Background
Ovarian cancer (OVC) is the fifth most common cause of cancer-related death in women [
1], and results in over 14,000 deaths annually in the US. It is considered to be the most lethal malignancy of the female reproductive system, largely because it is usually diagnosed at an advanced stage [
2]. While the overall 5-year survival for patients in different stages of this malignancy is 45 %, the survival rate is as high as 90 % when the disease is diagnosed at an early stage (stage I/II) compared to only 11 % when diagnosed at an advanced stage. Unfortunately, because of the asymptomatic nature of the disease, nearly 80 % of new cases of OVC are diagnosed at advanced stages (III/IV). Thus, early detection of the disease is critical to reducing mortality. In addition to asymptomatic progression, early stage diagnosis has been difficult to achieve because OVC exhibits a wide range of morphological, clinical, and genetic variations during the course of tumor progression [
2,
3]. Robust biomarkers that are sensitive and specific for OVC are needed for effectively screening the general population. CA125 has been used for years as a gold standard for disease monitoring and for assessing relapse or response to treatment. However, CA125 has low specificity for OVC [
4] as well as less than optimal ability to detect all types of OVC; therefore, CA125 is not an optimal biomarker for early detection. More recently, the diagnostic value of CA125 has been shown to be improved when used in combination with other markers, including CA19-9, MCSF, OVX1, and HE4 [
5,
6]. Additionally, the OVA1 blood test is an FDA-cleared test that helps evaluate an ovarian mass for malignancy prior to surgery. Apart from evaluating levels of CA125 and Beta-2 microglobulin, which are expected to be up-regulated in malignant conditions; the test also measures levels of apolipoprotein A1, prealbumin, and transferrin, which are expected to be down-regulated. In 2011, the FDA approved the use of blood tests for HE4 and CA125 with the Risk of Ovarian Malignancy Algorithm (ROMA), which demonstrated higher accuracy in determining risk in pre- and post-menopausal women. Additional tests such as OVACheck, which is based on proteomic technology, and OvaSure, which includes CA125 among five other biomarkers, require further validation [
7]. Although these tests demonstrate that a combination of multiple markers can generate synergistic advantages over a single marker in a clinical setting, they are primarily based on upregulation of CA125, which does not always occur. Also, these tests are mostly used for further evaluation of women who have already been diagnosed with pelvic mass and are due for surgery, rather than for initial diagnosis. There are no current FDA-cleared biomarkers for OVC screening; markers are cleared only for the limited application of monitoring disease recurrence and therapeutic response.
Proteomics research over the last two decades has identified hundreds of potential biomarkers for OVC [
8], and subsequently, appropriate validation methods have identified biomarkers with high sensitivity and specificity for early detection of the disease (Table
1). We have used a bioinformatics-guided approach to pinpoint a set of potential biomarkers for OVC. Subsequent screening and validation have distinguished three biomarkers with clinical relevance: human kallikrein 6, kallikrein 7, and PRSS8 [
9‐
11]. Overexpression of these proteins is highly specific for OVC but not other cancer types; these proteins are significantly overexpressed in OVC cells, and are secreted into bodily fluids. We have recently published the potential of KLK6 and KLK7 as early detection biomarkers [
12]. The current study provides data and analysis that define the potential use of PRSS8 as a biomarker for early detection of OVC.
