Thinking big: Finding more and (more) genes influencing glycaemic and anthropometric traits. Mark McCarthy, Oxford

Transcription

1 Thinking big: Finding more and (more) genes influencing glycaemic and anthropometric traits Mark McCarthy, Oxford

2 Slow progress in finding multifactorial genes Glazier et al, Science, 2002

3 Why? biological complexity Genes familial clustering twin studies adoption studies migration studies admixture studies gene discovery secular trends migration studies twin studies transgenerational effects intervention studies Environment maternal genotype environments prenatal postnatal childhood adult individual genotype risk of T2D paternal genotype

4 Why? inadequate study design Low chance of finding true positives Low priors + Low power + Low thresholds for declaring significance High chance of finding false positives Most positives in the literature are false

5 Recent advances using GWA approach Better understanding of patterns of human sequence variation Advances in genotyping technology Sample collections of adequate size Genome wide association scans Samples of interest 3,000,000,000 bases in human genome ~10,000,000 positions commonly variant in Europeans 80% of these captured by typing ~500k test for evidence of association

6 Summary of talk # common variant loci implicated T2D BMI glucose

7 Type 2 diabetes

8 Atlas of diabetes susceptibility (2006) Effect size Large Monogenic forms PNDM TNDM Other rare syndromes other MODY HNF1A mt3243 no genes up here LMNA These genes explain only a small proportion of observed familiality PPARG TCF7L2 Small Rare Beyond the scope of genetics! LARS2 ACDC CAPN10 LMNA INS HNF4A KCNJ11 Common Allele frequency

9 Wellcome Trust Case Control Consortium 2000 T2D 2000 T1D Cases Controls SNPs UK (WTCCC) , CHD 2000 HT 3000 UK common controls 2000 bipolar 2000 RhA 2000 Crohns FUSION ,000 DGI (Broad/Malmo) ,000 Total ~5 billion genotypes Main study with national cases/controls Affymetrix 500k array Replication in ~20,000 samples WTCCC, Nature 2007

10 Q-Q plot 1924 T2D 2938 controls 393,453 SNPs TCF7L2 Zeggini et al, Science 2007 WTCCC, Nature 2007 FTO SLC30A8 CDKAL1 HHEX IGF2BP2 CDKN2A PPARG KCNJ11

11 Back to the other T2D genes Large sample size Sladek et al, Nature Zeggini et al, Science Wellcome Trust Case Control Consortium, Nature DGI Consortium, Science Scott et al, Science Steinthorsdottir et al, Nature Genetics Modest odds ratios Highly significant

12 Beta cell reigns supreme.. A. CDKAL C. HHEX p=0.001 C/C C/A A/A rs p=0.001 A/A A/G G/G rs B. CDKAL D. HHEX p< C/C C/A A/A rs p=0.019 A/A A/G G/G rs Diabetes risk genotype Beta cell: KCNJ11, TCF7L2, CDKAL1, CDKN2A/B, IGF2BP2, HHEX, SLC30A8, TCF2 Insulin action: PPARG, FTO,??WFS1 Pascoe et al, Diabetes, 2007 Grarup et al, Diabetes, 2007

13 Fine mapping: caution required... expression P value Gene A Gene B Gene C D Gene B Gene A

14 What do they do? FTO PPARG WFS1? KCNJ11 TCF7L2 IGF2BP2 SLC30A8 HHEX CDKAL1 CDKN2A/B TCF2 (HNF1B) Obesity? Beta cell dysfunction Reduced Betacell mass Insulin Resistance Reduced insulin secretion Type 2 diabetes

15 What do they do? FTO PPARG WFS1? KCNJ11 TCF7L2 IGF2BP2 SLC30A8 HHEX CDKAL1 CDKN2A/B TCF2 (HNF1B) Obesity? Beta cell dysfunction Reduced Betacell mass Wnt signalling in the islet Insulin Resistance Reduced insulin secretion Type 2 diabetes

16 What do they do? FTO PPARG WFS1? KCNJ11 TCF7L2 IGF2BP2 SLC30A8 HHEX CDKAL1 CDKN2A/B TCF2 (HNF1B) Obesity? Beta cell dysfunction Reduced Betacell mass Wnt signalling in the islet NA demethylation in hypothalamus? Insulin Resistance Reduced insulin secretion Zn transport in the islet Type 2 diabetes

