Supplementary Information

Transcription

1 Sulementary Information Figure S1. Regions of association defined with LD cluming. For each of the 33 regions with <.5, we used a two stage aroach to define the start and end (dashed lines) of the associated locus of SNPs in LD (r >.8) to the to SNP in the region. For each SNP, the strength of LD to the to associated SNP is shown on a red (high) to blue (low) gradient. Figure S. Ages of affected and unaffected samles. For the majority of samles (91% of greyhounds, 7% of rottweilers and 93% of IWH), we had the age of disease onset or the age last confirmed healthy. The IWH samle set had clear bimodal distribution with a higher roortion of young affected dogs, allowing us to comare young cases (< years) to older controls (> years) while maintaining a sufficiently large samle set for GWAS analysis. Figure S3. Phenotye variance exlained at different association thresholds. Variance exlained dros at more stringent association thresholds but is substantial even when just a handful of regions are included, suggesting a small number of strong loci underlie the high OS risk in these breeds. The henotye variance exlained (black line) and standard error (ink shaded area) are shown for each breed over a range of significance thresholds. The henotye variance exlained would be otimally estimated using just the causal loci, if known. Because of the small samle sizes and relatively small number of risk loci, estimating the variance using the whole genome gives large standard errors (.3,.8 and.1 in the greyhounds, IWH and rottweilers resectively). Here, and in figure, we mitigated this effect by analyzing large genomic regions around significant association eaks. Average region size is.3mb,.5mb and 3.5Mb in the greyhounds, IWH and rottweilers. The number of regions included is shown along x axis.

2 Figure S. The difference in genotye relative risk between cases and controls ersists at more stringent significance thresholds. We estimated the relative risk for each dog based on the genotyes and odd ratios of SNPs at increasing significant thresholds (x axis). In each breed, the median GRR of affected dogs (red line) is distinct from the control dogs (black line) even at more stringent association thresholds than the.5 used in figure. The number of SNPs included in the GRR is along the x axis. Shaded areas encomass the 5-75% ercentiles of affected (ink) or unaffected (grey) dogs. Figure S5. Cross breed meta-analysis of OS GWAS datasets. We combined the GWAS datasets in a random-effects meta-analysis using PLINK and found no evidence of an excess of high scoring SNPs when we comared the observed and exected values for a, the meta-analysis of all three breeds, or (b, c, d) each ossible air of breeds. Dashed lines show 95% confidence intervals, with the red line indicated the median exected value. Figure S. Transcrition factor motif analysis of to candidate variant. We tested both the wild-tye and the risk allele for our to candidate variant using FIMO, and found 7 significant for the wild-tye allele and 1 for the risk allele, using recommended significance thresholds. We tested all 8 with TOMTOM, and just one, for PAX5, was significantly detected by both tools and was secific to the non-risk allele (C, black box), suggesting that the risk allele (A) will disrut binding. Figure S7. Size distribution of fixed and RRV regions by breed. a, A substantial ortion of the genome in each breed is fixed (SNPs with MAF <.5). The IWH breed has substantially more and longer blocks of fixation than the greyhounds or rottweilers. b, in each breed, we identified reduced relative variability (RRVs), the 1% of the genome that had excetionally few segregating SNPs comared to 8 other breeds and includes regions of near, but not comlete, fixation. Because the statistic accounts for the genomic background of the breed, the IWH are not longer outliers (Vaysse 11).

3 Figure S8. CGH enetrance lots of dog and human OS. Array-based comarative genomic hybridization analysis (~kb-resolution) demonstrates the extensive karyotyic instability of OS in a, greyhounds and b, rottweilers, with a large number of widely distributed CNAs (genomic gains in blue, losses in red) exceeding 5% enetrance. c, the CGH rofiles of greyhound and rottweiler OS cases show remarkable global conservation. A two-tailed Fisher s Exact test of all CNAs with!% differential enetrance identified regions with significantly different (<.5) coy number status between breeds (indicated by horizontal bars underneath the corresonding region in the comarison rofile), but none are significant after multile testing correction. d, recoding dog CGH data into human genomic coordinates highlights gross conservation with human OS (Angstadt et al. 11), also evident at e, the to OS associated locus on near CDKNA and CDKNB on human chromosome 9. Figure S9. Dog GWAS results in human OS associated regions. We see no association in dogs (red circles) at two loci significant in a human GWAS of osteosarcoma (dashed lines). Fixed SNPs (minor allele frequency <.5) are shown as black dots. Closer examination of SNP minor allele frequencies in the breeds (filled grey circles) shows extensive fixation in the IWH for the second GWAS locus on human chr, a gene desert, and shorter blocks of fixation in Rottweilers near the GRM locus on human chr. Figure S1. Confidence interval comarison. We defined significant associations using emirically defined 95% confidence intervals (black lines) rather than the less conservative intervals exected for a uniform distribution (red dashed lines). Table S1. Association at to SNPs when sex is included as a covariate in the EMMAX genomewide analysis. Table S. To associated variants in the greyhound finemaing and imutation analysis.

