Target-Selected Inactivation of the Zebrafish rag1 Gene

See allHide authors and affiliations

Science  05 Jul 2002:
Vol. 297, Issue 5578, pp. 99-102
DOI: 10.1126/science.1071762


The zebrafish has become a favorite organism for genetic analysis of vertebrate development, but methods for generating mutants by reverse genetic approaches have been lacking. We report a method to obtain stable mutants of a gene based on knowledge of the gene sequence only. Parental fish were mutagenized withN-ethyl-N-nitrosourea; in 2679 F1fish, the rag1 gene was analyzed for heterozygous mutations by resequencing. In total, we found 15 mutations: 9 resulted in amino acid substitutions and 1 resulted in a premature stop codon. This truncation mutant was found to be homozygous viable and defective in V(D)J joining. Although presumably immune deficient, these homozygous rag1 mutant fish are able to reach adulthood and are fertile. As sperm samples from all 2679 F1 fish were collected and cryopreserved, we have in principle generated a mutant library from which mutants of most zebrafish genes can be isolated.

Forward genetic screens in the zebrafish have identified thousands of mutants defective in many kinds of biological processes (1, 2). Cloning of the affected genes is accelerated by both an extensive and rapidly growing genetic linkage map (3–5) and the zebrafish genome sequencing project. However, until now a major drawback in zebrafish genetic analysis has been the inability to specifically study gene function through reverse genetics. It has been demonstrated that protein levels in embryos can be temporally reduced by antisense morpholino oligonucleotides (6), but this effect is transient and is not applicable to later stages of development.

In the mouse, reverse genetics was made possible by homologous recombination in embryonic stem cells, eventually giving rise to germ line transmission of the mutant alleles (7,8). Recently, it has been shown that in zebrafish short-term embryonic stem cell cultures are able to produce germ line chimeras (9), but this has not yet been extended to targeted gene inactivation. Therefore, we have taken another approach: target-selected mutagenesis. This involves random mutagenesis, followed by targeted screening for induced mutations at the genomic DNA level. For example, in Caenorhabditis elegans, we and others (10, 11) have successfully performed ethylmethane sulfonate mutagenesis and have screened animals with a polymerase chain reaction (PCR) strategy for deletions in hundreds of target genes. In the case of zebrafish, we have now combinedN-ethyl-N-nitrosourea (ENU) mutagenesis with high-throughput resequencing of the target gene.

Random ENU mutagenesis in zebrafish is widely used in forward screens. Nonmosaic mutants are typically found at tester loci at an average frequency of 1 in 650 mutagenized genomes. Recessive embryonic lethal phenotypes occur at a similarly high rate, namely, an average loss-of-function frequency per locus of 1 in 1600 mutagenized genomes (1, 2). Therefore, we constructed a library of 2679 randomly ENU mutagenized F1 males as outlined in Fig. 1, making use of the Tübingen 2000 screen fish. Genomic DNA was isolated, and testis samples were cryopreserved (12), generating a permanent library that could be screened for heterozygous mutations. In principle, we have isolated sufficient DNA samples to screen for mutations in most zebrafish genes, and the library is comprehensive enough that most genes should be represented by at least one null allele.

Figure 1

Overview of target-selected mutagenesis in zebrafish. Ninety-nine adult male zebrafish were mutagenized by three to five consecutive treatments with 3 mM ENU, in accordance with (32). The mutagenized fish were crossed with wild-type females to give a nonmosaic F1 generation of fish. Sperm was isolated and cryopreserved from 2679 fertile F1 males. Genomic DNA was isolated, arrayed in PCR plates, and screened for mutations by nested PCR amplification of the target gene and subsequent DNA sequence analysis. After a particular mutation was identified, in vitro fertilizations (IVF) were performed to recover the F2 line carrying the mutation (12). Finally, mutations can be bred to homozygosity and analyzed for phenotypes.

To test this approach, we screened for mutations in therag1 gene (13) of each individual F1fish by DNA resequencing. By nested PCR, we amplified two fragments that included exons 2 and 3 of the coding region (Fig. 2A). Subsequent sequence analysis with five different primers (Fig. 2A; Table 1) was done on a 96-capillary ABI3700 DNA analyzer (12). Heterozygosity detection was done with the programs Phred (14), Phrap (15), and PolyPhred (16), and was visualized with Consed (17).

Figure 2

Resequencing the rag1 gene. (A) Genomic organization of the rag1 gene. Coding regions are represented as yellow boxes. Regions A to E indicate segments screened by resequencing. (B) Distribution of SNPs in the two haplotypes of the rag1 gene found in this study compared with the published sequence (13). Black bars indicate a SNP in one haplotype; red bars indicate a SNP in both haplotypes. (C) Spectrum of the ENU-induced mutations found in this screen compared with the mutations detected in previous forward screens. (D) Mutations changing the coding region of therag1 gene. RAG1 domain structure is adapted from (33). The DDE motif characteristic of recombinases (34) is indicated below the structure. Amino acid changes are indicated with black arrows above the structure: D, Asp; E, Glu; F, Phe; G, Gly; H, His; K, Lys; L, Leu; M, Met; N, Asn; R, Arg; V, Val. Purple arrow indicates stop codon. Gray arrows indicate silent mutations.

Table 1

Rag1 resequencing statistics. Fragments correspond to the fragments indicated in Fig. 2A. All sequencing reactions were analyzed with PolyPhred (version 3.5 beta) with a ratio of 0.65, a background ratio of 0.15, and an average Phred score of 25. The total number of base pairs (bp) screened is corrected for fragment overlap and primer sequences. Failed sequences were excluded. Average lengths include standard errors. One mutation was found in fragment D, as well as in the overlapping part of fragment E.

