Tol2 transposon-mediated enhancer trap to identify developmentally regulated zebrafish genes in vivo



We have used the Tol2 transposable element to design and perform effective enhancer trapping in zebrafish. Modified transposon DNA and transposase RNA were delivered into zebrafish embryos by microinjection to produce heritable insertions in the zebrafish genome. The enhancer trap construct carries the EGFP gene controlled by a partial epithelial promoter from the keratin8 gene. Enhanced green fluorescent protein (EGFP) is used as a marker to select F1 transgenic fish and as a reporter to trap enhancers. We have isolated 28 transgenic lines that were derived from the 37 GFP-positive F0 founders and displayed various specific EGFP expression patterns in addition to basal expression from the modified keratin 8 promoter. Analyses of expression by whole-mount RNA in situ hybridization demonstrated that these patterns could recapitulate the expression of the tagged genes to a variable extent and, therefore, confirmed that our construct worked effectively as an enhancer trap. Transgenic offspring from the 37 F0 EGFP-positive founders have been genetically analyzed up to the F2 generation. Flanking sequences from 65 separate transposon insertion sites were identified by thermal asymmetric interlaced polymerase chain reaction. Injection of the transposase RNA into transgenic embryos induced remobilization of genomic Tol2 copies producing novel insertions including some in the germ line. The approach has great potential for developmental and anatomical studies of teleosts. Developmental Dynamics 231:449–459, 2004. © 2004 Wiley-Liss, Inc.


Several unique attributes make zebrafish a popular model to study vertebrate development and human diseases (Dooley and Zon, 2000). These include transparency of embryos, rapid embryonic development, small size, short generation time, availability of mutants, genetic and physical maps, and successful advance of the zebrafish genome sequencing project. Despite these advantages, lack of simple and reliable transgenic techniques narrows the scope of useful applications and methodologies compared with other model species. Functional studies of genes in zebrafish have been performed predominantly using chemical mutants analyzed by candidate gene approach and positional cloning. Reverse genetic approaches have been generally limited to knock-down by morpholino antisense oligonucleotides (Nasevicius and Ekker, 2000) and sense- or antisense RNA injections. A successful large-scale insertional mutagenesis project using retroviral technique to produce stable insertions in zebrafish genome has been reported (Gaiano et al., 1996), and 75 mutant lines with characterized insertion sites have been released to the community (Golling et al., 2002). One of the important lessons of this screen was that it generated a set of mutants in genes largely nonoverlapping with those identified in chemical screens, validating the use of diverse mutagenic strategies. However, this approach requires specialized expertise and facilities beyond the scope of most zebrafish biologists.

Recent evidence of effective transposition of heterologous transposons Tol2 from Oryzias latipes (Kawakami et al., 2000) and Sleeping Beauty from salmonid (Davidson et al., 2003) in the zebrafish genome opened new opportunities for reverse genetics and transgenesis-based applications. Unlike retroviral insertions, transposons can be remobilized in the presence of transposase. It is known from studies of other transposons that the majority of transpositions occur close to the donor site. Therefore, any transposon insertion in the genome can potentially be used as a donor site (“launching pad”) to produce novel insertions preferentially into the closely linked genes (Smith et al., 1996). Thus, starting from an initially small pool of transposon insertions, a larger number of insertions into the surrounding regions of the genome may be generated. Remobilization can also facilitate identification of the tagged genes. Excision of a transposon from a donor site can restore function of an affected gene associated with rescue of mutant phenotype and generation of revertants used as the supporting evidence that mutant phenotype is caused by the transposon insertion (Grossniklaus et al., 1998).

Insertional elements can carry selectable markers significantly facilitating selection of transgenics and their subsequent genetic analyses. Furthermore, they provide unique opportunities to identify mutations affecting the development of gametes by quantitative analysis of mutant allele transmission to the next generation (Springer et al., 1995). Because the sequence of the insert is known, identification of the gene affected by the insertion is relatively straightforward, compared with positional cloning of N-ethyl-N-nitroso-urea mutants. Loss-of-function mutations of many genes do not produce obvious morphological phenotypes; therefore, they are usually omitted in classic mutant screens. Reverse genetics using insertional mutagenesis will provide valuable material to address this issue.

Yet another advantage of the insertional approach that has been widely used recently in several model species is its adaptation for gene and enhancer trapping. Gene trap constructs, for example, should contain multiple splicing acceptor sequences upstream of a promoterless reporter gene coding sequence to allow gene–reporter fusion upon insertion into the gene in the same orientation. An enhancer trap construct carries a reporter gene with minimal promoter that is induced once it lands under control of a chromosomal enhancer (Bellen, 1999). Enhancers can activate the transcription at considerable distance in orientation-independent manner; therefore, the approach is the most effective for monitoring gene activity throughout the genome. Because enhancers can activate the reporters outside of genes, the subsequent identification of the regulated gene can be more difficult. However, in many reported cases, enhancer trap lines did reflect the expression pattern of the genes correctly (Perrimon et al., 1991; Bellen, 1999).

In this study, we report the successful application of an enhancer trap approach in zebrafish using a modified Tol2 element from medaka (Kawakami et al., 2000; Koga et al., 2002). This element uses conservative (“cut and paste”) transposition mechanism, i.e., it is excised from the donor site when the whole element moves to the new position. We chose the enhanced green fluorescent protein (EGFP) as a reporter to use the optical clarity of live zebrafish embryos.

