Charles M. Perou, PhD Associate Professor Departments of Genetics and Pathology Carolina Center for Genome Sciences Lineberger Comprehensive Cancer Center University of North Carolina at Chapel Hill Identification of GBM Subtypes using Expression Profiling

Agilent Custom Affymetrix Human Affymetrix HT-HG- 244k Exon 1.0 Array U133A Array Microarray

Identify Samples and represented on all 3 platforms

Factor Analysis

Single unified measurement for each gene 202 patients 11,681 genes Identification of GBM Subtypes

Census Clustering of the 202 samples X 1740 Unsupervised clustering of 1740 genes suggests that 4 subtypes of GBM exist variably expressed genes selected using a unified gene expression measure across 3 expression platforms Core TCGA Samples (173) /Pathway: with Subtype-defining genes • ProNeural: 1. nervous system development ProNeural Normal-like EGFR Mesenchymal 2. neuron differentiation (SOXs) 3. = proliferation 4. cell adhesion molecules 5. ErbB signaling pathway

FBXO3 GABRB2 • Normal‐like: SNCG NTSR2 1. nucleotide metabolic process MBP 2. neurological system process

DLL3 3. axon NKX2-2, NRXN1, NRXN2 TOP2B, CDC7, MYB 4. neuron projection SOX2, SOX4, SOX10,SOX11 5. synaptic transmission ERBB3, PAK3, PAK7 NCAM1 OLIG2 • EGFR: 1. regulation of transcription 2. cell migration 3. nervous system development

FGFR3 4. cell proliferation PDGFRA 5. metal ion binding EGFR AKT2 GLI2 TGFB3 • Mesenchymal:

CASP8, CASP5, CASP4, CASP1 1. immune response COL8A2, COL5A1, COL1A1,COL1A2 2. receptor activity ILR4 CHI3L1 3. wound healing TRADD TLR2, TLR4, 4. cytokine and chemokine IGFBI mediated signaling pathway RELB 5. NF‐B Signaling Pathway Correlations between gene expression subtypes and clinical parameters

Survival Analysis of Subtypes

n=196 Subtypes are correlated with:

1. Age (ProNeural – early age)

2. Nuclear Atypia (Normal-like=lowest) probability 3. Cellularity (Normal-like=lowest)

4. Average % Necrosis (Normal- like=lowest, Mesenchymal=highest)

Subtypes are NOT correlated with:

Overall Survival (months) 1. Average % tumor nuclei (>90%) Core TCGA Combined Samples (173) Validation Set (174)

ProNeural Normal-like EGFR Mesenchymal ProNeural Normal-like EGFR Mesenchymal

The validation set contains samples from three publically available data sets. microRNA‐based Subtype Discovery ~534 miRNAs assayed using Agilent microarrays with unsupervised clustering analysis

microRNA

GeneExpress ProNeural miR630 miR204

Norm-like 115

ProNeural 29 3 3

EGFR 2719

Mes 31614

Fisher’s exact test p‐value = 9.716e‐13 miRNA Subtypes Association to Clinical data

Survival Analysis Age DistributionAge by Subtype

p-value= 0.048

Gender DistributionGender by Subtype miRNA cluster Gender miR204 miR630 Proneural FEMALE 17 17 6 MALE 23 18 29 p‐value = 0.01432 TP53 Pathway All GBMs RB Pathway CDKN2A CDKN2B CDKN2C Activated oncogenes Pathway p15 p18 homozygous homozygous homozygous deletion in 51% deletion in 48% deletion in 2% 86% CDKN2A Analysis ARF homozygous deletion in 49% CDK4 CCND2 CDK6 amplification amplification in 13%MDM2 amplification in 5% in 17% amplification amplification MDM4 in 2% in 3% deletion, RB1 TP53 mutation in 11% mutation, 77% Senescence G1/S progression deletion in 35%

RTK/RAS/PI‐3K signaling Network EGFR ERBB2 PDGFRA MET

mutation, amplification mutation amplification amplification in 46% in 7% In 14% in 3%

mutation, NF‐1 RAS PI‐3K PTEN mutation, Class I deletion in 33% deletion in 17% mutation in 2% mutation in 15%

AKT amplification in 2%

Proliferation 79% Survival Translation FOXO mutation in 2% TP53 Pathway RB Pathway CDKN2A CDKN2B CDKN2C Activated oncogenes p16 p15 p18 homozygous homozygous homozygous 94% Normal-like deletion in 63% deletion in 63% deletion in 0% CDKN2A ARF homozygous deletion in 63%

CDK4 CCND2 CDK6 amplification amplification in 19%MDM2 amplification in 0% in 0% amplification amplification MDM4 in 0% in 6% deletion, RB1 TP53 mutation in 6% mutation, Senescence deletion in 25% Apoptosis 88% G1/S progression

