Proc. Natl. Acad. Sci. USAVol. 92, pp. 5997-6001, June 1995Genetics

Efficient induction of point mutations allowing recovery ofspecific locus mutations in zebrafishBRUCE B. RILEY AND DAVID J. GRUNWALDDepartment of Human Genetics, University of Utah, Salt Lake City, UT 84112

Communicated by Mario R Capecchi, University of Utah, Salt Lake City, UT, February 8, 1995

ABSTRACT A technique is described that greatly in-creases the efficiency of recovering specific locus point mu-tations in zebrafish (Danio rerio). Founder individuals thatwere mosaic for point mutations were produced by mutageniz-ing postmeiotic gametes with the alkylating agent N-ethyl-N-nitrosourea. Under optimal conditions, each founder carriedan average of 10 mutations affecting genes required forembryogenesis. Moreover, "2% of these founders transmittednew mutations at any prespecified pigmentation locus. Anal-yses ofnew pigmentation mutations confirmed that most werelikely to be point mutations. Thus, mutagenesis of postmeioticgametes with N-ethyl-N-nitrosourea yielded frequencies ofpoint mutations at specific loci that were 10- to 15-fold higherthan previously achieved in zebrafish. Our procedure should,therefore, greatly facilitate recovery of multiple mutant allelesat any locus of interest.

Studies of zebrafish development are progressing rapidly due,in part, to techniques for inducing and recovering mutationsthat disrupt functions of individual genes. Saturation mutagen-esis screens of the zebrafish genome have identified a largefraction of the genes required for embryonic development (1,2). To fully investigate the functions of genes identified in suchscreens, we required more efficient techniques to recover avariety of mutant alleles of specific genes of interest. In aprevious mutagenesis technique, y-rays were shown to inducespecific locus mutations at high frequencies (3), but thistechnique is thought to induce large deletions and chromo-somal rearrangements and, therefore, may have limited appli-cability for analyzing functions of single genes. In contrast, theaforementioned saturation mutagenesis screens demonstratedthat treating adult males with N-ethyl-N-nitrosourea (ENU)induced point mutations in prespermatogonial stem cells, butonly 0.1-0.2% of mutation-bearing founders harbored a mu-tation at a prespecified locus (1, 2). Because the latter techniquerequired application of near-lethal levels of ENU, it seemedunlikely that significantly higher point mutation frequenciescould be achieved by premeiotic mutagenesis with ENU.An alternative ENU mutagenesis technique that has been

less thoroughly investigated is induction of mutations duringpostmeiotic stages of spermatogenesis. Postmeiotic mutagen-esis with ENU generates offspring that are genetically mosaic,not heterozygous, for newly induced mutations (1, 2, 4). Wereasoned that genetically mosaic offspring, which harbor bothmutant and wild-type cells in all tissues, should be able to toleratelarger numbers of mutations than heterozygous carriers of newmutations. Here we report a technique for ENU mutagenesis ofpostmeiotic gametes that yields point mutations at frequenciesthat are 10- to 15-fold higher than those previously reported forzebrafish. Our technique permits rapid recovery of new muta-tions at prespecified loci and efficient recovery of mutationsaffecting embryogenesis in open-ended screens.

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

MATERIALS AND METHODSFish Strains. Wild-type zebrafish strains were derived from

the AB line developed in Eugene, OR, or, alternatively, fromstocks purchased from a pet store supplier (Ekkwill) that weresubsequently bred for a minimum of four generations in ourlaboratory. Tester strains carrying recessive mutations at oneor more pigmentation loci (albino, brass, golden, and sparse)were derived from strains supplied by C. Kimmel, W. Driever,and N. Hopkins.

Mutagenesis. Stocks of 50 mM ENU were prepared bydissolving a 1-g Isopac of ENU (Sigma catalogue no. 3385) in171 ml of sterile aquarium water (distilled water containing0.008% Instant Ocean salts) buffered with 5mM Mes (pH 6.0).Protective clothing and fume hoods were always used whilehandling ENU, which is highly toxic. Stocks could be stored at-80°C for several months without apparent loss of mutagenicactivity. To mutagenize, two to four wild-type males wereincubated for 30-60 min at 28°C in disposable 1-liter beakerscontaining 500 ml of aquarium water buffered with 3 mM Mes(pH 6.2) and various concentrations of ENU. Mutagenizedfish were then rinsed in fresh water and sequestered in mousecages containing aquarium water for 1 h to permit recoveryand observation. Fish treated with ENU doses of 2 mM-h (2mM ENU for 1 h) or less usually recovered quickly and couldbreed immediately. After each experiment, ENU was disposedof as described (2).