By using this technique, GFP-positive transgenic families (generally containing multiple transposon insertions) from 37 founders were raised and analyzed. Sixty-five sequences flanking the insertion sites were identified using the thermal asymmetric interlaced polymerase chain reaction (TAIL-PCR). We isolated 28 enhancer trap (ET) lines that exhibited distinct EGFP expression patterns. We also demonstrated that remobilization of Tol2 integrated in zebrafish genome can produce novel insertions.


Design of the ET Element

A simple ET element must carry a reporter gene with weak or minimal promoter and should be able to integrate into the genome at random positions (Bellen, 1999). Once such a construct is inserted into genomic DNA near a chromosomal enhancer, it can produce tissue-specific expression of a reporter gene.

We have modified a Tol2 transposable element from medaka (Kawakami et al., 2000) for use as an ET in zebrafish. The ET element carries the EGFP gene driven by a 460-bp promoter from the zebrafish keratin 8 (krt8) gene (Gong et al., 2002; Fig. 1). The incomplete krt8 promoter retains weak expression in epithelial tissues and can readily be used as a selective marker for the presence of the transgene. The Tol2 construct contains a partially deleted transposase gene to render it inactive but has intact 5′ and 3′ ends that are required for transposition. We coinjected the enhancer trap construct DNA with in vitro synthesized Tol2-transposase mRNA into zebrafish embryos at one- to two-cell stage (Kawakami et al., 2000). In this scheme, transposition occurs as long as there is active transposase protein in the cells. Transpositions into genomic DNA that occur in the germ-line cells will be transmitted to the subsequent generations. Insertions transmitted to the F1 generation are stable, because there is no transposase in the cells.

Figure 1.

A: Schematic representation of enhancer trap (ET) construct. The Tol2 element contains the EGFP reporter gene under control of a short 460-bp krt8 promoter. Expression of the reporter is induced when it is inserted near an enhancer. B: Evaluation of the Tol2 insertion number in different F1 fish by using Southern blot hybridization. The fish showing ET patterns were randomly picked from different F1 families, out-crossed to the wt, and DNA from 20 green fluorescent protein (GFP) -positive embryos were used for Southern blot analysis. The DNAs were digested with EcoRI and hybridized with digoxigenin (DIG) -labeled probe for the EGFP sequence. Lane 1, (size marker) 1-kb DNA ladder (New England BioLabs); lane 2, GFP-negative control; lanes 3–11, ET3, ET4, ET11, ET-15-1, ET22-2, ET9-2, ET17, ET10, and ET10, respectively. The F1 fish demonstrating single inserts were out-crossed again to give rise to single-insertion ET lines. For example, fish represented on lane 11 has a single insert corresponding to ET10, compared with another fish from the same family that carries two inserts (lane 10). Line ET11 has two inserts in the same gene only 4 kb apart; therefore, they cannot be separated by out-cross. EGFP, enhanced GFP.

The majority of injected fish showed mosaic clones of GFP expression by 10–24 hr mainly in the cells of the enveloping layer (EVL). At 5 days, only 20–40% of injected fish showed GFP fluorescence in surface cells and also in various cells inside the embryo, while the remaining fish looked GFP negative by this stage. The germ-line transmission in this population was ∼30% (11 of 35 preselected F0 fish), higher then average 16% (see the next paragraph), suggesting that such preselection may be useful to minimize amount of labor required for the subsequent F1 screening.

Analysis of Expression in F1 Generation

A total of 230 injected fish were raised to adulthood and out-crossed to the wt fish, and their progeny were analyzed for GFP fluorescence. We obtained 37 founder fish (F0) that transmitted the actively expressed EGFP gene to their offspring (F1), corresponding to 16% germ-line transmission. These data are in line with the previous report by Kawakami et al. (2000) (12%; one of eight) and similar to the efficiency of Sleeping Beauty element in zebrafish (10% for 5-kb construct and 31% for a 2-kb construct) reported by Davidson et al. (2003). The ratio of fluorescent embryos among their offspring usually ranged from 0.2 to 40%, reflecting the mosaicism of the F0 germ line. Due to space limitations, we grew the GFP-positive progeny of each founder in single tanks (maximum 30–40 fish in per 2-liter tank) and identified them as F1 families. Many of these families contained more than one insertion. Of the 37 families raised, two were lost (all GFP-positive progeny of founders #12 and #18 died before 14 days postfertilization [dpf]; both female founders died too). All GFP-positive F1 fish exhibited various levels of GFP fluorescence early in cells of the EVL and later in skin epithelia, which were likely reflective of the activity of the krt8 minipromoter. Transgenic fish from 13 F1 families demonstrated this expression pattern alone. In these fish, GFP fluorescence was not detected before late gastrulation, became brightest at 24 hr, and drastically decreased at 3–4 days. In situ RNA hybridization usually did not detect EGFP RNA after 48 hr; however, EGFP protein was still detected by immunohistochemistry (data not shown).