4 Table S3. Meta analysis of 9 breeds at the to candidate variant Table S. GRAIL keywords overreresented among genes in more than one osteosarcoma GWAS region Table S5. Association between to GWAS SNPs and CGH gain/loss in matched germline tumor samles, controlling for breed clusters using Cochran-Mantel-Haenszel (CMH) test Table S. Osteosarcoma related microrna sets curated from literature for INRICH testing.

5 Figure S1 Greyhound #1 Greyhound # Greyhound #11 LD 1 log 1 Greyhound # Greyhound #7 Greyhound #1 log 1 log 1 log 1 log 1 log 1 log 1 log 1 log 1 log 1 log 1 Greyhound #3 13 Greyhound # Greyhound #5 IWH #3 Rottweiler #1 Rottweiler # Rottweiler #3 Rottweiler # Rottweiler #5 Greyhound #8 Greyhound #13 Greyhound #9 Greyhound #1 Greyhound #1 IWH #1 IWH # Rottweiler #9 IWH # Rottweiler #11 Rottweiler #1 Rottweiler #1 Rottweiler # Rottweiler #13 Rottweiler #7 Rottweiler #1 Rottweiler #8 Rottweiler #15

6 Figure S fraction of samles fraction of samles fraction of samles Greyhound Irish wolfhound Rottweiler unaffected affected age in years

7 Figure S3 1. Greyhound fraction variance exlained Irish wolfhound fraction variance exlained Rottweiler 1. fraction variance exlained association cutoff (-log 1 )

8 Figure S Greyhound genotye relative risk Irish wolfhound genotye relative risk Rottweiler 5 genotye relative risk association cutoff (-log 1 ).5 5

9 Figure S5 a Greyhound, rottweiler and IWH b Greyhound and IWH c Greyhound and rottweiler d Rottweiler and IWH

10 Figure S FIMO TOMTOM T.F. motif P reference C allele P risk A allele candidate variant P reference C allele CHD_disc 1.E-5 EF_disc5 1.E-5 SP1_disc 3.E-5 Pax-5_known3 7.E-5 3.8E-3 RFX5_known 9.3E-5 HNFA/MA E-5 1.1E- Rfxdc_1/PB5.1.5E-5 3.E-3 Rfx_1/PB55.1.E-5 7.8E-3

11 Figure S7 Fraction of genome.3..1 a) Fixed b) Reduced relative variability Greyhound.3 Irish wolfhound Rottweiler region size (kb) 3 region size (kb)

12 Figure S8 a +1% Greyhound (n=1) % -1% b +1% Rottweiler (n=1) % -1% c +1% Breed comarison: greyhound vs. rottweiler % -1% d +1% X dog OS (n=) dog chromosome % -1% +1% human OS (n=15) % -1% X human chromosome e +1% dog OS (n=) to greyhound GWAS SNP % -1% +1% human OS (n=15) % -1% osition on human chr 9

13 Figure S9 Greyhound OS GWAS eak on human chr (GRM) OS GWAS eak on human chr (gene desert) -log 1 minor allele frequency Irish wolfhound minor allele frequency -log 1.5. Rottweiler log 1 minor allele frequency osition on dog chr 1 (Mb) osition on dog chr 17 (Mb)

14 Figure S1 Greyhound Rottweiler Irish wolfhound observed -log 1 () observed -log 1 () observed -log 1 () exected -log 1 exected -log 1 exected -log 1

15 !"#$%&'()&!""#$%&'%#()&')'#*)+,-")./()"1)%")%($3)&")&)$#5&%&')%()'/)788!9) :(#;.%)&(&<"%"=)& SNP and osition Greyhound (sex as covariate) (sex as henotye) f(a) males f(u) males f(a) females f(u) females BICFP133 chr11:57.8e BICFP1179 chr8: E BICFS chr13: E BICFS3351 chr5: E BICFP597 chr1: E BICFP11977 chr5: E BICFG35189 chr19: e BICFG38139 chr1:39 3.1E BICFG chr15: E TIGRPP153 chr1: E TIGRPP3311 chr5: E BICFS351 chr1: e TIGRPP5171 chr3: E BICFP198 chr7:738.5e Rottweiler BICFP11153 chr1: e BICFP1135 chr: e BICFP113 chr1: E TIGRPP7733 chr35: e BICFP31331 chr9: E BICFP11987 chr38: E TIGRPP875 chr1: E BICFS chr17: E BICFG35938 chr3:5171.8e BICFP91 chr3: E TIGRPP71 chr15: e BICFP1185 chr1: E BICFS chr: E BICFG39557 chr5: E BICFP8153 chr: e Irish Wolfhound BICFS3753 chr5:15 1.1E BICFP135 chr18: E BICFP1538 chr1: E BICFP1153 chr9: e ) )

