Dipa Gurung (Presenter)
Imperial College London
Bio: I am a PhD student at Imperial College London, United Kingdom. I graduated with an honours degree in Biochemistry from Queen Mary, University of London in 2012. After graduation, I continued my studies at Kings’ College London by undertaking an MSc in Analytical Science for Industry. The course also involved an internship position at the Institute of Cancer Research with a project focused on shot-gun proteomics. Following this, I decided to pursue a PhD in mass spectrometry based cancer studies under the supervision of Professor Zoltan Takats and Professor Jeremy Nicholson. The main objective of my PhD project is to associate the lipidomic data of breast cancer samples obtained from desorption electrospray ionisation mass spectrometry imaging (DESI-MSI) with the clinical phenotype and the histological data to develop systematic stratification method for cancer biopsy specimens.
Authorship: Dipa Gurung (1), James McKenzie (1), Francesca Rosini (1), Luisa Doria (1), Anna Mroz (1), Rathi Ramakrishnan (1), Edward St. John (1), Jeremy Nicholson (1), Zoltan Takats (1).
(1) Imperial College London, London, UK
| Short Abstract Breast cancer is a highly heterogeneous disease and one of the most prevalent form of cancers in women worldwide. Accuracy of diagnosis is often compromised as it is dependent on the pathologists’ subjective interpretation. Over the past decade, desorption electrospray ionisation (DESI) has been increasingly used for imaging based cancer studies. DESI mass spectrometry imaging enables spatial visualisation of lipid species across tissue sections allowing direct correlation with morphological features. The technique combined with multivariate statistical analysis serves as an excellent complementary technique for histological diagnosis of breast cancers and to establish key diagnostic lipid species as potential biomarkers. |
Long Abstract
Introduction
Breast cancer is highly heterogeneous disease and is one of the most prevalent form of cancers in women worldwide. The clinical heterogeneity explains why similar tumours have different prognoses. When the presence of abnormal tissue is initially confirmed, a series of tests on the suspected tumour and nearby tissue are performed. This is then followed by biopsy operations which acts as one of key the diagnostics to confirm the presence of cancer. Manual histological evaluation of the stained biopsy tissue sections has been the gold standard method to establish the final diagnosis. However, the accuracy of this morphology based tissue diagnosis is often compromised as it is dependent on the pathologists’ interpretation resulting in high investigator-to-investigator variance compromising prognosis.
Over the past decade, mass spectrometry (MS) has been increasingly used in the field of lipidomics. The sensitivity and the high specificity associated with MS have been the major driving forces behind the success of MS based lipidomics studies. More recently mass spectrometry imaging (MSI) has gained a lot of momentum and has emerged as an invaluable tool for mass spectrometry based cancer studies. Desorption electrospray ionisation (DESI) is a novel ambient imaging technique that allows spatial localisation of lipid species across tissue sections enabling direct correlation of the lipidomic information with the morphological features. Therefore, breast cancer samples were analysed by DESI-MSI to understand the metabolic differences between these histologically similar breast tumours and correlate them with clinical phenotypes.
Method
35 breast carcinoma samples, 45 histologically validated normal tumour bed samples and 7 benign fibroadenoma samples were collected, cryosectioned and analysed by DESI-MSI in positive and negative ion modes using a home-built DESI ion source coupled to an orbitrap mass spectrometer. The same tissue sections were hematoxylin and eosin (H&E) stained, digitalised and underwent histopathological examination.
Raw DESI-MSI data were converted to imzML format and imported into MATLAB R2014a (MathWorks) environment for data analysis using in-house written function. The raw data from each sample and the combined sample set underwent binning and normalisation for multivariate statistical analysis. The digital and integrated MS ion images were co-aligned for precise correlation of morphological features with that of biochemical features.
Results
Each individual tissue section data (87 total) was subjected to unsupervised principal component analysis (PCA) to visualize differences between different tissue types. For the tumour samples, the first PC distinguished the tumour section of the tissue from rest of the section in both positive and negative ion modes. The second PC mostly distinguished the stroma/connective tissue section from rest of the region. Recursive maximum margin criterion (RMMC) analysis was used for supervised classification. Tissue types in each sample and their spatial distribution were determined by an independent histopathologist based on the H&E stained optical image. Based on this information small number of representative mass spectra per tissue were selected from the MS ion image to build sample specific RMMC model which was used to classify all pixels in different tissue types. This data was complemented by a cross validation plot which exceeded 98% accuracy in all samples for both ion modes.
Multivariate statistical models were created using tumour samples and the histologically validated normal tumour bed samples. The spectra were binned with a bin size of 0.01Da in the m/z range of 600-1000 to cover the phospholipid mass range. The resulting model was very strong with a clear separation between the tumour and tumour bed samples. Leave-one- patient-out cross-validation performance of more than 90% was achieved in the negative ion mode. Further multivariate statistical analysis were performed on the entire dataset. Supervised RMMC analysis was performed to observe the performance of the model with regards to the classification of sub-tissue types (glandular, stroma, adipose, malignant) found within the multiple breast samples. The leave-one-out cross validation results showed good overall classification results of approximately 80%.
The lipid profiles of the normal and the tumour samples obtained from the DESI spectra were investigated in order to elucidate the key lipids responsible for the separation between the two classes. The most predominant lipids found in negative ion mode were triglycerides (TG), phosphatidylserines (PS), phosphophatidylethanolamine (PE) and phosphatidylinositol (PI). The lipid classes contributing significantly towards the separation were predominantly PE's and PI's.
The tumour dataset was divided according to various clinical factors: histological grade, hormone receptor [(HR) - estrogen (ER) and progesterone (PR)] status and HER2 receptor status as it has been shown that these molecular subgroups display distinct molecular and clinical characteristics. Each of these individual clinical factors were tested independently. The DESI-MSI data from the negative ion mode dataset showed that DESI-MSI can be effectively used for discriminating between positive and negative HR status and also between histological grade statuses.
Conclusion
DESI-MSI is a highly promising novel technique for histology level diagnosis of breast cancer and to establish key diagnostic ions as potential biomarkers. The changes in the lipidomic profile follow the pattern reported previously on the literature. Future work will focus on building on stronger dataset by increasing the number of samples.
References & Acknowledgements:
1. Punnonen K, Hietanen E, Auvinen O and Punnen R. Phospholipids and fatty-acids in breast-cancer tissue. J Cancer Res Clin Oncol 1989: 15; 575.
2.Veselkov K, Mirnezami R, Strittmatter N, Goldin R, Kinross JM, Speller AV, et al. Chemo-informatic strategy for imaging mass spectrometry-based hyperspectral profiling of lipid signatures in colorectal cancer. Proc Natl Acad Sci U S A 2013: 111;1216.
3. Guenther S, Muirhead L, Speller A, Golf O, Strittmatter N, Ramakrishnan R, Goldin RD, Jones E, Veselkov K, Nicholson J, Darzi A, and Takats Z. Spatially Resolved Metabolic Phenotyping of Breast Cancer by Desorption Electrospray Ionization Mass Spectrometry, Cancer Research 1 May 2015: 75;1828.
| Description | Y/N | Source |
| Grants | no | |
| Salary | no | |
| Board Member | no | |
| Stock | no | |
| Expenses | no |
IP Royalty: no
| Planning to mention or discuss specific products or technology of the company(ies) listed above: | no |