Fine-Mapping Quantitative Trait Loci in mice and plants

Richard Mott, William Valdar, Jonathan Flint, Paula Kover1 and colleagues

Wellcome Trust Centre for Human GeneticsOxford University

1University of Manchester

Mosaic Crosses

Inbred founders G3 GN F20

mixing chopping up inbreeding

F2, dialleleHeterogeneous Stock, Advanced Intercross, Random Outbreds

RecombinantInbred Lines

Hidden Chromosome Structure (ancestral mosaic)

Observed chromosome structure (genotypes)

Analysis of Mosaic Crosses

Mott et al, 2000 PNAS

Hidden Markov Model

Mice

Pseudo-random matingfor 50 generations

Heterogeneous stocks

LP/J CBA/J A/J C57BL/6J BALB/cJAKR/J DBA/2J C3H/HeJ

Collaborative Cross Mice 8-way funnel

Brother-sister matingfor >20 generations

~ 1000 RecombinantInbred Lines

A/J C57BL6/J NOD/LtJ 129S1/SvImJ CAST/EiJNZO/HlLtJ PWK/PhJ WSB/EiJ

Genetic diversity of mouse populations

Roberts et al Mamm Genome (2007) 18:473-481

The diversity captured in the model resources is shown in their respective color as (BXD RI, blue; B.P CSS, gray; Northport HS, green; LSDP, orange; CC, red).

Study design

2,000 heterogeneous stock mice

15,000 diallelic markers

More than 100 phenotypes

each mouse subject to a battery of tests spread over weeks 5-9 of the animal’s life

101 Phenotypes

Anxiety (conditioned and unconditioned tasks) [24]

Asthma (plethysmography) [13]

Biochemistry [15]

Diabetes [16]

Haematology [15]

Immunology [9]

Weight/size related [8]

Wound Healing [1]

Bacterial Gut Flora

Mandible Shape

Gene Expression

Hippocampus, Liver, Lung

Intraperitoneal Glucose Tolerance Test

Solberg et al (2006) Mammalian Genome

How to select peaks: a simulated example

How to select peaks: a simulated example

Simulate 7 x 5% QTLs

(ie, 35% genetic effect)

+ 20% shared environment effect

+ 45% noise

= 100% variance

Simulated example: 1D scan

Peaks from 1D scan

phenotype ~ covariates + ?

1D scan: condition on 1 peak

phenotype ~ covariates + peak 1 + ?

1D scan: condition on 2 peaks

phenotype ~ covariates + peak 1 + peak 2 + ?

1D scan: condition on 3 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + ?

1D scan: condition on 4 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 +peak 4 + ?

1D scan: condition on 5 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + ?

1D scan: condition on 6 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + peak 6 + ?

1D scan: condition on 7 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + peak 6 + peak 7 + ?

1D scan: condition on 8 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + peak 6 + peak 7 + peak 8 + ?

1D scan: condition on 9 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + peak 6 + peak 7 + peak 8 + peak 9 + ?

1D scan: condition on 10 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + peak 6 + peak 7 + peak 8 + peak 9 + peak 10 + ?

1D scan: condition on 11 peaks

phenotype ~ covariates + peak 1 + peak 2 + peak 3 + peak 4 + peak 5 + peak 6 + peak 7 + peak 8 + peak 9 + peak 10 + peak 11 + ?

Peaks chosen by forward selection

Forward selection on a bootstrap sample

Forward selection on a bootstrap sample

Forward selection on a bootstrap sample

Bootstrap evidence mounts up…

In 1000 bootstraps…

Bootstrap Posterior Probability(BPP)

Model averaging by bootstrap aggregation

Choosing only one model:

very data-dependent, arbitrary

can’t get all the true QTLs in one model

Bootstrap aggregation averages over models

true QTLs get included more often than false ones

References:

Broman & Speed (2002)

Hackett et al (2001)