RESULTSWe hypothesized that a highly efficient method for inducingpoint mutations in zebrafish would be to mutagenize postmei-otic gametes with ENU. To accomplish this, we used the simpleand reliable mutagenesis technique of incubating adult malesin buffered aquarium water containing ENU (1, 2). Thistechnique induces mutations in both premeiotic stem cells andpostmeiotic gametes. However, mutations in premeiotic stemcells are not transmitted by mature sperm until at least 3 weeksafter treatment (1, 2). Hence, we measured the effects of ENUmutagenesis within the first 2 weeks after treatment.

Induction of Somatic Mutations. We wished to establishoptimal mutagenesis conditions that would give maximumyields of point mutations without inducing high levels ofdominant lethality. In previous studies, dominant lethality wasusually associated with overtly abnormal embryos (1, 2, 4).Such lethality could result from induction of large deletionsand chromosomal rearrangements. Therefore, F1 progenyproduced by males that had been mutagenized under a varietyof conditions were simultaneously subjected to two kinds ofscreens (Fig. 1): One to measure the production of newpigmentation mutations carried by F1 soma and the other tomeasure the production of abnormally developing embryos(monsters).To detect new recessive pigmentation mutations, mu-

tagenized wild-type males were crossed to females carrying a

Abbreviation: ENU, N-ethyl-N-nitrosourea.

5997

5998 Genetics: Riley and Grunwald

60.f

._

g 301

0.

100e 80-0

4l Oj

20

na

0.0

12

6

0

0.2 0.4 0.6 0.8 1.0 1.2 1.4

02 , +. 0 .4,1.1 . 1. 40.2 0.4 0.6 0.8 1.0 1.2 1.4

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

ENU dose, mM-h

FIG. 1. Induction of somatic mutations and monsters in Fl em-

bryos. Wild-type males were given single treatments with various dosesof ENU. Treatments were for 1 h (E) with the indicated concentrationsof ENU except for 1.5 mM ENU for 0.5 h (0), 1.0 mM for 0.75 h (0),and 1.2 mM ENU for 0.75 h (-), as indicated. Treated males were

crossed repeatedly for up to 2 weeks after mutagenesis to gol/golfemales. Resulting F1 embryos were scored at 2-3 days of developmentfor somatic gol mutations and for gross morphological defects. (A)Percentages of embryos showing pigmentless spots in the pigmentedretina (gol mosaics). Embryos with grossly deformed eyes wereexcluded from the total. (B) Percentages of embryos with morpho-logical defects (monsters). These include severely deformed embryosthat could not be scored for retinal pigmentation and embryos withmilder defects that were included in the initial retinal screen. (C)Percentages of gol mosaic embryos showing normal morphology. Allmonsters (including mildly deformed embryos that were included inthe initial retinal screen) were excluded from the total and percentagesof gol mosaic embryos were recalculated. Each data point shows themean and standard deviation of 5-12 experiments, and 100-150embryos were screened for each experiment.

recessive viable mutation at the golden (gol) locus, and F1progeny were screened at 48 h for pigmentless (golden) cellsin the pigmented retina (Fig. 1A). To obtain an initial dose-response relationship, males were treated for 1 h with con-

centrations ranging from 0.4 mM to 4.0 mM ENU. Thefrequencies of gol mosaic embryos produced by mutagenizedmales increased with ENU concentrations up to 1.3 mM, atwhich dosage the frequency ofgol mosaic embryos approached

50%. Higher doses of ENU were excessively harsh: Exposureto 2 mM ENU for 1 h yielded embryos that were too aberrantto screen for golden retinal cells. Furthermore, males treatedfor 1 h with either 3 mM or 4 mM ENU usually died withinseveral hours or minutes after treatment, respectively.To test a wider variety of moderate mutagenesis regimens,

we reduced the duration of exposure to intermediate concen-trations of ENU. As with the 1-h treatments described above,shorter treatments yielded frequencies of gol mosaic embryosin a dose-dependent manner (Fig. 1A). For example, threeregimens that delivered total ENU doses of 0.75-0.8 mM-h(1.5 mM ENU for 0.5 h, 1.0 mM ENU for 0.75 h, and 0.8 mMENU for 1.0 h) resulted in similar frequencies of gol mosaicembryos. These data indicate that the frequency of inducingsomatic mutations atgol is a sensitive function ofENU dosage.A similar dose-response relationship was observed for the

