LAUSR

OBJECTIVES Peritoneal implants of ovarian borderline serous tumors are diagnostically challenging. Distinguishing invasive from non-invasive cases is crucial for patient management. This study aims to develop a molecular signature to distinguish invasive implants with malignant potential from those with benign. METHODS Archival formalin-fixed paraffin embedded tissues were retrieved from 3 institutions, with consensus histologic review. Lesions were classified as a non-invasive implant (n = 10), invasive implant (n = 9) or high grade (HG) peritoneal metastasis from HG serous ovarian carcinoma (n = 4). The nCounter® GX Human Cancer Gene Reference Assay was used to profile expression of 230 cancer genes and 6 control genes. The DEGs in HG peritoneal metastases compared to non-invasive implants were identified using T-tests performed in the NanoString Diff package, then used to cluster cases using the Eisen cluster 3.0 package. Lasso in glmnet package was used to select the subset of genes that most strongly correlate with a malignant potential. RESULTS 37 genes were downregulated and 16 genes were upregulated in HG peritoneal metastases. Using this 53-gene signature, one of nine of the invasive implants clustered with the HG peritoneal metastasis. Expression of ABCB1, CDC2, CDKN1A, FAT1, MMP9, MSH2, NQO1 and TOP2A were sufficient to indicate malignant potential of implants. The HG peritoneal metastasis and one invasive implant exhibited a high malignant likelihood (>92%) whereas the non-invasive implants and eight invasive implants displayed a low malignant likelihood (≤0.1%). CONCLUSIONS Invasive implants are heterogenous and often morphologically indistinguishable lesions with transcriptomes that may be classified as malignant or not. Additional research is needed to determine the importance of these genes as drivers and/or surrogates of malignant potential, and their utility for triaging invasive implants.