|Home | About | Journals | Submit | Contact Us | Français|
Sloan–Kettering Institute, Program in Developmental Biology, New York, NY 10065
Site specific recombinases (SSRs) are powerful tools for genome manipulation, used in diverse organisms including Drosophila melanogaster, mouse, Arabidopsis, zebrafish, and human cultured cells. The integrase from the bacteriophage ΦC31 belongs to the large serine family of integrases, and in contrast to other widely used SSRs such as Cre and Flp, recombination is directional and therefore irreversible. We have developed a vector system for recombinase mediated cassette exchange (RMCE) in the zebrafish, allowing swapping of the coding sequence in an integrated transgene. Utilizing codon–optimized ΦC31 integrase RNA bearing the 3′UTR from the nanos1 gene, we replaced the egfp coding sequence of an integrated reporter transgene with mCherry coding sequence. Recombination was achieved at high efficiency in both somatic cells and in the germline. We demonstrate an effective approach to RMCE, increasing the repertoire of tools available to manipulate the zebrafish genome.
Multiple site–specific recombinases (SSRs) have been identified from bacteriophage and plasmids, and adapted as tools for genome manipulations in eukaryotes. These enzymes catalyze recombination events, typically the integration of the phage or plasmid into the host genome, between specific short sequences. The most widely used are Cre, from the P1 bacteriophage, and FLP, from the S. cerevisiae 2μ plasmid; both are members of the tyrosine family of recombinases and have similar biochemical properties and activities. More recently, the integrase from the bacteriophage ΦC31 has been adapted for use in a wide number of organisms, and has proven especially useful in D. melanogaster. In the recombination events mediated by Cre and FLP, the recognition sequences (LoxP and FRT respectively) are homologous, and the reactions therefore are reversible. In contrast, the recombination mediated by ΦC31 integrase is directional; the att sites recognized by the integrase are heterologous, and the hybrid sequences created by recombination cannot be used by the integrase as substrates. This property may make certain types of reactions, such as recombinations between an integrated transgene and an exogenous plasmid, proceed more efficiently, since even in the presence of large excess of plasmid the reaction is irreversible.
An increasing number of tools are being developed to facilitate manipulations of the zebrafish genome, including efficient methods for transgenesis (Kawakami et al., 2000; Thermes et al., 2002; Moens et al., 2008), forward genetic approaches to induce or screen for mutations in specific genes (Doyon et al., 2008; Meng et al., 2008; Moens et al., 2008), and the use of the Cre/LoxP system to conditionally activate transgenes (Langenau et al., 2005; Pan et al., 2005; Thummel et al., 2005). However, having another SSR that could be practically used in zebrafish would allow the development of additional applications, particularly in a genetic background where Cre was also being used. In D. melanogaster, systems employing ΦC31 allow high frequency transgenesis (Groth et al., 2004; Fish et al., 2007), particularly for very large transgenes (Venken et al., 2006); provide well-characterized universal “docking sites” for transgenes, to avoid position effects (Markstein et al., 2008); and carry out recombinase–mediated cassette exchange (RMCE), in which a portion of an integrated transgene is swapped for a sequence on a targeting plasmid (Bateman et al., 2006). All of these applications require efficient intermolecular recombination between an integrated transgene and an exogenous plasmid. Two groups have recently described the use of ΦC31 in zebrafish, and demonstrated activity in somatic (Lu et al., 2010) and germ cells (Lister, 2010). In both cases, the reaction catalyzed was an intramolecular recombination event within an integrated transgene, resulting in excision of part of the transgene and activation of a second coding sequence. These reactions are similar to those typically mediated by the Cre/LoxP system, and will allow conditional activation of transgenes, particularly useful in conjunction with Cre. Here we report the use of ΦC31 integrase to efficiently mediate intermolecular recombination in zebrafish somatic and germ cells. This makes feasible the development of systems, analogous to those used in D. melanogaster, for RMCE (as we demonstrate here), and also for facilitating transgene integration into well–characterized chromosomal sites.
As part of an ongoing large–scale screen for enhancer sequences, we have analyzed many putative enhancers via transgenesis in zebrafish, by assessing their ability to drive tissue–specific expression of egfp in conjunction with a heterologous minimal promoter. The general approach and the vector we use for analysis have been previously described (Fisher et al., 2006a; Fisher et al., 2006b). Through this analysis, we have generated and characterized many transgenic lines with tissue–specific patterns of egfp expression. It would greatly increase the utility of these lines if the egfp coding sequence could be easily replaced, for example with that for another fluorescent protein or for Cre, which would be expressed in the identical pattern. Therefore, we sought to design a system to accomplish recombinase–mediated cassette exchange (RMCE) in the zebrafish, using the ΦC31 integrase; our strategy for vector design is illustrated in Figure 1.
