Adaptation of GAL4 activators for GAL4 enhancer trapping in zebrafish

Authors


Abstract

An enhancer trap-based GAL4-UAS system in zebrafish requires strong GAL4 activators with minimal adverse effects. However, the activity of yeast GAL4 is too low in zebrafish, while a fusion protein of the GAL4 DNA-binding domain and the VP16 activation domain is toxic to embryonic development, even when expressed at low levels. To alleviate this toxicity, we developed variant GAL4 activators by fusing either multimeric forms of the VP16 minimal activation domain or the NF-κB activation domain to the GAL4 DNA-binding domain. These variant GAL4 activators are sufficiently innocuous and yet highly effective transactivators in developing zebrafish. Enhancer-trap vectors containing these GAL4 activators downstream of an appropriate weak promoter were randomly inserted into the zebrafish genome using the Sleeping Beauty transposon system. By the combination of these genetic elements, we have successfully developed enhancer trap lines that activate UAS-dependent reporter genes in a tissue-specific fashion that reflects trapped enhancer activities. Developmental Dynamics 238:641–655, 2009. © 2009 Wiley-Liss, Inc.

INTRODUCTION

A commonly used strategy for analyzing the function of a gene in the developing embryo is to explore the phenotypic consequences of up- or down-regulation of its activity by forced misexpression during development. Binary gene expression systems, such as the GAL4-UAS system developed for Drosophila genetics research, are particularly powerful tools for addressing tissue- and developmental stage-specific gene functions.

The GAL4-UAS system in Drosophila (for review, see Duffy,2002) utilizes stable transgenic lines (GAL4 driver lines) expressing the yeast GAL4 transcriptional activator under the control of a tissue-specific promoter/enhancer. These lines are crossed with another set of upstream activating sequence (UAS) transgenic lines (UAS lines) carrying a gene of interest under the control of GAL4-binding UAS sites. The progeny of such a cross express the UAS-dependent gene according to the pattern of GAL4 expression in the driver line. By crossing a single UAS line carrying a gene of interest with a specified panel of GAL4 driver lines, a variety of spatio-temporal patterns of gene expression can be achieved. Moreover, the advantage of the binary expression system over simple promoter/enhancer-dependent gene activation is its robustness, namely a higher level of target gene activation and flexible on/off regulation.

Zebrafish has become a popular vertebrate model for the study of embryonic development. One advantage of using zebrafish embryos and certain other teleosts is their optical transparency, which allows the facile visualization of fluorescent reporter proteins such as GFP and DsRed for marking specific group of cells and/or monitoring cell type–specific gene expression. In addition, the ease of injection of in vitro–synthesized RNA or antisense morpholino oligonucleotides into fertilized eggs facilitates gain- or loss-of-function analyses of genes. However, these types of analyses are limited in that the spatial or temporal effects of injected reagents cannot be easily regulated. Therefore, implementation of a reliable system analogous to the Drosophila GAL4-UAS system in zebrafish is required in order to achieve spatial and temporal regulation of specific gene expression.

Previous attempts have been made to apply the Drosophila GAL4-UAS system to teleost fish. The early reports of use of the GAL4-UAS method in zebrafish indicated that the native yeast GAL4 had a low transactivation potential in vertebrates (Köster and Fraser,2001; Scheer and Campos-Ortega,1999). In order to improve upon this weak activation potential, several research groups have employed fusion proteins consisting of the GAL4 DNA-binding domain (GAL4 DBD) fused to the activation domain of herpes simplex virus (HSV) VP16. Although VP16 was proven to be a strong activation domain in teleost embryo in these studies, its use was also associated with deleterious effects on embryonic development (Grabher and Wittbrodt,2004; Köster and Fraser,2001; Scott et al.,2007).

The transactivation domain of VP16 is one of the most potent activation domains known. This domain interacts with a number of key components of the transcriptional machinery through its various modules. It is well established that high intercellular levels of the VP16 activation domain in yeast or cultured mammalian cells can lead to non-specific inhibition of transcription (Gilbert et al.,1993; Sadowski et al.,1988; Triezenberg et al.,1988), which may result from titration of transcriptional machinery components by the VP16 activation domain. This inhibition of transcription thereby elicits abnormal gene regulation, which eventually arrests cell division (Berger et al.,1992; Gilbert et al.,1993) or induces cell death (Shockett et al.,1995), probably resulting in a range of deleterious effects upon embryogenesis. Hence, the development of GAL4 derivatives with a strong activation potential but minimal toxicity would overcome this major barrier to the full utilization of the GAL4-UAS system in zebrafish. This is the major focus of our current study.

During the final phases of preparing this report, it came to our attention that there has been confusion in description concerning “GAL4VP16” in the previous studies. Grabher and Wittbrodt (2004) employed the full-length VP16 activation domain in their fusion construct as the original GAL4VP16 construct (Sadowski et al.,1988). However, the other studies that employed “GAL4VP16” in zebrafish (Davison et al.,2007; Inbal et al.,2006; Köster and Fraser,2001; Sagasti et al.,2005; Sassa et al.,2007; Scott et al.,2007) actually used GAL4VP16413-470 (Barlev et al.,1995) instead of the original GAL4VP16 construct (Sadowski et al.,1988), although this fact was not explicit in these reports (personal communication from Dr. Köster, who first used GAL4VP16413-470 in zebrafish; Köster and Fraser,2001). In GAL4VP16413-470, the truncated VP16 activation domain (aa 413–470) was fused to GAL4 DBD in place of the full activation domain (aa 413–490). The C-terminal region of VP16 (aa 471–490) is important for its transcriptional activation function (Sullivan et al.,1998; Triezenberg et al.,1988), and hence GAL4VP16413-470 was actually shown to be approximately 10-fold less active than GAL4VP16 in yeast (Barlev et al.,1995). This prompted us to directly compare GAL4VP16 and GAL4VP16413-470 on their transactivation potentials and toxicity in zebrafish embryos in this study.

Another requirement for implementing the GAL4-UAS system in vertebrates is to develop an efficient enhancer/gene trap system analogous to the P-element transposon system in Drosophila. Retroviral vectors (Gaiano et al.,1996) and the transposable elements Sleeping Beauty (SB), a reconstructed Tc3-like transposon derived from the salmonid genome (Davidson et al.,2003; Ivics et al.,1997), and Tol2, which is a hAT family transposon derived from the medaka genome (Kawakami et al.,1998,2004), have recently been utilized in zebrafish as efficient single-copy random-insertion vectors for transgenesis. These have facilitated the wide application of trapping methodologies in fish utilizing fluorescent proteins as the reporter system (Balciunas et al.,2004; Ellingsen et al.,2005; Grabher et al.,2003; Kawakami et al.,2004; Parinov et al.,2004). More recently, Tol2 has been used for GAL4-UAS based gene/enhancer trapping in combination with GAL4VP16413-470 (Davison et al.,2007; Scott et al.,2007) or with another derivative GAL4FF (Asakawa et al.,2008). The SB transposon also enables gene/enhancer trapping in zebrafish (Balciunas et al.,2004; Clark et al.,2004) and medaka (Grabher et al.,2003). The availability of multiple transposon systems for GAL4-UAS based trapping methodologies is advantageous, given target site preferences of transposons (for a review, see Mates et al.,2007).

