- Email: [email protected]
A novel locus for canine osteosarcoma (OSA1) maps to CFA34, the canine orthologue of human 3q26
Genomics 96 (2010) 220–227
Contents lists available at ScienceDirect
Genomics j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / y g e n o
A novel locus for canine osteosarcoma (OSA1) maps to CFA34, the canine orthologue of human 3q26 Jeffrey C. Phillips ⁎, Luis Lembcke, Tamara Chamberlin The University of Tennessee, Small Animal Clinical Sciences, C247 Veterinary Teaching Hospital, Knoxville, TN 37996-4554, USA
a r t i c l e
i n f o
Article history: Received 18 February 2010 Accepted 2 July 2010 Available online 18 July 2010 Keywords: Cancer Osteosarcoma Canine Linkage Locus OSA1
a b s t r a c t Spontaneous tumors in dogs share many of the same features of their human orthologues including clinical behavior, response to treatment, and molecular defects. It is therefore natural to consider the use of dogs and their spontaneous malignancies in the study of complex disease such as cancer. Scottish Deerhounds are a giant breed of dogs that exhibit a high incidence of bone cancer. Our previous work suggested that osteosarcoma within this breed could be explained by the presence of a major gene of dominant effect. Herein, we use a whole genome mapping approach to evaluate a four-generation pedigree of Scottish Deerhounds for linkage of their osteosarcoma phenotype. Using this approach we found evidence of linkage (Zmax = 5.766) between their phenotype and markers located on CFA34, in a region syntenic to human chromosome 3q26. The identification of this locus provides novel insight into the genetic basis of osteosarcoma in both canines and humans. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Osteosarcoma (osteogenic sarcoma, OSA) is the most common (N85%) primary malignant bone tumor in both humans and canines. In the United States, approximately 10,000 canine and 800 human cases of osteosarcoma are diagnosed each year [1–3]. Of the cancers that occur in both pet dogs and humans, osteosarcoma is perhaps the most similar. In both species it is histologically composed of mesenchymal cells that produce osteoid matrix. The dog and human also share similar anatomic predilections including the distal femur, proximal tibia, and proximal humerus amongst others. The clinical behavior in both species is characterized as highly aggressive with very high rates of metastatic spread. In fact, 85–90% of both humans and pet dogs are considered to have microscopic metastatic disease at the time of initial diagnosis [3,4]. Humans and pet dogs also share a similar response to surgery and chemotherapy regimens. While dramatic progress has been made in the treatment and outcome in both species, there is limited understanding of the etiology of the disease. Because of the many similarities shared by humans and pet dogs with osteosarcoma, it is only natural to use dogs and their spontaneous malignancies to more fully understand disease pathogenesis. The genetic changes that occur in osteosarcoma have received considerable attention in recent years in an attempt to elucidate ⁎ Corresponding author. College of Veterinary Medicine, Department of Small Animal Clinical Sciences, 2407 River Drive, C247 Veterinary Teaching Hospital, Knoxville, TN 37996-4544, USA. Fax: + 1 865 974 5554. E-mail address: [email protected] (J.C. Phillips). 0888-7543/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.ygeno.2010.07.002
etiology in both humans and canids. In humans, loss of heterozygosity screening has identified several chromosomal regions with high frequencies of allelic loss including chromosomes 3q, 13q, 17p, and 18q [5–8]. Heritable cancer syndromes, although extremely rare, have been associated with two of these regions; namely chromosomes 13q (Retinoblastoma) and 17p (Li-Fraumeni) [9]. Comparative genomic hybridization has also been used to identify regions of genomic amplification in human-origin tumors, including the most common amplicons at 12q11–q15 and 17p11–13 [10]; while transcriptome profiles of osteosarcoma cell lines has provided information on the molecular variation between tumors [11]. In dogs, gene expression studies have identified profiles associated with both tumor progression [12] and survival [13]. These profiles demonstrate strong similarities between canine and human osteosarcoma along with potential biologic targets in both species. Comparative genomic hybridization of canine tumor tissues has also identified regions of recurrent cytogenetic changes within select breeds and correlated these regions with candidate tumor suppressor or oncogenes [14]. In spite of these advances made in understanding the tumor biology of both human and canine osteosarcoma, it is unclear what precursor (germline or heritable) lesions are critical for tumorigenesis. In the canine population, large and giant breeds of dogs are at increased risk for the development of OSA; together these breeds account for the majority (N80%) of reported cases [4]. The Scottish Deerhound is a giant breed dog first recognized by the American Kennel Club (AKC) in 1886. The overall breed-specific incidence of OSA in the Deerhound has been estimated to be greater than 150 cases per 1,000 dogs (compared to 7 cases per 100,000 dogs in the general dog population) [4,15]. Given the high incidence of OSA within a
J.C. Phillips et al. / Genomics 96 (2010) 220–227
relatively small, closely related population, it is likely that a major genetic modifier (locus) for the disease exists within this breed. In a recent report [16], we described a population of greater than 1,000 Scottish Deerhounds that were ascertained as part of a prospective health study. Statistical analysis of this population supported the presence of a “putative” high-risk allele associated with the development of osteosarcoma. By comparing models of disease transmission, we were able to show that a highly penetrant, dominant model was most probable. The frequency of the high-risk allele in this dominant model was estimated to be 0.128, and only genotype was found to be correlated with disease onset. The artificial selection that has led to this high, breed-specific, incidence of OSA can be further utilized to identify the genetic determinants which regulate disease onset. Specifically, the inbreeding that has led to this high-incidence of disease also serves to simplify its transmission properties. While in humans OSA is clearly a complex disease with no obvious transmission patterns, the canine orthologue in the Scottish Deerhound exhibits a simple dominant transmission. Furthermore, with the completion of the canine genome sequencing project and the availability of high-resolution canine genetic maps, linkage and association studies can be pursued in canine populations such as the Scottish Deerhound. Through these mapping studies candidate loci and genes can be identified and used to further our understanding of not only the canine but also the human disease. In this study, we describe the mapping of the osteosarcoma phenotype (OSA1) that is segregating in a four-generation pedigree of Scottish Deerhounds. A whole genome linkage approach is used to map OSA1 to CFA34, and to exclude other chromosomal regions. The identification of a novel locus for canine osteosarcoma provides important information on the location of candidate genes for both canine and human OSA. 2. Materials and methods
221
tumor location, and treatment option. All procedures involving animals in this study were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Tennessee. 2.2. DNA samples EDTA preserved whole blood samples were collected from 60 dogs in the four generation pedigree. DNA was purified using standard methods. 2.3. Markers An extended microsatellite screening set was used for the whole genome scan. The microsatellite set included markers previously identified in a mapping set referred to as MSS3 [19]; 103 additional markers (610 total) were selected throughout the genome to increase coverage of select regions. Marker positions are as noted in the CanFam 2.0 genome assembly [20]. Average physical spacing of markers in this extended set was 4.1 Mb. Genetic distances between markers were obtained from the Canine Genetic Linkage Map [21] and verified on this sample population using the software SimWalk2 [22]. 2.4. Genotyping Fluorescent-labeled oligonucleotides were used to amplify polymorphic markers used for linkage analysis. Samples were run on a MegaBace® 1500, 96-capillary array (GE Healthcare) and genotypes were analyzed using the Genetic Profiler software (version 2.2; GE Healthcare). All genotypes were manually scored with values converted to simple integers (i.e. 1–9) for statistical analysis. Genotypes for the 75 untyped dogs were scored as unknown.