Table 1
Biomarkers with high potential for early screening and diagnosis of OVC
ATP7B |
CA125
| CLEC3B |
KLK6
| TOP2A | ARID4B | CEA | ID2 | IGFBP2 | INHA |
PDGFA | HE4 |
DUSP1
| BSG | CLDN3 | REEP5 | MIF | IGF2BP1 | LGALS3BP | CDC25C |
BRCA2 | CA72-4 |
IL13RA2
| STAT3 | CLDN4 | CCT3 | AFP | IQGAP1 | MSLN | NME1 |
DNAJC15 | BARD 1 | PLK1 | RAET1E | COPS5 | CD47 | PRL | RHOC | ST14 | AKT2 |
KLK14 |
BCL2
| VIL2 | TITF1 | CSF1 | ETV4 | MUC 1 | RNASE2 | AMH | ANGPT2 |
KLK9 | IGFII |
APOD
| TFF1 | EFNB1 | MAGEA4 | AMH | SYCP1 | CDC25A | XIST |
WFDC2 | BAG1 | CD247 | SPINK1 | KLK11 | SCGB2A1 | WT 1 | TRIM25 | CSF1R | KLK10 |
ERCC1 | BAG3 | CDC25B |
PRSS8
| KLK13 | SIX5 | OGP |
P11
| GADD45A | KLK15 |
KLK8 | BAG 4 |
DAB2
| CCNE1 | MVP | ZNF217 |
CDX2
| CYP2A | HLA-G | KLK5 |
RBL2 | OPN | HMGA1 | CEACAM6 | PARP1 | EYA2 | SMRP | PTK2 | JUP |
KLK7
|
SKP2 | Maspin | HOXB7 | ETS1 |
VEGFC
| ELF1 | Bcl-xL | TACC3 | MLANA | SOD2 |
IGFBP5
| MSN | BCHE | EPHA2 | ASNS | MUC5AC | TNFRSF1B | | | |
Human prostasin, a trypsin-like proteinase (40 KDa), is a glycosyl-phosphatidyl-inositol (GPI)-anchored extracellular serine protease. It is encoded by PRSS8, which is located on chromosome 16p11.2. Prostasin is also known as Channel Activating Protease 1. It was first isolated from seminal fluid, and is normally produced by the prostate gland. It is expressed in epithelial cells and ducts of the prostate [
13]. It is also present in low levels on the apical surface of epithelial tissues such as lung, kidney, liver, bronchi, colon, and salivary glands, indicating that it may have roles in multiple biological processes [
11]. Prostasin is present in multiple tissues that absorb sodium [
14]. It acts as a proteolytic activator of the epithelial sodium channel in vitro, and plays a major role in regulating sodium balance [
15‐
17]. Aberrant expression of prostasin is associated with many cancer types such as urinary bladder, uterus, prostate and ovarian, as compared to its level in corresponding normal tissue [
18‐
20]. However, activation of epithelial sodium channels by prostasin suppressed in vitro invasiveness of both prostate and breast tumor cells [
13,
21,
22]. Loss of PRSS8 in bladder cancer is associated with epithelial to mesenchymal transition – a process during which epithelial cells are converted to migratory and invasive cells [
23]. However, in OVC, it is difficult to deduce the role of PRSS8 based on its expression. The levels of PRSS8 in ovarian carcinoma cell lines are elevated, and the protein level is increased in the serum of late stage OVC patients. It was suggested that prostasin cleaves the extracellular domain of epidermal growth factor on epithelial cells; consequently, the receptor remains continuously phosphorylated and can potentially trigger metastasis [
11].
Serum prostasin was measured by microarray technology in 64 OVC patients and in 137 normal individuals [
24]. The serum prostasin mean level of detection was 13.7 μg/ml in OVC patients compared to 7.5 μg/ml in control subjects. Sensitivity and specificity of PRSS8 as a biomarker was calculated as high as 92 and 94 %, respectively. Moreover, post-operation levels of PRSS8 declined significantly in the majority of patients, indicating that PRSS8 may be potentially used not only as a diagnostic but also as a prognostic biomarker [
24]. Similarly, levels of PRSS8 mRNA were evaluated in 12 OVC patients and normal prostate tissues by RT-PCR and immunostaining [
11]. It was found that PRSS8 levels were 120 to 410–fold higher in OVC patients than normal controls [
11]. PRSS8 levels in OVC cell lines were shown to be linked to regulation by zinc-finger protein 217 (ZNF217). This protein is commonly overexpressed during cancer progression, and promotes tumor cell survival. Silencing of the ZNF217 gene in the OVC cell line HO-8910 resulted in a nearly 8-fold down-regulation of 164 genes including PRSS8. WFDC2 (HE4), which is currently used as an early detection biomarker for OVC, was also found to be downregulated [
25]. Results from these studies placed PRSS8 on the list of potential biomarkers for early detection of OVC [
26].
