Background
In 2012, it is estimated that 22,280 women will be diagnosed with ovarian cancer and 15,500 will die of the disease (SEER fact sheet). Ovarian cancer is considered a “silent killer” because of the absence of specific symptoms until late in the disease when 75% of the cases are diagnosed, five year survival rates are less than 30%, and a 70% recurrence rate is expected. Early diagnosis, when the cancer is confined to the ovary, can increase the 5-year survival rate to 90%. Because of the high fatality rate and relatively low prevalence of the disease, a sensitive and specific screening tool for asymptomatic women is needed. As such, effective and reliable diagnostic assays need to be highly sensitive and specific for the screening and detection of early stage ovarian cancer, especially in asymptomatic women.
CA125, a membrane-associated mucin found on the apical membrane of epithelial cells of the ocular surface, respiratory tract and female reproductive tract, is elevated in approximately 80% of women with late-stage ovarian cancer. It is the gold standard diagnostic marker to detect recurrent ovarian cancer and monitor response to treatment. However, the usefulness of CA125 as a marker cannot be extended to diagnosis as 20% of ovarian cancers do not express CA125 [
1], and elevated levels are detected in only half of early stage patients. Further, CA125 is detected in many benign gynecological conditions and is particularly unreliable in detecting ovarian cancer in premenopausal women [
2‐
5]. Additional biomarkers with high sensitivity and specificity for detecting ovarian cancer in the early stages of the disease are sought to complement CA125. Two promising markers are human epididymis protein 4 (HE4) and mesothelin (MSLN). CA125, HE4 and MSLN have been approved by the United States Food and Drug Administration (FDA) as biomarkers for recurrent ovarian cancer (CA125 and HE4) and diagnosis of mesothelioma (MSLN).
Human epididymis protein 4 (HE4), normally expressed in the epididymis, endometrial glands and respiratory tract [
6,
7], is up-regulated in both early and late stage ovarian cancer [
6‐
9] including 90% of serous carcinoma, and adenocarcinomas of the lung and endometrium [
10,
11]. It is not expressed in mucinous carcinoma [
12]. It has been widely studied as a biomarker, alone and in combination with CA125, for the diagnosis and monitoring of recurrent disease as little or no expression is observed in benign conditions [
8,
10,
13,
14]. When combined, HE4 has been shown to increase the sensitivity and specificity over CA125 alone [
2,
9,
11,
15], allowing for better detection of early stage ovarian cancer [
9,
15]. HE4 levels have, however, been shown to increase with age [
16].
Soluble mesothelin (MSLN) has a history as a biomarker for mesothelioma diagnosis, prognosis and monitoring [
17‐
19]. It is a differentiating antigen derived from a precursor protein that when cleaved yields Megakaryocyte Potentiating Factor (MPF), a 32 kDa excreted soluble protein [
20,
21], and MSLN, a 40 kDa GPI-linked glycoprotein that is also shed as a soluble form into the blood stream by frameshift mutation and proteolytic cleavage [
22,
23]. MSLN is hypothesized to be involved in cell adhesion and signaling [
24] and to contribute to the metastasis of ovarian cancer to the peritoneum by binding CA125 [
25,
26]. It is highly expressed in mesothelioma, ovarian and pancreatic cancers and lung adenocarcinoma [
7,
23,
24,
27‐
29], but only expressed normally in mesothelial cells of the peritoneum, pericardium and pleura [
27]. Like HE4, it has shown promise in the detection of early stage ovarian cancer, especially in combination with CA125 [
1,
30,
31]. Improvements may be gained adding MSLN to a biomarker panel with CA125 and HE4 to detect early stage disease [
15,
31] although contradictory results have been reported [
32,
33]. Levels of detection of MSLN from serum and plasma for ovarian cancer were shown to be similar [
34]. One issue with measuring MSLN in serum is that levels can be affected by conditions such as age, body mass index (BMI) and glomerular filtration rate [
16]. Less expensive, facile screening can be achieved screening urine over serum or plasma. It is of interest, therefore, that MSLN was detected with more sensitivity in urine than serum for both early stage (42% vs. 12%, respectively) and late stage (75% and 48%, respectively) disease [
35].
Although the literature is sparser than that for MSLN, a few studies have looked at MPF as a biomarker for mesothelioma [
36], ovarian cancer [
23] and pancreatic cancer [
37]. As biomarkers in mesothelioma, MSLN and MPF have been shown to behave similarly [
17‐
19].
Panels of biomarkers, able to cover the molecular heterogeneity of ovarian cancer [
31,
38], or specific to high grade serous carcinoma [
12], will be the most effective way to detect early disease for fewer fatalities. To date, no panel has been identified that can achieve the sensitivity (>75%) and specificity (>99.6%) needed to meet the accepted criteria of no more than ten surgeries for every case of early stage ovarian cancer identified. The most desirable biomarkers to add to panels will be those expressed early in disease, and not expressed in normal tissue.
One such biomarker is folate receptor alpha (FRA), a glycosylphosphatidylinositol (GPI)-anchored protein involved in folate transport into cells that is expressed in breast, lung, clear cell renal, ovarian and endometrial carcinomas, and non-small cell lung adenocarcinoma [
39‐
58]. FRA is expressed in a high percentage of serous ovarian carcinomas in all stages and grades [
44,
46,
55,
59,
60], and levels of circulating FRA have been shown to be comparable between early and late stage disease [
54,
56]. Expression of FRA in normal tissues is restricted to the apical surfaces of some polarized epithelial cells [
40].