16 ) Table S. To associated variants in the greyhound finemaing and imutation analysis Position on dog chr 11 variant tye tying method human coordinate risk allele f (case) f (control) Fragment* nonrisk allele Can- Fam allele hg A,B SNP genotyed chr E-8 T.8.8 C C C C SNP genotyed chr E- A C A A C SNP genotyed chr E-7 G.9.77 A G G INDEL imuted nf 3.E-8 C.87.8 CG C nf F DEL & SNP imuted chr E-8 G*.87.8 A** A A 73 F SNP genotyed chr E-5 C T C nf 57 G SNP imuted & validated chr E-8 A.87.8 C C C * luciferase assay fragment covering region (figure 3C) **variant site consists of a combined 8 b deletion and a SNP nf = matched region not found in human genome build hg19

17 Table S3: Meta analysis of 9 breeds at to candidate variant breed # aff # unaff risk allele* Frisk (aff) F (unaff) P (controlling for breed) Greyhound A E-8 Leonberger 3 5 A Great Pyrenees 1 1 A Irish Wolfhound 7 31 A Rottweiler 9 77 A Mastiff 1 13 C Golden retreiver 37 3 C Labrador retriever 1 C Great Dane C.5.1. Risk breeds (greyhound, Leonberger, Great Pyrenees, Irish wolfhound, rottweiler) Risk breeds without greyhound (Leonberger, Great Pyrenees, Irish wolfhound, rottweiler) A 1.7E A.3 All 9 breeds 87 A 7.9E-5 All breeds excet greyhound (8 breeds) 3 35 A.8 odds ratio with 95% C.I. 3. (.3-.).1 (.88-.1). (.77-.) 1.1 (.8-.31) 1.51 (.- 5.7) 1.7 ( ) 1.59 ( ).3 (.7-8.8).87 ( ).5 ( ) 1.79 (1.5-3.) 1.73 ( ) 1.5 ( ) Breslow- Day (OR heterogeneity) E-3.1 * risk allele is defined as allele with higher frequency in cases than in controls

18 Table S. GRAIL keywords overreresented among genes in more than one osteosarcoma GWAS region GWAS region Greyhound Gene GRAIL #1 "bone" # "rank" #3 "bleomycin" # "rankl" #5 "neural" # "differentiation" OTX.1 X X X X X X 5 BMPER.1 X X X X X X X X 7 EN1.7 X X X X X X X X X 11 CCL.199 X X X X X FOSB.83 X X X X X 1 ERCC.1 X X X Irish wolfhound 1 BLID.198 X X VWC.7 X X X X IKZF1. X X X X X X X X X X DDC.115 X X X X X X X X X TNFRSF11 3 A.93 X X X X X X BCL. X X X X X X X X X X X Rottweiler DLL3.9 X X X X X X X X X MIA.15 X X X LGALS1.15 X X NUMBL.57 X X X X X X X 1 EID.159 X CYPA.158 X ITPKC.1585 X X X X AKT.158 X X X X X X X X X X X PRX.1589 X X C19orf7.18 X X KIAA1.57 X 5 BLMH.11 X X FAM5C.7 X X X X X X X X 7 NELL1.8 X X X X X X X 9 EMCN.7 X X X X X X X X X X 11 LTAH.7 X X X X X NTN.189 X X X X X 1 TCF1. X X X X X X X X X X X X X #7 "develoment" #8 "morhogenetic" #9 "locus" #1 "endothelial" #11 "engrailed" #1 "coronary" #13 "notch" #1 "sns" #15 "atterning" #1 "embryos" #17 "artery" #18 "signaling"

19 Table S5. Association between to GWAS SNPs and CGH gain/loss in matched germline tumor samles, controlling for breed clusters using Cochran-Mantel-Haenszel (CMH) test SNP risk allele CGH robe P (CMH) genes heno -tye breed # affect # unaff f (aff) f (unaff) breed chr (rottweiler region 5) A chr BLMH gain grey rott chr (greyhound region 1) C chr gain grey rott fixed chr (rottweiler region 1) G chr ERCC1, FOSB gain grey 7.5 rott chr (rottweiler region 5) A chr TMIGD1 loss grey 7.9 rott chr (rottweiler region ) C chr loss grey rott