Our study presents evidences to demonstrate that PRSS8 can be used as an early detection biomarker for OVC. CA125 is a common OVC biomarker used in the clinic; however, as discussed above, although it is widely expressed on tumor cells, CA125 demonstrates low sensitivity. However, in combination, CA125 and PRSS8 increased the sensitivity to 92 % and specificity to 94 % [
24]; whereas, sensitivity of CA125 and PRSS8 alone was 64.9 and 51.4 %, respectively. In the same study, it was also shown that there is a low correlation between expression of CA125 and PRSS8, which is consistent with their function in different pathways; therefore, as biomarkers, they may be complementary. In our recent review, we indicated that CA125 and PRSS8 signal through multiple signaling pathways, including PI3K, AKT, ERK [
27]. It will, therefore, be interesting to investigate the mechanism of the synergistic effect of PRSS8 on CA125 as an early detection biomarker. The results of our current study indicate that PRSS8 is absent in normal ovarian tissues at the gene and protein level, and is upregulated from very early stages of the disease. Thus, PRSS8 exhibits properties of a complementary biomarker to CA125 for early detection of OVC.
Methods
Ethics approval
The continuing review # CR00003202 for the ovarian study protocol # Pro00002901 was approved by the Institutional Review Board (IRB).
Cell culture
A library of OVC cell lines and corresponding normal ovarian cells were obtained and cultured in specified conditioned media as described previously [
12]. Briefly, OVC cells lines used were TOV112D, OV-90, CAOV3, SKOV3, PA-1, SW626, and ES-2 (ATCC, Manassas, VA); SKOV-1, IGROV-1, HEY, A2780, and 2008 (S. Howell, UCSD); UCI-101 and UCI-107 (P. Carpenter, UCI); DOV-13 (R. Bast, MD Anderson Cancer Center, TX); 2774 (J. Wolf, MD Anderson Cancer Center, TX); BG-1 (Dr. K. Korach, NIH, NC); normal ovarian epithelial cell lines FHIOSE118 (J. Cheng, Moffitt Cancer Center, FL) and IOSE523 (N. Auersperg, University of British Columbia, Canada).
Immunoblot
Serum samples were depleted of abundant proteins by Affinity column ProteoPrep Blue Albumin and IgG Depletion Kit as described by the manufacturer (Sigma-Aldrich, St. Louis, MO). Protein concentrations were determined by Bradford Assay (Bio-Rad Laboratory, Hercules, CA), and 20 μg of total protein was resolved by 12.5 % SDS-PAGE for immunoblot analysis. Sera from OVC patients were purchased from Proteogenex (Culver City, CA) and Bioserve (Beltsville, MD). Nine normal sera were pooled to represent normal donors. A custom-made PRSS8 antibody (Precision Antibodies, MD), HRP-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) and SuperSignal West Dura (Thermo Fisher Scientific, Rockford, IL) were used for visualization of protein bands.
Immunohistochemistry and In situ hybridization
Whole mount tissues and tissue arrays containing different stages and subtypes of OVC were purchased from US Biomax (Rockville, MD) and Proteogenex. Tissue sections were de-paraffinized using Histochoice clearing agent (Amresco, Solon, OH) for 5 min followed by hydration steps with 100, 90, 70, and 50 % ethanol for 5 min each. After equilibrating with PBS for 5 min, the tissues were incubated with high pH solution (Amresco) at 95 °C for 20 min to retrieve antigens. The sections were cooled and washed with PBS for 5 min, and endogenous peroxidases were blocked by incubating in 3 % H2O2 for 15 min. The sections were marked with a hydrophobic PAP pen (Vector Labs, Burlingame, CA), blocked for 30 min at 37 °C in 5 % BSA in PBS/0.1 % Tween-20, and then incubated with the primary antibody (SC-136272, Santa Cruz Biotechnology) overnight at 4 °C. The sections were washed thrice in PBS/0.1 % Tween-20 for 5 min each. The tissues were incubated with secondary antibody (715–506-151 Jackson Immuno Research Laboratories) for 30 min at RT and washed as above. A DAB kit (Vector Labs) was used to visualize the antigen. Color development was interrupted by washing with distilled water for 5 min. Hematoxylin (Amresco) was used as the counterstain. Sections were dehydrated in ethanol solutions in the sequence of 50, 70, 90, 100 for 5 min each and 5 min in Histochoice clearing agent. After mounting the tissues (Permount, Vector Labs), the stained tissues were photographed using an Axio Imager Microscope (Carl Zeiss, Thornwood, NY).
