M-CSF in a new biomarker panel with HE4 and CA 125 in the diagnostics of epithelial ovarian cancer patients

The search for an effective screening test for ovarian cancer has been the focus of intensive research efforts. Blood flows into and out of tumors and circulates tumor-specific protein profiles, making serum or plasma the ideal biological media for finding a screening biomarker [13]. The co-expression of M-CSF and its transmembrane tyrosine kinase receptor has been detected in epithelial ovarian carcinoma and could be involved in the autocrine growth stimulation of this type of cancer. M-CSF, known as a colony stimulating factor 1 (CSF-1), is also a potent chemoattractant for monocytes, which in turn, can produce factors that stimulate proliferation of ovarian tumor cells including interleukin-1 (IL-1), interleukin-6 (IL-6) and tumor necrosis factor (TNF). Clinical specimens from ovarian cancer metastases display strong immunostaining for both CSF-1 and its receptor in contrast to noninvasive borderline tumors and to benign ovarian tissue [27,29]. HE4 is a novel protein and one of the more promising biomarkers for improving the diagnostic performance in ovarian cancer detection. This is a precursor to the epididymal secretory protein E4. It is overexpressed in ovarian carcinomas, but there is minimal expression in normal ovarian tissue. HE4 promotes migration and adhesion of ovarian cancer cells. It was reported that HE4 could be used as a biomarker for ovarian cancer with a specificity higher than that of CA 125. This protein was particularly highly expressed in histologic subtypes of serous or endometrioid ovarian carcinoma [6-8].

In this study we investigated the diagnostic usefulness of M-CSF used alone and in combination with HE4 and CA 125 in patients with EOC before surgical intervention or chemo- and/or radiotherapy. Furthermore, we estimated the diagnostic utility of these biomarkers in correlation to the stage and histological type of ovarian cancer.

Our results show that M-CSF as well as HE4 and CA 125 plasma levels in the total group of ovarian cancer patients were statistically significantly higher in comparison to the group of healthy controls. Data regarding M-CSF obtained in the present study are also in agreement with our previous studies [10] and with research results of other authors who compared ovarian cancer patients with healthy volunteers, although there were differences in the number and the composition of the groups tested [32-34]. Significantly elevated levels of M-CSF have also been found in the sera of patients with malignancies of the reproductive organs [32,35,36], breast [37] or pancreatic and ampullary cancer [26]. Moreover, it was observed that M-CSF concentrations were statistically different in every group (the analysis related to the stage of advancement of OC) compared to the healthy subjects. Similar results for M-CSF were confirmed in our previous investigations [10] and in the studies of other authors [34], although their observations regarding the highest concentrations of M-CSF related to the early stages of the disease and concerned a larger number of sub-types of EOC than our current study (serous, endometrioid, mucinous, clear cell). It is worth noting that Suzuki and colleagues [38] found no significant difference in M-CSF levels between early and advanced stages of OC, and this is in opposition to the results of our study in which a group of postmenopausal women was studied [10]. It should also be emphasized that M-CSF and its receptor gene, c-fms, were expressed in gynecological malignancies, and the co-expression of M-CSF and c-fms might be related to progression to the metastatic state [27] and is associated with poor prognosis [30,38]. Furthermore, it was observed that 56% (14/25) of the patients with clinically evident OC and normal levels of CA125 had elevated levels of M-CSF [39]. Our results are consistent with the results of other authors who observed increased levels of HE4 and CA 125 in the group of ovarian cancer patients compared to the healthy women group [8,40], although the groups tested were far smaller (60 and 30 women respectively). Similar results were obtained by Molina and others [41] who observed higher concentrations of the aforementioned tumor markers in ovarian (111 patients) as well as in other active gynecological cancers (32 patients) or in different benign gynecological lesions (285 subjects). Furthermore, other researchers have found circulating concentrations of HE4 and CA 125 significantly higher in patients with early and late stages of cancer compared with healthy women, although there were differences in ethnical characteristics of the population selected [42,43]. The analysis of research results published by other authors has revealed a finding, almost identical to ours, regarding significant differences in concentration levels of both tumor markers which were clearly related to the stage of the disease, with significantly higher concentration levels in the advanced stages III-IV than in stages I-II [41,44].

In line with our previous report [10] and others publications [32,41,45] there was a difference in M-CSF concentrations between the EOC patients and the benign (cysts) tumors groups. Gaducci et al. [45] discovered that preoperative serum M-CSF levels were significantly elevated in patients with epithelial OC when compared to those with a benign ovarian disease (cases of serous, mucinous and endometriotic cysts and also of fibromas, thecoma and benign cystic teratomas were analysed). Suzuki et al. [38] confirmed these findings, although malignant germ cell tumors of the ovary and mature cystic teratomas were studied. Furthermore, Burke et al. [46] revealed mRNA expression for M-CSF and its receptor in a majority of malignant and benign ovarian biopsies with minimal to no expression in the normal epithelium. Takagi et al. [32] observed serum M-CSF levels in this type of cancer significantly different from those in patients with a benign ovarian tumor (p < 0.01) and with leiomyoma (p < 0.001) as well as between endometrial (stages Ib-III) and cervical (stages III-IV) cancer patients and patients with leiomyoma (p < 0.05). Comparable results were obtained in studies conducted on patients with breast malignancies [37]. The findings of the current study regarding comparative tumor markers correspond to the findings reported in the existing literature [43,47,48] and to our previous publications [10,36]. Significantly higher serum concentrations of HE4 and CA 125 (p = 0.005, p = 0.001, p < 0.0001, p < 0.001, respectively) [6,41,49,50] were found in OC patients than in those with benign diseases, although there were differences in the menopausal status and composition of the groups compared. In addition, HE4 showed a greater than CA 125, satisfactory ability to distinguish ovarian cancer from endometriosis [51].