We created the target transgene, intended to be integrated chromosomally, by modification of our standard enhancer analysis vector. To replace the entire coding sequence of the integrated transgene, two recombination events, at two flanking pairs of att sites, are required. For the att sites, we used the minimal sequences shown to have full recombination activity, as assayed in cultured human cells (Calos, 2006). The ΦC31 att sites are fairly short, less than 40 bp, and are likely to have little or no effect on transgene expression. However, to avoid any potential interference, we placed the attP sites outside of the cFos minimal promoter and polyadenylation sequence. To verify that enhancer function and transgene expression were unchanged, we incorporated a previously characterized enhancer, from upstream of the mouse Sox10 gene, into the modified construct. The enhancer regulates a distinct pattern of expression, in neural crest and oligodendrocytes (Antonellis et al., 2008). Following Tol2–mediated transgenesis, both the mosaic expression and that following germline transmission were indistinguishable from the original construct lacking the attP sites (data not shown).
Two groups have recently reported successful use of ΦC31 integrase in the zebrafish (Lister, 2010; Lu et al., 2010). However, in both cases the integrase mediated a single, intramolecular recombination event. For RMCE, two intermolecular recombinations must take place, and it is possible that the frequency will be lower and different optimal conditions will be required. To maximize integrase activity in the zebrafish, we used a version of the coding sequence (ΦC31o) that had been codon–optimized for the mouse (Raymond and Soriano, 2007). To test the feasibility of the required recombination and the activity of the integrase, we performed injections of the target transgene, with the Sox10 enhancer driving egfp, together with the mCherry (mCh) targeting plasmid, ΦC31o integrase RNA, and Tol2 transposase RNA. Although the mCh plasmid does contain a minimal promoter, and could be expressed if it integrates near a genomic regulatory sequence, we anticipate such an event would be quite rare, since the plasmid lacks the Tol2 repeat sequences. Only if the expected recombinations, mediated by ΦC31 integrase, took place would mCh fall under control of the Sox10 enhancer and display the appropriate tissue–specific expression. When all four components were injected, we frequently saw evidence of recombination in somatic cells, as seen by mCh expression in expected cell types, including neural crest and derivatives (Fig. 2A), often coexpressed with egfp. Such expression was not observed in the absence of ΦC31o integrase RNA (not shown). The resulting fish were raised, and with high frequency they transmitted the recombined transgene, expressing mCh under the control of the Sox10 enhancer (Table 1). Clearly the ΦC31 integrase is capable of carrying out the necessary recombinations in the zebrafish embryo. However, we cannot distinguish in this assay whether the recombination took place episomally or after integration of the target transgene.
To test the efficiency of recombination with an integrated transgene as a substrate, we isolated fish with the target transgene integrated chromosomally. Their GFP+ progeny were injected with the mCh targeting plasmid and ΦC31o integrase RNA. When >20 pg of ΦC31o RNA was injected per embryo, we began to observe a high frequency of abnormal embryos, and at 40 pg, most injected embryos were dead by 10 hours (data not shown). Therefore, we performed these injections with 20 pg RNA per embryo. As before, we saw evidence of mCh expression, and therefore recombination, in the appropriate somatic cells (not shown). In D. melanogaster, it has been shown that the frequency of recombination can vary at different chromosomal locations (Bischof et al., 2007). Therefore, we performed similar injections with fish carrying two other attP target transgenes, containing enhancers from the human AGGRECAN (ACAN) or TWIST1 loci. In these fish, we also observed somatic cells within the appropriate cell–specific pattern switch expression from egfp to mCh (Fig. 2B and data not shown).
To test for germline transmission of the recombination event, the Sox10:egfp embryos displaying mosaic mCh expression were raised to adulthood. We identified four independent adult founders that transmitted the altered transgene to a fraction of their progeny, which showed the full Sox 10–regulated pattern of mCh expression (Fig. 2E–H; Table 1; Table S1). Successful transmission of recombination was also observed for an additional founder carrying the TWIST1:egfp transgene. (Figure 1I–L; Table 1). However, the percentage of injected embryos that showed germline recombination as adults was low, making it impractical for a routine procedure. To attempt to increase the efficiency of recombination in the germline, we performed similar injections using ΦC31o integrase RNA altered by addition of the 3′UTR from the zebrafish nanos1 gene (nos3′UTR). This sequence enriches heterologous RNAs in the germline precursor cells of the zebrafish embryo (Koprunner et al., 2001), and we anticipated it would increase the effective concentration, and activity, of the integrase in the germline. GFP+ embryos were injected as above, with the targeting plasmid and the altered integrase RNA. We still observed some appropriate somatic expression of mCh in the injected embryos, suggesting that while the integrase RNA was enriched in the germ cells, it was not exclusively localized there. As in the previous injections, we selected injected embryos with somatic expression to raise; however, these typically comprised more than half of the viable embryos, so this selection was not likely to increase the frequency of identifying founders substantially. The injected embryos were raised, and a high percentage transmitted the expected expression pattern of mCherry to their progeny (Table 1; Table S1). Overall, we identified 37 founders from 74 screened adults in fish injected with the ΦC31o-nos3′UTR RNA, compared to 5/110 for the RNA lacking the nos3′UTR, indicating that inclusion of the 3′UTR substantially increased the frequency of recombination events in the germline.