2.1. Sample population
2.5. Statistical analyses
Dogs used in this study were all privately owned and AKC registered Scottish Deerhounds from a larger population previously described [16]. A four-generation pedigree of Scottish Deerhounds was selected from this larger population and included 135 dogs; 50 of whom were identified as affected with osteosarcoma. DNA samples were available from a total of 60 dogs in this pedigree, with the remaining dogs (i.e. 75 dogs) providing phenotypic information that allowed for the construction of a comprehensive pedigree. Current and historic health information on all dogs was collected from owners, breeders, and veterinarians. Affected status was determined by the presence of characteristic radiographic findings of aggressive bone disease and/or histologic examination. Pedigree information and medical histories were entered into a pedigree management program (Cyrillic 2.1.3, Cherwell Software). The program PEDINFO was used to determine descriptive characteristics of the population. This program is available as a component of the genetic analysis package SAGE 5.2.0 [17]. Inbreeding coefficients were estimated using the Kinship Matrices option of the genetic software MENDEL 6.01 [18]. Health information collected included a variety of qualitative and quantitative traits (i.e. covariates). Qualitative traits included OSA status, presence/absence of other diseases including cancers, and gender. OSA status had three possible outcomes: affected, unaffected, or unknown. Quantitative traits included age at exam, age of onset, height, weight, and length of time sexually intact (as a measure of cumulative sex hormone exposure). Age at exam was considered to be the age at last known contact or age at death where applicable. Age of onset documented the age at which definitive diagnosis of osteosarcoma was made in affected individuals. Time was measured in years, height in inches at the shoulder and weight in pounds. Other information collected (where applicable) included cause of death,
Osteosarcoma was considered a dichotomous trait in which individuals were classified as “unaffected” (at the time of exam), “affected”, or in some cases “unknown status”. Individuals listed with “unknown” phenotypes were included to provide a non-disjoint fourgeneration pedigree; the disease phenotypes for these individuals were ignored for each analysis. Multipoint linkage analysis was performed on a 64 bit (Xeon) PC computer using the SimWalk2 (v2.91) program [22,23]. SimWalk2 uses a Markov chain Monte Carlo (MCMC) algorithm and simulated annealing algorithms to avoid limitations due to sample size or inbreeding. The program was compiled from source code for usage on a 64-bit platform. The osteosarcoma phenotype (OSA1) segregating in this pedigree was modeled as both a dominant and recessive Mendelian trait with allele frequencies of qa = 0.128 and qa = 0.501, respectively [16]. Microsatellite marker allele frequencies were obtained from analysis of a larger population previously described [16]. Recombination frequencies were assumed to be equal between males and females due to software limitations. Based on our previous work, multipoint LOD-score calculations were performed at penetrances of 1.0 and 0.9, under the assumption of locus homogeneity (alpha = 1.0). All other parameters were set to their default values under SimWalk2. LOD scores (Zmax or location scores) of ≥3.0 were considered consistent with linkage. Conversely, LOD scores of ≤−2.0 could be used to exclude regions from consideration. For comparison, a LOD score of N3.0 is associated with a genome-wide significance level of b0.05 in favor of linkage [24] while a LOD score of ≤−2.0 corresponds to 100:1 odds against linkage [25]. Haplotype analysis was used to determine the minimal candidate region for location of the osteosarcoma gene. Haplotypes were estimated using SimWalk2 and illustrated using Cyrillic 2.1.3.
222
J.C. Phillips et al. / Genomics 96 (2010) 220–227
3. Results 3.1. Descriptive analysis Thirteen sibships comprising a total of 135 dogs were selected from a previously described population of 1,272 Scottish Deerhounds [16] (Table 1). These sibships could be arranged into one large four generation pedigree. DNA samples were available from 60 of the 135 dogs for use in a whole genome scan. The remaining 75 untyped dogs provided health information only and were used to construct the comprehensive pedigree. A total of 9 inbreeding loops were identified in this pedigree, with an average inbreeding coefficient of 0.086 estimated using the kinship matrices option of MENDEL 6.01. Due to the graphical complexity, only a small subset of the pedigree is shown in Fig. 1. This pedigree included 68 male dogs and 67 female dogs, ranging in age from 4.0 to 13.6 years. Of these dogs, 50 were confirmed affected, 83 were unaffected at the time of exam, while 2 had unknown health status or were lost to followup. A similar gender distribution was found in the subset of dogs for which DNA samples were available (32 female and 28 male dogs). This subset, however, was further enriched with affected individuals including 40 affected dogs and only 20 dogs who were unaffected at the time of exam. The absolute percentage of affected dogs amongst all dogs with known health status in the pedigree was 38% (50/133), which is higher than the population based prevalence previously reported (i.e. 21%) [16]. Affected dogs ranged in age from 4.9 to 13.6 years with a mean age of onset of 7.6 years (±1.9 yrs). Unaffected dogs were significantly older than affected dogs (8.5 ± 2.9 yrs, p = 0.02). This pedigree of Scottish Deerhounds compares favorably to the previously described larger population [16] with no significant difference in age of onset, gender distribution, or gender-specific prevalence of osteosarcoma (Table 1). Differences in disease prevalence are directly attributable to phenotypic based selection of this pedigree. In other words, this pedigree is
enriched with affected individuals which allows for the use of familial based linkage methodology. Finally, no correlation was found between height, weight, or duration of gonadal hormone exposure and osteosarcoma (data not shown); similar to the larger population [16]. 3.2. Linkage analysis The whole genome linkage data consisted of genotypes from the 60 dogs at the 610 microsatellites identified from the CanFam 2.0 genome assembly. Multipoint linkage analysis was performed using SimWalk2 under both dominant and recessive models. Allele frequencies for the disease allele were set to qa = 0.128 and qa = 0.501, for the dominant and recessive model, respectively (Table 1). These values were based on our previous populationwide segregation analysis of the disease phenotype (OSA1) [16]. The resultant parametric multipoint LOD scores are illustrated in Fig. 2. These scores are based under the assumption of no locus heterogeneity within the pedigree and a fully penetrant trait. Similar results were found with a reduced penetrance of 0.9 (data not shown). The majority of the genome was excluded from linkage on the basis of LOD scores b−3.0 (Fig. 2). The regions with LOD scores of N1 but b3 can neither be definitively excluded from linkage nor are considered to exhibit strong evidence of linkage. Table 2 shows all results reaching a significance threshold of equivalent logarithm of odds (LOD) score N1. Regions from chromosomes 1, 15, 22, and 25 exhibited weak evidence of linkage with LOD scores marginally N1 (Table 2). These regions contained markers with relatively low information content due to high levels of homozygosity. Given the acrocentric nature of canine chromosomes, each of these four regions are localized to the most centromeric portion of their respective chromosomes; regions typically associated with consistently low recombination rates and heterozygosity across breeds [26]. These centromeric regions also exhibited low levels of recombination and
Table 1 (A) Descriptive statistics for 135 dogs in 13 families analyzed in this report (current pedigree) compared to a previous report [16] (mean ± standard deviation); (B) Parameter values used for linkage analysis performed with the SimWalk2 [22] software package. (A) Descriptive statistics
Total individuals Males Females Affected Unaffected %Affected Missing Probands Sibships Sibship size # DNA samples available Inbreeding loops Inbreeding coefficient Age (years) Age of onset (years)
Current pedigree
Phillips et al, 2007 [16]
135 68 67 50 82 38% 2 1 13 1–9 (3.0 ± 2.4) 60 9 0.086 4.0–13.6 (7.7 ± 2.1) 4.9–13.6 (7.6 ± 1.9)
1057 479 578 211 794 21% 52 5 314 1–11 (3.0 ± 1.9) 672 187 0.032 0.5–14.9 (6.8 ± 2.8) 2–13.9 (7.0 ± 2.0)
(B) Parameter values Parameter Allele frequency Dominant Recessive Penetrance Complete Reduced Alphaa Mistyping probability a
Value 0.128 0.501 1.0 0.9 1.0 0.025
The a priori proportion of pedigrees that are segregating an affected gene linked to this set of marker loci. A value of 1.0 assumes genetic homogeneity.
J.C. Phillips et al. / Genomics 96 (2010) 220–227
223
Fig. 1. Deerhound pedigree. Representative subset pedigree of dogs used in this study with known affected status. Standard pedigree symbols are used. Affected individuals are represented by filled symbols. Multiple inbreeding (double lines) and mating loops are present. Not all relationships are shown.
