Enhancer trapping in zebrafish using the Sleeping Beauty transposon
© Balciunas et al. 2004
Received: 05 August 2004
Accepted: 03 September 2004
Published: 03 September 2004
Skip to main content
© Balciunas et al. 2004
Received: 05 August 2004
Accepted: 03 September 2004
Published: 03 September 2004
Among functional elements of a metazoan gene, enhancers are particularly difficult to find and annotate. Pioneering experiments in Drosophila have demonstrated the value of enhancer "trapping" using an invertebrate to address this functional genomics problem.
We modulated a Sleeping Beauty transposon-based transgenesis cassette to establish an enhancer trapping technique for use in a vertebrate model system, zebrafish Danio rerio. We established 9 lines of zebrafish with distinct tissue- or organ-specific GFP expression patterns from 90 founders that produced GFP-expressing progeny. We have molecularly characterized these lines and show that in each line, a specific GFP expression pattern is due to a single transposition event. Many of the insertions are into introns of zebrafish genes predicted in the current genome assembly. We have identified both previously characterized as well as novel expression patterns from this screen. For example, the ET7 line harbors a transposon insertion near the mkp3 locus and expresses GFP in the midbrain-hindbrain boundary, forebrain and the ventricle, matching a subset of the known FGF8-dependent mkp3 expression domain. The ET2 line, in contrast, expresses GFP specifically in caudal primary motoneurons due to an insertion into the poly(ADP-ribose) glycohydrolase (PARG) locus. This surprising expression pattern was confirmed using in situ hybridization techniques for the endogenous PARG mRNA, indicating the enhancer trap has replicated this unexpected and highly localized PARG expression with good fidelity. Finally, we show that it is possible to excise a Sleeping Beauty transposon from a genomic location in the zebrafish germline.
This genomics tool offers the opportunity for large-scale biological approaches combining both expression and genomic-level sequence analysis using as a template an entire vertebrate genome.
Human, mouse and rat genomes likely have less than 40 000 genes each [1–4]. This is only two to three times as many genes as in Caenorhabditis elegans and Drosophila melanogaster, and only six times as many as Saccharomyces cerevisiae [5–7]. The increased complexity of vertebrates therefore can not be simply accounted for by a larger gene number. A part of the increased complexity is thought to be accomplished by alternative splicing, RNA editing and the use of protein modifications to generate a variety of protein products from a single gene, but everything starts with increased complexity at the level of transcriptional regulation. While promoters are relatively simple and short in yeast, their complexity increases in multicellular organisms, making regulatory sequences ever harder to identify. In humans, enhancer elements can be located over a megabase away from the transcriptional start site . Furthermore, current gene prediction programs used to annotate genomes often fail to correctly identify the 5' start site of a transcription unit, making in silico analysis of the regulatory sequences even more complex. To further complicate the matter, enhancer sequences diverge in evolution, co-evolving with their respective transcription factors, and often do not work across large evolutionary distances - worm to fly, for example . This makes information from non-vertebrate model systems sometimes inapplicable to vertebrate sequences.
Enhancer detection ("trapping") using insertion site context vectors was popularized as a genomics tool in Drosophila. The first fly enhancer trap vectors were based on the P element transposon and often used the transposase's own promoter fused to the beta-galactosidase reporter gene for enhancer detection [10–12]. Several of the enhancer trap lines were shown to express the LacZ reporter in cells corresponding to the expression patterns of nearby genes, validating the approach [12, 13]. In other work, promoters such as engrailed, fushi tarazu and Hsp70 were successfully developed for enhancer trapping in the fruitfly [14–16]. Further modifications to the system included the implementation of a bipartite system with a Gal4 transactivator , green fluorescent protein (GFP) , and even a GFP-LacZ fusion protein  as reporters. In addition to the P element, other transposons such as hobo and piggyBac with insertion site preferences distinct from those of the P element have been used in Drosophila [20, 21]. The availability of a variety of transposons, promoters and reporters for enhancer trapping in the fruitfly enabled researchers to obtain enhancer trap insertions into a considerable fraction of Drosophila genes (reviewed by ) and allows an investigator to choose vectors most suitable for the problem at hand.
The ability to excise from a genomic location has been instrumental to the utility of P element based vectors. For mutation-causing insertions, reversion of the mutant phenotype by P element excision proves that a given insertion causes the mutation. Since the mutageniCity of Drosophila enhancer trap transposons is not significantly higher than the average 15% rate obtained with regular P elements, most insertions do not result in a mutation . In these instances, the P element's ability to induce genomic deletions by "imprecise excision" can be used to obtain a mutation in the neighboring gene(s) .
