Unlocking the bottleneck in forward genetics using whole-genome sequencing and identity by descent to isolate causative mutations

Abstract

Forward genetics screens with N-ethyl-N-nitrosourea (ENU) provide a powerful way to illuminate gene function and generate mouse models of human disease; however, the identification of causative mutations remains a limiting step. Current strategies depend on conventional mapping, so the propagation of affected mice requires non-lethal screens; accurate tracking of phenotypes through pedigrees is complex and uncertain; out-crossing can introduce unexpected modifiers; and Sanger sequencing of candidate genes is inefficient. Here we show how these problems can be efficiently overcome using whole-genome sequencing (WGS) to detect the ENU mutations and then identify regions that are identical by descent (IBD) in multiple affected mice. In this strategy, we use a modification of the Lander-Green algorithm to isolate causative recessive and dominant mutations, even at low coverage, on a pure strain background. Analysis of the IBD regions also allows us to calculate the ENU mutation rate (1.54 mutations per Mb) and to model future strategies for genetic screens in mice. The introduction of this approach will accelerate the discovery of causal variants, permit broader and more informative lethal screens to be used, reduce animal costs, and herald a new era for ENU mutagenesis.

Figure 1. Whole-Genome Sequencing Identifies the IBD Homozygous Region and Causative ENU Mutation.

(A) The structure of an ENU pedigree: two ENU treated males paired with WT B6 females generate founder G1 mice for the ENU16CH17a pedigree, and G3 mice exhibiting the phenotype are selected for WGS. Thus mice within the pedigree carry 4 possible haplotypes, ENU1, ENU2, WT1 and WT2. A yellow star illustrates the segregation of a causative variant. (B) Mice homozygous for the mutation exhibit B cell lymphopenia (here gating on blood lymphocytes). (C) Plots of homozygous filtered variants show the haplotype blocks across the chromosomes of each sequenced mouse. (D) Shared homozygous variants seen in all 3 sequenced mice cluster in an IBD region on Chromosome 4, containing exonic mutations in two genes, Lyn and Tlr4. (E) Confirmation of the Lyn A to G transition by Sanger sequencing. (F) The mutation lies in exon 12 within the catalytic domain.

Figure 2. Identification of IBD Regions using a Modified Lander-Green Algorithm.

(A) Graphical representation of the output of the algorithm, showing the genotypes for the 3 mice, based on combinations of the 4 haplotypes ENU1, ENU2, WT1 and WT2 inherited from the founder mice. WT1 and WT2 are genetically indistinguishable. Each mouse is represented by a vertical third of the plot for each chromosome, and color blocks represent unphased haplotype combinations for each mouse as indicated in the figure. ENU/ENU indicates homozygous ENU regions and ENU/WT indicates heterozygous regions for ENU 1 or ENU 2. (B) Graphical representation of the chromosomal IBD regions, showing shared heterozygous (blue) and homozygous (red) IBD regions. Regions are only IBD if all mice share alleles from a particular ENU founder, ENU1 or ENU2. Non homozygous IBD regions in which all mice carry at least one matching ENU allele are considered IBD heterozygous.

Figure 3. Characterization of the ENU Mutations.

(A) The effect of expanding or contracting the homozygous ENU regions on the estimate of the ENU mutation frequency. (B) The effect of simulated depth of coverage on the estimated ENU mutation frequency. (C) Transition transversion ratio in homozygous ENU variants compared to a large dataset of non-mutagen induced laboratory mouse variation from the Centre for Genome Dynamics Mouse SNP Database. (D) The distribution of ENU mutations, showing reference base pairs and substitutions (ref-sub). (E) The proportion of homozygous mutations that occur at AT sites in homozygous ENU and homozygous WT regions. In each graph, columns or points show mean values across the 3 sequenced ENU16CH17a mice.

Figure 4. The Effect of Reduced Coverage on the Assignment of Regions and Variants by IBD.

(A) The proportion of ENU homozygous and heterozygous IBD regions from the full 24× coverage dataset identified at simulated lower depths of coverage per mouse. (B) The proportion of homozygous and heterozygous IBD variants from the 24× coverage dataset identified at simulated lower depths of coverage per mouse. The validated variants are the coding or splice variants confirmed by Sanger sequencing (Table S1). (C) The number of IBD homozygous SNPs at different simulated coverage depths compared to the number of shared homozygous SNPs across all 3 mice. (D) The number of IBD heterozygous SNPs at different simulated coverage depths compared to the number of shared heterozygous SNPs across all 3 mice.

Figure 5. Modeling the Frequency of Mutants and the Power to Assign Causation by WGS.

(A) The frequency distribution for all mutations within a pedigree at the G3 level, based on a model pedigree of 48 G3 arising from 4 G2 pairs (Materials and Methods). In the specific case of mutations causing fully penetrant phenotypes, the histograms show the distribution of affected mice with recessive (2 allele) and dominant (1 allele) traits. (B) The number of IBD candidate mutations, defined as missense, stop or splice-variants, as a function of the number of sequenced affected G3 mice, based on our model.