BPP > 0.25 strong evidence for a QTL

Valdar et al (2006) Nature Geneticshttp://gscan.well.ox.ac.uk

81 diabetes QTLs

T2D p-valhistone deacetylase 2 0.0009594

Mouse gene Human gene Description Chr SNPWGA T2D

logPGeneSniffer

ABCB11 ABCB11 ATP binding cassette, sub family B (MDR/TAP), member 11

2 rs567074 1.13 327

G6pc2 G6PC2 glucose 6 phosphatase, catalytic, 2 2 rs567074 1.13 2929

Whsc1 WHSC1 Wolf Hirschhorn syndrome candidate 1 4 rs487903 1.80 51

Whsc2 WHSC2 Wolf Hirschhorn syndrome candidate 2 4 rs13141668 2.13 29.9

SIM1 SIM1 single minded homolog 1 (Drosophila) 6 rs10457849 1.02 808

Hdac2 HDAC2 histone deacetylase 2 6 rs2348779 3.02 477

tns3 TNS3 tensin 3 7 rs865603 2.43 6

D530005L17Rik C9orf39 9 rs2593402 2.09 6.4

TRAF6 TRAF6 TNF receptor associated factor 6 11 rs10734444 1.06 332

RAG1 RAG1 recombination activating gene 1 11 rs10734444 1.06 985

Usp12 USP12 ubiquitin specific peptidase 12 13 rs2479559 2.47 0.7

Wdfy2 WDFY2 WD repeat and FYVE domain containing 2

13 rs9526793 1.56 0

4932702D22Rik EXDL1 exonuclease 3" 5" domain like 1 15 rs7173954 1.19 0

Ganc GANC glucosidase, alpha; neutral C 15 rs10518763 2.03 316

BC030867 C17orf53 17 rs228787 1.87

Mouse gene Human gene Description Chr SNPWGA T2D

logPGeneSniffer

ABCB11 ABCB11 ATP binding cassette, sub family B (MDR/TAP), member 11

2 rs567074 1.13 327

G6pc2 G6PC2 glucose 6 phosphatase, catalytic, 2 2 rs567074 1.13 2929

Whsc1 WHSC1 Wolf Hirschhorn syndrome candidate 1 4 rs487903 1.80 51

Whsc2 WHSC2 Wolf Hirschhorn syndrome candidate 2 4 rs13141668 2.13 29.9

SIM1 SIM1 single minded homolog 1 (Drosophila) 6 rs10457849 1.02 808

Hdac2 HDAC2 histone deacetylase 2 6 rs2348779 3.02 477

tns3 TNS3 tensin 3 7 rs865603 2.43 6

D530005L17Rik C9orf39 9 rs2593402 2.09 6.4

TRAF6 TRAF6 TNF receptor associated factor 6 11 rs10734444 1.06 332

RAG1 RAG1 recombination activating gene 1 11 rs10734444 1.06 985

Usp12 USP12 ubiquitin specific peptidase 12 13 rs2479559 2.47 0.7

Wdfy2 WDFY2 WD repeat and FYVE domain containing 2

13 rs9526793 1.56 0

4932702D22Rik EXDL1 exonuclease 3" 5" domain like 1 15 rs7173954 1.19 0

Ganc GANC glucosidase, alpha; neutral C 15 rs10518763 2.03 316

BC030867 C17orf53 17 rs228787 1.87

Mouse and human T2D candidates

Zeggini et al (2007) Science

Bacterial Gut Flora

• Different inbred strains have different gut flora (David Threadgill)• Flora can be assayed from feacal boli (quantitative PCR)

– Bacteroides distasonis– Clostridium– Eubacterium plexicaudatum– Firmcutes– Flexistipes– Lactobacillus acidophilus– Lactobacillus salivarius

• Data available for 600 HS mice at present

Defensins

Beta-defensin cluster

QTLs and Gene Expression Networks

ComplexPhenotypes

Gene ExpressionPhenotypes

Genotypes

QTL eQTL

100 traits2000 HS mice

47,000 probes500 hippocampi

250 livers250 lungs

13800 SNPs2000 mice + 300 parents

Directcorrelation

eQTLs at different BPP thresholds

Lung Liver Hippocampus (1) Hippocampus (2)

BPP Cis Trans Cis Trans Cis Trans Cis Trans

0.1 2194 7868 1151 6799 2478 3923 3945 5217

0.2 2189 7264 1150 6510 2474 3749 3940 4904

0.3 2179 6780 1146 6280 2469 3614 3928 4675

0.4 2160 6313 1136 5945 2456 3442 3915 4447

0.5 2135 5796 1121 5557 2442 3210 3890 4202

0.6 2090 5084 1086 5087 2394 2818 3843 3901

0.7 2004 3617 1021 3184 2319 1901 3729 2565

0.8 1889 1611 946 1303 2207 862 3573 1295

0.9 1721 443 849 390 2038 355 3358 598

1 1288 103 564 45 1548 130 2703 249

Tissue specificity of eQTLs at different BPP thresholds

BPP 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Only Liver 363 363 362 363 360 349 309 280 255 169