Among the fish from the other 24 F1 families, we identified and isolated 28 independent lines that, in addition to the krt8-like pattern described above, exhibited novel distinct EGFP expression patterns (ET lines; Fig. 2, Supplementary Table 1, which can be viewed at Because the GFP fluorescence in the skin epithelia was usually quite weak, it did not hinder observation of specific expression in the internal organs. These results strongly suggest that our construct worked effectively as an enhancer trap. Each of the 28 expression patterns segregated as a single locus when out-crossed to wild-type, indicating that it was produced by a single insert or a limited number of very closely linked inserts. Some of these lines also contained additional loci that produce only krt8-specific expression.

Figure 2.

A–C: Enhanced green fluorescent protein expression patterns in the spinal cord, mid-trunk lateral view (A), brain, dorsal view (B), and neuromast (C). Anterior is to the left. ot, optic tectum; hb, hindbrain; dc, diencephalon. Scale bars = 40 μm in A, 100 μm in B, 20 μm in C.

GFP fluorescence was detected in a variety of tissues and organs, including the central nervous system (CNS), neural crest and its derivatives, notochord, heart, muscles, digestive organs, kidney, and so on (Supplementary Table S1). Specific expression of the reporter gene was often localized to single cells or cell groups easily distinguishing these cells from their neighbors (Fig. 2). For example, in two different lines, two distinct cell types contributing to the neuromast of the lateral line were tagged, which allowed us to examine three-dimensional structure of the neuromast (Fig. 2C). In the ET20, the external mantle cells of the neuromast were labeled, whereas in the ET4 line, the expression of the reporter gene was restricted only to the internal hair cells. Crossing of these lines allowed visualizing of both cell types simultaneously, demonstrating the relative positions of these cells (Fig. 2C).

Specific EGFP expression was detected at various stages of development from the embryogenesis to adulthood. However, because screening of F1 for ET patterns was usually performed until 5 dpf, we may have overlooked ET lines with only late (after 5 dpf) expression patterns. Strong maternal expression of EGFP was observed in several enhancer trap lines. All the eggs of the heterozygous female inherited the maternal EGFP. When such female was out-crossed to a wild-type fish, GFP fluorescence was detected in 100% of the offspring up to 48 hr. However, after 24 hr, maternal EGFP levels were decreased and the difference between Tol2-positive and -negative fish became obvious, allowing the selection of transgenic progeny. PCR with EGFP-specific primers did not detect the gfp gene in embryos showing only maternal expression.

We also analyzed pools of GFP-fluorescence–negative offspring from 11 of the 37 founders and 12 randomly selected GFP-negative F0 fish using PCR with EGFP-specific primers and confirmed that they were all negative for the EGFP gene (not shown). In two F1 families (#1 and #8), GFP fluorescence was extremely faint; however, it allowed accurate segregation analysis. With these two lines, selection has to be performed before 48 hr because krt8-like expression EGFP was hardly detectable at later stages.

Analysis of the Insertion Sites

We used TAIL-PCR (Liu and Whittier, 1995) to amplify genomic sequences flanking Tol2 insertions in the F1 families. Genomic DNA for TAIL-PCR was extracted from 3 to 10 pooled GFP-positive embryos for each F1 family that presumably contained multiple inserts. Our primary strategy was to select and propagate only fish lines with interesting ET patterns or insertions into interesting genes, considering the resources of a small to mid-size lab and limited fish facility space. Thus, already during F0 screening we could focus our efforts on a limited number of selected insertions and minimize the fish space required for raising the transgenic lines. Approximately 5 to 20 TAIL-PCR products per family were usually amplified. The size of the amplified fragments varied from 250 bp to 3,000 bp. Because we only ran one sequencing reaction using the TAIL-PCR primers specific to the 5′- and 3′-ends of Tol2, the sequences rarely exceeded 1,000 bp. The flanking sequences included 30–35 bp of Tol2-end sequences that were used to confirm the specificity of an amplified fragment.

The majority of Tol2 insertions were flanked by zebrafish genomic DNA and only five families contained the original vector sequence surrounding Tol2. Hence, Tol2 mostly inserted into genomic DNA using the specific transposition mechanism and not by random DNA integration. Only 8% (5 of 65 flanking sequences) of insertions were nontransposase-mediated, compared with 50% (1 of 2; selection was too small, and direct PCR detection test was possibly more accurate) from the original report by Kawakami et al. (2000) and 11% (1 of 9) for Sleeping Beauty element (Davidson et al., 2003). Over 80% (27 of 31) of insertions with both 3′- and 5′-flanking sequences identified were surrounded by the classic 8-bp target site duplication. In one case, a 12-bp deletion of the 5′-end of Tol2 end sequence was detected, whereas for all other insertions, the transposon end sequences were intact. We used TAIL-PCR results to estimate the approximate number of insertions. It suggested that, on average, two to three copies of Tol2 were present per F1 family. Notice that TAIL-PCR is not a quantitative method, and the actual number of insertions identified in the families is likely underestimated. However, Southern blot hybridization with randomly picked F1 fish shows that, already after out-cross of the founders, the majority of fish contained only one or two inserts (Fig. 1B). A maximum of six insertions was detected by TAIL-PCR in one family (#7). To evaluate the distribution of insertions in the genome and predict the candidate tagged genes, the flanking sequences were analyzed by using the “Ensembl” database (Zv3 assembly). However, this version still contains multiple misassemblies and gaps; therefore, our current analysis of position of insertion sites may have errors, but it can be revised and validated later, when a quality of “Ensembl” data improve.

