Development 109, 577-584 (1990)Printed in Great Britain © T h e Company of Biologists Limited 1990

577

Stable lines of transgenic zebrafish exhibit reproducible patterns of

transgene expression

GARY W. STUART1*, JUERGEN R. V1ELKIND2, JAMES V. McMURRAY1 and MONTE

WESTERFIELD1

1 Institute of Neuroscience, University of Oregon, Eugene, Oregon 97403, USA2B.C. Cancer Research Centre, Vancouver, British Columbia V5Z 1L3, Canada

* Present address: Department of Life Sciences, Indiana State University, Terre Haute, Indiana 47809, USA

Summary

To study the frequency of germ-line transformation andto examine the reproducibility of tissue-specific trans-gene expression, we produced several lines of transgeniczebrafish expressing a recombinant chloramphenicolacetyltransferase (CAT) gene. Supercoiled plastnidscontaining both Rous sarcoma virus and SV-40 pro-moter sequences upstream of the CAT coding regionwere injected into zebrafish embryos prior to firstcleavage. CAT activity could be detected in batches ofinjected embryos as early as 8 h and up to at least 12 dayspost-fertilization. Approximately 18 % of injected fishraised to maturity exhibited CAT activity in their fins,and approximately 5 % of injected fish became stablegerm-line transformants. Breeding studies indicatedthat although transgenic founder fish were frequentlygerm-line mosaics, transgenic individuals of subsequentgenerations were fully hemizygous for the transgene

marker. The transgenes present in the F] progeny offour independent lines were relatively well expressed infin and skin, while lower levels of expression wereobserved in heart, gill and muscle. Little or no CATexpression was observed in the brain, liver and gonad. Amonoclonal antibody directed against the CAT geneproduct consistently revealed variegated patterns ofCAT expression in ectodermally derived fin epidermalcells in three of these lines. These results show that it ispossible to efficiently produce stable germ-line trans-formants of the zebrafish and to observe reproducibletissue-specific patterns of transgene expression in thisorganism. Possible mechanisms for the variegated ex-pression observed within tissues are also considered.

Key words: microinjection, mosaicism, RSV-LTR,transgenes, variegated expression, zebrafish.

Introduction

The production of transgenic organisms has been ofgreat value in the study of the genetic basis of embry-onic development in both mammalian and invertebratespecies (Flytzanis etal. 1985; Jaenisch, 1988; Fire, 1986;Rubin, 1988); however, the application of this tech-nique to lower vertebrates has been rare (Gurdon andMelton, 1981; Etkin and Pearman, 1987). This isparticularly surprising in view of the long-standingpopularity of the lower vertebrate embryo for the studyof vertebrate embryology and development (Morgan,1895; Gurdon, 1987; Kimmel, 1989). Unlike mam-malian embryos, most lower vertebrate embryosdevelop rapidly, are experimentally accessible (i.e.develop outside the mother), and are easy to obtain inlarge numbers. Unfortunately, many lower vertebratespecies, including frogs, exhibit long generation times,a characteristic that would hinder the genetic analysis ofnew traits resulting from transgene inserts. The use of

germ-line transformation in the study of vertebrategene expression and development would be most effec-tive in an organism with a relatively short generationtime as well as a rapidly developing, accessible embryo.In addition, in order to facilitate mechanistic studies ofdevelopmentally regulated gene expression, these or-ganisms should also be capable of exhibiting reproduc-ible, tissue-specific patterns of expression that reflectconsistent recognition of the transgene and its regulat-ory sequences.

The zebrafish, Brachydanio rerio, is a simple ver-tebrate with the potential to fulfill all of these require-ments. As in the frog, fertilization is external and can beeasily manipulated to produce haploid and gynogeneticoffspring (Streisinger ef al. 1981; Kimmel, 1989). Zebra-fish embryos are optically transparent and developrapidly, hatching from their chorions at 2 to 3 days post-fertilization. Furthermore, the production, mainten-ance and analysis of mutant lines are aided by arelatively short 3 to 4 month generation time. Recently,

578 G. W. Stuart and others

we have shown that it is possible to produce transgeniczebrafish containing stable germ-line inserts of foreignDNA (Stuart et al. 1988).

In this report, we extend our observations by demon-strating that, following the cytoplasmic injection offoreign genes into early zebrafish embryos, as many asone in 20 survivors are germ-line transformants. Unlikethe transgenes in our previous line of transgenic fish,the recombinant chloramphenicol acetyltransferase(CAT) genes used in this study are expressed, bothtransiently in developing zebrafish embryos and fry,and continuously in stable transformants. This allowedus to examine the tissue-specific expression of theforeign genes, each of which contains both a Roussarcoma virus long terminal repeat (RSV-LTR) and anSV-40 early promoter. The results indicate that thisconstruct is reproducibly expressed in a similar subsetof tissues in four independent lines; however, ananalysis of gene expression at the cellular level reveals avariegated pattern of transgene expression.

