Preclinical models and molecular biomarkers. Tools to improve treatment in endometrial carcinoma

Background: Endometrial carcinoma is the fourth most common cancer in European women, and incidence is increasing. No major improvements in treatment or survival have been achieved over the last decades, and an individualized treatment approach may improve the likelihood of a beneficial outcome for endometrial carcinoma patients. Identifying robust biomarkers that can improve risk-stratification and help select patients likely to benefit from specific treatments is vital. There is also a need for representative preclinical models in order to discover and validate new targeted therapies. Aim: To investigate potential new biomarkers in endometrial carcinoma, and to develop robust and reliable preclinical models to improve translational research. Material and methods: Endometrial carcinoma cell lines or patient derived primary tumor cells were implanted in the uterus of female mice to generate orthotopic endometrial carcinoma mouse models. We then applied several advanced imaging techniques to monitor development of disease in these models, including optical molecular imaging (OMI), magnetic resonance imaging (MRI) and positron emission tomography/computed tomography (PET/CT). Patients treated for endometrial carcinoma have been prospectively entered in a biobank of gynaecological malignancies and clinicopathological data, imaging data and biological samples have been collected. Tissue samples were evaluated by immunohistochemistry and microarray in order to explore expression of asparaginase-like protein 1 (ASRGL1) in endometrial carcinoma, relating the level of ASRGL1 to survival as well as imaging- and clinicopathological parameters. Results: Bioluminescent imaging (BLI), MRI, and PET/CT using fluorine-18- fluorodeoxyglucose (18F-FDG) or fluorine-18-fluorothymidine (18F-FLT) as tracer were all found to be feasible methods for monitoring of disease in orthotopic mouse models of endometrial carcinoma. With the exception of BLI, these imaging modalities were also demonstrated to visualize uterine tumors in patient derived xenografts (PDX) (Paper I). Epithelial adherence molecule (EpCAM) was found to be highly expressed in both endometrial carcinoma cell lines and in hysterectomy specimens from patients diagnosed with endometrial carcinoma. Near infrared fluorescence (NIRF) imaging using an Alexa 680 fluorophore (AF680)-conjugated anti-EpCAM antibody and BLI generated comparable images of uterine tumors in cell line based orthotopic endometrial carcinoma mouse models. EpCAM-AF680 NIRF was however superior to BLI in early delineation of metastatic disease. EpCAM-AF680 NIRF imaging was found to visualize uterine tumors in multiple orthotopic PDX models, with better contrast and earlier detection of tumors compared to 18F-FDG PET/CT imaging. Additionally, in vivo EpCAM-AF680 NIRF imaging accurately depicted a nonsignificant therapeutic response following treatment with paclitaxel or trastuzumab in an orthotopic PDX model of endometrial carcinoma. (Paper II). Low ASRGL1 protein expression in endometrial carcinoma hysterectomy samples was validated as a strong prognostic biomarker with independent survival impact, both in the whole patient cohort and in patients diagnosed with endometrioid endometrial carcinoma. Low ASRGL1 mRNA level was found to be significantly associated with poor outcome. ASRGL1 expression was mainly intact in the precursor lesion complex atypical hyperplasia (CAH), while the majority of metastatic lesions had lost ASRGL1 expression (Paper III). Similar expression of ASRGL1 in corresponding curettage and hysterectomy samples were observed in 85% of patients where ASRGL1 status in post-operative specimens was known. Low expression of ASRGL1 in curettage was found to independently predict poor outcome in the whole patient cohort, as well as in patients with presumed low risk curettage histology. Significant associations between low ASRGL1 expression and preoperatively assessed features of aggressive disease were observed, including high-risk curettage histology, hormone receptor loss in curettage, and large tumor size on MRI. Low ASRGL1 expression in curettage was also found to be an independent predictor of lymph node metastases (Paper IV). Conclusions: We have successfully generated orthotopical mouse models of endometrial carcinoma, including PDXs with different histological backgrounds, and demonstrated that tumor development and response to treatment can be monitored by multiple advanced small animal imaging techniques. Low expression of ASRGL1 in hysterectomy samples has been validated as an independent prognostic biomarker in endometrial carcinoma, and we suggest ASRGL1 expression in curettage as a promising pre-operative biomarker with a potential to improve risk-stratification and surgical planning