Genes colored according to: = above average RTK/RAS/PI‐3K signaling Network = average expression EGFR ERBB2 PDGFRA MET = below average

mutation, amplification mutation amplification amplification in 63% in 19% In 19% In 0%

mutation, NF‐1 RAS PI‐3K PTEN mutation, deletion in 13% Class I deletion in 38% mutation in 0% mutation in 19%

AKT amplification in 6% 94 % Proliferation Survival FOXO mutation in 0% Translation TP53 Pathway RB Pathway CDKN2A CDKN2B CDKN2C Activated oncogenes PRONEURAL p16 p15 p19 homozygous homozygous homozygous 85% deletion in 30% deletion in 30% deletion in 5% CDKN2A ARF homozygous deletion in 30%

CDK4 CCND2 CDK6 amplification in 10%MDM2 amplification in 5% amplification amplification amplification MDM4 in 35% in 5% in 0% deletion, RB1 TP53 mutation in 5% mutation, Senescence Apoptosis 70% deletion in 60% G1/S progression

Therapies RTK/RAS/PI‐3K signaling Network PI-3K EGFR ERBB2 PDGFRA MET PDGFRA CDK4 inhibitors

mutation, amplification mutation amplification amplification In 35% in 10% In 40% in 5%

mutation, NF‐1 RAS PI‐3K PTEN mutation, deletion in 15% Class I deletion in 20% mutation in 5% mutation in 30% AKT amplification in 0%

75% Proliferation Survival FOXO mutation in 0% Translation TP53 Pathway RB Pathway CDKN2A CDKN2B CDKN2C Activated oncogenes p16 p15 p18 homozygous homozygous homozygous 100% EGFR deletion in 0% CDKN2A deletion in 92% deletion in 92% ARF homozygous deletion in 92% Del 9p21.3 CDK4 CCND2 CDK6 amplification amplification in 15%MDM2 amplification in 0% in 8% amplification amplification MDM4 in 0% in 0% deletion, RB1 TP53 mutation in 0% mutation, Senescence Apoptosis 92% G1/S progression deletion in 0% Amp 7p11.2 and Therapies RTK/RAS/PI‐3K signaling Network Del 9p21.3 occurred EGFR inhibitors in 32/36 tumors EGFR ERBB2 PDGFRA MET MDM2-Nutlins Restore ARF/p16

mutation, amplification mutation amplification amplification in 92% in 8% In 8% in 0%

mutation, NF‐1 RAS PI‐3K PTEN mutation, Class I deletion in 0% mutation in 0% mutation deletion in 46% in 0% AKT amplification in 0%

100% Proliferation Survival FOXO mutation in 8% Translation TP53 Pathway RB Pathway CDKN2A CDKN2B CDKN2C Activated oncogenes Mesenchymal p16 p15 p18 homozygous homozygous homozygous 78% deletion in 39% deletion in 39% deletion in 0% CDKN2A ARF homozygous deletion in 39%

CDK4 CCND2 CDK6 amplification amplification in 9%MDM2 amplification in 13% in 13% amplification amplification MDM4 in 0% in 0% deletion, RB1 TP53 mutation in 17% mutation, 70% Senescence Apoptosis G1/S progression deletion in 39% Therapies 5/5 tumors RTK/RAS/PI‐3K signaling Network with co-mutation Chemo (TP53-RB1) of PTEN+NF1 EGFR ERBB2 PDGFRA MET RAS/RAF/MEK were this subtype

mutation, amplification mutation amplification amplification in 26% in 4% In 9% in 4%

NF‐1 RAS PI‐3K PTEN mutation, mutation, Class I deletion in 39% mutation in 9% mutation deletion in 39% in 0% AKT amplification in 0%

87% Proliferation Survival FOXO mutation in 0% Translation Human Cancer Biospecimen Core Resource Genome Sequencing Centers •The International Genomics Consortium and •Broad Institute Sequencing Platform, The Broad Translational Genomics Research Institute, Phoenix, Institute of the Massachusetts Institute of Technology Ariz., and Harvard University •Washington University Genome Sequencing Center, Biospecimen Source Sites Washington University School of Medicine Glioblastoma multiforme biospecimens: • Sequencing Center, Baylor College •MD Anderson Cancer Center of Medicine •Henry Ford Hospital System, Department of Neurosurgery University of North Carolina at Chapel Hill •University of California at San Francisco Medical Center, Neil Hayes Joel Parker Department of Neurological Surgery Michael Topal Victor Weigman Katherine Hoadley Yufeng Liu Patients who agreed to participate in the study Matt Wilkerson Yan Shi Sai Balu Cancer Genome Characterization Centers Yuan Qi •Broad Institute of MIT and Harvard •Harvard Medical School and Brigham and Women's National Cancer Institute Hospital Ana Barker •Lawrence Berkeley National Laboratory Daniela Gerhard •Memorial Sloan-Kettering Cancer Center Joe Vockley •The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University National Human Genome Research Institute •The University of Southern California Norris Francis Collins Comprehensive Cancer Center •Stanford University School of Medicine •University of North Carolina Lineberger Comprehensive Cancer Center