Materials and methods

Zebrafish linesTransgenic zebrafish lines bl48, bl49, bl50, and bl51 rep-resent the descendants of the transgenic founder fish CATfishI, II, III and IV, respectively. Founder fish were produced bymicroinjection of foreign DNA into early embryos of the ABwild-type line of the Oregon colony as described below.CATfish 1 and II were injected with the pUSVCAT plasmid(Karlsson et al. 1985), while CATfish III and IV were injectedwith a closely related derivative, pSVeRSVCAT(see below).

MicroinjectionSupercoiled plasmid DNA was injected into recently fertilizedzebrafish embryos prior to first cleavage essentially as de-scribed previously (Stuart etal. 1988), except that the chorionswere removed by digestion with 0.5mgml~' pronase. Theinjection solution contained 0.5 % phenol red (used to esti-mate the volume of the injection) and 50/igmF1 ofpUSVCAT or pSVeRSVCAT. pUSVCAT consists of a pUC9vector containing, in order: (1) the BamHl-Taql fragmentfrom SV-40 which includes the late region of SV-40, the SV-40origin and the SV-40 early region promoter, (2) a 524 basepair Taql-Hindlll fragment containing the RSV-LTR, and(3) a Hindlll-BamHl fragment containing the CAT gene(Karlsson et al. 1985). pSVeRSVCAT is a smaller version ofpUSVCAT; it contains the same fusion gene including bothviral promoters, but lacks the BamHI to Kpnl fragmentencompasing the late region of SV-40.

CAT assaysFor the chromatographic assay, young fry, embryos or finbiopsies were washed two times in Ca /Mg+2-free PBS,homogenized in 100/il of 250 mM Tris (pH8.0) and freeze-thawed three times. 100/tl of cell extract were obtained aftercentrifugation in a microfuge at 4°C for 5min. To thesupernatant was added 20/tl of dH2O, 2.5 /il [14C]chloramphe-nicol (40-50 mCimmoF1, New England Nuclear) and 20 /.tl of4mM acetyl-CoA (Boehringer Mannheim). The reaction wasincubated for 90min at 37°C. Negative controls consisted of100/il of 250 mM Tris (pH8.0) or 100/tl of cell extract fromuninjected embryos; positive controls included 1 unit of chlor-

amphenicol acetyl transferase (Sigma). The chloramphenicoland its acetylated forms were extracted with 1 ml of ethylacetate, dried under vacuum and redissolved in 30/il of ethylacetate. The chloramphenicol (slowest migrating spot) and itsacetylated forms were separated by ascending thin layerchromatography (TLC plates, Baker Chemicals) in chloro-form: methanol (95:5) and visualized by autoradiography with24 h exposures (Kodak XAR5 film).

Extracts for the scintillation assay were obtained from thedorsal, anal and caudal fins of adult fish, from the heads of fishat 3 weeks of age, from individual fry at 1-2 weeks of age, orfrom groups of 12-20 embryos at 1-2 days of age byhomogenization in 50-100/tl of assay buffer (100 mM TrispH7.8, 5mM EDTA, and 0.5 mM chloramphenicol). Fish wereanesthetized briefly in 10 mM ethyl-m-aminobenzoate meth-ane sulfonate (Sigma), 50mM Tris pH7.2 prior to biopsy.Individual tissues and organs of adult fish were homogenizedin 75-300/tl of assay buffer depending on the mass of thesample. Undissolved materials were removed by centrifu-gation in a microfuge. The concentration of protein in thesupernatant was determined spectrophotometrically using theBradford reagent (Bradford, 1976) supplied by Bio-Rad. CATactivity in the supernatant was measured using a rapidscintillation vial assay employing 0.1/tCi [3H]acetyl-CoA pervial (Neumann et al. 1987; Eastman, 1987). Extracts wereheated to 60°C for 10 min in the presence of 0.5 % BSA priorto assay to remove any endogenous acetyl-CoA hydrolyzingactivity. Positive signals ranged between 1500 and85000ctsmin~' in a 3h assay; negative and control signalsranged between 200 and 600ctsmin~'.

Blot analysesDNA samples for dot blot and Southern analyses wereisolated and immobilized on nylon membranes as previouslydescribed (Stuart et al. 1988). Dot blots were probed with thepBR322 based plasmid pSV-hygro (Stuart et al. 1988) radiola-beled to a specific activity of approximately Sxl^ctsmin~' /tg~'. DNA copy number standards (stnds) were recon-structed from known amounts of the pUSVCAT plasmid(7.9 kb). Copy number estimates were made using a DNAcontent of 2xl09bp/haploid genome (Hinegardner andRosen, 1972).

For Southern analysis, digested and undigested samplesand standards were separated on a 0.8% agarose gel andblotted to a nylon membrane using the alkaline blottingprocedure (Reed and Mann, 1985). A 1.3kb EcoRl-BamHlfragment from the recombinant CAT gene of pUSVCAT wasradiolabeled to a specific activity of SxlO^ctsmin"1 //g~' byrandom hexamer priming (Feinberg and Vogelstein, 1984)and hybridized overnight to the blot in a solution containing4XSSPE, 50 % deionized formamide, 1 % SDS, 0.5 % non-fatdry milk, and 0.5 mgml"' sheared calf thymus DNA. The blotwas washed six times in 2xSSC/0.1% SDS at 65°C andexposed to Kodak XAR-5 film for 22 h.

