Unit:
Faculty of MedicineLibrary of the School of Health Sciences
Author:
Stroggilos Rafael
Dissertation committee:
Μαρία Ρουμπελάκη, Αναπληρώτρια Καθηγήτρια, Ιατρική Σχολή, ΕΚΠΑ
Αριστείδης Ηλιόπουλος, Καθηγητής, Ιατρική Σχολή, ΕΚΠΑ
Ευστάθιος Καστρίτης, Καθηγητής, Ιατρική Σχολή, ΕΚΠΑ
Ιερώνυμος Ζωιδάκης, Επίκουρος Καθηγητής, Τμήμα Βιολογίας, ΕΚΠΑ
Παναγιώτης Πολίτης, Ερευνητής Β', Τμήμα Βασικής Έρευνας, ΙΙΒΕΑΑ
Μαρία Γαζούλη, Αναπληρώτρια Καθηγήτρια, Ιατρική Σχολή, ΕΚΠA
Ελένη Κατσαντώνη, Ερευνήτρια Γ', Τμήμα Βασικής Έρευνας, ΙΙΒΕΑΑ
Original Title:
Ανάλυση προωτεομικών δεδομένων απο φασματομετρία μάζας και ενσωμάτωσή τους με άλλα κλινικά και μοριακά δεδομένα σε κλινικά δείγματα και καρκινικές σειρές
Translated title:
Analysis of mass spectrometry proteomics data and integration with publicly available transcriptomics and clinical data for the identification of molecular prognosticators in bladder cancer
Summary:
DNA/RNA-based classification of Bladder Cancer (BC) supports the existence of multiple molecular subtypes, while investigations at the protein level are scarce. The purpose of this study was to investigate if Non-Muscle Invasive Bladder Cancer (NMIBC) can be stratified to biologically meaningful proteomic groups, to establish associations between the proteomics subtypes and previous transcriptomics classification systems and to characterize the continuum of transcriptomics alterations observed in the different stages of the disease. Subsequently, tissue specimens from 117 patients at primary diagnosis (98 with NMIBC and 19 with MIBC), were processed for high resolution LC-MS/MS analysis. Protein quantification was conducted by utilizing the mean abundance of the top three most abundant unique peptides per protein. The proteomics output was subjected to unsupervised consensus clustering, principal component analysis (PCA), and investigation of subtype-specific features, pathways, and genesets, as well as for the construction and validation of a Random Forest based classifier. NMIBC patients were optimally stratified to 3 proteomic subtypes (classes), differing at size, clinico-pathological and molecular backgrounds: Class 1 (mostly high stage/grade/risk samples) was the smallest in size (17/98) and expressed an immune/inflammatory phenotype, along with features involved in cell proliferation, unfolded protein response and DNA damage response, whereas class 2 (mixed stage/grade/risk composition) presented with an infiltrated/mesenchymal profile. Class 3 was rich in luminal/differentiation markers, in line with its pathological composition (mostly low stage/grade/risk samples). PCA revealed a close proximity of class 1 and conversely, remoteness of class 3 to the proteome of MIBC. Samples from class 2 were distributed in a wider fashion at the rotated space. Comparative analysis with GSEA between the three proteomic classes and the three UROMOL subtypes indicated statistically significant associations between the proteomics class 1 and UROMOL subtype 2 (subtype with a bad prognosis) and also between the proteomics class 3 and UROMOL subtype 1 (subtype with the best prognosis). Utilizing a Random Forest based classifier, the predicted high- and low-risk phenotypes for the proteomic class 1 and class 3, were further supported by their classification into the “progressed” and “non-progressed” subtypes of the UROMOL study, respectively. Statistically significant proteins distinguishing these two extreme classes (1 and 3) and also MIBC from NMIBC samples were found to consistently differ at the mRNA levels between NMIBC “Progressors” and “Non-Progressors” groups of the UROMOL and LUND cohorts. Functional assessment of the observed molecular de-regulations suggested severe pathway alterations at unfolded protein response, cytokine and inferferone-γ signaling, antigen presentation, mRNA processing, post translational modifications and in cell growth/division. Collectively, this study identifies three proteomic NMIBC subtypes and following a cross-omics analysis using transcriptomic data from two independent cohorts, shortlists molecular features potentially driving non-invasive carcinogenesis, meriting further validation in clinical trials.
Main subject category:
Health Sciences
Keywords:
Bladder cancer, Molecular subtyping, Proteomics, Transcriptomics, Coexpression networks
Number of references:
174