In situ hybridization was performed as we previously described [
12]. The probe sequence for PRSS8 was; 5’- DIGN-GCAGTAAAACTCCTGACTCTCA.
qRT-PCR
Total RNA was extracted from cells using Trizol (Invitrogen, Carlsbad, CA), and cDNA was generated with the SuperScript III RTS First-Strand cDNA Synthesis Kit (Invitrogen). All primers were custom synthesized to be used with an ABI7900 RT-PCR instrument (Applied Biosystems, Foster City, CA) as recommended by the manufacturer. Primers were further validated using end-point PCR of cDNA generated from the normal ovarian cell lines; all primers produced a single band with the expected size as visualized on an agarose gel. The primers were typically 20-mers having a Tm of 58 °C. For qPCR, 43 ng cDNA, 10 pmole primers, and SYBR Green PCR Master Mix (Applied Biosystems) were combined in a 20 μl reaction volume. All qPCR were performed in MicroAmp Fast Optical 96-Well Reaction Plates with Barcode (Applied Biosystems) in the standard mode (1st denaturation at 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min). The qPCR data were normalized with GAPDH, and further analyzed using software provided with the ABI7900. TissueScan Cancer Survey Panels (OriGene, Rockville, MD) were used for screening 22 different human cancer types (over 380 biospecimens), and Ovarian Cancer Panels I-IV (OriGene) were used for determining the expression levels of genes at various stages, grades, and subtypes of OVC (over 190 biospecimens). TissueScan Cancer Survey Panels were purchased in a 96-well format with lyophilized cDNA from various patients with different cancer types. Each well of the plate contained 2–3 ng of cDNA, and the plate was divided to scan two genes. The reaction mix was transferred to a ‘Fast Plate’, compatible with the ABI 7900 HT RT-PCR instrument. After dividing each plate into two ‘Fast-Plates’, each reaction consisted of approximately 1–1.5 ng of cDNA. The conditions used were: 1st denaturation at 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. The data from TissueScan panels were normalized with beta-Actin. All cancer tissues used in these panels contained an average of 75 % cancer cells and 25 % surrounding stromal components.
Discussion
Ovarian cancer causes the death of over 125,000 women worldwide each year, which is more than all other gynecologic cancers combined. Women visiting the clinic with apparent symptoms are usually categorized with late stage (III-IV) OVC. Less than 20 % of all reported OVC cases are diagnosed in early stages, primarily because of the complexity of the disease and lack of specific biomarkers. In this report, we show that PRSS8 is a potential biomarker that is up-regulated in OVC at all stages, grades, and major subtypes.
More than a hundred potential biomarkers for OVC have been identified via multiple “-omics” methods (Table
1). In our work, to simplify screening without using precious biomaterials from OVC patients, a library of 21 ovarian cell lines (Table
2) was used in this initial phase to screen candidate biomarkers. PRSS8 was identified based on its robust and consistent overexpression in the majority of those OVC cell lines (Fig.
1). This robust overexpression signature was further validated in OVC patient samples, where we found differential expression of more than 100 fold compared with normal epithelial ovary tissues (Fig.
1a). PRSS8 was also significantly upregulated in urinary bladder cancer but downregulated in pancreatic and stomach cancer, suggesting that the expression of PRSS8 in tumors may be related to the specific cell or tissue type of tumor origin. The overexpression of PRSS8 and the abundance of prostasin in OVC tissues at early stages and low grades showed that both are excellent candidates as early detection biomarkers. We have previously demonstrated that KLK6 and KLK7 can serve as ovarian cancer-specific biomarkers. These also exhibited selective upregulation in OVC (12). It is likely that a combination of PRSS8, KLK6, and KLK7 can provide additional specificity and sensitivity for early detection of OVC.