ImmunohistochemistryThe CAT antigen was visualized in whole mounts of fin tissueusing the indirect peroxidase-antiperoxidase (PAP) method(Sternberger, 1986). Fin biopsies were fixed for 4 to 6 h in 4 %paraformaldehyde in 0.1M PO4 pH7.3 (80 mM Na2HPO4 and20 mM NaH2PO4) followed by 5 min washes in 0.1M PO4pH7.3 and then distilled water, all at room temperature. Thefixed tissue was permeabilized in acetone for Ih at —20°C,followed by 5 min washes in distilled water and then PBS. Thesamples were next treated overnight at room temperaturewith a hybridoma supernatant containing the monoclonalCAT-2 antibody (provided by C. Gorman, Genentech) and

1 % DMSO. After washing the samples 3 times with 1 %BSA/1% DMSO in PBS, the secondary antibody (goat anti-mouse) diluted 1:200 with the BSA/DMSO/PBS solution wasapplied for 1-5 h at 37°C. PAP was similarly applied followinganother wash in BSA/DMSO/PBS. Excess PAP wasremoved by washing 3 times in 0.1M PO4 pH7.3. CATimmunoreactivity was visualized with a 0.5mgmr' solutionof diamino benzidine (Sigma) in 1% DMSO/O. IM PO4pH7.3 to which 0.0025% H2O2 was added after a 15minpresoak. Stained fins were dehydrated in increasing concen-trations of ethanol, cleared in methyl salicylate and mountedin permount.

Results

Injected CAT genes are expressed in early embryos,young fry and adult fishWe injected several hundred fertilized zebrafish em-bryos prior to first cleavage with the plasmid pUSVCAT(Karlsson et al. 1985). The recombinant CAT gene inthis plasmid was expected to function in zebrafish due tothe presence of eukaryotic transcription initiation(RSV-LTR and SV-40 early region promoter) andtermination (SV-40) signals. Also, recently publishedresults indicate that the pUSVCAT plasmid is wellexpressed following microinjection into the embryos ofanother small fish species (Chong and Vielkind, 1989).The CAT gene product provides a particularity con-venient transformation marker because assays for CATactivity are both rapid and sensitive, and becauseendogenous CAT activity has not previously beendetected in eukaryotic cells (Gorman et al. 1982;Neuman et al. 1987; Eastman, 1987).

In order to verify the ability of the engineered genesto produce functional CAT enzyme in zebrafish,injected embryos were examined for transiently ex-pressed CAT activity. CAT expression was monitored ingroups of 15 injected embryos harvested at varioustimes within the first day of development. CAT enzymewas initially detected in the gastrula at 8 h post-fertiliz-ation and was strongly expressed after 24 h (Fig. 1A). Insubsequent experiments, CAT activity could bedetected for at least 12 days (data not shown). Evidencefor the retention of pUSVCAT DNA in somatic cells ofolder fish was obtained after growing injected fish tomaturity (4 months). Fin extracts were obtained fromsome of these fish and tested for CAT activity. Ten ofthe 55 fish tested exhibited detectable CAT activity intheir fins. Examples are shown in Fig. IB.

Functional CAT genes are inherited by the progeny ofsome injected fishThe presence or absence of foreign DNA in the germ-line of injected fish was determined by crossing thesefish to uninjected fish and examining their progeny forCAT activity and/or pUSVCAT DNA. In order toavoid pre-selecting for germ-line transformants thatexpress their transgenes in any particular tissue, CATassays were performed using extracts from individualwhole progeny at 1-3 weeks of age or from groups ofwhole sibling embryos at 1-2 days of age. At least 15

Transgenic zebrafish 579

CAT expression in injected fish

Embryos

4h 8h 24h 24h (-

Adultfins

1

Fig. 1. CAT expression in injected fish. Negative andpositive CAT enzyme controls are indicated by (—) and (+)respectively. (A) Embryos injected with pUSVCAT prior tofirst cleavage were harvested in groups of 15 at 4 (blastula),8 (gastrula) and 24 (motile embryos) hours post-fertilization(h). The 24h timepoint was done in duplicate. Extractsobtained from these embryos were tested for CAT activityas described in Materials and methods. (B) CAT assayswere performed on extracts obtained from fin biopsies ofindividual adult fish (4 months of age) injected as embryoswith pUSVCAT DNA. The results from six of these fish areshown; one (sample 4) showed a strong signal, a second(sample 5) exhibited a weak signal.

offspring from each of 87 injected fish (including the 55injected fish examined above) were tested. Four parentfish were found to produce offspring containing CATactivity. Interestingly, two of these founder germ-linetransformants had no detectable CAT activity in theirfins (Table 1).