In the present work, novel, sensitive electrochemiluminescent assays were developed for the soluble forms of FRA, MSLN and MPF and were evaluated in a large cohort of ovarian cancer patient samples. Further, the diagnostic utility of these markers was compared to CA125 and HE4 in a subset of serous ovarian cancer samples. Finally, a multi-marker logistic regression model was developed that demonstrates increased diagnostic performance relative to any single marker.
Discussion
CA125, first introduced in the mid-1980s, remains the gold standard with respect to detection and monitoring of ovarian cancer. In recent years other markers, in particular HE4 have gained acceptance, with somewhat limited utility [
64‐
67], and some work has been reported on combining CA125 and HE4 to further increase the diagnostic application of these serum markers [
2,
9,
11,
15]. Several other markers of ovarian cancer, including MSLN and, to a lesser extent, MPF have been described [
1,
15,
23,
30‐
33]. However, there are no reports to our knowledge that have performed as comprehensive an analysis of all of these markers as presented here.
FRA has been the subject of intense research as a potential therapeutic target in the last several years primarily because of its highly restricted expression profile in normal tissues [
40] and high levels expression in a number of cancers of epithelial origin, including serous ovarian cancer [
39‐
52,
54‐
58,
68,
69]. Several late-stage clinical trials in ovarian cancer and non-small cell lung adenocarcinoma are presently on-going [
68‐
70]. It is important, therefore, to develop robust assays for FRA both in tissue and in the circulation. With this in mind, the present work describes, for the first time, a specific and sensitive ECL-based assay using novel MAb reagents for the detection of FRA in serum, plasma and urine, allowing a comprehensive analysis of its diagnostic potential. Further, the potential clinical utility of serum MPF is not well documented. We therefore chose to develop a similar ECL-based assay for MPF for comparative studies not only to FRA, but to other more accepted markers of ovarian cancer – CA125, MSLN and HE4.
The described assays showed excellent limits of detection in the low pg/mL range and wide dynamic ranges up to at least 5000 pg/mL. From a practical point of view, such assays will require less in the way of repeat sample testing due to high marker levels. Importantly, FRA, MSLN and MPF were all shown to distribute equivalently between serum and plasma allowing flexibility in the choice of sample matrix. On the other hand, markers such as osteopontin are known to distribute more into the plasma fraction, restricting the sample type. FRA, MSLN and MPF were all shown to be detectable in urine samples from both healthy women and women with serous ovarian cancer. The clinical utility of MSLN measurement in urine for ovarian cancer has previously been described. In view of the ease of urine sample collection, the clinical utility of diagnostic assays assessing levels of FRA and MPF in urine is evident.
The data presented here demonstrates that for each of the markers analyzed – CA125, HE4, MSLN, MPF and FRA – there was a preferential expression in the serous histotype. Recent work from our laboratory using immunohistochemical techniques for the detection of FRA, have shown a similar preferential expression in serous carcinomas [
59]. Taken together, these data support the current understanding of the origin of the various histotypes of ovarian cancer with the most common serous histotype deriving from tubal fimbriae [
71]. Indeed, we recently demonstrated that FRA is highly expressed in tubal epithelium while normal ovary epithelium is devoid of FRA expression [
59].
The analyzed markers showed low to moderate correlations with each other. Surprisingly, CA125 and MSLN were not highly correlated even though CA125 has been described to be the ligand for MSLN and to be involved in the metastatic process in ovarian cancer. More surprisingly, perhaps, is the moderate correlation between MSLN and MPF since these molecules should be present in a 1:1 molar ratio given the fact that they derive from the same RNA and that MPF is simply a proteolytic product of the initial gene product. However, since MSLN is a GPI-anchored protein whereas MPF is soluble, these findings, as with the correlations of the other described markers, most likely reflect a combination of the route by which the markers enter the circulation and, importantly, the clearance from the circulation. For example, while MSLN, MPF and FRA are detectable in urine, CA125 is not. Further, FRA has been shown to bind to megalin, both in the kidney and liver and, as such, is removed from circulation [
72]. The glomerular filtration of at least some of these markers [
32,
35,
73,
74], as well as the potential for other clearance mechanisms, has been described previously and may be a confounding factor in the measurement and application of these markers. Ultimately, this may explain the performance of CA125 as the single best marker for ovarian cancer.
However, as described herein, combining markers with CA125 does increase the diagnostic performance of the marker panel over CA125 alone, allowing for the development of a multi-marker panel that increases the sensitivity of detection of early stage disease while retaining specificity, the ultimate goal in ovarian cancer diagnosis.
Conclusions
CA125 remains the best single biomarker for diagnosis and monitoring of ovarian cancer. However, additional markers are sought for use independently or in combination with CA125 to improve the sensitivity for ovarian cancer detection whilst retaining specificity. The current study presents data on the utility of a novel marker, folate receptor alpha, FRA; with respect to discrimination between ovarian cancer, for example, the serous histotype, and normal controls. Further, data was presented for additional markers including MSLN and MPF and the use of these markers in a multi-marker panel that outperforms CA125 alone.
Development of additional markers for use either individually or in a panel for the diagnosis or detection of ovarian cancer, especially early stage disease, is critical. The novel ECL assays described herein provide a powerful tool for such development.
Competing interest
The authors declare no conflicting interests in this study.
Authors’ contributions
DJO and ES conceived and designed the experiments. PO, RH and LM performed the experiments. RPT, LMP and DJO analyzed the data. LMP and DJO wrote the paper. All authors read and approved the final manuscript.