Many founders gave rise to individual embryos expressing both egfp and mCh in an identical pattern. In some cases, these are presumably the result of founders transmitting multiple independent insertions of the attP construct, as is frequently observed with Tol2–mediated transgenesis. It is also possible that the entire targeting plasmid could insert via recombination at a single attP site, creating an unresolved intermediate, as has been described during RCME in Drosophila (Bateman et al., 2006). We tested individual GFP+/mCh+ progeny from four different adults, and verified the presence of plasmid backbone in progeny from two of the adults (data not shown). The presence of backbone is consistent with insertion of the entire targeting plasmid. We are currently testing whether additional injections of ΦC31o RNA can induce the second recombination event in the unresolved intermediates, as is the case in Drosophila (Bateman et al., 2006).
Following a complete RCME event, the mCherry cassette will be flanked by attR sites. Additionally, from a founder with only one genomic insertion of the original transgene, we should obtain progeny with either egfp or mCh expression, adding up to a total of approximately 50% of embryos in an outcross clutch. To verify complete recombination, we analyzed embryos from one such founder in more detail. We amplified portions of the transgenes from genomic DNA isolated from individual GFP+ or mCh+ embryos. In each pair of primers, one was designed to hybridize with either the egfp or mCh coding sequence; only the appropriate specific bands were detected from each sample (Fig. 3B). The amplified fragments were sequenced; from the GFP+ embryos, the unaltered attP site was present between the egfp coding and downstream sequences, while in the mCh+ embryos, attR sequences, produced by recombination, were found on both sides of the mCherry cassette (Fig. 3C, D).
To map the locations of the egfp and mCh transgenes, we performed linker–mediated PCR on DNA from individual GFP+ and mCh+ embryos from the same founder From each embryo we obtained a single amplified band, indicating the presence of only one Tol2 transgene. The sequence flanking the Tol2 arm was too short to place unambiguously in the genome, but was identical for the GFP+ and mCh+ embryos, indicating that in each embryo the Tol2 transgene is at the same genomic location. Together with the above data, showing that the mCh cassette is flanked by attR sites, we can conclude that the mCh cassette replaced the egfp cassette in the original transgene via RMCE.
As with other SSRs, such as Cre and Flp, ΦC31 integrase can mediate intramolecular recombinations resulting in excision of a portion of a transgene. It has previously been reported that it can carry out this type of recombination in zebrafish (Lister, 2010; Lu et al., 2010), making it a viable alternative to Cre for applications such as long–term lineage tracing or targeted misexpression of a gene. As transgenic lines expressing ΦC31 integrase in specific tissues are made and become available to the community, the utility of these approaches will increase. Here, we report for the first time efficient intermolecular recombination, mediated by ΦC31 integrase, in the zebrafish germline. Although the frequency was low in our initial experiments, the use of integrase RNA altered by addition of the nanos1 3′UTR greatly increased the recombination frequency in the germline. We found that nearly one half of injected fish were transmitting a recombined transgene, to as many as 18% of their progeny, making this a highly efficient and very practical approach to alter integrated transgenes. In our system, the sequence replaced by RMCE is relatively small (~1 kb). However, the native ΦC31 bacteriophage is >43 kb, and in Drosophila inserts >100 kb can be inserted via recombination at a single att site (Venken et al., 2006). There is less data on RMCE, but at least 5 kb sequences can be inserted with high efficiency (Bateman et al., 2006). If the same is true in zebrafish, it will expand the applications for which ΦC31–mediated recombination can be used.