heterozygosity in the population studied here (data not shown); factors which likely contributed to the creation of what are essentially non-informative chromosomal regions and hampered our ability to statistically eliminate them. This is a relatively common finding, to note regions of extended homozygosity; which are often especially marked around centromeres [20,21,26]. A region with moderate evidence of linkage (Zmax = 2.246) was found between the osteosarcoma phenotype (OSA1) and markers located in the telomeric portion of chromosome 18 (Figs. 2 and 3). This region spans 7 cM between markers REN168H09 and FH2429 and is characterized by relatively high levels of recombination and heterozygosity across breeds [20]. In the Scottish Deerhound population, relatively low levels of recombination were found for nearly all of chromosome 18; however, a relative increase in recombination rate was noted in the center of the region of interest (Fig. 3A). This effect led to the creation of a conserved haplotype block seen within the affected dogs of this pedigree, which extends to marker FH2429 of chromosome 18. Significant linkage was found between markers located on chromosome 34 (CFA34) and the osteosarcoma phenotype in this pedigree (OSA1). This 40-cM region extends from markers REN243023 through REN238D23 (Fig. 2). The peak LOD score (Zmax = 5.766) was found at 46.8-cM near marker REN234E12 (Fig. 3B). Using SimWalk2, haplotypes were constructed for chromosome 34 (CFA34); estimated haplotypes from the most informative sibship were then used to identify the critical region for OSA1 (Fig. 4). Affected individuals shared a common haplotype for markers spanning this region. Critical recombination events in affected individuals occurred between markers FH3836 and REN234E12 and REN173D06 and REN44K21, defining a 10-cM critical interval extending from marker FH3836 to REN44K21. Several unaffected individuals demonstrate recombination events that could significantly narrow the critical region. However, because of the reduced penetrance of this condition [16], one or more of these individuals could still have the affected allele. To eliminate the potential ambiguity, the critical region has been defined by recombination events seen in individuals expressing the disease phenotype. This region corresponds to
4.5 Mb on chromosome 34q16.2–q17.1 (CFA34) extending from 34.9 Mb to 39.4 Mb on the CanFam 2.0 genome assembly. 4. Discussion Here we have mapped the major locus for osteosarcoma in this pedigree (i.e. OSA1) to a 10-cM region of chromosome 34. The pedigree used for this linkage study consisted of a large four generation pedigree. DNA samples were available for genotyping from 60 of the 135 individuals in this pedigree. Forty-one of these DNA samples were collected from a highly informative portion of the pedigree which allowed for haplotype estimation (Fig. 4A). The remaining 19 DNA samples were collected from more distantly related, smaller sibships and contained relatively little information content for linkage studies (data not shown). The overall pedigree compares favorably with a larger population of Scottish Deerhounds previously described [16]; although the following key differences should be noted: (1) this pedigree was relatively enriched with affected individuals; and, (2) individuals in this pedigree were on average older than dogs in our previous report. The ascertainment process that led to these differences can bias the detection of covariates that modify disease; in this case age and perhaps other factors. Furthermore, this age difference can impact the perceived penetrance of diseases which exhibit a delayed (i.e. age-related) onset such as osteosarcoma. While no methodology exists for correcting these sources of bias in canine linkage studies, association studies on distantly related and outbred populations can be used to identify covariates that modify disease. Results from the whole genome linkage analysis effectively excluded the majority of genomic regions with the exception of loci on chromosomes 1, 15, 18, 22, 25, and 34. Each of these loci were associated with LOD scores of N1.0. While statistically the regions on chromosomes 1, 15, 22, and 25 could not be excluded; haplotype analysis proximal to these loci identified several affected recombinant individuals (data not shown). Given the low levels of recombination observed in these centromeric regions, they are unlikely to be linked to the osteosarcoma phenotype segregating in this pedigree. On
224
J.C. Phillips et al. / Genomics 96 (2010) 220–227
J.C. Phillips et al. / Genomics 96 (2010) 220–227
225
Table 2 Parametric linkage analysis results (all regions with LOD N 1 under either a recessive or dominant model). Chromosome location
Genetic location, cM (θ)
Closest marker
LOD dominant (θ)
LOD recessive (θ)
1q11 15q11 22q11 25q11 18q25 34q16.2–q17.1
1.0 1.0 1.0 1.0 48.37 46.8
FH3325 FH5157 REN49F22 REN54E19 REN266I17 REN234E12
1.042 −9.436 1.008 −9.087 2.246 5.766
−4.716 1.047 −14.591 1.020 −27.288 −56.810
chromosome 18, a region exhibiting moderate statistical significance (Zmax = 2.246) was identified telomeric to marker REN168H09. Haplotype analysis identified 3 affected recombinant individuals at marker REN47J11 (i.e. proximal to the region) but could not be used to further exclude the region due to the lack of informative markers. Additional informative markers would be helpful to evaluate this region as a covariate affecting osteosarcoma in this pedigree. Statistical evidence for linkage between markers on chromosome 34q16.2–q17.1 and the osteosarcoma phenotype segregating in this pedigree (OSA1) was strong. The most consistent model assumed genetic homogeneity (alpha = 1.0) within the population and a fully penetrant dominant trait. In contrast, results of segregation analysis [16] suggested a reduced penetrance with minimal heterogeneity (alpha = 0.95) could best explain the osteosarcoma phenotype segregating within the Deerhound population. These minor differences are likely attributable to the selection process used for the pedigree studied here. As mentioned previously, this pedigree is enriched with closely related affected individuals and thus relatively homogeneous. Furthermore, for diseases that show an age-related onset it can be difficult to distinguish between non-carrier and nonpenetrant “unaffected” individuals. In this pedigree, the unaffected dogs were significantly older than the unaffected dogs in our previous report. This age difference would be expected to lower the likelihood of misclassification of unaffected individuals and thus result in an apparent increase in disease penetrance as seen herein. Haplotype analysis identified a 10-cM critical region on CFA34 between markers FH3836 and REN44K21. This 4.5 Mb region of canine chromosome 34q16.2–17.1 is syntenic to human chromosome 3q26 [27]. Interestingly, human chromosome 3q26 is associated with a high incidence of loss of heterozygosity in both primary and metastatic osteosarcoma tumor samples from humans [28,29]. Loss of heterozygosity (LOH) is a genetic event that is thought to result in tumorspecific inactivation of suppressor genes and can be detected by comparing polymorphic markers between tumor and normal tissues [30]. Consistent loss of heterozygosity at specific polymorphic markers is generally considered a good predictor of the presence of a tumor suppressor gene located close to these markers. Furthermore, it is well known that most familial cancer syndromes are related to defects in tumor suppressor genes [31]. By using loss of heterozygosity analysis, a 1-cM critical sub region of human chromosome 3q26 between markers D3S1212 and D3S1246 has been identified [28]. This region is syntenic to CFA34 between markers FH3836 and REN234E12; the location of the peak LOD score defined in this pedigree. In human osteosarcomas, other common sites of LOH include chromosomal arms 13q, 17p, and 18q [28,29]. Chromosomes 13q and 17p are known to contain RB1 and TP53, respectively; tumor suppressor genes important for osteosarcoma tumorigenesis [32]. Retinoblastoma and Li-Fraumeni syndrome are familial cancer syndromes associated with defects in these two genes. To date, no
Fig. 3. Significant LOD scores. LOD scores (blue) and recombination rates (red) for (A) chromosome 18, and (B) chromosome 34. Physical length (Mb) is plotted on the x-axis and is scaled to accommodate varying chromosomal size. Recombination rates and LOD scores are plotted on the left and right y-axis, respectively. Recombination rates were determined in sliding 5 Mb windows as previously described [21]. Ideograms of each chromosome (left) showing critical regions indicated with blue arrow.
familial cancer syndromes have been localized to either chromosome 3q or 18q. Several candidate tumor suppressor genes, however, have been localized to chromosome 3q26 [33–36]. One such gene is the recently identified tumor suppressor gene PPM1L [35]. PPM1L encodes a novel serine-threonine phosphatase in the TGF-beta and BMP signaling pathways and is thought to play a role in APC mutationnegative familial colorectal cancer. The canine homologue for PPM1L is localized within the larger candidate region identified on canine chromosome 34. Another interesting candidate located within this region of human chromosome 3q26 is the MECOM gene (EVI1 complex locus). MECOM encodes a nuclear transcription factor, essential for the proliferation/ maintenance of hematopoietic stem cells. Aberrant expression of the MECOM gene has been associated with a variety of human hematopoietic malignancies [36]. Interesting, in the Scottish Deerhound population previously described [16], and the pedigree herein, the probability for the development of a hematopoietic malignancy in the sibling of a dog affected with osteosarcoma is 9% (data not shown). Given the relatively low population prevalence, this finding suggests an extremely high relative risk for hematopoietic malignancies in siblings of affected dogs. The canine homologue of MECOM is localized within the critical region of CFA34 defined herein and is another excellent candidate gene for this heritable cancer syndrome (OSA1). Further evaluation of the Scottish Deerhound for causative mutations associated with the osteosarcoma phenotype and as a model of the orthologous human disease is ongoing. This effort will
Fig. 2. Multipoint LOD score results. Results of parametric whole genome linkage analysis for the OSA1 phenotype. Dominant transmission (blue) and recessive (red) transmission (red) models are shown from centromeric to telomeric end (left to right). Genetic distance (cM) is plotted on the x-axis and is scaled to accommodate the varying chromosomal size, LOD scores are plotted on the y-axis. Assumptions include complete penetrance and genetic homogeneity.