production of monsters (Fig. 1B), which served as an indicatorof nonpropagable damage caused by ENU. ENU doses of0.6-0.8 mM-h caused only modest increases in the incidence ofmonsters over control levels. However, ENU doses exceeding0.9 mM-h usually caused >50% of F1 embryos to develop asmonsters. Hence, even though ENU doses of 0.9-1.3 mM-hgave the highest frequencies of gol mosaic embryos, the highincidence of monsters suggested that these regimens might besuboptimal for generating point mutations.To identify the regimens likely to yield the highest frequen-

cies of point mutations, frequencies of gol mosaic embryoswere recalculated after excluding all abnormally developingembryos (Fig. 1 C). Overall frequencies of normal-appearinggol mosaic embryos were similar for regimens delivering ENUdoses of 0.75-1.3 mM-h. However, mean frequencies of golmosaic embryos peaked at 0.8 mM-h ENU and declined ateither higher or lower doses ofENU. Thus, excluding monstersrevealed a relatively narrow range of ENU doses that mightefficiently induce new point mutations carried by the germlines of F1 embryos.

Frequencies of both gol mosaic embryos and monstersproduced by mutagenized males remained relatively constantfor up to 2 weeks after mutagenesis but dropped sharply by 3weeks (data not shown). These results presumably reflectspermatogonial turnover and the eventual replacement ofheavily mutagenized sperm with sperm derived from lessheavily mutagenized stem cells (1, 2).

Induction of Pigmentation Mutations Carried by F1 GermLines. We tested four mutagenesis regimens that efficientlyinduced somatic mutations to establish whether specific locusmutations were transmitted by the germ lines of F1 offspring(Table 1). Males mutagenized by these regimens were repeat-edly crossed to wild-type females for up to 2 weeks aftermutagenesis. Surviving F1 offspring were tested for germ-linetransmission of new pigmentation mutations by crossing themto tester fish carrying recessive mutations at up to fourpigmentation loci: albino, brass, golden, and sparse. Thehighest dose (1.3 mM for 1 h) was not effective because, onaverage, only 1% of F1 embryos survived to adulthood. The

Table 1. Mutation frequencies at specific loci

% mean frequency ofmutations per locus per

F1 germ line% No. mutants/no. loci screened Total Point

Mutagenesis regimen survival Albino Brass Golden Sparse Total mutations mutations

1.3 mM ENU for 1 h 1 ± 1 ND ND ND ND ND ND ND1.0 mM ENU for 1 h 22 ± 16 2/42 0/63 3/81 1/81 6/267 2.2 1.40.8 mM ENU for 1 h 35 ± 14 2/40 2/72 0/85 2/85 6/282 2.1 2.11.5 mM ENU for 0.5 h 21 ± 18 1/64 0/69 0/69 0/69 1/271 0.4 ND

Data for percent survival are expressed as the mean ± SD. ND, not determined. Data for point mutations were calculatedby subtracting values for lethal mutations from values for total mutations.

A

X je,0

B

co 0

9~~~ip V

T-

Proc. Natl. Acad. Sci. USA 92 (1995)

.. '

Proc. Natl. Acad. Sci. USA 92 (1995) 5999

lowest dose (1.5 mM for 0.5 h) also proved to be relativelyineffective because survivors appeared to carry few pigmen-tation mutations: 98 F1 offspring were tested for transmissionof new mutations at 2 or 3 pigmentation loci each (a total of271 loci tested), but only 1 of these was found to carry a newpigmentation mutation in its germ line. This constitutes anaverage mutation frequency of 0.4% per locus per F1 germline, which is similar to the highest frequencies reported inprevious studies (2-4). In contrast, two regimens (0.8 mMENU for 1 h and 1 mM ENU for 1 h) gave average mutationfrequencies in excess of 2% per locus per F1 germ line: Of 96F1 offspring from the regimen of 0.8mM ENU for 1 h that werescreened at a total of 282 loci, 6 carried new pigmentationmutations in their germ lines, giving an average mutationfrequency of 2.1% per locus per F1 germ line. Similarly, of 81F1 offspring from the regimen of 1 mM ENU for 1 h that werescreened at a total of 267 loci, 6 carried new pigmentationmutations, giving an average mutation frequency of 2.2% perlocus per F1 germ line. These specific locus mutation frequen-cies are 10- to 15-fold higher than average frequencies re-ported in previous ENU mutagenesis studies.