In contrast to the results published by Suzuki et al. [29] we observed significantly higher plasma levels of M-CSF in patients with a benign ovarian tumor than in healthy controls. These findings are similar to the results obtained in our previous studies conducted on female patients with uterine myoma [36]. Our present observations concerning both tumor markers are in agreement with published evidence, although HE4 is reported to be less frequently elevated than CA 125 in benign gynecologic disorders [8,52,53]. Furthermore, patients with serous epithelial OC displayed significantly different concentrations of all the parameters studied than those with endometrioid epithelial OC. We also demonstrated significant differences in the concentrations of the parameters tested in every group of OC sub-types vs patients with cystis endometrioides and with cystis serous. Comparable results, but only for CA 125, were published by Nolen et al. [54], although benign cases included a broad spectrum of non-malignant lesions representing a variety of histological origins. Significant differences were also observed, but only for HE4 and CA 125, between the patients with cystis endometrioides and cystis serous. These data are at variance with other publications [28], and may have been partly influenced by study participant selection, although similar results have recently been reported [7,41,55,56]. Unfortunately, we could not confirm our findings regarding M-CSF in other publications, since no reports on the subject are available.

The correlation between M-CSF, HE4 and CA 125 levels was estimated by Spearman’s rank correlation test and most of the measured values tended to increase for every marker. However, the degree of the correlation was not particularly strong (R: 0.32–0.47), and there were some discordant results. This suggests that each marker was elevated concurrently or under different conditions, and these results also support the necessity of combining the three markers. We found similar results in the existing literature [8,10,52].

In our analysis of postmenopausal subjects, we found that M-CSF alone provided the highest SE, PPV and NPV, (68–88%, 91–93%, 75–87% respectively) of any individual biomarker tested (with the exception of HE4 - stage I and CA 125 - EOC endometrioid sub-type). Our results are in agreement with the investigation of Skates et al. [33], whose study group comprised only 60 patients - 50% fewer than our study group, and with the publications of other authors [57] as well as with our previous studies [10,58]. The diagnostic SP was very high and reached the value of 94% for M-CSF and HE4, and 92% for CA 125 and these results are in accordance with [36,59] or at variance with [10] our earlier papers and the existing literature [41,52] but they may have been partly influenced by study participant selection.

Nonetheless, a multi-marker approach appears to hold promise for detecting early stage ovarian cancers [6]. Despite a decrease in specificity (86%) and PPV (85%), when M-CSF was combined with conventional markers, sensitivity levels improved to 80% - stage I, to 92–100% in stages II-IV and to 91% or 96% in every histological sub-type of OC. Moreover, in the present study we obtained a better outcome for the diagnostic SE than that attained by Skates et al. [33] and Zhang et al. [34] who found that multiple-marker panels (M-CSF, CA72-4 (carbohydrate antigen 72–4) and CA 125 - 70%; M-CSF, CA72-4, CA 125 and CA15-3 (carbohydrate antigen 15–3) - 71% respectively) significantly increased preoperative early-stage sensitivity. A comparable increase in the diagnostic SE was previously observed, although it regarded the combined analysis of VEGF with CA 125 and HE4 [59]. Similarly, NPV levels were improved to 74% - stage I, to 91–100% in stages II-IV and to 96% or 92% in the serous and endometrioid sub-types, substantially higher than NPV levels provided by either marker used alone. These observations are in opposition to the results of other authors who observed distinctly higher values of these diagnostic criteria for HE4 than for CA 125 in premenopausal subjects [60], or higher values of NPV for HE4 or CA 125 alone (72% and 92%, respectively) in a group of Asian women [61]. We were unable to compare our findings regarding the aforementioned panel of biomarkers with the findings of other authors, since no reports on the subject are available. The results of the current study support our previous findings regarding the usefulness of M-CSF and CA 125 in the diagnostics of this malignancy [10].

The AUC numerically describes the overall performance of a marker, with the AUC of 1 indicating a perfect SE and SP. We observed that M-CSF had the highest, comparable to conventional markers, diagnostic power in the groups with stages I-III of OC (0.7676, 0.8662, 0.9336, respectively) and in the patients with serous sub-type of EOC (0.9058). We demonstrated that the utilization of a combination of all the markers under investigation consistently provided the highest sensitivity for the detection of every stage, and particularly early stages, of EOC with the AUC value of 0.9069–0.9894, which was superior to the value of either marker used alone. Furthermore, we observed that the areas under the curve for the cytokine tested and the comparative markers were statistically significantly larger compared to AUC = 0,5 - borderline of diagnostic usefulness of the test. Test data published by Zhang et al. [34] are at variance with our present results (AUC for: CA 125–0.908, M-CSF - 0.792), although in a few previous studies the AUC values of CA 125 and HE4 for differentiating EOC were 0.82–0.95 and 0.85–0.96, respectively [6,49,50,52,56,62] and were significantly higher when the markers were used in combination [50], the results similar to our findings.