226
J.C. Phillips et al. / Genomics 96 (2010) 220–227
Fig. 4. Haplotype analysis from CFA34. Results are shown centromeric to telomeric end (top to bottom). Affected individuals are shown as blackened circles. The conserved haplotype shared by affected individuals is illustrated in red. Haplotypes from the most informative portion of the pedigree are shown in “A.”
involve genome-wide and regional association studies in other outbred canine populations with the goal of identifying additional covariates affecting penetrance (genome-wide) and further narrowing the critical region of CFA34 (regional SNP evaluation). Acknowledgments Special thanks to Ms. Lyn Robb of Tannochbrae Deerhounds and Ms. Rachel Matthews for their help in collecting health information and blood samples. The author would also like to acknowledge the Canine Hereditary Cancer Consortium along with investigators at the National Cancer Institute for their ongoing efforts to investigate the underlying causes of canine osteosarcoma. This project was funded by the UTCVM Center of Excellence in Livestock Diseases and Animal Health.
References [1] U.S. Cancer Statistics Working Group, United States Cancer Statistics: 1999-2002 Incidence and Mortality Web-based Report Version, Atlanta: Department of Health and Human Services, Centers for Disease Control and Prevention, and National Cancer Institute, 2005. [2] W. Misdorp, A.A. Hart, Some prognostic and epidemiologic factors in canine osteosarcoma, J. Natl Cancer Inst. 62 (3) (1979) 537–545. [3] P.A. Meyers, Gorlick R. Osteosarcoma, Pediatr. Clin. North Am. 44 (1997) 973–989. [4] G. Ru, G. Terracini, L. Glickman, Host Related Risk Factors for Canine Osteosarcoma, Vet. J. 156 (1998) 31–39. [5] I. Pasic, A. Shlien, A. Durbin, et al., Recurrent focal copy-number changes and loss of heterozygosity implicate two noncoding RNAs and one tumor suppressor gene at chromosome 3q13.31 in osteosarcoma, Cancer Res. 70 (1) (2010) 60–71. [6] R.P. Kruzelock, E.C. Murphy, L.C. Strong, S.L. Naylor, M.F. Hansen, Localization of a Novel Tumor Suppressor Locus on Human Chromosome 3q Important in Osteosarcoma Tumorigenesis, Cancer Res. 57 (1997) 106–109. [7] T. Yamaguchi, et al., Allelotype analysis in osteosarcomas: frequent allele loss on 3q, 13q, 17p, and 18q, Cancer Res. 52 (1992) 2419–2423.
J.C. Phillips et al. / Genomics 96 (2010) 220–227 [8] T. Johnson-Pais, M. Nellissery, D. Ammerman, et al., Determination of a minimal region of loss of heterozygosity on chromosome 18q21.33 in osteosarcoma, Int. J. Cancer 105 (2) (2003) 285–288. [9] G. Ottaviani, N. Jaffe, The etiology of osteosarcoma, Cancer Treat. Res. 152 (2010) 15–32. [10] J. Atiye, M. Wolf, S. Kaur, et al., Gene amplifications in osteosarcoma-CGH microarray analysis, Genes Chromosom. Cancer 42 (2) (2005) 158–163. [11] I. Trougakos, N. Chondrogianni, I. Amarantos, et al., Genome-wide transcriptome profile of the human osteosarcoma SA OS and U-2 OS cell lines, Cancer Genet. Cytogenet. 196 (2) (2010) 109–118. [12] M. Paoloni, S. Davis, S. Lana, et al., Canine tumor cross-species genomics uncovers targets linked to osteosarcoma progression, BMC Genomics 10 (2009) 625. [13] G. Selvarajah, J. Kirpensteijn, M. Van Wolferen, et al., Gene expression profiling of canine osteosarcoma reveals genes associated with short and long survival times, Mol. Cancer 8 (2009) 72. [14] R. Thomas, W. Huixia, T. Pei-Chien, et al., Evidence for breed-associated cytogenetic aberrations in canine appendicular osteosarcoma, Chromosome Res. 17 (2009) 365–377. [15] J. Dillberger, Deerhound health survey. Part 1. Lifespan and causes of death, The Claymore, 1996, pp. 36–44, Mar/Apr. [16] J.C. Phillips, B. Stephenson, M. Hauck, J. Dillberger, Heritability and segregation analysis of osteosarcoma in the Scottish Deerhound, Genomics 90 (3) (2007) 354–363. [17] S.A.G.E., Statistical Analysis for Genetic Epidemiology, Computer program package available from Statistical Solutions Ltd, Cork, Ireland, Release 5.2.0, 2006. [18] K. Lange, et al., MENDEL Version 4.0: A Complete Package for the Exact Genetic Analysis of Discrete Traits in Pedigree and Population Data Sets, Am. J. Hum. Genet. 69(Supplement) (2001) A1886. [19] D.R. Sargan, J. Aguirre-Hernandez, F. Galibert, E. Ostrander, An extended microsatellite set for linkage mapping in the domestic dog, J. Hered. 98 (3) (2007) 221–231. [20] K. Lindblad-Toh, et al., Genome sequence, comparative analysis and haplotype structure of the domestic dog, Nature 438 (7069) (2005) 803–819. [21] A.K. Wong, et al., A Comprehensive Linkage Map of the Dog Genetics, Genetics 184 (2010) 595–605. [22] E. Sobel, K. Lange, Descent graphs in pedigree analysis: applications to haplotyping, location scores, and marker sharing statistics, Am. J. Hum. Genet. 58 (1996) 1323–1337.
227
[23] E. Sobel, H. Sengul, D.E. Weeks, Multipoint estimation of identity-by-descent probabilities at arbitrary positions among marker loci on general pedigrees, Hum. Hered. 52 (2001) 121–131. [24] Lue Ping Zhao, Ross Prentice, Fumin Shen, Li. Hsu, On the Assessment of Statistical Significance in Disease-Gene Discovery, Am. J. Hum. Genet., 64, 1999, pp. 1739–1753. [25] H.W. Deng, LOD score exclusion analyses for candidate QTLs using random population samples, Genetica 119 (3) (2003) 303–315. [26] D. Curtis, A.E. Vine, J. Knight, Study of Regions of Extended Homozygosity Provides Powerful Method to Explore Haplotype Structure of Human Populations, Ann. Hum. Genet. 72 (2) (2008) 261–278. [27] C. Hitte, et al., Facilitating genome navigation: survey sequencing and dense radiation-hybrid gene mapping, Nat. Rev. Genet. 6 (8) (2005) 643–648. [28] R.P. Kruzelock, E.C. Murphy, L.C. Strong, S.L. Naylor, M.F. Hansen, Localization of a Novel Tumor Suppressor Locus on Human Chromosome 3q Important in Osteosarcoma Tumorigenesis, Cancer Res. 57 (1997) 106–109. [29] T. Yamaguchi, et al., Allelotype analysis in osteosarcomas: frequent allele loss on 3q, 13q, 17p, and 18q, Cancer Res. 52 (1992) 2419–2423. [30] M.F. Hansen, W.K. Cavenee, Tumor suppressors: recessive mutations that lead to cancer, Cell 53 (1988) 173–174. [31] A.G. Knudson, Genetics of Human Cancer, Ann. Rev. Genet. 20 (1986) 231–251. [32] G. Letson, C. Muro-Cacho, Genetic and Molecular Abnormalities in Tumors of the Bone and Soft Tissues, Cancer Control 8 (3) (2001) 239–251. [33] A.J. Bass, et al., SOX2 is an amplified lineage-survival oncogene in lung and esophageal squamous cell carcinomas, Nat. Genet. 41 (11) (2009) 1238–1242. [34] J. Kang, S.H. Koo, K.C. Kwon, J.W. Park, J.M. Kim, Identification of novel candidate target genes, including PEHB3, MASP1, and SST at 3q26.2-q29 in squamous cell carcinoma of the lung, BMC Cancer 16 (9) (2009) 237–241. [35] L.F. Thean, C. Loi, K.S. Ho, P.K. Koh, K.W. Eu, P.Y. Cheah, Genome-wide scan identifies a copy number variable region at 3q26 that regulates PPM1L in APC mutation-negative familial colorectal cancer patients, Genes Chromosom. Cancer 49 (2) (2010) 99–106. [36] S. Goyama, M. Kurokawa, Pathogenetic significance of ecotropic viral integration site-1 in hematological malignancies, Cancer Sci. 100 (6) (2009) 990–995.