Approximately 35 % of the Fj progeny obtained fromCATfish I inherited CAT genes (Table 1). Transmissionfrequencies observed using the enzyme assay (38%,n=87) and the DNA assay (34%, n=62) were virtuallyidentical, indicating that most if not all individuals thatinherited the foreign DNA also expressed CAT activity.

Table 1. Mosaicism in CATfish linesPresence of CAT in:

Fo fins Fj progeny* F2 progeny*

CATfish 1 (bl48)CATfish II (bl49)CATfish III (bl50)CATfish IV (bl51)

54/149 (36%)*53/98(54%)*5/41(12%)

14/57(25%)

30/57 (53%)*24/50(48%)28/50(56%)17/32(53%)

* All progeny were obtained by crossing identified (CAT +)individuals to wild-type stocks

*• Includes results from DNA hybridization assays as well asenzyme assays

580 G. W. Stuart and others

A CAT expression in individual F^ progeny

a b c d e f g h i k I

m n o p q r s t u v w

B Inheritance of CAT DNA in individual F-i progeny

a b c d e f g h i j k

m n o p q r s t u v w

Stnds(copies/cell) 0 6 30 300

In cases where individual F] progeny were tested usingboth assays, we observed a perfect correlation betweenthe presence of strong CAT activity and the inheritanceof approximately 100 copies per cell of the foreign gene(Fig. 2). A chi-square analysis indicates that the 36%transmission rate observed for CATfish I is significantlylower (95% confidence level) from the 50% trans-mission rate expected for a dominant marker. Hence,CATfish I appears to contain a mosaic distribution ofthe foreign DNA in her germ cells. On the other hand,the transmission frequency observed in F2 progeny(53%, n=57) suggests that the F: generation was fullyhemizygous for the CAT gene and provides indirectevidence that the foreign gene sequences were stablyintegrated into the zebrafish genome.

Approximately 50 % of the ¥{ progeny obtained fromCATfish II were found to express CAT (Table 1). Eachof these fish inherited approximately 10 copies per cellof the foreign gene (data not shown). The two remain-ing germ-line transformants have not yet been fully

Fig. 2. CAT expression and inheritanceof pUSVCAT DNA in individual F,siblings. Progeny were obtained bycrossing CATfish I to wild-type males.Negative and positive CAT enzymecontrols are indicated by (—) and (+)respectively. 23 individual F, fish at3 weeks of age (a-w) were tested bothfor CAT activity (A) and pUSVCATDNA (B). The CAT enzyme positivecontrol (+) indicates the signal obtainedfrom 1 unit of purified CAT enzyme.DNA copy number standards (stnds)were reconstructed from knownamounts of the pUSVCAT plasmid(7.9 kb). The pBR322 sequences ofpUSVCAT were detected usingradiolabeled pSV-hygro DNA, whichalso contains pBR322 sequences (Stuartet al. 1988). DNA copy numberestimates were made using a DNAcontent of 2xl09bp/haploid zebrafishgenome (Hinegardner and Rosen,1972). Nine individuals inherited bothhigh levels of CAT activity andapproximately 100 copies per cell of theinjected plasmid. Two individuals(samples n and s) exhibited very weakCAT activity, but no inherited foreignDNA. Subsequent blots sensitiveenough to detect single copy sequencesalso failed to reveal foreign DNA inthese two offspring.

characterized; however, breeding studies indicate that,like CATfish I, CATfish III and CATfish IV produce lessthan 50% CAT-positive progeny and are likely to bemosaic transformants. As observed with the F2 progenyof CATfish I, roughly 50% of the F2 progeny ofCATfish II, III, and IV inherited the CAT gene(Table 1).

Inherited CAT genes are integrated into the zebrafishgenomeStable transmission of foreign DNA through the germ-line, as demonstrated above, provides strong evidenceof genomic integration. In addition, the presence ofjunction fragments of reproducible size on Southernblots can provide further support for this conclusion(see Stuart et al. 1988, for a critical analysis of inte-gration evidence). To this end, a Southern blot wasperformed with /fywl-digested DNA obtained fromindividual FL progeny of CATfish I. Since the original

Transgenic zebrafish 581

injected plasmid contained a single Kpnl site, bands ofunit plasmid length on a Southern would reveal thepresence of plasmid multimers oriented as head-to-tailrepeats, most likely generated by homologous recombi-nation of the injected (supercoiled) plasmid. One ortwo bands of a different size might represent junctionfragments. The presence of additional bands wouldpresumbly reflect either multiple integration sites, orundefined rearrangements of the foreign DNA and/orendogenous DNA at the site of integration.