In this study, we report the successful adaptation of GAL4 activators for the GAL4-UAS-based enhancer trapping in zebrafish. Rather than using a fusion of the full-length VP16 activation domain to GAL4 DBD, we employed the VP16 minimal activation domain and the NF-κB p65 activation domain to minimize the possible deleterious effects of GAL4 activators. Enhancer-trap vectors containing these variant GAL4 activators downstream of an appropriate weak promoter were distributed in the zebrafish genome using the SB transposon system. By utilizing these unique combinations of genetic elements, we have successfully applied the GAL4-UAS system to zebrafish. Moreover, these new GAL4 activators adapted to the vertebrate system should prove useful in a variety of experimental systems using vertebrate cells and embryos.

RESULTS

Design of Enhancer Trap Vectors for GAL4 Drivers Utilizing Sleeping Beauty Transposase-Mediated Insertion

To generate GAL4-driver zebrafish lines with a variety of spatio-temporal specificities, a GAL4 enhancer-trap strategy was employed. The first set of enhancer-trap vectors, although potentially toxic, carried a GAL4VP16 fusion gene driven by a weak promoter. The transcription unit of this fusion gene was flanked by optimized IR/DR sequences of the SB transposon vector pT2 (Cui et al.,2002), in a truncated pUC19 plasmid backbone (pT2.2) (Fig. 1A, B).

Figure 1.

GAL4 activators and enhancer-trap constructs. A: Schematic representation of the GAL4 activators. The full-length yeast GAL4 contains the DNA-binding domain (DBD) in its N-terminal region and two separable transcriptional activation regions (AR1 and AR2). GAL4(ΔCR) lacks the central region that includes the inhibitory domain (ID) and the glucose-response domain (GRD). In fusion activators, GAL4 DBD was fused to either the full-length VP16 activation domain (amino acids [aa] 413–490); the activation domain (aa 365–550) of the mouse transcription factor NF-κB p65; or either tetrameric, trimeric, or dimeric forms of the minimal activation domain of VP16 (VPmad). VPmad corresponds to aa 436–447 of VP16 (the amino acid sequence is shown at right). B: Schematic representation of the enhancer-trap vectors for GAL4VP16 using three different promoters: the HSV tk promoter (TK), the hybrid promoter of tk and carp β-actin (TKBA), and the zebrafish gata2 promoter (GATA2). The shortened intron (si) sequence derived from the first intron of carp β-actin was placed between the promoter and the GAL4VP16 coding sequence. Transcription is terminated by the ocean pout AFP poly(A) signal (shown by black circle). These GAL4VP16 transcription units are flanked by the SB IR/DR sequences of the pT2.2 SB transposon vector. C: Measurement of the relative strength of the promoters TK, TKBA, and GATA2 in zebrafish embryos. The luciferase activity generated by the TK-Venusluc vector was arbitrarily assigned a value of 1. Data are the mean values with standard errors from two independent injection experiments. D: Schematic representation of the UAS-linked reporter vectors. The UAS:DsRedEx, UAS:DsRedExDR, and UAS:nlsVenus were used for generating the GAL4-responsive reporter lines. DsRedExDR has a short half-life due to the PEST degradation sequence. nlsVenus is a variant of YFP (Venus), which is fused to the SV40 large T antigen nuclear localization signal (nls). The UAS-reporter vectors contained 10 copies of the GAL4 binding site (10×UAS), the carp β-actin minimal promoter with the shortened first intron, the reporter genes, and the ocean pout AFP poly(A) signal. E: Schematic representations of the enhancer-trap vectors using the modified GAL4 activators. These constructs were essentially the same as the GAL4VP16-based trapping constructs, except that the GAL4 activator was either GAL4VPmad2, GAL4VPmad3, or GAL4NFkB, and the shortened intron sequence was removed.

We evaluated three promoters, TK, TKBA, and GATA2, for their utility in enhancer trapping for the following reasons: (1) The activity of the promoter in an enhancer-trap construct needs to be sufficiently weak to avoid background expression and at the same time must respond to genomic enhancers in proximity to the insertion site. Enhancers at a distance are unable to activate transcription from minimal or core promoters without the cooperation of promoter proximal elements (Bertolino and Singh,2002; and references therein). TK, TKBA, and GATA2 all include promoter proximal elements upstream of their minimal promoter regions. (2) The TK promoter, derived from the thymidine kinase gene of the human herpes simplex virus (HSV), satisfies the conditions of low background and good responsiveness to enhancers in chicken cells and embryos (Uchikawa et al.,2003; Matsumata et al.,2005). (3) Based upon a previous report suggesting that the mammalian TK promoter had a lower activity in fish cells than in amniote cells (Perz-Edwards et al.,2001), we also prepared a hybrid promoter, TKBA, which consists of the TK promoter upstream region and the carp β-actin promoter core region (see Experimental Procedures section). (4) In addition, we tested the zebrafish gata2 promoter (Meng et al.,1997), which has been successfully employed in monitoring retinoic acid-responsive gene expression in zebrafish (Perz-Edwards et al.,2001).

To assess the activity of these promoters in zebrafish embryos, we linked each one to a Venus-luciferase (Venusluc) fusion reporter gene and injected the resulting constructs into fertilized eggs. Luciferase activity was then measured in the developing embryos. The results indicated that the basal activities of the TK and GATA2 promoters were similar, whereas that of the TKBA promoter was one order of magnitude higher (Fig. 1C). We also tested effects of the Xenopus EF1a enhancer on these promoters in embryos and observed that the TK and TKBA promoters were activated ∼5-fold by the EF1a enhancer, but its effect on the GATA2 promoter was marginal, suggesting some difference in the compatibility between promoters and enhancers (see Supp. Fig. 1, which is available online).

To visualize the expression of the GAL4 activators in living zebrafish embryos, we established the GAL4-responsive fluorescent protein reporter fish lines UAS:DsRedEx, UAS:DsRedExDR, and UAS:nlsVenus (Fig. 1D; see also Experimental Procedures section). Activation of these reporter genes in the UAS-reporter lines was confirmed in embryos by injection of GAL4VP16 mRNA (data not shown). Due to the PEST degradation sequence, the UAS:DsRedExDR line gave rise to slightly weaker red fluorescence intensities than the UAS:DsRedEx line (described below).

Enhancer Trapping by GAL4VP16 Vectors Causes Deleterious Effects in the F1 Generation

The three SB enhancer-trap vectors containing GAL4VP16 (Fig. 1B), along with in vitro synthesized mRNA encoding the improved SB transposase (SB11; Geurts et al.,2003), were injected into one-cell-stage wild type embryos, which were raised to adulthood. A total of 74 surviving adult fish were crossed with the UAS:DsRedExDR reporter fish, and then screened for reporter expression at one day post-fertilization (dpf). Twelve out of 74 fish transmitted the SB transposon insertions to the F1 progeny (Table 1, Supp. Table 1). In this screen, there were no F1 embryos that were positive for the transposon but negative for UAS:DsRedExDR reporter expression, and all of these double transgenic embryos showed UAS:DsRedExDR reporter expression in a region-specific manner (Supp. Fig. 2), indicating that enhancer trapping was successful using our vectors. The high rate of trapping by SB transposon insertion suggested that this transposon preferentially integrates into a transcriptionally competent, open chromatin domain on the zebrafish chromosomes. The F0 germline mosaicism rates of these SB transposon insertions (i.e., rates of SB transposon-positive F1 progeny with particular UAS-reporter expression) were measured as 0.3–10% (Supp. Table 1). These values are consistent with a previous report (Davidson et al.,2003).

Table 1. Summary of Screens for Enhancer-trap Lines
Trap construct (Promoter-Activator)# of screened fish (# of injectionsa)# of founder (F0) fish (SB transgenesis rateb)# of F0 that generate reporter- positive F1 (enhancer trap ratec)Founder named (# of inserts that give UAS-reporter expressione)Total # of inserts that give reporter expression
  • a

    The screened fish were derived from the indicated number of independent injections.