The absence of PRSS8 and prostasin in normal epithelium and stroma indicates that gene and protein expression are tightly regulated in non-cancerous tissues. The significant overexpression profile at the onset of OVC and maintenance of this signature throughout OVC progression suggest that prostasin function may be required for maintaining the OVC phenotype (Fig.
2). We did not find significant differences between different stages and grades of OVC in the samples tested, indicating that PRSS8 and prostasin can be used as screening biomarkers for every stage and grade, including late stages and high grades, of most OVC subtypes. PRSS8 overexpression in borderline OVC may indicate that PRSS8 can also be used to detect low-incident OVC subtypes (<15 % in US). Median levels of PRSS8 gene expression were highest in borderline and clear cell OVC, followed by serous, papillary serous, and endometriod subtypes, indicating that PRSS8 expression is cell type-dependent within OVC subtypes (Fig.
3). We observed a robust overexpression profile of the PRSS8 gene in all OVC subtypes. The median overexpression was more than 100 fold suggesting that PRSS8 is an excellent candidate for early detection of OVC. The PRSS8 overexpression profile was largely maintained and translated into high protein expression in all stages and grades of OVC (Fig.
4), indicating that PRSS8 and prostasin can be used in OVC biopsy and small size samples. Using the combination of PRSS8/prostasin, KLK6, and KLK7 may provide a valuable diagnostic tool applicable for use on small OVC tissue samples available from clinical procedures. An additional analysis was that age does not influence upregulation of PRSS8 across the normal, benign and OVC tissue samples (Table
3). This further contributes to the overall strength of PRSS8 as a universal biomarker for early detection of OVC.
The immunohistochemical analysis in this report indicates that prostasin is downregulated not only in normal ovaries but also in several types of cancerous tissues that are in close proximity to the ovaries, such as the omentum and the uterus. In tissues tested, prostasin detected in normal and benign tissues was located in the membrane, but in OVC tissues prostasin was localized in the cytoplasm and nucleus, suggesting that cellular translocation of prostasin may be involved in OVC progression. In these test settings, the majority of benign tissues analyzed were of theca cell tumors (data not shown). Theca cells are endocrine cells that play an important role in fertility by producing androgen substrates that are key to estrogen biosynthesis [
30]. Endocrine infertility is commonly caused by excessive proliferation of theca cells and ovarian hyper-androgenism, indicating that PRSS8 levels may be affected by hormonal changes and balance. In a genome-wide study aimed at identifying estrogen response elements (ERE), it was shown that these elements are also found in the coding sequence of PRSS8; the presence of a high-affinity binding site for estrogen suggests that estrogen may control PRSS8 expression [
31], thus, elevating the level of prostasin in tissues. In a recent study, a regulatory network analysis of the estrogen receptor in a model of renal cell carcinoma indicated that estrogen may be involved in regulation of oncogenes and tumor suppressor genes, including PRSS8 [
32].
Although the cohort was small, our analysis of serum indicated that prostasin was largely absent in normal donor sera and benign OVC serum samples but was frequently abundant in OVC serum samples. In our analysis, samples that showed positive results were derived from thecoma patients, and the role of prostasin in benign ovarian samples is not well determined. PRSS8 was previously suggested as a potential biomarker for OVC at benign stages [
33], and our study further validated these findings. We would also emphasize that early detection methods for screening the general population should be non-invasive or minimally invasive method because the population without clinical symptoms most likely will not participate in any invasive clinical procedures. Blood tests are ideal for screening of asymptomatic patients during routine clinic visits. prostasin is a known secreted protein, and is detected in multiple human biological fluids, including peripheral blood, and thus, would be an excellent candidate as a serum biomarker for early stage OVC.