Three of the 15 progeny analyzed in this way wereshown to have inherited the foreign sequences, whilethe remaining 12 showed no evidence of pUSVCATDNA (Fig. 3). Each positive fish exhibited a stronglyhybridizing band (50-100 copies per cell) identical insize to the standard pUSVCAT plasmid linearized withKpnl. In addition, each positive fish inherited anidentical set of weaker hybridizing bands (1 to 5 copiesper cell). The observation that all CAT-positiveprogeny obtained in this cross inherited the samepattern of Kpnl fragments suggests that these fish

Southern analysis of F, 'CATfish'

u 1 2 3 4 5 6 7 8 9

23.1 —

9.4 —

6.6—

4.3 —

ifi

Fig. 3. Southern analysis of F[ progeny from an identifiedgerm-line transformant, CATfish I. Lanes 1-6 contained10 j.ig samples of Kpn\ digested DNA from each of six F,fish. Three of these fish (4,i5, and 6) showed no evidence ofthe foreign DNA. Nine additional siblings analyzed on thesame blot also failed to inherit the foreign DNA (notshown). Lanes 7, 8 and 9 contained the equivalent of 0.3, 3and 30 copies/cell, respectively, of the plasmid pUSVCATlinearized with Kpnl. An undigested lO jg sample of DNAderived from the same fish as that of lane 1 is also shown(u). The relative migrations of Lambda Hindlll sizestandards are shown on the left.

inherited the same insert of foreign DNA. Futhermore,all copies of the foreign plasmid appear to be geneticallylinked to each other and probably reside at a singlechromosomal locus. Hence, the presence of more thantwo weakly hybridizing Kpnl fragments is more likelyto reflect limited rearrangements of plasmid sequencesat a single locus rather than multiple unlinked inte-grations. The hypothesis of genomic integration isfurther supported by the observation that the inheritedforeign DNA migrates with high molecular weight(genomic) DNA when left undigested (lane u, Fig. 3).In summary, the data from breeding experiments andSouthern analysis suggest that a single foreign DNAinsert was present in the germ-line of CATfish I, andthat this insert was largely composed of a pUSVCATmultimer containing head-to-tail repeats of the intactplasmid.

Inherited foreign genes show reproducible patterns ofexpression in different tissuesThe tissue-specific expression of the foreign genespresent in all four lines of transgenic fish was examinedby performing CAT assays on selected adult tissues andorgans obtained from F] progeny by dissection(Table 2). The overall pattern of CAT gene expressionwas similar but not identical in these four lines. CATactivity was detected in fin, skin, heart, gill, skeletalmuscle and eye of each line, but not in brain, liver orgonad. In addition, CAT activity was weak or absent ingut tissues. However, while CAT activity was bestexpressed in the fins and skin of CATfish I, III and IVprogeny, expression in CATfish II progeny was greatestin the heart. CAT activity was generally expressed moreefficiently in the various tissues of CATfish I progenythan in those of CATfish II, III and IV, but there arealso some exceptions to this generalization (Table 2).

A monoclonal antibody that specifically recognizesCAT (isolated from a hybridoma cell line generouslyprovided by C. Gorman, Genentech) was used todetermine which cell types expressed the foreign geneinserts. In general, the antibody-staining procedureappeared to be less sensitive than the enzymatic CATassay; CAT immunoreactivity was difficult to detect inmost tissues from CATfish I progeny and in all tissues of

Table 2. Tissue-specific CAT' expression^ in individualFj siblings representing four lines of transgenic fish

CATfish:

FinsSkinHeartMuscleGillEyesGutBrainLiverGonad

t Measured•Less than

total protein

la

1351651

881166945

7***

Ib

154464669

11175

105*

1 as units/gram2folc1 over the

Ha

249120263

149

378

*

*

total |

lib

10953

20116152714***

proteinbackground

Ilia

56229069153834*

**

value

Illb

34399

135*3010**

*

IVa

663363

3617

14512***

of 3 units/gram

rvb

239302

*862017****

582 G. W. Stuart and others

Fig. 4. CAT immunoreactivity in wholemounts of fins from transgenic fish. Arepresentative area of an adult dorsal finfrom an individual F, offspring of CATfishI is shown as seen under differentialinterference contrast (DIC) optics. Scalebar represents 25 j.im.

CATfish II progeny, although CAT activity was ob-served in many of these tissues (Table 2). However,sections of CATfish I progeny at 12 days of ageoccasionally exhibited high levels of CAT immunoreac-tivity in isolated myotubes (not shown). In addition,CAT immunoreactivity appeared in a variegated pat-tern in skin cells from identified hemizygous progeny ofCATfish I, CATfish III and CATfish IV. This variegatedexpression pattern was quite obvious in whole mountsof fin tissue, in which a subset of the epithelial cellsappeared to express variable amounts of the CATantigen (Fig. 4). Similar staining patterns were ob-served in fins obtained from the homozygous F2progeny of CATfish I. In this case, homozygosity wasverified by 100% transmission of the transgene tooutcross progeny (data not shown).

Discussion

Cytoplasmic microlnjection produces germ-linetransformants at reasonable frequenciesWe have succeeded in efficiently producing germ-linetransformants of the zebrafish by injecting a con-venient, easily detectable marker gene into the cyto-plasm of newly fertilized eggs. These experimentsproduced four germ-line transformants out of 87 fishinjected as embryos with foreign CAT genes, a successrate approaching 5 %. Since strong CAT activity wasfound in the adult fins of nearly 20 % of injected fish, itappears that an even higher fraction of individualsretain injected sequences, perhaps in an integratedstate, in at least some somatic cells.