  • b

    SB transgenesis rates were obtained as the percentage of founder F0 fish that transmitted the transposon trap vector to progeny, the integration of which was mediated by SB transposition, regardless of UAS-reporter expression in F1 embryos. F0 fish that transmitted the trap vector integrated through non-SB transposition were not counted (17 fish out of 384 screened fish [4.4%]).

  • c

    Enhancer trap rates are the percentage of founder F0 fish that generated UAS-reporter-positive F1 fish among transposon positive F0 fish.

  • d

    m, male; f, female.

  • e

    Insert numbers were estimated by the number of different patterns of UAS-reporter expression observed in F1 embryos.

TKsi-GAL4VP1614(1)1(10%)1(100%)#9m (1)1
TKBAsi-GAL4VP1626(1)4(20%)4(100%)#3f (2), #4m (2), #14m (1), #22m (3)8
GATA2si-GAL4VP1634(3)7(20%)7(100%)#1m (1), #1f (1), #6m (1), #7m (1), #28m (2), #30m (2), #42m (2)10
TK-GAL4VPmad230(2)10(33%)3(30%)#1f (1), #4f (1), #5m (5)7
TKBA-GAL4VPmad236(2)8(20%)6(80%)#1f (5), #4m (2), #4f (1), #6f (5), #12m (4), #19f (1)18
GATA2-GAL4VPmad267(3)8(10%)2(30%)#21m (3), #32m (2)5
TK-GAL4VPmad348(2)8(20%)4(50%)#1f (1), #4f (1), #24f (1), #29f (1)4
TKBA-GAL4VPmad351(3)6(10%)4(70%)#7f (2), #14m (1), #16m (2), #19m (1)6
TK-GAL4NFkB46(2)12(26%)5(40%)#2m (1), #7m (1), #12f (3), #27m (1), #41m (1)7
TKBA-GAL4NFkB32(2)9(30%)6(70%)#10f (4), #14f (1), #19f (3), #29m (1), #34m (4), #36m (1)14

These double transgenic F1 embryos, however, did not survive beyond 5 dpf with the exception of those derived from one founder out of 12 (Supp. Table 1). Moreover, embryos that expressed DsRedExDR at high levels showed developmental delay and/or morphological abnormalities (Supp. Fig. 2, Supp. Table 1). These observations indicated that GAL4VP16 was deleterious to the development of zebrafish embryos when its expression was presumably activated under the influence of genomic enhancers.

Modified GAL4 Activators With Lower Toxicity Show a High Transactivation Potential

To minimize the deleterious effects likely caused by the use of GAL4VP16, we constructed GAL4 activator variants that might be better tolerated by zebrafish embryos, even at high expression levels (Fig. 1A). In one variant, the full-length VP16 activation domain was replaced with multimeric forms of its minimal activation domain (core sequence of N-terminal activation domain of VP16). This strategy was used successfully in the development of tetracycline-controlled transactivators that were tolerated at high concentrations in cultured cells (Baron et al.,1997). Dimeric, trimeric, and tetrameric forms of the VP16 minimal activation domain (designated as VPmad2, -3, and -4, respectively) were fused with GAL4 DBD to generate variants of the GAL4 activator. The recently reported GAL4 variant Gal4FF (Asakawa et al.,2008) utilized a dimeric form of the same minimal activation domain and is almost identical to GAL4VPmad2. Another variant used the NF-B p65 carboxyl terminus activation domain (Schmitz and Baeuerle,1991), which has been reported to be as potent as the VP16 activation domain but less toxic (Rivera,1998; Urlinger et al.,2000).

To assess the transactivation potential and toxicity of these variant GAL4 activators in zebrafish embryos, we injected about 5 pg of in vitro–synthesized mRNAs (at 5 pg/nl) encoding these GAL4 activators into one-cell-stage embryos of the UAS:nlsVenus reporter line, and examined both nlsVenus fluorescence intensity and the occurrence of developmental abnormalities in comparison with GAL4VP16 and the original yeast GAL4 protein. GAL4VPmad3 and GAL4VPmad4 activated nlsVenus expression at levels comparable to those of GAL4VP16 whereas the GAL4NFkB-injected embryos showed slightly weaker nlsVenus expression (Fig. 2D–G). GAL4VPmad2-injected embryos showed much weaker nlsVenus expression, but after increasing the amount of mRNA injected from about 5 to 20–30 pg the nlsVenus expression level became comparable to GAL4VPmad3 and GAL4VPmad4 at 5 pg (Fig. 2H,I). The native GAL4 protein did not significantly activate reporter expression in this assay (Fig. 2B), confirming a previous observation that it was two orders of magnitude less active than GAL4VP16 (Sadowski et al.,1988). We also tested GAL4(ΔCR), a truncated version of GAL4 consisting of only the DNA binding domain and two activation domains (Fig. 1A), since an analogous truncation of GAL4 (GAL4Δ) caused a stronger activation potential in Drosophila (Viktorinova and Wimmer,2007). Consistently, GAL4(ΔCR) injection caused a slightly stronger nlsVenus expression than GAL4 (Fig. 2C). However, the resulting activation was still less than that observed with GAL4VPmad2.

Figure 2.

Comparison of transactivation potentials and deleterious effects of the variant GAL4 activators in zebrafish embryos. AH: nlsVenus expression and morphology of embryos injected with the variant GAL4 activator mRNAs. Homozygous UAS:nlsVenus reporter embryos were injected at the one-cell stage with approximately 1 nl solution containing the activator mRNAs (5 pg/nl) and observed at 6.5, 1, and 25 hpf. Native GAL4 mRNA was injected at 13.5 pg/nl, which was equimolar to 5 pg/nl of GAL4VP16. I: GAL4VPmad2 mRNA was injected at concentrations of 10, 20, and 30 pg/nl. The fluorescent images were captured with fixed exposure times of 2, 1.5, and 1 sec at 6.5, 10, and 25 hpf, respectively. Representative images were taken from at least two independent injections, which gave consistent results.

No developmental defects were noted in embryos that were injected with 5 pg of these mRNAs at least until 25 hr post-fertilization (hpf), with the exception of GAL4VP16 (Fig. 2). GAL4VP16-injected embryos showed high nlsVenus expression, but virtually all embryos had abnormal morphologies after 10 hpf (Fig. 2D). In these embryos, gastrulation movements appeared to be impaired and cell death was also often observed in the brain at later stages.

Increasing the injected amount of GAL4VP16 mRNA to 10 pg caused severe morphological abnormalities and/or lethality in nearly all injected embryos (Supp. Fig. 3A). Injection of this level of GAL4VPmad3 mRNA caused some morphological abnormalities in about 60% of the embryos, whereas injection of GAL4VPmad2 resulted in no appreciable effects upon embryonic development at this concentration. When 20 pg of mRNAs was injected, GAL4VPmad2 and GAL4NFkB affected about 50% of the embryos. GAL4NFkB primarily affected tail elongation without significantly affecting anterior development (Supp. Fig. 3B). This phenotype was unique to GAL4NFkB and rarely observed when using GAL4VPmad2 or GAL4VPmad3.