Similar mutation frequencies were observed at each of thefour pigmentation loci. Even though the albino locus appearedto be roughly twice as mutable as the other loci, the frequencyof albino mutations was not statistically different from themean of all mutations combined (P > 0.2). In any case, ourmutagenesis technique clearly induced new mutations at allfour loci with high efficiency. Hence, the technique should begenerally useful for inducing mutations at other loci in thezebrafish genome.The high specific locus mutation frequencies obtained with

the two most efficient mutagenesis regimens could result froma large number of point mutations and/or a small number oflarge lesions. Analysis of mutant phenotypes indicated thatmost of the pigmentation mutations induced by these regimenswere likely to be point mutations: 5 of the 12 pigmentationmutations were "weak" alleles whose phenotypes were inter-mediate between those associated with putative null mutationsand the wild type. Weak alleles often arise from induction ofsmall intragenic lesions that reduce, but do not eliminate,activities of mutant gene products. A similar proportion ofweak pigmentation alleles was obtained in a mutagenesis study(1) in which >90% of the mutations were reported to be pointmutations. To further analyze the pigmentation mutationsfrom the current study, we attempted to propagate fish car-rying these mutations to determine whether they were ho-mozygous viable (consistent with their being point mutations)or lethal (likely to be large lesions that disrupted linkedessential genes). Of the 6 pigmentation mutations from theregimen of 1 mM ENU for 1 h, 2 were dominant lethalmutations and 3 others were found to be homozygous viable.The sixth mutation was not analyzed due to death of the F1carrier soon after its initial screening. Hence, 3 of 5 of thetested mutations behaved genetically as point mutations. Thenumber of mutants analyzed is too small to determine whetherinduction of dominant lethal mutations will be a generalfeature of 1-mM h mutagenesis. Of the 6 pigmentation muta-tions from the regimen of 0.8 mM ENU for 1 h, 5 werehomozygous viable and the sixth was not recovered. The latter,however, was a weak allele and, therefore, likely to be a pointmutation. Thus, these results suggest that the regimen of 0.8mM for 1 h gave the highest frequencies of point mutations(Table 1).The dose-response relationship for induction of point mu-

tations carried by F1 germ lines (Table 1) was similar to thedose-response relationship for induction of mutations carriedby F1 soma (Fig. 1 C). However, somatic mutation frequencieswere consistently 4- to 10-fold higher than germ-line mutationfrequencies, as reported (1, 2, 4). One explanation for therelative paucity of germ-line mutations is that the zebrafish

germ line may be derived from as few as 1 to 5 progenitor cells(5). Thus, there is a substantial probability that cells carryingany given mutation will fail to colonize the germ line. Incontrast, the pigmented retina is derived from 40 to 50progenitor cells (6) and will almost always be colonized by cellscarrying a given mutation.

Screens for Embryonic Lethal Mutations. The high fre-quency of pigmentation mutations reported above suggestedthat each F1 fish was likely to carry a relatively large numberof independent mutations in essential genes. To test thispossibility, we conducted three-generation mutant screens thatwere similar to those used in previous zebrafish screens exceptfor one modification: Mutations harbored by F1 offspring werepropagated by generating F2 families as was done (1, 2), butmutations were then uncovered by backcrossing F2 offspring totheir F1 parents. Mutations were identified by the appearanceof syndromes (developmental abnormalities shared by three ormore F3 progeny).One consequence of screening by backcrossing is that most

syndromes were expressed by relatively small fractions of F3progeny due to the germ-line mosaicism of F1 parents. Nev-ertheless, the appearance of syndromes proved to be a reliableindicator for the presence of new mutations. Moreover, onceidentified in a backcross, mutations were easily recoveredbecause F2 carriers, unlike their F1 parents, were heterozygousfor new mutations and transmitted those mutations at standardMendelian frequencies.