20 Table S. Osteosarcoma related microrna sets curated from literature for INRICH testing. Source # genes genes (autosomal genes maed to canfam) genes in 3 breed fixed regions* INRICH INRICH corrected Jones 1 8 MIR1B, MIR1, MIR1, MIR15, MIR15B, MIR1-, MIR181C, MIR19, MIR19A, MIR195, MIR1, MIR1, MIRB, MIR7A, MIR3, MIR335, MIR3, MIR51, MIR83, MIR8, MIR87A, MIR88, MIR57, MIR1, MIR5, MIR57, MIR3, MIRLET7G MIR15, MIR335, MIR3, MIR5, MIR Lulla MIR1, MIR135B, MIR1, MIR1, MIR18A, MIR15, MIR18A, MIR198, MIRB, MIR1, MIR31B, MIR51, MIR5, MIR55, MIR511-1 MIR1, MIR Maire Sarver 1 1 Thayanithy 1 8 MIR1, MIR1, MIR17, MIR137, MIR1, MIR18A, MIR15, MIR181A1, MIR181A, MIR195, MIR199B, MIR18-1, MIR18-, MIR99, MIR31, MIR39-1, MIR39-, MIR335, MIR37A1, MIR37A, MIR37C, MIR377, MIR38, MIR9, MIR1, MIR3, MIR51, MIR93, MIR95, MIR97, MIR53, MIR5, MIR758 MIR13B, MIR1, MIR15, MIR17, MIR18A, MIRA, MIR39, MIR377, MIR381, MIR9, MIR31, MIR9A, MIR93, MIR539, MIR1, MIR, MIRA, MIR53, MIR8, MIRLET7C, MIRLET7E MIR1B, MIR17, MIR15B1, MIR13B, MIR13, MIR19, MIR15, MIR15A, MIR17, MIR198, MIR19A, MIR3, MIR33A, MIR33B, MIR35, MIR39, MIR37, MIR38, MIR9, MIR5, MIR3, MIR33, MIR9A, MIR87A, MIR511-1, MIR539, MIR5, MIR5 MIR MIR Thayanithy 1 (downregulated x) 1 MIR15B1, MIR15, MIR198, MIR39, MIR37, MIR37A1, MIR38, MIR9, MIR3, MIR33, MIR87A, MIR511-1, MIR539, MIR5 MIR Thayanithy 1 (uregulated x) 1 MIR1B, MIR17, MIR13B, MIR19, MIR15A, MIR17, MIR19A, MIR19-1, MIR3, MIR33A, MIR33B, MIR35, MIR5, MIR9A all OS micrornas 17 MIR1, MIR1B, MIR17, MIR1B, MIR15B1, MIR1, MIR17, MIR13B, MIR13, MIR13, MIR135B, MIR137, MIR1, MIR1, MIR13, MIR1, MIR18A, MIR19, MIR15, MIR151, MIR15, MIR15A, MIR15B, MIR1-, MIR17, MIR181A1, MIR181A, MIR181C, MIR18A, MIR19, MIR19A, MIR191, MIR193A, MIR195, MIR198, MIR199A1, MIR199B, MIR19A, MIRB, MIRA, MIR1, MIR1, MIR1, MIR18-1, MIR18-, MIR19-1, MIRB, MIR7A, MIR99, MIR31B, MIR31, MIR3, MIR3, MIR39-1, MIR39-, MIR335, MIR33A, MIR33B, MIR3, MIR35, MIR39, MIR37, MIR37A1, MIR37A, MIR37C, MIR377, MIR381, MIR38, MIR9, MIR1, MIRA, MIR5, MIR31, MIR3, MIR33, MIR9A, MIR51, MIR5, MIR55, MIR83, MIR8, MIR87A, MIR88, MIR93, MIR95, MIR97, MIR511-1, MIR539, MIR53, MIR5, MIR57, MIR1, MIR1, MIR, MIRA, MIR5, MIR5, MIR53, MIR5, MIR57, MIR3, MIR8, MIR758, MIR99B, MIRLET7C, MIRLET7E, MIRLET7G MIR1, MIR15, MIR335, MIR3, MIR5, MIR3.5.5 * regions of overla of fixed regions (regions > 5kb where all SNPs have MAF <.5) from greyhounds, rottweilers and IWH