To quantify the transactivation potentials of the variant GAL4 activators, their encoding mRNAs were injected into one-cell embryos together with a UAS-dependent luciferase reporter vector (UAS:Venusluc). At 10 hpf, the injected embryos were lysed and luciferase activities were measured (Fig. 3). GAL4VPmad3, GAL4VPmad4, and GAL4NFkB activated luciferase expression comparably to GAL4VP16 by injection of 5 pg mRNA. Although GAL4VPmad2 caused much lower activation when injected at 5 pg, luciferase activity was similar or even stronger when GAL4VPmad2 was injected at 10–30 pg. These results were consistent with data on UAS:nlsVenus expression.

Figure 3.

Relative strengths of the transactivation potentials of variant GAL4 activators in zebrafish embryos. A: Schematic representation of the constructs used in a co-injection assay. B: Transactivation potentials of variant GAL4 activators in zebrafish embryos indicated by the co-injection assay. The indicated amounts of activator mRNAs were injected together with the UAS:Venusluc vector into wild type, one-cell-stage embryos. Injected embryos were lysed at 10 hpf and luciferase activities were measured. The luciferase activity generated by the UAS:Venusluc vector alone without exogenous activator was arbitrarily assigned a value of 1. The mean values with standard errors from at least two independent injection experiments are shown.

As stated in the Introduction section, we became aware that the previous studies employing “GAL4VP16” in zebrafish (Davison et al.,2007; Inbal et al.,2006; Köster and Fraser,2001; Sagasti et al.,2005; Sassa et al.,2007; Scott et al.,2007) actually used GAL4VP16413-470 (Barlev et al.,1995) instead of the original GAL4VP16 construct (Sadowski et al.,1988). Direct comparison of GAL4VP16413-470 with the GAL4 variants tested in this study revealed that GAL4VP16413-470 was less active and less toxic than GAL4VP16 and GAL4VPmad3, but more active than GAL4VPmad2 in zebrafish embryos (Supp. Fig. 3). Toxicity of GAL4VP16413-470 was generally low, but the embryos injected with GAL4VP16413-470 often displayed developmental defects in the anterior central nervous system (CNS) and the eye without significantly affecting posterior development (Supp. Fig. 3B). These characteristics of GAL4VP16413-470 should have allowed production of viable transgenic zebrafish lines with this GAL4 variant.

Overall, the variant GAL4 activators either with multimers of the VP16 minimal activation domain or the NF-κB p65 activation domain resulted in lower toxicity than GAL4VP16 while still attaining a transactivation potential comparable to GAL4VP16. In addition, the subregion of VP16 (aa 413–470) was found to be a moderate activation domain. The GAL4 variants including GAL4VPmad2, GAL4NFkB, and GAL4VP16413-470 appeared to be suitable for use in zebrafish embryos due to their minimal adverse effects to embryo morphogenesis and development.

Generation of Enhancer-Trap Lines Using Improved GAL4 Activators

To generate enhancer-trap lines using the newly generated variant GAL4 activators, we constructed seven SB enhancer-trap vectors in which one of three variant GAL4 activators (either GAL4VPmad2, GAL4VPmad3, or GAL4NFkB) was placed downstream of one of three promoters (either TK, TKBA, or GATA2) in the combinations indicated in Figure 1E. One-cell-stage embryos were injected with these enhancer-trap vectors plus SB11 transposase mRNA, raised to adulthood and then screened for enhancer trapping by crossing with reporter lines (Table 1, Supp. Table 1).

When the TK promoter was used in combination with either GAL4VPmad2 or GAL4NFkB, 30–40% of the founders crossed with the UAS:DsRedExDR reporter line yielded F1 embryos that expressed DsRedExDR in a region-specific manner (Table 1; Fig. 4A, F). Combination of the GATA2 promoter with GAL4VPmad2 yielded positive founders at a slightly reduced rate (Table 1), and DsRedExDR expression tended to be weak (Fig. 4C). In contrast, when the stronger TKBA promoter was combined with GAL4VPmad2, GAL4VPmad3, or GAL4NFkB, approximately 70% of founders generated DsRedExDR-positive F1 embryos (Table 1). In these embryos, DsRedExDR was expressed in region-specific patterns, but weak patchy expression frequently occurred throughout the embryo (Fig. 4B, E, G). These observations suggest that when GAL4 activators are activated by TK or GATA2 promoter alone without the aid of enhancers, the promoters are not strong enough to activate UAS-dependent reporter expression. In contrast, the TKBA promoter often gives rise to a low level of GAL4 activator expression due to its high basal activity, under which weak UAS-reporter expression may occur. However, the increased rate of enhancer trapping with the TKBA promoter suggested that use of a stronger promoter in the trap vector may allow detection of weaker enhancers owing to achievement of a high expression level of GAL4 activators. In a transient assay using zebrafish embryos, the expression level of TKBA-luciferase activated by the EF1a enhancer of moderate strength was very high, and that of TK-luciferase remained below the basal level of TKBA-luciferase (Supp. Fig. 1).

Figure 4.

Patterns of UAS:DsRedExDR reporter expression in enhancer-trap lines with modified GAL4 activators. AG: DsRedExDR expression patterns observed at 30–36 hpf for selected enhancer-trap lines are shown grouped according to the specific trap constructs. In Be, a 24-hpf embryo is presented for its clear reporter expression at this stage. In some panels, the embryo is outlined. A short description of the expression patterns for selected enhancer-trap lines is presented in Table 2.

Table 2. Summary of the Expression Patterns for Selected Enhancer-trap Lines
Enhancer trap lineTissues where prominent reporter expression was observed at 30-36 hpfStages when reporter expression was observed between 10 hpf and 5 dpf
TK-GAL4VPmad2 #4fEpidermis12 h to 4 d
TK-GAL4VPmad2 #5m-1Trigeminal ganglion, ventral hindbrain and hypochord17 h to 2 d
TK-GAL4VPmad2 #5m-3Diencephalon, midbrain, hindbrain and spinal cord14 h to >5 d
TK-GAL4VPmad2 #5m-5Diencephalon and retina10 h to 3 d
TKBA-GAL4VPmad2 #1f-2Telencephalon14 h to 3 d
TKBA-GAL4VPmad2 #6f-3Hindbrain motor neuron24 h to 2 d
TKBA-GAL4VPmad2 #12m-1Diencephalon and hindbrain15 h to 4 d
TKBA-GAL4VPmad2 #12m-3Mesoderm15 h to 4 d
GATA2-GAL4VPmad2 #21m-1Nasal placode and otic vesicle14 h to 2 d
TK-GAL4NFkB #2mMidbrain and hindbrain18 h to 4 d
TK-GAL4NFkB #12f-1Notochord30 h to >5 d
TKBA-GAL4NFkB #19f-2Central nervous system16 h to 2 d
TKBA-GAL4NFkB #34m-2Telencephalon, lens, hindbrain and spinal cordTelencephalon, 22 h to >5 d All others, 30 h to >5 d

When the GAL4VPmad3 activator was employed, genomic insertion of either TK-GAL4VPmad3 or TKBA-GAL4VPmad3 yielded DsRedExDR-positive founders at moderate to high rates (50–70%; Table 1). These combinations were often associated with weak and patchy yet region-unrestricted DsRedExDR expression in F1 embryos (Fig. 4D, E), presumably reflecting the high transactivation potential of GAL4VPmad3. Combination of the GAL4VPmad3 activator and the TKBA promoter may result in deleterious levels of activator expression, depending on the vector insertion sites. For instance, F1 embryos of TKBA-GAL4VPmad3 #7f-1 and TKBA-GAL4VPmad3 #16m-1 were morphologically normal but did not survive to adulthood (Supp. Table 1), which was indicative of adverse effects following activator expression.

