The zebrafish (Danio rerio) AB/Tübingen (AB/TU) wild-type strain, and the transgenic line Tg[Rx3:Gal4:UAS:RFP] (Weiss et al., 2012) were maintained under standard conditions at 28°C. Embryos and larvae were kept in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM MgSO4) supplemented with Methylene Blue (Sigma) at 28°C and staged according to somite number and morphology (Kimmel et al., 1995). Animal experiments were carried out according to ethical regulations. Experimental protocols have been approved by the Animal Experimentation Ethics Committees at the Pablo de Olavide University and CSIC (license number 02/04/2018/041).

To develop the different approaches carried out in this work, two different constructs were built, based on the Tol2 transposition system (Kawakami et al., 2004): Tol2_trap:TRAP and Tol2_UAS:TRAP. Both constructs were obtained through modifications of the Tol2-zTRAP plasmid, which contains the eGFP-rpl10a fusion gene flanked by Tol2 sites (Tryon et al., 2013). The Tol2_trap:TRAP construct was generated by inserting the gata2p minimal promoter from the ZED vector (Bessa et al., 2009) into the Tol2-zTRAP multiple cloning site using SalI and BamHI restriction enzymes. The Tol2_UAS:TRAP plasmid was built by inserting the UAS fragment from the Tol2kit p5E_UAS plasmid (Kwan et al., 2007) into the Tol2-zTRAP multiple cloning site using BamHI and NheI. To obtain the Tol2_Vsx2.2:TRAP plasmid, a 5 Kb region of the vsx2.2 promoter (Nicolas-Perez et al., 2016) was also introduced in the Tol2-zTRAP multiple cloning site using NheI.

The polyribosome-attached RNA isolation was performed as previously described (Heiman et al., 2008) with some modifications. As starting material, 300 Tg(vsx2:TRAP) positive zebrafish embryos were selected at 22 hpf. Immunoprecipitation was performed using Dynabeads protein G (Invitrogen) and anti-GFP monoclonal antibodies 19C8 and 19F7 (Heiman et al., 2008). Finally, RNA was purified using the Arcturus Picopure RNA isolation kit (Thermo Fisher) and the final yield was determined by Qubit 4 fluorometer. Three independent libraries were prepared from the RNA sample and sequenced with an Illumina HiSeq 2500 (2 × 125 cycles). At least 33 M reads were obtained for each library. Quality of reads was assessed by FastQC (v.0.72) and sequences corresponding to rRNA were removed using sortmerna-2.1. The remaining reads were aligned through Hisat2 (v.2.2.1) to the Danio rerio GRCz11 genome assembly. Counts were obtained using featureCounts (v.2.0.1). Finally, mRNA abundance was calculated in Reads Per Kilo base per Million mapped reads (RPKM) allowing the comparison among different samples. For the comparative expression analysis between whole-embryo (GEO accession number: GSM4521281) and vsx2:TRAP transcripts, we examined RPKM values for a list of transcription factor (TF) encoding genes that have been shown as enriched at the neural retina domain vs the RPE in cell populations isolated by FACS (Buono et al., 2021). To reduce noise, we included in our analysis only those genes with a RPKM value above 5 (89 TFs).

The different transgenic lines were generated on the (AB/TU) background by microinjection of each vector (Tol2_Vsx2.2:TRAP, Tol2_trap:TRAP or Tol2_UAS:TRAP) into one-cell stage embryos. Following the Tol2 transposon/transposase transgenesis method (Kawakami et al., 2000), 100–200 pg of the plasmid were injected together with 100–200 pg of the Tol2 transposase mRNA.

The eGFP:rpl10a expression pattern displayed by the embryos of each trap-TRAP line was examined and photographed using an Olympus fluorescence micro stereoscope equipped with a camera Nikon. Pictures were taken at 24, 40, 48, and 72 hpf. To prevent embryo pigmentation and facilitate the analysis of their expression pattern, E3 medium was supplemented with 0.2 mM 1-phenyl-2-thiourea (PTU). In the case of the UAS:TRAP and the different Gal4 strains crosses, pictures of the progeny were taken at 24 h post fertilization. In all cases, the embryos were dechorionated with forceps if needed and mounted in 2% methylcellulose.

