Transcript Two-Stage Association Mapping in Dogs Identifies Coat Color Locus

Regions of the Canine Genome Associated
with Osteosarcoma Identified by a Whole Genome
Case-Control Association Study
Kenine E Comstock, Elinor Karlsson, Claire Wade, Evan T Keller, Eric Lander,
Laurence Baker, Kerstin Lindblad-Toh
University of Michigan, Ann Arbor & Broad Institute, Boston, USA
Osteosarcoma in pet dogs and humans
Common Features
1.
2.
3.
4.
5.
Primary tumor occurs on weight
bearing limbs
Tumor growth with an intact immune
system
Similarities in histology
Location of metastases
Common environment
Differences
1. Age of onset
2. Incidence in population
Dogs and humans: Similarities in cancer genetics
The first example: Renal cancer syndrome
Canine Hereditary Multifocal Renal
Cystadenocarcinoma and Nodular Dermatofibrosis (RCND)
•German Shepherds
•Autosomal dominant
•Similar in phenotype to human
Birt-Hogg Dube syndrome.
•We showed that mutations in the
same gene in human and dogs
were associated with both
diseases.
Lingaas and Comstock, et al. Hum Mol Genet. 2003 12(23):3043-53
RCND-affected
RCND-unaffected
Cancer in dogs is highly breed-specific
Cancer
Breed
Increased risk
Hemangiosarcoma
Golden Retriever
5
Lymphoma, T-cell
Boxer
4
Melanoma
Scottish Terrier
6
Mammary Tumor
Springer Spaniel
5
Osteosarcoma
Greyhound
Rottweiler
10
5
Advantages of canine models for mapping cancer genes
•
Large families, inbred, good genealogical records, relatively short
generation times
•
Disease progression accelerated.
•
Genetic similarity within breeds: cancer is associated with fewer
mutations than in the species as a whole.
Methods for mapping genes
1. Linkage mapping using families
Best for rare, high penetrance diseases
2. Association mapping
For complex traits
Compares cases vs controls
No need for families
Markers: SNPs
100,000 SNPs or more for humans
Only ~30,000 SNPs in dogs!
Can narrow associated regions to kilobases
Resources available for
Canine Disease Gene Mapping
•7x Canine Genome
Sequence-released July
2005*
•Comparative maps of
human and dog
chromosomes
•SNP mapping set
*Lindblad-Toh et al. Nature (2005) 438:803-19
2.5 million SNPs discovered (1 SNP/kb)
• SNP discovery in 11 breeds
and 5 wild canids
Boxer
Alaskan Malamute
Beagle
Bedlington Terrier
California Coyote
English Shepherd
German Shepherd
Italian Greyhound
China gray wolf
Alaska gray wolf
India gray wolf
Spanish gray wolf
Poodle
Labrador Retriever
Portuguese Waterdog
Rottweiler
Dog SNP Chip for whole genome scans
Designed and tested by Michigan/Broad group
in collaboration with Affymetrix
~26,500 optimized SNP Mapping Set
Now commercially available
Coverage
Chromosome
1Mb ≥ 5 SNPs
Autosomes
97%
Chrom X
42%
Power analysis:
Prediction of sample size needed
Trait
Dogs
Power (LOD>5)
Recessive
20 + 20
100%
Dominant
50 + 50
100%
5-fold risk
100 + 100 *
98%
*In humans, would need approx 800 + 800
Can we map real traits using this SNP marker set?
Mendelian Trait
Cases+Controls
LOD
Region Size (kb)
Genes
Co-Dominant
Recessive
Dominant
10+9
7+11
27+40
5
3
5
800
800
300
1
7
3
Yes! Power calculations appear correct
•Predicted low sample numbers work
• One significant associated region
*Elinor Karlsson and Claire Wade, Broad Institute
Objectives:
1. To identify genes
involved with
osteosarcoma
susceptibility in dogs
2. To map a complex trait
using the new SNP set.
Large and giant breed dogs have a higher risk
of osteosarcoma
1) Mastiff-type breeds
Rottweiler-5 X increased risk
Labrador Retriever
Flat-coated Retriever
Golden Retriever
Mastiff
Bullmastiff
Newfoundland
2) Long-limbed breeds
Greyhound-5 to 10 X
Irish Wolfhound
Great Dane
Scottish Deerhound
Rhodesian Ridgeback
Great Pyrenees
Borzoi
Saint Bernard
Eligibility Criteria
Cases: Dogs diagnosed with osteosarcoma
Cases were confirmed by either radiography and/or by biopsy.
All cases are being confirmed by a single independent pathologist.
Controls: Healthy, cancer-free dogs
10 years and older for Greyhounds,
8 years and older for all other breeds.
Owners of control dogs are contacted regularly.
All dogs of a single breed in the study are unrelated
through the grandparents.
Study participants enrolled
Breed
Breed group
Table 1: Samples collected as of May 1, 2006
OSA
cases
Controls
(8 years and older)
Rottweiler
Mastiff
91
74
Golden retriever
Mastiff
29
370
Leonberger
Mastiff
26
51
Greyhound
Long-limbed
82
151 (*10 yrs and older)
Irish Wolfhound
Long-limbed
12
11
Great Dane
Long-limbed
22
33
All other breeds
NA
49
189
Relationship between haplotype size and breed
Haplotype  The series of alleles
along a chromosome inherited by an
individual from one parent.
Corgi
Irish wolfhound
Number of
generations
since common
ancestor
Within a breed
(Corgi-Corgi)
Between breeds
(Corgi-Irish
wolfhound
Number of
founders
Shared
Haplotypes
few
few
large
many
many
small
Stage 1: OSA susceptibility gene mapping
in each breed separately
x x
x x
1/
2Mb-1Mb
We’ve started with Rottweilers:
Searched genome for 0.5-1 Mb disease-associated haplotypes
OSA+
OSA+
Stage 2: Fine mapping across dog breeds within a breed group
10-20kb
Add related breeds to narrow the region to a smaller, shared ancestral haplotype
(~10-20kb). We are just getting started with this!


Golden retriever
(OSA+)

“OSA”
gene


Leonberger
(OSA+)

Likelihood of an association
Whole Genome Scan in Rottweilers:
Location in the genome
Regions associated with osteosarcoma in Rottweilers
B
C
-log(10) of the p-value
A
SNPs
SNPs grouped into 4 gamete “haplotype” blocks
•
•
Localized to three haplotype blocks
Association performed using Haploview
•
p< 0.01 with 100,000 permutations.
Currently ongoing…
•Screening for mutations
•Genome scan in greyhounds
•Fine mapping in related breeds
Acknowledgements
Michigan
Sarah Mandelbaum
Evan Keller
Larry Baker
Broad Institute
Elinor Karlsson
Claire Wade
Tara Biagi
Sara Fryc
Noriko Tonomura
Kerstin Lindblad-Toh
Collaborators
Frode Lingaas (Oslo)
Guillermo Couto (Ohio)
Jaime Modiano (Denver)
Elaine Ostrander (NHGRI)
Bill Kisseberth (Ohio)
Cheryl London (Ohio)
Elizabeth McNiel (Minnesota)
Uppsala
Nicolette Hillbertz
Leif Andersson
Göran Andersson
Åke Hedhammar
Olle Kämpe
Henrik von Euler
Gerli Pielberg
Barbara Kitchell (Michigan State)
Karen Michalski (Serenity Animal Hosp)
Chieko Azuma (Tufts)
David Vail (Colorado)
Chand Khanna (NCI)
This study was funded in part by the Canine Health Foundation.
Dog photos were provided by Warren Photographic.