Taken together, these observations suggested that the TK and GATA2 promoters were suitable for GAL4 enhancer trapping due to their low basal activity. In combination with these promoters, the moderate activators GAL4VPmad2 and GAL4NFkB were useful in generating enhancer-trap lines for region/tissue-specific GAL4 activator expression without undesirable deleterious effects. The stronger GAL4VPmad3 activator may also be useful in combination with these weak promoters for the detection of weak enhancers. In situations where the background expression of a gene under UAS does not matter, the TKBA promoter may be combined with GAL4VPmad2 and GAL4NFkB to increase the trapping rate and expression level.

Characterization of Enhancer-Trap Lines Expressing Variant GAL4 Activators

The UAS:DsRedExDR expression activated by the GAL4-based activators in the enhancer-trap lines was detectable from early developmental stages to over 5 dpf in various tissues (Fig. 5, Table 2). For instance, UAS:DsRedExDR expression in the TK-GAL4VPmad2 #5m-5 line was activated in the anterior neural plate at the tailbud stage (10 hpf) (Fig. 5 Aa) and then became localized to the retina primordia and diencephalon at 15 hpf (Fig. 5 Ab–c). In the reporter-crossed TK-GAL4VPmad2 #5m-3 line, DsRedExDR expression was observed in the diencephalon, midbrain, and hindbrain of the embryo (Fig. 5 Ad–f), which became detectable at 14 hpf and was more prominent at 15 hpf. Embryos of the reporter and TKBA-GAL4VPmad2 #1f-2 line showed DsRedExDR expression in a small portion of the telencephalon with similar time course (Fig. 5 Ag–i).

Figure 5.

Spatio-temporal reporter expression patterns generated by variant GAL4 activators in selected enhancer-trap lines. A: UAS:DsRedExDR expression in embryos of the TK-GAL4VPmad2 #5m-5 (ac), #5m-3 (df), and TKBA-GAL4VPmad2 #1f-2 (gi) lines at 11 hpf (a), 15 hpf (b,d,e,g,h), and 24 hpf (c,f,i). Lateral views of whole embryos (a,d,g). Dorsal views of whole embryos (b,e,h). Lateral views with anterior to the left (c,f,i). B: UAS:nlsVenus expression in the TK-GAL4VPmad2 #5m-5 line (b) and the TK-GAL4VPmad2 #5m-3 line (d) at 15 hpf is shown for comparison with UAS:DsRedExDR expression (a,c). Dorsal views with anterior to the bottom. C: Higher magnification images of TK-GAL4VPmad2 #5m-1 (a) and TKBA-GAL4VPmad2 #6f-3 (b,c) highlighting neuronal expression at 36 hpf. The UAS:DsRedEx line was used to obtain brighter red fluorescence in the embryo shown in c. Dorsolateral (a,b) and lateral (c) views of the head region.

When the TK-GAL4VPmad2 #5m-5 and #5m-3 lines were crossed with the UAS:nlsVenus reporter line, nlsVenus expression was detected with a time course and spatial pattern similar to that observed with DsRedExDR expression. However, nlsVenus displayed a more localized pattern, reflecting the nuclear accumulation of Venus fluorescence (Fig. 5 Ba–d).

We also obtained variant GAL4 trap lines that label small subsets of neurons in the embryo. Specific examples are the TK-GAL4VPmad2 #5m-1 and TKBA-GAL4VPmad2 #6f-3 lines (Fig. 5C). These lines will be useful in future investigations of the function of specific neuronal subsets and of the molecular mechanisms of neuronal cell-type specification. For these investigations, use of the UAS:DsRedEx reporter, which produces a more stable DsRed fluorescence, may be beneficial for the visualization of axonal pathways (Fig. 5 Cc).

These data indicate that gene expression under UAS control accurately reflects the pattern of GAL4 activator expression and is independent of reporter transgenic lines. Thus, the UAS:nlsVenus reporter, which is suitable for marking individual cells, and the UAS:DsRedExDR and UAS:DsRedEx reporters, which produce brighter fluorescence, can be used interchangeably, depending on the goals of the experiments.

GAL4 activator-mediated UAS-reporter expression was maintained at least through F4 generation in all our GAL4 lines that were studied in detail. This indicates that use of the transposon-based trap vectors in zebrafish gives consistent expression of GAL4 activators without suffering from epigenetic silencing, a notion consistent with other zebrafish studies using Tol2 (Scott et al.,2007; Davison et al.,2007).

Analysis of the Vector Insertion Sites of Enhancer-Trap Lines

To confirm that GAL4 activator expression in the enhancer trap lines was activated by genomic enhancers and to characterize these enhancers, the insertion sites of the SB transposon-based vectors were determined in several enhancer-trap lines (Table 3). These SB transposon insertions were all flanked by TA dinucleotides and lacked pUC19 sequences present in the original plasmids (Suppl. Table 2), showing single-copy insertion of the SB transposon as reported previously (Dupuy et al.,2002; Davidson et al.,2003; Ivics et al.,1997). Cosegregation of the GAL4 expression pattern and transposon insertion in a particular locus was confirmed in the next generation by PCR for the junction sequence (Supp. Table 2, data not shown).

Table 3. Insertion Sites of Enhancer-trap Lines
Enhancer trap lineInsertion siteaNearest gene(s)Insertion position in relation to nearest gene
  • a

    Position of T in the target dinucleotide TA in the zebrafish Zv7 genome sequence.

TK-GAL4VPmad2 #5m-2Chr. 9: 25,041,138G-protein coupled receptor 183 (gpr183/ebi2)5 kb downstream
TK-GAL4VPmad2 #5m-3Chr. 18: 23,456,647Nuclear receptor subfamily 2, group F, member 2 (nr2f2/COUPTFβ)189 kb upstream
TKBA-GAL4VPmad2 #12m-3Chr. 20: 26,945,391Exocyst complex component 2 (exoc2)Forkhead gene cluster (foxq1, foxf2, foxc1b)0.5 kb upstream59-78 kb upstream
GATA2-GAL4VPmad2 #21m-1Chr. 6: 58,649,035Sp7/Osterix transcription factor (sp7)17 kb upstream

The insertion sites in the trap lines were distributed in a variety of genomic locations (Table 3). In the TK-GAL4VPmad2 #5m-3 line, the vector was inserted 189 kb upstream of the nr2f2/svp40/COUPTFβ gene that encodes an orphan nuclear receptor implicated in neural and eye development (Fjose et al.,1995). In the GATA2-GAL4VPmad2 #21m-1 line, the insertion was located 17 kb upstream of the sp7/osterix transcription factor gene that is highly expressed in the otic vesicle (Thisse and Thisse,2005). In these lines, UAS reporter expression mimicked those of nearby genes (Figs. 4 Ad, 4 Ca, 5 Ad–f; also see below for nr2f2).

In the TKBA-GAL4VPmad2 #12m-3 line, vector insertion was mapped 0.5 kb upstream of the exoc2 gene but the activation of UAS reporters in the mesoderm (Fig. 4 Be) resembled the expression pattern of the foxc1b (forkhead box c1b) gene (Topczewska et al.,2001) located in the fox cluster, 59–78 kb from the vector insertion site.

Correlation of Expression Patterns Among the Trapped Gene, GAL4 Activator, UAS-Linked Reporter Transcript, and Reporter Fluorescence

Although in the TK-GAL4VPmad2 #5m-3 line the vector insertion site was 189 kb away from the nr2f2 gene, UAS-mediated reporter expression occurred in a pattern very similar to the nr2f2 expression pattern in the CNS (Fjose et al.,1995). This strongly suggested that GAL4VPmad2 expression in the #5m-3 line was regulated by neural enhancers of nr2f2. We observed DsRedExDR fluorescence with a 2-hr time lag following nr2f2 expression. This expression lag time roughly corresponded to the maturation period of DsRedExDR, although two-step gene expression in the binary GAL4-UAS system may have also contributed to this temporal lag.

