Significant gene content variation characterizes the genomes of inbred mouse strains

Abstract

The contribution to genetic diversity of genomic segmental copy number variations (CNVs) is less well understood than that of single-nucleotide polymorphisms (SNPs). While less frequent than SNPs, CNVs have greater potential to affect phenotype. In this study, we have performed the most comprehensive survey to date of CNVs in mice, analyzing the genomes of 42 Mouse Phenome Consortium priority strains. This microarray comparative genomic hybridization (CGH)-based analysis has identified 2094 putative CNVs, with an average of 10 Mb of DNA in 51 CNVs when individual mouse strains were compared to the reference strain C57BL/6J. This amount of variation results in gene content that can differ by hundreds of genes between strains. These genes include members of large families such as the major histocompatibility and pheromone receptor genes, but there are also many singleton genes including genes with expected phenotypic consequences from their deletion or amplification. Using a whole-genome association analysis, we demonstrate that complex multigenic phenotypes, such as food intake, can be associated with specific copy number changes.

Figure 1.

All mouse CNVs. Predicted CNVs for each inbred mouse strain are displayed based on their genomic position. Amplifications are shown in red above the baseline for each strain, deletions are shown in green below. The strains are: (1) 129X1/SvJ; (2) 129S1/SvImJ; (3) AKR/J; (4) A/J; (5) BTBR T+ tf/J; (6) BUB/BnJ; (7) BALB/cJ; (8) C3H/HeJ; (9) C57BLKS/J; (10) C57BL/10J; (11) C57BR/cdJ; (12) C57L/J; (13) C58/J; (14) CAST/EiJ; (15) CBA/J; (16) CE/J; (17) CZECHII/EiJ; (18) DBA/1J; (19) DBA/2J; (20) FVB/Ntac; (21) I/LnJ; (22) JF1/Ms; (23) KK/HIJ; (24) LP/J; (25) MA/MyJ; (26) MOLF/EiJ; (27) MSM/Ms; (28) NOD/LtJ; (29) NON/LtJ; (30) NZB/BINJ; (31) NZW/LacJ; (32) PERA/EiJ; (33) PL/J; (34) PWK/PhJ; (35) RIIIS/J; (36) SEA/GnJ; (37) SJL/J; (38) SM/J; (39) SPRET/EiJ; (40) SWR/J; (41) WSB/EiJ.

Figure 2.

Distribution of CNV lengths and amplitudes. (A) A histogram of the density of the log₁₀ lengths in base pairs for all amplification (gray) and deletion (blue) CNVs is shown. (B) A histogram of the density of the log₂ amplitudes of all CNVs in the data set is plotted along with a smoothed curve fit to the histogram. Parenthesized fractions on the X-axis show the positions of expected ratios of test to reference strain copy numbers. Inset is a plot of the observed peaks of the smoothed histogram curve versus the expected copy number ratio positions and a fitted linear regression line. The truncated bar at log₂(amplitude) = −4.32 has a height of 3.7.

Figure 3.

Deletion excess does not correlate with SNP content. A bar plot shows the fraction of all CNVs per strain which are deletions by number (black bars) or total length (white bars), sorted by decreasing deletion number fraction. The line plot shows the fraction of known SNPs between each strain and the reference, which vary.

Figure 4.

QPCR analysis provides validation for CGH-derived CNV data. QPCR was performed on 42 strains, generating eight different PCR products spanning three genomic regions. The mean of the log₂-transformed ratio between the QPCR signal for each strain and the C57BL/6J QPCR signal is plotted against the log₂-transformed median fold-change for all the probes within the corresponding CNVs. QPCR was performed for loci on chromosomes 7, 9, and 17, plotted as red circles, blue crosses, and green triangles, respectively. Error bars show ±1 SD for the QPCR data. The expected correlation of the two data sets is indicated by the solid line with slope = 1.0; the observed correlation, slope = 0.86, is indicated by the dashed line.

Figure 5.

Deletions in five strains remove a highly conserved element on chromosome 6. (A) The log₂ values of CGH probe fold-change values are shown as points along with their running mean (window size = 3) as an orange line for each of five strains. Extreme fold-changes are truncated at ±20 (log₂ of ±4.2). The locations of the calculated deletions for each strain are shown as green bars. The bar graph at the bottom of the plot shows the conservation score for this genomic locus. (B) A detailed view of the chromosome 6 highly conserved element deleted in five mouse strains shows extensive cross-species homology as plotted by the UCSC genome browser (Kent et al. 2002). The conservation track, based on scores from the phastCons algorithm (Siepel et al. 2005), along with individual species alignments using the MultiZ algorithm (Blanchette et al. 2004) are shown. Known genes, miRNA, mouse EST, and RepeatMasker tracks are also displayed, but none are present in this region.

Figure 6.

Whole genome CNV association with food intake. (A) Association scores are plotted for all tested CNV loci. CNV loci are plotted in genomic order along the X-axis with the height of each bar representing the z-score-transformed association score for that locus. Horizontal lines indicate z-score percentiles from randomized data. (B) Daily food intake shows a relationship to genomic amplification levels. Food intake from the Seburn1 MPD data set (Seburn 2001) is measured in gram of food per 30-g body weight. Genomic amplification levels are measured as median fold-change of probe intensities compared to C57BL/6J values for probes on chromosome 17 between positions 30,627,006 and 30,650,272.