The million mutation project: a new approach to genetics in Caenorhabditis elegans

Abstract

We have created a library of 2007 mutagenized Caenorhabditis elegans strains, each sequenced to a target depth of 15-fold coverage, to provide the research community with mutant alleles for each of the worm's more than 20,000 genes. The library contains over 800,000 unique single nucleotide variants (SNVs) with an average of eight nonsynonymous changes per gene and more than 16,000 insertion/deletion (indel) and copy number changes, providing an unprecedented genetic resource for this multicellular organism. To supplement this collection, we also sequenced 40 wild isolates, identifying more than 630,000 unique SNVs and 220,000 indels. Comparison of the two sets demonstrates that the mutant collection has a much richer array of both nonsense and missense mutations than the wild isolate set. We also find a wide range of rDNA and telomere repeat copy number in both sets. Scanning the mutant collection for molecular phenotypes reveals a nonsense suppressor as well as strains with higher levels of indels that harbor mutations in DNA repair genes and strains with abundant males associated with him mutations. All the strains are available through the Caenorhabditis Genetics Center and all the sequence changes have been deposited in WormBase and are available through an interactive website.

Figure 1.

Experimental design. Standard mutagenesis protocols were modified by (1) selecting for unc-22/+ animals (twitching in 1% nicotine) in the F1 generation to ensure that the gametes had been effectively exposed to the mutagen; (2) counter-selecting in the F2 generation to generate an unc-22(+) background for the mutant strains; and (3) clonally propagating independent lines for 10 generations, driving strains to homozygosity to simplify variant calling and to create uniform, stable strains for further manipulations.

Figure 2.

Distribution of mutations along chromosomes. (A) SNV density is plotted across chromosomes I and X for the aggregate variants for mutant strains (Mu) and the wild isolates (WI). The SNV density is essentially uniform along the length of both the autosome and X chromosome in the mutant strains. In contrast, for the wild isolates SNV density on chromosome I is much higher on the arms, where recombination is high and gene density is in general lower than in the centers, where recombination is low and gene density is higher. SNV density along the wild isolates' X chromosome is more uniform. (B) Similar density plots for all short indels and (C) only those outside of homopolymer runs. There is enrichment for short indels on the autosomal arms of both the WI and mutant strains, although this is much more pronounced for the WI. However, the distribution in the mutant strains is much more uniform when only considering the small indel events that do not involve homopolymer runs. The bias toward higher density on the autosomal arms remains for the WI.

Figure 3.

Large duplication and deletion events. (A) A large duplication of the right half of chromosome II and a deletion within that region (red circles) were detected by increased (duplication) or decreased (deletion) read depth (see Methods for details). Dashed vertical lines demark the change from normal copy number. The boundaries of the arms are indicated by the asterisks along the x-axis. (B) The fraction of reads containing a variant base is plotted across chromosome II for the same strain. Although on the left half of the chromosome, SNVs show essentially 100% of the reads with the variant base, in the region of the duplication the fraction drops <80% for most sites, reflecting the heterogeneity of the sites. The underlying event may represent a translocation event (the right end of chromosome I in this strain is also duplicated) maintained because it provides a function lost from a lethal mutation on the normal diploid chromosome. (C) Distribution of duplication (blue) and deletion (red) events on chromosome II in the 2007 mutant strains. Each segment represents a distinct event present in one or more of the strains. The duplication events predominate and both duplications and deletions are broadly distributed across the chromosomes (see Supplemental Fig. 8 for similar plots for all chromosomes).

Figure 4.

Mutation effects (SNVs) in mutant strains and wild isolates. The inferred effects of the SNVs in the mutant strains and wild isolates are plotted, showing the disparity in the fraction of mutations affecting coding sequence. Mutations resulting in nonsynonymous, nonsense, and splice site changes are three to 10-fold less frequent in the wild isolates, whereas synonymous changes are similar in number between the two. In addition, the number of mutations in intronic regions is similar between mutant and WI collections, but both UTR and intergenic sites are less abundant in the WI strains.

Figure 5.

Relative fractions of SNVs in different annotated features in the mutants and wild isolates. (A) The fraction of events observed in different annotated features compared with expected is shown for the mutant collection and the wild isolates. To facilitate comparison, the events have been normalized to synonymous mutations in both sets. In the mutant strains only the nonsense mutants are appreciably depressed as a fraction of expected. In contrast, in the wild isolates, nonsense SNVs are even more severely depressed and missense, splice junction, and to some extent even intergenic SNVs are depressed relative to the expected. (B) The fraction of events in different annotated features for a set of essential and nonessential genes is shown for the mutant set and wild isolates. Genes with deletion alleles produced by the The C. elegans Deletion Mutant Consortium (The C. elegans Deletion Mutant Consortium 2012) were divided into essential (lethal) and nonessential (viable) groups. The fractions of events in essential and nonessential genes within the mutant set are quite similar for all features but nonsense SNVs. The same is true for the wild isolates, but as expected from A the overall proportions are reduced.

Figure 6.

rDNA copy number varies between strains. The number of copies of the 18S–28S repeat and the 5S repeat is illustrated as estimated by the fraction of reads mapping to the regions. Ninety-two percent of the strains have between 55 and 130 copies of the 18S–28S repeat and 93% have between 130 and 210 copies of the 5S repeat, with extreme outliers for both regions (33–245 and 39–438, respectively).

Figure 7.

Deletions and duplications in the mitochondrial genome. The coverage of the mitochondrial genome in five mutant strains is shown in a UCSC Genome Browser display, along with the gene content below. Although there are hundreds of copies of the mitochondrial genome in the zygote, both deletions (red bars) and duplications (green bars) were detected in the mutant strains. None of these were fixed in the strains, with both normal and altered length molecules detected in all cases.