Because the embryos carrying the integrated target transgene were derived via Tol2–mediated transgenesis, many of them were carrying multiple independent insertions, as could be seen from the total transmission frequencies of transgenic embryos, which were >50% for many fish (Table S1). In this scenario, it is difficult to determine if there are differences in efficiency of ΦC31–induced recombination at different chromosomal locations, although it has been described in Drosophila (Bischof et al., 2007). In the future, it would be highly beneficial in zebrafish to isolate a set of characterized insertion sites with integrated attP sequences that support stable transgene expression and high recombination frequencies. Such a collection could facilitate several applications, including high–frequency transgenesis with large constructs, and the quantitative comparison of expression levels among multiple transgenes, such as when testing mutated versions of enhancer sequences.
Wild–type fish of the AB strain were maintained by standard husbandry practices (Westerfield, 1995), and in accordance with institutional guidelines.
We modified our previously described vector for enhancer analysis, pGW-cFos-egfp (Fisher et al., 2006a; Fisher et al., 2006b) by the addition of two ΦC31 attP sites, flanking the cFos minimal promoter, egfp coding sequence, and poly–adenylation sequence. Insertion of the sites was achieved through PCR–mediated, ligase–independent mutagenesis (Chiu et al., 2004). The primers used to insert the two attP sites are listed in Table S2. After the separate insertion of the two attP sites, the vectors were digested with XhoI and ClaI and the resulting fragments ligated to obtain the final vector with two attP sites. To create the transgene target for in vivo tests, we inserted via the Gateway cassette a previously described enhancer from upstream of the mouse Sox10 gene (Antonellis et al., 2008), which regulates robust expression in the neural crest and oligodendrocytes of transgenic fish.
The targeting plasmid consisted of attB sites flanking the cFos minimal promoter, mCherry coding sequence, and poly–adenylation sequence, in a pBluescript vector backbone. It was constructed in the following steps: 1) The attB sites were inserted separately into pBluescript, as above, using the attB1 and attB2 primers listed in Table S2. 2) The two attB–containing plasmids were cut with Notl and Scal, and the two attB–containing bands ligated to obtain a plasmid with both attB sites. 3) The egfp coding sequence was replaced in the original pGW-cFos-egfp plasmid through PCR–mediated mutagenesis, using the mCh primers listed in Table S2. 4) From the resulting vector, the entire cFos-mCherry-pA cassette was amplified and inserted into the attB–containing pBluescript plasmid, using the cFos primers in Table S2. Zebrafish injections and transgenesis. A coding sequence for ΦC31 integrase, codon–optimized for expression in mouse (Raymond and Soriano, 2007) was cloned into the pCS2+ plasmid to allow efficient in vitro transcription of RNA for injections. For some experiments, the RNA was further modified by addition of the 3′UTR from the zebrafish nanos1 gene, which enriches transcripts in the germ cells of the early embryo (Koprunner et al., 2001).
For in vivo testing of integrase activity between episomal plasmids, the target transgene (~12 pg) and mCherry donor plasmid (~6 pg) were injected into embryos at the one–cell stage, together with RNAs encoding the Tol2 transposase and ΦC31o integrase (20 pg each); RNA and DNA were prepared, and injections performed, as previously described (Fisher et al., 2006a; Fisher et al., 2006b). To test RCME on an integrated transgene, fish carrying only the target transgene were generated by standard Tol2–mediate transgenesis and screened for germline transmission. The resulting GFP+ embryos were injected with the mCherry donor plasmid (10 pg) and ΦC31o integrase RNA (20 pg), either with or without the nanos1 3′UTR. PCR genotyping. A PCR assay was designed to distinguish the original target transgene (containing egfp) from the product of recombination (containing mCherry), using the primers listed in Table S2; the expected size of amplification products were 968 bp (egfpF-Tol2); 1172 bp (egfpR-Sox10); 1157 bp (mCherryF-Tol2); and 1230 bp (mCherryR-Sox10). DNA was isolated from individual embryos using standard protocols, subject to PCR, and the PCR products cloned for sequencing.
To map the insertion sites of the target transgene in the genome, we employed linker–mediated PCR, essentially as described previously (Davison et al., 2007; Akitake et al., 2011), using genomic DNA prepared from individual embryos. The insert–specific primers are complimentary to the Tol2 arms (Table S2), and hybridize to both the original egfp transgene and to the locus altered by recombination with the mCh donor plasmid.
The authors thank Haibo Zhang for helpful input on cloning strategies; Marnie Halpern for useful discussions and thoughtful comments on the manuscript; and Paula Roy and Liping Sun for excellent fish care. The work was supported by a grant from the National Human Genome Research Institute (R01HG005039) to SF, and MGG was supported by the Damon Runyon Cancer Research Foundation as a Damon Runyon Fellow (DRG#1945-07) and by a grant from the National Institute of Child Health and Human Development (R01HD058530) to Marnie Halpern.