Only flanking sequences that were identical to “Ensembl” sequences were analyzed, and matches with less than 95% identity were omitted. Insertions into repeats were more difficult to interpret, because they produced many high-score BLAST alignments with minor differences. Such insertions were also not analyzed further. Wherever possible, inserts were mapped onto chromosomes (Supplementary Table S1). A maximum of five independent inserts were found on the chromosome 5. They were evenly distributed along the chromosome with a minimal distance of 2.7 Mbp between adjacent insertions. Furthermore, insertions from independent F1 families were not found in the same Ensembl contigs with one exception: insertions in F1 families #25 and #34 were only 200 kb apart in ctg22447, which may be an indication of a “hot spot.” Presence of such “hot spots” can be a drawback for a large-scale screen. Nevertheless, the data available so far suggest that there are likely no major transposition hot spots that can be a serious disadvantage for small to medium scale screens. Larger number of insertions has to be obtained and analyzed to address this issue.

In two instances, we also identified pairs of closely linked insertions within the same family, which were separated only by 0.2 and 4 kb in ET5 and ET11 and segregated together as a single locus. By analogy to other transposable elements, we assume that these linked pairs likely resulted from retransposition. Thus, although Tol2 is a cut-and-paste transposon, transposition shortly after replication of a donor site can give rise to a chromatid carrying two transposons in both recipient and donor sites (Chen et al., 1992). Moreover, the recipient site of other transposons is frequently found close to the donor site (Smith et al., 1996).

It has been demonstrated for other transposons like Drosophila P element and Ds element in Arabidopsis that insertions preferentially occur at the 5′-end of the transcribed regions, increasing the frequency of gene tagging (Spradling et al. 1995; Parinov et al., 1999). So far, our data do not suggest such a preference of Tol2 element in zebrafish. Only 8 of 42 “Ensembl” database-matching insertions, including 20 insertions into genes, have been found in the vicinity of 5′-gene ends (Table 1). However, due to the limited scale of our screen, these conclusions should be considered as preliminary.

Table 1. Screen Overview*
  • a

    Only ET lines showing specific GFP expression before 5 dpf were selected.

  • b

    Based on the exon–intron structures predicted by the “Ensembl” database Zv3.

  • *

    ET, enhancer trap; GFP, green fluorescent protein; UTR, untranslated region; dpf, days postfertilization.

F0 fish screened230
Transgenic F1 families37
 Transgenesis rate(16%)
ET linesa28
 ET ratio of total F0fish(12%)
 ET ratio of transgenic F0fish(75%)
Flanking sequences65
 Perfect BLAST match42
 No BLAST match18
 Vector sequence5
Insertions in intronsb16
Insertions in coding sequences1
Insertions in 3′ UTRs2
Insertions near of 5′ ends8

EGFP Expression Patterns in Enhancer Trap Lines Reflect the Tissue-Specific Expression of the Tagged Genes

To investigate whether EGFP expression in the ET lines reflects the expression patterns of the tagged genes, we performed whole-mount in situ hybridization with RNA probes for a few candidate genes using the EGFP RNA probe as a control. The ET33 line carried a single Tol2 insertion in the 3′UTR (50 bp downstream of the stop codon and 800 nucleotides upstream of the polyA signal) of the zic6 gene (GenBank accession no. AY572956), which has not been characterized previously in vertebrates. GFP fluorescence in this line was detected in the neural tube through the entire rostrocaudal extent with higher expression detected in the hindbrain, cerebellum, midbrain, and ventral forebrain (Figs. 2A, 3A, 4A(−Tpase)).

Figure 3.

A,B: Comparison of enhanced green fluorescent protein (EGFP) RNA and the endogenous transcript expression in ET33 (A1,2) and ET7 (B). Whole-mount RNA in situ hybridization with EGFP probe in ET33 (A1) and zic6 probe in wild-type embryo (A2). A3: Whole-mount immunohistochemistry using anti-GFP antibodies. A4: GFP fluorescence in ET33 embryos in vivo.

Figure 4.

A: Transposase injection into ET33 line (containing insertion into 3′-untranslated region of zic6) produces somatic transposition that is illustrated by corresponding muscle-specific expression pattern. An example of F1 fish that carry a retransposed Tol2 element is also shown. In contrast to the ET33 parent, this fish demonstrates strong expression in the flour plate and no expression in other brain regions. B: Evidence of germ-line transposition. Integration sites corresponding to three different fish from the offspring of the injected ET33 and the original sequence are shown. Lower case indicates Tol2 end sequences. The classic 8-bp direct repeat of target DNA associated with transposition is in boldface. C: Footprint sequences of the donor site after excision of the Tol2. wt, wild-type.

