Clinical and Pathological Significance of Epigenomic Changes in Colorectal Cancer Shuji Ogino, M.D., Ph.D. Associate Professor of Pathology Harvard Medical School Brigham and Women s Hospital Dana-Farber Cancer Institute I have no conflict of interest
Brief overview of our epidemiology projects
Multistep Carcinogenesis Epigenetic alterations (p16, MGMT, MLH1, etc ) p27 loss APC mutation KRAS mutation p53 mutation 18q loss normal cell aberrant crypt foci adenoma I adenoma II adenoma III Colon Cancer???????? Etiologic Factors (environmental or genetic)
Colon cancer risk factors Excess alcohol Lack of vegetable (folate) Red meat Obesity, diabetes, lack of exercise Smoking Family history We have very limited knowledge about mechanisms
Multistep Carcinogenesis Epigenetic alterations (p16, MGMT, MLH1, etc ) p27 loss APC mutation KRAS mutation p53 mutation 18q loss We try to link normal cell aberrant crypt foci adenoma I adenoma II adenoma III Colon Cancer Colon cancer risk factors Excess alcohol Lack of vegetable (folate) Red meat Obesity, diabetes, lack of exercise Smoking Family history
Multistep Carcinogenesis Epigenetic alterations (p16, MGMT, MLH1, etc ) APC mutation KRAS mutation p27 loss p53 mutation 18q loss We try to link normal cell aberrant crypt foci adenoma I adenoma II Colon cancer risk factors Excess alcohol adenoma III Colon Cancer Lack of vegetable (folate) Molecular Epidemiologic Pathology Red meat Obesity, diabetes, lack of exercise Smoking Family history
Prospective cohorts Nurses Health Study (121,700 Women) 1976 1986 Health Professionals Follow-up Study (51,500 Men)
Cohorts 120,000 Women 52,000 Men Exposures (diet, lifestyle, etc.) Plasma biomarkers Family history SNPs Correlations 1000 CRCs 1000 adenomas Molecular pathology Pathogenesis Clinical outcomes Prevention
Cohorts 120,000 Women 52,000 Men Exposures (diet, lifestyle, etc.) Plasma biomarkers Family history SNPs Tissue biomarkers are very Correlations important!!! 1000 CRCs 1000 adenomas Molecular pathology Pathogenesis Clinical outcomes Prevention
Epigenomic aberrations in colorectal cancer
Epigenomic aberrations Genome-wide DNA hypomethylation Proto-oncogene activation Genomic instability CpG island methylation Tumor suppressor silencing Mutually exclusive Ogino et al. Int J Cancer 2008
Conventional pathway APC β-catenin WNT Aberrant crypt foci KRAS p53 p21 Conventional adenoma CIN Colon cancer Normal Genome-wide hypomethylation CpG island methylation Hyperplastic polyp BRAF Serrated pathway Serrated polyp CIMP-high colon ca.
Genome-wide DNA Hypomethylation
LINE-1 LINE-1 (long interspersed nucleotide element-1) repetitive DNA elements have numerous CpG sites CH 3 CH 3 TCGTACGA TCGTACGA DNA methylation occurs at CpG (CG) sites
LINE-1 LINE-1 methylation level correlates with genome-wide DNA methylation
27% LINE-1 methylation T:73.7% C:26.3% Pyrosequencing T:74.5% C:25.5% T:72.9% C:27.1% T:65.4% C:34.6% 1300 E S A C T C A G T G T G T C A G T C A G T T A G T C T G 5 10 15 20 25 73% LINE-1 methylation T:28.2% C:71.8% T:30.8% C:69.2% T:26.3% C:73.7% T:21.0% C:79.0% 1400 1300 E S A C T C A G T G T G T C A G T C A G T T A G T C T G 5 10 15 20 25
0 50 100 150 200 250 LINE-1 methylation 20 l 24 30 l 34 40 l 44 50 l 54 60 l 64 70 l 74 80 l 84 90 l 94 25 l 29 35 l 39 45 l 49 55 l 59 65 l 69 75 l 79 85 l 89 No. of cases
75% LINE-1 methylated 60-75% methylated 45-60% methylated P=0.0001 <45% methylated Colon cancer-specific survival (years) Ogino et al. J Natl Cancer Inst 2008
LINE-1 hypomethylation high mortality Hazard ratio (HR) 10 1 P(trend) <0.0001 Colon cancer-specific mortality 92 84 76 68 60 52 44 36 28 20(%) LINE-1 methylation Ogino et al. J Natl Cancer Inst 2008