Transgenic fish are frequently mosaicsWe previously observed both germ-line and somaticmosaicism in the founder female of a line of transgeniczebrafish that failed to express its transgene (Stuart etal. 1988). In the present study, mosaicism in injectedfish is again suggested by the frequent lack of corre-lation between individuals expressing CAT activity in

their fins, and those capable of passing this trait on totheir progeny. Only two of the ten fish that expresseddetectable CAT activity in their fins proved to be germ-line transformants. Conversely, two germ-line transfor-mants were identified that failed to express detectablelevels of CAT activity in their fins, although CATactivity was easily detected in the fins of their progeny.Although it is possible that the CAT genes in these fishwere uniformly distributed but differentially repressedin either their fins or their progeny, a mosaic distri-bution of the injected foreign DNA is a more likelyalternative. This interpretation is supported by themosaic distribution of CAT genes observed within thegerm-line of transgenic founder fish (see Table 1 andResults).

Transgenes exhibit reproducible patterns of transgeneexpressionAll four lines of transgenic fish examined in this studyshow very similar patterns of tissue-restricted transgeneexpression, exhibiting, on the one hand, relatively highlevels of CAT activity in the fin and skin, and little or noactivity in the brain, gonad and liver (Table 2). Thissimilarity in expression extends even to the cellularlevel, as evidenced by the variegated expression ob-served in fin epithelial cells (Fig. 4). These resultsindicate that the integrated pUSVCAT and pSVeRSV-CAT fusion genes are recognized and utilized by thezebrafish transcriptional apparatus in a reproducibleway, resulting in consistent tissue-specific patterns oftransgene expression.

In some transgenic organisms, transgene expressionmay be influenced or even determined by endogenouspromoter elements or enhancers near the position ofintegration (Hazelrigg, 1984; Palmiter and Brinster,1986). Recently, gene constructs containing com-promised or weak promoters have been used in trans-genic organisms to detect endogenous sequences withthe potential to direct the specific expression of nearbygenes (O'Kane and Gehring, 1987; Allen et al. 1988;Kothary et al. 1988). Position effects like these may in

Transgenic zebrafish 583

fact be responsible for some of the differences intransgene expression observed among lines in thisstudy. However, the overall similarity of transgeneexpression patterns observed is unlikely to be due toposition effects, as this would require that the trans-genes in all four independent lines had integrated at thesame locus (or very similar loci). From a statistical pointof view, this is unlikely. Furthermore, the line-specificdifferences in transgene expression observed suggestdistinct (albeit weak) position effects and, therefore,integration at independent sites. Position effects may belimited in these experiments by the relatively strongviral promoters of the pUSVCAT construct.

The gene employed in these studies contained twopromoter regions upstream of the CAT coding se-quences: the RSV-LTR promoter immediatelyupstream, and the SV-40 early region promoterupstream of the RSV-LTR (Karlsson et al. 1985). Thetissue-specific activities of both promoters have beeninvestigated separately in transgenic mice. The SV-40promoter most frequently directed expression to thechoroid plexus of the brain, while the RSV-LTR wasmost efficiently utilized in skeletal muscle, heart muscleand limbs (Brinster et al. 1984; Overbeek et al. 1986;Swain et al. 1987). The latter specificity was interpretedas a tendency of the RSV-LTR to direct expression totissues of mesodermal origin (muscle, bone and connec-tive tissue).

The expression pattern we have observed with thepUSVCAT construct in various tissues of transgeniczebrafish is partially consistent with the view that thetissue-specific properties of the RSV-LTR dominateover those of the SV-40 early promoter located furtherupstream, and that this specificity directs expression tomesodermally derived tissues. CAT activity appearedreproducibly in skeletal muscle, heart muscle and fins,while little or no CAT expression occurred in the brain.However, at the cellular level, CAT immunoreactivityin the fins of CATfish I, CATfish III and CATfish IVoccurred primarily in epithelial cells, cells presumablyderived from ectoderm rather than mesoderm (Fig. 4).Whether expression in epithelial cells results from aspecificity inherent in the RSV-LTR, the SV-40 earlypromoter region, the prokaryotic vector DNA, or somecombinatorial effect of these sequences is unknown.The complex nature of the compound promoter pre-cludes the unambiguous assignment of tissue-specificityto individual promoters or promoter elements.

Transgenes exhibit variegated expressionAlthough the CAT-specific antibody revealed CATexpression in epithelial cells of hemizygous CATfishprogeny, the level of expression in individual cellsvaried greatly. In fact, CAT immunoreactivity wasundetectable in a relatively large fraction of these cells(Fig. 4). This uneven, or variegated, expression patterncould result from one of two general processes: somaticmutation (e.g. transgene instability) or mosaic geneactivation or inactivation. Transgene instability leadingto the complete loss of the insert has not been detected

in the germ-line of CATfish I or its progeny (seeResults). Hence, this relatively extreme form of insta-bility, if it exists, would necessarily be limited tosomatic tissues in this line. Although we cannot at thepresent time eliminate the possibility that numeroussomatic mutation events occurred relatively late indevelopment resulting in reduced (or elevated) levels oftransgene expression in some cells, it seems more likelythat the variegated expression observed in these exper-iments results from the selective activation or re-pression of stable transgene inserts. In Drosophila,variegated patterns of expression have been correlatedwith transgene integrations near euchromatic/heterochromatic borders (Hazelrigg et al. 1984). Sincewe have observed a variegated pattern of transgeneexpression in at least three out of four independent linesof transgenic fish, as well as in one newly isolated line(results not shown), it seems unlikely that this patternresults from a position effect. Instead, it would appearthat variegated expression is somehow specified by theinserted foreign DNA.