The expression of zic6 was restricted to the most medial cluster of cells in the dorsal neural tube, occupying the roof plate of the spinal cord, the optic tectum in the midbrain, and the most posterior rows of cells in the cerebellum (Fig. 3A-2). Expression of EGFP RNA (Fig. 3A-1) is practically identical to the expression of zic6 RNA (Fig. 3A-2). The GFP fluorescence in the ET33 and anti-GFP immunohistochemical staining showed broader distribution of GFP protein in comparison to that of RNA. The EGFP-positive cells occupied the entire dorsal area of the optic tectum, cerebellum, and also spread to the lateral sides of the hindbrain and the spinal cord (Fig. 3C,D). Unlike the RNA, which is present mainly in the perinuclear cytoplasm, EGFP protein is distributed evenly in the cytoplasm reaching distant cellular projections and revealing their complexity. This finding demonstrates the advantage of detecting the transgene at the protein level rather than RNA level in cells of complex architecture.

Out-cross of the ET7-1 line produced 50% GFP-positive offspring with identical expression, suggesting that it carried a single insertion. It was confirmed by Southern blot hybridization with GFP-specific probe (Supplementary Fig. S1). Analysis of the flanking sequences obtained by TAIL-PCR revealed that the insertion was localized within an intron of a novel gene, corresponding to the expressed sequence tag (EST) clone fd54c09 (AW019543; Clark et al., 2001). This EST encodes a zebrafish homologue of the human solute carrier SLC41A1 (Wabakken et al., 2003), which is 80% identical to the human gene at the amino acid level. The expression of this gene was detected in the olfactory bulb, ventral midbrain, anterior hindbrain, liver, pharynx, and gut (Fig. 3B). Of interest, this gene was also expressed in the small spherical glandular bilateral organs in the posterior part of the trunk above the pronephric ducts (shown by arrowheads, Fig. 3B). These most probably represent the corpuscles of Stannius, a little studied teleost-specific endocrine organ, involved in regulation of Ca2+ balance (Matty, 1985). Between 36 and 72 hr, this gene is predominantly expressed in the liver, reaching maximal level at 60 hr. At 4 days, its expression became almost undetectable in the liver, while the expression in the stomach and intestine significantly intensified and remained high for at least a week. The EGFP expression pattern matched the pattern of the cDNA clone fd54c09 in most details, excluding the spinal cord (Fig. 3B). Furthermore, GFP expression in the brain is more restricted compared with that of fd54c09 RNA. Thus, the enhancer trap approach is useful in revealing the details of anatomical organization in a zebrafish embryo that are not always possible by RNA hybridization. Furthermore, because it allows observations in vivo, it eliminates expensive and laborious whole-mount in situ hybridization and fulfills requirements ideal for physiological study.

Genetic Analysis of F2 Generation

F2 fish were raised and out-crossed to wild-type fish to identify carriers of single insertions. If 50% of embryos were GFP-positive, that finding would correspond to a single locus containing one insertion or a limited number of closely linked inserts, whereas higher ratios would suggest presence of unlinked transposon insertions. This test would also indicate mutations affecting genes required for the development of haploid gametes, insertions in which would result in lower than 50% EGFP segregation ratios. All out-crosses yielded 1:1 or higher GFP segregation ratios. Several F2 fish with multiple insertion were out-crossed to wild-type fish and similar segregation analysis was repeated in the F3 generation, but no fish with segregation ratios lower than 1:1 were recovered. Genetic analysis of the ET lines was significantly simplified: only fish with the specific ET patterns characteristic for each line was counted, fish with skin-specific expression were considered as ET-negative. Of the 28 ET lines, 18 did not produce any additional GFP patterns that were different from the parental pattern and segregated in the 1:1 ratio (ET pattern:GFP negative) when out-crossed to the wild-type fish. Another 10 ET lines still contained unlinked insertions with krt8-like EGFP expression, but the ET pattern segregated in a 1:1 ratio (ET pattern:krt8-like pattern plus GFP negative).

All 28 ET lines and 19 single insertion non-ET F2 lines were further analyzed by in-crossing. These crosses produced Mendelian segregation ratio (3:1) of EGFP (single insertion lines) or ET pattern (ET lines). Five in-crosses produced early developmental mutant phenotypes with Mendelian segregation (1 mutant:3 wild-type). However, only in one line (ET22-2) the recessive mutant phenotype cosegregated with EGFP. All 695 mutant offspring from the in-cross expressed EGFP, whereas 705 of 2,137 embryos with wild-type phenotype were EGFP-negative, indicating that the GFP transgene and the mutant locus are linked. Thus, analysis of GFP segregation is a convenient first step in filtering out a large number of unlinked mutations. However, this test only shows low level of linkage. To demonstrate tight linkage of an insertion and a phenotype, mutants must be shown to be invariably homozygous (Singer et al., 2002).

