A targeted gene expression system using the tryptophan repressor in zebrafish shows no silencing in subsequent generations

Bipartite gene expression systems allow selective gene expression in a tissue-specific manner in vivo (reviewed by del Valle Rodríguez et al., 2012; Elliott, 2008; Halpern et al., 2008). They consist of two parts: a driver line and an effector/reporter line. In the driver line, a tissue-specific promoter drives a transcriptional activator, while in the effector/reporter line a target gene is placed under control of the binding site [upstream activation sequence (UAS)] of the transcriptional activator. The cross between the driver and effector/reporter lines allows expression of target genes exclusively in tissues in which the specific promoter is functional. The advantage of these bipartite systems is that the same effector can be expressed in different tissues simply by crossing to different driver lines. Similarly, the same driver can easily be used to promote the expression of various effectors. In addition, effector lines for potentially deleterious gene products can be maintained without expression until crossed to driver lines.

Gal4/UAS was the first gene expression system to be developed in Drosophila (Brand and Perrimon, 1993). It uses the yeast Gal4 transcription factor, which coordinates the expression of genes needed for utilization of galactose through a common UAS (gUAS; Fig. 1A,B). The Gal4 transcriptional activator was integrated randomly in the Drosophila genome, landing at times adjacent to enhancers expressed in specific tissues and cell populations, thereby creating Drosophila lines in which Gal4 was expressed in a tissue-specific manner, which was termed ‘enhancer trapping’. Additionally, reporter/effector lines were generated, in which either lacZ (a reporter gene) or even-skipped (an effector gene) was placed downstream of gUAS. By crossing the enhancer lines with the reporter/effector lines it became possible not only to visualize the ‘trapped’ cell populations but also to misexpress even-skipped in specific cell populations in an effort to determine its role in their development. Since its establishment, the Gal4/gUAS system has facilitated a wide variety of techniques, including gene overexpression and misexpression, targeted gene knockouts, targeted cell ablation, disruption of neuronal synaptic transmission, and in vivo cell tracing followed by time-lapse microscopy during development (del Valle Rodríguez et al., 2012; Elliott, 2008; Halpern et al., 2008). Owing to its usefulness, this system has been adopted in several other model organisms, such as Arabidopsis (Engineer et al., 2005), Xenopus (Hartley et al., 2002), Medaka (Grabher and Wittbrodt, 2004), zebrafish (Asakawa et al., 2008; Scheer and Campos-Ortega, 1999; Scott et al., 2007), mouse (Hu et al., 2004; Ornitz et al., 1991; Rowitch et al., 1999) and human cell culture (Webster et al., 1988).

A serious disadvantage of the Gal4/gUAS system is that the UAS is silenced in subsequent generations in vertebrates due to methylation at CpG nucleotides (Akitake et al., 2011; Goll et al., 2009) (Fig. 1B). This leads to the silencing of the UAS-regulated effector/reporter gene as early as the first (F1) generation and necessitates continual reestablishment of these lines. To circumvent this problem, we have developed a new bipartite gene expression system that employs the tryptophan repressor (TrpR) and its UAS (tUAS), which are responsible for tryptophan biogenesis in E. coli (Gunsalus and Yanofsky, 1980). The minimal tUAS lacks CpGs (Fig. 1D) (Li et al., 1995), suggesting that it would not be silenced by methylation. We created tUAS effector/reporter zebrafish lines and found no indication of silencing as far as the fourth (F4) generation. Taking advantage of the wealth of data on the structure and function of TrpR, we identified TrpR mutants with reduced transcriptional activity in zebrafish, for use in cases where lower levels of effector protein expression are desired. Finally, we found that the TrpR/tUAS system works well in mammalian cell culture, demonstrating that this approach will be broadly applicable. The TrpR/tUAS system is an excellent alternative to the Gal4/gUAS system, and it can also be combined with Gal4/gUAS to permit combinatorial regulation of effector expression in vivo.

To make the driver construct, we fused the entire TrpR coding region to the Gal4 activator domain (G4AD) (Fig. 1C), which can be inhibited by Gal80 (Fujimoto et al., 2011; Traven et al., 2006) and thus affords an additional level of transcriptional regulation (see below). We added a nuclear localization signal (nls) and placed the nlsTrpR-Gal4AD fusion construct under the CMV promoter (pCMV:nlsTrpR-G4AD). For the reporter construct, we used a multimerized (3×) TrpR UAS (tUAS) (Fig. 1D) to drive DsRed fluorescent protein (ptUAS:DsRed). When we transfected the constructs into HEK 293 cells or injected them into zebrafish embryos, we found that TrpR/tUAS constructs were able to induce transcription of the DsRed reporter gene both in the cell line and in embryos (data not shown). This indicated that it would be feasible to create transgenic animals and test this expression system in stable lines.

