The genomic architecture of segmental duplications and associated copy number variants in dogs

Abstract

Structural variation is an important and abundant source of genetic and phenotypic variation. Here we describe the first systematic and genome-wide analysis of segmental duplications and associated copy number variants (CNVs) in the modern domesticated dog, Canis familiaris, which exhibits considerable morphological, physiological, and behavioral variation. Through computational analyses of the publicly available canine reference sequence, we estimate that segmental duplications comprise approximately 4.21% of the canine genome. Segmental duplications overlap 841 genes and are significantly enriched for specific biological functions such as immunity and defense and KRAB box transcription factors. We designed high-density tiling arrays spanning all predicted segmental duplications and performed aCGH in a panel of 17 breeds and a gray wolf. In total, we identified 3583 CNVs, approximately 68% of which were found in two or more samples that map to 678 unique regions. CNVs span 429 genes that are involved in a wide variety of biological processes such as olfaction, immunity, and gene regulation. Our results provide insight into mechanisms of canine genome evolution and generate a valuable resource for future evolutionary and phenotypic studies.

Figure 1.

The genomic architecture of canine segmental duplications and CNVs. Black lines represent all 38 canine autosomes, the X chromosome, and the uncharacterized chromosome (Un). Duplicated bases predicted by WGAC and WSSD are plotted as orange and blue rectangles below and above, respectively, each chromosome. Over 80% of chrUn contains duplicated bases. Of the autosomes, chromosomes 9, 16, and 26 possess the highest percentages of duplicated bases (over 4% of each chromosome), while chromosomes 12, 30, and 33 show the least amount of duplicated bases (<0.35% of each chromosome). Unique CNV regions (see text) are denoted by red rectangles.

Figure 2.

Validation of duplicons by FISH analysis. (A) Example of an interchromosomal duplication detected with clone CH82-381N09. (B,C) Two representative examples of tandom intrachromosomal duplication detected with clones CH82-381N09 and CH82-331L01, respectively.

Figure 3.

Heatmap representation of CNVs. Each row represents one of the 678 unique CNV regions and columns correspond to dogs. For each CNV region, boxes are colored as black, magenta, and green, depending on whether the individual showed no copy number variation, a loss, or a gain, respectively. CNV regions that show both a loss and a gain within an individual dog (see text) are colored yellow. Horizontal white lines separate CNV calls from single copy control regions, and CNVs that exhibit only losses, gains, or both gains and losses. Within each class, CNV regions are sorted from low to high frequency and from left to right dogs are sorted by decreasing number of CNVs. Dog breeds are abbreviated as follows: Basenji (BAS), Shetland Sheepdog (SHS), German Shepherd (GSH), Siberian Husky (SBH), Wolf (WLF), Labrador Retriever (LBR), Doberman Pinscher (DOB), Standard Poodle (STP), Belgian Shepherd Tervuren (BST), West Highland White Terrier (WST), Yorkshire Terrier (YRK), Boxer (Tasha) (TSH), Afghan Hound (AFH), Golden Retriever (GLR), Rottweiler (ROT), Pug (PUG), Whippet (WHP), German Short Haired Pointer (GSP), and Self-Self hybridization (SLF).

Figure 4.

qPCR of GCKR and PHYH regions. Each plot shows the relative copy number in comparison to the reference (y-axis) for each breed (x-axis). The reference sample was the same Boxer that was used as the reference in the aCGH experiments. Note that because the GCKR and PHYH regions are located in segmental duplications, a gain or loss is not expected to yield a relative copy change of 2 and 0.5, respectively. For example, if the reference sample contains three copies, a gain in the test sample would result in an expected relative copy number of 1.33. Vertical bars delimit 95% confidence intervals based on six independent replicates. False positives (CNVs predicted in the aCGH data but not confirmed by qPCR) are colored in red. Breed abbreviations are described in Figure 3.

Figure 5.

Example of a complex CNV region. Distribution of log₂ probe intensities across a 400-kb region of chromosome 17 that shows substantial variation in breakpoints across individuals and spatial heterogeneity of copy number within individuals. Black, green, and magenta indicate regions called as no copy number variation, gains, and losses, respectively. The gene structure of GCKR is shown, with black lines and red boxes corresponding to introns and exons, respectively. For clarity, additional RefSeq genes are not shown. Regions with significant WGAC or WSSD scores are indicated by gray rectangles. Breeds are abbreviated as described in Figure 3.