To examine the relationship between endogenous nr2f2 expression and GAL4VPmad2 expression more closely in the #5m-3 line, expression of their mRNAs was compared by in situ hybridization in 24-hpf embryos. Expression of DsRedExDR mRNA and DsRedExDR fluorescence were also compared in order to confirm the correlation between the activator expression and reporter activation. GAL4VPmad2 mRNA expression in the #5m-3 line mimicked nr2f2 expression in the CNS (Fig. 6), but there were some disparities in a part of the ventral diencephalon (Fig. 6A,C; arrowhead) and regions of the dorsal diencephalon and midbrain (Fig. 6E, arrow). These observations suggested that not all nr2f2 enhancers are equally detected by the GAL4VPmad2 vector insertion in the #5m-3 line.

Figure 6.

Comparison of expression patterns among the nr2f2 gene, the GAL4VPmad2 activator, and the UAS-linked reporter transcript and its fluorescence in the TK-GAL4VPmad2 #5m-3 line. AH: Whole-mount in situ hybridization for detecting nr2f2, GAL4VPmad2, and DsRedExDR mRNAs at 24 hpf. The specimens in C and D were processed for color development four times longer than A and B in order to visualize nr2f2 expression in the spinal cord. I,J: DsRedExDR fluorescence in a TK-GAL4VPmad2 #5m-3 embryo at 24 hpf.

The distribution of DsRedExDR mRNA in the #5m-3 line correlated well with that of GAL4VPmad2 mRNA (Fig. 6E–H), indicating that UAS:DsRedExDR expression faithfully reflects GAL4VPmad2 expression.

The hybridization signal of DsRedExDR was stronger than that of GAL4VPmad2, suggesting that mRNA levels are successfully amplified through the GAL4-UAS system (Fig. 6E–H). Moreover, DsRedExDR expression seemed to be in greater contrast than that observed with GAL4VPmad2, suggesting non-linear amplification, which is expected when using a binary expression system. Thus, the intensity distribution of DsRedExDR fluorescence (Fig. 6I,J) is most likely the outcome of non-linear amplification of the effect of enhancers that are derived from the nr2f2 gene and trapped by the GAL4VPmad2 vector.

In summary, GAL4-UAS interaction-dependent gene regulation was successfully adopted in the zebrafish model system through the use of vertebrate-adapted GAL4 activators and SB transposon-based single-copy insertion vectors.

DISCUSSION

Alleviation of GAL4VP16 Toxicity Through Alterations in the Activation Domain

To implement an enhancer trap-based GAL4-UAS system in zebrafish and widely in vertebrates, the construction of strong GAL4-based activators with minimal adverse effects even at high expression levels is required. This was the major focus of our current study. Our data confirmed previous indications that the native yeast GAL4 is too weak to use as an activator in zebrafish (Köster and Fraser,2001; Figs. 2 and 3). A fusion of the full-length VP16 activation domain to GAL4 DBD resulted in deleterious effects upon embryonic development, precluding the efficient production of viable enhancer-trap lines with the GAL4VP16-containing vectors. In the majority of cases, developmental arrest and morphological abnormalities occurred at the F1 generation (Supp. Fig. 2, Supp. Table 1). This finding suggested that GAL4VP16 exerts serious adverse effects on the developing embryo, even at moderate expression levels, when driven by endogenous zebrafish genomic enhancers. In support of this possibility, injection of a very low amount of GAL4VP16 mRNA (5 pg) was sufficient to cause developmental abnormalities in injected embryos (Fig. 2D).

We hypothesized that the toxicity of GAL4VP16 stemmed mainly from an overly robust interaction between the full-length VP16 activation domain and various components of the basal transcriptional machinery. Our attempts to ameliorate GAL4-based activators were successful by employing either multimers of the minimal activation domain of VP16 (VPmad2 and VPmad3) or the activation domain of NF-κB p65. Enhancer-trap lines expressing GAL4VPmad2, GAL4VPmad3, or GAL4NFkB were generated at a high frequency and were free of deleterious effects. These variant GAL4 activators, especially GAL4VPmad2 and GAL4NFkB, appeared to be well tolerated by zebrafish even when strong genomic enhancers activated their expression. During the preparation of this report, Asakawa et al. (2008) also reported that Gal4FF, which is essentially the same construct as GAL4VPmad2, was less toxic than GAL4VP16 in zebrafish embryos and useful for trapping experiments, thus corroborating our findings and conclusions.

The direct comparison of GAL4VP16413-470 with the other GAL4 variants showed that GAL4VP16413-470 was similar to GAL4VPmad2 regarding the transactivation potentials and toxicity, although they displayed adverse effects in different tissues (Supp. Fig. 3). Toxicity has not been described in transgenic zebrafish lines that express GAL4VP16413-470 under the control of tissue-specific regulatory elements from the genes isl1, gsc, gad2, and zic1 (Inbal et al.,2006; Sagasti et al.,2005; Sassa et al.,2007). Davison et al. (2007) also used GAL4VP16413-470 in their gene-trap constructs without observing deleterious effects. In contrast, Scott et al. (2007) observed lethality in a small fraction (4%) of their enhancer-trap lines expressing GAL4VP16413-470.

In medaka also, embryonic development is retarded and/or morphological abnormalities occur when the GAL4VP16 construct is expressed under the control of a non-specific, strong promoter or following injection of GAL4VP16 mRNA (Grabher and Wittbrodt,2004), consistent with our current observations.

Taken together, the current evidence indicates that the original GAL4VP16 cannot be tolerated by zebrafish unless expressed at very low levels, and thus is not suitable for GAL4 enhancer trapping. In contrast, GAL4VPmad2, GAL4NFkB, and GAL4VP16413-470 can be safely used for enhancer trapping.

GAL4VP16 has been employed in model animal systems other than fish and its deleterious effects following high-level expression were also noted. For example, the dose-dependent adverse effects of GAL4VP16 have been observed in transgenic mice that express GAL4VP16 in ocular tissues under the control of the Pax6 promoter (Govindarajan et al.,2005). Moreover, the toxicity of GAL4VP16 is suggested in Drosophila embryos because it was difficult to obtain active GAL4VP16 lines (Viktorinova and Wimmer,2007). We suggest that the variant GAL4 activators developed in this study, namely GAL4VPmad2, GAL4VPmad3, and GAL4NFkB, will be beneficial to the utilization of GAL4-UAS-based gene expression systems in a variety of model animal systems.

Optimum Promoter and Activator Combinations for Enhancer-Trap-Dependent GAL4 Drivers in Zebrafish

The enhancer-trap lines that were generated using the variant GAL4 activator-containing trap vectors (TK-GAL4VPmad2, GATA2-GAL4VPmad2, and TK-GAL4NFkB) gave strict tissue-specific expression of UAS reporters without background expression, reflecting the weak promoter activities of TK and GATA2 and the moderate transactivation potentials of GAL4VPmad2 and GAL4NFkB. The hsp70 promoter was also successfully used in enhancer trapping by other investigators, although its use resulted in low-level, enhancer-independent expression in the heart, skeletal muscle, and lens (Asakawa et al.,2008; Scott et al.,2007). Considering the lower toxicity of GAL4VPmad2, GAL4NFkB, and GAL4VP16413-470, combinations of these moderate activators and weak promoters are probably suitable for developing enhancer-trap lines that can activate genes in a tissue-specific manner. The stronger activator GAL4VPmad3 may also be employed in combination with these weak promoters, but only with caution, considering its toxic effects at high expression levels.