To build the constructs used for generating zebrafish transgenic lines, we used the Tol2 Gateway cloning system (Kwan et al., 2007; Villefranc et al., 2007), which will make it easy to swap promoters and effector/reporter genes in the future (Table 1). For the driver construct, we chose myosin 6b (myo6b) (Obholzer et al., 2008) and ribeye A (ribA; ctbp2a - Zebrafish Information Network) (Odermatt et al., 2012) promoters to drive nlsTrpR-G4AD (pmyo6b:nlsTrpR-G4AD and pribA:nlsTrpR-G4AD, respectively). In zebrafish, the myo6b promoter is expressed in auditory, vestibular and lateral line hair cells, whereas ribA is also expressed in these plus photoreceptor, bipolar and pineal cells. Since the transgenic driver lines are not visible until crossed to transgenic reporter lines, we added to the destination constructs the alpha A crystallin promoter (Hesselson et al., 2009) driving the expression of the red fluorescent protein Cherry (cryaa:Cherry, abbreviated CC), which promotes the expression of Cherry in the lens, allowing us to easily identify transgenic fish. For the reporter constructs, we used tUAS (Fig. 1D) to drive tagRFP (ptUAS:tRFP) (Merzlyak et al., 2007) or the nuclear-localized photoconvertible protein nlsEos (ptUAS:nlsEos) (Curran et al., 2010; Wiedenmann et al., 2004), which is particularly useful for lineage tracing (McGraw et al., 2012). We added cryaa:Venus (abbreviated CV) to identify transgenic carriers using yellow fluorescent protein expression in the lens. We injected each DNA construct together with Tol2 transposase RNA into one-cell stage zebrafish embryos (Asakawa et al., 2008), grew the embryos to adulthood and screened for stable insertions.

We isolated one insertion for Tg(CV,tUAS:tRFP)^w80 and several insertions for each of Tg(CV,tUAS:nlsEos)^w81, Tg(CC,myo6b:nlsTrpR-G4AD)^w83 and Tg(CC,ribA:nlsTrpR-G4AD)^w85 (Fig. 2). Both tRFP and nlsEos appeared to be ubiquitously expressed when 5 pg nlsTrpR-G4AD mRNA was injected into one-cell stage transgenic embryos (Fig. 2A,B). However, when the Tg(CV,tUAS:tRFP)^w80 reporter line was crossed to the Tg(CC,myo6b:nlsTrpR-G4AD)^w83 driver line, only a subset of hair cells was labeled in double-transgenic embryos (Fig. 2E, Fig. 4A). Since this pattern was not observed when the Tg(CV,tUAS:nlsEos)^w81 reporter line was crossed to Tg(CC,myo6b:nlsTrpR-G4AD)^w83 (Fig. 2D), we concluded that Tg(CV,tUAS:tRFP)^w80 is variegated due to an insertion-specific position effect. We also noticed that Tg(CC,ribA:nlsTrpR-G4AD)^w85 could induce expression of nlsEos when crossed with Tg(CV,tUAS:nlsEos)^w81 (Fig. 2F), but was unable to induce tRFP expression when crossed with Tg(CV,tUAS:tRFP)^w80 (data not shown). Since the ribA promoter is weak, we suspect that the insertion site of ptUAS:tRFP in our reporter line requires relatively high levels of TrpR-G4AD to activate transcription. We are in the process of screening for new Tg(CV,tUAS:tRFP)^w80 founders that lack insertion-specific effects.

