EOC, combined with invasive nature of first-line treatment (surgical cytoreduction combined with chemotherapy), require a very high specificity, suggested to be at least 99.6% and a sensitivity of at least 75% to yield a positive predictive value of 10 [
21]. Nonetheless, our studies suggest that, with further study, determination of IGFBP-4 levels could provide use in three clinical settings.
First, as shown in Figure
1, IGFBP-4 serum levels can be significantly increased in cases of early- and late-stage disease even when CA125 is within normal limits. Notably, three of the six early-stage cases with normal CA125 levels had increased IGFBP-4 levels. Therefore, combining IGFBP-4 and CA125 could increase the sensitivity for detecting EOC, especially, early stage disease. Future studies will be required.
Second, it has now been recommended that women with a suspicious adnexal mass should be referred to a gynecologic oncologist for evaluation since the early distinction between a benign and malignant mass represents an important clinical decision point. [reviewed in [
22]].. In our patient cohort, malignant masses were associated with average IGFBP-4 levels ~3× higher than benign masses (Figure
2). When assay specificity is set at 94%-the highest we could achieve, given the number of samples in our dataset -, sensitivity is 73%. Increasing the sample sizes for these studies will increase the power of the analysis and will allow us to better analyze the overlap between and variability within groups. It will therefore be of future interest to increase sample sizes and re-evaluate the clinical utility of IGFBP-4 alone or in combination with other markers for distinguishing between benign and malignant masses.
Finally, we investigated the potential use of IGFBP-4 as a biomarker to monitor disease recurrence and resistance to treatment in patients receiving chemotherapy. While levels did not reach statistical significance, there was a trend for NED patients to have lower average IGFPB4 levels compared to AWD patients. Although levels in chemotherapy patients did not always remain below the ROC-determined threshold of 1000 ng/ml, those whose cumulative IGFBP-4 level average was less than the threshold were more likely to be in the NED group than those with higher averages (Figure
3). Moreover, the percentage of serial IGFBP-4 readings above threshold was higher in the AWD group compared to the NED group, again suggesting that high serum levels of IGFBP-4 may be indicative of disease state. Although differences in age and stages between the two groups (Additional File
2: Table S2) may contribute to this difference, we believe it unlikely given that in our larger diagnostic data set there was no difference between early EOC IGFBP-4 levels and later stage disease levels (early average 1334.5 ng/ml, late average 1305.4 ng/ml, p = 0.91, Figure
2). Additionally, we did not find any correlation between age and IGFBP-4 levels in either cases or controls (Additional File
4: Figure S1, R
2 = 0.003), although a positive correlation has been previously reported in healthy individuals [
23]. Finally, it should be noted that the chemotherapy regimens received by women with disease recurrence were not always uniform between the groups. In such a small sample set, and in such a novel study, it is unknown at this time what effect, if any, these differences in agents may have had on IGFBP-4 levels.
Future studies are planned primarily to increase sample size and diversity of patients. Unexpectedly, we noted that Hispanic cases have significantly higher serum IGFPB-4 levels compared to other non-Hispanic cases or all controls (Additional File
5: Table S4). We are unaware of previous reports or studies suggesting a biologic basis for this finding. Thus, this intriguing finding will be specifically explored in future studies.
The IGF pathway has been implicated in carcinogenesis [
16] and the role of IGFBP-4 has been studied in a number of human malignancies, including lung, endocrine (thyroid and adrenal), breast, prostate, and hepatocellular cancers [
17]. Increased serum IGFBP-4 levels have also been associated with breast cancer, melanoma, and acute lymphoblastic leukemia [
17,
24]. While no previous reports have examined IGFBP-4 serum levels in ovarian cancer patients, IGFBP-4 was one of 52 proteins identified in a proteomic analysis of EOC ascites fluid although serum levels were neither evaluated nor compared with control samples [
25]. At this time it is unclear how increased levels of IGFBP-4 may relate to initiation or progression of ovarian cancer. Our findings may initially seem counterintuitive, as the understood role of IGFBP-4 is to bind to and inhibit IGF-I and IGF-II, thereby suppressing cell growth and proliferation. In cancer settings, however, it is suggested to do exactly the opposite [
17,
24]. One possible explanation pertains to the hormone responsiveness of IGFBP-4. In an estrogen-rich environment that may occur as a result of ovarian cancer, IGFBP-4 is thereby over stimulated, and could serve as a marker for this subset of cancers. In a second scenario, involving the known role of IGFBP-4 in follicle stimulation, IGFBP-4 becomes constitutively expressed over years of repeated ovulation, and this continued expression might drive overgrowth of the epithelial cell layer of the ovary, contributing directly to the growth of the tumor. At best, these hypotheses are speculative, and biologic proof is required. However, given the suggestion of diagnostic and prognostic significance, as well as the potential as a therapeutic or preventative target, we believe these future studies worthwhile. These studies may work specifically to address issues of samples size, adjust for patient ethnicity and investigate the molecular role of IGFBP-4 is carcinogenesis of progression.
Competing interests
Eugene Chudin and Mark Chee are employees and shareholders of Prognosys Biosciences, Inc. All other authors declare no conflicts of interest.
Authors' contributions
RAM participated in sample collection, selection, sequencing analysis, and all molecular studies and manuscript drafting. ML participated in all molecular studies. ES participated in sample collection and sequencing analysis. HS provided bioinformatics support and analysis. SC supplied samples and clinical information. EC provided bioinformatics support and analysis. RF, SM, and MD supplied samples and clinical information and participated in study design. RS provided bioinformatics support and analysis as well as data interpretation. PD supplied samples and clinical information and participated in study design. JAM participated in overall study design, sample selection, sequencing analysis, and manuscript drafting. All Authors reviewed and approved the final version of the manuscript.