From each analysed line (TT15, TT21, TT37, and TT42), genomic DNA was isolated from 5 individual 72 hpf zebrafish embryos using Chelex 100 sodium form (Sigma Aldrich). Briefly, embryos were incubated in 45 µl of Chelex 5% together with 5 µl of proteinase K (10 mg/ml) for a minimum of 2 h. Afterwards, proteinase K was heat-inactivated (95°C, 10 min). To prepare PCR templates, 5 µl (approximately 250 ng) of the extracted DNA was first digested using the restriction enzyme DpnII (NEB), diluted to 100 µl and ligated with T4 ligase (NEB). The resultant ligation was used as a template for inverse PCRs. Two different nested PCRs were performed to amplify the junction fragments containing both Tol2 ends and their adjacent genomic DNA. The first PCR product was diluted to 200 µl with ddH2O before the second PCR. The primers and PCR programs used were described in a previous study (Bessa et al., 2014). The common band amplified from the five embryos was purified and cloned in TOPO (Invitrogen, pCR8/GW/TOPO TA cloning KIT). The resulting clones were sequenced using the T7 primer and the sequences obtained were subsequently examined to identify the adjacent fragments, which were aligned to the zebrafish genome (GRCz10/danRer10) using the UCSC BLAT Search Genome tool (https://genome.ucsc.edu/cgi-bin/hgBlat) (Supplementary Table S1). The expression profile of nearby genes was then compared them with that of the corresponding line.

All information on the different trap:TRAP lines was uploaded to the zebrafish TRAP database website: trap-TRAP database: (https://amfermin.wixsite.com/website). Photographs of the different trap:TRAP lines at different developmental stages were included together with their descriptions and the trap:TRAP methodology.

To expand the range of potential applications of TRAP methods in zebrafish we designed two different Tol2-based vectors (Kawakami et al., 2000). The first, Tol2_trap:TRAP (Figure 1A), allows combining translating ribosome affinity purification (TRAP) with a standard enhancer trap approach. To this end, the TRAP cassette (eGFP-rpl10a) was fused to the gata2p minimal promoter (Bessa et al., 2009). The random integration of this vector in the zebrafish genome allowed capturing the activity of nearby cis-regulatory elements (i.e., enhancers), leading to the expression of eGFP-rpl10a in a tissue-specific manner. The second vector, Tol2_UAS:TRAP (Figure 1C), was designed to take advantage of the collection of Gal4 lines available in zebrafish (Abe et al., 2011) as drivers for the TRAP cassette. Both vectors were then tested in transgenic assays as described in the following sections.

To further examine the eGFP-rpl10a reported capacity to efficiently pull-down tissue-specific transcripts in zebrafish (Tryon et al., 2013), we fused this cassette to the neural-retina specific promoter vsx2.2 (Nicolas-Perez et al., 2016; Buono et al., 2021). The resulting construct Tol2_vsx2.2:TRAP was microinjected into one-cell embryos to generate the corresponding zebrafish line Tg[vsx2.2:TRAP] (Figures 1D,E). Then, polysomes from 22 hpf Tg[vsx2.2:TRAP] embryos were harvested using a standard TRAP protocol, and captured RNA (120 ng) was deep sequenced (see Methods). To assess the retrieval efficiency of specific transcripts in that sample, we focused on the expression of a retina-enriched cluster of TFs (cluster 1), identified in our group in vsx2.2-positive retinal cells sorted by FACS at stage 23 hpf (Buono et al., 2021). We compared TFs relative mRNA abundance (RPKM) between affinity purified RNAs from vsx2.2:TRAP embryos at 22 hpf (TRAP-seq), and control RNA-seq samples obtained from whole embryos also at 22 hpf (GSM4521281). This analysis revealed a very significant enrichment (p = 0.0004; two-tailed paired t-test) of neural retina-specific transcription factors in the vsx2.2:TRAP sample (Figure 1F). Particularly high RPKM fold changes were observed for genes that belong to the core gene regulatory network of neural retina specification. This was the case for TF encoding-genes such as vsx2 itself (22.2-fold), rx2 (8.9-fold), rx1 (6.8-fold), six3b (4.8-fold), foxd1 (3.5-fold) or six6b (3.5-fold) (Figure 1F; Supplementary Table S2). These results confirm the functionality of the eGFP-rpl10a cassette for affinity purification of tissue-specific transcripts.

To test TRAP compatibility with an enhancer trap approach we carried out a pilot screen in zebrafish. The Tol2_trap:TRAP vector was injected in one-cell stage zebrafish embryos together with Tol2 transposase mRNA synthesized in vitro. Embryos showing any fluorescence at 24 hpf were selected and raised. A total of 300 adult fish were screened for eGFP-rpl10a expression by outcrossing them with wild-type animals. A minimum of 100 F1 embryos from each cross was examined at 24 and 48 hpf. We identified 53 positive founders, indicating an enhancer trapping efficiency of 17.5% (Figure 1B). This efficiency rate falls within the range of previously reported trapping screens using the Tol2 transposition system: 12% for enhancer trapping (Parinov et al., 2004) and 23% for gene trapping (Kawakami et al., 2004). Among all founders, 33 (62% of total) showed tissue-specific expression patterns often restricted to a single domain. Each of these founders was isolated and outcrossed to generate stable lines, which were designated as TT followed by an identifier number. In most cases, the eGFP-rpl10a distribution indicated that the integration event was unique. However, some founders showed segregation of expression patterns in their progeny, indicating multiple integration events. This was the case for the lines TT1, which is derived from the same founder that produced the line TT56. Similarly, the 5 lines TT37, TT42, TT57, TT58, and TT59 derive from insertions found in the same founder; as it is also the case for the pairs TT48/TT50; and TT53/TT60 which are derived from common founders respectively. Among the stable lines, eGFP-rpl10a signal was detectable in a variety of tissues from different embryonic origin (i.e., from different germ layers) and located in different positions along the embryo axes. This observation indicates that the eGFP-rpl10a fusion does not compromise the random integration of the trapping cassette.