The success of enhancer trapping in Drosophila prompted application of this approach in the mouse [24–27]. As was the case in Drosophila, the lacZ reporter was shown to be expressed in part of the target gene's expression domain . Despite the considerable success of these early experiments, enhancer detection as an experimental approach in mouse was not explored further, giving way to different versions of gene traps (for a review, see ).
We believe the success of enhancer trapping in Drosophila can be largely attributed to the advantages of this experimental system over mouse. In Drosophila, large numbers of transgenic organisms can be readily generated and screened for gene expression patterns. It is far less practical in the mouse. This is partly due to the availability of efficient and precise transgene delivery tools in the fruitfly: the native P element, hobo and piggyBac transposons. In contrast, early mouse experiments were carried out by non-facilitated DNA transgenesis. This approach is less efficient and prone to induce deletions and other genome rearrangements in the recipient locus, as noted in the first published mouse enhancer trap locus [25, 30]. The compact nature of the Drosophila genome also contributed to the success of enhancer trapping, making the path from an enhancer trap insertion to the identification of the affected gene straightforward, especially once the Drosophila genome was sequenced.
The zebrafish Danio rerio is a vertebrate model system that provides many of the advantages found in invertebrates. A few hundred transparent, externally developing embryos can be obtained from a single pair of fish per week. The zebrafish genome is about two-fold smaller than the mouse genome, and its sequencing and annotation are nearing completion. Finally, transposon tools for efficient and precise transgene delivery into the zebrafish genome are available. We focused our research on the Sleeping Beauty (SB) transposon system [31, 32]. While not as efficient as the highest titer retrovirus used in zebrafish [33, 34], the Sleeping Beauty transposon system offers advantages in expression as well as ease of construction and testing of diverse vectors that can be done using basic molecular biology tools. Furthermore, the SB system offers the possibility of transposase-induced excision out of the genome to induce local deletions or to revert possible mutant phenotypes.
In this report, we investigated the potential of the SB transposon system for enhancer detection in zebrafish. Our results indicate that zebrafish enhancer trap lines with diverse GFP expression patterns can be readily generated using the SB system. Most of the obtained lines harbor a single transposon insertion event, facilitating the rapid identification of transposon insertion sites responsible for specific GFP expression patterns. We show that two enhancer trap lines exhibit GFP expression patterns matching the expression patterns of the target genes, and that both expected and novel gene expression patterns can be identified using this genomics tool. We conclude that enhancer trapping using the Sleeping Beauty transposon system is a viable experimental approach using as template a vertebrate genome.
We have previously demonstrated the excision of a Sleeping Beauty transposon from the genome in somatic tissues of transposase-injected zebrafish embryos . We tested if such an excision event could be inherited by examining transposon excision in the germline. Embryos homozygous for the ET1 insertion were injected with SB10 transposase mRNA, and while some were used for a somatic excision assay the rest were raised to test for germline transmission of an excision event. A PCR reaction on genomic DNA from transposase-injected embryos with primers flanking ET1 insertion point produced two bands. A large band corresponded in size to the transposon insertion allele, and a small band corresponded to a transposon-less allele (data not shown). Both cannonical Sleeping Beauty transposon footprints (ATGTCAT and ATGACAT, [44, 45]) were obtained upon cloning and sequencing of the smaller band, indicating a transposase-mediated excision and DNA repair. 26 fish were screened for germline transmission (see Materials and Methods), and one was shown to transmit the expected excision footprint. We conclude that the Sleeping Beauty transposon can be excised from a genomic location in the zebrafish germline.
One tissue-specific expression pattern was recovered from our pilot screen. We sought to recover more patterns and to test if enhancer detection in zebrafish is amenable to scale-up. To that end, we co-injected 3248 zebrafish embryos with the pT2/S2EF1α-GM2 and SB10 transposase mRNA mix. 2102 embryos survived to day 3 for scoring, of which 848 were mosaic GFP positive and were selected to be raised. 330 survived to adulthood and were screened for germline transmission of GFP expression, primarily by sibling incrossing. This approach provided a lower estimate of the transgenesis and expression rate because it does not distinguish instances were both parents are transgenic. In this screen, at least 80 of the founder fish produced GFP-expressing progeny resulting in a minimum estimate of a 24% transgenesis rate. The actual transgenesis rate is closer to 30% because most of the fish were screened by incross, and if a pair produced GFP-expressing progeny, only one parent was counted as a transmitter. Eight of the GFP-expressing fish displayed distinct GFP expression patterns (Figure 3). Together with the pilot screen, 9 tissue-specific expression lines were obtained from 90 transgenic founder fish (10%) using the pT2/S2EF1α-GM2 transposon.