Only Lung 895 893 890 881 869 847 816 771 704 544

Hippocampus 1238 1238 1237 1230 1222 1202 1176 1130 1063 837

Liver + lung 170 170 168 163 157 153 143 129 119 72

Hippocampus + lung 622 619 616 616 616 608 574 540 500 388

Hippocampus + liver 112 111 112 110 111 102 98 88 77 39

Hippocampus + liver + lung 506 506 504 500 493 482 471 449 398 284

snps from http://mouse.perlegen.com/

Alkaline Phosphatase

Foreman, et al. Physiol. Genomics 23, 295–303 (2005), Valdar et al, Nat. Genet. .38, 879-887 (2006)

Eif4g3

Akp2

Mrto4 Ccdc21

Akp2

Akp2

Ccdc21

Sytl1

Sytl1

Mrto4

Eif4g3

Large effect cis eQTLs are often artefacts

Plants

Arabidopsis thaliana HSRIL16-way funnel

EcotypeCountry of Origin NASC Ecotype

Country of Origin NASC

Ler-0 Germany NW20 Kn-0 Lithuania N6762

Ct-1 Italy N6674 Po-0 Germany N6839

Wil-2 Russia N6689 Oy-0 Norway N6824

No-0 Germany N6805 Bur-0 Ireland N6643

Hi-0 Netherlands N6736 Sf-2 Spain N6857

Wu-0 Germany N6897 Zu-0 Switzerland N6902

Col-0 USA N6673 Ws-0 Russia N6891

Rsch-4 Russia N6850 Can-0 Canary Isles N6660

Mt-0 Libya N1380 Edi-0 Scotland N6688

Tsu-0 Japan N6874

LD decay in the 19 founders

Differences Between Founders

• A 5-SNP window (~500kb) separates founders into 7 groups on average

• A 10-SNP window (~1Mb) separates founders into 13 groups on average

Col vs Ler: 0.54 Oy vs Po: 0.89

Mean=0.71

A 5-SNP window (~500kb) separates founders into 7 groups on averageA 10-SNP window (~1Mb) separates founders into 13 groups on average

Allele Frequencies in the founders

Haplotype Mosaic of the HSRIL

Prediction based on minimising recombinants in 1540 SNPs

QTL Mapping: Simulations

QTL MappingProvisional Results of 200 HSRILs, genotypes not corrected

Resequencing?

• Resequence 19 founders using Solexa/FLX– 80-90% of each genome should be recovered, based on Bur-0

• Impute sequences of HSRILs from the mosaics• Association Mapping using imputed genomes to refine QTLs

– QTL mapping resolution ~200-500kb in HSRILs– LD between founders decays by 10kb

Acknowledgements

MouseFunded by The Wellcome Trust

William ValdarCarmen Arboleda-HitasAmarjit BhomraPeter BurnsRichard CopleyStuart DavidsonKate ElliotSimon FiddyJonathan FlintPolinka HernandezSue MillerRichard MottChela NunezSagiv ShifmanLeah SolbergAmy TaylorMartin TaylorJennifer TaylorJordana Tzenova-BellBinnaz YalcinDave BannermanShoumo BhattacharyaBill CooksonRob DeaconDominique Gauguier

ArabidopsisFunded by BBSRC

Paula KoverXiao WangWilliam ValdarRichard Mott

QTL networks are linked by genes with correlated expression

Compute correlation R between every pair of transcripts, first removing sex effects.

Pick Phenotype with N QTL, lengths L1,L2,..LN

Derive empirical distribution of maximum inter-QTL R

Do 1000 times:

Randomly rearrange these QTL segments across genome

Compute max R between transcripts under different segments

Compute segment network joined by transcripts > Rthreshold

Find networks of QTLs linked by strongly correlated transcripts

Correlations between genes under different QTLs are higher than expected by permutation

Hippocampus Liver

Large QTL networks are unlikely

Correlation Threshold

Fal

se P

ositi

ve R

ate

Based on mRNA co-expression networks

Liver

Liver Haematological QTL-transcript networks

Food hyponeophagia(reluctance to try new food)

Sort1 neurotenson receptorcelsr2 gpcr, cadherincrmp1 neuronal development, semaphoringrb2 growth factor receptorcamk2b schizophrenia/depressionslc24a3 neuronal synapsesBC034068 Nat8l, acetyl transferase, retinal development

Wound healing

Brp17 PKND = MR1 involved in fibrillogenesis in muscleNcor1 breast cancer cell proliferationIgfb2 carcinogenesis/cell proliferationZswim1 transcription factor of unknown function4732437J24Rik heat shock hsp40 1600023A02Rik wfdc2 contains domain WAP, that is involved in cancer in rats1700009P17Rik unknown function, contains domain pro1217, involved in cell wall strengthening in plantsAV028368 (9) d4ertd429e , unknown, contains Zn/cys6 domain, probably DNA-binding

(Rthresh = 0.75)Based on mRNA co-expression networks

Parent of Origin Effects

• Parent of Origin Effects occur when the parental ancestry of the allele affects the phenotype.