Remobilization of the Genomic Tol2 Insertions Produce Novel Insertions That Can Be Transmitted to the New Generation

To test whether the Tol2 insertions can be used as the donor sites for novel transpositions, we injected transposase mRNA into the F2 embryos of ET33 line that were heterozygous for the Tol2 insert (Fig. 4). By 48 hours postfertilization, we observed new expression patterns in somatic cells of the injected embryos, suggesting that the transposon had jumped to the new genomic locations (Fig. 4A). We also identified fish with novel expression patterns (different from the ET33 founder) appearing with various frequencies among the offspring of the injected fish, suggesting that some transpositions occurred in the germ line (Fig. 4A–F1). Of 128 F0 fish, 12 produced progeny with new expression patterns different from the original ET33 parental one (including three krt8-like lines and nine with novel ET), corresponding to a 10% retransposition rate. Notice, that new insertions remaining under the influence of the same enhancer as in ET33 cannot be detected in such simple scheme; therefore, the actual retransposition rate can actually be higher. Sequencing of TAIL-PCR products from all these fish confirmed the generation of novel integration sites that were not found in the original ET33 fish (three examples are shown in Fig. 4B). Insertion of Tol2 is usually accompanied by 8-bp target site duplication. The excision of a transposon may not necessarily remove the duplicated sequence, leaving footprints of various lengths. We amplified and sequenced the “donor site” fragment from the 12 independent F1 fish carrying retransposed Tol2 elements (one embryo from each of the 12 founders). Because ET33 F0 founders used in this experiment were heterozygous for the donor site, the PCR products should contain equimolar amounts of wt DNA and “restored” donor site, that would produce a mixed electropherogram starting from the footprint (if there were any). Of 12 sequences, four contained footprints (Fig. 4C). The other eight sequences were wild-type, which may be an indication of “precise excision” that occurred with retransposition. It is also possible that, in all these eight fish, the new insertion sites segregated away from the donor. The “precise excision” could be demonstrated unambiguously only if homozygous donors were used for retransposition. One of the excision footprints that we identified (Fig. 4C) was six nucleotides long. Such footprints do not produce frame shifts and may sometimes restore the function of the gene product interrupted by transposon after it is excised.


A variety of cis-regulatory elements control tissue-specific gene expression in eukaryotes. Enhancers bind sequence-specific transcription factors and can activate genes over long distances and different elements of a promoter can alter the promoter-enhancer specificity (Arnosti, 2001). Distinct promoters possess different abilities to detect enhancers (Perrimon et al., 1991). Hence, the research objectives should govern the choice and design of the promoter–reporter gene construct to increase the success rate of an enhancer trap project. The promoter of the keratin8 gene drives the expression in tissues of ectodermal origin (Gong et al., 2002). Because the major focus of our laboratory is on neural development, we expected that such a promoter would be more effective for tagging CNS-specific enhancers. In fact, the CNS became the main target for the reporter expression, but it was not the only target. One of the insights gained from our analysis was that this construct effectively detected expression patterns in tissues derived from all three germ layers: ectoderm, endoderm, and mesoderm. In vertebrates, the majority of genes are expressed in the nervous system (Gawantka et al., 1998), which is in line with the observed preferential expression of the reporter gene in the CNS in our screen. Because the 460-bp krt8 promoter partially retains the specificity of the original promoter, it also can be used as a selection marker for presence of the transgene, thus eliminating the need for an additional marker gene in the transposon construct.

Enhancer traps often reflect only a subset of expression domains of a tagged gene (Casares et al., 1997; Hatini and DiNardo, 2001). The gene for cDNA clone fd54c09 (Fig. 3B) and acinus (data not shown) were expressed in the brain at the RNA level more broadly compared with the expression pattern of the EGFP reporter in the corresponding enhancer trap lines. However, it brings an unexpected benefit, because the simplified patterns allow analysis with single-cell resolution, thereby revealing novel structural and developmental details, which may be obscured by complex patterns of expression. For example, this approach allowed distinguishing the network of cells in the spinal cord (Fig. 2A). In two different lines, two distinct cell types contributing to the neuromast of the lateral line were labeled (Fig. 2C), revealing its three-dimensional structure. In the ET20, EGFP tags the external supporting cells of the neuromast, while in the ET4 line, the expression of the reporter gene is restricted only to the internal hair cells. The GFP expression patterns also allow visualization of specific tissues and organs that are difficult to observe even in the transparent zebrafish embryo. In the ET7 line and two other lines, GFP was detected at 48 hr in the presumptive corpuscles of Stannius (Fig. 3B). To our knowledge, this organ has not yet been described in zebrafish literature; therefore, further research is required to verify its identity. This organ is practically undetectable in the embryo by using Nomarski optics, but because it maintains the GFP expression until adult stage, it can now be analyzed in detail. Furthermore, having three transgenic lines expressing the reporter in this organ, it should be possible to analyze the genetic mechanisms involved in its formation and function. The specific expression patterns of the reporter gene in the ET lines also simplify genetic analyses when working with multiple insertion lines, because it allows distinguishing individual insertions visually.

Transposons are being successfully used for insertional mutagenesis in various model species. However, because coding sequences occupy only a small portion of zebrafish genome, the efficiency of inactivating genes by random mutagenesis is expected to be quite low, although insertions into introns and untranscribed regions can sometimes affect gene function by various mechanisms, including aberrant splicing and changed transcript levels. Furthermore, the gene redundancy would reduce the number of phenotypically manifesting mutations even further. Our results demonstrated that the great majority of random insertions of the 5-kb-long construct (including 21 insertions inside genes) did not visibly affect early development (only 1 of 47 inbred lines exhibited recessive mutant phenotype). One strategy to resolve this problem is to produce multiple insertions by increasing the insertion frequency (Gaiano et al., 1996; Golling et al., 2002). However, the subsequent genetic analysis and especially the maintenance of multiple insertion lines can be complicated and labor intensive.