Since the minimal tUAS has no CpG dinucleotides (Fig. 1D), we predicted that it would not be silenced by methylation. To determine whether silencing occurs in our transgenic reporter lines, we raised four generations of Tg(CV,tUAS:tRFP)^w80 and three generations of Tg(CV,tUAS:nlsEos)^w81 from our F0 founders. We then tested for silencing in two ways. First, we outcrossed different individuals from the F3 generation of Tg(CV,tUAS:tRFP)^w80 to individuals from the F2 generation of the driver line Tg(CC,myo6b:nlsTrpR-G4AD)^w83, and scored the progeny for the presence of transgenic markers: the lens markers (indicating the presence of the transgenes) and hair cell expression (indicating tissue-specific expression of the reporter gene). We reasoned that if the tUAS of the reporter line was silenced, we would see no hair cell expression in some of the progeny that are positive for both the red and green lens markers (embryos transgenic for both the driver and reporter line). Conversely, if we always saw hair cell expression in embryos that were double positive for red and green lens markers that would indicate that tUAS is not silenced. With this in mind, we outcrossed 18 F3 Tg(CV,tUAS:tRFP)^w80 adults to F2 Tg(CC,myo6b:nlsTrpR-G4AD)^w83 adults and found that 24.8% of the combined progeny (171/690) had red and green lens expression and were also positive for hair cell expression (Fig. 3). We found no cases of progeny positive for both lens markers but negative for hair cell expression, demonstrating that no silencing of the reporter line had occurred in the F3 generation of the reporter line.

Second, we assessed silencing by looking at individual lateral line neuromasts, which consist of hair cell clusters, within an embryo and by scoring hair cell expression in outcrosses of adults from different generations of the reporter line. We crossed a single adult carrier from each F1-F4 generation of Tg(CV,tUAS:tRFP)^w80 or F1-F3 generation of Tg(CV,tUAS:nlsEos)^w81 to an F2 or F3 adult carrier from the Tg(CC,myo6b:nlsTrpR-G4AD)^w83 line (Fig. 4). We sorted for double-transgenic embryos, as indicated by the presence of red and green lenses, and scored for tRFP (Fig. 4A-D) or nlsEos (Fig. 4F-H) expression in neuromast hair cells. We additionally stained for Parvalbumin, which labels all mature hair cells within the neuromasts. We found that tRFP was consistently expressed in a subset of mature neuromast hair cells due to the insertion-specific effect of the pTol2-CV,tUAS:tRFP construct. However, the ratio of hair cells expressing tRFP to the total number of neuromast hair cells (tRFP/Parvalbumin) across the four Tg(CV,tUAS:tRFP)^w80 generations was not statistically different (Fig. 4E), confirming that tUAS is not silenced. Similarly, when we assessed nlsEos expression across three Tg(CV,tUAS:nlsEos)^w81 generations, we found that all mature hair cells (Parvalbumin positive) also expressed nlsEos, again proving that tUAS is not silenced (Fig. 4H). We noted that some nlsEos-positive cells were negative for Parvalbumin, as shown by an nlsEos/Parv ratio greater than 100% in the F2 and F3 crosses, probably indicating the presence of some immature hair cells in the neuromasts. We assessed the variability of expression within a clutch of embryos in each generation and obtained similar results (supplementary material Fig. S1). From these experiments we conclude that tUAS reporter lines are not silenced across generations.

An advantage to using the Gal4 activator domain in nlsTrpR-G4AD is that we can regulate its ability to activate transcription by employing the Gal4 inhibitor Gal80 (Carrozza et al., 2002; Traven et al., 2006; Wu et al., 1996). This would be a useful feature in cases in which the manipulation of subgroups of cells is needed. For example, using a promoter that drives Gal80 expression in a partially overlapping manner with that used to drive nlsTrpR-G4AD, it would be possible to limit the tissues in which nlsTrpR-G4AD activates expression (Fujimoto et al., 2011). To determine whether this is feasible in our system, we crossed the Tg(CC,myo6b:nlsTrpR-G4AD)^w83 line to the Tg(CV,tUAS:nlsEos)^w81 line. We injected some of the one-cell stage embryos with pTol2-CG2,myo6b:Gal80IREStRFP DNA, a construct that expresses Gal80 under the myo6b promoter and marks the cells that received the plasmid using an internal ribosomal entry site (IRES) sequence to drive tRFP expression (Fig. 5A-B′). As predicted, cells expressing Gal80, as indicated by the presence of cytoplasmic tRFP, show inhibition of Eos expression in the nucleus, whereas cells negative for cytoplasmic tRFP show Eos expression in the nucleus (Fig. 5B-B′). Note that, because the larvae were not reared in the dark, nlsEos is detected both in the green and red channel due to some level of protein photoconversion. This experiment demonstrates that Gal80 can be used together with the TrpR/tUAS system to further control gene expression.