The expression patterns of the 10 most relevant transgenic lines (i.e. those exhibiting higher tissue specificity at 40–48 hpf) are shown in Figure 2. These include specific lines for hindbrain and spinal cord (TT1, Figures 2A, 2A′); jaw, branchial arches and pectoral fin buds (TT5, Figures 2B, 2B′); skeletal muscles (TT6, Figure 2C); central nervous system (TT7, Figure 2D); lateral line system (TT15, Figure 2E); rhombomere 5 (TT21, Figures 2F, 2F′); hindbrain and pectoral fin buds (TT28, Figures 2G, 2G′); retina (TT37, Figures 2H, 2H′); midbrain stripe (TT42, Figures 2I, 2I′) and spinal cord/pronephros (TT50, Figure 2J). For remaining lines, descriptions of their eGFP-rpl10a distribution patterns are shown in Supplementary Figures S1, S2. All the information on the trap-TRAP transgenic lines has been uploaded to a web-based database: trap-TRAP database. This website displays TT lines pictures and descriptions of their expression pattern at different developmental stages, together with complementary information about the trap-TRAP technology.

To investigate if the random insertions of the trap-TRAP cassette in the genome recapitulate the activity of nearby cis-regulatory elements, we mapped the targeted loci of a few representative lines by inverse PCR and compared their expression patterns with those of neighbouring genes. The analysis provided conclusive mapping results for 3 out of 4 lines examined (Supplementary Table S1), the lines TT15, TT21, and TT42. In these cases, we successfully identified insertion sites in the vicinity of genes with an expression similar to that found for the corresponding strains (Figure 3). The line TT15, which showed expression in the lateral line, contains an insertion in the chromosome 1 (Figures 3A, 3A′). This region is 5′ distal to lef1, whose expression pattern includes a lateral line domain similar to that observed in TT15 embryos (Neelathi et al., 2018). Similarly, the analysis carried out on the TT21 line, which shows expression in rhombomere number 5, indicated that the insertion occurred on chromosome 23 (Figures 3B, 3B′). In the vicinity of the integration site is mafba, whose expression has been described in rhombomeres 5 and 6 (Moens et al., 1998). Finally, examination of the TT42 line identified an insertion on chromosome 8 (Figure 3C). This line shows a very specific expression stripe restricted to the midbrain (Figure 3C′). Accordingly, the neighbouring gene her3 also shows expression in this domain (Thisse and Thisse, 2005). In all cases, the integration loci were compared with available data on chromatin accessibility, (ATAC-seq), as well as H3K27ac and H3K4me3 epigenetic marks (i.e., to highlight active enhancers and promoters respectively) previously obtained in 48 hpf zebrafish embryos (Bogdanovic et al., 2012; Marlétaz et al., 2018). These comparisons indicate that the eGFP-rpl10a cassette is able to capture the activity of nearby cis-regulatory elements.

As an alternative strategy to direct eGFP-rpl10a expression to tissues or cell types of interest, we combined the TRAP methodology with the Gal4-UAS system, which allows the spatiotemporal control of gene expression (Brand and Perrimon, 1993; Scheer and Campos-Ortega, 1999; Köster and Fraser, 2001). To this end, the fusion gene was placed under the control of the UAS element to generate the Tol2_UAS:TRAP vector (Figure 1C). This construct was injected in one-cell stage zebrafish embryos together with in vitro -synthesized Tol2 transposase mRNA. To screen for potential adult founders, F0 animals were outcrossed with fish from a Gal4-expressing strain: Tg[Rx3:Gal4], whose expression is restricted to the retina as previously reported (Rembold et al., 2006; Weiss et al., 2012). As a reference control, we first verified retinal expression in the progeny of a Tg[Rx3:Gal4] X Tg[UAS:RFP] cross (Figure 4A). Then this retina-specific pattern was further confirmed in the founders progeny crossed to the Gal4 reference line: Tg[Rx3:Gal4] X Tg[UAS:TRAP] (Figure 4B). The fact that we did not observe toxicity, optic cup malformations, or developmental delays in these crosses indicates that Gal4 drivers can be successfully combined with the UAS:TRAP line to specifically perform TRAP analysis in a broad variety of tissues and cell types in zebrafish.