Background
Earlier diagnosis of ovarian cancer could significantly increase survival. The general 5-year survival rate of ovarian cancer patients is approximately 90% at stage I but only 5–60% at stage II or above [
1]. Located within a deep pelvic cavity, this epithelial malignancy usually arises with no typical bodily signs or symptoms. Therefore, the current modalities for detecting early-stage ovarian cancer rely heavily on transvaginal ultrasonography (TVU) and serum biomarkers, such as CA125 [
2]. Unfortunately, a malignant ovarian cyst with a diameter < 5 cm is difficult to diagnose via TVU and is usually morphologically indistinguishable from a benign cyst, e.g., serous/mucinous cystadenoma, endometriotic and simple cysts of the ovary [
3,
4]. Additionally, the cancer biomarker CA125 has limitations for discriminating early-stage ovarian cancer from benign diseases [
4,
5]. For example, it is difficult to determine whether elevated serum CA125 within the range of 35–65 U/mL is due to an unidentified ovarian cancer or a common CA125-secreting gynecological disease, such as a pelvic inflammatory cyst, endometrioma or cystadenoma [
5]. Hence, there is a risk of missing the substantial progression of existing cancer and, subsequently, a poor prognosis. Notably, both ovarian endometriotic cysts and cystadenoma have odds of developing into endometrioid cancer (or clear cell cancer; incidence rate: 0.2–0.8%) [
6] or cystadenocarcinoma (primarily serous carcinoma; incidence rate: 0.4–0.6%) [
7], which complicates the clinical settings for early diagnosis and places more obstacles on the TVU and CA125-based screening programs.
Efforts to improve the efficacy of TVU and CA125-based screening include the adoption of more effective subsidiary/surrogate cancer biomarkers [
8]. Of them, human epididymis protein 4 (HE4), an ectopically expressed sperm maturation-related protein, was found to be a potent substitute for traditional CA125 and is currently gaining popularity in gynecological oncology clinics [
9]. Pre-clinical studies have indicated that serum HE4 exhibited higher sensitivity and specificity for detecting ovarian cancer than CA125 [
10]. However, as more large-scale clinical trials were undertaken, the complexity of ovarian cancer and the variety of patient groups and reagent types need be considered in evaluating the efficacy of a serum HE4 test [
11]. In a Quality Assessment of Diagnostic Accuracy Studies-2 (QUADA-2)-based meta-analysis by Li et al. it was revealed that HE4 just outperformed CA125 in diagnostic specificity (93% vs. 78%) while the sensitivities of both were similar (79% vs. 79%) [
12]. More importantly, the area under the receiver operating characteristic (ROC) curve (AUC) was greater for serum CA125 than for serum HE4, which indicates that CA125 may have a superior diagnostic capacity if the cut-off is properly adjusted [
12]. Additionally, Jacob et al. asserted that the clinical benefit of a combined serum HE4 and CA125 test is not significant if the additional cost is accounted for [
13]. Therefore, the primary task for gynecological oncology researchers remains unchanged: to continue to explore more low cost, convenient and efficient subsidiary/surrogate biomarkers for early-stage ovarian cancer.
Both the synthesis and secretion rates of CA125 in ovarian cancer cells are keenly influenced by extracellular signals from circulating cytokines. Konishi et al. reported that the secretory levels of CA125 in cells cultured in vitro can be increased upon administration of epithelial growth factor (EGF) [
14]. We, therefore, hypothesized that the CA125 glycoproteins produced in benign and malignant conditions might be distinguishable from one another according to their responses to stimulatory cytokines. In the human body, the insulin release pulse, which is induced by elevated postprandial blood glucose levels, comprises a natural extracellular cytokine signal source. Previous works have clarified that the survival, division and proliferation abilities of ovarian cancer cells are all affected by circulating insulin levels [
15‐
17]. However, little is known about whether the postprandial pulse of insulin can induce a unique instant/transient waveform of serum CA125, by which the arsenal of diagnostic tools for clinicians to discriminate between benign and malignant ovarian cysts can be upgraded. Therefore, the aim of this study was to profile the characteristic patterns of postprandial fluctuations of CA125 in the peripheral blood associated with endometriosis and benign and malignant epithelial tumors of the ovary and to assess the applicability of utilizing postprandial increases in serum CA125 to detect early-stage ovarian cancer in a complex pathological condition.