There is potentially another way to address the efficiency problem of random mutagenesis. Unlike retroviral insertions, transposons can be remobilized to another genomic location once the transposase is delivered into the cell. To this point, we have demonstrated that genomic Tol2 insertions could be remobilized. Because retranspositions occur preferentially to the closely linked sites they are expected to hit genes closest to the original insertion with much higher efficiency than random mutagenesis (Smith et al., 1996). Therefore, any transposon in the genome can become a potential donor site for effective mutagenesis of the surrounding genes. Moreover, transposition can induce deletions of various lengths, thus increasing their mutagenic activity with respect to surrounding genes (Shalev and Levy, 1997; Xiao et al., 2000). Currently, analysis is in progress to evaluate whether these events take place in our lines.

Gene trap mutagenesis produces higher mutagenicity rates and can be used to downscale the number of progeny that should be raised for the mutant screen by increasing the selective pressure (Stanford et al., 2001). However, it has lower trapping efficiency. In contrast, our screen demonstrated that the enhancer trap construct could produce high trapping frequency. High efficiency and high specificity of the transposon-mediated enhancer trap approach might be beneficial for further development of inducible expression systems such as the GAL4/UAS system that will allow such experiments as controlled gain-of-function, misexpression, and targeted cell ablation (Brand and Perrimon, 1993; Lin et al., 1995; Koster and Fraser, 2001; Scheer et al., 2001).

Finally, our screen has demonstrated an opportunity to gain novel information from a small-scale screen, which would be appropriate in a small to midsize laboratory. Thus, enhancer trap insertional mutagenesis is a very useful tool for random unbiased exploration of the zebrafish genome.


Zebrafish was maintained according to established protocols (Westerfield, 1995).

Constructs and Injections

We have cloned the Tol2 transposon from medaka DNA by using PCR primers designed based on Tol2 sequence (GenBank accession no. D84375). Cloning of Tol2 was performed in two steps using PCR primers: CATGCGGGCCCAGAGGTGTAAAGTACTTGAGTA and GCGCAAGCGGCCGCTTGAGACTAGGTTAAGTA for the 5′ end of the element and CGCGCTCGAGCAGAGGTGTAAAAAGTACTCAA and ATATAGGCGGCCGCCTGTGTTTCAGACACCA for the 3′ end. The first PCR product was cloned into pBK-CMV (Stratagene, USA) using ApaI and NotI restriction sites incorporated into the primers, followed by cloning the 3′-end product into the construct using NotI and XhoI restriction sites so that both PCR fragments were joined at the internal NotI site. The construct containing EGFP (Clontech Laboratories, USA) under 2.2-kb promoter of keratin 8 (krt8) gene (GenBank accession no. AF440690) was obtained from Dr. Zhiyuan Gong from the National University of Singapore. A fragment containing the EGFP gene and only 460 bp of the krt8 promoter was amplified using this plasmid as a template and primers ACAATGCAACTGTTCAGCTCA and ATGGCTGATTATGATCTAGAG. The product was digested by NotIand cloned into the Tol2 construct between EcoRV and NotI sites. The final enhancer trap construct (Fig. 1) consists of 2,671 bp of Tol2 5′ end, 529 bp Tol2 3′ end, and the mini-krt8 promoter-EGFP fragment (1,200 bp long). Transposase cDNA was cloned from medaka cDNA using the primer: ACGTGAGCTCACATCTATTACCACAATGCAC. We used mMESSAGE mMACHINE T7 kit (Ambion) for synthesis of in vitro transcribed capped transposase RNA and RNeasy Mini Kit (QIAGEN, Germany) for RNA cleanup. A total of 5–10 pg of plasmid DNA with 25–50 pg of in vitro synthesized transposase mRNA were coinjected into zebrafish embryos at the one- to two-cell stage. The actual concentration of RNA was empirically adjusted to produce 50% embryo survival rate.



The following primer mixtures (containing 1.5 μM specific primer and 10 μM AD primer) were prepared: for primary PCR: Toil5′-1/AD-3, Toil5′-1/AD-5, Toil5′-1/AD-6, Toil5′-1/AD-11, Toil3′-1/AD-3, Toil3′-1/AD-5, Toil3′-1/AD-6, Toil3′-1/AD-11; for secondary PCR: Toil5′-2/ AD-3, Toil5′-2/AD-5, Toil5′-2/AD-6, Toil5′-2/AD-11, Toil3′-2/ AD-3, Toil3′-2/AD-5, Toil3′-2/AD-6, Toil3′-2/AD-11; for tertiary PCR: Toil5′-3/ AD-3, Toil5′-3/AD-5, Toil5′-3/AD-6, Toil5′-3/AD-11, Toil3′-3/AD-3, Toil3′-3/AD-5, Toil3′-3/AD-6, Toil3′-3/AD-11. A total of 4 μl of primer mixtures were added to PCR reaction (total volume 20 μl).