LINE-1 hypomethylation Poor prognosis Ogino et al. J Natl Cancer Inst 2008
CpG island methylator phenotype (CIMP)
CpG Island Methylater Phenotype (CIMP) Extensive promoter CpG island methylation Inactivates many tumor suppressor genes 20% of colorectal cancers show CIMP-high Right colon, elderly female, BRAF mutation
CIMP Causes Microsatellite Instability (MSI) MSI = altered lengths of microsatellites (short nucleotide repeats) in tumor 15% of CRCs show MSI-high 10% MLH1 methylation in CIMP-high tumors 2% Lynch syndrome (HNPCC) 3% sporadic MSI-high, but not CIMP-high
MSI-high 15% CIMP-high 20% 5% 10% 10% Non-MSI-high Non-CIMP-high 75%
Problems in CIMP Criteria Lack of validated consensus marker panel and criteria for CIMP-high
CIMP Panel Validation 900 colorectal cancers MethyLight (real-time PCR) 8-marker panel CACNA1G, CDKN2A (p16), CRABP1, IGF2, MLH1, NEUROG1, RUNX3, SOCS1 (including the new Laird s panel) Ogino et al. J Mol Diagn 2007
Unsupervised hierarchical clustering analysis of 16 methylation markers in 900 colorectal cancer Nosho et al. PLoS ONE 2008
Nosho et al. PLoS ONE 2008
Nosho et al. PLoS ONE 2008
Now we have good methods, markers and criteria for CIMP-high
What causes CIMP-high?
female old age? one-carb. nutrients SNPs proximal location CIMP-high? BRAF+ genome-wide hypom
Dietary intake Folate Dihydrofolate Deoxyuridylate dump Methionine Tetrahydrofolate Purine SAM betaine methyltransferase betaine Homocysteine DNA methylation Methionine synthase B12 TCN2 5,10-methylene THF MTHFR 5-methyl THF (Circulating folate) 10-formylTHF Thymidylate dtmp DNA synthesis
MTHFR and TCN2 SNPs and CIMP-high MTHFR 1298 WT hetero homo Multivariate OR=7.6 (1.3-43) TCN2 WT hetero homo Multivariate OR=3.8 (1.02-14) 0% 25% 50% 75% 100% Frequency of CIMP-high Hazra et al. unpublished
Germline genetic variation Somatic DNA methylation
DNMT3B female old age? one-carb. nutrients SNPs proximal location CIMP-high? BRAF+ genome-wide hypom
DNMT3B (DNA methyltransferase 3B) Important enzyme for de novo DNA methylation DNMT3B overexpression in mice induces CRC w CpG island methylation Linhart et al. Gene Dev 2007
DNMT3B+ (N=116) DNMT3B(-) (N=649) OR=3.4, P<0.0001 No. of methylated markers 0 markers 1 marker 2 markers 3 markers 4 markers 5 markers 6 markers 7 markers 8 markers 0% 25% 50% 75% 100% Nosho et al. unpublished
What causes CIMP-high? Probably multifactorial
CIMP and BRAF Clinical outcome
CIMP-high is good (by the new CIMP panel) Independent BRAF+ is bad Ogino et al. Gut 2009
Survival probability Colon cancer-specific survival Tumors were classified into 4 types (by CIMP and BRAF status) Ogino et al. Gut 2009 years
Survival probability CIMP-low/0 BRAF(-) CIMP-low/0 BRAF(+) BRAF+ is bad! Ogino et al. Gut 2009 years
Survival probability CIMP-high BRAF(-) CIMP-high BRAF(+) BRAF+ is bad! Ogino et al. Gut 2009 years
Survival probability CIMP-high BRAF(+) CIMP-low/0 BRAF(+) CIMP-high is good! Ogino et al. Gut 2009 years
Survival probability CIMP-high BRAF(-) CIMP-low/0 BRAF(-) CIMP-high is good! Ogino et al. Gut 2009 years
CIMP and BRAF are independent predictors CIMP-0 CIMP-low CIMP-high Multivariate hazard ratio (95% CI) 1 0.8 (0.5-1.1) 0.4 (0.2-0.9) BRAF(-) BRAF+ 1 2.0 (1.1-3.4) Ogino et al. Gut 2009
CIMP-low
CIMP-low (low-level CIMP) CpG island methylation is not as extensive as CIMP-high
CIMP-low New epigenomic subtype? Ogino et al. J Mol Diagn 2006 Ogino et al. BMC Cancer 2007 Ogino et al. Gut 2007 Ogino et al. Mod Pathol 2008 Shima et al. unpublished Or Mixture of CIMP-high and CIMP-0? Same as CIMP-neg (CIMP-0)?