GFP expression in ET1 can be first observed in the polster region at 7–8 somite stage (not shown). The expression is very pronounced between 20 and 40 hours post-fertilization (hpf), when it marks the hatching gland (Figure 3A). Expression disappears as the hatching gland is resorbed. Line ET3 represents a pattern with the earliest onset of expression. Anterior localization of GFP in the diencephalon is detected by 5–6 somite stage in this line (Figure 3B). Extremely bright anterior expression persists in the ventral diencephalon (Figure 3C) and by 6 days post-fertilizations (dpf) is restricted slightly more posterior in the midline. The onset of expression for ET4 is 18 hpf with a bilateral expression pattern in cranial sensory ganglia that remains strong until 2 dpf and is undetectable by 5 dpf. This anterior expression in ET4 seems to label the lateral line ganglia both anterior and posterior to the otic vesicle (Figure 3D), however, no expression is detected in the lateral line in the trunk. In ET5 a single bilateral patch of strong GFP expression in the hyoid arch is observed by 24 hpf (Figure 3E), that by 48 hpf marks a more anterior location in the embryo (Figure 3F). Expression in this line is greatly diminished by 3 dpf and is undetectable by 5 dpf. Strong GFP expression is observed in ET6 by 26 hpf as a bilateral expression pattern consisting of two distinct patches in a subset of cranial sensory ganglia/placodes (Figure 3G). The expression weakens by 2 dpf and is undetectable by 3 dpf. GFP expression in ET7 begins weakly in the midbrain-hindbrain boundary (MHB) at 12–14 somites with the most pronounced expression in the anterior side of the MHB detected by 26 hpf (Figure 3H). Robust expression in the heart is first detected at around 32 hpf and remains ventricle-specific through 5 dpf (Figure 3I), even though expression in the brain is no longer restricted to the MHB. GFP expression in ET8 is already localized by 10–12 somites and remains strong in the telencephalon, and posterior side of the MHB through 26 hpf (Figure 3J). By 3 dpf the localized anterior expression is undetectable over autofluorescence, however, caudal expression appears to be enhanced in the dorsal neural tube. The onset of expression in ET9 occurs around 22 hpf and is difficult to detect by 2 dpf. At 28–30 hpf (Figure 3K,3L), three distinct expression domains are apparent in the telencephalon, diencephalon and hindbrain of ET9.
pT2/S2EF1α-GM2 transposon insertion events in analyzed enhancer trap lines.
In this paper, we describe the first use of enhancer trapping, or enhancer detection, as an experimental approach in zebrafish. We show that Sleeping Beauty transposons can trap enhancers by testing an artificial enhancer trapping event in vivo. This approach is likely to also be useful in the construction and testing of other trap vectors: gene (5' exon) and polyA (3' exon) and other related constructs. We then constructed two further truncations to the S1EF1α promoter in the transgenesis cassette  and found one to be particularly suitable for enhancer trapping. Ten percent (9 of 90) of GFP-expressing transgenic fish generated lines with unique GFP expression patterns. All reagents described in this paper, including the enhancer trap fish strains, are readily available upon request http://beckmancenter.ahc.umn.edu.
Many of the obtained enhancer trap lines express GFP in the nervous system. This was previously observed with both mouse and Drosophila enhancer trap vectors and was speculated to stem from the transcriptional complexity of neural tissue [11, 28]. Several of our lines also exhibit some level of GFP expression in the eye. At least two explanations can be put forward to explain this observation. First, many genes are expressed in the developing eye. Thus, the eye expression that we see may reflect expression of the tagged genes in the eye. Second, optical properties of the tissues in the eye may permit detection of GFP expression that is lower that what would be required for detection in other tissues.
The ET2 line harbors a transposon insertion into the zebrafish gene for poly(ADP-ribose) glycohydrolase (PARG). We demonstrate that both PARG and GFP in ET2 line are expressed in caudal primary motoneurons of 23 hour old embryos. Thus, GFP expression in the ET2 line mimics that of an endogenous gene (PARG), indicating that transgene expression is under control of an endogenous enhancer. A very intriguing question is what the actual trapped enhancer sequence is, how far away from the genomic enhancer the trap can insert and still detect it, and weather artificial enhancer trap approach (Figure 1) can be used to answer these questions.