Methods
Study population
From January 2015 through June 2017, patients were enrolled at two tertiary medical centers in Shanghai, China: The Department of Gynecological Oncology, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University and The Department of Gynecology, Obstetrics and Gynecology Hospital, Fudan University. The inclusion criteria were (1) an ovarian (adnexal) tumor (cyst) ≤ 5 cm in diameter under ultrasonography; (2) a scheduled exploratory laparotomy. The exclusion criteria were (1) any elevated serum AFP, CA199, or CEA level; (2) an ultrasonography-confirmed dermoid cyst (teratoma) or other germ cell tumor; (3) computed tomography- or MRI-confirmed abdominal metastasis or metastatic gastrointestinal cancer; (4) late-stage or nonepithelial ovarian cancer demonstrated by postoperative pathology; (5) a history of diabetes mellitus.
Cell culturing
The ovarian cancer cell lines SKOV-3, HO8910, ES-2, and OVCAR-3 were purchased from the Cell Bank, Chinese Academy of Sciences (Shanghai, China). Cells were cultured in DMEM (glucose = 1 g/L, Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% FBS (Thermo Fisher, Waltham, MA, USA) in 5% CO2 at 37 °C. The final concentrations of insulin (SinoBio, Shanghai, China), LY294002 (Selleck, Houston, TX, USA), AktVIII (Selleck), AZD6244 (Selleck), and CA125 standard (Thermo Fisher) were 100 nM, 20, 1 mM, 5 μM and 200 U/mL, respectively. The high glucose condition referred to a glucose concentration of 2 g/L.
Xenograft animal model
To build the cancer xenograft model, ovarian cancer cells (1 × 10
6) were subcutaneously injected into specific pathogen-free (SPF)-grade Balb/c nude mice weighting between 20 and 22 g. As tumors grew to 1 cm in diameter, mice were divided into four groups. (1) In the control group, each mouse was administered a dose of 0.2 mL saline p.o., and 0.2 mL saline i.h.; (2) in the glucose group, each mouse was administered 0.2 mL of 10% glucose p.o. and 0.2 mL saline i.h.; (3) in the insulin group, each mouse was administered 0.2 mL saline p.o. and 0.2 mL insulin (5 mU) i.h.; (4) in the glucose plus insulin group, each mouse was administered 0.2 mL of 10% glucose p.o. and 0.2 mL insulin (5 mU) i.h. The type I diabetes mouse model was built by administration of alloxan (Sigma-Aldrich) [
18]. Specifically, for each tumor-bearing nude mouse, 120 mg/kg alloxan was first administered via intraperitoneal injection, followed by a second dose of 80 mg/kg alloxan after a 24-h interval. The blood samples of treated mice were collected on day 3 via the tail-cutting method after a 12-h fasting period. Plasma glucose ≥ 11.1 mmol/L indicated that the diabetes nude mouse model was successfully generated.
CA125 assay
CA125 concentrations were detected in serum samples (fasting, postprandial or post-75 g oral glucose) from the patients (or animals) and in the supernatant (or cellular content) from cell cultures using the same reaction plate from a CA125 ELISA Kit (Thermo Fisher) according to the manufacturer’s instruction.
Protein labeling and deglycosylation
The standard CA125 protein (1000 U/mL, Thermo Fisher) was labeled with Alexa Fluor 488 NHS Ester (Thermo Fisher) according to the manufacturer’s instruction. The labeled CA125 was visualized at 517 nm using a 488-nm excitation wave. Removal of N-linked oligosaccharides of CA125 protein was achieved by digesting the protein with PNGaseF (New England Biolabs, Ipswitch, MA, USA) for 1 h at 37 °C.
RNA interfering
Mesothelin (MSLN)-specific and negative control (NC) siRNAs were purchased from GenePharma (Shanghai, China). MSLN-siRNA-1, forward: 5′-ccauuggaccugcugcuautt-3′, reverse: 5′-auagcagcagguccaauggtt-3′; MSLN-siRNA-2, forward: 5′-gccucaucuucuacaagaatt-3′, reverse: 5′-uucuuguagaagaugaggctt-3′. NC-siRNA, forward: 5′-uucuccgaacgugucacgutt-3′, reverse: 5′-acgugacacguucggagaatt-3′. siRNAs were transfected with Lipofectamine RNAiMAX (Thermo Fisher) according to the manufacturer’s instruction.