Contents lists available at ScienceDirect
Genomics j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / y g e n o
A novel locus for canine osteosarcoma (OSA1) maps to CFA34, the canine orthologue of human 3q26 Jeffrey C. Phillips ⁎, Luis Lembcke, Tamara Chamberlin The University of Tennessee, Small Animal Clinical Sciences, C247 Veterinary Teaching Hospital, Knoxville, TN 37996-4554, USA
a r t i c l e
i n f o
Article history: Received 18 February 2010 Accepted 2 July 2010 Available online 18 July 2010 Keywords: Cancer Osteosarcoma Canine Linkage Locus OSA1
a b s t r a c t Spontaneous tumors in dogs share many of the same features of their human orthologues including clinical behavior, response to treatment, and molecular defects. It is therefore natural to consider the use of dogs and their spontaneous malignancies in the study of complex disease such as cancer. Scottish Deerhounds are a giant breed of dogs that exhibit a high incidence of bone cancer. Our previous work suggested that osteosarcoma within this breed could be explained by the presence of a major gene of dominant effect. Herein, we use a whole genome mapping approach to evaluate a four-generation pedigree of Scottish Deerhounds for linkage of their osteosarcoma phenotype. Using this approach we found evidence of linkage (Zmax = 5.766) between their phenotype and markers located on CFA34, in a region syntenic to human chromosome 3q26. The identification of this locus provides novel insight into the genetic basis of osteosarcoma in both canines and humans. © 2010 Elsevier Inc. All rights reserved.
1. Introduction Osteosarcoma (osteogenic sarcoma, OSA) is the most common (N85%) primary malignant bone tumor in both humans and canines. In the United States, approximately 10,000 canine and 800 human cases of osteosarcoma are diagnosed each year [1–3]. Of the cancers that occur in both pet dogs and humans, osteosarcoma is perhaps the most similar. In both species it is histologically composed of mesenchymal cells that produce osteoid matrix. The dog and human also share similar anatomic predilections including the distal femur, proximal tibia, and proximal humerus amongst others. The clinical behavior in both species is characterized as highly aggressive with very high rates of metastatic spread. In fact, 85–90% of both humans and pet dogs are considered to have microscopic metastatic disease at the time of initial diagnosis [3,4]. Humans and pet dogs also share a similar response to surgery and chemotherapy regimens. While dramatic progress has been made in the treatment and outcome in both species, there is limited understanding of the etiology of the disease. Because of the many similarities shared by humans and pet dogs with osteosarcoma, it is only natural to use dogs and their spontaneous malignancies to more fully understand disease pathogenesis. The genetic changes that occur in osteosarcoma have received considerable attention in recent years in an attempt to elucidate ⁎ Corresponding author. College of Veterinary Medicine, Department of Small Animal Clinical Sciences, 2407 River Drive, C247 Veterinary Teaching Hospital, Knoxville, TN 37996-4544, USA. Fax: + 1 865 974 5554. E-mail address: [email protected] (J.C. Phillips). 0888-7543/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.ygeno.2010.07.002
etiology in both humans and canids. In humans, loss of heterozygosity screening has identified several chromosomal regions with high frequencies of allelic loss including chromosomes 3q, 13q, 17p, and 18q [5–8]. Heritable cancer syndromes, although extremely rare, have been associated with two of these regions; namely chromosomes 13q (Retinoblastoma) and 17p (Li-Fraumeni) [9]. Comparative genomic hybridization has also been used to identify regions of genomic amplification in human-origin tumors, including the most common amplicons at 12q11–q15 and 17p11–13 [10]; while transcriptome profiles of osteosarcoma cell lines has provided information on the molecular variation between tumors [11]. In dogs, gene expression studies have identified profiles associated with both tumor progression [12] and survival [13]. These profiles demonstrate strong similarities between canine and human osteosarcoma along with potential biologic targets in both species. Comparative genomic hybridization of canine tumor tissues has also identified regions of recurrent cytogenetic changes within select breeds and correlated these regions with candidate tumor suppressor or oncogenes [14]. In spite of these advances made in understanding the tumor biology of both human and canine osteosarcoma, it is unclear what precursor (germline or heritable) lesions are critical for tumorigenesis. In the canine population, large and giant breeds of dogs are at increased risk for the development of OSA; together these breeds account for the majority (N80%) of reported cases [4]. The Scottish Deerhound is a giant breed dog first recognized by the American Kennel Club (AKC) in 1886. The overall breed-specific incidence of OSA in the Deerhound has been estimated to be greater than 150 cases per 1,000 dogs (compared to 7 cases per 100,000 dogs in the general dog population) [4,15]. Given the high incidence of OSA within a
J.C. Phillips et al. / Genomics 96 (2010) 220–227
relatively small, closely related population, it is likely that a major genetic modifier (locus) for the disease exists within this breed. In a recent report [16], we described a population of greater than 1,000 Scottish Deerhounds that were ascertained as part of a prospective health study. Statistical analysis of this population supported the presence of a “putative” high-risk allele associated with the development of osteosarcoma. By comparing models of disease transmission, we were able to show that a highly penetrant, dominant model was most probable. The frequency of the high-risk allele in this dominant model was estimated to be 0.128, and only genotype was found to be correlated with disease onset. The artificial selection that has led to this high, breed-specific, incidence of OSA can be further utilized to identify the genetic determinants which regulate disease onset. Specifically, the inbreeding that has led to this high-incidence of disease also serves to simplify its transmission properties. While in humans OSA is clearly a complex disease with no obvious transmission patterns, the canine orthologue in the Scottish Deerhound exhibits a simple dominant transmission. Furthermore, with the completion of the canine genome sequencing project and the availability of high-resolution canine genetic maps, linkage and association studies can be pursued in canine populations such as the Scottish Deerhound. Through these mapping studies candidate loci and genes can be identified and used to further our understanding of not only the canine but also the human disease. In this study, we describe the mapping of the osteosarcoma phenotype (OSA1) that is segregating in a four-generation pedigree of Scottish Deerhounds. A whole genome linkage approach is used to map OSA1 to CFA34, and to exclude other chromosomal regions. The identification of a novel locus for canine osteosarcoma provides important information on the location of candidate genes for both canine and human OSA. 2. Materials and methods
221
tumor location, and treatment option. All procedures involving animals in this study were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Tennessee. 2.2. DNA samples EDTA preserved whole blood samples were collected from 60 dogs in the four generation pedigree. DNA was purified using standard methods. 2.3. Markers An extended microsatellite screening set was used for the whole genome scan. The microsatellite set included markers previously identified in a mapping set referred to as MSS3 [19]; 103 additional markers (610 total) were selected throughout the genome to increase coverage of select regions. Marker positions are as noted in the CanFam 2.0 genome assembly [20]. Average physical spacing of markers in this extended set was 4.1 Mb. Genetic distances between markers were obtained from the Canine Genetic Linkage Map [21] and verified on this sample population using the software SimWalk2 [22]. 2.4. Genotyping Fluorescent-labeled oligonucleotides were used to amplify polymorphic markers used for linkage analysis. Samples were run on a MegaBace® 1500, 96-capillary array (GE Healthcare) and genotypes were analyzed using the Genetic Profiler software (version 2.2; GE Healthcare). All genotypes were manually scored with values converted to simple integers (i.e. 1–9) for statistical analysis. Genotypes for the 75 untyped dogs were scored as unknown.