The ET7 line has a transposon insertion into a predicted novel gene 30 kb downstream of the zebrafish mkp3 locus. GFP expression in that line closely resembles part of the mkp3 expression domain, suggesting that the enhancer trap transposon in that line is under control of a subset of mkp3 enhancer elements.
Zebrafish enhancer trap lines will be valuable in future developmental genetics studies, be it classical mutagenesis or morpholino "knockdown" screening . GFP expression can be used as a sensitive marker for certain tissue or cell types. For example, the ET1 line expresses GFP in the hatching gland. The expression of the hgg1 gene is specific to the polster and hatching gland depends on nodal signaling and is absent in one-eyed-pinhead mutants . We phenocopied the one-eyed-pinhead mutation by morpholino injection in ET1 homozygotes and observed a complete loss of hatching gland-specific GFP expression (data not shown). While the ET1 line expresses GFP in an organ that can be readily observed using regular light microscopy techniques, other lines visualize tissues that are not nearly as easily morphologically accessible. In particular, the ET2 line visualizes the position of primary motoneuron cell bodies and axonal trajectory. Morpholinos against known genes or new members of the zebrafish secretome  can be screened for effects on neuronal cell body position or axonal pathfinding in the developing embryos by injection into ET2 line embryos. The ET7 line may provide a fluorescent readout of FGF8 signaling, thus facilitating the identification of genes involved in that signaling pathway.
A further utility offered by the transposon system is the possibility to revert a mutant phenotype or to generate localized deletions by transposon excision . We successfully excised the transposon in the germline of the ET1 line, resulting in the expected transposon footprint. It has been shown that excision of the Sleeping Beauty transposon from a plasmid results in local deletions with fairly high frequency which is dependent on the cell type or tissue used . Furthermore, the frequency of imprecise excision of Sleeping Beauty transposons significantly increases in cells with a compromised DNA repair pathway [62, 63]. It remains to be determined how frequently the excision of a Sleeping Beauty transposon from a genomic location in zebrafish germ line is accompanied by a deletion of flanking genomic DNA, and it should be possible to compromise the embryo's DNA repair machinery to induce such deletions at a high frequency.
Our experiments indicate that enhancer detection using Sleeping Beauty transposons is an easily scalable and efficient experimental technique in zebrafish. Obtaining fish with different GFP expression patterns is not the rate limiting step in this process. Preliminary molecular analysis of the insertion site is also straightforward using inverse PCR techniques. Identification of candidate genes should benefit from the progress in zebrafish genome sequencing and annotation. The main bottleneck step is the detailed biological analysis of GFP and the corresponding candidate gene expression profile.
In Drosophila, the generation of transposase-expressing lines of flies made enhancer detection and P-element mutagenesis in general a mainstream approach. Even without a similar gain in efficiency, transposase expressing fish lines would make enhancer trapping as well as related gene- and poly(A)-trap methodologies even more accessible for high-throughput functional analysis of the vertebrate genome.
pT2/S1EF1α-GFP (pDB358) was previously published . To make γ Cry/pT2/S1EF1α-GFP (pDB375), a BamHI-HindIII fragment from Cry1-GFP3  containing part of the X. laevis γ-Cry1 promoter was cloned into the Ecl136II site of pDB358. To produce pT2/S2EF1α-GFP (pDB371), a part of the EF1α promoter was deleted from pDB358 by ligation of the Bst1107I-NcoI and NheI-NcoI fragments of pDB358. Similarly, the EcoRI-NcoI and NheI-NcoI fragments of pDB358 were ligated to produce pT2/S3EF1α-GFP (pDB372).