17 What do they do? FTO PPARG WFS1? KCNJ11 TCF7L2 IGF2BP2 SLC30A8 HHEX CDKAL1 CDKN2A/B TCF2 (HNF1B) Obesity? Beta cell dysfunction Reduced Betacell mass Wnt signalling in the islet NA demethylation in hypothalamus? Insulin Resistance Reduced insulin secretion Zn transport in the islet Cell cycle regulation in islet Type 2 diabetes

18 It is worth finding more genes? Biological value number Many small effects Likely distribution of effect sizes Several medium effects Few larger effects effect size Predictive value Small effects (if true) offer as much biological insight into mechanisms as large effects

19 More T2D genes. WTCCC: 2000 cases, 3000 controls DGI: 1500 cases, 1500 controls FUSION: 1200 cases, 1200 controls Affymetrix 500k Affymetrix 500k Illumina 317k IMPUTE MACH ~1.9M SNPs 10k+individuals 58 SNPs individuals Q Q plot for 3 way m/a 11 known genes removed 11 SNPs 5814 individuals (DECODE) 90,000+ individuals in all 11 SNPs ~55k individuals (including KORA) Zeggini et al, NG, in press 6 definite new genes (p<5x10 8 ) T2D genes #12 17

20 New T2D genes Chr risk allele frequency JAZF CDC123/CAMK1D TSPAN8/LGR THADA ADAMTS NOTCH2 Stage 1 (DGI, FUSION, WTCCC) All data nearest gene(s) OR (95%CI) P value n eff OR (95%CI) P value 1.14 ( ) 1.15 ( ) 1.18 ( ) 1.25 ( ) 1.13 ( ) 1.30 ( ) 1.5E 04 59, E 04 62, E 05 62, E 04 60, E 04 62, E 04 58, ( ) 1.11 ( ) 1.09 ( ) 1.15 ( ) 1.09 ( ) 1.13 ( ) 5.0E E E E E E 08 Zeggini et al, NG 2008

21 Allelic Odds ratio TCF7L2 FTO* PPARG NOTCH2 CDC123 MC4R* CDKAL1 KCNJ11 JAZF1 IGF2BP2 TSPAN8 TCF2 HHEX WFS1 ADAMTS9 SLC30A8 CDKN2A/B THADA Risk allele frequency

22 Keeping count... Type 1 diabetes HLA INS CTLA4 PTPN22 IL2RA IFIH1 Up to Type 2 diabetes n=11 IL2 12q24 12q13 PTPN2 KIAA PPARG KCNJ11 TCF7L2 HHEX/IDE SLC30A8 N~20 FTO* CDKAL1 CDKN2A/B IGF2BP2 WFS1 TCF2 JAZF1 CDC123 ADAMTS9 THADA TSPAN8 NOTCH2 MC4R* Candidate gene approaches High throughput LD mapping Genome wide LD mapping * Via primary effect on adiposity

23 Ongoing work Larger metaanalyses Other ethnic groups Deeper replication Functional overlay More SNP signals Other signals Global CNV screens Resequencing Rare signal mining Epi SNPs Causal variants Resequencing Fine map genotyping Multiple ethnic groups Translation Epidemiology joint effects Prediction New biology New drugs Functional Cellular studies Animal models Integrative physiology Systems biology

24 Type 2 diabetes genes and cancer (PLEIOTROPY)... and birthweight (COMPLEXITY; NON LINEARITY)

25 T2D and coronary artery disease signals.. Aneurysms A T2D CAD T2D chr9 rs rs rs rs E 0- B MTAP 1 C CDKN2A CDKN2B ANRIL (Mb) cancer (melanoma)

26 The yin and yang of diabetes and cancer increased CDK activity oncogenesis CDKN2A loss of function reduced beta cell regeneration reduced cell division melanoma and other tumours reduced CDK4 activity CDKN2A gain of function

27 Cancer and diabetes Cdkn2a expression in islet increases with age (effector of senescence) Cdkn2a knockouts enhanced islet survival Cdkn2a overexpression/cdk4 knockouts reduced islet regeneration, islet hypoplasia, T2D

28 Pleiotropy prostate cancer common common prostate TCF2 rare cancer common common 8q24 Type 2 diabetes MODY renal cysts urogenital abnormalities Type 2 diabetes