View this table:

Single nucleotide polymorphisms (SNPs) were encountered at a high frequency; they could be distinguished from ENU-induced mutations because they were detected in multiple fish. As shown in Fig. 2B, we detected two haplotypes that differ from the published sequence of rag1 at about 1 in 116 nucleotides; five of these SNPs result in coding changes in the rag1 gene. The two haplotypes indicate that the fish used for this screen, the so-called Tübingen strain, is not homogeneous at this locus. Other loci also have been found to show heterogeneity within this library of F1 fish (18).

After we filtered the SNPs, we detected a total of 15 ENU-induced mutations (Table 1). Each of these mutations was verified by independent PCR and resequencing. Because most mutation spectra are based on the detection bias of the forward screen, we can use these data to get an initial unbiased picture of the mutation spectrum of ENU in the zebrafish. As shown in Fig. 2C, the distribution of the mutations found in rag1 is similar to those found in previous forward screens. The mutation rate is 15 mutations in 7,268,581 base pairs among 2679 fish (Table 1), which corresponds to a frequency of 2.1 × 10−6 per base pair.

Of the 15 mutations in rag1, 3 are silent, 2 are in intronic sequences (some introns were partially covered), and 10 result in changes in the coding sequence, of which 9 are missense mutations and 1 is a premature stop (Fig. 2D). The latter is in the middle of the catalytic domain and, therefore, is expected to be a complete null allele. We genotyped the F2 progeny of the fish that carry this allele (rag1 t26683) by isolating genomic DNA from fin clips, nested PCR, and sequencing. Crossing two F2 heterozygotes gave Mendelian segregation of the mutant allele. The homozygous F3 mutant embryos show rag1messenger RNA (mRNA) expression in the thymus (19) as seen by whole-mount in situ hybridizations (18). Adult mutant F3 fish are viable and fertile in our standard nonsterile aquarium facilities.

A functional RAG1 protein is required for V(D)J recombination (20). Therefore, we examined functionality of the truncated RAG1 protein in mutant adults by a PCR assay (21) that checks for V(D)J rearrangements of the immunoglobulin heavy-chain genes (22) as outlined in Fig. 3A. Whereas heterozygous rag1 t26683 mutant fish show joining of four different classes of heavy-chain genes, homozygousrag1 t26683 mutant fish were found to be deficient in V(D)J joining (Fig. 3B). Apparently, as in other vertebrates, there is only one functional rag1 gene; loss of function at this locus results in a complete block of immunoglobulin gene assembly and, presumably, in immunodeficiency.

Figure 3

Functionality of different RAG1 mutants. (A) Outline of the PCR assay used to test for V(D)J recombination of the immunoglobulin heavy-chain gene (12). PCR primers are represented as arrows. Semi-nested PCRs with an additional internal primer in the V segments were done to obtain a bright signal. C, constant domain. (B) V(D)J recombination in tail tissue from adult fish carrying different rag1 alleles. The mutations tested are as follows: R797stop (rag1 t26683), D251E (rag1 t26702), L653F (rag1 t26703), and M138L (rag1 t26693). V-J, PCRs with primers in V segments and J segments. V-V', PCRs with primers in V segments only. VH, variable heavy chain.

Some of the missense mutations are at conserved amino acid positions; therefore, it is possible for these to affect gene function. Three other alleles have been tested for loss of function as heteroalleles over the null allele. These were not defective in V(D)J joining (Fig. 3B). Although we have assayed only for a loss-of-function phenotype, production of an allelic series is a major advantage of this method over knockout strategies that rely on homologous disruption or deletions. In some cases, complete loss of function may be lethal, and missense mutations may have informative viable phenotypes.

The clonal immune system was first acquired in evolution by teleost fish. Here we describe experimental inactivation of the immune system in a teleost fish. These presumably immunodeficient fish survive to adulthood without obvious signs of infectious disease in a nonsterile environment. Whether homozygous mutant lines can be bred without acquiring diseases at a higher rate than control cases remains to be seen. The rag1 t26683 mutant is now available for detailed analysis of the role of the clonal immune system in resistance to infections and possibly in other processes.

We expect that the first-generation zebrafish mutant library described here will be followed by others, if only because the mutation efficiency can possibly be improved and the size of the library increased. Nevertheless, the current library has recently been used successfully to detect mutations in other genes (18), and new screens are ongoing. Once the DNA and the testes library were established, the time and reagent costs for the rag1 screen were about 2 months and $12,500, respectively (12); these are currently being reduced by further automation and alternative mutation detection methods. With establishment of this technology, the zebrafish is now the second vertebrate model organism in which reverse genetics can be applied on a large scale.

Target-selected mutagenesis has also been successfully developed in the two major invertebrate model organisms,Drosophila (23–25) and C. elegans(26, 27), as well as inArabidopsis (28, 29) and, recently, the mouse (30, 31). The approach described here for zebrafish differs in that random mutagenesis is followed directly by straightforward, high-throughput resequencing of the target gene. This approach is insensitive to the presence of frequent SNPs in outbred organisms, because these are filtered out in the data analysis step. Therefore, this approach could be applicable for any organism on the condition that it can be bred and mutagenized efficiently. The advantage of multiple alleles, including partial loss-of-function mutants, may make it the preferred approach even for organisms for which gene knockout strategies already exist, such asDrosophila and C. elegans.

  • * To whom correspondence should be addressed. E-mail: plasterk{at}


View Abstract

Navigate This Article