Cycle settings were as follows. Primary: (1) 94°C, 2 min; (2) 94°C, 30 sec; (3) 62°C, 1 min; (4) 72°C, 2.5 min; (5) go to “cycle 2” 4 times; (6) 94°C, 30 sec; (7) 25°C, 3 min; (8) ramping 0.3°/sec to 72°C; (9) 72°C, 2.5 min; (10) 94°C, 10 sec; (11) 61°C, 1 min; (12) 72°C, 2.5 min; (13) 94°C, 10 sec; (14) 61°C, 1 min; (15) 72°C, 2.5 min; (16) 94°C, 10 sec; (17) 44°C, 1 min; (18) 72°C, 2.5 min; (19) go to “cycle 10” 14 times; (20) 72°C, 5 min.

A total of 2 μl of the primary reaction was diluted in 25 μl of water and 2 μl of the mixture was added to the secondary reaction. Secondary: (1) 94°C, 10 sec; (2) 61°C, 1 min; (3) 72°C, 2.5 min; (4) 94°C, 10 sec; (5) 61°C, 1 min; (6) 72°C, 2.5 min; (7) 94°C, 10 sec; (8) 44°C, 1 min; (9) ramping 1.5°/sec to 72°C; (10) 72°C, 2.5 min; (11) go to “cycle 1” 14 times; (12) 72°C, 5 min.

A total of 2 μl of the secondary reaction was diluted in 25 μl of water, and 2 μl of the mixture was added to the tertiary reaction. Tertiary: (1) 94°C, 15 sec; (2) 44°C, 1 min; (3) ramping 1.5°/sec to 72°C; (4) 72°C, 2.5 min; (5) go to “cycle 1” 29 times; (6) 72°C, 5 min.

Products of the secondary and tertiary reactions were separated by using 1.8% agarose gel. The individual bands from the “band shift” pairs were cut from the gel and purified by using QIAquick Gel Extraction Kit (QIAGEN, Germany), and sequenced by using ABI Cycle Sequencing chemistry (PE Applied Biosystems, CA) and an ABI Prism 310 Genetic Analyzer with Data Collection Software (PE Applied Biosystems, Foster City, CA) supplied by the producer.

Analysis of Flanking Sequences

Flanking sequences were analyzed using BLAST and the current assembly (Zv3) of “Ensembl” database (Clamp et al., 2003) available at and NCBI database (vector sequences were detected using this database). We used the exon–intron prediction for candidate genes directly from Ensembl database, except for a few cases when the cDNA sequence was available. Sequences with less than 95% identity to database sequences were assumed “no match” and were not analyzed further.

RNA In Situ Hybridization

Whole-mount in situ hybridization using antisense RNA probes labeled with digoxigenin (Roche, USA) was carried out as previously described by Oxtoby and Jowett (1993). Zic6 and fd54c09 cDNA templates, used to generate RNA probes, were amplified using OneStep RT-PCR kit (Qiagen, Germany) with the following pairs of primers correspondingly: zic5′-GGGACAAATCTGTCAGCAGCA and T7zic3′-ATTGTAATACGACTCACTATAGGTATGCCACAAACCTATCAACT; psc5′-CACATACAGACATCATCTTTGCAC and T7psc3′-ATATAATACGACTCACTATAGGGATGCAGCCAGTGGCTGTAA. After in situ hybridization, stained embryos were mounted in phosphate buffered saline–glycerol and viewed under AxioPlan 2 microscope (Zeiss, Germany). Images were taken with AxioCam digital camera (Zeiss, Germany) and processed using Adobe Photoshop 5.5 software.

Southern Blot Hybridization

After digestion by BamHI or EcoRI, DNA was transferred to positively charged nylon membrane (Roche Applied Science) by capillary blotting (Sambrook et al., 1989) and cross-linked by ultraviolet irradiation.

The DNA probe for EGFP was labeled with digoxigenin (Roche Applied Science) using PCR digoxigenin (DIG) synthesis kit. We used DIG EasyHyb DIG Wash and Block Buffer Set for hybridization, alkaline phosphatase labeled anti-DIG antibody, and CDP-Star chemiluminescent substrate (Roche Applied Science, USA) for detection of the hybridized probe. Hybridization and detection was carried out as described in the manufacturer's User's Guide.

Analysis of Tol2 Excision Sites

Heterozygous ET33 embryos were injected with 25–50 pg of transposase mRNA and raised to maturity. The adult fish were out-crossed to the wt, and their progeny (F1) were analyzed for new GFP patterns. Individual embryos carrying retransposed element (1 from each of the 12 founders) were used for PCR with primers on either site of the original insertion: GAGTCAGCTCAGTACTCATG and TACTTCATCTGCCGGGTCTT.

The PCR product (464 bp in the wt) was sequenced in both directions. Because founders were heterozygous for the donor, the individual progeny containing retransposed element could be either heterozygous for the “recovered” donor site or may not contain the donor site due to segregation. Fish containing a footprint produced a mixed sequencing electropherogram starting from the footprint. Because there is no difference between wt sequence and the product of “precise precision,” obtaining the wt sequence in such test could not be directly used to confirm “precise excision” (it could be due to segregation).


We thank Dr. Zhiyuan Gong for providing krt8 promoter-EGFP construct, Megan Griffith and Karuna Sampath for comments on the manuscript, Stephen Johnson for fruitful discussion and suggestions, Tatiana Kolesnik for technical help, the fish communities of The Institute of Molecular and Cell Biology and Temasek Life Sciences Laboratory for support.