2.1. Sample population
2.5. Statistical analyses
Dogs used in this study were all privately owned and AKC registered Scottish Deerhounds from a larger population previously described [16]. A four-generation pedigree of Scottish Deerhounds was selected from this larger population and included 135 dogs; 50 of whom were identified as affected with osteosarcoma. DNA samples were available from a total of 60 dogs in this pedigree, with the remaining dogs (i.e. 75 dogs) providing phenotypic information that allowed for the construction of a comprehensive pedigree. Current and historic health information on all dogs was collected from owners, breeders, and veterinarians. Affected status was determined by the presence of characteristic radiographic findings of aggressive bone disease and/or histologic examination. Pedigree information and medical histories were entered into a pedigree management program (Cyrillic 2.1.3, Cherwell Software). The program PEDINFO was used to determine descriptive characteristics of the population. This program is available as a component of the genetic analysis package SAGE 5.2.0 [17]. Inbreeding coefficients were estimated using the Kinship Matrices option of the genetic software MENDEL 6.01 [18]. Health information collected included a variety of qualitative and quantitative traits (i.e. covariates). Qualitative traits included OSA status, presence/absence of other diseases including cancers, and gender. OSA status had three possible outcomes: affected, unaffected, or unknown. Quantitative traits included age at exam, age of onset, height, weight, and length of time sexually intact (as a measure of cumulative sex hormone exposure). Age at exam was considered to be the age at last known contact or age at death where applicable. Age of onset documented the age at which definitive diagnosis of osteosarcoma was made in affected individuals. Time was measured in years, height in inches at the shoulder and weight in pounds. Other information collected (where applicable) included cause of death,
Osteosarcoma was considered a dichotomous trait in which individuals were classified as “unaffected” (at the time of exam), “affected”, or in some cases “unknown status”. Individuals listed with “unknown” phenotypes were included to provide a non-disjoint fourgeneration pedigree; the disease phenotypes for these individuals were ignored for each analysis. Multipoint linkage analysis was performed on a 64 bit (Xeon) PC computer using the SimWalk2 (v2.91) program [22,23]. SimWalk2 uses a Markov chain Monte Carlo (MCMC) algorithm and simulated annealing algorithms to avoid limitations due to sample size or inbreeding. The program was compiled from source code for usage on a 64-bit platform. The osteosarcoma phenotype (OSA1) segregating in this pedigree was modeled as both a dominant and recessive Mendelian trait with allele frequencies of qa = 0.128 and qa = 0.501, respectively [16]. Microsatellite marker allele frequencies were obtained from analysis of a larger population previously described [16]. Recombination frequencies were assumed to be equal between males and females due to software limitations. Based on our previous work, multipoint LOD-score calculations were performed at penetrances of 1.0 and 0.9, under the assumption of locus homogeneity (alpha = 1.0). All other parameters were set to their default values under SimWalk2. LOD scores (Zmax or location scores) of ≥3.0 were considered consistent with linkage. Conversely, LOD scores of ≤−2.0 could be used to exclude regions from consideration. For comparison, a LOD score of N3.0 is associated with a genome-wide significance level of b0.05 in favor of linkage [24] while a LOD score of ≤−2.0 corresponds to 100:1 odds against linkage [25]. Haplotype analysis was used to determine the minimal candidate region for location of the osteosarcoma gene. Haplotypes were estimated using SimWalk2 and illustrated using Cyrillic 2.1.3.
222
J.C. Phillips et al. / Genomics 96 (2010) 220–227
3. Results 3.1. Descriptive analysis Thirteen sibships comprising a total of 135 dogs were selected from a previously described population of 1,272 Scottish Deerhounds [16] (Table 1). These sibships could be arranged into one large four generation pedigree. DNA samples were available from 60 of the 135 dogs for use in a whole genome scan. The remaining 75 untyped dogs provided health information only and were used to construct the comprehensive pedigree. A total of 9 inbreeding loops were identified in this pedigree, with an average inbreeding coefficient of 0.086 estimated using the kinship matrices option of MENDEL 6.01. Due to the graphical complexity, only a small subset of the pedigree is shown in Fig. 1. This pedigree included 68 male dogs and 67 female dogs, ranging in age from 4.0 to 13.6 years. Of these dogs, 50 were confirmed affected, 83 were unaffected at the time of exam, while 2 had unknown health status or were lost to followup. A similar gender distribution was found in the subset of dogs for which DNA samples were available (32 female and 28 male dogs). This subset, however, was further enriched with affected individuals including 40 affected dogs and only 20 dogs who were unaffected at the time of exam. The absolute percentage of affected dogs amongst all dogs with known health status in the pedigree was 38% (50/133), which is higher than the population based prevalence previously reported (i.e. 21%) [16]. Affected dogs ranged in age from 4.9 to 13.6 years with a mean age of onset of 7.6 years (±1.9 yrs). Unaffected dogs were significantly older than affected dogs (8.5 ± 2.9 yrs, p = 0.02). This pedigree of Scottish Deerhounds compares favorably to the previously described larger population [16] with no significant difference in age of onset, gender distribution, or gender-specific prevalence of osteosarcoma (Table 1). Differences in disease prevalence are directly attributable to phenotypic based selection of this pedigree. In other words, this pedigree is
enriched with affected individuals which allows for the use of familial based linkage methodology. Finally, no correlation was found between height, weight, or duration of gonadal hormone exposure and osteosarcoma (data not shown); similar to the larger population [16]. 3.2. Linkage analysis The whole genome linkage data consisted of genotypes from the 60 dogs at the 610 microsatellites identified from the CanFam 2.0 genome assembly. Multipoint linkage analysis was performed using SimWalk2 under both dominant and recessive models. Allele frequencies for the disease allele were set to qa = 0.128 and qa = 0.501, for the dominant and recessive model, respectively (Table 1). These values were based on our previous populationwide segregation analysis of the disease phenotype (OSA1) [16]. The resultant parametric multipoint LOD scores are illustrated in Fig. 2. These scores are based under the assumption of no locus heterogeneity within the pedigree and a fully penetrant trait. Similar results were found with a reduced penetrance of 0.9 (data not shown). The majority of the genome was excluded from linkage on the basis of LOD scores b−3.0 (Fig. 2). The regions with LOD scores of N1 but b3 can neither be definitively excluded from linkage nor are considered to exhibit strong evidence of linkage. Table 2 shows all results reaching a significance threshold of equivalent logarithm of odds (LOD) score N1. Regions from chromosomes 1, 15, 22, and 25 exhibited weak evidence of linkage with LOD scores marginally N1 (Table 2). These regions contained markers with relatively low information content due to high levels of homozygosity. Given the acrocentric nature of canine chromosomes, each of these four regions are localized to the most centromeric portion of their respective chromosomes; regions typically associated with consistently low recombination rates and heterozygosity across breeds [26]. These centromeric regions also exhibited low levels of recombination and
Table 1 (A) Descriptive statistics for 135 dogs in 13 families analyzed in this report (current pedigree) compared to a previous report [16] (mean ± standard deviation); (B) Parameter values used for linkage analysis performed with the SimWalk2 [22] software package. (A) Descriptive statistics
Total individuals Males Females Affected Unaffected %Affected Missing Probands Sibships Sibship size # DNA samples available Inbreeding loops Inbreeding coefficient Age (years) Age of onset (years)
Current pedigree
Phillips et al, 2007 [16]
135 68 67 50 82 38% 2 1 13 1–9 (3.0 ± 2.4) 60 9 0.086 4.0–13.6 (7.7 ± 2.1) 4.9–13.6 (7.6 ± 1.9)
1057 479 578 211 794 21% 52 5 314 1–11 (3.0 ± 1.9) 672 187 0.032 0.5–14.9 (6.8 ± 2.8) 2–13.9 (7.0 ± 2.0)
(B) Parameter values Parameter Allele frequency Dominant Recessive Penetrance Complete Reduced Alphaa Mistyping probability a
Value 0.128 0.501 1.0 0.9 1.0 0.025
The a priori proportion of pedigrees that are segregating an affected gene linked to this set of marker loci. A value of 1.0 assumes genetic homogeneity.
J.C. Phillips et al. / Genomics 96 (2010) 220–227
223
Fig. 1. Deerhound pedigree. Representative subset pedigree of dogs used in this study with known affected status. Standard pedigree symbols are used. Affected individuals are represented by filled symbols. Multiple inbreeding (double lines) and mating loops are present. Not all relationships are shown.
