We sought to identify candidate serum biomarkers for the detection and surveillance of EOC. Based on RNA-Seq transcriptome analysis of patient-derived tumors, highly expressed secreted proteins were identified using a bioinformatic approach.
RNA-Seq was used to quantify papillary serous ovarian cancer transcriptomes. Paired end sequencing of 22 flash frozen tumors was performed. Sequence alignments were processed with the program ELAND, expression levels with ERANGE and then bioinformatically screened for secreted protein signatures. Serum samples from women with benign and malignant pelvic masses and serial samples from women during chemotherapy regimens were measured for IGFBP-4 by ELISA. Student's t Test, ANOVA, and ROC curves were used for statistical analysis.
Insulin-like growth factor binding protein (IGFBP-4) was consistently present in the top 7.5% of all expressed genes in all tumor samples. We then screened serum samples to determine if increased tumor expression correlated with serum expression. In an initial discovery set of 21 samples, IGFBP-4 levels were found to be elevated in patients, including those with early stage disease and normal CA125 levels. In a larger and independent validation set (82 controls, 78 cases), IGFBP-4 levels were significantly increased (p < 5 × 10-5). IGFBP-4 levels were ~3× greater in women with malignant pelvic masses compared to women with benign masses. ROC sensitivity was 73% at 93% specificity (AUC 0.816). In women receiving chemotherapy, average IGFBP-4 levels were below the ROC-determined threshold and lower in NED patients compared to AWD patients.
This study, the first to our knowledge to use RNA-Seq for biomarker discovery, identified IGFBP-4 as overexpressed in ovarian cancer patients. Beyond this, these studies identified two additional intriguing findings. First, IGFBP-4 can be elevated in early stage disease without elevated CA125. Second, IGFBP-4 levels are significantly elevated with malignant versus benign disease. These findings provide the rationale for future validation studies.