Authorship:
Luisa Doria(1), David Phelps(1), Anna Mroz(1), Abigail Speller(1), Kirill Velsekov(1), Sadaf Ghaem-Maghami(1), Zoltan Takats(1)
(1)Imperial College of London
| Short Abstract Ovarian cancer is the fifth most common cancer among women and one of the causes is the poor and vague prognosis and diagnosis. REIMS and DESI-MS are two mass spectrometry techniques with great potential to characterize and discriminate different cancer types and stages. Both techniques are excellent to complement and support histology, REIMS can identify different ovarian cancer types in real time and DESI can provide detailed spatial information within the sample giving the opportunity to investigate tumour biology from an entirely new perspective with accurate biochemical information about each tissue type. |
Long Abstract
Introduction
Ovarian cancer is the fifth most common cancer among women, and it causes more deaths than any other type of female reproductive cancer. Currently, in an oncological surgery, the resected tissue is sent to a pathology laboratory for intraoperative histological examination while the patient remains under general anaesthesia which takes time (20 to 30 min), and is costly. A definitive diagnosis is then established based on the histological information provided by a pathologist and complemented with medical tests. As histological information is based on human interpretation, it can lead to some uncertainty and sometimes to a vague prognosis.
REIMS (Rapid evaporative ionization mass spectrometry) and DESI-MS (desorption electrospray ionization mass spectrometry) are two mass spectrometry techniques with great potential to characterize and discriminate different cancer types and stages. Furthermore they can reveal the lipidomic profile of the tissue analysed, giving additional information on tumour biochemistry important in cancer biology.
REIMS is a very recent technique which has the potential of being used in real-time surgery, taking advantage of the aerosol “smoke” produced by the surgical knife and analysing at the same time, permitting real time, intraoperative characterization of tissue.
Studies investigating breast cancer and colorectal cancer with DESI-MS have shown that this technique is capable not just of discriminating between different cancer types and stages but also providing detailed spatial information within the sample. The ability of directly correlating the mass spectral image created with histological features can provide topographically localised biochemical information to supplement conventional histological classification systems.
Therefore, with the combination of both techniques, we can accomplish robust tissue recognition and identification of a new pool of cancer biomarkers and tissue-specific molecular ion patterns which could lead us to a new method in cancer diagnosis and prognosis at a histology-level.
Methods
20 samples of serous ovarian carcinomas, 10 samples of endometroid ovarian carcinoma together with 10 different borderline and benign ovarian cancer samples and 10 healthy ovarian samples were collected. Each sample was divided into two, frozen, and analysed by two different techniques; one portion was analysed using DESI-MS and the other using REIMS in negative ion mode.
The samples analysed by REIMS were defrosted and analysed, and then preserved in formalin to be stained and examined by histopathology professionals in order to confirm the patient diagnosis. The data was also subjected to pre-processing steps, i.e. background subtraction, normalization and log-based transformation.
The samples analysed by DESI were cryosectioned, imaged and then stained with haematoxylin and eosin (H&E), digitalised and examined by histopathology professionals, in order to align the optical and MSI image for precise selection of regions of interest (ROI). The acquired DESI-MSI data were pre-processed to account for common bio-analytical complexities using total ion count normalization and log-based variance stabilizing transformation. The processed MSI data were subjected to supervised multivariate statistical analysis using Recursive Maximum Margin Criterion (RMMC) discriminant analysis to identify patterns of diagnostic and prognostic importance.
Results and Discussion
Both DESI and REIMS are excellent techniques to ionize and analyse lipid species, giving the potential to characterize the lipid metabolism and changes in cancer biology.
To ensure we can separate distinct epithelial ovarian cancer types and also cancer from healthy ovary, multivariate statistical models were created using the entire dataset from REIMS. The range of the spectrum used was between 600 and 1000Da, which covers the phospholipids present in the cells, mostly in the cell membrane. Supervised MMC analysis clearly differentiated all tissue types with a leave one patient out cross-validation performance of ~80%. This analysis allowed us to show that this MS technique is able to discriminate between different cancer types and also distinguish cancerous from healthy tissue in ovarian cancer. This application in a real time surgery will allow to overcome the time and the cost of an intraoperative histological examination and to identify the tumour margins at the same time.
DESI data was analysed with the same workflow as REIMS data, to confirm the separation between the different cancer types and also from healthy tissue. The same results were obtained with a clear separation of the different sample types. In addition, multivariate statistical analysis was performed in each individual sample to identify different tissue types based on the corresponding histological image and performing supervised analysis using maximum margin criteria (MMC). Cross-validation results for these areas within individual samples always exceeded 98% accuracy.
The lipids found in the DESI spectra were identified to better understand the ovarian cancer biochemistry. The predominant lipids found in negative ion mode were phosphatidylethanolamines (PE), phosphatidylserines (PS) and phosphatidylinositol (PI). Examination of the FA composition of these lipids revealed that short chain fatty acids are replaced with longer chain fatty acids in cancerous tissue which have previously been shown to be involved in inflammatory processes. A more detailed feature selection method will be used to evaluate which lipids contribute most to the characterization of each tissue type and to better understand the lipid metabolism in ovarian cancer.
Overall REIMS can identify different ovarian cancer types in real time and DESI can provide the opportunity to investigate tumour biology from an entirely new perspective, both techniques complementing and supporting histology with accurate biochemical information about each tissue type.