• Parental Genotypes were collected for the HS: It was possible to estimate if alleles were descended from dam or sire.

• 11% of QTLs investigated showed evidence of POE, defined as the maximum –log10P for the POE model was at least 4 log10P units greater than the maximum -log10P for the null POE model, over the QTL’s 95% confidence interval. This is a conservative estimate.

• Body-weight and Immunological phenotypes were particularly rich in POE QTL. Behavioural POE QTL were rare.

Parent of Origin Effects

The most significant POE QTL are listed below

-log10P Phenotype Locus

poe poe+ null null diff

Imm CD4 Intensity 11:rs3659504-gnf11.098.633 20.5 23.9 7.4 16.5

Imm %CD8+ 1:rs6363233-rs6224524 16.2 36.8 23.8 12.9

Imm %CD8+ 12:rs13481411-rs13481412 12.7 28.3 17.1 11.2

Weight 6 weeks 2:rs13476640-rs4223268 11.3 15.8 5.8 9.9

Imm %CD8+/CD3+ 3:rs6274760-rs3701904 12.1 19.9 10.3 9.6

Imm %B220+ 12:rs13481568-rs13481583 11.6 21.2 11.6 9.6

Haem.Red Cell Distribution Width

2:rs8279354-rs3696744 10.8 18.9 9.4 9.5

Imm %CD8+ 2:rs6286688-rs13476391 11.1 32.3 22.8 9.4

Weight.10weeks (g) 2:rs6230107-rs3674264 11.7 17.6 8.9 8.7

Imm %CD3+ 12:rs4229511-rs13481634 15.1 21.4 12.7 8.7

Parent of Origin expression QTL

• 60 Hippocampus cis-eQTL showed evidence for POE at logP threshold = 4.

• The most significant hippocampal POE eQTL is logP = 38 chromosome 7, syntenic to the imprinted human Prader-Willi locus, Some of the POE eQTL are known to be imprinted but many appear to be novel.

• In the lung, 42 POE eQTL were identified, the most significant being Fads (fatty acid desaturase 3) on chromosome 19, logP=23.

• In the liver, 19 POE eQTL were identified, the most significant being Ugt1a13, (UDP glycosyltransferase 1 family, polypeptide A10) on chromosome 19 at logP=11.

• There was considerable overlap between the sets of liver and lung POE eQTL, but little overlap with the hippocampal set.

Joint POE QTL and POE eQTL

• We identified 12 POE QTL that overlapped with a POE eQTL. There are several interesting candidate genes suggested by this analysis. These include:

• Pldn (pallidin, platelet storage pool deficiency) for Red Cell Distribution Width.

• Scn2b (sodium channel, voltage-gated, type II, beta) for Open field total activity.

• ENSMUSESTG00003715632 and D7Ert715e for Weight at 6 weeks; syntenic to human Prader-Willi syndrome.

Ultra fine mapping in highly outbred stocks (Binnaz Yalcin)

2003: MF1 outbred population from Harlan

Yalcin et al, 2004 Nature Genetics

Ultra fine mapping in highly outbred stocks (Binnaz Yalcin)

2003: MF1 outbred population from Harlan

2007: Binnaz is searching for alternative wild

populations

Yalcin et al, 2004 Nature Genetics

http://gscan.well.ox.ac.uk

Extra Slides

Estimating genotypes for unobserved snps

The plan was to use the same MF1 mice from Harlan UK (as we did a few

years ago) to refine large effect QTLs (ALP, MCV, CD4 and HDL) found in

the HS project. In the MF1 pilot project, we genotyped and phenotyped 400

of these mice. Unexpectedly almost all markers used were homozygous and

Jonathan almost had a heart attack!

Over the last few months, I have embarked upon finding another population

of outbred mice, commercially available. I am currently in the process of

genetically characterising over 40 different colonies of outbred mice in

order to pick up the most suitable one for QTL mapping. The smallest the

LD block the best.

single marker association

single marker associationHAPPY method

Insulin.15

Insulin.Delta

Insulin.30

Insulin.AUC

Insulin locus on chromosome 11

Insulin locus on chromosome 11

Insulin.15

Annotatedgenes

Insulin locus on chromosome 11

Insulin.15

Annotatedgenes

tensin 3