heterozygosity in the population studied here (data not shown); factors which likely contributed to the creation of what are essentially non-informative chromosomal regions and hampered our ability to statistically eliminate them. This is a relatively common finding, to note regions of extended homozygosity; which are often especially marked around centromeres [20,21,26]. A region with moderate evidence of linkage (Zmax = 2.246) was found between the osteosarcoma phenotype (OSA1) and markers located in the telomeric portion of chromosome 18 (Figs. 2 and 3). This region spans 7 cM between markers REN168H09 and FH2429 and is characterized by relatively high levels of recombination and heterozygosity across breeds [20]. In the Scottish Deerhound population, relatively low levels of recombination were found for nearly all of chromosome 18; however, a relative increase in recombination rate was noted in the center of the region of interest (Fig. 3A). This effect led to the creation of a conserved haplotype block seen within the affected dogs of this pedigree, which extends to marker FH2429 of chromosome 18. Significant linkage was found between markers located on chromosome 34 (CFA34) and the osteosarcoma phenotype in this pedigree (OSA1). This 40-cM region extends from markers REN243023 through REN238D23 (Fig. 2). The peak LOD score (Zmax = 5.766) was found at 46.8-cM near marker REN234E12 (Fig. 3B). Using SimWalk2, haplotypes were constructed for chromosome 34 (CFA34); estimated haplotypes from the most informative sibship were then used to identify the critical region for OSA1 (Fig. 4). Affected individuals shared a common haplotype for markers spanning this region. Critical recombination events in affected individuals occurred between markers FH3836 and REN234E12 and REN173D06 and REN44K21, defining a 10-cM critical interval extending from marker FH3836 to REN44K21. Several unaffected individuals demonstrate recombination events that could significantly narrow the critical region. However, because of the reduced penetrance of this condition [16], one or more of these individuals could still have the affected allele. To eliminate the potential ambiguity, the critical region has been defined by recombination events seen in individuals expressing the disease phenotype. This region corresponds to
4.5 Mb on chromosome 34q16.2–q17.1 (CFA34) extending from 34.9 Mb to 39.4 Mb on the CanFam 2.0 genome assembly. 4. Discussion Here we have mapped the major locus for osteosarcoma in this pedigree (i.e. OSA1) to a 10-cM region of chromosome 34. The pedigree used for this linkage study consisted of a large four generation pedigree. DNA samples were available for genotyping from 60 of the 135 individuals in this pedigree. Forty-one of these DNA samples were collected from a highly informative portion of the pedigree which allowed for haplotype estimation (Fig. 4A). The remaining 19 DNA samples were collected from more distantly related, smaller sibships and contained relatively little information content for linkage studies (data not shown). The overall pedigree compares favorably with a larger population of Scottish Deerhounds previously described [16]; although the following key differences should be noted: (1) this pedigree was relatively enriched with affected individuals; and, (2) individuals in this pedigree were on average older than dogs in our previous report. The ascertainment process that led to these differences can bias the detection of covariates that modify disease; in this case age and perhaps other factors. Furthermore, this age difference can impact the perceived penetrance of diseases which exhibit a delayed (i.e. age-related) onset such as osteosarcoma. While no methodology exists for correcting these sources of bias in canine linkage studies, association studies on distantly related and outbred populations can be used to identify covariates that modify disease. Results from the whole genome linkage analysis effectively excluded the majority of genomic regions with the exception of loci on chromosomes 1, 15, 18, 22, 25, and 34. Each of these loci were associated with LOD scores of N1.0. While statistically the regions on chromosomes 1, 15, 22, and 25 could not be excluded; haplotype analysis proximal to these loci identified several affected recombinant individuals (data not shown). Given the low levels of recombination observed in these centromeric regions, they are unlikely to be linked to the osteosarcoma phenotype segregating in this pedigree. On
224
J.C. Phillips et al. / Genomics 96 (2010) 220–227
J.C. Phillips et al. / Genomics 96 (2010) 220–227
225
Table 2 Parametric linkage analysis results (all regions with LOD N 1 under either a recessive or dominant model). Chromosome location
Genetic location, cM (θ)
Closest marker
LOD dominant (θ)
LOD recessive (θ)
1q11 15q11 22q11 25q11 18q25 34q16.2–q17.1
1.0 1.0 1.0 1.0 48.37 46.8
FH3325 FH5157 REN49F22 REN54E19 REN266I17 REN234E12
1.042 −9.436 1.008 −9.087 2.246 5.766
−4.716 1.047 −14.591 1.020 −27.288 −56.810
chromosome 18, a region exhibiting moderate statistical significance (Zmax = 2.246) was identified telomeric to marker REN168H09. Haplotype analysis identified 3 affected recombinant individuals at marker REN47J11 (i.e. proximal to the region) but could not be used to further exclude the region due to the lack of informative markers. Additional informative markers would be helpful to evaluate this region as a covariate affecting osteosarcoma in this pedigree. Statistical evidence for linkage between markers on chromosome 34q16.2–q17.1 and the osteosarcoma phenotype segregating in this pedigree (OSA1) was strong. The most consistent model assumed genetic homogeneity (alpha = 1.0) within the population and a fully penetrant dominant trait. In contrast, results of segregation analysis [16] suggested a reduced penetrance with minimal heterogeneity (alpha = 0.95) could best explain the osteosarcoma phenotype segregating within the Deerhound population. These minor differences are likely attributable to the selection process used for the pedigree studied here. As mentioned previously, this pedigree is enriched with closely related affected individuals and thus relatively homogeneous. Furthermore, for diseases that show an age-related onset it can be difficult to distinguish between non-carrier and nonpenetrant “unaffected” individuals. In this pedigree, the unaffected dogs were significantly older than the unaffected dogs in our previous report. This age difference would be expected to lower the likelihood of misclassification of unaffected individuals and thus result in an apparent increase in disease penetrance as seen herein. Haplotype analysis identified a 10-cM critical region on CFA34 between markers FH3836 and REN44K21. This 4.5 Mb region of canine chromosome 34q16.2–17.1 is syntenic to human chromosome 3q26 [27]. Interestingly, human chromosome 3q26 is associated with a high incidence of loss of heterozygosity in both primary and metastatic osteosarcoma tumor samples from humans [28,29]. Loss of heterozygosity (LOH) is a genetic event that is thought to result in tumorspecific inactivation of suppressor genes and can be detected by comparing polymorphic markers between tumor and normal tissues [30]. Consistent loss of heterozygosity at specific polymorphic markers is generally considered a good predictor of the presence of a tumor suppressor gene located close to these markers. Furthermore, it is well known that most familial cancer syndromes are related to defects in tumor suppressor genes [31]. By using loss of heterozygosity analysis, a 1-cM critical sub region of human chromosome 3q26 between markers D3S1212 and D3S1246 has been identified [28]. This region is syntenic to CFA34 between markers FH3836 and REN234E12; the location of the peak LOD score defined in this pedigree. In human osteosarcomas, other common sites of LOH include chromosomal arms 13q, 17p, and 18q [28,29]. Chromosomes 13q and 17p are known to contain RB1 and TP53, respectively; tumor suppressor genes important for osteosarcoma tumorigenesis [32]. Retinoblastoma and Li-Fraumeni syndrome are familial cancer syndromes associated with defects in these two genes. To date, no
Fig. 3. Significant LOD scores. LOD scores (blue) and recombination rates (red) for (A) chromosome 18, and (B) chromosome 34. Physical length (Mb) is plotted on the x-axis and is scaled to accommodate varying chromosomal size. Recombination rates and LOD scores are plotted on the left and right y-axis, respectively. Recombination rates were determined in sliding 5 Mb windows as previously described [21]. Ideograms of each chromosome (left) showing critical regions indicated with blue arrow.
familial cancer syndromes have been localized to either chromosome 3q or 18q. Several candidate tumor suppressor genes, however, have been localized to chromosome 3q26 [33–36]. One such gene is the recently identified tumor suppressor gene PPM1L [35]. PPM1L encodes a novel serine-threonine phosphatase in the TGF-beta and BMP signaling pathways and is thought to play a role in APC mutationnegative familial colorectal cancer. The canine homologue for PPM1L is localized within the larger candidate region identified on canine chromosome 34. Another interesting candidate located within this region of human chromosome 3q26 is the MECOM gene (EVI1 complex locus). MECOM encodes a nuclear transcription factor, essential for the proliferation/ maintenance of hematopoietic stem cells. Aberrant expression of the MECOM gene has been associated with a variety of human hematopoietic malignancies [36]. Interesting, in the Scottish Deerhound population previously described [16], and the pedigree herein, the probability for the development of a hematopoietic malignancy in the sibling of a dog affected with osteosarcoma is 9% (data not shown). Given the relatively low population prevalence, this finding suggests an extremely high relative risk for hematopoietic malignancies in siblings of affected dogs. The canine homologue of MECOM is localized within the critical region of CFA34 defined herein and is another excellent candidate gene for this heritable cancer syndrome (OSA1). Further evaluation of the Scottish Deerhound for causative mutations associated with the osteosarcoma phenotype and as a model of the orthologous human disease is ongoing. This effort will
Fig. 2. Multipoint LOD score results. Results of parametric whole genome linkage analysis for the OSA1 phenotype. Dominant transmission (blue) and recessive (red) transmission (red) models are shown from centromeric to telomeric end (left to right). Genetic distance (cM) is plotted on the x-axis and is scaled to accommodate the varying chromosomal size, LOD scores are plotted on the y-axis. Assumptions include complete penetrance and genetic homogeneity.
