A high-resolution map of segmental DNA copy number variation in the mouse genome

Abstract

Submicroscopic (less than 2 Mb) segmental DNA copy number changes are a recently recognized source of genetic variability between individuals. The biological consequences of copy number variants (CNVs) are largely undefined. In some cases, CNVs that cause gene dosage effects have been implicated in phenotypic variation. CNVs have been detected in diverse species, including mice and humans. Published studies in mice have been limited by resolution and strain selection. We chose to study 21 well-characterized inbred mouse strains that are the focus of an international effort to measure, catalog, and disseminate phenotype data. We performed comparative genomic hybridization using long oligomer arrays to characterize CNVs in these strains. This technique increased the resolution of CNV detection by more than an order of magnitude over previous methodologies. The CNVs range in size from 21 to 2,002 kb. Clustering strains by CNV profile recapitulates aspects of the known ancestry of these strains. Most of the CNVs (77.5%) contain annotated genes, and many (47.5%) colocalize with previously mapped segmental duplications in the mouse genome. We demonstrate that this technique can identify copy number differences associated with known polymorphic traits. The phenotype of previously uncharacterized strains can be predicted based on their copy number at these loci. Annotation of CNVs in the mouse genome combined with sequence-based analysis provides an important resource that will help define the genetic basis of complex traits.

Figure 1. Representative Germline CNVs in Mice Identified by High-Resolution aCGH

The log₂ ratios of signal intensity for C57BL/6J (reference) versus 20 test strains are shown. Inset, an expanded view of the CNVs in NOD/LtJ and A/J from (A) and (B). Scale, 500 kb. (A) A 135.6-kb segment of reduced copy number (mean log₂ = −1.02) on Chromosome 14 is present in most strains. (B) A 61.7-kb amplified segment (mean log₂ = +1.01) on Chromosome 1 is present in most strains.

Figure 2. Genome-wide Distribution of CNVs

The ideograms depict chromosomal locations of copy number gains (green arrows), losses (red arrows), and gains or losses (blue arrows) relative to C57BL/6J in autosomes from 20 inbred strains.

Figure 3. Relationship between Signal Intensity and Probe Uniqueness

(A) Absolute signal intensity of the reference strain increases as a function of BHC (p < 10⁻¹⁵); 97.9% of autosomal probe sequences are present at single copy in the C57BL/6J genome. (B) Probes falling in “high confidence” CNVs are unique (1,005 of 1,313 probes with BHC = 1). (C) Probes falling in high signal CNVs are duplicated (356 of 420 probes with BHC > 1).

Figure 4. Validation of Copy Number Changes Identified by aCGH

(A) Log₂ ratio plot demonstrates a 109.2-kb segment of copy number loss on Chromosome 6 in C57L/J, compared to C57BL/6J. (B) qPCR using a primer/probe set in the altered region demonstrates normal copy number (normalized to a relative copy number of one in C57BL/6J) in unaffected strains and significantly reduced copy number in four affected strains (inset, zoom-in view of y-axis). (C) qPCR fails to generate an amplicon of expected size in the altered region from affected strains. B6, C57BL/6J reference strain. (D) Log₂ ratio plot demonstrates a 473.7-kb segment of copy number gain on Chromosome 17 in BALB/cByJ compared to C57BL/6J. (E) qPCR demonstrates heterogeneity of copy number (normalized to a relative copy number of one in C57BL/6J) in this region among 20 strains. (F) Copy number estimates from aCGH and qPCR are highly concordant (p < 0.0001).

Figure 5. Heatmap Representation of Copy Number Changes in Mice

Unsupervised clustering of segmental gains (green) and losses (red) yields a dendrogram that recapitulates features of the known genealogy of these strains. Clustering in the vertical axis demonstrates three clusters: segments amplified in most strains, segments reduced in most strains, and a third cluster containing either singleton CNVs or mixtures of amplifications and deletions.

Figure 6. Relationship between Genomic Distance and Overlap between Segmental Duplications and CNVs

The number of CNVs that overlap at least one segmental duplication was calculated for a range of margin sizes. At a margin size of zero (complete overlap with CNV), 38 of 80 observed CNVs overlap segmental duplications. The extent of overlap between CNVs and segmental duplications (black solid line) increases with margin size. The red dotted line (expected CNVs) indicates the colocalization of segmental duplications with randomly permuted genomic regions of lengths equal to the observed CNVs. Each point of the permuted data was calculated by determining the 95th percentile of the overlap counts. The association between CNVs and segmental duplications remains significant to the 2-Mb window size (p < 0.01) and is highlighted in the yellow rectangle.