Clinical, epigenetic and phospholipid biomarkers for diagnosis, Prognosis and treatment stratification in ovarian cancer surgery

Abstract

Background: Ovarian cancer (OC) presents late when disease is widespread, as early detection via screening is not possible. Primary treatments rely on surgery and chemotherapy irrespective of tumour biology or the likelihood of a good prognosis. Preoperative diagnosis of OC is notoriously difficult without tissue biopsy. Consequently, some women undergo major surgery only to discover they have non-invasive tumours, or ovarian metastases from a distant primary cancer, resulting in inappropriate operations. This research endeavours to use tumour biology and clinical characteristics to retreat from blanket treatments for all when personalised treatments should be favoured in the new era of precision medicine.

Results: Analysis of historical data from Hammersmith Hospital revealed strong survival associations with readily available clinical and haematological features. In 647 women with high-stage OC between 2001-2014, no residual disease was strongly associated with improved overall survival (OS) (HR 2.31, 95% C.I. 1.68, 3.19). In keeping with previous studies grade 2 and 3 tumours conferred almost identical survival patterns, in contrast to low-grade tumours. Similarly, survival differences between stage 3C and 4 were negligible with stage 3A associating with significantly improved survival. Moderate or severe anaemia was highly predictive of poor OS (HR 0.61, 95% C.I. 0.42, 0.87), as was hypoalbuminaemia (HR 0.45, 95% C.I. 0.27, 0.73).

Tumour DNA methylation was used as a biomarker and found to be highly predictive of prognosis in OC. Six DNA methylation loci (FGF4, FGF21, MYLK2, MYLK3, MYL7, ITGAE) were associated with survival in women who achieved optimal surgical debulking. DNA methylation of myosin light chain kinase 3 (MYLK3) was validated in three independent cohorts as a highly prognostic biomarker, associated with OS, especially those with the least residual disease post-surgery (Hammersmith (HR 0.51, 0.31-0.84; P=0.01), Charité (0.46, 0.21-1.01; P=0.05), TCGA (0.64, 0.44-0.93; P=0.02), n=436). Women lost the survival advantage of surgery when MYLK3 methylation was below the median value (<84.9%).

The rapid evaporative ionisation mass spectrometry (REIMS) surgical intelligent knife (iKnife) analyses surgical diathermy aerosols to diagnose tissue type intra-operatively in real-time. The iKnife was used to sample 335 tissue samples from 198 women, yielding 3384 mass-spectra. Cross-validated OC versus normal tissue diagnosis was high (97.4% sensitivity, 100% specificity). Borderline ovarian tumours were readily distinguishable from OC (sensitivity 90.5%, specificity 89.7%) and benign tumours (sensitivity 90.5%, specificity 91.3%). Fresh tissue validation lead to excellent OC detection (100% accuracy). Histological agreement between iKnife and histology was very good (kappa 0.84, P<0.001, z=3.3) and performed better than the surgeon (kappa 0.60, P = 0.0088, z = 2.62). Five predominantly phosphatidic acid (PA(36:2)) and phosphatidylethanolamine (PE(34:2)) lipid species were identified as being significantly more abundant in OC compared to normal tissue or borderline tumours (P<0.001, q<0.001).

Conclusions: DNA methylation is an ideal biomarker and MYLK3 methylation could potentially be used to determine survival after surgery. Future work on MYLK3 will focus on measuring methylation of this locus in blood, so this marker can be more easily translated into clinical use. The REIMS iKnife accurately distinguishes gynaecological tissues by analysing mass-spectrometry-derived lipidomes and may improve surgical care when histology is unknown, leading to personalised operations tailored to the individual.