Efficient target-selected mutagenesis in Caenorhabditis elegans: toward a knockout for every gene

Abstract

Reverse genetic or gene-driven knockout approaches have contributed significantly to the success of model organisms for fundamental and biomedical research. Although various technologies are available for C. elegans, none of them scale very well for genome-wide application. To address this, we implemented a target-selected knockout approach that is based on random chemical mutagenesis and detection of single nucleotide mutations in genes of interest using high-throughput resequencing. A clonal library of 6144 EMS-mutagenized worms was established and screened, resulting in the identification of 1044 induced mutations in 109 Mbp, which translates into an average spacing between exonic mutations in the library of only 17 bp. We covered 25% of the open reading frames of 32 genes and identified one or more inactivating mutations (nonsense or splice site) in 84% of them. Extrapolation of our results indicates that nonsense mutations for >90% of all C. elegans genes are present in the library. To identify all of these mutations, one only needs to inspect those positions that--given the known specificity of the mutagen--can result in the introduction of a stop codon. We define these positions as nonsense introducing mutations (NIMs). The genome-wide collection of possible NIMs can be calculated for any organism with a sequenced genome and reduces the screening complexity by 200- to 2000-fold, depending on the organism and mutagen. For EMS-mutagenized C. elegans, there are only approximately 500,000 NIMs. We show that a NIM genotyping approach employing high-density microarrays can, in principle, be used for the genome-wide identification of C. elegans knockouts.

Figure 1.

Schematic outline of the mutant C.elegans library generation and screening. (A) Library generation: P0 worms were treated with EMS and F₁ progeny were singled out and expanded clonally until the F₂ to F₃ generation. Cultures were split in two, where the first half was used for cryopreservation in duplicate and the other half for DNA isolation and subsequent mutation detection in genes of interest. (B) Library screening: Genic regions of interest were amplified by a nested PCR and screened for induced mutations by dideoxy resequencing. Mutations were scored by PolyPhred 5.0 (Nickerson et al. 1997) and stored and annotated in LIMSTILL (http://limstill.niob.knaw.nl; V. Guryev and E. Cullen, unpubl.). When an interesting mutation was identified, it was reconfirmed in an independent assay and the mutant was retrieved from the frozen archive, outcrossed, and analyzed for phenotypic consequences. (C) Nested PCR setup: In the first PCR multiple regions of interest are amplified in a multiplex of various oligo 1 and 4 combinations. The second PCR was performed with a single pair of M13-tailed oligos 2 and 3. The resulting products were sequenced using universal M13F and/or M13R oligonucleotides.

Figure 2.

Schematic overview of the screening results for C01G5.2. The coding exons are shown in red and the five amplicons used to screen the majority of the gene are shown in black (first PCR product) and gray (second PCR product). All mutations (118 in total) are annotated on the bottom: intronic (8, black); silent (39, green); missense conserved (34, blue); missense nonconserved (33, pink); stop (3, red); and splice site (1, orange).

Figure 3.

Sequence context bias around the mutated G position. The ratio was calculated between the observed nucleotide occurrence around the G/C to A/T mutations in our screen (n = 973) and the expected frequency for a random G/C base pair in the complete coding part of the 32 genes that were screened (n = 11,740). Sequence data was analyzed with the G base centered on the 0 position of the positive DNA strand. *P < 0.001, ^#P < 0.05.

Figure 4.

Nonsense introducing mutations (NIMs) and genome-wide NIM-based knockout screening. (A) A NIM is defined by those positions in the genome that introduce a premature stop codon in the open reading frame of a gene upon mutation. In this example, only a small fraction of the positions that can be mutated by EMS results in the introduction of a stop codon (boxed positions). (B) Success rate for genome-wide knockout retrieval. The percentage of all C. elegans genes with an expected nonsense mutation is plotted as a function of the library size screened. Rates are plotted for all NIMs (all), the first 25%, 50%, and 75% of the NIMs per gene, and the first maximum of 22 NIMs per gene (first 22). Only G/C-to-A/T mutations (mutation frequency one per 46,000 GC-base) are taken into account for this calculation. The vertical line indicates the size of the current library.

Figure 5.

Array-based NIM analysis. The signal of the nonsense (vertical) versus the wild-type (horizontal) allele is plotted for wild-type worms (N2) (A) and pk2351, containing an EMS-induced stopcodon in ZK858.1 (B). For nonmutated positions, the nonsense signal (vertical axis) is expected to be background, whereas the wild-type signal spreads over the X-axis, most likely reflecting GC content differences. A heterozygous mutation would have similar intensities for both the nonsense and wild-type allele, and these are thus expected to be located close to the diagonal in the figure. Microarrays with 390,000 features were used to assay part of the C. elegans genome-wide NIM collection. Every NIM was represented by four probes that were temperature-normalized (T_m = 72°C, probe length between 25 and 40 nt), representing the wild-type and mutant alleles for both DNA strands. The stop codon in ZK858.1 was represented 10 times (red, plus strand probes; blue, minus strand probes).