29 JAZF1 Zeggini et al., Nat Gen 2008 chr TAX1BP Mb JAZF1 rs864745(t) OR = 1.26 (95%CI: ) p = 2.14x cM/Mb rs (g) OR = 1.10 (95%CI: ) p = 5x10 14 Thomas et al., Nat Gen 2008

30 T2D vs Ca associations at proven variants Using GWA data from DIAGRAM and CGEMS HNF1 (TCF2) 8q24 (near MYC)

31 Birthweight Non-genetic explanations Altered birth weight Genetic explanations Fetal origins hypothesis Low BW diabetes risk Diabetic uterus effect High BW diabetes risk T2D risk alters early growth affects diabetes risk Genetic variant Genes influencing insulin secretion are particularly good candidates for such a shared role Glucokinase = proof of principle

32 Variable effect on birthweight TCF7L2 diabetes risk allele increased BW (Freathy et al, AJHG 2007) CDKAL1 diabetes risk allele decreased BW (Freathy et al, submitted) Other diabetes risk alleles no effect on BW

33 What s going on? Risk of diabetes GDM mother Genetic variant decreasing insulin secretion e.g. Risk allele at TCF7L2 or CDKAL1 fetus Indirect effect to BW TCF7L2 Direct effect to BW CDKAL1 Direction of effect on BW is a bioassay of the timing of the beta cell defect

34 Continuous glycemic traits

35 MAGIC consortium GWA based discovery in ~50,000 DeCODE genetics Northern Finland Birth Cohort 1966 Netherlands Twin Resource/NESDA Rotterdam cohort KORA EUROSPAN, Sorbs CoLaus Twins UK Framingham Heart Study Diabetes Genetics Initiative FUSION SardiNIA Baltimore Study of Ageing (BLSA) CHS Inchianti Procardis GEMS Finrisk 2007 Health 2000 BWHHS Segovia Caerphilly Twins UK Oxford Biobank Neth Twin Resource GHRAS GENDAI Ely Fenland EYHS FUSION METSIM DIAGEN PIVUS ULSAM DARTS NFBC86 DESIR Replication studies in ~70,000

36 More MAGIC GLUCOSE ~46,820 GWA scans 2,327,768 SNPs HOMA B ~37,340 GWA scans 2,454,849 SNPs

37 Physiology and pathology PHYSIOLOGICAL VARIATION IN THE FASTING GLUCOSE SETPOINT GCK GKRP G6PC2 Little or no effect on T2D risk vs PATHOLOGICAL VARIATION IN THE BETA CELL RESPONSE TO INSULIN RESISTANCE CDKN2A CDKAL1 HHEX JAZF1 THADA... Little or no effect on fasting glucose levels (provided subclinical T2D excluded) Except MTNR1B

38 Weight, adiposity and obesity Trashidang Monastery, Sikkim

39 Q-Q plot TCF7L T2D 2938 controls 393,453 SNPs FTO SLC30A8 CDKAL1 HHEX IGF2BP2 CDKN2A PPARG KCNJ11

40 FTO variants influence adult weight 30,081 adults from 13 studies P=3x10 35 Frayling et al, Science % of the population who are homozygous for risk allele ~2 3kg heavier Independent discoveries subsequently reported by French and US groups

41 Clues to what FTO does arcuate

42 Fto and Ftm are coordinately regulated Fto Ftm Stratigopolous et al, Am J Physiol 2008

43 More obesity genes 4 population based N=11, ,062 SNPs + 3 case series N=5, ,883 SNPs Red = meta4 Blue = meta7 10 population based N=40,701 3 case series N=3, population based N=13,244 2 case series N=2,638 1 childhood cohort N=5, childhood case control N~7,000 Loos et al, NG, 2008

44 Data from 32,000 GWA scans...

45 70,000 adults... P~10-16 Half the effect of FTO Approx 0.3% variance 5% of a SD

46 GIANT consortium Genetic Investigation of ANThropometric Traits Joel Hirschhorn + 11 groups (Oxford, Cambridge, Exeter) EPIC Norfolk CoLaus 1958BC UKBS WTCCC-T2D WTCCC-HT WTCCC-CAD FUSION SardiNIA KORA DGI NHS PLCO N=32,754 8 more BMI signals ~50 more height signals ~5 signals for central adiposity

47 More genes... Combine GWAs from individuals Check top hits in + GWAs from + 44,000 individuals Do more genotyping in 40,000 samples Willer et al, NG in revision

