A GENOME-WIDE APPROACH TO PREDICT OUTCOME IN OSTEOSARCOMA
Nalan Gokgoz, Taiqiang Yan, Michelle Ghert, Mona Gill, Shelley B Bull, Robert S Bell, Jay S Wunder, Irene L Andrulis
Mount Sinai Hospital and Samuel Lunenfeld Research Institute Toronto, Ontario, Canada
OSTEOSARCOMA
Treatment involves (neo)adjuvant chemotherapy
and wide surgical resection Patients without Metastases at Diagnosis: 5 year disease-free survival 50-78% Patients with Metastases at Diagnosis:
5 year disease-free survival 10-20%.
Few accurate clinical predictors of outcome Molecular markers ( e.g. p53, RB, cdk4 ,SAS): not prognostic
An Emerging Molecular Paradigm
Analysis of global gene expression
Classification of OSA tumors Prediction of disease outcome.
Microarray Analysis
PATIENTS
High-grade Intramedullary 63 patients
A No Metastasis at Diagnosis 46 patients A1
B Metastasis at Diagnosis 17 patients
No Metastasis 4 years post Dx. (29 patients)
A2
Metastasis within 4 years Dx. (17 patients)
TUMOR SAMPLES •63 fresh frozen, primary,high-grade intramedullary osteosarcoma samples
•Tumor specimens from open biopsies obtained prior to chemotherapy.
•Tumor specimen chosen based on frozen section histological analysis. •Minimum follow-up 4years or metastasis
Clinical Charactersitics of Patients Presenting with Non-metastatic OSA
Microarray Analysis
19K cDNA microarrays
Image Acquisition : Axon Scanner Spot Analysis : GenePix Pro5 Data Storage: IobianTM Gene Traffic
Statistical Analysis
Quality Control
Reproducibility
Aim 1: Outcome of the Patients Presenting with no Metastases
No Metastases 4 years post Dx vs Metastases within 4 years Dx 18981 cDNAs T-statistic p<0.001 (BrB Array Tools)
n=50 genes for tumor classification/clustering
50 Most Significant Genes
No Mets 4 yrs post Dx. Mets within 4 yrs Dx.
STATISTICAL VALIDATION
Leave-One Out (LOO) cross-validation method Diagonal Linear Discriminant Analysis (DLDA) Class Prediction
Prediction Accuracy 74%
Differentially expressed genes that are higher in metastasis group
RB1-inducible coiled-coil 1 (RB1CC1) HBV preS1-transactivated protein 4 (PS1TP4) Hypothetical protein FLJ11184 (FLJ11184) Yippee-like 3 (Drosophila) (YPEL3) AP1 gamma subunit binding protein 1 (AP1GBP1) Protein phosphatase 2, regulatory subunit B', beta isoform (PPP2R5B) Tubulin folding cofactor A (TBCA) EP400 N-terminal like (EP400NL) GTP-binding protein 10 (putative) (GTPBP10) Melanoma cell adhesion molecule (MCAM) Potassium channel tetramerisation domain containing 20 (KCTD20) Pentatricopeptide repeat domain 3 (PTCD3) Adenosine deaminase-like (ADAL) Leucine rich repeat containing 3B (LRRC3B) Flotillin 2 (FLOT2) 12 ESTs
Differentially expressed genes that are lower in metastasis group
Adaptor protein, phosphotyrosine interaction, PH domain and leucine zipper containing 2 (APPL2) Hypothetical protein MGC39715 (MGC39715) DIP2 disco-interacting protein 2 homolog B (Drosophila) (DIP2B) PHD finger protein 19 (PHF19) Solute carrier family 6 (neurotransmitter transporter, creatine), member 8 (SLC6A8) Ras-associated protein Rap1 (RBJ) Muscleblind-like (Drosophila) (MBNL1) Fc fragment of IgG, low affinity IIIa, receptor (CD16a) (FCGR3A) Glial cells missing homolog 2 (Drosophila) (GCM2) Chromosome 9 open reading frame 123C9orf123 Chromosome 2 open reading frame 29 (C2orf29) Phospholipase D2 (PLD2) Ribosomal protein L27a (RPL27) Hypothetical protein LOC339400 (LOC339400) Chromosome 12 open reading frame 49 (C12orf49) Platelet-activating factor acetylhydrolase 2, 40kDa (PAFAH2) Solute carrier family 5 (sodium-dependent vitamin transporter), member 6(SLC5A6) 7 ESTs
Aim 2: Analysis of gene expression profiles of OSA patients presenting with metastasis Metastases at Dx vs No Metastases at Dx
18981 cDNAs
T-statistics
p<0.001 (BrB Array Tools)
n=2161 genes for tumor classification/clustering
DLDA Class Prediction 94% Prediction Accuracy
MOLECULAR VALIDATION by REAL TIME PCR
Gene Mean Expression Levels in Tumors of Patients with: No Metastasis at Metastasis at Diagnosis Diagnosis p-value
D4, Zinc and Double PHD Fingers Family 2 (DPF2) Secretory Carrier Membrane Protein 5 (SCAMP5) Signal-Induced ProliferationAssociated Gene 1 (SIPA1) Secreted Protein, Acidic, CysteineRich (SPARC) Hydroxyacylglutat hione Hydrolase (HAGH)
0.27
0.15
0.0149
0.36
0.22
0.0966
0.34
0.21
0.0790
5.02
1.64
0.0041
0.31
0.21
0.1156
STAM2 was selected as the internal control gene after assessing 6 housekeeping genes by a statistical model described by Szabo et.al.(2004).
DPF2 (Requiem)
member of the d4 domain family with a Kruppel type zinc-finger Functions as a transcription factor for the apoptotic response Induction of apoptosis by extracellular signals Examples: Deprivation of survival factors in myeloid cells Drug treatment in OS cells?
Work in Progress U2OS, SaOS, HOS Cells Knock down the DPF2 gene by SiRNA
Drug Treatment
Investigate the effect for the Apotosis
CONCLUSIONS
The use of this genome-wide approach identified a number of genes that may play a role in osteosarcoma. Genes and pathways not previously implicated in osteosarcoma have been elucidated by this study.
FUTURE STUDIES
Identify pathways related to genes in the classifier
Protein-Protein Interactions found by Pathway Studio for 50 Significant Genes in A1vs A2 groups
FUTURE STUDIES
Online Predicted Human Interactive Database (OPHID)
Protein-Protein Interactions found in OPHID for Significant Genes in A vs B groups
Acknowledgement
IL Andrulis
JS Wunder Mount Sinai Hospital Orthopedic Surgeons
RS Bell T.Yan M. Ghert Mona Gill
Hospital for Sick Children D.Malkin
Vancouver General Hospital C.Beauchamp University of Washington E.Conrad III R Kandel Royal Orthopedic Hospital R.Grimer
S Bull
W He R Parkes
Memorial Sloan-Kettering J.Healey Mayo Clinic M.Rock/ L.Wold
Acknowledgements
• Ontario Cancer Research Network (OCRN) • National Cancer Institute of Canada (NCIC) • Canadian Institute of Health Research (CIHR) Interdisciplinary Health Research Team (IHRT) in Musculoskeletal Neoplasia • Rubinoff-Gross Chair in Orthopaedic Oncology at Mount Sinai Hospital, University of Toronto