226
J.C. Phillips et al. / Genomics 96 (2010) 220–227
Fig. 4. Haplotype analysis from CFA34. Results are shown centromeric to telomeric end (top to bottom). Affected individuals are shown as blackened circles. The conserved haplotype shared by affected individuals is illustrated in red. Haplotypes from the most informative portion of the pedigree are shown in “A.”
involve genome-wide and regional association studies in other outbred canine populations with the goal of identifying additional covariates affecting penetrance (genome-wide) and further narrowing the critical region of CFA34 (regional SNP evaluation). Acknowledgments Special thanks to Ms. Lyn Robb of Tannochbrae Deerhounds and Ms. Rachel Matthews for their help in collecting health information and blood samples. The author would also like to acknowledge the Canine Hereditary Cancer Consortium along with investigators at the National Cancer Institute for their ongoing efforts to investigate the underlying causes of canine osteosarcoma. This project was funded by the UTCVM Center of Excellence in Livestock Diseases and Animal Health.
References [1] U.S. Cancer Statistics Working Group, United States Cancer Statistics: 1999-2002 Incidence and Mortality Web-based Report Version, Atlanta: Department of Health and Human Services, Centers for Disease Control and Prevention, and National Cancer Institute, 2005. [2] W. Misdorp, A.A. Hart, Some prognostic and epidemiologic factors in canine osteosarcoma, J. Natl Cancer Inst. 62 (3) (1979) 537–545. [3] P.A. Meyers, Gorlick R. Osteosarcoma, Pediatr. Clin. North Am. 44 (1997) 973–989. [4] G. Ru, G. Terracini, L. Glickman, Host Related Risk Factors for Canine Osteosarcoma, Vet. J. 156 (1998) 31–39. [5] I. Pasic, A. Shlien, A. Durbin, et al., Recurrent focal copy-number changes and loss of heterozygosity implicate two noncoding RNAs and one tumor suppressor gene at chromosome 3q13.31 in osteosarcoma, Cancer Res. 70 (1) (2010) 60–71. [6] R.P. Kruzelock, E.C. Murphy, L.C. Strong, S.L. Naylor, M.F. Hansen, Localization of a Novel Tumor Suppressor Locus on Human Chromosome 3q Important in Osteosarcoma Tumorigenesis, Cancer Res. 57 (1997) 106–109. [7] T. Yamaguchi, et al., Allelotype analysis in osteosarcomas: frequent allele loss on 3q, 13q, 17p, and 18q, Cancer Res. 52 (1992) 2419–2423.
J.C. Phillips et al. / Genomics 96 (2010) 220–227 [8] T. Johnson-Pais, M. Nellissery, D. Ammerman, et al., Determination of a minimal region of loss of heterozygosity on chromosome 18q21.33 in osteosarcoma, Int. J. Cancer 105 (2) (2003) 285–288. [9] G. Ottaviani, N. Jaffe, The etiology of osteosarcoma, Cancer Treat. Res. 152 (2010) 15–32. [10] J. Atiye, M. Wolf, S. Kaur, et al., Gene amplifications in osteosarcoma-CGH microarray analysis, Genes Chromosom. Cancer 42 (2) (2005) 158–163. [11] I. Trougakos, N. Chondrogianni, I. Amarantos, et al., Genome-wide transcriptome profile of the human osteosarcoma SA OS and U-2 OS cell lines, Cancer Genet. Cytogenet. 196 (2) (2010) 109–118. [12] M. Paoloni, S. Davis, S. Lana, et al., Canine tumor cross-species genomics uncovers targets linked to osteosarcoma progression, BMC Genomics 10 (2009) 625. [13] G. Selvarajah, J. Kirpensteijn, M. Van Wolferen, et al., Gene expression profiling of canine osteosarcoma reveals genes associated with short and long survival times, Mol. Cancer 8 (2009) 72. [14] R. Thomas, W. Huixia, T. Pei-Chien, et al., Evidence for breed-associated cytogenetic aberrations in canine appendicular osteosarcoma, Chromosome Res. 17 (2009) 365–377. [15] J. Dillberger, Deerhound health survey. Part 1. Lifespan and causes of death, The Claymore, 1996, pp. 36–44, Mar/Apr. [16] J.C. Phillips, B. Stephenson, M. Hauck, J. Dillberger, Heritability and segregation analysis of osteosarcoma in the Scottish Deerhound, Genomics 90 (3) (2007) 354–363. [17] S.A.G.E., Statistical Analysis for Genetic Epidemiology, Computer program package available from Statistical Solutions Ltd, Cork, Ireland, Release 5.2.0, 2006. [18] K. Lange, et al., MENDEL Version 4.0: A Complete Package for the Exact Genetic Analysis of Discrete Traits in Pedigree and Population Data Sets, Am. J. Hum. Genet. 69(Supplement) (2001) A1886. [19] D.R. Sargan, J. Aguirre-Hernandez, F. Galibert, E. Ostrander, An extended microsatellite set for linkage mapping in the domestic dog, J. Hered. 98 (3) (2007) 221–231. [20] K. Lindblad-Toh, et al., Genome sequence, comparative analysis and haplotype structure of the domestic dog, Nature 438 (7069) (2005) 803–819. [21] A.K. Wong, et al., A Comprehensive Linkage Map of the Dog Genetics, Genetics 184 (2010) 595–605. [22] E. Sobel, K. Lange, Descent graphs in pedigree analysis: applications to haplotyping, location scores, and marker sharing statistics, Am. J. Hum. Genet. 58 (1996) 1323–1337.
227
[23] E. Sobel, H. Sengul, D.E. Weeks, Multipoint estimation of identity-by-descent probabilities at arbitrary positions among marker loci on general pedigrees, Hum. Hered. 52 (2001) 121–131. [24] Lue Ping Zhao, Ross Prentice, Fumin Shen, Li. Hsu, On the Assessment of Statistical Significance in Disease-Gene Discovery, Am. J. Hum. Genet., 64, 1999, pp. 1739–1753. [25] H.W. Deng, LOD score exclusion analyses for candidate QTLs using random population samples, Genetica 119 (3) (2003) 303–315. [26] D. Curtis, A.E. Vine, J. Knight, Study of Regions of Extended Homozygosity Provides Powerful Method to Explore Haplotype Structure of Human Populations, Ann. Hum. Genet. 72 (2) (2008) 261–278. [27] C. Hitte, et al., Facilitating genome navigation: survey sequencing and dense radiation-hybrid gene mapping, Nat. Rev. Genet. 6 (8) (2005) 643–648. [28] R.P. Kruzelock, E.C. Murphy, L.C. Strong, S.L. Naylor, M.F. Hansen, Localization of a Novel Tumor Suppressor Locus on Human Chromosome 3q Important in Osteosarcoma Tumorigenesis, Cancer Res. 57 (1997) 106–109. [29] T. Yamaguchi, et al., Allelotype analysis in osteosarcomas: frequent allele loss on 3q, 13q, 17p, and 18q, Cancer Res. 52 (1992) 2419–2423. [30] M.F. Hansen, W.K. Cavenee, Tumor suppressors: recessive mutations that lead to cancer, Cell 53 (1988) 173–174. [31] A.G. Knudson, Genetics of Human Cancer, Ann. Rev. Genet. 20 (1986) 231–251. [32] G. Letson, C. Muro-Cacho, Genetic and Molecular Abnormalities in Tumors of the Bone and Soft Tissues, Cancer Control 8 (3) (2001) 239–251. [33] A.J. Bass, et al., SOX2 is an amplified lineage-survival oncogene in lung and esophageal squamous cell carcinomas, Nat. Genet. 41 (11) (2009) 1238–1242. [34] J. Kang, S.H. Koo, K.C. Kwon, J.W. Park, J.M. Kim, Identification of novel candidate target genes, including PEHB3, MASP1, and SST at 3q26.2-q29 in squamous cell carcinoma of the lung, BMC Cancer 16 (9) (2009) 237–241. [35] L.F. Thean, C. Loi, K.S. Ho, P.K. Koh, K.W. Eu, P.Y. Cheah, Genome-wide scan identifies a copy number variable region at 3q26 that regulates PPM1L in APC mutation-negative familial colorectal cancer patients, Genes Chromosom. Cancer 49 (2) (2010) 99–106. [36] S. Goyama, M. Kurokawa, Pathogenetic significance of ecotropic viral integration site-1 in hematological malignancies, Cancer Sci. 100 (6) (2009) 990–995.