The levels of soluble cMET ectodomain in the blood of patients with ovarian cancer are an independent prognostic biomarker

Abstract

The tyrosine kinase mesenchymal-epithelial transition (cMET) is typically overexpressed in up to 75% of patients with ovarian cancer, and cMET overexpression has been associated with poor prognosis. The proteolytic release of the soluble cMET (sMET) ectodomain by metalloproteases, a process called ectodomain shedding, reflects the malignant potential of tumour cells. sMET can be detected in the human circulation and has been proposed as biomarker in several cancers. However, the clinical relevance of sMET in ovarian cancer as blood-based biomarker is unknown and was therefore investigated in this study. sMET levels were determined by enzyme-linked immunosorbent assay in a set of 432 serum samples from 85 healthy controls and 86 patients with ovarian cancer (87% FIGO III/IV). Samples were collected at primary diagnosis, at four longitudinal follow-up time points during the course of treatment and at disease recurrence. Although there was no significant difference between median sMET levels at primary diagnosis of ovarian cancer vs. healthy controls, increased sMET levels at primary diagnosis were an independent predictor of shorter PFS (HR = 0.354, 95% CI: 0.130-0.968, P = 0.043) and shorter OS (HR = 0.217, 95% CI: 0.064-0.734, P = 0.014). In the follow-up samples, sMET levels were prognostically most informative after the first three cycles of chemotherapy, with high sMET levels being an independent predictor of shorter PFS (HR = 0.245, 95% CI: 0.100-0.602, P = 0.002). This is the first study to suggest that sMET levels in the blood can be used as an independent prognostic biomarker for ovarian cancer. Patients at high risk of recurrence and with poor prognosis, as identified based on sMET levels in the blood, could potentially benefit from cMET-directed therapies or other targeted regimes, such as PARP inhibitors or immunotherapy.

Keywords: biomarker; ovarian cancer; prognosis; soluble cMET; therapy monitoring.

Fig. 1

sMET levels among longitudinal sampling in ovarian cancer patients. Scatter plot showing sMET levels in healthy controls (n = 85), in ovarian cancer patients at primary diagnosis (n = 86), one week after primary surgery (n = 56), before platinum‐based chemotherapy (n = 67), after three cycles of chemotherapy (n = 56), after completion of chemotherapy (After Ctx, n = 68) and at disease relapse (n = 14). The black horizontal lines indicate the median sMET level in each group, with error bars showing the 95% confidence interval. P‐values correspond to the nonparametric two‐tailed Mann–Whitney U‐test for independent samples.

Fig. 2

Association of sMET levels with clinicopathological data of ovarian cancer patients. Scatter plots comparing sMET levels between (A) FIGOI‐IIIA vs. FIGO IIIB‐IV ovarian cancer (n _total = 86) and serous histology vs. nonserous histology (n _total = 86) (B) patients with and without residual tumour left after primary debulking at primary diagnosis (n _total = 86) or after surgery (n _total = 56). The black horizontal lines indicate the median sMET level in each group, with error bars showing the 95% confidence interval (CI). P‐values, according to the nonparametric two‐tailed Mann–Whitney U‐test for independent samples, are indicated (C + D). Spearman correlation analysis between sMET levels and (C) age (n = 86) or (D) serum CA125 levels (n = 84 patients with matching CA125 values at primary diagnosis) with simple linear regression is shown.

Fig. 3

Prognostic relevance of sMET at primary diagnosis to predict primary platinum resistance. (A) Results are shown from univariate and multivariate generalized linear model analyses to predict platinum resistance including odds ratio (OR), 95% CIs. The cut‐off (663 ng·mL⁻¹, P = 0.247, $n_{(>663\text{ng}·{\text{mL}}^{‐1})}$ = 13, $n_{(<663\text{ng}·{\text{mL}}^{‐1})}$ = 70) was determined as described in the Patients and methods section. (B) Scatter plot comparing sMET levels between primary platinum‐resistant ovarian cancer (n = 10) vs. primary platinum‐sensitive ovarian cancer (n = 73). The black horizontal lines indicate the median sMET level in each group, with error bars showing the 95% confidence interval (CI). P‐value according to the nonparametric two‐tailed Mann–Whitney U‐test for independent samples.

Fig. 4

Prognostic relevance of sMET at primary diagnosis and in the course of platinum‐based chemotherapy. Kaplan–Meier analysis comparing progression‐free survival (PFS) and overall survival (OS) of patients with a high sMET level vs. patients with a low sMET level (A) at primary diagnosis (n _total = 86) (B) before the onset of platinum‐based chemotherapy (n _total = 67) and (C) PFS after the first three cycles of platinum‐based chemotherapy (n _total = 56). P‐values (log‐rank, Mantel–Cox) and hazard ratio (HR) (Mantel–Haenszel) were calculated as described in the Patients and methods section. The following cut‐offs were used as follows: primary diagnosis (PFS) = 246 ng·mL⁻¹, primary diagnosis (OS) = 308.2 ng·mL⁻¹, before chemotherapy (PFS) 267.7 ng·mL⁻¹, before chemotherapy (OS) 567.1 ng·mL⁻¹ and after three cycles of chemotherapy (PFS) = 792.8 ng·mL⁻¹. Patient grouped into low sMET vs. high sMET as indicated.

Fig. 5

Prognostic relevance of the patient individual sMET dynamics. Patient's dynamic curves showing the progression of sMET levels between primary diagnosis and the completion of chemotherapy. (A) Example of an individual patient with a high AUC and (B) example of an individual patient with a low AUC. (C) Kaplan–Meier analysis comparing progression‐free survival (PFS) and overall survival (OS) of ovarian cancer patients (n _total = 56) in the AUC high group vs. AUC low group. The median was used as cut‐off. P‐values (log‐rank, Mantel–Cox) and hazard ratio (HR) (Mantel–Haenszel) were calculated as described in the Patients and methods section.

Fig. 6

sMET level in recurrent ovarian cancer. (A) Spearman correlation analysis between sMET levels at primary diagnosis vs. at disease relapse (n = 14) with linear regression shown (red line). (B) Kaplan–Meier analysis comparing progression‐free survival (PFS) and overall survival (OS) of patients with a high sMET serum level vs. patients with a low sMET serum level at the time of first relapse (n _total = 14). The median was used as cut‐off. P‐values (log‐rank, Mantel–Cox) and hazard ratio (HR) (Mantel–Haenszel) were calculated as described in the Patients and methods section.