Several examples of embryonic lethal mutations recoveredfrom our screen are shown in Fig. 2. These are representativeof mutations that cause well-defined morphological alter-ations. Such patterning mutations made up 30-40% of all

A c E

*~~~~~~~~~~~~~~~~~~~~~c

FIG. 2. Embryonic mutants recovered from backcross screens.Mutations initially identified in backcrosses between F2 offspring andtheir F1 parents were propagated by crossing F2 offspring to wild-typemates to generate F3 carriers. All mutant embryos depicted weregenerated by intercrossing F3 carriers except for the heterozygoteshown in E, which was generated by crossing an F3 carrier to awild-type mate. For embryos in B-E, anterior is to the left and dorsalis up. (A) Several 48-h embryos that are homozygous for a recessivelethal mutation causing dorsal curvature of the tail. Their sharedabnormalities constitute a clearly defined syndrome. (B) A 30-hembryo that is homozygous for a recessive lethal mutation affectingbrain patterning. The mutant displays a well-formed midbrain-hindbrain border (mhb) and an abnormal structure (*) in the middleof the hindbrain. (C) Trunk region of a 48-h embryo that is homozy-gous for a recessive lethal mutation that perturbs neuraxial develop-ment. The mutant displays a relatively normal notochord (nc) and aslightly reduced spinal cord, but the floor plate (fp) of the spinal cordis not present. (D) Trunk region of a 48-h wild-type embryo showingthe notochord (nc) and the floor plate (fp) of the spinal cord. (E) A24-h wild-type embryo (lower embryo) and a mutant sibling (upperembryo) that is heterozygous for a recessive lethal mutation with adominant defect in inner ear development. In the wild-type embryo,two otoliths (arrows) are clearly visible in the inner ear, one in thefuture utricle (anterior) and another in the future saccule (posterior).In contrast, the mutant embryo lacks the anterior otolith. Heterozy-gotes are rarely affected in both ears and usually grow to adulthoodnormally. Homozygotes usually show bilateral loss of utricular oto-liths, display severe equilibrium deficits, and die by day 10 of larvaldevelopment. Saccular otoliths are always present in both heterozy-gotes and homozygotes.

Genetics: Riley and Grunwald

6000 Genetics: Riley and Grunwald

syndromes observed in our screen. The remainder, which wedid not attempt to propagate, caused nonspecific necrosis ordegeneration. A similar ratio of specific patterning mutationsto nonspecific degeneration mutations was reported in a largesaturation mutagenesis screen of the zebrafish genome (1).To determine the number of backcrosses required to un-

cover most F1 mutations, it was important to determine therange of transmission frequencies of F1 mutations. Fig. 3Ashows the distribution of transmission frequencies of >100 F1mutations. The mean frequency of transmission was 28%, inclose agreement with the expected average of 25% (1, 2, 4).Because ENU ethylates only one strand of the sperm DNA,subsequent replication in the zygote generates two kinds ofpaternally derived chromosomes, one mutant and one wildtype, that segregate into different cells. When cells withdifferent paternal chromosomes contribute equally to thegerm line of an F1 offspring, 25% of its gametes will carry thepaternal mutation. However, because the zebrafish germ lineis derived from <10 progenitor cells (5), the proportion ofmutant cells that randomly contribute to the germ line couldvary greatly. Accordingly, we observed a wide range of germ-line transmission frequencies of F1 mutations, from 2 to 70%(Fig. 3A).From the frequency distribution of F1 mutations (Fig. 3A),

we calculated the proportion of mutations that could berecovered after a given number of backcrosses (Fig. 3B).About 50% of all mutations harbored by F1 parents will beuncovered in only two backcrosses, and two-thirds of allmutations will be uncovered in four backcrosses. The effi-ciency of uncovering further mutations drops dramatically withsubsequent backcrosses. For example, doubling the number ofbackcrosses from four to eight increases the yield of mutationsby only 20% (from two of three to four of five of all mutations).Hence, an optimal screening strategy might be to conductfewer backcrosses (e.g., four backcrosses) with a larger numberof F1 parents.

Screening families derived from the two most efficientmutagenesis regimens (0.8 mM ENU for 1 h and 1.0mM ENUfor 1 h) revealed that each F1 parent harbored numerousembryonic lethal mutations. On average, nearly two syn-dromes were observed in each backcross (Table 2), reflectingthe average number of mutations carried by each F2 offspring.By extrapolating from the recovery curve shown in Fig. 3B, weestimated that each F1 parent carried an average of 9 or 10embryonic lethal mutations (Table 2). This is -10-fold higherthan the frequencies of embryonic lethal mutations reportedin previous screens (1, 2), an increase that agrees closely withthe 10- to 15-fold increase in the frequencies of pigmentationmutations (Table 1). In contrast, screening families derivedfrom a suboptimal regimen (1.5 mM ENU for 0.5 h) revealedvery few embryonic lethal mutations (Table 2), which wasconsistent with the low incidence of pigmentation mutationsobtained with that regimen (Table 1).

