A large-scale zebrafish gene knockout resource for the genome-wide study of gene function

Abstract

With the completion of the zebrafish genome sequencing project, it becomes possible to analyze the function of zebrafish genes in a systematic way. The first step in such an analysis is to inactivate each protein-coding gene by targeted or random mutation. Here we describe a streamlined pipeline using proviral insertions coupled with high-throughput sequencing and mapping technologies to widely mutagenize genes in the zebrafish genome. We also report the first 6144 mutagenized and archived F1's predicted to carry up to 3776 mutations in annotated genes. Using in vitro fertilization, we have rescued and characterized ~0.5% of the predicted mutations, showing mutation efficacy and a variety of phenotypes relevant to both developmental processes and human genetic diseases. Mutagenized fish lines are being made freely available to the public through the Zebrafish International Resource Center. These fish lines establish an important milestone for zebrafish genetics research and should greatly facilitate systematic functional studies of the vertebrate genome.

Figure 1.

Overview of the retroviral mutagenesis pipeline. The pseudotyped murine leukemia virus (A) is injected into 1000–2000 cell stage blastula embryos (B). The infection rate is determined by quantitative PCR (qPCR), and founder fish with high infection rates are raised to adults. The founders are crossed to wild-type (T/AB) fish, and F₁ male fish are used for sperm cryopreservation and fin biopsies. Integrations are amplified and mapped from gDNA isolated from the fin biopsies. Mapped integrations are assigned to the corresponding sperm samples, and desired mutations are recovered by in vitro fertilization.

Figure 2.

Overview of high-throughput strategy to identify retroviral integrations using a next-generation sequencing platform. Genomic DNAs corresponding to individual F₁ fish were digested with three sets of restriction enzymes in parallel. After heat-inactivation of the restriction enzymes, the digested samples were then pooled together and ligated with DNA linkers, each containing a unique 6-bp barcode that indexes the F₁ fish. The linker ligated DNA fragments were amplified by linker-mediated PCR using linker and viral LTR specific primers to amplify the adjacent genomic DNA sequences. The LTR/gDNA/linker amplicons are subsequently ligated to Illumina paired-end adapters and sequenced using the Illumina sequencing platform.

Figure 3.

Strategies for mapping retroviral integrations. Paired-end sequencing was performed to capture the site of the retroviral integration (designated by LTR–retroviral 3′ long terminal repeat) and the linker cassette (LC) that contains the “barcode” identifier for the specific sample. Two strategies were used to map the integrations as this proved to be less error-prone than either strategy alone. In Strategy A, pairwise alignment of paired-end reads was performed to create contigs, and the resulting contigs were mapped to the zebrafish genome. Only contigs that mapped unambiguously were considered for identifying integrations. In Strategy B, each read from corresponding paired-ends was mapped independently, and colocalization in the correct orientation (pointing at each other) was used as the criterion for correct mapping. Integrations that mapped to the same genomic coordinates by both strategies were used for identification of integration events.

Figure 4.

Summary of proviral integrations from 6144 F₁ fish. (A) Distribution of the 6933 retroviral integrations in introns; 40% of integrations (2813/6933) are in the first intron. (B) Distribution of 963 integrations in exons. (C) Number of hits per gene based on integrations with unique genomic coordinates; 72% of genes have only one integration. (D) Distribution of 15,223 integrations across all chromosomes.

Figure 5.

Representative embryonic, larval, and adult phenotypes from selected retroviral insertional alleles. (A) Insertion in the wee1 gene led to an early phenotype of cellular necrosis starting at the gastrulation stage. Images here show a wild-type and mutant embryo at the 12 somite stage. (B) Insertion in the eif3i gene led to a vascular defect in homozygous mutants. The upper panel shows bright field images and the lower panel shows the lack of intersegmental vessels labeled by the flk-gfp transgenic marker in the eif3s3^−/− background at 1 day post-fertilization (dpf). (C) An insertion in the snapc1b gene causes embryonic phenotypes including jaw defects and a small liver visible at 5 dpf. Arrows point to the reduced jaw structures in the mutant, dashed lines demarcate the liver. (D) Homozygous rpa1 mutants at 2 dpf have small and necrotic heads, small eyes, and tails curling dorsally. These homozygous phenotypes are weaker but observable at 1 dpf. All homozygotes die at ∼5 dpf. (E) Insertion in a novel gene (zgc:194470) led to the larval phenotype of a larger body at day 12 of development. The mutant is homozygous viable, and the body sizes become the same as that of wild type when reaching the adult stage. One-hundred percent of the homozygous mutants show the larger larval phenotype (N = 200). (F) Slc7a5^−/− fish showed no observable embryonic defects, but they are 40% smaller than their wild-type or heterozygous siblings at 4 mo of age. Slc7a5 is a small subunit of the L-type amino acid transporter 1. (G) 6-mo-old adult tg^−/− (thyroglobulin) fish showed red swelling under the chins (black arrows), a phenotype reminiscent of human thyroid goiters. Tg^−/− fish are fertile and showed no observable embryonic defects.