For inverse PCR experiments, zebrafish genomic DNA was digested and ligated as described . 1 and 2.5 microliters of the ligation reaction were used for the first PCR reaction with RP1/LP1 or RP1/GFP-R1 primers in total volume of 25 μl. 1 μl of the first PCR reaction was used as a template for the second (nested) PCR reaction with primer pairs RP2/LP2 or RP2/GFP-R2, respectively. Expand Hi Fi PCR system (Roche) was used for all PCR reactions. A MJ Research PTC-100 PCR machine was used for PCR with the following program : 92°C 4 min., 92°C 10 sec., 60°C 30 sec., 68°C 6 min., 30 cycles. Starting at cycle 11, 20 sec. per cycle was added to the extension time. The same PCR reaction with an annealing temperature 55°C was used for amplification with primers flanking transposon insertion sites, and for amplification of partial PARG cDNA from a maternal cDNA library. Primer sequences are: LP1 GTGTCATGCACAAAGTAGATGTCC ; LP2 ACTGACTTGCCAAAACTATTGTTTG; nRP1 CTAGGATTAAATGTCAGGAATTGTG; RP2 GTGAGTTTAAATGTATTTGGCTAAG; GFP-R1 TTCGGGCATGGCACTCTTG; GFP-R2 TATGATCTGGGTATCTCGCAA; NeuroC1-F1 CGTAAAGATGCCTTGTTCAGAA; NeuroC1-R1 ATTCCGTGACTCTCCTGAAATA; NeuroIns-F1 GGCTTGCATACATGACTAATG; NeuroIns-R1 GAAGACTGAAGTCCTCAAACT; HG1-1 ACATTGAGCCACTAAGCATTG; HG-2 TGTGTGCACTTAAGGGGCGA. Mkp3-F1 AGTGTTGCATTCTCCAGGATA; Mkp3-R1 TGACACAGAACTTCCCTGAAC; EF1a-F2 TTCCTGCAGGTCGACTCT; GFP-R0 GTGTAATCCCAGCAGCTG. Information about other primers is available from the authors upon request.
A partial sequence for the zebrafish poly(ADP-ribose) glycohydrolase cDNA was amplified using primers neuroC1-F1 and neuroC1-R1 and cloned using a Topo TA cloning kit (Invitrogen) to make pDB376. To make antisense RNA probe, pDB376 was digested with SpeI and transcribed with T7 RNA polymerase (Promega) and DIG labeling kit (Roche). GFP probe was made by amplifying GFP with 46 base pairs of EF1α promoter from pT2/S1EF1α-GM2 using primers EF1a-F2 and GFP-R0, and cloning it into Topo TA cloning kit resulting in pSS100. pSS100 was linearized with SpeI and transcribed with T7 RNA polymerase using DIG labeling kit. To make mkp3 antisense probe, mkp3 cDNA was amplified from maternal cDNA library with primers Mkp3-F1 and Mkp3-R1 and cloned into Topo TA cloning kit to produce pDB528. The plasmid was linearized with SpeI and transcribed with T7 polymerase using DIG labeling kit.
Embryos injected with SB10 transposase mRNA and transposon DNA mix were raised as described [32, 64]. In pilot screens, adult fish were outcrossed to brass for ease of husbandry. All collected embryos were screened for GFP expression at 1 day post fertilization (dpf) and 3 dpf. We set an arbitrary 200 embryo cutoff for screening, meaning that if less that 200 embryos were obtained from a founder, an additional cross was set up and to obtain additional embryos for screening. Analysis of transgenesis data from pilot screens indicated that 10% of transgenic lines would have been missed if cutoff was set at 100 embryos, and this less stringent coverage protocol was used in scale up screen. Also, we decided to limit screening to 1 dpf since none of the transgenics would have been missed in the pilot screens without the 3 dpf screening.
Homozygous ET1 embryos were injected with SB10 transposase mRNA, raised and screened for loss of hatching gland specific GFP expression, or for a change in the GFP expression pattern. Twenty six fish were screened (R0, for Remobilization), and 2 gave GFP negative embryos, with an additional 2 giving ubiquitously GFP positive embryos, suggesting that germline remobilization events may have occurred in as many as 15% of transposase injected embryos. Ubiquitous GFP positive embryos (one in each of the two R0) did not survive. Of the two R0's that gave GFP negative embryos, one gave mosaic hatching gland expression in the next generation. PCR with transposon specific and flanking primers did not show any changes in the locus. The second R0 produced 19 embryos that were GFP negative from the total of 671 embryos obtainted. An R1 adult was outcrossed, embryo DNA was prepared, and PCR with primers HG1-1 and HG1-2 was conducted. The resulting PCR fragment was cloned using PCR 4 Topo cloning kit (Invitrogen). Plasmids were sequenced using M13 Forward primer, and one clone with a transposon footprint was identified. To confirm that it was not due to PCR contamination, a second clutch of embryos was obtained, the procedure was repeated, and the same footprint was obtained (data not shown).
We thank Paul Phelps, Sandra Leo, Amanda Mahoney and Tessa Hodapp for help with fish screening, and Aubrey Nielsen, Rachel Bowers, Dan Carlson and Pat Cliff for fish maintenance and Perry Hackett for critical reading of the manuscript. We thank all members of the Arnold and Mabel Beckman Center for Transposon Research for valuable discussions. This research was supported by the Arnold and Mabel Beckman Foundation and the National Institutes of Health (DA14546).
This article is published under license to BioMed Central Ltd. This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.