DISCUSSIONAttempts to increase mutation frequencies are usually con-founded by increased morbidity and lethality among carriers of

30A

*^ 20

48 10

0

100.

a

I0

cN- N m XT m~ £O t0%0 0 0 0 0% 0%

N I t

% of Germline Carrying a Mutation

0 4 8Number of Backcrosses

12

FIG. 3. Germ-line transmission frequencies and efficiency of re-covery of newly induced Fl mutations. (A) Percentage of F1 germ linecarrying a new mutation (frequency distribution). F2 offspring werebackcrossed to their F1 parents and F3 progeny were scored forsyndromes. Inheritance patterns showed that most syndromes resultedfrom recessive mutations, so the fraction of F3 progeny affected by asyndrome (putative homozygotes) was assumed to be 50% of thefraction of F1 gametes carrying the mutation. Data were derived fromfrequencies of 103 syndromes in addition to the 13 pigmentationmutations presented in Table 1. (B) Percent recovery of F1 mutations.To simplify calculations, mutations were grouped into germ-linefrequency classes as shown inA. The mean germ-line frequency (Lc) ofeach class was used to calculate the probability of recovering mutationsfrom that class after a given number of backcrosses. The percentage(p,,) of all mutations recovered after n backcrosses was then calculatedby summing the probable recoveries of mutations from individualfrequency classes, which were weighted by the relative number ofmutations contained in each (mc). Thus, pn = E mc [1 - (1 -c)n].

newly induced mutations. The ability to effect the high pointmutation frequencies observed in this study probably relies onthe genetic mosaicism caused by postmeiotic mutagenesis withENU. All tissues of mosaic F1 offspring carry significantproportions of wild-type cells, which may serve to complementpotential defects in mutagenized cells. Genetic mosaicism alsoameliorates a second problem associated with high mutagen-esis levels: cosegregation of multiple newly induced mutations.Even though each mosaic F1 parent carries an average of 10embryonic lethal mutations, each F2 offspring often inheritsonly 1 or 2 mutations. Such efficient segregation makes itpossible to isolate individual mutations in as little as a singlegeneration.The high specific locus mutation frequencies reported here

were not the result of inducing large lesions, as was shown tobe the case for a technique that used y-rays (3). Several linesof evidence suggest that the mutagenesis technique reportedhere primarily induces point mutations. (i) The majority of new

Table 2. Identification of embryonic lethal mutations

Mean no. of Mean no. of ExtrapolatedNo. of Mean no. of syndromes mutations no. of

F1 backcrosses observed recovered mutationsparents per F1 per per F1 per F1 germ

Mutagenesis regimen screened parent backcross parent line1.0 mM ENU for 1 h 18 4.1 1.7 5.8 10 ± 3.50.8 mM ENU for 1 h 6 2.2 1.9 4.0 9 ± 3.61.5 mM ENU for 0.5 h 4 3.0 0.4 1.0 1.7 ± 0Data for extrapolated number of mutations are the mean ± SD.

Proc. Natl. Acad. Sci. USA 92 (1995)

Proc. Natl. Acad. Sci. USA 92 (1995) 6001

pigmentation mutations were viable, indicating that they didnot affect linked essential genes. (ii) A large proportion of newpigmentation mutations were weak alleles, which typicallyresult from small intragenic lesions. (iii) The ratio of totalembryonic mutations to specific locus mutations was similar inour study and in previous studies (1, 2) that used ENU toinduce point mutations in premeiotic stem cells, indicating thataverage lesion sizes are comparable for postmeiotically andpremeiotically induced mutations. Thus, these results indicatethat, under optimal conditions, postmeiotic mutagenesis withENU increases the efficiency of inducing and recovering pointmutations by an order of magnitude over previous techniques.The mutagenesis technique reported here will facilitate