Earlier studies demonstrated upregulation of PRSS8 in early stages of OVC [
9,
21], but some reports showed conflicting data [
34,
35]. The abundance of prostasin in serum of OVC patients showed significant potential to be used as OVC biomarker but required strict maintenance of standardized conditions for accurate analyses [
33,
36]. Prostasin level does not change in urine before and after menopause; oral contraceptives or estrogen and progesterone therapy tend to increase PRSS8 levels albeit not significantly [
37]. However, an inaccurate diagnosis may be made in circumstances where abnormal hormone levels occur but are associated with stress. In a study where a large cohort (
n = 500) of OVC samples was assessed [
33], the need for standardized conditions was emphasized. In that study, prostasin was one of nine selected OVC serum biomarkers, and presented the highest discriminatory value (
P < 0.001) compared to benign cases. Similarly, our normal donor controls were procured during typical routine ‘clinical visits’. To further increase the sensitivity and specificity of prostasin detection, we generated a custom-made anti-prostasin antibody against prostasin (Additional file
1). The antiserum included a high titer antibody with high specificity to prostasin in serum samples derived from normal donors, benign, and OVC serum samples. Elevated prostasin levels in early stage OVC serum samples indicated that the prostasin secretion pathway was active, and that significant overexpression of the PRSS8 gene was translated to elevated prostasin levels in circulation. Thus, the abundance of prostasin correlates with overexpression of PRSS8 in early stage OVC. This correlation may be useful for initial population screening by prostasin, and for further clinical evaluation by PRSS8/prostasin analyses of ovarian biopsies.
Cancer Antigen 125 (CA125) is widely used in the clinic as a serum biomarker for OVC because it is elevated significantly in late stages. However, CA125 lacks specificity and sensitivity for early stage OVC as it detects less than 23 % of cases in stage I, while detecting greater than 80 % in late-stage OVC [
38,
39]. CA125 is also frequently upregulated in benign conditions (e.g., endometriosis, fibroids, etc.) and during ovulation; thus, CA125 lacks accurate diagnostic value for early stage disease in pre-menopausal women. Human Epididymis Protein 4 (HE4) has better sensitivity and specificity than CA125 for early detection of OVC [
40]. A combination of HE4, CA125, carcinoembryonic antigen (CEA), and vascular cell adhesion molecule (VCAM)-1 in an assay panel has been tested for detecting early stage OVC versus benign tumors, and achieved 86 % sensitivity [
41]. OVA1 and other OVC biomarker tests represent an effort to increase statistical power of early detection of OVC. In a recent study, PRSS8 showed significant synergy for increasing sensitivity and specificity when it was combined with OVA1 and tested against tissue samples from all stages of OVC [
33]. That study tested over 200 biomarkers (including CA125 and HE4) but none were sufficiently informative to be sole biomarkers for the broad applications and subtypes of OVC presented. Similarly, PRSS8 did not discriminate among OVC subtypes in our study, but the expression levels were significantly lower in clear cell and mucinous subtypes. In general, PRSS8 presented a poor correlation with CA125 and a moderate correlation with HE4 (
p = 0.463), further supporting the idea that HE4 is a better early detection OVC biomarker than CA125. Both panels of the 5 (OVA1) and the 9 (including PRSS8) biomarkers performed better (as measured by AUC values) for post-menopausal women compared to pre-menopausal women [
33].
A recent trend in early detection of OVC cancer is to measure longitudinal individual changes in levels of potential biomarkers [
42]. For example, the United Kingdom collaborative trial of OVC screening (UKCTOCS) followed more than 200,000 women, 50-years old and older, and compared the impact of screening by detecting CA125 levels vs. ultrasound, and then correlated the findings with OVC disease outcomes. The data were encouraging, but the conclusion was that additional biomarkers should be added to CA125-based screening to achieve a better clinical outcome. Prostasin can potentially be used for population screening by serum testing. Prostasin-positive patients could be guided to the clinic for further evaluation, where PRSS8 gene levels and prostasin protein levels could be measured in ovarian biopsies for diagnosis. Based on our current data on PRSS8/prostasin and on our previous report on KLK6 and KLK7 [
12] we believe that these members of the kallikrein family can potentially be combined with PRSS8/prostasin for early detection and screening for OVC. Future studies should include a large cohort of OVC tissues and serum samples to further validate the use of PRSS8 and prostasin, especially with KLK6, KLK7, HE4, OVA1, and CA125.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AT and UP designed the study and wrote the manuscript; AT, UP, AH, SB, AA, JD, DS and SY performed the experiments; TT provided the clinical samples; AT, AG, SB and UP analyzed the data; AG and AP reviewed the paper and provided advice. KSS designed the study, supervised and wrote the manuscript. All authors read and approved the final manuscript.