No. of methylated markers 0/8 (N=411) 1/8 (N=159) 2/8 (N=82) 3/8 (N=32) 4/8 (N=32) 5/8 (N=27) 6/8 (N=27) 7/8 (N=45) 8/8 (N=59) Total (N=874) KRAS mutation BRAF mutation CIMP-0 CIMP-low KRAS+ CIMP-high 0% 20% 40% 60% 80% 100% Ogino et al. J Mol Diagn 2006
0 (N=110) 1 (N=146) 2 (N=124) 3 (N=90) 4 (N=84) 5 (N=52) 6 (N=42) 7 (N=30) 8 (N=19) 9 (N=16) 10 (N=24) 11 (N=20) 12 (N=19) 13 (N=29) 14 (N=28) 15 (N=33) 16 (N=12) 100% 80% 60% 40% 20% 0% KRAS(+) BRAF(-) KRAS(-) BRAF(+) BRAF(+) CIMP-low No. of methylated markers (0 to 16) Mutation frequency
KRAS mutation is associated with CIMP-low, but not clustered with any of the methylation markers Nosho et al. PLoS ONE 2008
KRAS mutation BRAF mutation Random pattern Non-random pattern CIMP-low CIMP-high
CIMP-high vs. CIMP-low vs. CIMP-0 BRAF+ Old age Female Proximal Inactive WNT KRAS+ KRAS/BRAF WT Distal Genome-wide hypom CIN
Summary 1 CpG island methylator phenotype (CIMPhigh) Non-random methylation pattern (links to BRAF mutation) inversely associated with genome-wide hypomethylation DNMT3B may contribute to CIMP-high KRAS+ is associated with random methylation pattern (= CIMP-low)
Summary 2 CIMP-high good prognosis BRAF+ bad prognosis LINE-1 (i.e., genome-wide) hypomethylation bad prognosis
DFCI Ogino Lab Katsuhiko Nosho Kaori Shima Natsumi Irahara Shoko Kure Yoshifumi Baba Saori Toyoda Takako Kawasaki Past lab members DFCI / Brigham and Women s Hosp Charles Fuchs Massimo Loda Jeffrey Meyerhardt Ron Firestein William Hahn John Quackenbush G Mike Makrigiorgos Janina Longtine Jonathan Glickman Brian Wolpin Li Chen Beth Israel Deaconess Medical Center Lewis Cantley NIH/NCI Bennett Family Fund Entertainment Industry Foundation Japanese Society for Promotion of Science Japanese Foundation for Multidisciplinary Treatment of Cancer Special Thanks! NHS (n=121,700) and HPFS (n=51,500) prospective cohort participants Various US hospitals/pathology departments Harvard Sch Public Health / BWH Channing Lab Frank Speizer Walter Willett Susan Hankinson Meir Stampfer Graham Colditz Gregory Kirkner David Hunter Donna Spiegelman Peter Kraft Aditi Hazra Eva Schernhammer Many other staff members Massachusetts General Hospital Andrew Chan Mari Mino-Kenudson Jeffrey Engelman Kevin Haigis MIT Whitehead Institute Rudolf Jaenisch Sumita Gokhale University of Southern California Peter Laird Dan Weisenberger UT MD Anderson Cancer Center Jean-Pierre Issa Lanlan Shen