A high-resolution integrated map of copy number polymorphisms within and between breeds of the modern domesticated dog

BMC Genomics A high-resolution integrated map of copy number polymorphisms within and between breeds of the modern domesticated dog Thomas J Nicholas 0 Carl Baker 0 Evan E Eichler 0 1 Joshua M Akey 0 0 Department of Genome Sciences, University of Washington , 1705 NE Pacific, Seattle, WA. 98195 , USA 1 Howard Hughes Medical Institute , Seattle, WA , USA Background: Structural variation contributes to the rich genetic and phenotypic diversity of the modern domestic dog, Canis lupus familiaris, although compared to other organisms, catalogs of canine copy number variants (CNVs) are poorly defined. To this end, we developed a customized high-density tiling array across the canine genome and used it to discover CNVs in nine genetically diverse dogs and a gray wolf. Results: In total, we identified 403 CNVs that overlap 401 genes, which are enriched for defense/immunity, oxidoreductase, protease, receptor, signaling molecule and transporter genes. Furthermore, we performed detailed comparisons between CNVs located within versus outside of segmental duplications (SDs) and find that CNVs in SDs are enriched for gene content and complexity. Finally, we compiled all known dog CNV regions and genotyped them with a custom aCGH chip in 61 dogs from 12 diverse breeds. These data allowed us to perform the first population genetics analysis of canine structural variation and identify CNVs that potentially contribute to breed specific traits. Conclusions: Our comprehensive analysis of canine CNVs will be an important resource in genetically dissecting canine phenotypic and behavioral variation. - Background The domestication of the modern dog from their wolf ancestors has resulted in an extraordinary amount of diversity in canine form and function. As such, dogs are poised to provide unique insights into the genetic architecture of phenotypic variation and the mechanistic basis of strong artificial selection. A number of canine genomics resources have been developed to facilitate genotype-phenotype inferences, including a high-quality whole genome sequence and a dense catalog of SNPs discovered in a wide variety of breeds [1-3]. These genomics resources have been successfully used to identify an increasing number of genes that influence hallmark breed characteristics such as size, coat texture, and skin wrinkling [4-6]. Additionally, SNP data has been used to investigate patterns of genetic variation within and between breeds, establish timing and geography of domestication, examine relatedness among breeds, and identify signatures of artificial selection [4,7-9]. In addition to SNPs, it is important to characterize additional components of canine genomic variation in order to comprehensively assess the genetic basis of phenotypic diversity. For example, structural variation in general, and copy number variants (CNVs) in particular, has emerged as an important source of genetic variation in a wide range of organisms including dogs [10-18]. Duplications and deletions of genomic sequence can have significant impacts on a wide range of phenotypes including breed-defining traits. For example, a duplication of a set of FGF genes in Rhodesian and Thai Ridgebacks leads to the breeds characteristic dorsal hair ridge [19]. Although the FGF duplication provides a vivid example of the phenotypic consequences of structural variation in dogs, it remains unknown whether CNVs are an appreciable source of variation in morphological, behavioral, and physiological traits within and between breeds. Comprehensive discovery of structural variation in a diverse panel of breeds is an important first step in more systemically delimiting the contribution of CNVs to canine phenotypic variation. Previously, we used a customized aCGH chip to identify nearly 700 CNV regions located in segmental duplications (SDs) [17]. However, SDs only cover approximately 5% of the dog genome and thus a large fraction of total genomic space was unexplored. An additional study using a genomewide tiling array from NimbleGen identified approximately 60 CNV regions outside of SDs [10]. However, the low probe density (~1 probe every 5 kb), limited the number and size of CNVs that could be identified. In an effort to more comprehensively interrogate the canine genome for CNVs, we used a high-density (~1 probe every 1 kb) genome-wide tiling array to discover additional CNVs in a panel of nine genetically and phenotypically diverse dogs. In total, we discover over 400 new CNV regions. Moreover, we designed a custom aCGH chip to genotype all known canine CNVs in 61 dogs from 12 diverse breeds, allowing the first population genetics analysis of structural variation in dogs to be performed. The comprehensive CNV resources that we have developed will be important tools in genetically dissecting canine phenotypic variation. Results and Discussion Genome-wide identification of CNVs using a high-density aCGH chip We performed aCGH using a high-density tiling array in nine breeds (Table 1), a gray wolf, and a self-self hybridization. These nine breeds and gray wolf samples were previously studied using a custom array that exclusively targeted regions containing SDs [17]. In all of the aCGH hybridizations we used the same reference sample (a female Boxer distinct from Tasha, the Boxer used for generating the canine reference sequence), which was Table 1 Summary of CNVs identified with the genomewide aCGH chip Number of CNVs Total Gain Loss Average Size (kb) Genes also the reference in our prior SD experiments [17]. The aCGH chip consists of over 2.1 million probes distributed across the genome (not including the uncharacterized chromosome, chrUn) with an average probe density of 1 kb. CNVs were identified using a circular binary segmentation algorithm implemented in the program segMNT, part of NimbleGens NimbleScan software package. These calls were filtered by log2 values and number of probes using an adaptive threshold algorithm where the specific filtering criteria were a function of the size of the CNV (see Methods). We identified 1,008 CNVs in 403 unique CNV regions spanning 30.5 Mb of genomic sequence (Table 1). In the self-self hybridization, no CNVs were called using the same analysis and filters. The average number of CNVs per individual was 101, ranging from 86 (Shetland Sheepdog and Siberian Husky) to 136 (Gray Wolf). The average CNV size was approximately 81 kb (Table 1), and the largest CNV region was located on CFA 34 and spans 3.9 Mb. In total, these 403 CNV regions overlap or contain 401 protein coding genes. After assigning all genes PANTHER Molecular Function terms, we found that the most enriched gene classes are similar to those identified in SDs, namely, defense/immunity, and receptor genes, but also included oxidoreductase, protease, signaling molecule, and transporter genes (Additional file 1). Figure 1 summarizes the location and characteristics of all known dog CNVs derived from this and previous studies (...truncated)