On the other hand, the four combinations that contained the TKBA promoter and/or the GAL4VPmad3 activator (TKBA-GAL4VPmad2, TK-GAL4VPmad3, TKBA-GAL4VPmad3, and TKBA-GAL4NFkB) resulted in non-specific and weak expression at various frequencies in addition to its tissue-specific expression. Because the weak expression patterns from these promoter-activator combinations were similar to one another, these expression patterns are likely to be independent of enhancer trapping. Probably as a result of the high activity of the TKBA promoter and/or the relatively strong transactivation potential of GAL4VPmad3, these combinations led to weak enhancer-independent GAL4 activator expression, which subsequently caused weak, widespread expression of UAS- dependent reporter expression. However, with the exception TKBA-GAL4VPmad3, the enhancer-trap vectors utilizing these combinations of promoters and activators may be useful for trapping weak enhancers because success rates of establishing enhancer-trap lines were high using these vector constructs (Table 1). The combination of a stronger promoter and a stronger activator, exemplified by TKBA-GAL4VPmad3, appears to occasionally result in an intolerable level of GAL4 activator expression and thus is suggested not to be suitable for enhancer trapping.

Binary Expression System-Based Enhancer Trapping in Zebrafish

Using the enhancer-trap-dependent variant GAL4-expressing vectors, we performed a pilot screening of enhancer-trap zebrafish lines. The vectors appeared to trap genomic enhancer activities that included early neural enhancers and neuron-specific enhancers as shown in Figure 5. Among the trap lines, activator expression showed variation in tissue specificity, including cephalic placodes, notochord, hypochord, etc (Fig. 4, Table 2). These results hold promise for the future generation of a large variety of trap lines through large-scale trapping experiments.

We employed the Sleeping Beauty transposon, a member of the Tc1/mariner family, in our GAL4-based enhancer-trap constructs in order to assure efficient and single-copy insertion of the trap vector. Scott et al. (2007) and Asakawa et al. (2008) successfully utilized the Tol2 transposon, a member of the hAT family. These data suggest that GAL4-based enhancer trapping can be performed with various transposon and retrovirus systems in zebrafish. As variability of the target site preference is indicated for different transposition systems (reviewed in Mates et al.,2007), the use of various types of transposons may expand the specificity repertoires of trap lines. Moreover, the use of different promoters in the enhancer-trap vectors may further facilitate the trapping of various regulatory sequences in the genome by creating preferable conditions for enhancer-promoter interactions. Thus, the use of different combinations of transposon systems and promoters in enhancer-trap vectors will enhance the availability of GAL4-based driver lines for future application of the GAL4-UAS system.

The binary expression systems utilizing other DNA-binding domains have also been used in zebrafish. Emelyanov and Parinov (2008) have recently reported the bacterial LexA-based expression system, in which they achieved mifepristone-inducible expression in zebrafish embryos by using a LexA fusion with the progesterone receptor ligand-binding domain (LexPR). In this report, several LexPR driver lines were obtained through enhancer trapping with the maize transposon system. In addition, tetracycline (Tet)-On expression system was successfully used for heart-specific activation of the GFP reporter in a ligand-dependent manner (Huang et al.,2005). These binary expression systems would possibly be used in combination with the GAL4-UAS system in zebrafish. In fact, GAL4 induction can be performed with the Tet-On system in Drosophila (Duffy,2002). These combined systems would allow us more flexibly to design experiments for control of gene expression.

EXPERIMENTAL PROCEDURES

Plasmid Construction

The complete nucleotide sequences of the plasmids used in this study are available in a GenBank format upon request.

GAL4 Derivatives

The following coding sequences of GAL4 derivatives were cloned into pCBA3 (Kamachi et al.,2008). Full-length GAL4 was amplified by PCR from the yeast genome using the primer set 5′-ggggatccaccATGAAGCTACTGTCTTCTATCGAAC (sense) and 5′-gggaattcTTACTCTTTTTTTGGGTTTGGTGG (antisense; lowercase letters indicate linker sequences to create restriction sites) and cloned between the BamHI and EcoRI sites of pCBA3. GAL4(ΔCR) was constructed by combining N-terminal and C-terminal fragments amplified from pCBA3-GAL4 using the additional internal PCR primers 5′-gggaattcGGATCTAGAAGCCAACGTGTATCT (antisense for the N-terminal fragment) and 5′-gggaattcAATTTTAATCAAAGTGGGAATATTGC (sense for the C-terminal fragment). GAL4VP16 was derived from pSGVP (Sadowski et al.,1988), the fragment of which was amplified using the primers 5′-ggggatccACCATGAAGCTACTGTCTTC- TATC and 5′-ggtctagaCTACCCACCGTACTCGTCA and cloned between the BamHI and XbaI sites of pCBA3. The VP16 minimal activation domain used in this study has been previously described (Baron et al.,1997). Sequences of synthetic oligonucleotides encoding this peptide were: sense strand, 5′-aattcGCCGACGCCCTGGACGATTTCGATCTGGACATGC- TGc; antisense strand, 5′-aattgCAGCATGTCCAGATCGAAATCGTCC- AGGGCGTCGGCg, adapted to codon usage in zebrafish. The protruding 5′- and 3′-ends of the double-stranded oligonucleotides were compatible to the cleavage sites of restriction enzymes EcoRI and MfeI, respectively. To construct GAL4VPmad2, GAL4VPmad3, and GAL4VPmad4, two, three, and four copies of these oligonucleotides, respectively, were tandemly inserted into pCBA3 together with the sequence encoding GAL4 DBD. To generate GAL4NFkB, the sequence encoding the mouse NF-κB p65 activation domain was amplified by PCR from pCMV-AD (Stratagene) using the primers 5′-gggaattcCCTTCTGGGCAAATCTCAAACC and 5′-ggtctagaTTATTGGCCGCTGGAGC- TGATCT.

Enhancer-Trap Vectors

Enhancer-trap vectors were constructed using pT2.2, a modified version of the pT2 SB transposon vector (Cui et al.,2002), in which a truncated 2-kb pUC19 plasmid backbone (PvuII to AatII) was used based on the observation that a decrease in the DNA length outside of the SB transposon generally increases the transposition efficiency (Izsvak et al.,2000). The enhancer trap cassettes were placed between the IR/DR sequences of pT2.2 and contained one of three promoters (TK [−200 to +31, human HSV type I], TKBA [hybrid promoter of TK {−200 to −27} and carp β-actin {from −27 to +73}], or GATA2 [562 bp]); one of the GAL4-derived activators (GAL4VP16, GAL4VPmad2, GAL4VPmad3, or GAL4NFkB); and the ocean pout AFP poly(A) signal. The GATA2 promoter fragment was amplified by PCR from the zebrafish genome using following primers: 5′- ggggatccCGATATTCATTAATAGAA- TAGAGGCAT and 5′-ggggtcgacTCCACAGCCCGATTGTTAAAGTCTC. The 5′-end of this fragement corresponds to that of P5-GM2 described in Meng et al. (1997). The first set of trap constructs also included the shortened intron 1 sequence of carp β-actin, in which the sequence from the HindIII to BglII sites was removed in order to minimize a vector size, upstream of GAL4VP16. As a weak cryptic promoter was found to be present in the intron, we removed the shortened intron 1 sequence from the second set of trap vectors shown in Figure 1E.