It is possible to envision several mechanisms thatcould account for variegated expression. For example,non-uniform expression may be an indirect result of theuse of heterologous DNA sequences. The heterologousnature of the DNA-protein interactions required forthe recognition of chicken (or monkey) viral promotersin transgenic mice or fish might contribute to a reducedefficiency of gene activation in a given cell-type, pro-ducing a variegated pattern of transgene expression.Moreover, prokaryotic gene and/or vector sequencesmight also interfere with the process by which trans-fected genes (or their corresponding transcripts) arerecognized in particular cells (Townes et al. 1985;Kothary et al. 1989). Differential DNA methylation hasbeen implicated in the non-uniform expression of RSV-containing genes in transfected mammalian cell clones(MacGregor et al. 1987), and could be a primary orsecondary effect of the incomplete or inappropriaterecognition postulated above.

Alternatively, variegated gene expression may rep-resent a 'legitimate' form of gene regulation. Varie-gated patterns of gene expression similar to thosedescribed here have been postulated to explain the verylow levels of tissue-specific gene transcripts detectablein ectopic locations using PCR (Chelly et al. 1989;Sarkar and Sommer, 1989). Further speculation hasproduced the suggestion that 'ectopic' gene expressionin the thymus could mediate antigenic self-tolerance(Linsk et al. 1989). Hence, it is possible that thevariegated gene expression we have observed may notbe the result of incomplete or inappropriate DNA/RNA recognition as suggested above, but may actuallyreflect a legitimate host mechanism. In this case, theviral sequences used in our construct may, individuallyor in combination, specify a variegated pattern ofexpression dependent upon some unknown but variableproperty of the cells in which it is expressed. Amongother things, this property could be related to the age ofthe cell, its health or its position within the cell cycle(Chelly et al. 1989).

584 G. W. Stuart and others

We acknowledge the contributions of Ellen Johnson,Sherry O'Shea, Dorothy Schell, Marge Kuhn, CharlineWalker, and Molly Rothman to the results presented above.We also thank Pat Lambert and Thorn Montgomery for helpraising and maintaining fish stocks, Ruth Breimiller forsectioning and antibody staining, and Charles Kimmel, WaltMetcalfe, Eric Selker, John Postlethwait and Gina Stuart forhelpful comments on various versions of the manuscript. Thiswork was supported by NIH grants NS01065, GM22731,HD22486, and CA 38138, as well as a grant from the MedicalResearch Council of Canada. J.V. is a Scholar of the MedicalResearch Council. G.S. was supported by grants from theAmerican Heart Association, Oregon Affiliate and theAmyotrophic Lateral Sclerosis Association.

References

ALLEN, N. D., CRAM, D. G., BARTON, S. C , KETTLE, S., REIK, W.

AND SURANI, M. K. (1988). Transgenes as probes for activechromosomal domains in mouse development. Nature 333,852-855.

BRADFORD, M. (1976). Anal. Biochem. 72, 248-254.BRINSTER, R. L., CHEN, H. Y., MESSING, A., VAN DYKE, T.,

LEVINE, A. J. AND PALMITER, R. D. (1984). Transgenic miceharboring SV-40 T antigen genes develop characteristic braintumors. Cell 37, 367-379.

CHELLY, J., CONCORDET, J-P., KAPLAN, J-C. AND KAHN, A. (1989).

Illegitimate transcription: transcription of any gene in any celltype. Proc. natn. Acad. Sci. U.S.A. 86, 2617-2621.

CHONG, S. S. C. AND VIELKIND, J. R. (1989). Expression and fateof CAT reporter gene microinjected into fertilized medaka(Oryzias latipes) eggs in the form of plasmid DNA, recombinantphage particles and its DNA. Theor Appl. Genet. 78, 369-380.

EASTMAN, A. (1987). An improvement to the novel rapid assay forchloramphenicol acetyltransferase gene expression.BioTechniques 5, 730-732.

ETKIN, L. D. AND PEARMAN, B. (1987). Distribution, expression,and germ-line transmission of exogenous DNA sequencesfollowing microinjection into Xenopus laevis. eggs. Development99, 15-23.

FEINBERG, A. P. AND VOGELSTEIN, B. (1984). Addendum to atechnique for radio labeling DNA restriction endonucleasefragments to a high specific activity. Anal. Biochem. 137,266-267.

FIRE, A. (1986). Integrative transformation of C. elegans. EMBO J.5, 2673-2680.