48 GIANT consortium Genetic Investigation of ANThropometric Traits Joel Hirschhorn + 11 groups (Oxford, Cambridge, Exeter) EPIC Norfolk CoLaus 1958BC UKBS WTCCC-T2D WTCCC-HT WTCCC-CAD FUSION SardiNIA KORA DGI NHS PLCO N=32,754 DECODE Finnish Birth Cohort Other European Other US groups N=100,000+

49 Translation Clinical Translation New biological insights relevant to T2D in general Better measures of individual aetiology Clinical advances for everyone New therapeutic targets New biomarkers New preventative measures Personalized medicine Prognostics Diagnostics Therapeutic optimisation

50 Number and size For most diseases Common variants of modest effect no evidence of departure from additivity For most loci Causal variants not yet known

51 Individualised prediction 1.8% 8.2% OR 2.3 OR 4.2 Weedon et al, PLOS, 2007 Lango et al, Diabetes % 1.2% Big effects of possible clinical value restricted to just a small proportion Limited empirical evidence that this information would translate into better outcomes...

52 Individualised prediction 11 variants ( FTO), BMI, age AUC = 80% BMI and age AUC = 78% 12 variants AUC = 59% Sibling relative risk (overall) 3.0 Sibling relative risk (18 genes) 1.07

53 So where s the dark matter? Mismatch between visible mass of the universe & that inferred from gravitational effects on visible matter dark matter Mismatch between genes found & that inferred from measures of heritability dark heritability

54 So where s the dark matter? More of the same More More of the common same variant loci Finding the causal variants Looking in other ethnic groups Other types of variation structural variants low frequency variants What explains Weird this stuff dark inheritance? more common epistatic SNP interactions variants structural epigenetics variants inherited epigenetic effects epistasis (ie emergent effects of different genes) most likely, low frequency, medium penetrance variants yet to be discovered...

55 Atlas of susceptibility Penetrance High Intermediate alleles causing Mendelian disease Low frequency variants with intermediate penetrance rare examples of high penetrance common variants influencing common disease Modest Low Ver y rare rare variants of small effect very hard to identify by genetic means Rare Uncommon most common variants implicated in common disease by GWA Common Allele frequency

56 Atlas of diabetes susceptibility Effect size Large PNDM TNDM other MODY HNF1A Other rare syndromes mt3243 LMNA TCF7L2 LARS2 ACDC CAPN10 LMNA FTO PPARG CDKAL1 IGF2BP2 CDKN2A SLC30A8 KCNJ11 WFS1 HHEX HNF4A TCF2 INS Small Rare Common Allele frequency

57 Consider the following variant. MAF of 1% GRR of 3 For disease prevalence of 5% Penetrance of het = 0.14 Penetrance of homoz = 0.43 Too rare for GWA studies Penetrance too low for linkage Lambda(s) such loci would give lambda(s) > such variants in a single locus would linkage signal ROC curve 20 variants AUC variants AUC 0.77 Find by: Resequencing Previously implicated genes Exons/promoters etc Association testing Power 90%, alpha = unselected case control pairs 559 pairs if cases are selected from sibpairs

58 Atlas of diabetes susceptibility Effect size Large PNDM TNDM other MODY HNF1A Other rare syndromes mt3243 LMNA TCF7L2 LARS2 ACDC CAPN10 LMNA FTO PPARG CDKAL1 IGF2BP2 CDKN2A SLC30A8 KCNJ11 WFS1 HHEX HNF4A TCF2 INS Small Rare Common Allele frequency

59 Summary GWA studies demonstrate the contribution of the common disease/common variant concept Significant differences in tractability of different phenotypes Common variants so far explain only a small proportion of the variance in these traits CNVs, low frequency variants, epistasis to be measured Other major ethnic groups almost completely unexplored Challenges: resequencing, fine mapping, functional studies, epidemiology, translation... Many novel insights into disease biology Opportunities for translation

60 Acknowledgements Too numerous to mention... DIAGRAM consortium Case Control Consortium

61 Acknowledgements Ele Zeggini John Perry Cecilia Lindgren Tim Frayling Rachel Freathy Will Rayner Nic Timpson Inga Prokopenko Amanda Bennett Andy Usher Kate Elliott Chris Groves Hana Lango Mike Weedon

62 Thanks for the invitation, and your attention