fine-level genetics studies that were previously consideredimpractical or impossible in vertebrate species. For example, asshown here for pigmentation mutations, it is now feasible toobtain multiple alleles of previously identified mutations in ashort period of time. This could provide a rich variety of allelictypes, including weak alleles, gain of function alleles, andconditional null alleles, which are required to fully analyzegene function. Weak alleles can uncover subtle aspects of genefunction, such as functional gradients, that may be obscured bynull alleles (7, 8). In addition, phenotypic analysis of differentallele types often reveals diverse stage-specific and tissue-specific functions of individual genes (9-11). Structural anal-ysis of different alleles with discrete phenotypes can alsoidentify functional domains of the gene product in question(12, 13).As a special case of obtaining multiple alleles, it should be

possible to efficiently screen for point mutations that fail tocomplement -y-ray-induced deletions of previously cloned se-quences. Deletions provide an important test of gene functionby guaranteeing a null phenotype, but y-ray-induced deletionsare usually so large (3) that it is difficult to correlate a mutantphenotype with loss of a single gene. Therefore, point muta-tions that fail to complement large deletions covering knownsequences could provide a means to test functions of clonedgenes.The ability to induce point mutations with high efficiency

should facilitate genetic screens to elucidate networks ofinteracting genes that control specific developmental path-ways. For example, weak alleles that partially impair develop-mental pathways can be used to screen for second site muta-tions that either fully disrupt the pathway or that rescue the

pathway (14-20). Such screens can demonstrate interactionsbetween mutant genes whose phenotypes by themselves wouldnot immediately suggest participation in a common pathway.

Finally, our mutagenesis technique greatly reduces the spacerequired to screen relatively large numbers of mutations. Thus,it will be feasible to conduct relatively large mutant screens inlaboratories with limited space for fish tanks.

We thank Wolfgang Driever, Nancy Hopkins, and Chuck Kimmelfor supplying various fish strains, and Sharon Johnson and Eric Mullenfor assistance in breeding and maintaining the strains. This work wassupported by National Institutes of Health Grant 1 R03 RR07832-OlAl, National Science Foundation Grant MCB-9420984, and a grantfrom the University of Utah. B.B.R. was supported by NationalInstitutes of Health Fellowship 5F32 GM14797-02 and AmericanCancer Society Fellowship PF-04186-01.

1. Mullins, M. C., Hammerschmidt, M., Haffter, P. & Niisslein-Volhard, C. (1994) Curr. Bio. 4, 189-201.

2. Solnica-Krezel, L., Schier, A. F. & Driever, W. (1994) Genetics136, 1401-1420.

3. Chakrabarti, S., Streisinger, G., Singer, F. & Walker, C. (1983)Genetics 103, 109-123.

4. Grunwald, D. J. & Streisinger, G. (1992) Genet. Res. 59, 103-116.5. Walker, C. & Streisinger, G. (1983) Genetics 103, 125-136.6. Streisinger, G., Coale, F., Taggart, C., Walker, C. & Grunwald,

D. J. (1989) Dev. Bio. 131, 60-69.7. Gergen, J. P. & Wieschaus, E. (1986) Cell 45, 289-299.8. Beddington, R. S. P., Rashbass, P. & Wilson, V. (1992) Devel-

opment (Cambridge, UK) Suppl., 157-165.9. Austin, J. & Kimble, J. (1987) Cell 51, 589-599.

10. Preiss, J. R., Schnabel, H. & Schnabel, R. (1987) Cell 51, 601-611.11. Clifford, R. J. & Schupbach, T. (1989) Genetics 123, 771-787.12. Barbacid, M. (1987) Annu. Rev. Biochem. 56, 779-787.13. Henkemeyer, M., West, S. R., Gertler, F. B. & Hoffmann, F. M.

(1990) Cell 63, 949-960.14. Gertler, F. B., Bennett, R. L., Clark, M. J. & Hoffmann, F. M.

(1989) Cell 58, 103-113.15. Beitel, G., Clark, S. & Horvitz, H. R. (1990) Nature (London)

348, 503-509.16. Han, M., Aroian, R. & Stemnberg, P. W. (1990) Genetics 126,

899-913.17. Aroian, R. V. & Sternberg, P. W. (1991) Genetics 128, 251-267.18. Rogge, R. D., Karlovich, C. A. & Banerjee, U. (1991) Cell 64,

39-48.19. Simon, M. A., Bowtell, D. D. L., Dodson, G. S., Laverty, T. R. &

Rubin, G. M. (1991) Cell 67, 701-716.20. Cline, T. W. (1993) Trends Genet. 9, 385-390.

Genetics: Riley and Grunwald