FLYTZANIS, C. N., MCMAHON, A. P., HOUGH-EVANS, B. R.,

KATULA, K. S., BRITTEN, R. J. AND DAVIDSON, E. H. (1985).Persistence and integration of cloned DNA in post-embryonic seaurchins. Devi Biol. 108, 431-442.

GORMAN, C. M., MOFFAT, L. F. AND HOWARD, B. H. (1982).Recombinant genomes which express chloramphenicolacetyltransferase in mammalian cells. Mol. Cell. Biol. 2,1044-1051.

GURDON, J. B. (1987). Embryonic induction - molecular prospects.Development 99, 285-306.

GURDON, J. B. AND MELTON, D. (1981). Gene transfer inamphibian eggs and oocytes. Ann. Rev. Gen. 15, 189-218.

HAZELRIGG, T., LEVIS, R. AND RUBIN, G. M. (1984).Transformation of white locus DNA in Drosophila: dosagecompensation, zeste interaction, and position effects. Cell 36,469-481.

HINEGARDNER, R. AND ROSEN, D. E. (1972). Cellular DNA contentand the evolution of teleostean fishes. Am. Natur. 106, 311-319.

JAENISCH, R. (1988). Transgenic animals. Science. 240, 1468-1474.KARLSSON, S., HUMPHRIES, R. K., GLUZMAN, Y. AND NEINHUIS, A.

W. (1985). Transfer of genes into hematopoietic cells usingrecombinant DNA viruses. Proc. natn. Acad. Sci. U.S.A. 82,158-162.

KIMMEL, C. B. (1989). Genetics and early development of thezebrafish. Trends in Genet. 5, 283-288.

KOTHARY, R., CLAPOFF, S., BROWN, A., CAMPBELL, R., PETERSON,

A. AND ROSSANT, J. (1988). A transgene containing lac-z insertedinto the dystonia locus is expressed in neural tube. Nature 335,435-437.

KOTHARY, R., CLAPOFF, S., DARLING, S., PERRY, M. D., MORAN, L.

A. AND ROSSANT, J. (1989). Inducible expression of an hsp6S-laczhybrid gene in transgenic mice. Development 105, 707-714.

LINSK, R., GOTTESMAN, M., PERNIS, B., SOMMER, S. S. AND

SARKAR, G. (1989). Are tissues a patch quilt of ectopic geneexpression? Science 246, 261.

MACGREGOR, G. R., MOGG, A. E., BURKE, J. F. AND CASKEY, C.

T. (1987). Histochemical staining of clonal mammalian cell linesexpressing E. Coli galactosidase indicates heterogeneousexpression of the bacterial gene. Som. Cell and Mol. Genetics 13,253-265.

MORGAN, T. H. (1895). The formation of the fish embryo. J.Morph. 10, 419-472.

NEUMANN, J. R., MORENCY, C. A. AND RUSSIAN, V. (1987). A

novel rapid assay for chloramphenicol acetyltransferase geneexpression. BioTechniques 5, 444-447.

O'KANE, C. J. AND GEHRING, W. (1987). Detection in situ ofgenomic regulatory elements in Drosophila. Proc. natn. Acad.Set. U.S.A. 84,9123-9127.

OVERBEEK, P. A., LAI, S., VAN QUIL, K. R. AND WESTPHAL, H.

(1986). Tissue-specific expression in transgenic mice of a fusedgene containing RSV terminal sequences. Science 231,1574-1577.

PALMITER, R. D. AND BRINSTER, R. L. (1986). Germ-linetransformation of mice. Ann. Rev Genet. 20, 465-499.

REED, K. C. AND MANN, D. A. (1985). Rapid transfer of DNAfrom agarose gels to nylon membranes. Nucl. Acids Res. 13,7207-7221.

RUBIN, G. M. (1988). Drosophila melanogaster as an experimentalorganism. Science 240, 1453-1459.

SARKAR, G. AND SOMMER, S. S. (1989). Excess to mRNA sequenceor its protein product is not limited by tissue or speciesspecificity. Science 244, 331-334.

STERNBERGER, L. A. (1986). Immunocytochemistry. New York:Wiley.

STREISINGER, G., WALKER, C , DOWER, N., KNAUBER, D. AND

SINGER, F. (1981). Production of clones of homozygous diploidzebra fish. Nature 291, 293-296.

STUART, G. W., MCMURRAY, J. V. AND WESTERFIELD, M. (1988).

Replication, integration, and stable germ-line transmission offoreign sequences injected into early zebrafish embryos.Development 103, 403-412.

SWAIN, J. L., STEWART, T. A. AND LEDER, P. (1987). Parentallegacy determines methylation and expression of an autosomaltransgene: A molecular mechanism for parental imprinting. Cell50, 719-727.

TOWNES, T. M., LINGREL, J. B., CHEN, H. Y., BRINSTER, R. L. AND

PALMITER, R. D. (1985). Erythroid-specific expression of human^-globin genes in transgenic mice. EMBO J. 4, 1715-1723.

(Accepted 20 March 1990)