There are many cases in which it is advantageous to drive an effector protein at submaximal levels, e.g. if a biological sensor affects the cell when expressed at higher levels. Fortunately, there is a large body of work analyzing mutations in TrpR. In one study, a key residue in the DNA-binding domain [threonine 81 (T81)] was sequentially changed to 19 alternative amino acids and the activity of TrpR then measured in E. coli (Pfau et al., 1994). This produced a series of TrpR mutants with varying activity from mildly affected to virtually inactive. We tested a variety of these mutants in our system. In general, we found that the effects of each of the mutations were considerably less severe in zebrafish than reported for E. coli, although the same trend was observed (data not shown). From this analysis we were able to identify two mutants that had reduced activity compared with wild type but still retained significant activity using a luciferase assay in early zebrafish embryos. We found that T81M was ∼5× less active than wild type, and that T81A was ∼11× less active (Fig. 6A,B). These results demonstrate that the ability of TrpR to regulate transcription can be ‘tuned’ using mutations that affect DNA binding.

The problem of Gal4 UAS methylation exists not only in zebrafish but also in all vertebrates and plants. To explore whether TrpR could be used in systems other than zebrafish, we examined the ability of TrpR to drive luciferase expression in human HEK 293 cells. As shown in Fig. 7, TrpR is a very potent activator of transcription in mammalian cells. Similar to results obtained in zebrafish embryos, the T81M mutant was ∼5× less effective and the T81A mutant was ∼14× less effective than wild type. Thus, the TrpR system will be effective in other vertebrate systems in addition to zebrafish.

We describe a new bipartite gene expression system that relies on the use of the tryptophan repressor (TrpR). Unlike the commonly used Gal4/gUAS system, our TrpR system is not subject to gene silencing as its UAS does not contain CG sequences. We have demonstrated that expression is stable for three to four generations in two zebrafish reporter lines, even when expression is examined at the cellular level. Furthermore, we have shown that TrpR works in mammalian cell culture; consequently, we expect that it will be useful in all vertebrate systems, and there is no reason to believe that it will not work in invertebrates as well. The activity of TrpR can be ‘tuned’ using mutations in the DNA-binding region, which is particularly valuable for proteins such as biological sensors that have to be expressed at a level at which they are useful but which does not disrupt development. Finally, owing to use of the Gal4AD, Gal80 can be used to shut off gene expression temporally in specific spatial domains.

The existence of an alternative bipartite system is also useful for combinatorial experiments. For example, using one promoter to drive Gal4 and another to drive TrpR it is possible to express combinations of effectors in overlapping domains. We note that TrpR is not the only other bipartite system available. The Q-system, which utilizes the regulatory genes normally needed for quinic acid catabolism in Neurospora crassa, has also been studied (Potter et al., 2010), although its UAS also contains CpG sites for methylation (GGGTAATCGCTTATCC). LexA has been used in zebrafish to drive expression (Emelyanov and Parinov, 2008) and, in principle, the lexA UAS should not be silenced, although this system has not been studied over multiple generations as we have done here.

Although the TrpR/tUAS system worked well with a moderate level promoter, such as myo6b, we observed problems with the very strong promoter neuroD (McGraw et al., 2012). The embryos showed correct reporter expression at earlier developmental stages; however, we observed neuronal cell death and a general pericardial and periocular edema by 5 days post-fertilization. Although this does not preclude studies of the first few days of development using transient injections of driver constructs, it is problematic for the establishment of stable driver lines using strong promoters and analysis of larvae and adults. The observed toxicity could be due to a general titration of transcription factors from strong transcriptional activators known as ‘squelching’ (Gill and Ptashne, 1988; Habets et al., 2003) or to specific toxicity resulting from the TrpR protein spuriously binding a sequence in the zebrafish genome and activating a gene that is toxic to the embryo. When we generated fusion constructs of different transcriptional activators [Gal4AD and partially crippled VP16 (Asakawa et al., 2008; Distel et al., 2009)], with either TrpR or Gal4 and injected the in vitro transcribed mRNAs into embryos, we saw that TrpR mRNA was consistently more toxic than Gal4 mRNA regardless of which transcriptional activator was fused to it (data not shown). This suggests that the TrpR transcriptional activator is toxic to cells when expressed from strong promoters and implies that, currently, TrpR/tUAS is compatible only with moderate or weak promoters. Since there is wide variability in the expression strength of constructs produced using the Tol2 system according to the site of integration, one solution to the problem of strong promoters is to screen F0 carriers for those that do not show toxicity due to lower levels of expression from the promoter. Alternatively, a suboptimal translation initiation site (Kozak sequence) could be placed in front of TrpR with strong promoters, thereby reducing TrpR levels. Although this does place some limitations on the system, many promoters are moderate to weak and will be well suited for this approach, and solutions are available with promoters that are naturally strong. This might also be a peculiarity of zebrafish, as we did not observe toxicity when expressing TrpR in HEK 293 cells.