UAS-Reporter Vectors

The UAS-reporter vectors UAS:DsRedEx, UAS:DsRedExDR, UAS:nlsVenus, and UAS:Venusluc were constructed by inserting the coding sequences of the fluorescent proteins DsRedExpress, DsRedExpressDR (Clontech), nlsVenus (Venus [Nagai et al.,2002] fused to the SV40 large T antigen nuclear localization signal) and Venusluc (Venus fused to the firefly luciferase from pGL3-Basic [Promega]) individually into pG10-BA53si. pG10-BA53si contains 10 copies of the GAL4 binding sequence, derived from pG5luc (Promega), upstream of the β-actin minimal promoter (from −53) and the shortened intron 1 sequence of the carp β-actin gene. UAS:DsRedEx/I-SceI was constructed by inserting the reporter cassette into the I-SceI-pBSII vector (Thermes et al.,2002).

Promoter-Linked Venusluc Vectors

To assess the activity of the TK, TKBA, and GATA2 promoters, each promoter sequence was placed upstream of the Venusluc fusion gene.

Synthetic mRNAs for Embryo Injection

mRNAs for injection were transcribed in vitro using the mMessage mMachine SP6 kit (Ambion) and purified with MEGAclear (Ambion). SB11 transposase mRNA was transcribed from a pCBA3-SB11 plasmid in which the coding sequence of SB11 from pCMV-SB11 (Geurts et al.,2003) was cloned into pCBA3 (Kamachi et al.,2008). mRNAs of GAL4 and the GAL4 fusion activators were transcribed from the pCBA3-GAL4 activator plasmids. GAL4VP16413-470 mRNA was transcribed from pCS2-GAL4VP16413-470 (pCMV-GVP; Köster and Fraser,2001).

Generation of Enhancer Trap Lines

Zebrafish embryos were obtained by natural mating of wild-type TL fish and reared at 28.5°C in 0.03% Red Sea salt solution. One-cell-stage embryos were injected through the chorion with approximately 1 nl of a solution containing the enhancer-trap transposon vectors (20 pg/nl), SB11 transposase mRNA (50 pg/nl), and 0.05% Phenol Red (Sigma).

Injected embryos were raised to adulthood and crossed with UAS:DsRedExDR reporter fish. At least 100 embryos were usually collected for each founder and screened for DsRedExDR expression at 1 dpf. In addition, to estimate the success rate of vector insertion and enhancer trapping, DsRedExDR-negative embryos were checked for GAL4-derived activator sequences by genomic PCR using the following primers: forward 5′-GGTCAAAGACAGTTGACTGTAT- CGCC (GAL4VP16, GAL4VPmad2, and GAL4VPmad3) or 5′-TGA ACCAGGGTGTGTCCATGTCTCA (GAL4NFkB) and reverse 5′- CATCAAATCTCAACAGTCTCCACAGG.

F1 embryos that displayed interesting patterns of reporter expression were raised to adulthood in order to establish stable line stocks. These F1 adult fish were crossed with wild-type TL fish, and DsRedExDR-expressing embryos (embryos harboring both activator and reporter) were selected as F2 progeny.

Generation of UAS-Reporter Lines

Wild-type TL embryos were injected at the one-cell stage with approximately 1 nl of a solution containing the SfiI-SfiI fragment of UAS:DsRedExDR or UAS:nlsVenus vector (30 pg/nl) and 0.05% Phenol Red. Injected embryos were raised to adulthood and crossed with wild-type AB fish. F1 embryos were collected and screened by PCR for the presence of DsRedExDR or nlsVenus reporter sequences. Reporter sequence-positive founder fish were crossed with wild-type AB fish and F1 embryos were raised to adulthood. The reporter sequence-positive F1 fish were then crossed with each other and reporter-homozygous fish among the F2 progeny were selected and maintained. The UAS:DsRedEx zebrafish line was established by using the I-SceI meganuclease method (Thermes et al.,2002). In this latter case, TL embryos were injected with approximately 1 nl of a solution containing UAS:DsRedEx/I-SceI (25 pg/nl), I-SceI (0.5 unit/μl, NEB), BAS (100 ng/μl), and 0.05% Phenol Red in 1× I-SceI buffer.

To validate GAL4-activator-dependent expression of UAS-reporter genes, embryos of the UAS-reporter lines were injected with GAL4-VP16 mRNA (2.5 pg). These embryos developed widespread fluorescence of DsRedEx, DsRedExDR, and nlsVenus.

Luciferase Assay

To assess the promoter activities of TK, TKBA, and GATA2, approximately 1 nl of a solution containing each of the promoter-linked Venusluc vectors (20 pg/nl) was injected with the reference Renilla luciferase vector (phRG-TK [Promega]) (5 pg/nl) into one-cell-stage embryos of wild-type TL zebrafish. For assessment of the transactivation potential of the GAL4 fusion activators, approximately 1 nl of a solution containing each GAL4 activator mRNA (5–30 pg/nl) was injected with the Venusluc reporter vector (UAS:Venusluc, 20 pg/nl) and the reference Renilla luciferase mRNA (1 pg/nl) into one-cell-stage embryos of wild-type TL zebrafish.

Injected embryos were collected at 10 hpf and lysed in Passive Lysis Buffer (Promega) at 20 μl per embryo. For one injection experiment, eight embryos were used as a group to make a lysate and at least two groups were examined to confirm the consistency of injections in one injection experiment. Luciferase activities of these lysates were measured using the Dual-Luciferase Reporter Assay System (Promega). Firefly luciferase activity was normalized using Renilla luciferase activity levels. Data are shown as the mean values of the relative luciferase activities obtained from at least two independent injection experiments.

DNA Sequence Analysis of Genomic Regions Flanking SB Transposon Insertions

Genomic DNA was obtained from the tails of zebrafish enhancer-trap lines using the ChargeSwitch gDNA Tissue Kit (Invitrogen). Genomic regions flanking SB transposon insertions were amplified from genomic DNA using splinkerette PCR (Dupuy et al.,2002). The PCR fragments were sequenced and insertion sites were identified using the Ensembl zebrafish genomic sequence database (Zv7) as a reference.

In Situ Hybridization

A partial sequence for the zebrafish nr2f2 cDNA was amplified by RT-PCR using the primer set 5′-CCCTCTA GACCATTCTACCTGATAAGACA- TCCC and 5′-GGGGATCCTGTTACAGACCGAAAAGGAGC, cloned between the BamHI and XbaI sites of pBSIISK(+), and used as a template for antisense RNA probe synthesis. pCBA3-GAL4VPmad2 and pCBA3-DsRedExDR were used as a template for preparation of RNA probes. Whole-mount in situ hybridization of zebrafish embryos was performed using a standard procedure (Schulte-Merker,2002).

Acknowledgements

We thank our laboratory members for stimulating discussions and are grateful to Shigemi Toyoshima for maintaining the zebrafish. We express special thanks to Perry B. Hackett for his guidance regarding the SB transposon system and for a critical reading of the manuscript. We also thank Yasuo Kawasaki for providing the yeast genomic DNA, Atsushi Miyawaki for the Venus vector, Jochen Wittbrodt for I-SceI-pBSII, and Reinhard Köster for pCMV-GVP (pCS2-GAL4VP16413-470) and information concerning this construct. This work was sponsored in part by The Ministry of Education, Culture, Sports, Science and Technology of Japan (19-5118 to E.O.; 16027229, 20017019, and 18570197 to Y.K.; and 17107005 to H.K.). E.O. and Y.O. are recipients of fellowships from the Japan Society for the Promotion of Science for Japanese Junior Scientists.

Ancillary