Reverse transcription (RT)-quantitative polymerase chain reaction (qPCR)
Total RNA was isolated from the cultured cells using TRIzol (Thermo Fisher) and then reverse-transcribed into cDNAs using a Universal cDNA Synthesis Kit (Exiqon, Vedbaek, Denmark). Real-time PCR was performed using the primer pairs for Mucin 16 (forward: 5′-ccccaaattccagaggtgaa-3′, reverse: 5′-tgacaaaggcgcactggtac-3′), MSLN (forward: 5′-cgccttgctttccagaacat-3′, reverse: 5′-attctgctgactgagcgcct-3′) and β-actin (ACTB, forward: 5′-agcgagcatcccccaaagtt-3′, reverse: 5′-gggcacgaaggctcatcatt-3′). SYBRGreen I dye (Thermo Fisher) served as a quantitative indicator in the qPCR reaction. An ABI PRISM 7900 Sequence Detection System [Applied Biosystems, Carlsbad, CA, USA) was used for PCR experiments. The cycle threshold (CT) of each qPCR assay was recorded, and the results were presented as 2−ΔCT (ΔCT = CT(target) − CT(ACTB)].
Western blotting
The primary antibodies used for recognizing the target proteins were as follows: mouse anti-human Mucin 16 monoclonal antibody (1:10, Cat. No. sc-365002, Santa Cruz, Dallas, TX, USA), mouse anti-human Mesothelin monoclonal antibody (1:10, Cat. No. sc-365324, Santa Cruz) and mouse anti-human β-actin monoclonal antibody (1:10, CW Biotechnology, Beijing, China), which were then detected using the secondary antibody provided by the Wes Mouse Master kit (ProteinSimple, San Jose, CA, USA). All the Western blotting experiments were performed using Wes (ProteinSimple). The intensities of the resultant bands were quantified and visualized using the “Compass for SW” software (ProteinSimple).
Statistics
Two-sided χ2 test/Fisher’s exact test and Student’s t test (paired-samples or independent-samples; and two-sided independent-samples t test was the default setting) were used for nominal and numerical data, respectively. A support vector machine (SVM) was generated using the “kernlap” package in R software (version 3.3.3). The sensitivity of a diagnostic criterion was defined as the percentage of correctly identified cases among all pathologically confirmed ovarian cancer cases. The specificity was defined as the percentage of correctly identified cases among all pathologically confirmed non-cancer cases. The positive predictive value (PPV) was defined as the percentage of correctly identified cases among all the cases that could be judged as ovarian cancer by a given diagnostic criterion. The negative predictive value (NPV) was defined as the percentage of correctly identified cases among all the cases that could be judged as benign diseases by a given diagnostic criterion. SPSS 18.0 software (IBM, Armonk, NY, USA) was used to complete the statistical analyses (including the ROC analysis). p ≤ 0.05 was considered significant. All animal and in vitro experiments were carried out in triplicate unless otherwise indicated.
Discussion
In this study, the enrolled population was a typical representation of patients encountered in a gynecological clinical condition. The gynecological diseases they suffered included four major classes of benign cysts, i.e., pelvic inflammatory cysts, adnexal retention cysts, ovarian endometrioma and benign/borderline cystadenoma, which need to be discriminated from early-stage ovarian cancer due to the high-risk manifestations, such as long-term abdominal pain (inflammatory cysts), progressive CA125 elevation (ovarian endometrioma, benign/borderline cystadenoma), rapid tumor enlargement (retention cysts) and postmenopausal pelvic mass (retention cysts) [
6,
7,
23,
24]. Among these diseases, for the first time, we demonstrated that ovarian cancer possesses a unique, distinctive pattern of postprandial increases in serum CA125. By using this increase as a surrogate biomarker, with or without the SVM-modified algorithm, early-stage ovarian cancer detection can be significantly improved (Table
1).
Accepted as a basic carcinogenesis theory, earlier studies have indicated that the PI3K-Akt pathway is frequently altered in cancer cells, which exhibits far stronger activity than it does in healthy cells [
25,
26]. For instance, in ovarian cancer, a gain-of-function PI3KCA single nucleotide alteration (SNA) and a PTEN copy number deletion are common gene mutations that cause enriched phosphatidylinositol 3,4,5-trisphosphate (PIP3) and intensified pAkt signaling [
26,
27]. Likewise, in our study, the data demonstrated that it is the PI3K-Akt pathway that invokes CA125 expression (Fig.
5a, b) and distinguishes the postprandial pattern of cancer-derived CA125 from those observed in benign diseases (Figs.
1 and
2). Mechanistically, the serum CA125 test works in a static mode, which provides rather limited information on the source of CA125. The only way to increase the specificity of this test is to increase the cut-off, e.g., 65 U/mL [
5]. However, for malignant cysts< 5 cm, this strategy is often less effective (Fig.
1a, b). In contrast, the CA125 postprandial increment test provides a dynamic strategy for discerning subtle differences of PI3K-Akt signaling pathway among various types of ovarian (adnexal) cysts. Therefore, the properties of unidentified cysts can be determined at a lower CA125 cut-off level, i.e., 20 U/mL, decreasing the missed diagnosis rate imposed by the original cut-off of 35 U/mL (Table
1).