In conclusion, the lack of silencing and the ability to tune transcriptional activation with different TrpR mutants makes TrpR/tUAS a valuable alternative to the existing Gal4/gUAS system. We have already produced two reporter lines (tRFP and nlsEOS), which will be accessible through the Zebrafish International Resource Center (ZIRC), and additional effector and driver lines can readily be made using our Gateway system clones.

Using pDsRed-Express-N1 (Clontech) as a template, the whole CMV promoter was replaced by a HindIII-BglII-EcoRI linker using (5′-3′) TAAGCTTAGATCTGAATTCA and CCGGTGAATTCAGATCTAAGCT followed by insertion of three copies of the tUAS linker using GATCTGTACTAGTTAACTAGTACTCAGTCAGTCAGT and GATCACTGACTGACTGAGTACTAGTTAACTAGTACA into the BglII site. A minimal CMV promoter containing only the TATA box was PCR amplified from pDsRed-Express-N1 using primers GGGGAATTCTAGGCGTGTACGGTGGGA and GGGACCGGTGCGATCTGACGGTTCACTAAA and inserted into the EcoRI-AgeI sites of multimerized (3×) TrpR UAS (tUAS)-containing construct.

The UAS region was amplified from ptUAS:DsRed using primers GGCGGCGGTACCGGATAACCGTATTACCGCCATGC and GGCGGCGGATCCGGTGGCGACCGGTGC, which add a KpnI site to the 5′ end and a BamHI site to the 3′ end, then cloned into the KpnI and BglII sites of pGL3-Basic.

TrpR was amplified from E. coli using primers (the initiation codon is underlined) CGAATTCAGGATGGCACCCAAGAAGAAGAGGAAGGCCCAACAATCACCCTATTCAGC and CGTCGACCCATCGCTTTTCAGCAACACCTCTTC, which add EcoRI and SalI sites along with a nls at the N-terminus. It was cloned into the CS2 vector, which contains a CMV promoter, after mutating an internal SalI site without changing the protein sequence. The Gal4AD along with an SV40 polyadenylation site (Distel et al., 2009) were inserted between the SalI and NotI sites to produce the fusion protein nlsTrpR-G4AD.

These plasmids were generated by mutagenizing pCMV:nlsTrpR-G4AD with CGGCGCAGGCATCGCGAtGATTACGCGTG and GGCGCAGGCATCGCGgCGATTACGCGTG (the lowercase base indicates the introduced mutation).

cryaa:venus (abbreviated CV) was PCR amplified from pins:cre_cryaa:venus plasmid (gift of D. Stainier, University of California, San Francisco, CA, USA) using primers GGCGGCAGATCTATTAATAGTGTGCATTCAGTGCAG and GGCGGCAGATCTCACCGCGGTGGCG, which add BglII sites, then cloned into the BglII sites of pDestTol2pA2 (Kwan et al., 2007).

cryaa:cherry (abbreviated CC) was PCR amplified from phsp70l:loxP-mCherry-STOP-loxP-H2B-GFP_cryaa:Cherry plasmid (gift of D. Stainier), then cloned as with pDestTol2CV using the same primers used for cryaa:venus.

A gift from K. Kwan and C. B. Chien (University of Utah, Salt Lake City, UT, USA) (Kwan et al., 2007).

Gifts form T. Nicolson (Oregon Health and Science University, Portland, OR, USA) (McGraw et al., 2012; Obholzer et al., 2008).

The ribA promoter (1.8 kb) was amplified from a pribeyeA:ribeyeCherry plasmid (Odermatt et al., 2012) (gift from L. Lagnado, University of Cambridge, Cambridge, UK) using primers GGGGACAACTTTGTATAGAAAAGTTGCCAGGCTTTGAAGTCGTCACTC and GGGGACTGCTTTTTTGTACAAACTTGCTATACCTTACTCACAGGGAAG and Gateway cloned into pDONRP4-P1R.

