Targeted deep sequencing of effusion cytology samples is feasible, informs spatiotemporal tumor evolution, and has clinical and diagnostic utility.

During the course of disease, many cancer patients eventually present with metastatic disease including peritoneal or pleural spread. In this context, cytology specimens derived from ascites or pleural effusion may help to differentiate malignant from benign conditions and sometimes yield diagnosis of a malignancy. However, even when supported by immunohistochemistry, cytological interpretation can be challenging, especially if tumor cellularity is low. Here, we investigated whether targeted deep sequencing of formalin-fixed and paraffin embedded (FFPE) cytology specimens of cancer patients is feasible, and has diagnostic and clinical impact. To this end, a cohort of 20 matched pairs was compiled, each comprising a cytology sample (FFPE cell block) and at least one biopsy/surgical resection specimen serving as benchmark. In addition, 5 non-malignant effusions were sequenced serving as negative-controls. All samples yielded sufficient libraries and were successfully subjected to targeted sequencing employing a semiconductor based next-generation sequencing platform. Using gene panels of different size and composition, including the Oncomine Comprehensive Assay, for targeted sequencing, somatic mutations were detected in the tissue of all 20 cases. Of these, 15 (75%) harbored mutations that were also detected in the corresponding cytology samples. In four of these cases (20%), additional private mutations were detected in either cytology or tissue samples, reflecting spatiotemporal tumor evolution. Of the five remaining cases, three (15%) showed wild type alleles in cytology material whereas tumor tissue had mutations in interrogated genes. Two cases were discordant, showing different private mutations in the cytology and in the tissue sample, respectively. In summary, sequencing of cytology specimens (FFPE cell block) reflecting spatiotemporal tumor evolution is feasible and yields adjunct genetic information that may be exploitable for diagnostics and therapy.