Molecular Characterization of Metastatic Endometrial Cancer by Mass Spectrometry

Abstract

One of the most reliable prognostic factors in endometrial cancer is the presence of lymph node metastasis. Clinicians presently face the challenge that radiological imaging and conventional surgical-pathological variables such as tumour size, depth of invasion and grade of disease are unreliable in determining if the endometrial cancer has metastasized. Although only 10% of endometrial cancer patients suffer from lymph node metastasis, the majority of them undergo lymphadenectomy, which can be associated with significant complications including lower extremity lymphedema. Based on the assumption that metastasis is mainly determined by the properties of the primary tumour and its interaction with the surrounding tissues, a tissue based proteomic approach combining two complementary methods, peptide matrix assisted laser desorption/ionisation mass spectrometry imaging (MALDI MSI) and liquid chromatographytandem mass spectrometry (LC-MS/MS) was undertaken to identify molecular discriminators in primary endometrial cancers which correlate with lymph node metastasis. In a discovery approach, MALDI MSI was carried out on two tissue micro arrays (TMA), containing a total of 43 patients. Upon data acquisition, a canonical correlation analysis (CCA) based method was applied to rank the acquired m/z values based on their power to discriminate the primary carcinomas with and without metastatic potential. The highly ranked m/z values were able to classify 38 out of 43 patients (88.4%) correctly. The top discriminative m/z values were identified using a combination of in situ sequencing and LC-MS/MS from digested tumour samples. The differential abundance of the two identified proteins, plectin and α-Actin- 2 was further validated using data independent acquisition LC-MS/MS and immunohistochemistry. In a targeted approach, we aimed to improve the prediction model for endometrial cancer metastasis preoperatively. From publically available data and published research, we compiled a list of 60 target proteins with the potential to display differential abundance between primary endometrial cancers with lymph node metastasis versus those without. Using data dependent acquisition LC-MS/MS, we were able to detect 23 of these proteins in an independent cohort of endometrial cancer patients. Using data independent acquisition LC-MS/MS, the differential abundance of 5 of those proteins was observed (p < 0.05). Upon validation by immunohistochemistry, our data indicates that annexin A2 is upregulated while annexin A1 and alpha actinin 4 were downregulated in primary endometrial cancers with lymph node metastasis versus those without. The results of this immunohistochemistry analysis were used to generate a predictive model of endometrial cancer metastasis. Additionally the predictive model using highly ranked m/z values identified by MALDI MSI was generated and compared with other models containing the histopathological variables. However, when compared the MALDI MSI model showed significantly higher predictive accuracy than the model using immunohistochemistry data. Our results showed that the highly ranked m/z values identified from MALDI MSI data serve as new independent prognostic information beyond the established risk factors. The developed molecular classification tool has the potential to predict which tumours have metastasized and which patients would therefore benefit from radical surgery while avoiding those who will not benefit from it and consequently decreasing the risk of post-surgical morbidity. In conclusion, these findings demonstrate a successful application of MALDI MSI for the identification of protein biomarkers of endometrial cancer metastasis.