Our animal and cancer cell models both revealed that, insulin, only when combined with a high-glucose condition can invoke the active release of CA125; otherwise, it promotes re-absorption of CA125 (Figs.
3 and
4). The high-glucose condition, hence, plays a critical role in switching the function of insulin. Theoretically, insulin facilitates the recycling of GLUT4, a key glucose transporter, in ovarian cancer [
28]; in turn, the accumulated intracellular glucose, through glycolysis and the Kreb’s cycle, offers an abundant reservoir of glucose-6-phosphate and energy molecules to the glycosyltransferase system that is responsible for CA125-core glycosylation [
29,
30]. The accelerated CA125 synthesis/secretion at the protein level, therefore, is attainable only under a high-glucose condition. This fact was manifested by the post-75-g oral glucose tests in the enrolled population, which revealed an ascending region of serum CA125 coincidentally accompanying the elevated insulin level and the consumption of blood glucose resources (Fig.
1d). Mesothelin, however, as another factor affected by insulin, contributes in the opposite manner (Fig.
5d). Previous studies demonstrated that a blockade of the interaction between Mesothelin and CA125 (using an MORAb-009 antibody) led to a 1.3- to 13.75-fold increase in CA125 in patient sera [
31]. We, herein, demonstrated that overexpressed Mesothelin mediates 1.1- to 19.4-fold greater CA125 re-absorption in insulin-treated cancer cells compared with that in control cells (Fig.
5e), while Mesothelin overexpression itself is also modulated by the PI3K-Akt pathway (Fig.
5a, b). Moreover, there was a characteristic over-drop segment of postprandial CA125 observed in the non-cancer patients (Fig.
1d); and in patients with a pelvic inflammatory cyst, an inverse relationship between average fasting and postprandial CA125 was established (Fig.
1b). With the above Mesothelin-based mechanism, we can reasonably attribute these phenomena to the vast Mesothelin-expressing system of peritoneal mesothelium, where CA125 re-absorption events can actively occur, and the glucose and insulin-invoked responses for CA125 synthesis/secretion may be insufficient [
32]. Additionally, considering that a number of monocytes and resident macrophages may infiltrate the inflammatory cyst tissue [
33], Siglec family proteins (i.e., Siglec-2, 7 and 9), which are an additional panel of monocyte/macrophage-specific CA125 receptors [
34‐
36], may also skew the CA125 secretion/re-absorption balance toward a decrease in extracellular CA125. As supporting evidence, we have confirmed an effect of glucose and insulin on the induction of Siglec-9 overexpression in the peritoneal resident macrophages (data not shown).
Although in this work, by including the postprandial increase in the serum CA125 test, the CA125 cut-off was successfully lowered and the efficacy of early-stage ovarian cancer detection was accordingly improved, we still noted that there were cancer cases missed, even after adopting a postprandial increment-based diagnostic criterion with the highest efficacy (i.e., CA125 increment ≥ 10%). Most of these cases (5/6) were characterized by CA125 > 105 U/mL (Fig.
2e), which means that the CA125 increment ≥ 10% rule is not applicable to the entire CA125 level space, especially for higher CA125 values. In other words, the relationship between fasting and postprandial CA125 levels should have abided by a non-linear rule in early-stage ovarian cancer. These conditions necessitated the adoption of an SVM in our study. The kernel function is the foundation of the SVM, and we selected the RBF, which requires fewer input parameters and exhibits higher stability [
20]. Our results demonstrated the utility of the SVM in improving diagnostic quality of postprandial increases in CA125 (Table
1 and Fig.
2f) and an ideal usage of these two detection tools in the clinical practice.
A limitation of this work is that the food-intake of each patient was randomized, which has not yet been standardized. Certainly, a quantitative glucose uptake test, such as an OGTT and an intravenous glucose tolerance test (IVGTT), would be beneficial to gain a more stable, reliable, and informative dataset on the postprandial CA125 fluctuation patterns in cancer and non-cancer patients. Therefore, further efforts should focus on investigating the applicability of standard OGTT (or IVGTT)-associated fluctuation pattern(s) of cancer-derived serum CA125 in large-scale patient populations.
Authors’ contributions
ZWG, MC and KQS collected serum samples and medical records from the enrolled patients. YFH, YZ and YTH completed the statistical analysis. YFH and QL carried out the animal and cancer cell-related in vitro and in vivo experiments. QL performed the laboratorial work of qPCR, CA125 assay and Western blotting. WD approved the study design and supervised the work progression. YFH wrote the manuscript. All authors read and approved the final manuscript.