Multimerized (3×) TrpR UAS (tUAS) was PCR amplified from ptUAS:DsRed using primers GGGGACAACTTTGTATAGAAAAGTTGGGATGCATTAGTTATTAAGCTTAGATC and GACGTTCTCGGAGGAGGCCTGCAGGGCGACCGGTGCGATCTGA, which add HindIII and PstI sites, and cloned into p5E MCS (Kwan et al., 2007) using HindIII and PstI.

nlsTrpR-G4AD was PCR amplified from pCMV:nlsTrpR-G4AD using primers GGGGACAAGTTTGTACAAAAAAGCAGGCTGGACCATGGCACCCAAGAAG and GGGGACCACTTTGTACAAGAAAGCTGGGTGTGGTTTGTCCAAACTCATCAATG and Gateway cloned into pDONR221.

A gift from J. Bonkowsky (University of Utah, Salt Lake City, UT, USA) (Fujimoto et al., 2011).

These were generated by site-directed mutagenesis of pME-nlsTrpR-G4AD as described above.

This construct was described previously (Prendergast et al., 2012).

Gifts from K. Kwan and C. B. Chien (Kwan et al., 2007).

pTol2-CV,tUAS:tRFP, pTol2-CV,tUAS:nlsEos, pTol2-CC,myo6b:nlsTrpR-G4AD, pTol2-CC,ribA:nlsTrpR-G4AD, pTol2-CC,neuroD:nlsTrpRG4AD, pTol2-CC,neuroD:nlsTrpR_T81M-G4AD, pTol2-CC,neuroD:nlsTrpR_T81A-G4AD and pTol2-CG2,myo6b:gal80IREStRFP were generated using the constructs above and Gateway technology (Kwan et al., 2007; Villefranc et al., 2007).

One-cell zebrafish embryos were microinjected with 25 pg DNA constructs and 25 pg Tol2 transposase RNA to generate Tg(CV,tUAS:tRFP)^w80, Tg(CV,tUAS:nlsEos)^w81, Tg(CC,myo6b:nlsTrpR-G4AD)^w83 and Tg(CC,ribA:nlsTrpR-G4AD)^w85 germline transgenics as previously described (Fisher et al., 2006). Tg(CG2,myo6b:gfp) was kindly provided by A. Coffin (Washington State University-Vancouver, Vancouver, WA, USA) and will be described elsewhere.

Dual-Luciferase Reporter Assay (Promega) was used to assess the efficiency of TrpR modulators. One-cell stage zebrafish embryos were injected with 10 pg driver construct (pCMV:nlsTrpR-G4AD, pCMV:nlsTrpR_T81M-G4AD or pCMV:nlsTrpR_T81A-G4AD), 20 pg ptUAS:firefly luciferase and 5 pg pTK:Renilla luciferase. They were grown to 70% epiboly, ground and luminescence was measured using a Victor plate reader (PerkinElmer) sequentially after application of firefly substrate and Renilla substrate.

The HEK 293 cell line was transfected in 24-well dishes using Lipofectamine 2000 (Life Technologies) and standard procedures with 1 ng driver construct (pCMV:nlsTrpR-G4AD, pCMV:nlsTrpR_T81M-G4AD or pCMV:nlsTrpR_T81A-G4AD) and 50 ng ptUAS:firefly luciferase reporter construct per well plated with ∼100,000 cells per well the day before transfection. Luminescence was measured as described above.

Larvae were fixed with 4% paraformaldehyde for 2 hours at room temperature or overnight at 4°C, washed three times for 20 minutes each with PBST (0.1% Tween 20 in PBS) and incubated for 1 hour in distilled water. They were placed in block solution (1% BSA, 1% DMSO and 0.02% sodium azide in PBST, 10% normal goat serum) for 1 hour and then incubated with anti-Parvalbumin antibody (Millipore MAB1572; 1:400) overnight at 4°C. After four 20 minutes washes with PBST, they were incubated with secondary antibody (mouse anti-IgG1 Alexa488; Invitrogen) for 3 hours at room temperature, washed four times for 10 minutes each in PBST, and cleared in 50% glycerol/PBS. Embryos were imaged using an Olympus FV1000 confocal microscope.

We thank K. Kwan, J. Bonkowsky, C. B. Chien, L. Lagnado, D. Stainier and T. Nicolson for generously providing plasmids; A. Coffin for providing the transgenic line Tg(CG2,myo6b:GFP); and H. Stickney and C. Bouldin for helpful comments on the manuscript.